1 This is doc/gccint.info, produced by makeinfo version 4.8 from 2 /tmp/android-build-b1a4f38d56038d5f3847fea7c8c86b90/src/build/../gcc/gcc-4.4.3/gcc/doc/gccint.texi. 3 4 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 5 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free 6 Software Foundation, Inc. 7 8 Permission is granted to copy, distribute and/or modify this document 9 under the terms of the GNU Free Documentation License, Version 1.2 or 10 any later version published by the Free Software Foundation; with the 11 Invariant Sections being "Funding Free Software", the Front-Cover Texts 12 being (a) (see below), and with the Back-Cover Texts being (b) (see 13 below). A copy of the license is included in the section entitled "GNU 14 Free Documentation License". 15 16 (a) The FSF's Front-Cover Text is: 17 18 A GNU Manual 19 20 (b) The FSF's Back-Cover Text is: 21 22 You have freedom to copy and modify this GNU Manual, like GNU 23 software. Copies published by the Free Software Foundation raise 24 funds for GNU development. 25 26 INFO-DIR-SECTION Software development 27 START-INFO-DIR-ENTRY 28 * gccint: (gccint). Internals of the GNU Compiler Collection. 29 END-INFO-DIR-ENTRY 30 This file documents the internals of the GNU compilers. 31 32 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 33 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free 34 Software Foundation, Inc. 35 36 Permission is granted to copy, distribute and/or modify this document 37 under the terms of the GNU Free Documentation License, Version 1.2 or 38 any later version published by the Free Software Foundation; with the 39 Invariant Sections being "Funding Free Software", the Front-Cover Texts 40 being (a) (see below), and with the Back-Cover Texts being (b) (see 41 below). A copy of the license is included in the section entitled "GNU 42 Free Documentation License". 43 44 (a) The FSF's Front-Cover Text is: 45 46 A GNU Manual 47 48 (b) The FSF's Back-Cover Text is: 49 50 You have freedom to copy and modify this GNU Manual, like GNU 51 software. Copies published by the Free Software Foundation raise 52 funds for GNU development. 53 54 55 56 File: gccint.info, Node: Top, Next: Contributing, Up: (DIR) 57 58 Introduction 59 ************ 60 61 This manual documents the internals of the GNU compilers, including how 62 to port them to new targets and some information about how to write 63 front ends for new languages. It corresponds to the compilers 64 (GCC) version 4.4.3. The use of the GNU compilers is documented in a 65 separate manual. *Note Introduction: (gcc)Top. 66 67 This manual is mainly a reference manual rather than a tutorial. It 68 discusses how to contribute to GCC (*note Contributing::), the 69 characteristics of the machines supported by GCC as hosts and targets 70 (*note Portability::), how GCC relates to the ABIs on such systems 71 (*note Interface::), and the characteristics of the languages for which 72 GCC front ends are written (*note Languages::). It then describes the 73 GCC source tree structure and build system, some of the interfaces to 74 GCC front ends, and how support for a target system is implemented in 75 GCC. 76 77 Additional tutorial information is linked to from 78 `http://gcc.gnu.org/readings.html'. 79 80 * Menu: 81 82 * Contributing:: How to contribute to testing and developing GCC. 83 * Portability:: Goals of GCC's portability features. 84 * Interface:: Function-call interface of GCC output. 85 * Libgcc:: Low-level runtime library used by GCC. 86 * Languages:: Languages for which GCC front ends are written. 87 * Source Tree:: GCC source tree structure and build system. 88 * Options:: Option specification files. 89 * Passes:: Order of passes, what they do, and what each file is for. 90 * Trees:: The source representation used by the C and C++ front ends. 91 * GENERIC:: Language-independent representation generated by Front Ends 92 * GIMPLE:: Tuple representation used by Tree SSA optimizers 93 * Tree SSA:: Analysis and optimization of GIMPLE 94 * RTL:: Machine-dependent low-level intermediate representation. 95 * Control Flow:: Maintaining and manipulating the control flow graph. 96 * Loop Analysis and Representation:: Analysis and representation of loops 97 * Machine Desc:: How to write machine description instruction patterns. 98 * Target Macros:: How to write the machine description C macros and functions. 99 * Host Config:: Writing the `xm-MACHINE.h' file. 100 * Fragments:: Writing the `t-TARGET' and `x-HOST' files. 101 * Collect2:: How `collect2' works; how it finds `ld'. 102 * Header Dirs:: Understanding the standard header file directories. 103 * Type Information:: GCC's memory management; generating type information. 104 * Plugins:: Extending the compiler with plugins. 105 106 * Funding:: How to help assure funding for free software. 107 * GNU Project:: The GNU Project and GNU/Linux. 108 109 * Copying:: GNU General Public License says 110 how you can copy and share GCC. 111 * GNU Free Documentation License:: How you can copy and share this manual. 112 * Contributors:: People who have contributed to GCC. 113 114 * Option Index:: Index to command line options. 115 * Concept Index:: Index of concepts and symbol names. 116 117 118 File: gccint.info, Node: Contributing, Next: Portability, Prev: Top, Up: Top 119 120 1 Contributing to GCC Development 121 ********************************* 122 123 If you would like to help pretest GCC releases to assure they work well, 124 current development sources are available by SVN (see 125 `http://gcc.gnu.org/svn.html'). Source and binary snapshots are also 126 available for FTP; see `http://gcc.gnu.org/snapshots.html'. 127 128 If you would like to work on improvements to GCC, please read the 129 advice at these URLs: 130 131 `http://gcc.gnu.org/contribute.html' 132 `http://gcc.gnu.org/contributewhy.html' 133 134 for information on how to make useful contributions and avoid 135 duplication of effort. Suggested projects are listed at 136 `http://gcc.gnu.org/projects/'. 137 138 139 File: gccint.info, Node: Portability, Next: Interface, Prev: Contributing, Up: Top 140 141 2 GCC and Portability 142 ********************* 143 144 GCC itself aims to be portable to any machine where `int' is at least a 145 32-bit type. It aims to target machines with a flat (non-segmented) 146 byte addressed data address space (the code address space can be 147 separate). Target ABIs may have 8, 16, 32 or 64-bit `int' type. `char' 148 can be wider than 8 bits. 149 150 GCC gets most of the information about the target machine from a 151 machine description which gives an algebraic formula for each of the 152 machine's instructions. This is a very clean way to describe the 153 target. But when the compiler needs information that is difficult to 154 express in this fashion, ad-hoc parameters have been defined for 155 machine descriptions. The purpose of portability is to reduce the 156 total work needed on the compiler; it was not of interest for its own 157 sake. 158 159 GCC does not contain machine dependent code, but it does contain code 160 that depends on machine parameters such as endianness (whether the most 161 significant byte has the highest or lowest address of the bytes in a 162 word) and the availability of autoincrement addressing. In the 163 RTL-generation pass, it is often necessary to have multiple strategies 164 for generating code for a particular kind of syntax tree, strategies 165 that are usable for different combinations of parameters. Often, not 166 all possible cases have been addressed, but only the common ones or 167 only the ones that have been encountered. As a result, a new target 168 may require additional strategies. You will know if this happens 169 because the compiler will call `abort'. Fortunately, the new 170 strategies can be added in a machine-independent fashion, and will 171 affect only the target machines that need them. 172 173 174 File: gccint.info, Node: Interface, Next: Libgcc, Prev: Portability, Up: Top 175 176 3 Interfacing to GCC Output 177 *************************** 178 179 GCC is normally configured to use the same function calling convention 180 normally in use on the target system. This is done with the 181 machine-description macros described (*note Target Macros::). 182 183 However, returning of structure and union values is done differently on 184 some target machines. As a result, functions compiled with PCC 185 returning such types cannot be called from code compiled with GCC, and 186 vice versa. This does not cause trouble often because few Unix library 187 routines return structures or unions. 188 189 GCC code returns structures and unions that are 1, 2, 4 or 8 bytes 190 long in the same registers used for `int' or `double' return values. 191 (GCC typically allocates variables of such types in registers also.) 192 Structures and unions of other sizes are returned by storing them into 193 an address passed by the caller (usually in a register). The target 194 hook `TARGET_STRUCT_VALUE_RTX' tells GCC where to pass this address. 195 196 By contrast, PCC on most target machines returns structures and unions 197 of any size by copying the data into an area of static storage, and then 198 returning the address of that storage as if it were a pointer value. 199 The caller must copy the data from that memory area to the place where 200 the value is wanted. This is slower than the method used by GCC, and 201 fails to be reentrant. 202 203 On some target machines, such as RISC machines and the 80386, the 204 standard system convention is to pass to the subroutine the address of 205 where to return the value. On these machines, GCC has been configured 206 to be compatible with the standard compiler, when this method is used. 207 It may not be compatible for structures of 1, 2, 4 or 8 bytes. 208 209 GCC uses the system's standard convention for passing arguments. On 210 some machines, the first few arguments are passed in registers; in 211 others, all are passed on the stack. It would be possible to use 212 registers for argument passing on any machine, and this would probably 213 result in a significant speedup. But the result would be complete 214 incompatibility with code that follows the standard convention. So this 215 change is practical only if you are switching to GCC as the sole C 216 compiler for the system. We may implement register argument passing on 217 certain machines once we have a complete GNU system so that we can 218 compile the libraries with GCC. 219 220 On some machines (particularly the SPARC), certain types of arguments 221 are passed "by invisible reference". This means that the value is 222 stored in memory, and the address of the memory location is passed to 223 the subroutine. 224 225 If you use `longjmp', beware of automatic variables. ISO C says that 226 automatic variables that are not declared `volatile' have undefined 227 values after a `longjmp'. And this is all GCC promises to do, because 228 it is very difficult to restore register variables correctly, and one 229 of GCC's features is that it can put variables in registers without 230 your asking it to. 231 232 233 File: gccint.info, Node: Libgcc, Next: Languages, Prev: Interface, Up: Top 234 235 4 The GCC low-level runtime library 236 *********************************** 237 238 GCC provides a low-level runtime library, `libgcc.a' or `libgcc_s.so.1' 239 on some platforms. GCC generates calls to routines in this library 240 automatically, whenever it needs to perform some operation that is too 241 complicated to emit inline code for. 242 243 Most of the routines in `libgcc' handle arithmetic operations that the 244 target processor cannot perform directly. This includes integer 245 multiply and divide on some machines, and all floating-point and 246 fixed-point operations on other machines. `libgcc' also includes 247 routines for exception handling, and a handful of miscellaneous 248 operations. 249 250 Some of these routines can be defined in mostly machine-independent C. 251 Others must be hand-written in assembly language for each processor 252 that needs them. 253 254 GCC will also generate calls to C library routines, such as `memcpy' 255 and `memset', in some cases. The set of routines that GCC may possibly 256 use is documented in *Note Other Builtins: (gcc)Other Builtins. 257 258 These routines take arguments and return values of a specific machine 259 mode, not a specific C type. *Note Machine Modes::, for an explanation 260 of this concept. For illustrative purposes, in this chapter the 261 floating point type `float' is assumed to correspond to `SFmode'; 262 `double' to `DFmode'; and `long double' to both `TFmode' and `XFmode'. 263 Similarly, the integer types `int' and `unsigned int' correspond to 264 `SImode'; `long' and `unsigned long' to `DImode'; and `long long' and 265 `unsigned long long' to `TImode'. 266 267 * Menu: 268 269 * Integer library routines:: 270 * Soft float library routines:: 271 * Decimal float library routines:: 272 * Fixed-point fractional library routines:: 273 * Exception handling routines:: 274 * Miscellaneous routines:: 275 276 277 File: gccint.info, Node: Integer library routines, Next: Soft float library routines, Up: Libgcc 278 279 4.1 Routines for integer arithmetic 280 =================================== 281 282 The integer arithmetic routines are used on platforms that don't provide 283 hardware support for arithmetic operations on some modes. 284 285 4.1.1 Arithmetic functions 286 -------------------------- 287 288 -- Runtime Function: int __ashlsi3 (int A, int B) 289 -- Runtime Function: long __ashldi3 (long A, int B) 290 -- Runtime Function: long long __ashlti3 (long long A, int B) 291 These functions return the result of shifting A left by B bits. 292 293 -- Runtime Function: int __ashrsi3 (int A, int B) 294 -- Runtime Function: long __ashrdi3 (long A, int B) 295 -- Runtime Function: long long __ashrti3 (long long A, int B) 296 These functions return the result of arithmetically shifting A 297 right by B bits. 298 299 -- Runtime Function: int __divsi3 (int A, int B) 300 -- Runtime Function: long __divdi3 (long A, long B) 301 -- Runtime Function: long long __divti3 (long long A, long long B) 302 These functions return the quotient of the signed division of A and 303 B. 304 305 -- Runtime Function: int __lshrsi3 (int A, int B) 306 -- Runtime Function: long __lshrdi3 (long A, int B) 307 -- Runtime Function: long long __lshrti3 (long long A, int B) 308 These functions return the result of logically shifting A right by 309 B bits. 310 311 -- Runtime Function: int __modsi3 (int A, int B) 312 -- Runtime Function: long __moddi3 (long A, long B) 313 -- Runtime Function: long long __modti3 (long long A, long long B) 314 These functions return the remainder of the signed division of A 315 and B. 316 317 -- Runtime Function: int __mulsi3 (int A, int B) 318 -- Runtime Function: long __muldi3 (long A, long B) 319 -- Runtime Function: long long __multi3 (long long A, long long B) 320 These functions return the product of A and B. 321 322 -- Runtime Function: long __negdi2 (long A) 323 -- Runtime Function: long long __negti2 (long long A) 324 These functions return the negation of A. 325 326 -- Runtime Function: unsigned int __udivsi3 (unsigned int A, unsigned 327 int B) 328 -- Runtime Function: unsigned long __udivdi3 (unsigned long A, 329 unsigned long B) 330 -- Runtime Function: unsigned long long __udivti3 (unsigned long long 331 A, unsigned long long B) 332 These functions return the quotient of the unsigned division of A 333 and B. 334 335 -- Runtime Function: unsigned long __udivmoddi3 (unsigned long A, 336 unsigned long B, unsigned long *C) 337 -- Runtime Function: unsigned long long __udivti3 (unsigned long long 338 A, unsigned long long B, unsigned long long *C) 339 These functions calculate both the quotient and remainder of the 340 unsigned division of A and B. The return value is the quotient, 341 and the remainder is placed in variable pointed to by C. 342 343 -- Runtime Function: unsigned int __umodsi3 (unsigned int A, unsigned 344 int B) 345 -- Runtime Function: unsigned long __umoddi3 (unsigned long A, 346 unsigned long B) 347 -- Runtime Function: unsigned long long __umodti3 (unsigned long long 348 A, unsigned long long B) 349 These functions return the remainder of the unsigned division of A 350 and B. 351 352 4.1.2 Comparison functions 353 -------------------------- 354 355 The following functions implement integral comparisons. These functions 356 implement a low-level compare, upon which the higher level comparison 357 operators (such as less than and greater than or equal to) can be 358 constructed. The returned values lie in the range zero to two, to allow 359 the high-level operators to be implemented by testing the returned 360 result using either signed or unsigned comparison. 361 362 -- Runtime Function: int __cmpdi2 (long A, long B) 363 -- Runtime Function: int __cmpti2 (long long A, long long B) 364 These functions perform a signed comparison of A and B. If A is 365 less than B, they return 0; if A is greater than B, they return 2; 366 and if A and B are equal they return 1. 367 368 -- Runtime Function: int __ucmpdi2 (unsigned long A, unsigned long B) 369 -- Runtime Function: int __ucmpti2 (unsigned long long A, unsigned 370 long long B) 371 These functions perform an unsigned comparison of A and B. If A 372 is less than B, they return 0; if A is greater than B, they return 373 2; and if A and B are equal they return 1. 374 375 4.1.3 Trapping arithmetic functions 376 ----------------------------------- 377 378 The following functions implement trapping arithmetic. These functions 379 call the libc function `abort' upon signed arithmetic overflow. 380 381 -- Runtime Function: int __absvsi2 (int A) 382 -- Runtime Function: long __absvdi2 (long A) 383 These functions return the absolute value of A. 384 385 -- Runtime Function: int __addvsi3 (int A, int B) 386 -- Runtime Function: long __addvdi3 (long A, long B) 387 These functions return the sum of A and B; that is `A + B'. 388 389 -- Runtime Function: int __mulvsi3 (int A, int B) 390 -- Runtime Function: long __mulvdi3 (long A, long B) 391 The functions return the product of A and B; that is `A * B'. 392 393 -- Runtime Function: int __negvsi2 (int A) 394 -- Runtime Function: long __negvdi2 (long A) 395 These functions return the negation of A; that is `-A'. 396 397 -- Runtime Function: int __subvsi3 (int A, int B) 398 -- Runtime Function: long __subvdi3 (long A, long B) 399 These functions return the difference between B and A; that is `A 400 - B'. 401 402 4.1.4 Bit operations 403 -------------------- 404 405 -- Runtime Function: int __clzsi2 (int A) 406 -- Runtime Function: int __clzdi2 (long A) 407 -- Runtime Function: int __clzti2 (long long A) 408 These functions return the number of leading 0-bits in A, starting 409 at the most significant bit position. If A is zero, the result is 410 undefined. 411 412 -- Runtime Function: int __ctzsi2 (int A) 413 -- Runtime Function: int __ctzdi2 (long A) 414 -- Runtime Function: int __ctzti2 (long long A) 415 These functions return the number of trailing 0-bits in A, starting 416 at the least significant bit position. If A is zero, the result is 417 undefined. 418 419 -- Runtime Function: int __ffsdi2 (long A) 420 -- Runtime Function: int __ffsti2 (long long A) 421 These functions return the index of the least significant 1-bit in 422 A, or the value zero if A is zero. The least significant bit is 423 index one. 424 425 -- Runtime Function: int __paritysi2 (int A) 426 -- Runtime Function: int __paritydi2 (long A) 427 -- Runtime Function: int __parityti2 (long long A) 428 These functions return the value zero if the number of bits set in 429 A is even, and the value one otherwise. 430 431 -- Runtime Function: int __popcountsi2 (int A) 432 -- Runtime Function: int __popcountdi2 (long A) 433 -- Runtime Function: int __popcountti2 (long long A) 434 These functions return the number of bits set in A. 435 436 -- Runtime Function: int32_t __bswapsi2 (int32_t A) 437 -- Runtime Function: int64_t __bswapdi2 (int64_t A) 438 These functions return the A byteswapped. 439 440 441 File: gccint.info, Node: Soft float library routines, Next: Decimal float library routines, Prev: Integer library routines, Up: Libgcc 442 443 4.2 Routines for floating point emulation 444 ========================================= 445 446 The software floating point library is used on machines which do not 447 have hardware support for floating point. It is also used whenever 448 `-msoft-float' is used to disable generation of floating point 449 instructions. (Not all targets support this switch.) 450 451 For compatibility with other compilers, the floating point emulation 452 routines can be renamed with the `DECLARE_LIBRARY_RENAMES' macro (*note 453 Library Calls::). In this section, the default names are used. 454 455 Presently the library does not support `XFmode', which is used for 456 `long double' on some architectures. 457 458 4.2.1 Arithmetic functions 459 -------------------------- 460 461 -- Runtime Function: float __addsf3 (float A, float B) 462 -- Runtime Function: double __adddf3 (double A, double B) 463 -- Runtime Function: long double __addtf3 (long double A, long double 464 B) 465 -- Runtime Function: long double __addxf3 (long double A, long double 466 B) 467 These functions return the sum of A and B. 468 469 -- Runtime Function: float __subsf3 (float A, float B) 470 -- Runtime Function: double __subdf3 (double A, double B) 471 -- Runtime Function: long double __subtf3 (long double A, long double 472 B) 473 -- Runtime Function: long double __subxf3 (long double A, long double 474 B) 475 These functions return the difference between B and A; that is, 476 A - B. 477 478 -- Runtime Function: float __mulsf3 (float A, float B) 479 -- Runtime Function: double __muldf3 (double A, double B) 480 -- Runtime Function: long double __multf3 (long double A, long double 481 B) 482 -- Runtime Function: long double __mulxf3 (long double A, long double 483 B) 484 These functions return the product of A and B. 485 486 -- Runtime Function: float __divsf3 (float A, float B) 487 -- Runtime Function: double __divdf3 (double A, double B) 488 -- Runtime Function: long double __divtf3 (long double A, long double 489 B) 490 -- Runtime Function: long double __divxf3 (long double A, long double 491 B) 492 These functions return the quotient of A and B; that is, A / B. 493 494 -- Runtime Function: float __negsf2 (float A) 495 -- Runtime Function: double __negdf2 (double A) 496 -- Runtime Function: long double __negtf2 (long double A) 497 -- Runtime Function: long double __negxf2 (long double A) 498 These functions return the negation of A. They simply flip the 499 sign bit, so they can produce negative zero and negative NaN. 500 501 4.2.2 Conversion functions 502 -------------------------- 503 504 -- Runtime Function: double __extendsfdf2 (float A) 505 -- Runtime Function: long double __extendsftf2 (float A) 506 -- Runtime Function: long double __extendsfxf2 (float A) 507 -- Runtime Function: long double __extenddftf2 (double A) 508 -- Runtime Function: long double __extenddfxf2 (double A) 509 These functions extend A to the wider mode of their return type. 510 511 -- Runtime Function: double __truncxfdf2 (long double A) 512 -- Runtime Function: double __trunctfdf2 (long double A) 513 -- Runtime Function: float __truncxfsf2 (long double A) 514 -- Runtime Function: float __trunctfsf2 (long double A) 515 -- Runtime Function: float __truncdfsf2 (double A) 516 These functions truncate A to the narrower mode of their return 517 type, rounding toward zero. 518 519 -- Runtime Function: int __fixsfsi (float A) 520 -- Runtime Function: int __fixdfsi (double A) 521 -- Runtime Function: int __fixtfsi (long double A) 522 -- Runtime Function: int __fixxfsi (long double A) 523 These functions convert A to a signed integer, rounding toward 524 zero. 525 526 -- Runtime Function: long __fixsfdi (float A) 527 -- Runtime Function: long __fixdfdi (double A) 528 -- Runtime Function: long __fixtfdi (long double A) 529 -- Runtime Function: long __fixxfdi (long double A) 530 These functions convert A to a signed long, rounding toward zero. 531 532 -- Runtime Function: long long __fixsfti (float A) 533 -- Runtime Function: long long __fixdfti (double A) 534 -- Runtime Function: long long __fixtfti (long double A) 535 -- Runtime Function: long long __fixxfti (long double A) 536 These functions convert A to a signed long long, rounding toward 537 zero. 538 539 -- Runtime Function: unsigned int __fixunssfsi (float A) 540 -- Runtime Function: unsigned int __fixunsdfsi (double A) 541 -- Runtime Function: unsigned int __fixunstfsi (long double A) 542 -- Runtime Function: unsigned int __fixunsxfsi (long double A) 543 These functions convert A to an unsigned integer, rounding toward 544 zero. Negative values all become zero. 545 546 -- Runtime Function: unsigned long __fixunssfdi (float A) 547 -- Runtime Function: unsigned long __fixunsdfdi (double A) 548 -- Runtime Function: unsigned long __fixunstfdi (long double A) 549 -- Runtime Function: unsigned long __fixunsxfdi (long double A) 550 These functions convert A to an unsigned long, rounding toward 551 zero. Negative values all become zero. 552 553 -- Runtime Function: unsigned long long __fixunssfti (float A) 554 -- Runtime Function: unsigned long long __fixunsdfti (double A) 555 -- Runtime Function: unsigned long long __fixunstfti (long double A) 556 -- Runtime Function: unsigned long long __fixunsxfti (long double A) 557 These functions convert A to an unsigned long long, rounding 558 toward zero. Negative values all become zero. 559 560 -- Runtime Function: float __floatsisf (int I) 561 -- Runtime Function: double __floatsidf (int I) 562 -- Runtime Function: long double __floatsitf (int I) 563 -- Runtime Function: long double __floatsixf (int I) 564 These functions convert I, a signed integer, to floating point. 565 566 -- Runtime Function: float __floatdisf (long I) 567 -- Runtime Function: double __floatdidf (long I) 568 -- Runtime Function: long double __floatditf (long I) 569 -- Runtime Function: long double __floatdixf (long I) 570 These functions convert I, a signed long, to floating point. 571 572 -- Runtime Function: float __floattisf (long long I) 573 -- Runtime Function: double __floattidf (long long I) 574 -- Runtime Function: long double __floattitf (long long I) 575 -- Runtime Function: long double __floattixf (long long I) 576 These functions convert I, a signed long long, to floating point. 577 578 -- Runtime Function: float __floatunsisf (unsigned int I) 579 -- Runtime Function: double __floatunsidf (unsigned int I) 580 -- Runtime Function: long double __floatunsitf (unsigned int I) 581 -- Runtime Function: long double __floatunsixf (unsigned int I) 582 These functions convert I, an unsigned integer, to floating point. 583 584 -- Runtime Function: float __floatundisf (unsigned long I) 585 -- Runtime Function: double __floatundidf (unsigned long I) 586 -- Runtime Function: long double __floatunditf (unsigned long I) 587 -- Runtime Function: long double __floatundixf (unsigned long I) 588 These functions convert I, an unsigned long, to floating point. 589 590 -- Runtime Function: float __floatuntisf (unsigned long long I) 591 -- Runtime Function: double __floatuntidf (unsigned long long I) 592 -- Runtime Function: long double __floatuntitf (unsigned long long I) 593 -- Runtime Function: long double __floatuntixf (unsigned long long I) 594 These functions convert I, an unsigned long long, to floating 595 point. 596 597 4.2.3 Comparison functions 598 -------------------------- 599 600 There are two sets of basic comparison functions. 601 602 -- Runtime Function: int __cmpsf2 (float A, float B) 603 -- Runtime Function: int __cmpdf2 (double A, double B) 604 -- Runtime Function: int __cmptf2 (long double A, long double B) 605 These functions calculate a <=> b. That is, if A is less than B, 606 they return -1; if A is greater than B, they return 1; and if A 607 and B are equal they return 0. If either argument is NaN they 608 return 1, but you should not rely on this; if NaN is a 609 possibility, use one of the higher-level comparison functions. 610 611 -- Runtime Function: int __unordsf2 (float A, float B) 612 -- Runtime Function: int __unorddf2 (double A, double B) 613 -- Runtime Function: int __unordtf2 (long double A, long double B) 614 These functions return a nonzero value if either argument is NaN, 615 otherwise 0. 616 617 There is also a complete group of higher level functions which 618 correspond directly to comparison operators. They implement the ISO C 619 semantics for floating-point comparisons, taking NaN into account. Pay 620 careful attention to the return values defined for each set. Under the 621 hood, all of these routines are implemented as 622 623 if (__unordXf2 (a, b)) 624 return E; 625 return __cmpXf2 (a, b); 626 627 where E is a constant chosen to give the proper behavior for NaN. 628 Thus, the meaning of the return value is different for each set. Do 629 not rely on this implementation; only the semantics documented below 630 are guaranteed. 631 632 -- Runtime Function: int __eqsf2 (float A, float B) 633 -- Runtime Function: int __eqdf2 (double A, double B) 634 -- Runtime Function: int __eqtf2 (long double A, long double B) 635 These functions return zero if neither argument is NaN, and A and 636 B are equal. 637 638 -- Runtime Function: int __nesf2 (float A, float B) 639 -- Runtime Function: int __nedf2 (double A, double B) 640 -- Runtime Function: int __netf2 (long double A, long double B) 641 These functions return a nonzero value if either argument is NaN, 642 or if A and B are unequal. 643 644 -- Runtime Function: int __gesf2 (float A, float B) 645 -- Runtime Function: int __gedf2 (double A, double B) 646 -- Runtime Function: int __getf2 (long double A, long double B) 647 These functions return a value greater than or equal to zero if 648 neither argument is NaN, and A is greater than or equal to B. 649 650 -- Runtime Function: int __ltsf2 (float A, float B) 651 -- Runtime Function: int __ltdf2 (double A, double B) 652 -- Runtime Function: int __lttf2 (long double A, long double B) 653 These functions return a value less than zero if neither argument 654 is NaN, and A is strictly less than B. 655 656 -- Runtime Function: int __lesf2 (float A, float B) 657 -- Runtime Function: int __ledf2 (double A, double B) 658 -- Runtime Function: int __letf2 (long double A, long double B) 659 These functions return a value less than or equal to zero if 660 neither argument is NaN, and A is less than or equal to B. 661 662 -- Runtime Function: int __gtsf2 (float A, float B) 663 -- Runtime Function: int __gtdf2 (double A, double B) 664 -- Runtime Function: int __gttf2 (long double A, long double B) 665 These functions return a value greater than zero if neither 666 argument is NaN, and A is strictly greater than B. 667 668 4.2.4 Other floating-point functions 669 ------------------------------------ 670 671 -- Runtime Function: float __powisf2 (float A, int B) 672 -- Runtime Function: double __powidf2 (double A, int B) 673 -- Runtime Function: long double __powitf2 (long double A, int B) 674 -- Runtime Function: long double __powixf2 (long double A, int B) 675 These functions convert raise A to the power B. 676 677 -- Runtime Function: complex float __mulsc3 (float A, float B, float 678 C, float D) 679 -- Runtime Function: complex double __muldc3 (double A, double B, 680 double C, double D) 681 -- Runtime Function: complex long double __multc3 (long double A, long 682 double B, long double C, long double D) 683 -- Runtime Function: complex long double __mulxc3 (long double A, long 684 double B, long double C, long double D) 685 These functions return the product of A + iB and C + iD, following 686 the rules of C99 Annex G. 687 688 -- Runtime Function: complex float __divsc3 (float A, float B, float 689 C, float D) 690 -- Runtime Function: complex double __divdc3 (double A, double B, 691 double C, double D) 692 -- Runtime Function: complex long double __divtc3 (long double A, long 693 double B, long double C, long double D) 694 -- Runtime Function: complex long double __divxc3 (long double A, long 695 double B, long double C, long double D) 696 These functions return the quotient of A + iB and C + iD (i.e., (A 697 + iB) / (C + iD)), following the rules of C99 Annex G. 698 699 700 File: gccint.info, Node: Decimal float library routines, Next: Fixed-point fractional library routines, Prev: Soft float library routines, Up: Libgcc 701 702 4.3 Routines for decimal floating point emulation 703 ================================================= 704 705 The software decimal floating point library implements IEEE 754-2008 706 decimal floating point arithmetic and is only activated on selected 707 targets. 708 709 The software decimal floating point library supports either DPD 710 (Densely Packed Decimal) or BID (Binary Integer Decimal) encoding as 711 selected at configure time. 712 713 4.3.1 Arithmetic functions 714 -------------------------- 715 716 -- Runtime Function: _Decimal32 __dpd_addsd3 (_Decimal32 A, _Decimal32 717 B) 718 -- Runtime Function: _Decimal32 __bid_addsd3 (_Decimal32 A, _Decimal32 719 B) 720 -- Runtime Function: _Decimal64 __dpd_adddd3 (_Decimal64 A, _Decimal64 721 B) 722 -- Runtime Function: _Decimal64 __bid_adddd3 (_Decimal64 A, _Decimal64 723 B) 724 -- Runtime Function: _Decimal128 __dpd_addtd3 (_Decimal128 A, 725 _Decimal128 B) 726 -- Runtime Function: _Decimal128 __bid_addtd3 (_Decimal128 A, 727 _Decimal128 B) 728 These functions return the sum of A and B. 729 730 -- Runtime Function: _Decimal32 __dpd_subsd3 (_Decimal32 A, _Decimal32 731 B) 732 -- Runtime Function: _Decimal32 __bid_subsd3 (_Decimal32 A, _Decimal32 733 B) 734 -- Runtime Function: _Decimal64 __dpd_subdd3 (_Decimal64 A, _Decimal64 735 B) 736 -- Runtime Function: _Decimal64 __bid_subdd3 (_Decimal64 A, _Decimal64 737 B) 738 -- Runtime Function: _Decimal128 __dpd_subtd3 (_Decimal128 A, 739 _Decimal128 B) 740 -- Runtime Function: _Decimal128 __bid_subtd3 (_Decimal128 A, 741 _Decimal128 B) 742 These functions return the difference between B and A; that is, 743 A - B. 744 745 -- Runtime Function: _Decimal32 __dpd_mulsd3 (_Decimal32 A, _Decimal32 746 B) 747 -- Runtime Function: _Decimal32 __bid_mulsd3 (_Decimal32 A, _Decimal32 748 B) 749 -- Runtime Function: _Decimal64 __dpd_muldd3 (_Decimal64 A, _Decimal64 750 B) 751 -- Runtime Function: _Decimal64 __bid_muldd3 (_Decimal64 A, _Decimal64 752 B) 753 -- Runtime Function: _Decimal128 __dpd_multd3 (_Decimal128 A, 754 _Decimal128 B) 755 -- Runtime Function: _Decimal128 __bid_multd3 (_Decimal128 A, 756 _Decimal128 B) 757 These functions return the product of A and B. 758 759 -- Runtime Function: _Decimal32 __dpd_divsd3 (_Decimal32 A, _Decimal32 760 B) 761 -- Runtime Function: _Decimal32 __bid_divsd3 (_Decimal32 A, _Decimal32 762 B) 763 -- Runtime Function: _Decimal64 __dpd_divdd3 (_Decimal64 A, _Decimal64 764 B) 765 -- Runtime Function: _Decimal64 __bid_divdd3 (_Decimal64 A, _Decimal64 766 B) 767 -- Runtime Function: _Decimal128 __dpd_divtd3 (_Decimal128 A, 768 _Decimal128 B) 769 -- Runtime Function: _Decimal128 __bid_divtd3 (_Decimal128 A, 770 _Decimal128 B) 771 These functions return the quotient of A and B; that is, A / B. 772 773 -- Runtime Function: _Decimal32 __dpd_negsd2 (_Decimal32 A) 774 -- Runtime Function: _Decimal32 __bid_negsd2 (_Decimal32 A) 775 -- Runtime Function: _Decimal64 __dpd_negdd2 (_Decimal64 A) 776 -- Runtime Function: _Decimal64 __bid_negdd2 (_Decimal64 A) 777 -- Runtime Function: _Decimal128 __dpd_negtd2 (_Decimal128 A) 778 -- Runtime Function: _Decimal128 __bid_negtd2 (_Decimal128 A) 779 These functions return the negation of A. They simply flip the 780 sign bit, so they can produce negative zero and negative NaN. 781 782 4.3.2 Conversion functions 783 -------------------------- 784 785 -- Runtime Function: _Decimal64 __dpd_extendsddd2 (_Decimal32 A) 786 -- Runtime Function: _Decimal64 __bid_extendsddd2 (_Decimal32 A) 787 -- Runtime Function: _Decimal128 __dpd_extendsdtd2 (_Decimal32 A) 788 -- Runtime Function: _Decimal128 __bid_extendsdtd2 (_Decimal32 A) 789 -- Runtime Function: _Decimal128 __dpd_extendddtd2 (_Decimal64 A) 790 -- Runtime Function: _Decimal128 __bid_extendddtd2 (_Decimal64 A) 791 -- Runtime Function: _Decimal32 __dpd_truncddsd2 (_Decimal64 A) 792 -- Runtime Function: _Decimal32 __bid_truncddsd2 (_Decimal64 A) 793 -- Runtime Function: _Decimal32 __dpd_trunctdsd2 (_Decimal128 A) 794 -- Runtime Function: _Decimal32 __bid_trunctdsd2 (_Decimal128 A) 795 -- Runtime Function: _Decimal64 __dpd_trunctddd2 (_Decimal128 A) 796 -- Runtime Function: _Decimal64 __bid_trunctddd2 (_Decimal128 A) 797 These functions convert the value A from one decimal floating type 798 to another. 799 800 -- Runtime Function: _Decimal64 __dpd_extendsfdd (float A) 801 -- Runtime Function: _Decimal64 __bid_extendsfdd (float A) 802 -- Runtime Function: _Decimal128 __dpd_extendsftd (float A) 803 -- Runtime Function: _Decimal128 __bid_extendsftd (float A) 804 -- Runtime Function: _Decimal128 __dpd_extenddftd (double A) 805 -- Runtime Function: _Decimal128 __bid_extenddftd (double A) 806 -- Runtime Function: _Decimal128 __dpd_extendxftd (long double A) 807 -- Runtime Function: _Decimal128 __bid_extendxftd (long double A) 808 -- Runtime Function: _Decimal32 __dpd_truncdfsd (double A) 809 -- Runtime Function: _Decimal32 __bid_truncdfsd (double A) 810 -- Runtime Function: _Decimal32 __dpd_truncxfsd (long double A) 811 -- Runtime Function: _Decimal32 __bid_truncxfsd (long double A) 812 -- Runtime Function: _Decimal32 __dpd_trunctfsd (long double A) 813 -- Runtime Function: _Decimal32 __bid_trunctfsd (long double A) 814 -- Runtime Function: _Decimal64 __dpd_truncxfdd (long double A) 815 -- Runtime Function: _Decimal64 __bid_truncxfdd (long double A) 816 -- Runtime Function: _Decimal64 __dpd_trunctfdd (long double A) 817 -- Runtime Function: _Decimal64 __bid_trunctfdd (long double A) 818 These functions convert the value of A from a binary floating type 819 to a decimal floating type of a different size. 820 821 -- Runtime Function: float __dpd_truncddsf (_Decimal64 A) 822 -- Runtime Function: float __bid_truncddsf (_Decimal64 A) 823 -- Runtime Function: float __dpd_trunctdsf (_Decimal128 A) 824 -- Runtime Function: float __bid_trunctdsf (_Decimal128 A) 825 -- Runtime Function: double __dpd_extendsddf (_Decimal32 A) 826 -- Runtime Function: double __bid_extendsddf (_Decimal32 A) 827 -- Runtime Function: double __dpd_trunctddf (_Decimal128 A) 828 -- Runtime Function: double __bid_trunctddf (_Decimal128 A) 829 -- Runtime Function: long double __dpd_extendsdxf (_Decimal32 A) 830 -- Runtime Function: long double __bid_extendsdxf (_Decimal32 A) 831 -- Runtime Function: long double __dpd_extendddxf (_Decimal64 A) 832 -- Runtime Function: long double __bid_extendddxf (_Decimal64 A) 833 -- Runtime Function: long double __dpd_trunctdxf (_Decimal128 A) 834 -- Runtime Function: long double __bid_trunctdxf (_Decimal128 A) 835 -- Runtime Function: long double __dpd_extendsdtf (_Decimal32 A) 836 -- Runtime Function: long double __bid_extendsdtf (_Decimal32 A) 837 -- Runtime Function: long double __dpd_extendddtf (_Decimal64 A) 838 -- Runtime Function: long double __bid_extendddtf (_Decimal64 A) 839 These functions convert the value of A from a decimal floating type 840 to a binary floating type of a different size. 841 842 -- Runtime Function: _Decimal32 __dpd_extendsfsd (float A) 843 -- Runtime Function: _Decimal32 __bid_extendsfsd (float A) 844 -- Runtime Function: _Decimal64 __dpd_extenddfdd (double A) 845 -- Runtime Function: _Decimal64 __bid_extenddfdd (double A) 846 -- Runtime Function: _Decimal128 __dpd_extendtftd (long double A) 847 -- Runtime Function: _Decimal128 __bid_extendtftd (long double A) 848 -- Runtime Function: float __dpd_truncsdsf (_Decimal32 A) 849 -- Runtime Function: float __bid_truncsdsf (_Decimal32 A) 850 -- Runtime Function: double __dpd_truncdddf (_Decimal64 A) 851 -- Runtime Function: double __bid_truncdddf (_Decimal64 A) 852 -- Runtime Function: long double __dpd_trunctdtf (_Decimal128 A) 853 -- Runtime Function: long double __bid_trunctdtf (_Decimal128 A) 854 These functions convert the value of A between decimal and binary 855 floating types of the same size. 856 857 -- Runtime Function: int __dpd_fixsdsi (_Decimal32 A) 858 -- Runtime Function: int __bid_fixsdsi (_Decimal32 A) 859 -- Runtime Function: int __dpd_fixddsi (_Decimal64 A) 860 -- Runtime Function: int __bid_fixddsi (_Decimal64 A) 861 -- Runtime Function: int __dpd_fixtdsi (_Decimal128 A) 862 -- Runtime Function: int __bid_fixtdsi (_Decimal128 A) 863 These functions convert A to a signed integer. 864 865 -- Runtime Function: long __dpd_fixsddi (_Decimal32 A) 866 -- Runtime Function: long __bid_fixsddi (_Decimal32 A) 867 -- Runtime Function: long __dpd_fixdddi (_Decimal64 A) 868 -- Runtime Function: long __bid_fixdddi (_Decimal64 A) 869 -- Runtime Function: long __dpd_fixtddi (_Decimal128 A) 870 -- Runtime Function: long __bid_fixtddi (_Decimal128 A) 871 These functions convert A to a signed long. 872 873 -- Runtime Function: unsigned int __dpd_fixunssdsi (_Decimal32 A) 874 -- Runtime Function: unsigned int __bid_fixunssdsi (_Decimal32 A) 875 -- Runtime Function: unsigned int __dpd_fixunsddsi (_Decimal64 A) 876 -- Runtime Function: unsigned int __bid_fixunsddsi (_Decimal64 A) 877 -- Runtime Function: unsigned int __dpd_fixunstdsi (_Decimal128 A) 878 -- Runtime Function: unsigned int __bid_fixunstdsi (_Decimal128 A) 879 These functions convert A to an unsigned integer. Negative values 880 all become zero. 881 882 -- Runtime Function: unsigned long __dpd_fixunssddi (_Decimal32 A) 883 -- Runtime Function: unsigned long __bid_fixunssddi (_Decimal32 A) 884 -- Runtime Function: unsigned long __dpd_fixunsdddi (_Decimal64 A) 885 -- Runtime Function: unsigned long __bid_fixunsdddi (_Decimal64 A) 886 -- Runtime Function: unsigned long __dpd_fixunstddi (_Decimal128 A) 887 -- Runtime Function: unsigned long __bid_fixunstddi (_Decimal128 A) 888 These functions convert A to an unsigned long. Negative values 889 all become zero. 890 891 -- Runtime Function: _Decimal32 __dpd_floatsisd (int I) 892 -- Runtime Function: _Decimal32 __bid_floatsisd (int I) 893 -- Runtime Function: _Decimal64 __dpd_floatsidd (int I) 894 -- Runtime Function: _Decimal64 __bid_floatsidd (int I) 895 -- Runtime Function: _Decimal128 __dpd_floatsitd (int I) 896 -- Runtime Function: _Decimal128 __bid_floatsitd (int I) 897 These functions convert I, a signed integer, to decimal floating 898 point. 899 900 -- Runtime Function: _Decimal32 __dpd_floatdisd (long I) 901 -- Runtime Function: _Decimal32 __bid_floatdisd (long I) 902 -- Runtime Function: _Decimal64 __dpd_floatdidd (long I) 903 -- Runtime Function: _Decimal64 __bid_floatdidd (long I) 904 -- Runtime Function: _Decimal128 __dpd_floatditd (long I) 905 -- Runtime Function: _Decimal128 __bid_floatditd (long I) 906 These functions convert I, a signed long, to decimal floating 907 point. 908 909 -- Runtime Function: _Decimal32 __dpd_floatunssisd (unsigned int I) 910 -- Runtime Function: _Decimal32 __bid_floatunssisd (unsigned int I) 911 -- Runtime Function: _Decimal64 __dpd_floatunssidd (unsigned int I) 912 -- Runtime Function: _Decimal64 __bid_floatunssidd (unsigned int I) 913 -- Runtime Function: _Decimal128 __dpd_floatunssitd (unsigned int I) 914 -- Runtime Function: _Decimal128 __bid_floatunssitd (unsigned int I) 915 These functions convert I, an unsigned integer, to decimal 916 floating point. 917 918 -- Runtime Function: _Decimal32 __dpd_floatunsdisd (unsigned long I) 919 -- Runtime Function: _Decimal32 __bid_floatunsdisd (unsigned long I) 920 -- Runtime Function: _Decimal64 __dpd_floatunsdidd (unsigned long I) 921 -- Runtime Function: _Decimal64 __bid_floatunsdidd (unsigned long I) 922 -- Runtime Function: _Decimal128 __dpd_floatunsditd (unsigned long I) 923 -- Runtime Function: _Decimal128 __bid_floatunsditd (unsigned long I) 924 These functions convert I, an unsigned long, to decimal floating 925 point. 926 927 4.3.3 Comparison functions 928 -------------------------- 929 930 -- Runtime Function: int __dpd_unordsd2 (_Decimal32 A, _Decimal32 B) 931 -- Runtime Function: int __bid_unordsd2 (_Decimal32 A, _Decimal32 B) 932 -- Runtime Function: int __dpd_unorddd2 (_Decimal64 A, _Decimal64 B) 933 -- Runtime Function: int __bid_unorddd2 (_Decimal64 A, _Decimal64 B) 934 -- Runtime Function: int __dpd_unordtd2 (_Decimal128 A, _Decimal128 B) 935 -- Runtime Function: int __bid_unordtd2 (_Decimal128 A, _Decimal128 B) 936 These functions return a nonzero value if either argument is NaN, 937 otherwise 0. 938 939 There is also a complete group of higher level functions which 940 correspond directly to comparison operators. They implement the ISO C 941 semantics for floating-point comparisons, taking NaN into account. Pay 942 careful attention to the return values defined for each set. Under the 943 hood, all of these routines are implemented as 944 945 if (__bid_unordXd2 (a, b)) 946 return E; 947 return __bid_cmpXd2 (a, b); 948 949 where E is a constant chosen to give the proper behavior for NaN. 950 Thus, the meaning of the return value is different for each set. Do 951 not rely on this implementation; only the semantics documented below 952 are guaranteed. 953 954 -- Runtime Function: int __dpd_eqsd2 (_Decimal32 A, _Decimal32 B) 955 -- Runtime Function: int __bid_eqsd2 (_Decimal32 A, _Decimal32 B) 956 -- Runtime Function: int __dpd_eqdd2 (_Decimal64 A, _Decimal64 B) 957 -- Runtime Function: int __bid_eqdd2 (_Decimal64 A, _Decimal64 B) 958 -- Runtime Function: int __dpd_eqtd2 (_Decimal128 A, _Decimal128 B) 959 -- Runtime Function: int __bid_eqtd2 (_Decimal128 A, _Decimal128 B) 960 These functions return zero if neither argument is NaN, and A and 961 B are equal. 962 963 -- Runtime Function: int __dpd_nesd2 (_Decimal32 A, _Decimal32 B) 964 -- Runtime Function: int __bid_nesd2 (_Decimal32 A, _Decimal32 B) 965 -- Runtime Function: int __dpd_nedd2 (_Decimal64 A, _Decimal64 B) 966 -- Runtime Function: int __bid_nedd2 (_Decimal64 A, _Decimal64 B) 967 -- Runtime Function: int __dpd_netd2 (_Decimal128 A, _Decimal128 B) 968 -- Runtime Function: int __bid_netd2 (_Decimal128 A, _Decimal128 B) 969 These functions return a nonzero value if either argument is NaN, 970 or if A and B are unequal. 971 972 -- Runtime Function: int __dpd_gesd2 (_Decimal32 A, _Decimal32 B) 973 -- Runtime Function: int __bid_gesd2 (_Decimal32 A, _Decimal32 B) 974 -- Runtime Function: int __dpd_gedd2 (_Decimal64 A, _Decimal64 B) 975 -- Runtime Function: int __bid_gedd2 (_Decimal64 A, _Decimal64 B) 976 -- Runtime Function: int __dpd_getd2 (_Decimal128 A, _Decimal128 B) 977 -- Runtime Function: int __bid_getd2 (_Decimal128 A, _Decimal128 B) 978 These functions return a value greater than or equal to zero if 979 neither argument is NaN, and A is greater than or equal to B. 980 981 -- Runtime Function: int __dpd_ltsd2 (_Decimal32 A, _Decimal32 B) 982 -- Runtime Function: int __bid_ltsd2 (_Decimal32 A, _Decimal32 B) 983 -- Runtime Function: int __dpd_ltdd2 (_Decimal64 A, _Decimal64 B) 984 -- Runtime Function: int __bid_ltdd2 (_Decimal64 A, _Decimal64 B) 985 -- Runtime Function: int __dpd_lttd2 (_Decimal128 A, _Decimal128 B) 986 -- Runtime Function: int __bid_lttd2 (_Decimal128 A, _Decimal128 B) 987 These functions return a value less than zero if neither argument 988 is NaN, and A is strictly less than B. 989 990 -- Runtime Function: int __dpd_lesd2 (_Decimal32 A, _Decimal32 B) 991 -- Runtime Function: int __bid_lesd2 (_Decimal32 A, _Decimal32 B) 992 -- Runtime Function: int __dpd_ledd2 (_Decimal64 A, _Decimal64 B) 993 -- Runtime Function: int __bid_ledd2 (_Decimal64 A, _Decimal64 B) 994 -- Runtime Function: int __dpd_letd2 (_Decimal128 A, _Decimal128 B) 995 -- Runtime Function: int __bid_letd2 (_Decimal128 A, _Decimal128 B) 996 These functions return a value less than or equal to zero if 997 neither argument is NaN, and A is less than or equal to B. 998 999 -- Runtime Function: int __dpd_gtsd2 (_Decimal32 A, _Decimal32 B) 1000 -- Runtime Function: int __bid_gtsd2 (_Decimal32 A, _Decimal32 B) 1001 -- Runtime Function: int __dpd_gtdd2 (_Decimal64 A, _Decimal64 B) 1002 -- Runtime Function: int __bid_gtdd2 (_Decimal64 A, _Decimal64 B) 1003 -- Runtime Function: int __dpd_gttd2 (_Decimal128 A, _Decimal128 B) 1004 -- Runtime Function: int __bid_gttd2 (_Decimal128 A, _Decimal128 B) 1005 These functions return a value greater than zero if neither 1006 argument is NaN, and A is strictly greater than B. 1007 1008 1009 File: gccint.info, Node: Fixed-point fractional library routines, Next: Exception handling routines, Prev: Decimal float library routines, Up: Libgcc 1010 1011 4.4 Routines for fixed-point fractional emulation 1012 ================================================= 1013 1014 The software fixed-point library implements fixed-point fractional 1015 arithmetic, and is only activated on selected targets. 1016 1017 For ease of comprehension `fract' is an alias for the `_Fract' type, 1018 `accum' an alias for `_Accum', and `sat' an alias for `_Sat'. 1019 1020 For illustrative purposes, in this section the fixed-point fractional 1021 type `short fract' is assumed to correspond to machine mode `QQmode'; 1022 `unsigned short fract' to `UQQmode'; `fract' to `HQmode'; 1023 `unsigned fract' to `UHQmode'; `long fract' to `SQmode'; 1024 `unsigned long fract' to `USQmode'; `long long fract' to `DQmode'; and 1025 `unsigned long long fract' to `UDQmode'. Similarly the fixed-point 1026 accumulator type `short accum' corresponds to `HAmode'; 1027 `unsigned short accum' to `UHAmode'; `accum' to `SAmode'; 1028 `unsigned accum' to `USAmode'; `long accum' to `DAmode'; 1029 `unsigned long accum' to `UDAmode'; `long long accum' to `TAmode'; and 1030 `unsigned long long accum' to `UTAmode'. 1031 1032 4.4.1 Arithmetic functions 1033 -------------------------- 1034 1035 -- Runtime Function: short fract __addqq3 (short fract A, short fract 1036 B) 1037 -- Runtime Function: fract __addhq3 (fract A, fract B) 1038 -- Runtime Function: long fract __addsq3 (long fract A, long fract B) 1039 -- Runtime Function: long long fract __adddq3 (long long fract A, long 1040 long fract B) 1041 -- Runtime Function: unsigned short fract __adduqq3 (unsigned short 1042 fract A, unsigned short fract B) 1043 -- Runtime Function: unsigned fract __adduhq3 (unsigned fract A, 1044 unsigned fract B) 1045 -- Runtime Function: unsigned long fract __addusq3 (unsigned long 1046 fract A, unsigned long fract B) 1047 -- Runtime Function: unsigned long long fract __addudq3 (unsigned long 1048 long fract A, unsigned long long fract B) 1049 -- Runtime Function: short accum __addha3 (short accum A, short accum 1050 B) 1051 -- Runtime Function: accum __addsa3 (accum A, accum B) 1052 -- Runtime Function: long accum __addda3 (long accum A, long accum B) 1053 -- Runtime Function: long long accum __addta3 (long long accum A, long 1054 long accum B) 1055 -- Runtime Function: unsigned short accum __adduha3 (unsigned short 1056 accum A, unsigned short accum B) 1057 -- Runtime Function: unsigned accum __addusa3 (unsigned accum A, 1058 unsigned accum B) 1059 -- Runtime Function: unsigned long accum __adduda3 (unsigned long 1060 accum A, unsigned long accum B) 1061 -- Runtime Function: unsigned long long accum __adduta3 (unsigned long 1062 long accum A, unsigned long long accum B) 1063 These functions return the sum of A and B. 1064 1065 -- Runtime Function: short fract __ssaddqq3 (short fract A, short 1066 fract B) 1067 -- Runtime Function: fract __ssaddhq3 (fract A, fract B) 1068 -- Runtime Function: long fract __ssaddsq3 (long fract A, long fract B) 1069 -- Runtime Function: long long fract __ssadddq3 (long long fract A, 1070 long long fract B) 1071 -- Runtime Function: short accum __ssaddha3 (short accum A, short 1072 accum B) 1073 -- Runtime Function: accum __ssaddsa3 (accum A, accum B) 1074 -- Runtime Function: long accum __ssaddda3 (long accum A, long accum B) 1075 -- Runtime Function: long long accum __ssaddta3 (long long accum A, 1076 long long accum B) 1077 These functions return the sum of A and B with signed saturation. 1078 1079 -- Runtime Function: unsigned short fract __usadduqq3 (unsigned short 1080 fract A, unsigned short fract B) 1081 -- Runtime Function: unsigned fract __usadduhq3 (unsigned fract A, 1082 unsigned fract B) 1083 -- Runtime Function: unsigned long fract __usaddusq3 (unsigned long 1084 fract A, unsigned long fract B) 1085 -- Runtime Function: unsigned long long fract __usaddudq3 (unsigned 1086 long long fract A, unsigned long long fract B) 1087 -- Runtime Function: unsigned short accum __usadduha3 (unsigned short 1088 accum A, unsigned short accum B) 1089 -- Runtime Function: unsigned accum __usaddusa3 (unsigned accum A, 1090 unsigned accum B) 1091 -- Runtime Function: unsigned long accum __usadduda3 (unsigned long 1092 accum A, unsigned long accum B) 1093 -- Runtime Function: unsigned long long accum __usadduta3 (unsigned 1094 long long accum A, unsigned long long accum B) 1095 These functions return the sum of A and B with unsigned saturation. 1096 1097 -- Runtime Function: short fract __subqq3 (short fract A, short fract 1098 B) 1099 -- Runtime Function: fract __subhq3 (fract A, fract B) 1100 -- Runtime Function: long fract __subsq3 (long fract A, long fract B) 1101 -- Runtime Function: long long fract __subdq3 (long long fract A, long 1102 long fract B) 1103 -- Runtime Function: unsigned short fract __subuqq3 (unsigned short 1104 fract A, unsigned short fract B) 1105 -- Runtime Function: unsigned fract __subuhq3 (unsigned fract A, 1106 unsigned fract B) 1107 -- Runtime Function: unsigned long fract __subusq3 (unsigned long 1108 fract A, unsigned long fract B) 1109 -- Runtime Function: unsigned long long fract __subudq3 (unsigned long 1110 long fract A, unsigned long long fract B) 1111 -- Runtime Function: short accum __subha3 (short accum A, short accum 1112 B) 1113 -- Runtime Function: accum __subsa3 (accum A, accum B) 1114 -- Runtime Function: long accum __subda3 (long accum A, long accum B) 1115 -- Runtime Function: long long accum __subta3 (long long accum A, long 1116 long accum B) 1117 -- Runtime Function: unsigned short accum __subuha3 (unsigned short 1118 accum A, unsigned short accum B) 1119 -- Runtime Function: unsigned accum __subusa3 (unsigned accum A, 1120 unsigned accum B) 1121 -- Runtime Function: unsigned long accum __subuda3 (unsigned long 1122 accum A, unsigned long accum B) 1123 -- Runtime Function: unsigned long long accum __subuta3 (unsigned long 1124 long accum A, unsigned long long accum B) 1125 These functions return the difference of A and B; that is, `A - B'. 1126 1127 -- Runtime Function: short fract __sssubqq3 (short fract A, short 1128 fract B) 1129 -- Runtime Function: fract __sssubhq3 (fract A, fract B) 1130 -- Runtime Function: long fract __sssubsq3 (long fract A, long fract B) 1131 -- Runtime Function: long long fract __sssubdq3 (long long fract A, 1132 long long fract B) 1133 -- Runtime Function: short accum __sssubha3 (short accum A, short 1134 accum B) 1135 -- Runtime Function: accum __sssubsa3 (accum A, accum B) 1136 -- Runtime Function: long accum __sssubda3 (long accum A, long accum B) 1137 -- Runtime Function: long long accum __sssubta3 (long long accum A, 1138 long long accum B) 1139 These functions return the difference of A and B with signed 1140 saturation; that is, `A - B'. 1141 1142 -- Runtime Function: unsigned short fract __ussubuqq3 (unsigned short 1143 fract A, unsigned short fract B) 1144 -- Runtime Function: unsigned fract __ussubuhq3 (unsigned fract A, 1145 unsigned fract B) 1146 -- Runtime Function: unsigned long fract __ussubusq3 (unsigned long 1147 fract A, unsigned long fract B) 1148 -- Runtime Function: unsigned long long fract __ussubudq3 (unsigned 1149 long long fract A, unsigned long long fract B) 1150 -- Runtime Function: unsigned short accum __ussubuha3 (unsigned short 1151 accum A, unsigned short accum B) 1152 -- Runtime Function: unsigned accum __ussubusa3 (unsigned accum A, 1153 unsigned accum B) 1154 -- Runtime Function: unsigned long accum __ussubuda3 (unsigned long 1155 accum A, unsigned long accum B) 1156 -- Runtime Function: unsigned long long accum __ussubuta3 (unsigned 1157 long long accum A, unsigned long long accum B) 1158 These functions return the difference of A and B with unsigned 1159 saturation; that is, `A - B'. 1160 1161 -- Runtime Function: short fract __mulqq3 (short fract A, short fract 1162 B) 1163 -- Runtime Function: fract __mulhq3 (fract A, fract B) 1164 -- Runtime Function: long fract __mulsq3 (long fract A, long fract B) 1165 -- Runtime Function: long long fract __muldq3 (long long fract A, long 1166 long fract B) 1167 -- Runtime Function: unsigned short fract __muluqq3 (unsigned short 1168 fract A, unsigned short fract B) 1169 -- Runtime Function: unsigned fract __muluhq3 (unsigned fract A, 1170 unsigned fract B) 1171 -- Runtime Function: unsigned long fract __mulusq3 (unsigned long 1172 fract A, unsigned long fract B) 1173 -- Runtime Function: unsigned long long fract __muludq3 (unsigned long 1174 long fract A, unsigned long long fract B) 1175 -- Runtime Function: short accum __mulha3 (short accum A, short accum 1176 B) 1177 -- Runtime Function: accum __mulsa3 (accum A, accum B) 1178 -- Runtime Function: long accum __mulda3 (long accum A, long accum B) 1179 -- Runtime Function: long long accum __multa3 (long long accum A, long 1180 long accum B) 1181 -- Runtime Function: unsigned short accum __muluha3 (unsigned short 1182 accum A, unsigned short accum B) 1183 -- Runtime Function: unsigned accum __mulusa3 (unsigned accum A, 1184 unsigned accum B) 1185 -- Runtime Function: unsigned long accum __muluda3 (unsigned long 1186 accum A, unsigned long accum B) 1187 -- Runtime Function: unsigned long long accum __muluta3 (unsigned long 1188 long accum A, unsigned long long accum B) 1189 These functions return the product of A and B. 1190 1191 -- Runtime Function: short fract __ssmulqq3 (short fract A, short 1192 fract B) 1193 -- Runtime Function: fract __ssmulhq3 (fract A, fract B) 1194 -- Runtime Function: long fract __ssmulsq3 (long fract A, long fract B) 1195 -- Runtime Function: long long fract __ssmuldq3 (long long fract A, 1196 long long fract B) 1197 -- Runtime Function: short accum __ssmulha3 (short accum A, short 1198 accum B) 1199 -- Runtime Function: accum __ssmulsa3 (accum A, accum B) 1200 -- Runtime Function: long accum __ssmulda3 (long accum A, long accum B) 1201 -- Runtime Function: long long accum __ssmulta3 (long long accum A, 1202 long long accum B) 1203 These functions return the product of A and B with signed 1204 saturation. 1205 1206 -- Runtime Function: unsigned short fract __usmuluqq3 (unsigned short 1207 fract A, unsigned short fract B) 1208 -- Runtime Function: unsigned fract __usmuluhq3 (unsigned fract A, 1209 unsigned fract B) 1210 -- Runtime Function: unsigned long fract __usmulusq3 (unsigned long 1211 fract A, unsigned long fract B) 1212 -- Runtime Function: unsigned long long fract __usmuludq3 (unsigned 1213 long long fract A, unsigned long long fract B) 1214 -- Runtime Function: unsigned short accum __usmuluha3 (unsigned short 1215 accum A, unsigned short accum B) 1216 -- Runtime Function: unsigned accum __usmulusa3 (unsigned accum A, 1217 unsigned accum B) 1218 -- Runtime Function: unsigned long accum __usmuluda3 (unsigned long 1219 accum A, unsigned long accum B) 1220 -- Runtime Function: unsigned long long accum __usmuluta3 (unsigned 1221 long long accum A, unsigned long long accum B) 1222 These functions return the product of A and B with unsigned 1223 saturation. 1224 1225 -- Runtime Function: short fract __divqq3 (short fract A, short fract 1226 B) 1227 -- Runtime Function: fract __divhq3 (fract A, fract B) 1228 -- Runtime Function: long fract __divsq3 (long fract A, long fract B) 1229 -- Runtime Function: long long fract __divdq3 (long long fract A, long 1230 long fract B) 1231 -- Runtime Function: short accum __divha3 (short accum A, short accum 1232 B) 1233 -- Runtime Function: accum __divsa3 (accum A, accum B) 1234 -- Runtime Function: long accum __divda3 (long accum A, long accum B) 1235 -- Runtime Function: long long accum __divta3 (long long accum A, long 1236 long accum B) 1237 These functions return the quotient of the signed division of A 1238 and B. 1239 1240 -- Runtime Function: unsigned short fract __udivuqq3 (unsigned short 1241 fract A, unsigned short fract B) 1242 -- Runtime Function: unsigned fract __udivuhq3 (unsigned fract A, 1243 unsigned fract B) 1244 -- Runtime Function: unsigned long fract __udivusq3 (unsigned long 1245 fract A, unsigned long fract B) 1246 -- Runtime Function: unsigned long long fract __udivudq3 (unsigned 1247 long long fract A, unsigned long long fract B) 1248 -- Runtime Function: unsigned short accum __udivuha3 (unsigned short 1249 accum A, unsigned short accum B) 1250 -- Runtime Function: unsigned accum __udivusa3 (unsigned accum A, 1251 unsigned accum B) 1252 -- Runtime Function: unsigned long accum __udivuda3 (unsigned long 1253 accum A, unsigned long accum B) 1254 -- Runtime Function: unsigned long long accum __udivuta3 (unsigned 1255 long long accum A, unsigned long long accum B) 1256 These functions return the quotient of the unsigned division of A 1257 and B. 1258 1259 -- Runtime Function: short fract __ssdivqq3 (short fract A, short 1260 fract B) 1261 -- Runtime Function: fract __ssdivhq3 (fract A, fract B) 1262 -- Runtime Function: long fract __ssdivsq3 (long fract A, long fract B) 1263 -- Runtime Function: long long fract __ssdivdq3 (long long fract A, 1264 long long fract B) 1265 -- Runtime Function: short accum __ssdivha3 (short accum A, short 1266 accum B) 1267 -- Runtime Function: accum __ssdivsa3 (accum A, accum B) 1268 -- Runtime Function: long accum __ssdivda3 (long accum A, long accum B) 1269 -- Runtime Function: long long accum __ssdivta3 (long long accum A, 1270 long long accum B) 1271 These functions return the quotient of the signed division of A 1272 and B with signed saturation. 1273 1274 -- Runtime Function: unsigned short fract __usdivuqq3 (unsigned short 1275 fract A, unsigned short fract B) 1276 -- Runtime Function: unsigned fract __usdivuhq3 (unsigned fract A, 1277 unsigned fract B) 1278 -- Runtime Function: unsigned long fract __usdivusq3 (unsigned long 1279 fract A, unsigned long fract B) 1280 -- Runtime Function: unsigned long long fract __usdivudq3 (unsigned 1281 long long fract A, unsigned long long fract B) 1282 -- Runtime Function: unsigned short accum __usdivuha3 (unsigned short 1283 accum A, unsigned short accum B) 1284 -- Runtime Function: unsigned accum __usdivusa3 (unsigned accum A, 1285 unsigned accum B) 1286 -- Runtime Function: unsigned long accum __usdivuda3 (unsigned long 1287 accum A, unsigned long accum B) 1288 -- Runtime Function: unsigned long long accum __usdivuta3 (unsigned 1289 long long accum A, unsigned long long accum B) 1290 These functions return the quotient of the unsigned division of A 1291 and B with unsigned saturation. 1292 1293 -- Runtime Function: short fract __negqq2 (short fract A) 1294 -- Runtime Function: fract __neghq2 (fract A) 1295 -- Runtime Function: long fract __negsq2 (long fract A) 1296 -- Runtime Function: long long fract __negdq2 (long long fract A) 1297 -- Runtime Function: unsigned short fract __neguqq2 (unsigned short 1298 fract A) 1299 -- Runtime Function: unsigned fract __neguhq2 (unsigned fract A) 1300 -- Runtime Function: unsigned long fract __negusq2 (unsigned long 1301 fract A) 1302 -- Runtime Function: unsigned long long fract __negudq2 (unsigned long 1303 long fract A) 1304 -- Runtime Function: short accum __negha2 (short accum A) 1305 -- Runtime Function: accum __negsa2 (accum A) 1306 -- Runtime Function: long accum __negda2 (long accum A) 1307 -- Runtime Function: long long accum __negta2 (long long accum A) 1308 -- Runtime Function: unsigned short accum __neguha2 (unsigned short 1309 accum A) 1310 -- Runtime Function: unsigned accum __negusa2 (unsigned accum A) 1311 -- Runtime Function: unsigned long accum __neguda2 (unsigned long 1312 accum A) 1313 -- Runtime Function: unsigned long long accum __neguta2 (unsigned long 1314 long accum A) 1315 These functions return the negation of A. 1316 1317 -- Runtime Function: short fract __ssnegqq2 (short fract A) 1318 -- Runtime Function: fract __ssneghq2 (fract A) 1319 -- Runtime Function: long fract __ssnegsq2 (long fract A) 1320 -- Runtime Function: long long fract __ssnegdq2 (long long fract A) 1321 -- Runtime Function: short accum __ssnegha2 (short accum A) 1322 -- Runtime Function: accum __ssnegsa2 (accum A) 1323 -- Runtime Function: long accum __ssnegda2 (long accum A) 1324 -- Runtime Function: long long accum __ssnegta2 (long long accum A) 1325 These functions return the negation of A with signed saturation. 1326 1327 -- Runtime Function: unsigned short fract __usneguqq2 (unsigned short 1328 fract A) 1329 -- Runtime Function: unsigned fract __usneguhq2 (unsigned fract A) 1330 -- Runtime Function: unsigned long fract __usnegusq2 (unsigned long 1331 fract A) 1332 -- Runtime Function: unsigned long long fract __usnegudq2 (unsigned 1333 long long fract A) 1334 -- Runtime Function: unsigned short accum __usneguha2 (unsigned short 1335 accum A) 1336 -- Runtime Function: unsigned accum __usnegusa2 (unsigned accum A) 1337 -- Runtime Function: unsigned long accum __usneguda2 (unsigned long 1338 accum A) 1339 -- Runtime Function: unsigned long long accum __usneguta2 (unsigned 1340 long long accum A) 1341 These functions return the negation of A with unsigned saturation. 1342 1343 -- Runtime Function: short fract __ashlqq3 (short fract A, int B) 1344 -- Runtime Function: fract __ashlhq3 (fract A, int B) 1345 -- Runtime Function: long fract __ashlsq3 (long fract A, int B) 1346 -- Runtime Function: long long fract __ashldq3 (long long fract A, int 1347 B) 1348 -- Runtime Function: unsigned short fract __ashluqq3 (unsigned short 1349 fract A, int B) 1350 -- Runtime Function: unsigned fract __ashluhq3 (unsigned fract A, int 1351 B) 1352 -- Runtime Function: unsigned long fract __ashlusq3 (unsigned long 1353 fract A, int B) 1354 -- Runtime Function: unsigned long long fract __ashludq3 (unsigned 1355 long long fract A, int B) 1356 -- Runtime Function: short accum __ashlha3 (short accum A, int B) 1357 -- Runtime Function: accum __ashlsa3 (accum A, int B) 1358 -- Runtime Function: long accum __ashlda3 (long accum A, int B) 1359 -- Runtime Function: long long accum __ashlta3 (long long accum A, int 1360 B) 1361 -- Runtime Function: unsigned short accum __ashluha3 (unsigned short 1362 accum A, int B) 1363 -- Runtime Function: unsigned accum __ashlusa3 (unsigned accum A, int 1364 B) 1365 -- Runtime Function: unsigned long accum __ashluda3 (unsigned long 1366 accum A, int B) 1367 -- Runtime Function: unsigned long long accum __ashluta3 (unsigned 1368 long long accum A, int B) 1369 These functions return the result of shifting A left by B bits. 1370 1371 -- Runtime Function: short fract __ashrqq3 (short fract A, int B) 1372 -- Runtime Function: fract __ashrhq3 (fract A, int B) 1373 -- Runtime Function: long fract __ashrsq3 (long fract A, int B) 1374 -- Runtime Function: long long fract __ashrdq3 (long long fract A, int 1375 B) 1376 -- Runtime Function: short accum __ashrha3 (short accum A, int B) 1377 -- Runtime Function: accum __ashrsa3 (accum A, int B) 1378 -- Runtime Function: long accum __ashrda3 (long accum A, int B) 1379 -- Runtime Function: long long accum __ashrta3 (long long accum A, int 1380 B) 1381 These functions return the result of arithmetically shifting A 1382 right by B bits. 1383 1384 -- Runtime Function: unsigned short fract __lshruqq3 (unsigned short 1385 fract A, int B) 1386 -- Runtime Function: unsigned fract __lshruhq3 (unsigned fract A, int 1387 B) 1388 -- Runtime Function: unsigned long fract __lshrusq3 (unsigned long 1389 fract A, int B) 1390 -- Runtime Function: unsigned long long fract __lshrudq3 (unsigned 1391 long long fract A, int B) 1392 -- Runtime Function: unsigned short accum __lshruha3 (unsigned short 1393 accum A, int B) 1394 -- Runtime Function: unsigned accum __lshrusa3 (unsigned accum A, int 1395 B) 1396 -- Runtime Function: unsigned long accum __lshruda3 (unsigned long 1397 accum A, int B) 1398 -- Runtime Function: unsigned long long accum __lshruta3 (unsigned 1399 long long accum A, int B) 1400 These functions return the result of logically shifting A right by 1401 B bits. 1402 1403 -- Runtime Function: fract __ssashlhq3 (fract A, int B) 1404 -- Runtime Function: long fract __ssashlsq3 (long fract A, int B) 1405 -- Runtime Function: long long fract __ssashldq3 (long long fract A, 1406 int B) 1407 -- Runtime Function: short accum __ssashlha3 (short accum A, int B) 1408 -- Runtime Function: accum __ssashlsa3 (accum A, int B) 1409 -- Runtime Function: long accum __ssashlda3 (long accum A, int B) 1410 -- Runtime Function: long long accum __ssashlta3 (long long accum A, 1411 int B) 1412 These functions return the result of shifting A left by B bits 1413 with signed saturation. 1414 1415 -- Runtime Function: unsigned short fract __usashluqq3 (unsigned short 1416 fract A, int B) 1417 -- Runtime Function: unsigned fract __usashluhq3 (unsigned fract A, 1418 int B) 1419 -- Runtime Function: unsigned long fract __usashlusq3 (unsigned long 1420 fract A, int B) 1421 -- Runtime Function: unsigned long long fract __usashludq3 (unsigned 1422 long long fract A, int B) 1423 -- Runtime Function: unsigned short accum __usashluha3 (unsigned short 1424 accum A, int B) 1425 -- Runtime Function: unsigned accum __usashlusa3 (unsigned accum A, 1426 int B) 1427 -- Runtime Function: unsigned long accum __usashluda3 (unsigned long 1428 accum A, int B) 1429 -- Runtime Function: unsigned long long accum __usashluta3 (unsigned 1430 long long accum A, int B) 1431 These functions return the result of shifting A left by B bits 1432 with unsigned saturation. 1433 1434 4.4.2 Comparison functions 1435 -------------------------- 1436 1437 The following functions implement fixed-point comparisons. These 1438 functions implement a low-level compare, upon which the higher level 1439 comparison operators (such as less than and greater than or equal to) 1440 can be constructed. The returned values lie in the range zero to two, 1441 to allow the high-level operators to be implemented by testing the 1442 returned result using either signed or unsigned comparison. 1443 1444 -- Runtime Function: int __cmpqq2 (short fract A, short fract B) 1445 -- Runtime Function: int __cmphq2 (fract A, fract B) 1446 -- Runtime Function: int __cmpsq2 (long fract A, long fract B) 1447 -- Runtime Function: int __cmpdq2 (long long fract A, long long fract 1448 B) 1449 -- Runtime Function: int __cmpuqq2 (unsigned short fract A, unsigned 1450 short fract B) 1451 -- Runtime Function: int __cmpuhq2 (unsigned fract A, unsigned fract B) 1452 -- Runtime Function: int __cmpusq2 (unsigned long fract A, unsigned 1453 long fract B) 1454 -- Runtime Function: int __cmpudq2 (unsigned long long fract A, 1455 unsigned long long fract B) 1456 -- Runtime Function: int __cmpha2 (short accum A, short accum B) 1457 -- Runtime Function: int __cmpsa2 (accum A, accum B) 1458 -- Runtime Function: int __cmpda2 (long accum A, long accum B) 1459 -- Runtime Function: int __cmpta2 (long long accum A, long long accum 1460 B) 1461 -- Runtime Function: int __cmpuha2 (unsigned short accum A, unsigned 1462 short accum B) 1463 -- Runtime Function: int __cmpusa2 (unsigned accum A, unsigned accum B) 1464 -- Runtime Function: int __cmpuda2 (unsigned long accum A, unsigned 1465 long accum B) 1466 -- Runtime Function: int __cmputa2 (unsigned long long accum A, 1467 unsigned long long accum B) 1468 These functions perform a signed or unsigned comparison of A and B 1469 (depending on the selected machine mode). If A is less than B, 1470 they return 0; if A is greater than B, they return 2; and if A and 1471 B are equal they return 1. 1472 1473 4.4.3 Conversion functions 1474 -------------------------- 1475 1476 -- Runtime Function: fract __fractqqhq2 (short fract A) 1477 -- Runtime Function: long fract __fractqqsq2 (short fract A) 1478 -- Runtime Function: long long fract __fractqqdq2 (short fract A) 1479 -- Runtime Function: short accum __fractqqha (short fract A) 1480 -- Runtime Function: accum __fractqqsa (short fract A) 1481 -- Runtime Function: long accum __fractqqda (short fract A) 1482 -- Runtime Function: long long accum __fractqqta (short fract A) 1483 -- Runtime Function: unsigned short fract __fractqquqq (short fract A) 1484 -- Runtime Function: unsigned fract __fractqquhq (short fract A) 1485 -- Runtime Function: unsigned long fract __fractqqusq (short fract A) 1486 -- Runtime Function: unsigned long long fract __fractqqudq (short 1487 fract A) 1488 -- Runtime Function: unsigned short accum __fractqquha (short fract A) 1489 -- Runtime Function: unsigned accum __fractqqusa (short fract A) 1490 -- Runtime Function: unsigned long accum __fractqquda (short fract A) 1491 -- Runtime Function: unsigned long long accum __fractqquta (short 1492 fract A) 1493 -- Runtime Function: signed char __fractqqqi (short fract A) 1494 -- Runtime Function: short __fractqqhi (short fract A) 1495 -- Runtime Function: int __fractqqsi (short fract A) 1496 -- Runtime Function: long __fractqqdi (short fract A) 1497 -- Runtime Function: long long __fractqqti (short fract A) 1498 -- Runtime Function: float __fractqqsf (short fract A) 1499 -- Runtime Function: double __fractqqdf (short fract A) 1500 -- Runtime Function: short fract __fracthqqq2 (fract A) 1501 -- Runtime Function: long fract __fracthqsq2 (fract A) 1502 -- Runtime Function: long long fract __fracthqdq2 (fract A) 1503 -- Runtime Function: short accum __fracthqha (fract A) 1504 -- Runtime Function: accum __fracthqsa (fract A) 1505 -- Runtime Function: long accum __fracthqda (fract A) 1506 -- Runtime Function: long long accum __fracthqta (fract A) 1507 -- Runtime Function: unsigned short fract __fracthquqq (fract A) 1508 -- Runtime Function: unsigned fract __fracthquhq (fract A) 1509 -- Runtime Function: unsigned long fract __fracthqusq (fract A) 1510 -- Runtime Function: unsigned long long fract __fracthqudq (fract A) 1511 -- Runtime Function: unsigned short accum __fracthquha (fract A) 1512 -- Runtime Function: unsigned accum __fracthqusa (fract A) 1513 -- Runtime Function: unsigned long accum __fracthquda (fract A) 1514 -- Runtime Function: unsigned long long accum __fracthquta (fract A) 1515 -- Runtime Function: signed char __fracthqqi (fract A) 1516 -- Runtime Function: short __fracthqhi (fract A) 1517 -- Runtime Function: int __fracthqsi (fract A) 1518 -- Runtime Function: long __fracthqdi (fract A) 1519 -- Runtime Function: long long __fracthqti (fract A) 1520 -- Runtime Function: float __fracthqsf (fract A) 1521 -- Runtime Function: double __fracthqdf (fract A) 1522 -- Runtime Function: short fract __fractsqqq2 (long fract A) 1523 -- Runtime Function: fract __fractsqhq2 (long fract A) 1524 -- Runtime Function: long long fract __fractsqdq2 (long fract A) 1525 -- Runtime Function: short accum __fractsqha (long fract A) 1526 -- Runtime Function: accum __fractsqsa (long fract A) 1527 -- Runtime Function: long accum __fractsqda (long fract A) 1528 -- Runtime Function: long long accum __fractsqta (long fract A) 1529 -- Runtime Function: unsigned short fract __fractsquqq (long fract A) 1530 -- Runtime Function: unsigned fract __fractsquhq (long fract A) 1531 -- Runtime Function: unsigned long fract __fractsqusq (long fract A) 1532 -- Runtime Function: unsigned long long fract __fractsqudq (long fract 1533 A) 1534 -- Runtime Function: unsigned short accum __fractsquha (long fract A) 1535 -- Runtime Function: unsigned accum __fractsqusa (long fract A) 1536 -- Runtime Function: unsigned long accum __fractsquda (long fract A) 1537 -- Runtime Function: unsigned long long accum __fractsquta (long fract 1538 A) 1539 -- Runtime Function: signed char __fractsqqi (long fract A) 1540 -- Runtime Function: short __fractsqhi (long fract A) 1541 -- Runtime Function: int __fractsqsi (long fract A) 1542 -- Runtime Function: long __fractsqdi (long fract A) 1543 -- Runtime Function: long long __fractsqti (long fract A) 1544 -- Runtime Function: float __fractsqsf (long fract A) 1545 -- Runtime Function: double __fractsqdf (long fract A) 1546 -- Runtime Function: short fract __fractdqqq2 (long long fract A) 1547 -- Runtime Function: fract __fractdqhq2 (long long fract A) 1548 -- Runtime Function: long fract __fractdqsq2 (long long fract A) 1549 -- Runtime Function: short accum __fractdqha (long long fract A) 1550 -- Runtime Function: accum __fractdqsa (long long fract A) 1551 -- Runtime Function: long accum __fractdqda (long long fract A) 1552 -- Runtime Function: long long accum __fractdqta (long long fract A) 1553 -- Runtime Function: unsigned short fract __fractdquqq (long long 1554 fract A) 1555 -- Runtime Function: unsigned fract __fractdquhq (long long fract A) 1556 -- Runtime Function: unsigned long fract __fractdqusq (long long fract 1557 A) 1558 -- Runtime Function: unsigned long long fract __fractdqudq (long long 1559 fract A) 1560 -- Runtime Function: unsigned short accum __fractdquha (long long 1561 fract A) 1562 -- Runtime Function: unsigned accum __fractdqusa (long long fract A) 1563 -- Runtime Function: unsigned long accum __fractdquda (long long fract 1564 A) 1565 -- Runtime Function: unsigned long long accum __fractdquta (long long 1566 fract A) 1567 -- Runtime Function: signed char __fractdqqi (long long fract A) 1568 -- Runtime Function: short __fractdqhi (long long fract A) 1569 -- Runtime Function: int __fractdqsi (long long fract A) 1570 -- Runtime Function: long __fractdqdi (long long fract A) 1571 -- Runtime Function: long long __fractdqti (long long fract A) 1572 -- Runtime Function: float __fractdqsf (long long fract A) 1573 -- Runtime Function: double __fractdqdf (long long fract A) 1574 -- Runtime Function: short fract __fracthaqq (short accum A) 1575 -- Runtime Function: fract __fracthahq (short accum A) 1576 -- Runtime Function: long fract __fracthasq (short accum A) 1577 -- Runtime Function: long long fract __fracthadq (short accum A) 1578 -- Runtime Function: accum __fracthasa2 (short accum A) 1579 -- Runtime Function: long accum __fracthada2 (short accum A) 1580 -- Runtime Function: long long accum __fracthata2 (short accum A) 1581 -- Runtime Function: unsigned short fract __fracthauqq (short accum A) 1582 -- Runtime Function: unsigned fract __fracthauhq (short accum A) 1583 -- Runtime Function: unsigned long fract __fracthausq (short accum A) 1584 -- Runtime Function: unsigned long long fract __fracthaudq (short 1585 accum A) 1586 -- Runtime Function: unsigned short accum __fracthauha (short accum A) 1587 -- Runtime Function: unsigned accum __fracthausa (short accum A) 1588 -- Runtime Function: unsigned long accum __fracthauda (short accum A) 1589 -- Runtime Function: unsigned long long accum __fracthauta (short 1590 accum A) 1591 -- Runtime Function: signed char __fracthaqi (short accum A) 1592 -- Runtime Function: short __fracthahi (short accum A) 1593 -- Runtime Function: int __fracthasi (short accum A) 1594 -- Runtime Function: long __fracthadi (short accum A) 1595 -- Runtime Function: long long __fracthati (short accum A) 1596 -- Runtime Function: float __fracthasf (short accum A) 1597 -- Runtime Function: double __fracthadf (short accum A) 1598 -- Runtime Function: short fract __fractsaqq (accum A) 1599 -- Runtime Function: fract __fractsahq (accum A) 1600 -- Runtime Function: long fract __fractsasq (accum A) 1601 -- Runtime Function: long long fract __fractsadq (accum A) 1602 -- Runtime Function: short accum __fractsaha2 (accum A) 1603 -- Runtime Function: long accum __fractsada2 (accum A) 1604 -- Runtime Function: long long accum __fractsata2 (accum A) 1605 -- Runtime Function: unsigned short fract __fractsauqq (accum A) 1606 -- Runtime Function: unsigned fract __fractsauhq (accum A) 1607 -- Runtime Function: unsigned long fract __fractsausq (accum A) 1608 -- Runtime Function: unsigned long long fract __fractsaudq (accum A) 1609 -- Runtime Function: unsigned short accum __fractsauha (accum A) 1610 -- Runtime Function: unsigned accum __fractsausa (accum A) 1611 -- Runtime Function: unsigned long accum __fractsauda (accum A) 1612 -- Runtime Function: unsigned long long accum __fractsauta (accum A) 1613 -- Runtime Function: signed char __fractsaqi (accum A) 1614 -- Runtime Function: short __fractsahi (accum A) 1615 -- Runtime Function: int __fractsasi (accum A) 1616 -- Runtime Function: long __fractsadi (accum A) 1617 -- Runtime Function: long long __fractsati (accum A) 1618 -- Runtime Function: float __fractsasf (accum A) 1619 -- Runtime Function: double __fractsadf (accum A) 1620 -- Runtime Function: short fract __fractdaqq (long accum A) 1621 -- Runtime Function: fract __fractdahq (long accum A) 1622 -- Runtime Function: long fract __fractdasq (long accum A) 1623 -- Runtime Function: long long fract __fractdadq (long accum A) 1624 -- Runtime Function: short accum __fractdaha2 (long accum A) 1625 -- Runtime Function: accum __fractdasa2 (long accum A) 1626 -- Runtime Function: long long accum __fractdata2 (long accum A) 1627 -- Runtime Function: unsigned short fract __fractdauqq (long accum A) 1628 -- Runtime Function: unsigned fract __fractdauhq (long accum A) 1629 -- Runtime Function: unsigned long fract __fractdausq (long accum A) 1630 -- Runtime Function: unsigned long long fract __fractdaudq (long accum 1631 A) 1632 -- Runtime Function: unsigned short accum __fractdauha (long accum A) 1633 -- Runtime Function: unsigned accum __fractdausa (long accum A) 1634 -- Runtime Function: unsigned long accum __fractdauda (long accum A) 1635 -- Runtime Function: unsigned long long accum __fractdauta (long accum 1636 A) 1637 -- Runtime Function: signed char __fractdaqi (long accum A) 1638 -- Runtime Function: short __fractdahi (long accum A) 1639 -- Runtime Function: int __fractdasi (long accum A) 1640 -- Runtime Function: long __fractdadi (long accum A) 1641 -- Runtime Function: long long __fractdati (long accum A) 1642 -- Runtime Function: float __fractdasf (long accum A) 1643 -- Runtime Function: double __fractdadf (long accum A) 1644 -- Runtime Function: short fract __fracttaqq (long long accum A) 1645 -- Runtime Function: fract __fracttahq (long long accum A) 1646 -- Runtime Function: long fract __fracttasq (long long accum A) 1647 -- Runtime Function: long long fract __fracttadq (long long accum A) 1648 -- Runtime Function: short accum __fracttaha2 (long long accum A) 1649 -- Runtime Function: accum __fracttasa2 (long long accum A) 1650 -- Runtime Function: long accum __fracttada2 (long long accum A) 1651 -- Runtime Function: unsigned short fract __fracttauqq (long long 1652 accum A) 1653 -- Runtime Function: unsigned fract __fracttauhq (long long accum A) 1654 -- Runtime Function: unsigned long fract __fracttausq (long long accum 1655 A) 1656 -- Runtime Function: unsigned long long fract __fracttaudq (long long 1657 accum A) 1658 -- Runtime Function: unsigned short accum __fracttauha (long long 1659 accum A) 1660 -- Runtime Function: unsigned accum __fracttausa (long long accum A) 1661 -- Runtime Function: unsigned long accum __fracttauda (long long accum 1662 A) 1663 -- Runtime Function: unsigned long long accum __fracttauta (long long 1664 accum A) 1665 -- Runtime Function: signed char __fracttaqi (long long accum A) 1666 -- Runtime Function: short __fracttahi (long long accum A) 1667 -- Runtime Function: int __fracttasi (long long accum A) 1668 -- Runtime Function: long __fracttadi (long long accum A) 1669 -- Runtime Function: long long __fracttati (long long accum A) 1670 -- Runtime Function: float __fracttasf (long long accum A) 1671 -- Runtime Function: double __fracttadf (long long accum A) 1672 -- Runtime Function: short fract __fractuqqqq (unsigned short fract A) 1673 -- Runtime Function: fract __fractuqqhq (unsigned short fract A) 1674 -- Runtime Function: long fract __fractuqqsq (unsigned short fract A) 1675 -- Runtime Function: long long fract __fractuqqdq (unsigned short 1676 fract A) 1677 -- Runtime Function: short accum __fractuqqha (unsigned short fract A) 1678 -- Runtime Function: accum __fractuqqsa (unsigned short fract A) 1679 -- Runtime Function: long accum __fractuqqda (unsigned short fract A) 1680 -- Runtime Function: long long accum __fractuqqta (unsigned short 1681 fract A) 1682 -- Runtime Function: unsigned fract __fractuqquhq2 (unsigned short 1683 fract A) 1684 -- Runtime Function: unsigned long fract __fractuqqusq2 (unsigned 1685 short fract A) 1686 -- Runtime Function: unsigned long long fract __fractuqqudq2 (unsigned 1687 short fract A) 1688 -- Runtime Function: unsigned short accum __fractuqquha (unsigned 1689 short fract A) 1690 -- Runtime Function: unsigned accum __fractuqqusa (unsigned short 1691 fract A) 1692 -- Runtime Function: unsigned long accum __fractuqquda (unsigned short 1693 fract A) 1694 -- Runtime Function: unsigned long long accum __fractuqquta (unsigned 1695 short fract A) 1696 -- Runtime Function: signed char __fractuqqqi (unsigned short fract A) 1697 -- Runtime Function: short __fractuqqhi (unsigned short fract A) 1698 -- Runtime Function: int __fractuqqsi (unsigned short fract A) 1699 -- Runtime Function: long __fractuqqdi (unsigned short fract A) 1700 -- Runtime Function: long long __fractuqqti (unsigned short fract A) 1701 -- Runtime Function: float __fractuqqsf (unsigned short fract A) 1702 -- Runtime Function: double __fractuqqdf (unsigned short fract A) 1703 -- Runtime Function: short fract __fractuhqqq (unsigned fract A) 1704 -- Runtime Function: fract __fractuhqhq (unsigned fract A) 1705 -- Runtime Function: long fract __fractuhqsq (unsigned fract A) 1706 -- Runtime Function: long long fract __fractuhqdq (unsigned fract A) 1707 -- Runtime Function: short accum __fractuhqha (unsigned fract A) 1708 -- Runtime Function: accum __fractuhqsa (unsigned fract A) 1709 -- Runtime Function: long accum __fractuhqda (unsigned fract A) 1710 -- Runtime Function: long long accum __fractuhqta (unsigned fract A) 1711 -- Runtime Function: unsigned short fract __fractuhquqq2 (unsigned 1712 fract A) 1713 -- Runtime Function: unsigned long fract __fractuhqusq2 (unsigned 1714 fract A) 1715 -- Runtime Function: unsigned long long fract __fractuhqudq2 (unsigned 1716 fract A) 1717 -- Runtime Function: unsigned short accum __fractuhquha (unsigned 1718 fract A) 1719 -- Runtime Function: unsigned accum __fractuhqusa (unsigned fract A) 1720 -- Runtime Function: unsigned long accum __fractuhquda (unsigned fract 1721 A) 1722 -- Runtime Function: unsigned long long accum __fractuhquta (unsigned 1723 fract A) 1724 -- Runtime Function: signed char __fractuhqqi (unsigned fract A) 1725 -- Runtime Function: short __fractuhqhi (unsigned fract A) 1726 -- Runtime Function: int __fractuhqsi (unsigned fract A) 1727 -- Runtime Function: long __fractuhqdi (unsigned fract A) 1728 -- Runtime Function: long long __fractuhqti (unsigned fract A) 1729 -- Runtime Function: float __fractuhqsf (unsigned fract A) 1730 -- Runtime Function: double __fractuhqdf (unsigned fract A) 1731 -- Runtime Function: short fract __fractusqqq (unsigned long fract A) 1732 -- Runtime Function: fract __fractusqhq (unsigned long fract A) 1733 -- Runtime Function: long fract __fractusqsq (unsigned long fract A) 1734 -- Runtime Function: long long fract __fractusqdq (unsigned long fract 1735 A) 1736 -- Runtime Function: short accum __fractusqha (unsigned long fract A) 1737 -- Runtime Function: accum __fractusqsa (unsigned long fract A) 1738 -- Runtime Function: long accum __fractusqda (unsigned long fract A) 1739 -- Runtime Function: long long accum __fractusqta (unsigned long fract 1740 A) 1741 -- Runtime Function: unsigned short fract __fractusquqq2 (unsigned 1742 long fract A) 1743 -- Runtime Function: unsigned fract __fractusquhq2 (unsigned long 1744 fract A) 1745 -- Runtime Function: unsigned long long fract __fractusqudq2 (unsigned 1746 long fract A) 1747 -- Runtime Function: unsigned short accum __fractusquha (unsigned long 1748 fract A) 1749 -- Runtime Function: unsigned accum __fractusqusa (unsigned long fract 1750 A) 1751 -- Runtime Function: unsigned long accum __fractusquda (unsigned long 1752 fract A) 1753 -- Runtime Function: unsigned long long accum __fractusquta (unsigned 1754 long fract A) 1755 -- Runtime Function: signed char __fractusqqi (unsigned long fract A) 1756 -- Runtime Function: short __fractusqhi (unsigned long fract A) 1757 -- Runtime Function: int __fractusqsi (unsigned long fract A) 1758 -- Runtime Function: long __fractusqdi (unsigned long fract A) 1759 -- Runtime Function: long long __fractusqti (unsigned long fract A) 1760 -- Runtime Function: float __fractusqsf (unsigned long fract A) 1761 -- Runtime Function: double __fractusqdf (unsigned long fract A) 1762 -- Runtime Function: short fract __fractudqqq (unsigned long long 1763 fract A) 1764 -- Runtime Function: fract __fractudqhq (unsigned long long fract A) 1765 -- Runtime Function: long fract __fractudqsq (unsigned long long fract 1766 A) 1767 -- Runtime Function: long long fract __fractudqdq (unsigned long long 1768 fract A) 1769 -- Runtime Function: short accum __fractudqha (unsigned long long 1770 fract A) 1771 -- Runtime Function: accum __fractudqsa (unsigned long long fract A) 1772 -- Runtime Function: long accum __fractudqda (unsigned long long fract 1773 A) 1774 -- Runtime Function: long long accum __fractudqta (unsigned long long 1775 fract A) 1776 -- Runtime Function: unsigned short fract __fractudquqq2 (unsigned 1777 long long fract A) 1778 -- Runtime Function: unsigned fract __fractudquhq2 (unsigned long long 1779 fract A) 1780 -- Runtime Function: unsigned long fract __fractudqusq2 (unsigned long 1781 long fract A) 1782 -- Runtime Function: unsigned short accum __fractudquha (unsigned long 1783 long fract A) 1784 -- Runtime Function: unsigned accum __fractudqusa (unsigned long long 1785 fract A) 1786 -- Runtime Function: unsigned long accum __fractudquda (unsigned long 1787 long fract A) 1788 -- Runtime Function: unsigned long long accum __fractudquta (unsigned 1789 long long fract A) 1790 -- Runtime Function: signed char __fractudqqi (unsigned long long 1791 fract A) 1792 -- Runtime Function: short __fractudqhi (unsigned long long fract A) 1793 -- Runtime Function: int __fractudqsi (unsigned long long fract A) 1794 -- Runtime Function: long __fractudqdi (unsigned long long fract A) 1795 -- Runtime Function: long long __fractudqti (unsigned long long fract 1796 A) 1797 -- Runtime Function: float __fractudqsf (unsigned long long fract A) 1798 -- Runtime Function: double __fractudqdf (unsigned long long fract A) 1799 -- Runtime Function: short fract __fractuhaqq (unsigned short accum A) 1800 -- Runtime Function: fract __fractuhahq (unsigned short accum A) 1801 -- Runtime Function: long fract __fractuhasq (unsigned short accum A) 1802 -- Runtime Function: long long fract __fractuhadq (unsigned short 1803 accum A) 1804 -- Runtime Function: short accum __fractuhaha (unsigned short accum A) 1805 -- Runtime Function: accum __fractuhasa (unsigned short accum A) 1806 -- Runtime Function: long accum __fractuhada (unsigned short accum A) 1807 -- Runtime Function: long long accum __fractuhata (unsigned short 1808 accum A) 1809 -- Runtime Function: unsigned short fract __fractuhauqq (unsigned 1810 short accum A) 1811 -- Runtime Function: unsigned fract __fractuhauhq (unsigned short 1812 accum A) 1813 -- Runtime Function: unsigned long fract __fractuhausq (unsigned short 1814 accum A) 1815 -- Runtime Function: unsigned long long fract __fractuhaudq (unsigned 1816 short accum A) 1817 -- Runtime Function: unsigned accum __fractuhausa2 (unsigned short 1818 accum A) 1819 -- Runtime Function: unsigned long accum __fractuhauda2 (unsigned 1820 short accum A) 1821 -- Runtime Function: unsigned long long accum __fractuhauta2 (unsigned 1822 short accum A) 1823 -- Runtime Function: signed char __fractuhaqi (unsigned short accum A) 1824 -- Runtime Function: short __fractuhahi (unsigned short accum A) 1825 -- Runtime Function: int __fractuhasi (unsigned short accum A) 1826 -- Runtime Function: long __fractuhadi (unsigned short accum A) 1827 -- Runtime Function: long long __fractuhati (unsigned short accum A) 1828 -- Runtime Function: float __fractuhasf (unsigned short accum A) 1829 -- Runtime Function: double __fractuhadf (unsigned short accum A) 1830 -- Runtime Function: short fract __fractusaqq (unsigned accum A) 1831 -- Runtime Function: fract __fractusahq (unsigned accum A) 1832 -- Runtime Function: long fract __fractusasq (unsigned accum A) 1833 -- Runtime Function: long long fract __fractusadq (unsigned accum A) 1834 -- Runtime Function: short accum __fractusaha (unsigned accum A) 1835 -- Runtime Function: accum __fractusasa (unsigned accum A) 1836 -- Runtime Function: long accum __fractusada (unsigned accum A) 1837 -- Runtime Function: long long accum __fractusata (unsigned accum A) 1838 -- Runtime Function: unsigned short fract __fractusauqq (unsigned 1839 accum A) 1840 -- Runtime Function: unsigned fract __fractusauhq (unsigned accum A) 1841 -- Runtime Function: unsigned long fract __fractusausq (unsigned accum 1842 A) 1843 -- Runtime Function: unsigned long long fract __fractusaudq (unsigned 1844 accum A) 1845 -- Runtime Function: unsigned short accum __fractusauha2 (unsigned 1846 accum A) 1847 -- Runtime Function: unsigned long accum __fractusauda2 (unsigned 1848 accum A) 1849 -- Runtime Function: unsigned long long accum __fractusauta2 (unsigned 1850 accum A) 1851 -- Runtime Function: signed char __fractusaqi (unsigned accum A) 1852 -- Runtime Function: short __fractusahi (unsigned accum A) 1853 -- Runtime Function: int __fractusasi (unsigned accum A) 1854 -- Runtime Function: long __fractusadi (unsigned accum A) 1855 -- Runtime Function: long long __fractusati (unsigned accum A) 1856 -- Runtime Function: float __fractusasf (unsigned accum A) 1857 -- Runtime Function: double __fractusadf (unsigned accum A) 1858 -- Runtime Function: short fract __fractudaqq (unsigned long accum A) 1859 -- Runtime Function: fract __fractudahq (unsigned long accum A) 1860 -- Runtime Function: long fract __fractudasq (unsigned long accum A) 1861 -- Runtime Function: long long fract __fractudadq (unsigned long accum 1862 A) 1863 -- Runtime Function: short accum __fractudaha (unsigned long accum A) 1864 -- Runtime Function: accum __fractudasa (unsigned long accum A) 1865 -- Runtime Function: long accum __fractudada (unsigned long accum A) 1866 -- Runtime Function: long long accum __fractudata (unsigned long accum 1867 A) 1868 -- Runtime Function: unsigned short fract __fractudauqq (unsigned long 1869 accum A) 1870 -- Runtime Function: unsigned fract __fractudauhq (unsigned long accum 1871 A) 1872 -- Runtime Function: unsigned long fract __fractudausq (unsigned long 1873 accum A) 1874 -- Runtime Function: unsigned long long fract __fractudaudq (unsigned 1875 long accum A) 1876 -- Runtime Function: unsigned short accum __fractudauha2 (unsigned 1877 long accum A) 1878 -- Runtime Function: unsigned accum __fractudausa2 (unsigned long 1879 accum A) 1880 -- Runtime Function: unsigned long long accum __fractudauta2 (unsigned 1881 long accum A) 1882 -- Runtime Function: signed char __fractudaqi (unsigned long accum A) 1883 -- Runtime Function: short __fractudahi (unsigned long accum A) 1884 -- Runtime Function: int __fractudasi (unsigned long accum A) 1885 -- Runtime Function: long __fractudadi (unsigned long accum A) 1886 -- Runtime Function: long long __fractudati (unsigned long accum A) 1887 -- Runtime Function: float __fractudasf (unsigned long accum A) 1888 -- Runtime Function: double __fractudadf (unsigned long accum A) 1889 -- Runtime Function: short fract __fractutaqq (unsigned long long 1890 accum A) 1891 -- Runtime Function: fract __fractutahq (unsigned long long accum A) 1892 -- Runtime Function: long fract __fractutasq (unsigned long long accum 1893 A) 1894 -- Runtime Function: long long fract __fractutadq (unsigned long long 1895 accum A) 1896 -- Runtime Function: short accum __fractutaha (unsigned long long 1897 accum A) 1898 -- Runtime Function: accum __fractutasa (unsigned long long accum A) 1899 -- Runtime Function: long accum __fractutada (unsigned long long accum 1900 A) 1901 -- Runtime Function: long long accum __fractutata (unsigned long long 1902 accum A) 1903 -- Runtime Function: unsigned short fract __fractutauqq (unsigned long 1904 long accum A) 1905 -- Runtime Function: unsigned fract __fractutauhq (unsigned long long 1906 accum A) 1907 -- Runtime Function: unsigned long fract __fractutausq (unsigned long 1908 long accum A) 1909 -- Runtime Function: unsigned long long fract __fractutaudq (unsigned 1910 long long accum A) 1911 -- Runtime Function: unsigned short accum __fractutauha2 (unsigned 1912 long long accum A) 1913 -- Runtime Function: unsigned accum __fractutausa2 (unsigned long long 1914 accum A) 1915 -- Runtime Function: unsigned long accum __fractutauda2 (unsigned long 1916 long accum A) 1917 -- Runtime Function: signed char __fractutaqi (unsigned long long 1918 accum A) 1919 -- Runtime Function: short __fractutahi (unsigned long long accum A) 1920 -- Runtime Function: int __fractutasi (unsigned long long accum A) 1921 -- Runtime Function: long __fractutadi (unsigned long long accum A) 1922 -- Runtime Function: long long __fractutati (unsigned long long accum 1923 A) 1924 -- Runtime Function: float __fractutasf (unsigned long long accum A) 1925 -- Runtime Function: double __fractutadf (unsigned long long accum A) 1926 -- Runtime Function: short fract __fractqiqq (signed char A) 1927 -- Runtime Function: fract __fractqihq (signed char A) 1928 -- Runtime Function: long fract __fractqisq (signed char A) 1929 -- Runtime Function: long long fract __fractqidq (signed char A) 1930 -- Runtime Function: short accum __fractqiha (signed char A) 1931 -- Runtime Function: accum __fractqisa (signed char A) 1932 -- Runtime Function: long accum __fractqida (signed char A) 1933 -- Runtime Function: long long accum __fractqita (signed char A) 1934 -- Runtime Function: unsigned short fract __fractqiuqq (signed char A) 1935 -- Runtime Function: unsigned fract __fractqiuhq (signed char A) 1936 -- Runtime Function: unsigned long fract __fractqiusq (signed char A) 1937 -- Runtime Function: unsigned long long fract __fractqiudq (signed 1938 char A) 1939 -- Runtime Function: unsigned short accum __fractqiuha (signed char A) 1940 -- Runtime Function: unsigned accum __fractqiusa (signed char A) 1941 -- Runtime Function: unsigned long accum __fractqiuda (signed char A) 1942 -- Runtime Function: unsigned long long accum __fractqiuta (signed 1943 char A) 1944 -- Runtime Function: short fract __fracthiqq (short A) 1945 -- Runtime Function: fract __fracthihq (short A) 1946 -- Runtime Function: long fract __fracthisq (short A) 1947 -- Runtime Function: long long fract __fracthidq (short A) 1948 -- Runtime Function: short accum __fracthiha (short A) 1949 -- Runtime Function: accum __fracthisa (short A) 1950 -- Runtime Function: long accum __fracthida (short A) 1951 -- Runtime Function: long long accum __fracthita (short A) 1952 -- Runtime Function: unsigned short fract __fracthiuqq (short A) 1953 -- Runtime Function: unsigned fract __fracthiuhq (short A) 1954 -- Runtime Function: unsigned long fract __fracthiusq (short A) 1955 -- Runtime Function: unsigned long long fract __fracthiudq (short A) 1956 -- Runtime Function: unsigned short accum __fracthiuha (short A) 1957 -- Runtime Function: unsigned accum __fracthiusa (short A) 1958 -- Runtime Function: unsigned long accum __fracthiuda (short A) 1959 -- Runtime Function: unsigned long long accum __fracthiuta (short A) 1960 -- Runtime Function: short fract __fractsiqq (int A) 1961 -- Runtime Function: fract __fractsihq (int A) 1962 -- Runtime Function: long fract __fractsisq (int A) 1963 -- Runtime Function: long long fract __fractsidq (int A) 1964 -- Runtime Function: short accum __fractsiha (int A) 1965 -- Runtime Function: accum __fractsisa (int A) 1966 -- Runtime Function: long accum __fractsida (int A) 1967 -- Runtime Function: long long accum __fractsita (int A) 1968 -- Runtime Function: unsigned short fract __fractsiuqq (int A) 1969 -- Runtime Function: unsigned fract __fractsiuhq (int A) 1970 -- Runtime Function: unsigned long fract __fractsiusq (int A) 1971 -- Runtime Function: unsigned long long fract __fractsiudq (int A) 1972 -- Runtime Function: unsigned short accum __fractsiuha (int A) 1973 -- Runtime Function: unsigned accum __fractsiusa (int A) 1974 -- Runtime Function: unsigned long accum __fractsiuda (int A) 1975 -- Runtime Function: unsigned long long accum __fractsiuta (int A) 1976 -- Runtime Function: short fract __fractdiqq (long A) 1977 -- Runtime Function: fract __fractdihq (long A) 1978 -- Runtime Function: long fract __fractdisq (long A) 1979 -- Runtime Function: long long fract __fractdidq (long A) 1980 -- Runtime Function: short accum __fractdiha (long A) 1981 -- Runtime Function: accum __fractdisa (long A) 1982 -- Runtime Function: long accum __fractdida (long A) 1983 -- Runtime Function: long long accum __fractdita (long A) 1984 -- Runtime Function: unsigned short fract __fractdiuqq (long A) 1985 -- Runtime Function: unsigned fract __fractdiuhq (long A) 1986 -- Runtime Function: unsigned long fract __fractdiusq (long A) 1987 -- Runtime Function: unsigned long long fract __fractdiudq (long A) 1988 -- Runtime Function: unsigned short accum __fractdiuha (long A) 1989 -- Runtime Function: unsigned accum __fractdiusa (long A) 1990 -- Runtime Function: unsigned long accum __fractdiuda (long A) 1991 -- Runtime Function: unsigned long long accum __fractdiuta (long A) 1992 -- Runtime Function: short fract __fracttiqq (long long A) 1993 -- Runtime Function: fract __fracttihq (long long A) 1994 -- Runtime Function: long fract __fracttisq (long long A) 1995 -- Runtime Function: long long fract __fracttidq (long long A) 1996 -- Runtime Function: short accum __fracttiha (long long A) 1997 -- Runtime Function: accum __fracttisa (long long A) 1998 -- Runtime Function: long accum __fracttida (long long A) 1999 -- Runtime Function: long long accum __fracttita (long long A) 2000 -- Runtime Function: unsigned short fract __fracttiuqq (long long A) 2001 -- Runtime Function: unsigned fract __fracttiuhq (long long A) 2002 -- Runtime Function: unsigned long fract __fracttiusq (long long A) 2003 -- Runtime Function: unsigned long long fract __fracttiudq (long long 2004 A) 2005 -- Runtime Function: unsigned short accum __fracttiuha (long long A) 2006 -- Runtime Function: unsigned accum __fracttiusa (long long A) 2007 -- Runtime Function: unsigned long accum __fracttiuda (long long A) 2008 -- Runtime Function: unsigned long long accum __fracttiuta (long long 2009 A) 2010 -- Runtime Function: short fract __fractsfqq (float A) 2011 -- Runtime Function: fract __fractsfhq (float A) 2012 -- Runtime Function: long fract __fractsfsq (float A) 2013 -- Runtime Function: long long fract __fractsfdq (float A) 2014 -- Runtime Function: short accum __fractsfha (float A) 2015 -- Runtime Function: accum __fractsfsa (float A) 2016 -- Runtime Function: long accum __fractsfda (float A) 2017 -- Runtime Function: long long accum __fractsfta (float A) 2018 -- Runtime Function: unsigned short fract __fractsfuqq (float A) 2019 -- Runtime Function: unsigned fract __fractsfuhq (float A) 2020 -- Runtime Function: unsigned long fract __fractsfusq (float A) 2021 -- Runtime Function: unsigned long long fract __fractsfudq (float A) 2022 -- Runtime Function: unsigned short accum __fractsfuha (float A) 2023 -- Runtime Function: unsigned accum __fractsfusa (float A) 2024 -- Runtime Function: unsigned long accum __fractsfuda (float A) 2025 -- Runtime Function: unsigned long long accum __fractsfuta (float A) 2026 -- Runtime Function: short fract __fractdfqq (double A) 2027 -- Runtime Function: fract __fractdfhq (double A) 2028 -- Runtime Function: long fract __fractdfsq (double A) 2029 -- Runtime Function: long long fract __fractdfdq (double A) 2030 -- Runtime Function: short accum __fractdfha (double A) 2031 -- Runtime Function: accum __fractdfsa (double A) 2032 -- Runtime Function: long accum __fractdfda (double A) 2033 -- Runtime Function: long long accum __fractdfta (double A) 2034 -- Runtime Function: unsigned short fract __fractdfuqq (double A) 2035 -- Runtime Function: unsigned fract __fractdfuhq (double A) 2036 -- Runtime Function: unsigned long fract __fractdfusq (double A) 2037 -- Runtime Function: unsigned long long fract __fractdfudq (double A) 2038 -- Runtime Function: unsigned short accum __fractdfuha (double A) 2039 -- Runtime Function: unsigned accum __fractdfusa (double A) 2040 -- Runtime Function: unsigned long accum __fractdfuda (double A) 2041 -- Runtime Function: unsigned long long accum __fractdfuta (double A) 2042 These functions convert from fractional and signed non-fractionals 2043 to fractionals and signed non-fractionals, without saturation. 2044 2045 -- Runtime Function: fract __satfractqqhq2 (short fract A) 2046 -- Runtime Function: long fract __satfractqqsq2 (short fract A) 2047 -- Runtime Function: long long fract __satfractqqdq2 (short fract A) 2048 -- Runtime Function: short accum __satfractqqha (short fract A) 2049 -- Runtime Function: accum __satfractqqsa (short fract A) 2050 -- Runtime Function: long accum __satfractqqda (short fract A) 2051 -- Runtime Function: long long accum __satfractqqta (short fract A) 2052 -- Runtime Function: unsigned short fract __satfractqquqq (short fract 2053 A) 2054 -- Runtime Function: unsigned fract __satfractqquhq (short fract A) 2055 -- Runtime Function: unsigned long fract __satfractqqusq (short fract 2056 A) 2057 -- Runtime Function: unsigned long long fract __satfractqqudq (short 2058 fract A) 2059 -- Runtime Function: unsigned short accum __satfractqquha (short fract 2060 A) 2061 -- Runtime Function: unsigned accum __satfractqqusa (short fract A) 2062 -- Runtime Function: unsigned long accum __satfractqquda (short fract 2063 A) 2064 -- Runtime Function: unsigned long long accum __satfractqquta (short 2065 fract A) 2066 -- Runtime Function: short fract __satfracthqqq2 (fract A) 2067 -- Runtime Function: long fract __satfracthqsq2 (fract A) 2068 -- Runtime Function: long long fract __satfracthqdq2 (fract A) 2069 -- Runtime Function: short accum __satfracthqha (fract A) 2070 -- Runtime Function: accum __satfracthqsa (fract A) 2071 -- Runtime Function: long accum __satfracthqda (fract A) 2072 -- Runtime Function: long long accum __satfracthqta (fract A) 2073 -- Runtime Function: unsigned short fract __satfracthquqq (fract A) 2074 -- Runtime Function: unsigned fract __satfracthquhq (fract A) 2075 -- Runtime Function: unsigned long fract __satfracthqusq (fract A) 2076 -- Runtime Function: unsigned long long fract __satfracthqudq (fract A) 2077 -- Runtime Function: unsigned short accum __satfracthquha (fract A) 2078 -- Runtime Function: unsigned accum __satfracthqusa (fract A) 2079 -- Runtime Function: unsigned long accum __satfracthquda (fract A) 2080 -- Runtime Function: unsigned long long accum __satfracthquta (fract A) 2081 -- Runtime Function: short fract __satfractsqqq2 (long fract A) 2082 -- Runtime Function: fract __satfractsqhq2 (long fract A) 2083 -- Runtime Function: long long fract __satfractsqdq2 (long fract A) 2084 -- Runtime Function: short accum __satfractsqha (long fract A) 2085 -- Runtime Function: accum __satfractsqsa (long fract A) 2086 -- Runtime Function: long accum __satfractsqda (long fract A) 2087 -- Runtime Function: long long accum __satfractsqta (long fract A) 2088 -- Runtime Function: unsigned short fract __satfractsquqq (long fract 2089 A) 2090 -- Runtime Function: unsigned fract __satfractsquhq (long fract A) 2091 -- Runtime Function: unsigned long fract __satfractsqusq (long fract A) 2092 -- Runtime Function: unsigned long long fract __satfractsqudq (long 2093 fract A) 2094 -- Runtime Function: unsigned short accum __satfractsquha (long fract 2095 A) 2096 -- Runtime Function: unsigned accum __satfractsqusa (long fract A) 2097 -- Runtime Function: unsigned long accum __satfractsquda (long fract A) 2098 -- Runtime Function: unsigned long long accum __satfractsquta (long 2099 fract A) 2100 -- Runtime Function: short fract __satfractdqqq2 (long long fract A) 2101 -- Runtime Function: fract __satfractdqhq2 (long long fract A) 2102 -- Runtime Function: long fract __satfractdqsq2 (long long fract A) 2103 -- Runtime Function: short accum __satfractdqha (long long fract A) 2104 -- Runtime Function: accum __satfractdqsa (long long fract A) 2105 -- Runtime Function: long accum __satfractdqda (long long fract A) 2106 -- Runtime Function: long long accum __satfractdqta (long long fract A) 2107 -- Runtime Function: unsigned short fract __satfractdquqq (long long 2108 fract A) 2109 -- Runtime Function: unsigned fract __satfractdquhq (long long fract A) 2110 -- Runtime Function: unsigned long fract __satfractdqusq (long long 2111 fract A) 2112 -- Runtime Function: unsigned long long fract __satfractdqudq (long 2113 long fract A) 2114 -- Runtime Function: unsigned short accum __satfractdquha (long long 2115 fract A) 2116 -- Runtime Function: unsigned accum __satfractdqusa (long long fract A) 2117 -- Runtime Function: unsigned long accum __satfractdquda (long long 2118 fract A) 2119 -- Runtime Function: unsigned long long accum __satfractdquta (long 2120 long fract A) 2121 -- Runtime Function: short fract __satfracthaqq (short accum A) 2122 -- Runtime Function: fract __satfracthahq (short accum A) 2123 -- Runtime Function: long fract __satfracthasq (short accum A) 2124 -- Runtime Function: long long fract __satfracthadq (short accum A) 2125 -- Runtime Function: accum __satfracthasa2 (short accum A) 2126 -- Runtime Function: long accum __satfracthada2 (short accum A) 2127 -- Runtime Function: long long accum __satfracthata2 (short accum A) 2128 -- Runtime Function: unsigned short fract __satfracthauqq (short accum 2129 A) 2130 -- Runtime Function: unsigned fract __satfracthauhq (short accum A) 2131 -- Runtime Function: unsigned long fract __satfracthausq (short accum 2132 A) 2133 -- Runtime Function: unsigned long long fract __satfracthaudq (short 2134 accum A) 2135 -- Runtime Function: unsigned short accum __satfracthauha (short accum 2136 A) 2137 -- Runtime Function: unsigned accum __satfracthausa (short accum A) 2138 -- Runtime Function: unsigned long accum __satfracthauda (short accum 2139 A) 2140 -- Runtime Function: unsigned long long accum __satfracthauta (short 2141 accum A) 2142 -- Runtime Function: short fract __satfractsaqq (accum A) 2143 -- Runtime Function: fract __satfractsahq (accum A) 2144 -- Runtime Function: long fract __satfractsasq (accum A) 2145 -- Runtime Function: long long fract __satfractsadq (accum A) 2146 -- Runtime Function: short accum __satfractsaha2 (accum A) 2147 -- Runtime Function: long accum __satfractsada2 (accum A) 2148 -- Runtime Function: long long accum __satfractsata2 (accum A) 2149 -- Runtime Function: unsigned short fract __satfractsauqq (accum A) 2150 -- Runtime Function: unsigned fract __satfractsauhq (accum A) 2151 -- Runtime Function: unsigned long fract __satfractsausq (accum A) 2152 -- Runtime Function: unsigned long long fract __satfractsaudq (accum A) 2153 -- Runtime Function: unsigned short accum __satfractsauha (accum A) 2154 -- Runtime Function: unsigned accum __satfractsausa (accum A) 2155 -- Runtime Function: unsigned long accum __satfractsauda (accum A) 2156 -- Runtime Function: unsigned long long accum __satfractsauta (accum A) 2157 -- Runtime Function: short fract __satfractdaqq (long accum A) 2158 -- Runtime Function: fract __satfractdahq (long accum A) 2159 -- Runtime Function: long fract __satfractdasq (long accum A) 2160 -- Runtime Function: long long fract __satfractdadq (long accum A) 2161 -- Runtime Function: short accum __satfractdaha2 (long accum A) 2162 -- Runtime Function: accum __satfractdasa2 (long accum A) 2163 -- Runtime Function: long long accum __satfractdata2 (long accum A) 2164 -- Runtime Function: unsigned short fract __satfractdauqq (long accum 2165 A) 2166 -- Runtime Function: unsigned fract __satfractdauhq (long accum A) 2167 -- Runtime Function: unsigned long fract __satfractdausq (long accum A) 2168 -- Runtime Function: unsigned long long fract __satfractdaudq (long 2169 accum A) 2170 -- Runtime Function: unsigned short accum __satfractdauha (long accum 2171 A) 2172 -- Runtime Function: unsigned accum __satfractdausa (long accum A) 2173 -- Runtime Function: unsigned long accum __satfractdauda (long accum A) 2174 -- Runtime Function: unsigned long long accum __satfractdauta (long 2175 accum A) 2176 -- Runtime Function: short fract __satfracttaqq (long long accum A) 2177 -- Runtime Function: fract __satfracttahq (long long accum A) 2178 -- Runtime Function: long fract __satfracttasq (long long accum A) 2179 -- Runtime Function: long long fract __satfracttadq (long long accum A) 2180 -- Runtime Function: short accum __satfracttaha2 (long long accum A) 2181 -- Runtime Function: accum __satfracttasa2 (long long accum A) 2182 -- Runtime Function: long accum __satfracttada2 (long long accum A) 2183 -- Runtime Function: unsigned short fract __satfracttauqq (long long 2184 accum A) 2185 -- Runtime Function: unsigned fract __satfracttauhq (long long accum A) 2186 -- Runtime Function: unsigned long fract __satfracttausq (long long 2187 accum A) 2188 -- Runtime Function: unsigned long long fract __satfracttaudq (long 2189 long accum A) 2190 -- Runtime Function: unsigned short accum __satfracttauha (long long 2191 accum A) 2192 -- Runtime Function: unsigned accum __satfracttausa (long long accum A) 2193 -- Runtime Function: unsigned long accum __satfracttauda (long long 2194 accum A) 2195 -- Runtime Function: unsigned long long accum __satfracttauta (long 2196 long accum A) 2197 -- Runtime Function: short fract __satfractuqqqq (unsigned short fract 2198 A) 2199 -- Runtime Function: fract __satfractuqqhq (unsigned short fract A) 2200 -- Runtime Function: long fract __satfractuqqsq (unsigned short fract 2201 A) 2202 -- Runtime Function: long long fract __satfractuqqdq (unsigned short 2203 fract A) 2204 -- Runtime Function: short accum __satfractuqqha (unsigned short fract 2205 A) 2206 -- Runtime Function: accum __satfractuqqsa (unsigned short fract A) 2207 -- Runtime Function: long accum __satfractuqqda (unsigned short fract 2208 A) 2209 -- Runtime Function: long long accum __satfractuqqta (unsigned short 2210 fract A) 2211 -- Runtime Function: unsigned fract __satfractuqquhq2 (unsigned short 2212 fract A) 2213 -- Runtime Function: unsigned long fract __satfractuqqusq2 (unsigned 2214 short fract A) 2215 -- Runtime Function: unsigned long long fract __satfractuqqudq2 2216 (unsigned short fract A) 2217 -- Runtime Function: unsigned short accum __satfractuqquha (unsigned 2218 short fract A) 2219 -- Runtime Function: unsigned accum __satfractuqqusa (unsigned short 2220 fract A) 2221 -- Runtime Function: unsigned long accum __satfractuqquda (unsigned 2222 short fract A) 2223 -- Runtime Function: unsigned long long accum __satfractuqquta 2224 (unsigned short fract A) 2225 -- Runtime Function: short fract __satfractuhqqq (unsigned fract A) 2226 -- Runtime Function: fract __satfractuhqhq (unsigned fract A) 2227 -- Runtime Function: long fract __satfractuhqsq (unsigned fract A) 2228 -- Runtime Function: long long fract __satfractuhqdq (unsigned fract A) 2229 -- Runtime Function: short accum __satfractuhqha (unsigned fract A) 2230 -- Runtime Function: accum __satfractuhqsa (unsigned fract A) 2231 -- Runtime Function: long accum __satfractuhqda (unsigned fract A) 2232 -- Runtime Function: long long accum __satfractuhqta (unsigned fract A) 2233 -- Runtime Function: unsigned short fract __satfractuhquqq2 (unsigned 2234 fract A) 2235 -- Runtime Function: unsigned long fract __satfractuhqusq2 (unsigned 2236 fract A) 2237 -- Runtime Function: unsigned long long fract __satfractuhqudq2 2238 (unsigned fract A) 2239 -- Runtime Function: unsigned short accum __satfractuhquha (unsigned 2240 fract A) 2241 -- Runtime Function: unsigned accum __satfractuhqusa (unsigned fract A) 2242 -- Runtime Function: unsigned long accum __satfractuhquda (unsigned 2243 fract A) 2244 -- Runtime Function: unsigned long long accum __satfractuhquta 2245 (unsigned fract A) 2246 -- Runtime Function: short fract __satfractusqqq (unsigned long fract 2247 A) 2248 -- Runtime Function: fract __satfractusqhq (unsigned long fract A) 2249 -- Runtime Function: long fract __satfractusqsq (unsigned long fract A) 2250 -- Runtime Function: long long fract __satfractusqdq (unsigned long 2251 fract A) 2252 -- Runtime Function: short accum __satfractusqha (unsigned long fract 2253 A) 2254 -- Runtime Function: accum __satfractusqsa (unsigned long fract A) 2255 -- Runtime Function: long accum __satfractusqda (unsigned long fract A) 2256 -- Runtime Function: long long accum __satfractusqta (unsigned long 2257 fract A) 2258 -- Runtime Function: unsigned short fract __satfractusquqq2 (unsigned 2259 long fract A) 2260 -- Runtime Function: unsigned fract __satfractusquhq2 (unsigned long 2261 fract A) 2262 -- Runtime Function: unsigned long long fract __satfractusqudq2 2263 (unsigned long fract A) 2264 -- Runtime Function: unsigned short accum __satfractusquha (unsigned 2265 long fract A) 2266 -- Runtime Function: unsigned accum __satfractusqusa (unsigned long 2267 fract A) 2268 -- Runtime Function: unsigned long accum __satfractusquda (unsigned 2269 long fract A) 2270 -- Runtime Function: unsigned long long accum __satfractusquta 2271 (unsigned long fract A) 2272 -- Runtime Function: short fract __satfractudqqq (unsigned long long 2273 fract A) 2274 -- Runtime Function: fract __satfractudqhq (unsigned long long fract A) 2275 -- Runtime Function: long fract __satfractudqsq (unsigned long long 2276 fract A) 2277 -- Runtime Function: long long fract __satfractudqdq (unsigned long 2278 long fract A) 2279 -- Runtime Function: short accum __satfractudqha (unsigned long long 2280 fract A) 2281 -- Runtime Function: accum __satfractudqsa (unsigned long long fract A) 2282 -- Runtime Function: long accum __satfractudqda (unsigned long long 2283 fract A) 2284 -- Runtime Function: long long accum __satfractudqta (unsigned long 2285 long fract A) 2286 -- Runtime Function: unsigned short fract __satfractudquqq2 (unsigned 2287 long long fract A) 2288 -- Runtime Function: unsigned fract __satfractudquhq2 (unsigned long 2289 long fract A) 2290 -- Runtime Function: unsigned long fract __satfractudqusq2 (unsigned 2291 long long fract A) 2292 -- Runtime Function: unsigned short accum __satfractudquha (unsigned 2293 long long fract A) 2294 -- Runtime Function: unsigned accum __satfractudqusa (unsigned long 2295 long fract A) 2296 -- Runtime Function: unsigned long accum __satfractudquda (unsigned 2297 long long fract A) 2298 -- Runtime Function: unsigned long long accum __satfractudquta 2299 (unsigned long long fract A) 2300 -- Runtime Function: short fract __satfractuhaqq (unsigned short accum 2301 A) 2302 -- Runtime Function: fract __satfractuhahq (unsigned short accum A) 2303 -- Runtime Function: long fract __satfractuhasq (unsigned short accum 2304 A) 2305 -- Runtime Function: long long fract __satfractuhadq (unsigned short 2306 accum A) 2307 -- Runtime Function: short accum __satfractuhaha (unsigned short accum 2308 A) 2309 -- Runtime Function: accum __satfractuhasa (unsigned short accum A) 2310 -- Runtime Function: long accum __satfractuhada (unsigned short accum 2311 A) 2312 -- Runtime Function: long long accum __satfractuhata (unsigned short 2313 accum A) 2314 -- Runtime Function: unsigned short fract __satfractuhauqq (unsigned 2315 short accum A) 2316 -- Runtime Function: unsigned fract __satfractuhauhq (unsigned short 2317 accum A) 2318 -- Runtime Function: unsigned long fract __satfractuhausq (unsigned 2319 short accum A) 2320 -- Runtime Function: unsigned long long fract __satfractuhaudq 2321 (unsigned short accum A) 2322 -- Runtime Function: unsigned accum __satfractuhausa2 (unsigned short 2323 accum A) 2324 -- Runtime Function: unsigned long accum __satfractuhauda2 (unsigned 2325 short accum A) 2326 -- Runtime Function: unsigned long long accum __satfractuhauta2 2327 (unsigned short accum A) 2328 -- Runtime Function: short fract __satfractusaqq (unsigned accum A) 2329 -- Runtime Function: fract __satfractusahq (unsigned accum A) 2330 -- Runtime Function: long fract __satfractusasq (unsigned accum A) 2331 -- Runtime Function: long long fract __satfractusadq (unsigned accum A) 2332 -- Runtime Function: short accum __satfractusaha (unsigned accum A) 2333 -- Runtime Function: accum __satfractusasa (unsigned accum A) 2334 -- Runtime Function: long accum __satfractusada (unsigned accum A) 2335 -- Runtime Function: long long accum __satfractusata (unsigned accum A) 2336 -- Runtime Function: unsigned short fract __satfractusauqq (unsigned 2337 accum A) 2338 -- Runtime Function: unsigned fract __satfractusauhq (unsigned accum A) 2339 -- Runtime Function: unsigned long fract __satfractusausq (unsigned 2340 accum A) 2341 -- Runtime Function: unsigned long long fract __satfractusaudq 2342 (unsigned accum A) 2343 -- Runtime Function: unsigned short accum __satfractusauha2 (unsigned 2344 accum A) 2345 -- Runtime Function: unsigned long accum __satfractusauda2 (unsigned 2346 accum A) 2347 -- Runtime Function: unsigned long long accum __satfractusauta2 2348 (unsigned accum A) 2349 -- Runtime Function: short fract __satfractudaqq (unsigned long accum 2350 A) 2351 -- Runtime Function: fract __satfractudahq (unsigned long accum A) 2352 -- Runtime Function: long fract __satfractudasq (unsigned long accum A) 2353 -- Runtime Function: long long fract __satfractudadq (unsigned long 2354 accum A) 2355 -- Runtime Function: short accum __satfractudaha (unsigned long accum 2356 A) 2357 -- Runtime Function: accum __satfractudasa (unsigned long accum A) 2358 -- Runtime Function: long accum __satfractudada (unsigned long accum A) 2359 -- Runtime Function: long long accum __satfractudata (unsigned long 2360 accum A) 2361 -- Runtime Function: unsigned short fract __satfractudauqq (unsigned 2362 long accum A) 2363 -- Runtime Function: unsigned fract __satfractudauhq (unsigned long 2364 accum A) 2365 -- Runtime Function: unsigned long fract __satfractudausq (unsigned 2366 long accum A) 2367 -- Runtime Function: unsigned long long fract __satfractudaudq 2368 (unsigned long accum A) 2369 -- Runtime Function: unsigned short accum __satfractudauha2 (unsigned 2370 long accum A) 2371 -- Runtime Function: unsigned accum __satfractudausa2 (unsigned long 2372 accum A) 2373 -- Runtime Function: unsigned long long accum __satfractudauta2 2374 (unsigned long accum A) 2375 -- Runtime Function: short fract __satfractutaqq (unsigned long long 2376 accum A) 2377 -- Runtime Function: fract __satfractutahq (unsigned long long accum A) 2378 -- Runtime Function: long fract __satfractutasq (unsigned long long 2379 accum A) 2380 -- Runtime Function: long long fract __satfractutadq (unsigned long 2381 long accum A) 2382 -- Runtime Function: short accum __satfractutaha (unsigned long long 2383 accum A) 2384 -- Runtime Function: accum __satfractutasa (unsigned long long accum A) 2385 -- Runtime Function: long accum __satfractutada (unsigned long long 2386 accum A) 2387 -- Runtime Function: long long accum __satfractutata (unsigned long 2388 long accum A) 2389 -- Runtime Function: unsigned short fract __satfractutauqq (unsigned 2390 long long accum A) 2391 -- Runtime Function: unsigned fract __satfractutauhq (unsigned long 2392 long accum A) 2393 -- Runtime Function: unsigned long fract __satfractutausq (unsigned 2394 long long accum A) 2395 -- Runtime Function: unsigned long long fract __satfractutaudq 2396 (unsigned long long accum A) 2397 -- Runtime Function: unsigned short accum __satfractutauha2 (unsigned 2398 long long accum A) 2399 -- Runtime Function: unsigned accum __satfractutausa2 (unsigned long 2400 long accum A) 2401 -- Runtime Function: unsigned long accum __satfractutauda2 (unsigned 2402 long long accum A) 2403 -- Runtime Function: short fract __satfractqiqq (signed char A) 2404 -- Runtime Function: fract __satfractqihq (signed char A) 2405 -- Runtime Function: long fract __satfractqisq (signed char A) 2406 -- Runtime Function: long long fract __satfractqidq (signed char A) 2407 -- Runtime Function: short accum __satfractqiha (signed char A) 2408 -- Runtime Function: accum __satfractqisa (signed char A) 2409 -- Runtime Function: long accum __satfractqida (signed char A) 2410 -- Runtime Function: long long accum __satfractqita (signed char A) 2411 -- Runtime Function: unsigned short fract __satfractqiuqq (signed char 2412 A) 2413 -- Runtime Function: unsigned fract __satfractqiuhq (signed char A) 2414 -- Runtime Function: unsigned long fract __satfractqiusq (signed char 2415 A) 2416 -- Runtime Function: unsigned long long fract __satfractqiudq (signed 2417 char A) 2418 -- Runtime Function: unsigned short accum __satfractqiuha (signed char 2419 A) 2420 -- Runtime Function: unsigned accum __satfractqiusa (signed char A) 2421 -- Runtime Function: unsigned long accum __satfractqiuda (signed char 2422 A) 2423 -- Runtime Function: unsigned long long accum __satfractqiuta (signed 2424 char A) 2425 -- Runtime Function: short fract __satfracthiqq (short A) 2426 -- Runtime Function: fract __satfracthihq (short A) 2427 -- Runtime Function: long fract __satfracthisq (short A) 2428 -- Runtime Function: long long fract __satfracthidq (short A) 2429 -- Runtime Function: short accum __satfracthiha (short A) 2430 -- Runtime Function: accum __satfracthisa (short A) 2431 -- Runtime Function: long accum __satfracthida (short A) 2432 -- Runtime Function: long long accum __satfracthita (short A) 2433 -- Runtime Function: unsigned short fract __satfracthiuqq (short A) 2434 -- Runtime Function: unsigned fract __satfracthiuhq (short A) 2435 -- Runtime Function: unsigned long fract __satfracthiusq (short A) 2436 -- Runtime Function: unsigned long long fract __satfracthiudq (short A) 2437 -- Runtime Function: unsigned short accum __satfracthiuha (short A) 2438 -- Runtime Function: unsigned accum __satfracthiusa (short A) 2439 -- Runtime Function: unsigned long accum __satfracthiuda (short A) 2440 -- Runtime Function: unsigned long long accum __satfracthiuta (short A) 2441 -- Runtime Function: short fract __satfractsiqq (int A) 2442 -- Runtime Function: fract __satfractsihq (int A) 2443 -- Runtime Function: long fract __satfractsisq (int A) 2444 -- Runtime Function: long long fract __satfractsidq (int A) 2445 -- Runtime Function: short accum __satfractsiha (int A) 2446 -- Runtime Function: accum __satfractsisa (int A) 2447 -- Runtime Function: long accum __satfractsida (int A) 2448 -- Runtime Function: long long accum __satfractsita (int A) 2449 -- Runtime Function: unsigned short fract __satfractsiuqq (int A) 2450 -- Runtime Function: unsigned fract __satfractsiuhq (int A) 2451 -- Runtime Function: unsigned long fract __satfractsiusq (int A) 2452 -- Runtime Function: unsigned long long fract __satfractsiudq (int A) 2453 -- Runtime Function: unsigned short accum __satfractsiuha (int A) 2454 -- Runtime Function: unsigned accum __satfractsiusa (int A) 2455 -- Runtime Function: unsigned long accum __satfractsiuda (int A) 2456 -- Runtime Function: unsigned long long accum __satfractsiuta (int A) 2457 -- Runtime Function: short fract __satfractdiqq (long A) 2458 -- Runtime Function: fract __satfractdihq (long A) 2459 -- Runtime Function: long fract __satfractdisq (long A) 2460 -- Runtime Function: long long fract __satfractdidq (long A) 2461 -- Runtime Function: short accum __satfractdiha (long A) 2462 -- Runtime Function: accum __satfractdisa (long A) 2463 -- Runtime Function: long accum __satfractdida (long A) 2464 -- Runtime Function: long long accum __satfractdita (long A) 2465 -- Runtime Function: unsigned short fract __satfractdiuqq (long A) 2466 -- Runtime Function: unsigned fract __satfractdiuhq (long A) 2467 -- Runtime Function: unsigned long fract __satfractdiusq (long A) 2468 -- Runtime Function: unsigned long long fract __satfractdiudq (long A) 2469 -- Runtime Function: unsigned short accum __satfractdiuha (long A) 2470 -- Runtime Function: unsigned accum __satfractdiusa (long A) 2471 -- Runtime Function: unsigned long accum __satfractdiuda (long A) 2472 -- Runtime Function: unsigned long long accum __satfractdiuta (long A) 2473 -- Runtime Function: short fract __satfracttiqq (long long A) 2474 -- Runtime Function: fract __satfracttihq (long long A) 2475 -- Runtime Function: long fract __satfracttisq (long long A) 2476 -- Runtime Function: long long fract __satfracttidq (long long A) 2477 -- Runtime Function: short accum __satfracttiha (long long A) 2478 -- Runtime Function: accum __satfracttisa (long long A) 2479 -- Runtime Function: long accum __satfracttida (long long A) 2480 -- Runtime Function: long long accum __satfracttita (long long A) 2481 -- Runtime Function: unsigned short fract __satfracttiuqq (long long A) 2482 -- Runtime Function: unsigned fract __satfracttiuhq (long long A) 2483 -- Runtime Function: unsigned long fract __satfracttiusq (long long A) 2484 -- Runtime Function: unsigned long long fract __satfracttiudq (long 2485 long A) 2486 -- Runtime Function: unsigned short accum __satfracttiuha (long long A) 2487 -- Runtime Function: unsigned accum __satfracttiusa (long long A) 2488 -- Runtime Function: unsigned long accum __satfracttiuda (long long A) 2489 -- Runtime Function: unsigned long long accum __satfracttiuta (long 2490 long A) 2491 -- Runtime Function: short fract __satfractsfqq (float A) 2492 -- Runtime Function: fract __satfractsfhq (float A) 2493 -- Runtime Function: long fract __satfractsfsq (float A) 2494 -- Runtime Function: long long fract __satfractsfdq (float A) 2495 -- Runtime Function: short accum __satfractsfha (float A) 2496 -- Runtime Function: accum __satfractsfsa (float A) 2497 -- Runtime Function: long accum __satfractsfda (float A) 2498 -- Runtime Function: long long accum __satfractsfta (float A) 2499 -- Runtime Function: unsigned short fract __satfractsfuqq (float A) 2500 -- Runtime Function: unsigned fract __satfractsfuhq (float A) 2501 -- Runtime Function: unsigned long fract __satfractsfusq (float A) 2502 -- Runtime Function: unsigned long long fract __satfractsfudq (float A) 2503 -- Runtime Function: unsigned short accum __satfractsfuha (float A) 2504 -- Runtime Function: unsigned accum __satfractsfusa (float A) 2505 -- Runtime Function: unsigned long accum __satfractsfuda (float A) 2506 -- Runtime Function: unsigned long long accum __satfractsfuta (float A) 2507 -- Runtime Function: short fract __satfractdfqq (double A) 2508 -- Runtime Function: fract __satfractdfhq (double A) 2509 -- Runtime Function: long fract __satfractdfsq (double A) 2510 -- Runtime Function: long long fract __satfractdfdq (double A) 2511 -- Runtime Function: short accum __satfractdfha (double A) 2512 -- Runtime Function: accum __satfractdfsa (double A) 2513 -- Runtime Function: long accum __satfractdfda (double A) 2514 -- Runtime Function: long long accum __satfractdfta (double A) 2515 -- Runtime Function: unsigned short fract __satfractdfuqq (double A) 2516 -- Runtime Function: unsigned fract __satfractdfuhq (double A) 2517 -- Runtime Function: unsigned long fract __satfractdfusq (double A) 2518 -- Runtime Function: unsigned long long fract __satfractdfudq (double 2519 A) 2520 -- Runtime Function: unsigned short accum __satfractdfuha (double A) 2521 -- Runtime Function: unsigned accum __satfractdfusa (double A) 2522 -- Runtime Function: unsigned long accum __satfractdfuda (double A) 2523 -- Runtime Function: unsigned long long accum __satfractdfuta (double 2524 A) 2525 The functions convert from fractional and signed non-fractionals to 2526 fractionals, with saturation. 2527 2528 -- Runtime Function: unsigned char __fractunsqqqi (short fract A) 2529 -- Runtime Function: unsigned short __fractunsqqhi (short fract A) 2530 -- Runtime Function: unsigned int __fractunsqqsi (short fract A) 2531 -- Runtime Function: unsigned long __fractunsqqdi (short fract A) 2532 -- Runtime Function: unsigned long long __fractunsqqti (short fract A) 2533 -- Runtime Function: unsigned char __fractunshqqi (fract A) 2534 -- Runtime Function: unsigned short __fractunshqhi (fract A) 2535 -- Runtime Function: unsigned int __fractunshqsi (fract A) 2536 -- Runtime Function: unsigned long __fractunshqdi (fract A) 2537 -- Runtime Function: unsigned long long __fractunshqti (fract A) 2538 -- Runtime Function: unsigned char __fractunssqqi (long fract A) 2539 -- Runtime Function: unsigned short __fractunssqhi (long fract A) 2540 -- Runtime Function: unsigned int __fractunssqsi (long fract A) 2541 -- Runtime Function: unsigned long __fractunssqdi (long fract A) 2542 -- Runtime Function: unsigned long long __fractunssqti (long fract A) 2543 -- Runtime Function: unsigned char __fractunsdqqi (long long fract A) 2544 -- Runtime Function: unsigned short __fractunsdqhi (long long fract A) 2545 -- Runtime Function: unsigned int __fractunsdqsi (long long fract A) 2546 -- Runtime Function: unsigned long __fractunsdqdi (long long fract A) 2547 -- Runtime Function: unsigned long long __fractunsdqti (long long 2548 fract A) 2549 -- Runtime Function: unsigned char __fractunshaqi (short accum A) 2550 -- Runtime Function: unsigned short __fractunshahi (short accum A) 2551 -- Runtime Function: unsigned int __fractunshasi (short accum A) 2552 -- Runtime Function: unsigned long __fractunshadi (short accum A) 2553 -- Runtime Function: unsigned long long __fractunshati (short accum A) 2554 -- Runtime Function: unsigned char __fractunssaqi (accum A) 2555 -- Runtime Function: unsigned short __fractunssahi (accum A) 2556 -- Runtime Function: unsigned int __fractunssasi (accum A) 2557 -- Runtime Function: unsigned long __fractunssadi (accum A) 2558 -- Runtime Function: unsigned long long __fractunssati (accum A) 2559 -- Runtime Function: unsigned char __fractunsdaqi (long accum A) 2560 -- Runtime Function: unsigned short __fractunsdahi (long accum A) 2561 -- Runtime Function: unsigned int __fractunsdasi (long accum A) 2562 -- Runtime Function: unsigned long __fractunsdadi (long accum A) 2563 -- Runtime Function: unsigned long long __fractunsdati (long accum A) 2564 -- Runtime Function: unsigned char __fractunstaqi (long long accum A) 2565 -- Runtime Function: unsigned short __fractunstahi (long long accum A) 2566 -- Runtime Function: unsigned int __fractunstasi (long long accum A) 2567 -- Runtime Function: unsigned long __fractunstadi (long long accum A) 2568 -- Runtime Function: unsigned long long __fractunstati (long long 2569 accum A) 2570 -- Runtime Function: unsigned char __fractunsuqqqi (unsigned short 2571 fract A) 2572 -- Runtime Function: unsigned short __fractunsuqqhi (unsigned short 2573 fract A) 2574 -- Runtime Function: unsigned int __fractunsuqqsi (unsigned short 2575 fract A) 2576 -- Runtime Function: unsigned long __fractunsuqqdi (unsigned short 2577 fract A) 2578 -- Runtime Function: unsigned long long __fractunsuqqti (unsigned 2579 short fract A) 2580 -- Runtime Function: unsigned char __fractunsuhqqi (unsigned fract A) 2581 -- Runtime Function: unsigned short __fractunsuhqhi (unsigned fract A) 2582 -- Runtime Function: unsigned int __fractunsuhqsi (unsigned fract A) 2583 -- Runtime Function: unsigned long __fractunsuhqdi (unsigned fract A) 2584 -- Runtime Function: unsigned long long __fractunsuhqti (unsigned 2585 fract A) 2586 -- Runtime Function: unsigned char __fractunsusqqi (unsigned long 2587 fract A) 2588 -- Runtime Function: unsigned short __fractunsusqhi (unsigned long 2589 fract A) 2590 -- Runtime Function: unsigned int __fractunsusqsi (unsigned long fract 2591 A) 2592 -- Runtime Function: unsigned long __fractunsusqdi (unsigned long 2593 fract A) 2594 -- Runtime Function: unsigned long long __fractunsusqti (unsigned long 2595 fract A) 2596 -- Runtime Function: unsigned char __fractunsudqqi (unsigned long long 2597 fract A) 2598 -- Runtime Function: unsigned short __fractunsudqhi (unsigned long 2599 long fract A) 2600 -- Runtime Function: unsigned int __fractunsudqsi (unsigned long long 2601 fract A) 2602 -- Runtime Function: unsigned long __fractunsudqdi (unsigned long long 2603 fract A) 2604 -- Runtime Function: unsigned long long __fractunsudqti (unsigned long 2605 long fract A) 2606 -- Runtime Function: unsigned char __fractunsuhaqi (unsigned short 2607 accum A) 2608 -- Runtime Function: unsigned short __fractunsuhahi (unsigned short 2609 accum A) 2610 -- Runtime Function: unsigned int __fractunsuhasi (unsigned short 2611 accum A) 2612 -- Runtime Function: unsigned long __fractunsuhadi (unsigned short 2613 accum A) 2614 -- Runtime Function: unsigned long long __fractunsuhati (unsigned 2615 short accum A) 2616 -- Runtime Function: unsigned char __fractunsusaqi (unsigned accum A) 2617 -- Runtime Function: unsigned short __fractunsusahi (unsigned accum A) 2618 -- Runtime Function: unsigned int __fractunsusasi (unsigned accum A) 2619 -- Runtime Function: unsigned long __fractunsusadi (unsigned accum A) 2620 -- Runtime Function: unsigned long long __fractunsusati (unsigned 2621 accum A) 2622 -- Runtime Function: unsigned char __fractunsudaqi (unsigned long 2623 accum A) 2624 -- Runtime Function: unsigned short __fractunsudahi (unsigned long 2625 accum A) 2626 -- Runtime Function: unsigned int __fractunsudasi (unsigned long accum 2627 A) 2628 -- Runtime Function: unsigned long __fractunsudadi (unsigned long 2629 accum A) 2630 -- Runtime Function: unsigned long long __fractunsudati (unsigned long 2631 accum A) 2632 -- Runtime Function: unsigned char __fractunsutaqi (unsigned long long 2633 accum A) 2634 -- Runtime Function: unsigned short __fractunsutahi (unsigned long 2635 long accum A) 2636 -- Runtime Function: unsigned int __fractunsutasi (unsigned long long 2637 accum A) 2638 -- Runtime Function: unsigned long __fractunsutadi (unsigned long long 2639 accum A) 2640 -- Runtime Function: unsigned long long __fractunsutati (unsigned long 2641 long accum A) 2642 -- Runtime Function: short fract __fractunsqiqq (unsigned char A) 2643 -- Runtime Function: fract __fractunsqihq (unsigned char A) 2644 -- Runtime Function: long fract __fractunsqisq (unsigned char A) 2645 -- Runtime Function: long long fract __fractunsqidq (unsigned char A) 2646 -- Runtime Function: short accum __fractunsqiha (unsigned char A) 2647 -- Runtime Function: accum __fractunsqisa (unsigned char A) 2648 -- Runtime Function: long accum __fractunsqida (unsigned char A) 2649 -- Runtime Function: long long accum __fractunsqita (unsigned char A) 2650 -- Runtime Function: unsigned short fract __fractunsqiuqq (unsigned 2651 char A) 2652 -- Runtime Function: unsigned fract __fractunsqiuhq (unsigned char A) 2653 -- Runtime Function: unsigned long fract __fractunsqiusq (unsigned 2654 char A) 2655 -- Runtime Function: unsigned long long fract __fractunsqiudq 2656 (unsigned char A) 2657 -- Runtime Function: unsigned short accum __fractunsqiuha (unsigned 2658 char A) 2659 -- Runtime Function: unsigned accum __fractunsqiusa (unsigned char A) 2660 -- Runtime Function: unsigned long accum __fractunsqiuda (unsigned 2661 char A) 2662 -- Runtime Function: unsigned long long accum __fractunsqiuta 2663 (unsigned char A) 2664 -- Runtime Function: short fract __fractunshiqq (unsigned short A) 2665 -- Runtime Function: fract __fractunshihq (unsigned short A) 2666 -- Runtime Function: long fract __fractunshisq (unsigned short A) 2667 -- Runtime Function: long long fract __fractunshidq (unsigned short A) 2668 -- Runtime Function: short accum __fractunshiha (unsigned short A) 2669 -- Runtime Function: accum __fractunshisa (unsigned short A) 2670 -- Runtime Function: long accum __fractunshida (unsigned short A) 2671 -- Runtime Function: long long accum __fractunshita (unsigned short A) 2672 -- Runtime Function: unsigned short fract __fractunshiuqq (unsigned 2673 short A) 2674 -- Runtime Function: unsigned fract __fractunshiuhq (unsigned short A) 2675 -- Runtime Function: unsigned long fract __fractunshiusq (unsigned 2676 short A) 2677 -- Runtime Function: unsigned long long fract __fractunshiudq 2678 (unsigned short A) 2679 -- Runtime Function: unsigned short accum __fractunshiuha (unsigned 2680 short A) 2681 -- Runtime Function: unsigned accum __fractunshiusa (unsigned short A) 2682 -- Runtime Function: unsigned long accum __fractunshiuda (unsigned 2683 short A) 2684 -- Runtime Function: unsigned long long accum __fractunshiuta 2685 (unsigned short A) 2686 -- Runtime Function: short fract __fractunssiqq (unsigned int A) 2687 -- Runtime Function: fract __fractunssihq (unsigned int A) 2688 -- Runtime Function: long fract __fractunssisq (unsigned int A) 2689 -- Runtime Function: long long fract __fractunssidq (unsigned int A) 2690 -- Runtime Function: short accum __fractunssiha (unsigned int A) 2691 -- Runtime Function: accum __fractunssisa (unsigned int A) 2692 -- Runtime Function: long accum __fractunssida (unsigned int A) 2693 -- Runtime Function: long long accum __fractunssita (unsigned int A) 2694 -- Runtime Function: unsigned short fract __fractunssiuqq (unsigned 2695 int A) 2696 -- Runtime Function: unsigned fract __fractunssiuhq (unsigned int A) 2697 -- Runtime Function: unsigned long fract __fractunssiusq (unsigned int 2698 A) 2699 -- Runtime Function: unsigned long long fract __fractunssiudq 2700 (unsigned int A) 2701 -- Runtime Function: unsigned short accum __fractunssiuha (unsigned 2702 int A) 2703 -- Runtime Function: unsigned accum __fractunssiusa (unsigned int A) 2704 -- Runtime Function: unsigned long accum __fractunssiuda (unsigned int 2705 A) 2706 -- Runtime Function: unsigned long long accum __fractunssiuta 2707 (unsigned int A) 2708 -- Runtime Function: short fract __fractunsdiqq (unsigned long A) 2709 -- Runtime Function: fract __fractunsdihq (unsigned long A) 2710 -- Runtime Function: long fract __fractunsdisq (unsigned long A) 2711 -- Runtime Function: long long fract __fractunsdidq (unsigned long A) 2712 -- Runtime Function: short accum __fractunsdiha (unsigned long A) 2713 -- Runtime Function: accum __fractunsdisa (unsigned long A) 2714 -- Runtime Function: long accum __fractunsdida (unsigned long A) 2715 -- Runtime Function: long long accum __fractunsdita (unsigned long A) 2716 -- Runtime Function: unsigned short fract __fractunsdiuqq (unsigned 2717 long A) 2718 -- Runtime Function: unsigned fract __fractunsdiuhq (unsigned long A) 2719 -- Runtime Function: unsigned long fract __fractunsdiusq (unsigned 2720 long A) 2721 -- Runtime Function: unsigned long long fract __fractunsdiudq 2722 (unsigned long A) 2723 -- Runtime Function: unsigned short accum __fractunsdiuha (unsigned 2724 long A) 2725 -- Runtime Function: unsigned accum __fractunsdiusa (unsigned long A) 2726 -- Runtime Function: unsigned long accum __fractunsdiuda (unsigned 2727 long A) 2728 -- Runtime Function: unsigned long long accum __fractunsdiuta 2729 (unsigned long A) 2730 -- Runtime Function: short fract __fractunstiqq (unsigned long long A) 2731 -- Runtime Function: fract __fractunstihq (unsigned long long A) 2732 -- Runtime Function: long fract __fractunstisq (unsigned long long A) 2733 -- Runtime Function: long long fract __fractunstidq (unsigned long 2734 long A) 2735 -- Runtime Function: short accum __fractunstiha (unsigned long long A) 2736 -- Runtime Function: accum __fractunstisa (unsigned long long A) 2737 -- Runtime Function: long accum __fractunstida (unsigned long long A) 2738 -- Runtime Function: long long accum __fractunstita (unsigned long 2739 long A) 2740 -- Runtime Function: unsigned short fract __fractunstiuqq (unsigned 2741 long long A) 2742 -- Runtime Function: unsigned fract __fractunstiuhq (unsigned long 2743 long A) 2744 -- Runtime Function: unsigned long fract __fractunstiusq (unsigned 2745 long long A) 2746 -- Runtime Function: unsigned long long fract __fractunstiudq 2747 (unsigned long long A) 2748 -- Runtime Function: unsigned short accum __fractunstiuha (unsigned 2749 long long A) 2750 -- Runtime Function: unsigned accum __fractunstiusa (unsigned long 2751 long A) 2752 -- Runtime Function: unsigned long accum __fractunstiuda (unsigned 2753 long long A) 2754 -- Runtime Function: unsigned long long accum __fractunstiuta 2755 (unsigned long long A) 2756 These functions convert from fractionals to unsigned 2757 non-fractionals; and from unsigned non-fractionals to fractionals, 2758 without saturation. 2759 2760 -- Runtime Function: short fract __satfractunsqiqq (unsigned char A) 2761 -- Runtime Function: fract __satfractunsqihq (unsigned char A) 2762 -- Runtime Function: long fract __satfractunsqisq (unsigned char A) 2763 -- Runtime Function: long long fract __satfractunsqidq (unsigned char 2764 A) 2765 -- Runtime Function: short accum __satfractunsqiha (unsigned char A) 2766 -- Runtime Function: accum __satfractunsqisa (unsigned char A) 2767 -- Runtime Function: long accum __satfractunsqida (unsigned char A) 2768 -- Runtime Function: long long accum __satfractunsqita (unsigned char 2769 A) 2770 -- Runtime Function: unsigned short fract __satfractunsqiuqq (unsigned 2771 char A) 2772 -- Runtime Function: unsigned fract __satfractunsqiuhq (unsigned char 2773 A) 2774 -- Runtime Function: unsigned long fract __satfractunsqiusq (unsigned 2775 char A) 2776 -- Runtime Function: unsigned long long fract __satfractunsqiudq 2777 (unsigned char A) 2778 -- Runtime Function: unsigned short accum __satfractunsqiuha (unsigned 2779 char A) 2780 -- Runtime Function: unsigned accum __satfractunsqiusa (unsigned char 2781 A) 2782 -- Runtime Function: unsigned long accum __satfractunsqiuda (unsigned 2783 char A) 2784 -- Runtime Function: unsigned long long accum __satfractunsqiuta 2785 (unsigned char A) 2786 -- Runtime Function: short fract __satfractunshiqq (unsigned short A) 2787 -- Runtime Function: fract __satfractunshihq (unsigned short A) 2788 -- Runtime Function: long fract __satfractunshisq (unsigned short A) 2789 -- Runtime Function: long long fract __satfractunshidq (unsigned short 2790 A) 2791 -- Runtime Function: short accum __satfractunshiha (unsigned short A) 2792 -- Runtime Function: accum __satfractunshisa (unsigned short A) 2793 -- Runtime Function: long accum __satfractunshida (unsigned short A) 2794 -- Runtime Function: long long accum __satfractunshita (unsigned short 2795 A) 2796 -- Runtime Function: unsigned short fract __satfractunshiuqq (unsigned 2797 short A) 2798 -- Runtime Function: unsigned fract __satfractunshiuhq (unsigned short 2799 A) 2800 -- Runtime Function: unsigned long fract __satfractunshiusq (unsigned 2801 short A) 2802 -- Runtime Function: unsigned long long fract __satfractunshiudq 2803 (unsigned short A) 2804 -- Runtime Function: unsigned short accum __satfractunshiuha (unsigned 2805 short A) 2806 -- Runtime Function: unsigned accum __satfractunshiusa (unsigned short 2807 A) 2808 -- Runtime Function: unsigned long accum __satfractunshiuda (unsigned 2809 short A) 2810 -- Runtime Function: unsigned long long accum __satfractunshiuta 2811 (unsigned short A) 2812 -- Runtime Function: short fract __satfractunssiqq (unsigned int A) 2813 -- Runtime Function: fract __satfractunssihq (unsigned int A) 2814 -- Runtime Function: long fract __satfractunssisq (unsigned int A) 2815 -- Runtime Function: long long fract __satfractunssidq (unsigned int A) 2816 -- Runtime Function: short accum __satfractunssiha (unsigned int A) 2817 -- Runtime Function: accum __satfractunssisa (unsigned int A) 2818 -- Runtime Function: long accum __satfractunssida (unsigned int A) 2819 -- Runtime Function: long long accum __satfractunssita (unsigned int A) 2820 -- Runtime Function: unsigned short fract __satfractunssiuqq (unsigned 2821 int A) 2822 -- Runtime Function: unsigned fract __satfractunssiuhq (unsigned int A) 2823 -- Runtime Function: unsigned long fract __satfractunssiusq (unsigned 2824 int A) 2825 -- Runtime Function: unsigned long long fract __satfractunssiudq 2826 (unsigned int A) 2827 -- Runtime Function: unsigned short accum __satfractunssiuha (unsigned 2828 int A) 2829 -- Runtime Function: unsigned accum __satfractunssiusa (unsigned int A) 2830 -- Runtime Function: unsigned long accum __satfractunssiuda (unsigned 2831 int A) 2832 -- Runtime Function: unsigned long long accum __satfractunssiuta 2833 (unsigned int A) 2834 -- Runtime Function: short fract __satfractunsdiqq (unsigned long A) 2835 -- Runtime Function: fract __satfractunsdihq (unsigned long A) 2836 -- Runtime Function: long fract __satfractunsdisq (unsigned long A) 2837 -- Runtime Function: long long fract __satfractunsdidq (unsigned long 2838 A) 2839 -- Runtime Function: short accum __satfractunsdiha (unsigned long A) 2840 -- Runtime Function: accum __satfractunsdisa (unsigned long A) 2841 -- Runtime Function: long accum __satfractunsdida (unsigned long A) 2842 -- Runtime Function: long long accum __satfractunsdita (unsigned long 2843 A) 2844 -- Runtime Function: unsigned short fract __satfractunsdiuqq (unsigned 2845 long A) 2846 -- Runtime Function: unsigned fract __satfractunsdiuhq (unsigned long 2847 A) 2848 -- Runtime Function: unsigned long fract __satfractunsdiusq (unsigned 2849 long A) 2850 -- Runtime Function: unsigned long long fract __satfractunsdiudq 2851 (unsigned long A) 2852 -- Runtime Function: unsigned short accum __satfractunsdiuha (unsigned 2853 long A) 2854 -- Runtime Function: unsigned accum __satfractunsdiusa (unsigned long 2855 A) 2856 -- Runtime Function: unsigned long accum __satfractunsdiuda (unsigned 2857 long A) 2858 -- Runtime Function: unsigned long long accum __satfractunsdiuta 2859 (unsigned long A) 2860 -- Runtime Function: short fract __satfractunstiqq (unsigned long long 2861 A) 2862 -- Runtime Function: fract __satfractunstihq (unsigned long long A) 2863 -- Runtime Function: long fract __satfractunstisq (unsigned long long 2864 A) 2865 -- Runtime Function: long long fract __satfractunstidq (unsigned long 2866 long A) 2867 -- Runtime Function: short accum __satfractunstiha (unsigned long long 2868 A) 2869 -- Runtime Function: accum __satfractunstisa (unsigned long long A) 2870 -- Runtime Function: long accum __satfractunstida (unsigned long long 2871 A) 2872 -- Runtime Function: long long accum __satfractunstita (unsigned long 2873 long A) 2874 -- Runtime Function: unsigned short fract __satfractunstiuqq (unsigned 2875 long long A) 2876 -- Runtime Function: unsigned fract __satfractunstiuhq (unsigned long 2877 long A) 2878 -- Runtime Function: unsigned long fract __satfractunstiusq (unsigned 2879 long long A) 2880 -- Runtime Function: unsigned long long fract __satfractunstiudq 2881 (unsigned long long A) 2882 -- Runtime Function: unsigned short accum __satfractunstiuha (unsigned 2883 long long A) 2884 -- Runtime Function: unsigned accum __satfractunstiusa (unsigned long 2885 long A) 2886 -- Runtime Function: unsigned long accum __satfractunstiuda (unsigned 2887 long long A) 2888 -- Runtime Function: unsigned long long accum __satfractunstiuta 2889 (unsigned long long A) 2890 These functions convert from unsigned non-fractionals to 2891 fractionals, with saturation. 2892 2893 2894 File: gccint.info, Node: Exception handling routines, Next: Miscellaneous routines, Prev: Fixed-point fractional library routines, Up: Libgcc 2895 2896 4.5 Language-independent routines for exception handling 2897 ======================================================== 2898 2899 document me! 2900 2901 _Unwind_DeleteException 2902 _Unwind_Find_FDE 2903 _Unwind_ForcedUnwind 2904 _Unwind_GetGR 2905 _Unwind_GetIP 2906 _Unwind_GetLanguageSpecificData 2907 _Unwind_GetRegionStart 2908 _Unwind_GetTextRelBase 2909 _Unwind_GetDataRelBase 2910 _Unwind_RaiseException 2911 _Unwind_Resume 2912 _Unwind_SetGR 2913 _Unwind_SetIP 2914 _Unwind_FindEnclosingFunction 2915 _Unwind_SjLj_Register 2916 _Unwind_SjLj_Unregister 2917 _Unwind_SjLj_RaiseException 2918 _Unwind_SjLj_ForcedUnwind 2919 _Unwind_SjLj_Resume 2920 __deregister_frame 2921 __deregister_frame_info 2922 __deregister_frame_info_bases 2923 __register_frame 2924 __register_frame_info 2925 __register_frame_info_bases 2926 __register_frame_info_table 2927 __register_frame_info_table_bases 2928 __register_frame_table 2929 2930 2931 File: gccint.info, Node: Miscellaneous routines, Prev: Exception handling routines, Up: Libgcc 2932 2933 4.6 Miscellaneous runtime library routines 2934 ========================================== 2935 2936 4.6.1 Cache control functions 2937 ----------------------------- 2938 2939 -- Runtime Function: void __clear_cache (char *BEG, char *END) 2940 This function clears the instruction cache between BEG and END. 2941 2942 2943 File: gccint.info, Node: Languages, Next: Source Tree, Prev: Libgcc, Up: Top 2944 2945 5 Language Front Ends in GCC 2946 **************************** 2947 2948 The interface to front ends for languages in GCC, and in particular the 2949 `tree' structure (*note Trees::), was initially designed for C, and 2950 many aspects of it are still somewhat biased towards C and C-like 2951 languages. It is, however, reasonably well suited to other procedural 2952 languages, and front ends for many such languages have been written for 2953 GCC. 2954 2955 Writing a compiler as a front end for GCC, rather than compiling 2956 directly to assembler or generating C code which is then compiled by 2957 GCC, has several advantages: 2958 2959 * GCC front ends benefit from the support for many different target 2960 machines already present in GCC. 2961 2962 * GCC front ends benefit from all the optimizations in GCC. Some of 2963 these, such as alias analysis, may work better when GCC is 2964 compiling directly from source code then when it is compiling from 2965 generated C code. 2966 2967 * Better debugging information is generated when compiling directly 2968 from source code than when going via intermediate generated C code. 2969 2970 Because of the advantages of writing a compiler as a GCC front end, 2971 GCC front ends have also been created for languages very different from 2972 those for which GCC was designed, such as the declarative 2973 logic/functional language Mercury. For these reasons, it may also be 2974 useful to implement compilers created for specialized purposes (for 2975 example, as part of a research project) as GCC front ends. 2976 2977 2978 File: gccint.info, Node: Source Tree, Next: Options, Prev: Languages, Up: Top 2979 2980 6 Source Tree Structure and Build System 2981 **************************************** 2982 2983 This chapter describes the structure of the GCC source tree, and how 2984 GCC is built. The user documentation for building and installing GCC 2985 is in a separate manual (`http://gcc.gnu.org/install/'), with which it 2986 is presumed that you are familiar. 2987 2988 * Menu: 2989 2990 * Configure Terms:: Configuration terminology and history. 2991 * Top Level:: The top level source directory. 2992 * gcc Directory:: The `gcc' subdirectory. 2993 * Testsuites:: The GCC testsuites. 2994 2995 2996 File: gccint.info, Node: Configure Terms, Next: Top Level, Up: Source Tree 2997 2998 6.1 Configure Terms and History 2999 =============================== 3000 3001 The configure and build process has a long and colorful history, and can 3002 be confusing to anyone who doesn't know why things are the way they are. 3003 While there are other documents which describe the configuration process 3004 in detail, here are a few things that everyone working on GCC should 3005 know. 3006 3007 There are three system names that the build knows about: the machine 3008 you are building on ("build"), the machine that you are building for 3009 ("host"), and the machine that GCC will produce code for ("target"). 3010 When you configure GCC, you specify these with `--build=', `--host=', 3011 and `--target='. 3012 3013 Specifying the host without specifying the build should be avoided, as 3014 `configure' may (and once did) assume that the host you specify is also 3015 the build, which may not be true. 3016 3017 If build, host, and target are all the same, this is called a 3018 "native". If build and host are the same but target is different, this 3019 is called a "cross". If build, host, and target are all different this 3020 is called a "canadian" (for obscure reasons dealing with Canada's 3021 political party and the background of the person working on the build 3022 at that time). If host and target are the same, but build is 3023 different, you are using a cross-compiler to build a native for a 3024 different system. Some people call this a "host-x-host", "crossed 3025 native", or "cross-built native". If build and target are the same, 3026 but host is different, you are using a cross compiler to build a cross 3027 compiler that produces code for the machine you're building on. This 3028 is rare, so there is no common way of describing it. There is a 3029 proposal to call this a "crossback". 3030 3031 If build and host are the same, the GCC you are building will also be 3032 used to build the target libraries (like `libstdc++'). If build and 3033 host are different, you must have already built and installed a cross 3034 compiler that will be used to build the target libraries (if you 3035 configured with `--target=foo-bar', this compiler will be called 3036 `foo-bar-gcc'). 3037 3038 In the case of target libraries, the machine you're building for is the 3039 machine you specified with `--target'. So, build is the machine you're 3040 building on (no change there), host is the machine you're building for 3041 (the target libraries are built for the target, so host is the target 3042 you specified), and target doesn't apply (because you're not building a 3043 compiler, you're building libraries). The configure/make process will 3044 adjust these variables as needed. It also sets `$with_cross_host' to 3045 the original `--host' value in case you need it. 3046 3047 The `libiberty' support library is built up to three times: once for 3048 the host, once for the target (even if they are the same), and once for 3049 the build if build and host are different. This allows it to be used 3050 by all programs which are generated in the course of the build process. 3051 3052 3053 File: gccint.info, Node: Top Level, Next: gcc Directory, Prev: Configure Terms, Up: Source Tree 3054 3055 6.2 Top Level Source Directory 3056 ============================== 3057 3058 The top level source directory in a GCC distribution contains several 3059 files and directories that are shared with other software distributions 3060 such as that of GNU Binutils. It also contains several subdirectories 3061 that contain parts of GCC and its runtime libraries: 3062 3063 `boehm-gc' 3064 The Boehm conservative garbage collector, used as part of the Java 3065 runtime library. 3066 3067 `contrib' 3068 Contributed scripts that may be found useful in conjunction with 3069 GCC. One of these, `contrib/texi2pod.pl', is used to generate man 3070 pages from Texinfo manuals as part of the GCC build process. 3071 3072 `fastjar' 3073 An implementation of the `jar' command, used with the Java front 3074 end. 3075 3076 `fixincludes' 3077 The support for fixing system headers to work with GCC. See 3078 `fixincludes/README' for more information. The headers fixed by 3079 this mechanism are installed in `LIBSUBDIR/include-fixed'. Along 3080 with those headers, `README-fixinc' is also installed, as 3081 `LIBSUBDIR/include-fixed/README'. 3082 3083 `gcc' 3084 The main sources of GCC itself (except for runtime libraries), 3085 including optimizers, support for different target architectures, 3086 language front ends, and testsuites. *Note The `gcc' 3087 Subdirectory: gcc Directory, for details. 3088 3089 `include' 3090 Headers for the `libiberty' library. 3091 3092 `intl' 3093 GNU `libintl', from GNU `gettext', for systems which do not 3094 include it in libc. 3095 3096 `libada' 3097 The Ada runtime library. 3098 3099 `libcpp' 3100 The C preprocessor library. 3101 3102 `libgfortran' 3103 The Fortran runtime library. 3104 3105 `libffi' 3106 The `libffi' library, used as part of the Java runtime library. 3107 3108 `libiberty' 3109 The `libiberty' library, used for portability and for some 3110 generally useful data structures and algorithms. *Note 3111 Introduction: (libiberty)Top, for more information about this 3112 library. 3113 3114 `libjava' 3115 The Java runtime library. 3116 3117 `libmudflap' 3118 The `libmudflap' library, used for instrumenting pointer and array 3119 dereferencing operations. 3120 3121 `libobjc' 3122 The Objective-C and Objective-C++ runtime library. 3123 3124 `libstdc++-v3' 3125 The C++ runtime library. 3126 3127 `maintainer-scripts' 3128 Scripts used by the `gccadmin' account on `gcc.gnu.org'. 3129 3130 `zlib' 3131 The `zlib' compression library, used by the Java front end and as 3132 part of the Java runtime library. 3133 3134 The build system in the top level directory, including how recursion 3135 into subdirectories works and how building runtime libraries for 3136 multilibs is handled, is documented in a separate manual, included with 3137 GNU Binutils. *Note GNU configure and build system: (configure)Top, 3138 for details. 3139 3140 3141 File: gccint.info, Node: gcc Directory, Next: Testsuites, Prev: Top Level, Up: Source Tree 3142 3143 6.3 The `gcc' Subdirectory 3144 ========================== 3145 3146 The `gcc' directory contains many files that are part of the C sources 3147 of GCC, other files used as part of the configuration and build 3148 process, and subdirectories including documentation and a testsuite. 3149 The files that are sources of GCC are documented in a separate chapter. 3150 *Note Passes and Files of the Compiler: Passes. 3151 3152 * Menu: 3153 3154 * Subdirectories:: Subdirectories of `gcc'. 3155 * Configuration:: The configuration process, and the files it uses. 3156 * Build:: The build system in the `gcc' directory. 3157 * Makefile:: Targets in `gcc/Makefile'. 3158 * Library Files:: Library source files and headers under `gcc/'. 3159 * Headers:: Headers installed by GCC. 3160 * Documentation:: Building documentation in GCC. 3161 * Front End:: Anatomy of a language front end. 3162 * Back End:: Anatomy of a target back end. 3163 3164 3165 File: gccint.info, Node: Subdirectories, Next: Configuration, Up: gcc Directory 3166 3167 6.3.1 Subdirectories of `gcc' 3168 ----------------------------- 3169 3170 The `gcc' directory contains the following subdirectories: 3171 3172 `LANGUAGE' 3173 Subdirectories for various languages. Directories containing a 3174 file `config-lang.in' are language subdirectories. The contents of 3175 the subdirectories `cp' (for C++), `objc' (for Objective-C) and 3176 `objcp' (for Objective-C++) are documented in this manual (*note 3177 Passes and Files of the Compiler: Passes.); those for other 3178 languages are not. *Note Anatomy of a Language Front End: Front 3179 End, for details of the files in these directories. 3180 3181 `config' 3182 Configuration files for supported architectures and operating 3183 systems. *Note Anatomy of a Target Back End: Back End, for 3184 details of the files in this directory. 3185 3186 `doc' 3187 Texinfo documentation for GCC, together with automatically 3188 generated man pages and support for converting the installation 3189 manual to HTML. *Note Documentation::. 3190 3191 `ginclude' 3192 System headers installed by GCC, mainly those required by the C 3193 standard of freestanding implementations. *Note Headers Installed 3194 by GCC: Headers, for details of when these and other headers are 3195 installed. 3196 3197 `po' 3198 Message catalogs with translations of messages produced by GCC into 3199 various languages, `LANGUAGE.po'. This directory also contains 3200 `gcc.pot', the template for these message catalogues, `exgettext', 3201 a wrapper around `gettext' to extract the messages from the GCC 3202 sources and create `gcc.pot', which is run by `make gcc.pot', and 3203 `EXCLUDES', a list of files from which messages should not be 3204 extracted. 3205 3206 `testsuite' 3207 The GCC testsuites (except for those for runtime libraries). 3208 *Note Testsuites::. 3209 3210 3211 File: gccint.info, Node: Configuration, Next: Build, Prev: Subdirectories, Up: gcc Directory 3212 3213 6.3.2 Configuration in the `gcc' Directory 3214 ------------------------------------------ 3215 3216 The `gcc' directory is configured with an Autoconf-generated script 3217 `configure'. The `configure' script is generated from `configure.ac' 3218 and `aclocal.m4'. From the files `configure.ac' and `acconfig.h', 3219 Autoheader generates the file `config.in'. The file `cstamp-h.in' is 3220 used as a timestamp. 3221 3222 * Menu: 3223 3224 * Config Fragments:: Scripts used by `configure'. 3225 * System Config:: The `config.build', `config.host', and 3226 `config.gcc' files. 3227 * Configuration Files:: Files created by running `configure'. 3228 3229 3230 File: gccint.info, Node: Config Fragments, Next: System Config, Up: Configuration 3231 3232 6.3.2.1 Scripts Used by `configure' 3233 ................................... 3234 3235 `configure' uses some other scripts to help in its work: 3236 3237 * The standard GNU `config.sub' and `config.guess' files, kept in 3238 the top level directory, are used. 3239 3240 * The file `config.gcc' is used to handle configuration specific to 3241 the particular target machine. The file `config.build' is used to 3242 handle configuration specific to the particular build machine. 3243 The file `config.host' is used to handle configuration specific to 3244 the particular host machine. (In general, these should only be 3245 used for features that cannot reasonably be tested in Autoconf 3246 feature tests.) *Note The `config.build'; `config.host'; and 3247 `config.gcc' Files: System Config, for details of the contents of 3248 these files. 3249 3250 * Each language subdirectory has a file `LANGUAGE/config-lang.in' 3251 that is used for front-end-specific configuration. *Note The 3252 Front End `config-lang.in' File: Front End Config, for details of 3253 this file. 3254 3255 * A helper script `configure.frag' is used as part of creating the 3256 output of `configure'. 3257 3258 3259 File: gccint.info, Node: System Config, Next: Configuration Files, Prev: Config Fragments, Up: Configuration 3260 3261 6.3.2.2 The `config.build'; `config.host'; and `config.gcc' Files 3262 ................................................................. 3263 3264 The `config.build' file contains specific rules for particular systems 3265 which GCC is built on. This should be used as rarely as possible, as 3266 the behavior of the build system can always be detected by autoconf. 3267 3268 The `config.host' file contains specific rules for particular systems 3269 which GCC will run on. This is rarely needed. 3270 3271 The `config.gcc' file contains specific rules for particular systems 3272 which GCC will generate code for. This is usually needed. 3273 3274 Each file has a list of the shell variables it sets, with 3275 descriptions, at the top of the file. 3276 3277 FIXME: document the contents of these files, and what variables should 3278 be set to control build, host and target configuration. 3279 3280 3281 File: gccint.info, Node: Configuration Files, Prev: System Config, Up: Configuration 3282 3283 6.3.2.3 Files Created by `configure' 3284 .................................... 3285 3286 Here we spell out what files will be set up by `configure' in the `gcc' 3287 directory. Some other files are created as temporary files in the 3288 configuration process, and are not used in the subsequent build; these 3289 are not documented. 3290 3291 * `Makefile' is constructed from `Makefile.in', together with the 3292 host and target fragments (*note Makefile Fragments: Fragments.) 3293 `t-TARGET' and `x-HOST' from `config', if any, and language 3294 Makefile fragments `LANGUAGE/Make-lang.in'. 3295 3296 * `auto-host.h' contains information about the host machine 3297 determined by `configure'. If the host machine is different from 3298 the build machine, then `auto-build.h' is also created, containing 3299 such information about the build machine. 3300 3301 * `config.status' is a script that may be run to recreate the 3302 current configuration. 3303 3304 * `configargs.h' is a header containing details of the arguments 3305 passed to `configure' to configure GCC, and of the thread model 3306 used. 3307 3308 * `cstamp-h' is used as a timestamp. 3309 3310 * `fixinc/Makefile' is constructed from `fixinc/Makefile.in'. 3311 3312 * `gccbug', a script for reporting bugs in GCC, is constructed from 3313 `gccbug.in'. 3314 3315 * `intl/Makefile' is constructed from `intl/Makefile.in'. 3316 3317 * If a language `config-lang.in' file (*note The Front End 3318 `config-lang.in' File: Front End Config.) sets `outputs', then the 3319 files listed in `outputs' there are also generated. 3320 3321 The following configuration headers are created from the Makefile, 3322 using `mkconfig.sh', rather than directly by `configure'. `config.h', 3323 `bconfig.h' and `tconfig.h' all contain the `xm-MACHINE.h' header, if 3324 any, appropriate to the host, build and target machines respectively, 3325 the configuration headers for the target, and some definitions; for the 3326 host and build machines, these include the autoconfigured headers 3327 generated by `configure'. The other configuration headers are 3328 determined by `config.gcc'. They also contain the typedefs for `rtx', 3329 `rtvec' and `tree'. 3330 3331 * `config.h', for use in programs that run on the host machine. 3332 3333 * `bconfig.h', for use in programs that run on the build machine. 3334 3335 * `tconfig.h', for use in programs and libraries for the target 3336 machine. 3337 3338 * `tm_p.h', which includes the header `MACHINE-protos.h' that 3339 contains prototypes for functions in the target `.c' file. FIXME: 3340 why is such a separate header necessary? 3341 3342 3343 File: gccint.info, Node: Build, Next: Makefile, Prev: Configuration, Up: gcc Directory 3344 3345 6.3.3 Build System in the `gcc' Directory 3346 ----------------------------------------- 3347 3348 FIXME: describe the build system, including what is built in what 3349 stages. Also list the various source files that are used in the build 3350 process but aren't source files of GCC itself and so aren't documented 3351 below (*note Passes::). 3352 3353 3354 File: gccint.info, Node: Makefile, Next: Library Files, Prev: Build, Up: gcc Directory 3355 3356 6.3.4 Makefile Targets 3357 ---------------------- 3358 3359 These targets are available from the `gcc' directory: 3360 3361 `all' 3362 This is the default target. Depending on what your 3363 build/host/target configuration is, it coordinates all the things 3364 that need to be built. 3365 3366 `doc' 3367 Produce info-formatted documentation and man pages. Essentially it 3368 calls `make man' and `make info'. 3369 3370 `dvi' 3371 Produce DVI-formatted documentation. 3372 3373 `pdf' 3374 Produce PDF-formatted documentation. 3375 3376 `html' 3377 Produce HTML-formatted documentation. 3378 3379 `man' 3380 Generate man pages. 3381 3382 `info' 3383 Generate info-formatted pages. 3384 3385 `mostlyclean' 3386 Delete the files made while building the compiler. 3387 3388 `clean' 3389 That, and all the other files built by `make all'. 3390 3391 `distclean' 3392 That, and all the files created by `configure'. 3393 3394 `maintainer-clean' 3395 Distclean plus any file that can be generated from other files. 3396 Note that additional tools may be required beyond what is normally 3397 needed to build gcc. 3398 3399 `srcextra' 3400 Generates files in the source directory that do not exist in CVS 3401 but should go into a release tarball. One example is 3402 `gcc/java/parse.c' which is generated from the CVS source file 3403 `gcc/java/parse.y'. 3404 3405 `srcinfo' 3406 `srcman' 3407 Copies the info-formatted and manpage documentation into the source 3408 directory usually for the purpose of generating a release tarball. 3409 3410 `install' 3411 Installs gcc. 3412 3413 `uninstall' 3414 Deletes installed files. 3415 3416 `check' 3417 Run the testsuite. This creates a `testsuite' subdirectory that 3418 has various `.sum' and `.log' files containing the results of the 3419 testing. You can run subsets with, for example, `make check-gcc'. 3420 You can specify specific tests by setting RUNTESTFLAGS to be the 3421 name of the `.exp' file, optionally followed by (for some tests) 3422 an equals and a file wildcard, like: 3423 3424 make check-gcc RUNTESTFLAGS="execute.exp=19980413-*" 3425 3426 Note that running the testsuite may require additional tools be 3427 installed, such as TCL or dejagnu. 3428 3429 The toplevel tree from which you start GCC compilation is not the GCC 3430 directory, but rather a complex Makefile that coordinates the various 3431 steps of the build, including bootstrapping the compiler and using the 3432 new compiler to build target libraries. 3433 3434 When GCC is configured for a native configuration, the default action 3435 for `make' is to do a full three-stage bootstrap. This means that GCC 3436 is built three times--once with the native compiler, once with the 3437 native-built compiler it just built, and once with the compiler it 3438 built the second time. In theory, the last two should produce the same 3439 results, which `make compare' can check. Each stage is configured 3440 separately and compiled into a separate directory, to minimize problems 3441 due to ABI incompatibilities between the native compiler and GCC. 3442 3443 If you do a change, rebuilding will also start from the first stage 3444 and "bubble" up the change through the three stages. Each stage is 3445 taken from its build directory (if it had been built previously), 3446 rebuilt, and copied to its subdirectory. This will allow you to, for 3447 example, continue a bootstrap after fixing a bug which causes the 3448 stage2 build to crash. It does not provide as good coverage of the 3449 compiler as bootstrapping from scratch, but it ensures that the new 3450 code is syntactically correct (e.g., that you did not use GCC extensions 3451 by mistake), and avoids spurious bootstrap comparison failures(1). 3452 3453 Other targets available from the top level include: 3454 3455 `bootstrap-lean' 3456 Like `bootstrap', except that the various stages are removed once 3457 they're no longer needed. This saves disk space. 3458 3459 `bootstrap2' 3460 `bootstrap2-lean' 3461 Performs only the first two stages of bootstrap. Unlike a 3462 three-stage bootstrap, this does not perform a comparison to test 3463 that the compiler is running properly. Note that the disk space 3464 required by a "lean" bootstrap is approximately independent of the 3465 number of stages. 3466 3467 `stageN-bubble (N = 1...4)' 3468 Rebuild all the stages up to N, with the appropriate flags, 3469 "bubbling" the changes as described above. 3470 3471 `all-stageN (N = 1...4)' 3472 Assuming that stage N has already been built, rebuild it with the 3473 appropriate flags. This is rarely needed. 3474 3475 `cleanstrap' 3476 Remove everything (`make clean') and rebuilds (`make bootstrap'). 3477 3478 `compare' 3479 Compares the results of stages 2 and 3. This ensures that the 3480 compiler is running properly, since it should produce the same 3481 object files regardless of how it itself was compiled. 3482 3483 `profiledbootstrap' 3484 Builds a compiler with profiling feedback information. For more 3485 information, see *Note Building with profile feedback: 3486 (gccinstall)Building. 3487 3488 `restrap' 3489 Restart a bootstrap, so that everything that was not built with 3490 the system compiler is rebuilt. 3491 3492 `stageN-start (N = 1...4)' 3493 For each package that is bootstrapped, rename directories so that, 3494 for example, `gcc' points to the stageN GCC, compiled with the 3495 stageN-1 GCC(2). 3496 3497 You will invoke this target if you need to test or debug the 3498 stageN GCC. If you only need to execute GCC (but you need not run 3499 `make' either to rebuild it or to run test suites), you should be 3500 able to work directly in the `stageN-gcc' directory. This makes 3501 it easier to debug multiple stages in parallel. 3502 3503 `stage' 3504 For each package that is bootstrapped, relocate its build directory 3505 to indicate its stage. For example, if the `gcc' directory points 3506 to the stage2 GCC, after invoking this target it will be renamed 3507 to `stage2-gcc'. 3508 3509 3510 If you wish to use non-default GCC flags when compiling the stage2 and 3511 stage3 compilers, set `BOOT_CFLAGS' on the command line when doing 3512 `make'. 3513 3514 Usually, the first stage only builds the languages that the compiler 3515 is written in: typically, C and maybe Ada. If you are debugging a 3516 miscompilation of a different stage2 front-end (for example, of the 3517 Fortran front-end), you may want to have front-ends for other languages 3518 in the first stage as well. To do so, set `STAGE1_LANGUAGES' on the 3519 command line when doing `make'. 3520 3521 For example, in the aforementioned scenario of debugging a Fortran 3522 front-end miscompilation caused by the stage1 compiler, you may need a 3523 command like 3524 3525 make stage2-bubble STAGE1_LANGUAGES=c,fortran 3526 3527 Alternatively, you can use per-language targets to build and test 3528 languages that are not enabled by default in stage1. For example, 3529 `make f951' will build a Fortran compiler even in the stage1 build 3530 directory. 3531 3532 ---------- Footnotes ---------- 3533 3534 (1) Except if the compiler was buggy and miscompiled some of the files 3535 that were not modified. In this case, it's best to use `make restrap'. 3536 3537 (2) Customarily, the system compiler is also termed the `stage0' GCC. 3538 3539 3540 File: gccint.info, Node: Library Files, Next: Headers, Prev: Makefile, Up: gcc Directory 3541 3542 6.3.5 Library Source Files and Headers under the `gcc' Directory 3543 ---------------------------------------------------------------- 3544 3545 FIXME: list here, with explanation, all the C source files and headers 3546 under the `gcc' directory that aren't built into the GCC executable but 3547 rather are part of runtime libraries and object files, such as 3548 `crtstuff.c' and `unwind-dw2.c'. *Note Headers Installed by GCC: 3549 Headers, for more information about the `ginclude' directory. 3550 3551 3552 File: gccint.info, Node: Headers, Next: Documentation, Prev: Library Files, Up: gcc Directory 3553 3554 6.3.6 Headers Installed by GCC 3555 ------------------------------ 3556 3557 In general, GCC expects the system C library to provide most of the 3558 headers to be used with it. However, GCC will fix those headers if 3559 necessary to make them work with GCC, and will install some headers 3560 required of freestanding implementations. These headers are installed 3561 in `LIBSUBDIR/include'. Headers for non-C runtime libraries are also 3562 installed by GCC; these are not documented here. (FIXME: document them 3563 somewhere.) 3564 3565 Several of the headers GCC installs are in the `ginclude' directory. 3566 These headers, `iso646.h', `stdarg.h', `stdbool.h', and `stddef.h', are 3567 installed in `LIBSUBDIR/include', unless the target Makefile fragment 3568 (*note Target Fragment::) overrides this by setting `USER_H'. 3569 3570 In addition to these headers and those generated by fixing system 3571 headers to work with GCC, some other headers may also be installed in 3572 `LIBSUBDIR/include'. `config.gcc' may set `extra_headers'; this 3573 specifies additional headers under `config' to be installed on some 3574 systems. 3575 3576 GCC installs its own version of `<float.h>', from `ginclude/float.h'. 3577 This is done to cope with command-line options that change the 3578 representation of floating point numbers. 3579 3580 GCC also installs its own version of `<limits.h>'; this is generated 3581 from `glimits.h', together with `limitx.h' and `limity.h' if the system 3582 also has its own version of `<limits.h>'. (GCC provides its own header 3583 because it is required of ISO C freestanding implementations, but needs 3584 to include the system header from its own header as well because other 3585 standards such as POSIX specify additional values to be defined in 3586 `<limits.h>'.) The system's `<limits.h>' header is used via 3587 `LIBSUBDIR/include/syslimits.h', which is copied from `gsyslimits.h' if 3588 it does not need fixing to work with GCC; if it needs fixing, 3589 `syslimits.h' is the fixed copy. 3590 3591 GCC can also install `<tgmath.h>'. It will do this when `config.gcc' 3592 sets `use_gcc_tgmath' to `yes'. 3593 3594 3595 File: gccint.info, Node: Documentation, Next: Front End, Prev: Headers, Up: gcc Directory 3596 3597 6.3.7 Building Documentation 3598 ---------------------------- 3599 3600 The main GCC documentation is in the form of manuals in Texinfo format. 3601 These are installed in Info format; DVI versions may be generated by 3602 `make dvi', PDF versions by `make pdf', and HTML versions by `make 3603 html'. In addition, some man pages are generated from the Texinfo 3604 manuals, there are some other text files with miscellaneous 3605 documentation, and runtime libraries have their own documentation 3606 outside the `gcc' directory. FIXME: document the documentation for 3607 runtime libraries somewhere. 3608 3609 * Menu: 3610 3611 * Texinfo Manuals:: GCC manuals in Texinfo format. 3612 * Man Page Generation:: Generating man pages from Texinfo manuals. 3613 * Miscellaneous Docs:: Miscellaneous text files with documentation. 3614 3615 3616 File: gccint.info, Node: Texinfo Manuals, Next: Man Page Generation, Up: Documentation 3617 3618 6.3.7.1 Texinfo Manuals 3619 ....................... 3620 3621 The manuals for GCC as a whole, and the C and C++ front ends, are in 3622 files `doc/*.texi'. Other front ends have their own manuals in files 3623 `LANGUAGE/*.texi'. Common files `doc/include/*.texi' are provided 3624 which may be included in multiple manuals; the following files are in 3625 `doc/include': 3626 3627 `fdl.texi' 3628 The GNU Free Documentation License. 3629 3630 `funding.texi' 3631 The section "Funding Free Software". 3632 3633 `gcc-common.texi' 3634 Common definitions for manuals. 3635 3636 `gpl.texi' 3637 `gpl_v3.texi' 3638 The GNU General Public License. 3639 3640 `texinfo.tex' 3641 A copy of `texinfo.tex' known to work with the GCC manuals. 3642 3643 DVI-formatted manuals are generated by `make dvi', which uses 3644 `texi2dvi' (via the Makefile macro `$(TEXI2DVI)'). PDF-formatted 3645 manuals are generated by `make pdf', which uses `texi2pdf' (via the 3646 Makefile macro `$(TEXI2PDF)'). HTML formatted manuals are generated by 3647 `make html'. Info manuals are generated by `make info' (which is run 3648 as part of a bootstrap); this generates the manuals in the source 3649 directory, using `makeinfo' via the Makefile macro `$(MAKEINFO)', and 3650 they are included in release distributions. 3651 3652 Manuals are also provided on the GCC web site, in both HTML and 3653 PostScript forms. This is done via the script 3654 `maintainer-scripts/update_web_docs'. Each manual to be provided 3655 online must be listed in the definition of `MANUALS' in that file; a 3656 file `NAME.texi' must only appear once in the source tree, and the 3657 output manual must have the same name as the source file. (However, 3658 other Texinfo files, included in manuals but not themselves the root 3659 files of manuals, may have names that appear more than once in the 3660 source tree.) The manual file `NAME.texi' should only include other 3661 files in its own directory or in `doc/include'. HTML manuals will be 3662 generated by `makeinfo --html', PostScript manuals by `texi2dvi' and 3663 `dvips', and PDF manuals by `texi2pdf'. All Texinfo files that are 3664 parts of manuals must be checked into SVN, even if they are generated 3665 files, for the generation of online manuals to work. 3666 3667 The installation manual, `doc/install.texi', is also provided on the 3668 GCC web site. The HTML version is generated by the script 3669 `doc/install.texi2html'. 3670 3671 3672 File: gccint.info, Node: Man Page Generation, Next: Miscellaneous Docs, Prev: Texinfo Manuals, Up: Documentation 3673 3674 6.3.7.2 Man Page Generation 3675 ........................... 3676 3677 Because of user demand, in addition to full Texinfo manuals, man pages 3678 are provided which contain extracts from those manuals. These man 3679 pages are generated from the Texinfo manuals using 3680 `contrib/texi2pod.pl' and `pod2man'. (The man page for `g++', 3681 `cp/g++.1', just contains a `.so' reference to `gcc.1', but all the 3682 other man pages are generated from Texinfo manuals.) 3683 3684 Because many systems may not have the necessary tools installed to 3685 generate the man pages, they are only generated if the `configure' 3686 script detects that recent enough tools are installed, and the 3687 Makefiles allow generating man pages to fail without aborting the 3688 build. Man pages are also included in release distributions. They are 3689 generated in the source directory. 3690 3691 Magic comments in Texinfo files starting `@c man' control what parts 3692 of a Texinfo file go into a man page. Only a subset of Texinfo is 3693 supported by `texi2pod.pl', and it may be necessary to add support for 3694 more Texinfo features to this script when generating new man pages. To 3695 improve the man page output, some special Texinfo macros are provided 3696 in `doc/include/gcc-common.texi' which `texi2pod.pl' understands: 3697 3698 `@gcctabopt' 3699 Use in the form `@table @gcctabopt' for tables of options, where 3700 for printed output the effect of `@code' is better than that of 3701 `@option' but for man page output a different effect is wanted. 3702 3703 `@gccoptlist' 3704 Use for summary lists of options in manuals. 3705 3706 `@gol' 3707 Use at the end of each line inside `@gccoptlist'. This is 3708 necessary to avoid problems with differences in how the 3709 `@gccoptlist' macro is handled by different Texinfo formatters. 3710 3711 FIXME: describe the `texi2pod.pl' input language and magic comments in 3712 more detail. 3713 3714 3715 File: gccint.info, Node: Miscellaneous Docs, Prev: Man Page Generation, Up: Documentation 3716 3717 6.3.7.3 Miscellaneous Documentation 3718 ................................... 3719 3720 In addition to the formal documentation that is installed by GCC, there 3721 are several other text files with miscellaneous documentation: 3722 3723 `ABOUT-GCC-NLS' 3724 Notes on GCC's Native Language Support. FIXME: this should be 3725 part of this manual rather than a separate file. 3726 3727 `ABOUT-NLS' 3728 Notes on the Free Translation Project. 3729 3730 `COPYING' 3731 The GNU General Public License. 3732 3733 `COPYING.LIB' 3734 The GNU Lesser General Public License. 3735 3736 `*ChangeLog*' 3737 `*/ChangeLog*' 3738 Change log files for various parts of GCC. 3739 3740 `LANGUAGES' 3741 Details of a few changes to the GCC front-end interface. FIXME: 3742 the information in this file should be part of general 3743 documentation of the front-end interface in this manual. 3744 3745 `ONEWS' 3746 Information about new features in old versions of GCC. (For recent 3747 versions, the information is on the GCC web site.) 3748 3749 `README.Portability' 3750 Information about portability issues when writing code in GCC. 3751 FIXME: why isn't this part of this manual or of the GCC Coding 3752 Conventions? 3753 3754 FIXME: document such files in subdirectories, at least `config', `cp', 3755 `objc', `testsuite'. 3756 3757 3758 File: gccint.info, Node: Front End, Next: Back End, Prev: Documentation, Up: gcc Directory 3759 3760 6.3.8 Anatomy of a Language Front End 3761 ------------------------------------- 3762 3763 A front end for a language in GCC has the following parts: 3764 3765 * A directory `LANGUAGE' under `gcc' containing source files for 3766 that front end. *Note The Front End `LANGUAGE' Directory: Front 3767 End Directory, for details. 3768 3769 * A mention of the language in the list of supported languages in 3770 `gcc/doc/install.texi'. 3771 3772 * A mention of the name under which the language's runtime library is 3773 recognized by `--enable-shared=PACKAGE' in the documentation of 3774 that option in `gcc/doc/install.texi'. 3775 3776 * A mention of any special prerequisites for building the front end 3777 in the documentation of prerequisites in `gcc/doc/install.texi'. 3778 3779 * Details of contributors to that front end in 3780 `gcc/doc/contrib.texi'. If the details are in that front end's 3781 own manual then there should be a link to that manual's list in 3782 `contrib.texi'. 3783 3784 * Information about support for that language in 3785 `gcc/doc/frontends.texi'. 3786 3787 * Information about standards for that language, and the front end's 3788 support for them, in `gcc/doc/standards.texi'. This may be a link 3789 to such information in the front end's own manual. 3790 3791 * Details of source file suffixes for that language and `-x LANG' 3792 options supported, in `gcc/doc/invoke.texi'. 3793 3794 * Entries in `default_compilers' in `gcc.c' for source file suffixes 3795 for that language. 3796 3797 * Preferably testsuites, which may be under `gcc/testsuite' or 3798 runtime library directories. FIXME: document somewhere how to 3799 write testsuite harnesses. 3800 3801 * Probably a runtime library for the language, outside the `gcc' 3802 directory. FIXME: document this further. 3803 3804 * Details of the directories of any runtime libraries in 3805 `gcc/doc/sourcebuild.texi'. 3806 3807 If the front end is added to the official GCC source repository, the 3808 following are also necessary: 3809 3810 * At least one Bugzilla component for bugs in that front end and 3811 runtime libraries. This category needs to be mentioned in 3812 `gcc/gccbug.in', as well as being added to the Bugzilla database. 3813 3814 * Normally, one or more maintainers of that front end listed in 3815 `MAINTAINERS'. 3816 3817 * Mentions on the GCC web site in `index.html' and `frontends.html', 3818 with any relevant links on `readings.html'. (Front ends that are 3819 not an official part of GCC may also be listed on 3820 `frontends.html', with relevant links.) 3821 3822 * A news item on `index.html', and possibly an announcement on the 3823 <gcc-announce (a] gcc.gnu.org> mailing list. 3824 3825 * The front end's manuals should be mentioned in 3826 `maintainer-scripts/update_web_docs' (*note Texinfo Manuals::) and 3827 the online manuals should be linked to from 3828 `onlinedocs/index.html'. 3829 3830 * Any old releases or CVS repositories of the front end, before its 3831 inclusion in GCC, should be made available on the GCC FTP site 3832 `ftp://gcc.gnu.org/pub/gcc/old-releases/'. 3833 3834 * The release and snapshot script `maintainer-scripts/gcc_release' 3835 should be updated to generate appropriate tarballs for this front 3836 end. The associated `maintainer-scripts/snapshot-README' and 3837 `maintainer-scripts/snapshot-index.html' files should be updated 3838 to list the tarballs and diffs for this front end. 3839 3840 * If this front end includes its own version files that include the 3841 current date, `maintainer-scripts/update_version' should be 3842 updated accordingly. 3843 3844 * Menu: 3845 3846 * Front End Directory:: The front end `LANGUAGE' directory. 3847 * Front End Config:: The front end `config-lang.in' file. 3848 3849 3850 File: gccint.info, Node: Front End Directory, Next: Front End Config, Up: Front End 3851 3852 6.3.8.1 The Front End `LANGUAGE' Directory 3853 .......................................... 3854 3855 A front end `LANGUAGE' directory contains the source files of that 3856 front end (but not of any runtime libraries, which should be outside 3857 the `gcc' directory). This includes documentation, and possibly some 3858 subsidiary programs build alongside the front end. Certain files are 3859 special and other parts of the compiler depend on their names: 3860 3861 `config-lang.in' 3862 This file is required in all language subdirectories. *Note The 3863 Front End `config-lang.in' File: Front End Config, for details of 3864 its contents 3865 3866 `Make-lang.in' 3867 This file is required in all language subdirectories. It contains 3868 targets `LANG.HOOK' (where `LANG' is the setting of `language' in 3869 `config-lang.in') for the following values of `HOOK', and any 3870 other Makefile rules required to build those targets (which may if 3871 necessary use other Makefiles specified in `outputs' in 3872 `config-lang.in', although this is deprecated). It also adds any 3873 testsuite targets that can use the standard rule in 3874 `gcc/Makefile.in' to the variable `lang_checks'. 3875 3876 `all.cross' 3877 `start.encap' 3878 `rest.encap' 3879 FIXME: exactly what goes in each of these targets? 3880 3881 `tags' 3882 Build an `etags' `TAGS' file in the language subdirectory in 3883 the source tree. 3884 3885 `info' 3886 Build info documentation for the front end, in the build 3887 directory. This target is only called by `make bootstrap' if 3888 a suitable version of `makeinfo' is available, so does not 3889 need to check for this, and should fail if an error occurs. 3890 3891 `dvi' 3892 Build DVI documentation for the front end, in the build 3893 directory. This should be done using `$(TEXI2DVI)', with 3894 appropriate `-I' arguments pointing to directories of 3895 included files. 3896 3897 `pdf' 3898 Build PDF documentation for the front end, in the build 3899 directory. This should be done using `$(TEXI2PDF)', with 3900 appropriate `-I' arguments pointing to directories of 3901 included files. 3902 3903 `html' 3904 Build HTML documentation for the front end, in the build 3905 directory. 3906 3907 `man' 3908 Build generated man pages for the front end from Texinfo 3909 manuals (*note Man Page Generation::), in the build 3910 directory. This target is only called if the necessary tools 3911 are available, but should ignore errors so as not to stop the 3912 build if errors occur; man pages are optional and the tools 3913 involved may be installed in a broken way. 3914 3915 `install-common' 3916 Install everything that is part of the front end, apart from 3917 the compiler executables listed in `compilers' in 3918 `config-lang.in'. 3919 3920 `install-info' 3921 Install info documentation for the front end, if it is 3922 present in the source directory. This target should have 3923 dependencies on info files that should be installed. 3924 3925 `install-man' 3926 Install man pages for the front end. This target should 3927 ignore errors. 3928 3929 `install-plugin' 3930 Install headers needed for plugins. 3931 3932 `srcextra' 3933 Copies its dependencies into the source directory. This 3934 generally should be used for generated files such as Bison 3935 output files which are not present in CVS, but should be 3936 included in any release tarballs. This target will be 3937 executed during a bootstrap if 3938 `--enable-generated-files-in-srcdir' was specified as a 3939 `configure' option. 3940 3941 `srcinfo' 3942 `srcman' 3943 Copies its dependencies into the source directory. These 3944 targets will be executed during a bootstrap if 3945 `--enable-generated-files-in-srcdir' was specified as a 3946 `configure' option. 3947 3948 `uninstall' 3949 Uninstall files installed by installing the compiler. This is 3950 currently documented not to be supported, so the hook need 3951 not do anything. 3952 3953 `mostlyclean' 3954 `clean' 3955 `distclean' 3956 `maintainer-clean' 3957 The language parts of the standard GNU `*clean' targets. 3958 *Note Standard Targets for Users: (standards)Standard 3959 Targets, for details of the standard targets. For GCC, 3960 `maintainer-clean' should delete all generated files in the 3961 source directory that are not checked into CVS, but should 3962 not delete anything checked into CVS. 3963 3964 `Make-lang.in' must also define a variable `LANG_OBJS' to a list 3965 of host object files that are used by that language. 3966 3967 `lang.opt' 3968 This file registers the set of switches that the front end accepts 3969 on the command line, and their `--help' text. *Note Options::. 3970 3971 `lang-specs.h' 3972 This file provides entries for `default_compilers' in `gcc.c' 3973 which override the default of giving an error that a compiler for 3974 that language is not installed. 3975 3976 `LANGUAGE-tree.def' 3977 This file, which need not exist, defines any language-specific tree 3978 codes. 3979 3980 3981 File: gccint.info, Node: Front End Config, Prev: Front End Directory, Up: Front End 3982 3983 6.3.8.2 The Front End `config-lang.in' File 3984 ........................................... 3985 3986 Each language subdirectory contains a `config-lang.in' file. In 3987 addition the main directory contains `c-config-lang.in', which contains 3988 limited information for the C language. This file is a shell script 3989 that may define some variables describing the language: 3990 3991 `language' 3992 This definition must be present, and gives the name of the language 3993 for some purposes such as arguments to `--enable-languages'. 3994 3995 `lang_requires' 3996 If defined, this variable lists (space-separated) language front 3997 ends other than C that this front end requires to be enabled (with 3998 the names given being their `language' settings). For example, the 3999 Java front end depends on the C++ front end, so sets 4000 `lang_requires=c++'. 4001 4002 `subdir_requires' 4003 If defined, this variable lists (space-separated) front end 4004 directories other than C that this front end requires to be 4005 present. For example, the Objective-C++ front end uses source 4006 files from the C++ and Objective-C front ends, so sets 4007 `subdir_requires="cp objc"'. 4008 4009 `target_libs' 4010 If defined, this variable lists (space-separated) targets in the 4011 top level `Makefile' to build the runtime libraries for this 4012 language, such as `target-libobjc'. 4013 4014 `lang_dirs' 4015 If defined, this variable lists (space-separated) top level 4016 directories (parallel to `gcc'), apart from the runtime libraries, 4017 that should not be configured if this front end is not built. 4018 4019 `build_by_default' 4020 If defined to `no', this language front end is not built unless 4021 enabled in a `--enable-languages' argument. Otherwise, front ends 4022 are built by default, subject to any special logic in 4023 `configure.ac' (as is present to disable the Ada front end if the 4024 Ada compiler is not already installed). 4025 4026 `boot_language' 4027 If defined to `yes', this front end is built in stage 1 of the 4028 bootstrap. This is only relevant to front ends written in their 4029 own languages. 4030 4031 `compilers' 4032 If defined, a space-separated list of compiler executables that 4033 will be run by the driver. The names here will each end with 4034 `\$(exeext)'. 4035 4036 `outputs' 4037 If defined, a space-separated list of files that should be 4038 generated by `configure' substituting values in them. This 4039 mechanism can be used to create a file `LANGUAGE/Makefile' from 4040 `LANGUAGE/Makefile.in', but this is deprecated, building 4041 everything from the single `gcc/Makefile' is preferred. 4042 4043 `gtfiles' 4044 If defined, a space-separated list of files that should be scanned 4045 by gengtype.c to generate the garbage collection tables and 4046 routines for this language. This excludes the files that are 4047 common to all front ends. *Note Type Information::. 4048 4049 4050 4051 File: gccint.info, Node: Back End, Prev: Front End, Up: gcc Directory 4052 4053 6.3.9 Anatomy of a Target Back End 4054 ---------------------------------- 4055 4056 A back end for a target architecture in GCC has the following parts: 4057 4058 * A directory `MACHINE' under `gcc/config', containing a machine 4059 description `MACHINE.md' file (*note Machine Descriptions: Machine 4060 Desc.), header files `MACHINE.h' and `MACHINE-protos.h' and a 4061 source file `MACHINE.c' (*note Target Description Macros and 4062 Functions: Target Macros.), possibly a target Makefile fragment 4063 `t-MACHINE' (*note The Target Makefile Fragment: Target 4064 Fragment.), and maybe some other files. The names of these files 4065 may be changed from the defaults given by explicit specifications 4066 in `config.gcc'. 4067 4068 * If necessary, a file `MACHINE-modes.def' in the `MACHINE' 4069 directory, containing additional machine modes to represent 4070 condition codes. *Note Condition Code::, for further details. 4071 4072 * An optional `MACHINE.opt' file in the `MACHINE' directory, 4073 containing a list of target-specific options. You can also add 4074 other option files using the `extra_options' variable in 4075 `config.gcc'. *Note Options::. 4076 4077 * Entries in `config.gcc' (*note The `config.gcc' File: System 4078 Config.) for the systems with this target architecture. 4079 4080 * Documentation in `gcc/doc/invoke.texi' for any command-line 4081 options supported by this target (*note Run-time Target 4082 Specification: Run-time Target.). This means both entries in the 4083 summary table of options and details of the individual options. 4084 4085 * Documentation in `gcc/doc/extend.texi' for any target-specific 4086 attributes supported (*note Defining target-specific uses of 4087 `__attribute__': Target Attributes.), including where the same 4088 attribute is already supported on some targets, which are 4089 enumerated in the manual. 4090 4091 * Documentation in `gcc/doc/extend.texi' for any target-specific 4092 pragmas supported. 4093 4094 * Documentation in `gcc/doc/extend.texi' of any target-specific 4095 built-in functions supported. 4096 4097 * Documentation in `gcc/doc/extend.texi' of any target-specific 4098 format checking styles supported. 4099 4100 * Documentation in `gcc/doc/md.texi' of any target-specific 4101 constraint letters (*note Constraints for Particular Machines: 4102 Machine Constraints.). 4103 4104 * A note in `gcc/doc/contrib.texi' under the person or people who 4105 contributed the target support. 4106 4107 * Entries in `gcc/doc/install.texi' for all target triplets 4108 supported with this target architecture, giving details of any 4109 special notes about installation for this target, or saying that 4110 there are no special notes if there are none. 4111 4112 * Possibly other support outside the `gcc' directory for runtime 4113 libraries. FIXME: reference docs for this. The libstdc++ porting 4114 manual needs to be installed as info for this to work, or to be a 4115 chapter of this manual. 4116 4117 If the back end is added to the official GCC source repository, the 4118 following are also necessary: 4119 4120 * An entry for the target architecture in `readings.html' on the GCC 4121 web site, with any relevant links. 4122 4123 * Details of the properties of the back end and target architecture 4124 in `backends.html' on the GCC web site. 4125 4126 * A news item about the contribution of support for that target 4127 architecture, in `index.html' on the GCC web site. 4128 4129 * Normally, one or more maintainers of that target listed in 4130 `MAINTAINERS'. Some existing architectures may be unmaintained, 4131 but it would be unusual to add support for a target that does not 4132 have a maintainer when support is added. 4133 4134 4135 File: gccint.info, Node: Testsuites, Prev: gcc Directory, Up: Source Tree 4136 4137 6.4 Testsuites 4138 ============== 4139 4140 GCC contains several testsuites to help maintain compiler quality. 4141 Most of the runtime libraries and language front ends in GCC have 4142 testsuites. Currently only the C language testsuites are documented 4143 here; FIXME: document the others. 4144 4145 * Menu: 4146 4147 * Test Idioms:: Idioms used in testsuite code. 4148 * Test Directives:: Directives used within DejaGnu tests. 4149 * Ada Tests:: The Ada language testsuites. 4150 * C Tests:: The C language testsuites. 4151 * libgcj Tests:: The Java library testsuites. 4152 * gcov Testing:: Support for testing gcov. 4153 * profopt Testing:: Support for testing profile-directed optimizations. 4154 * compat Testing:: Support for testing binary compatibility. 4155 * Torture Tests:: Support for torture testing using multiple options. 4156 4157 4158 File: gccint.info, Node: Test Idioms, Next: Test Directives, Up: Testsuites 4159 4160 6.4.1 Idioms Used in Testsuite Code 4161 ----------------------------------- 4162 4163 In general, C testcases have a trailing `-N.c', starting with `-1.c', 4164 in case other testcases with similar names are added later. If the 4165 test is a test of some well-defined feature, it should have a name 4166 referring to that feature such as `FEATURE-1.c'. If it does not test a 4167 well-defined feature but just happens to exercise a bug somewhere in 4168 the compiler, and a bug report has been filed for this bug in the GCC 4169 bug database, `prBUG-NUMBER-1.c' is the appropriate form of name. 4170 Otherwise (for miscellaneous bugs not filed in the GCC bug database), 4171 and previously more generally, test cases are named after the date on 4172 which they were added. This allows people to tell at a glance whether 4173 a test failure is because of a recently found bug that has not yet been 4174 fixed, or whether it may be a regression, but does not give any other 4175 information about the bug or where discussion of it may be found. Some 4176 other language testsuites follow similar conventions. 4177 4178 In the `gcc.dg' testsuite, it is often necessary to test that an error 4179 is indeed a hard error and not just a warning--for example, where it is 4180 a constraint violation in the C standard, which must become an error 4181 with `-pedantic-errors'. The following idiom, where the first line 4182 shown is line LINE of the file and the line that generates the error, 4183 is used for this: 4184 4185 /* { dg-bogus "warning" "warning in place of error" } */ 4186 /* { dg-error "REGEXP" "MESSAGE" { target *-*-* } LINE } */ 4187 4188 It may be necessary to check that an expression is an integer constant 4189 expression and has a certain value. To check that `E' has value `V', 4190 an idiom similar to the following is used: 4191 4192 char x[((E) == (V) ? 1 : -1)]; 4193 4194 In `gcc.dg' tests, `__typeof__' is sometimes used to make assertions 4195 about the types of expressions. See, for example, 4196 `gcc.dg/c99-condexpr-1.c'. The more subtle uses depend on the exact 4197 rules for the types of conditional expressions in the C standard; see, 4198 for example, `gcc.dg/c99-intconst-1.c'. 4199 4200 It is useful to be able to test that optimizations are being made 4201 properly. This cannot be done in all cases, but it can be done where 4202 the optimization will lead to code being optimized away (for example, 4203 where flow analysis or alias analysis should show that certain code 4204 cannot be called) or to functions not being called because they have 4205 been expanded as built-in functions. Such tests go in 4206 `gcc.c-torture/execute'. Where code should be optimized away, a call 4207 to a nonexistent function such as `link_failure ()' may be inserted; a 4208 definition 4209 4210 #ifndef __OPTIMIZE__ 4211 void 4212 link_failure (void) 4213 { 4214 abort (); 4215 } 4216 #endif 4217 4218 will also be needed so that linking still succeeds when the test is run 4219 without optimization. When all calls to a built-in function should 4220 have been optimized and no calls to the non-built-in version of the 4221 function should remain, that function may be defined as `static' to 4222 call `abort ()' (although redeclaring a function as static may not work 4223 on all targets). 4224 4225 All testcases must be portable. Target-specific testcases must have 4226 appropriate code to avoid causing failures on unsupported systems; 4227 unfortunately, the mechanisms for this differ by directory. 4228 4229 FIXME: discuss non-C testsuites here. 4230 4231 4232 File: gccint.info, Node: Test Directives, Next: Ada Tests, Prev: Test Idioms, Up: Testsuites 4233 4234 6.4.2 Directives used within DejaGnu tests 4235 ------------------------------------------ 4236 4237 Test directives appear within comments in a test source file and begin 4238 with `dg-'. Some of these are defined within DejaGnu and others are 4239 local to the GCC testsuite. 4240 4241 The order in which test directives appear in a test can be important: 4242 directives local to GCC sometimes override information used by the 4243 DejaGnu directives, which know nothing about the GCC directives, so the 4244 DejaGnu directives must precede GCC directives. 4245 4246 Several test directives include selectors which are usually preceded by 4247 the keyword `target' or `xfail'. A selector is: one or more target 4248 triplets, possibly including wildcard characters; a single 4249 effective-target keyword; or a logical expression. Depending on the 4250 context, the selector specifies whether a test is skipped and reported 4251 as unsupported or is expected to fail. Use `*-*-*' to match any target. 4252 Effective-target keywords are defined in `target-supports.exp' in the 4253 GCC testsuite. 4254 4255 A selector expression appears within curly braces and uses a single 4256 logical operator: one of `!', `&&', or `||'. An operand is another 4257 selector expression, an effective-target keyword, a single target 4258 triplet, or a list of target triplets within quotes or curly braces. 4259 For example: 4260 4261 { target { ! "hppa*-*-* ia64*-*-*" } } 4262 { target { powerpc*-*-* && lp64 } } 4263 { xfail { lp64 || vect_no_align } } 4264 4265 `{ dg-do DO-WHAT-KEYWORD [{ target/xfail SELECTOR }] }' 4266 DO-WHAT-KEYWORD specifies how the test is compiled and whether it 4267 is executed. It is one of: 4268 4269 `preprocess' 4270 Compile with `-E' to run only the preprocessor. 4271 4272 `compile' 4273 Compile with `-S' to produce an assembly code file. 4274 4275 `assemble' 4276 Compile with `-c' to produce a relocatable object file. 4277 4278 `link' 4279 Compile, assemble, and link to produce an executable file. 4280 4281 `run' 4282 Produce and run an executable file, which is expected to 4283 return an exit code of 0. 4284 4285 The default is `compile'. That can be overridden for a set of 4286 tests by redefining `dg-do-what-default' within the `.exp' file 4287 for those tests. 4288 4289 If the directive includes the optional `{ target SELECTOR }' then 4290 the test is skipped unless the target system is included in the 4291 list of target triplets or matches the effective-target keyword. 4292 4293 If `do-what-keyword' is `run' and the directive includes the 4294 optional `{ xfail SELECTOR }' and the selector is met then the 4295 test is expected to fail. The `xfail' clause is ignored for other 4296 values of `do-what-keyword'; those tests can use directive 4297 `dg-xfail-if'. 4298 4299 `{ dg-options OPTIONS [{ target SELECTOR }] }' 4300 This DejaGnu directive provides a list of compiler options, to be 4301 used if the target system matches SELECTOR, that replace the 4302 default options used for this set of tests. 4303 4304 `{ dg-add-options FEATURE ... }' 4305 Add any compiler options that are needed to access certain 4306 features. This directive does nothing on targets that enable the 4307 features by default, or that don't provide them at all. It must 4308 come after all `dg-options' directives. 4309 4310 The supported values of FEATURE are: 4311 `c99_runtime' 4312 The target's C99 runtime (both headers and libraries). 4313 4314 `mips16_attribute' 4315 `mips16' function attributes. Only MIPS targets support this 4316 feature, and only then in certain modes. 4317 4318 `{ dg-timeout N [{target SELECTOR }] }' 4319 Set the time limit for the compilation and for the execution of 4320 the test to the specified number of seconds. 4321 4322 `{ dg-timeout-factor X [{ target SELECTOR }] }' 4323 Multiply the normal time limit for compilation and execution of 4324 the test by the specified floating-point factor. The normal 4325 timeout limit, in seconds, is found by searching the following in 4326 order: 4327 4328 * the value defined by an earlier `dg-timeout' directive in the 4329 test 4330 4331 * variable TOOL_TIMEOUT defined by the set of tests 4332 4333 * GCC,TIMEOUT set in the target board 4334 4335 * 300 4336 4337 `{ dg-skip-if COMMENT { SELECTOR } { INCLUDE-OPTS } { EXCLUDE-OPTS } }' 4338 Skip the test if the test system is included in SELECTOR and if 4339 each of the options in INCLUDE-OPTS is in the set of options with 4340 which the test would be compiled and if none of the options in 4341 EXCLUDE-OPTS is in the set of options with which the test would be 4342 compiled. 4343 4344 Use `"*"' for an empty INCLUDE-OPTS list and `""' for an empty 4345 EXCLUDE-OPTS list. 4346 4347 `{ dg-xfail-if COMMENT { SELECTOR } { INCLUDE-OPTS } { EXCLUDE-OPTS } }' 4348 Expect the test to fail if the conditions (which are the same as 4349 for `dg-skip-if') are met. This does not affect the execute step. 4350 4351 `{ dg-xfail-run-if COMMENT { SELECTOR } { INCLUDE-OPTS } { EXCLUDE-OPTS } }' 4352 Expect the execute step of a test to fail if the conditions (which 4353 are the same as for `dg-skip-if') and `dg-xfail-if') are met. 4354 4355 `{ dg-require-SUPPORT args }' 4356 Skip the test if the target does not provide the required support; 4357 see `gcc-dg.exp' in the GCC testsuite for the actual directives. 4358 These directives must appear after any `dg-do' directive in the 4359 test and before any `dg-additional-sources' directive. They 4360 require at least one argument, which can be an empty string if the 4361 specific procedure does not examine the argument. 4362 4363 `{ dg-require-effective-target KEYWORD }' 4364 Skip the test if the test target, including current multilib flags, 4365 is not covered by the effective-target keyword. This directive 4366 must appear after any `dg-do' directive in the test and before any 4367 `dg-additional-sources' directive. 4368 4369 `{ dg-shouldfail COMMENT { SELECTOR } { INCLUDE-OPTS } { EXCLUDE-OPTS } }' 4370 Expect the test executable to return a nonzero exit status if the 4371 conditions (which are the same as for `dg-skip-if') are met. 4372 4373 `{ dg-error REGEXP [COMMENT [{ target/xfail SELECTOR } [LINE] }]] }' 4374 This DejaGnu directive appears on a source line that is expected 4375 to get an error message, or else specifies the source line 4376 associated with the message. If there is no message for that line 4377 or if the text of that message is not matched by REGEXP then the 4378 check fails and COMMENT is included in the `FAIL' message. The 4379 check does not look for the string `"error"' unless it is part of 4380 REGEXP. 4381 4382 `{ dg-warning REGEXP [COMMENT [{ target/xfail SELECTOR } [LINE] }]] }' 4383 This DejaGnu directive appears on a source line that is expected 4384 to get a warning message, or else specifies the source line 4385 associated with the message. If there is no message for that line 4386 or if the text of that message is not matched by REGEXP then the 4387 check fails and COMMENT is included in the `FAIL' message. The 4388 check does not look for the string `"warning"' unless it is part 4389 of REGEXP. 4390 4391 `{ dg-message REGEXP [COMMENT [{ target/xfail SELECTOR } [LINE] }]] }' 4392 The line is expected to get a message other than an error or 4393 warning. If there is no message for that line or if the text of 4394 that message is not matched by REGEXP then the check fails and 4395 COMMENT is included in the `FAIL' message. 4396 4397 `{ dg-bogus REGEXP [COMMENT [{ target/xfail SELECTOR } [LINE] }]] }' 4398 This DejaGnu directive appears on a source line that should not 4399 get a message matching REGEXP, or else specifies the source line 4400 associated with the bogus message. It is usually used with `xfail' 4401 to indicate that the message is a known problem for a particular 4402 set of targets. 4403 4404 `{ dg-excess-errors COMMENT [{ target/xfail SELECTOR }] }' 4405 This DejaGnu directive indicates that the test is expected to fail 4406 due to compiler messages that are not handled by `dg-error', 4407 `dg-warning' or `dg-bogus'. For this directive `xfail' has the 4408 same effect as `target'. 4409 4410 `{ dg-output REGEXP [{ target/xfail SELECTOR }] }' 4411 This DejaGnu directive compares REGEXP to the combined output that 4412 the test executable writes to `stdout' and `stderr'. 4413 4414 `{ dg-prune-output REGEXP }' 4415 Prune messages matching REGEXP from test output. 4416 4417 `{ dg-additional-files "FILELIST" }' 4418 Specify additional files, other than source files, that must be 4419 copied to the system where the compiler runs. 4420 4421 `{ dg-additional-sources "FILELIST" }' 4422 Specify additional source files to appear in the compile line 4423 following the main test file. 4424 4425 `{ dg-final { LOCAL-DIRECTIVE } }' 4426 This DejaGnu directive is placed within a comment anywhere in the 4427 source file and is processed after the test has been compiled and 4428 run. Multiple `dg-final' commands are processed in the order in 4429 which they appear in the source file. 4430 4431 The GCC testsuite defines the following directives to be used 4432 within `dg-final'. 4433 4434 `cleanup-coverage-files' 4435 Removes coverage data files generated for this test. 4436 4437 `cleanup-repo-files' 4438 Removes files generated for this test for `-frepo'. 4439 4440 `cleanup-rtl-dump SUFFIX' 4441 Removes RTL dump files generated for this test. 4442 4443 `cleanup-tree-dump SUFFIX' 4444 Removes tree dump files matching SUFFIX which were generated 4445 for this test. 4446 4447 `cleanup-saved-temps' 4448 Removes files for the current test which were kept for 4449 `--save-temps'. 4450 4451 `scan-file FILENAME REGEXP [{ target/xfail SELECTOR }]' 4452 Passes if REGEXP matches text in FILENAME. 4453 4454 `scan-file-not FILENAME REGEXP [{ target/xfail SELECTOR }]' 4455 Passes if REGEXP does not match text in FILENAME. 4456 4457 `scan-hidden SYMBOL [{ target/xfail SELECTOR }]' 4458 Passes if SYMBOL is defined as a hidden symbol in the test's 4459 assembly output. 4460 4461 `scan-not-hidden SYMBOL [{ target/xfail SELECTOR }]' 4462 Passes if SYMBOL is not defined as a hidden symbol in the 4463 test's assembly output. 4464 4465 `scan-assembler-times REGEX NUM [{ target/xfail SELECTOR }]' 4466 Passes if REGEX is matched exactly NUM times in the test's 4467 assembler output. 4468 4469 `scan-assembler REGEX [{ target/xfail SELECTOR }]' 4470 Passes if REGEX matches text in the test's assembler output. 4471 4472 `scan-assembler-not REGEX [{ target/xfail SELECTOR }]' 4473 Passes if REGEX does not match text in the test's assembler 4474 output. 4475 4476 `scan-assembler-dem REGEX [{ target/xfail SELECTOR }]' 4477 Passes if REGEX matches text in the test's demangled 4478 assembler output. 4479 4480 `scan-assembler-dem-not REGEX [{ target/xfail SELECTOR }]' 4481 Passes if REGEX does not match text in the test's demangled 4482 assembler output. 4483 4484 `scan-tree-dump-times REGEX NUM SUFFIX [{ target/xfail SELECTOR }]' 4485 Passes if REGEX is found exactly NUM times in the dump file 4486 with suffix SUFFIX. 4487 4488 `scan-tree-dump REGEX SUFFIX [{ target/xfail SELECTOR }]' 4489 Passes if REGEX matches text in the dump file with suffix 4490 SUFFIX. 4491 4492 `scan-tree-dump-not REGEX SUFFIX [{ target/xfail SELECTOR }]' 4493 Passes if REGEX does not match text in the dump file with 4494 suffix SUFFIX. 4495 4496 `scan-tree-dump-dem REGEX SUFFIX [{ target/xfail SELECTOR }]' 4497 Passes if REGEX matches demangled text in the dump file with 4498 suffix SUFFIX. 4499 4500 `scan-tree-dump-dem-not REGEX SUFFIX [{ target/xfail SELECTOR }]' 4501 Passes if REGEX does not match demangled text in the dump 4502 file with suffix SUFFIX. 4503 4504 `output-exists [{ target/xfail SELECTOR }]' 4505 Passes if compiler output file exists. 4506 4507 `output-exists-not [{ target/xfail SELECTOR }]' 4508 Passes if compiler output file does not exist. 4509 4510 `run-gcov SOURCEFILE' 4511 Check line counts in `gcov' tests. 4512 4513 `run-gcov [branches] [calls] { OPTS SOURCEFILE }' 4514 Check branch and/or call counts, in addition to line counts, 4515 in `gcov' tests. 4516 4517 4518 File: gccint.info, Node: Ada Tests, Next: C Tests, Prev: Test Directives, Up: Testsuites 4519 4520 6.4.3 Ada Language Testsuites 4521 ----------------------------- 4522 4523 The Ada testsuite includes executable tests from the ACATS 2.5 4524 testsuite, publicly available at 4525 `http://www.adaic.org/compilers/acats/2.5' 4526 4527 These tests are integrated in the GCC testsuite in the 4528 `gcc/testsuite/ada/acats' directory, and enabled automatically when 4529 running `make check', assuming the Ada language has been enabled when 4530 configuring GCC. 4531 4532 You can also run the Ada testsuite independently, using `make 4533 check-ada', or run a subset of the tests by specifying which chapter to 4534 run, e.g.: 4535 4536 $ make check-ada CHAPTERS="c3 c9" 4537 4538 The tests are organized by directory, each directory corresponding to 4539 a chapter of the Ada Reference Manual. So for example, c9 corresponds 4540 to chapter 9, which deals with tasking features of the language. 4541 4542 There is also an extra chapter called `gcc' containing a template for 4543 creating new executable tests. 4544 4545 The tests are run using two `sh' scripts: `run_acats' and 4546 `run_all.sh'. To run the tests using a simulator or a cross target, 4547 see the small customization section at the top of `run_all.sh'. 4548 4549 These tests are run using the build tree: they can be run without doing 4550 a `make install'. 4551 4552 4553 File: gccint.info, Node: C Tests, Next: libgcj Tests, Prev: Ada Tests, Up: Testsuites 4554 4555 6.4.4 C Language Testsuites 4556 --------------------------- 4557 4558 GCC contains the following C language testsuites, in the 4559 `gcc/testsuite' directory: 4560 4561 `gcc.dg' 4562 This contains tests of particular features of the C compiler, 4563 using the more modern `dg' harness. Correctness tests for various 4564 compiler features should go here if possible. 4565 4566 Magic comments determine whether the file is preprocessed, 4567 compiled, linked or run. In these tests, error and warning 4568 message texts are compared against expected texts or regular 4569 expressions given in comments. These tests are run with the 4570 options `-ansi -pedantic' unless other options are given in the 4571 test. Except as noted below they are not run with multiple 4572 optimization options. 4573 4574 `gcc.dg/compat' 4575 This subdirectory contains tests for binary compatibility using 4576 `compat.exp', which in turn uses the language-independent support 4577 (*note Support for testing binary compatibility: compat Testing.). 4578 4579 `gcc.dg/cpp' 4580 This subdirectory contains tests of the preprocessor. 4581 4582 `gcc.dg/debug' 4583 This subdirectory contains tests for debug formats. Tests in this 4584 subdirectory are run for each debug format that the compiler 4585 supports. 4586 4587 `gcc.dg/format' 4588 This subdirectory contains tests of the `-Wformat' format 4589 checking. Tests in this directory are run with and without 4590 `-DWIDE'. 4591 4592 `gcc.dg/noncompile' 4593 This subdirectory contains tests of code that should not compile 4594 and does not need any special compilation options. They are run 4595 with multiple optimization options, since sometimes invalid code 4596 crashes the compiler with optimization. 4597 4598 `gcc.dg/special' 4599 FIXME: describe this. 4600 4601 `gcc.c-torture' 4602 This contains particular code fragments which have historically 4603 broken easily. These tests are run with multiple optimization 4604 options, so tests for features which only break at some 4605 optimization levels belong here. This also contains tests to 4606 check that certain optimizations occur. It might be worthwhile to 4607 separate the correctness tests cleanly from the code quality 4608 tests, but it hasn't been done yet. 4609 4610 `gcc.c-torture/compat' 4611 FIXME: describe this. 4612 4613 This directory should probably not be used for new tests. 4614 4615 `gcc.c-torture/compile' 4616 This testsuite contains test cases that should compile, but do not 4617 need to link or run. These test cases are compiled with several 4618 different combinations of optimization options. All warnings are 4619 disabled for these test cases, so this directory is not suitable if 4620 you wish to test for the presence or absence of compiler warnings. 4621 While special options can be set, and tests disabled on specific 4622 platforms, by the use of `.x' files, mostly these test cases 4623 should not contain platform dependencies. FIXME: discuss how 4624 defines such as `NO_LABEL_VALUES' and `STACK_SIZE' are used. 4625 4626 `gcc.c-torture/execute' 4627 This testsuite contains test cases that should compile, link and 4628 run; otherwise the same comments as for `gcc.c-torture/compile' 4629 apply. 4630 4631 `gcc.c-torture/execute/ieee' 4632 This contains tests which are specific to IEEE floating point. 4633 4634 `gcc.c-torture/unsorted' 4635 FIXME: describe this. 4636 4637 This directory should probably not be used for new tests. 4638 4639 `gcc.c-torture/misc-tests' 4640 This directory contains C tests that require special handling. 4641 Some of these tests have individual expect files, and others share 4642 special-purpose expect files: 4643 4644 ``bprob*.c'' 4645 Test `-fbranch-probabilities' using `bprob.exp', which in 4646 turn uses the generic, language-independent framework (*note 4647 Support for testing profile-directed optimizations: profopt 4648 Testing.). 4649 4650 ``dg-*.c'' 4651 Test the testsuite itself using `dg-test.exp'. 4652 4653 ``gcov*.c'' 4654 Test `gcov' output using `gcov.exp', which in turn uses the 4655 language-independent support (*note Support for testing gcov: 4656 gcov Testing.). 4657 4658 ``i386-pf-*.c'' 4659 Test i386-specific support for data prefetch using 4660 `i386-prefetch.exp'. 4661 4662 4663 FIXME: merge in `testsuite/README.gcc' and discuss the format of test 4664 cases and magic comments more. 4665 4666 4667 File: gccint.info, Node: libgcj Tests, Next: gcov Testing, Prev: C Tests, Up: Testsuites 4668 4669 6.4.5 The Java library testsuites. 4670 ---------------------------------- 4671 4672 Runtime tests are executed via `make check' in the 4673 `TARGET/libjava/testsuite' directory in the build tree. Additional 4674 runtime tests can be checked into this testsuite. 4675 4676 Regression testing of the core packages in libgcj is also covered by 4677 the Mauve testsuite. The Mauve Project develops tests for the Java 4678 Class Libraries. These tests are run as part of libgcj testing by 4679 placing the Mauve tree within the libjava testsuite sources at 4680 `libjava/testsuite/libjava.mauve/mauve', or by specifying the location 4681 of that tree when invoking `make', as in `make MAUVEDIR=~/mauve check'. 4682 4683 To detect regressions, a mechanism in `mauve.exp' compares the 4684 failures for a test run against the list of expected failures in 4685 `libjava/testsuite/libjava.mauve/xfails' from the source hierarchy. 4686 Update this file when adding new failing tests to Mauve, or when fixing 4687 bugs in libgcj that had caused Mauve test failures. 4688 4689 We encourage developers to contribute test cases to Mauve. 4690 4691 4692 File: gccint.info, Node: gcov Testing, Next: profopt Testing, Prev: libgcj Tests, Up: Testsuites 4693 4694 6.4.6 Support for testing `gcov' 4695 -------------------------------- 4696 4697 Language-independent support for testing `gcov', and for checking that 4698 branch profiling produces expected values, is provided by the expect 4699 file `gcov.exp'. `gcov' tests also rely on procedures in `gcc.dg.exp' 4700 to compile and run the test program. A typical `gcov' test contains 4701 the following DejaGnu commands within comments: 4702 4703 { dg-options "-fprofile-arcs -ftest-coverage" } 4704 { dg-do run { target native } } 4705 { dg-final { run-gcov sourcefile } } 4706 4707 Checks of `gcov' output can include line counts, branch percentages, 4708 and call return percentages. All of these checks are requested via 4709 commands that appear in comments in the test's source file. Commands 4710 to check line counts are processed by default. Commands to check 4711 branch percentages and call return percentages are processed if the 4712 `run-gcov' command has arguments `branches' or `calls', respectively. 4713 For example, the following specifies checking both, as well as passing 4714 `-b' to `gcov': 4715 4716 { dg-final { run-gcov branches calls { -b sourcefile } } } 4717 4718 A line count command appears within a comment on the source line that 4719 is expected to get the specified count and has the form `count(CNT)'. 4720 A test should only check line counts for lines that will get the same 4721 count for any architecture. 4722 4723 Commands to check branch percentages (`branch') and call return 4724 percentages (`returns') are very similar to each other. A beginning 4725 command appears on or before the first of a range of lines that will 4726 report the percentage, and the ending command follows that range of 4727 lines. The beginning command can include a list of percentages, all of 4728 which are expected to be found within the range. A range is terminated 4729 by the next command of the same kind. A command `branch(end)' or 4730 `returns(end)' marks the end of a range without starting a new one. 4731 For example: 4732 4733 if (i > 10 && j > i && j < 20) /* branch(27 50 75) */ 4734 /* branch(end) */ 4735 foo (i, j); 4736 4737 For a call return percentage, the value specified is the percentage of 4738 calls reported to return. For a branch percentage, the value is either 4739 the expected percentage or 100 minus that value, since the direction of 4740 a branch can differ depending on the target or the optimization level. 4741 4742 Not all branches and calls need to be checked. A test should not 4743 check for branches that might be optimized away or replaced with 4744 predicated instructions. Don't check for calls inserted by the 4745 compiler or ones that might be inlined or optimized away. 4746 4747 A single test can check for combinations of line counts, branch 4748 percentages, and call return percentages. The command to check a line 4749 count must appear on the line that will report that count, but commands 4750 to check branch percentages and call return percentages can bracket the 4751 lines that report them. 4752 4753 4754 File: gccint.info, Node: profopt Testing, Next: compat Testing, Prev: gcov Testing, Up: Testsuites 4755 4756 6.4.7 Support for testing profile-directed optimizations 4757 -------------------------------------------------------- 4758 4759 The file `profopt.exp' provides language-independent support for 4760 checking correct execution of a test built with profile-directed 4761 optimization. This testing requires that a test program be built and 4762 executed twice. The first time it is compiled to generate profile 4763 data, and the second time it is compiled to use the data that was 4764 generated during the first execution. The second execution is to 4765 verify that the test produces the expected results. 4766 4767 To check that the optimization actually generated better code, a test 4768 can be built and run a third time with normal optimizations to verify 4769 that the performance is better with the profile-directed optimizations. 4770 `profopt.exp' has the beginnings of this kind of support. 4771 4772 `profopt.exp' provides generic support for profile-directed 4773 optimizations. Each set of tests that uses it provides information 4774 about a specific optimization: 4775 4776 `tool' 4777 tool being tested, e.g., `gcc' 4778 4779 `profile_option' 4780 options used to generate profile data 4781 4782 `feedback_option' 4783 options used to optimize using that profile data 4784 4785 `prof_ext' 4786 suffix of profile data files 4787 4788 `PROFOPT_OPTIONS' 4789 list of options with which to run each test, similar to the lists 4790 for torture tests 4791 4792 4793 File: gccint.info, Node: compat Testing, Next: Torture Tests, Prev: profopt Testing, Up: Testsuites 4794 4795 6.4.8 Support for testing binary compatibility 4796 ---------------------------------------------- 4797 4798 The file `compat.exp' provides language-independent support for binary 4799 compatibility testing. It supports testing interoperability of two 4800 compilers that follow the same ABI, or of multiple sets of compiler 4801 options that should not affect binary compatibility. It is intended to 4802 be used for testsuites that complement ABI testsuites. 4803 4804 A test supported by this framework has three parts, each in a separate 4805 source file: a main program and two pieces that interact with each 4806 other to split up the functionality being tested. 4807 4808 `TESTNAME_main.SUFFIX' 4809 Contains the main program, which calls a function in file 4810 `TESTNAME_x.SUFFIX'. 4811 4812 `TESTNAME_x.SUFFIX' 4813 Contains at least one call to a function in `TESTNAME_y.SUFFIX'. 4814 4815 `TESTNAME_y.SUFFIX' 4816 Shares data with, or gets arguments from, `TESTNAME_x.SUFFIX'. 4817 4818 Within each test, the main program and one functional piece are 4819 compiled by the GCC under test. The other piece can be compiled by an 4820 alternate compiler. If no alternate compiler is specified, then all 4821 three source files are all compiled by the GCC under test. You can 4822 specify pairs of sets of compiler options. The first element of such a 4823 pair specifies options used with the GCC under test, and the second 4824 element of the pair specifies options used with the alternate compiler. 4825 Each test is compiled with each pair of options. 4826 4827 `compat.exp' defines default pairs of compiler options. These can be 4828 overridden by defining the environment variable `COMPAT_OPTIONS' as: 4829 4830 COMPAT_OPTIONS="[list [list {TST1} {ALT1}] 4831 ...[list {TSTN} {ALTN}]]" 4832 4833 where TSTI and ALTI are lists of options, with TSTI used by the 4834 compiler under test and ALTI used by the alternate compiler. For 4835 example, with `[list [list {-g -O0} {-O3}] [list {-fpic} {-fPIC -O2}]]', 4836 the test is first built with `-g -O0' by the compiler under test and 4837 with `-O3' by the alternate compiler. The test is built a second time 4838 using `-fpic' by the compiler under test and `-fPIC -O2' by the 4839 alternate compiler. 4840 4841 An alternate compiler is specified by defining an environment variable 4842 to be the full pathname of an installed compiler; for C define 4843 `ALT_CC_UNDER_TEST', and for C++ define `ALT_CXX_UNDER_TEST'. These 4844 will be written to the `site.exp' file used by DejaGnu. The default is 4845 to build each test with the compiler under test using the first of each 4846 pair of compiler options from `COMPAT_OPTIONS'. When 4847 `ALT_CC_UNDER_TEST' or `ALT_CXX_UNDER_TEST' is `same', each test is 4848 built using the compiler under test but with combinations of the 4849 options from `COMPAT_OPTIONS'. 4850 4851 To run only the C++ compatibility suite using the compiler under test 4852 and another version of GCC using specific compiler options, do the 4853 following from `OBJDIR/gcc': 4854 4855 rm site.exp 4856 make -k \ 4857 ALT_CXX_UNDER_TEST=${alt_prefix}/bin/g++ \ 4858 COMPAT_OPTIONS="lists as shown above" \ 4859 check-c++ \ 4860 RUNTESTFLAGS="compat.exp" 4861 4862 A test that fails when the source files are compiled with different 4863 compilers, but passes when the files are compiled with the same 4864 compiler, demonstrates incompatibility of the generated code or runtime 4865 support. A test that fails for the alternate compiler but passes for 4866 the compiler under test probably tests for a bug that was fixed in the 4867 compiler under test but is present in the alternate compiler. 4868 4869 The binary compatibility tests support a small number of test framework 4870 commands that appear within comments in a test file. 4871 4872 `dg-require-*' 4873 These commands can be used in `TESTNAME_main.SUFFIX' to skip the 4874 test if specific support is not available on the target. 4875 4876 `dg-options' 4877 The specified options are used for compiling this particular source 4878 file, appended to the options from `COMPAT_OPTIONS'. When this 4879 command appears in `TESTNAME_main.SUFFIX' the options are also 4880 used to link the test program. 4881 4882 `dg-xfail-if' 4883 This command can be used in a secondary source file to specify that 4884 compilation is expected to fail for particular options on 4885 particular targets. 4886 4887 4888 File: gccint.info, Node: Torture Tests, Prev: compat Testing, Up: Testsuites 4889 4890 6.4.9 Support for torture testing using multiple options 4891 -------------------------------------------------------- 4892 4893 Throughout the compiler testsuite there are several directories whose 4894 tests are run multiple times, each with a different set of options. 4895 These are known as torture tests. 4896 `gcc/testsuite/lib/torture-options.exp' defines procedures to set up 4897 these lists: 4898 4899 `torture-init' 4900 Initialize use of torture lists. 4901 4902 `set-torture-options' 4903 Set lists of torture options to use for tests with and without 4904 loops. Optionally combine a set of torture options with a set of 4905 other options, as is done with Objective-C runtime options. 4906 4907 `torture-finish' 4908 Finalize use of torture lists. 4909 4910 The `.exp' file for a set of tests that use torture options must 4911 include calls to these three procedures if: 4912 4913 * It calls `gcc-dg-runtest' and overrides DG_TORTURE_OPTIONS. 4914 4915 * It calls ${TOOL}`-torture' or ${TOOL}`-torture-execute', where 4916 TOOL is `c', `fortran', or `objc'. 4917 4918 * It calls `dg-pch'. 4919 4920 It is not necessary for a `.exp' file that calls `gcc-dg-runtest' to 4921 call the torture procedures if the tests should use the list in 4922 DG_TORTURE_OPTIONS defined in `gcc-dg.exp'. 4923 4924 Most uses of torture options can override the default lists by defining 4925 TORTURE_OPTIONS or add to the default list by defining 4926 ADDITIONAL_TORTURE_OPTIONS. Define these in a `.dejagnurc' file or add 4927 them to the `site.exp' file; for example 4928 4929 set ADDITIONAL_TORTURE_OPTIONS [list \ 4930 { -O2 -ftree-loop-linear } \ 4931 { -O2 -fpeel-loops } ] 4932 4933 4934 File: gccint.info, Node: Options, Next: Passes, Prev: Source Tree, Up: Top 4935 4936 7 Option specification files 4937 **************************** 4938 4939 Most GCC command-line options are described by special option 4940 definition files, the names of which conventionally end in `.opt'. 4941 This chapter describes the format of these files. 4942 4943 * Menu: 4944 4945 * Option file format:: The general layout of the files 4946 * Option properties:: Supported option properties 4947 4948 4949 File: gccint.info, Node: Option file format, Next: Option properties, Up: Options 4950 4951 7.1 Option file format 4952 ====================== 4953 4954 Option files are a simple list of records in which each field occupies 4955 its own line and in which the records themselves are separated by blank 4956 lines. Comments may appear on their own line anywhere within the file 4957 and are preceded by semicolons. Whitespace is allowed before the 4958 semicolon. 4959 4960 The files can contain the following types of record: 4961 4962 * A language definition record. These records have two fields: the 4963 string `Language' and the name of the language. Once a language 4964 has been declared in this way, it can be used as an option 4965 property. *Note Option properties::. 4966 4967 * A target specific save record to save additional information. These 4968 records have two fields: the string `TargetSave', and a 4969 declaration type to go in the `cl_target_option' structure. 4970 4971 * An option definition record. These records have the following 4972 fields: 4973 1. the name of the option, with the leading "-" removed 4974 4975 2. a space-separated list of option properties (*note Option 4976 properties::) 4977 4978 3. the help text to use for `--help' (omitted if the second field 4979 contains the `Undocumented' property). 4980 4981 By default, all options beginning with "f", "W" or "m" are 4982 implicitly assumed to take a "no-" form. This form should not be 4983 listed separately. If an option beginning with one of these 4984 letters does not have a "no-" form, you can use the 4985 `RejectNegative' property to reject it. 4986 4987 The help text is automatically line-wrapped before being displayed. 4988 Normally the name of the option is printed on the left-hand side of 4989 the output and the help text is printed on the right. However, if 4990 the help text contains a tab character, the text to the left of 4991 the tab is used instead of the option's name and the text to the 4992 right of the tab forms the help text. This allows you to 4993 elaborate on what type of argument the option takes. 4994 4995 * A target mask record. These records have one field of the form 4996 `Mask(X)'. The options-processing script will automatically 4997 allocate a bit in `target_flags' (*note Run-time Target::) for 4998 each mask name X and set the macro `MASK_X' to the appropriate 4999 bitmask. It will also declare a `TARGET_X' macro that has the 5000 value 1 when bit `MASK_X' is set and 0 otherwise. 5001 5002 They are primarily intended to declare target masks that are not 5003 associated with user options, either because these masks represent 5004 internal switches or because the options are not available on all 5005 configurations and yet the masks always need to be defined. 5006 5007 5008 File: gccint.info, Node: Option properties, Prev: Option file format, Up: Options 5009 5010 7.2 Option properties 5011 ===================== 5012 5013 The second field of an option record can specify the following 5014 properties: 5015 5016 `Common' 5017 The option is available for all languages and targets. 5018 5019 `Target' 5020 The option is available for all languages but is target-specific. 5021 5022 `LANGUAGE' 5023 The option is available when compiling for the given language. 5024 5025 It is possible to specify several different languages for the same 5026 option. Each LANGUAGE must have been declared by an earlier 5027 `Language' record. *Note Option file format::. 5028 5029 `RejectNegative' 5030 The option does not have a "no-" form. All options beginning with 5031 "f", "W" or "m" are assumed to have a "no-" form unless this 5032 property is used. 5033 5034 `Negative(OTHERNAME)' 5035 The option will turn off another option OTHERNAME, which is the 5036 the option name with the leading "-" removed. This chain action 5037 will propagate through the `Negative' property of the option to be 5038 turned off. 5039 5040 `Joined' 5041 `Separate' 5042 The option takes a mandatory argument. `Joined' indicates that 5043 the option and argument can be included in the same `argv' entry 5044 (as with `-mflush-func=NAME', for example). `Separate' indicates 5045 that the option and argument can be separate `argv' entries (as 5046 with `-o'). An option is allowed to have both of these properties. 5047 5048 `JoinedOrMissing' 5049 The option takes an optional argument. If the argument is given, 5050 it will be part of the same `argv' entry as the option itself. 5051 5052 This property cannot be used alongside `Joined' or `Separate'. 5053 5054 `UInteger' 5055 The option's argument is a non-negative integer. The option parser 5056 will check and convert the argument before passing it to the 5057 relevant option handler. `UInteger' should also be used on 5058 options like `-falign-loops' where both `-falign-loops' and 5059 `-falign-loops'=N are supported to make sure the saved options are 5060 given a full integer. 5061 5062 `Var(VAR)' 5063 The state of this option should be stored in variable VAR. The 5064 way that the state is stored depends on the type of option: 5065 5066 * If the option uses the `Mask' or `InverseMask' properties, 5067 VAR is the integer variable that contains the mask. 5068 5069 * If the option is a normal on/off switch, VAR is an integer 5070 variable that is nonzero when the option is enabled. The 5071 options parser will set the variable to 1 when the positive 5072 form of the option is used and 0 when the "no-" form is used. 5073 5074 * If the option takes an argument and has the `UInteger' 5075 property, VAR is an integer variable that stores the value of 5076 the argument. 5077 5078 * Otherwise, if the option takes an argument, VAR is a pointer 5079 to the argument string. The pointer will be null if the 5080 argument is optional and wasn't given. 5081 5082 The option-processing script will usually declare VAR in 5083 `options.c' and leave it to be zero-initialized at start-up time. 5084 You can modify this behavior using `VarExists' and `Init'. 5085 5086 `Var(VAR, SET)' 5087 The option controls an integer variable VAR and is active when VAR 5088 equals SET. The option parser will set VAR to SET when the 5089 positive form of the option is used and `!SET' when the "no-" form 5090 is used. 5091 5092 VAR is declared in the same way as for the single-argument form 5093 described above. 5094 5095 `VarExists' 5096 The variable specified by the `Var' property already exists. No 5097 definition should be added to `options.c' in response to this 5098 option record. 5099 5100 You should use this property only if the variable is declared 5101 outside `options.c'. 5102 5103 `Init(VALUE)' 5104 The variable specified by the `Var' property should be statically 5105 initialized to VALUE. 5106 5107 `Mask(NAME)' 5108 The option is associated with a bit in the `target_flags' variable 5109 (*note Run-time Target::) and is active when that bit is set. You 5110 may also specify `Var' to select a variable other than 5111 `target_flags'. 5112 5113 The options-processing script will automatically allocate a unique 5114 bit for the option. If the option is attached to `target_flags', 5115 the script will set the macro `MASK_NAME' to the appropriate 5116 bitmask. It will also declare a `TARGET_NAME' macro that has the 5117 value 1 when the option is active and 0 otherwise. If you use 5118 `Var' to attach the option to a different variable, the associated 5119 macros are called `OPTION_MASK_NAME' and `OPTION_NAME' 5120 respectively. 5121 5122 You can disable automatic bit allocation using `MaskExists'. 5123 5124 `InverseMask(OTHERNAME)' 5125 `InverseMask(OTHERNAME, THISNAME)' 5126 The option is the inverse of another option that has the 5127 `Mask(OTHERNAME)' property. If THISNAME is given, the 5128 options-processing script will declare a `TARGET_THISNAME' macro 5129 that is 1 when the option is active and 0 otherwise. 5130 5131 `MaskExists' 5132 The mask specified by the `Mask' property already exists. No 5133 `MASK' or `TARGET' definitions should be added to `options.h' in 5134 response to this option record. 5135 5136 The main purpose of this property is to support synonymous options. 5137 The first option should use `Mask(NAME)' and the others should use 5138 `Mask(NAME) MaskExists'. 5139 5140 `Report' 5141 The state of the option should be printed by `-fverbose-asm'. 5142 5143 `Undocumented' 5144 The option is deliberately missing documentation and should not be 5145 included in the `--help' output. 5146 5147 `Condition(COND)' 5148 The option should only be accepted if preprocessor condition COND 5149 is true. Note that any C declarations associated with the option 5150 will be present even if COND is false; COND simply controls 5151 whether the option is accepted and whether it is printed in the 5152 `--help' output. 5153 5154 `Save' 5155 Build the `cl_target_option' structure to hold a copy of the 5156 option, add the functions `cl_target_option_save' and 5157 `cl_target_option_restore' to save and restore the options. 5158 5159 5160 File: gccint.info, Node: Passes, Next: Trees, Prev: Options, Up: Top 5161 5162 8 Passes and Files of the Compiler 5163 ********************************** 5164 5165 This chapter is dedicated to giving an overview of the optimization and 5166 code generation passes of the compiler. In the process, it describes 5167 some of the language front end interface, though this description is no 5168 where near complete. 5169 5170 * Menu: 5171 5172 * Parsing pass:: The language front end turns text into bits. 5173 * Gimplification pass:: The bits are turned into something we can optimize. 5174 * Pass manager:: Sequencing the optimization passes. 5175 * Tree SSA passes:: Optimizations on a high-level representation. 5176 * RTL passes:: Optimizations on a low-level representation. 5177 5178 5179 File: gccint.info, Node: Parsing pass, Next: Gimplification pass, Up: Passes 5180 5181 8.1 Parsing pass 5182 ================ 5183 5184 The language front end is invoked only once, via 5185 `lang_hooks.parse_file', to parse the entire input. The language front 5186 end may use any intermediate language representation deemed 5187 appropriate. The C front end uses GENERIC trees (CROSSREF), plus a 5188 double handful of language specific tree codes defined in 5189 `c-common.def'. The Fortran front end uses a completely different 5190 private representation. 5191 5192 At some point the front end must translate the representation used in 5193 the front end to a representation understood by the language-independent 5194 portions of the compiler. Current practice takes one of two forms. 5195 The C front end manually invokes the gimplifier (CROSSREF) on each 5196 function, and uses the gimplifier callbacks to convert the 5197 language-specific tree nodes directly to GIMPLE (CROSSREF) before 5198 passing the function off to be compiled. The Fortran front end 5199 converts from a private representation to GENERIC, which is later 5200 lowered to GIMPLE when the function is compiled. Which route to choose 5201 probably depends on how well GENERIC (plus extensions) can be made to 5202 match up with the source language and necessary parsing data structures. 5203 5204 BUG: Gimplification must occur before nested function lowering, and 5205 nested function lowering must be done by the front end before passing 5206 the data off to cgraph. 5207 5208 TODO: Cgraph should control nested function lowering. It would only 5209 be invoked when it is certain that the outer-most function is used. 5210 5211 TODO: Cgraph needs a gimplify_function callback. It should be invoked 5212 when (1) it is certain that the function is used, (2) warning flags 5213 specified by the user require some amount of compilation in order to 5214 honor, (3) the language indicates that semantic analysis is not 5215 complete until gimplification occurs. Hum... this sounds overly 5216 complicated. Perhaps we should just have the front end gimplify 5217 always; in most cases it's only one function call. 5218 5219 The front end needs to pass all function definitions and top level 5220 declarations off to the middle-end so that they can be compiled and 5221 emitted to the object file. For a simple procedural language, it is 5222 usually most convenient to do this as each top level declaration or 5223 definition is seen. There is also a distinction to be made between 5224 generating functional code and generating complete debug information. 5225 The only thing that is absolutely required for functional code is that 5226 function and data _definitions_ be passed to the middle-end. For 5227 complete debug information, function, data and type declarations should 5228 all be passed as well. 5229 5230 In any case, the front end needs each complete top-level function or 5231 data declaration, and each data definition should be passed to 5232 `rest_of_decl_compilation'. Each complete type definition should be 5233 passed to `rest_of_type_compilation'. Each function definition should 5234 be passed to `cgraph_finalize_function'. 5235 5236 TODO: I know rest_of_compilation currently has all sorts of RTL 5237 generation semantics. I plan to move all code generation bits (both 5238 Tree and RTL) to compile_function. Should we hide cgraph from the 5239 front ends and move back to rest_of_compilation as the official 5240 interface? Possibly we should rename all three interfaces such that 5241 the names match in some meaningful way and that is more descriptive 5242 than "rest_of". 5243 5244 The middle-end will, at its option, emit the function and data 5245 definitions immediately or queue them for later processing. 5246 5247 5248 File: gccint.info, Node: Gimplification pass, Next: Pass manager, Prev: Parsing pass, Up: Passes 5249 5250 8.2 Gimplification pass 5251 ======================= 5252 5253 "Gimplification" is a whimsical term for the process of converting the 5254 intermediate representation of a function into the GIMPLE language 5255 (CROSSREF). The term stuck, and so words like "gimplification", 5256 "gimplify", "gimplifier" and the like are sprinkled throughout this 5257 section of code. 5258 5259 While a front end may certainly choose to generate GIMPLE directly if 5260 it chooses, this can be a moderately complex process unless the 5261 intermediate language used by the front end is already fairly simple. 5262 Usually it is easier to generate GENERIC trees plus extensions and let 5263 the language-independent gimplifier do most of the work. 5264 5265 The main entry point to this pass is `gimplify_function_tree' located 5266 in `gimplify.c'. From here we process the entire function gimplifying 5267 each statement in turn. The main workhorse for this pass is 5268 `gimplify_expr'. Approximately everything passes through here at least 5269 once, and it is from here that we invoke the `lang_hooks.gimplify_expr' 5270 callback. 5271 5272 The callback should examine the expression in question and return 5273 `GS_UNHANDLED' if the expression is not a language specific construct 5274 that requires attention. Otherwise it should alter the expression in 5275 some way to such that forward progress is made toward producing valid 5276 GIMPLE. If the callback is certain that the transformation is complete 5277 and the expression is valid GIMPLE, it should return `GS_ALL_DONE'. 5278 Otherwise it should return `GS_OK', which will cause the expression to 5279 be processed again. If the callback encounters an error during the 5280 transformation (because the front end is relying on the gimplification 5281 process to finish semantic checks), it should return `GS_ERROR'. 5282 5283 5284 File: gccint.info, Node: Pass manager, Next: Tree SSA passes, Prev: Gimplification pass, Up: Passes 5285 5286 8.3 Pass manager 5287 ================ 5288 5289 The pass manager is located in `passes.c', `tree-optimize.c' and 5290 `tree-pass.h'. Its job is to run all of the individual passes in the 5291 correct order, and take care of standard bookkeeping that applies to 5292 every pass. 5293 5294 The theory of operation is that each pass defines a structure that 5295 represents everything we need to know about that pass--when it should 5296 be run, how it should be run, what intermediate language form or 5297 on-the-side data structures it needs. We register the pass to be run 5298 in some particular order, and the pass manager arranges for everything 5299 to happen in the correct order. 5300 5301 The actuality doesn't completely live up to the theory at present. 5302 Command-line switches and `timevar_id_t' enumerations must still be 5303 defined elsewhere. The pass manager validates constraints but does not 5304 attempt to (re-)generate data structures or lower intermediate language 5305 form based on the requirements of the next pass. Nevertheless, what is 5306 present is useful, and a far sight better than nothing at all. 5307 5308 Each pass may have its own dump file (for GCC debugging purposes). 5309 Passes without any names, or with a name starting with a star, do not 5310 dump anything. 5311 5312 TODO: describe the global variables set up by the pass manager, and a 5313 brief description of how a new pass should use it. I need to look at 5314 what info RTL passes use first.... 5315 5316 5317 File: gccint.info, Node: Tree SSA passes, Next: RTL passes, Prev: Pass manager, Up: Passes 5318 5319 8.4 Tree SSA passes 5320 =================== 5321 5322 The following briefly describes the Tree optimization passes that are 5323 run after gimplification and what source files they are located in. 5324 5325 * Remove useless statements 5326 5327 This pass is an extremely simple sweep across the gimple code in 5328 which we identify obviously dead code and remove it. Here we do 5329 things like simplify `if' statements with constant conditions, 5330 remove exception handling constructs surrounding code that 5331 obviously cannot throw, remove lexical bindings that contain no 5332 variables, and other assorted simplistic cleanups. The idea is to 5333 get rid of the obvious stuff quickly rather than wait until later 5334 when it's more work to get rid of it. This pass is located in 5335 `tree-cfg.c' and described by `pass_remove_useless_stmts'. 5336 5337 * Mudflap declaration registration 5338 5339 If mudflap (*note -fmudflap -fmudflapth -fmudflapir: (gcc)Optimize 5340 Options.) is enabled, we generate code to register some variable 5341 declarations with the mudflap runtime. Specifically, the runtime 5342 tracks the lifetimes of those variable declarations that have 5343 their addresses taken, or whose bounds are unknown at compile time 5344 (`extern'). This pass generates new exception handling constructs 5345 (`try'/`finally'), and so must run before those are lowered. In 5346 addition, the pass enqueues declarations of static variables whose 5347 lifetimes extend to the entire program. The pass is located in 5348 `tree-mudflap.c' and is described by `pass_mudflap_1'. 5349 5350 * OpenMP lowering 5351 5352 If OpenMP generation (`-fopenmp') is enabled, this pass lowers 5353 OpenMP constructs into GIMPLE. 5354 5355 Lowering of OpenMP constructs involves creating replacement 5356 expressions for local variables that have been mapped using data 5357 sharing clauses, exposing the control flow of most synchronization 5358 directives and adding region markers to facilitate the creation of 5359 the control flow graph. The pass is located in `omp-low.c' and is 5360 described by `pass_lower_omp'. 5361 5362 * OpenMP expansion 5363 5364 If OpenMP generation (`-fopenmp') is enabled, this pass expands 5365 parallel regions into their own functions to be invoked by the 5366 thread library. The pass is located in `omp-low.c' and is 5367 described by `pass_expand_omp'. 5368 5369 * Lower control flow 5370 5371 This pass flattens `if' statements (`COND_EXPR') and moves lexical 5372 bindings (`BIND_EXPR') out of line. After this pass, all `if' 5373 statements will have exactly two `goto' statements in its `then' 5374 and `else' arms. Lexical binding information for each statement 5375 will be found in `TREE_BLOCK' rather than being inferred from its 5376 position under a `BIND_EXPR'. This pass is found in 5377 `gimple-low.c' and is described by `pass_lower_cf'. 5378 5379 * Lower exception handling control flow 5380 5381 This pass decomposes high-level exception handling constructs 5382 (`TRY_FINALLY_EXPR' and `TRY_CATCH_EXPR') into a form that 5383 explicitly represents the control flow involved. After this pass, 5384 `lookup_stmt_eh_region' will return a non-negative number for any 5385 statement that may have EH control flow semantics; examine 5386 `tree_can_throw_internal' or `tree_can_throw_external' for exact 5387 semantics. Exact control flow may be extracted from 5388 `foreach_reachable_handler'. The EH region nesting tree is defined 5389 in `except.h' and built in `except.c'. The lowering pass itself 5390 is in `tree-eh.c' and is described by `pass_lower_eh'. 5391 5392 * Build the control flow graph 5393 5394 This pass decomposes a function into basic blocks and creates all 5395 of the edges that connect them. It is located in `tree-cfg.c' and 5396 is described by `pass_build_cfg'. 5397 5398 * Find all referenced variables 5399 5400 This pass walks the entire function and collects an array of all 5401 variables referenced in the function, `referenced_vars'. The 5402 index at which a variable is found in the array is used as a UID 5403 for the variable within this function. This data is needed by the 5404 SSA rewriting routines. The pass is located in `tree-dfa.c' and 5405 is described by `pass_referenced_vars'. 5406 5407 * Enter static single assignment form 5408 5409 This pass rewrites the function such that it is in SSA form. After 5410 this pass, all `is_gimple_reg' variables will be referenced by 5411 `SSA_NAME', and all occurrences of other variables will be 5412 annotated with `VDEFS' and `VUSES'; PHI nodes will have been 5413 inserted as necessary for each basic block. This pass is located 5414 in `tree-ssa.c' and is described by `pass_build_ssa'. 5415 5416 * Warn for uninitialized variables 5417 5418 This pass scans the function for uses of `SSA_NAME's that are fed 5419 by default definition. For non-parameter variables, such uses are 5420 uninitialized. The pass is run twice, before and after 5421 optimization (if turned on). In the first pass we only warn for 5422 uses that are positively uninitialized; in the second pass we warn 5423 for uses that are possibly uninitialized. The pass is located in 5424 `tree-ssa.c' and is defined by `pass_early_warn_uninitialized' and 5425 `pass_late_warn_uninitialized'. 5426 5427 * Dead code elimination 5428 5429 This pass scans the function for statements without side effects 5430 whose result is unused. It does not do memory life analysis, so 5431 any value that is stored in memory is considered used. The pass 5432 is run multiple times throughout the optimization process. It is 5433 located in `tree-ssa-dce.c' and is described by `pass_dce'. 5434 5435 * Dominator optimizations 5436 5437 This pass performs trivial dominator-based copy and constant 5438 propagation, expression simplification, and jump threading. It is 5439 run multiple times throughout the optimization process. It it 5440 located in `tree-ssa-dom.c' and is described by `pass_dominator'. 5441 5442 * Forward propagation of single-use variables 5443 5444 This pass attempts to remove redundant computation by substituting 5445 variables that are used once into the expression that uses them and 5446 seeing if the result can be simplified. It is located in 5447 `tree-ssa-forwprop.c' and is described by `pass_forwprop'. 5448 5449 * Copy Renaming 5450 5451 This pass attempts to change the name of compiler temporaries 5452 involved in copy operations such that SSA->normal can coalesce the 5453 copy away. When compiler temporaries are copies of user 5454 variables, it also renames the compiler temporary to the user 5455 variable resulting in better use of user symbols. It is located 5456 in `tree-ssa-copyrename.c' and is described by `pass_copyrename'. 5457 5458 * PHI node optimizations 5459 5460 This pass recognizes forms of PHI inputs that can be represented as 5461 conditional expressions and rewrites them into straight line code. 5462 It is located in `tree-ssa-phiopt.c' and is described by 5463 `pass_phiopt'. 5464 5465 * May-alias optimization 5466 5467 This pass performs a flow sensitive SSA-based points-to analysis. 5468 The resulting may-alias, must-alias, and escape analysis 5469 information is used to promote variables from in-memory 5470 addressable objects to non-aliased variables that can be renamed 5471 into SSA form. We also update the `VDEF'/`VUSE' memory tags for 5472 non-renameable aggregates so that we get fewer false kills. The 5473 pass is located in `tree-ssa-alias.c' and is described by 5474 `pass_may_alias'. 5475 5476 Interprocedural points-to information is located in 5477 `tree-ssa-structalias.c' and described by `pass_ipa_pta'. 5478 5479 * Profiling 5480 5481 This pass rewrites the function in order to collect runtime block 5482 and value profiling data. Such data may be fed back into the 5483 compiler on a subsequent run so as to allow optimization based on 5484 expected execution frequencies. The pass is located in 5485 `predict.c' and is described by `pass_profile'. 5486 5487 * Lower complex arithmetic 5488 5489 This pass rewrites complex arithmetic operations into their 5490 component scalar arithmetic operations. The pass is located in 5491 `tree-complex.c' and is described by `pass_lower_complex'. 5492 5493 * Scalar replacement of aggregates 5494 5495 This pass rewrites suitable non-aliased local aggregate variables 5496 into a set of scalar variables. The resulting scalar variables are 5497 rewritten into SSA form, which allows subsequent optimization 5498 passes to do a significantly better job with them. The pass is 5499 located in `tree-sra.c' and is described by `pass_sra'. 5500 5501 * Dead store elimination 5502 5503 This pass eliminates stores to memory that are subsequently 5504 overwritten by another store, without any intervening loads. The 5505 pass is located in `tree-ssa-dse.c' and is described by `pass_dse'. 5506 5507 * Tail recursion elimination 5508 5509 This pass transforms tail recursion into a loop. It is located in 5510 `tree-tailcall.c' and is described by `pass_tail_recursion'. 5511 5512 * Forward store motion 5513 5514 This pass sinks stores and assignments down the flowgraph closer 5515 to their use point. The pass is located in `tree-ssa-sink.c' and 5516 is described by `pass_sink_code'. 5517 5518 * Partial redundancy elimination 5519 5520 This pass eliminates partially redundant computations, as well as 5521 performing load motion. The pass is located in `tree-ssa-pre.c' 5522 and is described by `pass_pre'. 5523 5524 Just before partial redundancy elimination, if 5525 `-funsafe-math-optimizations' is on, GCC tries to convert 5526 divisions to multiplications by the reciprocal. The pass is 5527 located in `tree-ssa-math-opts.c' and is described by 5528 `pass_cse_reciprocal'. 5529 5530 * Full redundancy elimination 5531 5532 This is a simpler form of PRE that only eliminates redundancies 5533 that occur an all paths. It is located in `tree-ssa-pre.c' and 5534 described by `pass_fre'. 5535 5536 * Loop optimization 5537 5538 The main driver of the pass is placed in `tree-ssa-loop.c' and 5539 described by `pass_loop'. 5540 5541 The optimizations performed by this pass are: 5542 5543 Loop invariant motion. This pass moves only invariants that would 5544 be hard to handle on RTL level (function calls, operations that 5545 expand to nontrivial sequences of insns). With `-funswitch-loops' 5546 it also moves operands of conditions that are invariant out of the 5547 loop, so that we can use just trivial invariantness analysis in 5548 loop unswitching. The pass also includes store motion. The pass 5549 is implemented in `tree-ssa-loop-im.c'. 5550 5551 Canonical induction variable creation. This pass creates a simple 5552 counter for number of iterations of the loop and replaces the exit 5553 condition of the loop using it, in case when a complicated 5554 analysis is necessary to determine the number of iterations. 5555 Later optimizations then may determine the number easily. The 5556 pass is implemented in `tree-ssa-loop-ivcanon.c'. 5557 5558 Induction variable optimizations. This pass performs standard 5559 induction variable optimizations, including strength reduction, 5560 induction variable merging and induction variable elimination. 5561 The pass is implemented in `tree-ssa-loop-ivopts.c'. 5562 5563 Loop unswitching. This pass moves the conditional jumps that are 5564 invariant out of the loops. To achieve this, a duplicate of the 5565 loop is created for each possible outcome of conditional jump(s). 5566 The pass is implemented in `tree-ssa-loop-unswitch.c'. This pass 5567 should eventually replace the RTL level loop unswitching in 5568 `loop-unswitch.c', but currently the RTL level pass is not 5569 completely redundant yet due to deficiencies in tree level alias 5570 analysis. 5571 5572 The optimizations also use various utility functions contained in 5573 `tree-ssa-loop-manip.c', `cfgloop.c', `cfgloopanal.c' and 5574 `cfgloopmanip.c'. 5575 5576 Vectorization. This pass transforms loops to operate on vector 5577 types instead of scalar types. Data parallelism across loop 5578 iterations is exploited to group data elements from consecutive 5579 iterations into a vector and operate on them in parallel. 5580 Depending on available target support the loop is conceptually 5581 unrolled by a factor `VF' (vectorization factor), which is the 5582 number of elements operated upon in parallel in each iteration, 5583 and the `VF' copies of each scalar operation are fused to form a 5584 vector operation. Additional loop transformations such as peeling 5585 and versioning may take place to align the number of iterations, 5586 and to align the memory accesses in the loop. The pass is 5587 implemented in `tree-vectorizer.c' (the main driver and general 5588 utilities), `tree-vect-analyze.c' and `tree-vect-transform.c'. 5589 Analysis of data references is in `tree-data-ref.c'. 5590 5591 Autoparallelization. This pass splits the loop iteration space to 5592 run into several threads. The pass is implemented in 5593 `tree-parloops.c'. 5594 5595 * Tree level if-conversion for vectorizer 5596 5597 This pass applies if-conversion to simple loops to help vectorizer. 5598 We identify if convertible loops, if-convert statements and merge 5599 basic blocks in one big block. The idea is to present loop in such 5600 form so that vectorizer can have one to one mapping between 5601 statements and available vector operations. This patch 5602 re-introduces COND_EXPR at GIMPLE level. This pass is located in 5603 `tree-if-conv.c' and is described by `pass_if_conversion'. 5604 5605 * Conditional constant propagation 5606 5607 This pass relaxes a lattice of values in order to identify those 5608 that must be constant even in the presence of conditional branches. 5609 The pass is located in `tree-ssa-ccp.c' and is described by 5610 `pass_ccp'. 5611 5612 A related pass that works on memory loads and stores, and not just 5613 register values, is located in `tree-ssa-ccp.c' and described by 5614 `pass_store_ccp'. 5615 5616 * Conditional copy propagation 5617 5618 This is similar to constant propagation but the lattice of values 5619 is the "copy-of" relation. It eliminates redundant copies from the 5620 code. The pass is located in `tree-ssa-copy.c' and described by 5621 `pass_copy_prop'. 5622 5623 A related pass that works on memory copies, and not just register 5624 copies, is located in `tree-ssa-copy.c' and described by 5625 `pass_store_copy_prop'. 5626 5627 * Value range propagation 5628 5629 This transformation is similar to constant propagation but instead 5630 of propagating single constant values, it propagates known value 5631 ranges. The implementation is based on Patterson's range 5632 propagation algorithm (Accurate Static Branch Prediction by Value 5633 Range Propagation, J. R. C. Patterson, PLDI '95). In contrast to 5634 Patterson's algorithm, this implementation does not propagate 5635 branch probabilities nor it uses more than a single range per SSA 5636 name. This means that the current implementation cannot be used 5637 for branch prediction (though adapting it would not be difficult). 5638 The pass is located in `tree-vrp.c' and is described by 5639 `pass_vrp'. 5640 5641 * Folding built-in functions 5642 5643 This pass simplifies built-in functions, as applicable, with 5644 constant arguments or with inferable string lengths. It is 5645 located in `tree-ssa-ccp.c' and is described by 5646 `pass_fold_builtins'. 5647 5648 * Split critical edges 5649 5650 This pass identifies critical edges and inserts empty basic blocks 5651 such that the edge is no longer critical. The pass is located in 5652 `tree-cfg.c' and is described by `pass_split_crit_edges'. 5653 5654 * Control dependence dead code elimination 5655 5656 This pass is a stronger form of dead code elimination that can 5657 eliminate unnecessary control flow statements. It is located in 5658 `tree-ssa-dce.c' and is described by `pass_cd_dce'. 5659 5660 * Tail call elimination 5661 5662 This pass identifies function calls that may be rewritten into 5663 jumps. No code transformation is actually applied here, but the 5664 data and control flow problem is solved. The code transformation 5665 requires target support, and so is delayed until RTL. In the 5666 meantime `CALL_EXPR_TAILCALL' is set indicating the possibility. 5667 The pass is located in `tree-tailcall.c' and is described by 5668 `pass_tail_calls'. The RTL transformation is handled by 5669 `fixup_tail_calls' in `calls.c'. 5670 5671 * Warn for function return without value 5672 5673 For non-void functions, this pass locates return statements that do 5674 not specify a value and issues a warning. Such a statement may 5675 have been injected by falling off the end of the function. This 5676 pass is run last so that we have as much time as possible to prove 5677 that the statement is not reachable. It is located in 5678 `tree-cfg.c' and is described by `pass_warn_function_return'. 5679 5680 * Mudflap statement annotation 5681 5682 If mudflap is enabled, we rewrite some memory accesses with code to 5683 validate that the memory access is correct. In particular, 5684 expressions involving pointer dereferences (`INDIRECT_REF', 5685 `ARRAY_REF', etc.) are replaced by code that checks the selected 5686 address range against the mudflap runtime's database of valid 5687 regions. This check includes an inline lookup into a 5688 direct-mapped cache, based on shift/mask operations of the pointer 5689 value, with a fallback function call into the runtime. The pass 5690 is located in `tree-mudflap.c' and is described by 5691 `pass_mudflap_2'. 5692 5693 * Leave static single assignment form 5694 5695 This pass rewrites the function such that it is in normal form. At 5696 the same time, we eliminate as many single-use temporaries as 5697 possible, so the intermediate language is no longer GIMPLE, but 5698 GENERIC. The pass is located in `tree-outof-ssa.c' and is 5699 described by `pass_del_ssa'. 5700 5701 * Merge PHI nodes that feed into one another 5702 5703 This is part of the CFG cleanup passes. It attempts to join PHI 5704 nodes from a forwarder CFG block into another block with PHI 5705 nodes. The pass is located in `tree-cfgcleanup.c' and is 5706 described by `pass_merge_phi'. 5707 5708 * Return value optimization 5709 5710 If a function always returns the same local variable, and that 5711 local variable is an aggregate type, then the variable is replaced 5712 with the return value for the function (i.e., the function's 5713 DECL_RESULT). This is equivalent to the C++ named return value 5714 optimization applied to GIMPLE. The pass is located in 5715 `tree-nrv.c' and is described by `pass_nrv'. 5716 5717 * Return slot optimization 5718 5719 If a function returns a memory object and is called as `var = 5720 foo()', this pass tries to change the call so that the address of 5721 `var' is sent to the caller to avoid an extra memory copy. This 5722 pass is located in `tree-nrv.c' and is described by 5723 `pass_return_slot'. 5724 5725 * Optimize calls to `__builtin_object_size' 5726 5727 This is a propagation pass similar to CCP that tries to remove 5728 calls to `__builtin_object_size' when the size of the object can be 5729 computed at compile-time. This pass is located in 5730 `tree-object-size.c' and is described by `pass_object_sizes'. 5731 5732 * Loop invariant motion 5733 5734 This pass removes expensive loop-invariant computations out of 5735 loops. The pass is located in `tree-ssa-loop.c' and described by 5736 `pass_lim'. 5737 5738 * Loop nest optimizations 5739 5740 This is a family of loop transformations that works on loop nests. 5741 It includes loop interchange, scaling, skewing and reversal and 5742 they are all geared to the optimization of data locality in array 5743 traversals and the removal of dependencies that hamper 5744 optimizations such as loop parallelization and vectorization. The 5745 pass is located in `tree-loop-linear.c' and described by 5746 `pass_linear_transform'. 5747 5748 * Removal of empty loops 5749 5750 This pass removes loops with no code in them. The pass is located 5751 in `tree-ssa-loop-ivcanon.c' and described by `pass_empty_loop'. 5752 5753 * Unrolling of small loops 5754 5755 This pass completely unrolls loops with few iterations. The pass 5756 is located in `tree-ssa-loop-ivcanon.c' and described by 5757 `pass_complete_unroll'. 5758 5759 * Predictive commoning 5760 5761 This pass makes the code reuse the computations from the previous 5762 iterations of the loops, especially loads and stores to memory. 5763 It does so by storing the values of these computations to a bank 5764 of temporary variables that are rotated at the end of loop. To 5765 avoid the need for this rotation, the loop is then unrolled and 5766 the copies of the loop body are rewritten to use the appropriate 5767 version of the temporary variable. This pass is located in 5768 `tree-predcom.c' and described by `pass_predcom'. 5769 5770 * Array prefetching 5771 5772 This pass issues prefetch instructions for array references inside 5773 loops. The pass is located in `tree-ssa-loop-prefetch.c' and 5774 described by `pass_loop_prefetch'. 5775 5776 * Reassociation 5777 5778 This pass rewrites arithmetic expressions to enable optimizations 5779 that operate on them, like redundancy elimination and 5780 vectorization. The pass is located in `tree-ssa-reassoc.c' and 5781 described by `pass_reassoc'. 5782 5783 * Optimization of `stdarg' functions 5784 5785 This pass tries to avoid the saving of register arguments into the 5786 stack on entry to `stdarg' functions. If the function doesn't use 5787 any `va_start' macros, no registers need to be saved. If 5788 `va_start' macros are used, the `va_list' variables don't escape 5789 the function, it is only necessary to save registers that will be 5790 used in `va_arg' macros. For instance, if `va_arg' is only used 5791 with integral types in the function, floating point registers 5792 don't need to be saved. This pass is located in `tree-stdarg.c' 5793 and described by `pass_stdarg'. 5794 5795 5796 5797 File: gccint.info, Node: RTL passes, Prev: Tree SSA passes, Up: Passes 5798 5799 8.5 RTL passes 5800 ============== 5801 5802 The following briefly describes the RTL generation and optimization 5803 passes that are run after the Tree optimization passes. 5804 5805 * RTL generation 5806 5807 The source files for RTL generation include `stmt.c', `calls.c', 5808 `expr.c', `explow.c', `expmed.c', `function.c', `optabs.c' and 5809 `emit-rtl.c'. Also, the file `insn-emit.c', generated from the 5810 machine description by the program `genemit', is used in this 5811 pass. The header file `expr.h' is used for communication within 5812 this pass. 5813 5814 The header files `insn-flags.h' and `insn-codes.h', generated from 5815 the machine description by the programs `genflags' and `gencodes', 5816 tell this pass which standard names are available for use and 5817 which patterns correspond to them. 5818 5819 * Generation of exception landing pads 5820 5821 This pass generates the glue that handles communication between the 5822 exception handling library routines and the exception handlers 5823 within the function. Entry points in the function that are 5824 invoked by the exception handling library are called "landing 5825 pads". The code for this pass is located in `except.c'. 5826 5827 * Control flow graph cleanup 5828 5829 This pass removes unreachable code, simplifies jumps to next, 5830 jumps to jump, jumps across jumps, etc. The pass is run multiple 5831 times. For historical reasons, it is occasionally referred to as 5832 the "jump optimization pass". The bulk of the code for this pass 5833 is in `cfgcleanup.c', and there are support routines in `cfgrtl.c' 5834 and `jump.c'. 5835 5836 * Forward propagation of single-def values 5837 5838 This pass attempts to remove redundant computation by substituting 5839 variables that come from a single definition, and seeing if the 5840 result can be simplified. It performs copy propagation and 5841 addressing mode selection. The pass is run twice, with values 5842 being propagated into loops only on the second run. The code is 5843 located in `fwprop.c'. 5844 5845 * Common subexpression elimination 5846 5847 This pass removes redundant computation within basic blocks, and 5848 optimizes addressing modes based on cost. The pass is run twice. 5849 The code for this pass is located in `cse.c'. 5850 5851 * Global common subexpression elimination 5852 5853 This pass performs two different types of GCSE depending on 5854 whether you are optimizing for size or not (LCM based GCSE tends 5855 to increase code size for a gain in speed, while Morel-Renvoise 5856 based GCSE does not). When optimizing for size, GCSE is done 5857 using Morel-Renvoise Partial Redundancy Elimination, with the 5858 exception that it does not try to move invariants out of 5859 loops--that is left to the loop optimization pass. If MR PRE 5860 GCSE is done, code hoisting (aka unification) is also done, as 5861 well as load motion. If you are optimizing for speed, LCM (lazy 5862 code motion) based GCSE is done. LCM is based on the work of 5863 Knoop, Ruthing, and Steffen. LCM based GCSE also does loop 5864 invariant code motion. We also perform load and store motion when 5865 optimizing for speed. Regardless of which type of GCSE is used, 5866 the GCSE pass also performs global constant and copy propagation. 5867 The source file for this pass is `gcse.c', and the LCM routines 5868 are in `lcm.c'. 5869 5870 * Loop optimization 5871 5872 This pass performs several loop related optimizations. The source 5873 files `cfgloopanal.c' and `cfgloopmanip.c' contain generic loop 5874 analysis and manipulation code. Initialization and finalization 5875 of loop structures is handled by `loop-init.c'. A loop invariant 5876 motion pass is implemented in `loop-invariant.c'. Basic block 5877 level optimizations--unrolling, peeling and unswitching loops-- 5878 are implemented in `loop-unswitch.c' and `loop-unroll.c'. 5879 Replacing of the exit condition of loops by special 5880 machine-dependent instructions is handled by `loop-doloop.c'. 5881 5882 * Jump bypassing 5883 5884 This pass is an aggressive form of GCSE that transforms the control 5885 flow graph of a function by propagating constants into conditional 5886 branch instructions. The source file for this pass is `gcse.c'. 5887 5888 * If conversion 5889 5890 This pass attempts to replace conditional branches and surrounding 5891 assignments with arithmetic, boolean value producing comparison 5892 instructions, and conditional move instructions. In the very last 5893 invocation after reload, it will generate predicated instructions 5894 when supported by the target. The code is located in `ifcvt.c'. 5895 5896 * Web construction 5897 5898 This pass splits independent uses of each pseudo-register. This 5899 can improve effect of the other transformation, such as CSE or 5900 register allocation. The code for this pass is located in `web.c'. 5901 5902 * Instruction combination 5903 5904 This pass attempts to combine groups of two or three instructions 5905 that are related by data flow into single instructions. It 5906 combines the RTL expressions for the instructions by substitution, 5907 simplifies the result using algebra, and then attempts to match 5908 the result against the machine description. The code is located 5909 in `combine.c'. 5910 5911 * Register movement 5912 5913 This pass looks for cases where matching constraints would force an 5914 instruction to need a reload, and this reload would be a 5915 register-to-register move. It then attempts to change the 5916 registers used by the instruction to avoid the move instruction. 5917 The code is located in `regmove.c'. 5918 5919 * Mode switching optimization 5920 5921 This pass looks for instructions that require the processor to be 5922 in a specific "mode" and minimizes the number of mode changes 5923 required to satisfy all users. What these modes are, and what 5924 they apply to are completely target-specific. The code for this 5925 pass is located in `mode-switching.c'. 5926 5927 * Modulo scheduling 5928 5929 This pass looks at innermost loops and reorders their instructions 5930 by overlapping different iterations. Modulo scheduling is 5931 performed immediately before instruction scheduling. The code for 5932 this pass is located in `modulo-sched.c'. 5933 5934 * Instruction scheduling 5935 5936 This pass looks for instructions whose output will not be 5937 available by the time that it is used in subsequent instructions. 5938 Memory loads and floating point instructions often have this 5939 behavior on RISC machines. It re-orders instructions within a 5940 basic block to try to separate the definition and use of items 5941 that otherwise would cause pipeline stalls. This pass is 5942 performed twice, before and after register allocation. The code 5943 for this pass is located in `haifa-sched.c', `sched-deps.c', 5944 `sched-ebb.c', `sched-rgn.c' and `sched-vis.c'. 5945 5946 * Register allocation 5947 5948 These passes make sure that all occurrences of pseudo registers are 5949 eliminated, either by allocating them to a hard register, replacing 5950 them by an equivalent expression (e.g. a constant) or by placing 5951 them on the stack. This is done in several subpasses: 5952 5953 * Register move optimizations. This pass makes some simple RTL 5954 code transformations which improve the subsequent register 5955 allocation. The source file is `regmove.c'. 5956 5957 * The integrated register allocator (IRA). It is called 5958 integrated because coalescing, register live range splitting, 5959 and hard register preferencing are done on-the-fly during 5960 coloring. It also has better integration with the reload 5961 pass. Pseudo-registers spilled by the allocator or the 5962 reload have still a chance to get hard-registers if the 5963 reload evicts some pseudo-registers from hard-registers. The 5964 allocator helps to choose better pseudos for spilling based 5965 on their live ranges and to coalesce stack slots allocated 5966 for the spilled pseudo-registers. IRA is a regional register 5967 allocator which is transformed into Chaitin-Briggs allocator 5968 if there is one region. By default, IRA chooses regions using 5969 register pressure but the user can force it to use one region 5970 or regions corresponding to all loops. 5971 5972 Source files of the allocator are `ira.c', `ira-build.c', 5973 `ira-costs.c', `ira-conflicts.c', `ira-color.c', 5974 `ira-emit.c', `ira-lives', plus header files `ira.h' and 5975 `ira-int.h' used for the communication between the allocator 5976 and the rest of the compiler and between the IRA files. 5977 5978 * Reloading. This pass renumbers pseudo registers with the 5979 hardware registers numbers they were allocated. Pseudo 5980 registers that did not get hard registers are replaced with 5981 stack slots. Then it finds instructions that are invalid 5982 because a value has failed to end up in a register, or has 5983 ended up in a register of the wrong kind. It fixes up these 5984 instructions by reloading the problematical values 5985 temporarily into registers. Additional instructions are 5986 generated to do the copying. 5987 5988 The reload pass also optionally eliminates the frame pointer 5989 and inserts instructions to save and restore call-clobbered 5990 registers around calls. 5991 5992 Source files are `reload.c' and `reload1.c', plus the header 5993 `reload.h' used for communication between them. 5994 5995 * Basic block reordering 5996 5997 This pass implements profile guided code positioning. If profile 5998 information is not available, various types of static analysis are 5999 performed to make the predictions normally coming from the profile 6000 feedback (IE execution frequency, branch probability, etc). It is 6001 implemented in the file `bb-reorder.c', and the various prediction 6002 routines are in `predict.c'. 6003 6004 * Variable tracking 6005 6006 This pass computes where the variables are stored at each position 6007 in code and generates notes describing the variable locations to 6008 RTL code. The location lists are then generated according to these 6009 notes to debug information if the debugging information format 6010 supports location lists. The code is located in `var-tracking.c'. 6011 6012 * Delayed branch scheduling 6013 6014 This optional pass attempts to find instructions that can go into 6015 the delay slots of other instructions, usually jumps and calls. 6016 The code for this pass is located in `reorg.c'. 6017 6018 * Branch shortening 6019 6020 On many RISC machines, branch instructions have a limited range. 6021 Thus, longer sequences of instructions must be used for long 6022 branches. In this pass, the compiler figures out what how far 6023 each instruction will be from each other instruction, and 6024 therefore whether the usual instructions, or the longer sequences, 6025 must be used for each branch. The code for this pass is located 6026 in `final.c'. 6027 6028 * Register-to-stack conversion 6029 6030 Conversion from usage of some hard registers to usage of a register 6031 stack may be done at this point. Currently, this is supported only 6032 for the floating-point registers of the Intel 80387 coprocessor. 6033 The code for this pass is located in `reg-stack.c'. 6034 6035 * Final 6036 6037 This pass outputs the assembler code for the function. The source 6038 files are `final.c' plus `insn-output.c'; the latter is generated 6039 automatically from the machine description by the tool `genoutput'. 6040 The header file `conditions.h' is used for communication between 6041 these files. If mudflap is enabled, the queue of deferred 6042 declarations and any addressed constants (e.g., string literals) 6043 is processed by `mudflap_finish_file' into a synthetic constructor 6044 function containing calls into the mudflap runtime. 6045 6046 * Debugging information output 6047 6048 This is run after final because it must output the stack slot 6049 offsets for pseudo registers that did not get hard registers. 6050 Source files are `dbxout.c' for DBX symbol table format, 6051 `sdbout.c' for SDB symbol table format, `dwarfout.c' for DWARF 6052 symbol table format, files `dwarf2out.c' and `dwarf2asm.c' for 6053 DWARF2 symbol table format, and `vmsdbgout.c' for VMS debug symbol 6054 table format. 6055 6056 6057 6058 File: gccint.info, Node: Trees, Next: GENERIC, Prev: Passes, Up: Top 6059 6060 9 Trees: The intermediate representation used by the C and C++ front ends 6061 ************************************************************************* 6062 6063 This chapter documents the internal representation used by GCC to 6064 represent C and C++ source programs. When presented with a C or C++ 6065 source program, GCC parses the program, performs semantic analysis 6066 (including the generation of error messages), and then produces the 6067 internal representation described here. This representation contains a 6068 complete representation for the entire translation unit provided as 6069 input to the front end. This representation is then typically processed 6070 by a code-generator in order to produce machine code, but could also be 6071 used in the creation of source browsers, intelligent editors, automatic 6072 documentation generators, interpreters, and any other programs needing 6073 the ability to process C or C++ code. 6074 6075 This chapter explains the internal representation. In particular, it 6076 documents the internal representation for C and C++ source constructs, 6077 and the macros, functions, and variables that can be used to access 6078 these constructs. The C++ representation is largely a superset of the 6079 representation used in the C front end. There is only one construct 6080 used in C that does not appear in the C++ front end and that is the GNU 6081 "nested function" extension. Many of the macros documented here do not 6082 apply in C because the corresponding language constructs do not appear 6083 in C. 6084 6085 If you are developing a "back end", be it is a code-generator or some 6086 other tool, that uses this representation, you may occasionally find 6087 that you need to ask questions not easily answered by the functions and 6088 macros available here. If that situation occurs, it is quite likely 6089 that GCC already supports the functionality you desire, but that the 6090 interface is simply not documented here. In that case, you should ask 6091 the GCC maintainers (via mail to <gcc (a] gcc.gnu.org>) about documenting 6092 the functionality you require. Similarly, if you find yourself writing 6093 functions that do not deal directly with your back end, but instead 6094 might be useful to other people using the GCC front end, you should 6095 submit your patches for inclusion in GCC. 6096 6097 * Menu: 6098 6099 * Deficiencies:: Topics net yet covered in this document. 6100 * Tree overview:: All about `tree's. 6101 * Types:: Fundamental and aggregate types. 6102 * Scopes:: Namespaces and classes. 6103 * Functions:: Overloading, function bodies, and linkage. 6104 * Declarations:: Type declarations and variables. 6105 * Attributes:: Declaration and type attributes. 6106 * Expression trees:: From `typeid' to `throw'. 6107 6108 6109 File: gccint.info, Node: Deficiencies, Next: Tree overview, Up: Trees 6110 6111 9.1 Deficiencies 6112 ================ 6113 6114 There are many places in which this document is incomplet and incorrekt. 6115 It is, as of yet, only _preliminary_ documentation. 6116 6117 6118 File: gccint.info, Node: Tree overview, Next: Types, Prev: Deficiencies, Up: Trees 6119 6120 9.2 Overview 6121 ============ 6122 6123 The central data structure used by the internal representation is the 6124 `tree'. These nodes, while all of the C type `tree', are of many 6125 varieties. A `tree' is a pointer type, but the object to which it 6126 points may be of a variety of types. From this point forward, we will 6127 refer to trees in ordinary type, rather than in `this font', except 6128 when talking about the actual C type `tree'. 6129 6130 You can tell what kind of node a particular tree is by using the 6131 `TREE_CODE' macro. Many, many macros take trees as input and return 6132 trees as output. However, most macros require a certain kind of tree 6133 node as input. In other words, there is a type-system for trees, but 6134 it is not reflected in the C type-system. 6135 6136 For safety, it is useful to configure GCC with `--enable-checking'. 6137 Although this results in a significant performance penalty (since all 6138 tree types are checked at run-time), and is therefore inappropriate in a 6139 release version, it is extremely helpful during the development process. 6140 6141 Many macros behave as predicates. Many, although not all, of these 6142 predicates end in `_P'. Do not rely on the result type of these macros 6143 being of any particular type. You may, however, rely on the fact that 6144 the type can be compared to `0', so that statements like 6145 if (TEST_P (t) && !TEST_P (y)) 6146 x = 1; 6147 and 6148 int i = (TEST_P (t) != 0); 6149 are legal. Macros that return `int' values now may be changed to 6150 return `tree' values, or other pointers in the future. Even those that 6151 continue to return `int' may return multiple nonzero codes where 6152 previously they returned only zero and one. Therefore, you should not 6153 write code like 6154 if (TEST_P (t) == 1) 6155 as this code is not guaranteed to work correctly in the future. 6156 6157 You should not take the address of values returned by the macros or 6158 functions described here. In particular, no guarantee is given that the 6159 values are lvalues. 6160 6161 In general, the names of macros are all in uppercase, while the names 6162 of functions are entirely in lowercase. There are rare exceptions to 6163 this rule. You should assume that any macro or function whose name is 6164 made up entirely of uppercase letters may evaluate its arguments more 6165 than once. You may assume that a macro or function whose name is made 6166 up entirely of lowercase letters will evaluate its arguments only once. 6167 6168 The `error_mark_node' is a special tree. Its tree code is 6169 `ERROR_MARK', but since there is only ever one node with that code, the 6170 usual practice is to compare the tree against `error_mark_node'. (This 6171 test is just a test for pointer equality.) If an error has occurred 6172 during front-end processing the flag `errorcount' will be set. If the 6173 front end has encountered code it cannot handle, it will issue a 6174 message to the user and set `sorrycount'. When these flags are set, 6175 any macro or function which normally returns a tree of a particular 6176 kind may instead return the `error_mark_node'. Thus, if you intend to 6177 do any processing of erroneous code, you must be prepared to deal with 6178 the `error_mark_node'. 6179 6180 Occasionally, a particular tree slot (like an operand to an expression, 6181 or a particular field in a declaration) will be referred to as 6182 "reserved for the back end". These slots are used to store RTL when 6183 the tree is converted to RTL for use by the GCC back end. However, if 6184 that process is not taking place (e.g., if the front end is being hooked 6185 up to an intelligent editor), then those slots may be used by the back 6186 end presently in use. 6187 6188 If you encounter situations that do not match this documentation, such 6189 as tree nodes of types not mentioned here, or macros documented to 6190 return entities of a particular kind that instead return entities of 6191 some different kind, you have found a bug, either in the front end or in 6192 the documentation. Please report these bugs as you would any other bug. 6193 6194 * Menu: 6195 6196 * Macros and Functions::Macros and functions that can be used with all trees. 6197 * Identifiers:: The names of things. 6198 * Containers:: Lists and vectors. 6199 6200 6201 File: gccint.info, Node: Macros and Functions, Next: Identifiers, Up: Tree overview 6202 6203 9.2.1 Trees 6204 ----------- 6205 6206 This section is not here yet. 6207 6208 6209 File: gccint.info, Node: Identifiers, Next: Containers, Prev: Macros and Functions, Up: Tree overview 6210 6211 9.2.2 Identifiers 6212 ----------------- 6213 6214 An `IDENTIFIER_NODE' represents a slightly more general concept that 6215 the standard C or C++ concept of identifier. In particular, an 6216 `IDENTIFIER_NODE' may contain a `$', or other extraordinary characters. 6217 6218 There are never two distinct `IDENTIFIER_NODE's representing the same 6219 identifier. Therefore, you may use pointer equality to compare 6220 `IDENTIFIER_NODE's, rather than using a routine like `strcmp'. 6221 6222 You can use the following macros to access identifiers: 6223 `IDENTIFIER_POINTER' 6224 The string represented by the identifier, represented as a 6225 `char*'. This string is always `NUL'-terminated, and contains no 6226 embedded `NUL' characters. 6227 6228 `IDENTIFIER_LENGTH' 6229 The length of the string returned by `IDENTIFIER_POINTER', not 6230 including the trailing `NUL'. This value of `IDENTIFIER_LENGTH 6231 (x)' is always the same as `strlen (IDENTIFIER_POINTER (x))'. 6232 6233 `IDENTIFIER_OPNAME_P' 6234 This predicate holds if the identifier represents the name of an 6235 overloaded operator. In this case, you should not depend on the 6236 contents of either the `IDENTIFIER_POINTER' or the 6237 `IDENTIFIER_LENGTH'. 6238 6239 `IDENTIFIER_TYPENAME_P' 6240 This predicate holds if the identifier represents the name of a 6241 user-defined conversion operator. In this case, the `TREE_TYPE' of 6242 the `IDENTIFIER_NODE' holds the type to which the conversion 6243 operator converts. 6244 6245 6246 6247 File: gccint.info, Node: Containers, Prev: Identifiers, Up: Tree overview 6248 6249 9.2.3 Containers 6250 ---------------- 6251 6252 Two common container data structures can be represented directly with 6253 tree nodes. A `TREE_LIST' is a singly linked list containing two trees 6254 per node. These are the `TREE_PURPOSE' and `TREE_VALUE' of each node. 6255 (Often, the `TREE_PURPOSE' contains some kind of tag, or additional 6256 information, while the `TREE_VALUE' contains the majority of the 6257 payload. In other cases, the `TREE_PURPOSE' is simply `NULL_TREE', 6258 while in still others both the `TREE_PURPOSE' and `TREE_VALUE' are of 6259 equal stature.) Given one `TREE_LIST' node, the next node is found by 6260 following the `TREE_CHAIN'. If the `TREE_CHAIN' is `NULL_TREE', then 6261 you have reached the end of the list. 6262 6263 A `TREE_VEC' is a simple vector. The `TREE_VEC_LENGTH' is an integer 6264 (not a tree) giving the number of nodes in the vector. The nodes 6265 themselves are accessed using the `TREE_VEC_ELT' macro, which takes two 6266 arguments. The first is the `TREE_VEC' in question; the second is an 6267 integer indicating which element in the vector is desired. The 6268 elements are indexed from zero. 6269 6270 6271 File: gccint.info, Node: Types, Next: Scopes, Prev: Tree overview, Up: Trees 6272 6273 9.3 Types 6274 ========= 6275 6276 All types have corresponding tree nodes. However, you should not assume 6277 that there is exactly one tree node corresponding to each type. There 6278 are often multiple nodes corresponding to the same type. 6279 6280 For the most part, different kinds of types have different tree codes. 6281 (For example, pointer types use a `POINTER_TYPE' code while arrays use 6282 an `ARRAY_TYPE' code.) However, pointers to member functions use the 6283 `RECORD_TYPE' code. Therefore, when writing a `switch' statement that 6284 depends on the code associated with a particular type, you should take 6285 care to handle pointers to member functions under the `RECORD_TYPE' 6286 case label. 6287 6288 In C++, an array type is not qualified; rather the type of the array 6289 elements is qualified. This situation is reflected in the intermediate 6290 representation. The macros described here will always examine the 6291 qualification of the underlying element type when applied to an array 6292 type. (If the element type is itself an array, then the recursion 6293 continues until a non-array type is found, and the qualification of this 6294 type is examined.) So, for example, `CP_TYPE_CONST_P' will hold of the 6295 type `const int ()[7]', denoting an array of seven `int's. 6296 6297 The following functions and macros deal with cv-qualification of types: 6298 `CP_TYPE_QUALS' 6299 This macro returns the set of type qualifiers applied to this type. 6300 This value is `TYPE_UNQUALIFIED' if no qualifiers have been 6301 applied. The `TYPE_QUAL_CONST' bit is set if the type is 6302 `const'-qualified. The `TYPE_QUAL_VOLATILE' bit is set if the 6303 type is `volatile'-qualified. The `TYPE_QUAL_RESTRICT' bit is set 6304 if the type is `restrict'-qualified. 6305 6306 `CP_TYPE_CONST_P' 6307 This macro holds if the type is `const'-qualified. 6308 6309 `CP_TYPE_VOLATILE_P' 6310 This macro holds if the type is `volatile'-qualified. 6311 6312 `CP_TYPE_RESTRICT_P' 6313 This macro holds if the type is `restrict'-qualified. 6314 6315 `CP_TYPE_CONST_NON_VOLATILE_P' 6316 This predicate holds for a type that is `const'-qualified, but 6317 _not_ `volatile'-qualified; other cv-qualifiers are ignored as 6318 well: only the `const'-ness is tested. 6319 6320 `TYPE_MAIN_VARIANT' 6321 This macro returns the unqualified version of a type. It may be 6322 applied to an unqualified type, but it is not always the identity 6323 function in that case. 6324 6325 A few other macros and functions are usable with all types: 6326 `TYPE_SIZE' 6327 The number of bits required to represent the type, represented as 6328 an `INTEGER_CST'. For an incomplete type, `TYPE_SIZE' will be 6329 `NULL_TREE'. 6330 6331 `TYPE_ALIGN' 6332 The alignment of the type, in bits, represented as an `int'. 6333 6334 `TYPE_NAME' 6335 This macro returns a declaration (in the form of a `TYPE_DECL') for 6336 the type. (Note this macro does _not_ return a `IDENTIFIER_NODE', 6337 as you might expect, given its name!) You can look at the 6338 `DECL_NAME' of the `TYPE_DECL' to obtain the actual name of the 6339 type. The `TYPE_NAME' will be `NULL_TREE' for a type that is not 6340 a built-in type, the result of a typedef, or a named class type. 6341 6342 `CP_INTEGRAL_TYPE' 6343 This predicate holds if the type is an integral type. Notice that 6344 in C++, enumerations are _not_ integral types. 6345 6346 `ARITHMETIC_TYPE_P' 6347 This predicate holds if the type is an integral type (in the C++ 6348 sense) or a floating point type. 6349 6350 `CLASS_TYPE_P' 6351 This predicate holds for a class-type. 6352 6353 `TYPE_BUILT_IN' 6354 This predicate holds for a built-in type. 6355 6356 `TYPE_PTRMEM_P' 6357 This predicate holds if the type is a pointer to data member. 6358 6359 `TYPE_PTR_P' 6360 This predicate holds if the type is a pointer type, and the 6361 pointee is not a data member. 6362 6363 `TYPE_PTRFN_P' 6364 This predicate holds for a pointer to function type. 6365 6366 `TYPE_PTROB_P' 6367 This predicate holds for a pointer to object type. Note however 6368 that it does not hold for the generic pointer to object type `void 6369 *'. You may use `TYPE_PTROBV_P' to test for a pointer to object 6370 type as well as `void *'. 6371 6372 `TYPE_CANONICAL' 6373 This macro returns the "canonical" type for the given type node. 6374 Canonical types are used to improve performance in the C++ and 6375 Objective-C++ front ends by allowing efficient comparison between 6376 two type nodes in `same_type_p': if the `TYPE_CANONICAL' values of 6377 the types are equal, the types are equivalent; otherwise, the types 6378 are not equivalent. The notion of equivalence for canonical types 6379 is the same as the notion of type equivalence in the language 6380 itself. For instance, 6381 6382 When `TYPE_CANONICAL' is `NULL_TREE', there is no canonical type 6383 for the given type node. In this case, comparison between this 6384 type and any other type requires the compiler to perform a deep, 6385 "structural" comparison to see if the two type nodes have the same 6386 form and properties. 6387 6388 The canonical type for a node is always the most fundamental type 6389 in the equivalence class of types. For instance, `int' is its own 6390 canonical type. A typedef `I' of `int' will have `int' as its 6391 canonical type. Similarly, `I*' and a typedef `IP' (defined to 6392 `I*') will has `int*' as their canonical type. When building a new 6393 type node, be sure to set `TYPE_CANONICAL' to the appropriate 6394 canonical type. If the new type is a compound type (built from 6395 other types), and any of those other types require structural 6396 equality, use `SET_TYPE_STRUCTURAL_EQUALITY' to ensure that the 6397 new type also requires structural equality. Finally, if for some 6398 reason you cannot guarantee that `TYPE_CANONICAL' will point to 6399 the canonical type, use `SET_TYPE_STRUCTURAL_EQUALITY' to make 6400 sure that the new type-and any type constructed based on 6401 it-requires structural equality. If you suspect that the canonical 6402 type system is miscomparing types, pass `--param 6403 verify-canonical-types=1' to the compiler or configure with 6404 `--enable-checking' to force the compiler to verify its 6405 canonical-type comparisons against the structural comparisons; the 6406 compiler will then print any warnings if the canonical types 6407 miscompare. 6408 6409 `TYPE_STRUCTURAL_EQUALITY_P' 6410 This predicate holds when the node requires structural equality 6411 checks, e.g., when `TYPE_CANONICAL' is `NULL_TREE'. 6412 6413 `SET_TYPE_STRUCTURAL_EQUALITY' 6414 This macro states that the type node it is given requires 6415 structural equality checks, e.g., it sets `TYPE_CANONICAL' to 6416 `NULL_TREE'. 6417 6418 `same_type_p' 6419 This predicate takes two types as input, and holds if they are the 6420 same type. For example, if one type is a `typedef' for the other, 6421 or both are `typedef's for the same type. This predicate also 6422 holds if the two trees given as input are simply copies of one 6423 another; i.e., there is no difference between them at the source 6424 level, but, for whatever reason, a duplicate has been made in the 6425 representation. You should never use `==' (pointer equality) to 6426 compare types; always use `same_type_p' instead. 6427 6428 Detailed below are the various kinds of types, and the macros that can 6429 be used to access them. Although other kinds of types are used 6430 elsewhere in G++, the types described here are the only ones that you 6431 will encounter while examining the intermediate representation. 6432 6433 `VOID_TYPE' 6434 Used to represent the `void' type. 6435 6436 `INTEGER_TYPE' 6437 Used to represent the various integral types, including `char', 6438 `short', `int', `long', and `long long'. This code is not used 6439 for enumeration types, nor for the `bool' type. The 6440 `TYPE_PRECISION' is the number of bits used in the representation, 6441 represented as an `unsigned int'. (Note that in the general case 6442 this is not the same value as `TYPE_SIZE'; suppose that there were 6443 a 24-bit integer type, but that alignment requirements for the ABI 6444 required 32-bit alignment. Then, `TYPE_SIZE' would be an 6445 `INTEGER_CST' for 32, while `TYPE_PRECISION' would be 24.) The 6446 integer type is unsigned if `TYPE_UNSIGNED' holds; otherwise, it 6447 is signed. 6448 6449 The `TYPE_MIN_VALUE' is an `INTEGER_CST' for the smallest integer 6450 that may be represented by this type. Similarly, the 6451 `TYPE_MAX_VALUE' is an `INTEGER_CST' for the largest integer that 6452 may be represented by this type. 6453 6454 `REAL_TYPE' 6455 Used to represent the `float', `double', and `long double' types. 6456 The number of bits in the floating-point representation is given 6457 by `TYPE_PRECISION', as in the `INTEGER_TYPE' case. 6458 6459 `FIXED_POINT_TYPE' 6460 Used to represent the `short _Fract', `_Fract', `long _Fract', 6461 `long long _Fract', `short _Accum', `_Accum', `long _Accum', and 6462 `long long _Accum' types. The number of bits in the fixed-point 6463 representation is given by `TYPE_PRECISION', as in the 6464 `INTEGER_TYPE' case. There may be padding bits, fractional bits 6465 and integral bits. The number of fractional bits is given by 6466 `TYPE_FBIT', and the number of integral bits is given by 6467 `TYPE_IBIT'. The fixed-point type is unsigned if `TYPE_UNSIGNED' 6468 holds; otherwise, it is signed. The fixed-point type is 6469 saturating if `TYPE_SATURATING' holds; otherwise, it is not 6470 saturating. 6471 6472 `COMPLEX_TYPE' 6473 Used to represent GCC built-in `__complex__' data types. The 6474 `TREE_TYPE' is the type of the real and imaginary parts. 6475 6476 `ENUMERAL_TYPE' 6477 Used to represent an enumeration type. The `TYPE_PRECISION' gives 6478 (as an `int'), the number of bits used to represent the type. If 6479 there are no negative enumeration constants, `TYPE_UNSIGNED' will 6480 hold. The minimum and maximum enumeration constants may be 6481 obtained with `TYPE_MIN_VALUE' and `TYPE_MAX_VALUE', respectively; 6482 each of these macros returns an `INTEGER_CST'. 6483 6484 The actual enumeration constants themselves may be obtained by 6485 looking at the `TYPE_VALUES'. This macro will return a 6486 `TREE_LIST', containing the constants. The `TREE_PURPOSE' of each 6487 node will be an `IDENTIFIER_NODE' giving the name of the constant; 6488 the `TREE_VALUE' will be an `INTEGER_CST' giving the value 6489 assigned to that constant. These constants will appear in the 6490 order in which they were declared. The `TREE_TYPE' of each of 6491 these constants will be the type of enumeration type itself. 6492 6493 `BOOLEAN_TYPE' 6494 Used to represent the `bool' type. 6495 6496 `POINTER_TYPE' 6497 Used to represent pointer types, and pointer to data member types. 6498 The `TREE_TYPE' gives the type to which this type points. If the 6499 type is a pointer to data member type, then `TYPE_PTRMEM_P' will 6500 hold. For a pointer to data member type of the form `T X::*', 6501 `TYPE_PTRMEM_CLASS_TYPE' will be the type `X', while 6502 `TYPE_PTRMEM_POINTED_TO_TYPE' will be the type `T'. 6503 6504 `REFERENCE_TYPE' 6505 Used to represent reference types. The `TREE_TYPE' gives the type 6506 to which this type refers. 6507 6508 `FUNCTION_TYPE' 6509 Used to represent the type of non-member functions and of static 6510 member functions. The `TREE_TYPE' gives the return type of the 6511 function. The `TYPE_ARG_TYPES' are a `TREE_LIST' of the argument 6512 types. The `TREE_VALUE' of each node in this list is the type of 6513 the corresponding argument; the `TREE_PURPOSE' is an expression 6514 for the default argument value, if any. If the last node in the 6515 list is `void_list_node' (a `TREE_LIST' node whose `TREE_VALUE' is 6516 the `void_type_node'), then functions of this type do not take 6517 variable arguments. Otherwise, they do take a variable number of 6518 arguments. 6519 6520 Note that in C (but not in C++) a function declared like `void f()' 6521 is an unprototyped function taking a variable number of arguments; 6522 the `TYPE_ARG_TYPES' of such a function will be `NULL'. 6523 6524 `METHOD_TYPE' 6525 Used to represent the type of a non-static member function. Like a 6526 `FUNCTION_TYPE', the return type is given by the `TREE_TYPE'. The 6527 type of `*this', i.e., the class of which functions of this type 6528 are a member, is given by the `TYPE_METHOD_BASETYPE'. The 6529 `TYPE_ARG_TYPES' is the parameter list, as for a `FUNCTION_TYPE', 6530 and includes the `this' argument. 6531 6532 `ARRAY_TYPE' 6533 Used to represent array types. The `TREE_TYPE' gives the type of 6534 the elements in the array. If the array-bound is present in the 6535 type, the `TYPE_DOMAIN' is an `INTEGER_TYPE' whose 6536 `TYPE_MIN_VALUE' and `TYPE_MAX_VALUE' will be the lower and upper 6537 bounds of the array, respectively. The `TYPE_MIN_VALUE' will 6538 always be an `INTEGER_CST' for zero, while the `TYPE_MAX_VALUE' 6539 will be one less than the number of elements in the array, i.e., 6540 the highest value which may be used to index an element in the 6541 array. 6542 6543 `RECORD_TYPE' 6544 Used to represent `struct' and `class' types, as well as pointers 6545 to member functions and similar constructs in other languages. 6546 `TYPE_FIELDS' contains the items contained in this type, each of 6547 which can be a `FIELD_DECL', `VAR_DECL', `CONST_DECL', or 6548 `TYPE_DECL'. You may not make any assumptions about the ordering 6549 of the fields in the type or whether one or more of them overlap. 6550 If `TYPE_PTRMEMFUNC_P' holds, then this type is a pointer-to-member 6551 type. In that case, the `TYPE_PTRMEMFUNC_FN_TYPE' is a 6552 `POINTER_TYPE' pointing to a `METHOD_TYPE'. The `METHOD_TYPE' is 6553 the type of a function pointed to by the pointer-to-member 6554 function. If `TYPE_PTRMEMFUNC_P' does not hold, this type is a 6555 class type. For more information, see *note Classes::. 6556 6557 `UNION_TYPE' 6558 Used to represent `union' types. Similar to `RECORD_TYPE' except 6559 that all `FIELD_DECL' nodes in `TYPE_FIELD' start at bit position 6560 zero. 6561 6562 `QUAL_UNION_TYPE' 6563 Used to represent part of a variant record in Ada. Similar to 6564 `UNION_TYPE' except that each `FIELD_DECL' has a `DECL_QUALIFIER' 6565 field, which contains a boolean expression that indicates whether 6566 the field is present in the object. The type will only have one 6567 field, so each field's `DECL_QUALIFIER' is only evaluated if none 6568 of the expressions in the previous fields in `TYPE_FIELDS' are 6569 nonzero. Normally these expressions will reference a field in the 6570 outer object using a `PLACEHOLDER_EXPR'. 6571 6572 `UNKNOWN_TYPE' 6573 This node is used to represent a type the knowledge of which is 6574 insufficient for a sound processing. 6575 6576 `OFFSET_TYPE' 6577 This node is used to represent a pointer-to-data member. For a 6578 data member `X::m' the `TYPE_OFFSET_BASETYPE' is `X' and the 6579 `TREE_TYPE' is the type of `m'. 6580 6581 `TYPENAME_TYPE' 6582 Used to represent a construct of the form `typename T::A'. The 6583 `TYPE_CONTEXT' is `T'; the `TYPE_NAME' is an `IDENTIFIER_NODE' for 6584 `A'. If the type is specified via a template-id, then 6585 `TYPENAME_TYPE_FULLNAME' yields a `TEMPLATE_ID_EXPR'. The 6586 `TREE_TYPE' is non-`NULL' if the node is implicitly generated in 6587 support for the implicit typename extension; in which case the 6588 `TREE_TYPE' is a type node for the base-class. 6589 6590 `TYPEOF_TYPE' 6591 Used to represent the `__typeof__' extension. The `TYPE_FIELDS' 6592 is the expression the type of which is being represented. 6593 6594 There are variables whose values represent some of the basic types. 6595 These include: 6596 `void_type_node' 6597 A node for `void'. 6598 6599 `integer_type_node' 6600 A node for `int'. 6601 6602 `unsigned_type_node.' 6603 A node for `unsigned int'. 6604 6605 `char_type_node.' 6606 A node for `char'. 6607 It may sometimes be useful to compare one of these variables with a 6608 type in hand, using `same_type_p'. 6609 6610 6611 File: gccint.info, Node: Scopes, Next: Functions, Prev: Types, Up: Trees 6612 6613 9.4 Scopes 6614 ========== 6615 6616 The root of the entire intermediate representation is the variable 6617 `global_namespace'. This is the namespace specified with `::' in C++ 6618 source code. All other namespaces, types, variables, functions, and so 6619 forth can be found starting with this namespace. 6620 6621 Besides namespaces, the other high-level scoping construct in C++ is 6622 the class. (Throughout this manual the term "class" is used to mean the 6623 types referred to in the ANSI/ISO C++ Standard as classes; these include 6624 types defined with the `class', `struct', and `union' keywords.) 6625 6626 * Menu: 6627 6628 * Namespaces:: Member functions, types, etc. 6629 * Classes:: Members, bases, friends, etc. 6630 6631 6632 File: gccint.info, Node: Namespaces, Next: Classes, Up: Scopes 6633 6634 9.4.1 Namespaces 6635 ---------------- 6636 6637 A namespace is represented by a `NAMESPACE_DECL' node. 6638 6639 However, except for the fact that it is distinguished as the root of 6640 the representation, the global namespace is no different from any other 6641 namespace. Thus, in what follows, we describe namespaces generally, 6642 rather than the global namespace in particular. 6643 6644 The following macros and functions can be used on a `NAMESPACE_DECL': 6645 6646 `DECL_NAME' 6647 This macro is used to obtain the `IDENTIFIER_NODE' corresponding to 6648 the unqualified name of the name of the namespace (*note 6649 Identifiers::). The name of the global namespace is `::', even 6650 though in C++ the global namespace is unnamed. However, you 6651 should use comparison with `global_namespace', rather than 6652 `DECL_NAME' to determine whether or not a namespace is the global 6653 one. An unnamed namespace will have a `DECL_NAME' equal to 6654 `anonymous_namespace_name'. Within a single translation unit, all 6655 unnamed namespaces will have the same name. 6656 6657 `DECL_CONTEXT' 6658 This macro returns the enclosing namespace. The `DECL_CONTEXT' for 6659 the `global_namespace' is `NULL_TREE'. 6660 6661 `DECL_NAMESPACE_ALIAS' 6662 If this declaration is for a namespace alias, then 6663 `DECL_NAMESPACE_ALIAS' is the namespace for which this one is an 6664 alias. 6665 6666 Do not attempt to use `cp_namespace_decls' for a namespace which is 6667 an alias. Instead, follow `DECL_NAMESPACE_ALIAS' links until you 6668 reach an ordinary, non-alias, namespace, and call 6669 `cp_namespace_decls' there. 6670 6671 `DECL_NAMESPACE_STD_P' 6672 This predicate holds if the namespace is the special `::std' 6673 namespace. 6674 6675 `cp_namespace_decls' 6676 This function will return the declarations contained in the 6677 namespace, including types, overloaded functions, other 6678 namespaces, and so forth. If there are no declarations, this 6679 function will return `NULL_TREE'. The declarations are connected 6680 through their `TREE_CHAIN' fields. 6681 6682 Although most entries on this list will be declarations, 6683 `TREE_LIST' nodes may also appear. In this case, the `TREE_VALUE' 6684 will be an `OVERLOAD'. The value of the `TREE_PURPOSE' is 6685 unspecified; back ends should ignore this value. As with the 6686 other kinds of declarations returned by `cp_namespace_decls', the 6687 `TREE_CHAIN' will point to the next declaration in this list. 6688 6689 For more information on the kinds of declarations that can occur 6690 on this list, *Note Declarations::. Some declarations will not 6691 appear on this list. In particular, no `FIELD_DECL', 6692 `LABEL_DECL', or `PARM_DECL' nodes will appear here. 6693 6694 This function cannot be used with namespaces that have 6695 `DECL_NAMESPACE_ALIAS' set. 6696 6697 6698 6699 File: gccint.info, Node: Classes, Prev: Namespaces, Up: Scopes 6700 6701 9.4.2 Classes 6702 ------------- 6703 6704 A class type is represented by either a `RECORD_TYPE' or a 6705 `UNION_TYPE'. A class declared with the `union' tag is represented by 6706 a `UNION_TYPE', while classes declared with either the `struct' or the 6707 `class' tag are represented by `RECORD_TYPE's. You can use the 6708 `CLASSTYPE_DECLARED_CLASS' macro to discern whether or not a particular 6709 type is a `class' as opposed to a `struct'. This macro will be true 6710 only for classes declared with the `class' tag. 6711 6712 Almost all non-function members are available on the `TYPE_FIELDS' 6713 list. Given one member, the next can be found by following the 6714 `TREE_CHAIN'. You should not depend in any way on the order in which 6715 fields appear on this list. All nodes on this list will be `DECL' 6716 nodes. A `FIELD_DECL' is used to represent a non-static data member, a 6717 `VAR_DECL' is used to represent a static data member, and a `TYPE_DECL' 6718 is used to represent a type. Note that the `CONST_DECL' for an 6719 enumeration constant will appear on this list, if the enumeration type 6720 was declared in the class. (Of course, the `TYPE_DECL' for the 6721 enumeration type will appear here as well.) There are no entries for 6722 base classes on this list. In particular, there is no `FIELD_DECL' for 6723 the "base-class portion" of an object. 6724 6725 The `TYPE_VFIELD' is a compiler-generated field used to point to 6726 virtual function tables. It may or may not appear on the `TYPE_FIELDS' 6727 list. However, back ends should handle the `TYPE_VFIELD' just like all 6728 the entries on the `TYPE_FIELDS' list. 6729 6730 The function members are available on the `TYPE_METHODS' list. Again, 6731 subsequent members are found by following the `TREE_CHAIN' field. If a 6732 function is overloaded, each of the overloaded functions appears; no 6733 `OVERLOAD' nodes appear on the `TYPE_METHODS' list. Implicitly 6734 declared functions (including default constructors, copy constructors, 6735 assignment operators, and destructors) will appear on this list as well. 6736 6737 Every class has an associated "binfo", which can be obtained with 6738 `TYPE_BINFO'. Binfos are used to represent base-classes. The binfo 6739 given by `TYPE_BINFO' is the degenerate case, whereby every class is 6740 considered to be its own base-class. The base binfos for a particular 6741 binfo are held in a vector, whose length is obtained with 6742 `BINFO_N_BASE_BINFOS'. The base binfos themselves are obtained with 6743 `BINFO_BASE_BINFO' and `BINFO_BASE_ITERATE'. To add a new binfo, use 6744 `BINFO_BASE_APPEND'. The vector of base binfos can be obtained with 6745 `BINFO_BASE_BINFOS', but normally you do not need to use that. The 6746 class type associated with a binfo is given by `BINFO_TYPE'. It is not 6747 always the case that `BINFO_TYPE (TYPE_BINFO (x))', because of typedefs 6748 and qualified types. Neither is it the case that `TYPE_BINFO 6749 (BINFO_TYPE (y))' is the same binfo as `y'. The reason is that if `y' 6750 is a binfo representing a base-class `B' of a derived class `D', then 6751 `BINFO_TYPE (y)' will be `B', and `TYPE_BINFO (BINFO_TYPE (y))' will be 6752 `B' as its own base-class, rather than as a base-class of `D'. 6753 6754 The access to a base type can be found with `BINFO_BASE_ACCESS'. This 6755 will produce `access_public_node', `access_private_node' or 6756 `access_protected_node'. If bases are always public, 6757 `BINFO_BASE_ACCESSES' may be `NULL'. 6758 6759 `BINFO_VIRTUAL_P' is used to specify whether the binfo is inherited 6760 virtually or not. The other flags, `BINFO_MARKED_P' and `BINFO_FLAG_1' 6761 to `BINFO_FLAG_6' can be used for language specific use. 6762 6763 The following macros can be used on a tree node representing a 6764 class-type. 6765 6766 `LOCAL_CLASS_P' 6767 This predicate holds if the class is local class _i.e._ declared 6768 inside a function body. 6769 6770 `TYPE_POLYMORPHIC_P' 6771 This predicate holds if the class has at least one virtual function 6772 (declared or inherited). 6773 6774 `TYPE_HAS_DEFAULT_CONSTRUCTOR' 6775 This predicate holds whenever its argument represents a class-type 6776 with default constructor. 6777 6778 `CLASSTYPE_HAS_MUTABLE' 6779 `TYPE_HAS_MUTABLE_P' 6780 These predicates hold for a class-type having a mutable data 6781 member. 6782 6783 `CLASSTYPE_NON_POD_P' 6784 This predicate holds only for class-types that are not PODs. 6785 6786 `TYPE_HAS_NEW_OPERATOR' 6787 This predicate holds for a class-type that defines `operator new'. 6788 6789 `TYPE_HAS_ARRAY_NEW_OPERATOR' 6790 This predicate holds for a class-type for which `operator new[]' 6791 is defined. 6792 6793 `TYPE_OVERLOADS_CALL_EXPR' 6794 This predicate holds for class-type for which the function call 6795 `operator()' is overloaded. 6796 6797 `TYPE_OVERLOADS_ARRAY_REF' 6798 This predicate holds for a class-type that overloads `operator[]' 6799 6800 `TYPE_OVERLOADS_ARROW' 6801 This predicate holds for a class-type for which `operator->' is 6802 overloaded. 6803 6804 6805 6806 File: gccint.info, Node: Declarations, Next: Attributes, Prev: Functions, Up: Trees 6807 6808 9.5 Declarations 6809 ================ 6810 6811 This section covers the various kinds of declarations that appear in the 6812 internal representation, except for declarations of functions 6813 (represented by `FUNCTION_DECL' nodes), which are described in *Note 6814 Functions::. 6815 6816 * Menu: 6817 6818 * Working with declarations:: Macros and functions that work on 6819 declarations. 6820 * Internal structure:: How declaration nodes are represented. 6821 6822 6823 File: gccint.info, Node: Working with declarations, Next: Internal structure, Up: Declarations 6824 6825 9.5.1 Working with declarations 6826 ------------------------------- 6827 6828 Some macros can be used with any kind of declaration. These include: 6829 `DECL_NAME' 6830 This macro returns an `IDENTIFIER_NODE' giving the name of the 6831 entity. 6832 6833 `TREE_TYPE' 6834 This macro returns the type of the entity declared. 6835 6836 `TREE_FILENAME' 6837 This macro returns the name of the file in which the entity was 6838 declared, as a `char*'. For an entity declared implicitly by the 6839 compiler (like `__builtin_memcpy'), this will be the string 6840 `"<internal>"'. 6841 6842 `TREE_LINENO' 6843 This macro returns the line number at which the entity was 6844 declared, as an `int'. 6845 6846 `DECL_ARTIFICIAL' 6847 This predicate holds if the declaration was implicitly generated 6848 by the compiler. For example, this predicate will hold of an 6849 implicitly declared member function, or of the `TYPE_DECL' 6850 implicitly generated for a class type. Recall that in C++ code 6851 like: 6852 struct S {}; 6853 is roughly equivalent to C code like: 6854 struct S {}; 6855 typedef struct S S; 6856 The implicitly generated `typedef' declaration is represented by a 6857 `TYPE_DECL' for which `DECL_ARTIFICIAL' holds. 6858 6859 `DECL_NAMESPACE_SCOPE_P' 6860 This predicate holds if the entity was declared at a namespace 6861 scope. 6862 6863 `DECL_CLASS_SCOPE_P' 6864 This predicate holds if the entity was declared at a class scope. 6865 6866 `DECL_FUNCTION_SCOPE_P' 6867 This predicate holds if the entity was declared inside a function 6868 body. 6869 6870 6871 The various kinds of declarations include: 6872 `LABEL_DECL' 6873 These nodes are used to represent labels in function bodies. For 6874 more information, see *Note Functions::. These nodes only appear 6875 in block scopes. 6876 6877 `CONST_DECL' 6878 These nodes are used to represent enumeration constants. The 6879 value of the constant is given by `DECL_INITIAL' which will be an 6880 `INTEGER_CST' with the same type as the `TREE_TYPE' of the 6881 `CONST_DECL', i.e., an `ENUMERAL_TYPE'. 6882 6883 `RESULT_DECL' 6884 These nodes represent the value returned by a function. When a 6885 value is assigned to a `RESULT_DECL', that indicates that the 6886 value should be returned, via bitwise copy, by the function. You 6887 can use `DECL_SIZE' and `DECL_ALIGN' on a `RESULT_DECL', just as 6888 with a `VAR_DECL'. 6889 6890 `TYPE_DECL' 6891 These nodes represent `typedef' declarations. The `TREE_TYPE' is 6892 the type declared to have the name given by `DECL_NAME'. In some 6893 cases, there is no associated name. 6894 6895 `VAR_DECL' 6896 These nodes represent variables with namespace or block scope, as 6897 well as static data members. The `DECL_SIZE' and `DECL_ALIGN' are 6898 analogous to `TYPE_SIZE' and `TYPE_ALIGN'. For a declaration, you 6899 should always use the `DECL_SIZE' and `DECL_ALIGN' rather than the 6900 `TYPE_SIZE' and `TYPE_ALIGN' given by the `TREE_TYPE', since 6901 special attributes may have been applied to the variable to give 6902 it a particular size and alignment. You may use the predicates 6903 `DECL_THIS_STATIC' or `DECL_THIS_EXTERN' to test whether the 6904 storage class specifiers `static' or `extern' were used to declare 6905 a variable. 6906 6907 If this variable is initialized (but does not require a 6908 constructor), the `DECL_INITIAL' will be an expression for the 6909 initializer. The initializer should be evaluated, and a bitwise 6910 copy into the variable performed. If the `DECL_INITIAL' is the 6911 `error_mark_node', there is an initializer, but it is given by an 6912 explicit statement later in the code; no bitwise copy is required. 6913 6914 GCC provides an extension that allows either automatic variables, 6915 or global variables, to be placed in particular registers. This 6916 extension is being used for a particular `VAR_DECL' if 6917 `DECL_REGISTER' holds for the `VAR_DECL', and if 6918 `DECL_ASSEMBLER_NAME' is not equal to `DECL_NAME'. In that case, 6919 `DECL_ASSEMBLER_NAME' is the name of the register into which the 6920 variable will be placed. 6921 6922 `PARM_DECL' 6923 Used to represent a parameter to a function. Treat these nodes 6924 similarly to `VAR_DECL' nodes. These nodes only appear in the 6925 `DECL_ARGUMENTS' for a `FUNCTION_DECL'. 6926 6927 The `DECL_ARG_TYPE' for a `PARM_DECL' is the type that will 6928 actually be used when a value is passed to this function. It may 6929 be a wider type than the `TREE_TYPE' of the parameter; for 6930 example, the ordinary type might be `short' while the 6931 `DECL_ARG_TYPE' is `int'. 6932 6933 `FIELD_DECL' 6934 These nodes represent non-static data members. The `DECL_SIZE' and 6935 `DECL_ALIGN' behave as for `VAR_DECL' nodes. The position of the 6936 field within the parent record is specified by a combination of 6937 three attributes. `DECL_FIELD_OFFSET' is the position, counting 6938 in bytes, of the `DECL_OFFSET_ALIGN'-bit sized word containing the 6939 bit of the field closest to the beginning of the structure. 6940 `DECL_FIELD_BIT_OFFSET' is the bit offset of the first bit of the 6941 field within this word; this may be nonzero even for fields that 6942 are not bit-fields, since `DECL_OFFSET_ALIGN' may be greater than 6943 the natural alignment of the field's type. 6944 6945 If `DECL_C_BIT_FIELD' holds, this field is a bit-field. In a 6946 bit-field, `DECL_BIT_FIELD_TYPE' also contains the type that was 6947 originally specified for it, while DECL_TYPE may be a modified 6948 type with lesser precision, according to the size of the bit field. 6949 6950 `NAMESPACE_DECL' 6951 *Note Namespaces::. 6952 6953 `TEMPLATE_DECL' 6954 These nodes are used to represent class, function, and variable 6955 (static data member) templates. The 6956 `DECL_TEMPLATE_SPECIALIZATIONS' are a `TREE_LIST'. The 6957 `TREE_VALUE' of each node in the list is a `TEMPLATE_DECL's or 6958 `FUNCTION_DECL's representing specializations (including 6959 instantiations) of this template. Back ends can safely ignore 6960 `TEMPLATE_DECL's, but should examine `FUNCTION_DECL' nodes on the 6961 specializations list just as they would ordinary `FUNCTION_DECL' 6962 nodes. 6963 6964 For a class template, the `DECL_TEMPLATE_INSTANTIATIONS' list 6965 contains the instantiations. The `TREE_VALUE' of each node is an 6966 instantiation of the class. The `DECL_TEMPLATE_SPECIALIZATIONS' 6967 contains partial specializations of the class. 6968 6969 `USING_DECL' 6970 Back ends can safely ignore these nodes. 6971 6972 6973 6974 File: gccint.info, Node: Internal structure, Prev: Working with declarations, Up: Declarations 6975 6976 9.5.2 Internal structure 6977 ------------------------ 6978 6979 `DECL' nodes are represented internally as a hierarchy of structures. 6980 6981 * Menu: 6982 6983 * Current structure hierarchy:: The current DECL node structure 6984 hierarchy. 6985 * Adding new DECL node types:: How to add a new DECL node to a 6986 frontend. 6987 6988 6989 File: gccint.info, Node: Current structure hierarchy, Next: Adding new DECL node types, Up: Internal structure 6990 6991 9.5.2.1 Current structure hierarchy 6992 ................................... 6993 6994 `struct tree_decl_minimal' 6995 This is the minimal structure to inherit from in order for common 6996 `DECL' macros to work. The fields it contains are a unique ID, 6997 source location, context, and name. 6998 6999 `struct tree_decl_common' 7000 This structure inherits from `struct tree_decl_minimal'. It 7001 contains fields that most `DECL' nodes need, such as a field to 7002 store alignment, machine mode, size, and attributes. 7003 7004 `struct tree_field_decl' 7005 This structure inherits from `struct tree_decl_common'. It is 7006 used to represent `FIELD_DECL'. 7007 7008 `struct tree_label_decl' 7009 This structure inherits from `struct tree_decl_common'. It is 7010 used to represent `LABEL_DECL'. 7011 7012 `struct tree_translation_unit_decl' 7013 This structure inherits from `struct tree_decl_common'. It is 7014 used to represent `TRANSLATION_UNIT_DECL'. 7015 7016 `struct tree_decl_with_rtl' 7017 This structure inherits from `struct tree_decl_common'. It 7018 contains a field to store the low-level RTL associated with a 7019 `DECL' node. 7020 7021 `struct tree_result_decl' 7022 This structure inherits from `struct tree_decl_with_rtl'. It is 7023 used to represent `RESULT_DECL'. 7024 7025 `struct tree_const_decl' 7026 This structure inherits from `struct tree_decl_with_rtl'. It is 7027 used to represent `CONST_DECL'. 7028 7029 `struct tree_parm_decl' 7030 This structure inherits from `struct tree_decl_with_rtl'. It is 7031 used to represent `PARM_DECL'. 7032 7033 `struct tree_decl_with_vis' 7034 This structure inherits from `struct tree_decl_with_rtl'. It 7035 contains fields necessary to store visibility information, as well 7036 as a section name and assembler name. 7037 7038 `struct tree_var_decl' 7039 This structure inherits from `struct tree_decl_with_vis'. It is 7040 used to represent `VAR_DECL'. 7041 7042 `struct tree_function_decl' 7043 This structure inherits from `struct tree_decl_with_vis'. It is 7044 used to represent `FUNCTION_DECL'. 7045 7046 7047 7048 File: gccint.info, Node: Adding new DECL node types, Prev: Current structure hierarchy, Up: Internal structure 7049 7050 9.5.2.2 Adding new DECL node types 7051 .................................. 7052 7053 Adding a new `DECL' tree consists of the following steps 7054 7055 Add a new tree code for the `DECL' node 7056 For language specific `DECL' nodes, there is a `.def' file in each 7057 frontend directory where the tree code should be added. For 7058 `DECL' nodes that are part of the middle-end, the code should be 7059 added to `tree.def'. 7060 7061 Create a new structure type for the `DECL' node 7062 These structures should inherit from one of the existing 7063 structures in the language hierarchy by using that structure as 7064 the first member. 7065 7066 struct tree_foo_decl 7067 { 7068 struct tree_decl_with_vis common; 7069 } 7070 7071 Would create a structure name `tree_foo_decl' that inherits from 7072 `struct tree_decl_with_vis'. 7073 7074 For language specific `DECL' nodes, this new structure type should 7075 go in the appropriate `.h' file. For `DECL' nodes that are part 7076 of the middle-end, the structure type should go in `tree.h'. 7077 7078 Add a member to the tree structure enumerator for the node 7079 For garbage collection and dynamic checking purposes, each `DECL' 7080 node structure type is required to have a unique enumerator value 7081 specified with it. For language specific `DECL' nodes, this new 7082 enumerator value should go in the appropriate `.def' file. For 7083 `DECL' nodes that are part of the middle-end, the enumerator 7084 values are specified in `treestruct.def'. 7085 7086 Update `union tree_node' 7087 In order to make your new structure type usable, it must be added 7088 to `union tree_node'. For language specific `DECL' nodes, a new 7089 entry should be added to the appropriate `.h' file of the form 7090 struct tree_foo_decl GTY ((tag ("TS_VAR_DECL"))) foo_decl; 7091 For `DECL' nodes that are part of the middle-end, the additional 7092 member goes directly into `union tree_node' in `tree.h'. 7093 7094 Update dynamic checking info 7095 In order to be able to check whether accessing a named portion of 7096 `union tree_node' is legal, and whether a certain `DECL' node 7097 contains one of the enumerated `DECL' node structures in the 7098 hierarchy, a simple lookup table is used. This lookup table needs 7099 to be kept up to date with the tree structure hierarchy, or else 7100 checking and containment macros will fail inappropriately. 7101 7102 For language specific `DECL' nodes, their is an `init_ts' function 7103 in an appropriate `.c' file, which initializes the lookup table. 7104 Code setting up the table for new `DECL' nodes should be added 7105 there. For each `DECL' tree code and enumerator value 7106 representing a member of the inheritance hierarchy, the table 7107 should contain 1 if that tree code inherits (directly or 7108 indirectly) from that member. Thus, a `FOO_DECL' node derived 7109 from `struct decl_with_rtl', and enumerator value `TS_FOO_DECL', 7110 would be set up as follows 7111 tree_contains_struct[FOO_DECL][TS_FOO_DECL] = 1; 7112 tree_contains_struct[FOO_DECL][TS_DECL_WRTL] = 1; 7113 tree_contains_struct[FOO_DECL][TS_DECL_COMMON] = 1; 7114 tree_contains_struct[FOO_DECL][TS_DECL_MINIMAL] = 1; 7115 7116 For `DECL' nodes that are part of the middle-end, the setup code 7117 goes into `tree.c'. 7118 7119 Add macros to access any new fields and flags 7120 Each added field or flag should have a macro that is used to access 7121 it, that performs appropriate checking to ensure only the right 7122 type of `DECL' nodes access the field. 7123 7124 These macros generally take the following form 7125 #define FOO_DECL_FIELDNAME(NODE) FOO_DECL_CHECK(NODE)->foo_decl.fieldname 7126 However, if the structure is simply a base class for further 7127 structures, something like the following should be used 7128 #define BASE_STRUCT_CHECK(T) CONTAINS_STRUCT_CHECK(T, TS_BASE_STRUCT) 7129 #define BASE_STRUCT_FIELDNAME(NODE) \ 7130 (BASE_STRUCT_CHECK(NODE)->base_struct.fieldname 7131 7132 7133 7134 File: gccint.info, Node: Functions, Next: Declarations, Prev: Scopes, Up: Trees 7135 7136 9.6 Functions 7137 ============= 7138 7139 A function is represented by a `FUNCTION_DECL' node. A set of 7140 overloaded functions is sometimes represented by a `OVERLOAD' node. 7141 7142 An `OVERLOAD' node is not a declaration, so none of the `DECL_' macros 7143 should be used on an `OVERLOAD'. An `OVERLOAD' node is similar to a 7144 `TREE_LIST'. Use `OVL_CURRENT' to get the function associated with an 7145 `OVERLOAD' node; use `OVL_NEXT' to get the next `OVERLOAD' node in the 7146 list of overloaded functions. The macros `OVL_CURRENT' and `OVL_NEXT' 7147 are actually polymorphic; you can use them to work with `FUNCTION_DECL' 7148 nodes as well as with overloads. In the case of a `FUNCTION_DECL', 7149 `OVL_CURRENT' will always return the function itself, and `OVL_NEXT' 7150 will always be `NULL_TREE'. 7151 7152 To determine the scope of a function, you can use the `DECL_CONTEXT' 7153 macro. This macro will return the class (either a `RECORD_TYPE' or a 7154 `UNION_TYPE') or namespace (a `NAMESPACE_DECL') of which the function 7155 is a member. For a virtual function, this macro returns the class in 7156 which the function was actually defined, not the base class in which 7157 the virtual declaration occurred. 7158 7159 If a friend function is defined in a class scope, the 7160 `DECL_FRIEND_CONTEXT' macro can be used to determine the class in which 7161 it was defined. For example, in 7162 class C { friend void f() {} }; 7163 the `DECL_CONTEXT' for `f' will be the `global_namespace', but the 7164 `DECL_FRIEND_CONTEXT' will be the `RECORD_TYPE' for `C'. 7165 7166 In C, the `DECL_CONTEXT' for a function maybe another function. This 7167 representation indicates that the GNU nested function extension is in 7168 use. For details on the semantics of nested functions, see the GCC 7169 Manual. The nested function can refer to local variables in its 7170 containing function. Such references are not explicitly marked in the 7171 tree structure; back ends must look at the `DECL_CONTEXT' for the 7172 referenced `VAR_DECL'. If the `DECL_CONTEXT' for the referenced 7173 `VAR_DECL' is not the same as the function currently being processed, 7174 and neither `DECL_EXTERNAL' nor `TREE_STATIC' hold, then the reference 7175 is to a local variable in a containing function, and the back end must 7176 take appropriate action. 7177 7178 * Menu: 7179 7180 * Function Basics:: Function names, linkage, and so forth. 7181 * Function Bodies:: The statements that make up a function body. 7182 7183 7184 File: gccint.info, Node: Function Basics, Next: Function Bodies, Up: Functions 7185 7186 9.6.1 Function Basics 7187 --------------------- 7188 7189 The following macros and functions can be used on a `FUNCTION_DECL': 7190 `DECL_MAIN_P' 7191 This predicate holds for a function that is the program entry point 7192 `::code'. 7193 7194 `DECL_NAME' 7195 This macro returns the unqualified name of the function, as an 7196 `IDENTIFIER_NODE'. For an instantiation of a function template, 7197 the `DECL_NAME' is the unqualified name of the template, not 7198 something like `f<int>'. The value of `DECL_NAME' is undefined 7199 when used on a constructor, destructor, overloaded operator, or 7200 type-conversion operator, or any function that is implicitly 7201 generated by the compiler. See below for macros that can be used 7202 to distinguish these cases. 7203 7204 `DECL_ASSEMBLER_NAME' 7205 This macro returns the mangled name of the function, also an 7206 `IDENTIFIER_NODE'. This name does not contain leading underscores 7207 on systems that prefix all identifiers with underscores. The 7208 mangled name is computed in the same way on all platforms; if 7209 special processing is required to deal with the object file format 7210 used on a particular platform, it is the responsibility of the 7211 back end to perform those modifications. (Of course, the back end 7212 should not modify `DECL_ASSEMBLER_NAME' itself.) 7213 7214 Using `DECL_ASSEMBLER_NAME' will cause additional memory to be 7215 allocated (for the mangled name of the entity) so it should be used 7216 only when emitting assembly code. It should not be used within the 7217 optimizers to determine whether or not two declarations are the 7218 same, even though some of the existing optimizers do use it in 7219 that way. These uses will be removed over time. 7220 7221 `DECL_EXTERNAL' 7222 This predicate holds if the function is undefined. 7223 7224 `TREE_PUBLIC' 7225 This predicate holds if the function has external linkage. 7226 7227 `DECL_LOCAL_FUNCTION_P' 7228 This predicate holds if the function was declared at block scope, 7229 even though it has a global scope. 7230 7231 `DECL_ANTICIPATED' 7232 This predicate holds if the function is a built-in function but its 7233 prototype is not yet explicitly declared. 7234 7235 `DECL_EXTERN_C_FUNCTION_P' 7236 This predicate holds if the function is declared as an ``extern 7237 "C"'' function. 7238 7239 `DECL_LINKONCE_P' 7240 This macro holds if multiple copies of this function may be 7241 emitted in various translation units. It is the responsibility of 7242 the linker to merge the various copies. Template instantiations 7243 are the most common example of functions for which 7244 `DECL_LINKONCE_P' holds; G++ instantiates needed templates in all 7245 translation units which require them, and then relies on the 7246 linker to remove duplicate instantiations. 7247 7248 FIXME: This macro is not yet implemented. 7249 7250 `DECL_FUNCTION_MEMBER_P' 7251 This macro holds if the function is a member of a class, rather 7252 than a member of a namespace. 7253 7254 `DECL_STATIC_FUNCTION_P' 7255 This predicate holds if the function a static member function. 7256 7257 `DECL_NONSTATIC_MEMBER_FUNCTION_P' 7258 This macro holds for a non-static member function. 7259 7260 `DECL_CONST_MEMFUNC_P' 7261 This predicate holds for a `const'-member function. 7262 7263 `DECL_VOLATILE_MEMFUNC_P' 7264 This predicate holds for a `volatile'-member function. 7265 7266 `DECL_CONSTRUCTOR_P' 7267 This macro holds if the function is a constructor. 7268 7269 `DECL_NONCONVERTING_P' 7270 This predicate holds if the constructor is a non-converting 7271 constructor. 7272 7273 `DECL_COMPLETE_CONSTRUCTOR_P' 7274 This predicate holds for a function which is a constructor for an 7275 object of a complete type. 7276 7277 `DECL_BASE_CONSTRUCTOR_P' 7278 This predicate holds for a function which is a constructor for a 7279 base class sub-object. 7280 7281 `DECL_COPY_CONSTRUCTOR_P' 7282 This predicate holds for a function which is a copy-constructor. 7283 7284 `DECL_DESTRUCTOR_P' 7285 This macro holds if the function is a destructor. 7286 7287 `DECL_COMPLETE_DESTRUCTOR_P' 7288 This predicate holds if the function is the destructor for an 7289 object a complete type. 7290 7291 `DECL_OVERLOADED_OPERATOR_P' 7292 This macro holds if the function is an overloaded operator. 7293 7294 `DECL_CONV_FN_P' 7295 This macro holds if the function is a type-conversion operator. 7296 7297 `DECL_GLOBAL_CTOR_P' 7298 This predicate holds if the function is a file-scope initialization 7299 function. 7300 7301 `DECL_GLOBAL_DTOR_P' 7302 This predicate holds if the function is a file-scope finalization 7303 function. 7304 7305 `DECL_THUNK_P' 7306 This predicate holds if the function is a thunk. 7307 7308 These functions represent stub code that adjusts the `this' pointer 7309 and then jumps to another function. When the jumped-to function 7310 returns, control is transferred directly to the caller, without 7311 returning to the thunk. The first parameter to the thunk is 7312 always the `this' pointer; the thunk should add `THUNK_DELTA' to 7313 this value. (The `THUNK_DELTA' is an `int', not an `INTEGER_CST'.) 7314 7315 Then, if `THUNK_VCALL_OFFSET' (an `INTEGER_CST') is nonzero the 7316 adjusted `this' pointer must be adjusted again. The complete 7317 calculation is given by the following pseudo-code: 7318 7319 this += THUNK_DELTA 7320 if (THUNK_VCALL_OFFSET) 7321 this += (*((ptrdiff_t **) this))[THUNK_VCALL_OFFSET] 7322 7323 Finally, the thunk should jump to the location given by 7324 `DECL_INITIAL'; this will always be an expression for the address 7325 of a function. 7326 7327 `DECL_NON_THUNK_FUNCTION_P' 7328 This predicate holds if the function is _not_ a thunk function. 7329 7330 `GLOBAL_INIT_PRIORITY' 7331 If either `DECL_GLOBAL_CTOR_P' or `DECL_GLOBAL_DTOR_P' holds, then 7332 this gives the initialization priority for the function. The 7333 linker will arrange that all functions for which 7334 `DECL_GLOBAL_CTOR_P' holds are run in increasing order of priority 7335 before `main' is called. When the program exits, all functions for 7336 which `DECL_GLOBAL_DTOR_P' holds are run in the reverse order. 7337 7338 `DECL_ARTIFICIAL' 7339 This macro holds if the function was implicitly generated by the 7340 compiler, rather than explicitly declared. In addition to 7341 implicitly generated class member functions, this macro holds for 7342 the special functions created to implement static initialization 7343 and destruction, to compute run-time type information, and so 7344 forth. 7345 7346 `DECL_ARGUMENTS' 7347 This macro returns the `PARM_DECL' for the first argument to the 7348 function. Subsequent `PARM_DECL' nodes can be obtained by 7349 following the `TREE_CHAIN' links. 7350 7351 `DECL_RESULT' 7352 This macro returns the `RESULT_DECL' for the function. 7353 7354 `TREE_TYPE' 7355 This macro returns the `FUNCTION_TYPE' or `METHOD_TYPE' for the 7356 function. 7357 7358 `TYPE_RAISES_EXCEPTIONS' 7359 This macro returns the list of exceptions that a (member-)function 7360 can raise. The returned list, if non `NULL', is comprised of nodes 7361 whose `TREE_VALUE' represents a type. 7362 7363 `TYPE_NOTHROW_P' 7364 This predicate holds when the exception-specification of its 7365 arguments is of the form ``()''. 7366 7367 `DECL_ARRAY_DELETE_OPERATOR_P' 7368 This predicate holds if the function an overloaded `operator 7369 delete[]'. 7370 7371 `DECL_FUNCTION_SPECIFIC_TARGET' 7372 This macro returns a tree node that holds the target options that 7373 are to be used to compile this particular function or `NULL_TREE' 7374 if the function is to be compiled with the target options 7375 specified on the command line. 7376 7377 `DECL_FUNCTION_SPECIFIC_OPTIMIZATION' 7378 This macro returns a tree node that holds the optimization options 7379 that are to be used to compile this particular function or 7380 `NULL_TREE' if the function is to be compiled with the 7381 optimization options specified on the command line. 7382 7383 7384 File: gccint.info, Node: Function Bodies, Prev: Function Basics, Up: Functions 7385 7386 9.6.2 Function Bodies 7387 --------------------- 7388 7389 A function that has a definition in the current translation unit will 7390 have a non-`NULL' `DECL_INITIAL'. However, back ends should not make 7391 use of the particular value given by `DECL_INITIAL'. 7392 7393 The `DECL_SAVED_TREE' macro will give the complete body of the 7394 function. 7395 7396 9.6.2.1 Statements 7397 .................. 7398 7399 There are tree nodes corresponding to all of the source-level statement 7400 constructs, used within the C and C++ frontends. These are enumerated 7401 here, together with a list of the various macros that can be used to 7402 obtain information about them. There are a few macros that can be used 7403 with all statements: 7404 7405 `STMT_IS_FULL_EXPR_P' 7406 In C++, statements normally constitute "full expressions"; 7407 temporaries created during a statement are destroyed when the 7408 statement is complete. However, G++ sometimes represents 7409 expressions by statements; these statements will not have 7410 `STMT_IS_FULL_EXPR_P' set. Temporaries created during such 7411 statements should be destroyed when the innermost enclosing 7412 statement with `STMT_IS_FULL_EXPR_P' set is exited. 7413 7414 7415 Here is the list of the various statement nodes, and the macros used to 7416 access them. This documentation describes the use of these nodes in 7417 non-template functions (including instantiations of template functions). 7418 In template functions, the same nodes are used, but sometimes in 7419 slightly different ways. 7420 7421 Many of the statements have substatements. For example, a `while' 7422 loop will have a body, which is itself a statement. If the substatement 7423 is `NULL_TREE', it is considered equivalent to a statement consisting 7424 of a single `;', i.e., an expression statement in which the expression 7425 has been omitted. A substatement may in fact be a list of statements, 7426 connected via their `TREE_CHAIN's. So, you should always process the 7427 statement tree by looping over substatements, like this: 7428 void process_stmt (stmt) 7429 tree stmt; 7430 { 7431 while (stmt) 7432 { 7433 switch (TREE_CODE (stmt)) 7434 { 7435 case IF_STMT: 7436 process_stmt (THEN_CLAUSE (stmt)); 7437 /* More processing here. */ 7438 break; 7439 7440 ... 7441 } 7442 7443 stmt = TREE_CHAIN (stmt); 7444 } 7445 } 7446 In other words, while the `then' clause of an `if' statement in C++ 7447 can be only one statement (although that one statement may be a 7448 compound statement), the intermediate representation will sometimes use 7449 several statements chained together. 7450 7451 `ASM_EXPR' 7452 Used to represent an inline assembly statement. For an inline 7453 assembly statement like: 7454 asm ("mov x, y"); 7455 The `ASM_STRING' macro will return a `STRING_CST' node for `"mov 7456 x, y"'. If the original statement made use of the 7457 extended-assembly syntax, then `ASM_OUTPUTS', `ASM_INPUTS', and 7458 `ASM_CLOBBERS' will be the outputs, inputs, and clobbers for the 7459 statement, represented as `STRING_CST' nodes. The 7460 extended-assembly syntax looks like: 7461 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle)); 7462 The first string is the `ASM_STRING', containing the instruction 7463 template. The next two strings are the output and inputs, 7464 respectively; this statement has no clobbers. As this example 7465 indicates, "plain" assembly statements are merely a special case 7466 of extended assembly statements; they have no cv-qualifiers, 7467 outputs, inputs, or clobbers. All of the strings will be 7468 `NUL'-terminated, and will contain no embedded `NUL'-characters. 7469 7470 If the assembly statement is declared `volatile', or if the 7471 statement was not an extended assembly statement, and is therefore 7472 implicitly volatile, then the predicate `ASM_VOLATILE_P' will hold 7473 of the `ASM_EXPR'. 7474 7475 `BREAK_STMT' 7476 Used to represent a `break' statement. There are no additional 7477 fields. 7478 7479 `CASE_LABEL_EXPR' 7480 Use to represent a `case' label, range of `case' labels, or a 7481 `default' label. If `CASE_LOW' is `NULL_TREE', then this is a 7482 `default' label. Otherwise, if `CASE_HIGH' is `NULL_TREE', then 7483 this is an ordinary `case' label. In this case, `CASE_LOW' is an 7484 expression giving the value of the label. Both `CASE_LOW' and 7485 `CASE_HIGH' are `INTEGER_CST' nodes. These values will have the 7486 same type as the condition expression in the switch statement. 7487 7488 Otherwise, if both `CASE_LOW' and `CASE_HIGH' are defined, the 7489 statement is a range of case labels. Such statements originate 7490 with the extension that allows users to write things of the form: 7491 case 2 ... 5: 7492 The first value will be `CASE_LOW', while the second will be 7493 `CASE_HIGH'. 7494 7495 `CLEANUP_STMT' 7496 Used to represent an action that should take place upon exit from 7497 the enclosing scope. Typically, these actions are calls to 7498 destructors for local objects, but back ends cannot rely on this 7499 fact. If these nodes are in fact representing such destructors, 7500 `CLEANUP_DECL' will be the `VAR_DECL' destroyed. Otherwise, 7501 `CLEANUP_DECL' will be `NULL_TREE'. In any case, the 7502 `CLEANUP_EXPR' is the expression to execute. The cleanups 7503 executed on exit from a scope should be run in the reverse order 7504 of the order in which the associated `CLEANUP_STMT's were 7505 encountered. 7506 7507 `CONTINUE_STMT' 7508 Used to represent a `continue' statement. There are no additional 7509 fields. 7510 7511 `CTOR_STMT' 7512 Used to mark the beginning (if `CTOR_BEGIN_P' holds) or end (if 7513 `CTOR_END_P' holds of the main body of a constructor. See also 7514 `SUBOBJECT' for more information on how to use these nodes. 7515 7516 `DECL_STMT' 7517 Used to represent a local declaration. The `DECL_STMT_DECL' macro 7518 can be used to obtain the entity declared. This declaration may 7519 be a `LABEL_DECL', indicating that the label declared is a local 7520 label. (As an extension, GCC allows the declaration of labels 7521 with scope.) In C, this declaration may be a `FUNCTION_DECL', 7522 indicating the use of the GCC nested function extension. For more 7523 information, *note Functions::. 7524 7525 `DO_STMT' 7526 Used to represent a `do' loop. The body of the loop is given by 7527 `DO_BODY' while the termination condition for the loop is given by 7528 `DO_COND'. The condition for a `do'-statement is always an 7529 expression. 7530 7531 `EMPTY_CLASS_EXPR' 7532 Used to represent a temporary object of a class with no data whose 7533 address is never taken. (All such objects are interchangeable.) 7534 The `TREE_TYPE' represents the type of the object. 7535 7536 `EXPR_STMT' 7537 Used to represent an expression statement. Use `EXPR_STMT_EXPR' to 7538 obtain the expression. 7539 7540 `FOR_STMT' 7541 Used to represent a `for' statement. The `FOR_INIT_STMT' is the 7542 initialization statement for the loop. The `FOR_COND' is the 7543 termination condition. The `FOR_EXPR' is the expression executed 7544 right before the `FOR_COND' on each loop iteration; often, this 7545 expression increments a counter. The body of the loop is given by 7546 `FOR_BODY'. Note that `FOR_INIT_STMT' and `FOR_BODY' return 7547 statements, while `FOR_COND' and `FOR_EXPR' return expressions. 7548 7549 `GOTO_EXPR' 7550 Used to represent a `goto' statement. The `GOTO_DESTINATION' will 7551 usually be a `LABEL_DECL'. However, if the "computed goto" 7552 extension has been used, the `GOTO_DESTINATION' will be an 7553 arbitrary expression indicating the destination. This expression 7554 will always have pointer type. 7555 7556 `HANDLER' 7557 Used to represent a C++ `catch' block. The `HANDLER_TYPE' is the 7558 type of exception that will be caught by this handler; it is equal 7559 (by pointer equality) to `NULL' if this handler is for all types. 7560 `HANDLER_PARMS' is the `DECL_STMT' for the catch parameter, and 7561 `HANDLER_BODY' is the code for the block itself. 7562 7563 `IF_STMT' 7564 Used to represent an `if' statement. The `IF_COND' is the 7565 expression. 7566 7567 If the condition is a `TREE_LIST', then the `TREE_PURPOSE' is a 7568 statement (usually a `DECL_STMT'). Each time the condition is 7569 evaluated, the statement should be executed. Then, the 7570 `TREE_VALUE' should be used as the conditional expression itself. 7571 This representation is used to handle C++ code like this: 7572 7573 if (int i = 7) ... 7574 7575 where there is a new local variable (or variables) declared within 7576 the condition. 7577 7578 The `THEN_CLAUSE' represents the statement given by the `then' 7579 condition, while the `ELSE_CLAUSE' represents the statement given 7580 by the `else' condition. 7581 7582 `LABEL_EXPR' 7583 Used to represent a label. The `LABEL_DECL' declared by this 7584 statement can be obtained with the `LABEL_EXPR_LABEL' macro. The 7585 `IDENTIFIER_NODE' giving the name of the label can be obtained from 7586 the `LABEL_DECL' with `DECL_NAME'. 7587 7588 `RETURN_STMT' 7589 Used to represent a `return' statement. The `RETURN_EXPR' is the 7590 expression returned; it will be `NULL_TREE' if the statement was 7591 just 7592 return; 7593 7594 `SUBOBJECT' 7595 In a constructor, these nodes are used to mark the point at which a 7596 subobject of `this' is fully constructed. If, after this point, an 7597 exception is thrown before a `CTOR_STMT' with `CTOR_END_P' set is 7598 encountered, the `SUBOBJECT_CLEANUP' must be executed. The 7599 cleanups must be executed in the reverse order in which they 7600 appear. 7601 7602 `SWITCH_STMT' 7603 Used to represent a `switch' statement. The `SWITCH_STMT_COND' is 7604 the expression on which the switch is occurring. See the 7605 documentation for an `IF_STMT' for more information on the 7606 representation used for the condition. The `SWITCH_STMT_BODY' is 7607 the body of the switch statement. The `SWITCH_STMT_TYPE' is the 7608 original type of switch expression as given in the source, before 7609 any compiler conversions. 7610 7611 `TRY_BLOCK' 7612 Used to represent a `try' block. The body of the try block is 7613 given by `TRY_STMTS'. Each of the catch blocks is a `HANDLER' 7614 node. The first handler is given by `TRY_HANDLERS'. Subsequent 7615 handlers are obtained by following the `TREE_CHAIN' link from one 7616 handler to the next. The body of the handler is given by 7617 `HANDLER_BODY'. 7618 7619 If `CLEANUP_P' holds of the `TRY_BLOCK', then the `TRY_HANDLERS' 7620 will not be a `HANDLER' node. Instead, it will be an expression 7621 that should be executed if an exception is thrown in the try 7622 block. It must rethrow the exception after executing that code. 7623 And, if an exception is thrown while the expression is executing, 7624 `terminate' must be called. 7625 7626 `USING_STMT' 7627 Used to represent a `using' directive. The namespace is given by 7628 `USING_STMT_NAMESPACE', which will be a NAMESPACE_DECL. This node 7629 is needed inside template functions, to implement using directives 7630 during instantiation. 7631 7632 `WHILE_STMT' 7633 Used to represent a `while' loop. The `WHILE_COND' is the 7634 termination condition for the loop. See the documentation for an 7635 `IF_STMT' for more information on the representation used for the 7636 condition. 7637 7638 The `WHILE_BODY' is the body of the loop. 7639 7640 7641 7642 File: gccint.info, Node: Attributes, Next: Expression trees, Prev: Declarations, Up: Trees 7643 7644 9.7 Attributes in trees 7645 ======================= 7646 7647 Attributes, as specified using the `__attribute__' keyword, are 7648 represented internally as a `TREE_LIST'. The `TREE_PURPOSE' is the 7649 name of the attribute, as an `IDENTIFIER_NODE'. The `TREE_VALUE' is a 7650 `TREE_LIST' of the arguments of the attribute, if any, or `NULL_TREE' 7651 if there are no arguments; the arguments are stored as the `TREE_VALUE' 7652 of successive entries in the list, and may be identifiers or 7653 expressions. The `TREE_CHAIN' of the attribute is the next attribute 7654 in a list of attributes applying to the same declaration or type, or 7655 `NULL_TREE' if there are no further attributes in the list. 7656 7657 Attributes may be attached to declarations and to types; these 7658 attributes may be accessed with the following macros. All attributes 7659 are stored in this way, and many also cause other changes to the 7660 declaration or type or to other internal compiler data structures. 7661 7662 -- Tree Macro: tree DECL_ATTRIBUTES (tree DECL) 7663 This macro returns the attributes on the declaration DECL. 7664 7665 -- Tree Macro: tree TYPE_ATTRIBUTES (tree TYPE) 7666 This macro returns the attributes on the type TYPE. 7667 7668 7669 File: gccint.info, Node: Expression trees, Prev: Attributes, Up: Trees 7670 7671 9.8 Expressions 7672 =============== 7673 7674 The internal representation for expressions is for the most part quite 7675 straightforward. However, there are a few facts that one must bear in 7676 mind. In particular, the expression "tree" is actually a directed 7677 acyclic graph. (For example there may be many references to the integer 7678 constant zero throughout the source program; many of these will be 7679 represented by the same expression node.) You should not rely on 7680 certain kinds of node being shared, nor should you rely on certain 7681 kinds of nodes being unshared. 7682 7683 The following macros can be used with all expression nodes: 7684 7685 `TREE_TYPE' 7686 Returns the type of the expression. This value may not be 7687 precisely the same type that would be given the expression in the 7688 original program. 7689 7690 In what follows, some nodes that one might expect to always have type 7691 `bool' are documented to have either integral or boolean type. At some 7692 point in the future, the C front end may also make use of this same 7693 intermediate representation, and at this point these nodes will 7694 certainly have integral type. The previous sentence is not meant to 7695 imply that the C++ front end does not or will not give these nodes 7696 integral type. 7697 7698 Below, we list the various kinds of expression nodes. Except where 7699 noted otherwise, the operands to an expression are accessed using the 7700 `TREE_OPERAND' macro. For example, to access the first operand to a 7701 binary plus expression `expr', use: 7702 7703 TREE_OPERAND (expr, 0) 7704 As this example indicates, the operands are zero-indexed. 7705 7706 All the expressions starting with `OMP_' represent directives and 7707 clauses used by the OpenMP API `http://www.openmp.org/'. 7708 7709 The table below begins with constants, moves on to unary expressions, 7710 then proceeds to binary expressions, and concludes with various other 7711 kinds of expressions: 7712 7713 `INTEGER_CST' 7714 These nodes represent integer constants. Note that the type of 7715 these constants is obtained with `TREE_TYPE'; they are not always 7716 of type `int'. In particular, `char' constants are represented 7717 with `INTEGER_CST' nodes. The value of the integer constant `e' is 7718 given by 7719 ((TREE_INT_CST_HIGH (e) << HOST_BITS_PER_WIDE_INT) 7720 + TREE_INST_CST_LOW (e)) 7721 HOST_BITS_PER_WIDE_INT is at least thirty-two on all platforms. 7722 Both `TREE_INT_CST_HIGH' and `TREE_INT_CST_LOW' return a 7723 `HOST_WIDE_INT'. The value of an `INTEGER_CST' is interpreted as 7724 a signed or unsigned quantity depending on the type of the 7725 constant. In general, the expression given above will overflow, 7726 so it should not be used to calculate the value of the constant. 7727 7728 The variable `integer_zero_node' is an integer constant with value 7729 zero. Similarly, `integer_one_node' is an integer constant with 7730 value one. The `size_zero_node' and `size_one_node' variables are 7731 analogous, but have type `size_t' rather than `int'. 7732 7733 The function `tree_int_cst_lt' is a predicate which holds if its 7734 first argument is less than its second. Both constants are 7735 assumed to have the same signedness (i.e., either both should be 7736 signed or both should be unsigned.) The full width of the 7737 constant is used when doing the comparison; the usual rules about 7738 promotions and conversions are ignored. Similarly, 7739 `tree_int_cst_equal' holds if the two constants are equal. The 7740 `tree_int_cst_sgn' function returns the sign of a constant. The 7741 value is `1', `0', or `-1' according on whether the constant is 7742 greater than, equal to, or less than zero. Again, the signedness 7743 of the constant's type is taken into account; an unsigned constant 7744 is never less than zero, no matter what its bit-pattern. 7745 7746 `REAL_CST' 7747 FIXME: Talk about how to obtain representations of this constant, 7748 do comparisons, and so forth. 7749 7750 `FIXED_CST' 7751 These nodes represent fixed-point constants. The type of these 7752 constants is obtained with `TREE_TYPE'. `TREE_FIXED_CST_PTR' 7753 points to to struct fixed_value; `TREE_FIXED_CST' returns the 7754 structure itself. Struct fixed_value contains `data' with the 7755 size of two HOST_BITS_PER_WIDE_INT and `mode' as the associated 7756 fixed-point machine mode for `data'. 7757 7758 `COMPLEX_CST' 7759 These nodes are used to represent complex number constants, that 7760 is a `__complex__' whose parts are constant nodes. The 7761 `TREE_REALPART' and `TREE_IMAGPART' return the real and the 7762 imaginary parts respectively. 7763 7764 `VECTOR_CST' 7765 These nodes are used to represent vector constants, whose parts are 7766 constant nodes. Each individual constant node is either an 7767 integer or a double constant node. The first operand is a 7768 `TREE_LIST' of the constant nodes and is accessed through 7769 `TREE_VECTOR_CST_ELTS'. 7770 7771 `STRING_CST' 7772 These nodes represent string-constants. The `TREE_STRING_LENGTH' 7773 returns the length of the string, as an `int'. The 7774 `TREE_STRING_POINTER' is a `char*' containing the string itself. 7775 The string may not be `NUL'-terminated, and it may contain 7776 embedded `NUL' characters. Therefore, the `TREE_STRING_LENGTH' 7777 includes the trailing `NUL' if it is present. 7778 7779 For wide string constants, the `TREE_STRING_LENGTH' is the number 7780 of bytes in the string, and the `TREE_STRING_POINTER' points to an 7781 array of the bytes of the string, as represented on the target 7782 system (that is, as integers in the target endianness). Wide and 7783 non-wide string constants are distinguished only by the `TREE_TYPE' 7784 of the `STRING_CST'. 7785 7786 FIXME: The formats of string constants are not well-defined when 7787 the target system bytes are not the same width as host system 7788 bytes. 7789 7790 `PTRMEM_CST' 7791 These nodes are used to represent pointer-to-member constants. The 7792 `PTRMEM_CST_CLASS' is the class type (either a `RECORD_TYPE' or 7793 `UNION_TYPE' within which the pointer points), and the 7794 `PTRMEM_CST_MEMBER' is the declaration for the pointed to object. 7795 Note that the `DECL_CONTEXT' for the `PTRMEM_CST_MEMBER' is in 7796 general different from the `PTRMEM_CST_CLASS'. For example, given: 7797 struct B { int i; }; 7798 struct D : public B {}; 7799 int D::*dp = &D::i; 7800 The `PTRMEM_CST_CLASS' for `&D::i' is `D', even though the 7801 `DECL_CONTEXT' for the `PTRMEM_CST_MEMBER' is `B', since `B::i' is 7802 a member of `B', not `D'. 7803 7804 `VAR_DECL' 7805 These nodes represent variables, including static data members. 7806 For more information, *note Declarations::. 7807 7808 `NEGATE_EXPR' 7809 These nodes represent unary negation of the single operand, for 7810 both integer and floating-point types. The type of negation can be 7811 determined by looking at the type of the expression. 7812 7813 The behavior of this operation on signed arithmetic overflow is 7814 controlled by the `flag_wrapv' and `flag_trapv' variables. 7815 7816 `ABS_EXPR' 7817 These nodes represent the absolute value of the single operand, for 7818 both integer and floating-point types. This is typically used to 7819 implement the `abs', `labs' and `llabs' builtins for integer 7820 types, and the `fabs', `fabsf' and `fabsl' builtins for floating 7821 point types. The type of abs operation can be determined by 7822 looking at the type of the expression. 7823 7824 This node is not used for complex types. To represent the modulus 7825 or complex abs of a complex value, use the `BUILT_IN_CABS', 7826 `BUILT_IN_CABSF' or `BUILT_IN_CABSL' builtins, as used to 7827 implement the C99 `cabs', `cabsf' and `cabsl' built-in functions. 7828 7829 `BIT_NOT_EXPR' 7830 These nodes represent bitwise complement, and will always have 7831 integral type. The only operand is the value to be complemented. 7832 7833 `TRUTH_NOT_EXPR' 7834 These nodes represent logical negation, and will always have 7835 integral (or boolean) type. The operand is the value being 7836 negated. The type of the operand and that of the result are 7837 always of `BOOLEAN_TYPE' or `INTEGER_TYPE'. 7838 7839 `PREDECREMENT_EXPR' 7840 `PREINCREMENT_EXPR' 7841 `POSTDECREMENT_EXPR' 7842 `POSTINCREMENT_EXPR' 7843 These nodes represent increment and decrement expressions. The 7844 value of the single operand is computed, and the operand 7845 incremented or decremented. In the case of `PREDECREMENT_EXPR' and 7846 `PREINCREMENT_EXPR', the value of the expression is the value 7847 resulting after the increment or decrement; in the case of 7848 `POSTDECREMENT_EXPR' and `POSTINCREMENT_EXPR' is the value before 7849 the increment or decrement occurs. The type of the operand, like 7850 that of the result, will be either integral, boolean, or 7851 floating-point. 7852 7853 `ADDR_EXPR' 7854 These nodes are used to represent the address of an object. (These 7855 expressions will always have pointer or reference type.) The 7856 operand may be another expression, or it may be a declaration. 7857 7858 As an extension, GCC allows users to take the address of a label. 7859 In this case, the operand of the `ADDR_EXPR' will be a 7860 `LABEL_DECL'. The type of such an expression is `void*'. 7861 7862 If the object addressed is not an lvalue, a temporary is created, 7863 and the address of the temporary is used. 7864 7865 `INDIRECT_REF' 7866 These nodes are used to represent the object pointed to by a 7867 pointer. The operand is the pointer being dereferenced; it will 7868 always have pointer or reference type. 7869 7870 `FIX_TRUNC_EXPR' 7871 These nodes represent conversion of a floating-point value to an 7872 integer. The single operand will have a floating-point type, while 7873 the complete expression will have an integral (or boolean) type. 7874 The operand is rounded towards zero. 7875 7876 `FLOAT_EXPR' 7877 These nodes represent conversion of an integral (or boolean) value 7878 to a floating-point value. The single operand will have integral 7879 type, while the complete expression will have a floating-point 7880 type. 7881 7882 FIXME: How is the operand supposed to be rounded? Is this 7883 dependent on `-mieee'? 7884 7885 `COMPLEX_EXPR' 7886 These nodes are used to represent complex numbers constructed from 7887 two expressions of the same (integer or real) type. The first 7888 operand is the real part and the second operand is the imaginary 7889 part. 7890 7891 `CONJ_EXPR' 7892 These nodes represent the conjugate of their operand. 7893 7894 `REALPART_EXPR' 7895 `IMAGPART_EXPR' 7896 These nodes represent respectively the real and the imaginary parts 7897 of complex numbers (their sole argument). 7898 7899 `NON_LVALUE_EXPR' 7900 These nodes indicate that their one and only operand is not an 7901 lvalue. A back end can treat these identically to the single 7902 operand. 7903 7904 `NOP_EXPR' 7905 These nodes are used to represent conversions that do not require 7906 any code-generation. For example, conversion of a `char*' to an 7907 `int*' does not require any code be generated; such a conversion is 7908 represented by a `NOP_EXPR'. The single operand is the expression 7909 to be converted. The conversion from a pointer to a reference is 7910 also represented with a `NOP_EXPR'. 7911 7912 `CONVERT_EXPR' 7913 These nodes are similar to `NOP_EXPR's, but are used in those 7914 situations where code may need to be generated. For example, if an 7915 `int*' is converted to an `int' code may need to be generated on 7916 some platforms. These nodes are never used for C++-specific 7917 conversions, like conversions between pointers to different 7918 classes in an inheritance hierarchy. Any adjustments that need to 7919 be made in such cases are always indicated explicitly. Similarly, 7920 a user-defined conversion is never represented by a 7921 `CONVERT_EXPR'; instead, the function calls are made explicit. 7922 7923 `FIXED_CONVERT_EXPR' 7924 These nodes are used to represent conversions that involve 7925 fixed-point values. For example, from a fixed-point value to 7926 another fixed-point value, from an integer to a fixed-point value, 7927 from a fixed-point value to an integer, from a floating-point 7928 value to a fixed-point value, or from a fixed-point value to a 7929 floating-point value. 7930 7931 `THROW_EXPR' 7932 These nodes represent `throw' expressions. The single operand is 7933 an expression for the code that should be executed to throw the 7934 exception. However, there is one implicit action not represented 7935 in that expression; namely the call to `__throw'. This function 7936 takes no arguments. If `setjmp'/`longjmp' exceptions are used, the 7937 function `__sjthrow' is called instead. The normal GCC back end 7938 uses the function `emit_throw' to generate this code; you can 7939 examine this function to see what needs to be done. 7940 7941 `LSHIFT_EXPR' 7942 `RSHIFT_EXPR' 7943 These nodes represent left and right shifts, respectively. The 7944 first operand is the value to shift; it will always be of integral 7945 type. The second operand is an expression for the number of bits 7946 by which to shift. Right shift should be treated as arithmetic, 7947 i.e., the high-order bits should be zero-filled when the 7948 expression has unsigned type and filled with the sign bit when the 7949 expression has signed type. Note that the result is undefined if 7950 the second operand is larger than or equal to the first operand's 7951 type size. 7952 7953 `BIT_IOR_EXPR' 7954 `BIT_XOR_EXPR' 7955 `BIT_AND_EXPR' 7956 These nodes represent bitwise inclusive or, bitwise exclusive or, 7957 and bitwise and, respectively. Both operands will always have 7958 integral type. 7959 7960 `TRUTH_ANDIF_EXPR' 7961 `TRUTH_ORIF_EXPR' 7962 These nodes represent logical "and" and logical "or", respectively. 7963 These operators are not strict; i.e., the second operand is 7964 evaluated only if the value of the expression is not determined by 7965 evaluation of the first operand. The type of the operands and 7966 that of the result are always of `BOOLEAN_TYPE' or `INTEGER_TYPE'. 7967 7968 `TRUTH_AND_EXPR' 7969 `TRUTH_OR_EXPR' 7970 `TRUTH_XOR_EXPR' 7971 These nodes represent logical and, logical or, and logical 7972 exclusive or. They are strict; both arguments are always 7973 evaluated. There are no corresponding operators in C or C++, but 7974 the front end will sometimes generate these expressions anyhow, if 7975 it can tell that strictness does not matter. The type of the 7976 operands and that of the result are always of `BOOLEAN_TYPE' or 7977 `INTEGER_TYPE'. 7978 7979 `POINTER_PLUS_EXPR' 7980 This node represents pointer arithmetic. The first operand is 7981 always a pointer/reference type. The second operand is always an 7982 unsigned integer type compatible with sizetype. This is the only 7983 binary arithmetic operand that can operate on pointer types. 7984 7985 `PLUS_EXPR' 7986 `MINUS_EXPR' 7987 `MULT_EXPR' 7988 These nodes represent various binary arithmetic operations. 7989 Respectively, these operations are addition, subtraction (of the 7990 second operand from the first) and multiplication. Their operands 7991 may have either integral or floating type, but there will never be 7992 case in which one operand is of floating type and the other is of 7993 integral type. 7994 7995 The behavior of these operations on signed arithmetic overflow is 7996 controlled by the `flag_wrapv' and `flag_trapv' variables. 7997 7998 `RDIV_EXPR' 7999 This node represents a floating point division operation. 8000 8001 `TRUNC_DIV_EXPR' 8002 `FLOOR_DIV_EXPR' 8003 `CEIL_DIV_EXPR' 8004 `ROUND_DIV_EXPR' 8005 These nodes represent integer division operations that return an 8006 integer result. `TRUNC_DIV_EXPR' rounds towards zero, 8007 `FLOOR_DIV_EXPR' rounds towards negative infinity, `CEIL_DIV_EXPR' 8008 rounds towards positive infinity and `ROUND_DIV_EXPR' rounds to 8009 the closest integer. Integer division in C and C++ is truncating, 8010 i.e. `TRUNC_DIV_EXPR'. 8011 8012 The behavior of these operations on signed arithmetic overflow, 8013 when dividing the minimum signed integer by minus one, is 8014 controlled by the `flag_wrapv' and `flag_trapv' variables. 8015 8016 `TRUNC_MOD_EXPR' 8017 `FLOOR_MOD_EXPR' 8018 `CEIL_MOD_EXPR' 8019 `ROUND_MOD_EXPR' 8020 These nodes represent the integer remainder or modulus operation. 8021 The integer modulus of two operands `a' and `b' is defined as `a - 8022 (a/b)*b' where the division calculated using the corresponding 8023 division operator. Hence for `TRUNC_MOD_EXPR' this definition 8024 assumes division using truncation towards zero, i.e. 8025 `TRUNC_DIV_EXPR'. Integer remainder in C and C++ uses truncating 8026 division, i.e. `TRUNC_MOD_EXPR'. 8027 8028 `EXACT_DIV_EXPR' 8029 The `EXACT_DIV_EXPR' code is used to represent integer divisions 8030 where the numerator is known to be an exact multiple of the 8031 denominator. This allows the backend to choose between the faster 8032 of `TRUNC_DIV_EXPR', `CEIL_DIV_EXPR' and `FLOOR_DIV_EXPR' for the 8033 current target. 8034 8035 `ARRAY_REF' 8036 These nodes represent array accesses. The first operand is the 8037 array; the second is the index. To calculate the address of the 8038 memory accessed, you must scale the index by the size of the type 8039 of the array elements. The type of these expressions must be the 8040 type of a component of the array. The third and fourth operands 8041 are used after gimplification to represent the lower bound and 8042 component size but should not be used directly; call 8043 `array_ref_low_bound' and `array_ref_element_size' instead. 8044 8045 `ARRAY_RANGE_REF' 8046 These nodes represent access to a range (or "slice") of an array. 8047 The operands are the same as that for `ARRAY_REF' and have the same 8048 meanings. The type of these expressions must be an array whose 8049 component type is the same as that of the first operand. The 8050 range of that array type determines the amount of data these 8051 expressions access. 8052 8053 `TARGET_MEM_REF' 8054 These nodes represent memory accesses whose address directly map to 8055 an addressing mode of the target architecture. The first argument 8056 is `TMR_SYMBOL' and must be a `VAR_DECL' of an object with a fixed 8057 address. The second argument is `TMR_BASE' and the third one is 8058 `TMR_INDEX'. The fourth argument is `TMR_STEP' and must be an 8059 `INTEGER_CST'. The fifth argument is `TMR_OFFSET' and must be an 8060 `INTEGER_CST'. Any of the arguments may be NULL if the 8061 appropriate component does not appear in the address. Address of 8062 the `TARGET_MEM_REF' is determined in the following way. 8063 8064 &TMR_SYMBOL + TMR_BASE + TMR_INDEX * TMR_STEP + TMR_OFFSET 8065 8066 The sixth argument is the reference to the original memory access, 8067 which is preserved for the purposes of the RTL alias analysis. 8068 The seventh argument is a tag representing the results of tree 8069 level alias analysis. 8070 8071 `LT_EXPR' 8072 `LE_EXPR' 8073 `GT_EXPR' 8074 `GE_EXPR' 8075 `EQ_EXPR' 8076 `NE_EXPR' 8077 These nodes represent the less than, less than or equal to, greater 8078 than, greater than or equal to, equal, and not equal comparison 8079 operators. The first and second operand with either be both of 8080 integral type or both of floating type. The result type of these 8081 expressions will always be of integral or boolean type. These 8082 operations return the result type's zero value for false, and the 8083 result type's one value for true. 8084 8085 For floating point comparisons, if we honor IEEE NaNs and either 8086 operand is NaN, then `NE_EXPR' always returns true and the 8087 remaining operators always return false. On some targets, 8088 comparisons against an IEEE NaN, other than equality and 8089 inequality, may generate a floating point exception. 8090 8091 `ORDERED_EXPR' 8092 `UNORDERED_EXPR' 8093 These nodes represent non-trapping ordered and unordered comparison 8094 operators. These operations take two floating point operands and 8095 determine whether they are ordered or unordered relative to each 8096 other. If either operand is an IEEE NaN, their comparison is 8097 defined to be unordered, otherwise the comparison is defined to be 8098 ordered. The result type of these expressions will always be of 8099 integral or boolean type. These operations return the result 8100 type's zero value for false, and the result type's one value for 8101 true. 8102 8103 `UNLT_EXPR' 8104 `UNLE_EXPR' 8105 `UNGT_EXPR' 8106 `UNGE_EXPR' 8107 `UNEQ_EXPR' 8108 `LTGT_EXPR' 8109 These nodes represent the unordered comparison operators. These 8110 operations take two floating point operands and determine whether 8111 the operands are unordered or are less than, less than or equal to, 8112 greater than, greater than or equal to, or equal respectively. For 8113 example, `UNLT_EXPR' returns true if either operand is an IEEE NaN 8114 or the first operand is less than the second. With the possible 8115 exception of `LTGT_EXPR', all of these operations are guaranteed 8116 not to generate a floating point exception. The result type of 8117 these expressions will always be of integral or boolean type. 8118 These operations return the result type's zero value for false, 8119 and the result type's one value for true. 8120 8121 `MODIFY_EXPR' 8122 These nodes represent assignment. The left-hand side is the first 8123 operand; the right-hand side is the second operand. The left-hand 8124 side will be a `VAR_DECL', `INDIRECT_REF', `COMPONENT_REF', or 8125 other lvalue. 8126 8127 These nodes are used to represent not only assignment with `=' but 8128 also compound assignments (like `+='), by reduction to `=' 8129 assignment. In other words, the representation for `i += 3' looks 8130 just like that for `i = i + 3'. 8131 8132 `INIT_EXPR' 8133 These nodes are just like `MODIFY_EXPR', but are used only when a 8134 variable is initialized, rather than assigned to subsequently. 8135 This means that we can assume that the target of the 8136 initialization is not used in computing its own value; any 8137 reference to the lhs in computing the rhs is undefined. 8138 8139 `COMPONENT_REF' 8140 These nodes represent non-static data member accesses. The first 8141 operand is the object (rather than a pointer to it); the second 8142 operand is the `FIELD_DECL' for the data member. The third 8143 operand represents the byte offset of the field, but should not be 8144 used directly; call `component_ref_field_offset' instead. 8145 8146 `COMPOUND_EXPR' 8147 These nodes represent comma-expressions. The first operand is an 8148 expression whose value is computed and thrown away prior to the 8149 evaluation of the second operand. The value of the entire 8150 expression is the value of the second operand. 8151 8152 `COND_EXPR' 8153 These nodes represent `?:' expressions. The first operand is of 8154 boolean or integral type. If it evaluates to a nonzero value, the 8155 second operand should be evaluated, and returned as the value of 8156 the expression. Otherwise, the third operand is evaluated, and 8157 returned as the value of the expression. 8158 8159 The second operand must have the same type as the entire 8160 expression, unless it unconditionally throws an exception or calls 8161 a noreturn function, in which case it should have void type. The 8162 same constraints apply to the third operand. This allows array 8163 bounds checks to be represented conveniently as `(i >= 0 && i < 8164 10) ? i : abort()'. 8165 8166 As a GNU extension, the C language front-ends allow the second 8167 operand of the `?:' operator may be omitted in the source. For 8168 example, `x ? : 3' is equivalent to `x ? x : 3', assuming that `x' 8169 is an expression without side-effects. In the tree 8170 representation, however, the second operand is always present, 8171 possibly protected by `SAVE_EXPR' if the first argument does cause 8172 side-effects. 8173 8174 `CALL_EXPR' 8175 These nodes are used to represent calls to functions, including 8176 non-static member functions. `CALL_EXPR's are implemented as 8177 expression nodes with a variable number of operands. Rather than 8178 using `TREE_OPERAND' to extract them, it is preferable to use the 8179 specialized accessor macros and functions that operate 8180 specifically on `CALL_EXPR' nodes. 8181 8182 `CALL_EXPR_FN' returns a pointer to the function to call; it is 8183 always an expression whose type is a `POINTER_TYPE'. 8184 8185 The number of arguments to the call is returned by 8186 `call_expr_nargs', while the arguments themselves can be accessed 8187 with the `CALL_EXPR_ARG' macro. The arguments are zero-indexed 8188 and numbered left-to-right. You can iterate over the arguments 8189 using `FOR_EACH_CALL_EXPR_ARG', as in: 8190 8191 tree call, arg; 8192 call_expr_arg_iterator iter; 8193 FOR_EACH_CALL_EXPR_ARG (arg, iter, call) 8194 /* arg is bound to successive arguments of call. */ 8195 ...; 8196 8197 For non-static member functions, there will be an operand 8198 corresponding to the `this' pointer. There will always be 8199 expressions corresponding to all of the arguments, even if the 8200 function is declared with default arguments and some arguments are 8201 not explicitly provided at the call sites. 8202 8203 `CALL_EXPR's also have a `CALL_EXPR_STATIC_CHAIN' operand that is 8204 used to implement nested functions. This operand is otherwise 8205 null. 8206 8207 `STMT_EXPR' 8208 These nodes are used to represent GCC's statement-expression 8209 extension. The statement-expression extension allows code like 8210 this: 8211 int f() { return ({ int j; j = 3; j + 7; }); } 8212 In other words, an sequence of statements may occur where a single 8213 expression would normally appear. The `STMT_EXPR' node represents 8214 such an expression. The `STMT_EXPR_STMT' gives the statement 8215 contained in the expression. The value of the expression is the 8216 value of the last sub-statement in the body. More precisely, the 8217 value is the value computed by the last statement nested inside 8218 `BIND_EXPR', `TRY_FINALLY_EXPR', or `TRY_CATCH_EXPR'. For 8219 example, in: 8220 ({ 3; }) 8221 the value is `3' while in: 8222 ({ if (x) { 3; } }) 8223 there is no value. If the `STMT_EXPR' does not yield a value, 8224 it's type will be `void'. 8225 8226 `BIND_EXPR' 8227 These nodes represent local blocks. The first operand is a list of 8228 variables, connected via their `TREE_CHAIN' field. These will 8229 never require cleanups. The scope of these variables is just the 8230 body of the `BIND_EXPR'. The body of the `BIND_EXPR' is the 8231 second operand. 8232 8233 `LOOP_EXPR' 8234 These nodes represent "infinite" loops. The `LOOP_EXPR_BODY' 8235 represents the body of the loop. It should be executed forever, 8236 unless an `EXIT_EXPR' is encountered. 8237 8238 `EXIT_EXPR' 8239 These nodes represent conditional exits from the nearest enclosing 8240 `LOOP_EXPR'. The single operand is the condition; if it is 8241 nonzero, then the loop should be exited. An `EXIT_EXPR' will only 8242 appear within a `LOOP_EXPR'. 8243 8244 `CLEANUP_POINT_EXPR' 8245 These nodes represent full-expressions. The single operand is an 8246 expression to evaluate. Any destructor calls engendered by the 8247 creation of temporaries during the evaluation of that expression 8248 should be performed immediately after the expression is evaluated. 8249 8250 `CONSTRUCTOR' 8251 These nodes represent the brace-enclosed initializers for a 8252 structure or array. The first operand is reserved for use by the 8253 back end. The second operand is a `TREE_LIST'. If the 8254 `TREE_TYPE' of the `CONSTRUCTOR' is a `RECORD_TYPE' or 8255 `UNION_TYPE', then the `TREE_PURPOSE' of each node in the 8256 `TREE_LIST' will be a `FIELD_DECL' and the `TREE_VALUE' of each 8257 node will be the expression used to initialize that field. 8258 8259 If the `TREE_TYPE' of the `CONSTRUCTOR' is an `ARRAY_TYPE', then 8260 the `TREE_PURPOSE' of each element in the `TREE_LIST' will be an 8261 `INTEGER_CST' or a `RANGE_EXPR' of two `INTEGER_CST's. A single 8262 `INTEGER_CST' indicates which element of the array (indexed from 8263 zero) is being assigned to. A `RANGE_EXPR' indicates an inclusive 8264 range of elements to initialize. In both cases the `TREE_VALUE' 8265 is the corresponding initializer. It is re-evaluated for each 8266 element of a `RANGE_EXPR'. If the `TREE_PURPOSE' is `NULL_TREE', 8267 then the initializer is for the next available array element. 8268 8269 In the front end, you should not depend on the fields appearing in 8270 any particular order. However, in the middle end, fields must 8271 appear in declaration order. You should not assume that all 8272 fields will be represented. Unrepresented fields will be set to 8273 zero. 8274 8275 `COMPOUND_LITERAL_EXPR' 8276 These nodes represent ISO C99 compound literals. The 8277 `COMPOUND_LITERAL_EXPR_DECL_STMT' is a `DECL_STMT' containing an 8278 anonymous `VAR_DECL' for the unnamed object represented by the 8279 compound literal; the `DECL_INITIAL' of that `VAR_DECL' is a 8280 `CONSTRUCTOR' representing the brace-enclosed list of initializers 8281 in the compound literal. That anonymous `VAR_DECL' can also be 8282 accessed directly by the `COMPOUND_LITERAL_EXPR_DECL' macro. 8283 8284 `SAVE_EXPR' 8285 A `SAVE_EXPR' represents an expression (possibly involving 8286 side-effects) that is used more than once. The side-effects should 8287 occur only the first time the expression is evaluated. Subsequent 8288 uses should just reuse the computed value. The first operand to 8289 the `SAVE_EXPR' is the expression to evaluate. The side-effects 8290 should be executed where the `SAVE_EXPR' is first encountered in a 8291 depth-first preorder traversal of the expression tree. 8292 8293 `TARGET_EXPR' 8294 A `TARGET_EXPR' represents a temporary object. The first operand 8295 is a `VAR_DECL' for the temporary variable. The second operand is 8296 the initializer for the temporary. The initializer is evaluated 8297 and, if non-void, copied (bitwise) into the temporary. If the 8298 initializer is void, that means that it will perform the 8299 initialization itself. 8300 8301 Often, a `TARGET_EXPR' occurs on the right-hand side of an 8302 assignment, or as the second operand to a comma-expression which is 8303 itself the right-hand side of an assignment, etc. In this case, 8304 we say that the `TARGET_EXPR' is "normal"; otherwise, we say it is 8305 "orphaned". For a normal `TARGET_EXPR' the temporary variable 8306 should be treated as an alias for the left-hand side of the 8307 assignment, rather than as a new temporary variable. 8308 8309 The third operand to the `TARGET_EXPR', if present, is a 8310 cleanup-expression (i.e., destructor call) for the temporary. If 8311 this expression is orphaned, then this expression must be executed 8312 when the statement containing this expression is complete. These 8313 cleanups must always be executed in the order opposite to that in 8314 which they were encountered. Note that if a temporary is created 8315 on one branch of a conditional operator (i.e., in the second or 8316 third operand to a `COND_EXPR'), the cleanup must be run only if 8317 that branch is actually executed. 8318 8319 See `STMT_IS_FULL_EXPR_P' for more information about running these 8320 cleanups. 8321 8322 `AGGR_INIT_EXPR' 8323 An `AGGR_INIT_EXPR' represents the initialization as the return 8324 value of a function call, or as the result of a constructor. An 8325 `AGGR_INIT_EXPR' will only appear as a full-expression, or as the 8326 second operand of a `TARGET_EXPR'. `AGGR_INIT_EXPR's have a 8327 representation similar to that of `CALL_EXPR's. You can use the 8328 `AGGR_INIT_EXPR_FN' and `AGGR_INIT_EXPR_ARG' macros to access the 8329 function to call and the arguments to pass. 8330 8331 If `AGGR_INIT_VIA_CTOR_P' holds of the `AGGR_INIT_EXPR', then the 8332 initialization is via a constructor call. The address of the 8333 `AGGR_INIT_EXPR_SLOT' operand, which is always a `VAR_DECL', is 8334 taken, and this value replaces the first argument in the argument 8335 list. 8336 8337 In either case, the expression is void. 8338 8339 `VA_ARG_EXPR' 8340 This node is used to implement support for the C/C++ variable 8341 argument-list mechanism. It represents expressions like `va_arg 8342 (ap, type)'. Its `TREE_TYPE' yields the tree representation for 8343 `type' and its sole argument yields the representation for `ap'. 8344 8345 `CHANGE_DYNAMIC_TYPE_EXPR' 8346 Indicates the special aliasing required by C++ placement new. It 8347 has two operands: a type and a location. It means that the 8348 dynamic type of the location is changing to be the specified type. 8349 The alias analysis code takes this into account when doing type 8350 based alias analysis. 8351 8352 `OMP_PARALLEL' 8353 Represents `#pragma omp parallel [clause1 ... clauseN]'. It has 8354 four operands: 8355 8356 Operand `OMP_PARALLEL_BODY' is valid while in GENERIC and High 8357 GIMPLE forms. It contains the body of code to be executed by all 8358 the threads. During GIMPLE lowering, this operand becomes `NULL' 8359 and the body is emitted linearly after `OMP_PARALLEL'. 8360 8361 Operand `OMP_PARALLEL_CLAUSES' is the list of clauses associated 8362 with the directive. 8363 8364 Operand `OMP_PARALLEL_FN' is created by `pass_lower_omp', it 8365 contains the `FUNCTION_DECL' for the function that will contain 8366 the body of the parallel region. 8367 8368 Operand `OMP_PARALLEL_DATA_ARG' is also created by 8369 `pass_lower_omp'. If there are shared variables to be communicated 8370 to the children threads, this operand will contain the `VAR_DECL' 8371 that contains all the shared values and variables. 8372 8373 `OMP_FOR' 8374 Represents `#pragma omp for [clause1 ... clauseN]'. It has 5 8375 operands: 8376 8377 Operand `OMP_FOR_BODY' contains the loop body. 8378 8379 Operand `OMP_FOR_CLAUSES' is the list of clauses associated with 8380 the directive. 8381 8382 Operand `OMP_FOR_INIT' is the loop initialization code of the form 8383 `VAR = N1'. 8384 8385 Operand `OMP_FOR_COND' is the loop conditional expression of the 8386 form `VAR {<,>,<=,>=} N2'. 8387 8388 Operand `OMP_FOR_INCR' is the loop index increment of the form 8389 `VAR {+=,-=} INCR'. 8390 8391 Operand `OMP_FOR_PRE_BODY' contains side-effect code from operands 8392 `OMP_FOR_INIT', `OMP_FOR_COND' and `OMP_FOR_INC'. These 8393 side-effects are part of the `OMP_FOR' block but must be evaluated 8394 before the start of loop body. 8395 8396 The loop index variable `VAR' must be a signed integer variable, 8397 which is implicitly private to each thread. Bounds `N1' and `N2' 8398 and the increment expression `INCR' are required to be loop 8399 invariant integer expressions that are evaluated without any 8400 synchronization. The evaluation order, frequency of evaluation and 8401 side-effects are unspecified by the standard. 8402 8403 `OMP_SECTIONS' 8404 Represents `#pragma omp sections [clause1 ... clauseN]'. 8405 8406 Operand `OMP_SECTIONS_BODY' contains the sections body, which in 8407 turn contains a set of `OMP_SECTION' nodes for each of the 8408 concurrent sections delimited by `#pragma omp section'. 8409 8410 Operand `OMP_SECTIONS_CLAUSES' is the list of clauses associated 8411 with the directive. 8412 8413 `OMP_SECTION' 8414 Section delimiter for `OMP_SECTIONS'. 8415 8416 `OMP_SINGLE' 8417 Represents `#pragma omp single'. 8418 8419 Operand `OMP_SINGLE_BODY' contains the body of code to be executed 8420 by a single thread. 8421 8422 Operand `OMP_SINGLE_CLAUSES' is the list of clauses associated 8423 with the directive. 8424 8425 `OMP_MASTER' 8426 Represents `#pragma omp master'. 8427 8428 Operand `OMP_MASTER_BODY' contains the body of code to be executed 8429 by the master thread. 8430 8431 `OMP_ORDERED' 8432 Represents `#pragma omp ordered'. 8433 8434 Operand `OMP_ORDERED_BODY' contains the body of code to be 8435 executed in the sequential order dictated by the loop index 8436 variable. 8437 8438 `OMP_CRITICAL' 8439 Represents `#pragma omp critical [name]'. 8440 8441 Operand `OMP_CRITICAL_BODY' is the critical section. 8442 8443 Operand `OMP_CRITICAL_NAME' is an optional identifier to label the 8444 critical section. 8445 8446 `OMP_RETURN' 8447 This does not represent any OpenMP directive, it is an artificial 8448 marker to indicate the end of the body of an OpenMP. It is used by 8449 the flow graph (`tree-cfg.c') and OpenMP region building code 8450 (`omp-low.c'). 8451 8452 `OMP_CONTINUE' 8453 Similarly, this instruction does not represent an OpenMP 8454 directive, it is used by `OMP_FOR' and `OMP_SECTIONS' to mark the 8455 place where the code needs to loop to the next iteration (in the 8456 case of `OMP_FOR') or the next section (in the case of 8457 `OMP_SECTIONS'). 8458 8459 In some cases, `OMP_CONTINUE' is placed right before `OMP_RETURN'. 8460 But if there are cleanups that need to occur right after the 8461 looping body, it will be emitted between `OMP_CONTINUE' and 8462 `OMP_RETURN'. 8463 8464 `OMP_ATOMIC' 8465 Represents `#pragma omp atomic'. 8466 8467 Operand 0 is the address at which the atomic operation is to be 8468 performed. 8469 8470 Operand 1 is the expression to evaluate. The gimplifier tries 8471 three alternative code generation strategies. Whenever possible, 8472 an atomic update built-in is used. If that fails, a 8473 compare-and-swap loop is attempted. If that also fails, a regular 8474 critical section around the expression is used. 8475 8476 `OMP_CLAUSE' 8477 Represents clauses associated with one of the `OMP_' directives. 8478 Clauses are represented by separate sub-codes defined in `tree.h'. 8479 Clauses codes can be one of: `OMP_CLAUSE_PRIVATE', 8480 `OMP_CLAUSE_SHARED', `OMP_CLAUSE_FIRSTPRIVATE', 8481 `OMP_CLAUSE_LASTPRIVATE', `OMP_CLAUSE_COPYIN', 8482 `OMP_CLAUSE_COPYPRIVATE', `OMP_CLAUSE_IF', 8483 `OMP_CLAUSE_NUM_THREADS', `OMP_CLAUSE_SCHEDULE', 8484 `OMP_CLAUSE_NOWAIT', `OMP_CLAUSE_ORDERED', `OMP_CLAUSE_DEFAULT', 8485 and `OMP_CLAUSE_REDUCTION'. Each code represents the 8486 corresponding OpenMP clause. 8487 8488 Clauses associated with the same directive are chained together 8489 via `OMP_CLAUSE_CHAIN'. Those clauses that accept a list of 8490 variables are restricted to exactly one, accessed with 8491 `OMP_CLAUSE_VAR'. Therefore, multiple variables under the same 8492 clause `C' need to be represented as multiple `C' clauses chained 8493 together. This facilitates adding new clauses during compilation. 8494 8495 `VEC_LSHIFT_EXPR' 8496 8497 `VEC_RSHIFT_EXPR' 8498 These nodes represent whole vector left and right shifts, 8499 respectively. The first operand is the vector to shift; it will 8500 always be of vector type. The second operand is an expression for 8501 the number of bits by which to shift. Note that the result is 8502 undefined if the second operand is larger than or equal to the 8503 first operand's type size. 8504 8505 `VEC_WIDEN_MULT_HI_EXPR' 8506 8507 `VEC_WIDEN_MULT_LO_EXPR' 8508 These nodes represent widening vector multiplication of the high 8509 and low parts of the two input vectors, respectively. Their 8510 operands are vectors that contain the same number of elements 8511 (`N') of the same integral type. The result is a vector that 8512 contains half as many elements, of an integral type whose size is 8513 twice as wide. In the case of `VEC_WIDEN_MULT_HI_EXPR' the high 8514 `N/2' elements of the two vector are multiplied to produce the 8515 vector of `N/2' products. In the case of `VEC_WIDEN_MULT_LO_EXPR' 8516 the low `N/2' elements of the two vector are multiplied to produce 8517 the vector of `N/2' products. 8518 8519 `VEC_UNPACK_HI_EXPR' 8520 8521 `VEC_UNPACK_LO_EXPR' 8522 These nodes represent unpacking of the high and low parts of the 8523 input vector, respectively. The single operand is a vector that 8524 contains `N' elements of the same integral or floating point type. 8525 The result is a vector that contains half as many elements, of an 8526 integral or floating point type whose size is twice as wide. In 8527 the case of `VEC_UNPACK_HI_EXPR' the high `N/2' elements of the 8528 vector are extracted and widened (promoted). In the case of 8529 `VEC_UNPACK_LO_EXPR' the low `N/2' elements of the vector are 8530 extracted and widened (promoted). 8531 8532 `VEC_UNPACK_FLOAT_HI_EXPR' 8533 8534 `VEC_UNPACK_FLOAT_LO_EXPR' 8535 These nodes represent unpacking of the high and low parts of the 8536 input vector, where the values are converted from fixed point to 8537 floating point. The single operand is a vector that contains `N' 8538 elements of the same integral type. The result is a vector that 8539 contains half as many elements of a floating point type whose size 8540 is twice as wide. In the case of `VEC_UNPACK_HI_EXPR' the high 8541 `N/2' elements of the vector are extracted, converted and widened. 8542 In the case of `VEC_UNPACK_LO_EXPR' the low `N/2' elements of the 8543 vector are extracted, converted and widened. 8544 8545 `VEC_PACK_TRUNC_EXPR' 8546 This node represents packing of truncated elements of the two 8547 input vectors into the output vector. Input operands are vectors 8548 that contain the same number of elements of the same integral or 8549 floating point type. The result is a vector that contains twice 8550 as many elements of an integral or floating point type whose size 8551 is half as wide. The elements of the two vectors are demoted and 8552 merged (concatenated) to form the output vector. 8553 8554 `VEC_PACK_SAT_EXPR' 8555 This node represents packing of elements of the two input vectors 8556 into the output vector using saturation. Input operands are 8557 vectors that contain the same number of elements of the same 8558 integral type. The result is a vector that contains twice as many 8559 elements of an integral type whose size is half as wide. The 8560 elements of the two vectors are demoted and merged (concatenated) 8561 to form the output vector. 8562 8563 `VEC_PACK_FIX_TRUNC_EXPR' 8564 This node represents packing of elements of the two input vectors 8565 into the output vector, where the values are converted from 8566 floating point to fixed point. Input operands are vectors that 8567 contain the same number of elements of a floating point type. The 8568 result is a vector that contains twice as many elements of an 8569 integral type whose size is half as wide. The elements of the two 8570 vectors are merged (concatenated) to form the output vector. 8571 8572 `VEC_EXTRACT_EVEN_EXPR' 8573 8574 `VEC_EXTRACT_ODD_EXPR' 8575 These nodes represent extracting of the even/odd elements of the 8576 two input vectors, respectively. Their operands and result are 8577 vectors that contain the same number of elements of the same type. 8578 8579 `VEC_INTERLEAVE_HIGH_EXPR' 8580 8581 `VEC_INTERLEAVE_LOW_EXPR' 8582 These nodes represent merging and interleaving of the high/low 8583 elements of the two input vectors, respectively. The operands and 8584 the result are vectors that contain the same number of elements 8585 (`N') of the same type. In the case of 8586 `VEC_INTERLEAVE_HIGH_EXPR', the high `N/2' elements of the first 8587 input vector are interleaved with the high `N/2' elements of the 8588 second input vector. In the case of `VEC_INTERLEAVE_LOW_EXPR', the 8589 low `N/2' elements of the first input vector are interleaved with 8590 the low `N/2' elements of the second input vector. 8591 8592 8593 8594 File: gccint.info, Node: RTL, Next: Control Flow, Prev: Tree SSA, Up: Top 8595 8596 10 RTL Representation 8597 ********************* 8598 8599 The last part of the compiler work is done on a low-level intermediate 8600 representation called Register Transfer Language. In this language, the 8601 instructions to be output are described, pretty much one by one, in an 8602 algebraic form that describes what the instruction does. 8603 8604 RTL is inspired by Lisp lists. It has both an internal form, made up 8605 of structures that point at other structures, and a textual form that 8606 is used in the machine description and in printed debugging dumps. The 8607 textual form uses nested parentheses to indicate the pointers in the 8608 internal form. 8609 8610 * Menu: 8611 8612 * RTL Objects:: Expressions vs vectors vs strings vs integers. 8613 * RTL Classes:: Categories of RTL expression objects, and their structure. 8614 * Accessors:: Macros to access expression operands or vector elts. 8615 * Special Accessors:: Macros to access specific annotations on RTL. 8616 * Flags:: Other flags in an RTL expression. 8617 * Machine Modes:: Describing the size and format of a datum. 8618 * Constants:: Expressions with constant values. 8619 * Regs and Memory:: Expressions representing register contents or memory. 8620 * Arithmetic:: Expressions representing arithmetic on other expressions. 8621 * Comparisons:: Expressions representing comparison of expressions. 8622 * Bit-Fields:: Expressions representing bit-fields in memory or reg. 8623 * Vector Operations:: Expressions involving vector datatypes. 8624 * Conversions:: Extending, truncating, floating or fixing. 8625 * RTL Declarations:: Declaring volatility, constancy, etc. 8626 * Side Effects:: Expressions for storing in registers, etc. 8627 * Incdec:: Embedded side-effects for autoincrement addressing. 8628 * Assembler:: Representing `asm' with operands. 8629 * Insns:: Expression types for entire insns. 8630 * Calls:: RTL representation of function call insns. 8631 * Sharing:: Some expressions are unique; others *must* be copied. 8632 * Reading RTL:: Reading textual RTL from a file. 8633 8634 8635 File: gccint.info, Node: RTL Objects, Next: RTL Classes, Up: RTL 8636 8637 10.1 RTL Object Types 8638 ===================== 8639 8640 RTL uses five kinds of objects: expressions, integers, wide integers, 8641 strings and vectors. Expressions are the most important ones. An RTL 8642 expression ("RTX", for short) is a C structure, but it is usually 8643 referred to with a pointer; a type that is given the typedef name `rtx'. 8644 8645 An integer is simply an `int'; their written form uses decimal digits. 8646 A wide integer is an integral object whose type is `HOST_WIDE_INT'; 8647 their written form uses decimal digits. 8648 8649 A string is a sequence of characters. In core it is represented as a 8650 `char *' in usual C fashion, and it is written in C syntax as well. 8651 However, strings in RTL may never be null. If you write an empty 8652 string in a machine description, it is represented in core as a null 8653 pointer rather than as a pointer to a null character. In certain 8654 contexts, these null pointers instead of strings are valid. Within RTL 8655 code, strings are most commonly found inside `symbol_ref' expressions, 8656 but they appear in other contexts in the RTL expressions that make up 8657 machine descriptions. 8658 8659 In a machine description, strings are normally written with double 8660 quotes, as you would in C. However, strings in machine descriptions may 8661 extend over many lines, which is invalid C, and adjacent string 8662 constants are not concatenated as they are in C. Any string constant 8663 may be surrounded with a single set of parentheses. Sometimes this 8664 makes the machine description easier to read. 8665 8666 There is also a special syntax for strings, which can be useful when C 8667 code is embedded in a machine description. Wherever a string can 8668 appear, it is also valid to write a C-style brace block. The entire 8669 brace block, including the outermost pair of braces, is considered to be 8670 the string constant. Double quote characters inside the braces are not 8671 special. Therefore, if you write string constants in the C code, you 8672 need not escape each quote character with a backslash. 8673 8674 A vector contains an arbitrary number of pointers to expressions. The 8675 number of elements in the vector is explicitly present in the vector. 8676 The written form of a vector consists of square brackets (`[...]') 8677 surrounding the elements, in sequence and with whitespace separating 8678 them. Vectors of length zero are not created; null pointers are used 8679 instead. 8680 8681 Expressions are classified by "expression codes" (also called RTX 8682 codes). The expression code is a name defined in `rtl.def', which is 8683 also (in uppercase) a C enumeration constant. The possible expression 8684 codes and their meanings are machine-independent. The code of an RTX 8685 can be extracted with the macro `GET_CODE (X)' and altered with 8686 `PUT_CODE (X, NEWCODE)'. 8687 8688 The expression code determines how many operands the expression 8689 contains, and what kinds of objects they are. In RTL, unlike Lisp, you 8690 cannot tell by looking at an operand what kind of object it is. 8691 Instead, you must know from its context--from the expression code of 8692 the containing expression. For example, in an expression of code 8693 `subreg', the first operand is to be regarded as an expression and the 8694 second operand as an integer. In an expression of code `plus', there 8695 are two operands, both of which are to be regarded as expressions. In 8696 a `symbol_ref' expression, there is one operand, which is to be 8697 regarded as a string. 8698 8699 Expressions are written as parentheses containing the name of the 8700 expression type, its flags and machine mode if any, and then the 8701 operands of the expression (separated by spaces). 8702 8703 Expression code names in the `md' file are written in lowercase, but 8704 when they appear in C code they are written in uppercase. In this 8705 manual, they are shown as follows: `const_int'. 8706 8707 In a few contexts a null pointer is valid where an expression is 8708 normally wanted. The written form of this is `(nil)'. 8709 8710 8711 File: gccint.info, Node: RTL Classes, Next: Accessors, Prev: RTL Objects, Up: RTL 8712 8713 10.2 RTL Classes and Formats 8714 ============================ 8715 8716 The various expression codes are divided into several "classes", which 8717 are represented by single characters. You can determine the class of 8718 an RTX code with the macro `GET_RTX_CLASS (CODE)'. Currently, 8719 `rtl.def' defines these classes: 8720 8721 `RTX_OBJ' 8722 An RTX code that represents an actual object, such as a register 8723 (`REG') or a memory location (`MEM', `SYMBOL_REF'). `LO_SUM') is 8724 also included; instead, `SUBREG' and `STRICT_LOW_PART' are not in 8725 this class, but in class `x'. 8726 8727 `RTX_CONST_OBJ' 8728 An RTX code that represents a constant object. `HIGH' is also 8729 included in this class. 8730 8731 `RTX_COMPARE' 8732 An RTX code for a non-symmetric comparison, such as `GEU' or `LT'. 8733 8734 `RTX_COMM_COMPARE' 8735 An RTX code for a symmetric (commutative) comparison, such as `EQ' 8736 or `ORDERED'. 8737 8738 `RTX_UNARY' 8739 An RTX code for a unary arithmetic operation, such as `NEG', 8740 `NOT', or `ABS'. This category also includes value extension 8741 (sign or zero) and conversions between integer and floating point. 8742 8743 `RTX_COMM_ARITH' 8744 An RTX code for a commutative binary operation, such as `PLUS' or 8745 `AND'. `NE' and `EQ' are comparisons, so they have class `<'. 8746 8747 `RTX_BIN_ARITH' 8748 An RTX code for a non-commutative binary operation, such as 8749 `MINUS', `DIV', or `ASHIFTRT'. 8750 8751 `RTX_BITFIELD_OPS' 8752 An RTX code for a bit-field operation. Currently only 8753 `ZERO_EXTRACT' and `SIGN_EXTRACT'. These have three inputs and 8754 are lvalues (so they can be used for insertion as well). *Note 8755 Bit-Fields::. 8756 8757 `RTX_TERNARY' 8758 An RTX code for other three input operations. Currently only 8759 `IF_THEN_ELSE' and `VEC_MERGE'. 8760 8761 `RTX_INSN' 8762 An RTX code for an entire instruction: `INSN', `JUMP_INSN', and 8763 `CALL_INSN'. *Note Insns::. 8764 8765 `RTX_MATCH' 8766 An RTX code for something that matches in insns, such as 8767 `MATCH_DUP'. These only occur in machine descriptions. 8768 8769 `RTX_AUTOINC' 8770 An RTX code for an auto-increment addressing mode, such as 8771 `POST_INC'. 8772 8773 `RTX_EXTRA' 8774 All other RTX codes. This category includes the remaining codes 8775 used only in machine descriptions (`DEFINE_*', etc.). It also 8776 includes all the codes describing side effects (`SET', `USE', 8777 `CLOBBER', etc.) and the non-insns that may appear on an insn 8778 chain, such as `NOTE', `BARRIER', and `CODE_LABEL'. `SUBREG' is 8779 also part of this class. 8780 8781 For each expression code, `rtl.def' specifies the number of contained 8782 objects and their kinds using a sequence of characters called the 8783 "format" of the expression code. For example, the format of `subreg' 8784 is `ei'. 8785 8786 These are the most commonly used format characters: 8787 8788 `e' 8789 An expression (actually a pointer to an expression). 8790 8791 `i' 8792 An integer. 8793 8794 `w' 8795 A wide integer. 8796 8797 `s' 8798 A string. 8799 8800 `E' 8801 A vector of expressions. 8802 8803 A few other format characters are used occasionally: 8804 8805 `u' 8806 `u' is equivalent to `e' except that it is printed differently in 8807 debugging dumps. It is used for pointers to insns. 8808 8809 `n' 8810 `n' is equivalent to `i' except that it is printed differently in 8811 debugging dumps. It is used for the line number or code number of 8812 a `note' insn. 8813 8814 `S' 8815 `S' indicates a string which is optional. In the RTL objects in 8816 core, `S' is equivalent to `s', but when the object is read, from 8817 an `md' file, the string value of this operand may be omitted. An 8818 omitted string is taken to be the null string. 8819 8820 `V' 8821 `V' indicates a vector which is optional. In the RTL objects in 8822 core, `V' is equivalent to `E', but when the object is read from 8823 an `md' file, the vector value of this operand may be omitted. An 8824 omitted vector is effectively the same as a vector of no elements. 8825 8826 `B' 8827 `B' indicates a pointer to basic block structure. 8828 8829 `0' 8830 `0' means a slot whose contents do not fit any normal category. 8831 `0' slots are not printed at all in dumps, and are often used in 8832 special ways by small parts of the compiler. 8833 8834 There are macros to get the number of operands and the format of an 8835 expression code: 8836 8837 `GET_RTX_LENGTH (CODE)' 8838 Number of operands of an RTX of code CODE. 8839 8840 `GET_RTX_FORMAT (CODE)' 8841 The format of an RTX of code CODE, as a C string. 8842 8843 Some classes of RTX codes always have the same format. For example, it 8844 is safe to assume that all comparison operations have format `ee'. 8845 8846 `1' 8847 All codes of this class have format `e'. 8848 8849 `<' 8850 `c' 8851 `2' 8852 All codes of these classes have format `ee'. 8853 8854 `b' 8855 `3' 8856 All codes of these classes have format `eee'. 8857 8858 `i' 8859 All codes of this class have formats that begin with `iuueiee'. 8860 *Note Insns::. Note that not all RTL objects linked onto an insn 8861 chain are of class `i'. 8862 8863 `o' 8864 `m' 8865 `x' 8866 You can make no assumptions about the format of these codes. 8867 8868 8869 File: gccint.info, Node: Accessors, Next: Special Accessors, Prev: RTL Classes, Up: RTL 8870 8871 10.3 Access to Operands 8872 ======================= 8873 8874 Operands of expressions are accessed using the macros `XEXP', `XINT', 8875 `XWINT' and `XSTR'. Each of these macros takes two arguments: an 8876 expression-pointer (RTX) and an operand number (counting from zero). 8877 Thus, 8878 8879 XEXP (X, 2) 8880 8881 accesses operand 2 of expression X, as an expression. 8882 8883 XINT (X, 2) 8884 8885 accesses the same operand as an integer. `XSTR', used in the same 8886 fashion, would access it as a string. 8887 8888 Any operand can be accessed as an integer, as an expression or as a 8889 string. You must choose the correct method of access for the kind of 8890 value actually stored in the operand. You would do this based on the 8891 expression code of the containing expression. That is also how you 8892 would know how many operands there are. 8893 8894 For example, if X is a `subreg' expression, you know that it has two 8895 operands which can be correctly accessed as `XEXP (X, 0)' and `XINT (X, 8896 1)'. If you did `XINT (X, 0)', you would get the address of the 8897 expression operand but cast as an integer; that might occasionally be 8898 useful, but it would be cleaner to write `(int) XEXP (X, 0)'. `XEXP 8899 (X, 1)' would also compile without error, and would return the second, 8900 integer operand cast as an expression pointer, which would probably 8901 result in a crash when accessed. Nothing stops you from writing `XEXP 8902 (X, 28)' either, but this will access memory past the end of the 8903 expression with unpredictable results. 8904 8905 Access to operands which are vectors is more complicated. You can use 8906 the macro `XVEC' to get the vector-pointer itself, or the macros 8907 `XVECEXP' and `XVECLEN' to access the elements and length of a vector. 8908 8909 `XVEC (EXP, IDX)' 8910 Access the vector-pointer which is operand number IDX in EXP. 8911 8912 `XVECLEN (EXP, IDX)' 8913 Access the length (number of elements) in the vector which is in 8914 operand number IDX in EXP. This value is an `int'. 8915 8916 `XVECEXP (EXP, IDX, ELTNUM)' 8917 Access element number ELTNUM in the vector which is in operand 8918 number IDX in EXP. This value is an RTX. 8919 8920 It is up to you to make sure that ELTNUM is not negative and is 8921 less than `XVECLEN (EXP, IDX)'. 8922 8923 All the macros defined in this section expand into lvalues and 8924 therefore can be used to assign the operands, lengths and vector 8925 elements as well as to access them. 8926 8927 8928 File: gccint.info, Node: Special Accessors, Next: Flags, Prev: Accessors, Up: RTL 8929 8930 10.4 Access to Special Operands 8931 =============================== 8932 8933 Some RTL nodes have special annotations associated with them. 8934 8935 `MEM' 8936 8937 `MEM_ALIAS_SET (X)' 8938 If 0, X is not in any alias set, and may alias anything. 8939 Otherwise, X can only alias `MEM's in a conflicting alias 8940 set. This value is set in a language-dependent manner in the 8941 front-end, and should not be altered in the back-end. In 8942 some front-ends, these numbers may correspond in some way to 8943 types, or other language-level entities, but they need not, 8944 and the back-end makes no such assumptions. These set 8945 numbers are tested with `alias_sets_conflict_p'. 8946 8947 `MEM_EXPR (X)' 8948 If this register is known to hold the value of some user-level 8949 declaration, this is that tree node. It may also be a 8950 `COMPONENT_REF', in which case this is some field reference, 8951 and `TREE_OPERAND (X, 0)' contains the declaration, or 8952 another `COMPONENT_REF', or null if there is no compile-time 8953 object associated with the reference. 8954 8955 `MEM_OFFSET (X)' 8956 The offset from the start of `MEM_EXPR' as a `CONST_INT' rtx. 8957 8958 `MEM_SIZE (X)' 8959 The size in bytes of the memory reference as a `CONST_INT' 8960 rtx. This is mostly relevant for `BLKmode' references as 8961 otherwise the size is implied by the mode. 8962 8963 `MEM_ALIGN (X)' 8964 The known alignment in bits of the memory reference. 8965 8966 `REG' 8967 8968 `ORIGINAL_REGNO (X)' 8969 This field holds the number the register "originally" had; 8970 for a pseudo register turned into a hard reg this will hold 8971 the old pseudo register number. 8972 8973 `REG_EXPR (X)' 8974 If this register is known to hold the value of some user-level 8975 declaration, this is that tree node. 8976 8977 `REG_OFFSET (X)' 8978 If this register is known to hold the value of some user-level 8979 declaration, this is the offset into that logical storage. 8980 8981 `SYMBOL_REF' 8982 8983 `SYMBOL_REF_DECL (X)' 8984 If the `symbol_ref' X was created for a `VAR_DECL' or a 8985 `FUNCTION_DECL', that tree is recorded here. If this value is 8986 null, then X was created by back end code generation routines, 8987 and there is no associated front end symbol table entry. 8988 8989 `SYMBOL_REF_DECL' may also point to a tree of class `'c'', 8990 that is, some sort of constant. In this case, the 8991 `symbol_ref' is an entry in the per-file constant pool; 8992 again, there is no associated front end symbol table entry. 8993 8994 `SYMBOL_REF_CONSTANT (X)' 8995 If `CONSTANT_POOL_ADDRESS_P (X)' is true, this is the constant 8996 pool entry for X. It is null otherwise. 8997 8998 `SYMBOL_REF_DATA (X)' 8999 A field of opaque type used to store `SYMBOL_REF_DECL' or 9000 `SYMBOL_REF_CONSTANT'. 9001 9002 `SYMBOL_REF_FLAGS (X)' 9003 In a `symbol_ref', this is used to communicate various 9004 predicates about the symbol. Some of these are common enough 9005 to be computed by common code, some are specific to the 9006 target. The common bits are: 9007 9008 `SYMBOL_FLAG_FUNCTION' 9009 Set if the symbol refers to a function. 9010 9011 `SYMBOL_FLAG_LOCAL' 9012 Set if the symbol is local to this "module". See 9013 `TARGET_BINDS_LOCAL_P'. 9014 9015 `SYMBOL_FLAG_EXTERNAL' 9016 Set if this symbol is not defined in this translation 9017 unit. Note that this is not the inverse of 9018 `SYMBOL_FLAG_LOCAL'. 9019 9020 `SYMBOL_FLAG_SMALL' 9021 Set if the symbol is located in the small data section. 9022 See `TARGET_IN_SMALL_DATA_P'. 9023 9024 `SYMBOL_REF_TLS_MODEL (X)' 9025 This is a multi-bit field accessor that returns the 9026 `tls_model' to be used for a thread-local storage 9027 symbol. It returns zero for non-thread-local symbols. 9028 9029 `SYMBOL_FLAG_HAS_BLOCK_INFO' 9030 Set if the symbol has `SYMBOL_REF_BLOCK' and 9031 `SYMBOL_REF_BLOCK_OFFSET' fields. 9032 9033 `SYMBOL_FLAG_ANCHOR' 9034 Set if the symbol is used as a section anchor. "Section 9035 anchors" are symbols that have a known position within 9036 an `object_block' and that can be used to access nearby 9037 members of that block. They are used to implement 9038 `-fsection-anchors'. 9039 9040 If this flag is set, then `SYMBOL_FLAG_HAS_BLOCK_INFO' 9041 will be too. 9042 9043 Bits beginning with `SYMBOL_FLAG_MACH_DEP' are available for 9044 the target's use. 9045 9046 `SYMBOL_REF_BLOCK (X)' 9047 If `SYMBOL_REF_HAS_BLOCK_INFO_P (X)', this is the `object_block' 9048 structure to which the symbol belongs, or `NULL' if it has not 9049 been assigned a block. 9050 9051 `SYMBOL_REF_BLOCK_OFFSET (X)' 9052 If `SYMBOL_REF_HAS_BLOCK_INFO_P (X)', this is the offset of X from 9053 the first object in `SYMBOL_REF_BLOCK (X)'. The value is negative 9054 if X has not yet been assigned to a block, or it has not been 9055 given an offset within that block. 9056 9057 9058 File: gccint.info, Node: Flags, Next: Machine Modes, Prev: Special Accessors, Up: RTL 9059 9060 10.5 Flags in an RTL Expression 9061 =============================== 9062 9063 RTL expressions contain several flags (one-bit bit-fields) that are 9064 used in certain types of expression. Most often they are accessed with 9065 the following macros, which expand into lvalues. 9066 9067 `CONSTANT_POOL_ADDRESS_P (X)' 9068 Nonzero in a `symbol_ref' if it refers to part of the current 9069 function's constant pool. For most targets these addresses are in 9070 a `.rodata' section entirely separate from the function, but for 9071 some targets the addresses are close to the beginning of the 9072 function. In either case GCC assumes these addresses can be 9073 addressed directly, perhaps with the help of base registers. 9074 Stored in the `unchanging' field and printed as `/u'. 9075 9076 `RTL_CONST_CALL_P (X)' 9077 In a `call_insn' indicates that the insn represents a call to a 9078 const function. Stored in the `unchanging' field and printed as 9079 `/u'. 9080 9081 `RTL_PURE_CALL_P (X)' 9082 In a `call_insn' indicates that the insn represents a call to a 9083 pure function. Stored in the `return_val' field and printed as 9084 `/i'. 9085 9086 `RTL_CONST_OR_PURE_CALL_P (X)' 9087 In a `call_insn', true if `RTL_CONST_CALL_P' or `RTL_PURE_CALL_P' 9088 is true. 9089 9090 `RTL_LOOPING_CONST_OR_PURE_CALL_P (X)' 9091 In a `call_insn' indicates that the insn represents a possibly 9092 infinite looping call to a const or pure function. Stored in the 9093 `call' field and printed as `/c'. Only true if one of 9094 `RTL_CONST_CALL_P' or `RTL_PURE_CALL_P' is true. 9095 9096 `INSN_ANNULLED_BRANCH_P (X)' 9097 In a `jump_insn', `call_insn', or `insn' indicates that the branch 9098 is an annulling one. See the discussion under `sequence' below. 9099 Stored in the `unchanging' field and printed as `/u'. 9100 9101 `INSN_DELETED_P (X)' 9102 In an `insn', `call_insn', `jump_insn', `code_label', `barrier', 9103 or `note', nonzero if the insn has been deleted. Stored in the 9104 `volatil' field and printed as `/v'. 9105 9106 `INSN_FROM_TARGET_P (X)' 9107 In an `insn' or `jump_insn' or `call_insn' in a delay slot of a 9108 branch, indicates that the insn is from the target of the branch. 9109 If the branch insn has `INSN_ANNULLED_BRANCH_P' set, this insn 9110 will only be executed if the branch is taken. For annulled 9111 branches with `INSN_FROM_TARGET_P' clear, the insn will be 9112 executed only if the branch is not taken. When 9113 `INSN_ANNULLED_BRANCH_P' is not set, this insn will always be 9114 executed. Stored in the `in_struct' field and printed as `/s'. 9115 9116 `LABEL_PRESERVE_P (X)' 9117 In a `code_label' or `note', indicates that the label is 9118 referenced by code or data not visible to the RTL of a given 9119 function. Labels referenced by a non-local goto will have this 9120 bit set. Stored in the `in_struct' field and printed as `/s'. 9121 9122 `LABEL_REF_NONLOCAL_P (X)' 9123 In `label_ref' and `reg_label' expressions, nonzero if this is a 9124 reference to a non-local label. Stored in the `volatil' field and 9125 printed as `/v'. 9126 9127 `MEM_IN_STRUCT_P (X)' 9128 In `mem' expressions, nonzero for reference to an entire structure, 9129 union or array, or to a component of one. Zero for references to a 9130 scalar variable or through a pointer to a scalar. If both this 9131 flag and `MEM_SCALAR_P' are clear, then we don't know whether this 9132 `mem' is in a structure or not. Both flags should never be 9133 simultaneously set. Stored in the `in_struct' field and printed 9134 as `/s'. 9135 9136 `MEM_KEEP_ALIAS_SET_P (X)' 9137 In `mem' expressions, 1 if we should keep the alias set for this 9138 mem unchanged when we access a component. Set to 1, for example, 9139 when we are already in a non-addressable component of an aggregate. 9140 Stored in the `jump' field and printed as `/j'. 9141 9142 `MEM_SCALAR_P (X)' 9143 In `mem' expressions, nonzero for reference to a scalar known not 9144 to be a member of a structure, union, or array. Zero for such 9145 references and for indirections through pointers, even pointers 9146 pointing to scalar types. If both this flag and `MEM_IN_STRUCT_P' 9147 are clear, then we don't know whether this `mem' is in a structure 9148 or not. Both flags should never be simultaneously set. Stored in 9149 the `return_val' field and printed as `/i'. 9150 9151 `MEM_VOLATILE_P (X)' 9152 In `mem', `asm_operands', and `asm_input' expressions, nonzero for 9153 volatile memory references. Stored in the `volatil' field and 9154 printed as `/v'. 9155 9156 `MEM_NOTRAP_P (X)' 9157 In `mem', nonzero for memory references that will not trap. 9158 Stored in the `call' field and printed as `/c'. 9159 9160 `MEM_POINTER (X)' 9161 Nonzero in a `mem' if the memory reference holds a pointer. 9162 Stored in the `frame_related' field and printed as `/f'. 9163 9164 `REG_FUNCTION_VALUE_P (X)' 9165 Nonzero in a `reg' if it is the place in which this function's 9166 value is going to be returned. (This happens only in a hard 9167 register.) Stored in the `return_val' field and printed as `/i'. 9168 9169 `REG_POINTER (X)' 9170 Nonzero in a `reg' if the register holds a pointer. Stored in the 9171 `frame_related' field and printed as `/f'. 9172 9173 `REG_USERVAR_P (X)' 9174 In a `reg', nonzero if it corresponds to a variable present in the 9175 user's source code. Zero for temporaries generated internally by 9176 the compiler. Stored in the `volatil' field and printed as `/v'. 9177 9178 The same hard register may be used also for collecting the values 9179 of functions called by this one, but `REG_FUNCTION_VALUE_P' is zero 9180 in this kind of use. 9181 9182 `RTX_FRAME_RELATED_P (X)' 9183 Nonzero in an `insn', `call_insn', `jump_insn', `barrier', or 9184 `set' which is part of a function prologue and sets the stack 9185 pointer, sets the frame pointer, or saves a register. This flag 9186 should also be set on an instruction that sets up a temporary 9187 register to use in place of the frame pointer. Stored in the 9188 `frame_related' field and printed as `/f'. 9189 9190 In particular, on RISC targets where there are limits on the sizes 9191 of immediate constants, it is sometimes impossible to reach the 9192 register save area directly from the stack pointer. In that case, 9193 a temporary register is used that is near enough to the register 9194 save area, and the Canonical Frame Address, i.e., DWARF2's logical 9195 frame pointer, register must (temporarily) be changed to be this 9196 temporary register. So, the instruction that sets this temporary 9197 register must be marked as `RTX_FRAME_RELATED_P'. 9198 9199 If the marked instruction is overly complex (defined in terms of 9200 what `dwarf2out_frame_debug_expr' can handle), you will also have 9201 to create a `REG_FRAME_RELATED_EXPR' note and attach it to the 9202 instruction. This note should contain a simple expression of the 9203 computation performed by this instruction, i.e., one that 9204 `dwarf2out_frame_debug_expr' can handle. 9205 9206 This flag is required for exception handling support on targets 9207 with RTL prologues. 9208 9209 `MEM_READONLY_P (X)' 9210 Nonzero in a `mem', if the memory is statically allocated and 9211 read-only. 9212 9213 Read-only in this context means never modified during the lifetime 9214 of the program, not necessarily in ROM or in write-disabled pages. 9215 A common example of the later is a shared library's global offset 9216 table. This table is initialized by the runtime loader, so the 9217 memory is technically writable, but after control is transfered 9218 from the runtime loader to the application, this memory will never 9219 be subsequently modified. 9220 9221 Stored in the `unchanging' field and printed as `/u'. 9222 9223 `SCHED_GROUP_P (X)' 9224 During instruction scheduling, in an `insn', `call_insn' or 9225 `jump_insn', indicates that the previous insn must be scheduled 9226 together with this insn. This is used to ensure that certain 9227 groups of instructions will not be split up by the instruction 9228 scheduling pass, for example, `use' insns before a `call_insn' may 9229 not be separated from the `call_insn'. Stored in the `in_struct' 9230 field and printed as `/s'. 9231 9232 `SET_IS_RETURN_P (X)' 9233 For a `set', nonzero if it is for a return. Stored in the `jump' 9234 field and printed as `/j'. 9235 9236 `SIBLING_CALL_P (X)' 9237 For a `call_insn', nonzero if the insn is a sibling call. Stored 9238 in the `jump' field and printed as `/j'. 9239 9240 `STRING_POOL_ADDRESS_P (X)' 9241 For a `symbol_ref' expression, nonzero if it addresses this 9242 function's string constant pool. Stored in the `frame_related' 9243 field and printed as `/f'. 9244 9245 `SUBREG_PROMOTED_UNSIGNED_P (X)' 9246 Returns a value greater then zero for a `subreg' that has 9247 `SUBREG_PROMOTED_VAR_P' nonzero if the object being referenced is 9248 kept zero-extended, zero if it is kept sign-extended, and less 9249 then zero if it is extended some other way via the `ptr_extend' 9250 instruction. Stored in the `unchanging' field and `volatil' 9251 field, printed as `/u' and `/v'. This macro may only be used to 9252 get the value it may not be used to change the value. Use 9253 `SUBREG_PROMOTED_UNSIGNED_SET' to change the value. 9254 9255 `SUBREG_PROMOTED_UNSIGNED_SET (X)' 9256 Set the `unchanging' and `volatil' fields in a `subreg' to reflect 9257 zero, sign, or other extension. If `volatil' is zero, then 9258 `unchanging' as nonzero means zero extension and as zero means 9259 sign extension. If `volatil' is nonzero then some other type of 9260 extension was done via the `ptr_extend' instruction. 9261 9262 `SUBREG_PROMOTED_VAR_P (X)' 9263 Nonzero in a `subreg' if it was made when accessing an object that 9264 was promoted to a wider mode in accord with the `PROMOTED_MODE' 9265 machine description macro (*note Storage Layout::). In this case, 9266 the mode of the `subreg' is the declared mode of the object and 9267 the mode of `SUBREG_REG' is the mode of the register that holds 9268 the object. Promoted variables are always either sign- or 9269 zero-extended to the wider mode on every assignment. Stored in 9270 the `in_struct' field and printed as `/s'. 9271 9272 `SYMBOL_REF_USED (X)' 9273 In a `symbol_ref', indicates that X has been used. This is 9274 normally only used to ensure that X is only declared external 9275 once. Stored in the `used' field. 9276 9277 `SYMBOL_REF_WEAK (X)' 9278 In a `symbol_ref', indicates that X has been declared weak. 9279 Stored in the `return_val' field and printed as `/i'. 9280 9281 `SYMBOL_REF_FLAG (X)' 9282 In a `symbol_ref', this is used as a flag for machine-specific 9283 purposes. Stored in the `volatil' field and printed as `/v'. 9284 9285 Most uses of `SYMBOL_REF_FLAG' are historic and may be subsumed by 9286 `SYMBOL_REF_FLAGS'. Certainly use of `SYMBOL_REF_FLAGS' is 9287 mandatory if the target requires more than one bit of storage. 9288 9289 These are the fields to which the above macros refer: 9290 9291 `call' 9292 In a `mem', 1 means that the memory reference will not trap. 9293 9294 In a `call', 1 means that this pure or const call may possibly 9295 infinite loop. 9296 9297 In an RTL dump, this flag is represented as `/c'. 9298 9299 `frame_related' 9300 In an `insn' or `set' expression, 1 means that it is part of a 9301 function prologue and sets the stack pointer, sets the frame 9302 pointer, saves a register, or sets up a temporary register to use 9303 in place of the frame pointer. 9304 9305 In `reg' expressions, 1 means that the register holds a pointer. 9306 9307 In `mem' expressions, 1 means that the memory reference holds a 9308 pointer. 9309 9310 In `symbol_ref' expressions, 1 means that the reference addresses 9311 this function's string constant pool. 9312 9313 In an RTL dump, this flag is represented as `/f'. 9314 9315 `in_struct' 9316 In `mem' expressions, it is 1 if the memory datum referred to is 9317 all or part of a structure or array; 0 if it is (or might be) a 9318 scalar variable. A reference through a C pointer has 0 because 9319 the pointer might point to a scalar variable. This information 9320 allows the compiler to determine something about possible cases of 9321 aliasing. 9322 9323 In `reg' expressions, it is 1 if the register has its entire life 9324 contained within the test expression of some loop. 9325 9326 In `subreg' expressions, 1 means that the `subreg' is accessing an 9327 object that has had its mode promoted from a wider mode. 9328 9329 In `label_ref' expressions, 1 means that the referenced label is 9330 outside the innermost loop containing the insn in which the 9331 `label_ref' was found. 9332 9333 In `code_label' expressions, it is 1 if the label may never be 9334 deleted. This is used for labels which are the target of 9335 non-local gotos. Such a label that would have been deleted is 9336 replaced with a `note' of type `NOTE_INSN_DELETED_LABEL'. 9337 9338 In an `insn' during dead-code elimination, 1 means that the insn is 9339 dead code. 9340 9341 In an `insn' or `jump_insn' during reorg for an insn in the delay 9342 slot of a branch, 1 means that this insn is from the target of the 9343 branch. 9344 9345 In an `insn' during instruction scheduling, 1 means that this insn 9346 must be scheduled as part of a group together with the previous 9347 insn. 9348 9349 In an RTL dump, this flag is represented as `/s'. 9350 9351 `return_val' 9352 In `reg' expressions, 1 means the register contains the value to 9353 be returned by the current function. On machines that pass 9354 parameters in registers, the same register number may be used for 9355 parameters as well, but this flag is not set on such uses. 9356 9357 In `mem' expressions, 1 means the memory reference is to a scalar 9358 known not to be a member of a structure, union, or array. 9359 9360 In `symbol_ref' expressions, 1 means the referenced symbol is weak. 9361 9362 In `call' expressions, 1 means the call is pure. 9363 9364 In an RTL dump, this flag is represented as `/i'. 9365 9366 `jump' 9367 In a `mem' expression, 1 means we should keep the alias set for 9368 this mem unchanged when we access a component. 9369 9370 In a `set', 1 means it is for a return. 9371 9372 In a `call_insn', 1 means it is a sibling call. 9373 9374 In an RTL dump, this flag is represented as `/j'. 9375 9376 `unchanging' 9377 In `reg' and `mem' expressions, 1 means that the value of the 9378 expression never changes. 9379 9380 In `subreg' expressions, it is 1 if the `subreg' references an 9381 unsigned object whose mode has been promoted to a wider mode. 9382 9383 In an `insn' or `jump_insn' in the delay slot of a branch 9384 instruction, 1 means an annulling branch should be used. 9385 9386 In a `symbol_ref' expression, 1 means that this symbol addresses 9387 something in the per-function constant pool. 9388 9389 In a `call_insn' 1 means that this instruction is a call to a const 9390 function. 9391 9392 In an RTL dump, this flag is represented as `/u'. 9393 9394 `used' 9395 This flag is used directly (without an access macro) at the end of 9396 RTL generation for a function, to count the number of times an 9397 expression appears in insns. Expressions that appear more than 9398 once are copied, according to the rules for shared structure 9399 (*note Sharing::). 9400 9401 For a `reg', it is used directly (without an access macro) by the 9402 leaf register renumbering code to ensure that each register is only 9403 renumbered once. 9404 9405 In a `symbol_ref', it indicates that an external declaration for 9406 the symbol has already been written. 9407 9408 `volatil' 9409 In a `mem', `asm_operands', or `asm_input' expression, it is 1 if 9410 the memory reference is volatile. Volatile memory references may 9411 not be deleted, reordered or combined. 9412 9413 In a `symbol_ref' expression, it is used for machine-specific 9414 purposes. 9415 9416 In a `reg' expression, it is 1 if the value is a user-level 9417 variable. 0 indicates an internal compiler temporary. 9418 9419 In an `insn', 1 means the insn has been deleted. 9420 9421 In `label_ref' and `reg_label' expressions, 1 means a reference to 9422 a non-local label. 9423 9424 In an RTL dump, this flag is represented as `/v'. 9425 9426 9427 File: gccint.info, Node: Machine Modes, Next: Constants, Prev: Flags, Up: RTL 9428 9429 10.6 Machine Modes 9430 ================== 9431 9432 A machine mode describes a size of data object and the representation 9433 used for it. In the C code, machine modes are represented by an 9434 enumeration type, `enum machine_mode', defined in `machmode.def'. Each 9435 RTL expression has room for a machine mode and so do certain kinds of 9436 tree expressions (declarations and types, to be precise). 9437 9438 In debugging dumps and machine descriptions, the machine mode of an RTL 9439 expression is written after the expression code with a colon to separate 9440 them. The letters `mode' which appear at the end of each machine mode 9441 name are omitted. For example, `(reg:SI 38)' is a `reg' expression 9442 with machine mode `SImode'. If the mode is `VOIDmode', it is not 9443 written at all. 9444 9445 Here is a table of machine modes. The term "byte" below refers to an 9446 object of `BITS_PER_UNIT' bits (*note Storage Layout::). 9447 9448 `BImode' 9449 "Bit" mode represents a single bit, for predicate registers. 9450 9451 `QImode' 9452 "Quarter-Integer" mode represents a single byte treated as an 9453 integer. 9454 9455 `HImode' 9456 "Half-Integer" mode represents a two-byte integer. 9457 9458 `PSImode' 9459 "Partial Single Integer" mode represents an integer which occupies 9460 four bytes but which doesn't really use all four. On some 9461 machines, this is the right mode to use for pointers. 9462 9463 `SImode' 9464 "Single Integer" mode represents a four-byte integer. 9465 9466 `PDImode' 9467 "Partial Double Integer" mode represents an integer which occupies 9468 eight bytes but which doesn't really use all eight. On some 9469 machines, this is the right mode to use for certain pointers. 9470 9471 `DImode' 9472 "Double Integer" mode represents an eight-byte integer. 9473 9474 `TImode' 9475 "Tetra Integer" (?) mode represents a sixteen-byte integer. 9476 9477 `OImode' 9478 "Octa Integer" (?) mode represents a thirty-two-byte integer. 9479 9480 `QFmode' 9481 "Quarter-Floating" mode represents a quarter-precision (single 9482 byte) floating point number. 9483 9484 `HFmode' 9485 "Half-Floating" mode represents a half-precision (two byte) 9486 floating point number. 9487 9488 `TQFmode' 9489 "Three-Quarter-Floating" (?) mode represents a 9490 three-quarter-precision (three byte) floating point number. 9491 9492 `SFmode' 9493 "Single Floating" mode represents a four byte floating point 9494 number. In the common case, of a processor with IEEE arithmetic 9495 and 8-bit bytes, this is a single-precision IEEE floating point 9496 number; it can also be used for double-precision (on processors 9497 with 16-bit bytes) and single-precision VAX and IBM types. 9498 9499 `DFmode' 9500 "Double Floating" mode represents an eight byte floating point 9501 number. In the common case, of a processor with IEEE arithmetic 9502 and 8-bit bytes, this is a double-precision IEEE floating point 9503 number. 9504 9505 `XFmode' 9506 "Extended Floating" mode represents an IEEE extended floating point 9507 number. This mode only has 80 meaningful bits (ten bytes). Some 9508 processors require such numbers to be padded to twelve bytes, 9509 others to sixteen; this mode is used for either. 9510 9511 `SDmode' 9512 "Single Decimal Floating" mode represents a four byte decimal 9513 floating point number (as distinct from conventional binary 9514 floating point). 9515 9516 `DDmode' 9517 "Double Decimal Floating" mode represents an eight byte decimal 9518 floating point number. 9519 9520 `TDmode' 9521 "Tetra Decimal Floating" mode represents a sixteen byte decimal 9522 floating point number all 128 of whose bits are meaningful. 9523 9524 `TFmode' 9525 "Tetra Floating" mode represents a sixteen byte floating point 9526 number all 128 of whose bits are meaningful. One common use is the 9527 IEEE quad-precision format. 9528 9529 `QQmode' 9530 "Quarter-Fractional" mode represents a single byte treated as a 9531 signed fractional number. The default format is "s.7". 9532 9533 `HQmode' 9534 "Half-Fractional" mode represents a two-byte signed fractional 9535 number. The default format is "s.15". 9536 9537 `SQmode' 9538 "Single Fractional" mode represents a four-byte signed fractional 9539 number. The default format is "s.31". 9540 9541 `DQmode' 9542 "Double Fractional" mode represents an eight-byte signed 9543 fractional number. The default format is "s.63". 9544 9545 `TQmode' 9546 "Tetra Fractional" mode represents a sixteen-byte signed 9547 fractional number. The default format is "s.127". 9548 9549 `UQQmode' 9550 "Unsigned Quarter-Fractional" mode represents a single byte 9551 treated as an unsigned fractional number. The default format is 9552 ".8". 9553 9554 `UHQmode' 9555 "Unsigned Half-Fractional" mode represents a two-byte unsigned 9556 fractional number. The default format is ".16". 9557 9558 `USQmode' 9559 "Unsigned Single Fractional" mode represents a four-byte unsigned 9560 fractional number. The default format is ".32". 9561 9562 `UDQmode' 9563 "Unsigned Double Fractional" mode represents an eight-byte unsigned 9564 fractional number. The default format is ".64". 9565 9566 `UTQmode' 9567 "Unsigned Tetra Fractional" mode represents a sixteen-byte unsigned 9568 fractional number. The default format is ".128". 9569 9570 `HAmode' 9571 "Half-Accumulator" mode represents a two-byte signed accumulator. 9572 The default format is "s8.7". 9573 9574 `SAmode' 9575 "Single Accumulator" mode represents a four-byte signed 9576 accumulator. The default format is "s16.15". 9577 9578 `DAmode' 9579 "Double Accumulator" mode represents an eight-byte signed 9580 accumulator. The default format is "s32.31". 9581 9582 `TAmode' 9583 "Tetra Accumulator" mode represents a sixteen-byte signed 9584 accumulator. The default format is "s64.63". 9585 9586 `UHAmode' 9587 "Unsigned Half-Accumulator" mode represents a two-byte unsigned 9588 accumulator. The default format is "8.8". 9589 9590 `USAmode' 9591 "Unsigned Single Accumulator" mode represents a four-byte unsigned 9592 accumulator. The default format is "16.16". 9593 9594 `UDAmode' 9595 "Unsigned Double Accumulator" mode represents an eight-byte 9596 unsigned accumulator. The default format is "32.32". 9597 9598 `UTAmode' 9599 "Unsigned Tetra Accumulator" mode represents a sixteen-byte 9600 unsigned accumulator. The default format is "64.64". 9601 9602 `CCmode' 9603 "Condition Code" mode represents the value of a condition code, 9604 which is a machine-specific set of bits used to represent the 9605 result of a comparison operation. Other machine-specific modes 9606 may also be used for the condition code. These modes are not used 9607 on machines that use `cc0' (see *note Condition Code::). 9608 9609 `BLKmode' 9610 "Block" mode represents values that are aggregates to which none of 9611 the other modes apply. In RTL, only memory references can have 9612 this mode, and only if they appear in string-move or vector 9613 instructions. On machines which have no such instructions, 9614 `BLKmode' will not appear in RTL. 9615 9616 `VOIDmode' 9617 Void mode means the absence of a mode or an unspecified mode. For 9618 example, RTL expressions of code `const_int' have mode `VOIDmode' 9619 because they can be taken to have whatever mode the context 9620 requires. In debugging dumps of RTL, `VOIDmode' is expressed by 9621 the absence of any mode. 9622 9623 `QCmode, HCmode, SCmode, DCmode, XCmode, TCmode' 9624 These modes stand for a complex number represented as a pair of 9625 floating point values. The floating point values are in `QFmode', 9626 `HFmode', `SFmode', `DFmode', `XFmode', and `TFmode', respectively. 9627 9628 `CQImode, CHImode, CSImode, CDImode, CTImode, COImode' 9629 These modes stand for a complex number represented as a pair of 9630 integer values. The integer values are in `QImode', `HImode', 9631 `SImode', `DImode', `TImode', and `OImode', respectively. 9632 9633 The machine description defines `Pmode' as a C macro which expands 9634 into the machine mode used for addresses. Normally this is the mode 9635 whose size is `BITS_PER_WORD', `SImode' on 32-bit machines. 9636 9637 The only modes which a machine description must support are `QImode', 9638 and the modes corresponding to `BITS_PER_WORD', `FLOAT_TYPE_SIZE' and 9639 `DOUBLE_TYPE_SIZE'. The compiler will attempt to use `DImode' for 9640 8-byte structures and unions, but this can be prevented by overriding 9641 the definition of `MAX_FIXED_MODE_SIZE'. Alternatively, you can have 9642 the compiler use `TImode' for 16-byte structures and unions. Likewise, 9643 you can arrange for the C type `short int' to avoid using `HImode'. 9644 9645 Very few explicit references to machine modes remain in the compiler 9646 and these few references will soon be removed. Instead, the machine 9647 modes are divided into mode classes. These are represented by the 9648 enumeration type `enum mode_class' defined in `machmode.h'. The 9649 possible mode classes are: 9650 9651 `MODE_INT' 9652 Integer modes. By default these are `BImode', `QImode', `HImode', 9653 `SImode', `DImode', `TImode', and `OImode'. 9654 9655 `MODE_PARTIAL_INT' 9656 The "partial integer" modes, `PQImode', `PHImode', `PSImode' and 9657 `PDImode'. 9658 9659 `MODE_FLOAT' 9660 Floating point modes. By default these are `QFmode', `HFmode', 9661 `TQFmode', `SFmode', `DFmode', `XFmode' and `TFmode'. 9662 9663 `MODE_DECIMAL_FLOAT' 9664 Decimal floating point modes. By default these are `SDmode', 9665 `DDmode' and `TDmode'. 9666 9667 `MODE_FRACT' 9668 Signed fractional modes. By default these are `QQmode', `HQmode', 9669 `SQmode', `DQmode' and `TQmode'. 9670 9671 `MODE_UFRACT' 9672 Unsigned fractional modes. By default these are `UQQmode', 9673 `UHQmode', `USQmode', `UDQmode' and `UTQmode'. 9674 9675 `MODE_ACCUM' 9676 Signed accumulator modes. By default these are `HAmode', 9677 `SAmode', `DAmode' and `TAmode'. 9678 9679 `MODE_UACCUM' 9680 Unsigned accumulator modes. By default these are `UHAmode', 9681 `USAmode', `UDAmode' and `UTAmode'. 9682 9683 `MODE_COMPLEX_INT' 9684 Complex integer modes. (These are not currently implemented). 9685 9686 `MODE_COMPLEX_FLOAT' 9687 Complex floating point modes. By default these are `QCmode', 9688 `HCmode', `SCmode', `DCmode', `XCmode', and `TCmode'. 9689 9690 `MODE_FUNCTION' 9691 Algol or Pascal function variables including a static chain. 9692 (These are not currently implemented). 9693 9694 `MODE_CC' 9695 Modes representing condition code values. These are `CCmode' plus 9696 any `CC_MODE' modes listed in the `MACHINE-modes.def'. *Note Jump 9697 Patterns::, also see *Note Condition Code::. 9698 9699 `MODE_RANDOM' 9700 This is a catchall mode class for modes which don't fit into the 9701 above classes. Currently `VOIDmode' and `BLKmode' are in 9702 `MODE_RANDOM'. 9703 9704 Here are some C macros that relate to machine modes: 9705 9706 `GET_MODE (X)' 9707 Returns the machine mode of the RTX X. 9708 9709 `PUT_MODE (X, NEWMODE)' 9710 Alters the machine mode of the RTX X to be NEWMODE. 9711 9712 `NUM_MACHINE_MODES' 9713 Stands for the number of machine modes available on the target 9714 machine. This is one greater than the largest numeric value of any 9715 machine mode. 9716 9717 `GET_MODE_NAME (M)' 9718 Returns the name of mode M as a string. 9719 9720 `GET_MODE_CLASS (M)' 9721 Returns the mode class of mode M. 9722 9723 `GET_MODE_WIDER_MODE (M)' 9724 Returns the next wider natural mode. For example, the expression 9725 `GET_MODE_WIDER_MODE (QImode)' returns `HImode'. 9726 9727 `GET_MODE_SIZE (M)' 9728 Returns the size in bytes of a datum of mode M. 9729 9730 `GET_MODE_BITSIZE (M)' 9731 Returns the size in bits of a datum of mode M. 9732 9733 `GET_MODE_IBIT (M)' 9734 Returns the number of integral bits of a datum of fixed-point mode 9735 M. 9736 9737 `GET_MODE_FBIT (M)' 9738 Returns the number of fractional bits of a datum of fixed-point 9739 mode M. 9740 9741 `GET_MODE_MASK (M)' 9742 Returns a bitmask containing 1 for all bits in a word that fit 9743 within mode M. This macro can only be used for modes whose 9744 bitsize is less than or equal to `HOST_BITS_PER_INT'. 9745 9746 `GET_MODE_ALIGNMENT (M)' 9747 Return the required alignment, in bits, for an object of mode M. 9748 9749 `GET_MODE_UNIT_SIZE (M)' 9750 Returns the size in bytes of the subunits of a datum of mode M. 9751 This is the same as `GET_MODE_SIZE' except in the case of complex 9752 modes. For them, the unit size is the size of the real or 9753 imaginary part. 9754 9755 `GET_MODE_NUNITS (M)' 9756 Returns the number of units contained in a mode, i.e., 9757 `GET_MODE_SIZE' divided by `GET_MODE_UNIT_SIZE'. 9758 9759 `GET_CLASS_NARROWEST_MODE (C)' 9760 Returns the narrowest mode in mode class C. 9761 9762 The global variables `byte_mode' and `word_mode' contain modes whose 9763 classes are `MODE_INT' and whose bitsizes are either `BITS_PER_UNIT' or 9764 `BITS_PER_WORD', respectively. On 32-bit machines, these are `QImode' 9765 and `SImode', respectively. 9766 9767 9768 File: gccint.info, Node: Constants, Next: Regs and Memory, Prev: Machine Modes, Up: RTL 9769 9770 10.7 Constant Expression Types 9771 ============================== 9772 9773 The simplest RTL expressions are those that represent constant values. 9774 9775 `(const_int I)' 9776 This type of expression represents the integer value I. I is 9777 customarily accessed with the macro `INTVAL' as in `INTVAL (EXP)', 9778 which is equivalent to `XWINT (EXP, 0)'. 9779 9780 Constants generated for modes with fewer bits than `HOST_WIDE_INT' 9781 must be sign extended to full width (e.g., with `gen_int_mode'). 9782 9783 There is only one expression object for the integer value zero; it 9784 is the value of the variable `const0_rtx'. Likewise, the only 9785 expression for integer value one is found in `const1_rtx', the only 9786 expression for integer value two is found in `const2_rtx', and the 9787 only expression for integer value negative one is found in 9788 `constm1_rtx'. Any attempt to create an expression of code 9789 `const_int' and value zero, one, two or negative one will return 9790 `const0_rtx', `const1_rtx', `const2_rtx' or `constm1_rtx' as 9791 appropriate. 9792 9793 Similarly, there is only one object for the integer whose value is 9794 `STORE_FLAG_VALUE'. It is found in `const_true_rtx'. If 9795 `STORE_FLAG_VALUE' is one, `const_true_rtx' and `const1_rtx' will 9796 point to the same object. If `STORE_FLAG_VALUE' is -1, 9797 `const_true_rtx' and `constm1_rtx' will point to the same object. 9798 9799 `(const_double:M I0 I1 ...)' 9800 Represents either a floating-point constant of mode M or an 9801 integer constant too large to fit into `HOST_BITS_PER_WIDE_INT' 9802 bits but small enough to fit within twice that number of bits (GCC 9803 does not provide a mechanism to represent even larger constants). 9804 In the latter case, M will be `VOIDmode'. 9805 9806 If M is `VOIDmode', the bits of the value are stored in I0 and I1. 9807 I0 is customarily accessed with the macro `CONST_DOUBLE_LOW' and 9808 I1 with `CONST_DOUBLE_HIGH'. 9809 9810 If the constant is floating point (regardless of its precision), 9811 then the number of integers used to store the value depends on the 9812 size of `REAL_VALUE_TYPE' (*note Floating Point::). The integers 9813 represent a floating point number, but not precisely in the target 9814 machine's or host machine's floating point format. To convert 9815 them to the precise bit pattern used by the target machine, use 9816 the macro `REAL_VALUE_TO_TARGET_DOUBLE' and friends (*note Data 9817 Output::). 9818 9819 `(const_fixed:M ...)' 9820 Represents a fixed-point constant of mode M. The operand is a 9821 data structure of type `struct fixed_value' and is accessed with 9822 the macro `CONST_FIXED_VALUE'. The high part of data is accessed 9823 with `CONST_FIXED_VALUE_HIGH'; the low part is accessed with 9824 `CONST_FIXED_VALUE_LOW'. 9825 9826 `(const_vector:M [X0 X1 ...])' 9827 Represents a vector constant. The square brackets stand for the 9828 vector containing the constant elements. X0, X1 and so on are the 9829 `const_int', `const_double' or `const_fixed' elements. 9830 9831 The number of units in a `const_vector' is obtained with the macro 9832 `CONST_VECTOR_NUNITS' as in `CONST_VECTOR_NUNITS (V)'. 9833 9834 Individual elements in a vector constant are accessed with the 9835 macro `CONST_VECTOR_ELT' as in `CONST_VECTOR_ELT (V, N)' where V 9836 is the vector constant and N is the element desired. 9837 9838 `(const_string STR)' 9839 Represents a constant string with value STR. Currently this is 9840 used only for insn attributes (*note Insn Attributes::) since 9841 constant strings in C are placed in memory. 9842 9843 `(symbol_ref:MODE SYMBOL)' 9844 Represents the value of an assembler label for data. SYMBOL is a 9845 string that describes the name of the assembler label. If it 9846 starts with a `*', the label is the rest of SYMBOL not including 9847 the `*'. Otherwise, the label is SYMBOL, usually prefixed with 9848 `_'. 9849 9850 The `symbol_ref' contains a mode, which is usually `Pmode'. 9851 Usually that is the only mode for which a symbol is directly valid. 9852 9853 `(label_ref:MODE LABEL)' 9854 Represents the value of an assembler label for code. It contains 9855 one operand, an expression, which must be a `code_label' or a 9856 `note' of type `NOTE_INSN_DELETED_LABEL' that appears in the 9857 instruction sequence to identify the place where the label should 9858 go. 9859 9860 The reason for using a distinct expression type for code label 9861 references is so that jump optimization can distinguish them. 9862 9863 The `label_ref' contains a mode, which is usually `Pmode'. 9864 Usually that is the only mode for which a label is directly valid. 9865 9866 `(const:M EXP)' 9867 Represents a constant that is the result of an assembly-time 9868 arithmetic computation. The operand, EXP, is an expression that 9869 contains only constants (`const_int', `symbol_ref' and `label_ref' 9870 expressions) combined with `plus' and `minus'. However, not all 9871 combinations are valid, since the assembler cannot do arbitrary 9872 arithmetic on relocatable symbols. 9873 9874 M should be `Pmode'. 9875 9876 `(high:M EXP)' 9877 Represents the high-order bits of EXP, usually a `symbol_ref'. 9878 The number of bits is machine-dependent and is normally the number 9879 of bits specified in an instruction that initializes the high 9880 order bits of a register. It is used with `lo_sum' to represent 9881 the typical two-instruction sequence used in RISC machines to 9882 reference a global memory location. 9883 9884 M should be `Pmode'. 9885 9886 The macro `CONST0_RTX (MODE)' refers to an expression with value 0 in 9887 mode MODE. If mode MODE is of mode class `MODE_INT', it returns 9888 `const0_rtx'. If mode MODE is of mode class `MODE_FLOAT', it returns a 9889 `CONST_DOUBLE' expression in mode MODE. Otherwise, it returns a 9890 `CONST_VECTOR' expression in mode MODE. Similarly, the macro 9891 `CONST1_RTX (MODE)' refers to an expression with value 1 in mode MODE 9892 and similarly for `CONST2_RTX'. The `CONST1_RTX' and `CONST2_RTX' 9893 macros are undefined for vector modes. 9894 9895 9896 File: gccint.info, Node: Regs and Memory, Next: Arithmetic, Prev: Constants, Up: RTL 9897 9898 10.8 Registers and Memory 9899 ========================= 9900 9901 Here are the RTL expression types for describing access to machine 9902 registers and to main memory. 9903 9904 `(reg:M N)' 9905 For small values of the integer N (those that are less than 9906 `FIRST_PSEUDO_REGISTER'), this stands for a reference to machine 9907 register number N: a "hard register". For larger values of N, it 9908 stands for a temporary value or "pseudo register". The compiler's 9909 strategy is to generate code assuming an unlimited number of such 9910 pseudo registers, and later convert them into hard registers or 9911 into memory references. 9912 9913 M is the machine mode of the reference. It is necessary because 9914 machines can generally refer to each register in more than one 9915 mode. For example, a register may contain a full word but there 9916 may be instructions to refer to it as a half word or as a single 9917 byte, as well as instructions to refer to it as a floating point 9918 number of various precisions. 9919 9920 Even for a register that the machine can access in only one mode, 9921 the mode must always be specified. 9922 9923 The symbol `FIRST_PSEUDO_REGISTER' is defined by the machine 9924 description, since the number of hard registers on the machine is 9925 an invariant characteristic of the machine. Note, however, that 9926 not all of the machine registers must be general registers. All 9927 the machine registers that can be used for storage of data are 9928 given hard register numbers, even those that can be used only in 9929 certain instructions or can hold only certain types of data. 9930 9931 A hard register may be accessed in various modes throughout one 9932 function, but each pseudo register is given a natural mode and is 9933 accessed only in that mode. When it is necessary to describe an 9934 access to a pseudo register using a nonnatural mode, a `subreg' 9935 expression is used. 9936 9937 A `reg' expression with a machine mode that specifies more than 9938 one word of data may actually stand for several consecutive 9939 registers. If in addition the register number specifies a 9940 hardware register, then it actually represents several consecutive 9941 hardware registers starting with the specified one. 9942 9943 Each pseudo register number used in a function's RTL code is 9944 represented by a unique `reg' expression. 9945 9946 Some pseudo register numbers, those within the range of 9947 `FIRST_VIRTUAL_REGISTER' to `LAST_VIRTUAL_REGISTER' only appear 9948 during the RTL generation phase and are eliminated before the 9949 optimization phases. These represent locations in the stack frame 9950 that cannot be determined until RTL generation for the function 9951 has been completed. The following virtual register numbers are 9952 defined: 9953 9954 `VIRTUAL_INCOMING_ARGS_REGNUM' 9955 This points to the first word of the incoming arguments 9956 passed on the stack. Normally these arguments are placed 9957 there by the caller, but the callee may have pushed some 9958 arguments that were previously passed in registers. 9959 9960 When RTL generation is complete, this virtual register is 9961 replaced by the sum of the register given by 9962 `ARG_POINTER_REGNUM' and the value of `FIRST_PARM_OFFSET'. 9963 9964 `VIRTUAL_STACK_VARS_REGNUM' 9965 If `FRAME_GROWS_DOWNWARD' is defined to a nonzero value, this 9966 points to immediately above the first variable on the stack. 9967 Otherwise, it points to the first variable on the stack. 9968 9969 `VIRTUAL_STACK_VARS_REGNUM' is replaced with the sum of the 9970 register given by `FRAME_POINTER_REGNUM' and the value 9971 `STARTING_FRAME_OFFSET'. 9972 9973 `VIRTUAL_STACK_DYNAMIC_REGNUM' 9974 This points to the location of dynamically allocated memory 9975 on the stack immediately after the stack pointer has been 9976 adjusted by the amount of memory desired. 9977 9978 This virtual register is replaced by the sum of the register 9979 given by `STACK_POINTER_REGNUM' and the value 9980 `STACK_DYNAMIC_OFFSET'. 9981 9982 `VIRTUAL_OUTGOING_ARGS_REGNUM' 9983 This points to the location in the stack at which outgoing 9984 arguments should be written when the stack is pre-pushed 9985 (arguments pushed using push insns should always use 9986 `STACK_POINTER_REGNUM'). 9987 9988 This virtual register is replaced by the sum of the register 9989 given by `STACK_POINTER_REGNUM' and the value 9990 `STACK_POINTER_OFFSET'. 9991 9992 `(subreg:M1 REG:M2 BYTENUM)' 9993 `subreg' expressions are used to refer to a register in a machine 9994 mode other than its natural one, or to refer to one register of a 9995 multi-part `reg' that actually refers to several registers. 9996 9997 Each pseudo register has a natural mode. If it is necessary to 9998 operate on it in a different mode, the register must be enclosed 9999 in a `subreg'. 10000 10001 There are currently three supported types for the first operand of 10002 a `subreg': 10003 * pseudo registers This is the most common case. Most 10004 `subreg's have pseudo `reg's as their first operand. 10005 10006 * mem `subreg's of `mem' were common in earlier versions of GCC 10007 and are still supported. During the reload pass these are 10008 replaced by plain `mem's. On machines that do not do 10009 instruction scheduling, use of `subreg's of `mem' are still 10010 used, but this is no longer recommended. Such `subreg's are 10011 considered to be `register_operand's rather than 10012 `memory_operand's before and during reload. Because of this, 10013 the scheduling passes cannot properly schedule instructions 10014 with `subreg's of `mem', so for machines that do scheduling, 10015 `subreg's of `mem' should never be used. To support this, 10016 the combine and recog passes have explicit code to inhibit 10017 the creation of `subreg's of `mem' when `INSN_SCHEDULING' is 10018 defined. 10019 10020 The use of `subreg's of `mem' after the reload pass is an area 10021 that is not well understood and should be avoided. There is 10022 still some code in the compiler to support this, but this 10023 code has possibly rotted. This use of `subreg's is 10024 discouraged and will most likely not be supported in the 10025 future. 10026 10027 * hard registers It is seldom necessary to wrap hard registers 10028 in `subreg's; such registers would normally reduce to a 10029 single `reg' rtx. This use of `subreg's is discouraged and 10030 may not be supported in the future. 10031 10032 10033 `subreg's of `subreg's are not supported. Using 10034 `simplify_gen_subreg' is the recommended way to avoid this problem. 10035 10036 `subreg's come in two distinct flavors, each having its own usage 10037 and rules: 10038 10039 Paradoxical subregs 10040 When M1 is strictly wider than M2, the `subreg' expression is 10041 called "paradoxical". The canonical test for this class of 10042 `subreg' is: 10043 10044 GET_MODE_SIZE (M1) > GET_MODE_SIZE (M2) 10045 10046 Paradoxical `subreg's can be used as both lvalues and rvalues. 10047 When used as an lvalue, the low-order bits of the source value 10048 are stored in REG and the high-order bits are discarded. 10049 When used as an rvalue, the low-order bits of the `subreg' are 10050 taken from REG while the high-order bits may or may not be 10051 defined. 10052 10053 The high-order bits of rvalues are in the following 10054 circumstances: 10055 10056 * `subreg's of `mem' When M2 is smaller than a word, the 10057 macro `LOAD_EXTEND_OP', can control how the high-order 10058 bits are defined. 10059 10060 * `subreg' of `reg's The upper bits are defined when 10061 `SUBREG_PROMOTED_VAR_P' is true. 10062 `SUBREG_PROMOTED_UNSIGNED_P' describes what the upper 10063 bits hold. Such subregs usually represent local 10064 variables, register variables and parameter pseudo 10065 variables that have been promoted to a wider mode. 10066 10067 10068 BYTENUM is always zero for a paradoxical `subreg', even on 10069 big-endian targets. 10070 10071 For example, the paradoxical `subreg': 10072 10073 (set (subreg:SI (reg:HI X) 0) Y) 10074 10075 stores the lower 2 bytes of Y in X and discards the upper 2 10076 bytes. A subsequent: 10077 10078 (set Z (subreg:SI (reg:HI X) 0)) 10079 10080 would set the lower two bytes of Z to Y and set the upper two 10081 bytes to an unknown value assuming `SUBREG_PROMOTED_VAR_P' is 10082 false. 10083 10084 Normal subregs 10085 When M1 is at least as narrow as M2 the `subreg' expression 10086 is called "normal". 10087 10088 Normal `subreg's restrict consideration to certain bits of 10089 REG. There are two cases. If M1 is smaller than a word, the 10090 `subreg' refers to the least-significant part (or "lowpart") 10091 of one word of REG. If M1 is word-sized or greater, the 10092 `subreg' refers to one or more complete words. 10093 10094 When used as an lvalue, `subreg' is a word-based accessor. 10095 Storing to a `subreg' modifies all the words of REG that 10096 overlap the `subreg', but it leaves the other words of REG 10097 alone. 10098 10099 When storing to a normal `subreg' that is smaller than a word, 10100 the other bits of the referenced word are usually left in an 10101 undefined state. This laxity makes it easier to generate 10102 efficient code for such instructions. To represent an 10103 instruction that preserves all the bits outside of those in 10104 the `subreg', use `strict_low_part' or `zero_extract' around 10105 the `subreg'. 10106 10107 BYTENUM must identify the offset of the first byte of the 10108 `subreg' from the start of REG, assuming that REG is laid out 10109 in memory order. The memory order of bytes is defined by two 10110 target macros, `WORDS_BIG_ENDIAN' and `BYTES_BIG_ENDIAN': 10111 10112 * `WORDS_BIG_ENDIAN', if set to 1, says that byte number 10113 zero is part of the most significant word; otherwise, it 10114 is part of the least significant word. 10115 10116 * `BYTES_BIG_ENDIAN', if set to 1, says that byte number 10117 zero is the most significant byte within a word; 10118 otherwise, it is the least significant byte within a 10119 word. 10120 10121 On a few targets, `FLOAT_WORDS_BIG_ENDIAN' disagrees with 10122 `WORDS_BIG_ENDIAN'. However, most parts of the compiler treat 10123 floating point values as if they had the same endianness as 10124 integer values. This works because they handle them solely 10125 as a collection of integer values, with no particular 10126 numerical value. Only real.c and the runtime libraries care 10127 about `FLOAT_WORDS_BIG_ENDIAN'. 10128 10129 Thus, 10130 10131 (subreg:HI (reg:SI X) 2) 10132 10133 on a `BYTES_BIG_ENDIAN', `UNITS_PER_WORD == 4' target is the 10134 same as 10135 10136 (subreg:HI (reg:SI X) 0) 10137 10138 on a little-endian, `UNITS_PER_WORD == 4' target. Both 10139 `subreg's access the lower two bytes of register X. 10140 10141 10142 A `MODE_PARTIAL_INT' mode behaves as if it were as wide as the 10143 corresponding `MODE_INT' mode, except that it has an unknown 10144 number of undefined bits. For example: 10145 10146 (subreg:PSI (reg:SI 0) 0) 10147 10148 accesses the whole of `(reg:SI 0)', but the exact relationship 10149 between the `PSImode' value and the `SImode' value is not defined. 10150 If we assume `UNITS_PER_WORD <= 4', then the following two 10151 `subreg's: 10152 10153 (subreg:PSI (reg:DI 0) 0) 10154 (subreg:PSI (reg:DI 0) 4) 10155 10156 represent independent 4-byte accesses to the two halves of 10157 `(reg:DI 0)'. Both `subreg's have an unknown number of undefined 10158 bits. 10159 10160 If `UNITS_PER_WORD <= 2' then these two `subreg's: 10161 10162 (subreg:HI (reg:PSI 0) 0) 10163 (subreg:HI (reg:PSI 0) 2) 10164 10165 represent independent 2-byte accesses that together span the whole 10166 of `(reg:PSI 0)'. Storing to the first `subreg' does not affect 10167 the value of the second, and vice versa. `(reg:PSI 0)' has an 10168 unknown number of undefined bits, so the assignment: 10169 10170 (set (subreg:HI (reg:PSI 0) 0) (reg:HI 4)) 10171 10172 does not guarantee that `(subreg:HI (reg:PSI 0) 0)' has the value 10173 `(reg:HI 4)'. 10174 10175 The rules above apply to both pseudo REGs and hard REGs. If the 10176 semantics are not correct for particular combinations of M1, M2 10177 and hard REG, the target-specific code must ensure that those 10178 combinations are never used. For example: 10179 10180 CANNOT_CHANGE_MODE_CLASS (M2, M1, CLASS) 10181 10182 must be true for every class CLASS that includes REG. 10183 10184 The first operand of a `subreg' expression is customarily accessed 10185 with the `SUBREG_REG' macro and the second operand is customarily 10186 accessed with the `SUBREG_BYTE' macro. 10187 10188 It has been several years since a platform in which 10189 `BYTES_BIG_ENDIAN' not equal to `WORDS_BIG_ENDIAN' has been 10190 tested. Anyone wishing to support such a platform in the future 10191 may be confronted with code rot. 10192 10193 `(scratch:M)' 10194 This represents a scratch register that will be required for the 10195 execution of a single instruction and not used subsequently. It is 10196 converted into a `reg' by either the local register allocator or 10197 the reload pass. 10198 10199 `scratch' is usually present inside a `clobber' operation (*note 10200 Side Effects::). 10201 10202 `(cc0)' 10203 This refers to the machine's condition code register. It has no 10204 operands and may not have a machine mode. There are two ways to 10205 use it: 10206 10207 * To stand for a complete set of condition code flags. This is 10208 best on most machines, where each comparison sets the entire 10209 series of flags. 10210 10211 With this technique, `(cc0)' may be validly used in only two 10212 contexts: as the destination of an assignment (in test and 10213 compare instructions) and in comparison operators comparing 10214 against zero (`const_int' with value zero; that is to say, 10215 `const0_rtx'). 10216 10217 * To stand for a single flag that is the result of a single 10218 condition. This is useful on machines that have only a 10219 single flag bit, and in which comparison instructions must 10220 specify the condition to test. 10221 10222 With this technique, `(cc0)' may be validly used in only two 10223 contexts: as the destination of an assignment (in test and 10224 compare instructions) where the source is a comparison 10225 operator, and as the first operand of `if_then_else' (in a 10226 conditional branch). 10227 10228 There is only one expression object of code `cc0'; it is the value 10229 of the variable `cc0_rtx'. Any attempt to create an expression of 10230 code `cc0' will return `cc0_rtx'. 10231 10232 Instructions can set the condition code implicitly. On many 10233 machines, nearly all instructions set the condition code based on 10234 the value that they compute or store. It is not necessary to 10235 record these actions explicitly in the RTL because the machine 10236 description includes a prescription for recognizing the 10237 instructions that do so (by means of the macro 10238 `NOTICE_UPDATE_CC'). *Note Condition Code::. Only instructions 10239 whose sole purpose is to set the condition code, and instructions 10240 that use the condition code, need mention `(cc0)'. 10241 10242 On some machines, the condition code register is given a register 10243 number and a `reg' is used instead of `(cc0)'. This is usually the 10244 preferable approach if only a small subset of instructions modify 10245 the condition code. Other machines store condition codes in 10246 general registers; in such cases a pseudo register should be used. 10247 10248 Some machines, such as the SPARC and RS/6000, have two sets of 10249 arithmetic instructions, one that sets and one that does not set 10250 the condition code. This is best handled by normally generating 10251 the instruction that does not set the condition code, and making a 10252 pattern that both performs the arithmetic and sets the condition 10253 code register (which would not be `(cc0)' in this case). For 10254 examples, search for `addcc' and `andcc' in `sparc.md'. 10255 10256 `(pc)' 10257 This represents the machine's program counter. It has no operands 10258 and may not have a machine mode. `(pc)' may be validly used only 10259 in certain specific contexts in jump instructions. 10260 10261 There is only one expression object of code `pc'; it is the value 10262 of the variable `pc_rtx'. Any attempt to create an expression of 10263 code `pc' will return `pc_rtx'. 10264 10265 All instructions that do not jump alter the program counter 10266 implicitly by incrementing it, but there is no need to mention 10267 this in the RTL. 10268 10269 `(mem:M ADDR ALIAS)' 10270 This RTX represents a reference to main memory at an address 10271 represented by the expression ADDR. M specifies how large a unit 10272 of memory is accessed. ALIAS specifies an alias set for the 10273 reference. In general two items are in different alias sets if 10274 they cannot reference the same memory address. 10275 10276 The construct `(mem:BLK (scratch))' is considered to alias all 10277 other memories. Thus it may be used as a memory barrier in 10278 epilogue stack deallocation patterns. 10279 10280 `(concatM RTX RTX)' 10281 This RTX represents the concatenation of two other RTXs. This is 10282 used for complex values. It should only appear in the RTL 10283 attached to declarations and during RTL generation. It should not 10284 appear in the ordinary insn chain. 10285 10286 `(concatnM [RTX ...])' 10287 This RTX represents the concatenation of all the RTX to make a 10288 single value. Like `concat', this should only appear in 10289 declarations, and not in the insn chain. 10290 10291 10292 File: gccint.info, Node: Arithmetic, Next: Comparisons, Prev: Regs and Memory, Up: RTL 10293 10294 10.9 RTL Expressions for Arithmetic 10295 =================================== 10296 10297 Unless otherwise specified, all the operands of arithmetic expressions 10298 must be valid for mode M. An operand is valid for mode M if it has 10299 mode M, or if it is a `const_int' or `const_double' and M is a mode of 10300 class `MODE_INT'. 10301 10302 For commutative binary operations, constants should be placed in the 10303 second operand. 10304 10305 `(plus:M X Y)' 10306 `(ss_plus:M X Y)' 10307 `(us_plus:M X Y)' 10308 These three expressions all represent the sum of the values 10309 represented by X and Y carried out in machine mode M. They differ 10310 in their behavior on overflow of integer modes. `plus' wraps 10311 round modulo the width of M; `ss_plus' saturates at the maximum 10312 signed value representable in M; `us_plus' saturates at the 10313 maximum unsigned value. 10314 10315 `(lo_sum:M X Y)' 10316 This expression represents the sum of X and the low-order bits of 10317 Y. It is used with `high' (*note Constants::) to represent the 10318 typical two-instruction sequence used in RISC machines to 10319 reference a global memory location. 10320 10321 The number of low order bits is machine-dependent but is normally 10322 the number of bits in a `Pmode' item minus the number of bits set 10323 by `high'. 10324 10325 M should be `Pmode'. 10326 10327 `(minus:M X Y)' 10328 `(ss_minus:M X Y)' 10329 `(us_minus:M X Y)' 10330 These three expressions represent the result of subtracting Y from 10331 X, carried out in mode M. Behavior on overflow is the same as for 10332 the three variants of `plus' (see above). 10333 10334 `(compare:M X Y)' 10335 Represents the result of subtracting Y from X for purposes of 10336 comparison. The result is computed without overflow, as if with 10337 infinite precision. 10338 10339 Of course, machines can't really subtract with infinite precision. 10340 However, they can pretend to do so when only the sign of the 10341 result will be used, which is the case when the result is stored 10342 in the condition code. And that is the _only_ way this kind of 10343 expression may validly be used: as a value to be stored in the 10344 condition codes, either `(cc0)' or a register. *Note 10345 Comparisons::. 10346 10347 The mode M is not related to the modes of X and Y, but instead is 10348 the mode of the condition code value. If `(cc0)' is used, it is 10349 `VOIDmode'. Otherwise it is some mode in class `MODE_CC', often 10350 `CCmode'. *Note Condition Code::. If M is `VOIDmode' or 10351 `CCmode', the operation returns sufficient information (in an 10352 unspecified format) so that any comparison operator can be applied 10353 to the result of the `COMPARE' operation. For other modes in 10354 class `MODE_CC', the operation only returns a subset of this 10355 information. 10356 10357 Normally, X and Y must have the same mode. Otherwise, `compare' 10358 is valid only if the mode of X is in class `MODE_INT' and Y is a 10359 `const_int' or `const_double' with mode `VOIDmode'. The mode of X 10360 determines what mode the comparison is to be done in; thus it must 10361 not be `VOIDmode'. 10362 10363 If one of the operands is a constant, it should be placed in the 10364 second operand and the comparison code adjusted as appropriate. 10365 10366 A `compare' specifying two `VOIDmode' constants is not valid since 10367 there is no way to know in what mode the comparison is to be 10368 performed; the comparison must either be folded during the 10369 compilation or the first operand must be loaded into a register 10370 while its mode is still known. 10371 10372 `(neg:M X)' 10373 `(ss_neg:M X)' 10374 `(us_neg:M X)' 10375 These two expressions represent the negation (subtraction from 10376 zero) of the value represented by X, carried out in mode M. They 10377 differ in the behavior on overflow of integer modes. In the case 10378 of `neg', the negation of the operand may be a number not 10379 representable in mode M, in which case it is truncated to M. 10380 `ss_neg' and `us_neg' ensure that an out-of-bounds result 10381 saturates to the maximum or minimum signed or unsigned value. 10382 10383 `(mult:M X Y)' 10384 `(ss_mult:M X Y)' 10385 `(us_mult:M X Y)' 10386 Represents the signed product of the values represented by X and Y 10387 carried out in machine mode M. `ss_mult' and `us_mult' ensure 10388 that an out-of-bounds result saturates to the maximum or minimum 10389 signed or unsigned value. 10390 10391 Some machines support a multiplication that generates a product 10392 wider than the operands. Write the pattern for this as 10393 10394 (mult:M (sign_extend:M X) (sign_extend:M Y)) 10395 10396 where M is wider than the modes of X and Y, which need not be the 10397 same. 10398 10399 For unsigned widening multiplication, use the same idiom, but with 10400 `zero_extend' instead of `sign_extend'. 10401 10402 `(div:M X Y)' 10403 `(ss_div:M X Y)' 10404 Represents the quotient in signed division of X by Y, carried out 10405 in machine mode M. If M is a floating point mode, it represents 10406 the exact quotient; otherwise, the integerized quotient. `ss_div' 10407 ensures that an out-of-bounds result saturates to the maximum or 10408 minimum signed value. 10409 10410 Some machines have division instructions in which the operands and 10411 quotient widths are not all the same; you should represent such 10412 instructions using `truncate' and `sign_extend' as in, 10413 10414 (truncate:M1 (div:M2 X (sign_extend:M2 Y))) 10415 10416 `(udiv:M X Y)' 10417 `(us_div:M X Y)' 10418 Like `div' but represents unsigned division. `us_div' ensures 10419 that an out-of-bounds result saturates to the maximum or minimum 10420 unsigned value. 10421 10422 `(mod:M X Y)' 10423 `(umod:M X Y)' 10424 Like `div' and `udiv' but represent the remainder instead of the 10425 quotient. 10426 10427 `(smin:M X Y)' 10428 `(smax:M X Y)' 10429 Represents the smaller (for `smin') or larger (for `smax') of X 10430 and Y, interpreted as signed values in mode M. When used with 10431 floating point, if both operands are zeros, or if either operand 10432 is `NaN', then it is unspecified which of the two operands is 10433 returned as the result. 10434 10435 `(umin:M X Y)' 10436 `(umax:M X Y)' 10437 Like `smin' and `smax', but the values are interpreted as unsigned 10438 integers. 10439 10440 `(not:M X)' 10441 Represents the bitwise complement of the value represented by X, 10442 carried out in mode M, which must be a fixed-point machine mode. 10443 10444 `(and:M X Y)' 10445 Represents the bitwise logical-and of the values represented by X 10446 and Y, carried out in machine mode M, which must be a fixed-point 10447 machine mode. 10448 10449 `(ior:M X Y)' 10450 Represents the bitwise inclusive-or of the values represented by X 10451 and Y, carried out in machine mode M, which must be a fixed-point 10452 mode. 10453 10454 `(xor:M X Y)' 10455 Represents the bitwise exclusive-or of the values represented by X 10456 and Y, carried out in machine mode M, which must be a fixed-point 10457 mode. 10458 10459 `(ashift:M X C)' 10460 `(ss_ashift:M X C)' 10461 `(us_ashift:M X C)' 10462 These three expressions represent the result of arithmetically 10463 shifting X left by C places. They differ in their behavior on 10464 overflow of integer modes. An `ashift' operation is a plain shift 10465 with no special behavior in case of a change in the sign bit; 10466 `ss_ashift' and `us_ashift' saturates to the minimum or maximum 10467 representable value if any of the bits shifted out differs from 10468 the final sign bit. 10469 10470 X have mode M, a fixed-point machine mode. C be a fixed-point 10471 mode or be a constant with mode `VOIDmode'; which mode is 10472 determined by the mode called for in the machine description entry 10473 for the left-shift instruction. For example, on the VAX, the mode 10474 of C is `QImode' regardless of M. 10475 10476 `(lshiftrt:M X C)' 10477 `(ashiftrt:M X C)' 10478 Like `ashift' but for right shift. Unlike the case for left shift, 10479 these two operations are distinct. 10480 10481 `(rotate:M X C)' 10482 `(rotatert:M X C)' 10483 Similar but represent left and right rotate. If C is a constant, 10484 use `rotate'. 10485 10486 `(abs:M X)' 10487 Represents the absolute value of X, computed in mode M. 10488 10489 `(sqrt:M X)' 10490 Represents the square root of X, computed in mode M. Most often M 10491 will be a floating point mode. 10492 10493 `(ffs:M X)' 10494 Represents one plus the index of the least significant 1-bit in X, 10495 represented as an integer of mode M. (The value is zero if X is 10496 zero.) The mode of X need not be M; depending on the target 10497 machine, various mode combinations may be valid. 10498 10499 `(clz:M X)' 10500 Represents the number of leading 0-bits in X, represented as an 10501 integer of mode M, starting at the most significant bit position. 10502 If X is zero, the value is determined by 10503 `CLZ_DEFINED_VALUE_AT_ZERO' (*note Misc::). Note that this is one 10504 of the few expressions that is not invariant under widening. The 10505 mode of X will usually be an integer mode. 10506 10507 `(ctz:M X)' 10508 Represents the number of trailing 0-bits in X, represented as an 10509 integer of mode M, starting at the least significant bit position. 10510 If X is zero, the value is determined by 10511 `CTZ_DEFINED_VALUE_AT_ZERO' (*note Misc::). Except for this case, 10512 `ctz(x)' is equivalent to `ffs(X) - 1'. The mode of X will 10513 usually be an integer mode. 10514 10515 `(popcount:M X)' 10516 Represents the number of 1-bits in X, represented as an integer of 10517 mode M. The mode of X will usually be an integer mode. 10518 10519 `(parity:M X)' 10520 Represents the number of 1-bits modulo 2 in X, represented as an 10521 integer of mode M. The mode of X will usually be an integer mode. 10522 10523 `(bswap:M X)' 10524 Represents the value X with the order of bytes reversed, carried 10525 out in mode M, which must be a fixed-point machine mode. 10526 10527 10528 File: gccint.info, Node: Comparisons, Next: Bit-Fields, Prev: Arithmetic, Up: RTL 10529 10530 10.10 Comparison Operations 10531 =========================== 10532 10533 Comparison operators test a relation on two operands and are considered 10534 to represent a machine-dependent nonzero value described by, but not 10535 necessarily equal to, `STORE_FLAG_VALUE' (*note Misc::) if the relation 10536 holds, or zero if it does not, for comparison operators whose results 10537 have a `MODE_INT' mode, `FLOAT_STORE_FLAG_VALUE' (*note Misc::) if the 10538 relation holds, or zero if it does not, for comparison operators that 10539 return floating-point values, and a vector of either 10540 `VECTOR_STORE_FLAG_VALUE' (*note Misc::) if the relation holds, or of 10541 zeros if it does not, for comparison operators that return vector 10542 results. The mode of the comparison operation is independent of the 10543 mode of the data being compared. If the comparison operation is being 10544 tested (e.g., the first operand of an `if_then_else'), the mode must be 10545 `VOIDmode'. 10546 10547 There are two ways that comparison operations may be used. The 10548 comparison operators may be used to compare the condition codes `(cc0)' 10549 against zero, as in `(eq (cc0) (const_int 0))'. Such a construct 10550 actually refers to the result of the preceding instruction in which the 10551 condition codes were set. The instruction setting the condition code 10552 must be adjacent to the instruction using the condition code; only 10553 `note' insns may separate them. 10554 10555 Alternatively, a comparison operation may directly compare two data 10556 objects. The mode of the comparison is determined by the operands; they 10557 must both be valid for a common machine mode. A comparison with both 10558 operands constant would be invalid as the machine mode could not be 10559 deduced from it, but such a comparison should never exist in RTL due to 10560 constant folding. 10561 10562 In the example above, if `(cc0)' were last set to `(compare X Y)', the 10563 comparison operation is identical to `(eq X Y)'. Usually only one style 10564 of comparisons is supported on a particular machine, but the combine 10565 pass will try to merge the operations to produce the `eq' shown in case 10566 it exists in the context of the particular insn involved. 10567 10568 Inequality comparisons come in two flavors, signed and unsigned. Thus, 10569 there are distinct expression codes `gt' and `gtu' for signed and 10570 unsigned greater-than. These can produce different results for the same 10571 pair of integer values: for example, 1 is signed greater-than -1 but not 10572 unsigned greater-than, because -1 when regarded as unsigned is actually 10573 `0xffffffff' which is greater than 1. 10574 10575 The signed comparisons are also used for floating point values. 10576 Floating point comparisons are distinguished by the machine modes of 10577 the operands. 10578 10579 `(eq:M X Y)' 10580 `STORE_FLAG_VALUE' if the values represented by X and Y are equal, 10581 otherwise 0. 10582 10583 `(ne:M X Y)' 10584 `STORE_FLAG_VALUE' if the values represented by X and Y are not 10585 equal, otherwise 0. 10586 10587 `(gt:M X Y)' 10588 `STORE_FLAG_VALUE' if the X is greater than Y. If they are 10589 fixed-point, the comparison is done in a signed sense. 10590 10591 `(gtu:M X Y)' 10592 Like `gt' but does unsigned comparison, on fixed-point numbers 10593 only. 10594 10595 `(lt:M X Y)' 10596 `(ltu:M X Y)' 10597 Like `gt' and `gtu' but test for "less than". 10598 10599 `(ge:M X Y)' 10600 `(geu:M X Y)' 10601 Like `gt' and `gtu' but test for "greater than or equal". 10602 10603 `(le:M X Y)' 10604 `(leu:M X Y)' 10605 Like `gt' and `gtu' but test for "less than or equal". 10606 10607 `(if_then_else COND THEN ELSE)' 10608 This is not a comparison operation but is listed here because it is 10609 always used in conjunction with a comparison operation. To be 10610 precise, COND is a comparison expression. This expression 10611 represents a choice, according to COND, between the value 10612 represented by THEN and the one represented by ELSE. 10613 10614 On most machines, `if_then_else' expressions are valid only to 10615 express conditional jumps. 10616 10617 `(cond [TEST1 VALUE1 TEST2 VALUE2 ...] DEFAULT)' 10618 Similar to `if_then_else', but more general. Each of TEST1, 10619 TEST2, ... is performed in turn. The result of this expression is 10620 the VALUE corresponding to the first nonzero test, or DEFAULT if 10621 none of the tests are nonzero expressions. 10622 10623 This is currently not valid for instruction patterns and is 10624 supported only for insn attributes. *Note Insn Attributes::. 10625 10626 10627 File: gccint.info, Node: Bit-Fields, Next: Vector Operations, Prev: Comparisons, Up: RTL 10628 10629 10.11 Bit-Fields 10630 ================ 10631 10632 Special expression codes exist to represent bit-field instructions. 10633 10634 `(sign_extract:M LOC SIZE POS)' 10635 This represents a reference to a sign-extended bit-field contained 10636 or starting in LOC (a memory or register reference). The bit-field 10637 is SIZE bits wide and starts at bit POS. The compilation option 10638 `BITS_BIG_ENDIAN' says which end of the memory unit POS counts 10639 from. 10640 10641 If LOC is in memory, its mode must be a single-byte integer mode. 10642 If LOC is in a register, the mode to use is specified by the 10643 operand of the `insv' or `extv' pattern (*note Standard Names::) 10644 and is usually a full-word integer mode, which is the default if 10645 none is specified. 10646 10647 The mode of POS is machine-specific and is also specified in the 10648 `insv' or `extv' pattern. 10649 10650 The mode M is the same as the mode that would be used for LOC if 10651 it were a register. 10652 10653 A `sign_extract' can not appear as an lvalue, or part thereof, in 10654 RTL. 10655 10656 `(zero_extract:M LOC SIZE POS)' 10657 Like `sign_extract' but refers to an unsigned or zero-extended 10658 bit-field. The same sequence of bits are extracted, but they are 10659 filled to an entire word with zeros instead of by sign-extension. 10660 10661 Unlike `sign_extract', this type of expressions can be lvalues in 10662 RTL; they may appear on the left side of an assignment, indicating 10663 insertion of a value into the specified bit-field. 10664 10665 10666 File: gccint.info, Node: Vector Operations, Next: Conversions, Prev: Bit-Fields, Up: RTL 10667 10668 10.12 Vector Operations 10669 ======================= 10670 10671 All normal RTL expressions can be used with vector modes; they are 10672 interpreted as operating on each part of the vector independently. 10673 Additionally, there are a few new expressions to describe specific 10674 vector operations. 10675 10676 `(vec_merge:M VEC1 VEC2 ITEMS)' 10677 This describes a merge operation between two vectors. The result 10678 is a vector of mode M; its elements are selected from either VEC1 10679 or VEC2. Which elements are selected is described by ITEMS, which 10680 is a bit mask represented by a `const_int'; a zero bit indicates 10681 the corresponding element in the result vector is taken from VEC2 10682 while a set bit indicates it is taken from VEC1. 10683 10684 `(vec_select:M VEC1 SELECTION)' 10685 This describes an operation that selects parts of a vector. VEC1 10686 is the source vector, SELECTION is a `parallel' that contains a 10687 `const_int' for each of the subparts of the result vector, giving 10688 the number of the source subpart that should be stored into it. 10689 10690 `(vec_concat:M VEC1 VEC2)' 10691 Describes a vector concat operation. The result is a 10692 concatenation of the vectors VEC1 and VEC2; its length is the sum 10693 of the lengths of the two inputs. 10694 10695 `(vec_duplicate:M VEC)' 10696 This operation converts a small vector into a larger one by 10697 duplicating the input values. The output vector mode must have 10698 the same submodes as the input vector mode, and the number of 10699 output parts must be an integer multiple of the number of input 10700 parts. 10701 10702 10703 10704 File: gccint.info, Node: Conversions, Next: RTL Declarations, Prev: Vector Operations, Up: RTL 10705 10706 10.13 Conversions 10707 ================= 10708 10709 All conversions between machine modes must be represented by explicit 10710 conversion operations. For example, an expression which is the sum of 10711 a byte and a full word cannot be written as `(plus:SI (reg:QI 34) 10712 (reg:SI 80))' because the `plus' operation requires two operands of the 10713 same machine mode. Therefore, the byte-sized operand is enclosed in a 10714 conversion operation, as in 10715 10716 (plus:SI (sign_extend:SI (reg:QI 34)) (reg:SI 80)) 10717 10718 The conversion operation is not a mere placeholder, because there may 10719 be more than one way of converting from a given starting mode to the 10720 desired final mode. The conversion operation code says how to do it. 10721 10722 For all conversion operations, X must not be `VOIDmode' because the 10723 mode in which to do the conversion would not be known. The conversion 10724 must either be done at compile-time or X must be placed into a register. 10725 10726 `(sign_extend:M X)' 10727 Represents the result of sign-extending the value X to machine 10728 mode M. M must be a fixed-point mode and X a fixed-point value of 10729 a mode narrower than M. 10730 10731 `(zero_extend:M X)' 10732 Represents the result of zero-extending the value X to machine 10733 mode M. M must be a fixed-point mode and X a fixed-point value of 10734 a mode narrower than M. 10735 10736 `(float_extend:M X)' 10737 Represents the result of extending the value X to machine mode M. 10738 M must be a floating point mode and X a floating point value of a 10739 mode narrower than M. 10740 10741 `(truncate:M X)' 10742 Represents the result of truncating the value X to machine mode M. 10743 M must be a fixed-point mode and X a fixed-point value of a mode 10744 wider than M. 10745 10746 `(ss_truncate:M X)' 10747 Represents the result of truncating the value X to machine mode M, 10748 using signed saturation in the case of overflow. Both M and the 10749 mode of X must be fixed-point modes. 10750 10751 `(us_truncate:M X)' 10752 Represents the result of truncating the value X to machine mode M, 10753 using unsigned saturation in the case of overflow. Both M and the 10754 mode of X must be fixed-point modes. 10755 10756 `(float_truncate:M X)' 10757 Represents the result of truncating the value X to machine mode M. 10758 M must be a floating point mode and X a floating point value of a 10759 mode wider than M. 10760 10761 `(float:M X)' 10762 Represents the result of converting fixed point value X, regarded 10763 as signed, to floating point mode M. 10764 10765 `(unsigned_float:M X)' 10766 Represents the result of converting fixed point value X, regarded 10767 as unsigned, to floating point mode M. 10768 10769 `(fix:M X)' 10770 When M is a floating-point mode, represents the result of 10771 converting floating point value X (valid for mode M) to an 10772 integer, still represented in floating point mode M, by rounding 10773 towards zero. 10774 10775 When M is a fixed-point mode, represents the result of converting 10776 floating point value X to mode M, regarded as signed. How 10777 rounding is done is not specified, so this operation may be used 10778 validly in compiling C code only for integer-valued operands. 10779 10780 `(unsigned_fix:M X)' 10781 Represents the result of converting floating point value X to 10782 fixed point mode M, regarded as unsigned. How rounding is done is 10783 not specified. 10784 10785 `(fract_convert:M X)' 10786 Represents the result of converting fixed-point value X to 10787 fixed-point mode M, signed integer value X to fixed-point mode M, 10788 floating-point value X to fixed-point mode M, fixed-point value X 10789 to integer mode M regarded as signed, or fixed-point value X to 10790 floating-point mode M. When overflows or underflows happen, the 10791 results are undefined. 10792 10793 `(sat_fract:M X)' 10794 Represents the result of converting fixed-point value X to 10795 fixed-point mode M, signed integer value X to fixed-point mode M, 10796 or floating-point value X to fixed-point mode M. When overflows 10797 or underflows happen, the results are saturated to the maximum or 10798 the minimum. 10799 10800 `(unsigned_fract_convert:M X)' 10801 Represents the result of converting fixed-point value X to integer 10802 mode M regarded as unsigned, or unsigned integer value X to 10803 fixed-point mode M. When overflows or underflows happen, the 10804 results are undefined. 10805 10806 `(unsigned_sat_fract:M X)' 10807 Represents the result of converting unsigned integer value X to 10808 fixed-point mode M. When overflows or underflows happen, the 10809 results are saturated to the maximum or the minimum. 10810 10811 10812 File: gccint.info, Node: RTL Declarations, Next: Side Effects, Prev: Conversions, Up: RTL 10813 10814 10.14 Declarations 10815 ================== 10816 10817 Declaration expression codes do not represent arithmetic operations but 10818 rather state assertions about their operands. 10819 10820 `(strict_low_part (subreg:M (reg:N R) 0))' 10821 This expression code is used in only one context: as the 10822 destination operand of a `set' expression. In addition, the 10823 operand of this expression must be a non-paradoxical `subreg' 10824 expression. 10825 10826 The presence of `strict_low_part' says that the part of the 10827 register which is meaningful in mode N, but is not part of mode M, 10828 is not to be altered. Normally, an assignment to such a subreg is 10829 allowed to have undefined effects on the rest of the register when 10830 M is less than a word. 10831 10832 10833 File: gccint.info, Node: Side Effects, Next: Incdec, Prev: RTL Declarations, Up: RTL 10834 10835 10.15 Side Effect Expressions 10836 ============================= 10837 10838 The expression codes described so far represent values, not actions. 10839 But machine instructions never produce values; they are meaningful only 10840 for their side effects on the state of the machine. Special expression 10841 codes are used to represent side effects. 10842 10843 The body of an instruction is always one of these side effect codes; 10844 the codes described above, which represent values, appear only as the 10845 operands of these. 10846 10847 `(set LVAL X)' 10848 Represents the action of storing the value of X into the place 10849 represented by LVAL. LVAL must be an expression representing a 10850 place that can be stored in: `reg' (or `subreg', `strict_low_part' 10851 or `zero_extract'), `mem', `pc', `parallel', or `cc0'. 10852 10853 If LVAL is a `reg', `subreg' or `mem', it has a machine mode; then 10854 X must be valid for that mode. 10855 10856 If LVAL is a `reg' whose machine mode is less than the full width 10857 of the register, then it means that the part of the register 10858 specified by the machine mode is given the specified value and the 10859 rest of the register receives an undefined value. Likewise, if 10860 LVAL is a `subreg' whose machine mode is narrower than the mode of 10861 the register, the rest of the register can be changed in an 10862 undefined way. 10863 10864 If LVAL is a `strict_low_part' of a subreg, then the part of the 10865 register specified by the machine mode of the `subreg' is given 10866 the value X and the rest of the register is not changed. 10867 10868 If LVAL is a `zero_extract', then the referenced part of the 10869 bit-field (a memory or register reference) specified by the 10870 `zero_extract' is given the value X and the rest of the bit-field 10871 is not changed. Note that `sign_extract' can not appear in LVAL. 10872 10873 If LVAL is `(cc0)', it has no machine mode, and X may be either a 10874 `compare' expression or a value that may have any mode. The 10875 latter case represents a "test" instruction. The expression `(set 10876 (cc0) (reg:M N))' is equivalent to `(set (cc0) (compare (reg:M N) 10877 (const_int 0)))'. Use the former expression to save space during 10878 the compilation. 10879 10880 If LVAL is a `parallel', it is used to represent the case of a 10881 function returning a structure in multiple registers. Each element 10882 of the `parallel' is an `expr_list' whose first operand is a `reg' 10883 and whose second operand is a `const_int' representing the offset 10884 (in bytes) into the structure at which the data in that register 10885 corresponds. The first element may be null to indicate that the 10886 structure is also passed partly in memory. 10887 10888 If LVAL is `(pc)', we have a jump instruction, and the 10889 possibilities for X are very limited. It may be a `label_ref' 10890 expression (unconditional jump). It may be an `if_then_else' 10891 (conditional jump), in which case either the second or the third 10892 operand must be `(pc)' (for the case which does not jump) and the 10893 other of the two must be a `label_ref' (for the case which does 10894 jump). X may also be a `mem' or `(plus:SI (pc) Y)', where Y may 10895 be a `reg' or a `mem'; these unusual patterns are used to 10896 represent jumps through branch tables. 10897 10898 If LVAL is neither `(cc0)' nor `(pc)', the mode of LVAL must not 10899 be `VOIDmode' and the mode of X must be valid for the mode of LVAL. 10900 10901 LVAL is customarily accessed with the `SET_DEST' macro and X with 10902 the `SET_SRC' macro. 10903 10904 `(return)' 10905 As the sole expression in a pattern, represents a return from the 10906 current function, on machines where this can be done with one 10907 instruction, such as VAXen. On machines where a multi-instruction 10908 "epilogue" must be executed in order to return from the function, 10909 returning is done by jumping to a label which precedes the 10910 epilogue, and the `return' expression code is never used. 10911 10912 Inside an `if_then_else' expression, represents the value to be 10913 placed in `pc' to return to the caller. 10914 10915 Note that an insn pattern of `(return)' is logically equivalent to 10916 `(set (pc) (return))', but the latter form is never used. 10917 10918 `(call FUNCTION NARGS)' 10919 Represents a function call. FUNCTION is a `mem' expression whose 10920 address is the address of the function to be called. NARGS is an 10921 expression which can be used for two purposes: on some machines it 10922 represents the number of bytes of stack argument; on others, it 10923 represents the number of argument registers. 10924 10925 Each machine has a standard machine mode which FUNCTION must have. 10926 The machine description defines macro `FUNCTION_MODE' to expand 10927 into the requisite mode name. The purpose of this mode is to 10928 specify what kind of addressing is allowed, on machines where the 10929 allowed kinds of addressing depend on the machine mode being 10930 addressed. 10931 10932 `(clobber X)' 10933 Represents the storing or possible storing of an unpredictable, 10934 undescribed value into X, which must be a `reg', `scratch', 10935 `parallel' or `mem' expression. 10936 10937 One place this is used is in string instructions that store 10938 standard values into particular hard registers. It may not be 10939 worth the trouble to describe the values that are stored, but it 10940 is essential to inform the compiler that the registers will be 10941 altered, lest it attempt to keep data in them across the string 10942 instruction. 10943 10944 If X is `(mem:BLK (const_int 0))' or `(mem:BLK (scratch))', it 10945 means that all memory locations must be presumed clobbered. If X 10946 is a `parallel', it has the same meaning as a `parallel' in a 10947 `set' expression. 10948 10949 Note that the machine description classifies certain hard 10950 registers as "call-clobbered". All function call instructions are 10951 assumed by default to clobber these registers, so there is no need 10952 to use `clobber' expressions to indicate this fact. Also, each 10953 function call is assumed to have the potential to alter any memory 10954 location, unless the function is declared `const'. 10955 10956 If the last group of expressions in a `parallel' are each a 10957 `clobber' expression whose arguments are `reg' or `match_scratch' 10958 (*note RTL Template::) expressions, the combiner phase can add the 10959 appropriate `clobber' expressions to an insn it has constructed 10960 when doing so will cause a pattern to be matched. 10961 10962 This feature can be used, for example, on a machine that whose 10963 multiply and add instructions don't use an MQ register but which 10964 has an add-accumulate instruction that does clobber the MQ 10965 register. Similarly, a combined instruction might require a 10966 temporary register while the constituent instructions might not. 10967 10968 When a `clobber' expression for a register appears inside a 10969 `parallel' with other side effects, the register allocator 10970 guarantees that the register is unoccupied both before and after 10971 that insn if it is a hard register clobber. For pseudo-register 10972 clobber, the register allocator and the reload pass do not assign 10973 the same hard register to the clobber and the input operands if 10974 there is an insn alternative containing the `&' constraint (*note 10975 Modifiers::) for the clobber and the hard register is in register 10976 classes of the clobber in the alternative. You can clobber either 10977 a specific hard register, a pseudo register, or a `scratch' 10978 expression; in the latter two cases, GCC will allocate a hard 10979 register that is available there for use as a temporary. 10980 10981 For instructions that require a temporary register, you should use 10982 `scratch' instead of a pseudo-register because this will allow the 10983 combiner phase to add the `clobber' when required. You do this by 10984 coding (`clobber' (`match_scratch' ...)). If you do clobber a 10985 pseudo register, use one which appears nowhere else--generate a 10986 new one each time. Otherwise, you may confuse CSE. 10987 10988 There is one other known use for clobbering a pseudo register in a 10989 `parallel': when one of the input operands of the insn is also 10990 clobbered by the insn. In this case, using the same pseudo 10991 register in the clobber and elsewhere in the insn produces the 10992 expected results. 10993 10994 `(use X)' 10995 Represents the use of the value of X. It indicates that the value 10996 in X at this point in the program is needed, even though it may 10997 not be apparent why this is so. Therefore, the compiler will not 10998 attempt to delete previous instructions whose only effect is to 10999 store a value in X. X must be a `reg' expression. 11000 11001 In some situations, it may be tempting to add a `use' of a 11002 register in a `parallel' to describe a situation where the value 11003 of a special register will modify the behavior of the instruction. 11004 An hypothetical example might be a pattern for an addition that can 11005 either wrap around or use saturating addition depending on the 11006 value of a special control register: 11007 11008 (parallel [(set (reg:SI 2) (unspec:SI [(reg:SI 3) 11009 (reg:SI 4)] 0)) 11010 (use (reg:SI 1))]) 11011 11012 This will not work, several of the optimizers only look at 11013 expressions locally; it is very likely that if you have multiple 11014 insns with identical inputs to the `unspec', they will be 11015 optimized away even if register 1 changes in between. 11016 11017 This means that `use' can _only_ be used to describe that the 11018 register is live. You should think twice before adding `use' 11019 statements, more often you will want to use `unspec' instead. The 11020 `use' RTX is most commonly useful to describe that a fixed 11021 register is implicitly used in an insn. It is also safe to use in 11022 patterns where the compiler knows for other reasons that the result 11023 of the whole pattern is variable, such as `movmemM' or `call' 11024 patterns. 11025 11026 During the reload phase, an insn that has a `use' as pattern can 11027 carry a reg_equal note. These `use' insns will be deleted before 11028 the reload phase exits. 11029 11030 During the delayed branch scheduling phase, X may be an insn. 11031 This indicates that X previously was located at this place in the 11032 code and its data dependencies need to be taken into account. 11033 These `use' insns will be deleted before the delayed branch 11034 scheduling phase exits. 11035 11036 `(parallel [X0 X1 ...])' 11037 Represents several side effects performed in parallel. The square 11038 brackets stand for a vector; the operand of `parallel' is a vector 11039 of expressions. X0, X1 and so on are individual side effect 11040 expressions--expressions of code `set', `call', `return', 11041 `clobber' or `use'. 11042 11043 "In parallel" means that first all the values used in the 11044 individual side-effects are computed, and second all the actual 11045 side-effects are performed. For example, 11046 11047 (parallel [(set (reg:SI 1) (mem:SI (reg:SI 1))) 11048 (set (mem:SI (reg:SI 1)) (reg:SI 1))]) 11049 11050 says unambiguously that the values of hard register 1 and the 11051 memory location addressed by it are interchanged. In both places 11052 where `(reg:SI 1)' appears as a memory address it refers to the 11053 value in register 1 _before_ the execution of the insn. 11054 11055 It follows that it is _incorrect_ to use `parallel' and expect the 11056 result of one `set' to be available for the next one. For 11057 example, people sometimes attempt to represent a jump-if-zero 11058 instruction this way: 11059 11060 (parallel [(set (cc0) (reg:SI 34)) 11061 (set (pc) (if_then_else 11062 (eq (cc0) (const_int 0)) 11063 (label_ref ...) 11064 (pc)))]) 11065 11066 But this is incorrect, because it says that the jump condition 11067 depends on the condition code value _before_ this instruction, not 11068 on the new value that is set by this instruction. 11069 11070 Peephole optimization, which takes place together with final 11071 assembly code output, can produce insns whose patterns consist of 11072 a `parallel' whose elements are the operands needed to output the 11073 resulting assembler code--often `reg', `mem' or constant 11074 expressions. This would not be well-formed RTL at any other stage 11075 in compilation, but it is ok then because no further optimization 11076 remains to be done. However, the definition of the macro 11077 `NOTICE_UPDATE_CC', if any, must deal with such insns if you 11078 define any peephole optimizations. 11079 11080 `(cond_exec [COND EXPR])' 11081 Represents a conditionally executed expression. The EXPR is 11082 executed only if the COND is nonzero. The COND expression must 11083 not have side-effects, but the EXPR may very well have 11084 side-effects. 11085 11086 `(sequence [INSNS ...])' 11087 Represents a sequence of insns. Each of the INSNS that appears in 11088 the vector is suitable for appearing in the chain of insns, so it 11089 must be an `insn', `jump_insn', `call_insn', `code_label', 11090 `barrier' or `note'. 11091 11092 A `sequence' RTX is never placed in an actual insn during RTL 11093 generation. It represents the sequence of insns that result from a 11094 `define_expand' _before_ those insns are passed to `emit_insn' to 11095 insert them in the chain of insns. When actually inserted, the 11096 individual sub-insns are separated out and the `sequence' is 11097 forgotten. 11098 11099 After delay-slot scheduling is completed, an insn and all the 11100 insns that reside in its delay slots are grouped together into a 11101 `sequence'. The insn requiring the delay slot is the first insn 11102 in the vector; subsequent insns are to be placed in the delay slot. 11103 11104 `INSN_ANNULLED_BRANCH_P' is set on an insn in a delay slot to 11105 indicate that a branch insn should be used that will conditionally 11106 annul the effect of the insns in the delay slots. In such a case, 11107 `INSN_FROM_TARGET_P' indicates that the insn is from the target of 11108 the branch and should be executed only if the branch is taken; 11109 otherwise the insn should be executed only if the branch is not 11110 taken. *Note Delay Slots::. 11111 11112 These expression codes appear in place of a side effect, as the body of 11113 an insn, though strictly speaking they do not always describe side 11114 effects as such: 11115 11116 `(asm_input S)' 11117 Represents literal assembler code as described by the string S. 11118 11119 `(unspec [OPERANDS ...] INDEX)' 11120 `(unspec_volatile [OPERANDS ...] INDEX)' 11121 Represents a machine-specific operation on OPERANDS. INDEX 11122 selects between multiple machine-specific operations. 11123 `unspec_volatile' is used for volatile operations and operations 11124 that may trap; `unspec' is used for other operations. 11125 11126 These codes may appear inside a `pattern' of an insn, inside a 11127 `parallel', or inside an expression. 11128 11129 `(addr_vec:M [LR0 LR1 ...])' 11130 Represents a table of jump addresses. The vector elements LR0, 11131 etc., are `label_ref' expressions. The mode M specifies how much 11132 space is given to each address; normally M would be `Pmode'. 11133 11134 `(addr_diff_vec:M BASE [LR0 LR1 ...] MIN MAX FLAGS)' 11135 Represents a table of jump addresses expressed as offsets from 11136 BASE. The vector elements LR0, etc., are `label_ref' expressions 11137 and so is BASE. The mode M specifies how much space is given to 11138 each address-difference. MIN and MAX are set up by branch 11139 shortening and hold a label with a minimum and a maximum address, 11140 respectively. FLAGS indicates the relative position of BASE, MIN 11141 and MAX to the containing insn and of MIN and MAX to BASE. See 11142 rtl.def for details. 11143 11144 `(prefetch:M ADDR RW LOCALITY)' 11145 Represents prefetch of memory at address ADDR. Operand RW is 1 if 11146 the prefetch is for data to be written, 0 otherwise; targets that 11147 do not support write prefetches should treat this as a normal 11148 prefetch. Operand LOCALITY specifies the amount of temporal 11149 locality; 0 if there is none or 1, 2, or 3 for increasing levels 11150 of temporal locality; targets that do not support locality hints 11151 should ignore this. 11152 11153 This insn is used to minimize cache-miss latency by moving data 11154 into a cache before it is accessed. It should use only 11155 non-faulting data prefetch instructions. 11156 11157 11158 File: gccint.info, Node: Incdec, Next: Assembler, Prev: Side Effects, Up: RTL 11159 11160 10.16 Embedded Side-Effects on Addresses 11161 ======================================== 11162 11163 Six special side-effect expression codes appear as memory addresses. 11164 11165 `(pre_dec:M X)' 11166 Represents the side effect of decrementing X by a standard amount 11167 and represents also the value that X has after being decremented. 11168 X must be a `reg' or `mem', but most machines allow only a `reg'. 11169 M must be the machine mode for pointers on the machine in use. 11170 The amount X is decremented by is the length in bytes of the 11171 machine mode of the containing memory reference of which this 11172 expression serves as the address. Here is an example of its use: 11173 11174 (mem:DF (pre_dec:SI (reg:SI 39))) 11175 11176 This says to decrement pseudo register 39 by the length of a 11177 `DFmode' value and use the result to address a `DFmode' value. 11178 11179 `(pre_inc:M X)' 11180 Similar, but specifies incrementing X instead of decrementing it. 11181 11182 `(post_dec:M X)' 11183 Represents the same side effect as `pre_dec' but a different 11184 value. The value represented here is the value X has before being 11185 decremented. 11186 11187 `(post_inc:M X)' 11188 Similar, but specifies incrementing X instead of decrementing it. 11189 11190 `(post_modify:M X Y)' 11191 Represents the side effect of setting X to Y and represents X 11192 before X is modified. X must be a `reg' or `mem', but most 11193 machines allow only a `reg'. M must be the machine mode for 11194 pointers on the machine in use. 11195 11196 The expression Y must be one of three forms: `(plus:M X Z)', 11197 `(minus:M X Z)', or `(plus:M X I)', where Z is an index register 11198 and I is a constant. 11199 11200 Here is an example of its use: 11201 11202 (mem:SF (post_modify:SI (reg:SI 42) (plus (reg:SI 42) 11203 (reg:SI 48)))) 11204 11205 This says to modify pseudo register 42 by adding the contents of 11206 pseudo register 48 to it, after the use of what ever 42 points to. 11207 11208 `(pre_modify:M X EXPR)' 11209 Similar except side effects happen before the use. 11210 11211 These embedded side effect expressions must be used with care. 11212 Instruction patterns may not use them. Until the `flow' pass of the 11213 compiler, they may occur only to represent pushes onto the stack. The 11214 `flow' pass finds cases where registers are incremented or decremented 11215 in one instruction and used as an address shortly before or after; 11216 these cases are then transformed to use pre- or post-increment or 11217 -decrement. 11218 11219 If a register used as the operand of these expressions is used in 11220 another address in an insn, the original value of the register is used. 11221 Uses of the register outside of an address are not permitted within the 11222 same insn as a use in an embedded side effect expression because such 11223 insns behave differently on different machines and hence must be treated 11224 as ambiguous and disallowed. 11225 11226 An instruction that can be represented with an embedded side effect 11227 could also be represented using `parallel' containing an additional 11228 `set' to describe how the address register is altered. This is not 11229 done because machines that allow these operations at all typically 11230 allow them wherever a memory address is called for. Describing them as 11231 additional parallel stores would require doubling the number of entries 11232 in the machine description. 11233 11234 11235 File: gccint.info, Node: Assembler, Next: Insns, Prev: Incdec, Up: RTL 11236 11237 10.17 Assembler Instructions as Expressions 11238 =========================================== 11239 11240 The RTX code `asm_operands' represents a value produced by a 11241 user-specified assembler instruction. It is used to represent an `asm' 11242 statement with arguments. An `asm' statement with a single output 11243 operand, like this: 11244 11245 asm ("foo %1,%2,%0" : "=a" (outputvar) : "g" (x + y), "di" (*z)); 11246 11247 is represented using a single `asm_operands' RTX which represents the 11248 value that is stored in `outputvar': 11249 11250 (set RTX-FOR-OUTPUTVAR 11251 (asm_operands "foo %1,%2,%0" "a" 0 11252 [RTX-FOR-ADDITION-RESULT RTX-FOR-*Z] 11253 [(asm_input:M1 "g") 11254 (asm_input:M2 "di")])) 11255 11256 Here the operands of the `asm_operands' RTX are the assembler template 11257 string, the output-operand's constraint, the index-number of the output 11258 operand among the output operands specified, a vector of input operand 11259 RTX's, and a vector of input-operand modes and constraints. The mode 11260 M1 is the mode of the sum `x+y'; M2 is that of `*z'. 11261 11262 When an `asm' statement has multiple output values, its insn has 11263 several such `set' RTX's inside of a `parallel'. Each `set' contains a 11264 `asm_operands'; all of these share the same assembler template and 11265 vectors, but each contains the constraint for the respective output 11266 operand. They are also distinguished by the output-operand index 11267 number, which is 0, 1, ... for successive output operands. 11268 11269 11270 File: gccint.info, Node: Insns, Next: Calls, Prev: Assembler, Up: RTL 11271 11272 10.18 Insns 11273 =========== 11274 11275 The RTL representation of the code for a function is a doubly-linked 11276 chain of objects called "insns". Insns are expressions with special 11277 codes that are used for no other purpose. Some insns are actual 11278 instructions; others represent dispatch tables for `switch' statements; 11279 others represent labels to jump to or various sorts of declarative 11280 information. 11281 11282 In addition to its own specific data, each insn must have a unique 11283 id-number that distinguishes it from all other insns in the current 11284 function (after delayed branch scheduling, copies of an insn with the 11285 same id-number may be present in multiple places in a function, but 11286 these copies will always be identical and will only appear inside a 11287 `sequence'), and chain pointers to the preceding and following insns. 11288 These three fields occupy the same position in every insn, independent 11289 of the expression code of the insn. They could be accessed with `XEXP' 11290 and `XINT', but instead three special macros are always used: 11291 11292 `INSN_UID (I)' 11293 Accesses the unique id of insn I. 11294 11295 `PREV_INSN (I)' 11296 Accesses the chain pointer to the insn preceding I. If I is the 11297 first insn, this is a null pointer. 11298 11299 `NEXT_INSN (I)' 11300 Accesses the chain pointer to the insn following I. If I is the 11301 last insn, this is a null pointer. 11302 11303 The first insn in the chain is obtained by calling `get_insns'; the 11304 last insn is the result of calling `get_last_insn'. Within the chain 11305 delimited by these insns, the `NEXT_INSN' and `PREV_INSN' pointers must 11306 always correspond: if INSN is not the first insn, 11307 11308 NEXT_INSN (PREV_INSN (INSN)) == INSN 11309 11310 is always true and if INSN is not the last insn, 11311 11312 PREV_INSN (NEXT_INSN (INSN)) == INSN 11313 11314 is always true. 11315 11316 After delay slot scheduling, some of the insns in the chain might be 11317 `sequence' expressions, which contain a vector of insns. The value of 11318 `NEXT_INSN' in all but the last of these insns is the next insn in the 11319 vector; the value of `NEXT_INSN' of the last insn in the vector is the 11320 same as the value of `NEXT_INSN' for the `sequence' in which it is 11321 contained. Similar rules apply for `PREV_INSN'. 11322 11323 This means that the above invariants are not necessarily true for insns 11324 inside `sequence' expressions. Specifically, if INSN is the first insn 11325 in a `sequence', `NEXT_INSN (PREV_INSN (INSN))' is the insn containing 11326 the `sequence' expression, as is the value of `PREV_INSN (NEXT_INSN 11327 (INSN))' if INSN is the last insn in the `sequence' expression. You 11328 can use these expressions to find the containing `sequence' expression. 11329 11330 Every insn has one of the following six expression codes: 11331 11332 `insn' 11333 The expression code `insn' is used for instructions that do not 11334 jump and do not do function calls. `sequence' expressions are 11335 always contained in insns with code `insn' even if one of those 11336 insns should jump or do function calls. 11337 11338 Insns with code `insn' have four additional fields beyond the three 11339 mandatory ones listed above. These four are described in a table 11340 below. 11341 11342 `jump_insn' 11343 The expression code `jump_insn' is used for instructions that may 11344 jump (or, more generally, may contain `label_ref' expressions to 11345 which `pc' can be set in that instruction). If there is an 11346 instruction to return from the current function, it is recorded as 11347 a `jump_insn'. 11348 11349 `jump_insn' insns have the same extra fields as `insn' insns, 11350 accessed in the same way and in addition contain a field 11351 `JUMP_LABEL' which is defined once jump optimization has completed. 11352 11353 For simple conditional and unconditional jumps, this field contains 11354 the `code_label' to which this insn will (possibly conditionally) 11355 branch. In a more complex jump, `JUMP_LABEL' records one of the 11356 labels that the insn refers to; other jump target labels are 11357 recorded as `REG_LABEL_TARGET' notes. The exception is `addr_vec' 11358 and `addr_diff_vec', where `JUMP_LABEL' is `NULL_RTX' and the only 11359 way to find the labels is to scan the entire body of the insn. 11360 11361 Return insns count as jumps, but since they do not refer to any 11362 labels, their `JUMP_LABEL' is `NULL_RTX'. 11363 11364 `call_insn' 11365 The expression code `call_insn' is used for instructions that may 11366 do function calls. It is important to distinguish these 11367 instructions because they imply that certain registers and memory 11368 locations may be altered unpredictably. 11369 11370 `call_insn' insns have the same extra fields as `insn' insns, 11371 accessed in the same way and in addition contain a field 11372 `CALL_INSN_FUNCTION_USAGE', which contains a list (chain of 11373 `expr_list' expressions) containing `use' and `clobber' 11374 expressions that denote hard registers and `MEM's used or 11375 clobbered by the called function. 11376 11377 A `MEM' generally points to a stack slots in which arguments passed 11378 to the libcall by reference (*note TARGET_PASS_BY_REFERENCE: 11379 Register Arguments.) are stored. If the argument is caller-copied 11380 (*note TARGET_CALLEE_COPIES: Register Arguments.), the stack slot 11381 will be mentioned in `CLOBBER' and `USE' entries; if it's 11382 callee-copied, only a `USE' will appear, and the `MEM' may point 11383 to addresses that are not stack slots. 11384 11385 `CLOBBER'ed registers in this list augment registers specified in 11386 `CALL_USED_REGISTERS' (*note Register Basics::). 11387 11388 `code_label' 11389 A `code_label' insn represents a label that a jump insn can jump 11390 to. It contains two special fields of data in addition to the 11391 three standard ones. `CODE_LABEL_NUMBER' is used to hold the 11392 "label number", a number that identifies this label uniquely among 11393 all the labels in the compilation (not just in the current 11394 function). Ultimately, the label is represented in the assembler 11395 output as an assembler label, usually of the form `LN' where N is 11396 the label number. 11397 11398 When a `code_label' appears in an RTL expression, it normally 11399 appears within a `label_ref' which represents the address of the 11400 label, as a number. 11401 11402 Besides as a `code_label', a label can also be represented as a 11403 `note' of type `NOTE_INSN_DELETED_LABEL'. 11404 11405 The field `LABEL_NUSES' is only defined once the jump optimization 11406 phase is completed. It contains the number of times this label is 11407 referenced in the current function. 11408 11409 The field `LABEL_KIND' differentiates four different types of 11410 labels: `LABEL_NORMAL', `LABEL_STATIC_ENTRY', 11411 `LABEL_GLOBAL_ENTRY', and `LABEL_WEAK_ENTRY'. The only labels 11412 that do not have type `LABEL_NORMAL' are "alternate entry points" 11413 to the current function. These may be static (visible only in the 11414 containing translation unit), global (exposed to all translation 11415 units), or weak (global, but can be overridden by another symbol 11416 with the same name). 11417 11418 Much of the compiler treats all four kinds of label identically. 11419 Some of it needs to know whether or not a label is an alternate 11420 entry point; for this purpose, the macro `LABEL_ALT_ENTRY_P' is 11421 provided. It is equivalent to testing whether `LABEL_KIND (label) 11422 == LABEL_NORMAL'. The only place that cares about the distinction 11423 between static, global, and weak alternate entry points, besides 11424 the front-end code that creates them, is the function 11425 `output_alternate_entry_point', in `final.c'. 11426 11427 To set the kind of a label, use the `SET_LABEL_KIND' macro. 11428 11429 `barrier' 11430 Barriers are placed in the instruction stream when control cannot 11431 flow past them. They are placed after unconditional jump 11432 instructions to indicate that the jumps are unconditional and 11433 after calls to `volatile' functions, which do not return (e.g., 11434 `exit'). They contain no information beyond the three standard 11435 fields. 11436 11437 `note' 11438 `note' insns are used to represent additional debugging and 11439 declarative information. They contain two nonstandard fields, an 11440 integer which is accessed with the macro `NOTE_LINE_NUMBER' and a 11441 string accessed with `NOTE_SOURCE_FILE'. 11442 11443 If `NOTE_LINE_NUMBER' is positive, the note represents the 11444 position of a source line and `NOTE_SOURCE_FILE' is the source 11445 file name that the line came from. These notes control generation 11446 of line number data in the assembler output. 11447 11448 Otherwise, `NOTE_LINE_NUMBER' is not really a line number but a 11449 code with one of the following values (and `NOTE_SOURCE_FILE' must 11450 contain a null pointer): 11451 11452 `NOTE_INSN_DELETED' 11453 Such a note is completely ignorable. Some passes of the 11454 compiler delete insns by altering them into notes of this 11455 kind. 11456 11457 `NOTE_INSN_DELETED_LABEL' 11458 This marks what used to be a `code_label', but was not used 11459 for other purposes than taking its address and was 11460 transformed to mark that no code jumps to it. 11461 11462 `NOTE_INSN_BLOCK_BEG' 11463 `NOTE_INSN_BLOCK_END' 11464 These types of notes indicate the position of the beginning 11465 and end of a level of scoping of variable names. They 11466 control the output of debugging information. 11467 11468 `NOTE_INSN_EH_REGION_BEG' 11469 `NOTE_INSN_EH_REGION_END' 11470 These types of notes indicate the position of the beginning 11471 and end of a level of scoping for exception handling. 11472 `NOTE_BLOCK_NUMBER' identifies which `CODE_LABEL' or `note' 11473 of type `NOTE_INSN_DELETED_LABEL' is associated with the 11474 given region. 11475 11476 `NOTE_INSN_LOOP_BEG' 11477 `NOTE_INSN_LOOP_END' 11478 These types of notes indicate the position of the beginning 11479 and end of a `while' or `for' loop. They enable the loop 11480 optimizer to find loops quickly. 11481 11482 `NOTE_INSN_LOOP_CONT' 11483 Appears at the place in a loop that `continue' statements 11484 jump to. 11485 11486 `NOTE_INSN_LOOP_VTOP' 11487 This note indicates the place in a loop where the exit test 11488 begins for those loops in which the exit test has been 11489 duplicated. This position becomes another virtual start of 11490 the loop when considering loop invariants. 11491 11492 `NOTE_INSN_FUNCTION_BEG' 11493 Appears at the start of the function body, after the function 11494 prologue. 11495 11496 11497 These codes are printed symbolically when they appear in debugging 11498 dumps. 11499 11500 The machine mode of an insn is normally `VOIDmode', but some phases 11501 use the mode for various purposes. 11502 11503 The common subexpression elimination pass sets the mode of an insn to 11504 `QImode' when it is the first insn in a block that has already been 11505 processed. 11506 11507 The second Haifa scheduling pass, for targets that can multiple issue, 11508 sets the mode of an insn to `TImode' when it is believed that the 11509 instruction begins an issue group. That is, when the instruction 11510 cannot issue simultaneously with the previous. This may be relied on 11511 by later passes, in particular machine-dependent reorg. 11512 11513 Here is a table of the extra fields of `insn', `jump_insn' and 11514 `call_insn' insns: 11515 11516 `PATTERN (I)' 11517 An expression for the side effect performed by this insn. This 11518 must be one of the following codes: `set', `call', `use', 11519 `clobber', `return', `asm_input', `asm_output', `addr_vec', 11520 `addr_diff_vec', `trap_if', `unspec', `unspec_volatile', 11521 `parallel', `cond_exec', or `sequence'. If it is a `parallel', 11522 each element of the `parallel' must be one these codes, except that 11523 `parallel' expressions cannot be nested and `addr_vec' and 11524 `addr_diff_vec' are not permitted inside a `parallel' expression. 11525 11526 `INSN_CODE (I)' 11527 An integer that says which pattern in the machine description 11528 matches this insn, or -1 if the matching has not yet been 11529 attempted. 11530 11531 Such matching is never attempted and this field remains -1 on an 11532 insn whose pattern consists of a single `use', `clobber', 11533 `asm_input', `addr_vec' or `addr_diff_vec' expression. 11534 11535 Matching is also never attempted on insns that result from an `asm' 11536 statement. These contain at least one `asm_operands' expression. 11537 The function `asm_noperands' returns a non-negative value for such 11538 insns. 11539 11540 In the debugging output, this field is printed as a number 11541 followed by a symbolic representation that locates the pattern in 11542 the `md' file as some small positive or negative offset from a 11543 named pattern. 11544 11545 `LOG_LINKS (I)' 11546 A list (chain of `insn_list' expressions) giving information about 11547 dependencies between instructions within a basic block. Neither a 11548 jump nor a label may come between the related insns. These are 11549 only used by the schedulers and by combine. This is a deprecated 11550 data structure. Def-use and use-def chains are now preferred. 11551 11552 `REG_NOTES (I)' 11553 A list (chain of `expr_list' and `insn_list' expressions) giving 11554 miscellaneous information about the insn. It is often information 11555 pertaining to the registers used in this insn. 11556 11557 The `LOG_LINKS' field of an insn is a chain of `insn_list' 11558 expressions. Each of these has two operands: the first is an insn, and 11559 the second is another `insn_list' expression (the next one in the 11560 chain). The last `insn_list' in the chain has a null pointer as second 11561 operand. The significant thing about the chain is which insns appear 11562 in it (as first operands of `insn_list' expressions). Their order is 11563 not significant. 11564 11565 This list is originally set up by the flow analysis pass; it is a null 11566 pointer until then. Flow only adds links for those data dependencies 11567 which can be used for instruction combination. For each insn, the flow 11568 analysis pass adds a link to insns which store into registers values 11569 that are used for the first time in this insn. 11570 11571 The `REG_NOTES' field of an insn is a chain similar to the `LOG_LINKS' 11572 field but it includes `expr_list' expressions in addition to 11573 `insn_list' expressions. There are several kinds of register notes, 11574 which are distinguished by the machine mode, which in a register note 11575 is really understood as being an `enum reg_note'. The first operand OP 11576 of the note is data whose meaning depends on the kind of note. 11577 11578 The macro `REG_NOTE_KIND (X)' returns the kind of register note. Its 11579 counterpart, the macro `PUT_REG_NOTE_KIND (X, NEWKIND)' sets the 11580 register note type of X to be NEWKIND. 11581 11582 Register notes are of three classes: They may say something about an 11583 input to an insn, they may say something about an output of an insn, or 11584 they may create a linkage between two insns. There are also a set of 11585 values that are only used in `LOG_LINKS'. 11586 11587 These register notes annotate inputs to an insn: 11588 11589 `REG_DEAD' 11590 The value in OP dies in this insn; that is to say, altering the 11591 value immediately after this insn would not affect the future 11592 behavior of the program. 11593 11594 It does not follow that the register OP has no useful value after 11595 this insn since OP is not necessarily modified by this insn. 11596 Rather, no subsequent instruction uses the contents of OP. 11597 11598 `REG_UNUSED' 11599 The register OP being set by this insn will not be used in a 11600 subsequent insn. This differs from a `REG_DEAD' note, which 11601 indicates that the value in an input will not be used subsequently. 11602 These two notes are independent; both may be present for the same 11603 register. 11604 11605 `REG_INC' 11606 The register OP is incremented (or decremented; at this level 11607 there is no distinction) by an embedded side effect inside this 11608 insn. This means it appears in a `post_inc', `pre_inc', 11609 `post_dec' or `pre_dec' expression. 11610 11611 `REG_NONNEG' 11612 The register OP is known to have a nonnegative value when this 11613 insn is reached. This is used so that decrement and branch until 11614 zero instructions, such as the m68k dbra, can be matched. 11615 11616 The `REG_NONNEG' note is added to insns only if the machine 11617 description has a `decrement_and_branch_until_zero' pattern. 11618 11619 `REG_LABEL_OPERAND' 11620 This insn uses OP, a `code_label' or a `note' of type 11621 `NOTE_INSN_DELETED_LABEL', but is not a `jump_insn', or it is a 11622 `jump_insn' that refers to the operand as an ordinary operand. 11623 The label may still eventually be a jump target, but if so in an 11624 indirect jump in a subsequent insn. The presence of this note 11625 allows jump optimization to be aware that OP is, in fact, being 11626 used, and flow optimization to build an accurate flow graph. 11627 11628 `REG_LABEL_TARGET' 11629 This insn is a `jump_insn' but not a `addr_vec' or 11630 `addr_diff_vec'. It uses OP, a `code_label' as a direct or 11631 indirect jump target. Its purpose is similar to that of 11632 `REG_LABEL_OPERAND'. This note is only present if the insn has 11633 multiple targets; the last label in the insn (in the highest 11634 numbered insn-field) goes into the `JUMP_LABEL' field and does not 11635 have a `REG_LABEL_TARGET' note. *Note JUMP_LABEL: Insns. 11636 11637 `REG_CROSSING_JUMP' 11638 This insn is an branching instruction (either an unconditional 11639 jump or an indirect jump) which crosses between hot and cold 11640 sections, which could potentially be very far apart in the 11641 executable. The presence of this note indicates to other 11642 optimizations that this branching instruction should not be 11643 "collapsed" into a simpler branching construct. It is used when 11644 the optimization to partition basic blocks into hot and cold 11645 sections is turned on. 11646 11647 `REG_SETJMP' 11648 Appears attached to each `CALL_INSN' to `setjmp' or a related 11649 function. 11650 11651 The following notes describe attributes of outputs of an insn: 11652 11653 `REG_EQUIV' 11654 `REG_EQUAL' 11655 This note is only valid on an insn that sets only one register and 11656 indicates that that register will be equal to OP at run time; the 11657 scope of this equivalence differs between the two types of notes. 11658 The value which the insn explicitly copies into the register may 11659 look different from OP, but they will be equal at run time. If the 11660 output of the single `set' is a `strict_low_part' expression, the 11661 note refers to the register that is contained in `SUBREG_REG' of 11662 the `subreg' expression. 11663 11664 For `REG_EQUIV', the register is equivalent to OP throughout the 11665 entire function, and could validly be replaced in all its 11666 occurrences by OP. ("Validly" here refers to the data flow of the 11667 program; simple replacement may make some insns invalid.) For 11668 example, when a constant is loaded into a register that is never 11669 assigned any other value, this kind of note is used. 11670 11671 When a parameter is copied into a pseudo-register at entry to a 11672 function, a note of this kind records that the register is 11673 equivalent to the stack slot where the parameter was passed. 11674 Although in this case the register may be set by other insns, it 11675 is still valid to replace the register by the stack slot 11676 throughout the function. 11677 11678 A `REG_EQUIV' note is also used on an instruction which copies a 11679 register parameter into a pseudo-register at entry to a function, 11680 if there is a stack slot where that parameter could be stored. 11681 Although other insns may set the pseudo-register, it is valid for 11682 the compiler to replace the pseudo-register by stack slot 11683 throughout the function, provided the compiler ensures that the 11684 stack slot is properly initialized by making the replacement in 11685 the initial copy instruction as well. This is used on machines 11686 for which the calling convention allocates stack space for 11687 register parameters. See `REG_PARM_STACK_SPACE' in *Note Stack 11688 Arguments::. 11689 11690 In the case of `REG_EQUAL', the register that is set by this insn 11691 will be equal to OP at run time at the end of this insn but not 11692 necessarily elsewhere in the function. In this case, OP is 11693 typically an arithmetic expression. For example, when a sequence 11694 of insns such as a library call is used to perform an arithmetic 11695 operation, this kind of note is attached to the insn that produces 11696 or copies the final value. 11697 11698 These two notes are used in different ways by the compiler passes. 11699 `REG_EQUAL' is used by passes prior to register allocation (such as 11700 common subexpression elimination and loop optimization) to tell 11701 them how to think of that value. `REG_EQUIV' notes are used by 11702 register allocation to indicate that there is an available 11703 substitute expression (either a constant or a `mem' expression for 11704 the location of a parameter on the stack) that may be used in 11705 place of a register if insufficient registers are available. 11706 11707 Except for stack homes for parameters, which are indicated by a 11708 `REG_EQUIV' note and are not useful to the early optimization 11709 passes and pseudo registers that are equivalent to a memory 11710 location throughout their entire life, which is not detected until 11711 later in the compilation, all equivalences are initially indicated 11712 by an attached `REG_EQUAL' note. In the early stages of register 11713 allocation, a `REG_EQUAL' note is changed into a `REG_EQUIV' note 11714 if OP is a constant and the insn represents the only set of its 11715 destination register. 11716 11717 Thus, compiler passes prior to register allocation need only check 11718 for `REG_EQUAL' notes and passes subsequent to register allocation 11719 need only check for `REG_EQUIV' notes. 11720 11721 These notes describe linkages between insns. They occur in pairs: one 11722 insn has one of a pair of notes that points to a second insn, which has 11723 the inverse note pointing back to the first insn. 11724 11725 `REG_CC_SETTER' 11726 `REG_CC_USER' 11727 On machines that use `cc0', the insns which set and use `cc0' set 11728 and use `cc0' are adjacent. However, when branch delay slot 11729 filling is done, this may no longer be true. In this case a 11730 `REG_CC_USER' note will be placed on the insn setting `cc0' to 11731 point to the insn using `cc0' and a `REG_CC_SETTER' note will be 11732 placed on the insn using `cc0' to point to the insn setting `cc0'. 11733 11734 These values are only used in the `LOG_LINKS' field, and indicate the 11735 type of dependency that each link represents. Links which indicate a 11736 data dependence (a read after write dependence) do not use any code, 11737 they simply have mode `VOIDmode', and are printed without any 11738 descriptive text. 11739 11740 `REG_DEP_TRUE' 11741 This indicates a true dependence (a read after write dependence). 11742 11743 `REG_DEP_OUTPUT' 11744 This indicates an output dependence (a write after write 11745 dependence). 11746 11747 `REG_DEP_ANTI' 11748 This indicates an anti dependence (a write after read dependence). 11749 11750 11751 These notes describe information gathered from gcov profile data. They 11752 are stored in the `REG_NOTES' field of an insn as an `expr_list'. 11753 11754 `REG_BR_PROB' 11755 This is used to specify the ratio of branches to non-branches of a 11756 branch insn according to the profile data. The value is stored as 11757 a value between 0 and REG_BR_PROB_BASE; larger values indicate a 11758 higher probability that the branch will be taken. 11759 11760 `REG_BR_PRED' 11761 These notes are found in JUMP insns after delayed branch scheduling 11762 has taken place. They indicate both the direction and the 11763 likelihood of the JUMP. The format is a bitmask of ATTR_FLAG_* 11764 values. 11765 11766 `REG_FRAME_RELATED_EXPR' 11767 This is used on an RTX_FRAME_RELATED_P insn wherein the attached 11768 expression is used in place of the actual insn pattern. This is 11769 done in cases where the pattern is either complex or misleading. 11770 11771 For convenience, the machine mode in an `insn_list' or `expr_list' is 11772 printed using these symbolic codes in debugging dumps. 11773 11774 The only difference between the expression codes `insn_list' and 11775 `expr_list' is that the first operand of an `insn_list' is assumed to 11776 be an insn and is printed in debugging dumps as the insn's unique id; 11777 the first operand of an `expr_list' is printed in the ordinary way as 11778 an expression. 11779 11780 11781 File: gccint.info, Node: Calls, Next: Sharing, Prev: Insns, Up: RTL 11782 11783 10.19 RTL Representation of Function-Call Insns 11784 =============================================== 11785 11786 Insns that call subroutines have the RTL expression code `call_insn'. 11787 These insns must satisfy special rules, and their bodies must use a 11788 special RTL expression code, `call'. 11789 11790 A `call' expression has two operands, as follows: 11791 11792 (call (mem:FM ADDR) NBYTES) 11793 11794 Here NBYTES is an operand that represents the number of bytes of 11795 argument data being passed to the subroutine, FM is a machine mode 11796 (which must equal as the definition of the `FUNCTION_MODE' macro in the 11797 machine description) and ADDR represents the address of the subroutine. 11798 11799 For a subroutine that returns no value, the `call' expression as shown 11800 above is the entire body of the insn, except that the insn might also 11801 contain `use' or `clobber' expressions. 11802 11803 For a subroutine that returns a value whose mode is not `BLKmode', the 11804 value is returned in a hard register. If this register's number is R, 11805 then the body of the call insn looks like this: 11806 11807 (set (reg:M R) 11808 (call (mem:FM ADDR) NBYTES)) 11809 11810 This RTL expression makes it clear (to the optimizer passes) that the 11811 appropriate register receives a useful value in this insn. 11812 11813 When a subroutine returns a `BLKmode' value, it is handled by passing 11814 to the subroutine the address of a place to store the value. So the 11815 call insn itself does not "return" any value, and it has the same RTL 11816 form as a call that returns nothing. 11817 11818 On some machines, the call instruction itself clobbers some register, 11819 for example to contain the return address. `call_insn' insns on these 11820 machines should have a body which is a `parallel' that contains both 11821 the `call' expression and `clobber' expressions that indicate which 11822 registers are destroyed. Similarly, if the call instruction requires 11823 some register other than the stack pointer that is not explicitly 11824 mentioned in its RTL, a `use' subexpression should mention that 11825 register. 11826 11827 Functions that are called are assumed to modify all registers listed in 11828 the configuration macro `CALL_USED_REGISTERS' (*note Register Basics::) 11829 and, with the exception of `const' functions and library calls, to 11830 modify all of memory. 11831 11832 Insns containing just `use' expressions directly precede the 11833 `call_insn' insn to indicate which registers contain inputs to the 11834 function. Similarly, if registers other than those in 11835 `CALL_USED_REGISTERS' are clobbered by the called function, insns 11836 containing a single `clobber' follow immediately after the call to 11837 indicate which registers. 11838 11839 11840 File: gccint.info, Node: Sharing, Next: Reading RTL, Prev: Calls, Up: RTL 11841 11842 10.20 Structure Sharing Assumptions 11843 =================================== 11844 11845 The compiler assumes that certain kinds of RTL expressions are unique; 11846 there do not exist two distinct objects representing the same value. 11847 In other cases, it makes an opposite assumption: that no RTL expression 11848 object of a certain kind appears in more than one place in the 11849 containing structure. 11850 11851 These assumptions refer to a single function; except for the RTL 11852 objects that describe global variables and external functions, and a 11853 few standard objects such as small integer constants, no RTL objects 11854 are common to two functions. 11855 11856 * Each pseudo-register has only a single `reg' object to represent 11857 it, and therefore only a single machine mode. 11858 11859 * For any symbolic label, there is only one `symbol_ref' object 11860 referring to it. 11861 11862 * All `const_int' expressions with equal values are shared. 11863 11864 * There is only one `pc' expression. 11865 11866 * There is only one `cc0' expression. 11867 11868 * There is only one `const_double' expression with value 0 for each 11869 floating point mode. Likewise for values 1 and 2. 11870 11871 * There is only one `const_vector' expression with value 0 for each 11872 vector mode, be it an integer or a double constant vector. 11873 11874 * No `label_ref' or `scratch' appears in more than one place in the 11875 RTL structure; in other words, it is safe to do a tree-walk of all 11876 the insns in the function and assume that each time a `label_ref' 11877 or `scratch' is seen it is distinct from all others that are seen. 11878 11879 * Only one `mem' object is normally created for each static variable 11880 or stack slot, so these objects are frequently shared in all the 11881 places they appear. However, separate but equal objects for these 11882 variables are occasionally made. 11883 11884 * When a single `asm' statement has multiple output operands, a 11885 distinct `asm_operands' expression is made for each output operand. 11886 However, these all share the vector which contains the sequence of 11887 input operands. This sharing is used later on to test whether two 11888 `asm_operands' expressions come from the same statement, so all 11889 optimizations must carefully preserve the sharing if they copy the 11890 vector at all. 11891 11892 * No RTL object appears in more than one place in the RTL structure 11893 except as described above. Many passes of the compiler rely on 11894 this by assuming that they can modify RTL objects in place without 11895 unwanted side-effects on other insns. 11896 11897 * During initial RTL generation, shared structure is freely 11898 introduced. After all the RTL for a function has been generated, 11899 all shared structure is copied by `unshare_all_rtl' in 11900 `emit-rtl.c', after which the above rules are guaranteed to be 11901 followed. 11902 11903 * During the combiner pass, shared structure within an insn can exist 11904 temporarily. However, the shared structure is copied before the 11905 combiner is finished with the insn. This is done by calling 11906 `copy_rtx_if_shared', which is a subroutine of `unshare_all_rtl'. 11907 11908 11909 File: gccint.info, Node: Reading RTL, Prev: Sharing, Up: RTL 11910 11911 10.21 Reading RTL 11912 ================= 11913 11914 To read an RTL object from a file, call `read_rtx'. It takes one 11915 argument, a stdio stream, and returns a single RTL object. This routine 11916 is defined in `read-rtl.c'. It is not available in the compiler 11917 itself, only the various programs that generate the compiler back end 11918 from the machine description. 11919 11920 People frequently have the idea of using RTL stored as text in a file 11921 as an interface between a language front end and the bulk of GCC. This 11922 idea is not feasible. 11923 11924 GCC was designed to use RTL internally only. Correct RTL for a given 11925 program is very dependent on the particular target machine. And the RTL 11926 does not contain all the information about the program. 11927 11928 The proper way to interface GCC to a new language front end is with 11929 the "tree" data structure, described in the files `tree.h' and 11930 `tree.def'. The documentation for this structure (*note Trees::) is 11931 incomplete. 11932 11933 11934 File: gccint.info, Node: GENERIC, Next: GIMPLE, Prev: Trees, Up: Top 11935 11936 11 GENERIC 11937 ********** 11938 11939 The purpose of GENERIC is simply to provide a language-independent way 11940 of representing an entire function in trees. To this end, it was 11941 necessary to add a few new tree codes to the back end, but most 11942 everything was already there. If you can express it with the codes in 11943 `gcc/tree.def', it's GENERIC. 11944 11945 Early on, there was a great deal of debate about how to think about 11946 statements in a tree IL. In GENERIC, a statement is defined as any 11947 expression whose value, if any, is ignored. A statement will always 11948 have `TREE_SIDE_EFFECTS' set (or it will be discarded), but a 11949 non-statement expression may also have side effects. A `CALL_EXPR', 11950 for instance. 11951 11952 It would be possible for some local optimizations to work on the 11953 GENERIC form of a function; indeed, the adapted tree inliner works fine 11954 on GENERIC, but the current compiler performs inlining after lowering 11955 to GIMPLE (a restricted form described in the next section). Indeed, 11956 currently the frontends perform this lowering before handing off to 11957 `tree_rest_of_compilation', but this seems inelegant. 11958 11959 If necessary, a front end can use some language-dependent tree codes 11960 in its GENERIC representation, so long as it provides a hook for 11961 converting them to GIMPLE and doesn't expect them to work with any 11962 (hypothetical) optimizers that run before the conversion to GIMPLE. The 11963 intermediate representation used while parsing C and C++ looks very 11964 little like GENERIC, but the C and C++ gimplifier hooks are perfectly 11965 happy to take it as input and spit out GIMPLE. 11966 11967 * Menu: 11968 11969 * Statements:: 11970 11971 11972 File: gccint.info, Node: Statements, Up: GENERIC 11973 11974 11.1 Statements 11975 =============== 11976 11977 Most statements in GIMPLE are assignment statements, represented by 11978 `GIMPLE_ASSIGN'. No other C expressions can appear at statement level; 11979 a reference to a volatile object is converted into a `GIMPLE_ASSIGN'. 11980 11981 There are also several varieties of complex statements. 11982 11983 * Menu: 11984 11985 * Blocks:: 11986 * Statement Sequences:: 11987 * Empty Statements:: 11988 * Jumps:: 11989 * Cleanups:: 11990 11991 11992 File: gccint.info, Node: Blocks, Next: Statement Sequences, Up: Statements 11993 11994 11.1.1 Blocks 11995 ------------- 11996 11997 Block scopes and the variables they declare in GENERIC are expressed 11998 using the `BIND_EXPR' code, which in previous versions of GCC was 11999 primarily used for the C statement-expression extension. 12000 12001 Variables in a block are collected into `BIND_EXPR_VARS' in 12002 declaration order. Any runtime initialization is moved out of 12003 `DECL_INITIAL' and into a statement in the controlled block. When 12004 gimplifying from C or C++, this initialization replaces the `DECL_STMT'. 12005 12006 Variable-length arrays (VLAs) complicate this process, as their size 12007 often refers to variables initialized earlier in the block. To handle 12008 this, we currently split the block at that point, and move the VLA into 12009 a new, inner `BIND_EXPR'. This strategy may change in the future. 12010 12011 A C++ program will usually contain more `BIND_EXPR's than there are 12012 syntactic blocks in the source code, since several C++ constructs have 12013 implicit scopes associated with them. On the other hand, although the 12014 C++ front end uses pseudo-scopes to handle cleanups for objects with 12015 destructors, these don't translate into the GIMPLE form; multiple 12016 declarations at the same level use the same `BIND_EXPR'. 12017 12018 12019 File: gccint.info, Node: Statement Sequences, Next: Empty Statements, Prev: Blocks, Up: Statements 12020 12021 11.1.2 Statement Sequences 12022 -------------------------- 12023 12024 Multiple statements at the same nesting level are collected into a 12025 `STATEMENT_LIST'. Statement lists are modified and traversed using the 12026 interface in `tree-iterator.h'. 12027 12028 12029 File: gccint.info, Node: Empty Statements, Next: Jumps, Prev: Statement Sequences, Up: Statements 12030 12031 11.1.3 Empty Statements 12032 ----------------------- 12033 12034 Whenever possible, statements with no effect are discarded. But if 12035 they are nested within another construct which cannot be discarded for 12036 some reason, they are instead replaced with an empty statement, 12037 generated by `build_empty_stmt'. Initially, all empty statements were 12038 shared, after the pattern of the Java front end, but this caused a lot 12039 of trouble in practice. 12040 12041 An empty statement is represented as `(void)0'. 12042 12043 12044 File: gccint.info, Node: Jumps, Next: Cleanups, Prev: Empty Statements, Up: Statements 12045 12046 11.1.4 Jumps 12047 ------------ 12048 12049 Other jumps are expressed by either `GOTO_EXPR' or `RETURN_EXPR'. 12050 12051 The operand of a `GOTO_EXPR' must be either a label or a variable 12052 containing the address to jump to. 12053 12054 The operand of a `RETURN_EXPR' is either `NULL_TREE', `RESULT_DECL', 12055 or a `MODIFY_EXPR' which sets the return value. It would be nice to 12056 move the `MODIFY_EXPR' into a separate statement, but the special 12057 return semantics in `expand_return' make that difficult. It may still 12058 happen in the future, perhaps by moving most of that logic into 12059 `expand_assignment'. 12060 12061 12062 File: gccint.info, Node: Cleanups, Prev: Jumps, Up: Statements 12063 12064 11.1.5 Cleanups 12065 --------------- 12066 12067 Destructors for local C++ objects and similar dynamic cleanups are 12068 represented in GIMPLE by a `TRY_FINALLY_EXPR'. `TRY_FINALLY_EXPR' has 12069 two operands, both of which are a sequence of statements to execute. 12070 The first sequence is executed. When it completes the second sequence 12071 is executed. 12072 12073 The first sequence may complete in the following ways: 12074 12075 1. Execute the last statement in the sequence and fall off the end. 12076 12077 2. Execute a goto statement (`GOTO_EXPR') to an ordinary label 12078 outside the sequence. 12079 12080 3. Execute a return statement (`RETURN_EXPR'). 12081 12082 4. Throw an exception. This is currently not explicitly represented 12083 in GIMPLE. 12084 12085 12086 The second sequence is not executed if the first sequence completes by 12087 calling `setjmp' or `exit' or any other function that does not return. 12088 The second sequence is also not executed if the first sequence 12089 completes via a non-local goto or a computed goto (in general the 12090 compiler does not know whether such a goto statement exits the first 12091 sequence or not, so we assume that it doesn't). 12092 12093 After the second sequence is executed, if it completes normally by 12094 falling off the end, execution continues wherever the first sequence 12095 would have continued, by falling off the end, or doing a goto, etc. 12096 12097 `TRY_FINALLY_EXPR' complicates the flow graph, since the cleanup needs 12098 to appear on every edge out of the controlled block; this reduces the 12099 freedom to move code across these edges. Therefore, the EH lowering 12100 pass which runs before most of the optimization passes eliminates these 12101 expressions by explicitly adding the cleanup to each edge. Rethrowing 12102 the exception is represented using `RESX_EXPR'. 12103 12104 12105 File: gccint.info, Node: GIMPLE, Next: Tree SSA, Prev: GENERIC, Up: Top 12106 12107 12 GIMPLE 12108 ********* 12109 12110 GIMPLE is a three-address representation derived from GENERIC by 12111 breaking down GENERIC expressions into tuples of no more than 3 12112 operands (with some exceptions like function calls). GIMPLE was 12113 heavily influenced by the SIMPLE IL used by the McCAT compiler project 12114 at McGill University, though we have made some different choices. For 12115 one thing, SIMPLE doesn't support `goto'. 12116 12117 Temporaries are introduced to hold intermediate values needed to 12118 compute complex expressions. Additionally, all the control structures 12119 used in GENERIC are lowered into conditional jumps, lexical scopes are 12120 removed and exception regions are converted into an on the side 12121 exception region tree. 12122 12123 The compiler pass which converts GENERIC into GIMPLE is referred to as 12124 the `gimplifier'. The gimplifier works recursively, generating GIMPLE 12125 tuples out of the original GENERIC expressions. 12126 12127 One of the early implementation strategies used for the GIMPLE 12128 representation was to use the same internal data structures used by 12129 front ends to represent parse trees. This simplified implementation 12130 because we could leverage existing functionality and interfaces. 12131 However, GIMPLE is a much more restrictive representation than abstract 12132 syntax trees (AST), therefore it does not require the full structural 12133 complexity provided by the main tree data structure. 12134 12135 The GENERIC representation of a function is stored in the 12136 `DECL_SAVED_TREE' field of the associated `FUNCTION_DECL' tree node. 12137 It is converted to GIMPLE by a call to `gimplify_function_tree'. 12138 12139 If a front end wants to include language-specific tree codes in the 12140 tree representation which it provides to the back end, it must provide a 12141 definition of `LANG_HOOKS_GIMPLIFY_EXPR' which knows how to convert the 12142 front end trees to GIMPLE. Usually such a hook will involve much of 12143 the same code for expanding front end trees to RTL. This function can 12144 return fully lowered GIMPLE, or it can return GENERIC trees and let the 12145 main gimplifier lower them the rest of the way; this is often simpler. 12146 GIMPLE that is not fully lowered is known as "High GIMPLE" and consists 12147 of the IL before the pass `pass_lower_cf'. High GIMPLE contains some 12148 container statements like lexical scopes (represented by `GIMPLE_BIND') 12149 and nested expressions (e.g., `GIMPLE_TRY'), while "Low GIMPLE" exposes 12150 all of the implicit jumps for control and exception expressions 12151 directly in the IL and EH region trees. 12152 12153 The C and C++ front ends currently convert directly from front end 12154 trees to GIMPLE, and hand that off to the back end rather than first 12155 converting to GENERIC. Their gimplifier hooks know about all the 12156 `_STMT' nodes and how to convert them to GENERIC forms. There was some 12157 work done on a genericization pass which would run first, but the 12158 existence of `STMT_EXPR' meant that in order to convert all of the C 12159 statements into GENERIC equivalents would involve walking the entire 12160 tree anyway, so it was simpler to lower all the way. This might change 12161 in the future if someone writes an optimization pass which would work 12162 better with higher-level trees, but currently the optimizers all expect 12163 GIMPLE. 12164 12165 You can request to dump a C-like representation of the GIMPLE form 12166 with the flag `-fdump-tree-gimple'. 12167 12168 * Menu: 12169 12170 * Tuple representation:: 12171 * GIMPLE instruction set:: 12172 * GIMPLE Exception Handling:: 12173 * Temporaries:: 12174 * Operands:: 12175 * Manipulating GIMPLE statements:: 12176 * Tuple specific accessors:: 12177 * GIMPLE sequences:: 12178 * Sequence iterators:: 12179 * Adding a new GIMPLE statement code:: 12180 * Statement and operand traversals:: 12181 12182 12183 File: gccint.info, Node: Tuple representation, Next: GIMPLE instruction set, Up: GIMPLE 12184 12185 12.1 Tuple representation 12186 ========================= 12187 12188 GIMPLE instructions are tuples of variable size divided in two groups: 12189 a header describing the instruction and its locations, and a variable 12190 length body with all the operands. Tuples are organized into a 12191 hierarchy with 3 main classes of tuples. 12192 12193 12.1.1 `gimple_statement_base' (gsbase) 12194 --------------------------------------- 12195 12196 This is the root of the hierarchy, it holds basic information needed by 12197 most GIMPLE statements. There are some fields that may not be relevant 12198 to every GIMPLE statement, but those were moved into the base structure 12199 to take advantage of holes left by other fields (thus making the 12200 structure more compact). The structure takes 4 words (32 bytes) on 64 12201 bit hosts: 12202 12203 Field Size (bits) 12204 `code' 8 12205 `subcode' 16 12206 `no_warning' 1 12207 `visited' 1 12208 `nontemporal_move' 1 12209 `plf' 2 12210 `modified' 1 12211 `has_volatile_ops' 1 12212 `references_memory_p' 1 12213 `uid' 32 12214 `location' 32 12215 `num_ops' 32 12216 `bb' 64 12217 `block' 63 12218 Total size 32 bytes 12219 12220 * `code' Main identifier for a GIMPLE instruction. 12221 12222 * `subcode' Used to distinguish different variants of the same basic 12223 instruction or provide flags applicable to a given code. The 12224 `subcode' flags field has different uses depending on the code of 12225 the instruction, but mostly it distinguishes instructions of the 12226 same family. The most prominent use of this field is in 12227 assignments, where subcode indicates the operation done on the RHS 12228 of the assignment. For example, a = b + c is encoded as 12229 `GIMPLE_ASSIGN <PLUS_EXPR, a, b, c>'. 12230 12231 * `no_warning' Bitflag to indicate whether a warning has already 12232 been issued on this statement. 12233 12234 * `visited' General purpose "visited" marker. Set and cleared by 12235 each pass when needed. 12236 12237 * `nontemporal_move' Bitflag used in assignments that represent 12238 non-temporal moves. Although this bitflag is only used in 12239 assignments, it was moved into the base to take advantage of the 12240 bit holes left by the previous fields. 12241 12242 * `plf' Pass Local Flags. This 2-bit mask can be used as general 12243 purpose markers by any pass. Passes are responsible for clearing 12244 and setting these two flags accordingly. 12245 12246 * `modified' Bitflag to indicate whether the statement has been 12247 modified. Used mainly by the operand scanner to determine when to 12248 re-scan a statement for operands. 12249 12250 * `has_volatile_ops' Bitflag to indicate whether this statement 12251 contains operands that have been marked volatile. 12252 12253 * `references_memory_p' Bitflag to indicate whether this statement 12254 contains memory references (i.e., its operands are either global 12255 variables, or pointer dereferences or anything that must reside in 12256 memory). 12257 12258 * `uid' This is an unsigned integer used by passes that want to 12259 assign IDs to every statement. These IDs must be assigned and used 12260 by each pass. 12261 12262 * `location' This is a `location_t' identifier to specify source code 12263 location for this statement. It is inherited from the front end. 12264 12265 * `num_ops' Number of operands that this statement has. This 12266 specifies the size of the operand vector embedded in the tuple. 12267 Only used in some tuples, but it is declared in the base tuple to 12268 take advantage of the 32-bit hole left by the previous fields. 12269 12270 * `bb' Basic block holding the instruction. 12271 12272 * `block' Lexical block holding this statement. Also used for debug 12273 information generation. 12274 12275 12.1.2 `gimple_statement_with_ops' 12276 ---------------------------------- 12277 12278 This tuple is actually split in two: `gimple_statement_with_ops_base' 12279 and `gimple_statement_with_ops'. This is needed to accommodate the way 12280 the operand vector is allocated. The operand vector is defined to be an 12281 array of 1 element. So, to allocate a dynamic number of operands, the 12282 memory allocator (`gimple_alloc') simply allocates enough memory to 12283 hold the structure itself plus `N - 1' operands which run "off the end" 12284 of the structure. For example, to allocate space for a tuple with 3 12285 operands, `gimple_alloc' reserves `sizeof (struct 12286 gimple_statement_with_ops) + 2 * sizeof (tree)' bytes. 12287 12288 On the other hand, several fields in this tuple need to be shared with 12289 the `gimple_statement_with_memory_ops' tuple. So, these common fields 12290 are placed in `gimple_statement_with_ops_base' which is then inherited 12291 from the other two tuples. 12292 12293 `gsbase' 256 12294 `addresses_taken' 64 12295 `def_ops' 64 12296 `use_ops' 64 12297 `op' `num_ops' * 64 12298 Total size 56 + 8 * `num_ops' bytes 12299 12300 * `gsbase' Inherited from `struct gimple_statement_base'. 12301 12302 * `addresses_taken' Bitmap holding the UIDs of all the `VAR_DECL's 12303 whose addresses are taken by this statement. For example, a 12304 statement of the form `p = &b' will have the UID for symbol `b' in 12305 this set. 12306 12307 * `def_ops' Array of pointers into the operand array indicating all 12308 the slots that contain a variable written-to by the statement. 12309 This array is also used for immediate use chaining. Note that it 12310 would be possible to not rely on this array, but the changes 12311 required to implement this are pretty invasive. 12312 12313 * `use_ops' Similar to `def_ops' but for variables read by the 12314 statement. 12315 12316 * `op' Array of trees with `num_ops' slots. 12317 12318 12.1.3 `gimple_statement_with_memory_ops' 12319 ----------------------------------------- 12320 12321 This tuple is essentially identical to `gimple_statement_with_ops', 12322 except that it contains 4 additional fields to hold vectors related 12323 memory stores and loads. Similar to the previous case, the structure 12324 is split in two to accommodate for the operand vector 12325 (`gimple_statement_with_memory_ops_base' and 12326 `gimple_statement_with_memory_ops'). 12327 12328 Field Size (bits) 12329 `gsbase' 256 12330 `addresses_taken' 64 12331 `def_ops' 64 12332 `use_ops' 64 12333 `vdef_ops' 64 12334 `vuse_ops' 64 12335 `stores' 64 12336 `loads' 64 12337 `op' `num_ops' * 64 12338 Total size 88 + 8 * `num_ops' bytes 12339 12340 * `vdef_ops' Similar to `def_ops' but for `VDEF' operators. There is 12341 one entry per memory symbol written by this statement. This is 12342 used to maintain the memory SSA use-def and def-def chains. 12343 12344 * `vuse_ops' Similar to `use_ops' but for `VUSE' operators. There is 12345 one entry per memory symbol loaded by this statement. This is used 12346 to maintain the memory SSA use-def chains. 12347 12348 * `stores' Bitset with all the UIDs for the symbols written-to by the 12349 statement. This is different than `vdef_ops' in that all the 12350 affected symbols are mentioned in this set. If memory 12351 partitioning is enabled, the `vdef_ops' vector will refer to memory 12352 partitions. Furthermore, no SSA information is stored in this set. 12353 12354 * `loads' Similar to `stores', but for memory loads. (Note that there 12355 is some amount of redundancy here, it should be possible to reduce 12356 memory utilization further by removing these sets). 12357 12358 All the other tuples are defined in terms of these three basic ones. 12359 Each tuple will add some fields. The main gimple type is defined to be 12360 the union of all these structures (`GTY' markers elided for clarity): 12361 12362 union gimple_statement_d 12363 { 12364 struct gimple_statement_base gsbase; 12365 struct gimple_statement_with_ops gsops; 12366 struct gimple_statement_with_memory_ops gsmem; 12367 struct gimple_statement_omp omp; 12368 struct gimple_statement_bind gimple_bind; 12369 struct gimple_statement_catch gimple_catch; 12370 struct gimple_statement_eh_filter gimple_eh_filter; 12371 struct gimple_statement_phi gimple_phi; 12372 struct gimple_statement_resx gimple_resx; 12373 struct gimple_statement_try gimple_try; 12374 struct gimple_statement_wce gimple_wce; 12375 struct gimple_statement_asm gimple_asm; 12376 struct gimple_statement_omp_critical gimple_omp_critical; 12377 struct gimple_statement_omp_for gimple_omp_for; 12378 struct gimple_statement_omp_parallel gimple_omp_parallel; 12379 struct gimple_statement_omp_task gimple_omp_task; 12380 struct gimple_statement_omp_sections gimple_omp_sections; 12381 struct gimple_statement_omp_single gimple_omp_single; 12382 struct gimple_statement_omp_continue gimple_omp_continue; 12383 struct gimple_statement_omp_atomic_load gimple_omp_atomic_load; 12384 struct gimple_statement_omp_atomic_store gimple_omp_atomic_store; 12385 }; 12386 12387 12388 File: gccint.info, Node: GIMPLE instruction set, Next: GIMPLE Exception Handling, Prev: Tuple representation, Up: GIMPLE 12389 12390 12.2 GIMPLE instruction set 12391 =========================== 12392 12393 The following table briefly describes the GIMPLE instruction set. 12394 12395 Instruction High GIMPLE Low GIMPLE 12396 `GIMPLE_ASM' x x 12397 `GIMPLE_ASSIGN' x x 12398 `GIMPLE_BIND' x 12399 `GIMPLE_CALL' x x 12400 `GIMPLE_CATCH' x 12401 `GIMPLE_CHANGE_DYNAMIC_TYPE' x x 12402 `GIMPLE_COND' x x 12403 `GIMPLE_EH_FILTER' x 12404 `GIMPLE_GOTO' x x 12405 `GIMPLE_LABEL' x x 12406 `GIMPLE_NOP' x x 12407 `GIMPLE_OMP_ATOMIC_LOAD' x x 12408 `GIMPLE_OMP_ATOMIC_STORE' x x 12409 `GIMPLE_OMP_CONTINUE' x x 12410 `GIMPLE_OMP_CRITICAL' x x 12411 `GIMPLE_OMP_FOR' x x 12412 `GIMPLE_OMP_MASTER' x x 12413 `GIMPLE_OMP_ORDERED' x x 12414 `GIMPLE_OMP_PARALLEL' x x 12415 `GIMPLE_OMP_RETURN' x x 12416 `GIMPLE_OMP_SECTION' x x 12417 `GIMPLE_OMP_SECTIONS' x x 12418 `GIMPLE_OMP_SECTIONS_SWITCH' x x 12419 `GIMPLE_OMP_SINGLE' x x 12420 `GIMPLE_PHI' x 12421 `GIMPLE_RESX' x 12422 `GIMPLE_RETURN' x x 12423 `GIMPLE_SWITCH' x x 12424 `GIMPLE_TRY' x 12425 12426 12427 File: gccint.info, Node: GIMPLE Exception Handling, Next: Temporaries, Prev: GIMPLE instruction set, Up: GIMPLE 12428 12429 12.3 Exception Handling 12430 ======================= 12431 12432 Other exception handling constructs are represented using 12433 `GIMPLE_TRY_CATCH'. `GIMPLE_TRY_CATCH' has two operands. The first 12434 operand is a sequence of statements to execute. If executing these 12435 statements does not throw an exception, then the second operand is 12436 ignored. Otherwise, if an exception is thrown, then the second operand 12437 of the `GIMPLE_TRY_CATCH' is checked. The second operand may have the 12438 following forms: 12439 12440 1. A sequence of statements to execute. When an exception occurs, 12441 these statements are executed, and then the exception is rethrown. 12442 12443 2. A sequence of `GIMPLE_CATCH' statements. Each `GIMPLE_CATCH' has 12444 a list of applicable exception types and handler code. If the 12445 thrown exception matches one of the caught types, the associated 12446 handler code is executed. If the handler code falls off the 12447 bottom, execution continues after the original `GIMPLE_TRY_CATCH'. 12448 12449 3. An `GIMPLE_EH_FILTER' statement. This has a list of permitted 12450 exception types, and code to handle a match failure. If the 12451 thrown exception does not match one of the allowed types, the 12452 associated match failure code is executed. If the thrown exception 12453 does match, it continues unwinding the stack looking for the next 12454 handler. 12455 12456 12457 Currently throwing an exception is not directly represented in GIMPLE, 12458 since it is implemented by calling a function. At some point in the 12459 future we will want to add some way to express that the call will throw 12460 an exception of a known type. 12461 12462 Just before running the optimizers, the compiler lowers the high-level 12463 EH constructs above into a set of `goto's, magic labels, and EH 12464 regions. Continuing to unwind at the end of a cleanup is represented 12465 with a `GIMPLE_RESX'. 12466 12467 12468 File: gccint.info, Node: Temporaries, Next: Operands, Prev: GIMPLE Exception Handling, Up: GIMPLE 12469 12470 12.4 Temporaries 12471 ================ 12472 12473 When gimplification encounters a subexpression that is too complex, it 12474 creates a new temporary variable to hold the value of the 12475 subexpression, and adds a new statement to initialize it before the 12476 current statement. These special temporaries are known as `expression 12477 temporaries', and are allocated using `get_formal_tmp_var'. The 12478 compiler tries to always evaluate identical expressions into the same 12479 temporary, to simplify elimination of redundant calculations. 12480 12481 We can only use expression temporaries when we know that it will not 12482 be reevaluated before its value is used, and that it will not be 12483 otherwise modified(1). Other temporaries can be allocated using 12484 `get_initialized_tmp_var' or `create_tmp_var'. 12485 12486 Currently, an expression like `a = b + 5' is not reduced any further. 12487 We tried converting it to something like 12488 T1 = b + 5; 12489 a = T1; 12490 but this bloated the representation for minimal benefit. However, a 12491 variable which must live in memory cannot appear in an expression; its 12492 value is explicitly loaded into a temporary first. Similarly, storing 12493 the value of an expression to a memory variable goes through a 12494 temporary. 12495 12496 ---------- Footnotes ---------- 12497 12498 (1) These restrictions are derived from those in Morgan 4.8. 12499 12500 12501 File: gccint.info, Node: Operands, Next: Manipulating GIMPLE statements, Prev: Temporaries, Up: GIMPLE 12502 12503 12.5 Operands 12504 ============= 12505 12506 In general, expressions in GIMPLE consist of an operation and the 12507 appropriate number of simple operands; these operands must either be a 12508 GIMPLE rvalue (`is_gimple_val'), i.e. a constant or a register 12509 variable. More complex operands are factored out into temporaries, so 12510 that 12511 a = b + c + d 12512 becomes 12513 T1 = b + c; 12514 a = T1 + d; 12515 12516 The same rule holds for arguments to a `GIMPLE_CALL'. 12517 12518 The target of an assignment is usually a variable, but can also be an 12519 `INDIRECT_REF' or a compound lvalue as described below. 12520 12521 * Menu: 12522 12523 * Compound Expressions:: 12524 * Compound Lvalues:: 12525 * Conditional Expressions:: 12526 * Logical Operators:: 12527 12528 12529 File: gccint.info, Node: Compound Expressions, Next: Compound Lvalues, Up: Operands 12530 12531 12.5.1 Compound Expressions 12532 --------------------------- 12533 12534 The left-hand side of a C comma expression is simply moved into a 12535 separate statement. 12536 12537 12538 File: gccint.info, Node: Compound Lvalues, Next: Conditional Expressions, Prev: Compound Expressions, Up: Operands 12539 12540 12.5.2 Compound Lvalues 12541 ----------------------- 12542 12543 Currently compound lvalues involving array and structure field 12544 references are not broken down; an expression like `a.b[2] = 42' is not 12545 reduced any further (though complex array subscripts are). This 12546 restriction is a workaround for limitations in later optimizers; if we 12547 were to convert this to 12548 12549 T1 = &a.b; 12550 T1[2] = 42; 12551 12552 alias analysis would not remember that the reference to `T1[2]' came 12553 by way of `a.b', so it would think that the assignment could alias 12554 another member of `a'; this broke `struct-alias-1.c'. Future optimizer 12555 improvements may make this limitation unnecessary. 12556 12557 12558 File: gccint.info, Node: Conditional Expressions, Next: Logical Operators, Prev: Compound Lvalues, Up: Operands 12559 12560 12.5.3 Conditional Expressions 12561 ------------------------------ 12562 12563 A C `?:' expression is converted into an `if' statement with each 12564 branch assigning to the same temporary. So, 12565 12566 a = b ? c : d; 12567 becomes 12568 if (b == 1) 12569 T1 = c; 12570 else 12571 T1 = d; 12572 a = T1; 12573 12574 The GIMPLE level if-conversion pass re-introduces `?:' expression, if 12575 appropriate. It is used to vectorize loops with conditions using vector 12576 conditional operations. 12577 12578 Note that in GIMPLE, `if' statements are represented using 12579 `GIMPLE_COND', as described below. 12580 12581 12582 File: gccint.info, Node: Logical Operators, Prev: Conditional Expressions, Up: Operands 12583 12584 12.5.4 Logical Operators 12585 ------------------------ 12586 12587 Except when they appear in the condition operand of a `GIMPLE_COND', 12588 logical `and' and `or' operators are simplified as follows: `a = b && 12589 c' becomes 12590 12591 T1 = (bool)b; 12592 if (T1 == true) 12593 T1 = (bool)c; 12594 a = T1; 12595 12596 Note that `T1' in this example cannot be an expression temporary, 12597 because it has two different assignments. 12598 12599 12.5.5 Manipulating operands 12600 ---------------------------- 12601 12602 All gimple operands are of type `tree'. But only certain types of 12603 trees are allowed to be used as operand tuples. Basic validation is 12604 controlled by the function `get_gimple_rhs_class', which given a tree 12605 code, returns an `enum' with the following values of type `enum 12606 gimple_rhs_class' 12607 12608 * `GIMPLE_INVALID_RHS' The tree cannot be used as a GIMPLE operand. 12609 12610 * `GIMPLE_BINARY_RHS' The tree is a valid GIMPLE binary operation. 12611 12612 * `GIMPLE_UNARY_RHS' The tree is a valid GIMPLE unary operation. 12613 12614 * `GIMPLE_SINGLE_RHS' The tree is a single object, that cannot be 12615 split into simpler operands (for instance, `SSA_NAME', `VAR_DECL', 12616 `COMPONENT_REF', etc). 12617 12618 This operand class also acts as an escape hatch for tree nodes 12619 that may be flattened out into the operand vector, but would need 12620 more than two slots on the RHS. For instance, a `COND_EXPR' 12621 expression of the form `(a op b) ? x : y' could be flattened out 12622 on the operand vector using 4 slots, but it would also require 12623 additional processing to distinguish `c = a op b' from `c = a op b 12624 ? x : y'. Something similar occurs with `ASSERT_EXPR'. In time, 12625 these special case tree expressions should be flattened into the 12626 operand vector. 12627 12628 For tree nodes in the categories `GIMPLE_BINARY_RHS' and 12629 `GIMPLE_UNARY_RHS', they cannot be stored inside tuples directly. They 12630 first need to be flattened and separated into individual components. 12631 For instance, given the GENERIC expression 12632 12633 a = b + c 12634 12635 its tree representation is: 12636 12637 MODIFY_EXPR <VAR_DECL <a>, PLUS_EXPR <VAR_DECL <b>, VAR_DECL <c>>> 12638 12639 In this case, the GIMPLE form for this statement is logically 12640 identical to its GENERIC form but in GIMPLE, the `PLUS_EXPR' on the RHS 12641 of the assignment is not represented as a tree, instead the two 12642 operands are taken out of the `PLUS_EXPR' sub-tree and flattened into 12643 the GIMPLE tuple as follows: 12644 12645 GIMPLE_ASSIGN <PLUS_EXPR, VAR_DECL <a>, VAR_DECL <b>, VAR_DECL <c>> 12646 12647 12.5.6 Operand vector allocation 12648 -------------------------------- 12649 12650 The operand vector is stored at the bottom of the three tuple 12651 structures that accept operands. This means, that depending on the code 12652 of a given statement, its operand vector will be at different offsets 12653 from the base of the structure. To access tuple operands use the 12654 following accessors 12655 12656 -- GIMPLE function: unsigned gimple_num_ops (gimple g) 12657 Returns the number of operands in statement G. 12658 12659 -- GIMPLE function: tree gimple_op (gimple g, unsigned i) 12660 Returns operand `I' from statement `G'. 12661 12662 -- GIMPLE function: tree *gimple_ops (gimple g) 12663 Returns a pointer into the operand vector for statement `G'. This 12664 is computed using an internal table called `gimple_ops_offset_'[]. 12665 This table is indexed by the gimple code of `G'. 12666 12667 When the compiler is built, this table is filled-in using the 12668 sizes of the structures used by each statement code defined in 12669 gimple.def. Since the operand vector is at the bottom of the 12670 structure, for a gimple code `C' the offset is computed as sizeof 12671 (struct-of `C') - sizeof (tree). 12672 12673 This mechanism adds one memory indirection to every access when 12674 using `gimple_op'(), if this becomes a bottleneck, a pass can 12675 choose to memoize the result from `gimple_ops'() and use that to 12676 access the operands. 12677 12678 12.5.7 Operand validation 12679 ------------------------- 12680 12681 When adding a new operand to a gimple statement, the operand will be 12682 validated according to what each tuple accepts in its operand vector. 12683 These predicates are called by the `gimple_<name>_set_...()'. Each 12684 tuple will use one of the following predicates (Note, this list is not 12685 exhaustive): 12686 12687 -- GIMPLE function: is_gimple_operand (tree t) 12688 This is the most permissive of the predicates. It essentially 12689 checks whether t has a `gimple_rhs_class' of `GIMPLE_SINGLE_RHS'. 12690 12691 -- GIMPLE function: is_gimple_val (tree t) 12692 Returns true if t is a "GIMPLE value", which are all the 12693 non-addressable stack variables (variables for which 12694 `is_gimple_reg' returns true) and constants (expressions for which 12695 `is_gimple_min_invariant' returns true). 12696 12697 -- GIMPLE function: is_gimple_addressable (tree t) 12698 Returns true if t is a symbol or memory reference whose address 12699 can be taken. 12700 12701 -- GIMPLE function: is_gimple_asm_val (tree t) 12702 Similar to `is_gimple_val' but it also accepts hard registers. 12703 12704 -- GIMPLE function: is_gimple_call_addr (tree t) 12705 Return true if t is a valid expression to use as the function 12706 called by a `GIMPLE_CALL'. 12707 12708 -- GIMPLE function: is_gimple_constant (tree t) 12709 Return true if t is a valid gimple constant. 12710 12711 -- GIMPLE function: is_gimple_min_invariant (tree t) 12712 Return true if t is a valid minimal invariant. This is different 12713 from constants, in that the specific value of t may not be known 12714 at compile time, but it is known that it doesn't change (e.g., the 12715 address of a function local variable). 12716 12717 -- GIMPLE function: is_gimple_min_invariant_address (tree t) 12718 Return true if t is an `ADDR_EXPR' that does not change once the 12719 program is running. 12720 12721 12.5.8 Statement validation 12722 --------------------------- 12723 12724 -- GIMPLE function: is_gimple_assign (gimple g) 12725 Return true if the code of g is `GIMPLE_ASSIGN'. 12726 12727 -- GIMPLE function: is_gimple_call (gimple g) 12728 Return true if the code of g is `GIMPLE_CALL' 12729 12730 -- GIMPLE function: gimple_assign_cast_p (gimple g) 12731 Return true if g is a `GIMPLE_ASSIGN' that performs a type cast 12732 operation 12733 12734 12735 File: gccint.info, Node: Manipulating GIMPLE statements, Next: Tuple specific accessors, Prev: Operands, Up: GIMPLE 12736 12737 12.6 Manipulating GIMPLE statements 12738 =================================== 12739 12740 This section documents all the functions available to handle each of 12741 the GIMPLE instructions. 12742 12743 12.6.1 Common accessors 12744 ----------------------- 12745 12746 The following are common accessors for gimple statements. 12747 12748 -- GIMPLE function: enum gimple_code gimple_code (gimple g) 12749 Return the code for statement `G'. 12750 12751 -- GIMPLE function: basic_block gimple_bb (gimple g) 12752 Return the basic block to which statement `G' belongs to. 12753 12754 -- GIMPLE function: tree gimple_block (gimple g) 12755 Return the lexical scope block holding statement `G'. 12756 12757 -- GIMPLE function: tree gimple_expr_type (gimple stmt) 12758 Return the type of the main expression computed by `STMT'. Return 12759 `void_type_node' if `STMT' computes nothing. This will only return 12760 something meaningful for `GIMPLE_ASSIGN', `GIMPLE_COND' and 12761 `GIMPLE_CALL'. For all other tuple codes, it will return 12762 `void_type_node'. 12763 12764 -- GIMPLE function: enum tree_code gimple_expr_code (gimple stmt) 12765 Return the tree code for the expression computed by `STMT'. This 12766 is only meaningful for `GIMPLE_CALL', `GIMPLE_ASSIGN' and 12767 `GIMPLE_COND'. If `STMT' is `GIMPLE_CALL', it will return 12768 `CALL_EXPR'. For `GIMPLE_COND', it returns the code of the 12769 comparison predicate. For `GIMPLE_ASSIGN' it returns the code of 12770 the operation performed by the `RHS' of the assignment. 12771 12772 -- GIMPLE function: void gimple_set_block (gimple g, tree block) 12773 Set the lexical scope block of `G' to `BLOCK'. 12774 12775 -- GIMPLE function: location_t gimple_locus (gimple g) 12776 Return locus information for statement `G'. 12777 12778 -- GIMPLE function: void gimple_set_locus (gimple g, location_t locus) 12779 Set locus information for statement `G'. 12780 12781 -- GIMPLE function: bool gimple_locus_empty_p (gimple g) 12782 Return true if `G' does not have locus information. 12783 12784 -- GIMPLE function: bool gimple_no_warning_p (gimple stmt) 12785 Return true if no warnings should be emitted for statement `STMT'. 12786 12787 -- GIMPLE function: void gimple_set_visited (gimple stmt, bool 12788 visited_p) 12789 Set the visited status on statement `STMT' to `VISITED_P'. 12790 12791 -- GIMPLE function: bool gimple_visited_p (gimple stmt) 12792 Return the visited status on statement `STMT'. 12793 12794 -- GIMPLE function: void gimple_set_plf (gimple stmt, enum plf_mask 12795 plf, bool val_p) 12796 Set pass local flag `PLF' on statement `STMT' to `VAL_P'. 12797 12798 -- GIMPLE function: unsigned int gimple_plf (gimple stmt, enum 12799 plf_mask plf) 12800 Return the value of pass local flag `PLF' on statement `STMT'. 12801 12802 -- GIMPLE function: bool gimple_has_ops (gimple g) 12803 Return true if statement `G' has register or memory operands. 12804 12805 -- GIMPLE function: bool gimple_has_mem_ops (gimple g) 12806 Return true if statement `G' has memory operands. 12807 12808 -- GIMPLE function: unsigned gimple_num_ops (gimple g) 12809 Return the number of operands for statement `G'. 12810 12811 -- GIMPLE function: tree *gimple_ops (gimple g) 12812 Return the array of operands for statement `G'. 12813 12814 -- GIMPLE function: tree gimple_op (gimple g, unsigned i) 12815 Return operand `I' for statement `G'. 12816 12817 -- GIMPLE function: tree *gimple_op_ptr (gimple g, unsigned i) 12818 Return a pointer to operand `I' for statement `G'. 12819 12820 -- GIMPLE function: void gimple_set_op (gimple g, unsigned i, tree op) 12821 Set operand `I' of statement `G' to `OP'. 12822 12823 -- GIMPLE function: bitmap gimple_addresses_taken (gimple stmt) 12824 Return the set of symbols that have had their address taken by 12825 `STMT'. 12826 12827 -- GIMPLE function: struct def_optype_d *gimple_def_ops (gimple g) 12828 Return the set of `DEF' operands for statement `G'. 12829 12830 -- GIMPLE function: void gimple_set_def_ops (gimple g, struct 12831 def_optype_d *def) 12832 Set `DEF' to be the set of `DEF' operands for statement `G'. 12833 12834 -- GIMPLE function: struct use_optype_d *gimple_use_ops (gimple g) 12835 Return the set of `USE' operands for statement `G'. 12836 12837 -- GIMPLE function: void gimple_set_use_ops (gimple g, struct 12838 use_optype_d *use) 12839 Set `USE' to be the set of `USE' operands for statement `G'. 12840 12841 -- GIMPLE function: struct voptype_d *gimple_vuse_ops (gimple g) 12842 Return the set of `VUSE' operands for statement `G'. 12843 12844 -- GIMPLE function: void gimple_set_vuse_ops (gimple g, struct 12845 voptype_d *ops) 12846 Set `OPS' to be the set of `VUSE' operands for statement `G'. 12847 12848 -- GIMPLE function: struct voptype_d *gimple_vdef_ops (gimple g) 12849 Return the set of `VDEF' operands for statement `G'. 12850 12851 -- GIMPLE function: void gimple_set_vdef_ops (gimple g, struct 12852 voptype_d *ops) 12853 Set `OPS' to be the set of `VDEF' operands for statement `G'. 12854 12855 -- GIMPLE function: bitmap gimple_loaded_syms (gimple g) 12856 Return the set of symbols loaded by statement `G'. Each element of 12857 the set is the `DECL_UID' of the corresponding symbol. 12858 12859 -- GIMPLE function: bitmap gimple_stored_syms (gimple g) 12860 Return the set of symbols stored by statement `G'. Each element of 12861 the set is the `DECL_UID' of the corresponding symbol. 12862 12863 -- GIMPLE function: bool gimple_modified_p (gimple g) 12864 Return true if statement `G' has operands and the modified field 12865 has been set. 12866 12867 -- GIMPLE function: bool gimple_has_volatile_ops (gimple stmt) 12868 Return true if statement `STMT' contains volatile operands. 12869 12870 -- GIMPLE function: void gimple_set_has_volatile_ops (gimple stmt, 12871 bool volatilep) 12872 Return true if statement `STMT' contains volatile operands. 12873 12874 -- GIMPLE function: void update_stmt (gimple s) 12875 Mark statement `S' as modified, and update it. 12876 12877 -- GIMPLE function: void update_stmt_if_modified (gimple s) 12878 Update statement `S' if it has been marked modified. 12879 12880 -- GIMPLE function: gimple gimple_copy (gimple stmt) 12881 Return a deep copy of statement `STMT'. 12882 12883 12884 File: gccint.info, Node: Tuple specific accessors, Next: GIMPLE sequences, Prev: Manipulating GIMPLE statements, Up: GIMPLE 12885 12886 12.7 Tuple specific accessors 12887 ============================= 12888 12889 * Menu: 12890 12891 * `GIMPLE_ASM':: 12892 * `GIMPLE_ASSIGN':: 12893 * `GIMPLE_BIND':: 12894 * `GIMPLE_CALL':: 12895 * `GIMPLE_CATCH':: 12896 * `GIMPLE_CHANGE_DYNAMIC_TYPE':: 12897 * `GIMPLE_COND':: 12898 * `GIMPLE_EH_FILTER':: 12899 * `GIMPLE_LABEL':: 12900 * `GIMPLE_NOP':: 12901 * `GIMPLE_OMP_ATOMIC_LOAD':: 12902 * `GIMPLE_OMP_ATOMIC_STORE':: 12903 * `GIMPLE_OMP_CONTINUE':: 12904 * `GIMPLE_OMP_CRITICAL':: 12905 * `GIMPLE_OMP_FOR':: 12906 * `GIMPLE_OMP_MASTER':: 12907 * `GIMPLE_OMP_ORDERED':: 12908 * `GIMPLE_OMP_PARALLEL':: 12909 * `GIMPLE_OMP_RETURN':: 12910 * `GIMPLE_OMP_SECTION':: 12911 * `GIMPLE_OMP_SECTIONS':: 12912 * `GIMPLE_OMP_SINGLE':: 12913 * `GIMPLE_PHI':: 12914 * `GIMPLE_RESX':: 12915 * `GIMPLE_RETURN':: 12916 * `GIMPLE_SWITCH':: 12917 * `GIMPLE_TRY':: 12918 * `GIMPLE_WITH_CLEANUP_EXPR':: 12919 12920 12921 File: gccint.info, Node: `GIMPLE_ASM', Next: `GIMPLE_ASSIGN', Up: Tuple specific accessors 12922 12923 12.7.1 `GIMPLE_ASM' 12924 ------------------- 12925 12926 -- GIMPLE function: gimple gimple_build_asm (const char *string, 12927 ninputs, noutputs, nclobbers, ...) 12928 Build a `GIMPLE_ASM' statement. This statement is used for 12929 building in-line assembly constructs. `STRING' is the assembly 12930 code. `NINPUT' is the number of register inputs. `NOUTPUT' is the 12931 number of register outputs. `NCLOBBERS' is the number of clobbered 12932 registers. The rest of the arguments trees for each input, 12933 output, and clobbered registers. 12934 12935 -- GIMPLE function: gimple gimple_build_asm_vec (const char *, 12936 VEC(tree,gc) *, VEC(tree,gc) *, VEC(tree,gc) *) 12937 Identical to gimple_build_asm, but the arguments are passed in 12938 VECs. 12939 12940 -- GIMPLE function: gimple_asm_ninputs (gimple g) 12941 Return the number of input operands for `GIMPLE_ASM' `G'. 12942 12943 -- GIMPLE function: gimple_asm_noutputs (gimple g) 12944 Return the number of output operands for `GIMPLE_ASM' `G'. 12945 12946 -- GIMPLE function: gimple_asm_nclobbers (gimple g) 12947 Return the number of clobber operands for `GIMPLE_ASM' `G'. 12948 12949 -- GIMPLE function: tree gimple_asm_input_op (gimple g, unsigned index) 12950 Return input operand `INDEX' of `GIMPLE_ASM' `G'. 12951 12952 -- GIMPLE function: void gimple_asm_set_input_op (gimple g, unsigned 12953 index, tree in_op) 12954 Set `IN_OP' to be input operand `INDEX' in `GIMPLE_ASM' `G'. 12955 12956 -- GIMPLE function: tree gimple_asm_output_op (gimple g, unsigned 12957 index) 12958 Return output operand `INDEX' of `GIMPLE_ASM' `G'. 12959 12960 -- GIMPLE function: void gimple_asm_set_output_op (gimple g, unsigned 12961 index, tree out_op) 12962 Set `OUT_OP' to be output operand `INDEX' in `GIMPLE_ASM' `G'. 12963 12964 -- GIMPLE function: tree gimple_asm_clobber_op (gimple g, unsigned 12965 index) 12966 Return clobber operand `INDEX' of `GIMPLE_ASM' `G'. 12967 12968 -- GIMPLE function: void gimple_asm_set_clobber_op (gimple g, unsigned 12969 index, tree clobber_op) 12970 Set `CLOBBER_OP' to be clobber operand `INDEX' in `GIMPLE_ASM' `G'. 12971 12972 -- GIMPLE function: const char *gimple_asm_string (gimple g) 12973 Return the string representing the assembly instruction in 12974 `GIMPLE_ASM' `G'. 12975 12976 -- GIMPLE function: bool gimple_asm_volatile_p (gimple g) 12977 Return true if `G' is an asm statement marked volatile. 12978 12979 -- GIMPLE function: void gimple_asm_set_volatile (gimple g) 12980 Mark asm statement `G' as volatile. 12981 12982 -- GIMPLE function: void gimple_asm_clear_volatile (gimple g) 12983 Remove volatile marker from asm statement `G'. 12984 12985 12986 File: gccint.info, Node: `GIMPLE_ASSIGN', Next: `GIMPLE_BIND', Prev: `GIMPLE_ASM', Up: Tuple specific accessors 12987 12988 12.7.2 `GIMPLE_ASSIGN' 12989 ---------------------- 12990 12991 -- GIMPLE function: gimple gimple_build_assign (tree lhs, tree rhs) 12992 Build a `GIMPLE_ASSIGN' statement. The left-hand side is an lvalue 12993 passed in lhs. The right-hand side can be either a unary or 12994 binary tree expression. The expression tree rhs will be flattened 12995 and its operands assigned to the corresponding operand slots in 12996 the new statement. This function is useful when you already have 12997 a tree expression that you want to convert into a tuple. However, 12998 try to avoid building expression trees for the sole purpose of 12999 calling this function. If you already have the operands in 13000 separate trees, it is better to use `gimple_build_assign_with_ops'. 13001 13002 -- GIMPLE function: gimple gimplify_assign (tree dst, tree src, 13003 gimple_seq *seq_p) 13004 Build a new `GIMPLE_ASSIGN' tuple and append it to the end of 13005 `*SEQ_P'. 13006 13007 `DST'/`SRC' are the destination and source respectively. You can pass 13008 ungimplified trees in `DST' or `SRC', in which case they will be 13009 converted to a gimple operand if necessary. 13010 13011 This function returns the newly created `GIMPLE_ASSIGN' tuple. 13012 13013 -- GIMPLE function: gimple gimple_build_assign_with_ops (enum 13014 tree_code subcode, tree lhs, tree op1, tree op2) 13015 This function is similar to `gimple_build_assign', but is used to 13016 build a `GIMPLE_ASSIGN' statement when the operands of the 13017 right-hand side of the assignment are already split into different 13018 operands. 13019 13020 The left-hand side is an lvalue passed in lhs. Subcode is the 13021 `tree_code' for the right-hand side of the assignment. Op1 and op2 13022 are the operands. If op2 is null, subcode must be a `tree_code' 13023 for a unary expression. 13024 13025 -- GIMPLE function: enum tree_code gimple_assign_rhs_code (gimple g) 13026 Return the code of the expression computed on the `RHS' of 13027 assignment statement `G'. 13028 13029 -- GIMPLE function: enum gimple_rhs_class gimple_assign_rhs_class 13030 (gimple g) 13031 Return the gimple rhs class of the code for the expression 13032 computed on the rhs of assignment statement `G'. This will never 13033 return `GIMPLE_INVALID_RHS'. 13034 13035 -- GIMPLE function: tree gimple_assign_lhs (gimple g) 13036 Return the `LHS' of assignment statement `G'. 13037 13038 -- GIMPLE function: tree *gimple_assign_lhs_ptr (gimple g) 13039 Return a pointer to the `LHS' of assignment statement `G'. 13040 13041 -- GIMPLE function: tree gimple_assign_rhs1 (gimple g) 13042 Return the first operand on the `RHS' of assignment statement `G'. 13043 13044 -- GIMPLE function: tree *gimple_assign_rhs1_ptr (gimple g) 13045 Return the address of the first operand on the `RHS' of assignment 13046 statement `G'. 13047 13048 -- GIMPLE function: tree gimple_assign_rhs2 (gimple g) 13049 Return the second operand on the `RHS' of assignment statement `G'. 13050 13051 -- GIMPLE function: tree *gimple_assign_rhs2_ptr (gimple g) 13052 Return the address of the second operand on the `RHS' of assignment 13053 statement `G'. 13054 13055 -- GIMPLE function: void gimple_assign_set_lhs (gimple g, tree lhs) 13056 Set `LHS' to be the `LHS' operand of assignment statement `G'. 13057 13058 -- GIMPLE function: void gimple_assign_set_rhs1 (gimple g, tree rhs) 13059 Set `RHS' to be the first operand on the `RHS' of assignment 13060 statement `G'. 13061 13062 -- GIMPLE function: tree gimple_assign_rhs2 (gimple g) 13063 Return the second operand on the `RHS' of assignment statement `G'. 13064 13065 -- GIMPLE function: tree *gimple_assign_rhs2_ptr (gimple g) 13066 Return a pointer to the second operand on the `RHS' of assignment 13067 statement `G'. 13068 13069 -- GIMPLE function: void gimple_assign_set_rhs2 (gimple g, tree rhs) 13070 Set `RHS' to be the second operand on the `RHS' of assignment 13071 statement `G'. 13072 13073 -- GIMPLE function: bool gimple_assign_cast_p (gimple s) 13074 Return true if `S' is an type-cast assignment. 13075 13076 13077 File: gccint.info, Node: `GIMPLE_BIND', Next: `GIMPLE_CALL', Prev: `GIMPLE_ASSIGN', Up: Tuple specific accessors 13078 13079 12.7.3 `GIMPLE_BIND' 13080 -------------------- 13081 13082 -- GIMPLE function: gimple gimple_build_bind (tree vars, gimple_seq 13083 body) 13084 Build a `GIMPLE_BIND' statement with a list of variables in `VARS' 13085 and a body of statements in sequence `BODY'. 13086 13087 -- GIMPLE function: tree gimple_bind_vars (gimple g) 13088 Return the variables declared in the `GIMPLE_BIND' statement `G'. 13089 13090 -- GIMPLE function: void gimple_bind_set_vars (gimple g, tree vars) 13091 Set `VARS' to be the set of variables declared in the `GIMPLE_BIND' 13092 statement `G'. 13093 13094 -- GIMPLE function: void gimple_bind_append_vars (gimple g, tree vars) 13095 Append `VARS' to the set of variables declared in the `GIMPLE_BIND' 13096 statement `G'. 13097 13098 -- GIMPLE function: gimple_seq gimple_bind_body (gimple g) 13099 Return the GIMPLE sequence contained in the `GIMPLE_BIND' statement 13100 `G'. 13101 13102 -- GIMPLE function: void gimple_bind_set_body (gimple g, gimple_seq 13103 seq) 13104 Set `SEQ' to be sequence contained in the `GIMPLE_BIND' statement 13105 `G'. 13106 13107 -- GIMPLE function: void gimple_bind_add_stmt (gimple gs, gimple stmt) 13108 Append a statement to the end of a `GIMPLE_BIND''s body. 13109 13110 -- GIMPLE function: void gimple_bind_add_seq (gimple gs, gimple_seq 13111 seq) 13112 Append a sequence of statements to the end of a `GIMPLE_BIND''s 13113 body. 13114 13115 -- GIMPLE function: tree gimple_bind_block (gimple g) 13116 Return the `TREE_BLOCK' node associated with `GIMPLE_BIND' 13117 statement `G'. This is analogous to the `BIND_EXPR_BLOCK' field in 13118 trees. 13119 13120 -- GIMPLE function: void gimple_bind_set_block (gimple g, tree block) 13121 Set `BLOCK' to be the `TREE_BLOCK' node associated with 13122 `GIMPLE_BIND' statement `G'. 13123 13124 13125 File: gccint.info, Node: `GIMPLE_CALL', Next: `GIMPLE_CATCH', Prev: `GIMPLE_BIND', Up: Tuple specific accessors 13126 13127 12.7.4 `GIMPLE_CALL' 13128 -------------------- 13129 13130 -- GIMPLE function: gimple gimple_build_call (tree fn, unsigned nargs, 13131 ...) 13132 Build a `GIMPLE_CALL' statement to function `FN'. The argument 13133 `FN' must be either a `FUNCTION_DECL' or a gimple call address as 13134 determined by `is_gimple_call_addr'. `NARGS' are the number of 13135 arguments. The rest of the arguments follow the argument `NARGS', 13136 and must be trees that are valid as rvalues in gimple (i.e., each 13137 operand is validated with `is_gimple_operand'). 13138 13139 -- GIMPLE function: gimple gimple_build_call_from_tree (tree call_expr) 13140 Build a `GIMPLE_CALL' from a `CALL_EXPR' node. The arguments and 13141 the function are taken from the expression directly. This routine 13142 assumes that `call_expr' is already in GIMPLE form. That is, its 13143 operands are GIMPLE values and the function call needs no further 13144 simplification. All the call flags in `call_expr' are copied over 13145 to the new `GIMPLE_CALL'. 13146 13147 -- GIMPLE function: gimple gimple_build_call_vec (tree fn, `VEC'(tree, 13148 heap) *args) 13149 Identical to `gimple_build_call' but the arguments are stored in a 13150 `VEC'(). 13151 13152 -- GIMPLE function: tree gimple_call_lhs (gimple g) 13153 Return the `LHS' of call statement `G'. 13154 13155 -- GIMPLE function: tree *gimple_call_lhs_ptr (gimple g) 13156 Return a pointer to the `LHS' of call statement `G'. 13157 13158 -- GIMPLE function: void gimple_call_set_lhs (gimple g, tree lhs) 13159 Set `LHS' to be the `LHS' operand of call statement `G'. 13160 13161 -- GIMPLE function: tree gimple_call_fn (gimple g) 13162 Return the tree node representing the function called by call 13163 statement `G'. 13164 13165 -- GIMPLE function: void gimple_call_set_fn (gimple g, tree fn) 13166 Set `FN' to be the function called by call statement `G'. This has 13167 to be a gimple value specifying the address of the called function. 13168 13169 -- GIMPLE function: tree gimple_call_fndecl (gimple g) 13170 If a given `GIMPLE_CALL''s callee is a `FUNCTION_DECL', return it. 13171 Otherwise return `NULL'. This function is analogous to 13172 `get_callee_fndecl' in `GENERIC'. 13173 13174 -- GIMPLE function: tree gimple_call_set_fndecl (gimple g, tree fndecl) 13175 Set the called function to `FNDECL'. 13176 13177 -- GIMPLE function: tree gimple_call_return_type (gimple g) 13178 Return the type returned by call statement `G'. 13179 13180 -- GIMPLE function: tree gimple_call_chain (gimple g) 13181 Return the static chain for call statement `G'. 13182 13183 -- GIMPLE function: void gimple_call_set_chain (gimple g, tree chain) 13184 Set `CHAIN' to be the static chain for call statement `G'. 13185 13186 -- GIMPLE function: gimple_call_num_args (gimple g) 13187 Return the number of arguments used by call statement `G'. 13188 13189 -- GIMPLE function: tree gimple_call_arg (gimple g, unsigned index) 13190 Return the argument at position `INDEX' for call statement `G'. 13191 The first argument is 0. 13192 13193 -- GIMPLE function: tree *gimple_call_arg_ptr (gimple g, unsigned 13194 index) 13195 Return a pointer to the argument at position `INDEX' for call 13196 statement `G'. 13197 13198 -- GIMPLE function: void gimple_call_set_arg (gimple g, unsigned 13199 index, tree arg) 13200 Set `ARG' to be the argument at position `INDEX' for call statement 13201 `G'. 13202 13203 -- GIMPLE function: void gimple_call_set_tail (gimple s) 13204 Mark call statement `S' as being a tail call (i.e., a call just 13205 before the exit of a function). These calls are candidate for tail 13206 call optimization. 13207 13208 -- GIMPLE function: bool gimple_call_tail_p (gimple s) 13209 Return true if `GIMPLE_CALL' `S' is marked as a tail call. 13210 13211 -- GIMPLE function: void gimple_call_mark_uninlinable (gimple s) 13212 Mark `GIMPLE_CALL' `S' as being uninlinable. 13213 13214 -- GIMPLE function: bool gimple_call_cannot_inline_p (gimple s) 13215 Return true if `GIMPLE_CALL' `S' cannot be inlined. 13216 13217 -- GIMPLE function: bool gimple_call_noreturn_p (gimple s) 13218 Return true if `S' is a noreturn call. 13219 13220 -- GIMPLE function: gimple gimple_call_copy_skip_args (gimple stmt, 13221 bitmap args_to_skip) 13222 Build a `GIMPLE_CALL' identical to `STMT' but skipping the 13223 arguments in the positions marked by the set `ARGS_TO_SKIP'. 13224 13225 13226 File: gccint.info, Node: `GIMPLE_CATCH', Next: `GIMPLE_CHANGE_DYNAMIC_TYPE', Prev: `GIMPLE_CALL', Up: Tuple specific accessors 13227 13228 12.7.5 `GIMPLE_CATCH' 13229 --------------------- 13230 13231 -- GIMPLE function: gimple gimple_build_catch (tree types, gimple_seq 13232 handler) 13233 Build a `GIMPLE_CATCH' statement. `TYPES' are the tree types this 13234 catch handles. `HANDLER' is a sequence of statements with the code 13235 for the handler. 13236 13237 -- GIMPLE function: tree gimple_catch_types (gimple g) 13238 Return the types handled by `GIMPLE_CATCH' statement `G'. 13239 13240 -- GIMPLE function: tree *gimple_catch_types_ptr (gimple g) 13241 Return a pointer to the types handled by `GIMPLE_CATCH' statement 13242 `G'. 13243 13244 -- GIMPLE function: gimple_seq gimple_catch_handler (gimple g) 13245 Return the GIMPLE sequence representing the body of the handler of 13246 `GIMPLE_CATCH' statement `G'. 13247 13248 -- GIMPLE function: void gimple_catch_set_types (gimple g, tree t) 13249 Set `T' to be the set of types handled by `GIMPLE_CATCH' `G'. 13250 13251 -- GIMPLE function: void gimple_catch_set_handler (gimple g, 13252 gimple_seq handler) 13253 Set `HANDLER' to be the body of `GIMPLE_CATCH' `G'. 13254 13255 13256 File: gccint.info, Node: `GIMPLE_CHANGE_DYNAMIC_TYPE', Next: `GIMPLE_COND', Prev: `GIMPLE_CATCH', Up: Tuple specific accessors 13257 13258 12.7.6 `GIMPLE_CHANGE_DYNAMIC_TYPE' 13259 ----------------------------------- 13260 13261 -- GIMPLE function: gimple gimple_build_cdt (tree type, tree ptr) 13262 Build a `GIMPLE_CHANGE_DYNAMIC_TYPE' statement. `TYPE' is the new 13263 type for the location `PTR'. 13264 13265 -- GIMPLE function: tree gimple_cdt_new_type (gimple g) 13266 Return the new type set by `GIMPLE_CHANGE_DYNAMIC_TYPE' statement 13267 `G'. 13268 13269 -- GIMPLE function: tree *gimple_cdt_new_type_ptr (gimple g) 13270 Return a pointer to the new type set by 13271 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13272 13273 -- GIMPLE function: void gimple_cdt_set_new_type (gimple g, tree 13274 new_type) 13275 Set `NEW_TYPE' to be the type returned by 13276 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13277 13278 -- GIMPLE function: tree gimple_cdt_location (gimple g) 13279 Return the location affected by `GIMPLE_CHANGE_DYNAMIC_TYPE' 13280 statement `G'. 13281 13282 -- GIMPLE function: tree *gimple_cdt_location_ptr (gimple g) 13283 Return a pointer to the location affected by 13284 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13285 13286 -- GIMPLE function: void gimple_cdt_set_location (gimple g, tree ptr) 13287 Set `PTR' to be the location affected by 13288 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13289 13290 13291 File: gccint.info, Node: `GIMPLE_COND', Next: `GIMPLE_EH_FILTER', Prev: `GIMPLE_CHANGE_DYNAMIC_TYPE', Up: Tuple specific accessors 13292 13293 12.7.7 `GIMPLE_COND' 13294 -------------------- 13295 13296 -- GIMPLE function: gimple gimple_build_cond (enum tree_code 13297 pred_code, tree lhs, tree rhs, tree t_label, tree f_label) 13298 Build a `GIMPLE_COND' statement. `A' `GIMPLE_COND' statement 13299 compares `LHS' and `RHS' and if the condition in `PRED_CODE' is 13300 true, jump to the label in `t_label', otherwise jump to the label 13301 in `f_label'. `PRED_CODE' are relational operator tree codes like 13302 `EQ_EXPR', `LT_EXPR', `LE_EXPR', `NE_EXPR', etc. 13303 13304 -- GIMPLE function: gimple gimple_build_cond_from_tree (tree cond, 13305 tree t_label, tree f_label) 13306 Build a `GIMPLE_COND' statement from the conditional expression 13307 tree `COND'. `T_LABEL' and `F_LABEL' are as in 13308 `gimple_build_cond'. 13309 13310 -- GIMPLE function: enum tree_code gimple_cond_code (gimple g) 13311 Return the code of the predicate computed by conditional statement 13312 `G'. 13313 13314 -- GIMPLE function: void gimple_cond_set_code (gimple g, enum 13315 tree_code code) 13316 Set `CODE' to be the predicate code for the conditional statement 13317 `G'. 13318 13319 -- GIMPLE function: tree gimple_cond_lhs (gimple g) 13320 Return the `LHS' of the predicate computed by conditional statement 13321 `G'. 13322 13323 -- GIMPLE function: void gimple_cond_set_lhs (gimple g, tree lhs) 13324 Set `LHS' to be the `LHS' operand of the predicate computed by 13325 conditional statement `G'. 13326 13327 -- GIMPLE function: tree gimple_cond_rhs (gimple g) 13328 Return the `RHS' operand of the predicate computed by conditional 13329 `G'. 13330 13331 -- GIMPLE function: void gimple_cond_set_rhs (gimple g, tree rhs) 13332 Set `RHS' to be the `RHS' operand of the predicate computed by 13333 conditional statement `G'. 13334 13335 -- GIMPLE function: tree gimple_cond_true_label (gimple g) 13336 Return the label used by conditional statement `G' when its 13337 predicate evaluates to true. 13338 13339 -- GIMPLE function: void gimple_cond_set_true_label (gimple g, tree 13340 label) 13341 Set `LABEL' to be the label used by conditional statement `G' when 13342 its predicate evaluates to true. 13343 13344 -- GIMPLE function: void gimple_cond_set_false_label (gimple g, tree 13345 label) 13346 Set `LABEL' to be the label used by conditional statement `G' when 13347 its predicate evaluates to false. 13348 13349 -- GIMPLE function: tree gimple_cond_false_label (gimple g) 13350 Return the label used by conditional statement `G' when its 13351 predicate evaluates to false. 13352 13353 -- GIMPLE function: void gimple_cond_make_false (gimple g) 13354 Set the conditional `COND_STMT' to be of the form 'if (1 == 0)'. 13355 13356 -- GIMPLE function: void gimple_cond_make_true (gimple g) 13357 Set the conditional `COND_STMT' to be of the form 'if (1 == 1)'. 13358 13359 13360 File: gccint.info, Node: `GIMPLE_EH_FILTER', Next: `GIMPLE_LABEL', Prev: `GIMPLE_COND', Up: Tuple specific accessors 13361 13362 12.7.8 `GIMPLE_EH_FILTER' 13363 ------------------------- 13364 13365 -- GIMPLE function: gimple gimple_build_eh_filter (tree types, 13366 gimple_seq failure) 13367 Build a `GIMPLE_EH_FILTER' statement. `TYPES' are the filter's 13368 types. `FAILURE' is a sequence with the filter's failure action. 13369 13370 -- GIMPLE function: tree gimple_eh_filter_types (gimple g) 13371 Return the types handled by `GIMPLE_EH_FILTER' statement `G'. 13372 13373 -- GIMPLE function: tree *gimple_eh_filter_types_ptr (gimple g) 13374 Return a pointer to the types handled by `GIMPLE_EH_FILTER' 13375 statement `G'. 13376 13377 -- GIMPLE function: gimple_seq gimple_eh_filter_failure (gimple g) 13378 Return the sequence of statement to execute when `GIMPLE_EH_FILTER' 13379 statement fails. 13380 13381 -- GIMPLE function: void gimple_eh_filter_set_types (gimple g, tree 13382 types) 13383 Set `TYPES' to be the set of types handled by `GIMPLE_EH_FILTER' 13384 `G'. 13385 13386 -- GIMPLE function: void gimple_eh_filter_set_failure (gimple g, 13387 gimple_seq failure) 13388 Set `FAILURE' to be the sequence of statements to execute on 13389 failure for `GIMPLE_EH_FILTER' `G'. 13390 13391 -- GIMPLE function: bool gimple_eh_filter_must_not_throw (gimple g) 13392 Return the `EH_FILTER_MUST_NOT_THROW' flag. 13393 13394 -- GIMPLE function: void gimple_eh_filter_set_must_not_throw (gimple 13395 g, bool mntp) 13396 Set the `EH_FILTER_MUST_NOT_THROW' flag. 13397 13398 13399 File: gccint.info, Node: `GIMPLE_LABEL', Next: `GIMPLE_NOP', Prev: `GIMPLE_EH_FILTER', Up: Tuple specific accessors 13400 13401 12.7.9 `GIMPLE_LABEL' 13402 --------------------- 13403 13404 -- GIMPLE function: gimple gimple_build_label (tree label) 13405 Build a `GIMPLE_LABEL' statement with corresponding to the tree 13406 label, `LABEL'. 13407 13408 -- GIMPLE function: tree gimple_label_label (gimple g) 13409 Return the `LABEL_DECL' node used by `GIMPLE_LABEL' statement `G'. 13410 13411 -- GIMPLE function: void gimple_label_set_label (gimple g, tree label) 13412 Set `LABEL' to be the `LABEL_DECL' node used by `GIMPLE_LABEL' 13413 statement `G'. 13414 13415 -- GIMPLE function: gimple gimple_build_goto (tree dest) 13416 Build a `GIMPLE_GOTO' statement to label `DEST'. 13417 13418 -- GIMPLE function: tree gimple_goto_dest (gimple g) 13419 Return the destination of the unconditional jump `G'. 13420 13421 -- GIMPLE function: void gimple_goto_set_dest (gimple g, tree dest) 13422 Set `DEST' to be the destination of the unconditional jump `G'. 13423 13424 13425 File: gccint.info, Node: `GIMPLE_NOP', Next: `GIMPLE_OMP_ATOMIC_LOAD', Prev: `GIMPLE_LABEL', Up: Tuple specific accessors 13426 13427 12.7.10 `GIMPLE_NOP' 13428 -------------------- 13429 13430 -- GIMPLE function: gimple gimple_build_nop (void) 13431 Build a `GIMPLE_NOP' statement. 13432 13433 -- GIMPLE function: bool gimple_nop_p (gimple g) 13434 Returns `TRUE' if statement `G' is a `GIMPLE_NOP'. 13435 13436 13437 File: gccint.info, Node: `GIMPLE_OMP_ATOMIC_LOAD', Next: `GIMPLE_OMP_ATOMIC_STORE', Prev: `GIMPLE_NOP', Up: Tuple specific accessors 13438 13439 12.7.11 `GIMPLE_OMP_ATOMIC_LOAD' 13440 -------------------------------- 13441 13442 -- GIMPLE function: gimple gimple_build_omp_atomic_load (tree lhs, 13443 tree rhs) 13444 Build a `GIMPLE_OMP_ATOMIC_LOAD' statement. `LHS' is the left-hand 13445 side of the assignment. `RHS' is the right-hand side of the 13446 assignment. 13447 13448 -- GIMPLE function: void gimple_omp_atomic_load_set_lhs (gimple g, 13449 tree lhs) 13450 Set the `LHS' of an atomic load. 13451 13452 -- GIMPLE function: tree gimple_omp_atomic_load_lhs (gimple g) 13453 Get the `LHS' of an atomic load. 13454 13455 -- GIMPLE function: void gimple_omp_atomic_load_set_rhs (gimple g, 13456 tree rhs) 13457 Set the `RHS' of an atomic set. 13458 13459 -- GIMPLE function: tree gimple_omp_atomic_load_rhs (gimple g) 13460 Get the `RHS' of an atomic set. 13461 13462 13463 File: gccint.info, Node: `GIMPLE_OMP_ATOMIC_STORE', Next: `GIMPLE_OMP_CONTINUE', Prev: `GIMPLE_OMP_ATOMIC_LOAD', Up: Tuple specific accessors 13464 13465 12.7.12 `GIMPLE_OMP_ATOMIC_STORE' 13466 --------------------------------- 13467 13468 -- GIMPLE function: gimple gimple_build_omp_atomic_store (tree val) 13469 Build a `GIMPLE_OMP_ATOMIC_STORE' statement. `VAL' is the value to 13470 be stored. 13471 13472 -- GIMPLE function: void gimple_omp_atomic_store_set_val (gimple g, 13473 tree val) 13474 Set the value being stored in an atomic store. 13475 13476 -- GIMPLE function: tree gimple_omp_atomic_store_val (gimple g) 13477 Return the value being stored in an atomic store. 13478 13479 13480 File: gccint.info, Node: `GIMPLE_OMP_CONTINUE', Next: `GIMPLE_OMP_CRITICAL', Prev: `GIMPLE_OMP_ATOMIC_STORE', Up: Tuple specific accessors 13481 13482 12.7.13 `GIMPLE_OMP_CONTINUE' 13483 ----------------------------- 13484 13485 -- GIMPLE function: gimple gimple_build_omp_continue (tree 13486 control_def, tree control_use) 13487 Build a `GIMPLE_OMP_CONTINUE' statement. `CONTROL_DEF' is the 13488 definition of the control variable. `CONTROL_USE' is the use of 13489 the control variable. 13490 13491 -- GIMPLE function: tree gimple_omp_continue_control_def (gimple s) 13492 Return the definition of the control variable on a 13493 `GIMPLE_OMP_CONTINUE' in `S'. 13494 13495 -- GIMPLE function: tree gimple_omp_continue_control_def_ptr (gimple s) 13496 Same as above, but return the pointer. 13497 13498 -- GIMPLE function: tree gimple_omp_continue_set_control_def (gimple s) 13499 Set the control variable definition for a `GIMPLE_OMP_CONTINUE' 13500 statement in `S'. 13501 13502 -- GIMPLE function: tree gimple_omp_continue_control_use (gimple s) 13503 Return the use of the control variable on a `GIMPLE_OMP_CONTINUE' 13504 in `S'. 13505 13506 -- GIMPLE function: tree gimple_omp_continue_control_use_ptr (gimple s) 13507 Same as above, but return the pointer. 13508 13509 -- GIMPLE function: tree gimple_omp_continue_set_control_use (gimple s) 13510 Set the control variable use for a `GIMPLE_OMP_CONTINUE' statement 13511 in `S'. 13512 13513 13514 File: gccint.info, Node: `GIMPLE_OMP_CRITICAL', Next: `GIMPLE_OMP_FOR', Prev: `GIMPLE_OMP_CONTINUE', Up: Tuple specific accessors 13515 13516 12.7.14 `GIMPLE_OMP_CRITICAL' 13517 ----------------------------- 13518 13519 -- GIMPLE function: gimple gimple_build_omp_critical (gimple_seq body, 13520 tree name) 13521 Build a `GIMPLE_OMP_CRITICAL' statement. `BODY' is the sequence of 13522 statements for which only one thread can execute. `NAME' is an 13523 optional identifier for this critical block. 13524 13525 -- GIMPLE function: tree gimple_omp_critical_name (gimple g) 13526 Return the name associated with `OMP_CRITICAL' statement `G'. 13527 13528 -- GIMPLE function: tree *gimple_omp_critical_name_ptr (gimple g) 13529 Return a pointer to the name associated with `OMP' critical 13530 statement `G'. 13531 13532 -- GIMPLE function: void gimple_omp_critical_set_name (gimple g, tree 13533 name) 13534 Set `NAME' to be the name associated with `OMP' critical statement 13535 `G'. 13536 13537 13538 File: gccint.info, Node: `GIMPLE_OMP_FOR', Next: `GIMPLE_OMP_MASTER', Prev: `GIMPLE_OMP_CRITICAL', Up: Tuple specific accessors 13539 13540 12.7.15 `GIMPLE_OMP_FOR' 13541 ------------------------ 13542 13543 -- GIMPLE function: gimple gimple_build_omp_for (gimple_seq body, tree 13544 clauses, tree index, tree initial, tree final, tree incr, 13545 gimple_seq pre_body, enum tree_code omp_for_cond) 13546 Build a `GIMPLE_OMP_FOR' statement. `BODY' is sequence of 13547 statements inside the for loop. `CLAUSES', are any of the `OMP' 13548 loop construct's clauses: private, firstprivate, lastprivate, 13549 reductions, ordered, schedule, and nowait. `PRE_BODY' is the 13550 sequence of statements that are loop invariant. `INDEX' is the 13551 index variable. `INITIAL' is the initial value of `INDEX'. 13552 `FINAL' is final value of `INDEX'. OMP_FOR_COND is the predicate 13553 used to compare `INDEX' and `FINAL'. `INCR' is the increment 13554 expression. 13555 13556 -- GIMPLE function: tree gimple_omp_for_clauses (gimple g) 13557 Return the clauses associated with `OMP_FOR' `G'. 13558 13559 -- GIMPLE function: tree *gimple_omp_for_clauses_ptr (gimple g) 13560 Return a pointer to the `OMP_FOR' `G'. 13561 13562 -- GIMPLE function: void gimple_omp_for_set_clauses (gimple g, tree 13563 clauses) 13564 Set `CLAUSES' to be the list of clauses associated with `OMP_FOR' 13565 `G'. 13566 13567 -- GIMPLE function: tree gimple_omp_for_index (gimple g) 13568 Return the index variable for `OMP_FOR' `G'. 13569 13570 -- GIMPLE function: tree *gimple_omp_for_index_ptr (gimple g) 13571 Return a pointer to the index variable for `OMP_FOR' `G'. 13572 13573 -- GIMPLE function: void gimple_omp_for_set_index (gimple g, tree 13574 index) 13575 Set `INDEX' to be the index variable for `OMP_FOR' `G'. 13576 13577 -- GIMPLE function: tree gimple_omp_for_initial (gimple g) 13578 Return the initial value for `OMP_FOR' `G'. 13579 13580 -- GIMPLE function: tree *gimple_omp_for_initial_ptr (gimple g) 13581 Return a pointer to the initial value for `OMP_FOR' `G'. 13582 13583 -- GIMPLE function: void gimple_omp_for_set_initial (gimple g, tree 13584 initial) 13585 Set `INITIAL' to be the initial value for `OMP_FOR' `G'. 13586 13587 -- GIMPLE function: tree gimple_omp_for_final (gimple g) 13588 Return the final value for `OMP_FOR' `G'. 13589 13590 -- GIMPLE function: tree *gimple_omp_for_final_ptr (gimple g) 13591 turn a pointer to the final value for `OMP_FOR' `G'. 13592 13593 -- GIMPLE function: void gimple_omp_for_set_final (gimple g, tree 13594 final) 13595 Set `FINAL' to be the final value for `OMP_FOR' `G'. 13596 13597 -- GIMPLE function: tree gimple_omp_for_incr (gimple g) 13598 Return the increment value for `OMP_FOR' `G'. 13599 13600 -- GIMPLE function: tree *gimple_omp_for_incr_ptr (gimple g) 13601 Return a pointer to the increment value for `OMP_FOR' `G'. 13602 13603 -- GIMPLE function: void gimple_omp_for_set_incr (gimple g, tree incr) 13604 Set `INCR' to be the increment value for `OMP_FOR' `G'. 13605 13606 -- GIMPLE function: gimple_seq gimple_omp_for_pre_body (gimple g) 13607 Return the sequence of statements to execute before the `OMP_FOR' 13608 statement `G' starts. 13609 13610 -- GIMPLE function: void gimple_omp_for_set_pre_body (gimple g, 13611 gimple_seq pre_body) 13612 Set `PRE_BODY' to be the sequence of statements to execute before 13613 the `OMP_FOR' statement `G' starts. 13614 13615 -- GIMPLE function: void gimple_omp_for_set_cond (gimple g, enum 13616 tree_code cond) 13617 Set `COND' to be the condition code for `OMP_FOR' `G'. 13618 13619 -- GIMPLE function: enum tree_code gimple_omp_for_cond (gimple g) 13620 Return the condition code associated with `OMP_FOR' `G'. 13621 13622 13623 File: gccint.info, Node: `GIMPLE_OMP_MASTER', Next: `GIMPLE_OMP_ORDERED', Prev: `GIMPLE_OMP_FOR', Up: Tuple specific accessors 13624 13625 12.7.16 `GIMPLE_OMP_MASTER' 13626 --------------------------- 13627 13628 -- GIMPLE function: gimple gimple_build_omp_master (gimple_seq body) 13629 Build a `GIMPLE_OMP_MASTER' statement. `BODY' is the sequence of 13630 statements to be executed by just the master. 13631 13632 13633 File: gccint.info, Node: `GIMPLE_OMP_ORDERED', Next: `GIMPLE_OMP_PARALLEL', Prev: `GIMPLE_OMP_MASTER', Up: Tuple specific accessors 13634 13635 12.7.17 `GIMPLE_OMP_ORDERED' 13636 ---------------------------- 13637 13638 -- GIMPLE function: gimple gimple_build_omp_ordered (gimple_seq body) 13639 Build a `GIMPLE_OMP_ORDERED' statement. 13640 13641 `BODY' is the sequence of statements inside a loop that will executed 13642 in sequence. 13643 13644 13645 File: gccint.info, Node: `GIMPLE_OMP_PARALLEL', Next: `GIMPLE_OMP_RETURN', Prev: `GIMPLE_OMP_ORDERED', Up: Tuple specific accessors 13646 13647 12.7.18 `GIMPLE_OMP_PARALLEL' 13648 ----------------------------- 13649 13650 -- GIMPLE function: gimple gimple_build_omp_parallel (gimple_seq body, 13651 tree clauses, tree child_fn, tree data_arg) 13652 Build a `GIMPLE_OMP_PARALLEL' statement. 13653 13654 `BODY' is sequence of statements which are executed in parallel. 13655 `CLAUSES', are the `OMP' parallel construct's clauses. `CHILD_FN' is 13656 the function created for the parallel threads to execute. `DATA_ARG' 13657 are the shared data argument(s). 13658 13659 -- GIMPLE function: bool gimple_omp_parallel_combined_p (gimple g) 13660 Return true if `OMP' parallel statement `G' has the 13661 `GF_OMP_PARALLEL_COMBINED' flag set. 13662 13663 -- GIMPLE function: void gimple_omp_parallel_set_combined_p (gimple g) 13664 Set the `GF_OMP_PARALLEL_COMBINED' field in `OMP' parallel 13665 statement `G'. 13666 13667 -- GIMPLE function: gimple_seq gimple_omp_body (gimple g) 13668 Return the body for the `OMP' statement `G'. 13669 13670 -- GIMPLE function: void gimple_omp_set_body (gimple g, gimple_seq 13671 body) 13672 Set `BODY' to be the body for the `OMP' statement `G'. 13673 13674 -- GIMPLE function: tree gimple_omp_parallel_clauses (gimple g) 13675 Return the clauses associated with `OMP_PARALLEL' `G'. 13676 13677 -- GIMPLE function: tree *gimple_omp_parallel_clauses_ptr (gimple g) 13678 Return a pointer to the clauses associated with `OMP_PARALLEL' `G'. 13679 13680 -- GIMPLE function: void gimple_omp_parallel_set_clauses (gimple g, 13681 tree clauses) 13682 Set `CLAUSES' to be the list of clauses associated with 13683 `OMP_PARALLEL' `G'. 13684 13685 -- GIMPLE function: tree gimple_omp_parallel_child_fn (gimple g) 13686 Return the child function used to hold the body of `OMP_PARALLEL' 13687 `G'. 13688 13689 -- GIMPLE function: tree *gimple_omp_parallel_child_fn_ptr (gimple g) 13690 Return a pointer to the child function used to hold the body of 13691 `OMP_PARALLEL' `G'. 13692 13693 -- GIMPLE function: void gimple_omp_parallel_set_child_fn (gimple g, 13694 tree child_fn) 13695 Set `CHILD_FN' to be the child function for `OMP_PARALLEL' `G'. 13696 13697 -- GIMPLE function: tree gimple_omp_parallel_data_arg (gimple g) 13698 Return the artificial argument used to send variables and values 13699 from the parent to the children threads in `OMP_PARALLEL' `G'. 13700 13701 -- GIMPLE function: tree *gimple_omp_parallel_data_arg_ptr (gimple g) 13702 Return a pointer to the data argument for `OMP_PARALLEL' `G'. 13703 13704 -- GIMPLE function: void gimple_omp_parallel_set_data_arg (gimple g, 13705 tree data_arg) 13706 Set `DATA_ARG' to be the data argument for `OMP_PARALLEL' `G'. 13707 13708 -- GIMPLE function: bool is_gimple_omp (gimple stmt) 13709 Returns true when the gimple statement `STMT' is any of the OpenMP 13710 types. 13711 13712 13713 File: gccint.info, Node: `GIMPLE_OMP_RETURN', Next: `GIMPLE_OMP_SECTION', Prev: `GIMPLE_OMP_PARALLEL', Up: Tuple specific accessors 13714 13715 12.7.19 `GIMPLE_OMP_RETURN' 13716 --------------------------- 13717 13718 -- GIMPLE function: gimple gimple_build_omp_return (bool wait_p) 13719 Build a `GIMPLE_OMP_RETURN' statement. `WAIT_P' is true if this is 13720 a non-waiting return. 13721 13722 -- GIMPLE function: void gimple_omp_return_set_nowait (gimple s) 13723 Set the nowait flag on `GIMPLE_OMP_RETURN' statement `S'. 13724 13725 -- GIMPLE function: bool gimple_omp_return_nowait_p (gimple g) 13726 Return true if `OMP' return statement `G' has the 13727 `GF_OMP_RETURN_NOWAIT' flag set. 13728 13729 13730 File: gccint.info, Node: `GIMPLE_OMP_SECTION', Next: `GIMPLE_OMP_SECTIONS', Prev: `GIMPLE_OMP_RETURN', Up: Tuple specific accessors 13731 13732 12.7.20 `GIMPLE_OMP_SECTION' 13733 ---------------------------- 13734 13735 -- GIMPLE function: gimple gimple_build_omp_section (gimple_seq body) 13736 Build a `GIMPLE_OMP_SECTION' statement for a sections statement. 13737 13738 `BODY' is the sequence of statements in the section. 13739 13740 -- GIMPLE function: bool gimple_omp_section_last_p (gimple g) 13741 Return true if `OMP' section statement `G' has the 13742 `GF_OMP_SECTION_LAST' flag set. 13743 13744 -- GIMPLE function: void gimple_omp_section_set_last (gimple g) 13745 Set the `GF_OMP_SECTION_LAST' flag on `G'. 13746 13747 13748 File: gccint.info, Node: `GIMPLE_OMP_SECTIONS', Next: `GIMPLE_OMP_SINGLE', Prev: `GIMPLE_OMP_SECTION', Up: Tuple specific accessors 13749 13750 12.7.21 `GIMPLE_OMP_SECTIONS' 13751 ----------------------------- 13752 13753 -- GIMPLE function: gimple gimple_build_omp_sections (gimple_seq body, 13754 tree clauses) 13755 Build a `GIMPLE_OMP_SECTIONS' statement. `BODY' is a sequence of 13756 section statements. `CLAUSES' are any of the `OMP' sections 13757 construct's clauses: private, firstprivate, lastprivate, 13758 reduction, and nowait. 13759 13760 -- GIMPLE function: gimple gimple_build_omp_sections_switch (void) 13761 Build a `GIMPLE_OMP_SECTIONS_SWITCH' statement. 13762 13763 -- GIMPLE function: tree gimple_omp_sections_control (gimple g) 13764 Return the control variable associated with the 13765 `GIMPLE_OMP_SECTIONS' in `G'. 13766 13767 -- GIMPLE function: tree *gimple_omp_sections_control_ptr (gimple g) 13768 Return a pointer to the clauses associated with the 13769 `GIMPLE_OMP_SECTIONS' in `G'. 13770 13771 -- GIMPLE function: void gimple_omp_sections_set_control (gimple g, 13772 tree control) 13773 Set `CONTROL' to be the set of clauses associated with the 13774 `GIMPLE_OMP_SECTIONS' in `G'. 13775 13776 -- GIMPLE function: tree gimple_omp_sections_clauses (gimple g) 13777 Return the clauses associated with `OMP_SECTIONS' `G'. 13778 13779 -- GIMPLE function: tree *gimple_omp_sections_clauses_ptr (gimple g) 13780 Return a pointer to the clauses associated with `OMP_SECTIONS' `G'. 13781 13782 -- GIMPLE function: void gimple_omp_sections_set_clauses (gimple g, 13783 tree clauses) 13784 Set `CLAUSES' to be the set of clauses associated with 13785 `OMP_SECTIONS' `G'. 13786 13787 13788 File: gccint.info, Node: `GIMPLE_OMP_SINGLE', Next: `GIMPLE_PHI', Prev: `GIMPLE_OMP_SECTIONS', Up: Tuple specific accessors 13789 13790 12.7.22 `GIMPLE_OMP_SINGLE' 13791 --------------------------- 13792 13793 -- GIMPLE function: gimple gimple_build_omp_single (gimple_seq body, 13794 tree clauses) 13795 Build a `GIMPLE_OMP_SINGLE' statement. `BODY' is the sequence of 13796 statements that will be executed once. `CLAUSES' are any of the 13797 `OMP' single construct's clauses: private, firstprivate, 13798 copyprivate, nowait. 13799 13800 -- GIMPLE function: tree gimple_omp_single_clauses (gimple g) 13801 Return the clauses associated with `OMP_SINGLE' `G'. 13802 13803 -- GIMPLE function: tree *gimple_omp_single_clauses_ptr (gimple g) 13804 Return a pointer to the clauses associated with `OMP_SINGLE' `G'. 13805 13806 -- GIMPLE function: void gimple_omp_single_set_clauses (gimple g, tree 13807 clauses) 13808 Set `CLAUSES' to be the clauses associated with `OMP_SINGLE' `G'. 13809 13810 13811 File: gccint.info, Node: `GIMPLE_PHI', Next: `GIMPLE_RESX', Prev: `GIMPLE_OMP_SINGLE', Up: Tuple specific accessors 13812 13813 12.7.23 `GIMPLE_PHI' 13814 -------------------- 13815 13816 -- GIMPLE function: gimple make_phi_node (tree var, int len) 13817 Build a `PHI' node with len argument slots for variable var. 13818 13819 -- GIMPLE function: unsigned gimple_phi_capacity (gimple g) 13820 Return the maximum number of arguments supported by `GIMPLE_PHI' 13821 `G'. 13822 13823 -- GIMPLE function: unsigned gimple_phi_num_args (gimple g) 13824 Return the number of arguments in `GIMPLE_PHI' `G'. This must 13825 always be exactly the number of incoming edges for the basic block 13826 holding `G'. 13827 13828 -- GIMPLE function: tree gimple_phi_result (gimple g) 13829 Return the `SSA' name created by `GIMPLE_PHI' `G'. 13830 13831 -- GIMPLE function: tree *gimple_phi_result_ptr (gimple g) 13832 Return a pointer to the `SSA' name created by `GIMPLE_PHI' `G'. 13833 13834 -- GIMPLE function: void gimple_phi_set_result (gimple g, tree result) 13835 Set `RESULT' to be the `SSA' name created by `GIMPLE_PHI' `G'. 13836 13837 -- GIMPLE function: struct phi_arg_d *gimple_phi_arg (gimple g, index) 13838 Return the `PHI' argument corresponding to incoming edge `INDEX' 13839 for `GIMPLE_PHI' `G'. 13840 13841 -- GIMPLE function: void gimple_phi_set_arg (gimple g, index, struct 13842 phi_arg_d * phiarg) 13843 Set `PHIARG' to be the argument corresponding to incoming edge 13844 `INDEX' for `GIMPLE_PHI' `G'. 13845 13846 13847 File: gccint.info, Node: `GIMPLE_RESX', Next: `GIMPLE_RETURN', Prev: `GIMPLE_PHI', Up: Tuple specific accessors 13848 13849 12.7.24 `GIMPLE_RESX' 13850 --------------------- 13851 13852 -- GIMPLE function: gimple gimple_build_resx (int region) 13853 Build a `GIMPLE_RESX' statement which is a statement. This 13854 statement is a placeholder for _Unwind_Resume before we know if a 13855 function call or a branch is needed. `REGION' is the exception 13856 region from which control is flowing. 13857 13858 -- GIMPLE function: int gimple_resx_region (gimple g) 13859 Return the region number for `GIMPLE_RESX' `G'. 13860 13861 -- GIMPLE function: void gimple_resx_set_region (gimple g, int region) 13862 Set `REGION' to be the region number for `GIMPLE_RESX' `G'. 13863 13864 13865 File: gccint.info, Node: `GIMPLE_RETURN', Next: `GIMPLE_SWITCH', Prev: `GIMPLE_RESX', Up: Tuple specific accessors 13866 13867 12.7.25 `GIMPLE_RETURN' 13868 ----------------------- 13869 13870 -- GIMPLE function: gimple gimple_build_return (tree retval) 13871 Build a `GIMPLE_RETURN' statement whose return value is retval. 13872 13873 -- GIMPLE function: tree gimple_return_retval (gimple g) 13874 Return the return value for `GIMPLE_RETURN' `G'. 13875 13876 -- GIMPLE function: void gimple_return_set_retval (gimple g, tree 13877 retval) 13878 Set `RETVAL' to be the return value for `GIMPLE_RETURN' `G'. 13879 13880 13881 File: gccint.info, Node: `GIMPLE_SWITCH', Next: `GIMPLE_TRY', Prev: `GIMPLE_RETURN', Up: Tuple specific accessors 13882 13883 12.7.26 `GIMPLE_SWITCH' 13884 ----------------------- 13885 13886 -- GIMPLE function: gimple gimple_build_switch ( nlabels, tree index, 13887 tree default_label, ...) 13888 Build a `GIMPLE_SWITCH' statement. `NLABELS' are the number of 13889 labels excluding the default label. The default label is passed 13890 in `DEFAULT_LABEL'. The rest of the arguments are trees 13891 representing the labels. Each label is a tree of code 13892 `CASE_LABEL_EXPR'. 13893 13894 -- GIMPLE function: gimple gimple_build_switch_vec (tree index, tree 13895 default_label, `VEC'(tree,heap) *args) 13896 This function is an alternate way of building `GIMPLE_SWITCH' 13897 statements. `INDEX' and `DEFAULT_LABEL' are as in 13898 gimple_build_switch. `ARGS' is a vector of `CASE_LABEL_EXPR' trees 13899 that contain the labels. 13900 13901 -- GIMPLE function: unsigned gimple_switch_num_labels (gimple g) 13902 Return the number of labels associated with the switch statement 13903 `G'. 13904 13905 -- GIMPLE function: void gimple_switch_set_num_labels (gimple g, 13906 unsigned nlabels) 13907 Set `NLABELS' to be the number of labels for the switch statement 13908 `G'. 13909 13910 -- GIMPLE function: tree gimple_switch_index (gimple g) 13911 Return the index variable used by the switch statement `G'. 13912 13913 -- GIMPLE function: void gimple_switch_set_index (gimple g, tree index) 13914 Set `INDEX' to be the index variable for switch statement `G'. 13915 13916 -- GIMPLE function: tree gimple_switch_label (gimple g, unsigned index) 13917 Return the label numbered `INDEX'. The default label is 0, followed 13918 by any labels in a switch statement. 13919 13920 -- GIMPLE function: void gimple_switch_set_label (gimple g, unsigned 13921 index, tree label) 13922 Set the label number `INDEX' to `LABEL'. 0 is always the default 13923 label. 13924 13925 -- GIMPLE function: tree gimple_switch_default_label (gimple g) 13926 Return the default label for a switch statement. 13927 13928 -- GIMPLE function: void gimple_switch_set_default_label (gimple g, 13929 tree label) 13930 Set the default label for a switch statement. 13931 13932 13933 File: gccint.info, Node: `GIMPLE_TRY', Next: `GIMPLE_WITH_CLEANUP_EXPR', Prev: `GIMPLE_SWITCH', Up: Tuple specific accessors 13934 13935 12.7.27 `GIMPLE_TRY' 13936 -------------------- 13937 13938 -- GIMPLE function: gimple gimple_build_try (gimple_seq eval, 13939 gimple_seq cleanup, unsigned int kind) 13940 Build a `GIMPLE_TRY' statement. `EVAL' is a sequence with the 13941 expression to evaluate. `CLEANUP' is a sequence of statements to 13942 run at clean-up time. `KIND' is the enumeration value 13943 `GIMPLE_TRY_CATCH' if this statement denotes a try/catch construct 13944 or `GIMPLE_TRY_FINALLY' if this statement denotes a try/finally 13945 construct. 13946 13947 -- GIMPLE function: enum gimple_try_flags gimple_try_kind (gimple g) 13948 Return the kind of try block represented by `GIMPLE_TRY' `G'. This 13949 is either `GIMPLE_TRY_CATCH' or `GIMPLE_TRY_FINALLY'. 13950 13951 -- GIMPLE function: bool gimple_try_catch_is_cleanup (gimple g) 13952 Return the `GIMPLE_TRY_CATCH_IS_CLEANUP' flag. 13953 13954 -- GIMPLE function: gimple_seq gimple_try_eval (gimple g) 13955 Return the sequence of statements used as the body for `GIMPLE_TRY' 13956 `G'. 13957 13958 -- GIMPLE function: gimple_seq gimple_try_cleanup (gimple g) 13959 Return the sequence of statements used as the cleanup body for 13960 `GIMPLE_TRY' `G'. 13961 13962 -- GIMPLE function: void gimple_try_set_catch_is_cleanup (gimple g, 13963 bool catch_is_cleanup) 13964 Set the `GIMPLE_TRY_CATCH_IS_CLEANUP' flag. 13965 13966 -- GIMPLE function: void gimple_try_set_eval (gimple g, gimple_seq 13967 eval) 13968 Set `EVAL' to be the sequence of statements to use as the body for 13969 `GIMPLE_TRY' `G'. 13970 13971 -- GIMPLE function: void gimple_try_set_cleanup (gimple g, gimple_seq 13972 cleanup) 13973 Set `CLEANUP' to be the sequence of statements to use as the 13974 cleanup body for `GIMPLE_TRY' `G'. 13975 13976 13977 File: gccint.info, Node: `GIMPLE_WITH_CLEANUP_EXPR', Prev: `GIMPLE_TRY', Up: Tuple specific accessors 13978 13979 12.7.28 `GIMPLE_WITH_CLEANUP_EXPR' 13980 ---------------------------------- 13981 13982 -- GIMPLE function: gimple gimple_build_wce (gimple_seq cleanup) 13983 Build a `GIMPLE_WITH_CLEANUP_EXPR' statement. `CLEANUP' is the 13984 clean-up expression. 13985 13986 -- GIMPLE function: gimple_seq gimple_wce_cleanup (gimple g) 13987 Return the cleanup sequence for cleanup statement `G'. 13988 13989 -- GIMPLE function: void gimple_wce_set_cleanup (gimple g, gimple_seq 13990 cleanup) 13991 Set `CLEANUP' to be the cleanup sequence for `G'. 13992 13993 -- GIMPLE function: bool gimple_wce_cleanup_eh_only (gimple g) 13994 Return the `CLEANUP_EH_ONLY' flag for a `WCE' tuple. 13995 13996 -- GIMPLE function: void gimple_wce_set_cleanup_eh_only (gimple g, 13997 bool eh_only_p) 13998 Set the `CLEANUP_EH_ONLY' flag for a `WCE' tuple. 13999 14000 14001 File: gccint.info, Node: GIMPLE sequences, Next: Sequence iterators, Prev: Tuple specific accessors, Up: GIMPLE 14002 14003 12.8 GIMPLE sequences 14004 ===================== 14005 14006 GIMPLE sequences are the tuple equivalent of `STATEMENT_LIST''s used in 14007 `GENERIC'. They are used to chain statements together, and when used 14008 in conjunction with sequence iterators, provide a framework for 14009 iterating through statements. 14010 14011 GIMPLE sequences are of type struct `gimple_sequence', but are more 14012 commonly passed by reference to functions dealing with sequences. The 14013 type for a sequence pointer is `gimple_seq' which is the same as struct 14014 `gimple_sequence' *. When declaring a local sequence, you can define a 14015 local variable of type struct `gimple_sequence'. When declaring a 14016 sequence allocated on the garbage collected heap, use the function 14017 `gimple_seq_alloc' documented below. 14018 14019 There are convenience functions for iterating through sequences in the 14020 section entitled Sequence Iterators. 14021 14022 Below is a list of functions to manipulate and query sequences. 14023 14024 -- GIMPLE function: void gimple_seq_add_stmt (gimple_seq *seq, gimple 14025 g) 14026 Link a gimple statement to the end of the sequence *`SEQ' if `G' is 14027 not `NULL'. If *`SEQ' is `NULL', allocate a sequence before 14028 linking. 14029 14030 -- GIMPLE function: void gimple_seq_add_seq (gimple_seq *dest, 14031 gimple_seq src) 14032 Append sequence `SRC' to the end of sequence *`DEST' if `SRC' is 14033 not `NULL'. If *`DEST' is `NULL', allocate a new sequence before 14034 appending. 14035 14036 -- GIMPLE function: gimple_seq gimple_seq_deep_copy (gimple_seq src) 14037 Perform a deep copy of sequence `SRC' and return the result. 14038 14039 -- GIMPLE function: gimple_seq gimple_seq_reverse (gimple_seq seq) 14040 Reverse the order of the statements in the sequence `SEQ'. Return 14041 `SEQ'. 14042 14043 -- GIMPLE function: gimple gimple_seq_first (gimple_seq s) 14044 Return the first statement in sequence `S'. 14045 14046 -- GIMPLE function: gimple gimple_seq_last (gimple_seq s) 14047 Return the last statement in sequence `S'. 14048 14049 -- GIMPLE function: void gimple_seq_set_last (gimple_seq s, gimple 14050 last) 14051 Set the last statement in sequence `S' to the statement in `LAST'. 14052 14053 -- GIMPLE function: void gimple_seq_set_first (gimple_seq s, gimple 14054 first) 14055 Set the first statement in sequence `S' to the statement in 14056 `FIRST'. 14057 14058 -- GIMPLE function: void gimple_seq_init (gimple_seq s) 14059 Initialize sequence `S' to an empty sequence. 14060 14061 -- GIMPLE function: gimple_seq gimple_seq_alloc (void) 14062 Allocate a new sequence in the garbage collected store and return 14063 it. 14064 14065 -- GIMPLE function: void gimple_seq_copy (gimple_seq dest, gimple_seq 14066 src) 14067 Copy the sequence `SRC' into the sequence `DEST'. 14068 14069 -- GIMPLE function: bool gimple_seq_empty_p (gimple_seq s) 14070 Return true if the sequence `S' is empty. 14071 14072 -- GIMPLE function: gimple_seq bb_seq (basic_block bb) 14073 Returns the sequence of statements in `BB'. 14074 14075 -- GIMPLE function: void set_bb_seq (basic_block bb, gimple_seq seq) 14076 Sets the sequence of statements in `BB' to `SEQ'. 14077 14078 -- GIMPLE function: bool gimple_seq_singleton_p (gimple_seq seq) 14079 Determine whether `SEQ' contains exactly one statement. 14080 14081 14082 File: gccint.info, Node: Sequence iterators, Next: Adding a new GIMPLE statement code, Prev: GIMPLE sequences, Up: GIMPLE 14083 14084 12.9 Sequence iterators 14085 ======================= 14086 14087 Sequence iterators are convenience constructs for iterating through 14088 statements in a sequence. Given a sequence `SEQ', here is a typical 14089 use of gimple sequence iterators: 14090 14091 gimple_stmt_iterator gsi; 14092 14093 for (gsi = gsi_start (seq); !gsi_end_p (gsi); gsi_next (&gsi)) 14094 { 14095 gimple g = gsi_stmt (gsi); 14096 /* Do something with gimple statement `G'. */ 14097 } 14098 14099 Backward iterations are possible: 14100 14101 for (gsi = gsi_last (seq); !gsi_end_p (gsi); gsi_prev (&gsi)) 14102 14103 Forward and backward iterations on basic blocks are possible with 14104 `gsi_start_bb' and `gsi_last_bb'. 14105 14106 In the documentation below we sometimes refer to enum 14107 `gsi_iterator_update'. The valid options for this enumeration are: 14108 14109 * `GSI_NEW_STMT' Only valid when a single statement is added. Move 14110 the iterator to it. 14111 14112 * `GSI_SAME_STMT' Leave the iterator at the same statement. 14113 14114 * `GSI_CONTINUE_LINKING' Move iterator to whatever position is 14115 suitable for linking other statements in the same direction. 14116 14117 Below is a list of the functions used to manipulate and use statement 14118 iterators. 14119 14120 -- GIMPLE function: gimple_stmt_iterator gsi_start (gimple_seq seq) 14121 Return a new iterator pointing to the sequence `SEQ''s first 14122 statement. If `SEQ' is empty, the iterator's basic block is 14123 `NULL'. Use `gsi_start_bb' instead when the iterator needs to 14124 always have the correct basic block set. 14125 14126 -- GIMPLE function: gimple_stmt_iterator gsi_start_bb (basic_block bb) 14127 Return a new iterator pointing to the first statement in basic 14128 block `BB'. 14129 14130 -- GIMPLE function: gimple_stmt_iterator gsi_last (gimple_seq seq) 14131 Return a new iterator initially pointing to the last statement of 14132 sequence `SEQ'. If `SEQ' is empty, the iterator's basic block is 14133 `NULL'. Use `gsi_last_bb' instead when the iterator needs to 14134 always have the correct basic block set. 14135 14136 -- GIMPLE function: gimple_stmt_iterator gsi_last_bb (basic_block bb) 14137 Return a new iterator pointing to the last statement in basic 14138 block `BB'. 14139 14140 -- GIMPLE function: bool gsi_end_p (gimple_stmt_iterator i) 14141 Return `TRUE' if at the end of `I'. 14142 14143 -- GIMPLE function: bool gsi_one_before_end_p (gimple_stmt_iterator i) 14144 Return `TRUE' if we're one statement before the end of `I'. 14145 14146 -- GIMPLE function: void gsi_next (gimple_stmt_iterator *i) 14147 Advance the iterator to the next gimple statement. 14148 14149 -- GIMPLE function: void gsi_prev (gimple_stmt_iterator *i) 14150 Advance the iterator to the previous gimple statement. 14151 14152 -- GIMPLE function: gimple gsi_stmt (gimple_stmt_iterator i) 14153 Return the current stmt. 14154 14155 -- GIMPLE function: gimple_stmt_iterator gsi_after_labels (basic_block 14156 bb) 14157 Return a block statement iterator that points to the first 14158 non-label statement in block `BB'. 14159 14160 -- GIMPLE function: gimple *gsi_stmt_ptr (gimple_stmt_iterator *i) 14161 Return a pointer to the current stmt. 14162 14163 -- GIMPLE function: basic_block gsi_bb (gimple_stmt_iterator i) 14164 Return the basic block associated with this iterator. 14165 14166 -- GIMPLE function: gimple_seq gsi_seq (gimple_stmt_iterator i) 14167 Return the sequence associated with this iterator. 14168 14169 -- GIMPLE function: void gsi_remove (gimple_stmt_iterator *i, bool 14170 remove_eh_info) 14171 Remove the current stmt from the sequence. The iterator is 14172 updated to point to the next statement. When `REMOVE_EH_INFO' is 14173 true we remove the statement pointed to by iterator `I' from the 14174 `EH' tables. Otherwise we do not modify the `EH' tables. 14175 Generally, `REMOVE_EH_INFO' should be true when the statement is 14176 going to be removed from the `IL' and not reinserted elsewhere. 14177 14178 -- GIMPLE function: void gsi_link_seq_before (gimple_stmt_iterator *i, 14179 gimple_seq seq, enum gsi_iterator_update mode) 14180 Links the sequence of statements `SEQ' before the statement pointed 14181 by iterator `I'. `MODE' indicates what to do with the iterator 14182 after insertion (see `enum gsi_iterator_update' above). 14183 14184 -- GIMPLE function: void gsi_link_before (gimple_stmt_iterator *i, 14185 gimple g, enum gsi_iterator_update mode) 14186 Links statement `G' before the statement pointed-to by iterator 14187 `I'. Updates iterator `I' according to `MODE'. 14188 14189 -- GIMPLE function: void gsi_link_seq_after (gimple_stmt_iterator *i, 14190 gimple_seq seq, enum gsi_iterator_update mode) 14191 Links sequence `SEQ' after the statement pointed-to by iterator 14192 `I'. `MODE' is as in `gsi_insert_after'. 14193 14194 -- GIMPLE function: void gsi_link_after (gimple_stmt_iterator *i, 14195 gimple g, enum gsi_iterator_update mode) 14196 Links statement `G' after the statement pointed-to by iterator `I'. 14197 `MODE' is as in `gsi_insert_after'. 14198 14199 -- GIMPLE function: gimple_seq gsi_split_seq_after 14200 (gimple_stmt_iterator i) 14201 Move all statements in the sequence after `I' to a new sequence. 14202 Return this new sequence. 14203 14204 -- GIMPLE function: gimple_seq gsi_split_seq_before 14205 (gimple_stmt_iterator *i) 14206 Move all statements in the sequence before `I' to a new sequence. 14207 Return this new sequence. 14208 14209 -- GIMPLE function: void gsi_replace (gimple_stmt_iterator *i, gimple 14210 stmt, bool update_eh_info) 14211 Replace the statement pointed-to by `I' to `STMT'. If 14212 `UPDATE_EH_INFO' is true, the exception handling information of 14213 the original statement is moved to the new statement. 14214 14215 -- GIMPLE function: void gsi_insert_before (gimple_stmt_iterator *i, 14216 gimple stmt, enum gsi_iterator_update mode) 14217 Insert statement `STMT' before the statement pointed-to by iterator 14218 `I', update `STMT''s basic block and scan it for new operands. 14219 `MODE' specifies how to update iterator `I' after insertion (see 14220 enum `gsi_iterator_update'). 14221 14222 -- GIMPLE function: void gsi_insert_seq_before (gimple_stmt_iterator 14223 *i, gimple_seq seq, enum gsi_iterator_update mode) 14224 Like `gsi_insert_before', but for all the statements in `SEQ'. 14225 14226 -- GIMPLE function: void gsi_insert_after (gimple_stmt_iterator *i, 14227 gimple stmt, enum gsi_iterator_update mode) 14228 Insert statement `STMT' after the statement pointed-to by iterator 14229 `I', update `STMT''s basic block and scan it for new operands. 14230 `MODE' specifies how to update iterator `I' after insertion (see 14231 enum `gsi_iterator_update'). 14232 14233 -- GIMPLE function: void gsi_insert_seq_after (gimple_stmt_iterator 14234 *i, gimple_seq seq, enum gsi_iterator_update mode) 14235 Like `gsi_insert_after', but for all the statements in `SEQ'. 14236 14237 -- GIMPLE function: gimple_stmt_iterator gsi_for_stmt (gimple stmt) 14238 Finds iterator for `STMT'. 14239 14240 -- GIMPLE function: void gsi_move_after (gimple_stmt_iterator *from, 14241 gimple_stmt_iterator *to) 14242 Move the statement at `FROM' so it comes right after the statement 14243 at `TO'. 14244 14245 -- GIMPLE function: void gsi_move_before (gimple_stmt_iterator *from, 14246 gimple_stmt_iterator *to) 14247 Move the statement at `FROM' so it comes right before the statement 14248 at `TO'. 14249 14250 -- GIMPLE function: void gsi_move_to_bb_end (gimple_stmt_iterator 14251 *from, basic_block bb) 14252 Move the statement at `FROM' to the end of basic block `BB'. 14253 14254 -- GIMPLE function: void gsi_insert_on_edge (edge e, gimple stmt) 14255 Add `STMT' to the pending list of edge `E'. No actual insertion is 14256 made until a call to `gsi_commit_edge_inserts'() is made. 14257 14258 -- GIMPLE function: void gsi_insert_seq_on_edge (edge e, gimple_seq 14259 seq) 14260 Add the sequence of statements in `SEQ' to the pending list of edge 14261 `E'. No actual insertion is made until a call to 14262 `gsi_commit_edge_inserts'() is made. 14263 14264 -- GIMPLE function: basic_block gsi_insert_on_edge_immediate (edge e, 14265 gimple stmt) 14266 Similar to `gsi_insert_on_edge'+`gsi_commit_edge_inserts'. If a 14267 new block has to be created, it is returned. 14268 14269 -- GIMPLE function: void gsi_commit_one_edge_insert (edge e, 14270 basic_block *new_bb) 14271 Commit insertions pending at edge `E'. If a new block is created, 14272 set `NEW_BB' to this block, otherwise set it to `NULL'. 14273 14274 -- GIMPLE function: void gsi_commit_edge_inserts (void) 14275 This routine will commit all pending edge insertions, creating any 14276 new basic blocks which are necessary. 14277 14278 14279 File: gccint.info, Node: Adding a new GIMPLE statement code, Next: Statement and operand traversals, Prev: Sequence iterators, Up: GIMPLE 14280 14281 12.10 Adding a new GIMPLE statement code 14282 ======================================== 14283 14284 The first step in adding a new GIMPLE statement code, is modifying the 14285 file `gimple.def', which contains all the GIMPLE codes. Then you must 14286 add a corresponding structure, and an entry in `union 14287 gimple_statement_d', both of which are located in `gimple.h'. This in 14288 turn, will require you to add a corresponding `GTY' tag in 14289 `gsstruct.def', and code to handle this tag in `gss_for_code' which is 14290 located in `gimple.c'. 14291 14292 In order for the garbage collector to know the size of the structure 14293 you created in `gimple.h', you need to add a case to handle your new 14294 GIMPLE statement in `gimple_size' which is located in `gimple.c'. 14295 14296 You will probably want to create a function to build the new gimple 14297 statement in `gimple.c'. The function should be called 14298 `gimple_build_<`NEW_TUPLE_NAME'>', and should return the new tuple of 14299 type gimple. 14300 14301 If your new statement requires accessors for any members or operands 14302 it may have, put simple inline accessors in `gimple.h' and any 14303 non-trivial accessors in `gimple.c' with a corresponding prototype in 14304 `gimple.h'. 14305 14306 14307 File: gccint.info, Node: Statement and operand traversals, Prev: Adding a new GIMPLE statement code, Up: GIMPLE 14308 14309 12.11 Statement and operand traversals 14310 ====================================== 14311 14312 There are two functions available for walking statements and sequences: 14313 `walk_gimple_stmt' and `walk_gimple_seq', accordingly, and a third 14314 function for walking the operands in a statement: `walk_gimple_op'. 14315 14316 -- GIMPLE function: tree walk_gimple_stmt (gimple_stmt_iterator *gsi, 14317 walk_stmt_fn callback_stmt, walk_tree_fn callback_op, struct 14318 walk_stmt_info *wi) 14319 This function is used to walk the current statement in `GSI', 14320 optionally using traversal state stored in `WI'. If `WI' is 14321 `NULL', no state is kept during the traversal. 14322 14323 The callback `CALLBACK_STMT' is called. If `CALLBACK_STMT' returns 14324 true, it means that the callback function has handled all the 14325 operands of the statement and it is not necessary to walk its 14326 operands. 14327 14328 If `CALLBACK_STMT' is `NULL' or it returns false, `CALLBACK_OP' is 14329 called on each operand of the statement via `walk_gimple_op'. If 14330 `walk_gimple_op' returns non-`NULL' for any operand, the remaining 14331 operands are not scanned. 14332 14333 The return value is that returned by the last call to 14334 `walk_gimple_op', or `NULL_TREE' if no `CALLBACK_OP' is specified. 14335 14336 -- GIMPLE function: tree walk_gimple_op (gimple stmt, walk_tree_fn 14337 callback_op, struct walk_stmt_info *wi) 14338 Use this function to walk the operands of statement `STMT'. Every 14339 operand is walked via `walk_tree' with optional state information 14340 in `WI'. 14341 14342 `CALLBACK_OP' is called on each operand of `STMT' via `walk_tree'. 14343 Additional parameters to `walk_tree' must be stored in `WI'. For 14344 each operand `OP', `walk_tree' is called as: 14345 14346 walk_tree (&`OP', `CALLBACK_OP', `WI', `WI'- `PSET') 14347 14348 If `CALLBACK_OP' returns non-`NULL' for an operand, the remaining 14349 operands are not scanned. The return value is that returned by 14350 the last call to `walk_tree', or `NULL_TREE' if no `CALLBACK_OP' is 14351 specified. 14352 14353 -- GIMPLE function: tree walk_gimple_seq (gimple_seq seq, walk_stmt_fn 14354 callback_stmt, walk_tree_fn callback_op, struct 14355 walk_stmt_info *wi) 14356 This function walks all the statements in the sequence `SEQ' 14357 calling `walk_gimple_stmt' on each one. `WI' is as in 14358 `walk_gimple_stmt'. If `walk_gimple_stmt' returns non-`NULL', the 14359 walk is stopped and the value returned. Otherwise, all the 14360 statements are walked and `NULL_TREE' returned. 14361 14362 14363 File: gccint.info, Node: Tree SSA, Next: RTL, Prev: GIMPLE, Up: Top 14364 14365 13 Analysis and Optimization of GIMPLE tuples 14366 ********************************************* 14367 14368 GCC uses three main intermediate languages to represent the program 14369 during compilation: GENERIC, GIMPLE and RTL. GENERIC is a 14370 language-independent representation generated by each front end. It is 14371 used to serve as an interface between the parser and optimizer. 14372 GENERIC is a common representation that is able to represent programs 14373 written in all the languages supported by GCC. 14374 14375 GIMPLE and RTL are used to optimize the program. GIMPLE is used for 14376 target and language independent optimizations (e.g., inlining, constant 14377 propagation, tail call elimination, redundancy elimination, etc). Much 14378 like GENERIC, GIMPLE is a language independent, tree based 14379 representation. However, it differs from GENERIC in that the GIMPLE 14380 grammar is more restrictive: expressions contain no more than 3 14381 operands (except function calls), it has no control flow structures and 14382 expressions with side-effects are only allowed on the right hand side 14383 of assignments. See the chapter describing GENERIC and GIMPLE for more 14384 details. 14385 14386 This chapter describes the data structures and functions used in the 14387 GIMPLE optimizers (also known as "tree optimizers" or "middle end"). 14388 In particular, it focuses on all the macros, data structures, functions 14389 and programming constructs needed to implement optimization passes for 14390 GIMPLE. 14391 14392 * Menu: 14393 14394 * Annotations:: Attributes for variables. 14395 * SSA Operands:: SSA names referenced by GIMPLE statements. 14396 * SSA:: Static Single Assignment representation. 14397 * Alias analysis:: Representing aliased loads and stores. 14398 14399 14400 File: gccint.info, Node: Annotations, Next: SSA Operands, Up: Tree SSA 14401 14402 13.1 Annotations 14403 ================ 14404 14405 The optimizers need to associate attributes with variables during the 14406 optimization process. For instance, we need to know whether a variable 14407 has aliases. All these attributes are stored in data structures called 14408 annotations which are then linked to the field `ann' in `struct 14409 tree_common'. 14410 14411 Presently, we define annotations for variables (`var_ann_t'). 14412 Annotations are defined and documented in `tree-flow.h'. 14413 14414 14415 File: gccint.info, Node: SSA Operands, Next: SSA, Prev: Annotations, Up: Tree SSA 14416 14417 13.2 SSA Operands 14418 ================= 14419 14420 Almost every GIMPLE statement will contain a reference to a variable or 14421 memory location. Since statements come in different shapes and sizes, 14422 their operands are going to be located at various spots inside the 14423 statement's tree. To facilitate access to the statement's operands, 14424 they are organized into lists associated inside each statement's 14425 annotation. Each element in an operand list is a pointer to a 14426 `VAR_DECL', `PARM_DECL' or `SSA_NAME' tree node. This provides a very 14427 convenient way of examining and replacing operands. 14428 14429 Data flow analysis and optimization is done on all tree nodes 14430 representing variables. Any node for which `SSA_VAR_P' returns nonzero 14431 is considered when scanning statement operands. However, not all 14432 `SSA_VAR_P' variables are processed in the same way. For the purposes 14433 of optimization, we need to distinguish between references to local 14434 scalar variables and references to globals, statics, structures, 14435 arrays, aliased variables, etc. The reason is simple, the compiler can 14436 gather complete data flow information for a local scalar. On the other 14437 hand, a global variable may be modified by a function call, it may not 14438 be possible to keep track of all the elements of an array or the fields 14439 of a structure, etc. 14440 14441 The operand scanner gathers two kinds of operands: "real" and 14442 "virtual". An operand for which `is_gimple_reg' returns true is 14443 considered real, otherwise it is a virtual operand. We also 14444 distinguish between uses and definitions. An operand is used if its 14445 value is loaded by the statement (e.g., the operand at the RHS of an 14446 assignment). If the statement assigns a new value to the operand, the 14447 operand is considered a definition (e.g., the operand at the LHS of an 14448 assignment). 14449 14450 Virtual and real operands also have very different data flow 14451 properties. Real operands are unambiguous references to the full 14452 object that they represent. For instance, given 14453 14454 { 14455 int a, b; 14456 a = b 14457 } 14458 14459 Since `a' and `b' are non-aliased locals, the statement `a = b' will 14460 have one real definition and one real use because variable `b' is 14461 completely modified with the contents of variable `a'. Real definition 14462 are also known as "killing definitions". Similarly, the use of `a' 14463 reads all its bits. 14464 14465 In contrast, virtual operands are used with variables that can have a 14466 partial or ambiguous reference. This includes structures, arrays, 14467 globals, and aliased variables. In these cases, we have two types of 14468 definitions. For globals, structures, and arrays, we can determine from 14469 a statement whether a variable of these types has a killing definition. 14470 If the variable does, then the statement is marked as having a "must 14471 definition" of that variable. However, if a statement is only defining 14472 a part of the variable (i.e. a field in a structure), or if we know 14473 that a statement might define the variable but we cannot say for sure, 14474 then we mark that statement as having a "may definition". For 14475 instance, given 14476 14477 { 14478 int a, b, *p; 14479 14480 if (...) 14481 p = &a; 14482 else 14483 p = &b; 14484 *p = 5; 14485 return *p; 14486 } 14487 14488 The assignment `*p = 5' may be a definition of `a' or `b'. If we 14489 cannot determine statically where `p' is pointing to at the time of the 14490 store operation, we create virtual definitions to mark that statement 14491 as a potential definition site for `a' and `b'. Memory loads are 14492 similarly marked with virtual use operands. Virtual operands are shown 14493 in tree dumps right before the statement that contains them. To 14494 request a tree dump with virtual operands, use the `-vops' option to 14495 `-fdump-tree': 14496 14497 { 14498 int a, b, *p; 14499 14500 if (...) 14501 p = &a; 14502 else 14503 p = &b; 14504 # a = VDEF <a> 14505 # b = VDEF <b> 14506 *p = 5; 14507 14508 # VUSE <a> 14509 # VUSE <b> 14510 return *p; 14511 } 14512 14513 Notice that `VDEF' operands have two copies of the referenced 14514 variable. This indicates that this is not a killing definition of that 14515 variable. In this case we refer to it as a "may definition" or 14516 "aliased store". The presence of the second copy of the variable in 14517 the `VDEF' operand will become important when the function is converted 14518 into SSA form. This will be used to link all the non-killing 14519 definitions to prevent optimizations from making incorrect assumptions 14520 about them. 14521 14522 Operands are updated as soon as the statement is finished via a call 14523 to `update_stmt'. If statement elements are changed via `SET_USE' or 14524 `SET_DEF', then no further action is required (i.e., those macros take 14525 care of updating the statement). If changes are made by manipulating 14526 the statement's tree directly, then a call must be made to 14527 `update_stmt' when complete. Calling one of the `bsi_insert' routines 14528 or `bsi_replace' performs an implicit call to `update_stmt'. 14529 14530 13.2.1 Operand Iterators And Access Routines 14531 -------------------------------------------- 14532 14533 Operands are collected by `tree-ssa-operands.c'. They are stored 14534 inside each statement's annotation and can be accessed through either 14535 the operand iterators or an access routine. 14536 14537 The following access routines are available for examining operands: 14538 14539 1. `SINGLE_SSA_{USE,DEF,TREE}_OPERAND': These accessors will return 14540 NULL unless there is exactly one operand matching the specified 14541 flags. If there is exactly one operand, the operand is returned 14542 as either a `tree', `def_operand_p', or `use_operand_p'. 14543 14544 tree t = SINGLE_SSA_TREE_OPERAND (stmt, flags); 14545 use_operand_p u = SINGLE_SSA_USE_OPERAND (stmt, SSA_ALL_VIRTUAL_USES); 14546 def_operand_p d = SINGLE_SSA_DEF_OPERAND (stmt, SSA_OP_ALL_DEFS); 14547 14548 2. `ZERO_SSA_OPERANDS': This macro returns true if there are no 14549 operands matching the specified flags. 14550 14551 if (ZERO_SSA_OPERANDS (stmt, SSA_OP_ALL_VIRTUALS)) 14552 return; 14553 14554 3. `NUM_SSA_OPERANDS': This macro Returns the number of operands 14555 matching 'flags'. This actually executes a loop to perform the 14556 count, so only use this if it is really needed. 14557 14558 int count = NUM_SSA_OPERANDS (stmt, flags) 14559 14560 If you wish to iterate over some or all operands, use the 14561 `FOR_EACH_SSA_{USE,DEF,TREE}_OPERAND' iterator. For example, to print 14562 all the operands for a statement: 14563 14564 void 14565 print_ops (tree stmt) 14566 { 14567 ssa_op_iter; 14568 tree var; 14569 14570 FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_ALL_OPERANDS) 14571 print_generic_expr (stderr, var, TDF_SLIM); 14572 } 14573 14574 How to choose the appropriate iterator: 14575 14576 1. Determine whether you are need to see the operand pointers, or 14577 just the trees, and choose the appropriate macro: 14578 14579 Need Macro: 14580 ---- ------- 14581 use_operand_p FOR_EACH_SSA_USE_OPERAND 14582 def_operand_p FOR_EACH_SSA_DEF_OPERAND 14583 tree FOR_EACH_SSA_TREE_OPERAND 14584 14585 2. You need to declare a variable of the type you are interested in, 14586 and an ssa_op_iter structure which serves as the loop controlling 14587 variable. 14588 14589 3. Determine which operands you wish to use, and specify the flags of 14590 those you are interested in. They are documented in 14591 `tree-ssa-operands.h': 14592 14593 #define SSA_OP_USE 0x01 /* Real USE operands. */ 14594 #define SSA_OP_DEF 0x02 /* Real DEF operands. */ 14595 #define SSA_OP_VUSE 0x04 /* VUSE operands. */ 14596 #define SSA_OP_VMAYUSE 0x08 /* USE portion of VDEFS. */ 14597 #define SSA_OP_VDEF 0x10 /* DEF portion of VDEFS. */ 14598 14599 /* These are commonly grouped operand flags. */ 14600 #define SSA_OP_VIRTUAL_USES (SSA_OP_VUSE | SSA_OP_VMAYUSE) 14601 #define SSA_OP_VIRTUAL_DEFS (SSA_OP_VDEF) 14602 #define SSA_OP_ALL_USES (SSA_OP_VIRTUAL_USES | SSA_OP_USE) 14603 #define SSA_OP_ALL_DEFS (SSA_OP_VIRTUAL_DEFS | SSA_OP_DEF) 14604 #define SSA_OP_ALL_OPERANDS (SSA_OP_ALL_USES | SSA_OP_ALL_DEFS) 14605 14606 So if you want to look at the use pointers for all the `USE' and 14607 `VUSE' operands, you would do something like: 14608 14609 use_operand_p use_p; 14610 ssa_op_iter iter; 14611 14612 FOR_EACH_SSA_USE_OPERAND (use_p, stmt, iter, (SSA_OP_USE | SSA_OP_VUSE)) 14613 { 14614 process_use_ptr (use_p); 14615 } 14616 14617 The `TREE' macro is basically the same as the `USE' and `DEF' macros, 14618 only with the use or def dereferenced via `USE_FROM_PTR (use_p)' and 14619 `DEF_FROM_PTR (def_p)'. Since we aren't using operand pointers, use 14620 and defs flags can be mixed. 14621 14622 tree var; 14623 ssa_op_iter iter; 14624 14625 FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_VUSE) 14626 { 14627 print_generic_expr (stderr, var, TDF_SLIM); 14628 } 14629 14630 `VDEF's are broken into two flags, one for the `DEF' portion 14631 (`SSA_OP_VDEF') and one for the USE portion (`SSA_OP_VMAYUSE'). If all 14632 you want to look at are the `VDEF's together, there is a fourth 14633 iterator macro for this, which returns both a def_operand_p and a 14634 use_operand_p for each `VDEF' in the statement. Note that you don't 14635 need any flags for this one. 14636 14637 use_operand_p use_p; 14638 def_operand_p def_p; 14639 ssa_op_iter iter; 14640 14641 FOR_EACH_SSA_MAYDEF_OPERAND (def_p, use_p, stmt, iter) 14642 { 14643 my_code; 14644 } 14645 14646 There are many examples in the code as well, as well as the 14647 documentation in `tree-ssa-operands.h'. 14648 14649 There are also a couple of variants on the stmt iterators regarding PHI 14650 nodes. 14651 14652 `FOR_EACH_PHI_ARG' Works exactly like `FOR_EACH_SSA_USE_OPERAND', 14653 except it works over `PHI' arguments instead of statement operands. 14654 14655 /* Look at every virtual PHI use. */ 14656 FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_VIRTUAL_USES) 14657 { 14658 my_code; 14659 } 14660 14661 /* Look at every real PHI use. */ 14662 FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_USES) 14663 my_code; 14664 14665 /* Look at every PHI use. */ 14666 FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_ALL_USES) 14667 my_code; 14668 14669 `FOR_EACH_PHI_OR_STMT_{USE,DEF}' works exactly like 14670 `FOR_EACH_SSA_{USE,DEF}_OPERAND', except it will function on either a 14671 statement or a `PHI' node. These should be used when it is appropriate 14672 but they are not quite as efficient as the individual `FOR_EACH_PHI' 14673 and `FOR_EACH_SSA' routines. 14674 14675 FOR_EACH_PHI_OR_STMT_USE (use_operand_p, stmt, iter, flags) 14676 { 14677 my_code; 14678 } 14679 14680 FOR_EACH_PHI_OR_STMT_DEF (def_operand_p, phi, iter, flags) 14681 { 14682 my_code; 14683 } 14684 14685 13.2.2 Immediate Uses 14686 --------------------- 14687 14688 Immediate use information is now always available. Using the immediate 14689 use iterators, you may examine every use of any `SSA_NAME'. For 14690 instance, to change each use of `ssa_var' to `ssa_var2' and call 14691 fold_stmt on each stmt after that is done: 14692 14693 use_operand_p imm_use_p; 14694 imm_use_iterator iterator; 14695 tree ssa_var, stmt; 14696 14697 14698 FOR_EACH_IMM_USE_STMT (stmt, iterator, ssa_var) 14699 { 14700 FOR_EACH_IMM_USE_ON_STMT (imm_use_p, iterator) 14701 SET_USE (imm_use_p, ssa_var_2); 14702 fold_stmt (stmt); 14703 } 14704 14705 There are 2 iterators which can be used. `FOR_EACH_IMM_USE_FAST' is 14706 used when the immediate uses are not changed, i.e., you are looking at 14707 the uses, but not setting them. 14708 14709 If they do get changed, then care must be taken that things are not 14710 changed under the iterators, so use the `FOR_EACH_IMM_USE_STMT' and 14711 `FOR_EACH_IMM_USE_ON_STMT' iterators. They attempt to preserve the 14712 sanity of the use list by moving all the uses for a statement into a 14713 controlled position, and then iterating over those uses. Then the 14714 optimization can manipulate the stmt when all the uses have been 14715 processed. This is a little slower than the FAST version since it adds 14716 a placeholder element and must sort through the list a bit for each 14717 statement. This placeholder element must be also be removed if the 14718 loop is terminated early. The macro `BREAK_FROM_IMM_USE_SAFE' is 14719 provided to do this : 14720 14721 FOR_EACH_IMM_USE_STMT (stmt, iterator, ssa_var) 14722 { 14723 if (stmt == last_stmt) 14724 BREAK_FROM_SAFE_IMM_USE (iter); 14725 14726 FOR_EACH_IMM_USE_ON_STMT (imm_use_p, iterator) 14727 SET_USE (imm_use_p, ssa_var_2); 14728 fold_stmt (stmt); 14729 } 14730 14731 There are checks in `verify_ssa' which verify that the immediate use 14732 list is up to date, as well as checking that an optimization didn't 14733 break from the loop without using this macro. It is safe to simply 14734 'break'; from a `FOR_EACH_IMM_USE_FAST' traverse. 14735 14736 Some useful functions and macros: 14737 1. `has_zero_uses (ssa_var)' : Returns true if there are no uses of 14738 `ssa_var'. 14739 14740 2. `has_single_use (ssa_var)' : Returns true if there is only a 14741 single use of `ssa_var'. 14742 14743 3. `single_imm_use (ssa_var, use_operand_p *ptr, tree *stmt)' : 14744 Returns true if there is only a single use of `ssa_var', and also 14745 returns the use pointer and statement it occurs in, in the second 14746 and third parameters. 14747 14748 4. `num_imm_uses (ssa_var)' : Returns the number of immediate uses of 14749 `ssa_var'. It is better not to use this if possible since it simply 14750 utilizes a loop to count the uses. 14751 14752 5. `PHI_ARG_INDEX_FROM_USE (use_p)' : Given a use within a `PHI' 14753 node, return the index number for the use. An assert is triggered 14754 if the use isn't located in a `PHI' node. 14755 14756 6. `USE_STMT (use_p)' : Return the statement a use occurs in. 14757 14758 Note that uses are not put into an immediate use list until their 14759 statement is actually inserted into the instruction stream via a 14760 `bsi_*' routine. 14761 14762 It is also still possible to utilize lazy updating of statements, but 14763 this should be used only when absolutely required. Both alias analysis 14764 and the dominator optimizations currently do this. 14765 14766 When lazy updating is being used, the immediate use information is out 14767 of date and cannot be used reliably. Lazy updating is achieved by 14768 simply marking statements modified via calls to `mark_stmt_modified' 14769 instead of `update_stmt'. When lazy updating is no longer required, 14770 all the modified statements must have `update_stmt' called in order to 14771 bring them up to date. This must be done before the optimization is 14772 finished, or `verify_ssa' will trigger an abort. 14773 14774 This is done with a simple loop over the instruction stream: 14775 block_stmt_iterator bsi; 14776 basic_block bb; 14777 FOR_EACH_BB (bb) 14778 { 14779 for (bsi = bsi_start (bb); !bsi_end_p (bsi); bsi_next (&bsi)) 14780 update_stmt_if_modified (bsi_stmt (bsi)); 14781 } 14782 14783 14784 File: gccint.info, Node: SSA, Next: Alias analysis, Prev: SSA Operands, Up: Tree SSA 14785 14786 13.3 Static Single Assignment 14787 ============================= 14788 14789 Most of the tree optimizers rely on the data flow information provided 14790 by the Static Single Assignment (SSA) form. We implement the SSA form 14791 as described in `R. Cytron, J. Ferrante, B. Rosen, M. Wegman, and K. 14792 Zadeck. Efficiently Computing Static Single Assignment Form and the 14793 Control Dependence Graph. ACM Transactions on Programming Languages 14794 and Systems, 13(4):451-490, October 1991'. 14795 14796 The SSA form is based on the premise that program variables are 14797 assigned in exactly one location in the program. Multiple assignments 14798 to the same variable create new versions of that variable. Naturally, 14799 actual programs are seldom in SSA form initially because variables tend 14800 to be assigned multiple times. The compiler modifies the program 14801 representation so that every time a variable is assigned in the code, a 14802 new version of the variable is created. Different versions of the same 14803 variable are distinguished by subscripting the variable name with its 14804 version number. Variables used in the right-hand side of expressions 14805 are renamed so that their version number matches that of the most 14806 recent assignment. 14807 14808 We represent variable versions using `SSA_NAME' nodes. The renaming 14809 process in `tree-ssa.c' wraps every real and virtual operand with an 14810 `SSA_NAME' node which contains the version number and the statement 14811 that created the `SSA_NAME'. Only definitions and virtual definitions 14812 may create new `SSA_NAME' nodes. 14813 14814 Sometimes, flow of control makes it impossible to determine the most 14815 recent version of a variable. In these cases, the compiler inserts an 14816 artificial definition for that variable called "PHI function" or "PHI 14817 node". This new definition merges all the incoming versions of the 14818 variable to create a new name for it. For instance, 14819 14820 if (...) 14821 a_1 = 5; 14822 else if (...) 14823 a_2 = 2; 14824 else 14825 a_3 = 13; 14826 14827 # a_4 = PHI <a_1, a_2, a_3> 14828 return a_4; 14829 14830 Since it is not possible to determine which of the three branches will 14831 be taken at runtime, we don't know which of `a_1', `a_2' or `a_3' to 14832 use at the return statement. So, the SSA renamer creates a new version 14833 `a_4' which is assigned the result of "merging" `a_1', `a_2' and `a_3'. 14834 Hence, PHI nodes mean "one of these operands. I don't know which". 14835 14836 The following macros can be used to examine PHI nodes 14837 14838 -- Macro: PHI_RESULT (PHI) 14839 Returns the `SSA_NAME' created by PHI node PHI (i.e., PHI's LHS). 14840 14841 -- Macro: PHI_NUM_ARGS (PHI) 14842 Returns the number of arguments in PHI. This number is exactly 14843 the number of incoming edges to the basic block holding PHI. 14844 14845 -- Macro: PHI_ARG_ELT (PHI, I) 14846 Returns a tuple representing the Ith argument of PHI. Each 14847 element of this tuple contains an `SSA_NAME' VAR and the incoming 14848 edge through which VAR flows. 14849 14850 -- Macro: PHI_ARG_EDGE (PHI, I) 14851 Returns the incoming edge for the Ith argument of PHI. 14852 14853 -- Macro: PHI_ARG_DEF (PHI, I) 14854 Returns the `SSA_NAME' for the Ith argument of PHI. 14855 14856 13.3.1 Preserving the SSA form 14857 ------------------------------ 14858 14859 Some optimization passes make changes to the function that invalidate 14860 the SSA property. This can happen when a pass has added new symbols or 14861 changed the program so that variables that were previously aliased 14862 aren't anymore. Whenever something like this happens, the affected 14863 symbols must be renamed into SSA form again. Transformations that emit 14864 new code or replicate existing statements will also need to update the 14865 SSA form. 14866 14867 Since GCC implements two different SSA forms for register and virtual 14868 variables, keeping the SSA form up to date depends on whether you are 14869 updating register or virtual names. In both cases, the general idea 14870 behind incremental SSA updates is similar: when new SSA names are 14871 created, they typically are meant to replace other existing names in 14872 the program. 14873 14874 For instance, given the following code: 14875 14876 1 L0: 14877 2 x_1 = PHI (0, x_5) 14878 3 if (x_1 < 10) 14879 4 if (x_1 > 7) 14880 5 y_2 = 0 14881 6 else 14882 7 y_3 = x_1 + x_7 14883 8 endif 14884 9 x_5 = x_1 + 1 14885 10 goto L0; 14886 11 endif 14887 14888 Suppose that we insert new names `x_10' and `x_11' (lines `4' and `8'). 14889 14890 1 L0: 14891 2 x_1 = PHI (0, x_5) 14892 3 if (x_1 < 10) 14893 4 x_10 = ... 14894 5 if (x_1 > 7) 14895 6 y_2 = 0 14896 7 else 14897 8 x_11 = ... 14898 9 y_3 = x_1 + x_7 14899 10 endif 14900 11 x_5 = x_1 + 1 14901 12 goto L0; 14902 13 endif 14903 14904 We want to replace all the uses of `x_1' with the new definitions of 14905 `x_10' and `x_11'. Note that the only uses that should be replaced are 14906 those at lines `5', `9' and `11'. Also, the use of `x_7' at line `9' 14907 should _not_ be replaced (this is why we cannot just mark symbol `x' for 14908 renaming). 14909 14910 Additionally, we may need to insert a PHI node at line `11' because 14911 that is a merge point for `x_10' and `x_11'. So the use of `x_1' at 14912 line `11' will be replaced with the new PHI node. The insertion of PHI 14913 nodes is optional. They are not strictly necessary to preserve the SSA 14914 form, and depending on what the caller inserted, they may not even be 14915 useful for the optimizers. 14916 14917 Updating the SSA form is a two step process. First, the pass has to 14918 identify which names need to be updated and/or which symbols need to be 14919 renamed into SSA form for the first time. When new names are 14920 introduced to replace existing names in the program, the mapping 14921 between the old and the new names are registered by calling 14922 `register_new_name_mapping' (note that if your pass creates new code by 14923 duplicating basic blocks, the call to `tree_duplicate_bb' will set up 14924 the necessary mappings automatically). On the other hand, if your pass 14925 exposes a new symbol that should be put in SSA form for the first time, 14926 the new symbol should be registered with `mark_sym_for_renaming'. 14927 14928 After the replacement mappings have been registered and new symbols 14929 marked for renaming, a call to `update_ssa' makes the registered 14930 changes. This can be done with an explicit call or by creating `TODO' 14931 flags in the `tree_opt_pass' structure for your pass. There are 14932 several `TODO' flags that control the behavior of `update_ssa': 14933 14934 * `TODO_update_ssa'. Update the SSA form inserting PHI nodes for 14935 newly exposed symbols and virtual names marked for updating. When 14936 updating real names, only insert PHI nodes for a real name `O_j' 14937 in blocks reached by all the new and old definitions for `O_j'. 14938 If the iterated dominance frontier for `O_j' is not pruned, we may 14939 end up inserting PHI nodes in blocks that have one or more edges 14940 with no incoming definition for `O_j'. This would lead to 14941 uninitialized warnings for `O_j''s symbol. 14942 14943 * `TODO_update_ssa_no_phi'. Update the SSA form without inserting 14944 any new PHI nodes at all. This is used by passes that have either 14945 inserted all the PHI nodes themselves or passes that need only to 14946 patch use-def and def-def chains for virtuals (e.g., DCE). 14947 14948 * `TODO_update_ssa_full_phi'. Insert PHI nodes everywhere they are 14949 needed. No pruning of the IDF is done. This is used by passes 14950 that need the PHI nodes for `O_j' even if it means that some 14951 arguments will come from the default definition of `O_j''s symbol 14952 (e.g., `pass_linear_transform'). 14953 14954 WARNING: If you need to use this flag, chances are that your pass 14955 may be doing something wrong. Inserting PHI nodes for an old name 14956 where not all edges carry a new replacement may lead to silent 14957 codegen errors or spurious uninitialized warnings. 14958 14959 * `TODO_update_ssa_only_virtuals'. Passes that update the SSA form 14960 on their own may want to delegate the updating of virtual names to 14961 the generic updater. Since FUD chains are easier to maintain, 14962 this simplifies the work they need to do. NOTE: If this flag is 14963 used, any OLD->NEW mappings for real names are explicitly 14964 destroyed and only the symbols marked for renaming are processed. 14965 14966 13.3.2 Preserving the virtual SSA form 14967 -------------------------------------- 14968 14969 The virtual SSA form is harder to preserve than the non-virtual SSA form 14970 mainly because the set of virtual operands for a statement may change at 14971 what some would consider unexpected times. In general, statement 14972 modifications should be bracketed between calls to `push_stmt_changes' 14973 and `pop_stmt_changes'. For example, 14974 14975 munge_stmt (tree stmt) 14976 { 14977 push_stmt_changes (&stmt); 14978 ... rewrite STMT ... 14979 pop_stmt_changes (&stmt); 14980 } 14981 14982 The call to `push_stmt_changes' saves the current state of the 14983 statement operands and the call to `pop_stmt_changes' compares the 14984 saved state with the current one and does the appropriate symbol 14985 marking for the SSA renamer. 14986 14987 It is possible to modify several statements at a time, provided that 14988 `push_stmt_changes' and `pop_stmt_changes' are called in LIFO order, as 14989 when processing a stack of statements. 14990 14991 Additionally, if the pass discovers that it did not need to make 14992 changes to the statement after calling `push_stmt_changes', it can 14993 simply discard the topmost change buffer by calling 14994 `discard_stmt_changes'. This will avoid the expensive operand re-scan 14995 operation and the buffer comparison that determines if symbols need to 14996 be marked for renaming. 14997 14998 13.3.3 Examining `SSA_NAME' nodes 14999 --------------------------------- 15000 15001 The following macros can be used to examine `SSA_NAME' nodes 15002 15003 -- Macro: SSA_NAME_DEF_STMT (VAR) 15004 Returns the statement S that creates the `SSA_NAME' VAR. If S is 15005 an empty statement (i.e., `IS_EMPTY_STMT (S)' returns `true'), it 15006 means that the first reference to this variable is a USE or a VUSE. 15007 15008 -- Macro: SSA_NAME_VERSION (VAR) 15009 Returns the version number of the `SSA_NAME' object VAR. 15010 15011 13.3.4 Walking use-def chains 15012 ----------------------------- 15013 15014 -- Tree SSA function: void walk_use_def_chains (VAR, FN, DATA) 15015 Walks use-def chains starting at the `SSA_NAME' node VAR. Calls 15016 function FN at each reaching definition found. Function FN takes 15017 three arguments: VAR, its defining statement (DEF_STMT) and a 15018 generic pointer to whatever state information that FN may want to 15019 maintain (DATA). Function FN is able to stop the walk by 15020 returning `true', otherwise in order to continue the walk, FN 15021 should return `false'. 15022 15023 Note, that if DEF_STMT is a `PHI' node, the semantics are slightly 15024 different. For each argument ARG of the PHI node, this function 15025 will: 15026 15027 1. Walk the use-def chains for ARG. 15028 15029 2. Call `FN (ARG, PHI, DATA)'. 15030 15031 Note how the first argument to FN is no longer the original 15032 variable VAR, but the PHI argument currently being examined. If 15033 FN wants to get at VAR, it should call `PHI_RESULT' (PHI). 15034 15035 13.3.5 Walking the dominator tree 15036 --------------------------------- 15037 15038 -- Tree SSA function: void walk_dominator_tree (WALK_DATA, BB) 15039 This function walks the dominator tree for the current CFG calling 15040 a set of callback functions defined in STRUCT DOM_WALK_DATA in 15041 `domwalk.h'. The call back functions you need to define give you 15042 hooks to execute custom code at various points during traversal: 15043 15044 1. Once to initialize any local data needed while processing BB 15045 and its children. This local data is pushed into an internal 15046 stack which is automatically pushed and popped as the walker 15047 traverses the dominator tree. 15048 15049 2. Once before traversing all the statements in the BB. 15050 15051 3. Once for every statement inside BB. 15052 15053 4. Once after traversing all the statements and before recursing 15054 into BB's dominator children. 15055 15056 5. It then recurses into all the dominator children of BB. 15057 15058 6. After recursing into all the dominator children of BB it can, 15059 optionally, traverse every statement in BB again (i.e., 15060 repeating steps 2 and 3). 15061 15062 7. Once after walking the statements in BB and BB's dominator 15063 children. At this stage, the block local data stack is 15064 popped. 15065 15066 15067 File: gccint.info, Node: Alias analysis, Prev: SSA, Up: Tree SSA 15068 15069 13.4 Alias analysis 15070 =================== 15071 15072 Alias analysis proceeds in 4 main phases: 15073 15074 1. Structural alias analysis. 15075 15076 This phase walks the types for structure variables, and determines 15077 which of the fields can overlap using offset and size of each 15078 field. For each field, a "subvariable" called a "Structure field 15079 tag" (SFT) is created, which represents that field as a separate 15080 variable. All accesses that could possibly overlap with a given 15081 field will have virtual operands for the SFT of that field. 15082 15083 struct foo 15084 { 15085 int a; 15086 int b; 15087 } 15088 struct foo temp; 15089 int bar (void) 15090 { 15091 int tmp1, tmp2, tmp3; 15092 SFT.0_2 = VDEF <SFT.0_1> 15093 temp.a = 5; 15094 SFT.1_4 = VDEF <SFT.1_3> 15095 temp.b = 6; 15096 15097 VUSE <SFT.1_4> 15098 tmp1_5 = temp.b; 15099 VUSE <SFT.0_2> 15100 tmp2_6 = temp.a; 15101 15102 tmp3_7 = tmp1_5 + tmp2_6; 15103 return tmp3_7; 15104 } 15105 15106 If you copy the symbol tag for a variable for some reason, you 15107 probably also want to copy the subvariables for that variable. 15108 15109 2. Points-to and escape analysis. 15110 15111 This phase walks the use-def chains in the SSA web looking for 15112 three things: 15113 15114 * Assignments of the form `P_i = &VAR' 15115 15116 * Assignments of the form P_i = malloc() 15117 15118 * Pointers and ADDR_EXPR that escape the current function. 15119 15120 The concept of `escaping' is the same one used in the Java world. 15121 When a pointer or an ADDR_EXPR escapes, it means that it has been 15122 exposed outside of the current function. So, assignment to global 15123 variables, function arguments and returning a pointer are all 15124 escape sites. 15125 15126 This is where we are currently limited. Since not everything is 15127 renamed into SSA, we lose track of escape properties when a 15128 pointer is stashed inside a field in a structure, for instance. 15129 In those cases, we are assuming that the pointer does escape. 15130 15131 We use escape analysis to determine whether a variable is 15132 call-clobbered. Simply put, if an ADDR_EXPR escapes, then the 15133 variable is call-clobbered. If a pointer P_i escapes, then all 15134 the variables pointed-to by P_i (and its memory tag) also escape. 15135 15136 3. Compute flow-sensitive aliases 15137 15138 We have two classes of memory tags. Memory tags associated with 15139 the pointed-to data type of the pointers in the program. These 15140 tags are called "symbol memory tag" (SMT). The other class are 15141 those associated with SSA_NAMEs, called "name memory tag" (NMT). 15142 The basic idea is that when adding operands for an INDIRECT_REF 15143 *P_i, we will first check whether P_i has a name tag, if it does 15144 we use it, because that will have more precise aliasing 15145 information. Otherwise, we use the standard symbol tag. 15146 15147 In this phase, we go through all the pointers we found in 15148 points-to analysis and create alias sets for the name memory tags 15149 associated with each pointer P_i. If P_i escapes, we mark 15150 call-clobbered the variables it points to and its tag. 15151 15152 4. Compute flow-insensitive aliases 15153 15154 This pass will compare the alias set of every symbol memory tag and 15155 every addressable variable found in the program. Given a symbol 15156 memory tag SMT and an addressable variable V. If the alias sets 15157 of SMT and V conflict (as computed by may_alias_p), then V is 15158 marked as an alias tag and added to the alias set of SMT. 15159 15160 Every language that wishes to perform language-specific alias 15161 analysis should define a function that computes, given a `tree' 15162 node, an alias set for the node. Nodes in different alias sets 15163 are not allowed to alias. For an example, see the C front-end 15164 function `c_get_alias_set'. 15165 15166 For instance, consider the following function: 15167 15168 foo (int i) 15169 { 15170 int *p, *q, a, b; 15171 15172 if (i > 10) 15173 p = &a; 15174 else 15175 q = &b; 15176 15177 *p = 3; 15178 *q = 5; 15179 a = b + 2; 15180 return *p; 15181 } 15182 15183 After aliasing analysis has finished, the symbol memory tag for 15184 pointer `p' will have two aliases, namely variables `a' and `b'. Every 15185 time pointer `p' is dereferenced, we want to mark the operation as a 15186 potential reference to `a' and `b'. 15187 15188 foo (int i) 15189 { 15190 int *p, a, b; 15191 15192 if (i_2 > 10) 15193 p_4 = &a; 15194 else 15195 p_6 = &b; 15196 # p_1 = PHI <p_4(1), p_6(2)>; 15197 15198 # a_7 = VDEF <a_3>; 15199 # b_8 = VDEF <b_5>; 15200 *p_1 = 3; 15201 15202 # a_9 = VDEF <a_7> 15203 # VUSE <b_8> 15204 a_9 = b_8 + 2; 15205 15206 # VUSE <a_9>; 15207 # VUSE <b_8>; 15208 return *p_1; 15209 } 15210 15211 In certain cases, the list of may aliases for a pointer may grow too 15212 large. This may cause an explosion in the number of virtual operands 15213 inserted in the code. Resulting in increased memory consumption and 15214 compilation time. 15215 15216 When the number of virtual operands needed to represent aliased loads 15217 and stores grows too large (configurable with `--param 15218 max-aliased-vops'), alias sets are grouped to avoid severe compile-time 15219 slow downs and memory consumption. The alias grouping heuristic 15220 proceeds as follows: 15221 15222 1. Sort the list of pointers in decreasing number of contributed 15223 virtual operands. 15224 15225 2. Take the first pointer from the list and reverse the role of the 15226 memory tag and its aliases. Usually, whenever an aliased variable 15227 Vi is found to alias with a memory tag T, we add Vi to the 15228 may-aliases set for T. Meaning that after alias analysis, we will 15229 have: 15230 15231 may-aliases(T) = { V1, V2, V3, ..., Vn } 15232 15233 This means that every statement that references T, will get `n' 15234 virtual operands for each of the Vi tags. But, when alias 15235 grouping is enabled, we make T an alias tag and add it to the 15236 alias set of all the Vi variables: 15237 15238 may-aliases(V1) = { T } 15239 may-aliases(V2) = { T } 15240 ... 15241 may-aliases(Vn) = { T } 15242 15243 This has two effects: (a) statements referencing T will only get a 15244 single virtual operand, and, (b) all the variables Vi will now 15245 appear to alias each other. So, we lose alias precision to 15246 improve compile time. But, in theory, a program with such a high 15247 level of aliasing should not be very optimizable in the first 15248 place. 15249 15250 3. Since variables may be in the alias set of more than one memory 15251 tag, the grouping done in step (2) needs to be extended to all the 15252 memory tags that have a non-empty intersection with the 15253 may-aliases set of tag T. For instance, if we originally had 15254 these may-aliases sets: 15255 15256 may-aliases(T) = { V1, V2, V3 } 15257 may-aliases(R) = { V2, V4 } 15258 15259 In step (2) we would have reverted the aliases for T as: 15260 15261 may-aliases(V1) = { T } 15262 may-aliases(V2) = { T } 15263 may-aliases(V3) = { T } 15264 15265 But note that now V2 is no longer aliased with R. We could add R 15266 to may-aliases(V2), but we are in the process of grouping aliases 15267 to reduce virtual operands so what we do is add V4 to the grouping 15268 to obtain: 15269 15270 may-aliases(V1) = { T } 15271 may-aliases(V2) = { T } 15272 may-aliases(V3) = { T } 15273 may-aliases(V4) = { T } 15274 15275 4. If the total number of virtual operands due to aliasing is still 15276 above the threshold set by max-alias-vops, go back to (2). 15277 15278 15279 File: gccint.info, Node: Loop Analysis and Representation, Next: Machine Desc, Prev: Control Flow, Up: Top 15280 15281 14 Analysis and Representation of Loops 15282 *************************************** 15283 15284 GCC provides extensive infrastructure for work with natural loops, i.e., 15285 strongly connected components of CFG with only one entry block. This 15286 chapter describes representation of loops in GCC, both on GIMPLE and in 15287 RTL, as well as the interfaces to loop-related analyses (induction 15288 variable analysis and number of iterations analysis). 15289 15290 * Menu: 15291 15292 * Loop representation:: Representation and analysis of loops. 15293 * Loop querying:: Getting information about loops. 15294 * Loop manipulation:: Loop manipulation functions. 15295 * LCSSA:: Loop-closed SSA form. 15296 * Scalar evolutions:: Induction variables on GIMPLE. 15297 * loop-iv:: Induction variables on RTL. 15298 * Number of iterations:: Number of iterations analysis. 15299 * Dependency analysis:: Data dependency analysis. 15300 * Lambda:: Linear loop transformations framework. 15301 * Omega:: A solver for linear programming problems. 15302 15303 15304 File: gccint.info, Node: Loop representation, Next: Loop querying, Up: Loop Analysis and Representation 15305 15306 14.1 Loop representation 15307 ======================== 15308 15309 This chapter describes the representation of loops in GCC, and functions 15310 that can be used to build, modify and analyze this representation. Most 15311 of the interfaces and data structures are declared in `cfgloop.h'. At 15312 the moment, loop structures are analyzed and this information is 15313 updated only by the optimization passes that deal with loops, but some 15314 efforts are being made to make it available throughout most of the 15315 optimization passes. 15316 15317 In general, a natural loop has one entry block (header) and possibly 15318 several back edges (latches) leading to the header from the inside of 15319 the loop. Loops with several latches may appear if several loops share 15320 a single header, or if there is a branching in the middle of the loop. 15321 The representation of loops in GCC however allows only loops with a 15322 single latch. During loop analysis, headers of such loops are split and 15323 forwarder blocks are created in order to disambiguate their structures. 15324 Heuristic based on profile information and structure of the induction 15325 variables in the loops is used to determine whether the latches 15326 correspond to sub-loops or to control flow in a single loop. This means 15327 that the analysis sometimes changes the CFG, and if you run it in the 15328 middle of an optimization pass, you must be able to deal with the new 15329 blocks. You may avoid CFG changes by passing 15330 `LOOPS_MAY_HAVE_MULTIPLE_LATCHES' flag to the loop discovery, note 15331 however that most other loop manipulation functions will not work 15332 correctly for loops with multiple latch edges (the functions that only 15333 query membership of blocks to loops and subloop relationships, or 15334 enumerate and test loop exits, can be expected to work). 15335 15336 Body of the loop is the set of blocks that are dominated by its header, 15337 and reachable from its latch against the direction of edges in CFG. The 15338 loops are organized in a containment hierarchy (tree) such that all the 15339 loops immediately contained inside loop L are the children of L in the 15340 tree. This tree is represented by the `struct loops' structure. The 15341 root of this tree is a fake loop that contains all blocks in the 15342 function. Each of the loops is represented in a `struct loop' 15343 structure. Each loop is assigned an index (`num' field of the `struct 15344 loop' structure), and the pointer to the loop is stored in the 15345 corresponding field of the `larray' vector in the loops structure. The 15346 indices do not have to be continuous, there may be empty (`NULL') 15347 entries in the `larray' created by deleting loops. Also, there is no 15348 guarantee on the relative order of a loop and its subloops in the 15349 numbering. The index of a loop never changes. 15350 15351 The entries of the `larray' field should not be accessed directly. 15352 The function `get_loop' returns the loop description for a loop with 15353 the given index. `number_of_loops' function returns number of loops in 15354 the function. To traverse all loops, use `FOR_EACH_LOOP' macro. The 15355 `flags' argument of the macro is used to determine the direction of 15356 traversal and the set of loops visited. Each loop is guaranteed to be 15357 visited exactly once, regardless of the changes to the loop tree, and 15358 the loops may be removed during the traversal. The newly created loops 15359 are never traversed, if they need to be visited, this must be done 15360 separately after their creation. The `FOR_EACH_LOOP' macro allocates 15361 temporary variables. If the `FOR_EACH_LOOP' loop were ended using 15362 break or goto, they would not be released; `FOR_EACH_LOOP_BREAK' macro 15363 must be used instead. 15364 15365 Each basic block contains the reference to the innermost loop it 15366 belongs to (`loop_father'). For this reason, it is only possible to 15367 have one `struct loops' structure initialized at the same time for each 15368 CFG. The global variable `current_loops' contains the `struct loops' 15369 structure. Many of the loop manipulation functions assume that 15370 dominance information is up-to-date. 15371 15372 The loops are analyzed through `loop_optimizer_init' function. The 15373 argument of this function is a set of flags represented in an integer 15374 bitmask. These flags specify what other properties of the loop 15375 structures should be calculated/enforced and preserved later: 15376 15377 * `LOOPS_MAY_HAVE_MULTIPLE_LATCHES': If this flag is set, no changes 15378 to CFG will be performed in the loop analysis, in particular, 15379 loops with multiple latch edges will not be disambiguated. If a 15380 loop has multiple latches, its latch block is set to NULL. Most of 15381 the loop manipulation functions will not work for loops in this 15382 shape. No other flags that require CFG changes can be passed to 15383 loop_optimizer_init. 15384 15385 * `LOOPS_HAVE_PREHEADERS': Forwarder blocks are created in such a 15386 way that each loop has only one entry edge, and additionally, the 15387 source block of this entry edge has only one successor. This 15388 creates a natural place where the code can be moved out of the 15389 loop, and ensures that the entry edge of the loop leads from its 15390 immediate super-loop. 15391 15392 * `LOOPS_HAVE_SIMPLE_LATCHES': Forwarder blocks are created to force 15393 the latch block of each loop to have only one successor. This 15394 ensures that the latch of the loop does not belong to any of its 15395 sub-loops, and makes manipulation with the loops significantly 15396 easier. Most of the loop manipulation functions assume that the 15397 loops are in this shape. Note that with this flag, the "normal" 15398 loop without any control flow inside and with one exit consists of 15399 two basic blocks. 15400 15401 * `LOOPS_HAVE_MARKED_IRREDUCIBLE_REGIONS': Basic blocks and edges in 15402 the strongly connected components that are not natural loops (have 15403 more than one entry block) are marked with `BB_IRREDUCIBLE_LOOP' 15404 and `EDGE_IRREDUCIBLE_LOOP' flags. The flag is not set for blocks 15405 and edges that belong to natural loops that are in such an 15406 irreducible region (but it is set for the entry and exit edges of 15407 such a loop, if they lead to/from this region). 15408 15409 * `LOOPS_HAVE_RECORDED_EXITS': The lists of exits are recorded and 15410 updated for each loop. This makes some functions (e.g., 15411 `get_loop_exit_edges') more efficient. Some functions (e.g., 15412 `single_exit') can be used only if the lists of exits are recorded. 15413 15414 These properties may also be computed/enforced later, using functions 15415 `create_preheaders', `force_single_succ_latches', 15416 `mark_irreducible_loops' and `record_loop_exits'. 15417 15418 The memory occupied by the loops structures should be freed with 15419 `loop_optimizer_finalize' function. 15420 15421 The CFG manipulation functions in general do not update loop 15422 structures. Specialized versions that additionally do so are provided 15423 for the most common tasks. On GIMPLE, `cleanup_tree_cfg_loop' function 15424 can be used to cleanup CFG while updating the loops structures if 15425 `current_loops' is set. 15426 15427 15428 File: gccint.info, Node: Loop querying, Next: Loop manipulation, Prev: Loop representation, Up: Loop Analysis and Representation 15429 15430 14.2 Loop querying 15431 ================== 15432 15433 The functions to query the information about loops are declared in 15434 `cfgloop.h'. Some of the information can be taken directly from the 15435 structures. `loop_father' field of each basic block contains the 15436 innermost loop to that the block belongs. The most useful fields of 15437 loop structure (that are kept up-to-date at all times) are: 15438 15439 * `header', `latch': Header and latch basic blocks of the loop. 15440 15441 * `num_nodes': Number of basic blocks in the loop (including the 15442 basic blocks of the sub-loops). 15443 15444 * `depth': The depth of the loop in the loops tree, i.e., the number 15445 of super-loops of the loop. 15446 15447 * `outer', `inner', `next': The super-loop, the first sub-loop, and 15448 the sibling of the loop in the loops tree. 15449 15450 There are other fields in the loop structures, many of them used only 15451 by some of the passes, or not updated during CFG changes; in general, 15452 they should not be accessed directly. 15453 15454 The most important functions to query loop structures are: 15455 15456 * `flow_loops_dump': Dumps the information about loops to a file. 15457 15458 * `verify_loop_structure': Checks consistency of the loop structures. 15459 15460 * `loop_latch_edge': Returns the latch edge of a loop. 15461 15462 * `loop_preheader_edge': If loops have preheaders, returns the 15463 preheader edge of a loop. 15464 15465 * `flow_loop_nested_p': Tests whether loop is a sub-loop of another 15466 loop. 15467 15468 * `flow_bb_inside_loop_p': Tests whether a basic block belongs to a 15469 loop (including its sub-loops). 15470 15471 * `find_common_loop': Finds the common super-loop of two loops. 15472 15473 * `superloop_at_depth': Returns the super-loop of a loop with the 15474 given depth. 15475 15476 * `tree_num_loop_insns', `num_loop_insns': Estimates the number of 15477 insns in the loop, on GIMPLE and on RTL. 15478 15479 * `loop_exit_edge_p': Tests whether edge is an exit from a loop. 15480 15481 * `mark_loop_exit_edges': Marks all exit edges of all loops with 15482 `EDGE_LOOP_EXIT' flag. 15483 15484 * `get_loop_body', `get_loop_body_in_dom_order', 15485 `get_loop_body_in_bfs_order': Enumerates the basic blocks in the 15486 loop in depth-first search order in reversed CFG, ordered by 15487 dominance relation, and breath-first search order, respectively. 15488 15489 * `single_exit': Returns the single exit edge of the loop, or `NULL' 15490 if the loop has more than one exit. You can only use this 15491 function if LOOPS_HAVE_MARKED_SINGLE_EXITS property is used. 15492 15493 * `get_loop_exit_edges': Enumerates the exit edges of a loop. 15494 15495 * `just_once_each_iteration_p': Returns true if the basic block is 15496 executed exactly once during each iteration of a loop (that is, it 15497 does not belong to a sub-loop, and it dominates the latch of the 15498 loop). 15499 15500 15501 File: gccint.info, Node: Loop manipulation, Next: LCSSA, Prev: Loop querying, Up: Loop Analysis and Representation 15502 15503 14.3 Loop manipulation 15504 ====================== 15505 15506 The loops tree can be manipulated using the following functions: 15507 15508 * `flow_loop_tree_node_add': Adds a node to the tree. 15509 15510 * `flow_loop_tree_node_remove': Removes a node from the tree. 15511 15512 * `add_bb_to_loop': Adds a basic block to a loop. 15513 15514 * `remove_bb_from_loops': Removes a basic block from loops. 15515 15516 Most low-level CFG functions update loops automatically. The following 15517 functions handle some more complicated cases of CFG manipulations: 15518 15519 * `remove_path': Removes an edge and all blocks it dominates. 15520 15521 * `split_loop_exit_edge': Splits exit edge of the loop, ensuring 15522 that PHI node arguments remain in the loop (this ensures that 15523 loop-closed SSA form is preserved). Only useful on GIMPLE. 15524 15525 Finally, there are some higher-level loop transformations implemented. 15526 While some of them are written so that they should work on non-innermost 15527 loops, they are mostly untested in that case, and at the moment, they 15528 are only reliable for the innermost loops: 15529 15530 * `create_iv': Creates a new induction variable. Only works on 15531 GIMPLE. `standard_iv_increment_position' can be used to find a 15532 suitable place for the iv increment. 15533 15534 * `duplicate_loop_to_header_edge', 15535 `tree_duplicate_loop_to_header_edge': These functions (on RTL and 15536 on GIMPLE) duplicate the body of the loop prescribed number of 15537 times on one of the edges entering loop header, thus performing 15538 either loop unrolling or loop peeling. `can_duplicate_loop_p' 15539 (`can_unroll_loop_p' on GIMPLE) must be true for the duplicated 15540 loop. 15541 15542 * `loop_version', `tree_ssa_loop_version': These function create a 15543 copy of a loop, and a branch before them that selects one of them 15544 depending on the prescribed condition. This is useful for 15545 optimizations that need to verify some assumptions in runtime (one 15546 of the copies of the loop is usually left unchanged, while the 15547 other one is transformed in some way). 15548 15549 * `tree_unroll_loop': Unrolls the loop, including peeling the extra 15550 iterations to make the number of iterations divisible by unroll 15551 factor, updating the exit condition, and removing the exits that 15552 now cannot be taken. Works only on GIMPLE. 15553 15554 15555 File: gccint.info, Node: LCSSA, Next: Scalar evolutions, Prev: Loop manipulation, Up: Loop Analysis and Representation 15556 15557 14.4 Loop-closed SSA form 15558 ========================= 15559 15560 Throughout the loop optimizations on tree level, one extra condition is 15561 enforced on the SSA form: No SSA name is used outside of the loop in 15562 that it is defined. The SSA form satisfying this condition is called 15563 "loop-closed SSA form" - LCSSA. To enforce LCSSA, PHI nodes must be 15564 created at the exits of the loops for the SSA names that are used 15565 outside of them. Only the real operands (not virtual SSA names) are 15566 held in LCSSA, in order to save memory. 15567 15568 There are various benefits of LCSSA: 15569 15570 * Many optimizations (value range analysis, final value replacement) 15571 are interested in the values that are defined in the loop and used 15572 outside of it, i.e., exactly those for that we create new PHI 15573 nodes. 15574 15575 * In induction variable analysis, it is not necessary to specify the 15576 loop in that the analysis should be performed - the scalar 15577 evolution analysis always returns the results with respect to the 15578 loop in that the SSA name is defined. 15579 15580 * It makes updating of SSA form during loop transformations simpler. 15581 Without LCSSA, operations like loop unrolling may force creation 15582 of PHI nodes arbitrarily far from the loop, while in LCSSA, the 15583 SSA form can be updated locally. However, since we only keep real 15584 operands in LCSSA, we cannot use this advantage (we could have 15585 local updating of real operands, but it is not much more efficient 15586 than to use generic SSA form updating for it as well; the amount 15587 of changes to SSA is the same). 15588 15589 However, it also means LCSSA must be updated. This is usually 15590 straightforward, unless you create a new value in loop and use it 15591 outside, or unless you manipulate loop exit edges (functions are 15592 provided to make these manipulations simple). 15593 `rewrite_into_loop_closed_ssa' is used to rewrite SSA form to LCSSA, 15594 and `verify_loop_closed_ssa' to check that the invariant of LCSSA is 15595 preserved. 15596 15597 15598 File: gccint.info, Node: Scalar evolutions, Next: loop-iv, Prev: LCSSA, Up: Loop Analysis and Representation 15599 15600 14.5 Scalar evolutions 15601 ====================== 15602 15603 Scalar evolutions (SCEV) are used to represent results of induction 15604 variable analysis on GIMPLE. They enable us to represent variables with 15605 complicated behavior in a simple and consistent way (we only use it to 15606 express values of polynomial induction variables, but it is possible to 15607 extend it). The interfaces to SCEV analysis are declared in 15608 `tree-scalar-evolution.h'. To use scalar evolutions analysis, 15609 `scev_initialize' must be used. To stop using SCEV, `scev_finalize' 15610 should be used. SCEV analysis caches results in order to save time and 15611 memory. This cache however is made invalid by most of the loop 15612 transformations, including removal of code. If such a transformation 15613 is performed, `scev_reset' must be called to clean the caches. 15614 15615 Given an SSA name, its behavior in loops can be analyzed using the 15616 `analyze_scalar_evolution' function. The returned SCEV however does 15617 not have to be fully analyzed and it may contain references to other 15618 SSA names defined in the loop. To resolve these (potentially 15619 recursive) references, `instantiate_parameters' or `resolve_mixers' 15620 functions must be used. `instantiate_parameters' is useful when you 15621 use the results of SCEV only for some analysis, and when you work with 15622 whole nest of loops at once. It will try replacing all SSA names by 15623 their SCEV in all loops, including the super-loops of the current loop, 15624 thus providing a complete information about the behavior of the 15625 variable in the loop nest. `resolve_mixers' is useful if you work with 15626 only one loop at a time, and if you possibly need to create code based 15627 on the value of the induction variable. It will only resolve the SSA 15628 names defined in the current loop, leaving the SSA names defined 15629 outside unchanged, even if their evolution in the outer loops is known. 15630 15631 The SCEV is a normal tree expression, except for the fact that it may 15632 contain several special tree nodes. One of them is `SCEV_NOT_KNOWN', 15633 used for SSA names whose value cannot be expressed. The other one is 15634 `POLYNOMIAL_CHREC'. Polynomial chrec has three arguments - base, step 15635 and loop (both base and step may contain further polynomial chrecs). 15636 Type of the expression and of base and step must be the same. A 15637 variable has evolution `POLYNOMIAL_CHREC(base, step, loop)' if it is 15638 (in the specified loop) equivalent to `x_1' in the following example 15639 15640 while (...) 15641 { 15642 x_1 = phi (base, x_2); 15643 x_2 = x_1 + step; 15644 } 15645 15646 Note that this includes the language restrictions on the operations. 15647 For example, if we compile C code and `x' has signed type, then the 15648 overflow in addition would cause undefined behavior, and we may assume 15649 that this does not happen. Hence, the value with this SCEV cannot 15650 overflow (which restricts the number of iterations of such a loop). 15651 15652 In many cases, one wants to restrict the attention just to affine 15653 induction variables. In this case, the extra expressive power of SCEV 15654 is not useful, and may complicate the optimizations. In this case, 15655 `simple_iv' function may be used to analyze a value - the result is a 15656 loop-invariant base and step. 15657 15658 15659 File: gccint.info, Node: loop-iv, Next: Number of iterations, Prev: Scalar evolutions, Up: Loop Analysis and Representation 15660 15661 14.6 IV analysis on RTL 15662 ======================= 15663 15664 The induction variable on RTL is simple and only allows analysis of 15665 affine induction variables, and only in one loop at once. The interface 15666 is declared in `cfgloop.h'. Before analyzing induction variables in a 15667 loop L, `iv_analysis_loop_init' function must be called on L. After 15668 the analysis (possibly calling `iv_analysis_loop_init' for several 15669 loops) is finished, `iv_analysis_done' should be called. The following 15670 functions can be used to access the results of the analysis: 15671 15672 * `iv_analyze': Analyzes a single register used in the given insn. 15673 If no use of the register in this insn is found, the following 15674 insns are scanned, so that this function can be called on the insn 15675 returned by get_condition. 15676 15677 * `iv_analyze_result': Analyzes result of the assignment in the 15678 given insn. 15679 15680 * `iv_analyze_expr': Analyzes a more complicated expression. All 15681 its operands are analyzed by `iv_analyze', and hence they must be 15682 used in the specified insn or one of the following insns. 15683 15684 The description of the induction variable is provided in `struct 15685 rtx_iv'. In order to handle subregs, the representation is a bit 15686 complicated; if the value of the `extend' field is not `UNKNOWN', the 15687 value of the induction variable in the i-th iteration is 15688 15689 delta + mult * extend_{extend_mode} (subreg_{mode} (base + i * step)), 15690 15691 with the following exception: if `first_special' is true, then the 15692 value in the first iteration (when `i' is zero) is `delta + mult * 15693 base'. However, if `extend' is equal to `UNKNOWN', then 15694 `first_special' must be false, `delta' 0, `mult' 1 and the value in the 15695 i-th iteration is 15696 15697 subreg_{mode} (base + i * step) 15698 15699 The function `get_iv_value' can be used to perform these calculations. 15700 15701 15702 File: gccint.info, Node: Number of iterations, Next: Dependency analysis, Prev: loop-iv, Up: Loop Analysis and Representation 15703 15704 14.7 Number of iterations analysis 15705 ================================== 15706 15707 Both on GIMPLE and on RTL, there are functions available to determine 15708 the number of iterations of a loop, with a similar interface. The 15709 number of iterations of a loop in GCC is defined as the number of 15710 executions of the loop latch. In many cases, it is not possible to 15711 determine the number of iterations unconditionally - the determined 15712 number is correct only if some assumptions are satisfied. The analysis 15713 tries to verify these conditions using the information contained in the 15714 program; if it fails, the conditions are returned together with the 15715 result. The following information and conditions are provided by the 15716 analysis: 15717 15718 * `assumptions': If this condition is false, the rest of the 15719 information is invalid. 15720 15721 * `noloop_assumptions' on RTL, `may_be_zero' on GIMPLE: If this 15722 condition is true, the loop exits in the first iteration. 15723 15724 * `infinite': If this condition is true, the loop is infinite. This 15725 condition is only available on RTL. On GIMPLE, conditions for 15726 finiteness of the loop are included in `assumptions'. 15727 15728 * `niter_expr' on RTL, `niter' on GIMPLE: The expression that gives 15729 number of iterations. The number of iterations is defined as the 15730 number of executions of the loop latch. 15731 15732 Both on GIMPLE and on RTL, it necessary for the induction variable 15733 analysis framework to be initialized (SCEV on GIMPLE, loop-iv on RTL). 15734 On GIMPLE, the results are stored to `struct tree_niter_desc' 15735 structure. Number of iterations before the loop is exited through a 15736 given exit can be determined using `number_of_iterations_exit' 15737 function. On RTL, the results are returned in `struct niter_desc' 15738 structure. The corresponding function is named `check_simple_exit'. 15739 There are also functions that pass through all the exits of a loop and 15740 try to find one with easy to determine number of iterations - 15741 `find_loop_niter' on GIMPLE and `find_simple_exit' on RTL. Finally, 15742 there are functions that provide the same information, but additionally 15743 cache it, so that repeated calls to number of iterations are not so 15744 costly - `number_of_latch_executions' on GIMPLE and 15745 `get_simple_loop_desc' on RTL. 15746 15747 Note that some of these functions may behave slightly differently than 15748 others - some of them return only the expression for the number of 15749 iterations, and fail if there are some assumptions. The function 15750 `number_of_latch_executions' works only for single-exit loops. The 15751 function `number_of_cond_exit_executions' can be used to determine 15752 number of executions of the exit condition of a single-exit loop (i.e., 15753 the `number_of_latch_executions' increased by one). 15754 15755 15756 File: gccint.info, Node: Dependency analysis, Next: Lambda, Prev: Number of iterations, Up: Loop Analysis and Representation 15757 15758 14.8 Data Dependency Analysis 15759 ============================= 15760 15761 The code for the data dependence analysis can be found in 15762 `tree-data-ref.c' and its interface and data structures are described 15763 in `tree-data-ref.h'. The function that computes the data dependences 15764 for all the array and pointer references for a given loop is 15765 `compute_data_dependences_for_loop'. This function is currently used 15766 by the linear loop transform and the vectorization passes. Before 15767 calling this function, one has to allocate two vectors: a first vector 15768 will contain the set of data references that are contained in the 15769 analyzed loop body, and the second vector will contain the dependence 15770 relations between the data references. Thus if the vector of data 15771 references is of size `n', the vector containing the dependence 15772 relations will contain `n*n' elements. However if the analyzed loop 15773 contains side effects, such as calls that potentially can interfere 15774 with the data references in the current analyzed loop, the analysis 15775 stops while scanning the loop body for data references, and inserts a 15776 single `chrec_dont_know' in the dependence relation array. 15777 15778 The data references are discovered in a particular order during the 15779 scanning of the loop body: the loop body is analyzed in execution order, 15780 and the data references of each statement are pushed at the end of the 15781 data reference array. Two data references syntactically occur in the 15782 program in the same order as in the array of data references. This 15783 syntactic order is important in some classical data dependence tests, 15784 and mapping this order to the elements of this array avoids costly 15785 queries to the loop body representation. 15786 15787 Three types of data references are currently handled: ARRAY_REF, 15788 INDIRECT_REF and COMPONENT_REF. The data structure for the data 15789 reference is `data_reference', where `data_reference_p' is a name of a 15790 pointer to the data reference structure. The structure contains the 15791 following elements: 15792 15793 * `base_object_info': Provides information about the base object of 15794 the data reference and its access functions. These access functions 15795 represent the evolution of the data reference in the loop relative 15796 to its base, in keeping with the classical meaning of the data 15797 reference access function for the support of arrays. For example, 15798 for a reference `a.b[i][j]', the base object is `a.b' and the 15799 access functions, one for each array subscript, are: `{i_init, + 15800 i_step}_1, {j_init, +, j_step}_2'. 15801 15802 * `first_location_in_loop': Provides information about the first 15803 location accessed by the data reference in the loop and about the 15804 access function used to represent evolution relative to this 15805 location. This data is used to support pointers, and is not used 15806 for arrays (for which we have base objects). Pointer accesses are 15807 represented as a one-dimensional access that starts from the first 15808 location accessed in the loop. For example: 15809 15810 for1 i 15811 for2 j 15812 *((int *)p + i + j) = a[i][j]; 15813 15814 The access function of the pointer access is `{0, + 4B}_for2' 15815 relative to `p + i'. The access functions of the array are 15816 `{i_init, + i_step}_for1' and `{j_init, +, j_step}_for2' relative 15817 to `a'. 15818 15819 Usually, the object the pointer refers to is either unknown, or we 15820 can't prove that the access is confined to the boundaries of a 15821 certain object. 15822 15823 Two data references can be compared only if at least one of these 15824 two representations has all its fields filled for both data 15825 references. 15826 15827 The current strategy for data dependence tests is as follows: If 15828 both `a' and `b' are represented as arrays, compare 15829 `a.base_object' and `b.base_object'; if they are equal, apply 15830 dependence tests (use access functions based on base_objects). 15831 Else if both `a' and `b' are represented as pointers, compare 15832 `a.first_location' and `b.first_location'; if they are equal, 15833 apply dependence tests (use access functions based on first 15834 location). However, if `a' and `b' are represented differently, 15835 only try to prove that the bases are definitely different. 15836 15837 * Aliasing information. 15838 15839 * Alignment information. 15840 15841 The structure describing the relation between two data references is 15842 `data_dependence_relation' and the shorter name for a pointer to such a 15843 structure is `ddr_p'. This structure contains: 15844 15845 * a pointer to each data reference, 15846 15847 * a tree node `are_dependent' that is set to `chrec_known' if the 15848 analysis has proved that there is no dependence between these two 15849 data references, `chrec_dont_know' if the analysis was not able to 15850 determine any useful result and potentially there could exist a 15851 dependence between these data references, and `are_dependent' is 15852 set to `NULL_TREE' if there exist a dependence relation between the 15853 data references, and the description of this dependence relation is 15854 given in the `subscripts', `dir_vects', and `dist_vects' arrays, 15855 15856 * a boolean that determines whether the dependence relation can be 15857 represented by a classical distance vector, 15858 15859 * an array `subscripts' that contains a description of each 15860 subscript of the data references. Given two array accesses a 15861 subscript is the tuple composed of the access functions for a given 15862 dimension. For example, given `A[f1][f2][f3]' and 15863 `B[g1][g2][g3]', there are three subscripts: `(f1, g1), (f2, g2), 15864 (f3, g3)'. 15865 15866 * two arrays `dir_vects' and `dist_vects' that contain classical 15867 representations of the data dependences under the form of 15868 direction and distance dependence vectors, 15869 15870 * an array of loops `loop_nest' that contains the loops to which the 15871 distance and direction vectors refer to. 15872 15873 Several functions for pretty printing the information extracted by the 15874 data dependence analysis are available: `dump_ddrs' prints with a 15875 maximum verbosity the details of a data dependence relations array, 15876 `dump_dist_dir_vectors' prints only the classical distance and 15877 direction vectors for a data dependence relations array, and 15878 `dump_data_references' prints the details of the data references 15879 contained in a data reference array. 15880 15881 15882 File: gccint.info, Node: Lambda, Next: Omega, Prev: Dependency analysis, Up: Loop Analysis and Representation 15883 15884 14.9 Linear loop transformations framework 15885 ========================================== 15886 15887 Lambda is a framework that allows transformations of loops using 15888 non-singular matrix based transformations of the iteration space and 15889 loop bounds. This allows compositions of skewing, scaling, interchange, 15890 and reversal transformations. These transformations are often used to 15891 improve cache behavior or remove inner loop dependencies to allow 15892 parallelization and vectorization to take place. 15893 15894 To perform these transformations, Lambda requires that the loopnest be 15895 converted into a internal form that can be matrix transformed easily. 15896 To do this conversion, the function `gcc_loopnest_to_lambda_loopnest' 15897 is provided. If the loop cannot be transformed using lambda, this 15898 function will return NULL. 15899 15900 Once a `lambda_loopnest' is obtained from the conversion function, it 15901 can be transformed by using `lambda_loopnest_transform', which takes a 15902 transformation matrix to apply. Note that it is up to the caller to 15903 verify that the transformation matrix is legal to apply to the loop 15904 (dependence respecting, etc). Lambda simply applies whatever matrix it 15905 is told to provide. It can be extended to make legal matrices out of 15906 any non-singular matrix, but this is not currently implemented. 15907 Legality of a matrix for a given loopnest can be verified using 15908 `lambda_transform_legal_p'. 15909 15910 Given a transformed loopnest, conversion back into gcc IR is done by 15911 `lambda_loopnest_to_gcc_loopnest'. This function will modify the loops 15912 so that they match the transformed loopnest. 15913 15914 15915 File: gccint.info, Node: Omega, Prev: Lambda, Up: Loop Analysis and Representation 15916 15917 14.10 Omega a solver for linear programming problems 15918 ==================================================== 15919 15920 The data dependence analysis contains several solvers triggered 15921 sequentially from the less complex ones to the more sophisticated. For 15922 ensuring the consistency of the results of these solvers, a data 15923 dependence check pass has been implemented based on two different 15924 solvers. The second method that has been integrated to GCC is based on 15925 the Omega dependence solver, written in the 1990's by William Pugh and 15926 David Wonnacott. Data dependence tests can be formulated using a 15927 subset of the Presburger arithmetics that can be translated to linear 15928 constraint systems. These linear constraint systems can then be solved 15929 using the Omega solver. 15930 15931 The Omega solver is using Fourier-Motzkin's algorithm for variable 15932 elimination: a linear constraint system containing `n' variables is 15933 reduced to a linear constraint system with `n-1' variables. The Omega 15934 solver can also be used for solving other problems that can be 15935 expressed under the form of a system of linear equalities and 15936 inequalities. The Omega solver is known to have an exponential worst 15937 case, also known under the name of "omega nightmare" in the literature, 15938 but in practice, the omega test is known to be efficient for the common 15939 data dependence tests. 15940 15941 The interface used by the Omega solver for describing the linear 15942 programming problems is described in `omega.h', and the solver is 15943 `omega_solve_problem'. 15944 15945 15946 File: gccint.info, Node: Control Flow, Next: Loop Analysis and Representation, Prev: RTL, Up: Top 15947 15948 15 Control Flow Graph 15949 ********************* 15950 15951 A control flow graph (CFG) is a data structure built on top of the 15952 intermediate code representation (the RTL or `tree' instruction stream) 15953 abstracting the control flow behavior of a function that is being 15954 compiled. The CFG is a directed graph where the vertices represent 15955 basic blocks and edges represent possible transfer of control flow from 15956 one basic block to another. The data structures used to represent the 15957 control flow graph are defined in `basic-block.h'. 15958 15959 * Menu: 15960 15961 * Basic Blocks:: The definition and representation of basic blocks. 15962 * Edges:: Types of edges and their representation. 15963 * Profile information:: Representation of frequencies and probabilities. 15964 * Maintaining the CFG:: Keeping the control flow graph and up to date. 15965 * Liveness information:: Using and maintaining liveness information. 15966 15967 15968 File: gccint.info, Node: Basic Blocks, Next: Edges, Up: Control Flow 15969 15970 15.1 Basic Blocks 15971 ================= 15972 15973 A basic block is a straight-line sequence of code with only one entry 15974 point and only one exit. In GCC, basic blocks are represented using 15975 the `basic_block' data type. 15976 15977 Two pointer members of the `basic_block' structure are the pointers 15978 `next_bb' and `prev_bb'. These are used to keep doubly linked chain of 15979 basic blocks in the same order as the underlying instruction stream. 15980 The chain of basic blocks is updated transparently by the provided API 15981 for manipulating the CFG. The macro `FOR_EACH_BB' can be used to visit 15982 all the basic blocks in lexicographical order. Dominator traversals 15983 are also possible using `walk_dominator_tree'. Given two basic blocks 15984 A and B, block A dominates block B if A is _always_ executed before B. 15985 15986 The `BASIC_BLOCK' array contains all basic blocks in an unspecified 15987 order. Each `basic_block' structure has a field that holds a unique 15988 integer identifier `index' that is the index of the block in the 15989 `BASIC_BLOCK' array. The total number of basic blocks in the function 15990 is `n_basic_blocks'. Both the basic block indices and the total number 15991 of basic blocks may vary during the compilation process, as passes 15992 reorder, create, duplicate, and destroy basic blocks. The index for 15993 any block should never be greater than `last_basic_block'. 15994 15995 Special basic blocks represent possible entry and exit points of a 15996 function. These blocks are called `ENTRY_BLOCK_PTR' and 15997 `EXIT_BLOCK_PTR'. These blocks do not contain any code, and are not 15998 elements of the `BASIC_BLOCK' array. Therefore they have been assigned 15999 unique, negative index numbers. 16000 16001 Each `basic_block' also contains pointers to the first instruction 16002 (the "head") and the last instruction (the "tail") or "end" of the 16003 instruction stream contained in a basic block. In fact, since the 16004 `basic_block' data type is used to represent blocks in both major 16005 intermediate representations of GCC (`tree' and RTL), there are 16006 pointers to the head and end of a basic block for both representations. 16007 16008 For RTL, these pointers are `rtx head, end'. In the RTL function 16009 representation, the head pointer always points either to a 16010 `NOTE_INSN_BASIC_BLOCK' or to a `CODE_LABEL', if present. In the RTL 16011 representation of a function, the instruction stream contains not only 16012 the "real" instructions, but also "notes". Any function that moves or 16013 duplicates the basic blocks needs to take care of updating of these 16014 notes. Many of these notes expect that the instruction stream consists 16015 of linear regions, making such updates difficult. The 16016 `NOTE_INSN_BASIC_BLOCK' note is the only kind of note that may appear 16017 in the instruction stream contained in a basic block. The instruction 16018 stream of a basic block always follows a `NOTE_INSN_BASIC_BLOCK', but 16019 zero or more `CODE_LABEL' nodes can precede the block note. A basic 16020 block ends by control flow instruction or last instruction before 16021 following `CODE_LABEL' or `NOTE_INSN_BASIC_BLOCK'. A `CODE_LABEL' 16022 cannot appear in the instruction stream of a basic block. 16023 16024 In addition to notes, the jump table vectors are also represented as 16025 "pseudo-instructions" inside the insn stream. These vectors never 16026 appear in the basic block and should always be placed just after the 16027 table jump instructions referencing them. After removing the 16028 table-jump it is often difficult to eliminate the code computing the 16029 address and referencing the vector, so cleaning up these vectors is 16030 postponed until after liveness analysis. Thus the jump table vectors 16031 may appear in the insn stream unreferenced and without any purpose. 16032 Before any edge is made "fall-thru", the existence of such construct in 16033 the way needs to be checked by calling `can_fallthru' function. 16034 16035 For the `tree' representation, the head and end of the basic block are 16036 being pointed to by the `stmt_list' field, but this special `tree' 16037 should never be referenced directly. Instead, at the tree level 16038 abstract containers and iterators are used to access statements and 16039 expressions in basic blocks. These iterators are called "block 16040 statement iterators" (BSIs). Grep for `^bsi' in the various `tree-*' 16041 files. The following snippet will pretty-print all the statements of 16042 the program in the GIMPLE representation. 16043 16044 FOR_EACH_BB (bb) 16045 { 16046 block_stmt_iterator si; 16047 16048 for (si = bsi_start (bb); !bsi_end_p (si); bsi_next (&si)) 16049 { 16050 tree stmt = bsi_stmt (si); 16051 print_generic_stmt (stderr, stmt, 0); 16052 } 16053 } 16054 16055 16056 File: gccint.info, Node: Edges, Next: Profile information, Prev: Basic Blocks, Up: Control Flow 16057 16058 15.2 Edges 16059 ========== 16060 16061 Edges represent possible control flow transfers from the end of some 16062 basic block A to the head of another basic block B. We say that A is a 16063 predecessor of B, and B is a successor of A. Edges are represented in 16064 GCC with the `edge' data type. Each `edge' acts as a link between two 16065 basic blocks: the `src' member of an edge points to the predecessor 16066 basic block of the `dest' basic block. The members `preds' and `succs' 16067 of the `basic_block' data type point to type-safe vectors of edges to 16068 the predecessors and successors of the block. 16069 16070 When walking the edges in an edge vector, "edge iterators" should be 16071 used. Edge iterators are constructed using the `edge_iterator' data 16072 structure and several methods are available to operate on them: 16073 16074 `ei_start' 16075 This function initializes an `edge_iterator' that points to the 16076 first edge in a vector of edges. 16077 16078 `ei_last' 16079 This function initializes an `edge_iterator' that points to the 16080 last edge in a vector of edges. 16081 16082 `ei_end_p' 16083 This predicate is `true' if an `edge_iterator' represents the last 16084 edge in an edge vector. 16085 16086 `ei_one_before_end_p' 16087 This predicate is `true' if an `edge_iterator' represents the 16088 second last edge in an edge vector. 16089 16090 `ei_next' 16091 This function takes a pointer to an `edge_iterator' and makes it 16092 point to the next edge in the sequence. 16093 16094 `ei_prev' 16095 This function takes a pointer to an `edge_iterator' and makes it 16096 point to the previous edge in the sequence. 16097 16098 `ei_edge' 16099 This function returns the `edge' currently pointed to by an 16100 `edge_iterator'. 16101 16102 `ei_safe_safe' 16103 This function returns the `edge' currently pointed to by an 16104 `edge_iterator', but returns `NULL' if the iterator is pointing at 16105 the end of the sequence. This function has been provided for 16106 existing code makes the assumption that a `NULL' edge indicates 16107 the end of the sequence. 16108 16109 16110 The convenience macro `FOR_EACH_EDGE' can be used to visit all of the 16111 edges in a sequence of predecessor or successor edges. It must not be 16112 used when an element might be removed during the traversal, otherwise 16113 elements will be missed. Here is an example of how to use the macro: 16114 16115 edge e; 16116 edge_iterator ei; 16117 16118 FOR_EACH_EDGE (e, ei, bb->succs) 16119 { 16120 if (e->flags & EDGE_FALLTHRU) 16121 break; 16122 } 16123 16124 There are various reasons why control flow may transfer from one block 16125 to another. One possibility is that some instruction, for example a 16126 `CODE_LABEL', in a linearized instruction stream just always starts a 16127 new basic block. In this case a "fall-thru" edge links the basic block 16128 to the first following basic block. But there are several other 16129 reasons why edges may be created. The `flags' field of the `edge' data 16130 type is used to store information about the type of edge we are dealing 16131 with. Each edge is of one of the following types: 16132 16133 _jump_ 16134 No type flags are set for edges corresponding to jump instructions. 16135 These edges are used for unconditional or conditional jumps and in 16136 RTL also for table jumps. They are the easiest to manipulate as 16137 they may be freely redirected when the flow graph is not in SSA 16138 form. 16139 16140 _fall-thru_ 16141 Fall-thru edges are present in case where the basic block may 16142 continue execution to the following one without branching. These 16143 edges have the `EDGE_FALLTHRU' flag set. Unlike other types of 16144 edges, these edges must come into the basic block immediately 16145 following in the instruction stream. The function 16146 `force_nonfallthru' is available to insert an unconditional jump 16147 in the case that redirection is needed. Note that this may 16148 require creation of a new basic block. 16149 16150 _exception handling_ 16151 Exception handling edges represent possible control transfers from 16152 a trapping instruction to an exception handler. The definition of 16153 "trapping" varies. In C++, only function calls can throw, but for 16154 Java, exceptions like division by zero or segmentation fault are 16155 defined and thus each instruction possibly throwing this kind of 16156 exception needs to be handled as control flow instruction. 16157 Exception edges have the `EDGE_ABNORMAL' and `EDGE_EH' flags set. 16158 16159 When updating the instruction stream it is easy to change possibly 16160 trapping instruction to non-trapping, by simply removing the 16161 exception edge. The opposite conversion is difficult, but should 16162 not happen anyway. The edges can be eliminated via 16163 `purge_dead_edges' call. 16164 16165 In the RTL representation, the destination of an exception edge is 16166 specified by `REG_EH_REGION' note attached to the insn. In case 16167 of a trapping call the `EDGE_ABNORMAL_CALL' flag is set too. In 16168 the `tree' representation, this extra flag is not set. 16169 16170 In the RTL representation, the predicate `may_trap_p' may be used 16171 to check whether instruction still may trap or not. For the tree 16172 representation, the `tree_could_trap_p' predicate is available, 16173 but this predicate only checks for possible memory traps, as in 16174 dereferencing an invalid pointer location. 16175 16176 _sibling calls_ 16177 Sibling calls or tail calls terminate the function in a 16178 non-standard way and thus an edge to the exit must be present. 16179 `EDGE_SIBCALL' and `EDGE_ABNORMAL' are set in such case. These 16180 edges only exist in the RTL representation. 16181 16182 _computed jumps_ 16183 Computed jumps contain edges to all labels in the function 16184 referenced from the code. All those edges have `EDGE_ABNORMAL' 16185 flag set. The edges used to represent computed jumps often cause 16186 compile time performance problems, since functions consisting of 16187 many taken labels and many computed jumps may have _very_ dense 16188 flow graphs, so these edges need to be handled with special care. 16189 During the earlier stages of the compilation process, GCC tries to 16190 avoid such dense flow graphs by factoring computed jumps. For 16191 example, given the following series of jumps, 16192 16193 goto *x; 16194 [ ... ] 16195 16196 goto *x; 16197 [ ... ] 16198 16199 goto *x; 16200 [ ... ] 16201 16202 factoring the computed jumps results in the following code sequence 16203 which has a much simpler flow graph: 16204 16205 goto y; 16206 [ ... ] 16207 16208 goto y; 16209 [ ... ] 16210 16211 goto y; 16212 [ ... ] 16213 16214 y: 16215 goto *x; 16216 16217 However, the classic problem with this transformation is that it 16218 has a runtime cost in there resulting code: An extra jump. 16219 Therefore, the computed jumps are un-factored in the later passes 16220 of the compiler. Be aware of that when you work on passes in that 16221 area. There have been numerous examples already where the compile 16222 time for code with unfactored computed jumps caused some serious 16223 headaches. 16224 16225 _nonlocal goto handlers_ 16226 GCC allows nested functions to return into caller using a `goto' 16227 to a label passed to as an argument to the callee. The labels 16228 passed to nested functions contain special code to cleanup after 16229 function call. Such sections of code are referred to as "nonlocal 16230 goto receivers". If a function contains such nonlocal goto 16231 receivers, an edge from the call to the label is created with the 16232 `EDGE_ABNORMAL' and `EDGE_ABNORMAL_CALL' flags set. 16233 16234 _function entry points_ 16235 By definition, execution of function starts at basic block 0, so 16236 there is always an edge from the `ENTRY_BLOCK_PTR' to basic block 16237 0. There is no `tree' representation for alternate entry points at 16238 this moment. In RTL, alternate entry points are specified by 16239 `CODE_LABEL' with `LABEL_ALTERNATE_NAME' defined. This feature is 16240 currently used for multiple entry point prologues and is limited 16241 to post-reload passes only. This can be used by back-ends to emit 16242 alternate prologues for functions called from different contexts. 16243 In future full support for multiple entry functions defined by 16244 Fortran 90 needs to be implemented. 16245 16246 _function exits_ 16247 In the pre-reload representation a function terminates after the 16248 last instruction in the insn chain and no explicit return 16249 instructions are used. This corresponds to the fall-thru edge 16250 into exit block. After reload, optimal RTL epilogues are used 16251 that use explicit (conditional) return instructions that are 16252 represented by edges with no flags set. 16253 16254 16255 16256 File: gccint.info, Node: Profile information, Next: Maintaining the CFG, Prev: Edges, Up: Control Flow 16257 16258 15.3 Profile information 16259 ======================== 16260 16261 In many cases a compiler must make a choice whether to trade speed in 16262 one part of code for speed in another, or to trade code size for code 16263 speed. In such cases it is useful to know information about how often 16264 some given block will be executed. That is the purpose for maintaining 16265 profile within the flow graph. GCC can handle profile information 16266 obtained through "profile feedback", but it can also estimate branch 16267 probabilities based on statics and heuristics. 16268 16269 The feedback based profile is produced by compiling the program with 16270 instrumentation, executing it on a train run and reading the numbers of 16271 executions of basic blocks and edges back to the compiler while 16272 re-compiling the program to produce the final executable. This method 16273 provides very accurate information about where a program spends most of 16274 its time on the train run. Whether it matches the average run of 16275 course depends on the choice of train data set, but several studies 16276 have shown that the behavior of a program usually changes just 16277 marginally over different data sets. 16278 16279 When profile feedback is not available, the compiler may be asked to 16280 attempt to predict the behavior of each branch in the program using a 16281 set of heuristics (see `predict.def' for details) and compute estimated 16282 frequencies of each basic block by propagating the probabilities over 16283 the graph. 16284 16285 Each `basic_block' contains two integer fields to represent profile 16286 information: `frequency' and `count'. The `frequency' is an estimation 16287 how often is basic block executed within a function. It is represented 16288 as an integer scaled in the range from 0 to `BB_FREQ_BASE'. The most 16289 frequently executed basic block in function is initially set to 16290 `BB_FREQ_BASE' and the rest of frequencies are scaled accordingly. 16291 During optimization, the frequency of the most frequent basic block can 16292 both decrease (for instance by loop unrolling) or grow (for instance by 16293 cross-jumping optimization), so scaling sometimes has to be performed 16294 multiple times. 16295 16296 The `count' contains hard-counted numbers of execution measured during 16297 training runs and is nonzero only when profile feedback is available. 16298 This value is represented as the host's widest integer (typically a 64 16299 bit integer) of the special type `gcov_type'. 16300 16301 Most optimization passes can use only the frequency information of a 16302 basic block, but a few passes may want to know hard execution counts. 16303 The frequencies should always match the counts after scaling, however 16304 during updating of the profile information numerical error may 16305 accumulate into quite large errors. 16306 16307 Each edge also contains a branch probability field: an integer in the 16308 range from 0 to `REG_BR_PROB_BASE'. It represents probability of 16309 passing control from the end of the `src' basic block to the `dest' 16310 basic block, i.e. the probability that control will flow along this 16311 edge. The `EDGE_FREQUENCY' macro is available to compute how 16312 frequently a given edge is taken. There is a `count' field for each 16313 edge as well, representing same information as for a basic block. 16314 16315 The basic block frequencies are not represented in the instruction 16316 stream, but in the RTL representation the edge frequencies are 16317 represented for conditional jumps (via the `REG_BR_PROB' macro) since 16318 they are used when instructions are output to the assembly file and the 16319 flow graph is no longer maintained. 16320 16321 The probability that control flow arrives via a given edge to its 16322 destination basic block is called "reverse probability" and is not 16323 directly represented, but it may be easily computed from frequencies of 16324 basic blocks. 16325 16326 Updating profile information is a delicate task that can unfortunately 16327 not be easily integrated with the CFG manipulation API. Many of the 16328 functions and hooks to modify the CFG, such as 16329 `redirect_edge_and_branch', do not have enough information to easily 16330 update the profile, so updating it is in the majority of cases left up 16331 to the caller. It is difficult to uncover bugs in the profile updating 16332 code, because they manifest themselves only by producing worse code, 16333 and checking profile consistency is not possible because of numeric 16334 error accumulation. Hence special attention needs to be given to this 16335 issue in each pass that modifies the CFG. 16336 16337 It is important to point out that `REG_BR_PROB_BASE' and 16338 `BB_FREQ_BASE' are both set low enough to be possible to compute second 16339 power of any frequency or probability in the flow graph, it is not 16340 possible to even square the `count' field, as modern CPUs are fast 16341 enough to execute $2^32$ operations quickly. 16342 16343 16344 File: gccint.info, Node: Maintaining the CFG, Next: Liveness information, Prev: Profile information, Up: Control Flow 16345 16346 15.4 Maintaining the CFG 16347 ======================== 16348 16349 An important task of each compiler pass is to keep both the control 16350 flow graph and all profile information up-to-date. Reconstruction of 16351 the control flow graph after each pass is not an option, since it may be 16352 very expensive and lost profile information cannot be reconstructed at 16353 all. 16354 16355 GCC has two major intermediate representations, and both use the 16356 `basic_block' and `edge' data types to represent control flow. Both 16357 representations share as much of the CFG maintenance code as possible. 16358 For each representation, a set of "hooks" is defined so that each 16359 representation can provide its own implementation of CFG manipulation 16360 routines when necessary. These hooks are defined in `cfghooks.h'. 16361 There are hooks for almost all common CFG manipulations, including 16362 block splitting and merging, edge redirection and creating and deleting 16363 basic blocks. These hooks should provide everything you need to 16364 maintain and manipulate the CFG in both the RTL and `tree' 16365 representation. 16366 16367 At the moment, the basic block boundaries are maintained transparently 16368 when modifying instructions, so there rarely is a need to move them 16369 manually (such as in case someone wants to output instruction outside 16370 basic block explicitly). Often the CFG may be better viewed as 16371 integral part of instruction chain, than structure built on the top of 16372 it. However, in principle the control flow graph for the `tree' 16373 representation is _not_ an integral part of the representation, in that 16374 a function tree may be expanded without first building a flow graph 16375 for the `tree' representation at all. This happens when compiling 16376 without any `tree' optimization enabled. When the `tree' optimizations 16377 are enabled and the instruction stream is rewritten in SSA form, the 16378 CFG is very tightly coupled with the instruction stream. In 16379 particular, statement insertion and removal has to be done with care. 16380 In fact, the whole `tree' representation can not be easily used or 16381 maintained without proper maintenance of the CFG simultaneously. 16382 16383 In the RTL representation, each instruction has a `BLOCK_FOR_INSN' 16384 value that represents pointer to the basic block that contains the 16385 instruction. In the `tree' representation, the function `bb_for_stmt' 16386 returns a pointer to the basic block containing the queried statement. 16387 16388 When changes need to be applied to a function in its `tree' 16389 representation, "block statement iterators" should be used. These 16390 iterators provide an integrated abstraction of the flow graph and the 16391 instruction stream. Block statement iterators are constructed using 16392 the `block_stmt_iterator' data structure and several modifier are 16393 available, including the following: 16394 16395 `bsi_start' 16396 This function initializes a `block_stmt_iterator' that points to 16397 the first non-empty statement in a basic block. 16398 16399 `bsi_last' 16400 This function initializes a `block_stmt_iterator' that points to 16401 the last statement in a basic block. 16402 16403 `bsi_end_p' 16404 This predicate is `true' if a `block_stmt_iterator' represents the 16405 end of a basic block. 16406 16407 `bsi_next' 16408 This function takes a `block_stmt_iterator' and makes it point to 16409 its successor. 16410 16411 `bsi_prev' 16412 This function takes a `block_stmt_iterator' and makes it point to 16413 its predecessor. 16414 16415 `bsi_insert_after' 16416 This function inserts a statement after the `block_stmt_iterator' 16417 passed in. The final parameter determines whether the statement 16418 iterator is updated to point to the newly inserted statement, or 16419 left pointing to the original statement. 16420 16421 `bsi_insert_before' 16422 This function inserts a statement before the `block_stmt_iterator' 16423 passed in. The final parameter determines whether the statement 16424 iterator is updated to point to the newly inserted statement, or 16425 left pointing to the original statement. 16426 16427 `bsi_remove' 16428 This function removes the `block_stmt_iterator' passed in and 16429 rechains the remaining statements in a basic block, if any. 16430 16431 In the RTL representation, the macros `BB_HEAD' and `BB_END' may be 16432 used to get the head and end `rtx' of a basic block. No abstract 16433 iterators are defined for traversing the insn chain, but you can just 16434 use `NEXT_INSN' and `PREV_INSN' instead. See *Note Insns::. 16435 16436 Usually a code manipulating pass simplifies the instruction stream and 16437 the flow of control, possibly eliminating some edges. This may for 16438 example happen when a conditional jump is replaced with an 16439 unconditional jump, but also when simplifying possibly trapping 16440 instruction to non-trapping while compiling Java. Updating of edges is 16441 not transparent and each optimization pass is required to do so 16442 manually. However only few cases occur in practice. The pass may call 16443 `purge_dead_edges' on a given basic block to remove superfluous edges, 16444 if any. 16445 16446 Another common scenario is redirection of branch instructions, but 16447 this is best modeled as redirection of edges in the control flow graph 16448 and thus use of `redirect_edge_and_branch' is preferred over more low 16449 level functions, such as `redirect_jump' that operate on RTL chain 16450 only. The CFG hooks defined in `cfghooks.h' should provide the 16451 complete API required for manipulating and maintaining the CFG. 16452 16453 It is also possible that a pass has to insert control flow instruction 16454 into the middle of a basic block, thus creating an entry point in the 16455 middle of the basic block, which is impossible by definition: The block 16456 must be split to make sure it only has one entry point, i.e. the head 16457 of the basic block. The CFG hook `split_block' may be used when an 16458 instruction in the middle of a basic block has to become the target of 16459 a jump or branch instruction. 16460 16461 For a global optimizer, a common operation is to split edges in the 16462 flow graph and insert instructions on them. In the RTL representation, 16463 this can be easily done using the `insert_insn_on_edge' function that 16464 emits an instruction "on the edge", caching it for a later 16465 `commit_edge_insertions' call that will take care of moving the 16466 inserted instructions off the edge into the instruction stream 16467 contained in a basic block. This includes the creation of new basic 16468 blocks where needed. In the `tree' representation, the equivalent 16469 functions are `bsi_insert_on_edge' which inserts a block statement 16470 iterator on an edge, and `bsi_commit_edge_inserts' which flushes the 16471 instruction to actual instruction stream. 16472 16473 While debugging the optimization pass, an `verify_flow_info' function 16474 may be useful to find bugs in the control flow graph updating code. 16475 16476 Note that at present, the representation of control flow in the `tree' 16477 representation is discarded before expanding to RTL. Long term the CFG 16478 should be maintained and "expanded" to the RTL representation along 16479 with the function `tree' itself. 16480 16481 16482 File: gccint.info, Node: Liveness information, Prev: Maintaining the CFG, Up: Control Flow 16483 16484 15.5 Liveness information 16485 ========================= 16486 16487 Liveness information is useful to determine whether some register is 16488 "live" at given point of program, i.e. that it contains a value that 16489 may be used at a later point in the program. This information is used, 16490 for instance, during register allocation, as the pseudo registers only 16491 need to be assigned to a unique hard register or to a stack slot if 16492 they are live. The hard registers and stack slots may be freely reused 16493 for other values when a register is dead. 16494 16495 Liveness information is available in the back end starting with 16496 `pass_df_initialize' and ending with `pass_df_finish'. Three flavors 16497 of live analysis are available: With `LR', it is possible to determine 16498 at any point `P' in the function if the register may be used on some 16499 path from `P' to the end of the function. With `UR', it is possible to 16500 determine if there is a path from the beginning of the function to `P' 16501 that defines the variable. `LIVE' is the intersection of the `LR' and 16502 `UR' and a variable is live at `P' if there is both an assignment that 16503 reaches it from the beginning of the function and a uses that can be 16504 reached on some path from `P' to the end of the function. 16505 16506 In general `LIVE' is the most useful of the three. The macros 16507 `DF_[LR,UR,LIVE]_[IN,OUT]' can be used to access this information. The 16508 macros take a basic block number and return a bitmap that is indexed by 16509 the register number. This information is only guaranteed to be up to 16510 date after calls are made to `df_analyze'. See the file `df-core.c' 16511 for details on using the dataflow. 16512 16513 The liveness information is stored partly in the RTL instruction stream 16514 and partly in the flow graph. Local information is stored in the 16515 instruction stream: Each instruction may contain `REG_DEAD' notes 16516 representing that the value of a given register is no longer needed, or 16517 `REG_UNUSED' notes representing that the value computed by the 16518 instruction is never used. The second is useful for instructions 16519 computing multiple values at once. 16520 16521 16522 File: gccint.info, Node: Machine Desc, Next: Target Macros, Prev: Loop Analysis and Representation, Up: Top 16523 16524 16 Machine Descriptions 16525 *********************** 16526 16527 A machine description has two parts: a file of instruction patterns 16528 (`.md' file) and a C header file of macro definitions. 16529 16530 The `.md' file for a target machine contains a pattern for each 16531 instruction that the target machine supports (or at least each 16532 instruction that is worth telling the compiler about). It may also 16533 contain comments. A semicolon causes the rest of the line to be a 16534 comment, unless the semicolon is inside a quoted string. 16535 16536 See the next chapter for information on the C header file. 16537 16538 * Menu: 16539 16540 * Overview:: How the machine description is used. 16541 * Patterns:: How to write instruction patterns. 16542 * Example:: An explained example of a `define_insn' pattern. 16543 * RTL Template:: The RTL template defines what insns match a pattern. 16544 * Output Template:: The output template says how to make assembler code 16545 from such an insn. 16546 * Output Statement:: For more generality, write C code to output 16547 the assembler code. 16548 * Predicates:: Controlling what kinds of operands can be used 16549 for an insn. 16550 * Constraints:: Fine-tuning operand selection. 16551 * Standard Names:: Names mark patterns to use for code generation. 16552 * Pattern Ordering:: When the order of patterns makes a difference. 16553 * Dependent Patterns:: Having one pattern may make you need another. 16554 * Jump Patterns:: Special considerations for patterns for jump insns. 16555 * Looping Patterns:: How to define patterns for special looping insns. 16556 * Insn Canonicalizations::Canonicalization of Instructions 16557 * Expander Definitions::Generating a sequence of several RTL insns 16558 for a standard operation. 16559 * Insn Splitting:: Splitting Instructions into Multiple Instructions. 16560 * Including Patterns:: Including Patterns in Machine Descriptions. 16561 * Peephole Definitions::Defining machine-specific peephole optimizations. 16562 * Insn Attributes:: Specifying the value of attributes for generated insns. 16563 * Conditional Execution::Generating `define_insn' patterns for 16564 predication. 16565 * Constant Definitions::Defining symbolic constants that can be used in the 16566 md file. 16567 * Iterators:: Using iterators to generate patterns from a template. 16568 16569 16570 File: gccint.info, Node: Overview, Next: Patterns, Up: Machine Desc 16571 16572 16.1 Overview of How the Machine Description is Used 16573 ==================================================== 16574 16575 There are three main conversions that happen in the compiler: 16576 16577 1. The front end reads the source code and builds a parse tree. 16578 16579 2. The parse tree is used to generate an RTL insn list based on named 16580 instruction patterns. 16581 16582 3. The insn list is matched against the RTL templates to produce 16583 assembler code. 16584 16585 16586 For the generate pass, only the names of the insns matter, from either 16587 a named `define_insn' or a `define_expand'. The compiler will choose 16588 the pattern with the right name and apply the operands according to the 16589 documentation later in this chapter, without regard for the RTL 16590 template or operand constraints. Note that the names the compiler looks 16591 for are hard-coded in the compiler--it will ignore unnamed patterns and 16592 patterns with names it doesn't know about, but if you don't provide a 16593 named pattern it needs, it will abort. 16594 16595 If a `define_insn' is used, the template given is inserted into the 16596 insn list. If a `define_expand' is used, one of three things happens, 16597 based on the condition logic. The condition logic may manually create 16598 new insns for the insn list, say via `emit_insn()', and invoke `DONE'. 16599 For certain named patterns, it may invoke `FAIL' to tell the compiler 16600 to use an alternate way of performing that task. If it invokes neither 16601 `DONE' nor `FAIL', the template given in the pattern is inserted, as if 16602 the `define_expand' were a `define_insn'. 16603 16604 Once the insn list is generated, various optimization passes convert, 16605 replace, and rearrange the insns in the insn list. This is where the 16606 `define_split' and `define_peephole' patterns get used, for example. 16607 16608 Finally, the insn list's RTL is matched up with the RTL templates in 16609 the `define_insn' patterns, and those patterns are used to emit the 16610 final assembly code. For this purpose, each named `define_insn' acts 16611 like it's unnamed, since the names are ignored. 16612 16613 16614 File: gccint.info, Node: Patterns, Next: Example, Prev: Overview, Up: Machine Desc 16615 16616 16.2 Everything about Instruction Patterns 16617 ========================================== 16618 16619 Each instruction pattern contains an incomplete RTL expression, with 16620 pieces to be filled in later, operand constraints that restrict how the 16621 pieces can be filled in, and an output pattern or C code to generate 16622 the assembler output, all wrapped up in a `define_insn' expression. 16623 16624 A `define_insn' is an RTL expression containing four or five operands: 16625 16626 1. An optional name. The presence of a name indicate that this 16627 instruction pattern can perform a certain standard job for the 16628 RTL-generation pass of the compiler. This pass knows certain 16629 names and will use the instruction patterns with those names, if 16630 the names are defined in the machine description. 16631 16632 The absence of a name is indicated by writing an empty string 16633 where the name should go. Nameless instruction patterns are never 16634 used for generating RTL code, but they may permit several simpler 16635 insns to be combined later on. 16636 16637 Names that are not thus known and used in RTL-generation have no 16638 effect; they are equivalent to no name at all. 16639 16640 For the purpose of debugging the compiler, you may also specify a 16641 name beginning with the `*' character. Such a name is used only 16642 for identifying the instruction in RTL dumps; it is entirely 16643 equivalent to having a nameless pattern for all other purposes. 16644 16645 2. The "RTL template" (*note RTL Template::) is a vector of incomplete 16646 RTL expressions which show what the instruction should look like. 16647 It is incomplete because it may contain `match_operand', 16648 `match_operator', and `match_dup' expressions that stand for 16649 operands of the instruction. 16650 16651 If the vector has only one element, that element is the template 16652 for the instruction pattern. If the vector has multiple elements, 16653 then the instruction pattern is a `parallel' expression containing 16654 the elements described. 16655 16656 3. A condition. This is a string which contains a C expression that 16657 is the final test to decide whether an insn body matches this 16658 pattern. 16659 16660 For a named pattern, the condition (if present) may not depend on 16661 the data in the insn being matched, but only the 16662 target-machine-type flags. The compiler needs to test these 16663 conditions during initialization in order to learn exactly which 16664 named instructions are available in a particular run. 16665 16666 For nameless patterns, the condition is applied only when matching 16667 an individual insn, and only after the insn has matched the 16668 pattern's recognition template. The insn's operands may be found 16669 in the vector `operands'. For an insn where the condition has 16670 once matched, it can't be used to control register allocation, for 16671 example by excluding certain hard registers or hard register 16672 combinations. 16673 16674 4. The "output template": a string that says how to output matching 16675 insns as assembler code. `%' in this string specifies where to 16676 substitute the value of an operand. *Note Output Template::. 16677 16678 When simple substitution isn't general enough, you can specify a 16679 piece of C code to compute the output. *Note Output Statement::. 16680 16681 5. Optionally, a vector containing the values of attributes for insns 16682 matching this pattern. *Note Insn Attributes::. 16683 16684 16685 File: gccint.info, Node: Example, Next: RTL Template, Prev: Patterns, Up: Machine Desc 16686 16687 16.3 Example of `define_insn' 16688 ============================= 16689 16690 Here is an actual example of an instruction pattern, for the 16691 68000/68020. 16692 16693 (define_insn "tstsi" 16694 [(set (cc0) 16695 (match_operand:SI 0 "general_operand" "rm"))] 16696 "" 16697 "* 16698 { 16699 if (TARGET_68020 || ! ADDRESS_REG_P (operands[0])) 16700 return \"tstl %0\"; 16701 return \"cmpl #0,%0\"; 16702 }") 16703 16704 This can also be written using braced strings: 16705 16706 (define_insn "tstsi" 16707 [(set (cc0) 16708 (match_operand:SI 0 "general_operand" "rm"))] 16709 "" 16710 { 16711 if (TARGET_68020 || ! ADDRESS_REG_P (operands[0])) 16712 return "tstl %0"; 16713 return "cmpl #0,%0"; 16714 }) 16715 16716 This is an instruction that sets the condition codes based on the 16717 value of a general operand. It has no condition, so any insn whose RTL 16718 description has the form shown may be handled according to this 16719 pattern. The name `tstsi' means "test a `SImode' value" and tells the 16720 RTL generation pass that, when it is necessary to test such a value, an 16721 insn to do so can be constructed using this pattern. 16722 16723 The output control string is a piece of C code which chooses which 16724 output template to return based on the kind of operand and the specific 16725 type of CPU for which code is being generated. 16726 16727 `"rm"' is an operand constraint. Its meaning is explained below. 16728 16729 16730 File: gccint.info, Node: RTL Template, Next: Output Template, Prev: Example, Up: Machine Desc 16731 16732 16.4 RTL Template 16733 ================= 16734 16735 The RTL template is used to define which insns match the particular 16736 pattern and how to find their operands. For named patterns, the RTL 16737 template also says how to construct an insn from specified operands. 16738 16739 Construction involves substituting specified operands into a copy of 16740 the template. Matching involves determining the values that serve as 16741 the operands in the insn being matched. Both of these activities are 16742 controlled by special expression types that direct matching and 16743 substitution of the operands. 16744 16745 `(match_operand:M N PREDICATE CONSTRAINT)' 16746 This expression is a placeholder for operand number N of the insn. 16747 When constructing an insn, operand number N will be substituted 16748 at this point. When matching an insn, whatever appears at this 16749 position in the insn will be taken as operand number N; but it 16750 must satisfy PREDICATE or this instruction pattern will not match 16751 at all. 16752 16753 Operand numbers must be chosen consecutively counting from zero in 16754 each instruction pattern. There may be only one `match_operand' 16755 expression in the pattern for each operand number. Usually 16756 operands are numbered in the order of appearance in `match_operand' 16757 expressions. In the case of a `define_expand', any operand numbers 16758 used only in `match_dup' expressions have higher values than all 16759 other operand numbers. 16760 16761 PREDICATE is a string that is the name of a function that accepts 16762 two arguments, an expression and a machine mode. *Note 16763 Predicates::. During matching, the function will be called with 16764 the putative operand as the expression and M as the mode argument 16765 (if M is not specified, `VOIDmode' will be used, which normally 16766 causes PREDICATE to accept any mode). If it returns zero, this 16767 instruction pattern fails to match. PREDICATE may be an empty 16768 string; then it means no test is to be done on the operand, so 16769 anything which occurs in this position is valid. 16770 16771 Most of the time, PREDICATE will reject modes other than M--but 16772 not always. For example, the predicate `address_operand' uses M 16773 as the mode of memory ref that the address should be valid for. 16774 Many predicates accept `const_int' nodes even though their mode is 16775 `VOIDmode'. 16776 16777 CONSTRAINT controls reloading and the choice of the best register 16778 class to use for a value, as explained later (*note Constraints::). 16779 If the constraint would be an empty string, it can be omitted. 16780 16781 People are often unclear on the difference between the constraint 16782 and the predicate. The predicate helps decide whether a given 16783 insn matches the pattern. The constraint plays no role in this 16784 decision; instead, it controls various decisions in the case of an 16785 insn which does match. 16786 16787 `(match_scratch:M N CONSTRAINT)' 16788 This expression is also a placeholder for operand number N and 16789 indicates that operand must be a `scratch' or `reg' expression. 16790 16791 When matching patterns, this is equivalent to 16792 16793 (match_operand:M N "scratch_operand" PRED) 16794 16795 but, when generating RTL, it produces a (`scratch':M) expression. 16796 16797 If the last few expressions in a `parallel' are `clobber' 16798 expressions whose operands are either a hard register or 16799 `match_scratch', the combiner can add or delete them when 16800 necessary. *Note Side Effects::. 16801 16802 `(match_dup N)' 16803 This expression is also a placeholder for operand number N. It is 16804 used when the operand needs to appear more than once in the insn. 16805 16806 In construction, `match_dup' acts just like `match_operand': the 16807 operand is substituted into the insn being constructed. But in 16808 matching, `match_dup' behaves differently. It assumes that operand 16809 number N has already been determined by a `match_operand' 16810 appearing earlier in the recognition template, and it matches only 16811 an identical-looking expression. 16812 16813 Note that `match_dup' should not be used to tell the compiler that 16814 a particular register is being used for two operands (example: 16815 `add' that adds one register to another; the second register is 16816 both an input operand and the output operand). Use a matching 16817 constraint (*note Simple Constraints::) for those. `match_dup' is 16818 for the cases where one operand is used in two places in the 16819 template, such as an instruction that computes both a quotient and 16820 a remainder, where the opcode takes two input operands but the RTL 16821 template has to refer to each of those twice; once for the 16822 quotient pattern and once for the remainder pattern. 16823 16824 `(match_operator:M N PREDICATE [OPERANDS...])' 16825 This pattern is a kind of placeholder for a variable RTL expression 16826 code. 16827 16828 When constructing an insn, it stands for an RTL expression whose 16829 expression code is taken from that of operand N, and whose 16830 operands are constructed from the patterns OPERANDS. 16831 16832 When matching an expression, it matches an expression if the 16833 function PREDICATE returns nonzero on that expression _and_ the 16834 patterns OPERANDS match the operands of the expression. 16835 16836 Suppose that the function `commutative_operator' is defined as 16837 follows, to match any expression whose operator is one of the 16838 commutative arithmetic operators of RTL and whose mode is MODE: 16839 16840 int 16841 commutative_integer_operator (x, mode) 16842 rtx x; 16843 enum machine_mode mode; 16844 { 16845 enum rtx_code code = GET_CODE (x); 16846 if (GET_MODE (x) != mode) 16847 return 0; 16848 return (GET_RTX_CLASS (code) == RTX_COMM_ARITH 16849 || code == EQ || code == NE); 16850 } 16851 16852 Then the following pattern will match any RTL expression consisting 16853 of a commutative operator applied to two general operands: 16854 16855 (match_operator:SI 3 "commutative_operator" 16856 [(match_operand:SI 1 "general_operand" "g") 16857 (match_operand:SI 2 "general_operand" "g")]) 16858 16859 Here the vector `[OPERANDS...]' contains two patterns because the 16860 expressions to be matched all contain two operands. 16861 16862 When this pattern does match, the two operands of the commutative 16863 operator are recorded as operands 1 and 2 of the insn. (This is 16864 done by the two instances of `match_operand'.) Operand 3 of the 16865 insn will be the entire commutative expression: use `GET_CODE 16866 (operands[3])' to see which commutative operator was used. 16867 16868 The machine mode M of `match_operator' works like that of 16869 `match_operand': it is passed as the second argument to the 16870 predicate function, and that function is solely responsible for 16871 deciding whether the expression to be matched "has" that mode. 16872 16873 When constructing an insn, argument 3 of the gen-function will 16874 specify the operation (i.e. the expression code) for the 16875 expression to be made. It should be an RTL expression, whose 16876 expression code is copied into a new expression whose operands are 16877 arguments 1 and 2 of the gen-function. The subexpressions of 16878 argument 3 are not used; only its expression code matters. 16879 16880 When `match_operator' is used in a pattern for matching an insn, 16881 it usually best if the operand number of the `match_operator' is 16882 higher than that of the actual operands of the insn. This improves 16883 register allocation because the register allocator often looks at 16884 operands 1 and 2 of insns to see if it can do register tying. 16885 16886 There is no way to specify constraints in `match_operator'. The 16887 operand of the insn which corresponds to the `match_operator' 16888 never has any constraints because it is never reloaded as a whole. 16889 However, if parts of its OPERANDS are matched by `match_operand' 16890 patterns, those parts may have constraints of their own. 16891 16892 `(match_op_dup:M N[OPERANDS...])' 16893 Like `match_dup', except that it applies to operators instead of 16894 operands. When constructing an insn, operand number N will be 16895 substituted at this point. But in matching, `match_op_dup' behaves 16896 differently. It assumes that operand number N has already been 16897 determined by a `match_operator' appearing earlier in the 16898 recognition template, and it matches only an identical-looking 16899 expression. 16900 16901 `(match_parallel N PREDICATE [SUBPAT...])' 16902 This pattern is a placeholder for an insn that consists of a 16903 `parallel' expression with a variable number of elements. This 16904 expression should only appear at the top level of an insn pattern. 16905 16906 When constructing an insn, operand number N will be substituted at 16907 this point. When matching an insn, it matches if the body of the 16908 insn is a `parallel' expression with at least as many elements as 16909 the vector of SUBPAT expressions in the `match_parallel', if each 16910 SUBPAT matches the corresponding element of the `parallel', _and_ 16911 the function PREDICATE returns nonzero on the `parallel' that is 16912 the body of the insn. It is the responsibility of the predicate 16913 to validate elements of the `parallel' beyond those listed in the 16914 `match_parallel'. 16915 16916 A typical use of `match_parallel' is to match load and store 16917 multiple expressions, which can contain a variable number of 16918 elements in a `parallel'. For example, 16919 16920 (define_insn "" 16921 [(match_parallel 0 "load_multiple_operation" 16922 [(set (match_operand:SI 1 "gpc_reg_operand" "=r") 16923 (match_operand:SI 2 "memory_operand" "m")) 16924 (use (reg:SI 179)) 16925 (clobber (reg:SI 179))])] 16926 "" 16927 "loadm 0,0,%1,%2") 16928 16929 This example comes from `a29k.md'. The function 16930 `load_multiple_operation' is defined in `a29k.c' and checks that 16931 subsequent elements in the `parallel' are the same as the `set' in 16932 the pattern, except that they are referencing subsequent registers 16933 and memory locations. 16934 16935 An insn that matches this pattern might look like: 16936 16937 (parallel 16938 [(set (reg:SI 20) (mem:SI (reg:SI 100))) 16939 (use (reg:SI 179)) 16940 (clobber (reg:SI 179)) 16941 (set (reg:SI 21) 16942 (mem:SI (plus:SI (reg:SI 100) 16943 (const_int 4)))) 16944 (set (reg:SI 22) 16945 (mem:SI (plus:SI (reg:SI 100) 16946 (const_int 8))))]) 16947 16948 `(match_par_dup N [SUBPAT...])' 16949 Like `match_op_dup', but for `match_parallel' instead of 16950 `match_operator'. 16951 16952 16953 16954 File: gccint.info, Node: Output Template, Next: Output Statement, Prev: RTL Template, Up: Machine Desc 16955 16956 16.5 Output Templates and Operand Substitution 16957 ============================================== 16958 16959 The "output template" is a string which specifies how to output the 16960 assembler code for an instruction pattern. Most of the template is a 16961 fixed string which is output literally. The character `%' is used to 16962 specify where to substitute an operand; it can also be used to identify 16963 places where different variants of the assembler require different 16964 syntax. 16965 16966 In the simplest case, a `%' followed by a digit N says to output 16967 operand N at that point in the string. 16968 16969 `%' followed by a letter and a digit says to output an operand in an 16970 alternate fashion. Four letters have standard, built-in meanings 16971 described below. The machine description macro `PRINT_OPERAND' can 16972 define additional letters with nonstandard meanings. 16973 16974 `%cDIGIT' can be used to substitute an operand that is a constant 16975 value without the syntax that normally indicates an immediate operand. 16976 16977 `%nDIGIT' is like `%cDIGIT' except that the value of the constant is 16978 negated before printing. 16979 16980 `%aDIGIT' can be used to substitute an operand as if it were a memory 16981 reference, with the actual operand treated as the address. This may be 16982 useful when outputting a "load address" instruction, because often the 16983 assembler syntax for such an instruction requires you to write the 16984 operand as if it were a memory reference. 16985 16986 `%lDIGIT' is used to substitute a `label_ref' into a jump instruction. 16987 16988 `%=' outputs a number which is unique to each instruction in the 16989 entire compilation. This is useful for making local labels to be 16990 referred to more than once in a single template that generates multiple 16991 assembler instructions. 16992 16993 `%' followed by a punctuation character specifies a substitution that 16994 does not use an operand. Only one case is standard: `%%' outputs a `%' 16995 into the assembler code. Other nonstandard cases can be defined in the 16996 `PRINT_OPERAND' macro. You must also define which punctuation 16997 characters are valid with the `PRINT_OPERAND_PUNCT_VALID_P' macro. 16998 16999 The template may generate multiple assembler instructions. Write the 17000 text for the instructions, with `\;' between them. 17001 17002 When the RTL contains two operands which are required by constraint to 17003 match each other, the output template must refer only to the 17004 lower-numbered operand. Matching operands are not always identical, 17005 and the rest of the compiler arranges to put the proper RTL expression 17006 for printing into the lower-numbered operand. 17007 17008 One use of nonstandard letters or punctuation following `%' is to 17009 distinguish between different assembler languages for the same machine; 17010 for example, Motorola syntax versus MIT syntax for the 68000. Motorola 17011 syntax requires periods in most opcode names, while MIT syntax does 17012 not. For example, the opcode `movel' in MIT syntax is `move.l' in 17013 Motorola syntax. The same file of patterns is used for both kinds of 17014 output syntax, but the character sequence `%.' is used in each place 17015 where Motorola syntax wants a period. The `PRINT_OPERAND' macro for 17016 Motorola syntax defines the sequence to output a period; the macro for 17017 MIT syntax defines it to do nothing. 17018 17019 As a special case, a template consisting of the single character `#' 17020 instructs the compiler to first split the insn, and then output the 17021 resulting instructions separately. This helps eliminate redundancy in 17022 the output templates. If you have a `define_insn' that needs to emit 17023 multiple assembler instructions, and there is an matching `define_split' 17024 already defined, then you can simply use `#' as the output template 17025 instead of writing an output template that emits the multiple assembler 17026 instructions. 17027 17028 If the macro `ASSEMBLER_DIALECT' is defined, you can use construct of 17029 the form `{option0|option1|option2}' in the templates. These describe 17030 multiple variants of assembler language syntax. *Note Instruction 17031 Output::. 17032 17033 17034 File: gccint.info, Node: Output Statement, Next: Predicates, Prev: Output Template, Up: Machine Desc 17035 17036 16.6 C Statements for Assembler Output 17037 ====================================== 17038 17039 Often a single fixed template string cannot produce correct and 17040 efficient assembler code for all the cases that are recognized by a 17041 single instruction pattern. For example, the opcodes may depend on the 17042 kinds of operands; or some unfortunate combinations of operands may 17043 require extra machine instructions. 17044 17045 If the output control string starts with a `@', then it is actually a 17046 series of templates, each on a separate line. (Blank lines and leading 17047 spaces and tabs are ignored.) The templates correspond to the 17048 pattern's constraint alternatives (*note Multi-Alternative::). For 17049 example, if a target machine has a two-address add instruction `addr' 17050 to add into a register and another `addm' to add a register to memory, 17051 you might write this pattern: 17052 17053 (define_insn "addsi3" 17054 [(set (match_operand:SI 0 "general_operand" "=r,m") 17055 (plus:SI (match_operand:SI 1 "general_operand" "0,0") 17056 (match_operand:SI 2 "general_operand" "g,r")))] 17057 "" 17058 "@ 17059 addr %2,%0 17060 addm %2,%0") 17061 17062 If the output control string starts with a `*', then it is not an 17063 output template but rather a piece of C program that should compute a 17064 template. It should execute a `return' statement to return the 17065 template-string you want. Most such templates use C string literals, 17066 which require doublequote characters to delimit them. To include these 17067 doublequote characters in the string, prefix each one with `\'. 17068 17069 If the output control string is written as a brace block instead of a 17070 double-quoted string, it is automatically assumed to be C code. In that 17071 case, it is not necessary to put in a leading asterisk, or to escape the 17072 doublequotes surrounding C string literals. 17073 17074 The operands may be found in the array `operands', whose C data type 17075 is `rtx []'. 17076 17077 It is very common to select different ways of generating assembler code 17078 based on whether an immediate operand is within a certain range. Be 17079 careful when doing this, because the result of `INTVAL' is an integer 17080 on the host machine. If the host machine has more bits in an `int' 17081 than the target machine has in the mode in which the constant will be 17082 used, then some of the bits you get from `INTVAL' will be superfluous. 17083 For proper results, you must carefully disregard the values of those 17084 bits. 17085 17086 It is possible to output an assembler instruction and then go on to 17087 output or compute more of them, using the subroutine `output_asm_insn'. 17088 This receives two arguments: a template-string and a vector of 17089 operands. The vector may be `operands', or it may be another array of 17090 `rtx' that you declare locally and initialize yourself. 17091 17092 When an insn pattern has multiple alternatives in its constraints, 17093 often the appearance of the assembler code is determined mostly by 17094 which alternative was matched. When this is so, the C code can test 17095 the variable `which_alternative', which is the ordinal number of the 17096 alternative that was actually satisfied (0 for the first, 1 for the 17097 second alternative, etc.). 17098 17099 For example, suppose there are two opcodes for storing zero, `clrreg' 17100 for registers and `clrmem' for memory locations. Here is how a pattern 17101 could use `which_alternative' to choose between them: 17102 17103 (define_insn "" 17104 [(set (match_operand:SI 0 "general_operand" "=r,m") 17105 (const_int 0))] 17106 "" 17107 { 17108 return (which_alternative == 0 17109 ? "clrreg %0" : "clrmem %0"); 17110 }) 17111 17112 The example above, where the assembler code to generate was _solely_ 17113 determined by the alternative, could also have been specified as 17114 follows, having the output control string start with a `@': 17115 17116 (define_insn "" 17117 [(set (match_operand:SI 0 "general_operand" "=r,m") 17118 (const_int 0))] 17119 "" 17120 "@ 17121 clrreg %0 17122 clrmem %0") 17123 17124 17125 File: gccint.info, Node: Predicates, Next: Constraints, Prev: Output Statement, Up: Machine Desc 17126 17127 16.7 Predicates 17128 =============== 17129 17130 A predicate determines whether a `match_operand' or `match_operator' 17131 expression matches, and therefore whether the surrounding instruction 17132 pattern will be used for that combination of operands. GCC has a 17133 number of machine-independent predicates, and you can define 17134 machine-specific predicates as needed. By convention, predicates used 17135 with `match_operand' have names that end in `_operand', and those used 17136 with `match_operator' have names that end in `_operator'. 17137 17138 All predicates are Boolean functions (in the mathematical sense) of 17139 two arguments: the RTL expression that is being considered at that 17140 position in the instruction pattern, and the machine mode that the 17141 `match_operand' or `match_operator' specifies. In this section, the 17142 first argument is called OP and the second argument MODE. Predicates 17143 can be called from C as ordinary two-argument functions; this can be 17144 useful in output templates or other machine-specific code. 17145 17146 Operand predicates can allow operands that are not actually acceptable 17147 to the hardware, as long as the constraints give reload the ability to 17148 fix them up (*note Constraints::). However, GCC will usually generate 17149 better code if the predicates specify the requirements of the machine 17150 instructions as closely as possible. Reload cannot fix up operands 17151 that must be constants ("immediate operands"); you must use a predicate 17152 that allows only constants, or else enforce the requirement in the 17153 extra condition. 17154 17155 Most predicates handle their MODE argument in a uniform manner. If 17156 MODE is `VOIDmode' (unspecified), then OP can have any mode. If MODE 17157 is anything else, then OP must have the same mode, unless OP is a 17158 `CONST_INT' or integer `CONST_DOUBLE'. These RTL expressions always 17159 have `VOIDmode', so it would be counterproductive to check that their 17160 mode matches. Instead, predicates that accept `CONST_INT' and/or 17161 integer `CONST_DOUBLE' check that the value stored in the constant will 17162 fit in the requested mode. 17163 17164 Predicates with this behavior are called "normal". `genrecog' can 17165 optimize the instruction recognizer based on knowledge of how normal 17166 predicates treat modes. It can also diagnose certain kinds of common 17167 errors in the use of normal predicates; for instance, it is almost 17168 always an error to use a normal predicate without specifying a mode. 17169 17170 Predicates that do something different with their MODE argument are 17171 called "special". The generic predicates `address_operand' and 17172 `pmode_register_operand' are special predicates. `genrecog' does not 17173 do any optimizations or diagnosis when special predicates are used. 17174 17175 * Menu: 17176 17177 * Machine-Independent Predicates:: Predicates available to all back ends. 17178 * Defining Predicates:: How to write machine-specific predicate 17179 functions. 17180 17181 17182 File: gccint.info, Node: Machine-Independent Predicates, Next: Defining Predicates, Up: Predicates 17183 17184 16.7.1 Machine-Independent Predicates 17185 ------------------------------------- 17186 17187 These are the generic predicates available to all back ends. They are 17188 defined in `recog.c'. The first category of predicates allow only 17189 constant, or "immediate", operands. 17190 17191 -- Function: immediate_operand 17192 This predicate allows any sort of constant that fits in MODE. It 17193 is an appropriate choice for instructions that take operands that 17194 must be constant. 17195 17196 -- Function: const_int_operand 17197 This predicate allows any `CONST_INT' expression that fits in 17198 MODE. It is an appropriate choice for an immediate operand that 17199 does not allow a symbol or label. 17200 17201 -- Function: const_double_operand 17202 This predicate accepts any `CONST_DOUBLE' expression that has 17203 exactly MODE. If MODE is `VOIDmode', it will also accept 17204 `CONST_INT'. It is intended for immediate floating point 17205 constants. 17206 17207 The second category of predicates allow only some kind of machine 17208 register. 17209 17210 -- Function: register_operand 17211 This predicate allows any `REG' or `SUBREG' expression that is 17212 valid for MODE. It is often suitable for arithmetic instruction 17213 operands on a RISC machine. 17214 17215 -- Function: pmode_register_operand 17216 This is a slight variant on `register_operand' which works around 17217 a limitation in the machine-description reader. 17218 17219 (match_operand N "pmode_register_operand" CONSTRAINT) 17220 17221 means exactly what 17222 17223 (match_operand:P N "register_operand" CONSTRAINT) 17224 17225 would mean, if the machine-description reader accepted `:P' mode 17226 suffixes. Unfortunately, it cannot, because `Pmode' is an alias 17227 for some other mode, and might vary with machine-specific options. 17228 *Note Misc::. 17229 17230 -- Function: scratch_operand 17231 This predicate allows hard registers and `SCRATCH' expressions, 17232 but not pseudo-registers. It is used internally by 17233 `match_scratch'; it should not be used directly. 17234 17235 The third category of predicates allow only some kind of memory 17236 reference. 17237 17238 -- Function: memory_operand 17239 This predicate allows any valid reference to a quantity of mode 17240 MODE in memory, as determined by the weak form of 17241 `GO_IF_LEGITIMATE_ADDRESS' (*note Addressing Modes::). 17242 17243 -- Function: address_operand 17244 This predicate is a little unusual; it allows any operand that is a 17245 valid expression for the _address_ of a quantity of mode MODE, 17246 again determined by the weak form of `GO_IF_LEGITIMATE_ADDRESS'. 17247 To first order, if `(mem:MODE (EXP))' is acceptable to 17248 `memory_operand', then EXP is acceptable to `address_operand'. 17249 Note that EXP does not necessarily have the mode MODE. 17250 17251 -- Function: indirect_operand 17252 This is a stricter form of `memory_operand' which allows only 17253 memory references with a `general_operand' as the address 17254 expression. New uses of this predicate are discouraged, because 17255 `general_operand' is very permissive, so it's hard to tell what an 17256 `indirect_operand' does or does not allow. If a target has 17257 different requirements for memory operands for different 17258 instructions, it is better to define target-specific predicates 17259 which enforce the hardware's requirements explicitly. 17260 17261 -- Function: push_operand 17262 This predicate allows a memory reference suitable for pushing a 17263 value onto the stack. This will be a `MEM' which refers to 17264 `stack_pointer_rtx', with a side-effect in its address expression 17265 (*note Incdec::); which one is determined by the `STACK_PUSH_CODE' 17266 macro (*note Frame Layout::). 17267 17268 -- Function: pop_operand 17269 This predicate allows a memory reference suitable for popping a 17270 value off the stack. Again, this will be a `MEM' referring to 17271 `stack_pointer_rtx', with a side-effect in its address expression. 17272 However, this time `STACK_POP_CODE' is expected. 17273 17274 The fourth category of predicates allow some combination of the above 17275 operands. 17276 17277 -- Function: nonmemory_operand 17278 This predicate allows any immediate or register operand valid for 17279 MODE. 17280 17281 -- Function: nonimmediate_operand 17282 This predicate allows any register or memory operand valid for 17283 MODE. 17284 17285 -- Function: general_operand 17286 This predicate allows any immediate, register, or memory operand 17287 valid for MODE. 17288 17289 Finally, there is one generic operator predicate. 17290 17291 -- Function: comparison_operator 17292 This predicate matches any expression which performs an arithmetic 17293 comparison in MODE; that is, `COMPARISON_P' is true for the 17294 expression code. 17295 17296 17297 File: gccint.info, Node: Defining Predicates, Prev: Machine-Independent Predicates, Up: Predicates 17298 17299 16.7.2 Defining Machine-Specific Predicates 17300 ------------------------------------------- 17301 17302 Many machines have requirements for their operands that cannot be 17303 expressed precisely using the generic predicates. You can define 17304 additional predicates using `define_predicate' and 17305 `define_special_predicate' expressions. These expressions have three 17306 operands: 17307 17308 * The name of the predicate, as it will be referred to in 17309 `match_operand' or `match_operator' expressions. 17310 17311 * An RTL expression which evaluates to true if the predicate allows 17312 the operand OP, false if it does not. This expression can only use 17313 the following RTL codes: 17314 17315 `MATCH_OPERAND' 17316 When written inside a predicate expression, a `MATCH_OPERAND' 17317 expression evaluates to true if the predicate it names would 17318 allow OP. The operand number and constraint are ignored. 17319 Due to limitations in `genrecog', you can only refer to 17320 generic predicates and predicates that have already been 17321 defined. 17322 17323 `MATCH_CODE' 17324 This expression evaluates to true if OP or a specified 17325 subexpression of OP has one of a given list of RTX codes. 17326 17327 The first operand of this expression is a string constant 17328 containing a comma-separated list of RTX code names (in lower 17329 case). These are the codes for which the `MATCH_CODE' will 17330 be true. 17331 17332 The second operand is a string constant which indicates what 17333 subexpression of OP to examine. If it is absent or the empty 17334 string, OP itself is examined. Otherwise, the string constant 17335 must be a sequence of digits and/or lowercase letters. Each 17336 character indicates a subexpression to extract from the 17337 current expression; for the first character this is OP, for 17338 the second and subsequent characters it is the result of the 17339 previous character. A digit N extracts `XEXP (E, N)'; a 17340 letter L extracts `XVECEXP (E, 0, N)' where N is the 17341 alphabetic ordinal of L (0 for `a', 1 for 'b', and so on). 17342 The `MATCH_CODE' then examines the RTX code of the 17343 subexpression extracted by the complete string. It is not 17344 possible to extract components of an `rtvec' that is not at 17345 position 0 within its RTX object. 17346 17347 `MATCH_TEST' 17348 This expression has one operand, a string constant containing 17349 a C expression. The predicate's arguments, OP and MODE, are 17350 available with those names in the C expression. The 17351 `MATCH_TEST' evaluates to true if the C expression evaluates 17352 to a nonzero value. `MATCH_TEST' expressions must not have 17353 side effects. 17354 17355 `AND' 17356 `IOR' 17357 `NOT' 17358 `IF_THEN_ELSE' 17359 The basic `MATCH_' expressions can be combined using these 17360 logical operators, which have the semantics of the C operators 17361 `&&', `||', `!', and `? :' respectively. As in Common Lisp, 17362 you may give an `AND' or `IOR' expression an arbitrary number 17363 of arguments; this has exactly the same effect as writing a 17364 chain of two-argument `AND' or `IOR' expressions. 17365 17366 * An optional block of C code, which should execute `return true' if 17367 the predicate is found to match and `return false' if it does not. 17368 It must not have any side effects. The predicate arguments, OP 17369 and MODE, are available with those names. 17370 17371 If a code block is present in a predicate definition, then the RTL 17372 expression must evaluate to true _and_ the code block must execute 17373 `return true' for the predicate to allow the operand. The RTL 17374 expression is evaluated first; do not re-check anything in the 17375 code block that was checked in the RTL expression. 17376 17377 The program `genrecog' scans `define_predicate' and 17378 `define_special_predicate' expressions to determine which RTX codes are 17379 possibly allowed. You should always make this explicit in the RTL 17380 predicate expression, using `MATCH_OPERAND' and `MATCH_CODE'. 17381 17382 Here is an example of a simple predicate definition, from the IA64 17383 machine description: 17384 17385 ;; True if OP is a `SYMBOL_REF' which refers to the sdata section. 17386 (define_predicate "small_addr_symbolic_operand" 17387 (and (match_code "symbol_ref") 17388 (match_test "SYMBOL_REF_SMALL_ADDR_P (op)"))) 17389 17390 And here is another, showing the use of the C block. 17391 17392 ;; True if OP is a register operand that is (or could be) a GR reg. 17393 (define_predicate "gr_register_operand" 17394 (match_operand 0 "register_operand") 17395 { 17396 unsigned int regno; 17397 if (GET_CODE (op) == SUBREG) 17398 op = SUBREG_REG (op); 17399 17400 regno = REGNO (op); 17401 return (regno >= FIRST_PSEUDO_REGISTER || GENERAL_REGNO_P (regno)); 17402 }) 17403 17404 Predicates written with `define_predicate' automatically include a 17405 test that MODE is `VOIDmode', or OP has the same mode as MODE, or OP is 17406 a `CONST_INT' or `CONST_DOUBLE'. They do _not_ check specifically for 17407 integer `CONST_DOUBLE', nor do they test that the value of either kind 17408 of constant fits in the requested mode. This is because 17409 target-specific predicates that take constants usually have to do more 17410 stringent value checks anyway. If you need the exact same treatment of 17411 `CONST_INT' or `CONST_DOUBLE' that the generic predicates provide, use 17412 a `MATCH_OPERAND' subexpression to call `const_int_operand', 17413 `const_double_operand', or `immediate_operand'. 17414 17415 Predicates written with `define_special_predicate' do not get any 17416 automatic mode checks, and are treated as having special mode handling 17417 by `genrecog'. 17418 17419 The program `genpreds' is responsible for generating code to test 17420 predicates. It also writes a header file containing function 17421 declarations for all machine-specific predicates. It is not necessary 17422 to declare these predicates in `CPU-protos.h'. 17423 17424 17425 File: gccint.info, Node: Constraints, Next: Standard Names, Prev: Predicates, Up: Machine Desc 17426 17427 16.8 Operand Constraints 17428 ======================== 17429 17430 Each `match_operand' in an instruction pattern can specify constraints 17431 for the operands allowed. The constraints allow you to fine-tune 17432 matching within the set of operands allowed by the predicate. 17433 17434 Constraints can say whether an operand may be in a register, and which 17435 kinds of register; whether the operand can be a memory reference, and 17436 which kinds of address; whether the operand may be an immediate 17437 constant, and which possible values it may have. Constraints can also 17438 require two operands to match. 17439 17440 * Menu: 17441 17442 * Simple Constraints:: Basic use of constraints. 17443 * Multi-Alternative:: When an insn has two alternative constraint-patterns. 17444 * Class Preferences:: Constraints guide which hard register to put things in. 17445 * Modifiers:: More precise control over effects of constraints. 17446 * Disable Insn Alternatives:: Disable insn alternatives using the `enabled' attribute. 17447 * Machine Constraints:: Existing constraints for some particular machines. 17448 * Define Constraints:: How to define machine-specific constraints. 17449 * C Constraint Interface:: How to test constraints from C code. 17450 17451 17452 File: gccint.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints 17453 17454 16.8.1 Simple Constraints 17455 ------------------------- 17456 17457 The simplest kind of constraint is a string full of letters, each of 17458 which describes one kind of operand that is permitted. Here are the 17459 letters that are allowed: 17460 17461 whitespace 17462 Whitespace characters are ignored and can be inserted at any 17463 position except the first. This enables each alternative for 17464 different operands to be visually aligned in the machine 17465 description even if they have different number of constraints and 17466 modifiers. 17467 17468 `m' 17469 A memory operand is allowed, with any kind of address that the 17470 machine supports in general. Note that the letter used for the 17471 general memory constraint can be re-defined by a back end using 17472 the `TARGET_MEM_CONSTRAINT' macro. 17473 17474 `o' 17475 A memory operand is allowed, but only if the address is 17476 "offsettable". This means that adding a small integer (actually, 17477 the width in bytes of the operand, as determined by its machine 17478 mode) may be added to the address and the result is also a valid 17479 memory address. 17480 17481 For example, an address which is constant is offsettable; so is an 17482 address that is the sum of a register and a constant (as long as a 17483 slightly larger constant is also within the range of 17484 address-offsets supported by the machine); but an autoincrement or 17485 autodecrement address is not offsettable. More complicated 17486 indirect/indexed addresses may or may not be offsettable depending 17487 on the other addressing modes that the machine supports. 17488 17489 Note that in an output operand which can be matched by another 17490 operand, the constraint letter `o' is valid only when accompanied 17491 by both `<' (if the target machine has predecrement addressing) 17492 and `>' (if the target machine has preincrement addressing). 17493 17494 `V' 17495 A memory operand that is not offsettable. In other words, 17496 anything that would fit the `m' constraint but not the `o' 17497 constraint. 17498 17499 `<' 17500 A memory operand with autodecrement addressing (either 17501 predecrement or postdecrement) is allowed. 17502 17503 `>' 17504 A memory operand with autoincrement addressing (either 17505 preincrement or postincrement) is allowed. 17506 17507 `r' 17508 A register operand is allowed provided that it is in a general 17509 register. 17510 17511 `i' 17512 An immediate integer operand (one with constant value) is allowed. 17513 This includes symbolic constants whose values will be known only at 17514 assembly time or later. 17515 17516 `n' 17517 An immediate integer operand with a known numeric value is allowed. 17518 Many systems cannot support assembly-time constants for operands 17519 less than a word wide. Constraints for these operands should use 17520 `n' rather than `i'. 17521 17522 `I', `J', `K', ... `P' 17523 Other letters in the range `I' through `P' may be defined in a 17524 machine-dependent fashion to permit immediate integer operands with 17525 explicit integer values in specified ranges. For example, on the 17526 68000, `I' is defined to stand for the range of values 1 to 8. 17527 This is the range permitted as a shift count in the shift 17528 instructions. 17529 17530 `E' 17531 An immediate floating operand (expression code `const_double') is 17532 allowed, but only if the target floating point format is the same 17533 as that of the host machine (on which the compiler is running). 17534 17535 `F' 17536 An immediate floating operand (expression code `const_double' or 17537 `const_vector') is allowed. 17538 17539 `G', `H' 17540 `G' and `H' may be defined in a machine-dependent fashion to 17541 permit immediate floating operands in particular ranges of values. 17542 17543 `s' 17544 An immediate integer operand whose value is not an explicit 17545 integer is allowed. 17546 17547 This might appear strange; if an insn allows a constant operand 17548 with a value not known at compile time, it certainly must allow 17549 any known value. So why use `s' instead of `i'? Sometimes it 17550 allows better code to be generated. 17551 17552 For example, on the 68000 in a fullword instruction it is possible 17553 to use an immediate operand; but if the immediate value is between 17554 -128 and 127, better code results from loading the value into a 17555 register and using the register. This is because the load into 17556 the register can be done with a `moveq' instruction. We arrange 17557 for this to happen by defining the letter `K' to mean "any integer 17558 outside the range -128 to 127", and then specifying `Ks' in the 17559 operand constraints. 17560 17561 `g' 17562 Any register, memory or immediate integer operand is allowed, 17563 except for registers that are not general registers. 17564 17565 `X' 17566 Any operand whatsoever is allowed, even if it does not satisfy 17567 `general_operand'. This is normally used in the constraint of a 17568 `match_scratch' when certain alternatives will not actually 17569 require a scratch register. 17570 17571 `0', `1', `2', ... `9' 17572 An operand that matches the specified operand number is allowed. 17573 If a digit is used together with letters within the same 17574 alternative, the digit should come last. 17575 17576 This number is allowed to be more than a single digit. If multiple 17577 digits are encountered consecutively, they are interpreted as a 17578 single decimal integer. There is scant chance for ambiguity, 17579 since to-date it has never been desirable that `10' be interpreted 17580 as matching either operand 1 _or_ operand 0. Should this be 17581 desired, one can use multiple alternatives instead. 17582 17583 This is called a "matching constraint" and what it really means is 17584 that the assembler has only a single operand that fills two roles 17585 considered separate in the RTL insn. For example, an add insn has 17586 two input operands and one output operand in the RTL, but on most 17587 CISC machines an add instruction really has only two operands, one 17588 of them an input-output operand: 17589 17590 addl #35,r12 17591 17592 Matching constraints are used in these circumstances. More 17593 precisely, the two operands that match must include one input-only 17594 operand and one output-only operand. Moreover, the digit must be a 17595 smaller number than the number of the operand that uses it in the 17596 constraint. 17597 17598 For operands to match in a particular case usually means that they 17599 are identical-looking RTL expressions. But in a few special cases 17600 specific kinds of dissimilarity are allowed. For example, `*x' as 17601 an input operand will match `*x++' as an output operand. For 17602 proper results in such cases, the output template should always 17603 use the output-operand's number when printing the operand. 17604 17605 `p' 17606 An operand that is a valid memory address is allowed. This is for 17607 "load address" and "push address" instructions. 17608 17609 `p' in the constraint must be accompanied by `address_operand' as 17610 the predicate in the `match_operand'. This predicate interprets 17611 the mode specified in the `match_operand' as the mode of the memory 17612 reference for which the address would be valid. 17613 17614 OTHER-LETTERS 17615 Other letters can be defined in machine-dependent fashion to stand 17616 for particular classes of registers or other arbitrary operand 17617 types. `d', `a' and `f' are defined on the 68000/68020 to stand 17618 for data, address and floating point registers. 17619 17620 In order to have valid assembler code, each operand must satisfy its 17621 constraint. But a failure to do so does not prevent the pattern from 17622 applying to an insn. Instead, it directs the compiler to modify the 17623 code so that the constraint will be satisfied. Usually this is done by 17624 copying an operand into a register. 17625 17626 Contrast, therefore, the two instruction patterns that follow: 17627 17628 (define_insn "" 17629 [(set (match_operand:SI 0 "general_operand" "=r") 17630 (plus:SI (match_dup 0) 17631 (match_operand:SI 1 "general_operand" "r")))] 17632 "" 17633 "...") 17634 17635 which has two operands, one of which must appear in two places, and 17636 17637 (define_insn "" 17638 [(set (match_operand:SI 0 "general_operand" "=r") 17639 (plus:SI (match_operand:SI 1 "general_operand" "0") 17640 (match_operand:SI 2 "general_operand" "r")))] 17641 "" 17642 "...") 17643 17644 which has three operands, two of which are required by a constraint to 17645 be identical. If we are considering an insn of the form 17646 17647 (insn N PREV NEXT 17648 (set (reg:SI 3) 17649 (plus:SI (reg:SI 6) (reg:SI 109))) 17650 ...) 17651 17652 the first pattern would not apply at all, because this insn does not 17653 contain two identical subexpressions in the right place. The pattern 17654 would say, "That does not look like an add instruction; try other 17655 patterns". The second pattern would say, "Yes, that's an add 17656 instruction, but there is something wrong with it". It would direct 17657 the reload pass of the compiler to generate additional insns to make 17658 the constraint true. The results might look like this: 17659 17660 (insn N2 PREV N 17661 (set (reg:SI 3) (reg:SI 6)) 17662 ...) 17663 17664 (insn N N2 NEXT 17665 (set (reg:SI 3) 17666 (plus:SI (reg:SI 3) (reg:SI 109))) 17667 ...) 17668 17669 It is up to you to make sure that each operand, in each pattern, has 17670 constraints that can handle any RTL expression that could be present for 17671 that operand. (When multiple alternatives are in use, each pattern 17672 must, for each possible combination of operand expressions, have at 17673 least one alternative which can handle that combination of operands.) 17674 The constraints don't need to _allow_ any possible operand--when this is 17675 the case, they do not constrain--but they must at least point the way to 17676 reloading any possible operand so that it will fit. 17677 17678 * If the constraint accepts whatever operands the predicate permits, 17679 there is no problem: reloading is never necessary for this operand. 17680 17681 For example, an operand whose constraints permit everything except 17682 registers is safe provided its predicate rejects registers. 17683 17684 An operand whose predicate accepts only constant values is safe 17685 provided its constraints include the letter `i'. If any possible 17686 constant value is accepted, then nothing less than `i' will do; if 17687 the predicate is more selective, then the constraints may also be 17688 more selective. 17689 17690 * Any operand expression can be reloaded by copying it into a 17691 register. So if an operand's constraints allow some kind of 17692 register, it is certain to be safe. It need not permit all 17693 classes of registers; the compiler knows how to copy a register 17694 into another register of the proper class in order to make an 17695 instruction valid. 17696 17697 * A nonoffsettable memory reference can be reloaded by copying the 17698 address into a register. So if the constraint uses the letter 17699 `o', all memory references are taken care of. 17700 17701 * A constant operand can be reloaded by allocating space in memory to 17702 hold it as preinitialized data. Then the memory reference can be 17703 used in place of the constant. So if the constraint uses the 17704 letters `o' or `m', constant operands are not a problem. 17705 17706 * If the constraint permits a constant and a pseudo register used in 17707 an insn was not allocated to a hard register and is equivalent to 17708 a constant, the register will be replaced with the constant. If 17709 the predicate does not permit a constant and the insn is 17710 re-recognized for some reason, the compiler will crash. Thus the 17711 predicate must always recognize any objects allowed by the 17712 constraint. 17713 17714 If the operand's predicate can recognize registers, but the constraint 17715 does not permit them, it can make the compiler crash. When this 17716 operand happens to be a register, the reload pass will be stymied, 17717 because it does not know how to copy a register temporarily into memory. 17718 17719 If the predicate accepts a unary operator, the constraint applies to 17720 the operand. For example, the MIPS processor at ISA level 3 supports an 17721 instruction which adds two registers in `SImode' to produce a `DImode' 17722 result, but only if the registers are correctly sign extended. This 17723 predicate for the input operands accepts a `sign_extend' of an `SImode' 17724 register. Write the constraint to indicate the type of register that 17725 is required for the operand of the `sign_extend'. 17726 17727 17728 File: gccint.info, Node: Multi-Alternative, Next: Class Preferences, Prev: Simple Constraints, Up: Constraints 17729 17730 16.8.2 Multiple Alternative Constraints 17731 --------------------------------------- 17732 17733 Sometimes a single instruction has multiple alternative sets of possible 17734 operands. For example, on the 68000, a logical-or instruction can 17735 combine register or an immediate value into memory, or it can combine 17736 any kind of operand into a register; but it cannot combine one memory 17737 location into another. 17738 17739 These constraints are represented as multiple alternatives. An 17740 alternative can be described by a series of letters for each operand. 17741 The overall constraint for an operand is made from the letters for this 17742 operand from the first alternative, a comma, the letters for this 17743 operand from the second alternative, a comma, and so on until the last 17744 alternative. Here is how it is done for fullword logical-or on the 17745 68000: 17746 17747 (define_insn "iorsi3" 17748 [(set (match_operand:SI 0 "general_operand" "=m,d") 17749 (ior:SI (match_operand:SI 1 "general_operand" "%0,0") 17750 (match_operand:SI 2 "general_operand" "dKs,dmKs")))] 17751 ...) 17752 17753 The first alternative has `m' (memory) for operand 0, `0' for operand 17754 1 (meaning it must match operand 0), and `dKs' for operand 2. The 17755 second alternative has `d' (data register) for operand 0, `0' for 17756 operand 1, and `dmKs' for operand 2. The `=' and `%' in the 17757 constraints apply to all the alternatives; their meaning is explained 17758 in the next section (*note Class Preferences::). 17759 17760 If all the operands fit any one alternative, the instruction is valid. 17761 Otherwise, for each alternative, the compiler counts how many 17762 instructions must be added to copy the operands so that that 17763 alternative applies. The alternative requiring the least copying is 17764 chosen. If two alternatives need the same amount of copying, the one 17765 that comes first is chosen. These choices can be altered with the `?' 17766 and `!' characters: 17767 17768 `?' 17769 Disparage slightly the alternative that the `?' appears in, as a 17770 choice when no alternative applies exactly. The compiler regards 17771 this alternative as one unit more costly for each `?' that appears 17772 in it. 17773 17774 `!' 17775 Disparage severely the alternative that the `!' appears in. This 17776 alternative can still be used if it fits without reloading, but if 17777 reloading is needed, some other alternative will be used. 17778 17779 When an insn pattern has multiple alternatives in its constraints, 17780 often the appearance of the assembler code is determined mostly by which 17781 alternative was matched. When this is so, the C code for writing the 17782 assembler code can use the variable `which_alternative', which is the 17783 ordinal number of the alternative that was actually satisfied (0 for 17784 the first, 1 for the second alternative, etc.). *Note Output 17785 Statement::. 17786 17787 17788 File: gccint.info, Node: Class Preferences, Next: Modifiers, Prev: Multi-Alternative, Up: Constraints 17789 17790 16.8.3 Register Class Preferences 17791 --------------------------------- 17792 17793 The operand constraints have another function: they enable the compiler 17794 to decide which kind of hardware register a pseudo register is best 17795 allocated to. The compiler examines the constraints that apply to the 17796 insns that use the pseudo register, looking for the machine-dependent 17797 letters such as `d' and `a' that specify classes of registers. The 17798 pseudo register is put in whichever class gets the most "votes". The 17799 constraint letters `g' and `r' also vote: they vote in favor of a 17800 general register. The machine description says which registers are 17801 considered general. 17802 17803 Of course, on some machines all registers are equivalent, and no 17804 register classes are defined. Then none of this complexity is relevant. 17805 17806 17807 File: gccint.info, Node: Modifiers, Next: Disable Insn Alternatives, Prev: Class Preferences, Up: Constraints 17808 17809 16.8.4 Constraint Modifier Characters 17810 ------------------------------------- 17811 17812 Here are constraint modifier characters. 17813 17814 `=' 17815 Means that this operand is write-only for this instruction: the 17816 previous value is discarded and replaced by output data. 17817 17818 `+' 17819 Means that this operand is both read and written by the 17820 instruction. 17821 17822 When the compiler fixes up the operands to satisfy the constraints, 17823 it needs to know which operands are inputs to the instruction and 17824 which are outputs from it. `=' identifies an output; `+' 17825 identifies an operand that is both input and output; all other 17826 operands are assumed to be input only. 17827 17828 If you specify `=' or `+' in a constraint, you put it in the first 17829 character of the constraint string. 17830 17831 `&' 17832 Means (in a particular alternative) that this operand is an 17833 "earlyclobber" operand, which is modified before the instruction is 17834 finished using the input operands. Therefore, this operand may 17835 not lie in a register that is used as an input operand or as part 17836 of any memory address. 17837 17838 `&' applies only to the alternative in which it is written. In 17839 constraints with multiple alternatives, sometimes one alternative 17840 requires `&' while others do not. See, for example, the `movdf' 17841 insn of the 68000. 17842 17843 An input operand can be tied to an earlyclobber operand if its only 17844 use as an input occurs before the early result is written. Adding 17845 alternatives of this form often allows GCC to produce better code 17846 when only some of the inputs can be affected by the earlyclobber. 17847 See, for example, the `mulsi3' insn of the ARM. 17848 17849 `&' does not obviate the need to write `='. 17850 17851 `%' 17852 Declares the instruction to be commutative for this operand and the 17853 following operand. This means that the compiler may interchange 17854 the two operands if that is the cheapest way to make all operands 17855 fit the constraints. This is often used in patterns for addition 17856 instructions that really have only two operands: the result must 17857 go in one of the arguments. Here for example, is how the 68000 17858 halfword-add instruction is defined: 17859 17860 (define_insn "addhi3" 17861 [(set (match_operand:HI 0 "general_operand" "=m,r") 17862 (plus:HI (match_operand:HI 1 "general_operand" "%0,0") 17863 (match_operand:HI 2 "general_operand" "di,g")))] 17864 ...) 17865 GCC can only handle one commutative pair in an asm; if you use 17866 more, the compiler may fail. Note that you need not use the 17867 modifier if the two alternatives are strictly identical; this 17868 would only waste time in the reload pass. The modifier is not 17869 operational after register allocation, so the result of 17870 `define_peephole2' and `define_split's performed after reload 17871 cannot rely on `%' to make the intended insn match. 17872 17873 `#' 17874 Says that all following characters, up to the next comma, are to be 17875 ignored as a constraint. They are significant only for choosing 17876 register preferences. 17877 17878 `*' 17879 Says that the following character should be ignored when choosing 17880 register preferences. `*' has no effect on the meaning of the 17881 constraint as a constraint, and no effect on reloading. 17882 17883 Here is an example: the 68000 has an instruction to sign-extend a 17884 halfword in a data register, and can also sign-extend a value by 17885 copying it into an address register. While either kind of 17886 register is acceptable, the constraints on an address-register 17887 destination are less strict, so it is best if register allocation 17888 makes an address register its goal. Therefore, `*' is used so 17889 that the `d' constraint letter (for data register) is ignored when 17890 computing register preferences. 17891 17892 (define_insn "extendhisi2" 17893 [(set (match_operand:SI 0 "general_operand" "=*d,a") 17894 (sign_extend:SI 17895 (match_operand:HI 1 "general_operand" "0,g")))] 17896 ...) 17897 17898 17899 File: gccint.info, Node: Machine Constraints, Next: Define Constraints, Prev: Disable Insn Alternatives, Up: Constraints 17900 17901 16.8.5 Constraints for Particular Machines 17902 ------------------------------------------ 17903 17904 Whenever possible, you should use the general-purpose constraint letters 17905 in `asm' arguments, since they will convey meaning more readily to 17906 people reading your code. Failing that, use the constraint letters 17907 that usually have very similar meanings across architectures. The most 17908 commonly used constraints are `m' and `r' (for memory and 17909 general-purpose registers respectively; *note Simple Constraints::), and 17910 `I', usually the letter indicating the most common immediate-constant 17911 format. 17912 17913 Each architecture defines additional constraints. These constraints 17914 are used by the compiler itself for instruction generation, as well as 17915 for `asm' statements; therefore, some of the constraints are not 17916 particularly useful for `asm'. Here is a summary of some of the 17917 machine-dependent constraints available on some particular machines; it 17918 includes both constraints that are useful for `asm' and constraints 17919 that aren't. The compiler source file mentioned in the table heading 17920 for each architecture is the definitive reference for the meanings of 17921 that architecture's constraints. 17922 17923 _ARM family--`config/arm/arm.h'_ 17924 17925 `f' 17926 Floating-point register 17927 17928 `w' 17929 VFP floating-point register 17930 17931 `F' 17932 One of the floating-point constants 0.0, 0.5, 1.0, 2.0, 3.0, 17933 4.0, 5.0 or 10.0 17934 17935 `G' 17936 Floating-point constant that would satisfy the constraint `F' 17937 if it were negated 17938 17939 `I' 17940 Integer that is valid as an immediate operand in a data 17941 processing instruction. That is, an integer in the range 0 17942 to 255 rotated by a multiple of 2 17943 17944 `J' 17945 Integer in the range -4095 to 4095 17946 17947 `K' 17948 Integer that satisfies constraint `I' when inverted (ones 17949 complement) 17950 17951 `L' 17952 Integer that satisfies constraint `I' when negated (twos 17953 complement) 17954 17955 `M' 17956 Integer in the range 0 to 32 17957 17958 `Q' 17959 A memory reference where the exact address is in a single 17960 register (``m'' is preferable for `asm' statements) 17961 17962 `R' 17963 An item in the constant pool 17964 17965 `S' 17966 A symbol in the text segment of the current file 17967 17968 `Uv' 17969 A memory reference suitable for VFP load/store insns 17970 (reg+constant offset) 17971 17972 `Uy' 17973 A memory reference suitable for iWMMXt load/store 17974 instructions. 17975 17976 `Uq' 17977 A memory reference suitable for the ARMv4 ldrsb instruction. 17978 17979 _AVR family--`config/avr/constraints.md'_ 17980 17981 `l' 17982 Registers from r0 to r15 17983 17984 `a' 17985 Registers from r16 to r23 17986 17987 `d' 17988 Registers from r16 to r31 17989 17990 `w' 17991 Registers from r24 to r31. These registers can be used in 17992 `adiw' command 17993 17994 `e' 17995 Pointer register (r26-r31) 17996 17997 `b' 17998 Base pointer register (r28-r31) 17999 18000 `q' 18001 Stack pointer register (SPH:SPL) 18002 18003 `t' 18004 Temporary register r0 18005 18006 `x' 18007 Register pair X (r27:r26) 18008 18009 `y' 18010 Register pair Y (r29:r28) 18011 18012 `z' 18013 Register pair Z (r31:r30) 18014 18015 `I' 18016 Constant greater than -1, less than 64 18017 18018 `J' 18019 Constant greater than -64, less than 1 18020 18021 `K' 18022 Constant integer 2 18023 18024 `L' 18025 Constant integer 0 18026 18027 `M' 18028 Constant that fits in 8 bits 18029 18030 `N' 18031 Constant integer -1 18032 18033 `O' 18034 Constant integer 8, 16, or 24 18035 18036 `P' 18037 Constant integer 1 18038 18039 `G' 18040 A floating point constant 0.0 18041 18042 `R' 18043 Integer constant in the range -6 ... 5. 18044 18045 `Q' 18046 A memory address based on Y or Z pointer with displacement. 18047 18048 _CRX Architecture--`config/crx/crx.h'_ 18049 18050 `b' 18051 Registers from r0 to r14 (registers without stack pointer) 18052 18053 `l' 18054 Register r16 (64-bit accumulator lo register) 18055 18056 `h' 18057 Register r17 (64-bit accumulator hi register) 18058 18059 `k' 18060 Register pair r16-r17. (64-bit accumulator lo-hi pair) 18061 18062 `I' 18063 Constant that fits in 3 bits 18064 18065 `J' 18066 Constant that fits in 4 bits 18067 18068 `K' 18069 Constant that fits in 5 bits 18070 18071 `L' 18072 Constant that is one of -1, 4, -4, 7, 8, 12, 16, 20, 32, 48 18073 18074 `G' 18075 Floating point constant that is legal for store immediate 18076 18077 _Hewlett-Packard PA-RISC--`config/pa/pa.h'_ 18078 18079 `a' 18080 General register 1 18081 18082 `f' 18083 Floating point register 18084 18085 `q' 18086 Shift amount register 18087 18088 `x' 18089 Floating point register (deprecated) 18090 18091 `y' 18092 Upper floating point register (32-bit), floating point 18093 register (64-bit) 18094 18095 `Z' 18096 Any register 18097 18098 `I' 18099 Signed 11-bit integer constant 18100 18101 `J' 18102 Signed 14-bit integer constant 18103 18104 `K' 18105 Integer constant that can be deposited with a `zdepi' 18106 instruction 18107 18108 `L' 18109 Signed 5-bit integer constant 18110 18111 `M' 18112 Integer constant 0 18113 18114 `N' 18115 Integer constant that can be loaded with a `ldil' instruction 18116 18117 `O' 18118 Integer constant whose value plus one is a power of 2 18119 18120 `P' 18121 Integer constant that can be used for `and' operations in 18122 `depi' and `extru' instructions 18123 18124 `S' 18125 Integer constant 31 18126 18127 `U' 18128 Integer constant 63 18129 18130 `G' 18131 Floating-point constant 0.0 18132 18133 `A' 18134 A `lo_sum' data-linkage-table memory operand 18135 18136 `Q' 18137 A memory operand that can be used as the destination operand 18138 of an integer store instruction 18139 18140 `R' 18141 A scaled or unscaled indexed memory operand 18142 18143 `T' 18144 A memory operand for floating-point loads and stores 18145 18146 `W' 18147 A register indirect memory operand 18148 18149 _picoChip family--`picochip.h'_ 18150 18151 `k' 18152 Stack register. 18153 18154 `f' 18155 Pointer register. A register which can be used to access 18156 memory without supplying an offset. Any other register can 18157 be used to access memory, but will need a constant offset. 18158 In the case of the offset being zero, it is more efficient to 18159 use a pointer register, since this reduces code size. 18160 18161 `t' 18162 A twin register. A register which may be paired with an 18163 adjacent register to create a 32-bit register. 18164 18165 `a' 18166 Any absolute memory address (e.g., symbolic constant, symbolic 18167 constant + offset). 18168 18169 `I' 18170 4-bit signed integer. 18171 18172 `J' 18173 4-bit unsigned integer. 18174 18175 `K' 18176 8-bit signed integer. 18177 18178 `M' 18179 Any constant whose absolute value is no greater than 4-bits. 18180 18181 `N' 18182 10-bit signed integer 18183 18184 `O' 18185 16-bit signed integer. 18186 18187 18188 _PowerPC and IBM RS6000--`config/rs6000/rs6000.h'_ 18189 18190 `b' 18191 Address base register 18192 18193 `f' 18194 Floating point register 18195 18196 `v' 18197 Vector register 18198 18199 `h' 18200 `MQ', `CTR', or `LINK' register 18201 18202 `q' 18203 `MQ' register 18204 18205 `c' 18206 `CTR' register 18207 18208 `l' 18209 `LINK' register 18210 18211 `x' 18212 `CR' register (condition register) number 0 18213 18214 `y' 18215 `CR' register (condition register) 18216 18217 `z' 18218 `FPMEM' stack memory for FPR-GPR transfers 18219 18220 `I' 18221 Signed 16-bit constant 18222 18223 `J' 18224 Unsigned 16-bit constant shifted left 16 bits (use `L' 18225 instead for `SImode' constants) 18226 18227 `K' 18228 Unsigned 16-bit constant 18229 18230 `L' 18231 Signed 16-bit constant shifted left 16 bits 18232 18233 `M' 18234 Constant larger than 31 18235 18236 `N' 18237 Exact power of 2 18238 18239 `O' 18240 Zero 18241 18242 `P' 18243 Constant whose negation is a signed 16-bit constant 18244 18245 `G' 18246 Floating point constant that can be loaded into a register 18247 with one instruction per word 18248 18249 `H' 18250 Integer/Floating point constant that can be loaded into a 18251 register using three instructions 18252 18253 `Q' 18254 Memory operand that is an offset from a register (`m' is 18255 preferable for `asm' statements) 18256 18257 `Z' 18258 Memory operand that is an indexed or indirect from a register 18259 (`m' is preferable for `asm' statements) 18260 18261 `R' 18262 AIX TOC entry 18263 18264 `a' 18265 Address operand that is an indexed or indirect from a 18266 register (`p' is preferable for `asm' statements) 18267 18268 `S' 18269 Constant suitable as a 64-bit mask operand 18270 18271 `T' 18272 Constant suitable as a 32-bit mask operand 18273 18274 `U' 18275 System V Release 4 small data area reference 18276 18277 `t' 18278 AND masks that can be performed by two rldic{l, r} 18279 instructions 18280 18281 `W' 18282 Vector constant that does not require memory 18283 18284 18285 _Intel 386--`config/i386/constraints.md'_ 18286 18287 `R' 18288 Legacy register--the eight integer registers available on all 18289 i386 processors (`a', `b', `c', `d', `si', `di', `bp', `sp'). 18290 18291 `q' 18292 Any register accessible as `Rl'. In 32-bit mode, `a', `b', 18293 `c', and `d'; in 64-bit mode, any integer register. 18294 18295 `Q' 18296 Any register accessible as `Rh': `a', `b', `c', and `d'. 18297 18298 `l' 18299 Any register that can be used as the index in a base+index 18300 memory access: that is, any general register except the stack 18301 pointer. 18302 18303 `a' 18304 The `a' register. 18305 18306 `b' 18307 The `b' register. 18308 18309 `c' 18310 The `c' register. 18311 18312 `d' 18313 The `d' register. 18314 18315 `S' 18316 The `si' register. 18317 18318 `D' 18319 The `di' register. 18320 18321 `A' 18322 The `a' and `d' registers, as a pair (for instructions that 18323 return half the result in one and half in the other). 18324 18325 `f' 18326 Any 80387 floating-point (stack) register. 18327 18328 `t' 18329 Top of 80387 floating-point stack (`%st(0)'). 18330 18331 `u' 18332 Second from top of 80387 floating-point stack (`%st(1)'). 18333 18334 `y' 18335 Any MMX register. 18336 18337 `x' 18338 Any SSE register. 18339 18340 `Yz' 18341 First SSE register (`%xmm0'). 18342 18343 `Y2' 18344 Any SSE register, when SSE2 is enabled. 18345 18346 `Yi' 18347 Any SSE register, when SSE2 and inter-unit moves are enabled. 18348 18349 `Ym' 18350 Any MMX register, when inter-unit moves are enabled. 18351 18352 `I' 18353 Integer constant in the range 0 ... 31, for 32-bit shifts. 18354 18355 `J' 18356 Integer constant in the range 0 ... 63, for 64-bit shifts. 18357 18358 `K' 18359 Signed 8-bit integer constant. 18360 18361 `L' 18362 `0xFF' or `0xFFFF', for andsi as a zero-extending move. 18363 18364 `M' 18365 0, 1, 2, or 3 (shifts for the `lea' instruction). 18366 18367 `N' 18368 Unsigned 8-bit integer constant (for `in' and `out' 18369 instructions). 18370 18371 `O' 18372 Integer constant in the range 0 ... 127, for 128-bit shifts. 18373 18374 `G' 18375 Standard 80387 floating point constant. 18376 18377 `C' 18378 Standard SSE floating point constant. 18379 18380 `e' 18381 32-bit signed integer constant, or a symbolic reference known 18382 to fit that range (for immediate operands in sign-extending 18383 x86-64 instructions). 18384 18385 `Z' 18386 32-bit unsigned integer constant, or a symbolic reference 18387 known to fit that range (for immediate operands in 18388 zero-extending x86-64 instructions). 18389 18390 18391 _Intel IA-64--`config/ia64/ia64.h'_ 18392 18393 `a' 18394 General register `r0' to `r3' for `addl' instruction 18395 18396 `b' 18397 Branch register 18398 18399 `c' 18400 Predicate register (`c' as in "conditional") 18401 18402 `d' 18403 Application register residing in M-unit 18404 18405 `e' 18406 Application register residing in I-unit 18407 18408 `f' 18409 Floating-point register 18410 18411 `m' 18412 Memory operand. Remember that `m' allows postincrement and 18413 postdecrement which require printing with `%Pn' on IA-64. 18414 Use `S' to disallow postincrement and postdecrement. 18415 18416 `G' 18417 Floating-point constant 0.0 or 1.0 18418 18419 `I' 18420 14-bit signed integer constant 18421 18422 `J' 18423 22-bit signed integer constant 18424 18425 `K' 18426 8-bit signed integer constant for logical instructions 18427 18428 `L' 18429 8-bit adjusted signed integer constant for compare pseudo-ops 18430 18431 `M' 18432 6-bit unsigned integer constant for shift counts 18433 18434 `N' 18435 9-bit signed integer constant for load and store 18436 postincrements 18437 18438 `O' 18439 The constant zero 18440 18441 `P' 18442 0 or -1 for `dep' instruction 18443 18444 `Q' 18445 Non-volatile memory for floating-point loads and stores 18446 18447 `R' 18448 Integer constant in the range 1 to 4 for `shladd' instruction 18449 18450 `S' 18451 Memory operand except postincrement and postdecrement 18452 18453 _FRV--`config/frv/frv.h'_ 18454 18455 `a' 18456 Register in the class `ACC_REGS' (`acc0' to `acc7'). 18457 18458 `b' 18459 Register in the class `EVEN_ACC_REGS' (`acc0' to `acc7'). 18460 18461 `c' 18462 Register in the class `CC_REGS' (`fcc0' to `fcc3' and `icc0' 18463 to `icc3'). 18464 18465 `d' 18466 Register in the class `GPR_REGS' (`gr0' to `gr63'). 18467 18468 `e' 18469 Register in the class `EVEN_REGS' (`gr0' to `gr63'). Odd 18470 registers are excluded not in the class but through the use 18471 of a machine mode larger than 4 bytes. 18472 18473 `f' 18474 Register in the class `FPR_REGS' (`fr0' to `fr63'). 18475 18476 `h' 18477 Register in the class `FEVEN_REGS' (`fr0' to `fr63'). Odd 18478 registers are excluded not in the class but through the use 18479 of a machine mode larger than 4 bytes. 18480 18481 `l' 18482 Register in the class `LR_REG' (the `lr' register). 18483 18484 `q' 18485 Register in the class `QUAD_REGS' (`gr2' to `gr63'). 18486 Register numbers not divisible by 4 are excluded not in the 18487 class but through the use of a machine mode larger than 8 18488 bytes. 18489 18490 `t' 18491 Register in the class `ICC_REGS' (`icc0' to `icc3'). 18492 18493 `u' 18494 Register in the class `FCC_REGS' (`fcc0' to `fcc3'). 18495 18496 `v' 18497 Register in the class `ICR_REGS' (`cc4' to `cc7'). 18498 18499 `w' 18500 Register in the class `FCR_REGS' (`cc0' to `cc3'). 18501 18502 `x' 18503 Register in the class `QUAD_FPR_REGS' (`fr0' to `fr63'). 18504 Register numbers not divisible by 4 are excluded not in the 18505 class but through the use of a machine mode larger than 8 18506 bytes. 18507 18508 `z' 18509 Register in the class `SPR_REGS' (`lcr' and `lr'). 18510 18511 `A' 18512 Register in the class `QUAD_ACC_REGS' (`acc0' to `acc7'). 18513 18514 `B' 18515 Register in the class `ACCG_REGS' (`accg0' to `accg7'). 18516 18517 `C' 18518 Register in the class `CR_REGS' (`cc0' to `cc7'). 18519 18520 `G' 18521 Floating point constant zero 18522 18523 `I' 18524 6-bit signed integer constant 18525 18526 `J' 18527 10-bit signed integer constant 18528 18529 `L' 18530 16-bit signed integer constant 18531 18532 `M' 18533 16-bit unsigned integer constant 18534 18535 `N' 18536 12-bit signed integer constant that is negative--i.e. in the 18537 range of -2048 to -1 18538 18539 `O' 18540 Constant zero 18541 18542 `P' 18543 12-bit signed integer constant that is greater than 18544 zero--i.e. in the range of 1 to 2047. 18545 18546 18547 _Blackfin family--`config/bfin/constraints.md'_ 18548 18549 `a' 18550 P register 18551 18552 `d' 18553 D register 18554 18555 `z' 18556 A call clobbered P register. 18557 18558 `qN' 18559 A single register. If N is in the range 0 to 7, the 18560 corresponding D register. If it is `A', then the register P0. 18561 18562 `D' 18563 Even-numbered D register 18564 18565 `W' 18566 Odd-numbered D register 18567 18568 `e' 18569 Accumulator register. 18570 18571 `A' 18572 Even-numbered accumulator register. 18573 18574 `B' 18575 Odd-numbered accumulator register. 18576 18577 `b' 18578 I register 18579 18580 `v' 18581 B register 18582 18583 `f' 18584 M register 18585 18586 `c' 18587 Registers used for circular buffering, i.e. I, B, or L 18588 registers. 18589 18590 `C' 18591 The CC register. 18592 18593 `t' 18594 LT0 or LT1. 18595 18596 `k' 18597 LC0 or LC1. 18598 18599 `u' 18600 LB0 or LB1. 18601 18602 `x' 18603 Any D, P, B, M, I or L register. 18604 18605 `y' 18606 Additional registers typically used only in prologues and 18607 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and 18608 USP. 18609 18610 `w' 18611 Any register except accumulators or CC. 18612 18613 `Ksh' 18614 Signed 16 bit integer (in the range -32768 to 32767) 18615 18616 `Kuh' 18617 Unsigned 16 bit integer (in the range 0 to 65535) 18618 18619 `Ks7' 18620 Signed 7 bit integer (in the range -64 to 63) 18621 18622 `Ku7' 18623 Unsigned 7 bit integer (in the range 0 to 127) 18624 18625 `Ku5' 18626 Unsigned 5 bit integer (in the range 0 to 31) 18627 18628 `Ks4' 18629 Signed 4 bit integer (in the range -8 to 7) 18630 18631 `Ks3' 18632 Signed 3 bit integer (in the range -3 to 4) 18633 18634 `Ku3' 18635 Unsigned 3 bit integer (in the range 0 to 7) 18636 18637 `PN' 18638 Constant N, where N is a single-digit constant in the range 0 18639 to 4. 18640 18641 `PA' 18642 An integer equal to one of the MACFLAG_XXX constants that is 18643 suitable for use with either accumulator. 18644 18645 `PB' 18646 An integer equal to one of the MACFLAG_XXX constants that is 18647 suitable for use only with accumulator A1. 18648 18649 `M1' 18650 Constant 255. 18651 18652 `M2' 18653 Constant 65535. 18654 18655 `J' 18656 An integer constant with exactly a single bit set. 18657 18658 `L' 18659 An integer constant with all bits set except exactly one. 18660 18661 `H' 18662 18663 `Q' 18664 Any SYMBOL_REF. 18665 18666 _M32C--`config/m32c/m32c.c'_ 18667 18668 `Rsp' 18669 `Rfb' 18670 `Rsb' 18671 `$sp', `$fb', `$sb'. 18672 18673 `Rcr' 18674 Any control register, when they're 16 bits wide (nothing if 18675 control registers are 24 bits wide) 18676 18677 `Rcl' 18678 Any control register, when they're 24 bits wide. 18679 18680 `R0w' 18681 `R1w' 18682 `R2w' 18683 `R3w' 18684 $r0, $r1, $r2, $r3. 18685 18686 `R02' 18687 $r0 or $r2, or $r2r0 for 32 bit values. 18688 18689 `R13' 18690 $r1 or $r3, or $r3r1 for 32 bit values. 18691 18692 `Rdi' 18693 A register that can hold a 64 bit value. 18694 18695 `Rhl' 18696 $r0 or $r1 (registers with addressable high/low bytes) 18697 18698 `R23' 18699 $r2 or $r3 18700 18701 `Raa' 18702 Address registers 18703 18704 `Raw' 18705 Address registers when they're 16 bits wide. 18706 18707 `Ral' 18708 Address registers when they're 24 bits wide. 18709 18710 `Rqi' 18711 Registers that can hold QI values. 18712 18713 `Rad' 18714 Registers that can be used with displacements ($a0, $a1, $sb). 18715 18716 `Rsi' 18717 Registers that can hold 32 bit values. 18718 18719 `Rhi' 18720 Registers that can hold 16 bit values. 18721 18722 `Rhc' 18723 Registers chat can hold 16 bit values, including all control 18724 registers. 18725 18726 `Rra' 18727 $r0 through R1, plus $a0 and $a1. 18728 18729 `Rfl' 18730 The flags register. 18731 18732 `Rmm' 18733 The memory-based pseudo-registers $mem0 through $mem15. 18734 18735 `Rpi' 18736 Registers that can hold pointers (16 bit registers for r8c, 18737 m16c; 24 bit registers for m32cm, m32c). 18738 18739 `Rpa' 18740 Matches multiple registers in a PARALLEL to form a larger 18741 register. Used to match function return values. 18742 18743 `Is3' 18744 -8 ... 7 18745 18746 `IS1' 18747 -128 ... 127 18748 18749 `IS2' 18750 -32768 ... 32767 18751 18752 `IU2' 18753 0 ... 65535 18754 18755 `In4' 18756 -8 ... -1 or 1 ... 8 18757 18758 `In5' 18759 -16 ... -1 or 1 ... 16 18760 18761 `In6' 18762 -32 ... -1 or 1 ... 32 18763 18764 `IM2' 18765 -65536 ... -1 18766 18767 `Ilb' 18768 An 8 bit value with exactly one bit set. 18769 18770 `Ilw' 18771 A 16 bit value with exactly one bit set. 18772 18773 `Sd' 18774 The common src/dest memory addressing modes. 18775 18776 `Sa' 18777 Memory addressed using $a0 or $a1. 18778 18779 `Si' 18780 Memory addressed with immediate addresses. 18781 18782 `Ss' 18783 Memory addressed using the stack pointer ($sp). 18784 18785 `Sf' 18786 Memory addressed using the frame base register ($fb). 18787 18788 `Ss' 18789 Memory addressed using the small base register ($sb). 18790 18791 `S1' 18792 $r1h 18793 18794 _MIPS--`config/mips/constraints.md'_ 18795 18796 `d' 18797 An address register. This is equivalent to `r' unless 18798 generating MIPS16 code. 18799 18800 `f' 18801 A floating-point register (if available). 18802 18803 `h' 18804 Formerly the `hi' register. This constraint is no longer 18805 supported. 18806 18807 `l' 18808 The `lo' register. Use this register to store values that are 18809 no bigger than a word. 18810 18811 `x' 18812 The concatenated `hi' and `lo' registers. Use this register 18813 to store doubleword values. 18814 18815 `c' 18816 A register suitable for use in an indirect jump. This will 18817 always be `$25' for `-mabicalls'. 18818 18819 `v' 18820 Register `$3'. Do not use this constraint in new code; it is 18821 retained only for compatibility with glibc. 18822 18823 `y' 18824 Equivalent to `r'; retained for backwards compatibility. 18825 18826 `z' 18827 A floating-point condition code register. 18828 18829 `I' 18830 A signed 16-bit constant (for arithmetic instructions). 18831 18832 `J' 18833 Integer zero. 18834 18835 `K' 18836 An unsigned 16-bit constant (for logic instructions). 18837 18838 `L' 18839 A signed 32-bit constant in which the lower 16 bits are zero. 18840 Such constants can be loaded using `lui'. 18841 18842 `M' 18843 A constant that cannot be loaded using `lui', `addiu' or 18844 `ori'. 18845 18846 `N' 18847 A constant in the range -65535 to -1 (inclusive). 18848 18849 `O' 18850 A signed 15-bit constant. 18851 18852 `P' 18853 A constant in the range 1 to 65535 (inclusive). 18854 18855 `G' 18856 Floating-point zero. 18857 18858 `R' 18859 An address that can be used in a non-macro load or store. 18860 18861 _Motorola 680x0--`config/m68k/constraints.md'_ 18862 18863 `a' 18864 Address register 18865 18866 `d' 18867 Data register 18868 18869 `f' 18870 68881 floating-point register, if available 18871 18872 `I' 18873 Integer in the range 1 to 8 18874 18875 `J' 18876 16-bit signed number 18877 18878 `K' 18879 Signed number whose magnitude is greater than 0x80 18880 18881 `L' 18882 Integer in the range -8 to -1 18883 18884 `M' 18885 Signed number whose magnitude is greater than 0x100 18886 18887 `N' 18888 Range 24 to 31, rotatert:SI 8 to 1 expressed as rotate 18889 18890 `O' 18891 16 (for rotate using swap) 18892 18893 `P' 18894 Range 8 to 15, rotatert:HI 8 to 1 expressed as rotate 18895 18896 `R' 18897 Numbers that mov3q can handle 18898 18899 `G' 18900 Floating point constant that is not a 68881 constant 18901 18902 `S' 18903 Operands that satisfy 'm' when -mpcrel is in effect 18904 18905 `T' 18906 Operands that satisfy 's' when -mpcrel is not in effect 18907 18908 `Q' 18909 Address register indirect addressing mode 18910 18911 `U' 18912 Register offset addressing 18913 18914 `W' 18915 const_call_operand 18916 18917 `Cs' 18918 symbol_ref or const 18919 18920 `Ci' 18921 const_int 18922 18923 `C0' 18924 const_int 0 18925 18926 `Cj' 18927 Range of signed numbers that don't fit in 16 bits 18928 18929 `Cmvq' 18930 Integers valid for mvq 18931 18932 `Capsw' 18933 Integers valid for a moveq followed by a swap 18934 18935 `Cmvz' 18936 Integers valid for mvz 18937 18938 `Cmvs' 18939 Integers valid for mvs 18940 18941 `Ap' 18942 push_operand 18943 18944 `Ac' 18945 Non-register operands allowed in clr 18946 18947 18948 _Motorola 68HC11 & 68HC12 families--`config/m68hc11/m68hc11.h'_ 18949 18950 `a' 18951 Register `a' 18952 18953 `b' 18954 Register `b' 18955 18956 `d' 18957 Register `d' 18958 18959 `q' 18960 An 8-bit register 18961 18962 `t' 18963 Temporary soft register _.tmp 18964 18965 `u' 18966 A soft register _.d1 to _.d31 18967 18968 `w' 18969 Stack pointer register 18970 18971 `x' 18972 Register `x' 18973 18974 `y' 18975 Register `y' 18976 18977 `z' 18978 Pseudo register `z' (replaced by `x' or `y' at the end) 18979 18980 `A' 18981 An address register: x, y or z 18982 18983 `B' 18984 An address register: x or y 18985 18986 `D' 18987 Register pair (x:d) to form a 32-bit value 18988 18989 `L' 18990 Constants in the range -65536 to 65535 18991 18992 `M' 18993 Constants whose 16-bit low part is zero 18994 18995 `N' 18996 Constant integer 1 or -1 18997 18998 `O' 18999 Constant integer 16 19000 19001 `P' 19002 Constants in the range -8 to 2 19003 19004 19005 _SPARC--`config/sparc/sparc.h'_ 19006 19007 `f' 19008 Floating-point register on the SPARC-V8 architecture and 19009 lower floating-point register on the SPARC-V9 architecture. 19010 19011 `e' 19012 Floating-point register. It is equivalent to `f' on the 19013 SPARC-V8 architecture and contains both lower and upper 19014 floating-point registers on the SPARC-V9 architecture. 19015 19016 `c' 19017 Floating-point condition code register. 19018 19019 `d' 19020 Lower floating-point register. It is only valid on the 19021 SPARC-V9 architecture when the Visual Instruction Set is 19022 available. 19023 19024 `b' 19025 Floating-point register. It is only valid on the SPARC-V9 19026 architecture when the Visual Instruction Set is available. 19027 19028 `h' 19029 64-bit global or out register for the SPARC-V8+ architecture. 19030 19031 `D' 19032 A vector constant 19033 19034 `I' 19035 Signed 13-bit constant 19036 19037 `J' 19038 Zero 19039 19040 `K' 19041 32-bit constant with the low 12 bits clear (a constant that 19042 can be loaded with the `sethi' instruction) 19043 19044 `L' 19045 A constant in the range supported by `movcc' instructions 19046 19047 `M' 19048 A constant in the range supported by `movrcc' instructions 19049 19050 `N' 19051 Same as `K', except that it verifies that bits that are not 19052 in the lower 32-bit range are all zero. Must be used instead 19053 of `K' for modes wider than `SImode' 19054 19055 `O' 19056 The constant 4096 19057 19058 `G' 19059 Floating-point zero 19060 19061 `H' 19062 Signed 13-bit constant, sign-extended to 32 or 64 bits 19063 19064 `Q' 19065 Floating-point constant whose integral representation can be 19066 moved into an integer register using a single sethi 19067 instruction 19068 19069 `R' 19070 Floating-point constant whose integral representation can be 19071 moved into an integer register using a single mov instruction 19072 19073 `S' 19074 Floating-point constant whose integral representation can be 19075 moved into an integer register using a high/lo_sum 19076 instruction sequence 19077 19078 `T' 19079 Memory address aligned to an 8-byte boundary 19080 19081 `U' 19082 Even register 19083 19084 `W' 19085 Memory address for `e' constraint registers 19086 19087 `Y' 19088 Vector zero 19089 19090 19091 _SPU--`config/spu/spu.h'_ 19092 19093 `a' 19094 An immediate which can be loaded with the il/ila/ilh/ilhu 19095 instructions. const_int is treated as a 64 bit value. 19096 19097 `c' 19098 An immediate for and/xor/or instructions. const_int is 19099 treated as a 64 bit value. 19100 19101 `d' 19102 An immediate for the `iohl' instruction. const_int is 19103 treated as a 64 bit value. 19104 19105 `f' 19106 An immediate which can be loaded with `fsmbi'. 19107 19108 `A' 19109 An immediate which can be loaded with the il/ila/ilh/ilhu 19110 instructions. const_int is treated as a 32 bit value. 19111 19112 `B' 19113 An immediate for most arithmetic instructions. const_int is 19114 treated as a 32 bit value. 19115 19116 `C' 19117 An immediate for and/xor/or instructions. const_int is 19118 treated as a 32 bit value. 19119 19120 `D' 19121 An immediate for the `iohl' instruction. const_int is 19122 treated as a 32 bit value. 19123 19124 `I' 19125 A constant in the range [-64, 63] for shift/rotate 19126 instructions. 19127 19128 `J' 19129 An unsigned 7-bit constant for conversion/nop/channel 19130 instructions. 19131 19132 `K' 19133 A signed 10-bit constant for most arithmetic instructions. 19134 19135 `M' 19136 A signed 16 bit immediate for `stop'. 19137 19138 `N' 19139 An unsigned 16-bit constant for `iohl' and `fsmbi'. 19140 19141 `O' 19142 An unsigned 7-bit constant whose 3 least significant bits are 19143 0. 19144 19145 `P' 19146 An unsigned 3-bit constant for 16-byte rotates and shifts 19147 19148 `R' 19149 Call operand, reg, for indirect calls 19150 19151 `S' 19152 Call operand, symbol, for relative calls. 19153 19154 `T' 19155 Call operand, const_int, for absolute calls. 19156 19157 `U' 19158 An immediate which can be loaded with the il/ila/ilh/ilhu 19159 instructions. const_int is sign extended to 128 bit. 19160 19161 `W' 19162 An immediate for shift and rotate instructions. const_int is 19163 treated as a 32 bit value. 19164 19165 `Y' 19166 An immediate for and/xor/or instructions. const_int is sign 19167 extended as a 128 bit. 19168 19169 `Z' 19170 An immediate for the `iohl' instruction. const_int is sign 19171 extended to 128 bit. 19172 19173 19174 _S/390 and zSeries--`config/s390/s390.h'_ 19175 19176 `a' 19177 Address register (general purpose register except r0) 19178 19179 `c' 19180 Condition code register 19181 19182 `d' 19183 Data register (arbitrary general purpose register) 19184 19185 `f' 19186 Floating-point register 19187 19188 `I' 19189 Unsigned 8-bit constant (0-255) 19190 19191 `J' 19192 Unsigned 12-bit constant (0-4095) 19193 19194 `K' 19195 Signed 16-bit constant (-32768-32767) 19196 19197 `L' 19198 Value appropriate as displacement. 19199 `(0..4095)' 19200 for short displacement 19201 19202 `(-524288..524287)' 19203 for long displacement 19204 19205 `M' 19206 Constant integer with a value of 0x7fffffff. 19207 19208 `N' 19209 Multiple letter constraint followed by 4 parameter letters. 19210 `0..9:' 19211 number of the part counting from most to least 19212 significant 19213 19214 `H,Q:' 19215 mode of the part 19216 19217 `D,S,H:' 19218 mode of the containing operand 19219 19220 `0,F:' 19221 value of the other parts (F--all bits set) 19222 The constraint matches if the specified part of a constant 19223 has a value different from its other parts. 19224 19225 `Q' 19226 Memory reference without index register and with short 19227 displacement. 19228 19229 `R' 19230 Memory reference with index register and short displacement. 19231 19232 `S' 19233 Memory reference without index register but with long 19234 displacement. 19235 19236 `T' 19237 Memory reference with index register and long displacement. 19238 19239 `U' 19240 Pointer with short displacement. 19241 19242 `W' 19243 Pointer with long displacement. 19244 19245 `Y' 19246 Shift count operand. 19247 19248 19249 _Score family--`config/score/score.h'_ 19250 19251 `d' 19252 Registers from r0 to r32. 19253 19254 `e' 19255 Registers from r0 to r16. 19256 19257 `t' 19258 r8--r11 or r22--r27 registers. 19259 19260 `h' 19261 hi register. 19262 19263 `l' 19264 lo register. 19265 19266 `x' 19267 hi + lo register. 19268 19269 `q' 19270 cnt register. 19271 19272 `y' 19273 lcb register. 19274 19275 `z' 19276 scb register. 19277 19278 `a' 19279 cnt + lcb + scb register. 19280 19281 `c' 19282 cr0--cr15 register. 19283 19284 `b' 19285 cp1 registers. 19286 19287 `f' 19288 cp2 registers. 19289 19290 `i' 19291 cp3 registers. 19292 19293 `j' 19294 cp1 + cp2 + cp3 registers. 19295 19296 `I' 19297 High 16-bit constant (32-bit constant with 16 LSBs zero). 19298 19299 `J' 19300 Unsigned 5 bit integer (in the range 0 to 31). 19301 19302 `K' 19303 Unsigned 16 bit integer (in the range 0 to 65535). 19304 19305 `L' 19306 Signed 16 bit integer (in the range -32768 to 32767). 19307 19308 `M' 19309 Unsigned 14 bit integer (in the range 0 to 16383). 19310 19311 `N' 19312 Signed 14 bit integer (in the range -8192 to 8191). 19313 19314 `Z' 19315 Any SYMBOL_REF. 19316 19317 _Xstormy16--`config/stormy16/stormy16.h'_ 19318 19319 `a' 19320 Register r0. 19321 19322 `b' 19323 Register r1. 19324 19325 `c' 19326 Register r2. 19327 19328 `d' 19329 Register r8. 19330 19331 `e' 19332 Registers r0 through r7. 19333 19334 `t' 19335 Registers r0 and r1. 19336 19337 `y' 19338 The carry register. 19339 19340 `z' 19341 Registers r8 and r9. 19342 19343 `I' 19344 A constant between 0 and 3 inclusive. 19345 19346 `J' 19347 A constant that has exactly one bit set. 19348 19349 `K' 19350 A constant that has exactly one bit clear. 19351 19352 `L' 19353 A constant between 0 and 255 inclusive. 19354 19355 `M' 19356 A constant between -255 and 0 inclusive. 19357 19358 `N' 19359 A constant between -3 and 0 inclusive. 19360 19361 `O' 19362 A constant between 1 and 4 inclusive. 19363 19364 `P' 19365 A constant between -4 and -1 inclusive. 19366 19367 `Q' 19368 A memory reference that is a stack push. 19369 19370 `R' 19371 A memory reference that is a stack pop. 19372 19373 `S' 19374 A memory reference that refers to a constant address of known 19375 value. 19376 19377 `T' 19378 The register indicated by Rx (not implemented yet). 19379 19380 `U' 19381 A constant that is not between 2 and 15 inclusive. 19382 19383 `Z' 19384 The constant 0. 19385 19386 19387 _Xtensa--`config/xtensa/constraints.md'_ 19388 19389 `a' 19390 General-purpose 32-bit register 19391 19392 `b' 19393 One-bit boolean register 19394 19395 `A' 19396 MAC16 40-bit accumulator register 19397 19398 `I' 19399 Signed 12-bit integer constant, for use in MOVI instructions 19400 19401 `J' 19402 Signed 8-bit integer constant, for use in ADDI instructions 19403 19404 `K' 19405 Integer constant valid for BccI instructions 19406 19407 `L' 19408 Unsigned constant valid for BccUI instructions 19409 19410 19411 19412 19413 File: gccint.info, Node: Disable Insn Alternatives, Next: Machine Constraints, Prev: Modifiers, Up: Constraints 19414 19415 16.8.6 Disable insn alternatives using the `enabled' attribute 19416 -------------------------------------------------------------- 19417 19418 The `enabled' insn attribute may be used to disable certain insn 19419 alternatives for machine-specific reasons. This is useful when adding 19420 new instructions to an existing pattern which are only available for 19421 certain cpu architecture levels as specified with the `-march=' option. 19422 19423 If an insn alternative is disabled, then it will never be used. The 19424 compiler treats the constraints for the disabled alternative as 19425 unsatisfiable. 19426 19427 In order to make use of the `enabled' attribute a back end has to add 19428 in the machine description files: 19429 19430 1. A definition of the `enabled' insn attribute. The attribute is 19431 defined as usual using the `define_attr' command. This definition 19432 should be based on other insn attributes and/or target flags. The 19433 `enabled' attribute is a numeric attribute and should evaluate to 19434 `(const_int 1)' for an enabled alternative and to `(const_int 0)' 19435 otherwise. 19436 19437 2. A definition of another insn attribute used to describe for what 19438 reason an insn alternative might be available or not. E.g. 19439 `cpu_facility' as in the example below. 19440 19441 3. An assignment for the second attribute to each insn definition 19442 combining instructions which are not all available under the same 19443 circumstances. (Note: It obviously only makes sense for 19444 definitions with more than one alternative. Otherwise the insn 19445 pattern should be disabled or enabled using the insn condition.) 19446 19447 E.g. the following two patterns could easily be merged using the 19448 `enabled' attribute: 19449 19450 19451 (define_insn "*movdi_old" 19452 [(set (match_operand:DI 0 "register_operand" "=d") 19453 (match_operand:DI 1 "register_operand" " d"))] 19454 "!TARGET_NEW" 19455 "lgr %0,%1") 19456 19457 (define_insn "*movdi_new" 19458 [(set (match_operand:DI 0 "register_operand" "=d,f,d") 19459 (match_operand:DI 1 "register_operand" " d,d,f"))] 19460 "TARGET_NEW" 19461 "@ 19462 lgr %0,%1 19463 ldgr %0,%1 19464 lgdr %0,%1") 19465 19466 to: 19467 19468 19469 (define_insn "*movdi_combined" 19470 [(set (match_operand:DI 0 "register_operand" "=d,f,d") 19471 (match_operand:DI 1 "register_operand" " d,d,f"))] 19472 "" 19473 "@ 19474 lgr %0,%1 19475 ldgr %0,%1 19476 lgdr %0,%1" 19477 [(set_attr "cpu_facility" "*,new,new")]) 19478 19479 with the `enabled' attribute defined like this: 19480 19481 19482 (define_attr "cpu_facility" "standard,new" (const_string "standard")) 19483 19484 (define_attr "enabled" "" 19485 (cond [(eq_attr "cpu_facility" "standard") (const_int 1) 19486 (and (eq_attr "cpu_facility" "new") 19487 (ne (symbol_ref "TARGET_NEW") (const_int 0))) 19488 (const_int 1)] 19489 (const_int 0))) 19490 19491 19492 File: gccint.info, Node: Define Constraints, Next: C Constraint Interface, Prev: Machine Constraints, Up: Constraints 19493 19494 16.8.7 Defining Machine-Specific Constraints 19495 -------------------------------------------- 19496 19497 Machine-specific constraints fall into two categories: register and 19498 non-register constraints. Within the latter category, constraints 19499 which allow subsets of all possible memory or address operands should 19500 be specially marked, to give `reload' more information. 19501 19502 Machine-specific constraints can be given names of arbitrary length, 19503 but they must be entirely composed of letters, digits, underscores 19504 (`_'), and angle brackets (`< >'). Like C identifiers, they must begin 19505 with a letter or underscore. 19506 19507 In order to avoid ambiguity in operand constraint strings, no 19508 constraint can have a name that begins with any other constraint's 19509 name. For example, if `x' is defined as a constraint name, `xy' may 19510 not be, and vice versa. As a consequence of this rule, no constraint 19511 may begin with one of the generic constraint letters: `E F V X g i m n 19512 o p r s'. 19513 19514 Register constraints correspond directly to register classes. *Note 19515 Register Classes::. There is thus not much flexibility in their 19516 definitions. 19517 19518 -- MD Expression: define_register_constraint name regclass docstring 19519 All three arguments are string constants. NAME is the name of the 19520 constraint, as it will appear in `match_operand' expressions. If 19521 NAME is a multi-letter constraint its length shall be the same for 19522 all constraints starting with the same letter. REGCLASS can be 19523 either the name of the corresponding register class (*note 19524 Register Classes::), or a C expression which evaluates to the 19525 appropriate register class. If it is an expression, it must have 19526 no side effects, and it cannot look at the operand. The usual use 19527 of expressions is to map some register constraints to `NO_REGS' 19528 when the register class is not available on a given 19529 subarchitecture. 19530 19531 DOCSTRING is a sentence documenting the meaning of the constraint. 19532 Docstrings are explained further below. 19533 19534 Non-register constraints are more like predicates: the constraint 19535 definition gives a Boolean expression which indicates whether the 19536 constraint matches. 19537 19538 -- MD Expression: define_constraint name docstring exp 19539 The NAME and DOCSTRING arguments are the same as for 19540 `define_register_constraint', but note that the docstring comes 19541 immediately after the name for these expressions. EXP is an RTL 19542 expression, obeying the same rules as the RTL expressions in 19543 predicate definitions. *Note Defining Predicates::, for details. 19544 If it evaluates true, the constraint matches; if it evaluates 19545 false, it doesn't. Constraint expressions should indicate which 19546 RTL codes they might match, just like predicate expressions. 19547 19548 `match_test' C expressions have access to the following variables: 19549 19550 OP 19551 The RTL object defining the operand. 19552 19553 MODE 19554 The machine mode of OP. 19555 19556 IVAL 19557 `INTVAL (OP)', if OP is a `const_int'. 19558 19559 HVAL 19560 `CONST_DOUBLE_HIGH (OP)', if OP is an integer `const_double'. 19561 19562 LVAL 19563 `CONST_DOUBLE_LOW (OP)', if OP is an integer `const_double'. 19564 19565 RVAL 19566 `CONST_DOUBLE_REAL_VALUE (OP)', if OP is a floating-point 19567 `const_double'. 19568 19569 The *VAL variables should only be used once another piece of the 19570 expression has verified that OP is the appropriate kind of RTL 19571 object. 19572 19573 Most non-register constraints should be defined with 19574 `define_constraint'. The remaining two definition expressions are only 19575 appropriate for constraints that should be handled specially by 19576 `reload' if they fail to match. 19577 19578 -- MD Expression: define_memory_constraint name docstring exp 19579 Use this expression for constraints that match a subset of all 19580 memory operands: that is, `reload' can make them match by 19581 converting the operand to the form `(mem (reg X))', where X is a 19582 base register (from the register class specified by 19583 `BASE_REG_CLASS', *note Register Classes::). 19584 19585 For example, on the S/390, some instructions do not accept 19586 arbitrary memory references, but only those that do not make use 19587 of an index register. The constraint letter `Q' is defined to 19588 represent a memory address of this type. If `Q' is defined with 19589 `define_memory_constraint', a `Q' constraint can handle any memory 19590 operand, because `reload' knows it can simply copy the memory 19591 address into a base register if required. This is analogous to 19592 the way a `o' constraint can handle any memory operand. 19593 19594 The syntax and semantics are otherwise identical to 19595 `define_constraint'. 19596 19597 -- MD Expression: define_address_constraint name docstring exp 19598 Use this expression for constraints that match a subset of all 19599 address operands: that is, `reload' can make the constraint match 19600 by converting the operand to the form `(reg X)', again with X a 19601 base register. 19602 19603 Constraints defined with `define_address_constraint' can only be 19604 used with the `address_operand' predicate, or machine-specific 19605 predicates that work the same way. They are treated analogously to 19606 the generic `p' constraint. 19607 19608 The syntax and semantics are otherwise identical to 19609 `define_constraint'. 19610 19611 For historical reasons, names beginning with the letters `G H' are 19612 reserved for constraints that match only `const_double's, and names 19613 beginning with the letters `I J K L M N O P' are reserved for 19614 constraints that match only `const_int's. This may change in the 19615 future. For the time being, constraints with these names must be 19616 written in a stylized form, so that `genpreds' can tell you did it 19617 correctly: 19618 19619 (define_constraint "[GHIJKLMNOP]..." 19620 "DOC..." 19621 (and (match_code "const_int") ; `const_double' for G/H 19622 CONDITION...)) ; usually a `match_test' 19623 19624 It is fine to use names beginning with other letters for constraints 19625 that match `const_double's or `const_int's. 19626 19627 Each docstring in a constraint definition should be one or more 19628 complete sentences, marked up in Texinfo format. _They are currently 19629 unused._ In the future they will be copied into the GCC manual, in 19630 *Note Machine Constraints::, replacing the hand-maintained tables 19631 currently found in that section. Also, in the future the compiler may 19632 use this to give more helpful diagnostics when poor choice of `asm' 19633 constraints causes a reload failure. 19634 19635 If you put the pseudo-Texinfo directive `@internal' at the beginning 19636 of a docstring, then (in the future) it will appear only in the 19637 internals manual's version of the machine-specific constraint tables. 19638 Use this for constraints that should not appear in `asm' statements. 19639 19640 19641 File: gccint.info, Node: C Constraint Interface, Prev: Define Constraints, Up: Constraints 19642 19643 16.8.8 Testing constraints from C 19644 --------------------------------- 19645 19646 It is occasionally useful to test a constraint from C code rather than 19647 implicitly via the constraint string in a `match_operand'. The 19648 generated file `tm_p.h' declares a few interfaces for working with 19649 machine-specific constraints. None of these interfaces work with the 19650 generic constraints described in *Note Simple Constraints::. This may 19651 change in the future. 19652 19653 *Warning:* `tm_p.h' may declare other functions that operate on 19654 constraints, besides the ones documented here. Do not use those 19655 functions from machine-dependent code. They exist to implement the old 19656 constraint interface that machine-independent components of the 19657 compiler still expect. They will change or disappear in the future. 19658 19659 Some valid constraint names are not valid C identifiers, so there is a 19660 mangling scheme for referring to them from C. Constraint names that do 19661 not contain angle brackets or underscores are left unchanged. 19662 Underscores are doubled, each `<' is replaced with `_l', and each `>' 19663 with `_g'. Here are some examples: 19664 19665 *Original* *Mangled* 19666 `x' `x' 19667 `P42x' `P42x' 19668 `P4_x' `P4__x' 19669 `P4>x' `P4_gx' 19670 `P4>>' `P4_g_g' 19671 `P4_g>' `P4__g_g' 19672 19673 Throughout this section, the variable C is either a constraint in the 19674 abstract sense, or a constant from `enum constraint_num'; the variable 19675 M is a mangled constraint name (usually as part of a larger identifier). 19676 19677 -- Enum: constraint_num 19678 For each machine-specific constraint, there is a corresponding 19679 enumeration constant: `CONSTRAINT_' plus the mangled name of the 19680 constraint. Functions that take an `enum constraint_num' as an 19681 argument expect one of these constants. 19682 19683 Machine-independent constraints do not have associated constants. 19684 This may change in the future. 19685 19686 -- Function: inline bool satisfies_constraint_M (rtx EXP) 19687 For each machine-specific, non-register constraint M, there is one 19688 of these functions; it returns `true' if EXP satisfies the 19689 constraint. These functions are only visible if `rtl.h' was 19690 included before `tm_p.h'. 19691 19692 -- Function: bool constraint_satisfied_p (rtx EXP, enum constraint_num 19693 C) 19694 Like the `satisfies_constraint_M' functions, but the constraint to 19695 test is given as an argument, C. If C specifies a register 19696 constraint, this function will always return `false'. 19697 19698 -- Function: enum reg_class regclass_for_constraint (enum 19699 constraint_num C) 19700 Returns the register class associated with C. If C is not a 19701 register constraint, or those registers are not available for the 19702 currently selected subtarget, returns `NO_REGS'. 19703 19704 Here is an example use of `satisfies_constraint_M'. In peephole 19705 optimizations (*note Peephole Definitions::), operand constraint 19706 strings are ignored, so if there are relevant constraints, they must be 19707 tested in the C condition. In the example, the optimization is applied 19708 if operand 2 does _not_ satisfy the `K' constraint. (This is a 19709 simplified version of a peephole definition from the i386 machine 19710 description.) 19711 19712 (define_peephole2 19713 [(match_scratch:SI 3 "r") 19714 (set (match_operand:SI 0 "register_operand" "") 19715 (mult:SI (match_operand:SI 1 "memory_operand" "") 19716 (match_operand:SI 2 "immediate_operand" "")))] 19717 19718 "!satisfies_constraint_K (operands[2])" 19719 19720 [(set (match_dup 3) (match_dup 1)) 19721 (set (match_dup 0) (mult:SI (match_dup 3) (match_dup 2)))] 19722 19723 "") 19724 19725 19726 File: gccint.info, Node: Standard Names, Next: Pattern Ordering, Prev: Constraints, Up: Machine Desc 19727 19728 16.9 Standard Pattern Names For Generation 19729 ========================================== 19730 19731 Here is a table of the instruction names that are meaningful in the RTL 19732 generation pass of the compiler. Giving one of these names to an 19733 instruction pattern tells the RTL generation pass that it can use the 19734 pattern to accomplish a certain task. 19735 19736 `movM' 19737 Here M stands for a two-letter machine mode name, in lowercase. 19738 This instruction pattern moves data with that machine mode from 19739 operand 1 to operand 0. For example, `movsi' moves full-word data. 19740 19741 If operand 0 is a `subreg' with mode M of a register whose own 19742 mode is wider than M, the effect of this instruction is to store 19743 the specified value in the part of the register that corresponds 19744 to mode M. Bits outside of M, but which are within the same 19745 target word as the `subreg' are undefined. Bits which are outside 19746 the target word are left unchanged. 19747 19748 This class of patterns is special in several ways. First of all, 19749 each of these names up to and including full word size _must_ be 19750 defined, because there is no other way to copy a datum from one 19751 place to another. If there are patterns accepting operands in 19752 larger modes, `movM' must be defined for integer modes of those 19753 sizes. 19754 19755 Second, these patterns are not used solely in the RTL generation 19756 pass. Even the reload pass can generate move insns to copy values 19757 from stack slots into temporary registers. When it does so, one 19758 of the operands is a hard register and the other is an operand 19759 that can need to be reloaded into a register. 19760 19761 Therefore, when given such a pair of operands, the pattern must 19762 generate RTL which needs no reloading and needs no temporary 19763 registers--no registers other than the operands. For example, if 19764 you support the pattern with a `define_expand', then in such a 19765 case the `define_expand' mustn't call `force_reg' or any other such 19766 function which might generate new pseudo registers. 19767 19768 This requirement exists even for subword modes on a RISC machine 19769 where fetching those modes from memory normally requires several 19770 insns and some temporary registers. 19771 19772 During reload a memory reference with an invalid address may be 19773 passed as an operand. Such an address will be replaced with a 19774 valid address later in the reload pass. In this case, nothing may 19775 be done with the address except to use it as it stands. If it is 19776 copied, it will not be replaced with a valid address. No attempt 19777 should be made to make such an address into a valid address and no 19778 routine (such as `change_address') that will do so may be called. 19779 Note that `general_operand' will fail when applied to such an 19780 address. 19781 19782 The global variable `reload_in_progress' (which must be explicitly 19783 declared if required) can be used to determine whether such special 19784 handling is required. 19785 19786 The variety of operands that have reloads depends on the rest of 19787 the machine description, but typically on a RISC machine these can 19788 only be pseudo registers that did not get hard registers, while on 19789 other machines explicit memory references will get optional 19790 reloads. 19791 19792 If a scratch register is required to move an object to or from 19793 memory, it can be allocated using `gen_reg_rtx' prior to life 19794 analysis. 19795 19796 If there are cases which need scratch registers during or after 19797 reload, you must provide an appropriate secondary_reload target 19798 hook. 19799 19800 The macro `can_create_pseudo_p' can be used to determine if it is 19801 unsafe to create new pseudo registers. If this variable is 19802 nonzero, then it is unsafe to call `gen_reg_rtx' to allocate a new 19803 pseudo. 19804 19805 The constraints on a `movM' must permit moving any hard register 19806 to any other hard register provided that `HARD_REGNO_MODE_OK' 19807 permits mode M in both registers and `REGISTER_MOVE_COST' applied 19808 to their classes returns a value of 2. 19809 19810 It is obligatory to support floating point `movM' instructions 19811 into and out of any registers that can hold fixed point values, 19812 because unions and structures (which have modes `SImode' or 19813 `DImode') can be in those registers and they may have floating 19814 point members. 19815 19816 There may also be a need to support fixed point `movM' 19817 instructions in and out of floating point registers. 19818 Unfortunately, I have forgotten why this was so, and I don't know 19819 whether it is still true. If `HARD_REGNO_MODE_OK' rejects fixed 19820 point values in floating point registers, then the constraints of 19821 the fixed point `movM' instructions must be designed to avoid ever 19822 trying to reload into a floating point register. 19823 19824 `reload_inM' 19825 `reload_outM' 19826 These named patterns have been obsoleted by the target hook 19827 `secondary_reload'. 19828 19829 Like `movM', but used when a scratch register is required to move 19830 between operand 0 and operand 1. Operand 2 describes the scratch 19831 register. See the discussion of the `SECONDARY_RELOAD_CLASS' 19832 macro in *note Register Classes::. 19833 19834 There are special restrictions on the form of the `match_operand's 19835 used in these patterns. First, only the predicate for the reload 19836 operand is examined, i.e., `reload_in' examines operand 1, but not 19837 the predicates for operand 0 or 2. Second, there may be only one 19838 alternative in the constraints. Third, only a single register 19839 class letter may be used for the constraint; subsequent constraint 19840 letters are ignored. As a special exception, an empty constraint 19841 string matches the `ALL_REGS' register class. This may relieve 19842 ports of the burden of defining an `ALL_REGS' constraint letter 19843 just for these patterns. 19844 19845 `movstrictM' 19846 Like `movM' except that if operand 0 is a `subreg' with mode M of 19847 a register whose natural mode is wider, the `movstrictM' 19848 instruction is guaranteed not to alter any of the register except 19849 the part which belongs to mode M. 19850 19851 `movmisalignM' 19852 This variant of a move pattern is designed to load or store a value 19853 from a memory address that is not naturally aligned for its mode. 19854 For a store, the memory will be in operand 0; for a load, the 19855 memory will be in operand 1. The other operand is guaranteed not 19856 to be a memory, so that it's easy to tell whether this is a load 19857 or store. 19858 19859 This pattern is used by the autovectorizer, and when expanding a 19860 `MISALIGNED_INDIRECT_REF' expression. 19861 19862 `load_multiple' 19863 Load several consecutive memory locations into consecutive 19864 registers. Operand 0 is the first of the consecutive registers, 19865 operand 1 is the first memory location, and operand 2 is a 19866 constant: the number of consecutive registers. 19867 19868 Define this only if the target machine really has such an 19869 instruction; do not define this if the most efficient way of 19870 loading consecutive registers from memory is to do them one at a 19871 time. 19872 19873 On some machines, there are restrictions as to which consecutive 19874 registers can be stored into memory, such as particular starting or 19875 ending register numbers or only a range of valid counts. For those 19876 machines, use a `define_expand' (*note Expander Definitions::) and 19877 make the pattern fail if the restrictions are not met. 19878 19879 Write the generated insn as a `parallel' with elements being a 19880 `set' of one register from the appropriate memory location (you may 19881 also need `use' or `clobber' elements). Use a `match_parallel' 19882 (*note RTL Template::) to recognize the insn. See `rs6000.md' for 19883 examples of the use of this insn pattern. 19884 19885 `store_multiple' 19886 Similar to `load_multiple', but store several consecutive registers 19887 into consecutive memory locations. Operand 0 is the first of the 19888 consecutive memory locations, operand 1 is the first register, and 19889 operand 2 is a constant: the number of consecutive registers. 19890 19891 `vec_setM' 19892 Set given field in the vector value. Operand 0 is the vector to 19893 modify, operand 1 is new value of field and operand 2 specify the 19894 field index. 19895 19896 `vec_extractM' 19897 Extract given field from the vector value. Operand 1 is the 19898 vector, operand 2 specify field index and operand 0 place to store 19899 value into. 19900 19901 `vec_extract_evenM' 19902 Extract even elements from the input vectors (operand 1 and 19903 operand 2). The even elements of operand 2 are concatenated to 19904 the even elements of operand 1 in their original order. The result 19905 is stored in operand 0. The output and input vectors should have 19906 the same modes. 19907 19908 `vec_extract_oddM' 19909 Extract odd elements from the input vectors (operand 1 and operand 19910 2). The odd elements of operand 2 are concatenated to the odd 19911 elements of operand 1 in their original order. The result is 19912 stored in operand 0. The output and input vectors should have the 19913 same modes. 19914 19915 `vec_interleave_highM' 19916 Merge high elements of the two input vectors into the output 19917 vector. The output and input vectors should have the same modes 19918 (`N' elements). The high `N/2' elements of the first input vector 19919 are interleaved with the high `N/2' elements of the second input 19920 vector. 19921 19922 `vec_interleave_lowM' 19923 Merge low elements of the two input vectors into the output 19924 vector. The output and input vectors should have the same modes 19925 (`N' elements). The low `N/2' elements of the first input vector 19926 are interleaved with the low `N/2' elements of the second input 19927 vector. 19928 19929 `vec_initM' 19930 Initialize the vector to given values. Operand 0 is the vector to 19931 initialize and operand 1 is parallel containing values for 19932 individual fields. 19933 19934 `pushM1' 19935 Output a push instruction. Operand 0 is value to push. Used only 19936 when `PUSH_ROUNDING' is defined. For historical reason, this 19937 pattern may be missing and in such case an `mov' expander is used 19938 instead, with a `MEM' expression forming the push operation. The 19939 `mov' expander method is deprecated. 19940 19941 `addM3' 19942 Add operand 2 and operand 1, storing the result in operand 0. All 19943 operands must have mode M. This can be used even on two-address 19944 machines, by means of constraints requiring operands 1 and 0 to be 19945 the same location. 19946 19947 `ssaddM3', `usaddM3' 19948 19949 `subM3', `sssubM3', `ussubM3' 19950 19951 `mulM3', `ssmulM3', `usmulM3' 19952 `divM3', `ssdivM3' 19953 `udivM3', `usdivM3' 19954 `modM3', `umodM3' 19955 `uminM3', `umaxM3' 19956 `andM3', `iorM3', `xorM3' 19957 Similar, for other arithmetic operations. 19958 19959 `sminM3', `smaxM3' 19960 Signed minimum and maximum operations. When used with floating 19961 point, if both operands are zeros, or if either operand is `NaN', 19962 then it is unspecified which of the two operands is returned as 19963 the result. 19964 19965 `reduc_smin_M', `reduc_smax_M' 19966 Find the signed minimum/maximum of the elements of a vector. The 19967 vector is operand 1, and the scalar result is stored in the least 19968 significant bits of operand 0 (also a vector). The output and 19969 input vector should have the same modes. 19970 19971 `reduc_umin_M', `reduc_umax_M' 19972 Find the unsigned minimum/maximum of the elements of a vector. The 19973 vector is operand 1, and the scalar result is stored in the least 19974 significant bits of operand 0 (also a vector). The output and 19975 input vector should have the same modes. 19976 19977 `reduc_splus_M' 19978 Compute the sum of the signed elements of a vector. The vector is 19979 operand 1, and the scalar result is stored in the least 19980 significant bits of operand 0 (also a vector). The output and 19981 input vector should have the same modes. 19982 19983 `reduc_uplus_M' 19984 Compute the sum of the unsigned elements of a vector. The vector 19985 is operand 1, and the scalar result is stored in the least 19986 significant bits of operand 0 (also a vector). The output and 19987 input vector should have the same modes. 19988 19989 `sdot_prodM' 19990 19991 `udot_prodM' 19992 Compute the sum of the products of two signed/unsigned elements. 19993 Operand 1 and operand 2 are of the same mode. Their product, which 19994 is of a wider mode, is computed and added to operand 3. Operand 3 19995 is of a mode equal or wider than the mode of the product. The 19996 result is placed in operand 0, which is of the same mode as 19997 operand 3. 19998 19999 `ssum_widenM3' 20000 20001 `usum_widenM3' 20002 Operands 0 and 2 are of the same mode, which is wider than the 20003 mode of operand 1. Add operand 1 to operand 2 and place the 20004 widened result in operand 0. (This is used express accumulation of 20005 elements into an accumulator of a wider mode.) 20006 20007 `vec_shl_M', `vec_shr_M' 20008 Whole vector left/right shift in bits. Operand 1 is a vector to 20009 be shifted. Operand 2 is an integer shift amount in bits. 20010 Operand 0 is where the resulting shifted vector is stored. The 20011 output and input vectors should have the same modes. 20012 20013 `vec_pack_trunc_M' 20014 Narrow (demote) and merge the elements of two vectors. Operands 1 20015 and 2 are vectors of the same mode having N integral or floating 20016 point elements of size S. Operand 0 is the resulting vector in 20017 which 2*N elements of size N/2 are concatenated after narrowing 20018 them down using truncation. 20019 20020 `vec_pack_ssat_M', `vec_pack_usat_M' 20021 Narrow (demote) and merge the elements of two vectors. Operands 1 20022 and 2 are vectors of the same mode having N integral elements of 20023 size S. Operand 0 is the resulting vector in which the elements 20024 of the two input vectors are concatenated after narrowing them 20025 down using signed/unsigned saturating arithmetic. 20026 20027 `vec_pack_sfix_trunc_M', `vec_pack_ufix_trunc_M' 20028 Narrow, convert to signed/unsigned integral type and merge the 20029 elements of two vectors. Operands 1 and 2 are vectors of the same 20030 mode having N floating point elements of size S. Operand 0 is the 20031 resulting vector in which 2*N elements of size N/2 are 20032 concatenated. 20033 20034 `vec_unpacks_hi_M', `vec_unpacks_lo_M' 20035 Extract and widen (promote) the high/low part of a vector of signed 20036 integral or floating point elements. The input vector (operand 1) 20037 has N elements of size S. Widen (promote) the high/low elements 20038 of the vector using signed or floating point extension and place 20039 the resulting N/2 values of size 2*S in the output vector (operand 20040 0). 20041 20042 `vec_unpacku_hi_M', `vec_unpacku_lo_M' 20043 Extract and widen (promote) the high/low part of a vector of 20044 unsigned integral elements. The input vector (operand 1) has N 20045 elements of size S. Widen (promote) the high/low elements of the 20046 vector using zero extension and place the resulting N/2 values of 20047 size 2*S in the output vector (operand 0). 20048 20049 `vec_unpacks_float_hi_M', `vec_unpacks_float_lo_M' 20050 `vec_unpacku_float_hi_M', `vec_unpacku_float_lo_M' 20051 Extract, convert to floating point type and widen the high/low 20052 part of a vector of signed/unsigned integral elements. The input 20053 vector (operand 1) has N elements of size S. Convert the high/low 20054 elements of the vector using floating point conversion and place 20055 the resulting N/2 values of size 2*S in the output vector (operand 20056 0). 20057 20058 `vec_widen_umult_hi_M', `vec_widen_umult_lo_M' 20059 `vec_widen_smult_hi_M', `vec_widen_smult_lo_M' 20060 Signed/Unsigned widening multiplication. The two inputs (operands 20061 1 and 2) are vectors with N signed/unsigned elements of size S. 20062 Multiply the high/low elements of the two vectors, and put the N/2 20063 products of size 2*S in the output vector (operand 0). 20064 20065 `mulhisi3' 20066 Multiply operands 1 and 2, which have mode `HImode', and store a 20067 `SImode' product in operand 0. 20068 20069 `mulqihi3', `mulsidi3' 20070 Similar widening-multiplication instructions of other widths. 20071 20072 `umulqihi3', `umulhisi3', `umulsidi3' 20073 Similar widening-multiplication instructions that do unsigned 20074 multiplication. 20075 20076 `usmulqihi3', `usmulhisi3', `usmulsidi3' 20077 Similar widening-multiplication instructions that interpret the 20078 first operand as unsigned and the second operand as signed, then 20079 do a signed multiplication. 20080 20081 `smulM3_highpart' 20082 Perform a signed multiplication of operands 1 and 2, which have 20083 mode M, and store the most significant half of the product in 20084 operand 0. The least significant half of the product is discarded. 20085 20086 `umulM3_highpart' 20087 Similar, but the multiplication is unsigned. 20088 20089 `maddMN4' 20090 Multiply operands 1 and 2, sign-extend them to mode N, add operand 20091 3, and store the result in operand 0. Operands 1 and 2 have mode 20092 M and operands 0 and 3 have mode N. Both modes must be integer or 20093 fixed-point modes and N must be twice the size of M. 20094 20095 In other words, `maddMN4' is like `mulMN3' except that it also 20096 adds operand 3. 20097 20098 These instructions are not allowed to `FAIL'. 20099 20100 `umaddMN4' 20101 Like `maddMN4', but zero-extend the multiplication operands 20102 instead of sign-extending them. 20103 20104 `ssmaddMN4' 20105 Like `maddMN4', but all involved operations must be 20106 signed-saturating. 20107 20108 `usmaddMN4' 20109 Like `umaddMN4', but all involved operations must be 20110 unsigned-saturating. 20111 20112 `msubMN4' 20113 Multiply operands 1 and 2, sign-extend them to mode N, subtract the 20114 result from operand 3, and store the result in operand 0. 20115 Operands 1 and 2 have mode M and operands 0 and 3 have mode N. 20116 Both modes must be integer or fixed-point modes and N must be twice 20117 the size of M. 20118 20119 In other words, `msubMN4' is like `mulMN3' except that it also 20120 subtracts the result from operand 3. 20121 20122 These instructions are not allowed to `FAIL'. 20123 20124 `umsubMN4' 20125 Like `msubMN4', but zero-extend the multiplication operands 20126 instead of sign-extending them. 20127 20128 `ssmsubMN4' 20129 Like `msubMN4', but all involved operations must be 20130 signed-saturating. 20131 20132 `usmsubMN4' 20133 Like `umsubMN4', but all involved operations must be 20134 unsigned-saturating. 20135 20136 `divmodM4' 20137 Signed division that produces both a quotient and a remainder. 20138 Operand 1 is divided by operand 2 to produce a quotient stored in 20139 operand 0 and a remainder stored in operand 3. 20140 20141 For machines with an instruction that produces both a quotient and 20142 a remainder, provide a pattern for `divmodM4' but do not provide 20143 patterns for `divM3' and `modM3'. This allows optimization in the 20144 relatively common case when both the quotient and remainder are 20145 computed. 20146 20147 If an instruction that just produces a quotient or just a remainder 20148 exists and is more efficient than the instruction that produces 20149 both, write the output routine of `divmodM4' to call 20150 `find_reg_note' and look for a `REG_UNUSED' note on the quotient 20151 or remainder and generate the appropriate instruction. 20152 20153 `udivmodM4' 20154 Similar, but does unsigned division. 20155 20156 `ashlM3', `ssashlM3', `usashlM3' 20157 Arithmetic-shift operand 1 left by a number of bits specified by 20158 operand 2, and store the result in operand 0. Here M is the mode 20159 of operand 0 and operand 1; operand 2's mode is specified by the 20160 instruction pattern, and the compiler will convert the operand to 20161 that mode before generating the instruction. The meaning of 20162 out-of-range shift counts can optionally be specified by 20163 `TARGET_SHIFT_TRUNCATION_MASK'. *Note 20164 TARGET_SHIFT_TRUNCATION_MASK::. Operand 2 is always a scalar type. 20165 20166 `ashrM3', `lshrM3', `rotlM3', `rotrM3' 20167 Other shift and rotate instructions, analogous to the `ashlM3' 20168 instructions. Operand 2 is always a scalar type. 20169 20170 `vashlM3', `vashrM3', `vlshrM3', `vrotlM3', `vrotrM3' 20171 Vector shift and rotate instructions that take vectors as operand 2 20172 instead of a scalar type. 20173 20174 `negM2', `ssnegM2', `usnegM2' 20175 Negate operand 1 and store the result in operand 0. 20176 20177 `absM2' 20178 Store the absolute value of operand 1 into operand 0. 20179 20180 `sqrtM2' 20181 Store the square root of operand 1 into operand 0. 20182 20183 The `sqrt' built-in function of C always uses the mode which 20184 corresponds to the C data type `double' and the `sqrtf' built-in 20185 function uses the mode which corresponds to the C data type 20186 `float'. 20187 20188 `fmodM3' 20189 Store the remainder of dividing operand 1 by operand 2 into 20190 operand 0, rounded towards zero to an integer. 20191 20192 The `fmod' built-in function of C always uses the mode which 20193 corresponds to the C data type `double' and the `fmodf' built-in 20194 function uses the mode which corresponds to the C data type 20195 `float'. 20196 20197 `remainderM3' 20198 Store the remainder of dividing operand 1 by operand 2 into 20199 operand 0, rounded to the nearest integer. 20200 20201 The `remainder' built-in function of C always uses the mode which 20202 corresponds to the C data type `double' and the `remainderf' 20203 built-in function uses the mode which corresponds to the C data 20204 type `float'. 20205 20206 `cosM2' 20207 Store the cosine of operand 1 into operand 0. 20208 20209 The `cos' built-in function of C always uses the mode which 20210 corresponds to the C data type `double' and the `cosf' built-in 20211 function uses the mode which corresponds to the C data type 20212 `float'. 20213 20214 `sinM2' 20215 Store the sine of operand 1 into operand 0. 20216 20217 The `sin' built-in function of C always uses the mode which 20218 corresponds to the C data type `double' and the `sinf' built-in 20219 function uses the mode which corresponds to the C data type 20220 `float'. 20221 20222 `expM2' 20223 Store the exponential of operand 1 into operand 0. 20224 20225 The `exp' built-in function of C always uses the mode which 20226 corresponds to the C data type `double' and the `expf' built-in 20227 function uses the mode which corresponds to the C data type 20228 `float'. 20229 20230 `logM2' 20231 Store the natural logarithm of operand 1 into operand 0. 20232 20233 The `log' built-in function of C always uses the mode which 20234 corresponds to the C data type `double' and the `logf' built-in 20235 function uses the mode which corresponds to the C data type 20236 `float'. 20237 20238 `powM3' 20239 Store the value of operand 1 raised to the exponent operand 2 into 20240 operand 0. 20241 20242 The `pow' built-in function of C always uses the mode which 20243 corresponds to the C data type `double' and the `powf' built-in 20244 function uses the mode which corresponds to the C data type 20245 `float'. 20246 20247 `atan2M3' 20248 Store the arc tangent (inverse tangent) of operand 1 divided by 20249 operand 2 into operand 0, using the signs of both arguments to 20250 determine the quadrant of the result. 20251 20252 The `atan2' built-in function of C always uses the mode which 20253 corresponds to the C data type `double' and the `atan2f' built-in 20254 function uses the mode which corresponds to the C data type 20255 `float'. 20256 20257 `floorM2' 20258 Store the largest integral value not greater than argument. 20259 20260 The `floor' built-in function of C always uses the mode which 20261 corresponds to the C data type `double' and the `floorf' built-in 20262 function uses the mode which corresponds to the C data type 20263 `float'. 20264 20265 `btruncM2' 20266 Store the argument rounded to integer towards zero. 20267 20268 The `trunc' built-in function of C always uses the mode which 20269 corresponds to the C data type `double' and the `truncf' built-in 20270 function uses the mode which corresponds to the C data type 20271 `float'. 20272 20273 `roundM2' 20274 Store the argument rounded to integer away from zero. 20275 20276 The `round' built-in function of C always uses the mode which 20277 corresponds to the C data type `double' and the `roundf' built-in 20278 function uses the mode which corresponds to the C data type 20279 `float'. 20280 20281 `ceilM2' 20282 Store the argument rounded to integer away from zero. 20283 20284 The `ceil' built-in function of C always uses the mode which 20285 corresponds to the C data type `double' and the `ceilf' built-in 20286 function uses the mode which corresponds to the C data type 20287 `float'. 20288 20289 `nearbyintM2' 20290 Store the argument rounded according to the default rounding mode 20291 20292 The `nearbyint' built-in function of C always uses the mode which 20293 corresponds to the C data type `double' and the `nearbyintf' 20294 built-in function uses the mode which corresponds to the C data 20295 type `float'. 20296 20297 `rintM2' 20298 Store the argument rounded according to the default rounding mode 20299 and raise the inexact exception when the result differs in value 20300 from the argument 20301 20302 The `rint' built-in function of C always uses the mode which 20303 corresponds to the C data type `double' and the `rintf' built-in 20304 function uses the mode which corresponds to the C data type 20305 `float'. 20306 20307 `lrintMN2' 20308 Convert operand 1 (valid for floating point mode M) to fixed point 20309 mode N as a signed number according to the current rounding mode 20310 and store in operand 0 (which has mode N). 20311 20312 `lroundM2' 20313 Convert operand 1 (valid for floating point mode M) to fixed point 20314 mode N as a signed number rounding to nearest and away from zero 20315 and store in operand 0 (which has mode N). 20316 20317 `lfloorM2' 20318 Convert operand 1 (valid for floating point mode M) to fixed point 20319 mode N as a signed number rounding down and store in operand 0 20320 (which has mode N). 20321 20322 `lceilM2' 20323 Convert operand 1 (valid for floating point mode M) to fixed point 20324 mode N as a signed number rounding up and store in operand 0 20325 (which has mode N). 20326 20327 `copysignM3' 20328 Store a value with the magnitude of operand 1 and the sign of 20329 operand 2 into operand 0. 20330 20331 The `copysign' built-in function of C always uses the mode which 20332 corresponds to the C data type `double' and the `copysignf' 20333 built-in function uses the mode which corresponds to the C data 20334 type `float'. 20335 20336 `ffsM2' 20337 Store into operand 0 one plus the index of the least significant 20338 1-bit of operand 1. If operand 1 is zero, store zero. M is the 20339 mode of operand 0; operand 1's mode is specified by the instruction 20340 pattern, and the compiler will convert the operand to that mode 20341 before generating the instruction. 20342 20343 The `ffs' built-in function of C always uses the mode which 20344 corresponds to the C data type `int'. 20345 20346 `clzM2' 20347 Store into operand 0 the number of leading 0-bits in X, starting 20348 at the most significant bit position. If X is 0, the 20349 `CLZ_DEFINED_VALUE_AT_ZERO' (*note Misc::) macro defines if the 20350 result is undefined or has a useful value. M is the mode of 20351 operand 0; operand 1's mode is specified by the instruction 20352 pattern, and the compiler will convert the operand to that mode 20353 before generating the instruction. 20354 20355 `ctzM2' 20356 Store into operand 0 the number of trailing 0-bits in X, starting 20357 at the least significant bit position. If X is 0, the 20358 `CTZ_DEFINED_VALUE_AT_ZERO' (*note Misc::) macro defines if the 20359 result is undefined or has a useful value. M is the mode of 20360 operand 0; operand 1's mode is specified by the instruction 20361 pattern, and the compiler will convert the operand to that mode 20362 before generating the instruction. 20363 20364 `popcountM2' 20365 Store into operand 0 the number of 1-bits in X. M is the mode of 20366 operand 0; operand 1's mode is specified by the instruction 20367 pattern, and the compiler will convert the operand to that mode 20368 before generating the instruction. 20369 20370 `parityM2' 20371 Store into operand 0 the parity of X, i.e. the number of 1-bits in 20372 X modulo 2. M is the mode of operand 0; operand 1's mode is 20373 specified by the instruction pattern, and the compiler will convert 20374 the operand to that mode before generating the instruction. 20375 20376 `one_cmplM2' 20377 Store the bitwise-complement of operand 1 into operand 0. 20378 20379 `cmpM' 20380 Compare operand 0 and operand 1, and set the condition codes. The 20381 RTL pattern should look like this: 20382 20383 (set (cc0) (compare (match_operand:M 0 ...) 20384 (match_operand:M 1 ...))) 20385 20386 `tstM' 20387 Compare operand 0 against zero, and set the condition codes. The 20388 RTL pattern should look like this: 20389 20390 (set (cc0) (match_operand:M 0 ...)) 20391 20392 `tstM' patterns should not be defined for machines that do not use 20393 `(cc0)'. Doing so would confuse the optimizer since it would no 20394 longer be clear which `set' operations were comparisons. The 20395 `cmpM' patterns should be used instead. 20396 20397 `movmemM' 20398 Block move instruction. The destination and source blocks of 20399 memory are the first two operands, and both are `mem:BLK's with an 20400 address in mode `Pmode'. 20401 20402 The number of bytes to move is the third operand, in mode M. 20403 Usually, you specify `word_mode' for M. However, if you can 20404 generate better code knowing the range of valid lengths is smaller 20405 than those representable in a full word, you should provide a 20406 pattern with a mode corresponding to the range of values you can 20407 handle efficiently (e.g., `QImode' for values in the range 0-127; 20408 note we avoid numbers that appear negative) and also a pattern 20409 with `word_mode'. 20410 20411 The fourth operand is the known shared alignment of the source and 20412 destination, in the form of a `const_int' rtx. Thus, if the 20413 compiler knows that both source and destination are word-aligned, 20414 it may provide the value 4 for this operand. 20415 20416 Optional operands 5 and 6 specify expected alignment and size of 20417 block respectively. The expected alignment differs from alignment 20418 in operand 4 in a way that the blocks are not required to be 20419 aligned according to it in all cases. This expected alignment is 20420 also in bytes, just like operand 4. Expected size, when unknown, 20421 is set to `(const_int -1)'. 20422 20423 Descriptions of multiple `movmemM' patterns can only be beneficial 20424 if the patterns for smaller modes have fewer restrictions on their 20425 first, second and fourth operands. Note that the mode M in 20426 `movmemM' does not impose any restriction on the mode of 20427 individually moved data units in the block. 20428 20429 These patterns need not give special consideration to the 20430 possibility that the source and destination strings might overlap. 20431 20432 `movstr' 20433 String copy instruction, with `stpcpy' semantics. Operand 0 is an 20434 output operand in mode `Pmode'. The addresses of the destination 20435 and source strings are operands 1 and 2, and both are `mem:BLK's 20436 with addresses in mode `Pmode'. The execution of the expansion of 20437 this pattern should store in operand 0 the address in which the 20438 `NUL' terminator was stored in the destination string. 20439 20440 `setmemM' 20441 Block set instruction. The destination string is the first 20442 operand, given as a `mem:BLK' whose address is in mode `Pmode'. 20443 The number of bytes to set is the second operand, in mode M. The 20444 value to initialize the memory with is the third operand. Targets 20445 that only support the clearing of memory should reject any value 20446 that is not the constant 0. See `movmemM' for a discussion of the 20447 choice of mode. 20448 20449 The fourth operand is the known alignment of the destination, in 20450 the form of a `const_int' rtx. Thus, if the compiler knows that 20451 the destination is word-aligned, it may provide the value 4 for 20452 this operand. 20453 20454 Optional operands 5 and 6 specify expected alignment and size of 20455 block respectively. The expected alignment differs from alignment 20456 in operand 4 in a way that the blocks are not required to be 20457 aligned according to it in all cases. This expected alignment is 20458 also in bytes, just like operand 4. Expected size, when unknown, 20459 is set to `(const_int -1)'. 20460 20461 The use for multiple `setmemM' is as for `movmemM'. 20462 20463 `cmpstrnM' 20464 String compare instruction, with five operands. Operand 0 is the 20465 output; it has mode M. The remaining four operands are like the 20466 operands of `movmemM'. The two memory blocks specified are 20467 compared byte by byte in lexicographic order starting at the 20468 beginning of each string. The instruction is not allowed to 20469 prefetch more than one byte at a time since either string may end 20470 in the first byte and reading past that may access an invalid page 20471 or segment and cause a fault. The effect of the instruction is to 20472 store a value in operand 0 whose sign indicates the result of the 20473 comparison. 20474 20475 `cmpstrM' 20476 String compare instruction, without known maximum length. Operand 20477 0 is the output; it has mode M. The second and third operand are 20478 the blocks of memory to be compared; both are `mem:BLK' with an 20479 address in mode `Pmode'. 20480 20481 The fourth operand is the known shared alignment of the source and 20482 destination, in the form of a `const_int' rtx. Thus, if the 20483 compiler knows that both source and destination are word-aligned, 20484 it may provide the value 4 for this operand. 20485 20486 The two memory blocks specified are compared byte by byte in 20487 lexicographic order starting at the beginning of each string. The 20488 instruction is not allowed to prefetch more than one byte at a 20489 time since either string may end in the first byte and reading 20490 past that may access an invalid page or segment and cause a fault. 20491 The effect of the instruction is to store a value in operand 0 20492 whose sign indicates the result of the comparison. 20493 20494 `cmpmemM' 20495 Block compare instruction, with five operands like the operands of 20496 `cmpstrM'. The two memory blocks specified are compared byte by 20497 byte in lexicographic order starting at the beginning of each 20498 block. Unlike `cmpstrM' the instruction can prefetch any bytes in 20499 the two memory blocks. The effect of the instruction is to store 20500 a value in operand 0 whose sign indicates the result of the 20501 comparison. 20502 20503 `strlenM' 20504 Compute the length of a string, with three operands. Operand 0 is 20505 the result (of mode M), operand 1 is a `mem' referring to the 20506 first character of the string, operand 2 is the character to 20507 search for (normally zero), and operand 3 is a constant describing 20508 the known alignment of the beginning of the string. 20509 20510 `floatMN2' 20511 Convert signed integer operand 1 (valid for fixed point mode M) to 20512 floating point mode N and store in operand 0 (which has mode N). 20513 20514 `floatunsMN2' 20515 Convert unsigned integer operand 1 (valid for fixed point mode M) 20516 to floating point mode N and store in operand 0 (which has mode N). 20517 20518 `fixMN2' 20519 Convert operand 1 (valid for floating point mode M) to fixed point 20520 mode N as a signed number and store in operand 0 (which has mode 20521 N). This instruction's result is defined only when the value of 20522 operand 1 is an integer. 20523 20524 If the machine description defines this pattern, it also needs to 20525 define the `ftrunc' pattern. 20526 20527 `fixunsMN2' 20528 Convert operand 1 (valid for floating point mode M) to fixed point 20529 mode N as an unsigned number and store in operand 0 (which has 20530 mode N). This instruction's result is defined only when the value 20531 of operand 1 is an integer. 20532 20533 `ftruncM2' 20534 Convert operand 1 (valid for floating point mode M) to an integer 20535 value, still represented in floating point mode M, and store it in 20536 operand 0 (valid for floating point mode M). 20537 20538 `fix_truncMN2' 20539 Like `fixMN2' but works for any floating point value of mode M by 20540 converting the value to an integer. 20541 20542 `fixuns_truncMN2' 20543 Like `fixunsMN2' but works for any floating point value of mode M 20544 by converting the value to an integer. 20545 20546 `truncMN2' 20547 Truncate operand 1 (valid for mode M) to mode N and store in 20548 operand 0 (which has mode N). Both modes must be fixed point or 20549 both floating point. 20550 20551 `extendMN2' 20552 Sign-extend operand 1 (valid for mode M) to mode N and store in 20553 operand 0 (which has mode N). Both modes must be fixed point or 20554 both floating point. 20555 20556 `zero_extendMN2' 20557 Zero-extend operand 1 (valid for mode M) to mode N and store in 20558 operand 0 (which has mode N). Both modes must be fixed point. 20559 20560 `fractMN2' 20561 Convert operand 1 of mode M to mode N and store in operand 0 20562 (which has mode N). Mode M and mode N could be fixed-point to 20563 fixed-point, signed integer to fixed-point, fixed-point to signed 20564 integer, floating-point to fixed-point, or fixed-point to 20565 floating-point. When overflows or underflows happen, the results 20566 are undefined. 20567 20568 `satfractMN2' 20569 Convert operand 1 of mode M to mode N and store in operand 0 20570 (which has mode N). Mode M and mode N could be fixed-point to 20571 fixed-point, signed integer to fixed-point, or floating-point to 20572 fixed-point. When overflows or underflows happen, the instruction 20573 saturates the results to the maximum or the minimum. 20574 20575 `fractunsMN2' 20576 Convert operand 1 of mode M to mode N and store in operand 0 20577 (which has mode N). Mode M and mode N could be unsigned integer 20578 to fixed-point, or fixed-point to unsigned integer. When 20579 overflows or underflows happen, the results are undefined. 20580 20581 `satfractunsMN2' 20582 Convert unsigned integer operand 1 of mode M to fixed-point mode N 20583 and store in operand 0 (which has mode N). When overflows or 20584 underflows happen, the instruction saturates the results to the 20585 maximum or the minimum. 20586 20587 `extv' 20588 Extract a bit-field from operand 1 (a register or memory operand), 20589 where operand 2 specifies the width in bits and operand 3 the 20590 starting bit, and store it in operand 0. Operand 0 must have mode 20591 `word_mode'. Operand 1 may have mode `byte_mode' or `word_mode'; 20592 often `word_mode' is allowed only for registers. Operands 2 and 3 20593 must be valid for `word_mode'. 20594 20595 The RTL generation pass generates this instruction only with 20596 constants for operands 2 and 3 and the constant is never zero for 20597 operand 2. 20598 20599 The bit-field value is sign-extended to a full word integer before 20600 it is stored in operand 0. 20601 20602 `extzv' 20603 Like `extv' except that the bit-field value is zero-extended. 20604 20605 `insv' 20606 Store operand 3 (which must be valid for `word_mode') into a 20607 bit-field in operand 0, where operand 1 specifies the width in 20608 bits and operand 2 the starting bit. Operand 0 may have mode 20609 `byte_mode' or `word_mode'; often `word_mode' is allowed only for 20610 registers. Operands 1 and 2 must be valid for `word_mode'. 20611 20612 The RTL generation pass generates this instruction only with 20613 constants for operands 1 and 2 and the constant is never zero for 20614 operand 1. 20615 20616 `movMODEcc' 20617 Conditionally move operand 2 or operand 3 into operand 0 according 20618 to the comparison in operand 1. If the comparison is true, 20619 operand 2 is moved into operand 0, otherwise operand 3 is moved. 20620 20621 The mode of the operands being compared need not be the same as 20622 the operands being moved. Some machines, sparc64 for example, 20623 have instructions that conditionally move an integer value based 20624 on the floating point condition codes and vice versa. 20625 20626 If the machine does not have conditional move instructions, do not 20627 define these patterns. 20628 20629 `addMODEcc' 20630 Similar to `movMODEcc' but for conditional addition. Conditionally 20631 move operand 2 or (operands 2 + operand 3) into operand 0 20632 according to the comparison in operand 1. If the comparison is 20633 true, operand 2 is moved into operand 0, otherwise (operand 2 + 20634 operand 3) is moved. 20635 20636 `sCOND' 20637 Store zero or nonzero in the operand according to the condition 20638 codes. Value stored is nonzero iff the condition COND is true. 20639 COND is the name of a comparison operation expression code, such 20640 as `eq', `lt' or `leu'. 20641 20642 You specify the mode that the operand must have when you write the 20643 `match_operand' expression. The compiler automatically sees which 20644 mode you have used and supplies an operand of that mode. 20645 20646 The value stored for a true condition must have 1 as its low bit, 20647 or else must be negative. Otherwise the instruction is not 20648 suitable and you should omit it from the machine description. You 20649 describe to the compiler exactly which value is stored by defining 20650 the macro `STORE_FLAG_VALUE' (*note Misc::). If a description 20651 cannot be found that can be used for all the `sCOND' patterns, you 20652 should omit those operations from the machine description. 20653 20654 These operations may fail, but should do so only in relatively 20655 uncommon cases; if they would fail for common cases involving 20656 integer comparisons, it is best to omit these patterns. 20657 20658 If these operations are omitted, the compiler will usually 20659 generate code that copies the constant one to the target and 20660 branches around an assignment of zero to the target. If this code 20661 is more efficient than the potential instructions used for the 20662 `sCOND' pattern followed by those required to convert the result 20663 into a 1 or a zero in `SImode', you should omit the `sCOND' 20664 operations from the machine description. 20665 20666 `bCOND' 20667 Conditional branch instruction. Operand 0 is a `label_ref' that 20668 refers to the label to jump to. Jump if the condition codes meet 20669 condition COND. 20670 20671 Some machines do not follow the model assumed here where a 20672 comparison instruction is followed by a conditional branch 20673 instruction. In that case, the `cmpM' (and `tstM') patterns should 20674 simply store the operands away and generate all the required insns 20675 in a `define_expand' (*note Expander Definitions::) for the 20676 conditional branch operations. All calls to expand `bCOND' 20677 patterns are immediately preceded by calls to expand either a 20678 `cmpM' pattern or a `tstM' pattern. 20679 20680 Machines that use a pseudo register for the condition code value, 20681 or where the mode used for the comparison depends on the condition 20682 being tested, should also use the above mechanism. *Note Jump 20683 Patterns::. 20684 20685 The above discussion also applies to the `movMODEcc' and `sCOND' 20686 patterns. 20687 20688 `cbranchMODE4' 20689 Conditional branch instruction combined with a compare instruction. 20690 Operand 0 is a comparison operator. Operand 1 and operand 2 are 20691 the first and second operands of the comparison, respectively. 20692 Operand 3 is a `label_ref' that refers to the label to jump to. 20693 20694 `jump' 20695 A jump inside a function; an unconditional branch. Operand 0 is 20696 the `label_ref' of the label to jump to. This pattern name is 20697 mandatory on all machines. 20698 20699 `call' 20700 Subroutine call instruction returning no value. Operand 0 is the 20701 function to call; operand 1 is the number of bytes of arguments 20702 pushed as a `const_int'; operand 2 is the number of registers used 20703 as operands. 20704 20705 On most machines, operand 2 is not actually stored into the RTL 20706 pattern. It is supplied for the sake of some RISC machines which 20707 need to put this information into the assembler code; they can put 20708 it in the RTL instead of operand 1. 20709 20710 Operand 0 should be a `mem' RTX whose address is the address of the 20711 function. Note, however, that this address can be a `symbol_ref' 20712 expression even if it would not be a legitimate memory address on 20713 the target machine. If it is also not a valid argument for a call 20714 instruction, the pattern for this operation should be a 20715 `define_expand' (*note Expander Definitions::) that places the 20716 address into a register and uses that register in the call 20717 instruction. 20718 20719 `call_value' 20720 Subroutine call instruction returning a value. Operand 0 is the 20721 hard register in which the value is returned. There are three more 20722 operands, the same as the three operands of the `call' instruction 20723 (but with numbers increased by one). 20724 20725 Subroutines that return `BLKmode' objects use the `call' insn. 20726 20727 `call_pop', `call_value_pop' 20728 Similar to `call' and `call_value', except used if defined and if 20729 `RETURN_POPS_ARGS' is nonzero. They should emit a `parallel' that 20730 contains both the function call and a `set' to indicate the 20731 adjustment made to the frame pointer. 20732 20733 For machines where `RETURN_POPS_ARGS' can be nonzero, the use of 20734 these patterns increases the number of functions for which the 20735 frame pointer can be eliminated, if desired. 20736 20737 `untyped_call' 20738 Subroutine call instruction returning a value of any type. 20739 Operand 0 is the function to call; operand 1 is a memory location 20740 where the result of calling the function is to be stored; operand 20741 2 is a `parallel' expression where each element is a `set' 20742 expression that indicates the saving of a function return value 20743 into the result block. 20744 20745 This instruction pattern should be defined to support 20746 `__builtin_apply' on machines where special instructions are needed 20747 to call a subroutine with arbitrary arguments or to save the value 20748 returned. This instruction pattern is required on machines that 20749 have multiple registers that can hold a return value (i.e. 20750 `FUNCTION_VALUE_REGNO_P' is true for more than one register). 20751 20752 `return' 20753 Subroutine return instruction. This instruction pattern name 20754 should be defined only if a single instruction can do all the work 20755 of returning from a function. 20756 20757 Like the `movM' patterns, this pattern is also used after the RTL 20758 generation phase. In this case it is to support machines where 20759 multiple instructions are usually needed to return from a 20760 function, but some class of functions only requires one 20761 instruction to implement a return. Normally, the applicable 20762 functions are those which do not need to save any registers or 20763 allocate stack space. 20764 20765 For such machines, the condition specified in this pattern should 20766 only be true when `reload_completed' is nonzero and the function's 20767 epilogue would only be a single instruction. For machines with 20768 register windows, the routine `leaf_function_p' may be used to 20769 determine if a register window push is required. 20770 20771 Machines that have conditional return instructions should define 20772 patterns such as 20773 20774 (define_insn "" 20775 [(set (pc) 20776 (if_then_else (match_operator 20777 0 "comparison_operator" 20778 [(cc0) (const_int 0)]) 20779 (return) 20780 (pc)))] 20781 "CONDITION" 20782 "...") 20783 20784 where CONDITION would normally be the same condition specified on 20785 the named `return' pattern. 20786 20787 `untyped_return' 20788 Untyped subroutine return instruction. This instruction pattern 20789 should be defined to support `__builtin_return' on machines where 20790 special instructions are needed to return a value of any type. 20791 20792 Operand 0 is a memory location where the result of calling a 20793 function with `__builtin_apply' is stored; operand 1 is a 20794 `parallel' expression where each element is a `set' expression 20795 that indicates the restoring of a function return value from the 20796 result block. 20797 20798 `nop' 20799 No-op instruction. This instruction pattern name should always be 20800 defined to output a no-op in assembler code. `(const_int 0)' will 20801 do as an RTL pattern. 20802 20803 `indirect_jump' 20804 An instruction to jump to an address which is operand zero. This 20805 pattern name is mandatory on all machines. 20806 20807 `casesi' 20808 Instruction to jump through a dispatch table, including bounds 20809 checking. This instruction takes five operands: 20810 20811 1. The index to dispatch on, which has mode `SImode'. 20812 20813 2. The lower bound for indices in the table, an integer constant. 20814 20815 3. The total range of indices in the table--the largest index 20816 minus the smallest one (both inclusive). 20817 20818 4. A label that precedes the table itself. 20819 20820 5. A label to jump to if the index has a value outside the 20821 bounds. 20822 20823 The table is a `addr_vec' or `addr_diff_vec' inside of a 20824 `jump_insn'. The number of elements in the table is one plus the 20825 difference between the upper bound and the lower bound. 20826 20827 `tablejump' 20828 Instruction to jump to a variable address. This is a low-level 20829 capability which can be used to implement a dispatch table when 20830 there is no `casesi' pattern. 20831 20832 This pattern requires two operands: the address or offset, and a 20833 label which should immediately precede the jump table. If the 20834 macro `CASE_VECTOR_PC_RELATIVE' evaluates to a nonzero value then 20835 the first operand is an offset which counts from the address of 20836 the table; otherwise, it is an absolute address to jump to. In 20837 either case, the first operand has mode `Pmode'. 20838 20839 The `tablejump' insn is always the last insn before the jump table 20840 it uses. Its assembler code normally has no need to use the 20841 second operand, but you should incorporate it in the RTL pattern so 20842 that the jump optimizer will not delete the table as unreachable 20843 code. 20844 20845 `decrement_and_branch_until_zero' 20846 Conditional branch instruction that decrements a register and 20847 jumps if the register is nonzero. Operand 0 is the register to 20848 decrement and test; operand 1 is the label to jump to if the 20849 register is nonzero. *Note Looping Patterns::. 20850 20851 This optional instruction pattern is only used by the combiner, 20852 typically for loops reversed by the loop optimizer when strength 20853 reduction is enabled. 20854 20855 `doloop_end' 20856 Conditional branch instruction that decrements a register and 20857 jumps if the register is nonzero. This instruction takes five 20858 operands: Operand 0 is the register to decrement and test; operand 20859 1 is the number of loop iterations as a `const_int' or 20860 `const0_rtx' if this cannot be determined until run-time; operand 20861 2 is the actual or estimated maximum number of iterations as a 20862 `const_int'; operand 3 is the number of enclosed loops as a 20863 `const_int' (an innermost loop has a value of 1); operand 4 is the 20864 label to jump to if the register is nonzero. *Note Looping 20865 Patterns::. 20866 20867 This optional instruction pattern should be defined for machines 20868 with low-overhead looping instructions as the loop optimizer will 20869 try to modify suitable loops to utilize it. If nested 20870 low-overhead looping is not supported, use a `define_expand' 20871 (*note Expander Definitions::) and make the pattern fail if 20872 operand 3 is not `const1_rtx'. Similarly, if the actual or 20873 estimated maximum number of iterations is too large for this 20874 instruction, make it fail. 20875 20876 `doloop_begin' 20877 Companion instruction to `doloop_end' required for machines that 20878 need to perform some initialization, such as loading special 20879 registers used by a low-overhead looping instruction. If 20880 initialization insns do not always need to be emitted, use a 20881 `define_expand' (*note Expander Definitions::) and make it fail. 20882 20883 `canonicalize_funcptr_for_compare' 20884 Canonicalize the function pointer in operand 1 and store the result 20885 into operand 0. 20886 20887 Operand 0 is always a `reg' and has mode `Pmode'; operand 1 may be 20888 a `reg', `mem', `symbol_ref', `const_int', etc and also has mode 20889 `Pmode'. 20890 20891 Canonicalization of a function pointer usually involves computing 20892 the address of the function which would be called if the function 20893 pointer were used in an indirect call. 20894 20895 Only define this pattern if function pointers on the target machine 20896 can have different values but still call the same function when 20897 used in an indirect call. 20898 20899 `save_stack_block' 20900 `save_stack_function' 20901 `save_stack_nonlocal' 20902 `restore_stack_block' 20903 `restore_stack_function' 20904 `restore_stack_nonlocal' 20905 Most machines save and restore the stack pointer by copying it to 20906 or from an object of mode `Pmode'. Do not define these patterns on 20907 such machines. 20908 20909 Some machines require special handling for stack pointer saves and 20910 restores. On those machines, define the patterns corresponding to 20911 the non-standard cases by using a `define_expand' (*note Expander 20912 Definitions::) that produces the required insns. The three types 20913 of saves and restores are: 20914 20915 1. `save_stack_block' saves the stack pointer at the start of a 20916 block that allocates a variable-sized object, and 20917 `restore_stack_block' restores the stack pointer when the 20918 block is exited. 20919 20920 2. `save_stack_function' and `restore_stack_function' do a 20921 similar job for the outermost block of a function and are 20922 used when the function allocates variable-sized objects or 20923 calls `alloca'. Only the epilogue uses the restored stack 20924 pointer, allowing a simpler save or restore sequence on some 20925 machines. 20926 20927 3. `save_stack_nonlocal' is used in functions that contain labels 20928 branched to by nested functions. It saves the stack pointer 20929 in such a way that the inner function can use 20930 `restore_stack_nonlocal' to restore the stack pointer. The 20931 compiler generates code to restore the frame and argument 20932 pointer registers, but some machines require saving and 20933 restoring additional data such as register window information 20934 or stack backchains. Place insns in these patterns to save 20935 and restore any such required data. 20936 20937 When saving the stack pointer, operand 0 is the save area and 20938 operand 1 is the stack pointer. The mode used to allocate the 20939 save area defaults to `Pmode' but you can override that choice by 20940 defining the `STACK_SAVEAREA_MODE' macro (*note Storage Layout::). 20941 You must specify an integral mode, or `VOIDmode' if no save area 20942 is needed for a particular type of save (either because no save is 20943 needed or because a machine-specific save area can be used). 20944 Operand 0 is the stack pointer and operand 1 is the save area for 20945 restore operations. If `save_stack_block' is defined, operand 0 20946 must not be `VOIDmode' since these saves can be arbitrarily nested. 20947 20948 A save area is a `mem' that is at a constant offset from 20949 `virtual_stack_vars_rtx' when the stack pointer is saved for use by 20950 nonlocal gotos and a `reg' in the other two cases. 20951 20952 `allocate_stack' 20953 Subtract (or add if `STACK_GROWS_DOWNWARD' is undefined) operand 1 20954 from the stack pointer to create space for dynamically allocated 20955 data. 20956 20957 Store the resultant pointer to this space into operand 0. If you 20958 are allocating space from the main stack, do this by emitting a 20959 move insn to copy `virtual_stack_dynamic_rtx' to operand 0. If 20960 you are allocating the space elsewhere, generate code to copy the 20961 location of the space to operand 0. In the latter case, you must 20962 ensure this space gets freed when the corresponding space on the 20963 main stack is free. 20964 20965 Do not define this pattern if all that must be done is the 20966 subtraction. Some machines require other operations such as stack 20967 probes or maintaining the back chain. Define this pattern to emit 20968 those operations in addition to updating the stack pointer. 20969 20970 `check_stack' 20971 If stack checking cannot be done on your system by probing the 20972 stack with a load or store instruction (*note Stack Checking::), 20973 define this pattern to perform the needed check and signaling an 20974 error if the stack has overflowed. The single operand is the 20975 location in the stack furthest from the current stack pointer that 20976 you need to validate. Normally, on machines where this pattern is 20977 needed, you would obtain the stack limit from a global or 20978 thread-specific variable or register. 20979 20980 `nonlocal_goto' 20981 Emit code to generate a non-local goto, e.g., a jump from one 20982 function to a label in an outer function. This pattern has four 20983 arguments, each representing a value to be used in the jump. The 20984 first argument is to be loaded into the frame pointer, the second 20985 is the address to branch to (code to dispatch to the actual label), 20986 the third is the address of a location where the stack is saved, 20987 and the last is the address of the label, to be placed in the 20988 location for the incoming static chain. 20989 20990 On most machines you need not define this pattern, since GCC will 20991 already generate the correct code, which is to load the frame 20992 pointer and static chain, restore the stack (using the 20993 `restore_stack_nonlocal' pattern, if defined), and jump indirectly 20994 to the dispatcher. You need only define this pattern if this code 20995 will not work on your machine. 20996 20997 `nonlocal_goto_receiver' 20998 This pattern, if defined, contains code needed at the target of a 20999 nonlocal goto after the code already generated by GCC. You will 21000 not normally need to define this pattern. A typical reason why 21001 you might need this pattern is if some value, such as a pointer to 21002 a global table, must be restored when the frame pointer is 21003 restored. Note that a nonlocal goto only occurs within a 21004 unit-of-translation, so a global table pointer that is shared by 21005 all functions of a given module need not be restored. There are 21006 no arguments. 21007 21008 `exception_receiver' 21009 This pattern, if defined, contains code needed at the site of an 21010 exception handler that isn't needed at the site of a nonlocal 21011 goto. You will not normally need to define this pattern. A 21012 typical reason why you might need this pattern is if some value, 21013 such as a pointer to a global table, must be restored after 21014 control flow is branched to the handler of an exception. There 21015 are no arguments. 21016 21017 `builtin_setjmp_setup' 21018 This pattern, if defined, contains additional code needed to 21019 initialize the `jmp_buf'. You will not normally need to define 21020 this pattern. A typical reason why you might need this pattern is 21021 if some value, such as a pointer to a global table, must be 21022 restored. Though it is preferred that the pointer value be 21023 recalculated if possible (given the address of a label for 21024 instance). The single argument is a pointer to the `jmp_buf'. 21025 Note that the buffer is five words long and that the first three 21026 are normally used by the generic mechanism. 21027 21028 `builtin_setjmp_receiver' 21029 This pattern, if defined, contains code needed at the site of an 21030 built-in setjmp that isn't needed at the site of a nonlocal goto. 21031 You will not normally need to define this pattern. A typical 21032 reason why you might need this pattern is if some value, such as a 21033 pointer to a global table, must be restored. It takes one 21034 argument, which is the label to which builtin_longjmp transfered 21035 control; this pattern may be emitted at a small offset from that 21036 label. 21037 21038 `builtin_longjmp' 21039 This pattern, if defined, performs the entire action of the 21040 longjmp. You will not normally need to define this pattern unless 21041 you also define `builtin_setjmp_setup'. The single argument is a 21042 pointer to the `jmp_buf'. 21043 21044 `eh_return' 21045 This pattern, if defined, affects the way `__builtin_eh_return', 21046 and thence the call frame exception handling library routines, are 21047 built. It is intended to handle non-trivial actions needed along 21048 the abnormal return path. 21049 21050 The address of the exception handler to which the function should 21051 return is passed as operand to this pattern. It will normally 21052 need to copied by the pattern to some special register or memory 21053 location. If the pattern needs to determine the location of the 21054 target call frame in order to do so, it may use 21055 `EH_RETURN_STACKADJ_RTX', if defined; it will have already been 21056 assigned. 21057 21058 If this pattern is not defined, the default action will be to 21059 simply copy the return address to `EH_RETURN_HANDLER_RTX'. Either 21060 that macro or this pattern needs to be defined if call frame 21061 exception handling is to be used. 21062 21063 `prologue' 21064 This pattern, if defined, emits RTL for entry to a function. The 21065 function entry is responsible for setting up the stack frame, 21066 initializing the frame pointer register, saving callee saved 21067 registers, etc. 21068 21069 Using a prologue pattern is generally preferred over defining 21070 `TARGET_ASM_FUNCTION_PROLOGUE' to emit assembly code for the 21071 prologue. 21072 21073 The `prologue' pattern is particularly useful for targets which 21074 perform instruction scheduling. 21075 21076 `epilogue' 21077 This pattern emits RTL for exit from a function. The function 21078 exit is responsible for deallocating the stack frame, restoring 21079 callee saved registers and emitting the return instruction. 21080 21081 Using an epilogue pattern is generally preferred over defining 21082 `TARGET_ASM_FUNCTION_EPILOGUE' to emit assembly code for the 21083 epilogue. 21084 21085 The `epilogue' pattern is particularly useful for targets which 21086 perform instruction scheduling or which have delay slots for their 21087 return instruction. 21088 21089 `sibcall_epilogue' 21090 This pattern, if defined, emits RTL for exit from a function 21091 without the final branch back to the calling function. This 21092 pattern will be emitted before any sibling call (aka tail call) 21093 sites. 21094 21095 The `sibcall_epilogue' pattern must not clobber any arguments used 21096 for parameter passing or any stack slots for arguments passed to 21097 the current function. 21098 21099 `trap' 21100 This pattern, if defined, signals an error, typically by causing 21101 some kind of signal to be raised. Among other places, it is used 21102 by the Java front end to signal `invalid array index' exceptions. 21103 21104 `conditional_trap' 21105 Conditional trap instruction. Operand 0 is a piece of RTL which 21106 performs a comparison. Operand 1 is the trap code, an integer. 21107 21108 A typical `conditional_trap' pattern looks like 21109 21110 (define_insn "conditional_trap" 21111 [(trap_if (match_operator 0 "trap_operator" 21112 [(cc0) (const_int 0)]) 21113 (match_operand 1 "const_int_operand" "i"))] 21114 "" 21115 "...") 21116 21117 `prefetch' 21118 This pattern, if defined, emits code for a non-faulting data 21119 prefetch instruction. Operand 0 is the address of the memory to 21120 prefetch. Operand 1 is a constant 1 if the prefetch is preparing 21121 for a write to the memory address, or a constant 0 otherwise. 21122 Operand 2 is the expected degree of temporal locality of the data 21123 and is a value between 0 and 3, inclusive; 0 means that the data 21124 has no temporal locality, so it need not be left in the cache 21125 after the access; 3 means that the data has a high degree of 21126 temporal locality and should be left in all levels of cache 21127 possible; 1 and 2 mean, respectively, a low or moderate degree of 21128 temporal locality. 21129 21130 Targets that do not support write prefetches or locality hints can 21131 ignore the values of operands 1 and 2. 21132 21133 `blockage' 21134 This pattern defines a pseudo insn that prevents the instruction 21135 scheduler from moving instructions across the boundary defined by 21136 the blockage insn. Normally an UNSPEC_VOLATILE pattern. 21137 21138 `memory_barrier' 21139 If the target memory model is not fully synchronous, then this 21140 pattern should be defined to an instruction that orders both loads 21141 and stores before the instruction with respect to loads and stores 21142 after the instruction. This pattern has no operands. 21143 21144 `sync_compare_and_swapMODE' 21145 This pattern, if defined, emits code for an atomic compare-and-swap 21146 operation. Operand 1 is the memory on which the atomic operation 21147 is performed. Operand 2 is the "old" value to be compared against 21148 the current contents of the memory location. Operand 3 is the 21149 "new" value to store in the memory if the compare succeeds. 21150 Operand 0 is the result of the operation; it should contain the 21151 contents of the memory before the operation. If the compare 21152 succeeds, this should obviously be a copy of operand 2. 21153 21154 This pattern must show that both operand 0 and operand 1 are 21155 modified. 21156 21157 This pattern must issue any memory barrier instructions such that 21158 all memory operations before the atomic operation occur before the 21159 atomic operation and all memory operations after the atomic 21160 operation occur after the atomic operation. 21161 21162 `sync_compare_and_swap_ccMODE' 21163 This pattern is just like `sync_compare_and_swapMODE', except it 21164 should act as if compare part of the compare-and-swap were issued 21165 via `cmpM'. This comparison will only be used with `EQ' and `NE' 21166 branches and `setcc' operations. 21167 21168 Some targets do expose the success or failure of the 21169 compare-and-swap operation via the status flags. Ideally we 21170 wouldn't need a separate named pattern in order to take advantage 21171 of this, but the combine pass does not handle patterns with 21172 multiple sets, which is required by definition for 21173 `sync_compare_and_swapMODE'. 21174 21175 `sync_addMODE', `sync_subMODE' 21176 `sync_iorMODE', `sync_andMODE' 21177 `sync_xorMODE', `sync_nandMODE' 21178 These patterns emit code for an atomic operation on memory. 21179 Operand 0 is the memory on which the atomic operation is performed. 21180 Operand 1 is the second operand to the binary operator. 21181 21182 This pattern must issue any memory barrier instructions such that 21183 all memory operations before the atomic operation occur before the 21184 atomic operation and all memory operations after the atomic 21185 operation occur after the atomic operation. 21186 21187 If these patterns are not defined, the operation will be 21188 constructed from a compare-and-swap operation, if defined. 21189 21190 `sync_old_addMODE', `sync_old_subMODE' 21191 `sync_old_iorMODE', `sync_old_andMODE' 21192 `sync_old_xorMODE', `sync_old_nandMODE' 21193 These patterns are emit code for an atomic operation on memory, 21194 and return the value that the memory contained before the 21195 operation. Operand 0 is the result value, operand 1 is the memory 21196 on which the atomic operation is performed, and operand 2 is the 21197 second operand to the binary operator. 21198 21199 This pattern must issue any memory barrier instructions such that 21200 all memory operations before the atomic operation occur before the 21201 atomic operation and all memory operations after the atomic 21202 operation occur after the atomic operation. 21203 21204 If these patterns are not defined, the operation will be 21205 constructed from a compare-and-swap operation, if defined. 21206 21207 `sync_new_addMODE', `sync_new_subMODE' 21208 `sync_new_iorMODE', `sync_new_andMODE' 21209 `sync_new_xorMODE', `sync_new_nandMODE' 21210 These patterns are like their `sync_old_OP' counterparts, except 21211 that they return the value that exists in the memory location 21212 after the operation, rather than before the operation. 21213 21214 `sync_lock_test_and_setMODE' 21215 This pattern takes two forms, based on the capabilities of the 21216 target. In either case, operand 0 is the result of the operand, 21217 operand 1 is the memory on which the atomic operation is 21218 performed, and operand 2 is the value to set in the lock. 21219 21220 In the ideal case, this operation is an atomic exchange operation, 21221 in which the previous value in memory operand is copied into the 21222 result operand, and the value operand is stored in the memory 21223 operand. 21224 21225 For less capable targets, any value operand that is not the 21226 constant 1 should be rejected with `FAIL'. In this case the 21227 target may use an atomic test-and-set bit operation. The result 21228 operand should contain 1 if the bit was previously set and 0 if 21229 the bit was previously clear. The true contents of the memory 21230 operand are implementation defined. 21231 21232 This pattern must issue any memory barrier instructions such that 21233 the pattern as a whole acts as an acquire barrier, that is all 21234 memory operations after the pattern do not occur until the lock is 21235 acquired. 21236 21237 If this pattern is not defined, the operation will be constructed 21238 from a compare-and-swap operation, if defined. 21239 21240 `sync_lock_releaseMODE' 21241 This pattern, if defined, releases a lock set by 21242 `sync_lock_test_and_setMODE'. Operand 0 is the memory that 21243 contains the lock; operand 1 is the value to store in the lock. 21244 21245 If the target doesn't implement full semantics for 21246 `sync_lock_test_and_setMODE', any value operand which is not the 21247 constant 0 should be rejected with `FAIL', and the true contents 21248 of the memory operand are implementation defined. 21249 21250 This pattern must issue any memory barrier instructions such that 21251 the pattern as a whole acts as a release barrier, that is the lock 21252 is released only after all previous memory operations have 21253 completed. 21254 21255 If this pattern is not defined, then a `memory_barrier' pattern 21256 will be emitted, followed by a store of the value to the memory 21257 operand. 21258 21259 `stack_protect_set' 21260 This pattern, if defined, moves a `Pmode' value from the memory in 21261 operand 1 to the memory in operand 0 without leaving the value in 21262 a register afterward. This is to avoid leaking the value some 21263 place that an attacker might use to rewrite the stack guard slot 21264 after having clobbered it. 21265 21266 If this pattern is not defined, then a plain move pattern is 21267 generated. 21268 21269 `stack_protect_test' 21270 This pattern, if defined, compares a `Pmode' value from the memory 21271 in operand 1 with the memory in operand 0 without leaving the 21272 value in a register afterward and branches to operand 2 if the 21273 values weren't equal. 21274 21275 If this pattern is not defined, then a plain compare pattern and 21276 conditional branch pattern is used. 21277 21278 `clear_cache' 21279 This pattern, if defined, flushes the instruction cache for a 21280 region of memory. The region is bounded to by the Pmode pointers 21281 in operand 0 inclusive and operand 1 exclusive. 21282 21283 If this pattern is not defined, a call to the library function 21284 `__clear_cache' is used. 21285 21286 21287 21288 File: gccint.info, Node: Pattern Ordering, Next: Dependent Patterns, Prev: Standard Names, Up: Machine Desc 21289 21290 16.10 When the Order of Patterns Matters 21291 ======================================== 21292 21293 Sometimes an insn can match more than one instruction pattern. Then the 21294 pattern that appears first in the machine description is the one used. 21295 Therefore, more specific patterns (patterns that will match fewer 21296 things) and faster instructions (those that will produce better code 21297 when they do match) should usually go first in the description. 21298 21299 In some cases the effect of ordering the patterns can be used to hide 21300 a pattern when it is not valid. For example, the 68000 has an 21301 instruction for converting a fullword to floating point and another for 21302 converting a byte to floating point. An instruction converting an 21303 integer to floating point could match either one. We put the pattern 21304 to convert the fullword first to make sure that one will be used rather 21305 than the other. (Otherwise a large integer might be generated as a 21306 single-byte immediate quantity, which would not work.) Instead of 21307 using this pattern ordering it would be possible to make the pattern 21308 for convert-a-byte smart enough to deal properly with any constant 21309 value. 21310 21311 21312 File: gccint.info, Node: Dependent Patterns, Next: Jump Patterns, Prev: Pattern Ordering, Up: Machine Desc 21313 21314 16.11 Interdependence of Patterns 21315 ================================= 21316 21317 Every machine description must have a named pattern for each of the 21318 conditional branch names `bCOND'. The recognition template must always 21319 have the form 21320 21321 (set (pc) 21322 (if_then_else (COND (cc0) (const_int 0)) 21323 (label_ref (match_operand 0 "" "")) 21324 (pc))) 21325 21326 In addition, every machine description must have an anonymous pattern 21327 for each of the possible reverse-conditional branches. Their templates 21328 look like 21329 21330 (set (pc) 21331 (if_then_else (COND (cc0) (const_int 0)) 21332 (pc) 21333 (label_ref (match_operand 0 "" "")))) 21334 21335 They are necessary because jump optimization can turn direct-conditional 21336 branches into reverse-conditional branches. 21337 21338 It is often convenient to use the `match_operator' construct to reduce 21339 the number of patterns that must be specified for branches. For 21340 example, 21341 21342 (define_insn "" 21343 [(set (pc) 21344 (if_then_else (match_operator 0 "comparison_operator" 21345 [(cc0) (const_int 0)]) 21346 (pc) 21347 (label_ref (match_operand 1 "" ""))))] 21348 "CONDITION" 21349 "...") 21350 21351 In some cases machines support instructions identical except for the 21352 machine mode of one or more operands. For example, there may be 21353 "sign-extend halfword" and "sign-extend byte" instructions whose 21354 patterns are 21355 21356 (set (match_operand:SI 0 ...) 21357 (extend:SI (match_operand:HI 1 ...))) 21358 21359 (set (match_operand:SI 0 ...) 21360 (extend:SI (match_operand:QI 1 ...))) 21361 21362 Constant integers do not specify a machine mode, so an instruction to 21363 extend a constant value could match either pattern. The pattern it 21364 actually will match is the one that appears first in the file. For 21365 correct results, this must be the one for the widest possible mode 21366 (`HImode', here). If the pattern matches the `QImode' instruction, the 21367 results will be incorrect if the constant value does not actually fit 21368 that mode. 21369 21370 Such instructions to extend constants are rarely generated because 21371 they are optimized away, but they do occasionally happen in nonoptimized 21372 compilations. 21373 21374 If a constraint in a pattern allows a constant, the reload pass may 21375 replace a register with a constant permitted by the constraint in some 21376 cases. Similarly for memory references. Because of this substitution, 21377 you should not provide separate patterns for increment and decrement 21378 instructions. Instead, they should be generated from the same pattern 21379 that supports register-register add insns by examining the operands and 21380 generating the appropriate machine instruction. 21381 21382 21383 File: gccint.info, Node: Jump Patterns, Next: Looping Patterns, Prev: Dependent Patterns, Up: Machine Desc 21384 21385 16.12 Defining Jump Instruction Patterns 21386 ======================================== 21387 21388 For most machines, GCC assumes that the machine has a condition code. 21389 A comparison insn sets the condition code, recording the results of both 21390 signed and unsigned comparison of the given operands. A separate branch 21391 insn tests the condition code and branches or not according its value. 21392 The branch insns come in distinct signed and unsigned flavors. Many 21393 common machines, such as the VAX, the 68000 and the 32000, work this 21394 way. 21395 21396 Some machines have distinct signed and unsigned compare instructions, 21397 and only one set of conditional branch instructions. The easiest way 21398 to handle these machines is to treat them just like the others until 21399 the final stage where assembly code is written. At this time, when 21400 outputting code for the compare instruction, peek ahead at the 21401 following branch using `next_cc0_user (insn)'. (The variable `insn' 21402 refers to the insn being output, in the output-writing code in an 21403 instruction pattern.) If the RTL says that is an unsigned branch, 21404 output an unsigned compare; otherwise output a signed compare. When 21405 the branch itself is output, you can treat signed and unsigned branches 21406 identically. 21407 21408 The reason you can do this is that GCC always generates a pair of 21409 consecutive RTL insns, possibly separated by `note' insns, one to set 21410 the condition code and one to test it, and keeps the pair inviolate 21411 until the end. 21412 21413 To go with this technique, you must define the machine-description 21414 macro `NOTICE_UPDATE_CC' to do `CC_STATUS_INIT'; in other words, no 21415 compare instruction is superfluous. 21416 21417 Some machines have compare-and-branch instructions and no condition 21418 code. A similar technique works for them. When it is time to "output" 21419 a compare instruction, record its operands in two static variables. 21420 When outputting the branch-on-condition-code instruction that follows, 21421 actually output a compare-and-branch instruction that uses the 21422 remembered operands. 21423 21424 It also works to define patterns for compare-and-branch instructions. 21425 In optimizing compilation, the pair of compare and branch instructions 21426 will be combined according to these patterns. But this does not happen 21427 if optimization is not requested. So you must use one of the solutions 21428 above in addition to any special patterns you define. 21429 21430 In many RISC machines, most instructions do not affect the condition 21431 code and there may not even be a separate condition code register. On 21432 these machines, the restriction that the definition and use of the 21433 condition code be adjacent insns is not necessary and can prevent 21434 important optimizations. For example, on the IBM RS/6000, there is a 21435 delay for taken branches unless the condition code register is set three 21436 instructions earlier than the conditional branch. The instruction 21437 scheduler cannot perform this optimization if it is not permitted to 21438 separate the definition and use of the condition code register. 21439 21440 On these machines, do not use `(cc0)', but instead use a register to 21441 represent the condition code. If there is a specific condition code 21442 register in the machine, use a hard register. If the condition code or 21443 comparison result can be placed in any general register, or if there are 21444 multiple condition registers, use a pseudo register. 21445 21446 On some machines, the type of branch instruction generated may depend 21447 on the way the condition code was produced; for example, on the 68k and 21448 SPARC, setting the condition code directly from an add or subtract 21449 instruction does not clear the overflow bit the way that a test 21450 instruction does, so a different branch instruction must be used for 21451 some conditional branches. For machines that use `(cc0)', the set and 21452 use of the condition code must be adjacent (separated only by `note' 21453 insns) allowing flags in `cc_status' to be used. (*Note Condition 21454 Code::.) Also, the comparison and branch insns can be located from 21455 each other by using the functions `prev_cc0_setter' and `next_cc0_user'. 21456 21457 However, this is not true on machines that do not use `(cc0)'. On 21458 those machines, no assumptions can be made about the adjacency of the 21459 compare and branch insns and the above methods cannot be used. Instead, 21460 we use the machine mode of the condition code register to record 21461 different formats of the condition code register. 21462 21463 Registers used to store the condition code value should have a mode 21464 that is in class `MODE_CC'. Normally, it will be `CCmode'. If 21465 additional modes are required (as for the add example mentioned above in 21466 the SPARC), define them in `MACHINE-modes.def' (*note Condition 21467 Code::). Also define `SELECT_CC_MODE' to choose a mode given an 21468 operand of a compare. 21469 21470 If it is known during RTL generation that a different mode will be 21471 required (for example, if the machine has separate compare instructions 21472 for signed and unsigned quantities, like most IBM processors), they can 21473 be specified at that time. 21474 21475 If the cases that require different modes would be made by instruction 21476 combination, the macro `SELECT_CC_MODE' determines which machine mode 21477 should be used for the comparison result. The patterns should be 21478 written using that mode. To support the case of the add on the SPARC 21479 discussed above, we have the pattern 21480 21481 (define_insn "" 21482 [(set (reg:CC_NOOV 0) 21483 (compare:CC_NOOV 21484 (plus:SI (match_operand:SI 0 "register_operand" "%r") 21485 (match_operand:SI 1 "arith_operand" "rI")) 21486 (const_int 0)))] 21487 "" 21488 "...") 21489 21490 The `SELECT_CC_MODE' macro on the SPARC returns `CC_NOOVmode' for 21491 comparisons whose argument is a `plus'. 21492 21493 21494 File: gccint.info, Node: Looping Patterns, Next: Insn Canonicalizations, Prev: Jump Patterns, Up: Machine Desc 21495 21496 16.13 Defining Looping Instruction Patterns 21497 =========================================== 21498 21499 Some machines have special jump instructions that can be utilized to 21500 make loops more efficient. A common example is the 68000 `dbra' 21501 instruction which performs a decrement of a register and a branch if the 21502 result was greater than zero. Other machines, in particular digital 21503 signal processors (DSPs), have special block repeat instructions to 21504 provide low-overhead loop support. For example, the TI TMS320C3x/C4x 21505 DSPs have a block repeat instruction that loads special registers to 21506 mark the top and end of a loop and to count the number of loop 21507 iterations. This avoids the need for fetching and executing a 21508 `dbra'-like instruction and avoids pipeline stalls associated with the 21509 jump. 21510 21511 GCC has three special named patterns to support low overhead looping. 21512 They are `decrement_and_branch_until_zero', `doloop_begin', and 21513 `doloop_end'. The first pattern, `decrement_and_branch_until_zero', is 21514 not emitted during RTL generation but may be emitted during the 21515 instruction combination phase. This requires the assistance of the 21516 loop optimizer, using information collected during strength reduction, 21517 to reverse a loop to count down to zero. Some targets also require the 21518 loop optimizer to add a `REG_NONNEG' note to indicate that the 21519 iteration count is always positive. This is needed if the target 21520 performs a signed loop termination test. For example, the 68000 uses a 21521 pattern similar to the following for its `dbra' instruction: 21522 21523 (define_insn "decrement_and_branch_until_zero" 21524 [(set (pc) 21525 (if_then_else 21526 (ge (plus:SI (match_operand:SI 0 "general_operand" "+d*am") 21527 (const_int -1)) 21528 (const_int 0)) 21529 (label_ref (match_operand 1 "" "")) 21530 (pc))) 21531 (set (match_dup 0) 21532 (plus:SI (match_dup 0) 21533 (const_int -1)))] 21534 "find_reg_note (insn, REG_NONNEG, 0)" 21535 "...") 21536 21537 Note that since the insn is both a jump insn and has an output, it must 21538 deal with its own reloads, hence the `m' constraints. Also note that 21539 since this insn is generated by the instruction combination phase 21540 combining two sequential insns together into an implicit parallel insn, 21541 the iteration counter needs to be biased by the same amount as the 21542 decrement operation, in this case -1. Note that the following similar 21543 pattern will not be matched by the combiner. 21544 21545 (define_insn "decrement_and_branch_until_zero" 21546 [(set (pc) 21547 (if_then_else 21548 (ge (match_operand:SI 0 "general_operand" "+d*am") 21549 (const_int 1)) 21550 (label_ref (match_operand 1 "" "")) 21551 (pc))) 21552 (set (match_dup 0) 21553 (plus:SI (match_dup 0) 21554 (const_int -1)))] 21555 "find_reg_note (insn, REG_NONNEG, 0)" 21556 "...") 21557 21558 The other two special looping patterns, `doloop_begin' and 21559 `doloop_end', are emitted by the loop optimizer for certain 21560 well-behaved loops with a finite number of loop iterations using 21561 information collected during strength reduction. 21562 21563 The `doloop_end' pattern describes the actual looping instruction (or 21564 the implicit looping operation) and the `doloop_begin' pattern is an 21565 optional companion pattern that can be used for initialization needed 21566 for some low-overhead looping instructions. 21567 21568 Note that some machines require the actual looping instruction to be 21569 emitted at the top of the loop (e.g., the TMS320C3x/C4x DSPs). Emitting 21570 the true RTL for a looping instruction at the top of the loop can cause 21571 problems with flow analysis. So instead, a dummy `doloop' insn is 21572 emitted at the end of the loop. The machine dependent reorg pass checks 21573 for the presence of this `doloop' insn and then searches back to the 21574 top of the loop, where it inserts the true looping insn (provided there 21575 are no instructions in the loop which would cause problems). Any 21576 additional labels can be emitted at this point. In addition, if the 21577 desired special iteration counter register was not allocated, this 21578 machine dependent reorg pass could emit a traditional compare and jump 21579 instruction pair. 21580 21581 The essential difference between the `decrement_and_branch_until_zero' 21582 and the `doloop_end' patterns is that the loop optimizer allocates an 21583 additional pseudo register for the latter as an iteration counter. 21584 This pseudo register cannot be used within the loop (i.e., general 21585 induction variables cannot be derived from it), however, in many cases 21586 the loop induction variable may become redundant and removed by the 21587 flow pass. 21588 21589 21590 File: gccint.info, Node: Insn Canonicalizations, Next: Expander Definitions, Prev: Looping Patterns, Up: Machine Desc 21591 21592 16.14 Canonicalization of Instructions 21593 ====================================== 21594 21595 There are often cases where multiple RTL expressions could represent an 21596 operation performed by a single machine instruction. This situation is 21597 most commonly encountered with logical, branch, and multiply-accumulate 21598 instructions. In such cases, the compiler attempts to convert these 21599 multiple RTL expressions into a single canonical form to reduce the 21600 number of insn patterns required. 21601 21602 In addition to algebraic simplifications, following canonicalizations 21603 are performed: 21604 21605 * For commutative and comparison operators, a constant is always 21606 made the second operand. If a machine only supports a constant as 21607 the second operand, only patterns that match a constant in the 21608 second operand need be supplied. 21609 21610 * For associative operators, a sequence of operators will always 21611 chain to the left; for instance, only the left operand of an 21612 integer `plus' can itself be a `plus'. `and', `ior', `xor', 21613 `plus', `mult', `smin', `smax', `umin', and `umax' are associative 21614 when applied to integers, and sometimes to floating-point. 21615 21616 * For these operators, if only one operand is a `neg', `not', 21617 `mult', `plus', or `minus' expression, it will be the first 21618 operand. 21619 21620 * In combinations of `neg', `mult', `plus', and `minus', the `neg' 21621 operations (if any) will be moved inside the operations as far as 21622 possible. For instance, `(neg (mult A B))' is canonicalized as 21623 `(mult (neg A) B)', but `(plus (mult (neg A) B) C)' is 21624 canonicalized as `(minus A (mult B C))'. 21625 21626 * For the `compare' operator, a constant is always the second operand 21627 on machines where `cc0' is used (*note Jump Patterns::). On other 21628 machines, there are rare cases where the compiler might want to 21629 construct a `compare' with a constant as the first operand. 21630 However, these cases are not common enough for it to be worthwhile 21631 to provide a pattern matching a constant as the first operand 21632 unless the machine actually has such an instruction. 21633 21634 An operand of `neg', `not', `mult', `plus', or `minus' is made the 21635 first operand under the same conditions as above. 21636 21637 * `(ltu (plus A B) B)' is converted to `(ltu (plus A B) A)'. 21638 Likewise with `geu' instead of `ltu'. 21639 21640 * `(minus X (const_int N))' is converted to `(plus X (const_int 21641 -N))'. 21642 21643 * Within address computations (i.e., inside `mem'), a left shift is 21644 converted into the appropriate multiplication by a power of two. 21645 21646 * De Morgan's Law is used to move bitwise negation inside a bitwise 21647 logical-and or logical-or operation. If this results in only one 21648 operand being a `not' expression, it will be the first one. 21649 21650 A machine that has an instruction that performs a bitwise 21651 logical-and of one operand with the bitwise negation of the other 21652 should specify the pattern for that instruction as 21653 21654 (define_insn "" 21655 [(set (match_operand:M 0 ...) 21656 (and:M (not:M (match_operand:M 1 ...)) 21657 (match_operand:M 2 ...)))] 21658 "..." 21659 "...") 21660 21661 Similarly, a pattern for a "NAND" instruction should be written 21662 21663 (define_insn "" 21664 [(set (match_operand:M 0 ...) 21665 (ior:M (not:M (match_operand:M 1 ...)) 21666 (not:M (match_operand:M 2 ...))))] 21667 "..." 21668 "...") 21669 21670 In both cases, it is not necessary to include patterns for the many 21671 logically equivalent RTL expressions. 21672 21673 * The only possible RTL expressions involving both bitwise 21674 exclusive-or and bitwise negation are `(xor:M X Y)' and `(not:M 21675 (xor:M X Y))'. 21676 21677 * The sum of three items, one of which is a constant, will only 21678 appear in the form 21679 21680 (plus:M (plus:M X Y) CONSTANT) 21681 21682 * On machines that do not use `cc0', `(compare X (const_int 0))' 21683 will be converted to X. 21684 21685 * Equality comparisons of a group of bits (usually a single bit) 21686 with zero will be written using `zero_extract' rather than the 21687 equivalent `and' or `sign_extract' operations. 21688 21689 21690 Further canonicalization rules are defined in the function 21691 `commutative_operand_precedence' in `gcc/rtlanal.c'. 21692 21693 21694 File: gccint.info, Node: Expander Definitions, Next: Insn Splitting, Prev: Insn Canonicalizations, Up: Machine Desc 21695 21696 16.15 Defining RTL Sequences for Code Generation 21697 ================================================ 21698 21699 On some target machines, some standard pattern names for RTL generation 21700 cannot be handled with single insn, but a sequence of RTL insns can 21701 represent them. For these target machines, you can write a 21702 `define_expand' to specify how to generate the sequence of RTL. 21703 21704 A `define_expand' is an RTL expression that looks almost like a 21705 `define_insn'; but, unlike the latter, a `define_expand' is used only 21706 for RTL generation and it can produce more than one RTL insn. 21707 21708 A `define_expand' RTX has four operands: 21709 21710 * The name. Each `define_expand' must have a name, since the only 21711 use for it is to refer to it by name. 21712 21713 * The RTL template. This is a vector of RTL expressions representing 21714 a sequence of separate instructions. Unlike `define_insn', there 21715 is no implicit surrounding `PARALLEL'. 21716 21717 * The condition, a string containing a C expression. This 21718 expression is used to express how the availability of this pattern 21719 depends on subclasses of target machine, selected by command-line 21720 options when GCC is run. This is just like the condition of a 21721 `define_insn' that has a standard name. Therefore, the condition 21722 (if present) may not depend on the data in the insn being matched, 21723 but only the target-machine-type flags. The compiler needs to 21724 test these conditions during initialization in order to learn 21725 exactly which named instructions are available in a particular run. 21726 21727 * The preparation statements, a string containing zero or more C 21728 statements which are to be executed before RTL code is generated 21729 from the RTL template. 21730 21731 Usually these statements prepare temporary registers for use as 21732 internal operands in the RTL template, but they can also generate 21733 RTL insns directly by calling routines such as `emit_insn', etc. 21734 Any such insns precede the ones that come from the RTL template. 21735 21736 Every RTL insn emitted by a `define_expand' must match some 21737 `define_insn' in the machine description. Otherwise, the compiler will 21738 crash when trying to generate code for the insn or trying to optimize 21739 it. 21740 21741 The RTL template, in addition to controlling generation of RTL insns, 21742 also describes the operands that need to be specified when this pattern 21743 is used. In particular, it gives a predicate for each operand. 21744 21745 A true operand, which needs to be specified in order to generate RTL 21746 from the pattern, should be described with a `match_operand' in its 21747 first occurrence in the RTL template. This enters information on the 21748 operand's predicate into the tables that record such things. GCC uses 21749 the information to preload the operand into a register if that is 21750 required for valid RTL code. If the operand is referred to more than 21751 once, subsequent references should use `match_dup'. 21752 21753 The RTL template may also refer to internal "operands" which are 21754 temporary registers or labels used only within the sequence made by the 21755 `define_expand'. Internal operands are substituted into the RTL 21756 template with `match_dup', never with `match_operand'. The values of 21757 the internal operands are not passed in as arguments by the compiler 21758 when it requests use of this pattern. Instead, they are computed 21759 within the pattern, in the preparation statements. These statements 21760 compute the values and store them into the appropriate elements of 21761 `operands' so that `match_dup' can find them. 21762 21763 There are two special macros defined for use in the preparation 21764 statements: `DONE' and `FAIL'. Use them with a following semicolon, as 21765 a statement. 21766 21767 `DONE' 21768 Use the `DONE' macro to end RTL generation for the pattern. The 21769 only RTL insns resulting from the pattern on this occasion will be 21770 those already emitted by explicit calls to `emit_insn' within the 21771 preparation statements; the RTL template will not be generated. 21772 21773 `FAIL' 21774 Make the pattern fail on this occasion. When a pattern fails, it 21775 means that the pattern was not truly available. The calling 21776 routines in the compiler will try other strategies for code 21777 generation using other patterns. 21778 21779 Failure is currently supported only for binary (addition, 21780 multiplication, shifting, etc.) and bit-field (`extv', `extzv', 21781 and `insv') operations. 21782 21783 If the preparation falls through (invokes neither `DONE' nor `FAIL'), 21784 then the `define_expand' acts like a `define_insn' in that the RTL 21785 template is used to generate the insn. 21786 21787 The RTL template is not used for matching, only for generating the 21788 initial insn list. If the preparation statement always invokes `DONE' 21789 or `FAIL', the RTL template may be reduced to a simple list of 21790 operands, such as this example: 21791 21792 (define_expand "addsi3" 21793 [(match_operand:SI 0 "register_operand" "") 21794 (match_operand:SI 1 "register_operand" "") 21795 (match_operand:SI 2 "register_operand" "")] 21796 "" 21797 " 21798 { 21799 handle_add (operands[0], operands[1], operands[2]); 21800 DONE; 21801 }") 21802 21803 Here is an example, the definition of left-shift for the SPUR chip: 21804 21805 (define_expand "ashlsi3" 21806 [(set (match_operand:SI 0 "register_operand" "") 21807 (ashift:SI 21808 (match_operand:SI 1 "register_operand" "") 21809 (match_operand:SI 2 "nonmemory_operand" "")))] 21810 "" 21811 " 21812 21813 { 21814 if (GET_CODE (operands[2]) != CONST_INT 21815 || (unsigned) INTVAL (operands[2]) > 3) 21816 FAIL; 21817 }") 21818 21819 This example uses `define_expand' so that it can generate an RTL insn 21820 for shifting when the shift-count is in the supported range of 0 to 3 21821 but fail in other cases where machine insns aren't available. When it 21822 fails, the compiler tries another strategy using different patterns 21823 (such as, a library call). 21824 21825 If the compiler were able to handle nontrivial condition-strings in 21826 patterns with names, then it would be possible to use a `define_insn' 21827 in that case. Here is another case (zero-extension on the 68000) which 21828 makes more use of the power of `define_expand': 21829 21830 (define_expand "zero_extendhisi2" 21831 [(set (match_operand:SI 0 "general_operand" "") 21832 (const_int 0)) 21833 (set (strict_low_part 21834 (subreg:HI 21835 (match_dup 0) 21836 0)) 21837 (match_operand:HI 1 "general_operand" ""))] 21838 "" 21839 "operands[1] = make_safe_from (operands[1], operands[0]);") 21840 21841 Here two RTL insns are generated, one to clear the entire output operand 21842 and the other to copy the input operand into its low half. This 21843 sequence is incorrect if the input operand refers to [the old value of] 21844 the output operand, so the preparation statement makes sure this isn't 21845 so. The function `make_safe_from' copies the `operands[1]' into a 21846 temporary register if it refers to `operands[0]'. It does this by 21847 emitting another RTL insn. 21848 21849 Finally, a third example shows the use of an internal operand. 21850 Zero-extension on the SPUR chip is done by `and'-ing the result against 21851 a halfword mask. But this mask cannot be represented by a `const_int' 21852 because the constant value is too large to be legitimate on this 21853 machine. So it must be copied into a register with `force_reg' and 21854 then the register used in the `and'. 21855 21856 (define_expand "zero_extendhisi2" 21857 [(set (match_operand:SI 0 "register_operand" "") 21858 (and:SI (subreg:SI 21859 (match_operand:HI 1 "register_operand" "") 21860 0) 21861 (match_dup 2)))] 21862 "" 21863 "operands[2] 21864 = force_reg (SImode, GEN_INT (65535)); ") 21865 21866 _Note:_ If the `define_expand' is used to serve a standard binary or 21867 unary arithmetic operation or a bit-field operation, then the last insn 21868 it generates must not be a `code_label', `barrier' or `note'. It must 21869 be an `insn', `jump_insn' or `call_insn'. If you don't need a real insn 21870 at the end, emit an insn to copy the result of the operation into 21871 itself. Such an insn will generate no code, but it can avoid problems 21872 in the compiler. 21873 21874 21875 File: gccint.info, Node: Insn Splitting, Next: Including Patterns, Prev: Expander Definitions, Up: Machine Desc 21876 21877 16.16 Defining How to Split Instructions 21878 ======================================== 21879 21880 There are two cases where you should specify how to split a pattern 21881 into multiple insns. On machines that have instructions requiring 21882 delay slots (*note Delay Slots::) or that have instructions whose 21883 output is not available for multiple cycles (*note Processor pipeline 21884 description::), the compiler phases that optimize these cases need to 21885 be able to move insns into one-instruction delay slots. However, some 21886 insns may generate more than one machine instruction. These insns 21887 cannot be placed into a delay slot. 21888 21889 Often you can rewrite the single insn as a list of individual insns, 21890 each corresponding to one machine instruction. The disadvantage of 21891 doing so is that it will cause the compilation to be slower and require 21892 more space. If the resulting insns are too complex, it may also 21893 suppress some optimizations. The compiler splits the insn if there is a 21894 reason to believe that it might improve instruction or delay slot 21895 scheduling. 21896 21897 The insn combiner phase also splits putative insns. If three insns are 21898 merged into one insn with a complex expression that cannot be matched by 21899 some `define_insn' pattern, the combiner phase attempts to split the 21900 complex pattern into two insns that are recognized. Usually it can 21901 break the complex pattern into two patterns by splitting out some 21902 subexpression. However, in some other cases, such as performing an 21903 addition of a large constant in two insns on a RISC machine, the way to 21904 split the addition into two insns is machine-dependent. 21905 21906 The `define_split' definition tells the compiler how to split a 21907 complex insn into several simpler insns. It looks like this: 21908 21909 (define_split 21910 [INSN-PATTERN] 21911 "CONDITION" 21912 [NEW-INSN-PATTERN-1 21913 NEW-INSN-PATTERN-2 21914 ...] 21915 "PREPARATION-STATEMENTS") 21916 21917 INSN-PATTERN is a pattern that needs to be split and CONDITION is the 21918 final condition to be tested, as in a `define_insn'. When an insn 21919 matching INSN-PATTERN and satisfying CONDITION is found, it is replaced 21920 in the insn list with the insns given by NEW-INSN-PATTERN-1, 21921 NEW-INSN-PATTERN-2, etc. 21922 21923 The PREPARATION-STATEMENTS are similar to those statements that are 21924 specified for `define_expand' (*note Expander Definitions::) and are 21925 executed before the new RTL is generated to prepare for the generated 21926 code or emit some insns whose pattern is not fixed. Unlike those in 21927 `define_expand', however, these statements must not generate any new 21928 pseudo-registers. Once reload has completed, they also must not 21929 allocate any space in the stack frame. 21930 21931 Patterns are matched against INSN-PATTERN in two different 21932 circumstances. If an insn needs to be split for delay slot scheduling 21933 or insn scheduling, the insn is already known to be valid, which means 21934 that it must have been matched by some `define_insn' and, if 21935 `reload_completed' is nonzero, is known to satisfy the constraints of 21936 that `define_insn'. In that case, the new insn patterns must also be 21937 insns that are matched by some `define_insn' and, if `reload_completed' 21938 is nonzero, must also satisfy the constraints of those definitions. 21939 21940 As an example of this usage of `define_split', consider the following 21941 example from `a29k.md', which splits a `sign_extend' from `HImode' to 21942 `SImode' into a pair of shift insns: 21943 21944 (define_split 21945 [(set (match_operand:SI 0 "gen_reg_operand" "") 21946 (sign_extend:SI (match_operand:HI 1 "gen_reg_operand" "")))] 21947 "" 21948 [(set (match_dup 0) 21949 (ashift:SI (match_dup 1) 21950 (const_int 16))) 21951 (set (match_dup 0) 21952 (ashiftrt:SI (match_dup 0) 21953 (const_int 16)))] 21954 " 21955 { operands[1] = gen_lowpart (SImode, operands[1]); }") 21956 21957 When the combiner phase tries to split an insn pattern, it is always 21958 the case that the pattern is _not_ matched by any `define_insn'. The 21959 combiner pass first tries to split a single `set' expression and then 21960 the same `set' expression inside a `parallel', but followed by a 21961 `clobber' of a pseudo-reg to use as a scratch register. In these 21962 cases, the combiner expects exactly two new insn patterns to be 21963 generated. It will verify that these patterns match some `define_insn' 21964 definitions, so you need not do this test in the `define_split' (of 21965 course, there is no point in writing a `define_split' that will never 21966 produce insns that match). 21967 21968 Here is an example of this use of `define_split', taken from 21969 `rs6000.md': 21970 21971 (define_split 21972 [(set (match_operand:SI 0 "gen_reg_operand" "") 21973 (plus:SI (match_operand:SI 1 "gen_reg_operand" "") 21974 (match_operand:SI 2 "non_add_cint_operand" "")))] 21975 "" 21976 [(set (match_dup 0) (plus:SI (match_dup 1) (match_dup 3))) 21977 (set (match_dup 0) (plus:SI (match_dup 0) (match_dup 4)))] 21978 " 21979 { 21980 int low = INTVAL (operands[2]) & 0xffff; 21981 int high = (unsigned) INTVAL (operands[2]) >> 16; 21982 21983 if (low & 0x8000) 21984 high++, low |= 0xffff0000; 21985 21986 operands[3] = GEN_INT (high << 16); 21987 operands[4] = GEN_INT (low); 21988 }") 21989 21990 Here the predicate `non_add_cint_operand' matches any `const_int' that 21991 is _not_ a valid operand of a single add insn. The add with the 21992 smaller displacement is written so that it can be substituted into the 21993 address of a subsequent operation. 21994 21995 An example that uses a scratch register, from the same file, generates 21996 an equality comparison of a register and a large constant: 21997 21998 (define_split 21999 [(set (match_operand:CC 0 "cc_reg_operand" "") 22000 (compare:CC (match_operand:SI 1 "gen_reg_operand" "") 22001 (match_operand:SI 2 "non_short_cint_operand" ""))) 22002 (clobber (match_operand:SI 3 "gen_reg_operand" ""))] 22003 "find_single_use (operands[0], insn, 0) 22004 && (GET_CODE (*find_single_use (operands[0], insn, 0)) == EQ 22005 || GET_CODE (*find_single_use (operands[0], insn, 0)) == NE)" 22006 [(set (match_dup 3) (xor:SI (match_dup 1) (match_dup 4))) 22007 (set (match_dup 0) (compare:CC (match_dup 3) (match_dup 5)))] 22008 " 22009 { 22010 /* Get the constant we are comparing against, C, and see what it 22011 looks like sign-extended to 16 bits. Then see what constant 22012 could be XOR'ed with C to get the sign-extended value. */ 22013 22014 int c = INTVAL (operands[2]); 22015 int sextc = (c << 16) >> 16; 22016 int xorv = c ^ sextc; 22017 22018 operands[4] = GEN_INT (xorv); 22019 operands[5] = GEN_INT (sextc); 22020 }") 22021 22022 To avoid confusion, don't write a single `define_split' that accepts 22023 some insns that match some `define_insn' as well as some insns that 22024 don't. Instead, write two separate `define_split' definitions, one for 22025 the insns that are valid and one for the insns that are not valid. 22026 22027 The splitter is allowed to split jump instructions into sequence of 22028 jumps or create new jumps in while splitting non-jump instructions. As 22029 the central flowgraph and branch prediction information needs to be 22030 updated, several restriction apply. 22031 22032 Splitting of jump instruction into sequence that over by another jump 22033 instruction is always valid, as compiler expect identical behavior of 22034 new jump. When new sequence contains multiple jump instructions or new 22035 labels, more assistance is needed. Splitter is required to create only 22036 unconditional jumps, or simple conditional jump instructions. 22037 Additionally it must attach a `REG_BR_PROB' note to each conditional 22038 jump. A global variable `split_branch_probability' holds the 22039 probability of the original branch in case it was an simple conditional 22040 jump, -1 otherwise. To simplify recomputing of edge frequencies, the 22041 new sequence is required to have only forward jumps to the newly 22042 created labels. 22043 22044 For the common case where the pattern of a define_split exactly 22045 matches the pattern of a define_insn, use `define_insn_and_split'. It 22046 looks like this: 22047 22048 (define_insn_and_split 22049 [INSN-PATTERN] 22050 "CONDITION" 22051 "OUTPUT-TEMPLATE" 22052 "SPLIT-CONDITION" 22053 [NEW-INSN-PATTERN-1 22054 NEW-INSN-PATTERN-2 22055 ...] 22056 "PREPARATION-STATEMENTS" 22057 [INSN-ATTRIBUTES]) 22058 22059 INSN-PATTERN, CONDITION, OUTPUT-TEMPLATE, and INSN-ATTRIBUTES are used 22060 as in `define_insn'. The NEW-INSN-PATTERN vector and the 22061 PREPARATION-STATEMENTS are used as in a `define_split'. The 22062 SPLIT-CONDITION is also used as in `define_split', with the additional 22063 behavior that if the condition starts with `&&', the condition used for 22064 the split will be the constructed as a logical "and" of the split 22065 condition with the insn condition. For example, from i386.md: 22066 22067 (define_insn_and_split "zero_extendhisi2_and" 22068 [(set (match_operand:SI 0 "register_operand" "=r") 22069 (zero_extend:SI (match_operand:HI 1 "register_operand" "0"))) 22070 (clobber (reg:CC 17))] 22071 "TARGET_ZERO_EXTEND_WITH_AND && !optimize_size" 22072 "#" 22073 "&& reload_completed" 22074 [(parallel [(set (match_dup 0) 22075 (and:SI (match_dup 0) (const_int 65535))) 22076 (clobber (reg:CC 17))])] 22077 "" 22078 [(set_attr "type" "alu1")]) 22079 22080 In this case, the actual split condition will be 22081 `TARGET_ZERO_EXTEND_WITH_AND && !optimize_size && reload_completed'. 22082 22083 The `define_insn_and_split' construction provides exactly the same 22084 functionality as two separate `define_insn' and `define_split' 22085 patterns. It exists for compactness, and as a maintenance tool to 22086 prevent having to ensure the two patterns' templates match. 22087 22088 22089 File: gccint.info, Node: Including Patterns, Next: Peephole Definitions, Prev: Insn Splitting, Up: Machine Desc 22090 22091 16.17 Including Patterns in Machine Descriptions. 22092 ================================================= 22093 22094 The `include' pattern tells the compiler tools where to look for 22095 patterns that are in files other than in the file `.md'. This is used 22096 only at build time and there is no preprocessing allowed. 22097 22098 It looks like: 22099 22100 22101 (include 22102 PATHNAME) 22103 22104 For example: 22105 22106 22107 (include "filestuff") 22108 22109 Where PATHNAME is a string that specifies the location of the file, 22110 specifies the include file to be in `gcc/config/target/filestuff'. The 22111 directory `gcc/config/target' is regarded as the default directory. 22112 22113 Machine descriptions may be split up into smaller more manageable 22114 subsections and placed into subdirectories. 22115 22116 By specifying: 22117 22118 22119 (include "BOGUS/filestuff") 22120 22121 the include file is specified to be in 22122 `gcc/config/TARGET/BOGUS/filestuff'. 22123 22124 Specifying an absolute path for the include file such as; 22125 22126 (include "/u2/BOGUS/filestuff") 22127 is permitted but is not encouraged. 22128 22129 16.17.1 RTL Generation Tool Options for Directory Search 22130 -------------------------------------------------------- 22131 22132 The `-IDIR' option specifies directories to search for machine 22133 descriptions. For example: 22134 22135 22136 genrecog -I/p1/abc/proc1 -I/p2/abcd/pro2 target.md 22137 22138 Add the directory DIR to the head of the list of directories to be 22139 searched for header files. This can be used to override a system 22140 machine definition file, substituting your own version, since these 22141 directories are searched before the default machine description file 22142 directories. If you use more than one `-I' option, the directories are 22143 scanned in left-to-right order; the standard default directory come 22144 after. 22145 22146 22147 File: gccint.info, Node: Peephole Definitions, Next: Insn Attributes, Prev: Including Patterns, Up: Machine Desc 22148 22149 16.18 Machine-Specific Peephole Optimizers 22150 ========================================== 22151 22152 In addition to instruction patterns the `md' file may contain 22153 definitions of machine-specific peephole optimizations. 22154 22155 The combiner does not notice certain peephole optimizations when the 22156 data flow in the program does not suggest that it should try them. For 22157 example, sometimes two consecutive insns related in purpose can be 22158 combined even though the second one does not appear to use a register 22159 computed in the first one. A machine-specific peephole optimizer can 22160 detect such opportunities. 22161 22162 There are two forms of peephole definitions that may be used. The 22163 original `define_peephole' is run at assembly output time to match 22164 insns and substitute assembly text. Use of `define_peephole' is 22165 deprecated. 22166 22167 A newer `define_peephole2' matches insns and substitutes new insns. 22168 The `peephole2' pass is run after register allocation but before 22169 scheduling, which may result in much better code for targets that do 22170 scheduling. 22171 22172 * Menu: 22173 22174 * define_peephole:: RTL to Text Peephole Optimizers 22175 * define_peephole2:: RTL to RTL Peephole Optimizers 22176 22177 22178 File: gccint.info, Node: define_peephole, Next: define_peephole2, Up: Peephole Definitions 22179 22180 16.18.1 RTL to Text Peephole Optimizers 22181 --------------------------------------- 22182 22183 A definition looks like this: 22184 22185 (define_peephole 22186 [INSN-PATTERN-1 22187 INSN-PATTERN-2 22188 ...] 22189 "CONDITION" 22190 "TEMPLATE" 22191 "OPTIONAL-INSN-ATTRIBUTES") 22192 22193 The last string operand may be omitted if you are not using any 22194 machine-specific information in this machine description. If present, 22195 it must obey the same rules as in a `define_insn'. 22196 22197 In this skeleton, INSN-PATTERN-1 and so on are patterns to match 22198 consecutive insns. The optimization applies to a sequence of insns when 22199 INSN-PATTERN-1 matches the first one, INSN-PATTERN-2 matches the next, 22200 and so on. 22201 22202 Each of the insns matched by a peephole must also match a 22203 `define_insn'. Peepholes are checked only at the last stage just 22204 before code generation, and only optionally. Therefore, any insn which 22205 would match a peephole but no `define_insn' will cause a crash in code 22206 generation in an unoptimized compilation, or at various optimization 22207 stages. 22208 22209 The operands of the insns are matched with `match_operands', 22210 `match_operator', and `match_dup', as usual. What is not usual is that 22211 the operand numbers apply to all the insn patterns in the definition. 22212 So, you can check for identical operands in two insns by using 22213 `match_operand' in one insn and `match_dup' in the other. 22214 22215 The operand constraints used in `match_operand' patterns do not have 22216 any direct effect on the applicability of the peephole, but they will 22217 be validated afterward, so make sure your constraints are general enough 22218 to apply whenever the peephole matches. If the peephole matches but 22219 the constraints are not satisfied, the compiler will crash. 22220 22221 It is safe to omit constraints in all the operands of the peephole; or 22222 you can write constraints which serve as a double-check on the criteria 22223 previously tested. 22224 22225 Once a sequence of insns matches the patterns, the CONDITION is 22226 checked. This is a C expression which makes the final decision whether 22227 to perform the optimization (we do so if the expression is nonzero). If 22228 CONDITION is omitted (in other words, the string is empty) then the 22229 optimization is applied to every sequence of insns that matches the 22230 patterns. 22231 22232 The defined peephole optimizations are applied after register 22233 allocation is complete. Therefore, the peephole definition can check 22234 which operands have ended up in which kinds of registers, just by 22235 looking at the operands. 22236 22237 The way to refer to the operands in CONDITION is to write 22238 `operands[I]' for operand number I (as matched by `(match_operand I 22239 ...)'). Use the variable `insn' to refer to the last of the insns 22240 being matched; use `prev_active_insn' to find the preceding insns. 22241 22242 When optimizing computations with intermediate results, you can use 22243 CONDITION to match only when the intermediate results are not used 22244 elsewhere. Use the C expression `dead_or_set_p (INSN, OP)', where INSN 22245 is the insn in which you expect the value to be used for the last time 22246 (from the value of `insn', together with use of `prev_nonnote_insn'), 22247 and OP is the intermediate value (from `operands[I]'). 22248 22249 Applying the optimization means replacing the sequence of insns with 22250 one new insn. The TEMPLATE controls ultimate output of assembler code 22251 for this combined insn. It works exactly like the template of a 22252 `define_insn'. Operand numbers in this template are the same ones used 22253 in matching the original sequence of insns. 22254 22255 The result of a defined peephole optimizer does not need to match any 22256 of the insn patterns in the machine description; it does not even have 22257 an opportunity to match them. The peephole optimizer definition itself 22258 serves as the insn pattern to control how the insn is output. 22259 22260 Defined peephole optimizers are run as assembler code is being output, 22261 so the insns they produce are never combined or rearranged in any way. 22262 22263 Here is an example, taken from the 68000 machine description: 22264 22265 (define_peephole 22266 [(set (reg:SI 15) (plus:SI (reg:SI 15) (const_int 4))) 22267 (set (match_operand:DF 0 "register_operand" "=f") 22268 (match_operand:DF 1 "register_operand" "ad"))] 22269 "FP_REG_P (operands[0]) && ! FP_REG_P (operands[1])" 22270 { 22271 rtx xoperands[2]; 22272 xoperands[1] = gen_rtx_REG (SImode, REGNO (operands[1]) + 1); 22273 #ifdef MOTOROLA 22274 output_asm_insn ("move.l %1,(sp)", xoperands); 22275 output_asm_insn ("move.l %1,-(sp)", operands); 22276 return "fmove.d (sp)+,%0"; 22277 #else 22278 output_asm_insn ("movel %1,sp@", xoperands); 22279 output_asm_insn ("movel %1,sp@-", operands); 22280 return "fmoved sp@+,%0"; 22281 #endif 22282 }) 22283 22284 The effect of this optimization is to change 22285 22286 jbsr _foobar 22287 addql #4,sp 22288 movel d1,sp@- 22289 movel d0,sp@- 22290 fmoved sp@+,fp0 22291 22292 into 22293 22294 jbsr _foobar 22295 movel d1,sp@ 22296 movel d0,sp@- 22297 fmoved sp@+,fp0 22298 22299 INSN-PATTERN-1 and so on look _almost_ like the second operand of 22300 `define_insn'. There is one important difference: the second operand 22301 of `define_insn' consists of one or more RTX's enclosed in square 22302 brackets. Usually, there is only one: then the same action can be 22303 written as an element of a `define_peephole'. But when there are 22304 multiple actions in a `define_insn', they are implicitly enclosed in a 22305 `parallel'. Then you must explicitly write the `parallel', and the 22306 square brackets within it, in the `define_peephole'. Thus, if an insn 22307 pattern looks like this, 22308 22309 (define_insn "divmodsi4" 22310 [(set (match_operand:SI 0 "general_operand" "=d") 22311 (div:SI (match_operand:SI 1 "general_operand" "0") 22312 (match_operand:SI 2 "general_operand" "dmsK"))) 22313 (set (match_operand:SI 3 "general_operand" "=d") 22314 (mod:SI (match_dup 1) (match_dup 2)))] 22315 "TARGET_68020" 22316 "divsl%.l %2,%3:%0") 22317 22318 then the way to mention this insn in a peephole is as follows: 22319 22320 (define_peephole 22321 [... 22322 (parallel 22323 [(set (match_operand:SI 0 "general_operand" "=d") 22324 (div:SI (match_operand:SI 1 "general_operand" "0") 22325 (match_operand:SI 2 "general_operand" "dmsK"))) 22326 (set (match_operand:SI 3 "general_operand" "=d") 22327 (mod:SI (match_dup 1) (match_dup 2)))]) 22328 ...] 22329 ...) 22330 22331 22332 File: gccint.info, Node: define_peephole2, Prev: define_peephole, Up: Peephole Definitions 22333 22334 16.18.2 RTL to RTL Peephole Optimizers 22335 -------------------------------------- 22336 22337 The `define_peephole2' definition tells the compiler how to substitute 22338 one sequence of instructions for another sequence, what additional 22339 scratch registers may be needed and what their lifetimes must be. 22340 22341 (define_peephole2 22342 [INSN-PATTERN-1 22343 INSN-PATTERN-2 22344 ...] 22345 "CONDITION" 22346 [NEW-INSN-PATTERN-1 22347 NEW-INSN-PATTERN-2 22348 ...] 22349 "PREPARATION-STATEMENTS") 22350 22351 The definition is almost identical to `define_split' (*note Insn 22352 Splitting::) except that the pattern to match is not a single 22353 instruction, but a sequence of instructions. 22354 22355 It is possible to request additional scratch registers for use in the 22356 output template. If appropriate registers are not free, the pattern 22357 will simply not match. 22358 22359 Scratch registers are requested with a `match_scratch' pattern at the 22360 top level of the input pattern. The allocated register (initially) will 22361 be dead at the point requested within the original sequence. If the 22362 scratch is used at more than a single point, a `match_dup' pattern at 22363 the top level of the input pattern marks the last position in the input 22364 sequence at which the register must be available. 22365 22366 Here is an example from the IA-32 machine description: 22367 22368 (define_peephole2 22369 [(match_scratch:SI 2 "r") 22370 (parallel [(set (match_operand:SI 0 "register_operand" "") 22371 (match_operator:SI 3 "arith_or_logical_operator" 22372 [(match_dup 0) 22373 (match_operand:SI 1 "memory_operand" "")])) 22374 (clobber (reg:CC 17))])] 22375 "! optimize_size && ! TARGET_READ_MODIFY" 22376 [(set (match_dup 2) (match_dup 1)) 22377 (parallel [(set (match_dup 0) 22378 (match_op_dup 3 [(match_dup 0) (match_dup 2)])) 22379 (clobber (reg:CC 17))])] 22380 "") 22381 22382 This pattern tries to split a load from its use in the hopes that we'll 22383 be able to schedule around the memory load latency. It allocates a 22384 single `SImode' register of class `GENERAL_REGS' (`"r"') that needs to 22385 be live only at the point just before the arithmetic. 22386 22387 A real example requiring extended scratch lifetimes is harder to come 22388 by, so here's a silly made-up example: 22389 22390 (define_peephole2 22391 [(match_scratch:SI 4 "r") 22392 (set (match_operand:SI 0 "" "") (match_operand:SI 1 "" "")) 22393 (set (match_operand:SI 2 "" "") (match_dup 1)) 22394 (match_dup 4) 22395 (set (match_operand:SI 3 "" "") (match_dup 1))] 22396 "/* determine 1 does not overlap 0 and 2 */" 22397 [(set (match_dup 4) (match_dup 1)) 22398 (set (match_dup 0) (match_dup 4)) 22399 (set (match_dup 2) (match_dup 4))] 22400 (set (match_dup 3) (match_dup 4))] 22401 "") 22402 22403 If we had not added the `(match_dup 4)' in the middle of the input 22404 sequence, it might have been the case that the register we chose at the 22405 beginning of the sequence is killed by the first or second `set'. 22406 22407 22408 File: gccint.info, Node: Insn Attributes, Next: Conditional Execution, Prev: Peephole Definitions, Up: Machine Desc 22409 22410 16.19 Instruction Attributes 22411 ============================ 22412 22413 In addition to describing the instruction supported by the target 22414 machine, the `md' file also defines a group of "attributes" and a set of 22415 values for each. Every generated insn is assigned a value for each 22416 attribute. One possible attribute would be the effect that the insn 22417 has on the machine's condition code. This attribute can then be used 22418 by `NOTICE_UPDATE_CC' to track the condition codes. 22419 22420 * Menu: 22421 22422 * Defining Attributes:: Specifying attributes and their values. 22423 * Expressions:: Valid expressions for attribute values. 22424 * Tagging Insns:: Assigning attribute values to insns. 22425 * Attr Example:: An example of assigning attributes. 22426 * Insn Lengths:: Computing the length of insns. 22427 * Constant Attributes:: Defining attributes that are constant. 22428 * Delay Slots:: Defining delay slots required for a machine. 22429 * Processor pipeline description:: Specifying information for insn scheduling. 22430 22431 22432 File: gccint.info, Node: Defining Attributes, Next: Expressions, Up: Insn Attributes 22433 22434 16.19.1 Defining Attributes and their Values 22435 -------------------------------------------- 22436 22437 The `define_attr' expression is used to define each attribute required 22438 by the target machine. It looks like: 22439 22440 (define_attr NAME LIST-OF-VALUES DEFAULT) 22441 22442 NAME is a string specifying the name of the attribute being defined. 22443 22444 LIST-OF-VALUES is either a string that specifies a comma-separated 22445 list of values that can be assigned to the attribute, or a null string 22446 to indicate that the attribute takes numeric values. 22447 22448 DEFAULT is an attribute expression that gives the value of this 22449 attribute for insns that match patterns whose definition does not 22450 include an explicit value for this attribute. *Note Attr Example::, 22451 for more information on the handling of defaults. *Note Constant 22452 Attributes::, for information on attributes that do not depend on any 22453 particular insn. 22454 22455 For each defined attribute, a number of definitions are written to the 22456 `insn-attr.h' file. For cases where an explicit set of values is 22457 specified for an attribute, the following are defined: 22458 22459 * A `#define' is written for the symbol `HAVE_ATTR_NAME'. 22460 22461 * An enumerated class is defined for `attr_NAME' with elements of 22462 the form `UPPER-NAME_UPPER-VALUE' where the attribute name and 22463 value are first converted to uppercase. 22464 22465 * A function `get_attr_NAME' is defined that is passed an insn and 22466 returns the attribute value for that insn. 22467 22468 For example, if the following is present in the `md' file: 22469 22470 (define_attr "type" "branch,fp,load,store,arith" ...) 22471 22472 the following lines will be written to the file `insn-attr.h'. 22473 22474 #define HAVE_ATTR_type 22475 enum attr_type {TYPE_BRANCH, TYPE_FP, TYPE_LOAD, 22476 TYPE_STORE, TYPE_ARITH}; 22477 extern enum attr_type get_attr_type (); 22478 22479 If the attribute takes numeric values, no `enum' type will be defined 22480 and the function to obtain the attribute's value will return `int'. 22481 22482 There are attributes which are tied to a specific meaning. These 22483 attributes are not free to use for other purposes: 22484 22485 `length' 22486 The `length' attribute is used to calculate the length of emitted 22487 code chunks. This is especially important when verifying branch 22488 distances. *Note Insn Lengths::. 22489 22490 `enabled' 22491 The `enabled' attribute can be defined to prevent certain 22492 alternatives of an insn definition from being used during code 22493 generation. *Note Disable Insn Alternatives::. 22494 22495 22496 22497 File: gccint.info, Node: Expressions, Next: Tagging Insns, Prev: Defining Attributes, Up: Insn Attributes 22498 22499 16.19.2 Attribute Expressions 22500 ----------------------------- 22501 22502 RTL expressions used to define attributes use the codes described above 22503 plus a few specific to attribute definitions, to be discussed below. 22504 Attribute value expressions must have one of the following forms: 22505 22506 `(const_int I)' 22507 The integer I specifies the value of a numeric attribute. I must 22508 be non-negative. 22509 22510 The value of a numeric attribute can be specified either with a 22511 `const_int', or as an integer represented as a string in 22512 `const_string', `eq_attr' (see below), `attr', `symbol_ref', 22513 simple arithmetic expressions, and `set_attr' overrides on 22514 specific instructions (*note Tagging Insns::). 22515 22516 `(const_string VALUE)' 22517 The string VALUE specifies a constant attribute value. If VALUE 22518 is specified as `"*"', it means that the default value of the 22519 attribute is to be used for the insn containing this expression. 22520 `"*"' obviously cannot be used in the DEFAULT expression of a 22521 `define_attr'. 22522 22523 If the attribute whose value is being specified is numeric, VALUE 22524 must be a string containing a non-negative integer (normally 22525 `const_int' would be used in this case). Otherwise, it must 22526 contain one of the valid values for the attribute. 22527 22528 `(if_then_else TEST TRUE-VALUE FALSE-VALUE)' 22529 TEST specifies an attribute test, whose format is defined below. 22530 The value of this expression is TRUE-VALUE if TEST is true, 22531 otherwise it is FALSE-VALUE. 22532 22533 `(cond [TEST1 VALUE1 ...] DEFAULT)' 22534 The first operand of this expression is a vector containing an even 22535 number of expressions and consisting of pairs of TEST and VALUE 22536 expressions. The value of the `cond' expression is that of the 22537 VALUE corresponding to the first true TEST expression. If none of 22538 the TEST expressions are true, the value of the `cond' expression 22539 is that of the DEFAULT expression. 22540 22541 TEST expressions can have one of the following forms: 22542 22543 `(const_int I)' 22544 This test is true if I is nonzero and false otherwise. 22545 22546 `(not TEST)' 22547 `(ior TEST1 TEST2)' 22548 `(and TEST1 TEST2)' 22549 These tests are true if the indicated logical function is true. 22550 22551 `(match_operand:M N PRED CONSTRAINTS)' 22552 This test is true if operand N of the insn whose attribute value 22553 is being determined has mode M (this part of the test is ignored 22554 if M is `VOIDmode') and the function specified by the string PRED 22555 returns a nonzero value when passed operand N and mode M (this 22556 part of the test is ignored if PRED is the null string). 22557 22558 The CONSTRAINTS operand is ignored and should be the null string. 22559 22560 `(le ARITH1 ARITH2)' 22561 `(leu ARITH1 ARITH2)' 22562 `(lt ARITH1 ARITH2)' 22563 `(ltu ARITH1 ARITH2)' 22564 `(gt ARITH1 ARITH2)' 22565 `(gtu ARITH1 ARITH2)' 22566 `(ge ARITH1 ARITH2)' 22567 `(geu ARITH1 ARITH2)' 22568 `(ne ARITH1 ARITH2)' 22569 `(eq ARITH1 ARITH2)' 22570 These tests are true if the indicated comparison of the two 22571 arithmetic expressions is true. Arithmetic expressions are formed 22572 with `plus', `minus', `mult', `div', `mod', `abs', `neg', `and', 22573 `ior', `xor', `not', `ashift', `lshiftrt', and `ashiftrt' 22574 expressions. 22575 22576 `const_int' and `symbol_ref' are always valid terms (*note Insn 22577 Lengths::,for additional forms). `symbol_ref' is a string 22578 denoting a C expression that yields an `int' when evaluated by the 22579 `get_attr_...' routine. It should normally be a global variable. 22580 22581 `(eq_attr NAME VALUE)' 22582 NAME is a string specifying the name of an attribute. 22583 22584 VALUE is a string that is either a valid value for attribute NAME, 22585 a comma-separated list of values, or `!' followed by a value or 22586 list. If VALUE does not begin with a `!', this test is true if 22587 the value of the NAME attribute of the current insn is in the list 22588 specified by VALUE. If VALUE begins with a `!', this test is true 22589 if the attribute's value is _not_ in the specified list. 22590 22591 For example, 22592 22593 (eq_attr "type" "load,store") 22594 22595 is equivalent to 22596 22597 (ior (eq_attr "type" "load") (eq_attr "type" "store")) 22598 22599 If NAME specifies an attribute of `alternative', it refers to the 22600 value of the compiler variable `which_alternative' (*note Output 22601 Statement::) and the values must be small integers. For example, 22602 22603 (eq_attr "alternative" "2,3") 22604 22605 is equivalent to 22606 22607 (ior (eq (symbol_ref "which_alternative") (const_int 2)) 22608 (eq (symbol_ref "which_alternative") (const_int 3))) 22609 22610 Note that, for most attributes, an `eq_attr' test is simplified in 22611 cases where the value of the attribute being tested is known for 22612 all insns matching a particular pattern. This is by far the most 22613 common case. 22614 22615 `(attr_flag NAME)' 22616 The value of an `attr_flag' expression is true if the flag 22617 specified by NAME is true for the `insn' currently being scheduled. 22618 22619 NAME is a string specifying one of a fixed set of flags to test. 22620 Test the flags `forward' and `backward' to determine the direction 22621 of a conditional branch. Test the flags `very_likely', `likely', 22622 `very_unlikely', and `unlikely' to determine if a conditional 22623 branch is expected to be taken. 22624 22625 If the `very_likely' flag is true, then the `likely' flag is also 22626 true. Likewise for the `very_unlikely' and `unlikely' flags. 22627 22628 This example describes a conditional branch delay slot which can 22629 be nullified for forward branches that are taken (annul-true) or 22630 for backward branches which are not taken (annul-false). 22631 22632 (define_delay (eq_attr "type" "cbranch") 22633 [(eq_attr "in_branch_delay" "true") 22634 (and (eq_attr "in_branch_delay" "true") 22635 (attr_flag "forward")) 22636 (and (eq_attr "in_branch_delay" "true") 22637 (attr_flag "backward"))]) 22638 22639 The `forward' and `backward' flags are false if the current `insn' 22640 being scheduled is not a conditional branch. 22641 22642 The `very_likely' and `likely' flags are true if the `insn' being 22643 scheduled is not a conditional branch. The `very_unlikely' and 22644 `unlikely' flags are false if the `insn' being scheduled is not a 22645 conditional branch. 22646 22647 `attr_flag' is only used during delay slot scheduling and has no 22648 meaning to other passes of the compiler. 22649 22650 `(attr NAME)' 22651 The value of another attribute is returned. This is most useful 22652 for numeric attributes, as `eq_attr' and `attr_flag' produce more 22653 efficient code for non-numeric attributes. 22654 22655 22656 File: gccint.info, Node: Tagging Insns, Next: Attr Example, Prev: Expressions, Up: Insn Attributes 22657 22658 16.19.3 Assigning Attribute Values to Insns 22659 ------------------------------------------- 22660 22661 The value assigned to an attribute of an insn is primarily determined by 22662 which pattern is matched by that insn (or which `define_peephole' 22663 generated it). Every `define_insn' and `define_peephole' can have an 22664 optional last argument to specify the values of attributes for matching 22665 insns. The value of any attribute not specified in a particular insn 22666 is set to the default value for that attribute, as specified in its 22667 `define_attr'. Extensive use of default values for attributes permits 22668 the specification of the values for only one or two attributes in the 22669 definition of most insn patterns, as seen in the example in the next 22670 section. 22671 22672 The optional last argument of `define_insn' and `define_peephole' is a 22673 vector of expressions, each of which defines the value for a single 22674 attribute. The most general way of assigning an attribute's value is 22675 to use a `set' expression whose first operand is an `attr' expression 22676 giving the name of the attribute being set. The second operand of the 22677 `set' is an attribute expression (*note Expressions::) giving the value 22678 of the attribute. 22679 22680 When the attribute value depends on the `alternative' attribute (i.e., 22681 which is the applicable alternative in the constraint of the insn), the 22682 `set_attr_alternative' expression can be used. It allows the 22683 specification of a vector of attribute expressions, one for each 22684 alternative. 22685 22686 When the generality of arbitrary attribute expressions is not required, 22687 the simpler `set_attr' expression can be used, which allows specifying 22688 a string giving either a single attribute value or a list of attribute 22689 values, one for each alternative. 22690 22691 The form of each of the above specifications is shown below. In each 22692 case, NAME is a string specifying the attribute to be set. 22693 22694 `(set_attr NAME VALUE-STRING)' 22695 VALUE-STRING is either a string giving the desired attribute value, 22696 or a string containing a comma-separated list giving the values for 22697 succeeding alternatives. The number of elements must match the 22698 number of alternatives in the constraint of the insn pattern. 22699 22700 Note that it may be useful to specify `*' for some alternative, in 22701 which case the attribute will assume its default value for insns 22702 matching that alternative. 22703 22704 `(set_attr_alternative NAME [VALUE1 VALUE2 ...])' 22705 Depending on the alternative of the insn, the value will be one of 22706 the specified values. This is a shorthand for using a `cond' with 22707 tests on the `alternative' attribute. 22708 22709 `(set (attr NAME) VALUE)' 22710 The first operand of this `set' must be the special RTL expression 22711 `attr', whose sole operand is a string giving the name of the 22712 attribute being set. VALUE is the value of the attribute. 22713 22714 The following shows three different ways of representing the same 22715 attribute value specification: 22716 22717 (set_attr "type" "load,store,arith") 22718 22719 (set_attr_alternative "type" 22720 [(const_string "load") (const_string "store") 22721 (const_string "arith")]) 22722 22723 (set (attr "type") 22724 (cond [(eq_attr "alternative" "1") (const_string "load") 22725 (eq_attr "alternative" "2") (const_string "store")] 22726 (const_string "arith"))) 22727 22728 The `define_asm_attributes' expression provides a mechanism to specify 22729 the attributes assigned to insns produced from an `asm' statement. It 22730 has the form: 22731 22732 (define_asm_attributes [ATTR-SETS]) 22733 22734 where ATTR-SETS is specified the same as for both the `define_insn' and 22735 the `define_peephole' expressions. 22736 22737 These values will typically be the "worst case" attribute values. For 22738 example, they might indicate that the condition code will be clobbered. 22739 22740 A specification for a `length' attribute is handled specially. The 22741 way to compute the length of an `asm' insn is to multiply the length 22742 specified in the expression `define_asm_attributes' by the number of 22743 machine instructions specified in the `asm' statement, determined by 22744 counting the number of semicolons and newlines in the string. 22745 Therefore, the value of the `length' attribute specified in a 22746 `define_asm_attributes' should be the maximum possible length of a 22747 single machine instruction. 22748 22749 22750 File: gccint.info, Node: Attr Example, Next: Insn Lengths, Prev: Tagging Insns, Up: Insn Attributes 22751 22752 16.19.4 Example of Attribute Specifications 22753 ------------------------------------------- 22754 22755 The judicious use of defaulting is important in the efficient use of 22756 insn attributes. Typically, insns are divided into "types" and an 22757 attribute, customarily called `type', is used to represent this value. 22758 This attribute is normally used only to define the default value for 22759 other attributes. An example will clarify this usage. 22760 22761 Assume we have a RISC machine with a condition code and in which only 22762 full-word operations are performed in registers. Let us assume that we 22763 can divide all insns into loads, stores, (integer) arithmetic 22764 operations, floating point operations, and branches. 22765 22766 Here we will concern ourselves with determining the effect of an insn 22767 on the condition code and will limit ourselves to the following possible 22768 effects: The condition code can be set unpredictably (clobbered), not 22769 be changed, be set to agree with the results of the operation, or only 22770 changed if the item previously set into the condition code has been 22771 modified. 22772 22773 Here is part of a sample `md' file for such a machine: 22774 22775 (define_attr "type" "load,store,arith,fp,branch" (const_string "arith")) 22776 22777 (define_attr "cc" "clobber,unchanged,set,change0" 22778 (cond [(eq_attr "type" "load") 22779 (const_string "change0") 22780 (eq_attr "type" "store,branch") 22781 (const_string "unchanged") 22782 (eq_attr "type" "arith") 22783 (if_then_else (match_operand:SI 0 "" "") 22784 (const_string "set") 22785 (const_string "clobber"))] 22786 (const_string "clobber"))) 22787 22788 (define_insn "" 22789 [(set (match_operand:SI 0 "general_operand" "=r,r,m") 22790 (match_operand:SI 1 "general_operand" "r,m,r"))] 22791 "" 22792 "@ 22793 move %0,%1 22794 load %0,%1 22795 store %0,%1" 22796 [(set_attr "type" "arith,load,store")]) 22797 22798 Note that we assume in the above example that arithmetic operations 22799 performed on quantities smaller than a machine word clobber the 22800 condition code since they will set the condition code to a value 22801 corresponding to the full-word result. 22802 22803 22804 File: gccint.info, Node: Insn Lengths, Next: Constant Attributes, Prev: Attr Example, Up: Insn Attributes 22805 22806 16.19.5 Computing the Length of an Insn 22807 --------------------------------------- 22808 22809 For many machines, multiple types of branch instructions are provided, 22810 each for different length branch displacements. In most cases, the 22811 assembler will choose the correct instruction to use. However, when 22812 the assembler cannot do so, GCC can when a special attribute, the 22813 `length' attribute, is defined. This attribute must be defined to have 22814 numeric values by specifying a null string in its `define_attr'. 22815 22816 In the case of the `length' attribute, two additional forms of 22817 arithmetic terms are allowed in test expressions: 22818 22819 `(match_dup N)' 22820 This refers to the address of operand N of the current insn, which 22821 must be a `label_ref'. 22822 22823 `(pc)' 22824 This refers to the address of the _current_ insn. It might have 22825 been more consistent with other usage to make this the address of 22826 the _next_ insn but this would be confusing because the length of 22827 the current insn is to be computed. 22828 22829 For normal insns, the length will be determined by value of the 22830 `length' attribute. In the case of `addr_vec' and `addr_diff_vec' insn 22831 patterns, the length is computed as the number of vectors multiplied by 22832 the size of each vector. 22833 22834 Lengths are measured in addressable storage units (bytes). 22835 22836 The following macros can be used to refine the length computation: 22837 22838 `ADJUST_INSN_LENGTH (INSN, LENGTH)' 22839 If defined, modifies the length assigned to instruction INSN as a 22840 function of the context in which it is used. LENGTH is an lvalue 22841 that contains the initially computed length of the insn and should 22842 be updated with the correct length of the insn. 22843 22844 This macro will normally not be required. A case in which it is 22845 required is the ROMP. On this machine, the size of an `addr_vec' 22846 insn must be increased by two to compensate for the fact that 22847 alignment may be required. 22848 22849 The routine that returns `get_attr_length' (the value of the `length' 22850 attribute) can be used by the output routine to determine the form of 22851 the branch instruction to be written, as the example below illustrates. 22852 22853 As an example of the specification of variable-length branches, 22854 consider the IBM 360. If we adopt the convention that a register will 22855 be set to the starting address of a function, we can jump to labels 22856 within 4k of the start using a four-byte instruction. Otherwise, we 22857 need a six-byte sequence to load the address from memory and then 22858 branch to it. 22859 22860 On such a machine, a pattern for a branch instruction might be 22861 specified as follows: 22862 22863 (define_insn "jump" 22864 [(set (pc) 22865 (label_ref (match_operand 0 "" "")))] 22866 "" 22867 { 22868 return (get_attr_length (insn) == 4 22869 ? "b %l0" : "l r15,=a(%l0); br r15"); 22870 } 22871 [(set (attr "length") 22872 (if_then_else (lt (match_dup 0) (const_int 4096)) 22873 (const_int 4) 22874 (const_int 6)))]) 22875 22876 22877 File: gccint.info, Node: Constant Attributes, Next: Delay Slots, Prev: Insn Lengths, Up: Insn Attributes 22878 22879 16.19.6 Constant Attributes 22880 --------------------------- 22881 22882 A special form of `define_attr', where the expression for the default 22883 value is a `const' expression, indicates an attribute that is constant 22884 for a given run of the compiler. Constant attributes may be used to 22885 specify which variety of processor is used. For example, 22886 22887 (define_attr "cpu" "m88100,m88110,m88000" 22888 (const 22889 (cond [(symbol_ref "TARGET_88100") (const_string "m88100") 22890 (symbol_ref "TARGET_88110") (const_string "m88110")] 22891 (const_string "m88000")))) 22892 22893 (define_attr "memory" "fast,slow" 22894 (const 22895 (if_then_else (symbol_ref "TARGET_FAST_MEM") 22896 (const_string "fast") 22897 (const_string "slow")))) 22898 22899 The routine generated for constant attributes has no parameters as it 22900 does not depend on any particular insn. RTL expressions used to define 22901 the value of a constant attribute may use the `symbol_ref' form, but 22902 may not use either the `match_operand' form or `eq_attr' forms 22903 involving insn attributes. 22904 22905 22906 File: gccint.info, Node: Delay Slots, Next: Processor pipeline description, Prev: Constant Attributes, Up: Insn Attributes 22907 22908 16.19.7 Delay Slot Scheduling 22909 ----------------------------- 22910 22911 The insn attribute mechanism can be used to specify the requirements for 22912 delay slots, if any, on a target machine. An instruction is said to 22913 require a "delay slot" if some instructions that are physically after 22914 the instruction are executed as if they were located before it. 22915 Classic examples are branch and call instructions, which often execute 22916 the following instruction before the branch or call is performed. 22917 22918 On some machines, conditional branch instructions can optionally 22919 "annul" instructions in the delay slot. This means that the 22920 instruction will not be executed for certain branch outcomes. Both 22921 instructions that annul if the branch is true and instructions that 22922 annul if the branch is false are supported. 22923 22924 Delay slot scheduling differs from instruction scheduling in that 22925 determining whether an instruction needs a delay slot is dependent only 22926 on the type of instruction being generated, not on data flow between the 22927 instructions. See the next section for a discussion of data-dependent 22928 instruction scheduling. 22929 22930 The requirement of an insn needing one or more delay slots is indicated 22931 via the `define_delay' expression. It has the following form: 22932 22933 (define_delay TEST 22934 [DELAY-1 ANNUL-TRUE-1 ANNUL-FALSE-1 22935 DELAY-2 ANNUL-TRUE-2 ANNUL-FALSE-2 22936 ...]) 22937 22938 TEST is an attribute test that indicates whether this `define_delay' 22939 applies to a particular insn. If so, the number of required delay 22940 slots is determined by the length of the vector specified as the second 22941 argument. An insn placed in delay slot N must satisfy attribute test 22942 DELAY-N. ANNUL-TRUE-N is an attribute test that specifies which insns 22943 may be annulled if the branch is true. Similarly, ANNUL-FALSE-N 22944 specifies which insns in the delay slot may be annulled if the branch 22945 is false. If annulling is not supported for that delay slot, `(nil)' 22946 should be coded. 22947 22948 For example, in the common case where branch and call insns require a 22949 single delay slot, which may contain any insn other than a branch or 22950 call, the following would be placed in the `md' file: 22951 22952 (define_delay (eq_attr "type" "branch,call") 22953 [(eq_attr "type" "!branch,call") (nil) (nil)]) 22954 22955 Multiple `define_delay' expressions may be specified. In this case, 22956 each such expression specifies different delay slot requirements and 22957 there must be no insn for which tests in two `define_delay' expressions 22958 are both true. 22959 22960 For example, if we have a machine that requires one delay slot for 22961 branches but two for calls, no delay slot can contain a branch or call 22962 insn, and any valid insn in the delay slot for the branch can be 22963 annulled if the branch is true, we might represent this as follows: 22964 22965 (define_delay (eq_attr "type" "branch") 22966 [(eq_attr "type" "!branch,call") 22967 (eq_attr "type" "!branch,call") 22968 (nil)]) 22969 22970 (define_delay (eq_attr "type" "call") 22971 [(eq_attr "type" "!branch,call") (nil) (nil) 22972 (eq_attr "type" "!branch,call") (nil) (nil)]) 22973 22974 22975 File: gccint.info, Node: Processor pipeline description, Prev: Delay Slots, Up: Insn Attributes 22976 22977 16.19.8 Specifying processor pipeline description 22978 ------------------------------------------------- 22979 22980 To achieve better performance, most modern processors (super-pipelined, 22981 superscalar RISC, and VLIW processors) have many "functional units" on 22982 which several instructions can be executed simultaneously. An 22983 instruction starts execution if its issue conditions are satisfied. If 22984 not, the instruction is stalled until its conditions are satisfied. 22985 Such "interlock (pipeline) delay" causes interruption of the fetching 22986 of successor instructions (or demands nop instructions, e.g. for some 22987 MIPS processors). 22988 22989 There are two major kinds of interlock delays in modern processors. 22990 The first one is a data dependence delay determining "instruction 22991 latency time". The instruction execution is not started until all 22992 source data have been evaluated by prior instructions (there are more 22993 complex cases when the instruction execution starts even when the data 22994 are not available but will be ready in given time after the instruction 22995 execution start). Taking the data dependence delays into account is 22996 simple. The data dependence (true, output, and anti-dependence) delay 22997 between two instructions is given by a constant. In most cases this 22998 approach is adequate. The second kind of interlock delays is a 22999 reservation delay. The reservation delay means that two instructions 23000 under execution will be in need of shared processors resources, i.e. 23001 buses, internal registers, and/or functional units, which are reserved 23002 for some time. Taking this kind of delay into account is complex 23003 especially for modern RISC processors. 23004 23005 The task of exploiting more processor parallelism is solved by an 23006 instruction scheduler. For a better solution to this problem, the 23007 instruction scheduler has to have an adequate description of the 23008 processor parallelism (or "pipeline description"). GCC machine 23009 descriptions describe processor parallelism and functional unit 23010 reservations for groups of instructions with the aid of "regular 23011 expressions". 23012 23013 The GCC instruction scheduler uses a "pipeline hazard recognizer" to 23014 figure out the possibility of the instruction issue by the processor on 23015 a given simulated processor cycle. The pipeline hazard recognizer is 23016 automatically generated from the processor pipeline description. The 23017 pipeline hazard recognizer generated from the machine description is 23018 based on a deterministic finite state automaton (DFA): the instruction 23019 issue is possible if there is a transition from one automaton state to 23020 another one. This algorithm is very fast, and furthermore, its speed 23021 is not dependent on processor complexity(1). 23022 23023 The rest of this section describes the directives that constitute an 23024 automaton-based processor pipeline description. The order of these 23025 constructions within the machine description file is not important. 23026 23027 The following optional construction describes names of automata 23028 generated and used for the pipeline hazards recognition. Sometimes the 23029 generated finite state automaton used by the pipeline hazard recognizer 23030 is large. If we use more than one automaton and bind functional units 23031 to the automata, the total size of the automata is usually less than 23032 the size of the single automaton. If there is no one such 23033 construction, only one finite state automaton is generated. 23034 23035 (define_automaton AUTOMATA-NAMES) 23036 23037 AUTOMATA-NAMES is a string giving names of the automata. The names 23038 are separated by commas. All the automata should have unique names. 23039 The automaton name is used in the constructions `define_cpu_unit' and 23040 `define_query_cpu_unit'. 23041 23042 Each processor functional unit used in the description of instruction 23043 reservations should be described by the following construction. 23044 23045 (define_cpu_unit UNIT-NAMES [AUTOMATON-NAME]) 23046 23047 UNIT-NAMES is a string giving the names of the functional units 23048 separated by commas. Don't use name `nothing', it is reserved for 23049 other goals. 23050 23051 AUTOMATON-NAME is a string giving the name of the automaton with which 23052 the unit is bound. The automaton should be described in construction 23053 `define_automaton'. You should give "automaton-name", if there is a 23054 defined automaton. 23055 23056 The assignment of units to automata are constrained by the uses of the 23057 units in insn reservations. The most important constraint is: if a 23058 unit reservation is present on a particular cycle of an alternative for 23059 an insn reservation, then some unit from the same automaton must be 23060 present on the same cycle for the other alternatives of the insn 23061 reservation. The rest of the constraints are mentioned in the 23062 description of the subsequent constructions. 23063 23064 The following construction describes CPU functional units analogously 23065 to `define_cpu_unit'. The reservation of such units can be queried for 23066 an automaton state. The instruction scheduler never queries 23067 reservation of functional units for given automaton state. So as a 23068 rule, you don't need this construction. This construction could be 23069 used for future code generation goals (e.g. to generate VLIW insn 23070 templates). 23071 23072 (define_query_cpu_unit UNIT-NAMES [AUTOMATON-NAME]) 23073 23074 UNIT-NAMES is a string giving names of the functional units separated 23075 by commas. 23076 23077 AUTOMATON-NAME is a string giving the name of the automaton with which 23078 the unit is bound. 23079 23080 The following construction is the major one to describe pipeline 23081 characteristics of an instruction. 23082 23083 (define_insn_reservation INSN-NAME DEFAULT_LATENCY 23084 CONDITION REGEXP) 23085 23086 DEFAULT_LATENCY is a number giving latency time of the instruction. 23087 There is an important difference between the old description and the 23088 automaton based pipeline description. The latency time is used for all 23089 dependencies when we use the old description. In the automaton based 23090 pipeline description, the given latency time is only used for true 23091 dependencies. The cost of anti-dependencies is always zero and the 23092 cost of output dependencies is the difference between latency times of 23093 the producing and consuming insns (if the difference is negative, the 23094 cost is considered to be zero). You can always change the default 23095 costs for any description by using the target hook 23096 `TARGET_SCHED_ADJUST_COST' (*note Scheduling::). 23097 23098 INSN-NAME is a string giving the internal name of the insn. The 23099 internal names are used in constructions `define_bypass' and in the 23100 automaton description file generated for debugging. The internal name 23101 has nothing in common with the names in `define_insn'. It is a good 23102 practice to use insn classes described in the processor manual. 23103 23104 CONDITION defines what RTL insns are described by this construction. 23105 You should remember that you will be in trouble if CONDITION for two or 23106 more different `define_insn_reservation' constructions is TRUE for an 23107 insn. In this case what reservation will be used for the insn is not 23108 defined. Such cases are not checked during generation of the pipeline 23109 hazards recognizer because in general recognizing that two conditions 23110 may have the same value is quite difficult (especially if the conditions 23111 contain `symbol_ref'). It is also not checked during the pipeline 23112 hazard recognizer work because it would slow down the recognizer 23113 considerably. 23114 23115 REGEXP is a string describing the reservation of the cpu's functional 23116 units by the instruction. The reservations are described by a regular 23117 expression according to the following syntax: 23118 23119 regexp = regexp "," oneof 23120 | oneof 23121 23122 oneof = oneof "|" allof 23123 | allof 23124 23125 allof = allof "+" repeat 23126 | repeat 23127 23128 repeat = element "*" number 23129 | element 23130 23131 element = cpu_function_unit_name 23132 | reservation_name 23133 | result_name 23134 | "nothing" 23135 | "(" regexp ")" 23136 23137 * `,' is used for describing the start of the next cycle in the 23138 reservation. 23139 23140 * `|' is used for describing a reservation described by the first 23141 regular expression *or* a reservation described by the second 23142 regular expression *or* etc. 23143 23144 * `+' is used for describing a reservation described by the first 23145 regular expression *and* a reservation described by the second 23146 regular expression *and* etc. 23147 23148 * `*' is used for convenience and simply means a sequence in which 23149 the regular expression are repeated NUMBER times with cycle 23150 advancing (see `,'). 23151 23152 * `cpu_function_unit_name' denotes reservation of the named 23153 functional unit. 23154 23155 * `reservation_name' -- see description of construction 23156 `define_reservation'. 23157 23158 * `nothing' denotes no unit reservations. 23159 23160 Sometimes unit reservations for different insns contain common parts. 23161 In such case, you can simplify the pipeline description by describing 23162 the common part by the following construction 23163 23164 (define_reservation RESERVATION-NAME REGEXP) 23165 23166 RESERVATION-NAME is a string giving name of REGEXP. Functional unit 23167 names and reservation names are in the same name space. So the 23168 reservation names should be different from the functional unit names 23169 and can not be the reserved name `nothing'. 23170 23171 The following construction is used to describe exceptions in the 23172 latency time for given instruction pair. This is so called bypasses. 23173 23174 (define_bypass NUMBER OUT_INSN_NAMES IN_INSN_NAMES 23175 [GUARD]) 23176 23177 NUMBER defines when the result generated by the instructions given in 23178 string OUT_INSN_NAMES will be ready for the instructions given in 23179 string IN_INSN_NAMES. The instructions in the string are separated by 23180 commas. 23181 23182 GUARD is an optional string giving the name of a C function which 23183 defines an additional guard for the bypass. The function will get the 23184 two insns as parameters. If the function returns zero the bypass will 23185 be ignored for this case. The additional guard is necessary to 23186 recognize complicated bypasses, e.g. when the consumer is only an 23187 address of insn `store' (not a stored value). 23188 23189 If there are more one bypass with the same output and input insns, the 23190 chosen bypass is the first bypass with a guard in description whose 23191 guard function returns nonzero. If there is no such bypass, then 23192 bypass without the guard function is chosen. 23193 23194 The following five constructions are usually used to describe VLIW 23195 processors, or more precisely, to describe a placement of small 23196 instructions into VLIW instruction slots. They can be used for RISC 23197 processors, too. 23198 23199 (exclusion_set UNIT-NAMES UNIT-NAMES) 23200 (presence_set UNIT-NAMES PATTERNS) 23201 (final_presence_set UNIT-NAMES PATTERNS) 23202 (absence_set UNIT-NAMES PATTERNS) 23203 (final_absence_set UNIT-NAMES PATTERNS) 23204 23205 UNIT-NAMES is a string giving names of functional units separated by 23206 commas. 23207 23208 PATTERNS is a string giving patterns of functional units separated by 23209 comma. Currently pattern is one unit or units separated by 23210 white-spaces. 23211 23212 The first construction (`exclusion_set') means that each functional 23213 unit in the first string can not be reserved simultaneously with a unit 23214 whose name is in the second string and vice versa. For example, the 23215 construction is useful for describing processors (e.g. some SPARC 23216 processors) with a fully pipelined floating point functional unit which 23217 can execute simultaneously only single floating point insns or only 23218 double floating point insns. 23219 23220 The second construction (`presence_set') means that each functional 23221 unit in the first string can not be reserved unless at least one of 23222 pattern of units whose names are in the second string is reserved. 23223 This is an asymmetric relation. For example, it is useful for 23224 description that VLIW `slot1' is reserved after `slot0' reservation. 23225 We could describe it by the following construction 23226 23227 (presence_set "slot1" "slot0") 23228 23229 Or `slot1' is reserved only after `slot0' and unit `b0' reservation. 23230 In this case we could write 23231 23232 (presence_set "slot1" "slot0 b0") 23233 23234 The third construction (`final_presence_set') is analogous to 23235 `presence_set'. The difference between them is when checking is done. 23236 When an instruction is issued in given automaton state reflecting all 23237 current and planned unit reservations, the automaton state is changed. 23238 The first state is a source state, the second one is a result state. 23239 Checking for `presence_set' is done on the source state reservation, 23240 checking for `final_presence_set' is done on the result reservation. 23241 This construction is useful to describe a reservation which is actually 23242 two subsequent reservations. For example, if we use 23243 23244 (presence_set "slot1" "slot0") 23245 23246 the following insn will be never issued (because `slot1' requires 23247 `slot0' which is absent in the source state). 23248 23249 (define_reservation "insn_and_nop" "slot0 + slot1") 23250 23251 but it can be issued if we use analogous `final_presence_set'. 23252 23253 The forth construction (`absence_set') means that each functional unit 23254 in the first string can be reserved only if each pattern of units whose 23255 names are in the second string is not reserved. This is an asymmetric 23256 relation (actually `exclusion_set' is analogous to this one but it is 23257 symmetric). For example it might be useful in a VLIW description to 23258 say that `slot0' cannot be reserved after either `slot1' or `slot2' 23259 have been reserved. This can be described as: 23260 23261 (absence_set "slot0" "slot1, slot2") 23262 23263 Or `slot2' can not be reserved if `slot0' and unit `b0' are reserved 23264 or `slot1' and unit `b1' are reserved. In this case we could write 23265 23266 (absence_set "slot2" "slot0 b0, slot1 b1") 23267 23268 All functional units mentioned in a set should belong to the same 23269 automaton. 23270 23271 The last construction (`final_absence_set') is analogous to 23272 `absence_set' but checking is done on the result (state) reservation. 23273 See comments for `final_presence_set'. 23274 23275 You can control the generator of the pipeline hazard recognizer with 23276 the following construction. 23277 23278 (automata_option OPTIONS) 23279 23280 OPTIONS is a string giving options which affect the generated code. 23281 Currently there are the following options: 23282 23283 * "no-minimization" makes no minimization of the automaton. This is 23284 only worth to do when we are debugging the description and need to 23285 look more accurately at reservations of states. 23286 23287 * "time" means printing time statistics about the generation of 23288 automata. 23289 23290 * "stats" means printing statistics about the generated automata 23291 such as the number of DFA states, NDFA states and arcs. 23292 23293 * "v" means a generation of the file describing the result automata. 23294 The file has suffix `.dfa' and can be used for the description 23295 verification and debugging. 23296 23297 * "w" means a generation of warning instead of error for 23298 non-critical errors. 23299 23300 * "ndfa" makes nondeterministic finite state automata. This affects 23301 the treatment of operator `|' in the regular expressions. The 23302 usual treatment of the operator is to try the first alternative 23303 and, if the reservation is not possible, the second alternative. 23304 The nondeterministic treatment means trying all alternatives, some 23305 of them may be rejected by reservations in the subsequent insns. 23306 23307 * "progress" means output of a progress bar showing how many states 23308 were generated so far for automaton being processed. This is 23309 useful during debugging a DFA description. If you see too many 23310 generated states, you could interrupt the generator of the pipeline 23311 hazard recognizer and try to figure out a reason for generation of 23312 the huge automaton. 23313 23314 As an example, consider a superscalar RISC machine which can issue 23315 three insns (two integer insns and one floating point insn) on the 23316 cycle but can finish only two insns. To describe this, we define the 23317 following functional units. 23318 23319 (define_cpu_unit "i0_pipeline, i1_pipeline, f_pipeline") 23320 (define_cpu_unit "port0, port1") 23321 23322 All simple integer insns can be executed in any integer pipeline and 23323 their result is ready in two cycles. The simple integer insns are 23324 issued into the first pipeline unless it is reserved, otherwise they 23325 are issued into the second pipeline. Integer division and 23326 multiplication insns can be executed only in the second integer 23327 pipeline and their results are ready correspondingly in 8 and 4 cycles. 23328 The integer division is not pipelined, i.e. the subsequent integer 23329 division insn can not be issued until the current division insn 23330 finished. Floating point insns are fully pipelined and their results 23331 are ready in 3 cycles. Where the result of a floating point insn is 23332 used by an integer insn, an additional delay of one cycle is incurred. 23333 To describe all of this we could specify 23334 23335 (define_cpu_unit "div") 23336 23337 (define_insn_reservation "simple" 2 (eq_attr "type" "int") 23338 "(i0_pipeline | i1_pipeline), (port0 | port1)") 23339 23340 (define_insn_reservation "mult" 4 (eq_attr "type" "mult") 23341 "i1_pipeline, nothing*2, (port0 | port1)") 23342 23343 (define_insn_reservation "div" 8 (eq_attr "type" "div") 23344 "i1_pipeline, div*7, div + (port0 | port1)") 23345 23346 (define_insn_reservation "float" 3 (eq_attr "type" "float") 23347 "f_pipeline, nothing, (port0 | port1)) 23348 23349 (define_bypass 4 "float" "simple,mult,div") 23350 23351 To simplify the description we could describe the following reservation 23352 23353 (define_reservation "finish" "port0|port1") 23354 23355 and use it in all `define_insn_reservation' as in the following 23356 construction 23357 23358 (define_insn_reservation "simple" 2 (eq_attr "type" "int") 23359 "(i0_pipeline | i1_pipeline), finish") 23360 23361 ---------- Footnotes ---------- 23362 23363 (1) However, the size of the automaton depends on processor 23364 complexity. To limit this effect, machine descriptions can split 23365 orthogonal parts of the machine description among several automata: but 23366 then, since each of these must be stepped independently, this does 23367 cause a small decrease in the algorithm's performance. 23368 23369 23370 File: gccint.info, Node: Conditional Execution, Next: Constant Definitions, Prev: Insn Attributes, Up: Machine Desc 23371 23372 16.20 Conditional Execution 23373 =========================== 23374 23375 A number of architectures provide for some form of conditional 23376 execution, or predication. The hallmark of this feature is the ability 23377 to nullify most of the instructions in the instruction set. When the 23378 instruction set is large and not entirely symmetric, it can be quite 23379 tedious to describe these forms directly in the `.md' file. An 23380 alternative is the `define_cond_exec' template. 23381 23382 (define_cond_exec 23383 [PREDICATE-PATTERN] 23384 "CONDITION" 23385 "OUTPUT-TEMPLATE") 23386 23387 PREDICATE-PATTERN is the condition that must be true for the insn to 23388 be executed at runtime and should match a relational operator. One can 23389 use `match_operator' to match several relational operators at once. 23390 Any `match_operand' operands must have no more than one alternative. 23391 23392 CONDITION is a C expression that must be true for the generated 23393 pattern to match. 23394 23395 OUTPUT-TEMPLATE is a string similar to the `define_insn' output 23396 template (*note Output Template::), except that the `*' and `@' special 23397 cases do not apply. This is only useful if the assembly text for the 23398 predicate is a simple prefix to the main insn. In order to handle the 23399 general case, there is a global variable `current_insn_predicate' that 23400 will contain the entire predicate if the current insn is predicated, 23401 and will otherwise be `NULL'. 23402 23403 When `define_cond_exec' is used, an implicit reference to the 23404 `predicable' instruction attribute is made. *Note Insn Attributes::. 23405 This attribute must be boolean (i.e. have exactly two elements in its 23406 LIST-OF-VALUES). Further, it must not be used with complex 23407 expressions. That is, the default and all uses in the insns must be a 23408 simple constant, not dependent on the alternative or anything else. 23409 23410 For each `define_insn' for which the `predicable' attribute is true, a 23411 new `define_insn' pattern will be generated that matches a predicated 23412 version of the instruction. For example, 23413 23414 (define_insn "addsi" 23415 [(set (match_operand:SI 0 "register_operand" "r") 23416 (plus:SI (match_operand:SI 1 "register_operand" "r") 23417 (match_operand:SI 2 "register_operand" "r")))] 23418 "TEST1" 23419 "add %2,%1,%0") 23420 23421 (define_cond_exec 23422 [(ne (match_operand:CC 0 "register_operand" "c") 23423 (const_int 0))] 23424 "TEST2" 23425 "(%0)") 23426 23427 generates a new pattern 23428 23429 (define_insn "" 23430 [(cond_exec 23431 (ne (match_operand:CC 3 "register_operand" "c") (const_int 0)) 23432 (set (match_operand:SI 0 "register_operand" "r") 23433 (plus:SI (match_operand:SI 1 "register_operand" "r") 23434 (match_operand:SI 2 "register_operand" "r"))))] 23435 "(TEST2) && (TEST1)" 23436 "(%3) add %2,%1,%0") 23437 23438 23439 File: gccint.info, Node: Constant Definitions, Next: Iterators, Prev: Conditional Execution, Up: Machine Desc 23440 23441 16.21 Constant Definitions 23442 ========================== 23443 23444 Using literal constants inside instruction patterns reduces legibility 23445 and can be a maintenance problem. 23446 23447 To overcome this problem, you may use the `define_constants' 23448 expression. It contains a vector of name-value pairs. From that point 23449 on, wherever any of the names appears in the MD file, it is as if the 23450 corresponding value had been written instead. You may use 23451 `define_constants' multiple times; each appearance adds more constants 23452 to the table. It is an error to redefine a constant with a different 23453 value. 23454 23455 To come back to the a29k load multiple example, instead of 23456 23457 (define_insn "" 23458 [(match_parallel 0 "load_multiple_operation" 23459 [(set (match_operand:SI 1 "gpc_reg_operand" "=r") 23460 (match_operand:SI 2 "memory_operand" "m")) 23461 (use (reg:SI 179)) 23462 (clobber (reg:SI 179))])] 23463 "" 23464 "loadm 0,0,%1,%2") 23465 23466 You could write: 23467 23468 (define_constants [ 23469 (R_BP 177) 23470 (R_FC 178) 23471 (R_CR 179) 23472 (R_Q 180) 23473 ]) 23474 23475 (define_insn "" 23476 [(match_parallel 0 "load_multiple_operation" 23477 [(set (match_operand:SI 1 "gpc_reg_operand" "=r") 23478 (match_operand:SI 2 "memory_operand" "m")) 23479 (use (reg:SI R_CR)) 23480 (clobber (reg:SI R_CR))])] 23481 "" 23482 "loadm 0,0,%1,%2") 23483 23484 The constants that are defined with a define_constant are also output 23485 in the insn-codes.h header file as #defines. 23486 23487 23488 File: gccint.info, Node: Iterators, Prev: Constant Definitions, Up: Machine Desc 23489 23490 16.22 Iterators 23491 =============== 23492 23493 Ports often need to define similar patterns for more than one machine 23494 mode or for more than one rtx code. GCC provides some simple iterator 23495 facilities to make this process easier. 23496 23497 * Menu: 23498 23499 * Mode Iterators:: Generating variations of patterns for different modes. 23500 * Code Iterators:: Doing the same for codes. 23501 23502 23503 File: gccint.info, Node: Mode Iterators, Next: Code Iterators, Up: Iterators 23504 23505 16.22.1 Mode Iterators 23506 ---------------------- 23507 23508 Ports often need to define similar patterns for two or more different 23509 modes. For example: 23510 23511 * If a processor has hardware support for both single and double 23512 floating-point arithmetic, the `SFmode' patterns tend to be very 23513 similar to the `DFmode' ones. 23514 23515 * If a port uses `SImode' pointers in one configuration and `DImode' 23516 pointers in another, it will usually have very similar `SImode' 23517 and `DImode' patterns for manipulating pointers. 23518 23519 Mode iterators allow several patterns to be instantiated from one 23520 `.md' file template. They can be used with any type of rtx-based 23521 construct, such as a `define_insn', `define_split', or 23522 `define_peephole2'. 23523 23524 * Menu: 23525 23526 * Defining Mode Iterators:: Defining a new mode iterator. 23527 * Substitutions:: Combining mode iterators with substitutions 23528 * Examples:: Examples 23529 23530 23531 File: gccint.info, Node: Defining Mode Iterators, Next: Substitutions, Up: Mode Iterators 23532 23533 16.22.1.1 Defining Mode Iterators 23534 ................................. 23535 23536 The syntax for defining a mode iterator is: 23537 23538 (define_mode_iterator NAME [(MODE1 "COND1") ... (MODEN "CONDN")]) 23539 23540 This allows subsequent `.md' file constructs to use the mode suffix 23541 `:NAME'. Every construct that does so will be expanded N times, once 23542 with every use of `:NAME' replaced by `:MODE1', once with every use 23543 replaced by `:MODE2', and so on. In the expansion for a particular 23544 MODEI, every C condition will also require that CONDI be true. 23545 23546 For example: 23547 23548 (define_mode_iterator P [(SI "Pmode == SImode") (DI "Pmode == DImode")]) 23549 23550 defines a new mode suffix `:P'. Every construct that uses `:P' will 23551 be expanded twice, once with every `:P' replaced by `:SI' and once with 23552 every `:P' replaced by `:DI'. The `:SI' version will only apply if 23553 `Pmode == SImode' and the `:DI' version will only apply if `Pmode == 23554 DImode'. 23555 23556 As with other `.md' conditions, an empty string is treated as "always 23557 true". `(MODE "")' can also be abbreviated to `MODE'. For example: 23558 23559 (define_mode_iterator GPR [SI (DI "TARGET_64BIT")]) 23560 23561 means that the `:DI' expansion only applies if `TARGET_64BIT' but that 23562 the `:SI' expansion has no such constraint. 23563 23564 Iterators are applied in the order they are defined. This can be 23565 significant if two iterators are used in a construct that requires 23566 substitutions. *Note Substitutions::. 23567 23568 23569 File: gccint.info, Node: Substitutions, Next: Examples, Prev: Defining Mode Iterators, Up: Mode Iterators 23570 23571 16.22.1.2 Substitution in Mode Iterators 23572 ........................................ 23573 23574 If an `.md' file construct uses mode iterators, each version of the 23575 construct will often need slightly different strings or modes. For 23576 example: 23577 23578 * When a `define_expand' defines several `addM3' patterns (*note 23579 Standard Names::), each expander will need to use the appropriate 23580 mode name for M. 23581 23582 * When a `define_insn' defines several instruction patterns, each 23583 instruction will often use a different assembler mnemonic. 23584 23585 * When a `define_insn' requires operands with different modes, using 23586 an iterator for one of the operand modes usually requires a 23587 specific mode for the other operand(s). 23588 23589 GCC supports such variations through a system of "mode attributes". 23590 There are two standard attributes: `mode', which is the name of the 23591 mode in lower case, and `MODE', which is the same thing in upper case. 23592 You can define other attributes using: 23593 23594 (define_mode_attr NAME [(MODE1 "VALUE1") ... (MODEN "VALUEN")]) 23595 23596 where NAME is the name of the attribute and VALUEI is the value 23597 associated with MODEI. 23598 23599 When GCC replaces some :ITERATOR with :MODE, it will scan each string 23600 and mode in the pattern for sequences of the form `<ITERATOR:ATTR>', 23601 where ATTR is the name of a mode attribute. If the attribute is 23602 defined for MODE, the whole `<...>' sequence will be replaced by the 23603 appropriate attribute value. 23604 23605 For example, suppose an `.md' file has: 23606 23607 (define_mode_iterator P [(SI "Pmode == SImode") (DI "Pmode == DImode")]) 23608 (define_mode_attr load [(SI "lw") (DI "ld")]) 23609 23610 If one of the patterns that uses `:P' contains the string 23611 `"<P:load>\t%0,%1"', the `SI' version of that pattern will use 23612 `"lw\t%0,%1"' and the `DI' version will use `"ld\t%0,%1"'. 23613 23614 Here is an example of using an attribute for a mode: 23615 23616 (define_mode_iterator LONG [SI DI]) 23617 (define_mode_attr SHORT [(SI "HI") (DI "SI")]) 23618 (define_insn ... 23619 (sign_extend:LONG (match_operand:<LONG:SHORT> ...)) ...) 23620 23621 The `ITERATOR:' prefix may be omitted, in which case the substitution 23622 will be attempted for every iterator expansion. 23623 23624 23625 File: gccint.info, Node: Examples, Prev: Substitutions, Up: Mode Iterators 23626 23627 16.22.1.3 Mode Iterator Examples 23628 ................................ 23629 23630 Here is an example from the MIPS port. It defines the following modes 23631 and attributes (among others): 23632 23633 (define_mode_iterator GPR [SI (DI "TARGET_64BIT")]) 23634 (define_mode_attr d [(SI "") (DI "d")]) 23635 23636 and uses the following template to define both `subsi3' and `subdi3': 23637 23638 (define_insn "sub<mode>3" 23639 [(set (match_operand:GPR 0 "register_operand" "=d") 23640 (minus:GPR (match_operand:GPR 1 "register_operand" "d") 23641 (match_operand:GPR 2 "register_operand" "d")))] 23642 "" 23643 "<d>subu\t%0,%1,%2" 23644 [(set_attr "type" "arith") 23645 (set_attr "mode" "<MODE>")]) 23646 23647 This is exactly equivalent to: 23648 23649 (define_insn "subsi3" 23650 [(set (match_operand:SI 0 "register_operand" "=d") 23651 (minus:SI (match_operand:SI 1 "register_operand" "d") 23652 (match_operand:SI 2 "register_operand" "d")))] 23653 "" 23654 "subu\t%0,%1,%2" 23655 [(set_attr "type" "arith") 23656 (set_attr "mode" "SI")]) 23657 23658 (define_insn "subdi3" 23659 [(set (match_operand:DI 0 "register_operand" "=d") 23660 (minus:DI (match_operand:DI 1 "register_operand" "d") 23661 (match_operand:DI 2 "register_operand" "d")))] 23662 "" 23663 "dsubu\t%0,%1,%2" 23664 [(set_attr "type" "arith") 23665 (set_attr "mode" "DI")]) 23666 23667 23668 File: gccint.info, Node: Code Iterators, Prev: Mode Iterators, Up: Iterators 23669 23670 16.22.2 Code Iterators 23671 ---------------------- 23672 23673 Code iterators operate in a similar way to mode iterators. *Note Mode 23674 Iterators::. 23675 23676 The construct: 23677 23678 (define_code_iterator NAME [(CODE1 "COND1") ... (CODEN "CONDN")]) 23679 23680 defines a pseudo rtx code NAME that can be instantiated as CODEI if 23681 condition CONDI is true. Each CODEI must have the same rtx format. 23682 *Note RTL Classes::. 23683 23684 As with mode iterators, each pattern that uses NAME will be expanded N 23685 times, once with all uses of NAME replaced by CODE1, once with all uses 23686 replaced by CODE2, and so on. *Note Defining Mode Iterators::. 23687 23688 It is possible to define attributes for codes as well as for modes. 23689 There are two standard code attributes: `code', the name of the code in 23690 lower case, and `CODE', the name of the code in upper case. Other 23691 attributes are defined using: 23692 23693 (define_code_attr NAME [(CODE1 "VALUE1") ... (CODEN "VALUEN")]) 23694 23695 Here's an example of code iterators in action, taken from the MIPS 23696 port: 23697 23698 (define_code_iterator any_cond [unordered ordered unlt unge uneq ltgt unle ungt 23699 eq ne gt ge lt le gtu geu ltu leu]) 23700 23701 (define_expand "b<code>" 23702 [(set (pc) 23703 (if_then_else (any_cond:CC (cc0) 23704 (const_int 0)) 23705 (label_ref (match_operand 0 "")) 23706 (pc)))] 23707 "" 23708 { 23709 gen_conditional_branch (operands, <CODE>); 23710 DONE; 23711 }) 23712 23713 This is equivalent to: 23714 23715 (define_expand "bunordered" 23716 [(set (pc) 23717 (if_then_else (unordered:CC (cc0) 23718 (const_int 0)) 23719 (label_ref (match_operand 0 "")) 23720 (pc)))] 23721 "" 23722 { 23723 gen_conditional_branch (operands, UNORDERED); 23724 DONE; 23725 }) 23726 23727 (define_expand "bordered" 23728 [(set (pc) 23729 (if_then_else (ordered:CC (cc0) 23730 (const_int 0)) 23731 (label_ref (match_operand 0 "")) 23732 (pc)))] 23733 "" 23734 { 23735 gen_conditional_branch (operands, ORDERED); 23736 DONE; 23737 }) 23738 23739 ... 23740 23741 23742 File: gccint.info, Node: Target Macros, Next: Host Config, Prev: Machine Desc, Up: Top 23743 23744 17 Target Description Macros and Functions 23745 ****************************************** 23746 23747 In addition to the file `MACHINE.md', a machine description includes a 23748 C header file conventionally given the name `MACHINE.h' and a C source 23749 file named `MACHINE.c'. The header file defines numerous macros that 23750 convey the information about the target machine that does not fit into 23751 the scheme of the `.md' file. The file `tm.h' should be a link to 23752 `MACHINE.h'. The header file `config.h' includes `tm.h' and most 23753 compiler source files include `config.h'. The source file defines a 23754 variable `targetm', which is a structure containing pointers to 23755 functions and data relating to the target machine. `MACHINE.c' should 23756 also contain their definitions, if they are not defined elsewhere in 23757 GCC, and other functions called through the macros defined in the `.h' 23758 file. 23759 23760 * Menu: 23761 23762 * Target Structure:: The `targetm' variable. 23763 * Driver:: Controlling how the driver runs the compilation passes. 23764 * Run-time Target:: Defining `-m' options like `-m68000' and `-m68020'. 23765 * Per-Function Data:: Defining data structures for per-function information. 23766 * Storage Layout:: Defining sizes and alignments of data. 23767 * Type Layout:: Defining sizes and properties of basic user data types. 23768 * Registers:: Naming and describing the hardware registers. 23769 * Register Classes:: Defining the classes of hardware registers. 23770 * Old Constraints:: The old way to define machine-specific constraints. 23771 * Stack and Calling:: Defining which way the stack grows and by how much. 23772 * Varargs:: Defining the varargs macros. 23773 * Trampolines:: Code set up at run time to enter a nested function. 23774 * Library Calls:: Controlling how library routines are implicitly called. 23775 * Addressing Modes:: Defining addressing modes valid for memory operands. 23776 * Anchored Addresses:: Defining how `-fsection-anchors' should work. 23777 * Condition Code:: Defining how insns update the condition code. 23778 * Costs:: Defining relative costs of different operations. 23779 * Scheduling:: Adjusting the behavior of the instruction scheduler. 23780 * Sections:: Dividing storage into text, data, and other sections. 23781 * PIC:: Macros for position independent code. 23782 * Assembler Format:: Defining how to write insns and pseudo-ops to output. 23783 * Debugging Info:: Defining the format of debugging output. 23784 * Floating Point:: Handling floating point for cross-compilers. 23785 * Mode Switching:: Insertion of mode-switching instructions. 23786 * Target Attributes:: Defining target-specific uses of `__attribute__'. 23787 * Emulated TLS:: Emulated TLS support. 23788 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it. 23789 * PCH Target:: Validity checking for precompiled headers. 23790 * C++ ABI:: Controlling C++ ABI changes. 23791 * Misc:: Everything else. 23792 23793 23794 File: gccint.info, Node: Target Structure, Next: Driver, Up: Target Macros 23795 23796 17.1 The Global `targetm' Variable 23797 ================================== 23798 23799 -- Variable: struct gcc_target targetm 23800 The target `.c' file must define the global `targetm' variable 23801 which contains pointers to functions and data relating to the 23802 target machine. The variable is declared in `target.h'; 23803 `target-def.h' defines the macro `TARGET_INITIALIZER' which is 23804 used to initialize the variable, and macros for the default 23805 initializers for elements of the structure. The `.c' file should 23806 override those macros for which the default definition is 23807 inappropriate. For example: 23808 #include "target.h" 23809 #include "target-def.h" 23810 23811 /* Initialize the GCC target structure. */ 23812 23813 #undef TARGET_COMP_TYPE_ATTRIBUTES 23814 #define TARGET_COMP_TYPE_ATTRIBUTES MACHINE_comp_type_attributes 23815 23816 struct gcc_target targetm = TARGET_INITIALIZER; 23817 23818 Where a macro should be defined in the `.c' file in this manner to form 23819 part of the `targetm' structure, it is documented below as a "Target 23820 Hook" with a prototype. Many macros will change in future from being 23821 defined in the `.h' file to being part of the `targetm' structure. 23822 23823 23824 File: gccint.info, Node: Driver, Next: Run-time Target, Prev: Target Structure, Up: Target Macros 23825 23826 17.2 Controlling the Compilation Driver, `gcc' 23827 ============================================== 23828 23829 You can control the compilation driver. 23830 23831 -- Macro: SWITCH_TAKES_ARG (CHAR) 23832 A C expression which determines whether the option `-CHAR' takes 23833 arguments. The value should be the number of arguments that 23834 option takes-zero, for many options. 23835 23836 By default, this macro is defined as `DEFAULT_SWITCH_TAKES_ARG', 23837 which handles the standard options properly. You need not define 23838 `SWITCH_TAKES_ARG' unless you wish to add additional options which 23839 take arguments. Any redefinition should call 23840 `DEFAULT_SWITCH_TAKES_ARG' and then check for additional options. 23841 23842 -- Macro: WORD_SWITCH_TAKES_ARG (NAME) 23843 A C expression which determines whether the option `-NAME' takes 23844 arguments. The value should be the number of arguments that 23845 option takes-zero, for many options. This macro rather than 23846 `SWITCH_TAKES_ARG' is used for multi-character option names. 23847 23848 By default, this macro is defined as 23849 `DEFAULT_WORD_SWITCH_TAKES_ARG', which handles the standard options 23850 properly. You need not define `WORD_SWITCH_TAKES_ARG' unless you 23851 wish to add additional options which take arguments. Any 23852 redefinition should call `DEFAULT_WORD_SWITCH_TAKES_ARG' and then 23853 check for additional options. 23854 23855 -- Macro: SWITCH_CURTAILS_COMPILATION (CHAR) 23856 A C expression which determines whether the option `-CHAR' stops 23857 compilation before the generation of an executable. The value is 23858 boolean, nonzero if the option does stop an executable from being 23859 generated, zero otherwise. 23860 23861 By default, this macro is defined as 23862 `DEFAULT_SWITCH_CURTAILS_COMPILATION', which handles the standard 23863 options properly. You need not define 23864 `SWITCH_CURTAILS_COMPILATION' unless you wish to add additional 23865 options which affect the generation of an executable. Any 23866 redefinition should call `DEFAULT_SWITCH_CURTAILS_COMPILATION' and 23867 then check for additional options. 23868 23869 -- Macro: SWITCHES_NEED_SPACES 23870 A string-valued C expression which enumerates the options for which 23871 the linker needs a space between the option and its argument. 23872 23873 If this macro is not defined, the default value is `""'. 23874 23875 -- Macro: TARGET_OPTION_TRANSLATE_TABLE 23876 If defined, a list of pairs of strings, the first of which is a 23877 potential command line target to the `gcc' driver program, and the 23878 second of which is a space-separated (tabs and other whitespace 23879 are not supported) list of options with which to replace the first 23880 option. The target defining this list is responsible for assuring 23881 that the results are valid. Replacement options may not be the 23882 `--opt' style, they must be the `-opt' style. It is the intention 23883 of this macro to provide a mechanism for substitution that affects 23884 the multilibs chosen, such as one option that enables many 23885 options, some of which select multilibs. Example nonsensical 23886 definition, where `-malt-abi', `-EB', and `-mspoo' cause different 23887 multilibs to be chosen: 23888 23889 #define TARGET_OPTION_TRANSLATE_TABLE \ 23890 { "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" }, \ 23891 { "-compat", "-EB -malign=4 -mspoo" } 23892 23893 -- Macro: DRIVER_SELF_SPECS 23894 A list of specs for the driver itself. It should be a suitable 23895 initializer for an array of strings, with no surrounding braces. 23896 23897 The driver applies these specs to its own command line between 23898 loading default `specs' files (but not command-line specified 23899 ones) and choosing the multilib directory or running any 23900 subcommands. It applies them in the order given, so each spec can 23901 depend on the options added by earlier ones. It is also possible 23902 to remove options using `%<OPTION' in the usual way. 23903 23904 This macro can be useful when a port has several interdependent 23905 target options. It provides a way of standardizing the command 23906 line so that the other specs are easier to write. 23907 23908 Do not define this macro if it does not need to do anything. 23909 23910 -- Macro: OPTION_DEFAULT_SPECS 23911 A list of specs used to support configure-time default options 23912 (i.e. `--with' options) in the driver. It should be a suitable 23913 initializer for an array of structures, each containing two 23914 strings, without the outermost pair of surrounding braces. 23915 23916 The first item in the pair is the name of the default. This must 23917 match the code in `config.gcc' for the target. The second item is 23918 a spec to apply if a default with this name was specified. The 23919 string `%(VALUE)' in the spec will be replaced by the value of the 23920 default everywhere it occurs. 23921 23922 The driver will apply these specs to its own command line between 23923 loading default `specs' files and processing `DRIVER_SELF_SPECS', 23924 using the same mechanism as `DRIVER_SELF_SPECS'. 23925 23926 Do not define this macro if it does not need to do anything. 23927 23928 -- Macro: CPP_SPEC 23929 A C string constant that tells the GCC driver program options to 23930 pass to CPP. It can also specify how to translate options you 23931 give to GCC into options for GCC to pass to the CPP. 23932 23933 Do not define this macro if it does not need to do anything. 23934 23935 -- Macro: CPLUSPLUS_CPP_SPEC 23936 This macro is just like `CPP_SPEC', but is used for C++, rather 23937 than C. If you do not define this macro, then the value of 23938 `CPP_SPEC' (if any) will be used instead. 23939 23940 -- Macro: CC1_SPEC 23941 A C string constant that tells the GCC driver program options to 23942 pass to `cc1', `cc1plus', `f771', and the other language front 23943 ends. It can also specify how to translate options you give to 23944 GCC into options for GCC to pass to front ends. 23945 23946 Do not define this macro if it does not need to do anything. 23947 23948 -- Macro: CC1PLUS_SPEC 23949 A C string constant that tells the GCC driver program options to 23950 pass to `cc1plus'. It can also specify how to translate options 23951 you give to GCC into options for GCC to pass to the `cc1plus'. 23952 23953 Do not define this macro if it does not need to do anything. Note 23954 that everything defined in CC1_SPEC is already passed to `cc1plus' 23955 so there is no need to duplicate the contents of CC1_SPEC in 23956 CC1PLUS_SPEC. 23957 23958 -- Macro: ASM_SPEC 23959 A C string constant that tells the GCC driver program options to 23960 pass to the assembler. It can also specify how to translate 23961 options you give to GCC into options for GCC to pass to the 23962 assembler. See the file `sun3.h' for an example of this. 23963 23964 Do not define this macro if it does not need to do anything. 23965 23966 -- Macro: ASM_FINAL_SPEC 23967 A C string constant that tells the GCC driver program how to run 23968 any programs which cleanup after the normal assembler. Normally, 23969 this is not needed. See the file `mips.h' for an example of this. 23970 23971 Do not define this macro if it does not need to do anything. 23972 23973 -- Macro: AS_NEEDS_DASH_FOR_PIPED_INPUT 23974 Define this macro, with no value, if the driver should give the 23975 assembler an argument consisting of a single dash, `-', to 23976 instruct it to read from its standard input (which will be a pipe 23977 connected to the output of the compiler proper). This argument is 23978 given after any `-o' option specifying the name of the output file. 23979 23980 If you do not define this macro, the assembler is assumed to read 23981 its standard input if given no non-option arguments. If your 23982 assembler cannot read standard input at all, use a `%{pipe:%e}' 23983 construct; see `mips.h' for instance. 23984 23985 -- Macro: LINK_SPEC 23986 A C string constant that tells the GCC driver program options to 23987 pass to the linker. It can also specify how to translate options 23988 you give to GCC into options for GCC to pass to the linker. 23989 23990 Do not define this macro if it does not need to do anything. 23991 23992 -- Macro: LIB_SPEC 23993 Another C string constant used much like `LINK_SPEC'. The 23994 difference between the two is that `LIB_SPEC' is used at the end 23995 of the command given to the linker. 23996 23997 If this macro is not defined, a default is provided that loads the 23998 standard C library from the usual place. See `gcc.c'. 23999 24000 -- Macro: LIBGCC_SPEC 24001 Another C string constant that tells the GCC driver program how 24002 and when to place a reference to `libgcc.a' into the linker 24003 command line. This constant is placed both before and after the 24004 value of `LIB_SPEC'. 24005 24006 If this macro is not defined, the GCC driver provides a default 24007 that passes the string `-lgcc' to the linker. 24008 24009 -- Macro: REAL_LIBGCC_SPEC 24010 By default, if `ENABLE_SHARED_LIBGCC' is defined, the 24011 `LIBGCC_SPEC' is not directly used by the driver program but is 24012 instead modified to refer to different versions of `libgcc.a' 24013 depending on the values of the command line flags `-static', 24014 `-shared', `-static-libgcc', and `-shared-libgcc'. On targets 24015 where these modifications are inappropriate, define 24016 `REAL_LIBGCC_SPEC' instead. `REAL_LIBGCC_SPEC' tells the driver 24017 how to place a reference to `libgcc' on the link command line, 24018 but, unlike `LIBGCC_SPEC', it is used unmodified. 24019 24020 -- Macro: USE_LD_AS_NEEDED 24021 A macro that controls the modifications to `LIBGCC_SPEC' mentioned 24022 in `REAL_LIBGCC_SPEC'. If nonzero, a spec will be generated that 24023 uses -as-needed and the shared libgcc in place of the static 24024 exception handler library, when linking without any of `-static', 24025 `-static-libgcc', or `-shared-libgcc'. 24026 24027 -- Macro: LINK_EH_SPEC 24028 If defined, this C string constant is added to `LINK_SPEC'. When 24029 `USE_LD_AS_NEEDED' is zero or undefined, it also affects the 24030 modifications to `LIBGCC_SPEC' mentioned in `REAL_LIBGCC_SPEC'. 24031 24032 -- Macro: STARTFILE_SPEC 24033 Another C string constant used much like `LINK_SPEC'. The 24034 difference between the two is that `STARTFILE_SPEC' is used at the 24035 very beginning of the command given to the linker. 24036 24037 If this macro is not defined, a default is provided that loads the 24038 standard C startup file from the usual place. See `gcc.c'. 24039 24040 -- Macro: ENDFILE_SPEC 24041 Another C string constant used much like `LINK_SPEC'. The 24042 difference between the two is that `ENDFILE_SPEC' is used at the 24043 very end of the command given to the linker. 24044 24045 Do not define this macro if it does not need to do anything. 24046 24047 -- Macro: THREAD_MODEL_SPEC 24048 GCC `-v' will print the thread model GCC was configured to use. 24049 However, this doesn't work on platforms that are multilibbed on 24050 thread models, such as AIX 4.3. On such platforms, define 24051 `THREAD_MODEL_SPEC' such that it evaluates to a string without 24052 blanks that names one of the recognized thread models. `%*', the 24053 default value of this macro, will expand to the value of 24054 `thread_file' set in `config.gcc'. 24055 24056 -- Macro: SYSROOT_SUFFIX_SPEC 24057 Define this macro to add a suffix to the target sysroot when GCC is 24058 configured with a sysroot. This will cause GCC to search for 24059 usr/lib, et al, within sysroot+suffix. 24060 24061 -- Macro: SYSROOT_HEADERS_SUFFIX_SPEC 24062 Define this macro to add a headers_suffix to the target sysroot 24063 when GCC is configured with a sysroot. This will cause GCC to 24064 pass the updated sysroot+headers_suffix to CPP, causing it to 24065 search for usr/include, et al, within sysroot+headers_suffix. 24066 24067 -- Macro: EXTRA_SPECS 24068 Define this macro to provide additional specifications to put in 24069 the `specs' file that can be used in various specifications like 24070 `CC1_SPEC'. 24071 24072 The definition should be an initializer for an array of structures, 24073 containing a string constant, that defines the specification name, 24074 and a string constant that provides the specification. 24075 24076 Do not define this macro if it does not need to do anything. 24077 24078 `EXTRA_SPECS' is useful when an architecture contains several 24079 related targets, which have various `..._SPECS' which are similar 24080 to each other, and the maintainer would like one central place to 24081 keep these definitions. 24082 24083 For example, the PowerPC System V.4 targets use `EXTRA_SPECS' to 24084 define either `_CALL_SYSV' when the System V calling sequence is 24085 used or `_CALL_AIX' when the older AIX-based calling sequence is 24086 used. 24087 24088 The `config/rs6000/rs6000.h' target file defines: 24089 24090 #define EXTRA_SPECS \ 24091 { "cpp_sysv_default", CPP_SYSV_DEFAULT }, 24092 24093 #define CPP_SYS_DEFAULT "" 24094 24095 The `config/rs6000/sysv.h' target file defines: 24096 #undef CPP_SPEC 24097 #define CPP_SPEC \ 24098 "%{posix: -D_POSIX_SOURCE } \ 24099 %{mcall-sysv: -D_CALL_SYSV } \ 24100 %{!mcall-sysv: %(cpp_sysv_default) } \ 24101 %{msoft-float: -D_SOFT_FLOAT} %{mcpu=403: -D_SOFT_FLOAT}" 24102 24103 #undef CPP_SYSV_DEFAULT 24104 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV" 24105 24106 while the `config/rs6000/eabiaix.h' target file defines 24107 `CPP_SYSV_DEFAULT' as: 24108 24109 #undef CPP_SYSV_DEFAULT 24110 #define CPP_SYSV_DEFAULT "-D_CALL_AIX" 24111 24112 -- Macro: LINK_LIBGCC_SPECIAL_1 24113 Define this macro if the driver program should find the library 24114 `libgcc.a'. If you do not define this macro, the driver program 24115 will pass the argument `-lgcc' to tell the linker to do the search. 24116 24117 -- Macro: LINK_GCC_C_SEQUENCE_SPEC 24118 The sequence in which libgcc and libc are specified to the linker. 24119 By default this is `%G %L %G'. 24120 24121 -- Macro: LINK_COMMAND_SPEC 24122 A C string constant giving the complete command line need to 24123 execute the linker. When you do this, you will need to update 24124 your port each time a change is made to the link command line 24125 within `gcc.c'. Therefore, define this macro only if you need to 24126 completely redefine the command line for invoking the linker and 24127 there is no other way to accomplish the effect you need. 24128 Overriding this macro may be avoidable by overriding 24129 `LINK_GCC_C_SEQUENCE_SPEC' instead. 24130 24131 -- Macro: LINK_ELIMINATE_DUPLICATE_LDIRECTORIES 24132 A nonzero value causes `collect2' to remove duplicate 24133 `-LDIRECTORY' search directories from linking commands. Do not 24134 give it a nonzero value if removing duplicate search directories 24135 changes the linker's semantics. 24136 24137 -- Macro: MULTILIB_DEFAULTS 24138 Define this macro as a C expression for the initializer of an 24139 array of string to tell the driver program which options are 24140 defaults for this target and thus do not need to be handled 24141 specially when using `MULTILIB_OPTIONS'. 24142 24143 Do not define this macro if `MULTILIB_OPTIONS' is not defined in 24144 the target makefile fragment or if none of the options listed in 24145 `MULTILIB_OPTIONS' are set by default. *Note Target Fragment::. 24146 24147 -- Macro: RELATIVE_PREFIX_NOT_LINKDIR 24148 Define this macro to tell `gcc' that it should only translate a 24149 `-B' prefix into a `-L' linker option if the prefix indicates an 24150 absolute file name. 24151 24152 -- Macro: MD_EXEC_PREFIX 24153 If defined, this macro is an additional prefix to try after 24154 `STANDARD_EXEC_PREFIX'. `MD_EXEC_PREFIX' is not searched when the 24155 `-b' option is used, or the compiler is built as a cross compiler. 24156 If you define `MD_EXEC_PREFIX', then be sure to add it to the 24157 list of directories used to find the assembler in `configure.in'. 24158 24159 -- Macro: STANDARD_STARTFILE_PREFIX 24160 Define this macro as a C string constant if you wish to override 24161 the standard choice of `libdir' as the default prefix to try when 24162 searching for startup files such as `crt0.o'. 24163 `STANDARD_STARTFILE_PREFIX' is not searched when the compiler is 24164 built as a cross compiler. 24165 24166 -- Macro: STANDARD_STARTFILE_PREFIX_1 24167 Define this macro as a C string constant if you wish to override 24168 the standard choice of `/lib' as a prefix to try after the default 24169 prefix when searching for startup files such as `crt0.o'. 24170 `STANDARD_STARTFILE_PREFIX_1' is not searched when the compiler is 24171 built as a cross compiler. 24172 24173 -- Macro: STANDARD_STARTFILE_PREFIX_2 24174 Define this macro as a C string constant if you wish to override 24175 the standard choice of `/lib' as yet another prefix to try after 24176 the default prefix when searching for startup files such as 24177 `crt0.o'. `STANDARD_STARTFILE_PREFIX_2' is not searched when the 24178 compiler is built as a cross compiler. 24179 24180 -- Macro: MD_STARTFILE_PREFIX 24181 If defined, this macro supplies an additional prefix to try after 24182 the standard prefixes. `MD_EXEC_PREFIX' is not searched when the 24183 `-b' option is used, or when the compiler is built as a cross 24184 compiler. 24185 24186 -- Macro: MD_STARTFILE_PREFIX_1 24187 If defined, this macro supplies yet another prefix to try after the 24188 standard prefixes. It is not searched when the `-b' option is 24189 used, or when the compiler is built as a cross compiler. 24190 24191 -- Macro: INIT_ENVIRONMENT 24192 Define this macro as a C string constant if you wish to set 24193 environment variables for programs called by the driver, such as 24194 the assembler and loader. The driver passes the value of this 24195 macro to `putenv' to initialize the necessary environment 24196 variables. 24197 24198 -- Macro: LOCAL_INCLUDE_DIR 24199 Define this macro as a C string constant if you wish to override 24200 the standard choice of `/usr/local/include' as the default prefix 24201 to try when searching for local header files. `LOCAL_INCLUDE_DIR' 24202 comes before `SYSTEM_INCLUDE_DIR' in the search order. 24203 24204 Cross compilers do not search either `/usr/local/include' or its 24205 replacement. 24206 24207 -- Macro: MODIFY_TARGET_NAME 24208 Define this macro if you wish to define command-line switches that 24209 modify the default target name. 24210 24211 For each switch, you can include a string to be appended to the 24212 first part of the configuration name or a string to be deleted 24213 from the configuration name, if present. The definition should be 24214 an initializer for an array of structures. Each array element 24215 should have three elements: the switch name (a string constant, 24216 including the initial dash), one of the enumeration codes `ADD' or 24217 `DELETE' to indicate whether the string should be inserted or 24218 deleted, and the string to be inserted or deleted (a string 24219 constant). 24220 24221 For example, on a machine where `64' at the end of the 24222 configuration name denotes a 64-bit target and you want the `-32' 24223 and `-64' switches to select between 32- and 64-bit targets, you 24224 would code 24225 24226 #define MODIFY_TARGET_NAME \ 24227 { { "-32", DELETE, "64"}, \ 24228 {"-64", ADD, "64"}} 24229 24230 -- Macro: SYSTEM_INCLUDE_DIR 24231 Define this macro as a C string constant if you wish to specify a 24232 system-specific directory to search for header files before the 24233 standard directory. `SYSTEM_INCLUDE_DIR' comes before 24234 `STANDARD_INCLUDE_DIR' in the search order. 24235 24236 Cross compilers do not use this macro and do not search the 24237 directory specified. 24238 24239 -- Macro: STANDARD_INCLUDE_DIR 24240 Define this macro as a C string constant if you wish to override 24241 the standard choice of `/usr/include' as the default prefix to try 24242 when searching for header files. 24243 24244 Cross compilers ignore this macro and do not search either 24245 `/usr/include' or its replacement. 24246 24247 -- Macro: STANDARD_INCLUDE_COMPONENT 24248 The "component" corresponding to `STANDARD_INCLUDE_DIR'. See 24249 `INCLUDE_DEFAULTS', below, for the description of components. If 24250 you do not define this macro, no component is used. 24251 24252 -- Macro: INCLUDE_DEFAULTS 24253 Define this macro if you wish to override the entire default 24254 search path for include files. For a native compiler, the default 24255 search path usually consists of `GCC_INCLUDE_DIR', 24256 `LOCAL_INCLUDE_DIR', `SYSTEM_INCLUDE_DIR', 24257 `GPLUSPLUS_INCLUDE_DIR', and `STANDARD_INCLUDE_DIR'. In addition, 24258 `GPLUSPLUS_INCLUDE_DIR' and `GCC_INCLUDE_DIR' are defined 24259 automatically by `Makefile', and specify private search areas for 24260 GCC. The directory `GPLUSPLUS_INCLUDE_DIR' is used only for C++ 24261 programs. 24262 24263 The definition should be an initializer for an array of structures. 24264 Each array element should have four elements: the directory name (a 24265 string constant), the component name (also a string constant), a 24266 flag for C++-only directories, and a flag showing that the 24267 includes in the directory don't need to be wrapped in `extern `C'' 24268 when compiling C++. Mark the end of the array with a null element. 24269 24270 The component name denotes what GNU package the include file is 24271 part of, if any, in all uppercase letters. For example, it might 24272 be `GCC' or `BINUTILS'. If the package is part of a 24273 vendor-supplied operating system, code the component name as `0'. 24274 24275 For example, here is the definition used for VAX/VMS: 24276 24277 #define INCLUDE_DEFAULTS \ 24278 { \ 24279 { "GNU_GXX_INCLUDE:", "G++", 1, 1}, \ 24280 { "GNU_CC_INCLUDE:", "GCC", 0, 0}, \ 24281 { "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0}, \ 24282 { ".", 0, 0, 0}, \ 24283 { 0, 0, 0, 0} \ 24284 } 24285 24286 Here is the order of prefixes tried for exec files: 24287 24288 1. Any prefixes specified by the user with `-B'. 24289 24290 2. The environment variable `GCC_EXEC_PREFIX' or, if `GCC_EXEC_PREFIX' 24291 is not set and the compiler has not been installed in the 24292 configure-time PREFIX, the location in which the compiler has 24293 actually been installed. 24294 24295 3. The directories specified by the environment variable 24296 `COMPILER_PATH'. 24297 24298 4. The macro `STANDARD_EXEC_PREFIX', if the compiler has been 24299 installed in the configured-time PREFIX. 24300 24301 5. The location `/usr/libexec/gcc/', but only if this is a native 24302 compiler. 24303 24304 6. The location `/usr/lib/gcc/', but only if this is a native 24305 compiler. 24306 24307 7. The macro `MD_EXEC_PREFIX', if defined, but only if this is a 24308 native compiler. 24309 24310 Here is the order of prefixes tried for startfiles: 24311 24312 1. Any prefixes specified by the user with `-B'. 24313 24314 2. The environment variable `GCC_EXEC_PREFIX' or its automatically 24315 determined value based on the installed toolchain location. 24316 24317 3. The directories specified by the environment variable 24318 `LIBRARY_PATH' (or port-specific name; native only, cross 24319 compilers do not use this). 24320 24321 4. The macro `STANDARD_EXEC_PREFIX', but only if the toolchain is 24322 installed in the configured PREFIX or this is a native compiler. 24323 24324 5. The location `/usr/lib/gcc/', but only if this is a native 24325 compiler. 24326 24327 6. The macro `MD_EXEC_PREFIX', if defined, but only if this is a 24328 native compiler. 24329 24330 7. The macro `MD_STARTFILE_PREFIX', if defined, but only if this is a 24331 native compiler, or we have a target system root. 24332 24333 8. The macro `MD_STARTFILE_PREFIX_1', if defined, but only if this is 24334 a native compiler, or we have a target system root. 24335 24336 9. The macro `STANDARD_STARTFILE_PREFIX', with any sysroot 24337 modifications. If this path is relative it will be prefixed by 24338 `GCC_EXEC_PREFIX' and the machine suffix or `STANDARD_EXEC_PREFIX' 24339 and the machine suffix. 24340 24341 10. The macro `STANDARD_STARTFILE_PREFIX_1', but only if this is a 24342 native compiler, or we have a target system root. The default for 24343 this macro is `/lib/'. 24344 24345 11. The macro `STANDARD_STARTFILE_PREFIX_2', but only if this is a 24346 native compiler, or we have a target system root. The default for 24347 this macro is `/usr/lib/'. 24348 24349 24350 File: gccint.info, Node: Run-time Target, Next: Per-Function Data, Prev: Driver, Up: Target Macros 24351 24352 17.3 Run-time Target Specification 24353 ================================== 24354 24355 Here are run-time target specifications. 24356 24357 -- Macro: TARGET_CPU_CPP_BUILTINS () 24358 This function-like macro expands to a block of code that defines 24359 built-in preprocessor macros and assertions for the target CPU, 24360 using the functions `builtin_define', `builtin_define_std' and 24361 `builtin_assert'. When the front end calls this macro it provides 24362 a trailing semicolon, and since it has finished command line 24363 option processing your code can use those results freely. 24364 24365 `builtin_assert' takes a string in the form you pass to the 24366 command-line option `-A', such as `cpu=mips', and creates the 24367 assertion. `builtin_define' takes a string in the form accepted 24368 by option `-D' and unconditionally defines the macro. 24369 24370 `builtin_define_std' takes a string representing the name of an 24371 object-like macro. If it doesn't lie in the user's namespace, 24372 `builtin_define_std' defines it unconditionally. Otherwise, it 24373 defines a version with two leading underscores, and another version 24374 with two leading and trailing underscores, and defines the original 24375 only if an ISO standard was not requested on the command line. For 24376 example, passing `unix' defines `__unix', `__unix__' and possibly 24377 `unix'; passing `_mips' defines `__mips', `__mips__' and possibly 24378 `_mips', and passing `_ABI64' defines only `_ABI64'. 24379 24380 You can also test for the C dialect being compiled. The variable 24381 `c_language' is set to one of `clk_c', `clk_cplusplus' or 24382 `clk_objective_c'. Note that if we are preprocessing assembler, 24383 this variable will be `clk_c' but the function-like macro 24384 `preprocessing_asm_p()' will return true, so you might want to 24385 check for that first. If you need to check for strict ANSI, the 24386 variable `flag_iso' can be used. The function-like macro 24387 `preprocessing_trad_p()' can be used to check for traditional 24388 preprocessing. 24389 24390 -- Macro: TARGET_OS_CPP_BUILTINS () 24391 Similarly to `TARGET_CPU_CPP_BUILTINS' but this macro is optional 24392 and is used for the target operating system instead. 24393 24394 -- Macro: TARGET_OBJFMT_CPP_BUILTINS () 24395 Similarly to `TARGET_CPU_CPP_BUILTINS' but this macro is optional 24396 and is used for the target object format. `elfos.h' uses this 24397 macro to define `__ELF__', so you probably do not need to define 24398 it yourself. 24399 24400 -- Variable: extern int target_flags 24401 This variable is declared in `options.h', which is included before 24402 any target-specific headers. 24403 24404 -- Variable: Target Hook int TARGET_DEFAULT_TARGET_FLAGS 24405 This variable specifies the initial value of `target_flags'. Its 24406 default setting is 0. 24407 24408 -- Target Hook: bool TARGET_HANDLE_OPTION (size_t CODE, const char 24409 *ARG, int VALUE) 24410 This hook is called whenever the user specifies one of the 24411 target-specific options described by the `.opt' definition files 24412 (*note Options::). It has the opportunity to do some 24413 option-specific processing and should return true if the option is 24414 valid. The default definition does nothing but return true. 24415 24416 CODE specifies the `OPT_NAME' enumeration value associated with 24417 the selected option; NAME is just a rendering of the option name 24418 in which non-alphanumeric characters are replaced by underscores. 24419 ARG specifies the string argument and is null if no argument was 24420 given. If the option is flagged as a `UInteger' (*note Option 24421 properties::), VALUE is the numeric value of the argument. 24422 Otherwise VALUE is 1 if the positive form of the option was used 24423 and 0 if the "no-" form was. 24424 24425 -- Target Hook: bool TARGET_HANDLE_C_OPTION (size_t CODE, const char 24426 *ARG, int VALUE) 24427 This target hook is called whenever the user specifies one of the 24428 target-specific C language family options described by the `.opt' 24429 definition files(*note Options::). It has the opportunity to do 24430 some option-specific processing and should return true if the 24431 option is valid. The default definition does nothing but return 24432 false. 24433 24434 In general, you should use `TARGET_HANDLE_OPTION' to handle 24435 options. However, if processing an option requires routines that 24436 are only available in the C (and related language) front ends, 24437 then you should use `TARGET_HANDLE_C_OPTION' instead. 24438 24439 -- Macro: TARGET_VERSION 24440 This macro is a C statement to print on `stderr' a string 24441 describing the particular machine description choice. Every 24442 machine description should define `TARGET_VERSION'. For example: 24443 24444 #ifdef MOTOROLA 24445 #define TARGET_VERSION \ 24446 fprintf (stderr, " (68k, Motorola syntax)"); 24447 #else 24448 #define TARGET_VERSION \ 24449 fprintf (stderr, " (68k, MIT syntax)"); 24450 #endif 24451 24452 -- Macro: OVERRIDE_OPTIONS 24453 Sometimes certain combinations of command options do not make 24454 sense on a particular target machine. You can define a macro 24455 `OVERRIDE_OPTIONS' to take account of this. This macro, if 24456 defined, is executed once just after all the command options have 24457 been parsed. 24458 24459 Don't use this macro to turn on various extra optimizations for 24460 `-O'. That is what `OPTIMIZATION_OPTIONS' is for. 24461 24462 -- Macro: C_COMMON_OVERRIDE_OPTIONS 24463 This is similar to `OVERRIDE_OPTIONS' but is only used in the C 24464 language frontends (C, Objective-C, C++, Objective-C++) and so can 24465 be used to alter option flag variables which only exist in those 24466 frontends. 24467 24468 -- Macro: OPTIMIZATION_OPTIONS (LEVEL, SIZE) 24469 Some machines may desire to change what optimizations are 24470 performed for various optimization levels. This macro, if 24471 defined, is executed once just after the optimization level is 24472 determined and before the remainder of the command options have 24473 been parsed. Values set in this macro are used as the default 24474 values for the other command line options. 24475 24476 LEVEL is the optimization level specified; 2 if `-O2' is 24477 specified, 1 if `-O' is specified, and 0 if neither is specified. 24478 24479 SIZE is nonzero if `-Os' is specified and zero otherwise. 24480 24481 This macro is run once at program startup and when the optimization 24482 options are changed via `#pragma GCC optimize' or by using the 24483 `optimize' attribute. 24484 24485 *Do not examine `write_symbols' in this macro!* The debugging 24486 options are not supposed to alter the generated code. 24487 24488 -- Target Hook: bool TARGET_HELP (void) 24489 This hook is called in response to the user invoking 24490 `--target-help' on the command line. It gives the target a chance 24491 to display extra information on the target specific command line 24492 options found in its `.opt' file. 24493 24494 -- Macro: CAN_DEBUG_WITHOUT_FP 24495 Define this macro if debugging can be performed even without a 24496 frame pointer. If this macro is defined, GCC will turn on the 24497 `-fomit-frame-pointer' option whenever `-O' is specified. 24498 24499 24500 File: gccint.info, Node: Per-Function Data, Next: Storage Layout, Prev: Run-time Target, Up: Target Macros 24501 24502 17.4 Defining data structures for per-function information. 24503 =========================================================== 24504 24505 If the target needs to store information on a per-function basis, GCC 24506 provides a macro and a couple of variables to allow this. Note, just 24507 using statics to store the information is a bad idea, since GCC supports 24508 nested functions, so you can be halfway through encoding one function 24509 when another one comes along. 24510 24511 GCC defines a data structure called `struct function' which contains 24512 all of the data specific to an individual function. This structure 24513 contains a field called `machine' whose type is `struct 24514 machine_function *', which can be used by targets to point to their own 24515 specific data. 24516 24517 If a target needs per-function specific data it should define the type 24518 `struct machine_function' and also the macro `INIT_EXPANDERS'. This 24519 macro should be used to initialize the function pointer 24520 `init_machine_status'. This pointer is explained below. 24521 24522 One typical use of per-function, target specific data is to create an 24523 RTX to hold the register containing the function's return address. This 24524 RTX can then be used to implement the `__builtin_return_address' 24525 function, for level 0. 24526 24527 Note--earlier implementations of GCC used a single data area to hold 24528 all of the per-function information. Thus when processing of a nested 24529 function began the old per-function data had to be pushed onto a stack, 24530 and when the processing was finished, it had to be popped off the 24531 stack. GCC used to provide function pointers called 24532 `save_machine_status' and `restore_machine_status' to handle the saving 24533 and restoring of the target specific information. Since the single 24534 data area approach is no longer used, these pointers are no longer 24535 supported. 24536 24537 -- Macro: INIT_EXPANDERS 24538 Macro called to initialize any target specific information. This 24539 macro is called once per function, before generation of any RTL 24540 has begun. The intention of this macro is to allow the 24541 initialization of the function pointer `init_machine_status'. 24542 24543 -- Variable: void (*)(struct function *) init_machine_status 24544 If this function pointer is non-`NULL' it will be called once per 24545 function, before function compilation starts, in order to allow the 24546 target to perform any target specific initialization of the 24547 `struct function' structure. It is intended that this would be 24548 used to initialize the `machine' of that structure. 24549 24550 `struct machine_function' structures are expected to be freed by 24551 GC. Generally, any memory that they reference must be allocated 24552 by using `ggc_alloc', including the structure itself. 24553 24554 24555 File: gccint.info, Node: Storage Layout, Next: Type Layout, Prev: Per-Function Data, Up: Target Macros 24556 24557 17.5 Storage Layout 24558 =================== 24559 24560 Note that the definitions of the macros in this table which are sizes or 24561 alignments measured in bits do not need to be constant. They can be C 24562 expressions that refer to static variables, such as the `target_flags'. 24563 *Note Run-time Target::. 24564 24565 -- Macro: BITS_BIG_ENDIAN 24566 Define this macro to have the value 1 if the most significant bit 24567 in a byte has the lowest number; otherwise define it to have the 24568 value zero. This means that bit-field instructions count from the 24569 most significant bit. If the machine has no bit-field 24570 instructions, then this must still be defined, but it doesn't 24571 matter which value it is defined to. This macro need not be a 24572 constant. 24573 24574 This macro does not affect the way structure fields are packed into 24575 bytes or words; that is controlled by `BYTES_BIG_ENDIAN'. 24576 24577 -- Macro: BYTES_BIG_ENDIAN 24578 Define this macro to have the value 1 if the most significant byte 24579 in a word has the lowest number. This macro need not be a 24580 constant. 24581 24582 -- Macro: WORDS_BIG_ENDIAN 24583 Define this macro to have the value 1 if, in a multiword object, 24584 the most significant word has the lowest number. This applies to 24585 both memory locations and registers; GCC fundamentally assumes 24586 that the order of words in memory is the same as the order in 24587 registers. This macro need not be a constant. 24588 24589 -- Macro: LIBGCC2_WORDS_BIG_ENDIAN 24590 Define this macro if `WORDS_BIG_ENDIAN' is not constant. This 24591 must be a constant value with the same meaning as 24592 `WORDS_BIG_ENDIAN', which will be used only when compiling 24593 `libgcc2.c'. Typically the value will be set based on 24594 preprocessor defines. 24595 24596 -- Macro: FLOAT_WORDS_BIG_ENDIAN 24597 Define this macro to have the value 1 if `DFmode', `XFmode' or 24598 `TFmode' floating point numbers are stored in memory with the word 24599 containing the sign bit at the lowest address; otherwise define it 24600 to have the value 0. This macro need not be a constant. 24601 24602 You need not define this macro if the ordering is the same as for 24603 multi-word integers. 24604 24605 -- Macro: BITS_PER_UNIT 24606 Define this macro to be the number of bits in an addressable 24607 storage unit (byte). If you do not define this macro the default 24608 is 8. 24609 24610 -- Macro: BITS_PER_WORD 24611 Number of bits in a word. If you do not define this macro, the 24612 default is `BITS_PER_UNIT * UNITS_PER_WORD'. 24613 24614 -- Macro: MAX_BITS_PER_WORD 24615 Maximum number of bits in a word. If this is undefined, the 24616 default is `BITS_PER_WORD'. Otherwise, it is the constant value 24617 that is the largest value that `BITS_PER_WORD' can have at 24618 run-time. 24619 24620 -- Macro: UNITS_PER_WORD 24621 Number of storage units in a word; normally the size of a 24622 general-purpose register, a power of two from 1 or 8. 24623 24624 -- Macro: MIN_UNITS_PER_WORD 24625 Minimum number of units in a word. If this is undefined, the 24626 default is `UNITS_PER_WORD'. Otherwise, it is the constant value 24627 that is the smallest value that `UNITS_PER_WORD' can have at 24628 run-time. 24629 24630 -- Macro: UNITS_PER_SIMD_WORD (MODE) 24631 Number of units in the vectors that the vectorizer can produce for 24632 scalar mode MODE. The default is equal to `UNITS_PER_WORD', 24633 because the vectorizer can do some transformations even in absence 24634 of specialized SIMD hardware. 24635 24636 -- Macro: POINTER_SIZE 24637 Width of a pointer, in bits. You must specify a value no wider 24638 than the width of `Pmode'. If it is not equal to the width of 24639 `Pmode', you must define `POINTERS_EXTEND_UNSIGNED'. If you do 24640 not specify a value the default is `BITS_PER_WORD'. 24641 24642 -- Macro: POINTERS_EXTEND_UNSIGNED 24643 A C expression that determines how pointers should be extended from 24644 `ptr_mode' to either `Pmode' or `word_mode'. It is greater than 24645 zero if pointers should be zero-extended, zero if they should be 24646 sign-extended, and negative if some other sort of conversion is 24647 needed. In the last case, the extension is done by the target's 24648 `ptr_extend' instruction. 24649 24650 You need not define this macro if the `ptr_mode', `Pmode' and 24651 `word_mode' are all the same width. 24652 24653 -- Macro: PROMOTE_MODE (M, UNSIGNEDP, TYPE) 24654 A macro to update M and UNSIGNEDP when an object whose type is 24655 TYPE and which has the specified mode and signedness is to be 24656 stored in a register. This macro is only called when TYPE is a 24657 scalar type. 24658 24659 On most RISC machines, which only have operations that operate on 24660 a full register, define this macro to set M to `word_mode' if M is 24661 an integer mode narrower than `BITS_PER_WORD'. In most cases, 24662 only integer modes should be widened because wider-precision 24663 floating-point operations are usually more expensive than their 24664 narrower counterparts. 24665 24666 For most machines, the macro definition does not change UNSIGNEDP. 24667 However, some machines, have instructions that preferentially 24668 handle either signed or unsigned quantities of certain modes. For 24669 example, on the DEC Alpha, 32-bit loads from memory and 32-bit add 24670 instructions sign-extend the result to 64 bits. On such machines, 24671 set UNSIGNEDP according to which kind of extension is more 24672 efficient. 24673 24674 Do not define this macro if it would never modify M. 24675 24676 -- Macro: PROMOTE_FUNCTION_MODE 24677 Like `PROMOTE_MODE', but is applied to outgoing function arguments 24678 or function return values, as specified by 24679 `TARGET_PROMOTE_FUNCTION_ARGS' and 24680 `TARGET_PROMOTE_FUNCTION_RETURN', respectively. 24681 24682 The default is `PROMOTE_MODE'. 24683 24684 -- Target Hook: bool TARGET_PROMOTE_FUNCTION_ARGS (tree FNTYPE) 24685 This target hook should return `true' if the promotion described by 24686 `PROMOTE_FUNCTION_MODE' should be done for outgoing function 24687 arguments. 24688 24689 -- Target Hook: bool TARGET_PROMOTE_FUNCTION_RETURN (tree FNTYPE) 24690 This target hook should return `true' if the promotion described by 24691 `PROMOTE_FUNCTION_MODE' should be done for the return value of 24692 functions. 24693 24694 If this target hook returns `true', `TARGET_FUNCTION_VALUE' must 24695 perform the same promotions done by `PROMOTE_FUNCTION_MODE'. 24696 24697 -- Macro: PARM_BOUNDARY 24698 Normal alignment required for function parameters on the stack, in 24699 bits. All stack parameters receive at least this much alignment 24700 regardless of data type. On most machines, this is the same as the 24701 size of an integer. 24702 24703 -- Macro: STACK_BOUNDARY 24704 Define this macro to the minimum alignment enforced by hardware 24705 for the stack pointer on this machine. The definition is a C 24706 expression for the desired alignment (measured in bits). This 24707 value is used as a default if `PREFERRED_STACK_BOUNDARY' is not 24708 defined. On most machines, this should be the same as 24709 `PARM_BOUNDARY'. 24710 24711 -- Macro: PREFERRED_STACK_BOUNDARY 24712 Define this macro if you wish to preserve a certain alignment for 24713 the stack pointer, greater than what the hardware enforces. The 24714 definition is a C expression for the desired alignment (measured 24715 in bits). This macro must evaluate to a value equal to or larger 24716 than `STACK_BOUNDARY'. 24717 24718 -- Macro: INCOMING_STACK_BOUNDARY 24719 Define this macro if the incoming stack boundary may be different 24720 from `PREFERRED_STACK_BOUNDARY'. This macro must evaluate to a 24721 value equal to or larger than `STACK_BOUNDARY'. 24722 24723 -- Macro: FUNCTION_BOUNDARY 24724 Alignment required for a function entry point, in bits. 24725 24726 -- Macro: BIGGEST_ALIGNMENT 24727 Biggest alignment that any data type can require on this machine, 24728 in bits. Note that this is not the biggest alignment that is 24729 supported, just the biggest alignment that, when violated, may 24730 cause a fault. 24731 24732 -- Macro: MALLOC_ABI_ALIGNMENT 24733 Alignment, in bits, a C conformant malloc implementation has to 24734 provide. If not defined, the default value is `BITS_PER_WORD'. 24735 24736 -- Macro: ATTRIBUTE_ALIGNED_VALUE 24737 Alignment used by the `__attribute__ ((aligned))' construct. If 24738 not defined, the default value is `BIGGEST_ALIGNMENT'. 24739 24740 -- Macro: MINIMUM_ATOMIC_ALIGNMENT 24741 If defined, the smallest alignment, in bits, that can be given to 24742 an object that can be referenced in one operation, without 24743 disturbing any nearby object. Normally, this is `BITS_PER_UNIT', 24744 but may be larger on machines that don't have byte or half-word 24745 store operations. 24746 24747 -- Macro: BIGGEST_FIELD_ALIGNMENT 24748 Biggest alignment that any structure or union field can require on 24749 this machine, in bits. If defined, this overrides 24750 `BIGGEST_ALIGNMENT' for structure and union fields only, unless 24751 the field alignment has been set by the `__attribute__ ((aligned 24752 (N)))' construct. 24753 24754 -- Macro: ADJUST_FIELD_ALIGN (FIELD, COMPUTED) 24755 An expression for the alignment of a structure field FIELD if the 24756 alignment computed in the usual way (including applying of 24757 `BIGGEST_ALIGNMENT' and `BIGGEST_FIELD_ALIGNMENT' to the 24758 alignment) is COMPUTED. It overrides alignment only if the field 24759 alignment has not been set by the `__attribute__ ((aligned (N)))' 24760 construct. 24761 24762 -- Macro: MAX_STACK_ALIGNMENT 24763 Biggest stack alignment guaranteed by the backend. Use this macro 24764 to specify the maximum alignment of a variable on stack. 24765 24766 If not defined, the default value is `STACK_BOUNDARY'. 24767 24768 24769 -- Macro: MAX_OFILE_ALIGNMENT 24770 Biggest alignment supported by the object file format of this 24771 machine. Use this macro to limit the alignment which can be 24772 specified using the `__attribute__ ((aligned (N)))' construct. If 24773 not defined, the default value is `BIGGEST_ALIGNMENT'. 24774 24775 On systems that use ELF, the default (in `config/elfos.h') is the 24776 largest supported 32-bit ELF section alignment representable on a 24777 32-bit host e.g. `(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)'. On 24778 32-bit ELF the largest supported section alignment in bits is 24779 `(0x80000000 * 8)', but this is not representable on 32-bit hosts. 24780 24781 -- Macro: DATA_ALIGNMENT (TYPE, BASIC-ALIGN) 24782 If defined, a C expression to compute the alignment for a variable 24783 in the static store. TYPE is the data type, and BASIC-ALIGN is 24784 the alignment that the object would ordinarily have. The value of 24785 this macro is used instead of that alignment to align the object. 24786 24787 If this macro is not defined, then BASIC-ALIGN is used. 24788 24789 One use of this macro is to increase alignment of medium-size data 24790 to make it all fit in fewer cache lines. Another is to cause 24791 character arrays to be word-aligned so that `strcpy' calls that 24792 copy constants to character arrays can be done inline. 24793 24794 -- Macro: CONSTANT_ALIGNMENT (CONSTANT, BASIC-ALIGN) 24795 If defined, a C expression to compute the alignment given to a 24796 constant that is being placed in memory. CONSTANT is the constant 24797 and BASIC-ALIGN is the alignment that the object would ordinarily 24798 have. The value of this macro is used instead of that alignment to 24799 align the object. 24800 24801 If this macro is not defined, then BASIC-ALIGN is used. 24802 24803 The typical use of this macro is to increase alignment for string 24804 constants to be word aligned so that `strcpy' calls that copy 24805 constants can be done inline. 24806 24807 -- Macro: LOCAL_ALIGNMENT (TYPE, BASIC-ALIGN) 24808 If defined, a C expression to compute the alignment for a variable 24809 in the local store. TYPE is the data type, and BASIC-ALIGN is the 24810 alignment that the object would ordinarily have. The value of this 24811 macro is used instead of that alignment to align the object. 24812 24813 If this macro is not defined, then BASIC-ALIGN is used. 24814 24815 One use of this macro is to increase alignment of medium-size data 24816 to make it all fit in fewer cache lines. 24817 24818 -- Macro: STACK_SLOT_ALIGNMENT (TYPE, MODE, BASIC-ALIGN) 24819 If defined, a C expression to compute the alignment for stack slot. 24820 TYPE is the data type, MODE is the widest mode available, and 24821 BASIC-ALIGN is the alignment that the slot would ordinarily have. 24822 The value of this macro is used instead of that alignment to align 24823 the slot. 24824 24825 If this macro is not defined, then BASIC-ALIGN is used when TYPE 24826 is `NULL'. Otherwise, `LOCAL_ALIGNMENT' will be used. 24827 24828 This macro is to set alignment of stack slot to the maximum 24829 alignment of all possible modes which the slot may have. 24830 24831 -- Macro: LOCAL_DECL_ALIGNMENT (DECL) 24832 If defined, a C expression to compute the alignment for a local 24833 variable DECL. 24834 24835 If this macro is not defined, then `LOCAL_ALIGNMENT (TREE_TYPE 24836 (DECL), DECL_ALIGN (DECL))' is used. 24837 24838 One use of this macro is to increase alignment of medium-size data 24839 to make it all fit in fewer cache lines. 24840 24841 -- Macro: MINIMUM_ALIGNMENT (EXP, MODE, ALIGN) 24842 If defined, a C expression to compute the minimum required 24843 alignment for dynamic stack realignment purposes for EXP (a type 24844 or decl), MODE, assuming normal alignment ALIGN. 24845 24846 If this macro is not defined, then ALIGN will be used. 24847 24848 -- Macro: EMPTY_FIELD_BOUNDARY 24849 Alignment in bits to be given to a structure bit-field that 24850 follows an empty field such as `int : 0;'. 24851 24852 If `PCC_BITFIELD_TYPE_MATTERS' is true, it overrides this macro. 24853 24854 -- Macro: STRUCTURE_SIZE_BOUNDARY 24855 Number of bits which any structure or union's size must be a 24856 multiple of. Each structure or union's size is rounded up to a 24857 multiple of this. 24858 24859 If you do not define this macro, the default is the same as 24860 `BITS_PER_UNIT'. 24861 24862 -- Macro: STRICT_ALIGNMENT 24863 Define this macro to be the value 1 if instructions will fail to 24864 work if given data not on the nominal alignment. If instructions 24865 will merely go slower in that case, define this macro as 0. 24866 24867 -- Macro: PCC_BITFIELD_TYPE_MATTERS 24868 Define this if you wish to imitate the way many other C compilers 24869 handle alignment of bit-fields and the structures that contain 24870 them. 24871 24872 The behavior is that the type written for a named bit-field (`int', 24873 `short', or other integer type) imposes an alignment for the entire 24874 structure, as if the structure really did contain an ordinary 24875 field of that type. In addition, the bit-field is placed within 24876 the structure so that it would fit within such a field, not 24877 crossing a boundary for it. 24878 24879 Thus, on most machines, a named bit-field whose type is written as 24880 `int' would not cross a four-byte boundary, and would force 24881 four-byte alignment for the whole structure. (The alignment used 24882 may not be four bytes; it is controlled by the other alignment 24883 parameters.) 24884 24885 An unnamed bit-field will not affect the alignment of the 24886 containing structure. 24887 24888 If the macro is defined, its definition should be a C expression; 24889 a nonzero value for the expression enables this behavior. 24890 24891 Note that if this macro is not defined, or its value is zero, some 24892 bit-fields may cross more than one alignment boundary. The 24893 compiler can support such references if there are `insv', `extv', 24894 and `extzv' insns that can directly reference memory. 24895 24896 The other known way of making bit-fields work is to define 24897 `STRUCTURE_SIZE_BOUNDARY' as large as `BIGGEST_ALIGNMENT'. Then 24898 every structure can be accessed with fullwords. 24899 24900 Unless the machine has bit-field instructions or you define 24901 `STRUCTURE_SIZE_BOUNDARY' that way, you must define 24902 `PCC_BITFIELD_TYPE_MATTERS' to have a nonzero value. 24903 24904 If your aim is to make GCC use the same conventions for laying out 24905 bit-fields as are used by another compiler, here is how to 24906 investigate what the other compiler does. Compile and run this 24907 program: 24908 24909 struct foo1 24910 { 24911 char x; 24912 char :0; 24913 char y; 24914 }; 24915 24916 struct foo2 24917 { 24918 char x; 24919 int :0; 24920 char y; 24921 }; 24922 24923 main () 24924 { 24925 printf ("Size of foo1 is %d\n", 24926 sizeof (struct foo1)); 24927 printf ("Size of foo2 is %d\n", 24928 sizeof (struct foo2)); 24929 exit (0); 24930 } 24931 24932 If this prints 2 and 5, then the compiler's behavior is what you 24933 would get from `PCC_BITFIELD_TYPE_MATTERS'. 24934 24935 -- Macro: BITFIELD_NBYTES_LIMITED 24936 Like `PCC_BITFIELD_TYPE_MATTERS' except that its effect is limited 24937 to aligning a bit-field within the structure. 24938 24939 -- Target Hook: bool TARGET_ALIGN_ANON_BITFIELD (void) 24940 When `PCC_BITFIELD_TYPE_MATTERS' is true this hook will determine 24941 whether unnamed bitfields affect the alignment of the containing 24942 structure. The hook should return true if the structure should 24943 inherit the alignment requirements of an unnamed bitfield's type. 24944 24945 -- Target Hook: bool TARGET_NARROW_VOLATILE_BITFIELD (void) 24946 This target hook should return `true' if accesses to volatile 24947 bitfields should use the narrowest mode possible. It should 24948 return `false' if these accesses should use the bitfield container 24949 type. 24950 24951 The default is `!TARGET_STRICT_ALIGN'. 24952 24953 -- Macro: MEMBER_TYPE_FORCES_BLK (FIELD, MODE) 24954 Return 1 if a structure or array containing FIELD should be 24955 accessed using `BLKMODE'. 24956 24957 If FIELD is the only field in the structure, MODE is its mode, 24958 otherwise MODE is VOIDmode. MODE is provided in the case where 24959 structures of one field would require the structure's mode to 24960 retain the field's mode. 24961 24962 Normally, this is not needed. 24963 24964 -- Macro: ROUND_TYPE_ALIGN (TYPE, COMPUTED, SPECIFIED) 24965 Define this macro as an expression for the alignment of a type 24966 (given by TYPE as a tree node) if the alignment computed in the 24967 usual way is COMPUTED and the alignment explicitly specified was 24968 SPECIFIED. 24969 24970 The default is to use SPECIFIED if it is larger; otherwise, use 24971 the smaller of COMPUTED and `BIGGEST_ALIGNMENT' 24972 24973 -- Macro: MAX_FIXED_MODE_SIZE 24974 An integer expression for the size in bits of the largest integer 24975 machine mode that should actually be used. All integer machine 24976 modes of this size or smaller can be used for structures and 24977 unions with the appropriate sizes. If this macro is undefined, 24978 `GET_MODE_BITSIZE (DImode)' is assumed. 24979 24980 -- Macro: STACK_SAVEAREA_MODE (SAVE_LEVEL) 24981 If defined, an expression of type `enum machine_mode' that 24982 specifies the mode of the save area operand of a 24983 `save_stack_LEVEL' named pattern (*note Standard Names::). 24984 SAVE_LEVEL is one of `SAVE_BLOCK', `SAVE_FUNCTION', or 24985 `SAVE_NONLOCAL' and selects which of the three named patterns is 24986 having its mode specified. 24987 24988 You need not define this macro if it always returns `Pmode'. You 24989 would most commonly define this macro if the `save_stack_LEVEL' 24990 patterns need to support both a 32- and a 64-bit mode. 24991 24992 -- Macro: STACK_SIZE_MODE 24993 If defined, an expression of type `enum machine_mode' that 24994 specifies the mode of the size increment operand of an 24995 `allocate_stack' named pattern (*note Standard Names::). 24996 24997 You need not define this macro if it always returns `word_mode'. 24998 You would most commonly define this macro if the `allocate_stack' 24999 pattern needs to support both a 32- and a 64-bit mode. 25000 25001 -- Target Hook: enum machine_mode TARGET_LIBGCC_CMP_RETURN_MODE () 25002 This target hook should return the mode to be used for the return 25003 value of compare instructions expanded to libgcc calls. If not 25004 defined `word_mode' is returned which is the right choice for a 25005 majority of targets. 25006 25007 -- Target Hook: enum machine_mode TARGET_LIBGCC_SHIFT_COUNT_MODE () 25008 This target hook should return the mode to be used for the shift 25009 count operand of shift instructions expanded to libgcc calls. If 25010 not defined `word_mode' is returned which is the right choice for 25011 a majority of targets. 25012 25013 -- Macro: ROUND_TOWARDS_ZERO 25014 If defined, this macro should be true if the prevailing rounding 25015 mode is towards zero. 25016 25017 Defining this macro only affects the way `libgcc.a' emulates 25018 floating-point arithmetic. 25019 25020 Not defining this macro is equivalent to returning zero. 25021 25022 -- Macro: LARGEST_EXPONENT_IS_NORMAL (SIZE) 25023 This macro should return true if floats with SIZE bits do not have 25024 a NaN or infinity representation, but use the largest exponent for 25025 normal numbers instead. 25026 25027 Defining this macro only affects the way `libgcc.a' emulates 25028 floating-point arithmetic. 25029 25030 The default definition of this macro returns false for all sizes. 25031 25032 -- Target Hook: bool TARGET_VECTOR_OPAQUE_P (tree TYPE) 25033 This target hook should return `true' a vector is opaque. That 25034 is, if no cast is needed when copying a vector value of type TYPE 25035 into another vector lvalue of the same size. Vector opaque types 25036 cannot be initialized. The default is that there are no such 25037 types. 25038 25039 -- Target Hook: bool TARGET_MS_BITFIELD_LAYOUT_P (tree RECORD_TYPE) 25040 This target hook returns `true' if bit-fields in the given 25041 RECORD_TYPE are to be laid out following the rules of Microsoft 25042 Visual C/C++, namely: (i) a bit-field won't share the same storage 25043 unit with the previous bit-field if their underlying types have 25044 different sizes, and the bit-field will be aligned to the highest 25045 alignment of the underlying types of itself and of the previous 25046 bit-field; (ii) a zero-sized bit-field will affect the alignment of 25047 the whole enclosing structure, even if it is unnamed; except that 25048 (iii) a zero-sized bit-field will be disregarded unless it follows 25049 another bit-field of nonzero size. If this hook returns `true', 25050 other macros that control bit-field layout are ignored. 25051 25052 When a bit-field is inserted into a packed record, the whole size 25053 of the underlying type is used by one or more same-size adjacent 25054 bit-fields (that is, if its long:3, 32 bits is used in the record, 25055 and any additional adjacent long bit-fields are packed into the 25056 same chunk of 32 bits. However, if the size changes, a new field 25057 of that size is allocated). In an unpacked record, this is the 25058 same as using alignment, but not equivalent when packing. 25059 25060 If both MS bit-fields and `__attribute__((packed))' are used, the 25061 latter will take precedence. If `__attribute__((packed))' is used 25062 on a single field when MS bit-fields are in use, it will take 25063 precedence for that field, but the alignment of the rest of the 25064 structure may affect its placement. 25065 25066 -- Target Hook: bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void) 25067 Returns true if the target supports decimal floating point. 25068 25069 -- Target Hook: bool TARGET_FIXED_POINT_SUPPORTED_P (void) 25070 Returns true if the target supports fixed-point arithmetic. 25071 25072 -- Target Hook: void TARGET_EXPAND_TO_RTL_HOOK (void) 25073 This hook is called just before expansion into rtl, allowing the 25074 target to perform additional initializations or analysis before 25075 the expansion. For example, the rs6000 port uses it to allocate a 25076 scratch stack slot for use in copying SDmode values between memory 25077 and floating point registers whenever the function being expanded 25078 has any SDmode usage. 25079 25080 -- Target Hook: void TARGET_INSTANTIATE_DECLS (void) 25081 This hook allows the backend to perform additional instantiations 25082 on rtl that are not actually in any insns yet, but will be later. 25083 25084 -- Target Hook: const char * TARGET_MANGLE_TYPE (tree TYPE) 25085 If your target defines any fundamental types, or any types your 25086 target uses should be mangled differently from the default, define 25087 this hook to return the appropriate encoding for these types as 25088 part of a C++ mangled name. The TYPE argument is the tree 25089 structure representing the type to be mangled. The hook may be 25090 applied to trees which are not target-specific fundamental types; 25091 it should return `NULL' for all such types, as well as arguments 25092 it does not recognize. If the return value is not `NULL', it must 25093 point to a statically-allocated string constant. 25094 25095 Target-specific fundamental types might be new fundamental types or 25096 qualified versions of ordinary fundamental types. Encode new 25097 fundamental types as `u N NAME', where NAME is the name used for 25098 the type in source code, and N is the length of NAME in decimal. 25099 Encode qualified versions of ordinary types as `U N NAME CODE', 25100 where NAME is the name used for the type qualifier in source code, 25101 N is the length of NAME as above, and CODE is the code used to 25102 represent the unqualified version of this type. (See 25103 `write_builtin_type' in `cp/mangle.c' for the list of codes.) In 25104 both cases the spaces are for clarity; do not include any spaces 25105 in your string. 25106 25107 This hook is applied to types prior to typedef resolution. If the 25108 mangled name for a particular type depends only on that type's 25109 main variant, you can perform typedef resolution yourself using 25110 `TYPE_MAIN_VARIANT' before mangling. 25111 25112 The default version of this hook always returns `NULL', which is 25113 appropriate for a target that does not define any new fundamental 25114 types. 25115 25116 25117 File: gccint.info, Node: Type Layout, Next: Registers, Prev: Storage Layout, Up: Target Macros 25118 25119 17.6 Layout of Source Language Data Types 25120 ========================================= 25121 25122 These macros define the sizes and other characteristics of the standard 25123 basic data types used in programs being compiled. Unlike the macros in 25124 the previous section, these apply to specific features of C and related 25125 languages, rather than to fundamental aspects of storage layout. 25126 25127 -- Macro: INT_TYPE_SIZE 25128 A C expression for the size in bits of the type `int' on the 25129 target machine. If you don't define this, the default is one word. 25130 25131 -- Macro: SHORT_TYPE_SIZE 25132 A C expression for the size in bits of the type `short' on the 25133 target machine. If you don't define this, the default is half a 25134 word. (If this would be less than one storage unit, it is rounded 25135 up to one unit.) 25136 25137 -- Macro: LONG_TYPE_SIZE 25138 A C expression for the size in bits of the type `long' on the 25139 target machine. If you don't define this, the default is one word. 25140 25141 -- Macro: ADA_LONG_TYPE_SIZE 25142 On some machines, the size used for the Ada equivalent of the type 25143 `long' by a native Ada compiler differs from that used by C. In 25144 that situation, define this macro to be a C expression to be used 25145 for the size of that type. If you don't define this, the default 25146 is the value of `LONG_TYPE_SIZE'. 25147 25148 -- Macro: LONG_LONG_TYPE_SIZE 25149 A C expression for the size in bits of the type `long long' on the 25150 target machine. If you don't define this, the default is two 25151 words. If you want to support GNU Ada on your machine, the value 25152 of this macro must be at least 64. 25153 25154 -- Macro: CHAR_TYPE_SIZE 25155 A C expression for the size in bits of the type `char' on the 25156 target machine. If you don't define this, the default is 25157 `BITS_PER_UNIT'. 25158 25159 -- Macro: BOOL_TYPE_SIZE 25160 A C expression for the size in bits of the C++ type `bool' and C99 25161 type `_Bool' on the target machine. If you don't define this, and 25162 you probably shouldn't, the default is `CHAR_TYPE_SIZE'. 25163 25164 -- Macro: FLOAT_TYPE_SIZE 25165 A C expression for the size in bits of the type `float' on the 25166 target machine. If you don't define this, the default is one word. 25167 25168 -- Macro: DOUBLE_TYPE_SIZE 25169 A C expression for the size in bits of the type `double' on the 25170 target machine. If you don't define this, the default is two 25171 words. 25172 25173 -- Macro: LONG_DOUBLE_TYPE_SIZE 25174 A C expression for the size in bits of the type `long double' on 25175 the target machine. If you don't define this, the default is two 25176 words. 25177 25178 -- Macro: SHORT_FRACT_TYPE_SIZE 25179 A C expression for the size in bits of the type `short _Fract' on 25180 the target machine. If you don't define this, the default is 25181 `BITS_PER_UNIT'. 25182 25183 -- Macro: FRACT_TYPE_SIZE 25184 A C expression for the size in bits of the type `_Fract' on the 25185 target machine. If you don't define this, the default is 25186 `BITS_PER_UNIT * 2'. 25187 25188 -- Macro: LONG_FRACT_TYPE_SIZE 25189 A C expression for the size in bits of the type `long _Fract' on 25190 the target machine. If you don't define this, the default is 25191 `BITS_PER_UNIT * 4'. 25192 25193 -- Macro: LONG_LONG_FRACT_TYPE_SIZE 25194 A C expression for the size in bits of the type `long long _Fract' 25195 on the target machine. If you don't define this, the default is 25196 `BITS_PER_UNIT * 8'. 25197 25198 -- Macro: SHORT_ACCUM_TYPE_SIZE 25199 A C expression for the size in bits of the type `short _Accum' on 25200 the target machine. If you don't define this, the default is 25201 `BITS_PER_UNIT * 2'. 25202 25203 -- Macro: ACCUM_TYPE_SIZE 25204 A C expression for the size in bits of the type `_Accum' on the 25205 target machine. If you don't define this, the default is 25206 `BITS_PER_UNIT * 4'. 25207 25208 -- Macro: LONG_ACCUM_TYPE_SIZE 25209 A C expression for the size in bits of the type `long _Accum' on 25210 the target machine. If you don't define this, the default is 25211 `BITS_PER_UNIT * 8'. 25212 25213 -- Macro: LONG_LONG_ACCUM_TYPE_SIZE 25214 A C expression for the size in bits of the type `long long _Accum' 25215 on the target machine. If you don't define this, the default is 25216 `BITS_PER_UNIT * 16'. 25217 25218 -- Macro: LIBGCC2_LONG_DOUBLE_TYPE_SIZE 25219 Define this macro if `LONG_DOUBLE_TYPE_SIZE' is not constant or if 25220 you want routines in `libgcc2.a' for a size other than 25221 `LONG_DOUBLE_TYPE_SIZE'. If you don't define this, the default is 25222 `LONG_DOUBLE_TYPE_SIZE'. 25223 25224 -- Macro: LIBGCC2_HAS_DF_MODE 25225 Define this macro if neither `LIBGCC2_DOUBLE_TYPE_SIZE' nor 25226 `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is `DFmode' but you want `DFmode' 25227 routines in `libgcc2.a' anyway. If you don't define this and 25228 either `LIBGCC2_DOUBLE_TYPE_SIZE' or 25229 `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 64 then the default is 1, 25230 otherwise it is 0. 25231 25232 -- Macro: LIBGCC2_HAS_XF_MODE 25233 Define this macro if `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is not 25234 `XFmode' but you want `XFmode' routines in `libgcc2.a' anyway. If 25235 you don't define this and `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 80 25236 then the default is 1, otherwise it is 0. 25237 25238 -- Macro: LIBGCC2_HAS_TF_MODE 25239 Define this macro if `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is not 25240 `TFmode' but you want `TFmode' routines in `libgcc2.a' anyway. If 25241 you don't define this and `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 128 25242 then the default is 1, otherwise it is 0. 25243 25244 -- Macro: SF_SIZE 25245 -- Macro: DF_SIZE 25246 -- Macro: XF_SIZE 25247 -- Macro: TF_SIZE 25248 Define these macros to be the size in bits of the mantissa of 25249 `SFmode', `DFmode', `XFmode' and `TFmode' values, if the defaults 25250 in `libgcc2.h' are inappropriate. By default, `FLT_MANT_DIG' is 25251 used for `SF_SIZE', `LDBL_MANT_DIG' for `XF_SIZE' and `TF_SIZE', 25252 and `DBL_MANT_DIG' or `LDBL_MANT_DIG' for `DF_SIZE' according to 25253 whether `LIBGCC2_DOUBLE_TYPE_SIZE' or 25254 `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 64. 25255 25256 -- Macro: TARGET_FLT_EVAL_METHOD 25257 A C expression for the value for `FLT_EVAL_METHOD' in `float.h', 25258 assuming, if applicable, that the floating-point control word is 25259 in its default state. If you do not define this macro the value of 25260 `FLT_EVAL_METHOD' will be zero. 25261 25262 -- Macro: WIDEST_HARDWARE_FP_SIZE 25263 A C expression for the size in bits of the widest floating-point 25264 format supported by the hardware. If you define this macro, you 25265 must specify a value less than or equal to the value of 25266 `LONG_DOUBLE_TYPE_SIZE'. If you do not define this macro, the 25267 value of `LONG_DOUBLE_TYPE_SIZE' is the default. 25268 25269 -- Macro: DEFAULT_SIGNED_CHAR 25270 An expression whose value is 1 or 0, according to whether the type 25271 `char' should be signed or unsigned by default. The user can 25272 always override this default with the options `-fsigned-char' and 25273 `-funsigned-char'. 25274 25275 -- Target Hook: bool TARGET_DEFAULT_SHORT_ENUMS (void) 25276 This target hook should return true if the compiler should give an 25277 `enum' type only as many bytes as it takes to represent the range 25278 of possible values of that type. It should return false if all 25279 `enum' types should be allocated like `int'. 25280 25281 The default is to return false. 25282 25283 -- Macro: SIZE_TYPE 25284 A C expression for a string describing the name of the data type 25285 to use for size values. The typedef name `size_t' is defined 25286 using the contents of the string. 25287 25288 The string can contain more than one keyword. If so, separate 25289 them with spaces, and write first any length keyword, then 25290 `unsigned' if appropriate, and finally `int'. The string must 25291 exactly match one of the data type names defined in the function 25292 `init_decl_processing' in the file `c-decl.c'. You may not omit 25293 `int' or change the order--that would cause the compiler to crash 25294 on startup. 25295 25296 If you don't define this macro, the default is `"long unsigned 25297 int"'. 25298 25299 -- Macro: PTRDIFF_TYPE 25300 A C expression for a string describing the name of the data type 25301 to use for the result of subtracting two pointers. The typedef 25302 name `ptrdiff_t' is defined using the contents of the string. See 25303 `SIZE_TYPE' above for more information. 25304 25305 If you don't define this macro, the default is `"long int"'. 25306 25307 -- Macro: WCHAR_TYPE 25308 A C expression for a string describing the name of the data type 25309 to use for wide characters. The typedef name `wchar_t' is defined 25310 using the contents of the string. See `SIZE_TYPE' above for more 25311 information. 25312 25313 If you don't define this macro, the default is `"int"'. 25314 25315 -- Macro: WCHAR_TYPE_SIZE 25316 A C expression for the size in bits of the data type for wide 25317 characters. This is used in `cpp', which cannot make use of 25318 `WCHAR_TYPE'. 25319 25320 -- Macro: WINT_TYPE 25321 A C expression for a string describing the name of the data type to 25322 use for wide characters passed to `printf' and returned from 25323 `getwc'. The typedef name `wint_t' is defined using the contents 25324 of the string. See `SIZE_TYPE' above for more information. 25325 25326 If you don't define this macro, the default is `"unsigned int"'. 25327 25328 -- Macro: INTMAX_TYPE 25329 A C expression for a string describing the name of the data type 25330 that can represent any value of any standard or extended signed 25331 integer type. The typedef name `intmax_t' is defined using the 25332 contents of the string. See `SIZE_TYPE' above for more 25333 information. 25334 25335 If you don't define this macro, the default is the first of 25336 `"int"', `"long int"', or `"long long int"' that has as much 25337 precision as `long long int'. 25338 25339 -- Macro: UINTMAX_TYPE 25340 A C expression for a string describing the name of the data type 25341 that can represent any value of any standard or extended unsigned 25342 integer type. The typedef name `uintmax_t' is defined using the 25343 contents of the string. See `SIZE_TYPE' above for more 25344 information. 25345 25346 If you don't define this macro, the default is the first of 25347 `"unsigned int"', `"long unsigned int"', or `"long long unsigned 25348 int"' that has as much precision as `long long unsigned int'. 25349 25350 -- Macro: TARGET_PTRMEMFUNC_VBIT_LOCATION 25351 The C++ compiler represents a pointer-to-member-function with a 25352 struct that looks like: 25353 25354 struct { 25355 union { 25356 void (*fn)(); 25357 ptrdiff_t vtable_index; 25358 }; 25359 ptrdiff_t delta; 25360 }; 25361 25362 The C++ compiler must use one bit to indicate whether the function 25363 that will be called through a pointer-to-member-function is 25364 virtual. Normally, we assume that the low-order bit of a function 25365 pointer must always be zero. Then, by ensuring that the 25366 vtable_index is odd, we can distinguish which variant of the union 25367 is in use. But, on some platforms function pointers can be odd, 25368 and so this doesn't work. In that case, we use the low-order bit 25369 of the `delta' field, and shift the remainder of the `delta' field 25370 to the left. 25371 25372 GCC will automatically make the right selection about where to 25373 store this bit using the `FUNCTION_BOUNDARY' setting for your 25374 platform. However, some platforms such as ARM/Thumb have 25375 `FUNCTION_BOUNDARY' set such that functions always start at even 25376 addresses, but the lowest bit of pointers to functions indicate 25377 whether the function at that address is in ARM or Thumb mode. If 25378 this is the case of your architecture, you should define this 25379 macro to `ptrmemfunc_vbit_in_delta'. 25380 25381 In general, you should not have to define this macro. On 25382 architectures in which function addresses are always even, 25383 according to `FUNCTION_BOUNDARY', GCC will automatically define 25384 this macro to `ptrmemfunc_vbit_in_pfn'. 25385 25386 -- Macro: TARGET_VTABLE_USES_DESCRIPTORS 25387 Normally, the C++ compiler uses function pointers in vtables. This 25388 macro allows the target to change to use "function descriptors" 25389 instead. Function descriptors are found on targets for whom a 25390 function pointer is actually a small data structure. Normally the 25391 data structure consists of the actual code address plus a data 25392 pointer to which the function's data is relative. 25393 25394 If vtables are used, the value of this macro should be the number 25395 of words that the function descriptor occupies. 25396 25397 -- Macro: TARGET_VTABLE_ENTRY_ALIGN 25398 By default, the vtable entries are void pointers, the so the 25399 alignment is the same as pointer alignment. The value of this 25400 macro specifies the alignment of the vtable entry in bits. It 25401 should be defined only when special alignment is necessary. */ 25402 25403 -- Macro: TARGET_VTABLE_DATA_ENTRY_DISTANCE 25404 There are a few non-descriptor entries in the vtable at offsets 25405 below zero. If these entries must be padded (say, to preserve the 25406 alignment specified by `TARGET_VTABLE_ENTRY_ALIGN'), set this to 25407 the number of words in each data entry. 25408 25409 25410 File: gccint.info, Node: Registers, Next: Register Classes, Prev: Type Layout, Up: Target Macros 25411 25412 17.7 Register Usage 25413 =================== 25414 25415 This section explains how to describe what registers the target machine 25416 has, and how (in general) they can be used. 25417 25418 The description of which registers a specific instruction can use is 25419 done with register classes; see *Note Register Classes::. For 25420 information on using registers to access a stack frame, see *Note Frame 25421 Registers::. For passing values in registers, see *Note Register 25422 Arguments::. For returning values in registers, see *Note Scalar 25423 Return::. 25424 25425 * Menu: 25426 25427 * Register Basics:: Number and kinds of registers. 25428 * Allocation Order:: Order in which registers are allocated. 25429 * Values in Registers:: What kinds of values each reg can hold. 25430 * Leaf Functions:: Renumbering registers for leaf functions. 25431 * Stack Registers:: Handling a register stack such as 80387. 25432 25433 25434 File: gccint.info, Node: Register Basics, Next: Allocation Order, Up: Registers 25435 25436 17.7.1 Basic Characteristics of Registers 25437 ----------------------------------------- 25438 25439 Registers have various characteristics. 25440 25441 -- Macro: FIRST_PSEUDO_REGISTER 25442 Number of hardware registers known to the compiler. They receive 25443 numbers 0 through `FIRST_PSEUDO_REGISTER-1'; thus, the first 25444 pseudo register's number really is assigned the number 25445 `FIRST_PSEUDO_REGISTER'. 25446 25447 -- Macro: FIXED_REGISTERS 25448 An initializer that says which registers are used for fixed 25449 purposes all throughout the compiled code and are therefore not 25450 available for general allocation. These would include the stack 25451 pointer, the frame pointer (except on machines where that can be 25452 used as a general register when no frame pointer is needed), the 25453 program counter on machines where that is considered one of the 25454 addressable registers, and any other numbered register with a 25455 standard use. 25456 25457 This information is expressed as a sequence of numbers, separated 25458 by commas and surrounded by braces. The Nth number is 1 if 25459 register N is fixed, 0 otherwise. 25460 25461 The table initialized from this macro, and the table initialized by 25462 the following one, may be overridden at run time either 25463 automatically, by the actions of the macro 25464 `CONDITIONAL_REGISTER_USAGE', or by the user with the command 25465 options `-ffixed-REG', `-fcall-used-REG' and `-fcall-saved-REG'. 25466 25467 -- Macro: CALL_USED_REGISTERS 25468 Like `FIXED_REGISTERS' but has 1 for each register that is 25469 clobbered (in general) by function calls as well as for fixed 25470 registers. This macro therefore identifies the registers that are 25471 not available for general allocation of values that must live 25472 across function calls. 25473 25474 If a register has 0 in `CALL_USED_REGISTERS', the compiler 25475 automatically saves it on function entry and restores it on 25476 function exit, if the register is used within the function. 25477 25478 -- Macro: CALL_REALLY_USED_REGISTERS 25479 Like `CALL_USED_REGISTERS' except this macro doesn't require that 25480 the entire set of `FIXED_REGISTERS' be included. 25481 (`CALL_USED_REGISTERS' must be a superset of `FIXED_REGISTERS'). 25482 This macro is optional. If not specified, it defaults to the value 25483 of `CALL_USED_REGISTERS'. 25484 25485 -- Macro: HARD_REGNO_CALL_PART_CLOBBERED (REGNO, MODE) 25486 A C expression that is nonzero if it is not permissible to store a 25487 value of mode MODE in hard register number REGNO across a call 25488 without some part of it being clobbered. For most machines this 25489 macro need not be defined. It is only required for machines that 25490 do not preserve the entire contents of a register across a call. 25491 25492 -- Macro: CONDITIONAL_REGISTER_USAGE 25493 Zero or more C statements that may conditionally modify five 25494 variables `fixed_regs', `call_used_regs', `global_regs', 25495 `reg_names', and `reg_class_contents', to take into account any 25496 dependence of these register sets on target flags. The first three 25497 of these are of type `char []' (interpreted as Boolean vectors). 25498 `global_regs' is a `const char *[]', and `reg_class_contents' is a 25499 `HARD_REG_SET'. Before the macro is called, `fixed_regs', 25500 `call_used_regs', `reg_class_contents', and `reg_names' have been 25501 initialized from `FIXED_REGISTERS', `CALL_USED_REGISTERS', 25502 `REG_CLASS_CONTENTS', and `REGISTER_NAMES', respectively. 25503 `global_regs' has been cleared, and any `-ffixed-REG', 25504 `-fcall-used-REG' and `-fcall-saved-REG' command options have been 25505 applied. 25506 25507 You need not define this macro if it has no work to do. 25508 25509 If the usage of an entire class of registers depends on the target 25510 flags, you may indicate this to GCC by using this macro to modify 25511 `fixed_regs' and `call_used_regs' to 1 for each of the registers 25512 in the classes which should not be used by GCC. Also define the 25513 macro `REG_CLASS_FROM_LETTER' / `REG_CLASS_FROM_CONSTRAINT' to 25514 return `NO_REGS' if it is called with a letter for a class that 25515 shouldn't be used. 25516 25517 (However, if this class is not included in `GENERAL_REGS' and all 25518 of the insn patterns whose constraints permit this class are 25519 controlled by target switches, then GCC will automatically avoid 25520 using these registers when the target switches are opposed to 25521 them.) 25522 25523 -- Macro: INCOMING_REGNO (OUT) 25524 Define this macro if the target machine has register windows. 25525 This C expression returns the register number as seen by the 25526 called function corresponding to the register number OUT as seen 25527 by the calling function. Return OUT if register number OUT is not 25528 an outbound register. 25529 25530 -- Macro: OUTGOING_REGNO (IN) 25531 Define this macro if the target machine has register windows. 25532 This C expression returns the register number as seen by the 25533 calling function corresponding to the register number IN as seen 25534 by the called function. Return IN if register number IN is not an 25535 inbound register. 25536 25537 -- Macro: LOCAL_REGNO (REGNO) 25538 Define this macro if the target machine has register windows. 25539 This C expression returns true if the register is call-saved but 25540 is in the register window. Unlike most call-saved registers, such 25541 registers need not be explicitly restored on function exit or 25542 during non-local gotos. 25543 25544 -- Macro: PC_REGNUM 25545 If the program counter has a register number, define this as that 25546 register number. Otherwise, do not define it. 25547 25548 25549 File: gccint.info, Node: Allocation Order, Next: Values in Registers, Prev: Register Basics, Up: Registers 25550 25551 17.7.2 Order of Allocation of Registers 25552 --------------------------------------- 25553 25554 Registers are allocated in order. 25555 25556 -- Macro: REG_ALLOC_ORDER 25557 If defined, an initializer for a vector of integers, containing the 25558 numbers of hard registers in the order in which GCC should prefer 25559 to use them (from most preferred to least). 25560 25561 If this macro is not defined, registers are used lowest numbered 25562 first (all else being equal). 25563 25564 One use of this macro is on machines where the highest numbered 25565 registers must always be saved and the save-multiple-registers 25566 instruction supports only sequences of consecutive registers. On 25567 such machines, define `REG_ALLOC_ORDER' to be an initializer that 25568 lists the highest numbered allocable register first. 25569 25570 -- Macro: ADJUST_REG_ALLOC_ORDER 25571 A C statement (sans semicolon) to choose the order in which to 25572 allocate hard registers for pseudo-registers local to a basic 25573 block. 25574 25575 Store the desired register order in the array `reg_alloc_order'. 25576 Element 0 should be the register to allocate first; element 1, the 25577 next register; and so on. 25578 25579 The macro body should not assume anything about the contents of 25580 `reg_alloc_order' before execution of the macro. 25581 25582 On most machines, it is not necessary to define this macro. 25583 25584 -- Macro: HONOR_REG_ALLOC_ORDER 25585 Normally, IRA tries to estimate the costs for saving a register in 25586 the prologue and restoring it in the epilogue. This discourages 25587 it from using call-saved registers. If a machine wants to ensure 25588 that IRA allocates registers in the order given by REG_ALLOC_ORDER 25589 even if some call-saved registers appear earlier than call-used 25590 ones, this macro should be defined. 25591 25592 -- Macro: IRA_HARD_REGNO_ADD_COST_MULTIPLIER (REGNO) 25593 In some case register allocation order is not enough for the 25594 Integrated Register Allocator (IRA) to generate a good code. If 25595 this macro is defined, it should return a floating point value 25596 based on REGNO. The cost of using REGNO for a pseudo will be 25597 increased by approximately the pseudo's usage frequency times the 25598 value returned by this macro. Not defining this macro is 25599 equivalent to having it always return `0.0'. 25600 25601 On most machines, it is not necessary to define this macro. 25602 25603 25604 File: gccint.info, Node: Values in Registers, Next: Leaf Functions, Prev: Allocation Order, Up: Registers 25605 25606 17.7.3 How Values Fit in Registers 25607 ---------------------------------- 25608 25609 This section discusses the macros that describe which kinds of values 25610 (specifically, which machine modes) each register can hold, and how many 25611 consecutive registers are needed for a given mode. 25612 25613 -- Macro: HARD_REGNO_NREGS (REGNO, MODE) 25614 A C expression for the number of consecutive hard registers, 25615 starting at register number REGNO, required to hold a value of mode 25616 MODE. This macro must never return zero, even if a register 25617 cannot hold the requested mode - indicate that with 25618 HARD_REGNO_MODE_OK and/or CANNOT_CHANGE_MODE_CLASS instead. 25619 25620 On a machine where all registers are exactly one word, a suitable 25621 definition of this macro is 25622 25623 #define HARD_REGNO_NREGS(REGNO, MODE) \ 25624 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ 25625 / UNITS_PER_WORD) 25626 25627 -- Macro: HARD_REGNO_NREGS_HAS_PADDING (REGNO, MODE) 25628 A C expression that is nonzero if a value of mode MODE, stored in 25629 memory, ends with padding that causes it to take up more space than 25630 in registers starting at register number REGNO (as determined by 25631 multiplying GCC's notion of the size of the register when 25632 containing this mode by the number of registers returned by 25633 `HARD_REGNO_NREGS'). By default this is zero. 25634 25635 For example, if a floating-point value is stored in three 32-bit 25636 registers but takes up 128 bits in memory, then this would be 25637 nonzero. 25638 25639 This macros only needs to be defined if there are cases where 25640 `subreg_get_info' would otherwise wrongly determine that a 25641 `subreg' can be represented by an offset to the register number, 25642 when in fact such a `subreg' would contain some of the padding not 25643 stored in registers and so not be representable. 25644 25645 -- Macro: HARD_REGNO_NREGS_WITH_PADDING (REGNO, MODE) 25646 For values of REGNO and MODE for which 25647 `HARD_REGNO_NREGS_HAS_PADDING' returns nonzero, a C expression 25648 returning the greater number of registers required to hold the 25649 value including any padding. In the example above, the value 25650 would be four. 25651 25652 -- Macro: REGMODE_NATURAL_SIZE (MODE) 25653 Define this macro if the natural size of registers that hold values 25654 of mode MODE is not the word size. It is a C expression that 25655 should give the natural size in bytes for the specified mode. It 25656 is used by the register allocator to try to optimize its results. 25657 This happens for example on SPARC 64-bit where the natural size of 25658 floating-point registers is still 32-bit. 25659 25660 -- Macro: HARD_REGNO_MODE_OK (REGNO, MODE) 25661 A C expression that is nonzero if it is permissible to store a 25662 value of mode MODE in hard register number REGNO (or in several 25663 registers starting with that one). For a machine where all 25664 registers are equivalent, a suitable definition is 25665 25666 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1 25667 25668 You need not include code to check for the numbers of fixed 25669 registers, because the allocation mechanism considers them to be 25670 always occupied. 25671 25672 On some machines, double-precision values must be kept in even/odd 25673 register pairs. You can implement that by defining this macro to 25674 reject odd register numbers for such modes. 25675 25676 The minimum requirement for a mode to be OK in a register is that 25677 the `movMODE' instruction pattern support moves between the 25678 register and other hard register in the same class and that moving 25679 a value into the register and back out not alter it. 25680 25681 Since the same instruction used to move `word_mode' will work for 25682 all narrower integer modes, it is not necessary on any machine for 25683 `HARD_REGNO_MODE_OK' to distinguish between these modes, provided 25684 you define patterns `movhi', etc., to take advantage of this. This 25685 is useful because of the interaction between `HARD_REGNO_MODE_OK' 25686 and `MODES_TIEABLE_P'; it is very desirable for all integer modes 25687 to be tieable. 25688 25689 Many machines have special registers for floating point arithmetic. 25690 Often people assume that floating point machine modes are allowed 25691 only in floating point registers. This is not true. Any 25692 registers that can hold integers can safely _hold_ a floating 25693 point machine mode, whether or not floating arithmetic can be done 25694 on it in those registers. Integer move instructions can be used 25695 to move the values. 25696 25697 On some machines, though, the converse is true: fixed-point machine 25698 modes may not go in floating registers. This is true if the 25699 floating registers normalize any value stored in them, because 25700 storing a non-floating value there would garble it. In this case, 25701 `HARD_REGNO_MODE_OK' should reject fixed-point machine modes in 25702 floating registers. But if the floating registers do not 25703 automatically normalize, if you can store any bit pattern in one 25704 and retrieve it unchanged without a trap, then any machine mode 25705 may go in a floating register, so you can define this macro to say 25706 so. 25707 25708 The primary significance of special floating registers is rather 25709 that they are the registers acceptable in floating point arithmetic 25710 instructions. However, this is of no concern to 25711 `HARD_REGNO_MODE_OK'. You handle it by writing the proper 25712 constraints for those instructions. 25713 25714 On some machines, the floating registers are especially slow to 25715 access, so that it is better to store a value in a stack frame 25716 than in such a register if floating point arithmetic is not being 25717 done. As long as the floating registers are not in class 25718 `GENERAL_REGS', they will not be used unless some pattern's 25719 constraint asks for one. 25720 25721 -- Macro: HARD_REGNO_RENAME_OK (FROM, TO) 25722 A C expression that is nonzero if it is OK to rename a hard 25723 register FROM to another hard register TO. 25724 25725 One common use of this macro is to prevent renaming of a register 25726 to another register that is not saved by a prologue in an interrupt 25727 handler. 25728 25729 The default is always nonzero. 25730 25731 -- Macro: MODES_TIEABLE_P (MODE1, MODE2) 25732 A C expression that is nonzero if a value of mode MODE1 is 25733 accessible in mode MODE2 without copying. 25734 25735 If `HARD_REGNO_MODE_OK (R, MODE1)' and `HARD_REGNO_MODE_OK (R, 25736 MODE2)' are always the same for any R, then `MODES_TIEABLE_P 25737 (MODE1, MODE2)' should be nonzero. If they differ for any R, you 25738 should define this macro to return zero unless some other 25739 mechanism ensures the accessibility of the value in a narrower 25740 mode. 25741 25742 You should define this macro to return nonzero in as many cases as 25743 possible since doing so will allow GCC to perform better register 25744 allocation. 25745 25746 -- Target Hook: bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int REGNO) 25747 This target hook should return `true' if it is OK to use a hard 25748 register REGNO as scratch reg in peephole2. 25749 25750 One common use of this macro is to prevent using of a register that 25751 is not saved by a prologue in an interrupt handler. 25752 25753 The default version of this hook always returns `true'. 25754 25755 -- Macro: AVOID_CCMODE_COPIES 25756 Define this macro if the compiler should avoid copies to/from 25757 `CCmode' registers. You should only define this macro if support 25758 for copying to/from `CCmode' is incomplete. 25759 25760 25761 File: gccint.info, Node: Leaf Functions, Next: Stack Registers, Prev: Values in Registers, Up: Registers 25762 25763 17.7.4 Handling Leaf Functions 25764 ------------------------------ 25765 25766 On some machines, a leaf function (i.e., one which makes no calls) can 25767 run more efficiently if it does not make its own register window. 25768 Often this means it is required to receive its arguments in the 25769 registers where they are passed by the caller, instead of the registers 25770 where they would normally arrive. 25771 25772 The special treatment for leaf functions generally applies only when 25773 other conditions are met; for example, often they may use only those 25774 registers for its own variables and temporaries. We use the term "leaf 25775 function" to mean a function that is suitable for this special 25776 handling, so that functions with no calls are not necessarily "leaf 25777 functions". 25778 25779 GCC assigns register numbers before it knows whether the function is 25780 suitable for leaf function treatment. So it needs to renumber the 25781 registers in order to output a leaf function. The following macros 25782 accomplish this. 25783 25784 -- Macro: LEAF_REGISTERS 25785 Name of a char vector, indexed by hard register number, which 25786 contains 1 for a register that is allowable in a candidate for leaf 25787 function treatment. 25788 25789 If leaf function treatment involves renumbering the registers, 25790 then the registers marked here should be the ones before 25791 renumbering--those that GCC would ordinarily allocate. The 25792 registers which will actually be used in the assembler code, after 25793 renumbering, should not be marked with 1 in this vector. 25794 25795 Define this macro only if the target machine offers a way to 25796 optimize the treatment of leaf functions. 25797 25798 -- Macro: LEAF_REG_REMAP (REGNO) 25799 A C expression whose value is the register number to which REGNO 25800 should be renumbered, when a function is treated as a leaf 25801 function. 25802 25803 If REGNO is a register number which should not appear in a leaf 25804 function before renumbering, then the expression should yield -1, 25805 which will cause the compiler to abort. 25806 25807 Define this macro only if the target machine offers a way to 25808 optimize the treatment of leaf functions, and registers need to be 25809 renumbered to do this. 25810 25811 `TARGET_ASM_FUNCTION_PROLOGUE' and `TARGET_ASM_FUNCTION_EPILOGUE' must 25812 usually treat leaf functions specially. They can test the C variable 25813 `current_function_is_leaf' which is nonzero for leaf functions. 25814 `current_function_is_leaf' is set prior to local register allocation 25815 and is valid for the remaining compiler passes. They can also test the 25816 C variable `current_function_uses_only_leaf_regs' which is nonzero for 25817 leaf functions which only use leaf registers. 25818 `current_function_uses_only_leaf_regs' is valid after all passes that 25819 modify the instructions have been run and is only useful if 25820 `LEAF_REGISTERS' is defined. 25821 25822 25823 File: gccint.info, Node: Stack Registers, Prev: Leaf Functions, Up: Registers 25824 25825 17.7.5 Registers That Form a Stack 25826 ---------------------------------- 25827 25828 There are special features to handle computers where some of the 25829 "registers" form a stack. Stack registers are normally written by 25830 pushing onto the stack, and are numbered relative to the top of the 25831 stack. 25832 25833 Currently, GCC can only handle one group of stack-like registers, and 25834 they must be consecutively numbered. Furthermore, the existing support 25835 for stack-like registers is specific to the 80387 floating point 25836 coprocessor. If you have a new architecture that uses stack-like 25837 registers, you will need to do substantial work on `reg-stack.c' and 25838 write your machine description to cooperate with it, as well as 25839 defining these macros. 25840 25841 -- Macro: STACK_REGS 25842 Define this if the machine has any stack-like registers. 25843 25844 -- Macro: FIRST_STACK_REG 25845 The number of the first stack-like register. This one is the top 25846 of the stack. 25847 25848 -- Macro: LAST_STACK_REG 25849 The number of the last stack-like register. This one is the 25850 bottom of the stack. 25851 25852 25853 File: gccint.info, Node: Register Classes, Next: Old Constraints, Prev: Registers, Up: Target Macros 25854 25855 17.8 Register Classes 25856 ===================== 25857 25858 On many machines, the numbered registers are not all equivalent. For 25859 example, certain registers may not be allowed for indexed addressing; 25860 certain registers may not be allowed in some instructions. These 25861 machine restrictions are described to the compiler using "register 25862 classes". 25863 25864 You define a number of register classes, giving each one a name and 25865 saying which of the registers belong to it. Then you can specify 25866 register classes that are allowed as operands to particular instruction 25867 patterns. 25868 25869 In general, each register will belong to several classes. In fact, one 25870 class must be named `ALL_REGS' and contain all the registers. Another 25871 class must be named `NO_REGS' and contain no registers. Often the 25872 union of two classes will be another class; however, this is not 25873 required. 25874 25875 One of the classes must be named `GENERAL_REGS'. There is nothing 25876 terribly special about the name, but the operand constraint letters `r' 25877 and `g' specify this class. If `GENERAL_REGS' is the same as 25878 `ALL_REGS', just define it as a macro which expands to `ALL_REGS'. 25879 25880 Order the classes so that if class X is contained in class Y then X 25881 has a lower class number than Y. 25882 25883 The way classes other than `GENERAL_REGS' are specified in operand 25884 constraints is through machine-dependent operand constraint letters. 25885 You can define such letters to correspond to various classes, then use 25886 them in operand constraints. 25887 25888 You should define a class for the union of two classes whenever some 25889 instruction allows both classes. For example, if an instruction allows 25890 either a floating point (coprocessor) register or a general register 25891 for a certain operand, you should define a class `FLOAT_OR_GENERAL_REGS' 25892 which includes both of them. Otherwise you will get suboptimal code. 25893 25894 You must also specify certain redundant information about the register 25895 classes: for each class, which classes contain it and which ones are 25896 contained in it; for each pair of classes, the largest class contained 25897 in their union. 25898 25899 When a value occupying several consecutive registers is expected in a 25900 certain class, all the registers used must belong to that class. 25901 Therefore, register classes cannot be used to enforce a requirement for 25902 a register pair to start with an even-numbered register. The way to 25903 specify this requirement is with `HARD_REGNO_MODE_OK'. 25904 25905 Register classes used for input-operands of bitwise-and or shift 25906 instructions have a special requirement: each such class must have, for 25907 each fixed-point machine mode, a subclass whose registers can transfer 25908 that mode to or from memory. For example, on some machines, the 25909 operations for single-byte values (`QImode') are limited to certain 25910 registers. When this is so, each register class that is used in a 25911 bitwise-and or shift instruction must have a subclass consisting of 25912 registers from which single-byte values can be loaded or stored. This 25913 is so that `PREFERRED_RELOAD_CLASS' can always have a possible value to 25914 return. 25915 25916 -- Data type: enum reg_class 25917 An enumerated type that must be defined with all the register 25918 class names as enumerated values. `NO_REGS' must be first. 25919 `ALL_REGS' must be the last register class, followed by one more 25920 enumerated value, `LIM_REG_CLASSES', which is not a register class 25921 but rather tells how many classes there are. 25922 25923 Each register class has a number, which is the value of casting 25924 the class name to type `int'. The number serves as an index in 25925 many of the tables described below. 25926 25927 -- Macro: N_REG_CLASSES 25928 The number of distinct register classes, defined as follows: 25929 25930 #define N_REG_CLASSES (int) LIM_REG_CLASSES 25931 25932 -- Macro: REG_CLASS_NAMES 25933 An initializer containing the names of the register classes as C 25934 string constants. These names are used in writing some of the 25935 debugging dumps. 25936 25937 -- Macro: REG_CLASS_CONTENTS 25938 An initializer containing the contents of the register classes, as 25939 integers which are bit masks. The Nth integer specifies the 25940 contents of class N. The way the integer MASK is interpreted is 25941 that register R is in the class if `MASK & (1 << R)' is 1. 25942 25943 When the machine has more than 32 registers, an integer does not 25944 suffice. Then the integers are replaced by sub-initializers, 25945 braced groupings containing several integers. Each 25946 sub-initializer must be suitable as an initializer for the type 25947 `HARD_REG_SET' which is defined in `hard-reg-set.h'. In this 25948 situation, the first integer in each sub-initializer corresponds to 25949 registers 0 through 31, the second integer to registers 32 through 25950 63, and so on. 25951 25952 -- Macro: REGNO_REG_CLASS (REGNO) 25953 A C expression whose value is a register class containing hard 25954 register REGNO. In general there is more than one such class; 25955 choose a class which is "minimal", meaning that no smaller class 25956 also contains the register. 25957 25958 -- Macro: BASE_REG_CLASS 25959 A macro whose definition is the name of the class to which a valid 25960 base register must belong. A base register is one used in an 25961 address which is the register value plus a displacement. 25962 25963 -- Macro: MODE_BASE_REG_CLASS (MODE) 25964 This is a variation of the `BASE_REG_CLASS' macro which allows the 25965 selection of a base register in a mode dependent manner. If MODE 25966 is VOIDmode then it should return the same value as 25967 `BASE_REG_CLASS'. 25968 25969 -- Macro: MODE_BASE_REG_REG_CLASS (MODE) 25970 A C expression whose value is the register class to which a valid 25971 base register must belong in order to be used in a base plus index 25972 register address. You should define this macro if base plus index 25973 addresses have different requirements than other base register 25974 uses. 25975 25976 -- Macro: MODE_CODE_BASE_REG_CLASS (MODE, OUTER_CODE, INDEX_CODE) 25977 A C expression whose value is the register class to which a valid 25978 base register must belong. OUTER_CODE and INDEX_CODE define the 25979 context in which the base register occurs. OUTER_CODE is the code 25980 of the immediately enclosing expression (`MEM' for the top level 25981 of an address, `ADDRESS' for something that occurs in an 25982 `address_operand'). INDEX_CODE is the code of the corresponding 25983 index expression if OUTER_CODE is `PLUS'; `SCRATCH' otherwise. 25984 25985 -- Macro: INDEX_REG_CLASS 25986 A macro whose definition is the name of the class to which a valid 25987 index register must belong. An index register is one used in an 25988 address where its value is either multiplied by a scale factor or 25989 added to another register (as well as added to a displacement). 25990 25991 -- Macro: REGNO_OK_FOR_BASE_P (NUM) 25992 A C expression which is nonzero if register number NUM is suitable 25993 for use as a base register in operand addresses. It may be either 25994 a suitable hard register or a pseudo register that has been 25995 allocated such a hard register. 25996 25997 -- Macro: REGNO_MODE_OK_FOR_BASE_P (NUM, MODE) 25998 A C expression that is just like `REGNO_OK_FOR_BASE_P', except that 25999 that expression may examine the mode of the memory reference in 26000 MODE. You should define this macro if the mode of the memory 26001 reference affects whether a register may be used as a base 26002 register. If you define this macro, the compiler will use it 26003 instead of `REGNO_OK_FOR_BASE_P'. The mode may be `VOIDmode' for 26004 addresses that appear outside a `MEM', i.e., as an 26005 `address_operand'. 26006 26007 26008 -- Macro: REGNO_MODE_OK_FOR_REG_BASE_P (NUM, MODE) 26009 A C expression which is nonzero if register number NUM is suitable 26010 for use as a base register in base plus index operand addresses, 26011 accessing memory in mode MODE. It may be either a suitable hard 26012 register or a pseudo register that has been allocated such a hard 26013 register. You should define this macro if base plus index 26014 addresses have different requirements than other base register 26015 uses. 26016 26017 Use of this macro is deprecated; please use the more general 26018 `REGNO_MODE_CODE_OK_FOR_BASE_P'. 26019 26020 -- Macro: REGNO_MODE_CODE_OK_FOR_BASE_P (NUM, MODE, OUTER_CODE, 26021 INDEX_CODE) 26022 A C expression that is just like `REGNO_MODE_OK_FOR_BASE_P', except 26023 that that expression may examine the context in which the register 26024 appears in the memory reference. OUTER_CODE is the code of the 26025 immediately enclosing expression (`MEM' if at the top level of the 26026 address, `ADDRESS' for something that occurs in an 26027 `address_operand'). INDEX_CODE is the code of the corresponding 26028 index expression if OUTER_CODE is `PLUS'; `SCRATCH' otherwise. 26029 The mode may be `VOIDmode' for addresses that appear outside a 26030 `MEM', i.e., as an `address_operand'. 26031 26032 -- Macro: REGNO_OK_FOR_INDEX_P (NUM) 26033 A C expression which is nonzero if register number NUM is suitable 26034 for use as an index register in operand addresses. It may be 26035 either a suitable hard register or a pseudo register that has been 26036 allocated such a hard register. 26037 26038 The difference between an index register and a base register is 26039 that the index register may be scaled. If an address involves the 26040 sum of two registers, neither one of them scaled, then either one 26041 may be labeled the "base" and the other the "index"; but whichever 26042 labeling is used must fit the machine's constraints of which 26043 registers may serve in each capacity. The compiler will try both 26044 labelings, looking for one that is valid, and will reload one or 26045 both registers only if neither labeling works. 26046 26047 -- Macro: PREFERRED_RELOAD_CLASS (X, CLASS) 26048 A C expression that places additional restrictions on the register 26049 class to use when it is necessary to copy value X into a register 26050 in class CLASS. The value is a register class; perhaps CLASS, or 26051 perhaps another, smaller class. On many machines, the following 26052 definition is safe: 26053 26054 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS 26055 26056 Sometimes returning a more restrictive class makes better code. 26057 For example, on the 68000, when X is an integer constant that is 26058 in range for a `moveq' instruction, the value of this macro is 26059 always `DATA_REGS' as long as CLASS includes the data registers. 26060 Requiring a data register guarantees that a `moveq' will be used. 26061 26062 One case where `PREFERRED_RELOAD_CLASS' must not return CLASS is 26063 if X is a legitimate constant which cannot be loaded into some 26064 register class. By returning `NO_REGS' you can force X into a 26065 memory location. For example, rs6000 can load immediate values 26066 into general-purpose registers, but does not have an instruction 26067 for loading an immediate value into a floating-point register, so 26068 `PREFERRED_RELOAD_CLASS' returns `NO_REGS' when X is a 26069 floating-point constant. If the constant can't be loaded into any 26070 kind of register, code generation will be better if 26071 `LEGITIMATE_CONSTANT_P' makes the constant illegitimate instead of 26072 using `PREFERRED_RELOAD_CLASS'. 26073 26074 If an insn has pseudos in it after register allocation, reload 26075 will go through the alternatives and call repeatedly 26076 `PREFERRED_RELOAD_CLASS' to find the best one. Returning 26077 `NO_REGS', in this case, makes reload add a `!' in front of the 26078 constraint: the x86 back-end uses this feature to discourage usage 26079 of 387 registers when math is done in the SSE registers (and vice 26080 versa). 26081 26082 -- Macro: PREFERRED_OUTPUT_RELOAD_CLASS (X, CLASS) 26083 Like `PREFERRED_RELOAD_CLASS', but for output reloads instead of 26084 input reloads. If you don't define this macro, the default is to 26085 use CLASS, unchanged. 26086 26087 You can also use `PREFERRED_OUTPUT_RELOAD_CLASS' to discourage 26088 reload from using some alternatives, like `PREFERRED_RELOAD_CLASS'. 26089 26090 -- Macro: LIMIT_RELOAD_CLASS (MODE, CLASS) 26091 A C expression that places additional restrictions on the register 26092 class to use when it is necessary to be able to hold a value of 26093 mode MODE in a reload register for which class CLASS would 26094 ordinarily be used. 26095 26096 Unlike `PREFERRED_RELOAD_CLASS', this macro should be used when 26097 there are certain modes that simply can't go in certain reload 26098 classes. 26099 26100 The value is a register class; perhaps CLASS, or perhaps another, 26101 smaller class. 26102 26103 Don't define this macro unless the target machine has limitations 26104 which require the macro to do something nontrivial. 26105 26106 -- Target Hook: enum reg_class TARGET_SECONDARY_RELOAD (bool IN_P, rtx 26107 X, enum reg_class RELOAD_CLASS, enum machine_mode 26108 RELOAD_MODE, secondary_reload_info *SRI) 26109 Many machines have some registers that cannot be copied directly 26110 to or from memory or even from other types of registers. An 26111 example is the `MQ' register, which on most machines, can only be 26112 copied to or from general registers, but not memory. Below, we 26113 shall be using the term 'intermediate register' when a move 26114 operation cannot be performed directly, but has to be done by 26115 copying the source into the intermediate register first, and then 26116 copying the intermediate register to the destination. An 26117 intermediate register always has the same mode as source and 26118 destination. Since it holds the actual value being copied, reload 26119 might apply optimizations to re-use an intermediate register and 26120 eliding the copy from the source when it can determine that the 26121 intermediate register still holds the required value. 26122 26123 Another kind of secondary reload is required on some machines which 26124 allow copying all registers to and from memory, but require a 26125 scratch register for stores to some memory locations (e.g., those 26126 with symbolic address on the RT, and those with certain symbolic 26127 address on the SPARC when compiling PIC). Scratch registers need 26128 not have the same mode as the value being copied, and usually hold 26129 a different value that that being copied. Special patterns in the 26130 md file are needed to describe how the copy is performed with the 26131 help of the scratch register; these patterns also describe the 26132 number, register class(es) and mode(s) of the scratch register(s). 26133 26134 In some cases, both an intermediate and a scratch register are 26135 required. 26136 26137 For input reloads, this target hook is called with nonzero IN_P, 26138 and X is an rtx that needs to be copied to a register of class 26139 RELOAD_CLASS in RELOAD_MODE. For output reloads, this target hook 26140 is called with zero IN_P, and a register of class RELOAD_CLASS 26141 needs to be copied to rtx X in RELOAD_MODE. 26142 26143 If copying a register of RELOAD_CLASS from/to X requires an 26144 intermediate register, the hook `secondary_reload' should return 26145 the register class required for this intermediate register. If no 26146 intermediate register is required, it should return NO_REGS. If 26147 more than one intermediate register is required, describe the one 26148 that is closest in the copy chain to the reload register. 26149 26150 If scratch registers are needed, you also have to describe how to 26151 perform the copy from/to the reload register to/from this closest 26152 intermediate register. Or if no intermediate register is 26153 required, but still a scratch register is needed, describe the 26154 copy from/to the reload register to/from the reload operand X. 26155 26156 You do this by setting `sri->icode' to the instruction code of a 26157 pattern in the md file which performs the move. Operands 0 and 1 26158 are the output and input of this copy, respectively. Operands 26159 from operand 2 onward are for scratch operands. These scratch 26160 operands must have a mode, and a single-register-class output 26161 constraint. 26162 26163 When an intermediate register is used, the `secondary_reload' hook 26164 will be called again to determine how to copy the intermediate 26165 register to/from the reload operand X, so your hook must also have 26166 code to handle the register class of the intermediate operand. 26167 26168 X might be a pseudo-register or a `subreg' of a pseudo-register, 26169 which could either be in a hard register or in memory. Use 26170 `true_regnum' to find out; it will return -1 if the pseudo is in 26171 memory and the hard register number if it is in a register. 26172 26173 Scratch operands in memory (constraint `"=m"' / `"=&m"') are 26174 currently not supported. For the time being, you will have to 26175 continue to use `SECONDARY_MEMORY_NEEDED' for that purpose. 26176 26177 `copy_cost' also uses this target hook to find out how values are 26178 copied. If you want it to include some extra cost for the need to 26179 allocate (a) scratch register(s), set `sri->extra_cost' to the 26180 additional cost. Or if two dependent moves are supposed to have a 26181 lower cost than the sum of the individual moves due to expected 26182 fortuitous scheduling and/or special forwarding logic, you can set 26183 `sri->extra_cost' to a negative amount. 26184 26185 -- Macro: SECONDARY_RELOAD_CLASS (CLASS, MODE, X) 26186 -- Macro: SECONDARY_INPUT_RELOAD_CLASS (CLASS, MODE, X) 26187 -- Macro: SECONDARY_OUTPUT_RELOAD_CLASS (CLASS, MODE, X) 26188 These macros are obsolete, new ports should use the target hook 26189 `TARGET_SECONDARY_RELOAD' instead. 26190 26191 These are obsolete macros, replaced by the 26192 `TARGET_SECONDARY_RELOAD' target hook. Older ports still define 26193 these macros to indicate to the reload phase that it may need to 26194 allocate at least one register for a reload in addition to the 26195 register to contain the data. Specifically, if copying X to a 26196 register CLASS in MODE requires an intermediate register, you were 26197 supposed to define `SECONDARY_INPUT_RELOAD_CLASS' to return the 26198 largest register class all of whose registers can be used as 26199 intermediate registers or scratch registers. 26200 26201 If copying a register CLASS in MODE to X requires an intermediate 26202 or scratch register, `SECONDARY_OUTPUT_RELOAD_CLASS' was supposed 26203 to be defined be defined to return the largest register class 26204 required. If the requirements for input and output reloads were 26205 the same, the macro `SECONDARY_RELOAD_CLASS' should have been used 26206 instead of defining both macros identically. 26207 26208 The values returned by these macros are often `GENERAL_REGS'. 26209 Return `NO_REGS' if no spare register is needed; i.e., if X can be 26210 directly copied to or from a register of CLASS in MODE without 26211 requiring a scratch register. Do not define this macro if it 26212 would always return `NO_REGS'. 26213 26214 If a scratch register is required (either with or without an 26215 intermediate register), you were supposed to define patterns for 26216 `reload_inM' or `reload_outM', as required (*note Standard 26217 Names::. These patterns, which were normally implemented with a 26218 `define_expand', should be similar to the `movM' patterns, except 26219 that operand 2 is the scratch register. 26220 26221 These patterns need constraints for the reload register and scratch 26222 register that contain a single register class. If the original 26223 reload register (whose class is CLASS) can meet the constraint 26224 given in the pattern, the value returned by these macros is used 26225 for the class of the scratch register. Otherwise, two additional 26226 reload registers are required. Their classes are obtained from 26227 the constraints in the insn pattern. 26228 26229 X might be a pseudo-register or a `subreg' of a pseudo-register, 26230 which could either be in a hard register or in memory. Use 26231 `true_regnum' to find out; it will return -1 if the pseudo is in 26232 memory and the hard register number if it is in a register. 26233 26234 These macros should not be used in the case where a particular 26235 class of registers can only be copied to memory and not to another 26236 class of registers. In that case, secondary reload registers are 26237 not needed and would not be helpful. Instead, a stack location 26238 must be used to perform the copy and the `movM' pattern should use 26239 memory as an intermediate storage. This case often occurs between 26240 floating-point and general registers. 26241 26242 -- Macro: SECONDARY_MEMORY_NEEDED (CLASS1, CLASS2, M) 26243 Certain machines have the property that some registers cannot be 26244 copied to some other registers without using memory. Define this 26245 macro on those machines to be a C expression that is nonzero if 26246 objects of mode M in registers of CLASS1 can only be copied to 26247 registers of class CLASS2 by storing a register of CLASS1 into 26248 memory and loading that memory location into a register of CLASS2. 26249 26250 Do not define this macro if its value would always be zero. 26251 26252 -- Macro: SECONDARY_MEMORY_NEEDED_RTX (MODE) 26253 Normally when `SECONDARY_MEMORY_NEEDED' is defined, the compiler 26254 allocates a stack slot for a memory location needed for register 26255 copies. If this macro is defined, the compiler instead uses the 26256 memory location defined by this macro. 26257 26258 Do not define this macro if you do not define 26259 `SECONDARY_MEMORY_NEEDED'. 26260 26261 -- Macro: SECONDARY_MEMORY_NEEDED_MODE (MODE) 26262 When the compiler needs a secondary memory location to copy 26263 between two registers of mode MODE, it normally allocates 26264 sufficient memory to hold a quantity of `BITS_PER_WORD' bits and 26265 performs the store and load operations in a mode that many bits 26266 wide and whose class is the same as that of MODE. 26267 26268 This is right thing to do on most machines because it ensures that 26269 all bits of the register are copied and prevents accesses to the 26270 registers in a narrower mode, which some machines prohibit for 26271 floating-point registers. 26272 26273 However, this default behavior is not correct on some machines, 26274 such as the DEC Alpha, that store short integers in floating-point 26275 registers differently than in integer registers. On those 26276 machines, the default widening will not work correctly and you 26277 must define this macro to suppress that widening in some cases. 26278 See the file `alpha.h' for details. 26279 26280 Do not define this macro if you do not define 26281 `SECONDARY_MEMORY_NEEDED' or if widening MODE to a mode that is 26282 `BITS_PER_WORD' bits wide is correct for your machine. 26283 26284 -- Macro: SMALL_REGISTER_CLASSES 26285 On some machines, it is risky to let hard registers live across 26286 arbitrary insns. Typically, these machines have instructions that 26287 require values to be in specific registers (like an accumulator), 26288 and reload will fail if the required hard register is used for 26289 another purpose across such an insn. 26290 26291 Define `SMALL_REGISTER_CLASSES' to be an expression with a nonzero 26292 value on these machines. When this macro has a nonzero value, the 26293 compiler will try to minimize the lifetime of hard registers. 26294 26295 It is always safe to define this macro with a nonzero value, but 26296 if you unnecessarily define it, you will reduce the amount of 26297 optimizations that can be performed in some cases. If you do not 26298 define this macro with a nonzero value when it is required, the 26299 compiler will run out of spill registers and print a fatal error 26300 message. For most machines, you should not define this macro at 26301 all. 26302 26303 -- Macro: CLASS_LIKELY_SPILLED_P (CLASS) 26304 A C expression whose value is nonzero if pseudos that have been 26305 assigned to registers of class CLASS would likely be spilled 26306 because registers of CLASS are needed for spill registers. 26307 26308 The default value of this macro returns 1 if CLASS has exactly one 26309 register and zero otherwise. On most machines, this default 26310 should be used. Only define this macro to some other expression 26311 if pseudos allocated by `local-alloc.c' end up in memory because 26312 their hard registers were needed for spill registers. If this 26313 macro returns nonzero for those classes, those pseudos will only 26314 be allocated by `global.c', which knows how to reallocate the 26315 pseudo to another register. If there would not be another 26316 register available for reallocation, you should not change the 26317 definition of this macro since the only effect of such a 26318 definition would be to slow down register allocation. 26319 26320 -- Macro: CLASS_MAX_NREGS (CLASS, MODE) 26321 A C expression for the maximum number of consecutive registers of 26322 class CLASS needed to hold a value of mode MODE. 26323 26324 This is closely related to the macro `HARD_REGNO_NREGS'. In fact, 26325 the value of the macro `CLASS_MAX_NREGS (CLASS, MODE)' should be 26326 the maximum value of `HARD_REGNO_NREGS (REGNO, MODE)' for all 26327 REGNO values in the class CLASS. 26328 26329 This macro helps control the handling of multiple-word values in 26330 the reload pass. 26331 26332 -- Macro: CANNOT_CHANGE_MODE_CLASS (FROM, TO, CLASS) 26333 If defined, a C expression that returns nonzero for a CLASS for 26334 which a change from mode FROM to mode TO is invalid. 26335 26336 For the example, loading 32-bit integer or floating-point objects 26337 into floating-point registers on the Alpha extends them to 64 bits. 26338 Therefore loading a 64-bit object and then storing it as a 32-bit 26339 object does not store the low-order 32 bits, as would be the case 26340 for a normal register. Therefore, `alpha.h' defines 26341 `CANNOT_CHANGE_MODE_CLASS' as below: 26342 26343 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \ 26344 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \ 26345 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0) 26346 26347 -- Target Hook: const enum reg_class * TARGET_IRA_COVER_CLASSES () 26348 Return an array of cover classes for the Integrated Register 26349 Allocator (IRA). Cover classes are a set of non-intersecting 26350 register classes covering all hard registers used for register 26351 allocation purposes. If a move between two registers in the same 26352 cover class is possible, it should be cheaper than a load or store 26353 of the registers. The array is terminated by a `LIM_REG_CLASSES' 26354 element. 26355 26356 This hook is called once at compiler startup, after the 26357 command-line options have been processed. It is then re-examined 26358 by every call to `target_reinit'. 26359 26360 The default implementation returns `IRA_COVER_CLASSES', if defined, 26361 otherwise there is no default implementation. You must define 26362 either this macro or `IRA_COVER_CLASSES' in order to use the 26363 integrated register allocator with Chaitin-Briggs coloring. If the 26364 macro is not defined, the only available coloring algorithm is 26365 Chow's priority coloring. 26366 26367 -- Macro: IRA_COVER_CLASSES 26368 See the documentation for `TARGET_IRA_COVER_CLASSES'. 26369 26370 26371 File: gccint.info, Node: Old Constraints, Next: Stack and Calling, Prev: Register Classes, Up: Target Macros 26372 26373 17.9 Obsolete Macros for Defining Constraints 26374 ============================================= 26375 26376 Machine-specific constraints can be defined with these macros instead 26377 of the machine description constructs described in *Note Define 26378 Constraints::. This mechanism is obsolete. New ports should not use 26379 it; old ports should convert to the new mechanism. 26380 26381 -- Macro: CONSTRAINT_LEN (CHAR, STR) 26382 For the constraint at the start of STR, which starts with the 26383 letter C, return the length. This allows you to have register 26384 class / constant / extra constraints that are longer than a single 26385 letter; you don't need to define this macro if you can do with 26386 single-letter constraints only. The definition of this macro 26387 should use DEFAULT_CONSTRAINT_LEN for all the characters that you 26388 don't want to handle specially. There are some sanity checks in 26389 genoutput.c that check the constraint lengths for the md file, so 26390 you can also use this macro to help you while you are 26391 transitioning from a byzantine single-letter-constraint scheme: 26392 when you return a negative length for a constraint you want to 26393 re-use, genoutput will complain about every instance where it is 26394 used in the md file. 26395 26396 -- Macro: REG_CLASS_FROM_LETTER (CHAR) 26397 A C expression which defines the machine-dependent operand 26398 constraint letters for register classes. If CHAR is such a 26399 letter, the value should be the register class corresponding to 26400 it. Otherwise, the value should be `NO_REGS'. The register 26401 letter `r', corresponding to class `GENERAL_REGS', will not be 26402 passed to this macro; you do not need to handle it. 26403 26404 -- Macro: REG_CLASS_FROM_CONSTRAINT (CHAR, STR) 26405 Like `REG_CLASS_FROM_LETTER', but you also get the constraint 26406 string passed in STR, so that you can use suffixes to distinguish 26407 between different variants. 26408 26409 -- Macro: CONST_OK_FOR_LETTER_P (VALUE, C) 26410 A C expression that defines the machine-dependent operand 26411 constraint letters (`I', `J', `K', ... `P') that specify 26412 particular ranges of integer values. If C is one of those 26413 letters, the expression should check that VALUE, an integer, is in 26414 the appropriate range and return 1 if so, 0 otherwise. If C is 26415 not one of those letters, the value should be 0 regardless of 26416 VALUE. 26417 26418 -- Macro: CONST_OK_FOR_CONSTRAINT_P (VALUE, C, STR) 26419 Like `CONST_OK_FOR_LETTER_P', but you also get the constraint 26420 string passed in STR, so that you can use suffixes to distinguish 26421 between different variants. 26422 26423 -- Macro: CONST_DOUBLE_OK_FOR_LETTER_P (VALUE, C) 26424 A C expression that defines the machine-dependent operand 26425 constraint letters that specify particular ranges of 26426 `const_double' values (`G' or `H'). 26427 26428 If C is one of those letters, the expression should check that 26429 VALUE, an RTX of code `const_double', is in the appropriate range 26430 and return 1 if so, 0 otherwise. If C is not one of those 26431 letters, the value should be 0 regardless of VALUE. 26432 26433 `const_double' is used for all floating-point constants and for 26434 `DImode' fixed-point constants. A given letter can accept either 26435 or both kinds of values. It can use `GET_MODE' to distinguish 26436 between these kinds. 26437 26438 -- Macro: CONST_DOUBLE_OK_FOR_CONSTRAINT_P (VALUE, C, STR) 26439 Like `CONST_DOUBLE_OK_FOR_LETTER_P', but you also get the 26440 constraint string passed in STR, so that you can use suffixes to 26441 distinguish between different variants. 26442 26443 -- Macro: EXTRA_CONSTRAINT (VALUE, C) 26444 A C expression that defines the optional machine-dependent 26445 constraint letters that can be used to segregate specific types of 26446 operands, usually memory references, for the target machine. Any 26447 letter that is not elsewhere defined and not matched by 26448 `REG_CLASS_FROM_LETTER' / `REG_CLASS_FROM_CONSTRAINT' may be used. 26449 Normally this macro will not be defined. 26450 26451 If it is required for a particular target machine, it should 26452 return 1 if VALUE corresponds to the operand type represented by 26453 the constraint letter C. If C is not defined as an extra 26454 constraint, the value returned should be 0 regardless of VALUE. 26455 26456 For example, on the ROMP, load instructions cannot have their 26457 output in r0 if the memory reference contains a symbolic address. 26458 Constraint letter `Q' is defined as representing a memory address 26459 that does _not_ contain a symbolic address. An alternative is 26460 specified with a `Q' constraint on the input and `r' on the 26461 output. The next alternative specifies `m' on the input and a 26462 register class that does not include r0 on the output. 26463 26464 -- Macro: EXTRA_CONSTRAINT_STR (VALUE, C, STR) 26465 Like `EXTRA_CONSTRAINT', but you also get the constraint string 26466 passed in STR, so that you can use suffixes to distinguish between 26467 different variants. 26468 26469 -- Macro: EXTRA_MEMORY_CONSTRAINT (C, STR) 26470 A C expression that defines the optional machine-dependent 26471 constraint letters, amongst those accepted by `EXTRA_CONSTRAINT', 26472 that should be treated like memory constraints by the reload pass. 26473 26474 It should return 1 if the operand type represented by the 26475 constraint at the start of STR, the first letter of which is the 26476 letter C, comprises a subset of all memory references including 26477 all those whose address is simply a base register. This allows 26478 the reload pass to reload an operand, if it does not directly 26479 correspond to the operand type of C, by copying its address into a 26480 base register. 26481 26482 For example, on the S/390, some instructions do not accept 26483 arbitrary memory references, but only those that do not make use 26484 of an index register. The constraint letter `Q' is defined via 26485 `EXTRA_CONSTRAINT' as representing a memory address of this type. 26486 If the letter `Q' is marked as `EXTRA_MEMORY_CONSTRAINT', a `Q' 26487 constraint can handle any memory operand, because the reload pass 26488 knows it can be reloaded by copying the memory address into a base 26489 register if required. This is analogous to the way a `o' 26490 constraint can handle any memory operand. 26491 26492 -- Macro: EXTRA_ADDRESS_CONSTRAINT (C, STR) 26493 A C expression that defines the optional machine-dependent 26494 constraint letters, amongst those accepted by `EXTRA_CONSTRAINT' / 26495 `EXTRA_CONSTRAINT_STR', that should be treated like address 26496 constraints by the reload pass. 26497 26498 It should return 1 if the operand type represented by the 26499 constraint at the start of STR, which starts with the letter C, 26500 comprises a subset of all memory addresses including all those 26501 that consist of just a base register. This allows the reload pass 26502 to reload an operand, if it does not directly correspond to the 26503 operand type of STR, by copying it into a base register. 26504 26505 Any constraint marked as `EXTRA_ADDRESS_CONSTRAINT' can only be 26506 used with the `address_operand' predicate. It is treated 26507 analogously to the `p' constraint. 26508 26509 26510 File: gccint.info, Node: Stack and Calling, Next: Varargs, Prev: Old Constraints, Up: Target Macros 26511 26512 17.10 Stack Layout and Calling Conventions 26513 ========================================== 26514 26515 This describes the stack layout and calling conventions. 26516 26517 * Menu: 26518 26519 * Frame Layout:: 26520 * Exception Handling:: 26521 * Stack Checking:: 26522 * Frame Registers:: 26523 * Elimination:: 26524 * Stack Arguments:: 26525 * Register Arguments:: 26526 * Scalar Return:: 26527 * Aggregate Return:: 26528 * Caller Saves:: 26529 * Function Entry:: 26530 * Profiling:: 26531 * Tail Calls:: 26532 * Stack Smashing Protection:: 26533 26534 26535 File: gccint.info, Node: Frame Layout, Next: Exception Handling, Up: Stack and Calling 26536 26537 17.10.1 Basic Stack Layout 26538 -------------------------- 26539 26540 Here is the basic stack layout. 26541 26542 -- Macro: STACK_GROWS_DOWNWARD 26543 Define this macro if pushing a word onto the stack moves the stack 26544 pointer to a smaller address. 26545 26546 When we say, "define this macro if ...", it means that the 26547 compiler checks this macro only with `#ifdef' so the precise 26548 definition used does not matter. 26549 26550 -- Macro: STACK_PUSH_CODE 26551 This macro defines the operation used when something is pushed on 26552 the stack. In RTL, a push operation will be `(set (mem 26553 (STACK_PUSH_CODE (reg sp))) ...)' 26554 26555 The choices are `PRE_DEC', `POST_DEC', `PRE_INC', and `POST_INC'. 26556 Which of these is correct depends on the stack direction and on 26557 whether the stack pointer points to the last item on the stack or 26558 whether it points to the space for the next item on the stack. 26559 26560 The default is `PRE_DEC' when `STACK_GROWS_DOWNWARD' is defined, 26561 which is almost always right, and `PRE_INC' otherwise, which is 26562 often wrong. 26563 26564 -- Macro: FRAME_GROWS_DOWNWARD 26565 Define this macro to nonzero value if the addresses of local 26566 variable slots are at negative offsets from the frame pointer. 26567 26568 -- Macro: ARGS_GROW_DOWNWARD 26569 Define this macro if successive arguments to a function occupy 26570 decreasing addresses on the stack. 26571 26572 -- Macro: STARTING_FRAME_OFFSET 26573 Offset from the frame pointer to the first local variable slot to 26574 be allocated. 26575 26576 If `FRAME_GROWS_DOWNWARD', find the next slot's offset by 26577 subtracting the first slot's length from `STARTING_FRAME_OFFSET'. 26578 Otherwise, it is found by adding the length of the first slot to 26579 the value `STARTING_FRAME_OFFSET'. 26580 26581 -- Macro: STACK_ALIGNMENT_NEEDED 26582 Define to zero to disable final alignment of the stack during 26583 reload. The nonzero default for this macro is suitable for most 26584 ports. 26585 26586 On ports where `STARTING_FRAME_OFFSET' is nonzero or where there 26587 is a register save block following the local block that doesn't 26588 require alignment to `STACK_BOUNDARY', it may be beneficial to 26589 disable stack alignment and do it in the backend. 26590 26591 -- Macro: STACK_POINTER_OFFSET 26592 Offset from the stack pointer register to the first location at 26593 which outgoing arguments are placed. If not specified, the 26594 default value of zero is used. This is the proper value for most 26595 machines. 26596 26597 If `ARGS_GROW_DOWNWARD', this is the offset to the location above 26598 the first location at which outgoing arguments are placed. 26599 26600 -- Macro: FIRST_PARM_OFFSET (FUNDECL) 26601 Offset from the argument pointer register to the first argument's 26602 address. On some machines it may depend on the data type of the 26603 function. 26604 26605 If `ARGS_GROW_DOWNWARD', this is the offset to the location above 26606 the first argument's address. 26607 26608 -- Macro: STACK_DYNAMIC_OFFSET (FUNDECL) 26609 Offset from the stack pointer register to an item dynamically 26610 allocated on the stack, e.g., by `alloca'. 26611 26612 The default value for this macro is `STACK_POINTER_OFFSET' plus the 26613 length of the outgoing arguments. The default is correct for most 26614 machines. See `function.c' for details. 26615 26616 -- Macro: INITIAL_FRAME_ADDRESS_RTX 26617 A C expression whose value is RTL representing the address of the 26618 initial stack frame. This address is passed to `RETURN_ADDR_RTX' 26619 and `DYNAMIC_CHAIN_ADDRESS'. If you don't define this macro, a 26620 reasonable default value will be used. Define this macro in order 26621 to make frame pointer elimination work in the presence of 26622 `__builtin_frame_address (count)' and `__builtin_return_address 26623 (count)' for `count' not equal to zero. 26624 26625 -- Macro: DYNAMIC_CHAIN_ADDRESS (FRAMEADDR) 26626 A C expression whose value is RTL representing the address in a 26627 stack frame where the pointer to the caller's frame is stored. 26628 Assume that FRAMEADDR is an RTL expression for the address of the 26629 stack frame itself. 26630 26631 If you don't define this macro, the default is to return the value 26632 of FRAMEADDR--that is, the stack frame address is also the address 26633 of the stack word that points to the previous frame. 26634 26635 -- Macro: SETUP_FRAME_ADDRESSES 26636 If defined, a C expression that produces the machine-specific code 26637 to setup the stack so that arbitrary frames can be accessed. For 26638 example, on the SPARC, we must flush all of the register windows 26639 to the stack before we can access arbitrary stack frames. You 26640 will seldom need to define this macro. 26641 26642 -- Target Hook: bool TARGET_BUILTIN_SETJMP_FRAME_VALUE () 26643 This target hook should return an rtx that is used to store the 26644 address of the current frame into the built in `setjmp' buffer. 26645 The default value, `virtual_stack_vars_rtx', is correct for most 26646 machines. One reason you may need to define this target hook is if 26647 `hard_frame_pointer_rtx' is the appropriate value on your machine. 26648 26649 -- Macro: FRAME_ADDR_RTX (FRAMEADDR) 26650 A C expression whose value is RTL representing the value of the 26651 frame address for the current frame. FRAMEADDR is the frame 26652 pointer of the current frame. This is used for 26653 __builtin_frame_address. You need only define this macro if the 26654 frame address is not the same as the frame pointer. Most machines 26655 do not need to define it. 26656 26657 -- Macro: RETURN_ADDR_RTX (COUNT, FRAMEADDR) 26658 A C expression whose value is RTL representing the value of the 26659 return address for the frame COUNT steps up from the current 26660 frame, after the prologue. FRAMEADDR is the frame pointer of the 26661 COUNT frame, or the frame pointer of the COUNT - 1 frame if 26662 `RETURN_ADDR_IN_PREVIOUS_FRAME' is defined. 26663 26664 The value of the expression must always be the correct address when 26665 COUNT is zero, but may be `NULL_RTX' if there is no way to 26666 determine the return address of other frames. 26667 26668 -- Macro: RETURN_ADDR_IN_PREVIOUS_FRAME 26669 Define this if the return address of a particular stack frame is 26670 accessed from the frame pointer of the previous stack frame. 26671 26672 -- Macro: INCOMING_RETURN_ADDR_RTX 26673 A C expression whose value is RTL representing the location of the 26674 incoming return address at the beginning of any function, before 26675 the prologue. This RTL is either a `REG', indicating that the 26676 return value is saved in `REG', or a `MEM' representing a location 26677 in the stack. 26678 26679 You only need to define this macro if you want to support call 26680 frame debugging information like that provided by DWARF 2. 26681 26682 If this RTL is a `REG', you should also define 26683 `DWARF_FRAME_RETURN_COLUMN' to `DWARF_FRAME_REGNUM (REGNO)'. 26684 26685 -- Macro: DWARF_ALT_FRAME_RETURN_COLUMN 26686 A C expression whose value is an integer giving a DWARF 2 column 26687 number that may be used as an alternative return column. The 26688 column must not correspond to any gcc hard register (that is, it 26689 must not be in the range of `DWARF_FRAME_REGNUM'). 26690 26691 This macro can be useful if `DWARF_FRAME_RETURN_COLUMN' is set to a 26692 general register, but an alternative column needs to be used for 26693 signal frames. Some targets have also used different frame return 26694 columns over time. 26695 26696 -- Macro: DWARF_ZERO_REG 26697 A C expression whose value is an integer giving a DWARF 2 register 26698 number that is considered to always have the value zero. This 26699 should only be defined if the target has an architected zero 26700 register, and someone decided it was a good idea to use that 26701 register number to terminate the stack backtrace. New ports 26702 should avoid this. 26703 26704 -- Target Hook: void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char 26705 *LABEL, rtx PATTERN, int INDEX) 26706 This target hook allows the backend to emit frame-related insns 26707 that contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame 26708 debugging info engine will invoke it on insns of the form 26709 (set (reg) (unspec [...] UNSPEC_INDEX)) 26710 and 26711 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)). 26712 to let the backend emit the call frame instructions. LABEL is the 26713 CFI label attached to the insn, PATTERN is the pattern of the insn 26714 and INDEX is `UNSPEC_INDEX' or `UNSPECV_INDEX'. 26715 26716 -- Macro: INCOMING_FRAME_SP_OFFSET 26717 A C expression whose value is an integer giving the offset, in 26718 bytes, from the value of the stack pointer register to the top of 26719 the stack frame at the beginning of any function, before the 26720 prologue. The top of the frame is defined to be the value of the 26721 stack pointer in the previous frame, just before the call 26722 instruction. 26723 26724 You only need to define this macro if you want to support call 26725 frame debugging information like that provided by DWARF 2. 26726 26727 -- Macro: ARG_POINTER_CFA_OFFSET (FUNDECL) 26728 A C expression whose value is an integer giving the offset, in 26729 bytes, from the argument pointer to the canonical frame address 26730 (cfa). The final value should coincide with that calculated by 26731 `INCOMING_FRAME_SP_OFFSET'. Which is unfortunately not usable 26732 during virtual register instantiation. 26733 26734 The default value for this macro is `FIRST_PARM_OFFSET (fundecl)', 26735 which is correct for most machines; in general, the arguments are 26736 found immediately before the stack frame. Note that this is not 26737 the case on some targets that save registers into the caller's 26738 frame, such as SPARC and rs6000, and so such targets need to 26739 define this macro. 26740 26741 You only need to define this macro if the default is incorrect, 26742 and you want to support call frame debugging information like that 26743 provided by DWARF 2. 26744 26745 -- Macro: FRAME_POINTER_CFA_OFFSET (FUNDECL) 26746 If defined, a C expression whose value is an integer giving the 26747 offset in bytes from the frame pointer to the canonical frame 26748 address (cfa). The final value should coincide with that 26749 calculated by `INCOMING_FRAME_SP_OFFSET'. 26750 26751 Normally the CFA is calculated as an offset from the argument 26752 pointer, via `ARG_POINTER_CFA_OFFSET', but if the argument pointer 26753 is variable due to the ABI, this may not be possible. If this 26754 macro is defined, it implies that the virtual register 26755 instantiation should be based on the frame pointer instead of the 26756 argument pointer. Only one of `FRAME_POINTER_CFA_OFFSET' and 26757 `ARG_POINTER_CFA_OFFSET' should be defined. 26758 26759 -- Macro: CFA_FRAME_BASE_OFFSET (FUNDECL) 26760 If defined, a C expression whose value is an integer giving the 26761 offset in bytes from the canonical frame address (cfa) to the 26762 frame base used in DWARF 2 debug information. The default is 26763 zero. A different value may reduce the size of debug information 26764 on some ports. 26765 26766 26767 File: gccint.info, Node: Exception Handling, Next: Stack Checking, Prev: Frame Layout, Up: Stack and Calling 26768 26769 17.10.2 Exception Handling Support 26770 ---------------------------------- 26771 26772 -- Macro: EH_RETURN_DATA_REGNO (N) 26773 A C expression whose value is the Nth register number used for 26774 data by exception handlers, or `INVALID_REGNUM' if fewer than N 26775 registers are usable. 26776 26777 The exception handling library routines communicate with the 26778 exception handlers via a set of agreed upon registers. Ideally 26779 these registers should be call-clobbered; it is possible to use 26780 call-saved registers, but may negatively impact code size. The 26781 target must support at least 2 data registers, but should define 4 26782 if there are enough free registers. 26783 26784 You must define this macro if you want to support call frame 26785 exception handling like that provided by DWARF 2. 26786 26787 -- Macro: EH_RETURN_STACKADJ_RTX 26788 A C expression whose value is RTL representing a location in which 26789 to store a stack adjustment to be applied before function return. 26790 This is used to unwind the stack to an exception handler's call 26791 frame. It will be assigned zero on code paths that return 26792 normally. 26793 26794 Typically this is a call-clobbered hard register that is otherwise 26795 untouched by the epilogue, but could also be a stack slot. 26796 26797 Do not define this macro if the stack pointer is saved and restored 26798 by the regular prolog and epilog code in the call frame itself; in 26799 this case, the exception handling library routines will update the 26800 stack location to be restored in place. Otherwise, you must define 26801 this macro if you want to support call frame exception handling 26802 like that provided by DWARF 2. 26803 26804 -- Macro: EH_RETURN_HANDLER_RTX 26805 A C expression whose value is RTL representing a location in which 26806 to store the address of an exception handler to which we should 26807 return. It will not be assigned on code paths that return 26808 normally. 26809 26810 Typically this is the location in the call frame at which the 26811 normal return address is stored. For targets that return by 26812 popping an address off the stack, this might be a memory address 26813 just below the _target_ call frame rather than inside the current 26814 call frame. If defined, `EH_RETURN_STACKADJ_RTX' will have already 26815 been assigned, so it may be used to calculate the location of the 26816 target call frame. 26817 26818 Some targets have more complex requirements than storing to an 26819 address calculable during initial code generation. In that case 26820 the `eh_return' instruction pattern should be used instead. 26821 26822 If you want to support call frame exception handling, you must 26823 define either this macro or the `eh_return' instruction pattern. 26824 26825 -- Macro: RETURN_ADDR_OFFSET 26826 If defined, an integer-valued C expression for which rtl will be 26827 generated to add it to the exception handler address before it is 26828 searched in the exception handling tables, and to subtract it 26829 again from the address before using it to return to the exception 26830 handler. 26831 26832 -- Macro: ASM_PREFERRED_EH_DATA_FORMAT (CODE, GLOBAL) 26833 This macro chooses the encoding of pointers embedded in the 26834 exception handling sections. If at all possible, this should be 26835 defined such that the exception handling section will not require 26836 dynamic relocations, and so may be read-only. 26837 26838 CODE is 0 for data, 1 for code labels, 2 for function pointers. 26839 GLOBAL is true if the symbol may be affected by dynamic 26840 relocations. The macro should return a combination of the 26841 `DW_EH_PE_*' defines as found in `dwarf2.h'. 26842 26843 If this macro is not defined, pointers will not be encoded but 26844 represented directly. 26845 26846 -- Macro: ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (FILE, ENCODING, SIZE, 26847 ADDR, DONE) 26848 This macro allows the target to emit whatever special magic is 26849 required to represent the encoding chosen by 26850 `ASM_PREFERRED_EH_DATA_FORMAT'. Generic code takes care of 26851 pc-relative and indirect encodings; this must be defined if the 26852 target uses text-relative or data-relative encodings. 26853 26854 This is a C statement that branches to DONE if the format was 26855 handled. ENCODING is the format chosen, SIZE is the number of 26856 bytes that the format occupies, ADDR is the `SYMBOL_REF' to be 26857 emitted. 26858 26859 -- Macro: MD_UNWIND_SUPPORT 26860 A string specifying a file to be #include'd in unwind-dw2.c. The 26861 file so included typically defines `MD_FALLBACK_FRAME_STATE_FOR'. 26862 26863 -- Macro: MD_FALLBACK_FRAME_STATE_FOR (CONTEXT, FS) 26864 This macro allows the target to add CPU and operating system 26865 specific code to the call-frame unwinder for use when there is no 26866 unwind data available. The most common reason to implement this 26867 macro is to unwind through signal frames. 26868 26869 This macro is called from `uw_frame_state_for' in `unwind-dw2.c', 26870 `unwind-dw2-xtensa.c' and `unwind-ia64.c'. CONTEXT is an 26871 `_Unwind_Context'; FS is an `_Unwind_FrameState'. Examine 26872 `context->ra' for the address of the code being executed and 26873 `context->cfa' for the stack pointer value. If the frame can be 26874 decoded, the register save addresses should be updated in FS and 26875 the macro should evaluate to `_URC_NO_REASON'. If the frame 26876 cannot be decoded, the macro should evaluate to 26877 `_URC_END_OF_STACK'. 26878 26879 For proper signal handling in Java this macro is accompanied by 26880 `MAKE_THROW_FRAME', defined in `libjava/include/*-signal.h' 26881 headers. 26882 26883 -- Macro: MD_HANDLE_UNWABI (CONTEXT, FS) 26884 This macro allows the target to add operating system specific code 26885 to the call-frame unwinder to handle the IA-64 `.unwabi' unwinding 26886 directive, usually used for signal or interrupt frames. 26887 26888 This macro is called from `uw_update_context' in `unwind-ia64.c'. 26889 CONTEXT is an `_Unwind_Context'; FS is an `_Unwind_FrameState'. 26890 Examine `fs->unwabi' for the abi and context in the `.unwabi' 26891 directive. If the `.unwabi' directive can be handled, the 26892 register save addresses should be updated in FS. 26893 26894 -- Macro: TARGET_USES_WEAK_UNWIND_INFO 26895 A C expression that evaluates to true if the target requires unwind 26896 info to be given comdat linkage. Define it to be `1' if comdat 26897 linkage is necessary. The default is `0'. 26898 26899 26900 File: gccint.info, Node: Stack Checking, Next: Frame Registers, Prev: Exception Handling, Up: Stack and Calling 26901 26902 17.10.3 Specifying How Stack Checking is Done 26903 --------------------------------------------- 26904 26905 GCC will check that stack references are within the boundaries of the 26906 stack, if the option `-fstack-check' is specified, in one of three ways: 26907 26908 1. If the value of the `STACK_CHECK_BUILTIN' macro is nonzero, GCC 26909 will assume that you have arranged for full stack checking to be 26910 done at appropriate places in the configuration files. GCC will 26911 not do other special processing. 26912 26913 2. If `STACK_CHECK_BUILTIN' is zero and the value of the 26914 `STACK_CHECK_STATIC_BUILTIN' macro is nonzero, GCC will assume 26915 that you have arranged for static stack checking (checking of the 26916 static stack frame of functions) to be done at appropriate places 26917 in the configuration files. GCC will only emit code to do dynamic 26918 stack checking (checking on dynamic stack allocations) using the 26919 third approach below. 26920 26921 3. If neither of the above are true, GCC will generate code to 26922 periodically "probe" the stack pointer using the values of the 26923 macros defined below. 26924 26925 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is 26926 defined, GCC will change its allocation strategy for large objects if 26927 the option `-fstack-check' is specified: they will always be allocated 26928 dynamically if their size exceeds `STACK_CHECK_MAX_VAR_SIZE' bytes. 26929 26930 -- Macro: STACK_CHECK_BUILTIN 26931 A nonzero value if stack checking is done by the configuration 26932 files in a machine-dependent manner. You should define this macro 26933 if stack checking is require by the ABI of your machine or if you 26934 would like to do stack checking in some more efficient way than 26935 the generic approach. The default value of this macro is zero. 26936 26937 -- Macro: STACK_CHECK_STATIC_BUILTIN 26938 A nonzero value if static stack checking is done by the 26939 configuration files in a machine-dependent manner. You should 26940 define this macro if you would like to do static stack checking in 26941 some more efficient way than the generic approach. The default 26942 value of this macro is zero. 26943 26944 -- Macro: STACK_CHECK_PROBE_INTERVAL 26945 An integer representing the interval at which GCC must generate 26946 stack probe instructions. You will normally define this macro to 26947 be no larger than the size of the "guard pages" at the end of a 26948 stack area. The default value of 4096 is suitable for most 26949 systems. 26950 26951 -- Macro: STACK_CHECK_PROBE_LOAD 26952 An integer which is nonzero if GCC should perform the stack probe 26953 as a load instruction and zero if GCC should use a store 26954 instruction. The default is zero, which is the most efficient 26955 choice on most systems. 26956 26957 -- Macro: STACK_CHECK_PROTECT 26958 The number of bytes of stack needed to recover from a stack 26959 overflow, for languages where such a recovery is supported. The 26960 default value of 75 words should be adequate for most machines. 26961 26962 The following macros are relevant only if neither STACK_CHECK_BUILTIN 26963 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether 26964 in the opposite case. 26965 26966 -- Macro: STACK_CHECK_MAX_FRAME_SIZE 26967 The maximum size of a stack frame, in bytes. GCC will generate 26968 probe instructions in non-leaf functions to ensure at least this 26969 many bytes of stack are available. If a stack frame is larger 26970 than this size, stack checking will not be reliable and GCC will 26971 issue a warning. The default is chosen so that GCC only generates 26972 one instruction on most systems. You should normally not change 26973 the default value of this macro. 26974 26975 -- Macro: STACK_CHECK_FIXED_FRAME_SIZE 26976 GCC uses this value to generate the above warning message. It 26977 represents the amount of fixed frame used by a function, not 26978 including space for any callee-saved registers, temporaries and 26979 user variables. You need only specify an upper bound for this 26980 amount and will normally use the default of four words. 26981 26982 -- Macro: STACK_CHECK_MAX_VAR_SIZE 26983 The maximum size, in bytes, of an object that GCC will place in the 26984 fixed area of the stack frame when the user specifies 26985 `-fstack-check'. GCC computed the default from the values of the 26986 above macros and you will normally not need to override that 26987 default. 26988 26989 26990 File: gccint.info, Node: Frame Registers, Next: Elimination, Prev: Stack Checking, Up: Stack and Calling 26991 26992 17.10.4 Registers That Address the Stack Frame 26993 ---------------------------------------------- 26994 26995 This discusses registers that address the stack frame. 26996 26997 -- Macro: STACK_POINTER_REGNUM 26998 The register number of the stack pointer register, which must also 26999 be a fixed register according to `FIXED_REGISTERS'. On most 27000 machines, the hardware determines which register this is. 27001 27002 -- Macro: FRAME_POINTER_REGNUM 27003 The register number of the frame pointer register, which is used to 27004 access automatic variables in the stack frame. On some machines, 27005 the hardware determines which register this is. On other 27006 machines, you can choose any register you wish for this purpose. 27007 27008 -- Macro: HARD_FRAME_POINTER_REGNUM 27009 On some machines the offset between the frame pointer and starting 27010 offset of the automatic variables is not known until after register 27011 allocation has been done (for example, because the saved registers 27012 are between these two locations). On those machines, define 27013 `FRAME_POINTER_REGNUM' the number of a special, fixed register to 27014 be used internally until the offset is known, and define 27015 `HARD_FRAME_POINTER_REGNUM' to be the actual hard register number 27016 used for the frame pointer. 27017 27018 You should define this macro only in the very rare circumstances 27019 when it is not possible to calculate the offset between the frame 27020 pointer and the automatic variables until after register 27021 allocation has been completed. When this macro is defined, you 27022 must also indicate in your definition of `ELIMINABLE_REGS' how to 27023 eliminate `FRAME_POINTER_REGNUM' into either 27024 `HARD_FRAME_POINTER_REGNUM' or `STACK_POINTER_REGNUM'. 27025 27026 Do not define this macro if it would be the same as 27027 `FRAME_POINTER_REGNUM'. 27028 27029 -- Macro: ARG_POINTER_REGNUM 27030 The register number of the arg pointer register, which is used to 27031 access the function's argument list. On some machines, this is 27032 the same as the frame pointer register. On some machines, the 27033 hardware determines which register this is. On other machines, 27034 you can choose any register you wish for this purpose. If this is 27035 not the same register as the frame pointer register, then you must 27036 mark it as a fixed register according to `FIXED_REGISTERS', or 27037 arrange to be able to eliminate it (*note Elimination::). 27038 27039 -- Macro: RETURN_ADDRESS_POINTER_REGNUM 27040 The register number of the return address pointer register, which 27041 is used to access the current function's return address from the 27042 stack. On some machines, the return address is not at a fixed 27043 offset from the frame pointer or stack pointer or argument 27044 pointer. This register can be defined to point to the return 27045 address on the stack, and then be converted by `ELIMINABLE_REGS' 27046 into either the frame pointer or stack pointer. 27047 27048 Do not define this macro unless there is no other way to get the 27049 return address from the stack. 27050 27051 -- Macro: STATIC_CHAIN_REGNUM 27052 -- Macro: STATIC_CHAIN_INCOMING_REGNUM 27053 Register numbers used for passing a function's static chain 27054 pointer. If register windows are used, the register number as 27055 seen by the called function is `STATIC_CHAIN_INCOMING_REGNUM', 27056 while the register number as seen by the calling function is 27057 `STATIC_CHAIN_REGNUM'. If these registers are the same, 27058 `STATIC_CHAIN_INCOMING_REGNUM' need not be defined. 27059 27060 The static chain register need not be a fixed register. 27061 27062 If the static chain is passed in memory, these macros should not be 27063 defined; instead, the next two macros should be defined. 27064 27065 -- Macro: STATIC_CHAIN 27066 -- Macro: STATIC_CHAIN_INCOMING 27067 If the static chain is passed in memory, these macros provide rtx 27068 giving `mem' expressions that denote where they are stored. 27069 `STATIC_CHAIN' and `STATIC_CHAIN_INCOMING' give the locations as 27070 seen by the calling and called functions, respectively. Often the 27071 former will be at an offset from the stack pointer and the latter 27072 at an offset from the frame pointer. 27073 27074 The variables `stack_pointer_rtx', `frame_pointer_rtx', and 27075 `arg_pointer_rtx' will have been initialized prior to the use of 27076 these macros and should be used to refer to those items. 27077 27078 If the static chain is passed in a register, the two previous 27079 macros should be defined instead. 27080 27081 -- Macro: DWARF_FRAME_REGISTERS 27082 This macro specifies the maximum number of hard registers that can 27083 be saved in a call frame. This is used to size data structures 27084 used in DWARF2 exception handling. 27085 27086 Prior to GCC 3.0, this macro was needed in order to establish a 27087 stable exception handling ABI in the face of adding new hard 27088 registers for ISA extensions. In GCC 3.0 and later, the EH ABI is 27089 insulated from changes in the number of hard registers. 27090 Nevertheless, this macro can still be used to reduce the runtime 27091 memory requirements of the exception handling routines, which can 27092 be substantial if the ISA contains a lot of registers that are not 27093 call-saved. 27094 27095 If this macro is not defined, it defaults to 27096 `FIRST_PSEUDO_REGISTER'. 27097 27098 -- Macro: PRE_GCC3_DWARF_FRAME_REGISTERS 27099 This macro is similar to `DWARF_FRAME_REGISTERS', but is provided 27100 for backward compatibility in pre GCC 3.0 compiled code. 27101 27102 If this macro is not defined, it defaults to 27103 `DWARF_FRAME_REGISTERS'. 27104 27105 -- Macro: DWARF_REG_TO_UNWIND_COLUMN (REGNO) 27106 Define this macro if the target's representation for dwarf 27107 registers is different than the internal representation for unwind 27108 column. Given a dwarf register, this macro should return the 27109 internal unwind column number to use instead. 27110 27111 See the PowerPC's SPE target for an example. 27112 27113 -- Macro: DWARF_FRAME_REGNUM (REGNO) 27114 Define this macro if the target's representation for dwarf 27115 registers used in .eh_frame or .debug_frame is different from that 27116 used in other debug info sections. Given a GCC hard register 27117 number, this macro should return the .eh_frame register number. 27118 The default is `DBX_REGISTER_NUMBER (REGNO)'. 27119 27120 27121 -- Macro: DWARF2_FRAME_REG_OUT (REGNO, FOR_EH) 27122 Define this macro to map register numbers held in the call frame 27123 info that GCC has collected using `DWARF_FRAME_REGNUM' to those 27124 that should be output in .debug_frame (`FOR_EH' is zero) and 27125 .eh_frame (`FOR_EH' is nonzero). The default is to return `REGNO'. 27126 27127 27128 27129 File: gccint.info, Node: Elimination, Next: Stack Arguments, Prev: Frame Registers, Up: Stack and Calling 27130 27131 17.10.5 Eliminating Frame Pointer and Arg Pointer 27132 ------------------------------------------------- 27133 27134 This is about eliminating the frame pointer and arg pointer. 27135 27136 -- Macro: FRAME_POINTER_REQUIRED 27137 A C expression which is nonzero if a function must have and use a 27138 frame pointer. This expression is evaluated in the reload pass. 27139 If its value is nonzero the function will have a frame pointer. 27140 27141 The expression can in principle examine the current function and 27142 decide according to the facts, but on most machines the constant 0 27143 or the constant 1 suffices. Use 0 when the machine allows code to 27144 be generated with no frame pointer, and doing so saves some time 27145 or space. Use 1 when there is no possible advantage to avoiding a 27146 frame pointer. 27147 27148 In certain cases, the compiler does not know how to produce valid 27149 code without a frame pointer. The compiler recognizes those cases 27150 and automatically gives the function a frame pointer regardless of 27151 what `FRAME_POINTER_REQUIRED' says. You don't need to worry about 27152 them. 27153 27154 In a function that does not require a frame pointer, the frame 27155 pointer register can be allocated for ordinary usage, unless you 27156 mark it as a fixed register. See `FIXED_REGISTERS' for more 27157 information. 27158 27159 -- Macro: INITIAL_FRAME_POINTER_OFFSET (DEPTH-VAR) 27160 A C statement to store in the variable DEPTH-VAR the difference 27161 between the frame pointer and the stack pointer values immediately 27162 after the function prologue. The value would be computed from 27163 information such as the result of `get_frame_size ()' and the 27164 tables of registers `regs_ever_live' and `call_used_regs'. 27165 27166 If `ELIMINABLE_REGS' is defined, this macro will be not be used and 27167 need not be defined. Otherwise, it must be defined even if 27168 `FRAME_POINTER_REQUIRED' is defined to always be true; in that 27169 case, you may set DEPTH-VAR to anything. 27170 27171 -- Macro: ELIMINABLE_REGS 27172 If defined, this macro specifies a table of register pairs used to 27173 eliminate unneeded registers that point into the stack frame. If 27174 it is not defined, the only elimination attempted by the compiler 27175 is to replace references to the frame pointer with references to 27176 the stack pointer. 27177 27178 The definition of this macro is a list of structure 27179 initializations, each of which specifies an original and 27180 replacement register. 27181 27182 On some machines, the position of the argument pointer is not 27183 known until the compilation is completed. In such a case, a 27184 separate hard register must be used for the argument pointer. 27185 This register can be eliminated by replacing it with either the 27186 frame pointer or the argument pointer, depending on whether or not 27187 the frame pointer has been eliminated. 27188 27189 In this case, you might specify: 27190 #define ELIMINABLE_REGS \ 27191 {{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM}, \ 27192 {ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM}, \ 27193 {FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM}} 27194 27195 Note that the elimination of the argument pointer with the stack 27196 pointer is specified first since that is the preferred elimination. 27197 27198 -- Macro: CAN_ELIMINATE (FROM-REG, TO-REG) 27199 A C expression that returns nonzero if the compiler is allowed to 27200 try to replace register number FROM-REG with register number 27201 TO-REG. This macro need only be defined if `ELIMINABLE_REGS' is 27202 defined, and will usually be the constant 1, since most of the 27203 cases preventing register elimination are things that the compiler 27204 already knows about. 27205 27206 -- Macro: INITIAL_ELIMINATION_OFFSET (FROM-REG, TO-REG, OFFSET-VAR) 27207 This macro is similar to `INITIAL_FRAME_POINTER_OFFSET'. It 27208 specifies the initial difference between the specified pair of 27209 registers. This macro must be defined if `ELIMINABLE_REGS' is 27210 defined. 27211 27212 27213 File: gccint.info, Node: Stack Arguments, Next: Register Arguments, Prev: Elimination, Up: Stack and Calling 27214 27215 17.10.6 Passing Function Arguments on the Stack 27216 ----------------------------------------------- 27217 27218 The macros in this section control how arguments are passed on the 27219 stack. See the following section for other macros that control passing 27220 certain arguments in registers. 27221 27222 -- Target Hook: bool TARGET_PROMOTE_PROTOTYPES (tree FNTYPE) 27223 This target hook returns `true' if an argument declared in a 27224 prototype as an integral type smaller than `int' should actually be 27225 passed as an `int'. In addition to avoiding errors in certain 27226 cases of mismatch, it also makes for better code on certain 27227 machines. The default is to not promote prototypes. 27228 27229 -- Macro: PUSH_ARGS 27230 A C expression. If nonzero, push insns will be used to pass 27231 outgoing arguments. If the target machine does not have a push 27232 instruction, set it to zero. That directs GCC to use an alternate 27233 strategy: to allocate the entire argument block and then store the 27234 arguments into it. When `PUSH_ARGS' is nonzero, `PUSH_ROUNDING' 27235 must be defined too. 27236 27237 -- Macro: PUSH_ARGS_REVERSED 27238 A C expression. If nonzero, function arguments will be evaluated 27239 from last to first, rather than from first to last. If this macro 27240 is not defined, it defaults to `PUSH_ARGS' on targets where the 27241 stack and args grow in opposite directions, and 0 otherwise. 27242 27243 -- Macro: PUSH_ROUNDING (NPUSHED) 27244 A C expression that is the number of bytes actually pushed onto the 27245 stack when an instruction attempts to push NPUSHED bytes. 27246 27247 On some machines, the definition 27248 27249 #define PUSH_ROUNDING(BYTES) (BYTES) 27250 27251 will suffice. But on other machines, instructions that appear to 27252 push one byte actually push two bytes in an attempt to maintain 27253 alignment. Then the definition should be 27254 27255 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) 27256 27257 -- Macro: ACCUMULATE_OUTGOING_ARGS 27258 A C expression. If nonzero, the maximum amount of space required 27259 for outgoing arguments will be computed and placed into the 27260 variable `current_function_outgoing_args_size'. No space will be 27261 pushed onto the stack for each call; instead, the function 27262 prologue should increase the stack frame size by this amount. 27263 27264 Setting both `PUSH_ARGS' and `ACCUMULATE_OUTGOING_ARGS' is not 27265 proper. 27266 27267 -- Macro: REG_PARM_STACK_SPACE (FNDECL) 27268 Define this macro if functions should assume that stack space has 27269 been allocated for arguments even when their values are passed in 27270 registers. 27271 27272 The value of this macro is the size, in bytes, of the area 27273 reserved for arguments passed in registers for the function 27274 represented by FNDECL, which can be zero if GCC is calling a 27275 library function. The argument FNDECL can be the FUNCTION_DECL, 27276 or the type itself of the function. 27277 27278 This space can be allocated by the caller, or be a part of the 27279 machine-dependent stack frame: `OUTGOING_REG_PARM_STACK_SPACE' says 27280 which. 27281 27282 -- Macro: OUTGOING_REG_PARM_STACK_SPACE (FNTYPE) 27283 Define this to a nonzero value if it is the responsibility of the 27284 caller to allocate the area reserved for arguments passed in 27285 registers when calling a function of FNTYPE. FNTYPE may be NULL 27286 if the function called is a library function. 27287 27288 If `ACCUMULATE_OUTGOING_ARGS' is defined, this macro controls 27289 whether the space for these arguments counts in the value of 27290 `current_function_outgoing_args_size'. 27291 27292 -- Macro: STACK_PARMS_IN_REG_PARM_AREA 27293 Define this macro if `REG_PARM_STACK_SPACE' is defined, but the 27294 stack parameters don't skip the area specified by it. 27295 27296 Normally, when a parameter is not passed in registers, it is 27297 placed on the stack beyond the `REG_PARM_STACK_SPACE' area. 27298 Defining this macro suppresses this behavior and causes the 27299 parameter to be passed on the stack in its natural location. 27300 27301 -- Macro: RETURN_POPS_ARGS (FUNDECL, FUNTYPE, STACK-SIZE) 27302 A C expression that should indicate the number of bytes of its own 27303 arguments that a function pops on returning, or 0 if the function 27304 pops no arguments and the caller must therefore pop them all after 27305 the function returns. 27306 27307 FUNDECL is a C variable whose value is a tree node that describes 27308 the function in question. Normally it is a node of type 27309 `FUNCTION_DECL' that describes the declaration of the function. 27310 From this you can obtain the `DECL_ATTRIBUTES' of the function. 27311 27312 FUNTYPE is a C variable whose value is a tree node that describes 27313 the function in question. Normally it is a node of type 27314 `FUNCTION_TYPE' that describes the data type of the function. 27315 From this it is possible to obtain the data types of the value and 27316 arguments (if known). 27317 27318 When a call to a library function is being considered, FUNDECL 27319 will contain an identifier node for the library function. Thus, if 27320 you need to distinguish among various library functions, you can 27321 do so by their names. Note that "library function" in this 27322 context means a function used to perform arithmetic, whose name is 27323 known specially in the compiler and was not mentioned in the C 27324 code being compiled. 27325 27326 STACK-SIZE is the number of bytes of arguments passed on the 27327 stack. If a variable number of bytes is passed, it is zero, and 27328 argument popping will always be the responsibility of the calling 27329 function. 27330 27331 On the VAX, all functions always pop their arguments, so the 27332 definition of this macro is STACK-SIZE. On the 68000, using the 27333 standard calling convention, no functions pop their arguments, so 27334 the value of the macro is always 0 in this case. But an 27335 alternative calling convention is available in which functions 27336 that take a fixed number of arguments pop them but other functions 27337 (such as `printf') pop nothing (the caller pops all). When this 27338 convention is in use, FUNTYPE is examined to determine whether a 27339 function takes a fixed number of arguments. 27340 27341 -- Macro: CALL_POPS_ARGS (CUM) 27342 A C expression that should indicate the number of bytes a call 27343 sequence pops off the stack. It is added to the value of 27344 `RETURN_POPS_ARGS' when compiling a function call. 27345 27346 CUM is the variable in which all arguments to the called function 27347 have been accumulated. 27348 27349 On certain architectures, such as the SH5, a call trampoline is 27350 used that pops certain registers off the stack, depending on the 27351 arguments that have been passed to the function. Since this is a 27352 property of the call site, not of the called function, 27353 `RETURN_POPS_ARGS' is not appropriate. 27354 27355 27356 File: gccint.info, Node: Register Arguments, Next: Scalar Return, Prev: Stack Arguments, Up: Stack and Calling 27357 27358 17.10.7 Passing Arguments in Registers 27359 -------------------------------------- 27360 27361 This section describes the macros which let you control how various 27362 types of arguments are passed in registers or how they are arranged in 27363 the stack. 27364 27365 -- Macro: FUNCTION_ARG (CUM, MODE, TYPE, NAMED) 27366 A C expression that controls whether a function argument is passed 27367 in a register, and which register. 27368 27369 The arguments are CUM, which summarizes all the previous 27370 arguments; MODE, the machine mode of the argument; TYPE, the data 27371 type of the argument as a tree node or 0 if that is not known 27372 (which happens for C support library functions); and NAMED, which 27373 is 1 for an ordinary argument and 0 for nameless arguments that 27374 correspond to `...' in the called function's prototype. TYPE can 27375 be an incomplete type if a syntax error has previously occurred. 27376 27377 The value of the expression is usually either a `reg' RTX for the 27378 hard register in which to pass the argument, or zero to pass the 27379 argument on the stack. 27380 27381 For machines like the VAX and 68000, where normally all arguments 27382 are pushed, zero suffices as a definition. 27383 27384 The value of the expression can also be a `parallel' RTX. This is 27385 used when an argument is passed in multiple locations. The mode 27386 of the `parallel' should be the mode of the entire argument. The 27387 `parallel' holds any number of `expr_list' pairs; each one 27388 describes where part of the argument is passed. In each 27389 `expr_list' the first operand must be a `reg' RTX for the hard 27390 register in which to pass this part of the argument, and the mode 27391 of the register RTX indicates how large this part of the argument 27392 is. The second operand of the `expr_list' is a `const_int' which 27393 gives the offset in bytes into the entire argument of where this 27394 part starts. As a special exception the first `expr_list' in the 27395 `parallel' RTX may have a first operand of zero. This indicates 27396 that the entire argument is also stored on the stack. 27397 27398 The last time this macro is called, it is called with `MODE == 27399 VOIDmode', and its result is passed to the `call' or `call_value' 27400 pattern as operands 2 and 3 respectively. 27401 27402 The usual way to make the ISO library `stdarg.h' work on a machine 27403 where some arguments are usually passed in registers, is to cause 27404 nameless arguments to be passed on the stack instead. This is done 27405 by making `FUNCTION_ARG' return 0 whenever NAMED is 0. 27406 27407 You may use the hook `targetm.calls.must_pass_in_stack' in the 27408 definition of this macro to determine if this argument is of a 27409 type that must be passed in the stack. If `REG_PARM_STACK_SPACE' 27410 is not defined and `FUNCTION_ARG' returns nonzero for such an 27411 argument, the compiler will abort. If `REG_PARM_STACK_SPACE' is 27412 defined, the argument will be computed in the stack and then 27413 loaded into a register. 27414 27415 -- Target Hook: bool TARGET_MUST_PASS_IN_STACK (enum machine_mode 27416 MODE, tree TYPE) 27417 This target hook should return `true' if we should not pass TYPE 27418 solely in registers. The file `expr.h' defines a definition that 27419 is usually appropriate, refer to `expr.h' for additional 27420 documentation. 27421 27422 -- Macro: FUNCTION_INCOMING_ARG (CUM, MODE, TYPE, NAMED) 27423 Define this macro if the target machine has "register windows", so 27424 that the register in which a function sees an arguments is not 27425 necessarily the same as the one in which the caller passed the 27426 argument. 27427 27428 For such machines, `FUNCTION_ARG' computes the register in which 27429 the caller passes the value, and `FUNCTION_INCOMING_ARG' should be 27430 defined in a similar fashion to tell the function being called 27431 where the arguments will arrive. 27432 27433 If `FUNCTION_INCOMING_ARG' is not defined, `FUNCTION_ARG' serves 27434 both purposes. 27435 27436 -- Target Hook: int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *CUM, 27437 enum machine_mode MODE, tree TYPE, bool NAMED) 27438 This target hook returns the number of bytes at the beginning of an 27439 argument that must be put in registers. The value must be zero for 27440 arguments that are passed entirely in registers or that are 27441 entirely pushed on the stack. 27442 27443 On some machines, certain arguments must be passed partially in 27444 registers and partially in memory. On these machines, typically 27445 the first few words of arguments are passed in registers, and the 27446 rest on the stack. If a multi-word argument (a `double' or a 27447 structure) crosses that boundary, its first few words must be 27448 passed in registers and the rest must be pushed. This macro tells 27449 the compiler when this occurs, and how many bytes should go in 27450 registers. 27451 27452 `FUNCTION_ARG' for these arguments should return the first 27453 register to be used by the caller for this argument; likewise 27454 `FUNCTION_INCOMING_ARG', for the called function. 27455 27456 -- Target Hook: bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *CUM, 27457 enum machine_mode MODE, tree TYPE, bool NAMED) 27458 This target hook should return `true' if an argument at the 27459 position indicated by CUM should be passed by reference. This 27460 predicate is queried after target independent reasons for being 27461 passed by reference, such as `TREE_ADDRESSABLE (type)'. 27462 27463 If the hook returns true, a copy of that argument is made in 27464 memory and a pointer to the argument is passed instead of the 27465 argument itself. The pointer is passed in whatever way is 27466 appropriate for passing a pointer to that type. 27467 27468 -- Target Hook: bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *CUM, enum 27469 machine_mode MODE, tree TYPE, bool NAMED) 27470 The function argument described by the parameters to this hook is 27471 known to be passed by reference. The hook should return true if 27472 the function argument should be copied by the callee instead of 27473 copied by the caller. 27474 27475 For any argument for which the hook returns true, if it can be 27476 determined that the argument is not modified, then a copy need not 27477 be generated. 27478 27479 The default version of this hook always returns false. 27480 27481 -- Macro: CUMULATIVE_ARGS 27482 A C type for declaring a variable that is used as the first 27483 argument of `FUNCTION_ARG' and other related values. For some 27484 target machines, the type `int' suffices and can hold the number 27485 of bytes of argument so far. 27486 27487 There is no need to record in `CUMULATIVE_ARGS' anything about the 27488 arguments that have been passed on the stack. The compiler has 27489 other variables to keep track of that. For target machines on 27490 which all arguments are passed on the stack, there is no need to 27491 store anything in `CUMULATIVE_ARGS'; however, the data structure 27492 must exist and should not be empty, so use `int'. 27493 27494 -- Macro: OVERRIDE_ABI_FORMAT (FNDECL) 27495 If defined, this macro is called before generating any code for a 27496 function, but after the CFUN descriptor for the function has been 27497 created. The back end may use this macro to update CFUN to 27498 reflect an ABI other than that which would normally be used by 27499 default. If the compiler is generating code for a 27500 compiler-generated function, FNDECL may be `NULL'. 27501 27502 -- Macro: INIT_CUMULATIVE_ARGS (CUM, FNTYPE, LIBNAME, FNDECL, 27503 N_NAMED_ARGS) 27504 A C statement (sans semicolon) for initializing the variable CUM 27505 for the state at the beginning of the argument list. The variable 27506 has type `CUMULATIVE_ARGS'. The value of FNTYPE is the tree node 27507 for the data type of the function which will receive the args, or 27508 0 if the args are to a compiler support library function. For 27509 direct calls that are not libcalls, FNDECL contain the declaration 27510 node of the function. FNDECL is also set when 27511 `INIT_CUMULATIVE_ARGS' is used to find arguments for the function 27512 being compiled. N_NAMED_ARGS is set to the number of named 27513 arguments, including a structure return address if it is passed as 27514 a parameter, when making a call. When processing incoming 27515 arguments, N_NAMED_ARGS is set to -1. 27516 27517 When processing a call to a compiler support library function, 27518 LIBNAME identifies which one. It is a `symbol_ref' rtx which 27519 contains the name of the function, as a string. LIBNAME is 0 when 27520 an ordinary C function call is being processed. Thus, each time 27521 this macro is called, either LIBNAME or FNTYPE is nonzero, but 27522 never both of them at once. 27523 27524 -- Macro: INIT_CUMULATIVE_LIBCALL_ARGS (CUM, MODE, LIBNAME) 27525 Like `INIT_CUMULATIVE_ARGS' but only used for outgoing libcalls, 27526 it gets a `MODE' argument instead of FNTYPE, that would be `NULL'. 27527 INDIRECT would always be zero, too. If this macro is not 27528 defined, `INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, 0)' is 27529 used instead. 27530 27531 -- Macro: INIT_CUMULATIVE_INCOMING_ARGS (CUM, FNTYPE, LIBNAME) 27532 Like `INIT_CUMULATIVE_ARGS' but overrides it for the purposes of 27533 finding the arguments for the function being compiled. If this 27534 macro is undefined, `INIT_CUMULATIVE_ARGS' is used instead. 27535 27536 The value passed for LIBNAME is always 0, since library routines 27537 with special calling conventions are never compiled with GCC. The 27538 argument LIBNAME exists for symmetry with `INIT_CUMULATIVE_ARGS'. 27539 27540 -- Macro: FUNCTION_ARG_ADVANCE (CUM, MODE, TYPE, NAMED) 27541 A C statement (sans semicolon) to update the summarizer variable 27542 CUM to advance past an argument in the argument list. The values 27543 MODE, TYPE and NAMED describe that argument. Once this is done, 27544 the variable CUM is suitable for analyzing the _following_ 27545 argument with `FUNCTION_ARG', etc. 27546 27547 This macro need not do anything if the argument in question was 27548 passed on the stack. The compiler knows how to track the amount 27549 of stack space used for arguments without any special help. 27550 27551 -- Macro: FUNCTION_ARG_OFFSET (MODE, TYPE) 27552 If defined, a C expression that is the number of bytes to add to 27553 the offset of the argument passed in memory. This is needed for 27554 the SPU, which passes `char' and `short' arguments in the preferred 27555 slot that is in the middle of the quad word instead of starting at 27556 the top. 27557 27558 -- Macro: FUNCTION_ARG_PADDING (MODE, TYPE) 27559 If defined, a C expression which determines whether, and in which 27560 direction, to pad out an argument with extra space. The value 27561 should be of type `enum direction': either `upward' to pad above 27562 the argument, `downward' to pad below, or `none' to inhibit 27563 padding. 27564 27565 The _amount_ of padding is always just enough to reach the next 27566 multiple of `FUNCTION_ARG_BOUNDARY'; this macro does not control 27567 it. 27568 27569 This macro has a default definition which is right for most 27570 systems. For little-endian machines, the default is to pad 27571 upward. For big-endian machines, the default is to pad downward 27572 for an argument of constant size shorter than an `int', and upward 27573 otherwise. 27574 27575 -- Macro: PAD_VARARGS_DOWN 27576 If defined, a C expression which determines whether the default 27577 implementation of va_arg will attempt to pad down before reading 27578 the next argument, if that argument is smaller than its aligned 27579 space as controlled by `PARM_BOUNDARY'. If this macro is not 27580 defined, all such arguments are padded down if `BYTES_BIG_ENDIAN' 27581 is true. 27582 27583 -- Macro: BLOCK_REG_PADDING (MODE, TYPE, FIRST) 27584 Specify padding for the last element of a block move between 27585 registers and memory. FIRST is nonzero if this is the only 27586 element. Defining this macro allows better control of register 27587 function parameters on big-endian machines, without using 27588 `PARALLEL' rtl. In particular, `MUST_PASS_IN_STACK' need not test 27589 padding and mode of types in registers, as there is no longer a 27590 "wrong" part of a register; For example, a three byte aggregate 27591 may be passed in the high part of a register if so required. 27592 27593 -- Macro: FUNCTION_ARG_BOUNDARY (MODE, TYPE) 27594 If defined, a C expression that gives the alignment boundary, in 27595 bits, of an argument with the specified mode and type. If it is 27596 not defined, `PARM_BOUNDARY' is used for all arguments. 27597 27598 -- Macro: FUNCTION_ARG_REGNO_P (REGNO) 27599 A C expression that is nonzero if REGNO is the number of a hard 27600 register in which function arguments are sometimes passed. This 27601 does _not_ include implicit arguments such as the static chain and 27602 the structure-value address. On many machines, no registers can be 27603 used for this purpose since all function arguments are pushed on 27604 the stack. 27605 27606 -- Target Hook: bool TARGET_SPLIT_COMPLEX_ARG (tree TYPE) 27607 This hook should return true if parameter of type TYPE are passed 27608 as two scalar parameters. By default, GCC will attempt to pack 27609 complex arguments into the target's word size. Some ABIs require 27610 complex arguments to be split and treated as their individual 27611 components. For example, on AIX64, complex floats should be 27612 passed in a pair of floating point registers, even though a 27613 complex float would fit in one 64-bit floating point register. 27614 27615 The default value of this hook is `NULL', which is treated as 27616 always false. 27617 27618 -- Target Hook: tree TARGET_BUILD_BUILTIN_VA_LIST (void) 27619 This hook returns a type node for `va_list' for the target. The 27620 default version of the hook returns `void*'. 27621 27622 -- Target Hook: tree TARGET_FN_ABI_VA_LIST (tree FNDECL) 27623 This hook returns the va_list type of the calling convention 27624 specified by FNDECL. The default version of this hook returns 27625 `va_list_type_node'. 27626 27627 -- Target Hook: tree TARGET_CANONICAL_VA_LIST_TYPE (tree TYPE) 27628 This hook returns the va_list type of the calling convention 27629 specified by the type of TYPE. If TYPE is not a valid va_list 27630 type, it returns `NULL_TREE'. 27631 27632 -- Target Hook: tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree VALIST, tree 27633 TYPE, tree *PRE_P, tree *POST_P) 27634 This hook performs target-specific gimplification of 27635 `VA_ARG_EXPR'. The first two parameters correspond to the 27636 arguments to `va_arg'; the latter two are as in 27637 `gimplify.c:gimplify_expr'. 27638 27639 -- Target Hook: bool TARGET_VALID_POINTER_MODE (enum machine_mode MODE) 27640 Define this to return nonzero if the port can handle pointers with 27641 machine mode MODE. The default version of this hook returns true 27642 for both `ptr_mode' and `Pmode'. 27643 27644 -- Target Hook: bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode 27645 MODE) 27646 Define this to return nonzero if the port is prepared to handle 27647 insns involving scalar mode MODE. For a scalar mode to be 27648 considered supported, all the basic arithmetic and comparisons 27649 must work. 27650 27651 The default version of this hook returns true for any mode 27652 required to handle the basic C types (as defined by the port). 27653 Included here are the double-word arithmetic supported by the code 27654 in `optabs.c'. 27655 27656 -- Target Hook: bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode 27657 MODE) 27658 Define this to return nonzero if the port is prepared to handle 27659 insns involving vector mode MODE. At the very least, it must have 27660 move patterns for this mode. 27661 27662 27663 File: gccint.info, Node: Scalar Return, Next: Aggregate Return, Prev: Register Arguments, Up: Stack and Calling 27664 27665 17.10.8 How Scalar Function Values Are Returned 27666 ----------------------------------------------- 27667 27668 This section discusses the macros that control returning scalars as 27669 values--values that can fit in registers. 27670 27671 -- Target Hook: rtx TARGET_FUNCTION_VALUE (tree RET_TYPE, tree 27672 FN_DECL_OR_TYPE, bool OUTGOING) 27673 Define this to return an RTX representing the place where a 27674 function returns or receives a value of data type RET_TYPE, a tree 27675 node node representing a data type. FN_DECL_OR_TYPE is a tree node 27676 representing `FUNCTION_DECL' or `FUNCTION_TYPE' of a function 27677 being called. If OUTGOING is false, the hook should compute the 27678 register in which the caller will see the return value. 27679 Otherwise, the hook should return an RTX representing the place 27680 where a function returns a value. 27681 27682 On many machines, only `TYPE_MODE (RET_TYPE)' is relevant. 27683 (Actually, on most machines, scalar values are returned in the same 27684 place regardless of mode.) The value of the expression is usually 27685 a `reg' RTX for the hard register where the return value is stored. 27686 The value can also be a `parallel' RTX, if the return value is in 27687 multiple places. See `FUNCTION_ARG' for an explanation of the 27688 `parallel' form. Note that the callee will populate every 27689 location specified in the `parallel', but if the first element of 27690 the `parallel' contains the whole return value, callers will use 27691 that element as the canonical location and ignore the others. The 27692 m68k port uses this type of `parallel' to return pointers in both 27693 `%a0' (the canonical location) and `%d0'. 27694 27695 If `TARGET_PROMOTE_FUNCTION_RETURN' returns true, you must apply 27696 the same promotion rules specified in `PROMOTE_MODE' if VALTYPE is 27697 a scalar type. 27698 27699 If the precise function being called is known, FUNC is a tree node 27700 (`FUNCTION_DECL') for it; otherwise, FUNC is a null pointer. This 27701 makes it possible to use a different value-returning convention 27702 for specific functions when all their calls are known. 27703 27704 Some target machines have "register windows" so that the register 27705 in which a function returns its value is not the same as the one 27706 in which the caller sees the value. For such machines, you should 27707 return different RTX depending on OUTGOING. 27708 27709 `TARGET_FUNCTION_VALUE' is not used for return values with 27710 aggregate data types, because these are returned in another way. 27711 See `TARGET_STRUCT_VALUE_RTX' and related macros, below. 27712 27713 -- Macro: FUNCTION_VALUE (VALTYPE, FUNC) 27714 This macro has been deprecated. Use `TARGET_FUNCTION_VALUE' for a 27715 new target instead. 27716 27717 -- Macro: FUNCTION_OUTGOING_VALUE (VALTYPE, FUNC) 27718 This macro has been deprecated. Use `TARGET_FUNCTION_VALUE' for a 27719 new target instead. 27720 27721 -- Macro: LIBCALL_VALUE (MODE) 27722 A C expression to create an RTX representing the place where a 27723 library function returns a value of mode MODE. 27724 27725 Note that "library function" in this context means a compiler 27726 support routine, used to perform arithmetic, whose name is known 27727 specially by the compiler and was not mentioned in the C code being 27728 compiled. 27729 27730 -- Target Hook: rtx TARGET_LIBCALL_VALUE (enum machine_mode 27731 MODE, rtx FUN) Define this hook if the back-end needs to know the 27732 name of the libcall function in order to determine where the 27733 result should be returned. 27734 27735 The mode of the result is given by MODE and the name of the called 27736 library function is given by FUN. The hook should return an RTX 27737 representing the place where the library function result will be 27738 returned. 27739 27740 If this hook is not defined, then LIBCALL_VALUE will be used. 27741 27742 -- Macro: FUNCTION_VALUE_REGNO_P (REGNO) 27743 A C expression that is nonzero if REGNO is the number of a hard 27744 register in which the values of called function may come back. 27745 27746 A register whose use for returning values is limited to serving as 27747 the second of a pair (for a value of type `double', say) need not 27748 be recognized by this macro. So for most machines, this definition 27749 suffices: 27750 27751 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) 27752 27753 If the machine has register windows, so that the caller and the 27754 called function use different registers for the return value, this 27755 macro should recognize only the caller's register numbers. 27756 27757 -- Macro: TARGET_ENUM_VA_LIST (IDX, PNAME, PTYPE) 27758 This target macro is used in function `c_common_nodes_and_builtins' 27759 to iterate through the target specific builtin types for va_list. 27760 The variable IDX is used as iterator. PNAME has to be a pointer to 27761 a `const char *' and PTYPE a pointer to a `tree' typed variable. 27762 The arguments PNAME and PTYPE are used to store the result of this 27763 macro and are set to the name of the va_list builtin type and its 27764 internal type. If the return value of this macro is zero, then 27765 there is no more element. Otherwise the IDX should be increased 27766 for the next call of this macro to iterate through all types. 27767 27768 -- Macro: APPLY_RESULT_SIZE 27769 Define this macro if `untyped_call' and `untyped_return' need more 27770 space than is implied by `FUNCTION_VALUE_REGNO_P' for saving and 27771 restoring an arbitrary return value. 27772 27773 -- Target Hook: bool TARGET_RETURN_IN_MSB (tree TYPE) 27774 This hook should return true if values of type TYPE are returned 27775 at the most significant end of a register (in other words, if they 27776 are padded at the least significant end). You can assume that TYPE 27777 is returned in a register; the caller is required to check this. 27778 27779 Note that the register provided by `TARGET_FUNCTION_VALUE' must be 27780 able to hold the complete return value. For example, if a 1-, 2- 27781 or 3-byte structure is returned at the most significant end of a 27782 4-byte register, `TARGET_FUNCTION_VALUE' should provide an 27783 `SImode' rtx. 27784 27785 27786 File: gccint.info, Node: Aggregate Return, Next: Caller Saves, Prev: Scalar Return, Up: Stack and Calling 27787 27788 17.10.9 How Large Values Are Returned 27789 ------------------------------------- 27790 27791 When a function value's mode is `BLKmode' (and in some other cases), 27792 the value is not returned according to `TARGET_FUNCTION_VALUE' (*note 27793 Scalar Return::). Instead, the caller passes the address of a block of 27794 memory in which the value should be stored. This address is called the 27795 "structure value address". 27796 27797 This section describes how to control returning structure values in 27798 memory. 27799 27800 -- Target Hook: bool TARGET_RETURN_IN_MEMORY (tree TYPE, tree FNTYPE) 27801 This target hook should return a nonzero value to say to return the 27802 function value in memory, just as large structures are always 27803 returned. Here TYPE will be the data type of the value, and FNTYPE 27804 will be the type of the function doing the returning, or `NULL' for 27805 libcalls. 27806 27807 Note that values of mode `BLKmode' must be explicitly handled by 27808 this function. Also, the option `-fpcc-struct-return' takes 27809 effect regardless of this macro. On most systems, it is possible 27810 to leave the hook undefined; this causes a default definition to 27811 be used, whose value is the constant 1 for `BLKmode' values, and 0 27812 otherwise. 27813 27814 Do not use this hook to indicate that structures and unions should 27815 always be returned in memory. You should instead use 27816 `DEFAULT_PCC_STRUCT_RETURN' to indicate this. 27817 27818 -- Macro: DEFAULT_PCC_STRUCT_RETURN 27819 Define this macro to be 1 if all structure and union return values 27820 must be in memory. Since this results in slower code, this should 27821 be defined only if needed for compatibility with other compilers 27822 or with an ABI. If you define this macro to be 0, then the 27823 conventions used for structure and union return values are decided 27824 by the `TARGET_RETURN_IN_MEMORY' target hook. 27825 27826 If not defined, this defaults to the value 1. 27827 27828 -- Target Hook: rtx TARGET_STRUCT_VALUE_RTX (tree FNDECL, int INCOMING) 27829 This target hook should return the location of the structure value 27830 address (normally a `mem' or `reg'), or 0 if the address is passed 27831 as an "invisible" first argument. Note that FNDECL may be `NULL', 27832 for libcalls. You do not need to define this target hook if the 27833 address is always passed as an "invisible" first argument. 27834 27835 On some architectures the place where the structure value address 27836 is found by the called function is not the same place that the 27837 caller put it. This can be due to register windows, or it could 27838 be because the function prologue moves it to a different place. 27839 INCOMING is `1' or `2' when the location is needed in the context 27840 of the called function, and `0' in the context of the caller. 27841 27842 If INCOMING is nonzero and the address is to be found on the 27843 stack, return a `mem' which refers to the frame pointer. If 27844 INCOMING is `2', the result is being used to fetch the structure 27845 value address at the beginning of a function. If you need to emit 27846 adjusting code, you should do it at this point. 27847 27848 -- Macro: PCC_STATIC_STRUCT_RETURN 27849 Define this macro if the usual system convention on the target 27850 machine for returning structures and unions is for the called 27851 function to return the address of a static variable containing the 27852 value. 27853 27854 Do not define this if the usual system convention is for the 27855 caller to pass an address to the subroutine. 27856 27857 This macro has effect in `-fpcc-struct-return' mode, but it does 27858 nothing when you use `-freg-struct-return' mode. 27859 27860 27861 File: gccint.info, Node: Caller Saves, Next: Function Entry, Prev: Aggregate Return, Up: Stack and Calling 27862 27863 17.10.10 Caller-Saves Register Allocation 27864 ----------------------------------------- 27865 27866 If you enable it, GCC can save registers around function calls. This 27867 makes it possible to use call-clobbered registers to hold variables that 27868 must live across calls. 27869 27870 -- Macro: CALLER_SAVE_PROFITABLE (REFS, CALLS) 27871 A C expression to determine whether it is worthwhile to consider 27872 placing a pseudo-register in a call-clobbered hard register and 27873 saving and restoring it around each function call. The expression 27874 should be 1 when this is worth doing, and 0 otherwise. 27875 27876 If you don't define this macro, a default is used which is good on 27877 most machines: `4 * CALLS < REFS'. 27878 27879 -- Macro: HARD_REGNO_CALLER_SAVE_MODE (REGNO, NREGS) 27880 A C expression specifying which mode is required for saving NREGS 27881 of a pseudo-register in call-clobbered hard register REGNO. If 27882 REGNO is unsuitable for caller save, `VOIDmode' should be 27883 returned. For most machines this macro need not be defined since 27884 GCC will select the smallest suitable mode. 27885 27886 27887 File: gccint.info, Node: Function Entry, Next: Profiling, Prev: Caller Saves, Up: Stack and Calling 27888 27889 17.10.11 Function Entry and Exit 27890 -------------------------------- 27891 27892 This section describes the macros that output function entry 27893 ("prologue") and exit ("epilogue") code. 27894 27895 -- Target Hook: void TARGET_ASM_FUNCTION_PROLOGUE (FILE *FILE, 27896 HOST_WIDE_INT SIZE) 27897 If defined, a function that outputs the assembler code for entry 27898 to a function. The prologue is responsible for setting up the 27899 stack frame, initializing the frame pointer register, saving 27900 registers that must be saved, and allocating SIZE additional bytes 27901 of storage for the local variables. SIZE is an integer. FILE is 27902 a stdio stream to which the assembler code should be output. 27903 27904 The label for the beginning of the function need not be output by 27905 this macro. That has already been done when the macro is run. 27906 27907 To determine which registers to save, the macro can refer to the 27908 array `regs_ever_live': element R is nonzero if hard register R is 27909 used anywhere within the function. This implies the function 27910 prologue should save register R, provided it is not one of the 27911 call-used registers. (`TARGET_ASM_FUNCTION_EPILOGUE' must 27912 likewise use `regs_ever_live'.) 27913 27914 On machines that have "register windows", the function entry code 27915 does not save on the stack the registers that are in the windows, 27916 even if they are supposed to be preserved by function calls; 27917 instead it takes appropriate steps to "push" the register stack, 27918 if any non-call-used registers are used in the function. 27919 27920 On machines where functions may or may not have frame-pointers, the 27921 function entry code must vary accordingly; it must set up the frame 27922 pointer if one is wanted, and not otherwise. To determine whether 27923 a frame pointer is in wanted, the macro can refer to the variable 27924 `frame_pointer_needed'. The variable's value will be 1 at run 27925 time in a function that needs a frame pointer. *Note 27926 Elimination::. 27927 27928 The function entry code is responsible for allocating any stack 27929 space required for the function. This stack space consists of the 27930 regions listed below. In most cases, these regions are allocated 27931 in the order listed, with the last listed region closest to the 27932 top of the stack (the lowest address if `STACK_GROWS_DOWNWARD' is 27933 defined, and the highest address if it is not defined). You can 27934 use a different order for a machine if doing so is more convenient 27935 or required for compatibility reasons. Except in cases where 27936 required by standard or by a debugger, there is no reason why the 27937 stack layout used by GCC need agree with that used by other 27938 compilers for a machine. 27939 27940 -- Target Hook: void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *FILE) 27941 If defined, a function that outputs assembler code at the end of a 27942 prologue. This should be used when the function prologue is being 27943 emitted as RTL, and you have some extra assembler that needs to be 27944 emitted. *Note prologue instruction pattern::. 27945 27946 -- Target Hook: void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *FILE) 27947 If defined, a function that outputs assembler code at the start of 27948 an epilogue. This should be used when the function epilogue is 27949 being emitted as RTL, and you have some extra assembler that needs 27950 to be emitted. *Note epilogue instruction pattern::. 27951 27952 -- Target Hook: void TARGET_ASM_FUNCTION_EPILOGUE (FILE *FILE, 27953 HOST_WIDE_INT SIZE) 27954 If defined, a function that outputs the assembler code for exit 27955 from a function. The epilogue is responsible for restoring the 27956 saved registers and stack pointer to their values when the 27957 function was called, and returning control to the caller. This 27958 macro takes the same arguments as the macro 27959 `TARGET_ASM_FUNCTION_PROLOGUE', and the registers to restore are 27960 determined from `regs_ever_live' and `CALL_USED_REGISTERS' in the 27961 same way. 27962 27963 On some machines, there is a single instruction that does all the 27964 work of returning from the function. On these machines, give that 27965 instruction the name `return' and do not define the macro 27966 `TARGET_ASM_FUNCTION_EPILOGUE' at all. 27967 27968 Do not define a pattern named `return' if you want the 27969 `TARGET_ASM_FUNCTION_EPILOGUE' to be used. If you want the target 27970 switches to control whether return instructions or epilogues are 27971 used, define a `return' pattern with a validity condition that 27972 tests the target switches appropriately. If the `return' 27973 pattern's validity condition is false, epilogues will be used. 27974 27975 On machines where functions may or may not have frame-pointers, the 27976 function exit code must vary accordingly. Sometimes the code for 27977 these two cases is completely different. To determine whether a 27978 frame pointer is wanted, the macro can refer to the variable 27979 `frame_pointer_needed'. The variable's value will be 1 when 27980 compiling a function that needs a frame pointer. 27981 27982 Normally, `TARGET_ASM_FUNCTION_PROLOGUE' and 27983 `TARGET_ASM_FUNCTION_EPILOGUE' must treat leaf functions specially. 27984 The C variable `current_function_is_leaf' is nonzero for such a 27985 function. *Note Leaf Functions::. 27986 27987 On some machines, some functions pop their arguments on exit while 27988 others leave that for the caller to do. For example, the 68020 27989 when given `-mrtd' pops arguments in functions that take a fixed 27990 number of arguments. 27991 27992 Your definition of the macro `RETURN_POPS_ARGS' decides which 27993 functions pop their own arguments. `TARGET_ASM_FUNCTION_EPILOGUE' 27994 needs to know what was decided. The variable that is called 27995 `current_function_pops_args' is the number of bytes of its 27996 arguments that a function should pop. *Note Scalar Return::. 27997 27998 * A region of `current_function_pretend_args_size' bytes of 27999 uninitialized space just underneath the first argument arriving on 28000 the stack. (This may not be at the very start of the allocated 28001 stack region if the calling sequence has pushed anything else 28002 since pushing the stack arguments. But usually, on such machines, 28003 nothing else has been pushed yet, because the function prologue 28004 itself does all the pushing.) This region is used on machines 28005 where an argument may be passed partly in registers and partly in 28006 memory, and, in some cases to support the features in `<stdarg.h>'. 28007 28008 * An area of memory used to save certain registers used by the 28009 function. The size of this area, which may also include space for 28010 such things as the return address and pointers to previous stack 28011 frames, is machine-specific and usually depends on which registers 28012 have been used in the function. Machines with register windows 28013 often do not require a save area. 28014 28015 * A region of at least SIZE bytes, possibly rounded up to an 28016 allocation boundary, to contain the local variables of the 28017 function. On some machines, this region and the save area may 28018 occur in the opposite order, with the save area closer to the top 28019 of the stack. 28020 28021 * Optionally, when `ACCUMULATE_OUTGOING_ARGS' is defined, a region of 28022 `current_function_outgoing_args_size' bytes to be used for outgoing 28023 argument lists of the function. *Note Stack Arguments::. 28024 28025 -- Macro: EXIT_IGNORE_STACK 28026 Define this macro as a C expression that is nonzero if the return 28027 instruction or the function epilogue ignores the value of the stack 28028 pointer; in other words, if it is safe to delete an instruction to 28029 adjust the stack pointer before a return from the function. The 28030 default is 0. 28031 28032 Note that this macro's value is relevant only for functions for 28033 which frame pointers are maintained. It is never safe to delete a 28034 final stack adjustment in a function that has no frame pointer, 28035 and the compiler knows this regardless of `EXIT_IGNORE_STACK'. 28036 28037 -- Macro: EPILOGUE_USES (REGNO) 28038 Define this macro as a C expression that is nonzero for registers 28039 that are used by the epilogue or the `return' pattern. The stack 28040 and frame pointer registers are already assumed to be used as 28041 needed. 28042 28043 -- Macro: EH_USES (REGNO) 28044 Define this macro as a C expression that is nonzero for registers 28045 that are used by the exception handling mechanism, and so should 28046 be considered live on entry to an exception edge. 28047 28048 -- Macro: DELAY_SLOTS_FOR_EPILOGUE 28049 Define this macro if the function epilogue contains delay slots to 28050 which instructions from the rest of the function can be "moved". 28051 The definition should be a C expression whose value is an integer 28052 representing the number of delay slots there. 28053 28054 -- Macro: ELIGIBLE_FOR_EPILOGUE_DELAY (INSN, N) 28055 A C expression that returns 1 if INSN can be placed in delay slot 28056 number N of the epilogue. 28057 28058 The argument N is an integer which identifies the delay slot now 28059 being considered (since different slots may have different rules of 28060 eligibility). It is never negative and is always less than the 28061 number of epilogue delay slots (what `DELAY_SLOTS_FOR_EPILOGUE' 28062 returns). If you reject a particular insn for a given delay slot, 28063 in principle, it may be reconsidered for a subsequent delay slot. 28064 Also, other insns may (at least in principle) be considered for 28065 the so far unfilled delay slot. 28066 28067 The insns accepted to fill the epilogue delay slots are put in an 28068 RTL list made with `insn_list' objects, stored in the variable 28069 `current_function_epilogue_delay_list'. The insn for the first 28070 delay slot comes first in the list. Your definition of the macro 28071 `TARGET_ASM_FUNCTION_EPILOGUE' should fill the delay slots by 28072 outputting the insns in this list, usually by calling 28073 `final_scan_insn'. 28074 28075 You need not define this macro if you did not define 28076 `DELAY_SLOTS_FOR_EPILOGUE'. 28077 28078 -- Target Hook: void TARGET_ASM_OUTPUT_MI_THUNK (FILE *FILE, tree 28079 THUNK_FNDECL, HOST_WIDE_INT DELTA, HOST_WIDE_INT 28080 VCALL_OFFSET, tree FUNCTION) 28081 A function that outputs the assembler code for a thunk function, 28082 used to implement C++ virtual function calls with multiple 28083 inheritance. The thunk acts as a wrapper around a virtual 28084 function, adjusting the implicit object parameter before handing 28085 control off to the real function. 28086 28087 First, emit code to add the integer DELTA to the location that 28088 contains the incoming first argument. Assume that this argument 28089 contains a pointer, and is the one used to pass the `this' pointer 28090 in C++. This is the incoming argument _before_ the function 28091 prologue, e.g. `%o0' on a sparc. The addition must preserve the 28092 values of all other incoming arguments. 28093 28094 Then, if VCALL_OFFSET is nonzero, an additional adjustment should 28095 be made after adding `delta'. In particular, if P is the adjusted 28096 pointer, the following adjustment should be made: 28097 28098 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)] 28099 28100 After the additions, emit code to jump to FUNCTION, which is a 28101 `FUNCTION_DECL'. This is a direct pure jump, not a call, and does 28102 not touch the return address. Hence returning from FUNCTION will 28103 return to whoever called the current `thunk'. 28104 28105 The effect must be as if FUNCTION had been called directly with 28106 the adjusted first argument. This macro is responsible for 28107 emitting all of the code for a thunk function; 28108 `TARGET_ASM_FUNCTION_PROLOGUE' and `TARGET_ASM_FUNCTION_EPILOGUE' 28109 are not invoked. 28110 28111 The THUNK_FNDECL is redundant. (DELTA and FUNCTION have already 28112 been extracted from it.) It might possibly be useful on some 28113 targets, but probably not. 28114 28115 If you do not define this macro, the target-independent code in 28116 the C++ front end will generate a less efficient heavyweight thunk 28117 that calls FUNCTION instead of jumping to it. The generic 28118 approach does not support varargs. 28119 28120 -- Target Hook: bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (tree 28121 THUNK_FNDECL, HOST_WIDE_INT DELTA, HOST_WIDE_INT 28122 VCALL_OFFSET, tree FUNCTION) 28123 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would 28124 be able to output the assembler code for the thunk function 28125 specified by the arguments it is passed, and false otherwise. In 28126 the latter case, the generic approach will be used by the C++ 28127 front end, with the limitations previously exposed. 28128 28129 28130 File: gccint.info, Node: Profiling, Next: Tail Calls, Prev: Function Entry, Up: Stack and Calling 28131 28132 17.10.12 Generating Code for Profiling 28133 -------------------------------------- 28134 28135 These macros will help you generate code for profiling. 28136 28137 -- Macro: FUNCTION_PROFILER (FILE, LABELNO) 28138 A C statement or compound statement to output to FILE some 28139 assembler code to call the profiling subroutine `mcount'. 28140 28141 The details of how `mcount' expects to be called are determined by 28142 your operating system environment, not by GCC. To figure them out, 28143 compile a small program for profiling using the system's installed 28144 C compiler and look at the assembler code that results. 28145 28146 Older implementations of `mcount' expect the address of a counter 28147 variable to be loaded into some register. The name of this 28148 variable is `LP' followed by the number LABELNO, so you would 28149 generate the name using `LP%d' in a `fprintf'. 28150 28151 -- Macro: PROFILE_HOOK 28152 A C statement or compound statement to output to FILE some assembly 28153 code to call the profiling subroutine `mcount' even the target does 28154 not support profiling. 28155 28156 -- Macro: NO_PROFILE_COUNTERS 28157 Define this macro to be an expression with a nonzero value if the 28158 `mcount' subroutine on your system does not need a counter variable 28159 allocated for each function. This is true for almost all modern 28160 implementations. If you define this macro, you must not use the 28161 LABELNO argument to `FUNCTION_PROFILER'. 28162 28163 -- Macro: PROFILE_BEFORE_PROLOGUE 28164 Define this macro if the code for function profiling should come 28165 before the function prologue. Normally, the profiling code comes 28166 after. 28167 28168 28169 File: gccint.info, Node: Tail Calls, Next: Stack Smashing Protection, Prev: Profiling, Up: Stack and Calling 28170 28171 17.10.13 Permitting tail calls 28172 ------------------------------ 28173 28174 -- Target Hook: bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree DECL, tree 28175 EXP) 28176 True if it is ok to do sibling call optimization for the specified 28177 call expression EXP. DECL will be the called function, or `NULL' 28178 if this is an indirect call. 28179 28180 It is not uncommon for limitations of calling conventions to 28181 prevent tail calls to functions outside the current unit of 28182 translation, or during PIC compilation. The hook is used to 28183 enforce these restrictions, as the `sibcall' md pattern can not 28184 fail, or fall over to a "normal" call. The criteria for 28185 successful sibling call optimization may vary greatly between 28186 different architectures. 28187 28188 -- Target Hook: void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *REGS) 28189 Add any hard registers to REGS that are live on entry to the 28190 function. This hook only needs to be defined to provide registers 28191 that cannot be found by examination of FUNCTION_ARG_REGNO_P, the 28192 callee saved registers, STATIC_CHAIN_INCOMING_REGNUM, 28193 STATIC_CHAIN_REGNUM, TARGET_STRUCT_VALUE_RTX, 28194 FRAME_POINTER_REGNUM, EH_USES, FRAME_POINTER_REGNUM, 28195 ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM. 28196 28197 28198 File: gccint.info, Node: Stack Smashing Protection, Prev: Tail Calls, Up: Stack and Calling 28199 28200 17.10.14 Stack smashing protection 28201 ---------------------------------- 28202 28203 -- Target Hook: tree TARGET_STACK_PROTECT_GUARD (void) 28204 This hook returns a `DECL' node for the external variable to use 28205 for the stack protection guard. This variable is initialized by 28206 the runtime to some random value and is used to initialize the 28207 guard value that is placed at the top of the local stack frame. 28208 The type of this variable must be `ptr_type_node'. 28209 28210 The default version of this hook creates a variable called 28211 `__stack_chk_guard', which is normally defined in `libgcc2.c'. 28212 28213 -- Target Hook: tree TARGET_STACK_PROTECT_FAIL (void) 28214 This hook returns a tree expression that alerts the runtime that 28215 the stack protect guard variable has been modified. This 28216 expression should involve a call to a `noreturn' function. 28217 28218 The default version of this hook invokes a function called 28219 `__stack_chk_fail', taking no arguments. This function is 28220 normally defined in `libgcc2.c'. 28221 28222 28223 File: gccint.info, Node: Varargs, Next: Trampolines, Prev: Stack and Calling, Up: Target Macros 28224 28225 17.11 Implementing the Varargs Macros 28226 ===================================== 28227 28228 GCC comes with an implementation of `<varargs.h>' and `<stdarg.h>' that 28229 work without change on machines that pass arguments on the stack. 28230 Other machines require their own implementations of varargs, and the 28231 two machine independent header files must have conditionals to include 28232 it. 28233 28234 ISO `<stdarg.h>' differs from traditional `<varargs.h>' mainly in the 28235 calling convention for `va_start'. The traditional implementation 28236 takes just one argument, which is the variable in which to store the 28237 argument pointer. The ISO implementation of `va_start' takes an 28238 additional second argument. The user is supposed to write the last 28239 named argument of the function here. 28240 28241 However, `va_start' should not use this argument. The way to find the 28242 end of the named arguments is with the built-in functions described 28243 below. 28244 28245 -- Macro: __builtin_saveregs () 28246 Use this built-in function to save the argument registers in 28247 memory so that the varargs mechanism can access them. Both ISO 28248 and traditional versions of `va_start' must use 28249 `__builtin_saveregs', unless you use 28250 `TARGET_SETUP_INCOMING_VARARGS' (see below) instead. 28251 28252 On some machines, `__builtin_saveregs' is open-coded under the 28253 control of the target hook `TARGET_EXPAND_BUILTIN_SAVEREGS'. On 28254 other machines, it calls a routine written in assembler language, 28255 found in `libgcc2.c'. 28256 28257 Code generated for the call to `__builtin_saveregs' appears at the 28258 beginning of the function, as opposed to where the call to 28259 `__builtin_saveregs' is written, regardless of what the code is. 28260 This is because the registers must be saved before the function 28261 starts to use them for its own purposes. 28262 28263 -- Macro: __builtin_args_info (CATEGORY) 28264 Use this built-in function to find the first anonymous arguments in 28265 registers. 28266 28267 In general, a machine may have several categories of registers 28268 used for arguments, each for a particular category of data types. 28269 (For example, on some machines, floating-point registers are used 28270 for floating-point arguments while other arguments are passed in 28271 the general registers.) To make non-varargs functions use the 28272 proper calling convention, you have defined the `CUMULATIVE_ARGS' 28273 data type to record how many registers in each category have been 28274 used so far 28275 28276 `__builtin_args_info' accesses the same data structure of type 28277 `CUMULATIVE_ARGS' after the ordinary argument layout is finished 28278 with it, with CATEGORY specifying which word to access. Thus, the 28279 value indicates the first unused register in a given category. 28280 28281 Normally, you would use `__builtin_args_info' in the implementation 28282 of `va_start', accessing each category just once and storing the 28283 value in the `va_list' object. This is because `va_list' will 28284 have to update the values, and there is no way to alter the values 28285 accessed by `__builtin_args_info'. 28286 28287 -- Macro: __builtin_next_arg (LASTARG) 28288 This is the equivalent of `__builtin_args_info', for stack 28289 arguments. It returns the address of the first anonymous stack 28290 argument, as type `void *'. If `ARGS_GROW_DOWNWARD', it returns 28291 the address of the location above the first anonymous stack 28292 argument. Use it in `va_start' to initialize the pointer for 28293 fetching arguments from the stack. Also use it in `va_start' to 28294 verify that the second parameter LASTARG is the last named argument 28295 of the current function. 28296 28297 -- Macro: __builtin_classify_type (OBJECT) 28298 Since each machine has its own conventions for which data types are 28299 passed in which kind of register, your implementation of `va_arg' 28300 has to embody these conventions. The easiest way to categorize the 28301 specified data type is to use `__builtin_classify_type' together 28302 with `sizeof' and `__alignof__'. 28303 28304 `__builtin_classify_type' ignores the value of OBJECT, considering 28305 only its data type. It returns an integer describing what kind of 28306 type that is--integer, floating, pointer, structure, and so on. 28307 28308 The file `typeclass.h' defines an enumeration that you can use to 28309 interpret the values of `__builtin_classify_type'. 28310 28311 These machine description macros help implement varargs: 28312 28313 -- Target Hook: rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void) 28314 If defined, this hook produces the machine-specific code for a 28315 call to `__builtin_saveregs'. This code will be moved to the very 28316 beginning of the function, before any parameter access are made. 28317 The return value of this function should be an RTX that contains 28318 the value to use as the return of `__builtin_saveregs'. 28319 28320 -- Target Hook: void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS 28321 *ARGS_SO_FAR, enum machine_mode MODE, tree TYPE, int 28322 *PRETEND_ARGS_SIZE, int SECOND_TIME) 28323 This target hook offers an alternative to using 28324 `__builtin_saveregs' and defining the hook 28325 `TARGET_EXPAND_BUILTIN_SAVEREGS'. Use it to store the anonymous 28326 register arguments into the stack so that all the arguments appear 28327 to have been passed consecutively on the stack. Once this is 28328 done, you can use the standard implementation of varargs that 28329 works for machines that pass all their arguments on the stack. 28330 28331 The argument ARGS_SO_FAR points to the `CUMULATIVE_ARGS' data 28332 structure, containing the values that are obtained after 28333 processing the named arguments. The arguments MODE and TYPE 28334 describe the last named argument--its machine mode and its data 28335 type as a tree node. 28336 28337 The target hook should do two things: first, push onto the stack 28338 all the argument registers _not_ used for the named arguments, and 28339 second, store the size of the data thus pushed into the 28340 `int'-valued variable pointed to by PRETEND_ARGS_SIZE. The value 28341 that you store here will serve as additional offset for setting up 28342 the stack frame. 28343 28344 Because you must generate code to push the anonymous arguments at 28345 compile time without knowing their data types, 28346 `TARGET_SETUP_INCOMING_VARARGS' is only useful on machines that 28347 have just a single category of argument register and use it 28348 uniformly for all data types. 28349 28350 If the argument SECOND_TIME is nonzero, it means that the 28351 arguments of the function are being analyzed for the second time. 28352 This happens for an inline function, which is not actually 28353 compiled until the end of the source file. The hook 28354 `TARGET_SETUP_INCOMING_VARARGS' should not generate any 28355 instructions in this case. 28356 28357 -- Target Hook: bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS 28358 *CA) 28359 Define this hook to return `true' if the location where a function 28360 argument is passed depends on whether or not it is a named 28361 argument. 28362 28363 This hook controls how the NAMED argument to `FUNCTION_ARG' is set 28364 for varargs and stdarg functions. If this hook returns `true', 28365 the NAMED argument is always true for named arguments, and false 28366 for unnamed arguments. If it returns `false', but 28367 `TARGET_PRETEND_OUTGOING_VARARGS_NAMED' returns `true', then all 28368 arguments are treated as named. Otherwise, all named arguments 28369 except the last are treated as named. 28370 28371 You need not define this hook if it always returns zero. 28372 28373 -- Target Hook: bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED 28374 If you need to conditionally change ABIs so that one works with 28375 `TARGET_SETUP_INCOMING_VARARGS', but the other works like neither 28376 `TARGET_SETUP_INCOMING_VARARGS' nor 28377 `TARGET_STRICT_ARGUMENT_NAMING' was defined, then define this hook 28378 to return `true' if `TARGET_SETUP_INCOMING_VARARGS' is used, 28379 `false' otherwise. Otherwise, you should not define this hook. 28380 28381 28382 File: gccint.info, Node: Trampolines, Next: Library Calls, Prev: Varargs, Up: Target Macros 28383 28384 17.12 Trampolines for Nested Functions 28385 ====================================== 28386 28387 A "trampoline" is a small piece of code that is created at run time 28388 when the address of a nested function is taken. It normally resides on 28389 the stack, in the stack frame of the containing function. These macros 28390 tell GCC how to generate code to allocate and initialize a trampoline. 28391 28392 The instructions in the trampoline must do two things: load a constant 28393 address into the static chain register, and jump to the real address of 28394 the nested function. On CISC machines such as the m68k, this requires 28395 two instructions, a move immediate and a jump. Then the two addresses 28396 exist in the trampoline as word-long immediate operands. On RISC 28397 machines, it is often necessary to load each address into a register in 28398 two parts. Then pieces of each address form separate immediate 28399 operands. 28400 28401 The code generated to initialize the trampoline must store the variable 28402 parts--the static chain value and the function address--into the 28403 immediate operands of the instructions. On a CISC machine, this is 28404 simply a matter of copying each address to a memory reference at the 28405 proper offset from the start of the trampoline. On a RISC machine, it 28406 may be necessary to take out pieces of the address and store them 28407 separately. 28408 28409 -- Macro: TRAMPOLINE_TEMPLATE (FILE) 28410 A C statement to output, on the stream FILE, assembler code for a 28411 block of data that contains the constant parts of a trampoline. 28412 This code should not include a label--the label is taken care of 28413 automatically. 28414 28415 If you do not define this macro, it means no template is needed 28416 for the target. Do not define this macro on systems where the 28417 block move code to copy the trampoline into place would be larger 28418 than the code to generate it on the spot. 28419 28420 -- Macro: TRAMPOLINE_SECTION 28421 Return the section into which the trampoline template is to be 28422 placed (*note Sections::). The default value is 28423 `readonly_data_section'. 28424 28425 -- Macro: TRAMPOLINE_SIZE 28426 A C expression for the size in bytes of the trampoline, as an 28427 integer. 28428 28429 -- Macro: TRAMPOLINE_ALIGNMENT 28430 Alignment required for trampolines, in bits. 28431 28432 If you don't define this macro, the value of `BIGGEST_ALIGNMENT' 28433 is used for aligning trampolines. 28434 28435 -- Macro: INITIALIZE_TRAMPOLINE (ADDR, FNADDR, STATIC_CHAIN) 28436 A C statement to initialize the variable parts of a trampoline. 28437 ADDR is an RTX for the address of the trampoline; FNADDR is an RTX 28438 for the address of the nested function; STATIC_CHAIN is an RTX for 28439 the static chain value that should be passed to the function when 28440 it is called. 28441 28442 -- Macro: TRAMPOLINE_ADJUST_ADDRESS (ADDR) 28443 A C statement that should perform any machine-specific adjustment 28444 in the address of the trampoline. Its argument contains the 28445 address that was passed to `INITIALIZE_TRAMPOLINE'. In case the 28446 address to be used for a function call should be different from 28447 the address in which the template was stored, the different 28448 address should be assigned to ADDR. If this macro is not defined, 28449 ADDR will be used for function calls. 28450 28451 If this macro is not defined, by default the trampoline is 28452 allocated as a stack slot. This default is right for most 28453 machines. The exceptions are machines where it is impossible to 28454 execute instructions in the stack area. On such machines, you may 28455 have to implement a separate stack, using this macro in 28456 conjunction with `TARGET_ASM_FUNCTION_PROLOGUE' and 28457 `TARGET_ASM_FUNCTION_EPILOGUE'. 28458 28459 FP points to a data structure, a `struct function', which 28460 describes the compilation status of the immediate containing 28461 function of the function which the trampoline is for. The stack 28462 slot for the trampoline is in the stack frame of this containing 28463 function. Other allocation strategies probably must do something 28464 analogous with this information. 28465 28466 Implementing trampolines is difficult on many machines because they 28467 have separate instruction and data caches. Writing into a stack 28468 location fails to clear the memory in the instruction cache, so when 28469 the program jumps to that location, it executes the old contents. 28470 28471 Here are two possible solutions. One is to clear the relevant parts of 28472 the instruction cache whenever a trampoline is set up. The other is to 28473 make all trampolines identical, by having them jump to a standard 28474 subroutine. The former technique makes trampoline execution faster; the 28475 latter makes initialization faster. 28476 28477 To clear the instruction cache when a trampoline is initialized, define 28478 the following macro. 28479 28480 -- Macro: CLEAR_INSN_CACHE (BEG, END) 28481 If defined, expands to a C expression clearing the _instruction 28482 cache_ in the specified interval. The definition of this macro 28483 would typically be a series of `asm' statements. Both BEG and END 28484 are both pointer expressions. 28485 28486 The operating system may also require the stack to be made executable 28487 before calling the trampoline. To implement this requirement, define 28488 the following macro. 28489 28490 -- Macro: ENABLE_EXECUTE_STACK 28491 Define this macro if certain operations must be performed before 28492 executing code located on the stack. The macro should expand to a 28493 series of C file-scope constructs (e.g. functions) and provide a 28494 unique entry point named `__enable_execute_stack'. The target is 28495 responsible for emitting calls to the entry point in the code, for 28496 example from the `INITIALIZE_TRAMPOLINE' macro. 28497 28498 To use a standard subroutine, define the following macro. In addition, 28499 you must make sure that the instructions in a trampoline fill an entire 28500 cache line with identical instructions, or else ensure that the 28501 beginning of the trampoline code is always aligned at the same point in 28502 its cache line. Look in `m68k.h' as a guide. 28503 28504 -- Macro: TRANSFER_FROM_TRAMPOLINE 28505 Define this macro if trampolines need a special subroutine to do 28506 their work. The macro should expand to a series of `asm' 28507 statements which will be compiled with GCC. They go in a library 28508 function named `__transfer_from_trampoline'. 28509 28510 If you need to avoid executing the ordinary prologue code of a 28511 compiled C function when you jump to the subroutine, you can do so 28512 by placing a special label of your own in the assembler code. Use 28513 one `asm' statement to generate an assembler label, and another to 28514 make the label global. Then trampolines can use that label to 28515 jump directly to your special assembler code. 28516 28517 28518 File: gccint.info, Node: Library Calls, Next: Addressing Modes, Prev: Trampolines, Up: Target Macros 28519 28520 17.13 Implicit Calls to Library Routines 28521 ======================================== 28522 28523 Here is an explanation of implicit calls to library routines. 28524 28525 -- Macro: DECLARE_LIBRARY_RENAMES 28526 This macro, if defined, should expand to a piece of C code that 28527 will get expanded when compiling functions for libgcc.a. It can 28528 be used to provide alternate names for GCC's internal library 28529 functions if there are ABI-mandated names that the compiler should 28530 provide. 28531 28532 -- Target Hook: void TARGET_INIT_LIBFUNCS (void) 28533 This hook should declare additional library routines or rename 28534 existing ones, using the functions `set_optab_libfunc' and 28535 `init_one_libfunc' defined in `optabs.c'. `init_optabs' calls 28536 this macro after initializing all the normal library routines. 28537 28538 The default is to do nothing. Most ports don't need to define 28539 this hook. 28540 28541 -- Macro: FLOAT_LIB_COMPARE_RETURNS_BOOL (MODE, COMPARISON) 28542 This macro should return `true' if the library routine that 28543 implements the floating point comparison operator COMPARISON in 28544 mode MODE will return a boolean, and FALSE if it will return a 28545 tristate. 28546 28547 GCC's own floating point libraries return tristates from the 28548 comparison operators, so the default returns false always. Most 28549 ports don't need to define this macro. 28550 28551 -- Macro: TARGET_LIB_INT_CMP_BIASED 28552 This macro should evaluate to `true' if the integer comparison 28553 functions (like `__cmpdi2') return 0 to indicate that the first 28554 operand is smaller than the second, 1 to indicate that they are 28555 equal, and 2 to indicate that the first operand is greater than 28556 the second. If this macro evaluates to `false' the comparison 28557 functions return -1, 0, and 1 instead of 0, 1, and 2. If the 28558 target uses the routines in `libgcc.a', you do not need to define 28559 this macro. 28560 28561 -- Macro: US_SOFTWARE_GOFAST 28562 Define this macro if your system C library uses the US Software 28563 GOFAST library to provide floating point emulation. 28564 28565 In addition to defining this macro, your architecture must set 28566 `TARGET_INIT_LIBFUNCS' to `gofast_maybe_init_libfuncs', or else 28567 call that function from its version of that hook. It is defined 28568 in `config/gofast.h', which must be included by your 28569 architecture's `CPU.c' file. See `sparc/sparc.c' for an example. 28570 28571 If this macro is defined, the 28572 `TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL' target hook must return 28573 false for `SFmode' and `DFmode' comparisons. 28574 28575 -- Macro: TARGET_EDOM 28576 The value of `EDOM' on the target machine, as a C integer constant 28577 expression. If you don't define this macro, GCC does not attempt 28578 to deposit the value of `EDOM' into `errno' directly. Look in 28579 `/usr/include/errno.h' to find the value of `EDOM' on your system. 28580 28581 If you do not define `TARGET_EDOM', then compiled code reports 28582 domain errors by calling the library function and letting it 28583 report the error. If mathematical functions on your system use 28584 `matherr' when there is an error, then you should leave 28585 `TARGET_EDOM' undefined so that `matherr' is used normally. 28586 28587 -- Macro: GEN_ERRNO_RTX 28588 Define this macro as a C expression to create an rtl expression 28589 that refers to the global "variable" `errno'. (On certain systems, 28590 `errno' may not actually be a variable.) If you don't define this 28591 macro, a reasonable default is used. 28592 28593 -- Macro: TARGET_C99_FUNCTIONS 28594 When this macro is nonzero, GCC will implicitly optimize `sin' 28595 calls into `sinf' and similarly for other functions defined by C99 28596 standard. The default is zero because a number of existing 28597 systems lack support for these functions in their runtime so this 28598 macro needs to be redefined to one on systems that do support the 28599 C99 runtime. 28600 28601 -- Macro: TARGET_HAS_SINCOS 28602 When this macro is nonzero, GCC will implicitly optimize calls to 28603 `sin' and `cos' with the same argument to a call to `sincos'. The 28604 default is zero. The target has to provide the following 28605 functions: 28606 void sincos(double x, double *sin, double *cos); 28607 void sincosf(float x, float *sin, float *cos); 28608 void sincosl(long double x, long double *sin, long double *cos); 28609 28610 -- Macro: NEXT_OBJC_RUNTIME 28611 Define this macro to generate code for Objective-C message sending 28612 using the calling convention of the NeXT system. This calling 28613 convention involves passing the object, the selector and the 28614 method arguments all at once to the method-lookup library function. 28615 28616 The default calling convention passes just the object and the 28617 selector to the lookup function, which returns a pointer to the 28618 method. 28619 28620 28621 File: gccint.info, Node: Addressing Modes, Next: Anchored Addresses, Prev: Library Calls, Up: Target Macros 28622 28623 17.14 Addressing Modes 28624 ====================== 28625 28626 This is about addressing modes. 28627 28628 -- Macro: HAVE_PRE_INCREMENT 28629 -- Macro: HAVE_PRE_DECREMENT 28630 -- Macro: HAVE_POST_INCREMENT 28631 -- Macro: HAVE_POST_DECREMENT 28632 A C expression that is nonzero if the machine supports 28633 pre-increment, pre-decrement, post-increment, or post-decrement 28634 addressing respectively. 28635 28636 -- Macro: HAVE_PRE_MODIFY_DISP 28637 -- Macro: HAVE_POST_MODIFY_DISP 28638 A C expression that is nonzero if the machine supports pre- or 28639 post-address side-effect generation involving constants other than 28640 the size of the memory operand. 28641 28642 -- Macro: HAVE_PRE_MODIFY_REG 28643 -- Macro: HAVE_POST_MODIFY_REG 28644 A C expression that is nonzero if the machine supports pre- or 28645 post-address side-effect generation involving a register 28646 displacement. 28647 28648 -- Macro: CONSTANT_ADDRESS_P (X) 28649 A C expression that is 1 if the RTX X is a constant which is a 28650 valid address. On most machines, this can be defined as 28651 `CONSTANT_P (X)', but a few machines are more restrictive in which 28652 constant addresses are supported. 28653 28654 -- Macro: CONSTANT_P (X) 28655 `CONSTANT_P', which is defined by target-independent code, accepts 28656 integer-values expressions whose values are not explicitly known, 28657 such as `symbol_ref', `label_ref', and `high' expressions and 28658 `const' arithmetic expressions, in addition to `const_int' and 28659 `const_double' expressions. 28660 28661 -- Macro: MAX_REGS_PER_ADDRESS 28662 A number, the maximum number of registers that can appear in a 28663 valid memory address. Note that it is up to you to specify a 28664 value equal to the maximum number that `GO_IF_LEGITIMATE_ADDRESS' 28665 would ever accept. 28666 28667 -- Macro: GO_IF_LEGITIMATE_ADDRESS (MODE, X, LABEL) 28668 A C compound statement with a conditional `goto LABEL;' executed 28669 if X (an RTX) is a legitimate memory address on the target machine 28670 for a memory operand of mode MODE. 28671 28672 It usually pays to define several simpler macros to serve as 28673 subroutines for this one. Otherwise it may be too complicated to 28674 understand. 28675 28676 This macro must exist in two variants: a strict variant and a 28677 non-strict one. The strict variant is used in the reload pass. It 28678 must be defined so that any pseudo-register that has not been 28679 allocated a hard register is considered a memory reference. In 28680 contexts where some kind of register is required, a pseudo-register 28681 with no hard register must be rejected. 28682 28683 The non-strict variant is used in other passes. It must be 28684 defined to accept all pseudo-registers in every context where some 28685 kind of register is required. 28686 28687 Compiler source files that want to use the strict variant of this 28688 macro define the macro `REG_OK_STRICT'. You should use an `#ifdef 28689 REG_OK_STRICT' conditional to define the strict variant in that 28690 case and the non-strict variant otherwise. 28691 28692 Subroutines to check for acceptable registers for various purposes 28693 (one for base registers, one for index registers, and so on) are 28694 typically among the subroutines used to define 28695 `GO_IF_LEGITIMATE_ADDRESS'. Then only these subroutine macros 28696 need have two variants; the higher levels of macros may be the 28697 same whether strict or not. 28698 28699 Normally, constant addresses which are the sum of a `symbol_ref' 28700 and an integer are stored inside a `const' RTX to mark them as 28701 constant. Therefore, there is no need to recognize such sums 28702 specifically as legitimate addresses. Normally you would simply 28703 recognize any `const' as legitimate. 28704 28705 Usually `PRINT_OPERAND_ADDRESS' is not prepared to handle constant 28706 sums that are not marked with `const'. It assumes that a naked 28707 `plus' indicates indexing. If so, then you _must_ reject such 28708 naked constant sums as illegitimate addresses, so that none of 28709 them will be given to `PRINT_OPERAND_ADDRESS'. 28710 28711 On some machines, whether a symbolic address is legitimate depends 28712 on the section that the address refers to. On these machines, 28713 define the target hook `TARGET_ENCODE_SECTION_INFO' to store the 28714 information into the `symbol_ref', and then check for it here. 28715 When you see a `const', you will have to look inside it to find the 28716 `symbol_ref' in order to determine the section. *Note Assembler 28717 Format::. 28718 28719 -- Macro: TARGET_MEM_CONSTRAINT 28720 A single character to be used instead of the default `'m'' 28721 character for general memory addresses. This defines the 28722 constraint letter which matches the memory addresses accepted by 28723 `GO_IF_LEGITIMATE_ADDRESS_P'. Define this macro if you want to 28724 support new address formats in your back end without changing the 28725 semantics of the `'m'' constraint. This is necessary in order to 28726 preserve functionality of inline assembly constructs using the 28727 `'m'' constraint. 28728 28729 -- Macro: FIND_BASE_TERM (X) 28730 A C expression to determine the base term of address X, or to 28731 provide a simplified version of X from which `alias.c' can easily 28732 find the base term. This macro is used in only two places: 28733 `find_base_value' and `find_base_term' in `alias.c'. 28734 28735 It is always safe for this macro to not be defined. It exists so 28736 that alias analysis can understand machine-dependent addresses. 28737 28738 The typical use of this macro is to handle addresses containing a 28739 label_ref or symbol_ref within an UNSPEC. 28740 28741 -- Macro: LEGITIMIZE_ADDRESS (X, OLDX, MODE, WIN) 28742 A C compound statement that attempts to replace X with a valid 28743 memory address for an operand of mode MODE. WIN will be a C 28744 statement label elsewhere in the code; the macro definition may use 28745 28746 GO_IF_LEGITIMATE_ADDRESS (MODE, X, WIN); 28747 28748 to avoid further processing if the address has become legitimate. 28749 28750 X will always be the result of a call to `break_out_memory_refs', 28751 and OLDX will be the operand that was given to that function to 28752 produce X. 28753 28754 The code generated by this macro should not alter the substructure 28755 of X. If it transforms X into a more legitimate form, it should 28756 assign X (which will always be a C variable) a new value. 28757 28758 It is not necessary for this macro to come up with a legitimate 28759 address. The compiler has standard ways of doing so in all cases. 28760 In fact, it is safe to omit this macro. But often a 28761 machine-dependent strategy can generate better code. 28762 28763 -- Macro: LEGITIMIZE_RELOAD_ADDRESS (X, MODE, OPNUM, TYPE, IND_LEVELS, 28764 WIN) 28765 A C compound statement that attempts to replace X, which is an 28766 address that needs reloading, with a valid memory address for an 28767 operand of mode MODE. WIN will be a C statement label elsewhere 28768 in the code. It is not necessary to define this macro, but it 28769 might be useful for performance reasons. 28770 28771 For example, on the i386, it is sometimes possible to use a single 28772 reload register instead of two by reloading a sum of two pseudo 28773 registers into a register. On the other hand, for number of RISC 28774 processors offsets are limited so that often an intermediate 28775 address needs to be generated in order to address a stack slot. 28776 By defining `LEGITIMIZE_RELOAD_ADDRESS' appropriately, the 28777 intermediate addresses generated for adjacent some stack slots can 28778 be made identical, and thus be shared. 28779 28780 _Note_: This macro should be used with caution. It is necessary 28781 to know something of how reload works in order to effectively use 28782 this, and it is quite easy to produce macros that build in too 28783 much knowledge of reload internals. 28784 28785 _Note_: This macro must be able to reload an address created by a 28786 previous invocation of this macro. If it fails to handle such 28787 addresses then the compiler may generate incorrect code or abort. 28788 28789 The macro definition should use `push_reload' to indicate parts 28790 that need reloading; OPNUM, TYPE and IND_LEVELS are usually 28791 suitable to be passed unaltered to `push_reload'. 28792 28793 The code generated by this macro must not alter the substructure of 28794 X. If it transforms X into a more legitimate form, it should 28795 assign X (which will always be a C variable) a new value. This 28796 also applies to parts that you change indirectly by calling 28797 `push_reload'. 28798 28799 The macro definition may use `strict_memory_address_p' to test if 28800 the address has become legitimate. 28801 28802 If you want to change only a part of X, one standard way of doing 28803 this is to use `copy_rtx'. Note, however, that it unshares only a 28804 single level of rtl. Thus, if the part to be changed is not at the 28805 top level, you'll need to replace first the top level. It is not 28806 necessary for this macro to come up with a legitimate address; 28807 but often a machine-dependent strategy can generate better code. 28808 28809 -- Macro: GO_IF_MODE_DEPENDENT_ADDRESS (ADDR, LABEL) 28810 A C statement or compound statement with a conditional `goto 28811 LABEL;' executed if memory address X (an RTX) can have different 28812 meanings depending on the machine mode of the memory reference it 28813 is used for or if the address is valid for some modes but not 28814 others. 28815 28816 Autoincrement and autodecrement addresses typically have 28817 mode-dependent effects because the amount of the increment or 28818 decrement is the size of the operand being addressed. Some 28819 machines have other mode-dependent addresses. Many RISC machines 28820 have no mode-dependent addresses. 28821 28822 You may assume that ADDR is a valid address for the machine. 28823 28824 -- Macro: LEGITIMATE_CONSTANT_P (X) 28825 A C expression that is nonzero if X is a legitimate constant for 28826 an immediate operand on the target machine. You can assume that X 28827 satisfies `CONSTANT_P', so you need not check this. In fact, `1' 28828 is a suitable definition for this macro on machines where anything 28829 `CONSTANT_P' is valid. 28830 28831 -- Target Hook: rtx TARGET_DELEGITIMIZE_ADDRESS (rtx X) 28832 This hook is used to undo the possibly obfuscating effects of the 28833 `LEGITIMIZE_ADDRESS' and `LEGITIMIZE_RELOAD_ADDRESS' target 28834 macros. Some backend implementations of these macros wrap symbol 28835 references inside an `UNSPEC' rtx to represent PIC or similar 28836 addressing modes. This target hook allows GCC's optimizers to 28837 understand the semantics of these opaque `UNSPEC's by converting 28838 them back into their original form. 28839 28840 -- Target Hook: bool TARGET_CANNOT_FORCE_CONST_MEM (rtx X) 28841 This hook should return true if X is of a form that cannot (or 28842 should not) be spilled to the constant pool. The default version 28843 of this hook returns false. 28844 28845 The primary reason to define this hook is to prevent reload from 28846 deciding that a non-legitimate constant would be better reloaded 28847 from the constant pool instead of spilling and reloading a register 28848 holding the constant. This restriction is often true of addresses 28849 of TLS symbols for various targets. 28850 28851 -- Target Hook: bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum 28852 machine_mode MODE, rtx X) 28853 This hook should return true if pool entries for constant X can be 28854 placed in an `object_block' structure. MODE is the mode of X. 28855 28856 The default version returns false for all constants. 28857 28858 -- Target Hook: tree TARGET_BUILTIN_RECIPROCAL (enum tree_code FN, 28859 bool TM_FN, bool SQRT) 28860 This hook should return the DECL of a function that implements 28861 reciprocal of the builtin function with builtin function code FN, 28862 or `NULL_TREE' if such a function is not available. TM_FN is true 28863 when FN is a code of a machine-dependent builtin function. When 28864 SQRT is true, additional optimizations that apply only to the 28865 reciprocal of a square root function are performed, and only 28866 reciprocals of `sqrt' function are valid. 28867 28868 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void) 28869 This hook should return the DECL of a function F that given an 28870 address ADDR as an argument returns a mask M that can be used to 28871 extract from two vectors the relevant data that resides in ADDR in 28872 case ADDR is not properly aligned. 28873 28874 The autovectorizer, when vectorizing a load operation from an 28875 address ADDR that may be unaligned, will generate two vector loads 28876 from the two aligned addresses around ADDR. It then generates a 28877 `REALIGN_LOAD' operation to extract the relevant data from the two 28878 loaded vectors. The first two arguments to `REALIGN_LOAD', V1 and 28879 V2, are the two vectors, each of size VS, and the third argument, 28880 OFF, defines how the data will be extracted from these two 28881 vectors: if OFF is 0, then the returned vector is V2; otherwise, 28882 the returned vector is composed from the last VS-OFF elements of 28883 V1 concatenated to the first OFF elements of V2. 28884 28885 If this hook is defined, the autovectorizer will generate a call 28886 to F (using the DECL tree that this hook returns) and will use the 28887 return value of F as the argument OFF to `REALIGN_LOAD'. 28888 Therefore, the mask M returned by F should comply with the 28889 semantics expected by `REALIGN_LOAD' described above. If this 28890 hook is not defined, then ADDR will be used as the argument OFF to 28891 `REALIGN_LOAD', in which case the low log2(VS)-1 bits of ADDR will 28892 be considered. 28893 28894 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree X) 28895 This hook should return the DECL of a function F that implements 28896 widening multiplication of the even elements of two input vectors 28897 of type X. 28898 28899 If this hook is defined, the autovectorizer will use it along with 28900 the `TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD' target hook when 28901 vectorizing widening multiplication in cases that the order of the 28902 results does not have to be preserved (e.g. used only by a 28903 reduction computation). Otherwise, the `widen_mult_hi/lo' idioms 28904 will be used. 28905 28906 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree X) 28907 This hook should return the DECL of a function F that implements 28908 widening multiplication of the odd elements of two input vectors 28909 of type X. 28910 28911 If this hook is defined, the autovectorizer will use it along with 28912 the `TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN' target hook when 28913 vectorizing widening multiplication in cases that the order of the 28914 results does not have to be preserved (e.g. used only by a 28915 reduction computation). Otherwise, the `widen_mult_hi/lo' idioms 28916 will be used. 28917 28918 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum 28919 tree_code CODE, tree TYPE) 28920 This hook should return the DECL of a function that implements 28921 conversion of the input vector of type TYPE. If TYPE is an 28922 integral type, the result of the conversion is a vector of 28923 floating-point type of the same size. If TYPE is a floating-point 28924 type, the result of the conversion is a vector of integral type of 28925 the same size. CODE specifies how the conversion is to be applied 28926 (truncation, rounding, etc.). 28927 28928 If this hook is defined, the autovectorizer will use the 28929 `TARGET_VECTORIZE_BUILTIN_CONVERSION' target hook when vectorizing 28930 conversion. Otherwise, it will return `NULL_TREE'. 28931 28932 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION 28933 (enum built_in_function CODE, tree VEC_TYPE_OUT, tree 28934 VEC_TYPE_IN) 28935 This hook should return the decl of a function that implements the 28936 vectorized variant of the builtin function with builtin function 28937 code CODE or `NULL_TREE' if such a function is not available. The 28938 return type of the vectorized function shall be of vector type 28939 VEC_TYPE_OUT and the argument types should be VEC_TYPE_IN. 28940 28941 28942 File: gccint.info, Node: Anchored Addresses, Next: Condition Code, Prev: Addressing Modes, Up: Target Macros 28943 28944 17.15 Anchored Addresses 28945 ======================== 28946 28947 GCC usually addresses every static object as a separate entity. For 28948 example, if we have: 28949 28950 static int a, b, c; 28951 int foo (void) { return a + b + c; } 28952 28953 the code for `foo' will usually calculate three separate symbolic 28954 addresses: those of `a', `b' and `c'. On some targets, it would be 28955 better to calculate just one symbolic address and access the three 28956 variables relative to it. The equivalent pseudocode would be something 28957 like: 28958 28959 int foo (void) 28960 { 28961 register int *xr = &x; 28962 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 28963 } 28964 28965 (which isn't valid C). We refer to shared addresses like `x' as 28966 "section anchors". Their use is controlled by `-fsection-anchors'. 28967 28968 The hooks below describe the target properties that GCC needs to know 28969 in order to make effective use of section anchors. It won't use 28970 section anchors at all unless either `TARGET_MIN_ANCHOR_OFFSET' or 28971 `TARGET_MAX_ANCHOR_OFFSET' is set to a nonzero value. 28972 28973 -- Variable: Target Hook HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET 28974 The minimum offset that should be applied to a section anchor. On 28975 most targets, it should be the smallest offset that can be applied 28976 to a base register while still giving a legitimate address for 28977 every mode. The default value is 0. 28978 28979 -- Variable: Target Hook HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET 28980 Like `TARGET_MIN_ANCHOR_OFFSET', but the maximum (inclusive) 28981 offset that should be applied to section anchors. The default 28982 value is 0. 28983 28984 -- Target Hook: void TARGET_ASM_OUTPUT_ANCHOR (rtx X) 28985 Write the assembly code to define section anchor X, which is a 28986 `SYMBOL_REF' for which `SYMBOL_REF_ANCHOR_P (X)' is true. The 28987 hook is called with the assembly output position set to the 28988 beginning of `SYMBOL_REF_BLOCK (X)'. 28989 28990 If `ASM_OUTPUT_DEF' is available, the hook's default definition 28991 uses it to define the symbol as `. + SYMBOL_REF_BLOCK_OFFSET (X)'. 28992 If `ASM_OUTPUT_DEF' is not available, the hook's default definition 28993 is `NULL', which disables the use of section anchors altogether. 28994 28995 -- Target Hook: bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx X) 28996 Return true if GCC should attempt to use anchors to access 28997 `SYMBOL_REF' X. You can assume `SYMBOL_REF_HAS_BLOCK_INFO_P (X)' 28998 and `!SYMBOL_REF_ANCHOR_P (X)'. 28999 29000 The default version is correct for most targets, but you might 29001 need to intercept this hook to handle things like target-specific 29002 attributes or target-specific sections. 29003 29004 29005 File: gccint.info, Node: Condition Code, Next: Costs, Prev: Anchored Addresses, Up: Target Macros 29006 29007 17.16 Condition Code Status 29008 =========================== 29009 29010 This describes the condition code status. 29011 29012 The file `conditions.h' defines a variable `cc_status' to describe how 29013 the condition code was computed (in case the interpretation of the 29014 condition code depends on the instruction that it was set by). This 29015 variable contains the RTL expressions on which the condition code is 29016 currently based, and several standard flags. 29017 29018 Sometimes additional machine-specific flags must be defined in the 29019 machine description header file. It can also add additional 29020 machine-specific information by defining `CC_STATUS_MDEP'. 29021 29022 -- Macro: CC_STATUS_MDEP 29023 C code for a data type which is used for declaring the `mdep' 29024 component of `cc_status'. It defaults to `int'. 29025 29026 This macro is not used on machines that do not use `cc0'. 29027 29028 -- Macro: CC_STATUS_MDEP_INIT 29029 A C expression to initialize the `mdep' field to "empty". The 29030 default definition does nothing, since most machines don't use the 29031 field anyway. If you want to use the field, you should probably 29032 define this macro to initialize it. 29033 29034 This macro is not used on machines that do not use `cc0'. 29035 29036 -- Macro: NOTICE_UPDATE_CC (EXP, INSN) 29037 A C compound statement to set the components of `cc_status' 29038 appropriately for an insn INSN whose body is EXP. It is this 29039 macro's responsibility to recognize insns that set the condition 29040 code as a byproduct of other activity as well as those that 29041 explicitly set `(cc0)'. 29042 29043 This macro is not used on machines that do not use `cc0'. 29044 29045 If there are insns that do not set the condition code but do alter 29046 other machine registers, this macro must check to see whether they 29047 invalidate the expressions that the condition code is recorded as 29048 reflecting. For example, on the 68000, insns that store in address 29049 registers do not set the condition code, which means that usually 29050 `NOTICE_UPDATE_CC' can leave `cc_status' unaltered for such insns. 29051 But suppose that the previous insn set the condition code based 29052 on location `a4@(102)' and the current insn stores a new value in 29053 `a4'. Although the condition code is not changed by this, it will 29054 no longer be true that it reflects the contents of `a4@(102)'. 29055 Therefore, `NOTICE_UPDATE_CC' must alter `cc_status' in this case 29056 to say that nothing is known about the condition code value. 29057 29058 The definition of `NOTICE_UPDATE_CC' must be prepared to deal with 29059 the results of peephole optimization: insns whose patterns are 29060 `parallel' RTXs containing various `reg', `mem' or constants which 29061 are just the operands. The RTL structure of these insns is not 29062 sufficient to indicate what the insns actually do. What 29063 `NOTICE_UPDATE_CC' should do when it sees one is just to run 29064 `CC_STATUS_INIT'. 29065 29066 A possible definition of `NOTICE_UPDATE_CC' is to call a function 29067 that looks at an attribute (*note Insn Attributes::) named, for 29068 example, `cc'. This avoids having detailed information about 29069 patterns in two places, the `md' file and in `NOTICE_UPDATE_CC'. 29070 29071 -- Macro: SELECT_CC_MODE (OP, X, Y) 29072 Returns a mode from class `MODE_CC' to be used when comparison 29073 operation code OP is applied to rtx X and Y. For example, on the 29074 SPARC, `SELECT_CC_MODE' is defined as (see *note Jump Patterns:: 29075 for a description of the reason for this definition) 29076 29077 #define SELECT_CC_MODE(OP,X,Y) \ 29078 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ 29079 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \ 29080 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ 29081 || GET_CODE (X) == NEG) \ 29082 ? CC_NOOVmode : CCmode)) 29083 29084 You should define this macro if and only if you define extra CC 29085 modes in `MACHINE-modes.def'. 29086 29087 -- Macro: CANONICALIZE_COMPARISON (CODE, OP0, OP1) 29088 On some machines not all possible comparisons are defined, but you 29089 can convert an invalid comparison into a valid one. For example, 29090 the Alpha does not have a `GT' comparison, but you can use an `LT' 29091 comparison instead and swap the order of the operands. 29092 29093 On such machines, define this macro to be a C statement to do any 29094 required conversions. CODE is the initial comparison code and OP0 29095 and OP1 are the left and right operands of the comparison, 29096 respectively. You should modify CODE, OP0, and OP1 as required. 29097 29098 GCC will not assume that the comparison resulting from this macro 29099 is valid but will see if the resulting insn matches a pattern in 29100 the `md' file. 29101 29102 You need not define this macro if it would never change the 29103 comparison code or operands. 29104 29105 -- Macro: REVERSIBLE_CC_MODE (MODE) 29106 A C expression whose value is one if it is always safe to reverse a 29107 comparison whose mode is MODE. If `SELECT_CC_MODE' can ever 29108 return MODE for a floating-point inequality comparison, then 29109 `REVERSIBLE_CC_MODE (MODE)' must be zero. 29110 29111 You need not define this macro if it would always returns zero or 29112 if the floating-point format is anything other than 29113 `IEEE_FLOAT_FORMAT'. For example, here is the definition used on 29114 the SPARC, where floating-point inequality comparisons are always 29115 given `CCFPEmode': 29116 29117 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode) 29118 29119 -- Macro: REVERSE_CONDITION (CODE, MODE) 29120 A C expression whose value is reversed condition code of the CODE 29121 for comparison done in CC_MODE MODE. The macro is used only in 29122 case `REVERSIBLE_CC_MODE (MODE)' is nonzero. Define this macro in 29123 case machine has some non-standard way how to reverse certain 29124 conditionals. For instance in case all floating point conditions 29125 are non-trapping, compiler may freely convert unordered compares 29126 to ordered one. Then definition may look like: 29127 29128 #define REVERSE_CONDITION(CODE, MODE) \ 29129 ((MODE) != CCFPmode ? reverse_condition (CODE) \ 29130 : reverse_condition_maybe_unordered (CODE)) 29131 29132 -- Macro: REVERSE_CONDEXEC_PREDICATES_P (OP1, OP2) 29133 A C expression that returns true if the conditional execution 29134 predicate OP1, a comparison operation, is the inverse of OP2 and 29135 vice versa. Define this to return 0 if the target has conditional 29136 execution predicates that cannot be reversed safely. There is no 29137 need to validate that the arguments of op1 and op2 are the same, 29138 this is done separately. If no expansion is specified, this macro 29139 is defined as follows: 29140 29141 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \ 29142 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL)) 29143 29144 -- Target Hook: bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, 29145 unsigned int *) 29146 On targets which do not use `(cc0)', and which use a hard register 29147 rather than a pseudo-register to hold condition codes, the regular 29148 CSE passes are often not able to identify cases in which the hard 29149 register is set to a common value. Use this hook to enable a 29150 small pass which optimizes such cases. This hook should return 29151 true to enable this pass, and it should set the integers to which 29152 its arguments point to the hard register numbers used for 29153 condition codes. When there is only one such register, as is true 29154 on most systems, the integer pointed to by the second argument 29155 should be set to `INVALID_REGNUM'. 29156 29157 The default version of this hook returns false. 29158 29159 -- Target Hook: enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum 29160 machine_mode, enum machine_mode) 29161 On targets which use multiple condition code modes in class 29162 `MODE_CC', it is sometimes the case that a comparison can be 29163 validly done in more than one mode. On such a system, define this 29164 target hook to take two mode arguments and to return a mode in 29165 which both comparisons may be validly done. If there is no such 29166 mode, return `VOIDmode'. 29167 29168 The default version of this hook checks whether the modes are the 29169 same. If they are, it returns that mode. If they are different, 29170 it returns `VOIDmode'. 29171 29172 29173 File: gccint.info, Node: Costs, Next: Scheduling, Prev: Condition Code, Up: Target Macros 29174 29175 17.17 Describing Relative Costs of Operations 29176 ============================================= 29177 29178 These macros let you describe the relative speed of various operations 29179 on the target machine. 29180 29181 -- Macro: REGISTER_MOVE_COST (MODE, FROM, TO) 29182 A C expression for the cost of moving data of mode MODE from a 29183 register in class FROM to one in class TO. The classes are 29184 expressed using the enumeration values such as `GENERAL_REGS'. A 29185 value of 2 is the default; other values are interpreted relative to 29186 that. 29187 29188 It is not required that the cost always equal 2 when FROM is the 29189 same as TO; on some machines it is expensive to move between 29190 registers if they are not general registers. 29191 29192 If reload sees an insn consisting of a single `set' between two 29193 hard registers, and if `REGISTER_MOVE_COST' applied to their 29194 classes returns a value of 2, reload does not check to ensure that 29195 the constraints of the insn are met. Setting a cost of other than 29196 2 will allow reload to verify that the constraints are met. You 29197 should do this if the `movM' pattern's constraints do not allow 29198 such copying. 29199 29200 -- Macro: MEMORY_MOVE_COST (MODE, CLASS, IN) 29201 A C expression for the cost of moving data of mode MODE between a 29202 register of class CLASS and memory; IN is zero if the value is to 29203 be written to memory, nonzero if it is to be read in. This cost 29204 is relative to those in `REGISTER_MOVE_COST'. If moving between 29205 registers and memory is more expensive than between two registers, 29206 you should define this macro to express the relative cost. 29207 29208 If you do not define this macro, GCC uses a default cost of 4 plus 29209 the cost of copying via a secondary reload register, if one is 29210 needed. If your machine requires a secondary reload register to 29211 copy between memory and a register of CLASS but the reload 29212 mechanism is more complex than copying via an intermediate, define 29213 this macro to reflect the actual cost of the move. 29214 29215 GCC defines the function `memory_move_secondary_cost' if secondary 29216 reloads are needed. It computes the costs due to copying via a 29217 secondary register. If your machine copies from memory using a 29218 secondary register in the conventional way but the default base 29219 value of 4 is not correct for your machine, define this macro to 29220 add some other value to the result of that function. The 29221 arguments to that function are the same as to this macro. 29222 29223 -- Macro: BRANCH_COST (SPEED_P, PREDICTABLE_P) 29224 A C expression for the cost of a branch instruction. A value of 1 29225 is the default; other values are interpreted relative to that. 29226 Parameter SPEED_P is true when the branch in question should be 29227 optimized for speed. When it is false, `BRANCH_COST' should be 29228 returning value optimal for code size rather then performance 29229 considerations. PREDICTABLE_P is true for well predictable 29230 branches. On many architectures the `BRANCH_COST' can be reduced 29231 then. 29232 29233 Here are additional macros which do not specify precise relative costs, 29234 but only that certain actions are more expensive than GCC would 29235 ordinarily expect. 29236 29237 -- Macro: SLOW_BYTE_ACCESS 29238 Define this macro as a C expression which is nonzero if accessing 29239 less than a word of memory (i.e. a `char' or a `short') is no 29240 faster than accessing a word of memory, i.e., if such access 29241 require more than one instruction or if there is no difference in 29242 cost between byte and (aligned) word loads. 29243 29244 When this macro is not defined, the compiler will access a field by 29245 finding the smallest containing object; when it is defined, a 29246 fullword load will be used if alignment permits. Unless bytes 29247 accesses are faster than word accesses, using word accesses is 29248 preferable since it may eliminate subsequent memory access if 29249 subsequent accesses occur to other fields in the same word of the 29250 structure, but to different bytes. 29251 29252 -- Macro: SLOW_UNALIGNED_ACCESS (MODE, ALIGNMENT) 29253 Define this macro to be the value 1 if memory accesses described 29254 by the MODE and ALIGNMENT parameters have a cost many times greater 29255 than aligned accesses, for example if they are emulated in a trap 29256 handler. 29257 29258 When this macro is nonzero, the compiler will act as if 29259 `STRICT_ALIGNMENT' were nonzero when generating code for block 29260 moves. This can cause significantly more instructions to be 29261 produced. Therefore, do not set this macro nonzero if unaligned 29262 accesses only add a cycle or two to the time for a memory access. 29263 29264 If the value of this macro is always zero, it need not be defined. 29265 If this macro is defined, it should produce a nonzero value when 29266 `STRICT_ALIGNMENT' is nonzero. 29267 29268 -- Macro: MOVE_RATIO 29269 The threshold of number of scalar memory-to-memory move insns, 29270 _below_ which a sequence of insns should be generated instead of a 29271 string move insn or a library call. Increasing the value will 29272 always make code faster, but eventually incurs high cost in 29273 increased code size. 29274 29275 Note that on machines where the corresponding move insn is a 29276 `define_expand' that emits a sequence of insns, this macro counts 29277 the number of such sequences. 29278 29279 If you don't define this, a reasonable default is used. 29280 29281 -- Macro: MOVE_BY_PIECES_P (SIZE, ALIGNMENT) 29282 A C expression used to determine whether `move_by_pieces' will be 29283 used to copy a chunk of memory, or whether some other block move 29284 mechanism will be used. Defaults to 1 if `move_by_pieces_ninsns' 29285 returns less than `MOVE_RATIO'. 29286 29287 -- Macro: MOVE_MAX_PIECES 29288 A C expression used by `move_by_pieces' to determine the largest 29289 unit a load or store used to copy memory is. Defaults to 29290 `MOVE_MAX'. 29291 29292 -- Macro: CLEAR_RATIO 29293 The threshold of number of scalar move insns, _below_ which a 29294 sequence of insns should be generated to clear memory instead of a 29295 string clear insn or a library call. Increasing the value will 29296 always make code faster, but eventually incurs high cost in 29297 increased code size. 29298 29299 If you don't define this, a reasonable default is used. 29300 29301 -- Macro: CLEAR_BY_PIECES_P (SIZE, ALIGNMENT) 29302 A C expression used to determine whether `clear_by_pieces' will be 29303 used to clear a chunk of memory, or whether some other block clear 29304 mechanism will be used. Defaults to 1 if `move_by_pieces_ninsns' 29305 returns less than `CLEAR_RATIO'. 29306 29307 -- Macro: SET_RATIO 29308 The threshold of number of scalar move insns, _below_ which a 29309 sequence of insns should be generated to set memory to a constant 29310 value, instead of a block set insn or a library call. Increasing 29311 the value will always make code faster, but eventually incurs high 29312 cost in increased code size. 29313 29314 If you don't define this, it defaults to the value of `MOVE_RATIO'. 29315 29316 -- Macro: SET_BY_PIECES_P (SIZE, ALIGNMENT) 29317 A C expression used to determine whether `store_by_pieces' will be 29318 used to set a chunk of memory to a constant value, or whether some 29319 other mechanism will be used. Used by `__builtin_memset' when 29320 storing values other than constant zero. Defaults to 1 if 29321 `move_by_pieces_ninsns' returns less than `SET_RATIO'. 29322 29323 -- Macro: STORE_BY_PIECES_P (SIZE, ALIGNMENT) 29324 A C expression used to determine whether `store_by_pieces' will be 29325 used to set a chunk of memory to a constant string value, or 29326 whether some other mechanism will be used. Used by 29327 `__builtin_strcpy' when called with a constant source string. 29328 Defaults to 1 if `move_by_pieces_ninsns' returns less than 29329 `MOVE_RATIO'. 29330 29331 -- Macro: USE_LOAD_POST_INCREMENT (MODE) 29332 A C expression used to determine whether a load postincrement is a 29333 good thing to use for a given mode. Defaults to the value of 29334 `HAVE_POST_INCREMENT'. 29335 29336 -- Macro: USE_LOAD_POST_DECREMENT (MODE) 29337 A C expression used to determine whether a load postdecrement is a 29338 good thing to use for a given mode. Defaults to the value of 29339 `HAVE_POST_DECREMENT'. 29340 29341 -- Macro: USE_LOAD_PRE_INCREMENT (MODE) 29342 A C expression used to determine whether a load preincrement is a 29343 good thing to use for a given mode. Defaults to the value of 29344 `HAVE_PRE_INCREMENT'. 29345 29346 -- Macro: USE_LOAD_PRE_DECREMENT (MODE) 29347 A C expression used to determine whether a load predecrement is a 29348 good thing to use for a given mode. Defaults to the value of 29349 `HAVE_PRE_DECREMENT'. 29350 29351 -- Macro: USE_STORE_POST_INCREMENT (MODE) 29352 A C expression used to determine whether a store postincrement is 29353 a good thing to use for a given mode. Defaults to the value of 29354 `HAVE_POST_INCREMENT'. 29355 29356 -- Macro: USE_STORE_POST_DECREMENT (MODE) 29357 A C expression used to determine whether a store postdecrement is 29358 a good thing to use for a given mode. Defaults to the value of 29359 `HAVE_POST_DECREMENT'. 29360 29361 -- Macro: USE_STORE_PRE_INCREMENT (MODE) 29362 This macro is used to determine whether a store preincrement is a 29363 good thing to use for a given mode. Defaults to the value of 29364 `HAVE_PRE_INCREMENT'. 29365 29366 -- Macro: USE_STORE_PRE_DECREMENT (MODE) 29367 This macro is used to determine whether a store predecrement is a 29368 good thing to use for a given mode. Defaults to the value of 29369 `HAVE_PRE_DECREMENT'. 29370 29371 -- Macro: NO_FUNCTION_CSE 29372 Define this macro if it is as good or better to call a constant 29373 function address than to call an address kept in a register. 29374 29375 -- Macro: RANGE_TEST_NON_SHORT_CIRCUIT 29376 Define this macro if a non-short-circuit operation produced by 29377 `fold_range_test ()' is optimal. This macro defaults to true if 29378 `BRANCH_COST' is greater than or equal to the value 2. 29379 29380 -- Target Hook: bool TARGET_RTX_COSTS (rtx X, int CODE, int 29381 OUTER_CODE, int *TOTAL) 29382 This target hook describes the relative costs of RTL expressions. 29383 29384 The cost may depend on the precise form of the expression, which is 29385 available for examination in X, and the rtx code of the expression 29386 in which it is contained, found in OUTER_CODE. CODE is the 29387 expression code--redundant, since it can be obtained with 29388 `GET_CODE (X)'. 29389 29390 In implementing this hook, you can use the construct 29391 `COSTS_N_INSNS (N)' to specify a cost equal to N fast instructions. 29392 29393 On entry to the hook, `*TOTAL' contains a default estimate for the 29394 cost of the expression. The hook should modify this value as 29395 necessary. Traditionally, the default costs are `COSTS_N_INSNS 29396 (5)' for multiplications, `COSTS_N_INSNS (7)' for division and 29397 modulus operations, and `COSTS_N_INSNS (1)' for all other 29398 operations. 29399 29400 When optimizing for code size, i.e. when `optimize_size' is 29401 nonzero, this target hook should be used to estimate the relative 29402 size cost of an expression, again relative to `COSTS_N_INSNS'. 29403 29404 The hook returns true when all subexpressions of X have been 29405 processed, and false when `rtx_cost' should recurse. 29406 29407 -- Target Hook: int TARGET_ADDRESS_COST (rtx ADDRESS) 29408 This hook computes the cost of an addressing mode that contains 29409 ADDRESS. If not defined, the cost is computed from the ADDRESS 29410 expression and the `TARGET_RTX_COST' hook. 29411 29412 For most CISC machines, the default cost is a good approximation 29413 of the true cost of the addressing mode. However, on RISC 29414 machines, all instructions normally have the same length and 29415 execution time. Hence all addresses will have equal costs. 29416 29417 In cases where more than one form of an address is known, the form 29418 with the lowest cost will be used. If multiple forms have the 29419 same, lowest, cost, the one that is the most complex will be used. 29420 29421 For example, suppose an address that is equal to the sum of a 29422 register and a constant is used twice in the same basic block. 29423 When this macro is not defined, the address will be computed in a 29424 register and memory references will be indirect through that 29425 register. On machines where the cost of the addressing mode 29426 containing the sum is no higher than that of a simple indirect 29427 reference, this will produce an additional instruction and 29428 possibly require an additional register. Proper specification of 29429 this macro eliminates this overhead for such machines. 29430 29431 This hook is never called with an invalid address. 29432 29433 On machines where an address involving more than one register is as 29434 cheap as an address computation involving only one register, 29435 defining `TARGET_ADDRESS_COST' to reflect this can cause two 29436 registers to be live over a region of code where only one would 29437 have been if `TARGET_ADDRESS_COST' were not defined in that 29438 manner. This effect should be considered in the definition of 29439 this macro. Equivalent costs should probably only be given to 29440 addresses with different numbers of registers on machines with 29441 lots of registers. 29442 29443 29444 File: gccint.info, Node: Scheduling, Next: Sections, Prev: Costs, Up: Target Macros 29445 29446 17.18 Adjusting the Instruction Scheduler 29447 ========================================= 29448 29449 The instruction scheduler may need a fair amount of machine-specific 29450 adjustment in order to produce good code. GCC provides several target 29451 hooks for this purpose. It is usually enough to define just a few of 29452 them: try the first ones in this list first. 29453 29454 -- Target Hook: int TARGET_SCHED_ISSUE_RATE (void) 29455 This hook returns the maximum number of instructions that can ever 29456 issue at the same time on the target machine. The default is one. 29457 Although the insn scheduler can define itself the possibility of 29458 issue an insn on the same cycle, the value can serve as an 29459 additional constraint to issue insns on the same simulated 29460 processor cycle (see hooks `TARGET_SCHED_REORDER' and 29461 `TARGET_SCHED_REORDER2'). This value must be constant over the 29462 entire compilation. If you need it to vary depending on what the 29463 instructions are, you must use `TARGET_SCHED_VARIABLE_ISSUE'. 29464 29465 -- Target Hook: int TARGET_SCHED_VARIABLE_ISSUE (FILE *FILE, int 29466 VERBOSE, rtx INSN, int MORE) 29467 This hook is executed by the scheduler after it has scheduled an 29468 insn from the ready list. It should return the number of insns 29469 which can still be issued in the current cycle. The default is 29470 `MORE - 1' for insns other than `CLOBBER' and `USE', which 29471 normally are not counted against the issue rate. You should 29472 define this hook if some insns take more machine resources than 29473 others, so that fewer insns can follow them in the same cycle. 29474 FILE is either a null pointer, or a stdio stream to write any 29475 debug output to. VERBOSE is the verbose level provided by 29476 `-fsched-verbose-N'. INSN is the instruction that was scheduled. 29477 29478 -- Target Hook: int TARGET_SCHED_ADJUST_COST (rtx INSN, rtx LINK, rtx 29479 DEP_INSN, int COST) 29480 This function corrects the value of COST based on the relationship 29481 between INSN and DEP_INSN through the dependence LINK. It should 29482 return the new value. The default is to make no adjustment to 29483 COST. This can be used for example to specify to the scheduler 29484 using the traditional pipeline description that an output- or 29485 anti-dependence does not incur the same cost as a data-dependence. 29486 If the scheduler using the automaton based pipeline description, 29487 the cost of anti-dependence is zero and the cost of 29488 output-dependence is maximum of one and the difference of latency 29489 times of the first and the second insns. If these values are not 29490 acceptable, you could use the hook to modify them too. See also 29491 *note Processor pipeline description::. 29492 29493 -- Target Hook: int TARGET_SCHED_ADJUST_PRIORITY (rtx INSN, int 29494 PRIORITY) 29495 This hook adjusts the integer scheduling priority PRIORITY of 29496 INSN. It should return the new priority. Increase the priority to 29497 execute INSN earlier, reduce the priority to execute INSN later. 29498 Do not define this hook if you do not need to adjust the 29499 scheduling priorities of insns. 29500 29501 -- Target Hook: int TARGET_SCHED_REORDER (FILE *FILE, int VERBOSE, rtx 29502 *READY, int *N_READYP, int CLOCK) 29503 This hook is executed by the scheduler after it has scheduled the 29504 ready list, to allow the machine description to reorder it (for 29505 example to combine two small instructions together on `VLIW' 29506 machines). FILE is either a null pointer, or a stdio stream to 29507 write any debug output to. VERBOSE is the verbose level provided 29508 by `-fsched-verbose-N'. READY is a pointer to the ready list of 29509 instructions that are ready to be scheduled. N_READYP is a 29510 pointer to the number of elements in the ready list. The scheduler 29511 reads the ready list in reverse order, starting with 29512 READY[*N_READYP-1] and going to READY[0]. CLOCK is the timer tick 29513 of the scheduler. You may modify the ready list and the number of 29514 ready insns. The return value is the number of insns that can 29515 issue this cycle; normally this is just `issue_rate'. See also 29516 `TARGET_SCHED_REORDER2'. 29517 29518 -- Target Hook: int TARGET_SCHED_REORDER2 (FILE *FILE, int VERBOSE, 29519 rtx *READY, int *N_READY, CLOCK) 29520 Like `TARGET_SCHED_REORDER', but called at a different time. That 29521 function is called whenever the scheduler starts a new cycle. 29522 This one is called once per iteration over a cycle, immediately 29523 after `TARGET_SCHED_VARIABLE_ISSUE'; it can reorder the ready list 29524 and return the number of insns to be scheduled in the same cycle. 29525 Defining this hook can be useful if there are frequent situations 29526 where scheduling one insn causes other insns to become ready in 29527 the same cycle. These other insns can then be taken into account 29528 properly. 29529 29530 -- Target Hook: void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx 29531 HEAD, rtx TAIL) 29532 This hook is called after evaluation forward dependencies of insns 29533 in chain given by two parameter values (HEAD and TAIL 29534 correspondingly) but before insns scheduling of the insn chain. 29535 For example, it can be used for better insn classification if it 29536 requires analysis of dependencies. This hook can use backward and 29537 forward dependencies of the insn scheduler because they are already 29538 calculated. 29539 29540 -- Target Hook: void TARGET_SCHED_INIT (FILE *FILE, int VERBOSE, int 29541 MAX_READY) 29542 This hook is executed by the scheduler at the beginning of each 29543 block of instructions that are to be scheduled. FILE is either a 29544 null pointer, or a stdio stream to write any debug output to. 29545 VERBOSE is the verbose level provided by `-fsched-verbose-N'. 29546 MAX_READY is the maximum number of insns in the current scheduling 29547 region that can be live at the same time. This can be used to 29548 allocate scratch space if it is needed, e.g. by 29549 `TARGET_SCHED_REORDER'. 29550 29551 -- Target Hook: void TARGET_SCHED_FINISH (FILE *FILE, int VERBOSE) 29552 This hook is executed by the scheduler at the end of each block of 29553 instructions that are to be scheduled. It can be used to perform 29554 cleanup of any actions done by the other scheduling hooks. FILE 29555 is either a null pointer, or a stdio stream to write any debug 29556 output to. VERBOSE is the verbose level provided by 29557 `-fsched-verbose-N'. 29558 29559 -- Target Hook: void TARGET_SCHED_INIT_GLOBAL (FILE *FILE, int 29560 VERBOSE, int OLD_MAX_UID) 29561 This hook is executed by the scheduler after function level 29562 initializations. FILE is either a null pointer, or a stdio stream 29563 to write any debug output to. VERBOSE is the verbose level 29564 provided by `-fsched-verbose-N'. OLD_MAX_UID is the maximum insn 29565 uid when scheduling begins. 29566 29567 -- Target Hook: void TARGET_SCHED_FINISH_GLOBAL (FILE *FILE, int 29568 VERBOSE) 29569 This is the cleanup hook corresponding to 29570 `TARGET_SCHED_INIT_GLOBAL'. FILE is either a null pointer, or a 29571 stdio stream to write any debug output to. VERBOSE is the verbose 29572 level provided by `-fsched-verbose-N'. 29573 29574 -- Target Hook: int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void) 29575 The hook returns an RTL insn. The automaton state used in the 29576 pipeline hazard recognizer is changed as if the insn were scheduled 29577 when the new simulated processor cycle starts. Usage of the hook 29578 may simplify the automaton pipeline description for some VLIW 29579 processors. If the hook is defined, it is used only for the 29580 automaton based pipeline description. The default is not to 29581 change the state when the new simulated processor cycle starts. 29582 29583 -- Target Hook: void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void) 29584 The hook can be used to initialize data used by the previous hook. 29585 29586 -- Target Hook: int TARGET_SCHED_DFA_POST_CYCLE_INSN (void) 29587 The hook is analogous to `TARGET_SCHED_DFA_PRE_CYCLE_INSN' but used 29588 to changed the state as if the insn were scheduled when the new 29589 simulated processor cycle finishes. 29590 29591 -- Target Hook: void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void) 29592 The hook is analogous to `TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN' but 29593 used to initialize data used by the previous hook. 29594 29595 -- Target Hook: void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void) 29596 The hook to notify target that the current simulated cycle is 29597 about to finish. The hook is analogous to 29598 `TARGET_SCHED_DFA_PRE_CYCLE_INSN' but used to change the state in 29599 more complicated situations - e.g., when advancing state on a 29600 single insn is not enough. 29601 29602 -- Target Hook: void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void) 29603 The hook to notify target that new simulated cycle has just 29604 started. The hook is analogous to 29605 `TARGET_SCHED_DFA_POST_CYCLE_INSN' but used to change the state in 29606 more complicated situations - e.g., when advancing state on a 29607 single insn is not enough. 29608 29609 -- Target Hook: int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD 29610 (void) 29611 This hook controls better choosing an insn from the ready insn 29612 queue for the DFA-based insn scheduler. Usually the scheduler 29613 chooses the first insn from the queue. If the hook returns a 29614 positive value, an additional scheduler code tries all 29615 permutations of `TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD 29616 ()' subsequent ready insns to choose an insn whose issue will 29617 result in maximal number of issued insns on the same cycle. For 29618 the VLIW processor, the code could actually solve the problem of 29619 packing simple insns into the VLIW insn. Of course, if the rules 29620 of VLIW packing are described in the automaton. 29621 29622 This code also could be used for superscalar RISC processors. Let 29623 us consider a superscalar RISC processor with 3 pipelines. Some 29624 insns can be executed in pipelines A or B, some insns can be 29625 executed only in pipelines B or C, and one insn can be executed in 29626 pipeline B. The processor may issue the 1st insn into A and the 29627 2nd one into B. In this case, the 3rd insn will wait for freeing B 29628 until the next cycle. If the scheduler issues the 3rd insn the 29629 first, the processor could issue all 3 insns per cycle. 29630 29631 Actually this code demonstrates advantages of the automaton based 29632 pipeline hazard recognizer. We try quickly and easy many insn 29633 schedules to choose the best one. 29634 29635 The default is no multipass scheduling. 29636 29637 -- Target Hook: int 29638 TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx) 29639 This hook controls what insns from the ready insn queue will be 29640 considered for the multipass insn scheduling. If the hook returns 29641 zero for insn passed as the parameter, the insn will be not chosen 29642 to be issued. 29643 29644 The default is that any ready insns can be chosen to be issued. 29645 29646 -- Target Hook: int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, 29647 int, int *) 29648 This hook is called by the insn scheduler before issuing insn 29649 passed as the third parameter on given cycle. If the hook returns 29650 nonzero, the insn is not issued on given processors cycle. 29651 Instead of that, the processor cycle is advanced. If the value 29652 passed through the last parameter is zero, the insn ready queue is 29653 not sorted on the new cycle start as usually. The first parameter 29654 passes file for debugging output. The second one passes the 29655 scheduler verbose level of the debugging output. The forth and 29656 the fifth parameter values are correspondingly processor cycle on 29657 which the previous insn has been issued and the current processor 29658 cycle. 29659 29660 -- Target Hook: bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def 29661 *_DEP, int COST, int DISTANCE) 29662 This hook is used to define which dependences are considered 29663 costly by the target, so costly that it is not advisable to 29664 schedule the insns that are involved in the dependence too close 29665 to one another. The parameters to this hook are as follows: The 29666 first parameter _DEP is the dependence being evaluated. The 29667 second parameter COST is the cost of the dependence, and the third 29668 parameter DISTANCE is the distance in cycles between the two insns. 29669 The hook returns `true' if considering the distance between the two 29670 insns the dependence between them is considered costly by the 29671 target, and `false' otherwise. 29672 29673 Defining this hook can be useful in multiple-issue out-of-order 29674 machines, where (a) it's practically hopeless to predict the 29675 actual data/resource delays, however: (b) there's a better chance 29676 to predict the actual grouping that will be formed, and (c) 29677 correctly emulating the grouping can be very important. In such 29678 targets one may want to allow issuing dependent insns closer to 29679 one another--i.e., closer than the dependence distance; however, 29680 not in cases of "costly dependences", which this hooks allows to 29681 define. 29682 29683 -- Target Hook: void TARGET_SCHED_H_I_D_EXTENDED (void) 29684 This hook is called by the insn scheduler after emitting a new 29685 instruction to the instruction stream. The hook notifies a target 29686 backend to extend its per instruction data structures. 29687 29688 -- Target Hook: void * TARGET_SCHED_ALLOC_SCHED_CONTEXT (void) 29689 Return a pointer to a store large enough to hold target scheduling 29690 context. 29691 29692 -- Target Hook: void TARGET_SCHED_INIT_SCHED_CONTEXT (void *TC, bool 29693 CLEAN_P) 29694 Initialize store pointed to by TC to hold target scheduling 29695 context. It CLEAN_P is true then initialize TC as if scheduler is 29696 at the beginning of the block. Otherwise, make a copy of the 29697 current context in TC. 29698 29699 -- Target Hook: void TARGET_SCHED_SET_SCHED_CONTEXT (void *TC) 29700 Copy target scheduling context pointer to by TC to the current 29701 context. 29702 29703 -- Target Hook: void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *TC) 29704 Deallocate internal data in target scheduling context pointed to 29705 by TC. 29706 29707 -- Target Hook: void TARGET_SCHED_FREE_SCHED_CONTEXT (void *TC) 29708 Deallocate a store for target scheduling context pointed to by TC. 29709 29710 -- Target Hook: void * TARGET_SCHED_ALLOC_SCHED_CONTEXT (void) 29711 Return a pointer to a store large enough to hold target scheduling 29712 context. 29713 29714 -- Target Hook: void TARGET_SCHED_INIT_SCHED_CONTEXT (void *TC, bool 29715 CLEAN_P) 29716 Initialize store pointed to by TC to hold target scheduling 29717 context. It CLEAN_P is true then initialize TC as if scheduler is 29718 at the beginning of the block. Otherwise, make a copy of the 29719 current context in TC. 29720 29721 -- Target Hook: void TARGET_SCHED_SET_SCHED_CONTEXT (void *TC) 29722 Copy target scheduling context pointer to by TC to the current 29723 context. 29724 29725 -- Target Hook: void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *TC) 29726 Deallocate internal data in target scheduling context pointed to 29727 by TC. 29728 29729 -- Target Hook: void TARGET_SCHED_FREE_SCHED_CONTEXT (void *TC) 29730 Deallocate a store for target scheduling context pointed to by TC. 29731 29732 -- Target Hook: int TARGET_SCHED_SPECULATE_INSN (rtx INSN, int 29733 REQUEST, rtx *NEW_PAT) 29734 This hook is called by the insn scheduler when INSN has only 29735 speculative dependencies and therefore can be scheduled 29736 speculatively. The hook is used to check if the pattern of INSN 29737 has a speculative version and, in case of successful check, to 29738 generate that speculative pattern. The hook should return 1, if 29739 the instruction has a speculative form, or -1, if it doesn't. 29740 REQUEST describes the type of requested speculation. If the 29741 return value equals 1 then NEW_PAT is assigned the generated 29742 speculative pattern. 29743 29744 -- Target Hook: int TARGET_SCHED_NEEDS_BLOCK_P (rtx INSN) 29745 This hook is called by the insn scheduler during generation of 29746 recovery code for INSN. It should return nonzero, if the 29747 corresponding check instruction should branch to recovery code, or 29748 zero otherwise. 29749 29750 -- Target Hook: rtx TARGET_SCHED_GEN_CHECK (rtx INSN, rtx LABEL, int 29751 MUTATE_P) 29752 This hook is called by the insn scheduler to generate a pattern 29753 for recovery check instruction. If MUTATE_P is zero, then INSN is 29754 a speculative instruction for which the check should be generated. 29755 LABEL is either a label of a basic block, where recovery code 29756 should be emitted, or a null pointer, when requested check doesn't 29757 branch to recovery code (a simple check). If MUTATE_P is nonzero, 29758 then a pattern for a branchy check corresponding to a simple check 29759 denoted by INSN should be generated. In this case LABEL can't be 29760 null. 29761 29762 -- Target Hook: int 29763 TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx INSN) 29764 This hook is used as a workaround for 29765 `TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD' not being 29766 called on the first instruction of the ready list. The hook is 29767 used to discard speculative instruction that stand first in the 29768 ready list from being scheduled on the current cycle. For 29769 non-speculative instructions, the hook should always return 29770 nonzero. For example, in the ia64 backend the hook is used to 29771 cancel data speculative insns when the ALAT table is nearly full. 29772 29773 -- Target Hook: void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int 29774 *FLAGS, spec_info_t SPEC_INFO) 29775 This hook is used by the insn scheduler to find out what features 29776 should be enabled/used. FLAGS initially may have either the 29777 SCHED_RGN or SCHED_EBB bit set. This denotes the scheduler pass 29778 for which the data should be provided. The target backend should 29779 modify FLAGS by modifying the bits corresponding to the following 29780 features: USE_DEPS_LIST, USE_GLAT, DETACH_LIFE_INFO, and 29781 DO_SPECULATION. For the DO_SPECULATION feature an additional 29782 structure SPEC_INFO should be filled by the target. The structure 29783 describes speculation types that can be used in the scheduler. 29784 29785 -- Target Hook: int TARGET_SCHED_SMS_RES_MII (struct ddg *G) 29786 This hook is called by the swing modulo scheduler to calculate a 29787 resource-based lower bound which is based on the resources 29788 available in the machine and the resources required by each 29789 instruction. The target backend can use G to calculate such 29790 bound. A very simple lower bound will be used in case this hook 29791 is not implemented: the total number of instructions divided by 29792 the issue rate. 29793 29794 29795 File: gccint.info, Node: Sections, Next: PIC, Prev: Scheduling, Up: Target Macros 29796 29797 17.19 Dividing the Output into Sections (Texts, Data, ...) 29798 ========================================================== 29799 29800 An object file is divided into sections containing different types of 29801 data. In the most common case, there are three sections: the "text 29802 section", which holds instructions and read-only data; the "data 29803 section", which holds initialized writable data; and the "bss section", 29804 which holds uninitialized data. Some systems have other kinds of 29805 sections. 29806 29807 `varasm.c' provides several well-known sections, such as 29808 `text_section', `data_section' and `bss_section'. The normal way of 29809 controlling a `FOO_section' variable is to define the associated 29810 `FOO_SECTION_ASM_OP' macro, as described below. The macros are only 29811 read once, when `varasm.c' initializes itself, so their values must be 29812 run-time constants. They may however depend on command-line flags. 29813 29814 _Note:_ Some run-time files, such `crtstuff.c', also make use of the 29815 `FOO_SECTION_ASM_OP' macros, and expect them to be string literals. 29816 29817 Some assemblers require a different string to be written every time a 29818 section is selected. If your assembler falls into this category, you 29819 should define the `TARGET_ASM_INIT_SECTIONS' hook and use 29820 `get_unnamed_section' to set up the sections. 29821 29822 You must always create a `text_section', either by defining 29823 `TEXT_SECTION_ASM_OP' or by initializing `text_section' in 29824 `TARGET_ASM_INIT_SECTIONS'. The same is true of `data_section' and 29825 `DATA_SECTION_ASM_OP'. If you do not create a distinct 29826 `readonly_data_section', the default is to reuse `text_section'. 29827 29828 All the other `varasm.c' sections are optional, and are null if the 29829 target does not provide them. 29830 29831 -- Macro: TEXT_SECTION_ASM_OP 29832 A C expression whose value is a string, including spacing, 29833 containing the assembler operation that should precede 29834 instructions and read-only data. Normally `"\t.text"' is right. 29835 29836 -- Macro: HOT_TEXT_SECTION_NAME 29837 If defined, a C string constant for the name of the section 29838 containing most frequently executed functions of the program. If 29839 not defined, GCC will provide a default definition if the target 29840 supports named sections. 29841 29842 -- Macro: UNLIKELY_EXECUTED_TEXT_SECTION_NAME 29843 If defined, a C string constant for the name of the section 29844 containing unlikely executed functions in the program. 29845 29846 -- Macro: DATA_SECTION_ASM_OP 29847 A C expression whose value is a string, including spacing, 29848 containing the assembler operation to identify the following data 29849 as writable initialized data. Normally `"\t.data"' is right. 29850 29851 -- Macro: SDATA_SECTION_ASM_OP 29852 If defined, a C expression whose value is a string, including 29853 spacing, containing the assembler operation to identify the 29854 following data as initialized, writable small data. 29855 29856 -- Macro: READONLY_DATA_SECTION_ASM_OP 29857 A C expression whose value is a string, including spacing, 29858 containing the assembler operation to identify the following data 29859 as read-only initialized data. 29860 29861 -- Macro: BSS_SECTION_ASM_OP 29862 If defined, a C expression whose value is a string, including 29863 spacing, containing the assembler operation to identify the 29864 following data as uninitialized global data. If not defined, and 29865 neither `ASM_OUTPUT_BSS' nor `ASM_OUTPUT_ALIGNED_BSS' are defined, 29866 uninitialized global data will be output in the data section if 29867 `-fno-common' is passed, otherwise `ASM_OUTPUT_COMMON' will be 29868 used. 29869 29870 -- Macro: SBSS_SECTION_ASM_OP 29871 If defined, a C expression whose value is a string, including 29872 spacing, containing the assembler operation to identify the 29873 following data as uninitialized, writable small data. 29874 29875 -- Macro: INIT_SECTION_ASM_OP 29876 If defined, a C expression whose value is a string, including 29877 spacing, containing the assembler operation to identify the 29878 following data as initialization code. If not defined, GCC will 29879 assume such a section does not exist. This section has no 29880 corresponding `init_section' variable; it is used entirely in 29881 runtime code. 29882 29883 -- Macro: FINI_SECTION_ASM_OP 29884 If defined, a C expression whose value is a string, including 29885 spacing, containing the assembler operation to identify the 29886 following data as finalization code. If not defined, GCC will 29887 assume such a section does not exist. This section has no 29888 corresponding `fini_section' variable; it is used entirely in 29889 runtime code. 29890 29891 -- Macro: INIT_ARRAY_SECTION_ASM_OP 29892 If defined, a C expression whose value is a string, including 29893 spacing, containing the assembler operation to identify the 29894 following data as part of the `.init_array' (or equivalent) 29895 section. If not defined, GCC will assume such a section does not 29896 exist. Do not define both this macro and `INIT_SECTION_ASM_OP'. 29897 29898 -- Macro: FINI_ARRAY_SECTION_ASM_OP 29899 If defined, a C expression whose value is a string, including 29900 spacing, containing the assembler operation to identify the 29901 following data as part of the `.fini_array' (or equivalent) 29902 section. If not defined, GCC will assume such a section does not 29903 exist. Do not define both this macro and `FINI_SECTION_ASM_OP'. 29904 29905 -- Macro: CRT_CALL_STATIC_FUNCTION (SECTION_OP, FUNCTION) 29906 If defined, an ASM statement that switches to a different section 29907 via SECTION_OP, calls FUNCTION, and switches back to the text 29908 section. This is used in `crtstuff.c' if `INIT_SECTION_ASM_OP' or 29909 `FINI_SECTION_ASM_OP' to calls to initialization and finalization 29910 functions from the init and fini sections. By default, this macro 29911 uses a simple function call. Some ports need hand-crafted 29912 assembly code to avoid dependencies on registers initialized in 29913 the function prologue or to ensure that constant pools don't end 29914 up too far way in the text section. 29915 29916 -- Macro: TARGET_LIBGCC_SDATA_SECTION 29917 If defined, a string which names the section into which small 29918 variables defined in crtstuff and libgcc should go. This is useful 29919 when the target has options for optimizing access to small data, 29920 and you want the crtstuff and libgcc routines to be conservative 29921 in what they expect of your application yet liberal in what your 29922 application expects. For example, for targets with a `.sdata' 29923 section (like MIPS), you could compile crtstuff with `-G 0' so 29924 that it doesn't require small data support from your application, 29925 but use this macro to put small data into `.sdata' so that your 29926 application can access these variables whether it uses small data 29927 or not. 29928 29929 -- Macro: FORCE_CODE_SECTION_ALIGN 29930 If defined, an ASM statement that aligns a code section to some 29931 arbitrary boundary. This is used to force all fragments of the 29932 `.init' and `.fini' sections to have to same alignment and thus 29933 prevent the linker from having to add any padding. 29934 29935 -- Macro: JUMP_TABLES_IN_TEXT_SECTION 29936 Define this macro to be an expression with a nonzero value if jump 29937 tables (for `tablejump' insns) should be output in the text 29938 section, along with the assembler instructions. Otherwise, the 29939 readonly data section is used. 29940 29941 This macro is irrelevant if there is no separate readonly data 29942 section. 29943 29944 -- Target Hook: void TARGET_ASM_INIT_SECTIONS (void) 29945 Define this hook if you need to do something special to set up the 29946 `varasm.c' sections, or if your target has some special sections 29947 of its own that you need to create. 29948 29949 GCC calls this hook after processing the command line, but before 29950 writing any assembly code, and before calling any of the 29951 section-returning hooks described below. 29952 29953 -- Target Hook: TARGET_ASM_RELOC_RW_MASK (void) 29954 Return a mask describing how relocations should be treated when 29955 selecting sections. Bit 1 should be set if global relocations 29956 should be placed in a read-write section; bit 0 should be set if 29957 local relocations should be placed in a read-write section. 29958 29959 The default version of this function returns 3 when `-fpic' is in 29960 effect, and 0 otherwise. The hook is typically redefined when the 29961 target cannot support (some kinds of) dynamic relocations in 29962 read-only sections even in executables. 29963 29964 -- Target Hook: section * TARGET_ASM_SELECT_SECTION (tree EXP, int 29965 RELOC, unsigned HOST_WIDE_INT ALIGN) 29966 Return the section into which EXP should be placed. You can 29967 assume that EXP is either a `VAR_DECL' node or a constant of some 29968 sort. RELOC indicates whether the initial value of EXP requires 29969 link-time relocations. Bit 0 is set when variable contains local 29970 relocations only, while bit 1 is set for global relocations. 29971 ALIGN is the constant alignment in bits. 29972 29973 The default version of this function takes care of putting 29974 read-only variables in `readonly_data_section'. 29975 29976 See also USE_SELECT_SECTION_FOR_FUNCTIONS. 29977 29978 -- Macro: USE_SELECT_SECTION_FOR_FUNCTIONS 29979 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be 29980 called for `FUNCTION_DECL's as well as for variables and constants. 29981 29982 In the case of a `FUNCTION_DECL', RELOC will be zero if the 29983 function has been determined to be likely to be called, and 29984 nonzero if it is unlikely to be called. 29985 29986 -- Target Hook: void TARGET_ASM_UNIQUE_SECTION (tree DECL, int RELOC) 29987 Build up a unique section name, expressed as a `STRING_CST' node, 29988 and assign it to `DECL_SECTION_NAME (DECL)'. As with 29989 `TARGET_ASM_SELECT_SECTION', RELOC indicates whether the initial 29990 value of EXP requires link-time relocations. 29991 29992 The default version of this function appends the symbol name to the 29993 ELF section name that would normally be used for the symbol. For 29994 example, the function `foo' would be placed in `.text.foo'. 29995 Whatever the actual target object format, this is often good 29996 enough. 29997 29998 -- Target Hook: section * TARGET_ASM_FUNCTION_RODATA_SECTION (tree 29999 DECL) 30000 Return the readonly data section associated with 30001 `DECL_SECTION_NAME (DECL)'. The default version of this function 30002 selects `.gnu.linkonce.r.name' if the function's section is 30003 `.gnu.linkonce.t.name', `.rodata.name' if function is in 30004 `.text.name', and the normal readonly-data section otherwise. 30005 30006 -- Target Hook: section * TARGET_ASM_SELECT_RTX_SECTION (enum 30007 machine_mode MODE, rtx X, unsigned HOST_WIDE_INT ALIGN) 30008 Return the section into which a constant X, of mode MODE, should 30009 be placed. You can assume that X is some kind of constant in RTL. 30010 The argument MODE is redundant except in the case of a 30011 `const_int' rtx. ALIGN is the constant alignment in bits. 30012 30013 The default version of this function takes care of putting symbolic 30014 constants in `flag_pic' mode in `data_section' and everything else 30015 in `readonly_data_section'. 30016 30017 -- Target Hook: void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree DECL, 30018 tree ID) 30019 Define this hook if you need to postprocess the assembler name 30020 generated by target-independent code. The ID provided to this 30021 hook will be the computed name (e.g., the macro `DECL_NAME' of the 30022 DECL in C, or the mangled name of the DECL in C++). The return 30023 value of the hook is an `IDENTIFIER_NODE' for the appropriate 30024 mangled name on your target system. The default implementation of 30025 this hook just returns the ID provided. 30026 30027 -- Target Hook: void TARGET_ENCODE_SECTION_INFO (tree DECL, rtx RTL, 30028 int NEW_DECL_P) 30029 Define this hook if references to a symbol or a constant must be 30030 treated differently depending on something about the variable or 30031 function named by the symbol (such as what section it is in). 30032 30033 The hook is executed immediately after rtl has been created for 30034 DECL, which may be a variable or function declaration or an entry 30035 in the constant pool. In either case, RTL is the rtl in question. 30036 Do _not_ use `DECL_RTL (DECL)' in this hook; that field may not 30037 have been initialized yet. 30038 30039 In the case of a constant, it is safe to assume that the rtl is a 30040 `mem' whose address is a `symbol_ref'. Most decls will also have 30041 this form, but that is not guaranteed. Global register variables, 30042 for instance, will have a `reg' for their rtl. (Normally the 30043 right thing to do with such unusual rtl is leave it alone.) 30044 30045 The NEW_DECL_P argument will be true if this is the first time 30046 that `TARGET_ENCODE_SECTION_INFO' has been invoked on this decl. 30047 It will be false for subsequent invocations, which will happen for 30048 duplicate declarations. Whether or not anything must be done for 30049 the duplicate declaration depends on whether the hook examines 30050 `DECL_ATTRIBUTES'. NEW_DECL_P is always true when the hook is 30051 called for a constant. 30052 30053 The usual thing for this hook to do is to record flags in the 30054 `symbol_ref', using `SYMBOL_REF_FLAG' or `SYMBOL_REF_FLAGS'. 30055 Historically, the name string was modified if it was necessary to 30056 encode more than one bit of information, but this practice is now 30057 discouraged; use `SYMBOL_REF_FLAGS'. 30058 30059 The default definition of this hook, `default_encode_section_info' 30060 in `varasm.c', sets a number of commonly-useful bits in 30061 `SYMBOL_REF_FLAGS'. Check whether the default does what you need 30062 before overriding it. 30063 30064 -- Target Hook: const char *TARGET_STRIP_NAME_ENCODING (const char 30065 *name) 30066 Decode NAME and return the real name part, sans the characters 30067 that `TARGET_ENCODE_SECTION_INFO' may have added. 30068 30069 -- Target Hook: bool TARGET_IN_SMALL_DATA_P (tree EXP) 30070 Returns true if EXP should be placed into a "small data" section. 30071 The default version of this hook always returns false. 30072 30073 -- Variable: Target Hook bool TARGET_HAVE_SRODATA_SECTION 30074 Contains the value true if the target places read-only "small 30075 data" into a separate section. The default value is false. 30076 30077 -- Target Hook: bool TARGET_BINDS_LOCAL_P (tree EXP) 30078 Returns true if EXP names an object for which name resolution 30079 rules must resolve to the current "module" (dynamic shared library 30080 or executable image). 30081 30082 The default version of this hook implements the name resolution 30083 rules for ELF, which has a looser model of global name binding 30084 than other currently supported object file formats. 30085 30086 -- Variable: Target Hook bool TARGET_HAVE_TLS 30087 Contains the value true if the target supports thread-local 30088 storage. The default value is false. 30089 30090 30091 File: gccint.info, Node: PIC, Next: Assembler Format, Prev: Sections, Up: Target Macros 30092 30093 17.20 Position Independent Code 30094 =============================== 30095 30096 This section describes macros that help implement generation of position 30097 independent code. Simply defining these macros is not enough to 30098 generate valid PIC; you must also add support to the macros 30099 `GO_IF_LEGITIMATE_ADDRESS' and `PRINT_OPERAND_ADDRESS', as well as 30100 `LEGITIMIZE_ADDRESS'. You must modify the definition of `movsi' to do 30101 something appropriate when the source operand contains a symbolic 30102 address. You may also need to alter the handling of switch statements 30103 so that they use relative addresses. 30104 30105 -- Macro: PIC_OFFSET_TABLE_REGNUM 30106 The register number of the register used to address a table of 30107 static data addresses in memory. In some cases this register is 30108 defined by a processor's "application binary interface" (ABI). 30109 When this macro is defined, RTL is generated for this register 30110 once, as with the stack pointer and frame pointer registers. If 30111 this macro is not defined, it is up to the machine-dependent files 30112 to allocate such a register (if necessary). Note that this 30113 register must be fixed when in use (e.g. when `flag_pic' is true). 30114 30115 -- Macro: PIC_OFFSET_TABLE_REG_CALL_CLOBBERED 30116 Define this macro if the register defined by 30117 `PIC_OFFSET_TABLE_REGNUM' is clobbered by calls. Do not define 30118 this macro if `PIC_OFFSET_TABLE_REGNUM' is not defined. 30119 30120 -- Macro: LEGITIMATE_PIC_OPERAND_P (X) 30121 A C expression that is nonzero if X is a legitimate immediate 30122 operand on the target machine when generating position independent 30123 code. You can assume that X satisfies `CONSTANT_P', so you need 30124 not check this. You can also assume FLAG_PIC is true, so you need 30125 not check it either. You need not define this macro if all 30126 constants (including `SYMBOL_REF') can be immediate operands when 30127 generating position independent code. 30128 30129 30130 File: gccint.info, Node: Assembler Format, Next: Debugging Info, Prev: PIC, Up: Target Macros 30131 30132 17.21 Defining the Output Assembler Language 30133 ============================================ 30134 30135 This section describes macros whose principal purpose is to describe how 30136 to write instructions in assembler language--rather than what the 30137 instructions do. 30138 30139 * Menu: 30140 30141 * File Framework:: Structural information for the assembler file. 30142 * Data Output:: Output of constants (numbers, strings, addresses). 30143 * Uninitialized Data:: Output of uninitialized variables. 30144 * Label Output:: Output and generation of labels. 30145 * Initialization:: General principles of initialization 30146 and termination routines. 30147 * Macros for Initialization:: 30148 Specific macros that control the handling of 30149 initialization and termination routines. 30150 * Instruction Output:: Output of actual instructions. 30151 * Dispatch Tables:: Output of jump tables. 30152 * Exception Region Output:: Output of exception region code. 30153 * Alignment Output:: Pseudo ops for alignment and skipping data. 30154 30155 30156 File: gccint.info, Node: File Framework, Next: Data Output, Up: Assembler Format 30157 30158 17.21.1 The Overall Framework of an Assembler File 30159 -------------------------------------------------- 30160 30161 This describes the overall framework of an assembly file. 30162 30163 -- Target Hook: void TARGET_ASM_FILE_START () 30164 Output to `asm_out_file' any text which the assembler expects to 30165 find at the beginning of a file. The default behavior is 30166 controlled by two flags, documented below. Unless your target's 30167 assembler is quite unusual, if you override the default, you 30168 should call `default_file_start' at some point in your target 30169 hook. This lets other target files rely on these variables. 30170 30171 -- Target Hook: bool TARGET_ASM_FILE_START_APP_OFF 30172 If this flag is true, the text of the macro `ASM_APP_OFF' will be 30173 printed as the very first line in the assembly file, unless 30174 `-fverbose-asm' is in effect. (If that macro has been defined to 30175 the empty string, this variable has no effect.) With the normal 30176 definition of `ASM_APP_OFF', the effect is to notify the GNU 30177 assembler that it need not bother stripping comments or extra 30178 whitespace from its input. This allows it to work a bit faster. 30179 30180 The default is false. You should not set it to true unless you 30181 have verified that your port does not generate any extra 30182 whitespace or comments that will cause GAS to issue errors in 30183 NO_APP mode. 30184 30185 -- Target Hook: bool TARGET_ASM_FILE_START_FILE_DIRECTIVE 30186 If this flag is true, `output_file_directive' will be called for 30187 the primary source file, immediately after printing `ASM_APP_OFF' 30188 (if that is enabled). Most ELF assemblers expect this to be done. 30189 The default is false. 30190 30191 -- Target Hook: void TARGET_ASM_FILE_END () 30192 Output to `asm_out_file' any text which the assembler expects to 30193 find at the end of a file. The default is to output nothing. 30194 30195 -- Function: void file_end_indicate_exec_stack () 30196 Some systems use a common convention, the `.note.GNU-stack' 30197 special section, to indicate whether or not an object file relies 30198 on the stack being executable. If your system uses this 30199 convention, you should define `TARGET_ASM_FILE_END' to this 30200 function. If you need to do other things in that hook, have your 30201 hook function call this function. 30202 30203 -- Macro: ASM_COMMENT_START 30204 A C string constant describing how to begin a comment in the target 30205 assembler language. The compiler assumes that the comment will 30206 end at the end of the line. 30207 30208 -- Macro: ASM_APP_ON 30209 A C string constant for text to be output before each `asm' 30210 statement or group of consecutive ones. Normally this is 30211 `"#APP"', which is a comment that has no effect on most assemblers 30212 but tells the GNU assembler that it must check the lines that 30213 follow for all valid assembler constructs. 30214 30215 -- Macro: ASM_APP_OFF 30216 A C string constant for text to be output after each `asm' 30217 statement or group of consecutive ones. Normally this is 30218 `"#NO_APP"', which tells the GNU assembler to resume making the 30219 time-saving assumptions that are valid for ordinary compiler 30220 output. 30221 30222 -- Macro: ASM_OUTPUT_SOURCE_FILENAME (STREAM, NAME) 30223 A C statement to output COFF information or DWARF debugging 30224 information which indicates that filename NAME is the current 30225 source file to the stdio stream STREAM. 30226 30227 This macro need not be defined if the standard form of output for 30228 the file format in use is appropriate. 30229 30230 -- Macro: OUTPUT_QUOTED_STRING (STREAM, STRING) 30231 A C statement to output the string STRING to the stdio stream 30232 STREAM. If you do not call the function `output_quoted_string' in 30233 your config files, GCC will only call it to output filenames to 30234 the assembler source. So you can use it to canonicalize the format 30235 of the filename using this macro. 30236 30237 -- Macro: ASM_OUTPUT_IDENT (STREAM, STRING) 30238 A C statement to output something to the assembler file to handle a 30239 `#ident' directive containing the text STRING. If this macro is 30240 not defined, nothing is output for a `#ident' directive. 30241 30242 -- Target Hook: void TARGET_ASM_NAMED_SECTION (const char *NAME, 30243 unsigned int FLAGS, unsigned int ALIGN) 30244 Output assembly directives to switch to section NAME. The section 30245 should have attributes as specified by FLAGS, which is a bit mask 30246 of the `SECTION_*' flags defined in `output.h'. If ALIGN is 30247 nonzero, it contains an alignment in bytes to be used for the 30248 section, otherwise some target default should be used. Only 30249 targets that must specify an alignment within the section 30250 directive need pay attention to ALIGN - we will still use 30251 `ASM_OUTPUT_ALIGN'. 30252 30253 -- Target Hook: bool TARGET_HAVE_NAMED_SECTIONS 30254 This flag is true if the target supports 30255 `TARGET_ASM_NAMED_SECTION'. 30256 30257 -- Target Hook: bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS 30258 This flag is true if we can create zeroed data by switching to a 30259 BSS section and then using `ASM_OUTPUT_SKIP' to allocate the space. 30260 This is true on most ELF targets. 30261 30262 -- Target Hook: unsigned int TARGET_SECTION_TYPE_FLAGS (tree DECL, 30263 const char *NAME, int RELOC) 30264 Choose a set of section attributes for use by 30265 `TARGET_ASM_NAMED_SECTION' based on a variable or function decl, a 30266 section name, and whether or not the declaration's initializer may 30267 contain runtime relocations. DECL may be null, in which case 30268 read-write data should be assumed. 30269 30270 The default version of this function handles choosing code vs data, 30271 read-only vs read-write data, and `flag_pic'. You should only 30272 need to override this if your target has special flags that might 30273 be set via `__attribute__'. 30274 30275 -- Target Hook: int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type 30276 TYPE, const char * TEXT) 30277 Provides the target with the ability to record the gcc command line 30278 switches that have been passed to the compiler, and options that 30279 are enabled. The TYPE argument specifies what is being recorded. 30280 It can take the following values: 30281 30282 `SWITCH_TYPE_PASSED' 30283 TEXT is a command line switch that has been set by the user. 30284 30285 `SWITCH_TYPE_ENABLED' 30286 TEXT is an option which has been enabled. This might be as a 30287 direct result of a command line switch, or because it is 30288 enabled by default or because it has been enabled as a side 30289 effect of a different command line switch. For example, the 30290 `-O2' switch enables various different individual 30291 optimization passes. 30292 30293 `SWITCH_TYPE_DESCRIPTIVE' 30294 TEXT is either NULL or some descriptive text which should be 30295 ignored. If TEXT is NULL then it is being used to warn the 30296 target hook that either recording is starting or ending. The 30297 first time TYPE is SWITCH_TYPE_DESCRIPTIVE and TEXT is NULL, 30298 the warning is for start up and the second time the warning 30299 is for wind down. This feature is to allow the target hook 30300 to make any necessary preparations before it starts to record 30301 switches and to perform any necessary tidying up after it has 30302 finished recording switches. 30303 30304 `SWITCH_TYPE_LINE_START' 30305 This option can be ignored by this target hook. 30306 30307 `SWITCH_TYPE_LINE_END' 30308 This option can be ignored by this target hook. 30309 30310 The hook's return value must be zero. Other return values may be 30311 supported in the future. 30312 30313 By default this hook is set to NULL, but an example implementation 30314 is provided for ELF based targets. Called ELF_RECORD_GCC_SWITCHES, 30315 it records the switches as ASCII text inside a new, string 30316 mergeable section in the assembler output file. The name of the 30317 new section is provided by the 30318 `TARGET_ASM_RECORD_GCC_SWITCHES_SECTION' target hook. 30319 30320 -- Target Hook: const char * TARGET_ASM_RECORD_GCC_SWITCHES_SECTION 30321 This is the name of the section that will be created by the example 30322 ELF implementation of the `TARGET_ASM_RECORD_GCC_SWITCHES' target 30323 hook. 30324 30325 30326 File: gccint.info, Node: Data Output, Next: Uninitialized Data, Prev: File Framework, Up: Assembler Format 30327 30328 17.21.2 Output of Data 30329 ---------------------- 30330 30331 -- Target Hook: const char * TARGET_ASM_BYTE_OP 30332 -- Target Hook: const char * TARGET_ASM_ALIGNED_HI_OP 30333 -- Target Hook: const char * TARGET_ASM_ALIGNED_SI_OP 30334 -- Target Hook: const char * TARGET_ASM_ALIGNED_DI_OP 30335 -- Target Hook: const char * TARGET_ASM_ALIGNED_TI_OP 30336 -- Target Hook: const char * TARGET_ASM_UNALIGNED_HI_OP 30337 -- Target Hook: const char * TARGET_ASM_UNALIGNED_SI_OP 30338 -- Target Hook: const char * TARGET_ASM_UNALIGNED_DI_OP 30339 -- Target Hook: const char * TARGET_ASM_UNALIGNED_TI_OP 30340 These hooks specify assembly directives for creating certain kinds 30341 of integer object. The `TARGET_ASM_BYTE_OP' directive creates a 30342 byte-sized object, the `TARGET_ASM_ALIGNED_HI_OP' one creates an 30343 aligned two-byte object, and so on. Any of the hooks may be 30344 `NULL', indicating that no suitable directive is available. 30345 30346 The compiler will print these strings at the start of a new line, 30347 followed immediately by the object's initial value. In most cases, 30348 the string should contain a tab, a pseudo-op, and then another tab. 30349 30350 -- Target Hook: bool TARGET_ASM_INTEGER (rtx X, unsigned int SIZE, int 30351 ALIGNED_P) 30352 The `assemble_integer' function uses this hook to output an 30353 integer object. X is the object's value, SIZE is its size in 30354 bytes and ALIGNED_P indicates whether it is aligned. The function 30355 should return `true' if it was able to output the object. If it 30356 returns false, `assemble_integer' will try to split the object 30357 into smaller parts. 30358 30359 The default implementation of this hook will use the 30360 `TARGET_ASM_BYTE_OP' family of strings, returning `false' when the 30361 relevant string is `NULL'. 30362 30363 -- Macro: OUTPUT_ADDR_CONST_EXTRA (STREAM, X, FAIL) 30364 A C statement to recognize RTX patterns that `output_addr_const' 30365 can't deal with, and output assembly code to STREAM corresponding 30366 to the pattern X. This may be used to allow machine-dependent 30367 `UNSPEC's to appear within constants. 30368 30369 If `OUTPUT_ADDR_CONST_EXTRA' fails to recognize a pattern, it must 30370 `goto fail', so that a standard error message is printed. If it 30371 prints an error message itself, by calling, for example, 30372 `output_operand_lossage', it may just complete normally. 30373 30374 -- Macro: ASM_OUTPUT_ASCII (STREAM, PTR, LEN) 30375 A C statement to output to the stdio stream STREAM an assembler 30376 instruction to assemble a string constant containing the LEN bytes 30377 at PTR. PTR will be a C expression of type `char *' and LEN a C 30378 expression of type `int'. 30379 30380 If the assembler has a `.ascii' pseudo-op as found in the Berkeley 30381 Unix assembler, do not define the macro `ASM_OUTPUT_ASCII'. 30382 30383 -- Macro: ASM_OUTPUT_FDESC (STREAM, DECL, N) 30384 A C statement to output word N of a function descriptor for DECL. 30385 This must be defined if `TARGET_VTABLE_USES_DESCRIPTORS' is 30386 defined, and is otherwise unused. 30387 30388 -- Macro: CONSTANT_POOL_BEFORE_FUNCTION 30389 You may define this macro as a C expression. You should define the 30390 expression to have a nonzero value if GCC should output the 30391 constant pool for a function before the code for the function, or 30392 a zero value if GCC should output the constant pool after the 30393 function. If you do not define this macro, the usual case, GCC 30394 will output the constant pool before the function. 30395 30396 -- Macro: ASM_OUTPUT_POOL_PROLOGUE (FILE, FUNNAME, FUNDECL, SIZE) 30397 A C statement to output assembler commands to define the start of 30398 the constant pool for a function. FUNNAME is a string giving the 30399 name of the function. Should the return type of the function be 30400 required, it can be obtained via FUNDECL. SIZE is the size, in 30401 bytes, of the constant pool that will be written immediately after 30402 this call. 30403 30404 If no constant-pool prefix is required, the usual case, this macro 30405 need not be defined. 30406 30407 -- Macro: ASM_OUTPUT_SPECIAL_POOL_ENTRY (FILE, X, MODE, ALIGN, 30408 LABELNO, JUMPTO) 30409 A C statement (with or without semicolon) to output a constant in 30410 the constant pool, if it needs special treatment. (This macro 30411 need not do anything for RTL expressions that can be output 30412 normally.) 30413 30414 The argument FILE is the standard I/O stream to output the 30415 assembler code on. X is the RTL expression for the constant to 30416 output, and MODE is the machine mode (in case X is a `const_int'). 30417 ALIGN is the required alignment for the value X; you should 30418 output an assembler directive to force this much alignment. 30419 30420 The argument LABELNO is a number to use in an internal label for 30421 the address of this pool entry. The definition of this macro is 30422 responsible for outputting the label definition at the proper 30423 place. Here is how to do this: 30424 30425 `(*targetm.asm_out.internal_label)' (FILE, "LC", LABELNO); 30426 30427 When you output a pool entry specially, you should end with a 30428 `goto' to the label JUMPTO. This will prevent the same pool entry 30429 from being output a second time in the usual manner. 30430 30431 You need not define this macro if it would do nothing. 30432 30433 -- Macro: ASM_OUTPUT_POOL_EPILOGUE (FILE FUNNAME FUNDECL SIZE) 30434 A C statement to output assembler commands to at the end of the 30435 constant pool for a function. FUNNAME is a string giving the name 30436 of the function. Should the return type of the function be 30437 required, you can obtain it via FUNDECL. SIZE is the size, in 30438 bytes, of the constant pool that GCC wrote immediately before this 30439 call. 30440 30441 If no constant-pool epilogue is required, the usual case, you need 30442 not define this macro. 30443 30444 -- Macro: IS_ASM_LOGICAL_LINE_SEPARATOR (C, STR) 30445 Define this macro as a C expression which is nonzero if C is used 30446 as a logical line separator by the assembler. STR points to the 30447 position in the string where C was found; this can be used if a 30448 line separator uses multiple characters. 30449 30450 If you do not define this macro, the default is that only the 30451 character `;' is treated as a logical line separator. 30452 30453 -- Target Hook: const char * TARGET_ASM_OPEN_PAREN 30454 -- Target Hook: const char * TARGET_ASM_CLOSE_PAREN 30455 These target hooks are C string constants, describing the syntax 30456 in the assembler for grouping arithmetic expressions. If not 30457 overridden, they default to normal parentheses, which is correct 30458 for most assemblers. 30459 30460 These macros are provided by `real.h' for writing the definitions of 30461 `ASM_OUTPUT_DOUBLE' and the like: 30462 30463 -- Macro: REAL_VALUE_TO_TARGET_SINGLE (X, L) 30464 -- Macro: REAL_VALUE_TO_TARGET_DOUBLE (X, L) 30465 -- Macro: REAL_VALUE_TO_TARGET_LONG_DOUBLE (X, L) 30466 -- Macro: REAL_VALUE_TO_TARGET_DECIMAL32 (X, L) 30467 -- Macro: REAL_VALUE_TO_TARGET_DECIMAL64 (X, L) 30468 -- Macro: REAL_VALUE_TO_TARGET_DECIMAL128 (X, L) 30469 These translate X, of type `REAL_VALUE_TYPE', to the target's 30470 floating point representation, and store its bit pattern in the 30471 variable L. For `REAL_VALUE_TO_TARGET_SINGLE' and 30472 `REAL_VALUE_TO_TARGET_DECIMAL32', this variable should be a simple 30473 `long int'. For the others, it should be an array of `long int'. 30474 The number of elements in this array is determined by the size of 30475 the desired target floating point data type: 32 bits of it go in 30476 each `long int' array element. Each array element holds 32 bits 30477 of the result, even if `long int' is wider than 32 bits on the 30478 host machine. 30479 30480 The array element values are designed so that you can print them 30481 out using `fprintf' in the order they should appear in the target 30482 machine's memory. 30483 30484 30485 File: gccint.info, Node: Uninitialized Data, Next: Label Output, Prev: Data Output, Up: Assembler Format 30486 30487 17.21.3 Output of Uninitialized Variables 30488 ----------------------------------------- 30489 30490 Each of the macros in this section is used to do the whole job of 30491 outputting a single uninitialized variable. 30492 30493 -- Macro: ASM_OUTPUT_COMMON (STREAM, NAME, SIZE, ROUNDED) 30494 A C statement (sans semicolon) to output to the stdio stream 30495 STREAM the assembler definition of a common-label named NAME whose 30496 size is SIZE bytes. The variable ROUNDED is the size rounded up 30497 to whatever alignment the caller wants. 30498 30499 Use the expression `assemble_name (STREAM, NAME)' to output the 30500 name itself; before and after that, output the additional 30501 assembler syntax for defining the name, and a newline. 30502 30503 This macro controls how the assembler definitions of uninitialized 30504 common global variables are output. 30505 30506 -- Macro: ASM_OUTPUT_ALIGNED_COMMON (STREAM, NAME, SIZE, ALIGNMENT) 30507 Like `ASM_OUTPUT_COMMON' except takes the required alignment as a 30508 separate, explicit argument. If you define this macro, it is used 30509 in place of `ASM_OUTPUT_COMMON', and gives you more flexibility in 30510 handling the required alignment of the variable. The alignment is 30511 specified as the number of bits. 30512 30513 -- Macro: ASM_OUTPUT_ALIGNED_DECL_COMMON (STREAM, DECL, NAME, SIZE, 30514 ALIGNMENT) 30515 Like `ASM_OUTPUT_ALIGNED_COMMON' except that DECL of the variable 30516 to be output, if there is one, or `NULL_TREE' if there is no 30517 corresponding variable. If you define this macro, GCC will use it 30518 in place of both `ASM_OUTPUT_COMMON' and 30519 `ASM_OUTPUT_ALIGNED_COMMON'. Define this macro when you need to 30520 see the variable's decl in order to chose what to output. 30521 30522 -- Macro: ASM_OUTPUT_BSS (STREAM, DECL, NAME, SIZE, ROUNDED) 30523 A C statement (sans semicolon) to output to the stdio stream 30524 STREAM the assembler definition of uninitialized global DECL named 30525 NAME whose size is SIZE bytes. The variable ROUNDED is the size 30526 rounded up to whatever alignment the caller wants. 30527 30528 Try to use function `asm_output_bss' defined in `varasm.c' when 30529 defining this macro. If unable, use the expression `assemble_name 30530 (STREAM, NAME)' to output the name itself; before and after that, 30531 output the additional assembler syntax for defining the name, and 30532 a newline. 30533 30534 There are two ways of handling global BSS. One is to define either 30535 this macro or its aligned counterpart, `ASM_OUTPUT_ALIGNED_BSS'. 30536 The other is to have `TARGET_ASM_SELECT_SECTION' return a 30537 switchable BSS section (*note 30538 TARGET_HAVE_SWITCHABLE_BSS_SECTIONS::). You do not need to do 30539 both. 30540 30541 Some languages do not have `common' data, and require a non-common 30542 form of global BSS in order to handle uninitialized globals 30543 efficiently. C++ is one example of this. However, if the target 30544 does not support global BSS, the front end may choose to make 30545 globals common in order to save space in the object file. 30546 30547 -- Macro: ASM_OUTPUT_ALIGNED_BSS (STREAM, DECL, NAME, SIZE, ALIGNMENT) 30548 Like `ASM_OUTPUT_BSS' except takes the required alignment as a 30549 separate, explicit argument. If you define this macro, it is used 30550 in place of `ASM_OUTPUT_BSS', and gives you more flexibility in 30551 handling the required alignment of the variable. The alignment is 30552 specified as the number of bits. 30553 30554 Try to use function `asm_output_aligned_bss' defined in file 30555 `varasm.c' when defining this macro. 30556 30557 -- Macro: ASM_OUTPUT_LOCAL (STREAM, NAME, SIZE, ROUNDED) 30558 A C statement (sans semicolon) to output to the stdio stream 30559 STREAM the assembler definition of a local-common-label named NAME 30560 whose size is SIZE bytes. The variable ROUNDED is the size 30561 rounded up to whatever alignment the caller wants. 30562 30563 Use the expression `assemble_name (STREAM, NAME)' to output the 30564 name itself; before and after that, output the additional 30565 assembler syntax for defining the name, and a newline. 30566 30567 This macro controls how the assembler definitions of uninitialized 30568 static variables are output. 30569 30570 -- Macro: ASM_OUTPUT_ALIGNED_LOCAL (STREAM, NAME, SIZE, ALIGNMENT) 30571 Like `ASM_OUTPUT_LOCAL' except takes the required alignment as a 30572 separate, explicit argument. If you define this macro, it is used 30573 in place of `ASM_OUTPUT_LOCAL', and gives you more flexibility in 30574 handling the required alignment of the variable. The alignment is 30575 specified as the number of bits. 30576 30577 -- Macro: ASM_OUTPUT_ALIGNED_DECL_LOCAL (STREAM, DECL, NAME, SIZE, 30578 ALIGNMENT) 30579 Like `ASM_OUTPUT_ALIGNED_DECL' except that DECL of the variable to 30580 be output, if there is one, or `NULL_TREE' if there is no 30581 corresponding variable. If you define this macro, GCC will use it 30582 in place of both `ASM_OUTPUT_DECL' and `ASM_OUTPUT_ALIGNED_DECL'. 30583 Define this macro when you need to see the variable's decl in 30584 order to chose what to output. 30585 30586 30587 File: gccint.info, Node: Label Output, Next: Initialization, Prev: Uninitialized Data, Up: Assembler Format 30588 30589 17.21.4 Output and Generation of Labels 30590 --------------------------------------- 30591 30592 This is about outputting labels. 30593 30594 -- Macro: ASM_OUTPUT_LABEL (STREAM, NAME) 30595 A C statement (sans semicolon) to output to the stdio stream 30596 STREAM the assembler definition of a label named NAME. Use the 30597 expression `assemble_name (STREAM, NAME)' to output the name 30598 itself; before and after that, output the additional assembler 30599 syntax for defining the name, and a newline. A default definition 30600 of this macro is provided which is correct for most systems. 30601 30602 -- Macro: ASM_OUTPUT_INTERNAL_LABEL (STREAM, NAME) 30603 Identical to `ASM_OUTPUT_LABEL', except that NAME is known to 30604 refer to a compiler-generated label. The default definition uses 30605 `assemble_name_raw', which is like `assemble_name' except that it 30606 is more efficient. 30607 30608 -- Macro: SIZE_ASM_OP 30609 A C string containing the appropriate assembler directive to 30610 specify the size of a symbol, without any arguments. On systems 30611 that use ELF, the default (in `config/elfos.h') is `"\t.size\t"'; 30612 on other systems, the default is not to define this macro. 30613 30614 Define this macro only if it is correct to use the default 30615 definitions of `ASM_OUTPUT_SIZE_DIRECTIVE' and 30616 `ASM_OUTPUT_MEASURED_SIZE' for your system. If you need your own 30617 custom definitions of those macros, or if you do not need explicit 30618 symbol sizes at all, do not define this macro. 30619 30620 -- Macro: ASM_OUTPUT_SIZE_DIRECTIVE (STREAM, NAME, SIZE) 30621 A C statement (sans semicolon) to output to the stdio stream 30622 STREAM a directive telling the assembler that the size of the 30623 symbol NAME is SIZE. SIZE is a `HOST_WIDE_INT'. If you define 30624 `SIZE_ASM_OP', a default definition of this macro is provided. 30625 30626 -- Macro: ASM_OUTPUT_MEASURED_SIZE (STREAM, NAME) 30627 A C statement (sans semicolon) to output to the stdio stream 30628 STREAM a directive telling the assembler to calculate the size of 30629 the symbol NAME by subtracting its address from the current 30630 address. 30631 30632 If you define `SIZE_ASM_OP', a default definition of this macro is 30633 provided. The default assumes that the assembler recognizes a 30634 special `.' symbol as referring to the current address, and can 30635 calculate the difference between this and another symbol. If your 30636 assembler does not recognize `.' or cannot do calculations with 30637 it, you will need to redefine `ASM_OUTPUT_MEASURED_SIZE' to use 30638 some other technique. 30639 30640 -- Macro: TYPE_ASM_OP 30641 A C string containing the appropriate assembler directive to 30642 specify the type of a symbol, without any arguments. On systems 30643 that use ELF, the default (in `config/elfos.h') is `"\t.type\t"'; 30644 on other systems, the default is not to define this macro. 30645 30646 Define this macro only if it is correct to use the default 30647 definition of `ASM_OUTPUT_TYPE_DIRECTIVE' for your system. If you 30648 need your own custom definition of this macro, or if you do not 30649 need explicit symbol types at all, do not define this macro. 30650 30651 -- Macro: TYPE_OPERAND_FMT 30652 A C string which specifies (using `printf' syntax) the format of 30653 the second operand to `TYPE_ASM_OP'. On systems that use ELF, the 30654 default (in `config/elfos.h') is `"@%s"'; on other systems, the 30655 default is not to define this macro. 30656 30657 Define this macro only if it is correct to use the default 30658 definition of `ASM_OUTPUT_TYPE_DIRECTIVE' for your system. If you 30659 need your own custom definition of this macro, or if you do not 30660 need explicit symbol types at all, do not define this macro. 30661 30662 -- Macro: ASM_OUTPUT_TYPE_DIRECTIVE (STREAM, TYPE) 30663 A C statement (sans semicolon) to output to the stdio stream 30664 STREAM a directive telling the assembler that the type of the 30665 symbol NAME is TYPE. TYPE is a C string; currently, that string 30666 is always either `"function"' or `"object"', but you should not 30667 count on this. 30668 30669 If you define `TYPE_ASM_OP' and `TYPE_OPERAND_FMT', a default 30670 definition of this macro is provided. 30671 30672 -- Macro: ASM_DECLARE_FUNCTION_NAME (STREAM, NAME, DECL) 30673 A C statement (sans semicolon) to output to the stdio stream 30674 STREAM any text necessary for declaring the name NAME of a 30675 function which is being defined. This macro is responsible for 30676 outputting the label definition (perhaps using 30677 `ASM_OUTPUT_LABEL'). The argument DECL is the `FUNCTION_DECL' 30678 tree node representing the function. 30679 30680 If this macro is not defined, then the function name is defined in 30681 the usual manner as a label (by means of `ASM_OUTPUT_LABEL'). 30682 30683 You may wish to use `ASM_OUTPUT_TYPE_DIRECTIVE' in the definition 30684 of this macro. 30685 30686 -- Macro: ASM_DECLARE_FUNCTION_SIZE (STREAM, NAME, DECL) 30687 A C statement (sans semicolon) to output to the stdio stream 30688 STREAM any text necessary for declaring the size of a function 30689 which is being defined. The argument NAME is the name of the 30690 function. The argument DECL is the `FUNCTION_DECL' tree node 30691 representing the function. 30692 30693 If this macro is not defined, then the function size is not 30694 defined. 30695 30696 You may wish to use `ASM_OUTPUT_MEASURED_SIZE' in the definition 30697 of this macro. 30698 30699 -- Macro: ASM_DECLARE_OBJECT_NAME (STREAM, NAME, DECL) 30700 A C statement (sans semicolon) to output to the stdio stream 30701 STREAM any text necessary for declaring the name NAME of an 30702 initialized variable which is being defined. This macro must 30703 output the label definition (perhaps using `ASM_OUTPUT_LABEL'). 30704 The argument DECL is the `VAR_DECL' tree node representing the 30705 variable. 30706 30707 If this macro is not defined, then the variable name is defined in 30708 the usual manner as a label (by means of `ASM_OUTPUT_LABEL'). 30709 30710 You may wish to use `ASM_OUTPUT_TYPE_DIRECTIVE' and/or 30711 `ASM_OUTPUT_SIZE_DIRECTIVE' in the definition of this macro. 30712 30713 -- Macro: ASM_DECLARE_CONSTANT_NAME (STREAM, NAME, EXP, SIZE) 30714 A C statement (sans semicolon) to output to the stdio stream 30715 STREAM any text necessary for declaring the name NAME of a 30716 constant which is being defined. This macro is responsible for 30717 outputting the label definition (perhaps using 30718 `ASM_OUTPUT_LABEL'). The argument EXP is the value of the 30719 constant, and SIZE is the size of the constant in bytes. NAME 30720 will be an internal label. 30721 30722 If this macro is not defined, then the NAME is defined in the 30723 usual manner as a label (by means of `ASM_OUTPUT_LABEL'). 30724 30725 You may wish to use `ASM_OUTPUT_TYPE_DIRECTIVE' in the definition 30726 of this macro. 30727 30728 -- Macro: ASM_DECLARE_REGISTER_GLOBAL (STREAM, DECL, REGNO, NAME) 30729 A C statement (sans semicolon) to output to the stdio stream 30730 STREAM any text necessary for claiming a register REGNO for a 30731 global variable DECL with name NAME. 30732 30733 If you don't define this macro, that is equivalent to defining it 30734 to do nothing. 30735 30736 -- Macro: ASM_FINISH_DECLARE_OBJECT (STREAM, DECL, TOPLEVEL, ATEND) 30737 A C statement (sans semicolon) to finish up declaring a variable 30738 name once the compiler has processed its initializer fully and 30739 thus has had a chance to determine the size of an array when 30740 controlled by an initializer. This is used on systems where it's 30741 necessary to declare something about the size of the object. 30742 30743 If you don't define this macro, that is equivalent to defining it 30744 to do nothing. 30745 30746 You may wish to use `ASM_OUTPUT_SIZE_DIRECTIVE' and/or 30747 `ASM_OUTPUT_MEASURED_SIZE' in the definition of this macro. 30748 30749 -- Target Hook: void TARGET_ASM_GLOBALIZE_LABEL (FILE *STREAM, const 30750 char *NAME) 30751 This target hook is a function to output to the stdio stream 30752 STREAM some commands that will make the label NAME global; that 30753 is, available for reference from other files. 30754 30755 The default implementation relies on a proper definition of 30756 `GLOBAL_ASM_OP'. 30757 30758 -- Target Hook: void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *STREAM, 30759 tree DECL) 30760 This target hook is a function to output to the stdio stream 30761 STREAM some commands that will make the name associated with DECL 30762 global; that is, available for reference from other files. 30763 30764 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL 30765 target hook. 30766 30767 -- Macro: ASM_WEAKEN_LABEL (STREAM, NAME) 30768 A C statement (sans semicolon) to output to the stdio stream 30769 STREAM some commands that will make the label NAME weak; that is, 30770 available for reference from other files but only used if no other 30771 definition is available. Use the expression `assemble_name 30772 (STREAM, NAME)' to output the name itself; before and after that, 30773 output the additional assembler syntax for making that name weak, 30774 and a newline. 30775 30776 If you don't define this macro or `ASM_WEAKEN_DECL', GCC will not 30777 support weak symbols and you should not define the `SUPPORTS_WEAK' 30778 macro. 30779 30780 -- Macro: ASM_WEAKEN_DECL (STREAM, DECL, NAME, VALUE) 30781 Combines (and replaces) the function of `ASM_WEAKEN_LABEL' and 30782 `ASM_OUTPUT_WEAK_ALIAS', allowing access to the associated function 30783 or variable decl. If VALUE is not `NULL', this C statement should 30784 output to the stdio stream STREAM assembler code which defines 30785 (equates) the weak symbol NAME to have the value VALUE. If VALUE 30786 is `NULL', it should output commands to make NAME weak. 30787 30788 -- Macro: ASM_OUTPUT_WEAKREF (STREAM, DECL, NAME, VALUE) 30789 Outputs a directive that enables NAME to be used to refer to 30790 symbol VALUE with weak-symbol semantics. `decl' is the 30791 declaration of `name'. 30792 30793 -- Macro: SUPPORTS_WEAK 30794 A C expression which evaluates to true if the target supports weak 30795 symbols. 30796 30797 If you don't define this macro, `defaults.h' provides a default 30798 definition. If either `ASM_WEAKEN_LABEL' or `ASM_WEAKEN_DECL' is 30799 defined, the default definition is `1'; otherwise, it is `0'. 30800 Define this macro if you want to control weak symbol support with 30801 a compiler flag such as `-melf'. 30802 30803 -- Macro: MAKE_DECL_ONE_ONLY (DECL) 30804 A C statement (sans semicolon) to mark DECL to be emitted as a 30805 public symbol such that extra copies in multiple translation units 30806 will be discarded by the linker. Define this macro if your object 30807 file format provides support for this concept, such as the `COMDAT' 30808 section flags in the Microsoft Windows PE/COFF format, and this 30809 support requires changes to DECL, such as putting it in a separate 30810 section. 30811 30812 -- Macro: SUPPORTS_ONE_ONLY 30813 A C expression which evaluates to true if the target supports 30814 one-only semantics. 30815 30816 If you don't define this macro, `varasm.c' provides a default 30817 definition. If `MAKE_DECL_ONE_ONLY' is defined, the default 30818 definition is `1'; otherwise, it is `0'. Define this macro if you 30819 want to control one-only symbol support with a compiler flag, or if 30820 setting the `DECL_ONE_ONLY' flag is enough to mark a declaration to 30821 be emitted as one-only. 30822 30823 -- Target Hook: void TARGET_ASM_ASSEMBLE_VISIBILITY (tree DECL, const 30824 char *VISIBILITY) 30825 This target hook is a function to output to ASM_OUT_FILE some 30826 commands that will make the symbol(s) associated with DECL have 30827 hidden, protected or internal visibility as specified by 30828 VISIBILITY. 30829 30830 -- Macro: TARGET_WEAK_NOT_IN_ARCHIVE_TOC 30831 A C expression that evaluates to true if the target's linker 30832 expects that weak symbols do not appear in a static archive's 30833 table of contents. The default is `0'. 30834 30835 Leaving weak symbols out of an archive's table of contents means 30836 that, if a symbol will only have a definition in one translation 30837 unit and will have undefined references from other translation 30838 units, that symbol should not be weak. Defining this macro to be 30839 nonzero will thus have the effect that certain symbols that would 30840 normally be weak (explicit template instantiations, and vtables 30841 for polymorphic classes with noninline key methods) will instead 30842 be nonweak. 30843 30844 The C++ ABI requires this macro to be zero. Define this macro for 30845 targets where full C++ ABI compliance is impossible and where 30846 linker restrictions require weak symbols to be left out of a 30847 static archive's table of contents. 30848 30849 -- Macro: ASM_OUTPUT_EXTERNAL (STREAM, DECL, NAME) 30850 A C statement (sans semicolon) to output to the stdio stream 30851 STREAM any text necessary for declaring the name of an external 30852 symbol named NAME which is referenced in this compilation but not 30853 defined. The value of DECL is the tree node for the declaration. 30854 30855 This macro need not be defined if it does not need to output 30856 anything. The GNU assembler and most Unix assemblers don't 30857 require anything. 30858 30859 -- Target Hook: void TARGET_ASM_EXTERNAL_LIBCALL (rtx SYMREF) 30860 This target hook is a function to output to ASM_OUT_FILE an 30861 assembler pseudo-op to declare a library function name external. 30862 The name of the library function is given by SYMREF, which is a 30863 `symbol_ref'. 30864 30865 -- Target Hook: void TARGET_ASM_MARK_DECL_PRESERVED (tree DECL) 30866 This target hook is a function to output to ASM_OUT_FILE an 30867 assembler directive to annotate used symbol. Darwin target use 30868 .no_dead_code_strip directive. 30869 30870 -- Macro: ASM_OUTPUT_LABELREF (STREAM, NAME) 30871 A C statement (sans semicolon) to output to the stdio stream 30872 STREAM a reference in assembler syntax to a label named NAME. 30873 This should add `_' to the front of the name, if that is customary 30874 on your operating system, as it is in most Berkeley Unix systems. 30875 This macro is used in `assemble_name'. 30876 30877 -- Macro: ASM_OUTPUT_SYMBOL_REF (STREAM, SYM) 30878 A C statement (sans semicolon) to output a reference to 30879 `SYMBOL_REF' SYM. If not defined, `assemble_name' will be used to 30880 output the name of the symbol. This macro may be used to modify 30881 the way a symbol is referenced depending on information encoded by 30882 `TARGET_ENCODE_SECTION_INFO'. 30883 30884 -- Macro: ASM_OUTPUT_LABEL_REF (STREAM, BUF) 30885 A C statement (sans semicolon) to output a reference to BUF, the 30886 result of `ASM_GENERATE_INTERNAL_LABEL'. If not defined, 30887 `assemble_name' will be used to output the name of the symbol. 30888 This macro is not used by `output_asm_label', or the `%l' 30889 specifier that calls it; the intention is that this macro should 30890 be set when it is necessary to output a label differently when its 30891 address is being taken. 30892 30893 -- Target Hook: void TARGET_ASM_INTERNAL_LABEL (FILE *STREAM, const 30894 char *PREFIX, unsigned long LABELNO) 30895 A function to output to the stdio stream STREAM a label whose name 30896 is made from the string PREFIX and the number LABELNO. 30897 30898 It is absolutely essential that these labels be distinct from the 30899 labels used for user-level functions and variables. Otherwise, 30900 certain programs will have name conflicts with internal labels. 30901 30902 It is desirable to exclude internal labels from the symbol table 30903 of the object file. Most assemblers have a naming convention for 30904 labels that should be excluded; on many systems, the letter `L' at 30905 the beginning of a label has this effect. You should find out what 30906 convention your system uses, and follow it. 30907 30908 The default version of this function utilizes 30909 `ASM_GENERATE_INTERNAL_LABEL'. 30910 30911 -- Macro: ASM_OUTPUT_DEBUG_LABEL (STREAM, PREFIX, NUM) 30912 A C statement to output to the stdio stream STREAM a debug info 30913 label whose name is made from the string PREFIX and the number 30914 NUM. This is useful for VLIW targets, where debug info labels may 30915 need to be treated differently than branch target labels. On some 30916 systems, branch target labels must be at the beginning of 30917 instruction bundles, but debug info labels can occur in the middle 30918 of instruction bundles. 30919 30920 If this macro is not defined, then 30921 `(*targetm.asm_out.internal_label)' will be used. 30922 30923 -- Macro: ASM_GENERATE_INTERNAL_LABEL (STRING, PREFIX, NUM) 30924 A C statement to store into the string STRING a label whose name 30925 is made from the string PREFIX and the number NUM. 30926 30927 This string, when output subsequently by `assemble_name', should 30928 produce the output that `(*targetm.asm_out.internal_label)' would 30929 produce with the same PREFIX and NUM. 30930 30931 If the string begins with `*', then `assemble_name' will output 30932 the rest of the string unchanged. It is often convenient for 30933 `ASM_GENERATE_INTERNAL_LABEL' to use `*' in this way. If the 30934 string doesn't start with `*', then `ASM_OUTPUT_LABELREF' gets to 30935 output the string, and may change it. (Of course, 30936 `ASM_OUTPUT_LABELREF' is also part of your machine description, so 30937 you should know what it does on your machine.) 30938 30939 -- Macro: ASM_FORMAT_PRIVATE_NAME (OUTVAR, NAME, NUMBER) 30940 A C expression to assign to OUTVAR (which is a variable of type 30941 `char *') a newly allocated string made from the string NAME and 30942 the number NUMBER, with some suitable punctuation added. Use 30943 `alloca' to get space for the string. 30944 30945 The string will be used as an argument to `ASM_OUTPUT_LABELREF' to 30946 produce an assembler label for an internal static variable whose 30947 name is NAME. Therefore, the string must be such as to result in 30948 valid assembler code. The argument NUMBER is different each time 30949 this macro is executed; it prevents conflicts between 30950 similarly-named internal static variables in different scopes. 30951 30952 Ideally this string should not be a valid C identifier, to prevent 30953 any conflict with the user's own symbols. Most assemblers allow 30954 periods or percent signs in assembler symbols; putting at least 30955 one of these between the name and the number will suffice. 30956 30957 If this macro is not defined, a default definition will be provided 30958 which is correct for most systems. 30959 30960 -- Macro: ASM_OUTPUT_DEF (STREAM, NAME, VALUE) 30961 A C statement to output to the stdio stream STREAM assembler code 30962 which defines (equates) the symbol NAME to have the value VALUE. 30963 30964 If `SET_ASM_OP' is defined, a default definition is provided which 30965 is correct for most systems. 30966 30967 -- Macro: ASM_OUTPUT_DEF_FROM_DECLS (STREAM, DECL_OF_NAME, 30968 DECL_OF_VALUE) 30969 A C statement to output to the stdio stream STREAM assembler code 30970 which defines (equates) the symbol whose tree node is DECL_OF_NAME 30971 to have the value of the tree node DECL_OF_VALUE. This macro will 30972 be used in preference to `ASM_OUTPUT_DEF' if it is defined and if 30973 the tree nodes are available. 30974 30975 If `SET_ASM_OP' is defined, a default definition is provided which 30976 is correct for most systems. 30977 30978 -- Macro: TARGET_DEFERRED_OUTPUT_DEFS (DECL_OF_NAME, DECL_OF_VALUE) 30979 A C statement that evaluates to true if the assembler code which 30980 defines (equates) the symbol whose tree node is DECL_OF_NAME to 30981 have the value of the tree node DECL_OF_VALUE should be emitted 30982 near the end of the current compilation unit. The default is to 30983 not defer output of defines. This macro affects defines output by 30984 `ASM_OUTPUT_DEF' and `ASM_OUTPUT_DEF_FROM_DECLS'. 30985 30986 -- Macro: ASM_OUTPUT_WEAK_ALIAS (STREAM, NAME, VALUE) 30987 A C statement to output to the stdio stream STREAM assembler code 30988 which defines (equates) the weak symbol NAME to have the value 30989 VALUE. If VALUE is `NULL', it defines NAME as an undefined weak 30990 symbol. 30991 30992 Define this macro if the target only supports weak aliases; define 30993 `ASM_OUTPUT_DEF' instead if possible. 30994 30995 -- Macro: OBJC_GEN_METHOD_LABEL (BUF, IS_INST, CLASS_NAME, CAT_NAME, 30996 SEL_NAME) 30997 Define this macro to override the default assembler names used for 30998 Objective-C methods. 30999 31000 The default name is a unique method number followed by the name of 31001 the class (e.g. `_1_Foo'). For methods in categories, the name of 31002 the category is also included in the assembler name (e.g. 31003 `_1_Foo_Bar'). 31004 31005 These names are safe on most systems, but make debugging difficult 31006 since the method's selector is not present in the name. 31007 Therefore, particular systems define other ways of computing names. 31008 31009 BUF is an expression of type `char *' which gives you a buffer in 31010 which to store the name; its length is as long as CLASS_NAME, 31011 CAT_NAME and SEL_NAME put together, plus 50 characters extra. 31012 31013 The argument IS_INST specifies whether the method is an instance 31014 method or a class method; CLASS_NAME is the name of the class; 31015 CAT_NAME is the name of the category (or `NULL' if the method is 31016 not in a category); and SEL_NAME is the name of the selector. 31017 31018 On systems where the assembler can handle quoted names, you can 31019 use this macro to provide more human-readable names. 31020 31021 -- Macro: ASM_DECLARE_CLASS_REFERENCE (STREAM, NAME) 31022 A C statement (sans semicolon) to output to the stdio stream 31023 STREAM commands to declare that the label NAME is an Objective-C 31024 class reference. This is only needed for targets whose linkers 31025 have special support for NeXT-style runtimes. 31026 31027 -- Macro: ASM_DECLARE_UNRESOLVED_REFERENCE (STREAM, NAME) 31028 A C statement (sans semicolon) to output to the stdio stream 31029 STREAM commands to declare that the label NAME is an unresolved 31030 Objective-C class reference. This is only needed for targets 31031 whose linkers have special support for NeXT-style runtimes. 31032 31033 31034 File: gccint.info, Node: Initialization, Next: Macros for Initialization, Prev: Label Output, Up: Assembler Format 31035 31036 17.21.5 How Initialization Functions Are Handled 31037 ------------------------------------------------ 31038 31039 The compiled code for certain languages includes "constructors" (also 31040 called "initialization routines")--functions to initialize data in the 31041 program when the program is started. These functions need to be called 31042 before the program is "started"--that is to say, before `main' is 31043 called. 31044 31045 Compiling some languages generates "destructors" (also called 31046 "termination routines") that should be called when the program 31047 terminates. 31048 31049 To make the initialization and termination functions work, the compiler 31050 must output something in the assembler code to cause those functions to 31051 be called at the appropriate time. When you port the compiler to a new 31052 system, you need to specify how to do this. 31053 31054 There are two major ways that GCC currently supports the execution of 31055 initialization and termination functions. Each way has two variants. 31056 Much of the structure is common to all four variations. 31057 31058 The linker must build two lists of these functions--a list of 31059 initialization functions, called `__CTOR_LIST__', and a list of 31060 termination functions, called `__DTOR_LIST__'. 31061 31062 Each list always begins with an ignored function pointer (which may 31063 hold 0, -1, or a count of the function pointers after it, depending on 31064 the environment). This is followed by a series of zero or more function 31065 pointers to constructors (or destructors), followed by a function 31066 pointer containing zero. 31067 31068 Depending on the operating system and its executable file format, 31069 either `crtstuff.c' or `libgcc2.c' traverses these lists at startup 31070 time and exit time. Constructors are called in reverse order of the 31071 list; destructors in forward order. 31072 31073 The best way to handle static constructors works only for object file 31074 formats which provide arbitrarily-named sections. A section is set 31075 aside for a list of constructors, and another for a list of destructors. 31076 Traditionally these are called `.ctors' and `.dtors'. Each object file 31077 that defines an initialization function also puts a word in the 31078 constructor section to point to that function. The linker accumulates 31079 all these words into one contiguous `.ctors' section. Termination 31080 functions are handled similarly. 31081 31082 This method will be chosen as the default by `target-def.h' if 31083 `TARGET_ASM_NAMED_SECTION' is defined. A target that does not support 31084 arbitrary sections, but does support special designated constructor and 31085 destructor sections may define `CTORS_SECTION_ASM_OP' and 31086 `DTORS_SECTION_ASM_OP' to achieve the same effect. 31087 31088 When arbitrary sections are available, there are two variants, 31089 depending upon how the code in `crtstuff.c' is called. On systems that 31090 support a ".init" section which is executed at program startup, parts 31091 of `crtstuff.c' are compiled into that section. The program is linked 31092 by the `gcc' driver like this: 31093 31094 ld -o OUTPUT_FILE crti.o crtbegin.o ... -lgcc crtend.o crtn.o 31095 31096 The prologue of a function (`__init') appears in the `.init' section 31097 of `crti.o'; the epilogue appears in `crtn.o'. Likewise for the 31098 function `__fini' in the ".fini" section. Normally these files are 31099 provided by the operating system or by the GNU C library, but are 31100 provided by GCC for a few targets. 31101 31102 The objects `crtbegin.o' and `crtend.o' are (for most targets) 31103 compiled from `crtstuff.c'. They contain, among other things, code 31104 fragments within the `.init' and `.fini' sections that branch to 31105 routines in the `.text' section. The linker will pull all parts of a 31106 section together, which results in a complete `__init' function that 31107 invokes the routines we need at startup. 31108 31109 To use this variant, you must define the `INIT_SECTION_ASM_OP' macro 31110 properly. 31111 31112 If no init section is available, when GCC compiles any function called 31113 `main' (or more accurately, any function designated as a program entry 31114 point by the language front end calling `expand_main_function'), it 31115 inserts a procedure call to `__main' as the first executable code after 31116 the function prologue. The `__main' function is defined in `libgcc2.c' 31117 and runs the global constructors. 31118 31119 In file formats that don't support arbitrary sections, there are again 31120 two variants. In the simplest variant, the GNU linker (GNU `ld') and 31121 an `a.out' format must be used. In this case, `TARGET_ASM_CONSTRUCTOR' 31122 is defined to produce a `.stabs' entry of type `N_SETT', referencing 31123 the name `__CTOR_LIST__', and with the address of the void function 31124 containing the initialization code as its value. The GNU linker 31125 recognizes this as a request to add the value to a "set"; the values 31126 are accumulated, and are eventually placed in the executable as a 31127 vector in the format described above, with a leading (ignored) count 31128 and a trailing zero element. `TARGET_ASM_DESTRUCTOR' is handled 31129 similarly. Since no init section is available, the absence of 31130 `INIT_SECTION_ASM_OP' causes the compilation of `main' to call `__main' 31131 as above, starting the initialization process. 31132 31133 The last variant uses neither arbitrary sections nor the GNU linker. 31134 This is preferable when you want to do dynamic linking and when using 31135 file formats which the GNU linker does not support, such as `ECOFF'. In 31136 this case, `TARGET_HAVE_CTORS_DTORS' is false, initialization and 31137 termination functions are recognized simply by their names. This 31138 requires an extra program in the linkage step, called `collect2'. This 31139 program pretends to be the linker, for use with GCC; it does its job by 31140 running the ordinary linker, but also arranges to include the vectors of 31141 initialization and termination functions. These functions are called 31142 via `__main' as described above. In order to use this method, 31143 `use_collect2' must be defined in the target in `config.gcc'. 31144 31145 The following section describes the specific macros that control and 31146 customize the handling of initialization and termination functions. 31147 31148 31149 File: gccint.info, Node: Macros for Initialization, Next: Instruction Output, Prev: Initialization, Up: Assembler Format 31150 31151 17.21.6 Macros Controlling Initialization Routines 31152 -------------------------------------------------- 31153 31154 Here are the macros that control how the compiler handles initialization 31155 and termination functions: 31156 31157 -- Macro: INIT_SECTION_ASM_OP 31158 If defined, a C string constant, including spacing, for the 31159 assembler operation to identify the following data as 31160 initialization code. If not defined, GCC will assume such a 31161 section does not exist. When you are using special sections for 31162 initialization and termination functions, this macro also controls 31163 how `crtstuff.c' and `libgcc2.c' arrange to run the initialization 31164 functions. 31165 31166 -- Macro: HAS_INIT_SECTION 31167 If defined, `main' will not call `__main' as described above. 31168 This macro should be defined for systems that control start-up code 31169 on a symbol-by-symbol basis, such as OSF/1, and should not be 31170 defined explicitly for systems that support `INIT_SECTION_ASM_OP'. 31171 31172 -- Macro: LD_INIT_SWITCH 31173 If defined, a C string constant for a switch that tells the linker 31174 that the following symbol is an initialization routine. 31175 31176 -- Macro: LD_FINI_SWITCH 31177 If defined, a C string constant for a switch that tells the linker 31178 that the following symbol is a finalization routine. 31179 31180 -- Macro: COLLECT_SHARED_INIT_FUNC (STREAM, FUNC) 31181 If defined, a C statement that will write a function that can be 31182 automatically called when a shared library is loaded. The function 31183 should call FUNC, which takes no arguments. If not defined, and 31184 the object format requires an explicit initialization function, 31185 then a function called `_GLOBAL__DI' will be generated. 31186 31187 This function and the following one are used by collect2 when 31188 linking a shared library that needs constructors or destructors, 31189 or has DWARF2 exception tables embedded in the code. 31190 31191 -- Macro: COLLECT_SHARED_FINI_FUNC (STREAM, FUNC) 31192 If defined, a C statement that will write a function that can be 31193 automatically called when a shared library is unloaded. The 31194 function should call FUNC, which takes no arguments. If not 31195 defined, and the object format requires an explicit finalization 31196 function, then a function called `_GLOBAL__DD' will be generated. 31197 31198 -- Macro: INVOKE__main 31199 If defined, `main' will call `__main' despite the presence of 31200 `INIT_SECTION_ASM_OP'. This macro should be defined for systems 31201 where the init section is not actually run automatically, but is 31202 still useful for collecting the lists of constructors and 31203 destructors. 31204 31205 -- Macro: SUPPORTS_INIT_PRIORITY 31206 If nonzero, the C++ `init_priority' attribute is supported and the 31207 compiler should emit instructions to control the order of 31208 initialization of objects. If zero, the compiler will issue an 31209 error message upon encountering an `init_priority' attribute. 31210 31211 -- Target Hook: bool TARGET_HAVE_CTORS_DTORS 31212 This value is true if the target supports some "native" method of 31213 collecting constructors and destructors to be run at startup and 31214 exit. It is false if we must use `collect2'. 31215 31216 -- Target Hook: void TARGET_ASM_CONSTRUCTOR (rtx SYMBOL, int PRIORITY) 31217 If defined, a function that outputs assembler code to arrange to 31218 call the function referenced by SYMBOL at initialization time. 31219 31220 Assume that SYMBOL is a `SYMBOL_REF' for a function taking no 31221 arguments and with no return value. If the target supports 31222 initialization priorities, PRIORITY is a value between 0 and 31223 `MAX_INIT_PRIORITY'; otherwise it must be `DEFAULT_INIT_PRIORITY'. 31224 31225 If this macro is not defined by the target, a suitable default will 31226 be chosen if (1) the target supports arbitrary section names, (2) 31227 the target defines `CTORS_SECTION_ASM_OP', or (3) `USE_COLLECT2' 31228 is not defined. 31229 31230 -- Target Hook: void TARGET_ASM_DESTRUCTOR (rtx SYMBOL, int PRIORITY) 31231 This is like `TARGET_ASM_CONSTRUCTOR' but used for termination 31232 functions rather than initialization functions. 31233 31234 If `TARGET_HAVE_CTORS_DTORS' is true, the initialization routine 31235 generated for the generated object file will have static linkage. 31236 31237 If your system uses `collect2' as the means of processing 31238 constructors, then that program normally uses `nm' to scan an object 31239 file for constructor functions to be called. 31240 31241 On certain kinds of systems, you can define this macro to make 31242 `collect2' work faster (and, in some cases, make it work at all): 31243 31244 -- Macro: OBJECT_FORMAT_COFF 31245 Define this macro if the system uses COFF (Common Object File 31246 Format) object files, so that `collect2' can assume this format 31247 and scan object files directly for dynamic constructor/destructor 31248 functions. 31249 31250 This macro is effective only in a native compiler; `collect2' as 31251 part of a cross compiler always uses `nm' for the target machine. 31252 31253 -- Macro: REAL_NM_FILE_NAME 31254 Define this macro as a C string constant containing the file name 31255 to use to execute `nm'. The default is to search the path 31256 normally for `nm'. 31257 31258 If your system supports shared libraries and has a program to list 31259 the dynamic dependencies of a given library or executable, you can 31260 define these macros to enable support for running initialization 31261 and termination functions in shared libraries: 31262 31263 -- Macro: LDD_SUFFIX 31264 Define this macro to a C string constant containing the name of 31265 the program which lists dynamic dependencies, like `"ldd"' under 31266 SunOS 4. 31267 31268 -- Macro: PARSE_LDD_OUTPUT (PTR) 31269 Define this macro to be C code that extracts filenames from the 31270 output of the program denoted by `LDD_SUFFIX'. PTR is a variable 31271 of type `char *' that points to the beginning of a line of output 31272 from `LDD_SUFFIX'. If the line lists a dynamic dependency, the 31273 code must advance PTR to the beginning of the filename on that 31274 line. Otherwise, it must set PTR to `NULL'. 31275 31276 -- Macro: SHLIB_SUFFIX 31277 Define this macro to a C string constant containing the default 31278 shared library extension of the target (e.g., `".so"'). `collect2' 31279 strips version information after this suffix when generating global 31280 constructor and destructor names. This define is only needed on 31281 targets that use `collect2' to process constructors and 31282 destructors. 31283 31284 31285 File: gccint.info, Node: Instruction Output, Next: Dispatch Tables, Prev: Macros for Initialization, Up: Assembler Format 31286 31287 17.21.7 Output of Assembler Instructions 31288 ---------------------------------------- 31289 31290 This describes assembler instruction output. 31291 31292 -- Macro: REGISTER_NAMES 31293 A C initializer containing the assembler's names for the machine 31294 registers, each one as a C string constant. This is what 31295 translates register numbers in the compiler into assembler 31296 language. 31297 31298 -- Macro: ADDITIONAL_REGISTER_NAMES 31299 If defined, a C initializer for an array of structures containing 31300 a name and a register number. This macro defines additional names 31301 for hard registers, thus allowing the `asm' option in declarations 31302 to refer to registers using alternate names. 31303 31304 -- Macro: ASM_OUTPUT_OPCODE (STREAM, PTR) 31305 Define this macro if you are using an unusual assembler that 31306 requires different names for the machine instructions. 31307 31308 The definition is a C statement or statements which output an 31309 assembler instruction opcode to the stdio stream STREAM. The 31310 macro-operand PTR is a variable of type `char *' which points to 31311 the opcode name in its "internal" form--the form that is written 31312 in the machine description. The definition should output the 31313 opcode name to STREAM, performing any translation you desire, and 31314 increment the variable PTR to point at the end of the opcode so 31315 that it will not be output twice. 31316 31317 In fact, your macro definition may process less than the entire 31318 opcode name, or more than the opcode name; but if you want to 31319 process text that includes `%'-sequences to substitute operands, 31320 you must take care of the substitution yourself. Just be sure to 31321 increment PTR over whatever text should not be output normally. 31322 31323 If you need to look at the operand values, they can be found as the 31324 elements of `recog_data.operand'. 31325 31326 If the macro definition does nothing, the instruction is output in 31327 the usual way. 31328 31329 -- Macro: FINAL_PRESCAN_INSN (INSN, OPVEC, NOPERANDS) 31330 If defined, a C statement to be executed just prior to the output 31331 of assembler code for INSN, to modify the extracted operands so 31332 they will be output differently. 31333 31334 Here the argument OPVEC is the vector containing the operands 31335 extracted from INSN, and NOPERANDS is the number of elements of 31336 the vector which contain meaningful data for this insn. The 31337 contents of this vector are what will be used to convert the insn 31338 template into assembler code, so you can change the assembler 31339 output by changing the contents of the vector. 31340 31341 This macro is useful when various assembler syntaxes share a single 31342 file of instruction patterns; by defining this macro differently, 31343 you can cause a large class of instructions to be output 31344 differently (such as with rearranged operands). Naturally, 31345 variations in assembler syntax affecting individual insn patterns 31346 ought to be handled by writing conditional output routines in 31347 those patterns. 31348 31349 If this macro is not defined, it is equivalent to a null statement. 31350 31351 -- Macro: PRINT_OPERAND (STREAM, X, CODE) 31352 A C compound statement to output to stdio stream STREAM the 31353 assembler syntax for an instruction operand X. X is an RTL 31354 expression. 31355 31356 CODE is a value that can be used to specify one of several ways of 31357 printing the operand. It is used when identical operands must be 31358 printed differently depending on the context. CODE comes from the 31359 `%' specification that was used to request printing of the 31360 operand. If the specification was just `%DIGIT' then CODE is 0; 31361 if the specification was `%LTR DIGIT' then CODE is the ASCII code 31362 for LTR. 31363 31364 If X is a register, this macro should print the register's name. 31365 The names can be found in an array `reg_names' whose type is `char 31366 *[]'. `reg_names' is initialized from `REGISTER_NAMES'. 31367 31368 When the machine description has a specification `%PUNCT' (a `%' 31369 followed by a punctuation character), this macro is called with a 31370 null pointer for X and the punctuation character for CODE. 31371 31372 -- Macro: PRINT_OPERAND_PUNCT_VALID_P (CODE) 31373 A C expression which evaluates to true if CODE is a valid 31374 punctuation character for use in the `PRINT_OPERAND' macro. If 31375 `PRINT_OPERAND_PUNCT_VALID_P' is not defined, it means that no 31376 punctuation characters (except for the standard one, `%') are used 31377 in this way. 31378 31379 -- Macro: PRINT_OPERAND_ADDRESS (STREAM, X) 31380 A C compound statement to output to stdio stream STREAM the 31381 assembler syntax for an instruction operand that is a memory 31382 reference whose address is X. X is an RTL expression. 31383 31384 On some machines, the syntax for a symbolic address depends on the 31385 section that the address refers to. On these machines, define the 31386 hook `TARGET_ENCODE_SECTION_INFO' to store the information into the 31387 `symbol_ref', and then check for it here. *Note Assembler 31388 Format::. 31389 31390 -- Macro: DBR_OUTPUT_SEQEND (FILE) 31391 A C statement, to be executed after all slot-filler instructions 31392 have been output. If necessary, call `dbr_sequence_length' to 31393 determine the number of slots filled in a sequence (zero if not 31394 currently outputting a sequence), to decide how many no-ops to 31395 output, or whatever. 31396 31397 Don't define this macro if it has nothing to do, but it is helpful 31398 in reading assembly output if the extent of the delay sequence is 31399 made explicit (e.g. with white space). 31400 31401 Note that output routines for instructions with delay slots must be 31402 prepared to deal with not being output as part of a sequence (i.e. when 31403 the scheduling pass is not run, or when no slot fillers could be 31404 found.) The variable `final_sequence' is null when not processing a 31405 sequence, otherwise it contains the `sequence' rtx being output. 31406 31407 -- Macro: REGISTER_PREFIX 31408 -- Macro: LOCAL_LABEL_PREFIX 31409 -- Macro: USER_LABEL_PREFIX 31410 -- Macro: IMMEDIATE_PREFIX 31411 If defined, C string expressions to be used for the `%R', `%L', 31412 `%U', and `%I' options of `asm_fprintf' (see `final.c'). These 31413 are useful when a single `md' file must support multiple assembler 31414 formats. In that case, the various `tm.h' files can define these 31415 macros differently. 31416 31417 -- Macro: ASM_FPRINTF_EXTENSIONS (FILE, ARGPTR, FORMAT) 31418 If defined this macro should expand to a series of `case' 31419 statements which will be parsed inside the `switch' statement of 31420 the `asm_fprintf' function. This allows targets to define extra 31421 printf formats which may useful when generating their assembler 31422 statements. Note that uppercase letters are reserved for future 31423 generic extensions to asm_fprintf, and so are not available to 31424 target specific code. The output file is given by the parameter 31425 FILE. The varargs input pointer is ARGPTR and the rest of the 31426 format string, starting the character after the one that is being 31427 switched upon, is pointed to by FORMAT. 31428 31429 -- Macro: ASSEMBLER_DIALECT 31430 If your target supports multiple dialects of assembler language 31431 (such as different opcodes), define this macro as a C expression 31432 that gives the numeric index of the assembler language dialect to 31433 use, with zero as the first variant. 31434 31435 If this macro is defined, you may use constructs of the form 31436 `{option0|option1|option2...}' 31437 in the output templates of patterns (*note Output Template::) or 31438 in the first argument of `asm_fprintf'. This construct outputs 31439 `option0', `option1', `option2', etc., if the value of 31440 `ASSEMBLER_DIALECT' is zero, one, two, etc. Any special characters 31441 within these strings retain their usual meaning. If there are 31442 fewer alternatives within the braces than the value of 31443 `ASSEMBLER_DIALECT', the construct outputs nothing. 31444 31445 If you do not define this macro, the characters `{', `|' and `}' 31446 do not have any special meaning when used in templates or operands 31447 to `asm_fprintf'. 31448 31449 Define the macros `REGISTER_PREFIX', `LOCAL_LABEL_PREFIX', 31450 `USER_LABEL_PREFIX' and `IMMEDIATE_PREFIX' if you can express the 31451 variations in assembler language syntax with that mechanism. 31452 Define `ASSEMBLER_DIALECT' and use the `{option0|option1}' syntax 31453 if the syntax variant are larger and involve such things as 31454 different opcodes or operand order. 31455 31456 -- Macro: ASM_OUTPUT_REG_PUSH (STREAM, REGNO) 31457 A C expression to output to STREAM some assembler code which will 31458 push hard register number REGNO onto the stack. The code need not 31459 be optimal, since this macro is used only when profiling. 31460 31461 -- Macro: ASM_OUTPUT_REG_POP (STREAM, REGNO) 31462 A C expression to output to STREAM some assembler code which will 31463 pop hard register number REGNO off of the stack. The code need 31464 not be optimal, since this macro is used only when profiling. 31465 31466 31467 File: gccint.info, Node: Dispatch Tables, Next: Exception Region Output, Prev: Instruction Output, Up: Assembler Format 31468 31469 17.21.8 Output of Dispatch Tables 31470 --------------------------------- 31471 31472 This concerns dispatch tables. 31473 31474 -- Macro: ASM_OUTPUT_ADDR_DIFF_ELT (STREAM, BODY, VALUE, REL) 31475 A C statement to output to the stdio stream STREAM an assembler 31476 pseudo-instruction to generate a difference between two labels. 31477 VALUE and REL are the numbers of two internal labels. The 31478 definitions of these labels are output using 31479 `(*targetm.asm_out.internal_label)', and they must be printed in 31480 the same way here. For example, 31481 31482 fprintf (STREAM, "\t.word L%d-L%d\n", 31483 VALUE, REL) 31484 31485 You must provide this macro on machines where the addresses in a 31486 dispatch table are relative to the table's own address. If 31487 defined, GCC will also use this macro on all machines when 31488 producing PIC. BODY is the body of the `ADDR_DIFF_VEC'; it is 31489 provided so that the mode and flags can be read. 31490 31491 -- Macro: ASM_OUTPUT_ADDR_VEC_ELT (STREAM, VALUE) 31492 This macro should be provided on machines where the addresses in a 31493 dispatch table are absolute. 31494 31495 The definition should be a C statement to output to the stdio 31496 stream STREAM an assembler pseudo-instruction to generate a 31497 reference to a label. VALUE is the number of an internal label 31498 whose definition is output using 31499 `(*targetm.asm_out.internal_label)'. For example, 31500 31501 fprintf (STREAM, "\t.word L%d\n", VALUE) 31502 31503 -- Macro: ASM_OUTPUT_CASE_LABEL (STREAM, PREFIX, NUM, TABLE) 31504 Define this if the label before a jump-table needs to be output 31505 specially. The first three arguments are the same as for 31506 `(*targetm.asm_out.internal_label)'; the fourth argument is the 31507 jump-table which follows (a `jump_insn' containing an `addr_vec' 31508 or `addr_diff_vec'). 31509 31510 This feature is used on system V to output a `swbeg' statement for 31511 the table. 31512 31513 If this macro is not defined, these labels are output with 31514 `(*targetm.asm_out.internal_label)'. 31515 31516 -- Macro: ASM_OUTPUT_CASE_END (STREAM, NUM, TABLE) 31517 Define this if something special must be output at the end of a 31518 jump-table. The definition should be a C statement to be executed 31519 after the assembler code for the table is written. It should write 31520 the appropriate code to stdio stream STREAM. The argument TABLE 31521 is the jump-table insn, and NUM is the label-number of the 31522 preceding label. 31523 31524 If this macro is not defined, nothing special is output at the end 31525 of the jump-table. 31526 31527 -- Target Hook: void TARGET_ASM_EMIT_UNWIND_LABEL (STREAM, DECL, 31528 FOR_EH, EMPTY) 31529 This target hook emits a label at the beginning of each FDE. It 31530 should be defined on targets where FDEs need special labels, and it 31531 should write the appropriate label, for the FDE associated with the 31532 function declaration DECL, to the stdio stream STREAM. The third 31533 argument, FOR_EH, is a boolean: true if this is for an exception 31534 table. The fourth argument, EMPTY, is a boolean: true if this is 31535 a placeholder label for an omitted FDE. 31536 31537 The default is that FDEs are not given nonlocal labels. 31538 31539 -- Target Hook: void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (STREAM) 31540 This target hook emits a label at the beginning of the exception 31541 table. It should be defined on targets where it is desirable for 31542 the table to be broken up according to function. 31543 31544 The default is that no label is emitted. 31545 31546 -- Target Hook: void TARGET_UNWIND_EMIT (FILE * STREAM, rtx INSN) 31547 This target hook emits and assembly directives required to unwind 31548 the given instruction. This is only used when TARGET_UNWIND_INFO 31549 is set. 31550 31551 31552 File: gccint.info, Node: Exception Region Output, Next: Alignment Output, Prev: Dispatch Tables, Up: Assembler Format 31553 31554 17.21.9 Assembler Commands for Exception Regions 31555 ------------------------------------------------ 31556 31557 This describes commands marking the start and the end of an exception 31558 region. 31559 31560 -- Macro: EH_FRAME_SECTION_NAME 31561 If defined, a C string constant for the name of the section 31562 containing exception handling frame unwind information. If not 31563 defined, GCC will provide a default definition if the target 31564 supports named sections. `crtstuff.c' uses this macro to switch 31565 to the appropriate section. 31566 31567 You should define this symbol if your target supports DWARF 2 frame 31568 unwind information and the default definition does not work. 31569 31570 -- Macro: EH_FRAME_IN_DATA_SECTION 31571 If defined, DWARF 2 frame unwind information will be placed in the 31572 data section even though the target supports named sections. This 31573 might be necessary, for instance, if the system linker does garbage 31574 collection and sections cannot be marked as not to be collected. 31575 31576 Do not define this macro unless `TARGET_ASM_NAMED_SECTION' is also 31577 defined. 31578 31579 -- Macro: EH_TABLES_CAN_BE_READ_ONLY 31580 Define this macro to 1 if your target is such that no frame unwind 31581 information encoding used with non-PIC code will ever require a 31582 runtime relocation, but the linker may not support merging 31583 read-only and read-write sections into a single read-write section. 31584 31585 -- Macro: MASK_RETURN_ADDR 31586 An rtx used to mask the return address found via 31587 `RETURN_ADDR_RTX', so that it does not contain any extraneous set 31588 bits in it. 31589 31590 -- Macro: DWARF2_UNWIND_INFO 31591 Define this macro to 0 if your target supports DWARF 2 frame unwind 31592 information, but it does not yet work with exception handling. 31593 Otherwise, if your target supports this information (if it defines 31594 `INCOMING_RETURN_ADDR_RTX' and either `UNALIGNED_INT_ASM_OP' or 31595 `OBJECT_FORMAT_ELF'), GCC will provide a default definition of 1. 31596 31597 If `TARGET_UNWIND_INFO' is defined, the target specific unwinder 31598 will be used in all cases. Defining this macro will enable the 31599 generation of DWARF 2 frame debugging information. 31600 31601 If `TARGET_UNWIND_INFO' is not defined, and this macro is defined 31602 to 1, the DWARF 2 unwinder will be the default exception handling 31603 mechanism; otherwise, the `setjmp'/`longjmp'-based scheme will be 31604 used by default. 31605 31606 -- Macro: TARGET_UNWIND_INFO 31607 Define this macro if your target has ABI specified unwind tables. 31608 Usually these will be output by `TARGET_UNWIND_EMIT'. 31609 31610 -- Variable: Target Hook bool TARGET_UNWIND_TABLES_DEFAULT 31611 This variable should be set to `true' if the target ABI requires 31612 unwinding tables even when exceptions are not used. 31613 31614 -- Macro: MUST_USE_SJLJ_EXCEPTIONS 31615 This macro need only be defined if `DWARF2_UNWIND_INFO' is 31616 runtime-variable. In that case, `except.h' cannot correctly 31617 determine the corresponding definition of 31618 `MUST_USE_SJLJ_EXCEPTIONS', so the target must provide it directly. 31619 31620 -- Macro: DONT_USE_BUILTIN_SETJMP 31621 Define this macro to 1 if the `setjmp'/`longjmp'-based scheme 31622 should use the `setjmp'/`longjmp' functions from the C library 31623 instead of the `__builtin_setjmp'/`__builtin_longjmp' machinery. 31624 31625 -- Macro: DWARF_CIE_DATA_ALIGNMENT 31626 This macro need only be defined if the target might save registers 31627 in the function prologue at an offset to the stack pointer that is 31628 not aligned to `UNITS_PER_WORD'. The definition should be the 31629 negative minimum alignment if `STACK_GROWS_DOWNWARD' is defined, 31630 and the positive minimum alignment otherwise. *Note SDB and 31631 DWARF::. Only applicable if the target supports DWARF 2 frame 31632 unwind information. 31633 31634 -- Variable: Target Hook bool TARGET_TERMINATE_DW2_EH_FRAME_INFO 31635 Contains the value true if the target should add a zero word onto 31636 the end of a Dwarf-2 frame info section when used for exception 31637 handling. Default value is false if `EH_FRAME_SECTION_NAME' is 31638 defined, and true otherwise. 31639 31640 -- Target Hook: rtx TARGET_DWARF_REGISTER_SPAN (rtx REG) 31641 Given a register, this hook should return a parallel of registers 31642 to represent where to find the register pieces. Define this hook 31643 if the register and its mode are represented in Dwarf in 31644 non-contiguous locations, or if the register should be represented 31645 in more than one register in Dwarf. Otherwise, this hook should 31646 return `NULL_RTX'. If not defined, the default is to return 31647 `NULL_RTX'. 31648 31649 -- Target Hook: void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree ADDRESS) 31650 If some registers are represented in Dwarf-2 unwind information in 31651 multiple pieces, define this hook to fill in information about the 31652 sizes of those pieces in the table used by the unwinder at runtime. 31653 It will be called by `expand_builtin_init_dwarf_reg_sizes' after 31654 filling in a single size corresponding to each hard register; 31655 ADDRESS is the address of the table. 31656 31657 -- Target Hook: bool TARGET_ASM_TTYPE (rtx SYM) 31658 This hook is used to output a reference from a frame unwinding 31659 table to the type_info object identified by SYM. It should return 31660 `true' if the reference was output. Returning `false' will cause 31661 the reference to be output using the normal Dwarf2 routines. 31662 31663 -- Target Hook: bool TARGET_ARM_EABI_UNWINDER 31664 This hook should be set to `true' on targets that use an ARM EABI 31665 based unwinding library, and `false' on other targets. This 31666 effects the format of unwinding tables, and how the unwinder in 31667 entered after running a cleanup. The default is `false'. 31668 31669 31670 File: gccint.info, Node: Alignment Output, Prev: Exception Region Output, Up: Assembler Format 31671 31672 17.21.10 Assembler Commands for Alignment 31673 ----------------------------------------- 31674 31675 This describes commands for alignment. 31676 31677 -- Macro: JUMP_ALIGN (LABEL) 31678 The alignment (log base 2) to put in front of LABEL, which is a 31679 common destination of jumps and has no fallthru incoming edge. 31680 31681 This macro need not be defined if you don't want any special 31682 alignment to be done at such a time. Most machine descriptions do 31683 not currently define the macro. 31684 31685 Unless it's necessary to inspect the LABEL parameter, it is better 31686 to set the variable ALIGN_JUMPS in the target's 31687 `OVERRIDE_OPTIONS'. Otherwise, you should try to honor the user's 31688 selection in ALIGN_JUMPS in a `JUMP_ALIGN' implementation. 31689 31690 -- Macro: LABEL_ALIGN_AFTER_BARRIER (LABEL) 31691 The alignment (log base 2) to put in front of LABEL, which follows 31692 a `BARRIER'. 31693 31694 This macro need not be defined if you don't want any special 31695 alignment to be done at such a time. Most machine descriptions do 31696 not currently define the macro. 31697 31698 -- Macro: LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP 31699 The maximum number of bytes to skip when applying 31700 `LABEL_ALIGN_AFTER_BARRIER'. This works only if 31701 `ASM_OUTPUT_MAX_SKIP_ALIGN' is defined. 31702 31703 -- Macro: LOOP_ALIGN (LABEL) 31704 The alignment (log base 2) to put in front of LABEL, which follows 31705 a `NOTE_INSN_LOOP_BEG' note. 31706 31707 This macro need not be defined if you don't want any special 31708 alignment to be done at such a time. Most machine descriptions do 31709 not currently define the macro. 31710 31711 Unless it's necessary to inspect the LABEL parameter, it is better 31712 to set the variable `align_loops' in the target's 31713 `OVERRIDE_OPTIONS'. Otherwise, you should try to honor the user's 31714 selection in `align_loops' in a `LOOP_ALIGN' implementation. 31715 31716 -- Macro: LOOP_ALIGN_MAX_SKIP 31717 The maximum number of bytes to skip when applying `LOOP_ALIGN'. 31718 This works only if `ASM_OUTPUT_MAX_SKIP_ALIGN' is defined. 31719 31720 -- Macro: LABEL_ALIGN (LABEL) 31721 The alignment (log base 2) to put in front of LABEL. If 31722 `LABEL_ALIGN_AFTER_BARRIER' / `LOOP_ALIGN' specify a different 31723 alignment, the maximum of the specified values is used. 31724 31725 Unless it's necessary to inspect the LABEL parameter, it is better 31726 to set the variable `align_labels' in the target's 31727 `OVERRIDE_OPTIONS'. Otherwise, you should try to honor the user's 31728 selection in `align_labels' in a `LABEL_ALIGN' implementation. 31729 31730 -- Macro: LABEL_ALIGN_MAX_SKIP 31731 The maximum number of bytes to skip when applying `LABEL_ALIGN'. 31732 This works only if `ASM_OUTPUT_MAX_SKIP_ALIGN' is defined. 31733 31734 -- Macro: ASM_OUTPUT_SKIP (STREAM, NBYTES) 31735 A C statement to output to the stdio stream STREAM an assembler 31736 instruction to advance the location counter by NBYTES bytes. 31737 Those bytes should be zero when loaded. NBYTES will be a C 31738 expression of type `unsigned HOST_WIDE_INT'. 31739 31740 -- Macro: ASM_NO_SKIP_IN_TEXT 31741 Define this macro if `ASM_OUTPUT_SKIP' should not be used in the 31742 text section because it fails to put zeros in the bytes that are 31743 skipped. This is true on many Unix systems, where the pseudo-op 31744 to skip bytes produces no-op instructions rather than zeros when 31745 used in the text section. 31746 31747 -- Macro: ASM_OUTPUT_ALIGN (STREAM, POWER) 31748 A C statement to output to the stdio stream STREAM an assembler 31749 command to advance the location counter to a multiple of 2 to the 31750 POWER bytes. POWER will be a C expression of type `int'. 31751 31752 -- Macro: ASM_OUTPUT_ALIGN_WITH_NOP (STREAM, POWER) 31753 Like `ASM_OUTPUT_ALIGN', except that the "nop" instruction is used 31754 for padding, if necessary. 31755 31756 -- Macro: ASM_OUTPUT_MAX_SKIP_ALIGN (STREAM, POWER, MAX_SKIP) 31757 A C statement to output to the stdio stream STREAM an assembler 31758 command to advance the location counter to a multiple of 2 to the 31759 POWER bytes, but only if MAX_SKIP or fewer bytes are needed to 31760 satisfy the alignment request. POWER and MAX_SKIP will be a C 31761 expression of type `int'. 31762 31763 31764 File: gccint.info, Node: Debugging Info, Next: Floating Point, Prev: Assembler Format, Up: Target Macros 31765 31766 17.22 Controlling Debugging Information Format 31767 ============================================== 31768 31769 This describes how to specify debugging information. 31770 31771 * Menu: 31772 31773 * All Debuggers:: Macros that affect all debugging formats uniformly. 31774 * DBX Options:: Macros enabling specific options in DBX format. 31775 * DBX Hooks:: Hook macros for varying DBX format. 31776 * File Names and DBX:: Macros controlling output of file names in DBX format. 31777 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. 31778 * VMS Debug:: Macros for VMS debug format. 31779 31780 31781 File: gccint.info, Node: All Debuggers, Next: DBX Options, Up: Debugging Info 31782 31783 17.22.1 Macros Affecting All Debugging Formats 31784 ---------------------------------------------- 31785 31786 These macros affect all debugging formats. 31787 31788 -- Macro: DBX_REGISTER_NUMBER (REGNO) 31789 A C expression that returns the DBX register number for the 31790 compiler register number REGNO. In the default macro provided, 31791 the value of this expression will be REGNO itself. But sometimes 31792 there are some registers that the compiler knows about and DBX 31793 does not, or vice versa. In such cases, some register may need to 31794 have one number in the compiler and another for DBX. 31795 31796 If two registers have consecutive numbers inside GCC, and they can 31797 be used as a pair to hold a multiword value, then they _must_ have 31798 consecutive numbers after renumbering with `DBX_REGISTER_NUMBER'. 31799 Otherwise, debuggers will be unable to access such a pair, because 31800 they expect register pairs to be consecutive in their own 31801 numbering scheme. 31802 31803 If you find yourself defining `DBX_REGISTER_NUMBER' in way that 31804 does not preserve register pairs, then what you must do instead is 31805 redefine the actual register numbering scheme. 31806 31807 -- Macro: DEBUGGER_AUTO_OFFSET (X) 31808 A C expression that returns the integer offset value for an 31809 automatic variable having address X (an RTL expression). The 31810 default computation assumes that X is based on the frame-pointer 31811 and gives the offset from the frame-pointer. This is required for 31812 targets that produce debugging output for DBX or COFF-style 31813 debugging output for SDB and allow the frame-pointer to be 31814 eliminated when the `-g' options is used. 31815 31816 -- Macro: DEBUGGER_ARG_OFFSET (OFFSET, X) 31817 A C expression that returns the integer offset value for an 31818 argument having address X (an RTL expression). The nominal offset 31819 is OFFSET. 31820 31821 -- Macro: PREFERRED_DEBUGGING_TYPE 31822 A C expression that returns the type of debugging output GCC should 31823 produce when the user specifies just `-g'. Define this if you 31824 have arranged for GCC to support more than one format of debugging 31825 output. Currently, the allowable values are `DBX_DEBUG', 31826 `SDB_DEBUG', `DWARF_DEBUG', `DWARF2_DEBUG', `XCOFF_DEBUG', 31827 `VMS_DEBUG', and `VMS_AND_DWARF2_DEBUG'. 31828 31829 When the user specifies `-ggdb', GCC normally also uses the value 31830 of this macro to select the debugging output format, but with two 31831 exceptions. If `DWARF2_DEBUGGING_INFO' is defined, GCC uses the 31832 value `DWARF2_DEBUG'. Otherwise, if `DBX_DEBUGGING_INFO' is 31833 defined, GCC uses `DBX_DEBUG'. 31834 31835 The value of this macro only affects the default debugging output; 31836 the user can always get a specific type of output by using 31837 `-gstabs', `-gcoff', `-gdwarf-2', `-gxcoff', or `-gvms'. 31838 31839 31840 File: gccint.info, Node: DBX Options, Next: DBX Hooks, Prev: All Debuggers, Up: Debugging Info 31841 31842 17.22.2 Specific Options for DBX Output 31843 --------------------------------------- 31844 31845 These are specific options for DBX output. 31846 31847 -- Macro: DBX_DEBUGGING_INFO 31848 Define this macro if GCC should produce debugging output for DBX 31849 in response to the `-g' option. 31850 31851 -- Macro: XCOFF_DEBUGGING_INFO 31852 Define this macro if GCC should produce XCOFF format debugging 31853 output in response to the `-g' option. This is a variant of DBX 31854 format. 31855 31856 -- Macro: DEFAULT_GDB_EXTENSIONS 31857 Define this macro to control whether GCC should by default generate 31858 GDB's extended version of DBX debugging information (assuming 31859 DBX-format debugging information is enabled at all). If you don't 31860 define the macro, the default is 1: always generate the extended 31861 information if there is any occasion to. 31862 31863 -- Macro: DEBUG_SYMS_TEXT 31864 Define this macro if all `.stabs' commands should be output while 31865 in the text section. 31866 31867 -- Macro: ASM_STABS_OP 31868 A C string constant, including spacing, naming the assembler 31869 pseudo op to use instead of `"\t.stabs\t"' to define an ordinary 31870 debugging symbol. If you don't define this macro, `"\t.stabs\t"' 31871 is used. This macro applies only to DBX debugging information 31872 format. 31873 31874 -- Macro: ASM_STABD_OP 31875 A C string constant, including spacing, naming the assembler 31876 pseudo op to use instead of `"\t.stabd\t"' to define a debugging 31877 symbol whose value is the current location. If you don't define 31878 this macro, `"\t.stabd\t"' is used. This macro applies only to 31879 DBX debugging information format. 31880 31881 -- Macro: ASM_STABN_OP 31882 A C string constant, including spacing, naming the assembler 31883 pseudo op to use instead of `"\t.stabn\t"' to define a debugging 31884 symbol with no name. If you don't define this macro, 31885 `"\t.stabn\t"' is used. This macro applies only to DBX debugging 31886 information format. 31887 31888 -- Macro: DBX_NO_XREFS 31889 Define this macro if DBX on your system does not support the 31890 construct `xsTAGNAME'. On some systems, this construct is used to 31891 describe a forward reference to a structure named TAGNAME. On 31892 other systems, this construct is not supported at all. 31893 31894 -- Macro: DBX_CONTIN_LENGTH 31895 A symbol name in DBX-format debugging information is normally 31896 continued (split into two separate `.stabs' directives) when it 31897 exceeds a certain length (by default, 80 characters). On some 31898 operating systems, DBX requires this splitting; on others, 31899 splitting must not be done. You can inhibit splitting by defining 31900 this macro with the value zero. You can override the default 31901 splitting-length by defining this macro as an expression for the 31902 length you desire. 31903 31904 -- Macro: DBX_CONTIN_CHAR 31905 Normally continuation is indicated by adding a `\' character to 31906 the end of a `.stabs' string when a continuation follows. To use 31907 a different character instead, define this macro as a character 31908 constant for the character you want to use. Do not define this 31909 macro if backslash is correct for your system. 31910 31911 -- Macro: DBX_STATIC_STAB_DATA_SECTION 31912 Define this macro if it is necessary to go to the data section 31913 before outputting the `.stabs' pseudo-op for a non-global static 31914 variable. 31915 31916 -- Macro: DBX_TYPE_DECL_STABS_CODE 31917 The value to use in the "code" field of the `.stabs' directive for 31918 a typedef. The default is `N_LSYM'. 31919 31920 -- Macro: DBX_STATIC_CONST_VAR_CODE 31921 The value to use in the "code" field of the `.stabs' directive for 31922 a static variable located in the text section. DBX format does not 31923 provide any "right" way to do this. The default is `N_FUN'. 31924 31925 -- Macro: DBX_REGPARM_STABS_CODE 31926 The value to use in the "code" field of the `.stabs' directive for 31927 a parameter passed in registers. DBX format does not provide any 31928 "right" way to do this. The default is `N_RSYM'. 31929 31930 -- Macro: DBX_REGPARM_STABS_LETTER 31931 The letter to use in DBX symbol data to identify a symbol as a 31932 parameter passed in registers. DBX format does not customarily 31933 provide any way to do this. The default is `'P''. 31934 31935 -- Macro: DBX_FUNCTION_FIRST 31936 Define this macro if the DBX information for a function and its 31937 arguments should precede the assembler code for the function. 31938 Normally, in DBX format, the debugging information entirely 31939 follows the assembler code. 31940 31941 -- Macro: DBX_BLOCKS_FUNCTION_RELATIVE 31942 Define this macro, with value 1, if the value of a symbol 31943 describing the scope of a block (`N_LBRAC' or `N_RBRAC') should be 31944 relative to the start of the enclosing function. Normally, GCC 31945 uses an absolute address. 31946 31947 -- Macro: DBX_LINES_FUNCTION_RELATIVE 31948 Define this macro, with value 1, if the value of a symbol 31949 indicating the current line number (`N_SLINE') should be relative 31950 to the start of the enclosing function. Normally, GCC uses an 31951 absolute address. 31952 31953 -- Macro: DBX_USE_BINCL 31954 Define this macro if GCC should generate `N_BINCL' and `N_EINCL' 31955 stabs for included header files, as on Sun systems. This macro 31956 also directs GCC to output a type number as a pair of a file 31957 number and a type number within the file. Normally, GCC does not 31958 generate `N_BINCL' or `N_EINCL' stabs, and it outputs a single 31959 number for a type number. 31960 31961 31962 File: gccint.info, Node: DBX Hooks, Next: File Names and DBX, Prev: DBX Options, Up: Debugging Info 31963 31964 17.22.3 Open-Ended Hooks for DBX Format 31965 --------------------------------------- 31966 31967 These are hooks for DBX format. 31968 31969 -- Macro: DBX_OUTPUT_LBRAC (STREAM, NAME) 31970 Define this macro to say how to output to STREAM the debugging 31971 information for the start of a scope level for variable names. The 31972 argument NAME is the name of an assembler symbol (for use with 31973 `assemble_name') whose value is the address where the scope begins. 31974 31975 -- Macro: DBX_OUTPUT_RBRAC (STREAM, NAME) 31976 Like `DBX_OUTPUT_LBRAC', but for the end of a scope level. 31977 31978 -- Macro: DBX_OUTPUT_NFUN (STREAM, LSCOPE_LABEL, DECL) 31979 Define this macro if the target machine requires special handling 31980 to output an `N_FUN' entry for the function DECL. 31981 31982 -- Macro: DBX_OUTPUT_SOURCE_LINE (STREAM, LINE, COUNTER) 31983 A C statement to output DBX debugging information before code for 31984 line number LINE of the current source file to the stdio stream 31985 STREAM. COUNTER is the number of time the macro was invoked, 31986 including the current invocation; it is intended to generate 31987 unique labels in the assembly output. 31988 31989 This macro should not be defined if the default output is correct, 31990 or if it can be made correct by defining 31991 `DBX_LINES_FUNCTION_RELATIVE'. 31992 31993 -- Macro: NO_DBX_FUNCTION_END 31994 Some stabs encapsulation formats (in particular ECOFF), cannot 31995 handle the `.stabs "",N_FUN,,0,0,Lscope-function-1' gdb dbx 31996 extension construct. On those machines, define this macro to turn 31997 this feature off without disturbing the rest of the gdb extensions. 31998 31999 -- Macro: NO_DBX_BNSYM_ENSYM 32000 Some assemblers cannot handle the `.stabd BNSYM/ENSYM,0,0' gdb dbx 32001 extension construct. On those machines, define this macro to turn 32002 this feature off without disturbing the rest of the gdb extensions. 32003 32004 32005 File: gccint.info, Node: File Names and DBX, Next: SDB and DWARF, Prev: DBX Hooks, Up: Debugging Info 32006 32007 17.22.4 File Names in DBX Format 32008 -------------------------------- 32009 32010 This describes file names in DBX format. 32011 32012 -- Macro: DBX_OUTPUT_MAIN_SOURCE_FILENAME (STREAM, NAME) 32013 A C statement to output DBX debugging information to the stdio 32014 stream STREAM, which indicates that file NAME is the main source 32015 file--the file specified as the input file for compilation. This 32016 macro is called only once, at the beginning of compilation. 32017 32018 This macro need not be defined if the standard form of output for 32019 DBX debugging information is appropriate. 32020 32021 It may be necessary to refer to a label equal to the beginning of 32022 the text section. You can use `assemble_name (stream, 32023 ltext_label_name)' to do so. If you do this, you must also set 32024 the variable USED_LTEXT_LABEL_NAME to `true'. 32025 32026 -- Macro: NO_DBX_MAIN_SOURCE_DIRECTORY 32027 Define this macro, with value 1, if GCC should not emit an 32028 indication of the current directory for compilation and current 32029 source language at the beginning of the file. 32030 32031 -- Macro: NO_DBX_GCC_MARKER 32032 Define this macro, with value 1, if GCC should not emit an 32033 indication that this object file was compiled by GCC. The default 32034 is to emit an `N_OPT' stab at the beginning of every source file, 32035 with `gcc2_compiled.' for the string and value 0. 32036 32037 -- Macro: DBX_OUTPUT_MAIN_SOURCE_FILE_END (STREAM, NAME) 32038 A C statement to output DBX debugging information at the end of 32039 compilation of the main source file NAME. Output should be 32040 written to the stdio stream STREAM. 32041 32042 If you don't define this macro, nothing special is output at the 32043 end of compilation, which is correct for most machines. 32044 32045 -- Macro: DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END 32046 Define this macro _instead of_ defining 32047 `DBX_OUTPUT_MAIN_SOURCE_FILE_END', if what needs to be output at 32048 the end of compilation is a `N_SO' stab with an empty string, 32049 whose value is the highest absolute text address in the file. 32050 32051 32052 File: gccint.info, Node: SDB and DWARF, Next: VMS Debug, Prev: File Names and DBX, Up: Debugging Info 32053 32054 17.22.5 Macros for SDB and DWARF Output 32055 --------------------------------------- 32056 32057 Here are macros for SDB and DWARF output. 32058 32059 -- Macro: SDB_DEBUGGING_INFO 32060 Define this macro if GCC should produce COFF-style debugging output 32061 for SDB in response to the `-g' option. 32062 32063 -- Macro: DWARF2_DEBUGGING_INFO 32064 Define this macro if GCC should produce dwarf version 2 format 32065 debugging output in response to the `-g' option. 32066 32067 -- Target Hook: int TARGET_DWARF_CALLING_CONVENTION (tree 32068 FUNCTION) 32069 Define this to enable the dwarf attribute 32070 `DW_AT_calling_convention' to be emitted for each function. 32071 Instead of an integer return the enum value for the `DW_CC_' 32072 tag. 32073 32074 To support optional call frame debugging information, you must also 32075 define `INCOMING_RETURN_ADDR_RTX' and either set 32076 `RTX_FRAME_RELATED_P' on the prologue insns if you use RTL for the 32077 prologue, or call `dwarf2out_def_cfa' and `dwarf2out_reg_save' as 32078 appropriate from `TARGET_ASM_FUNCTION_PROLOGUE' if you don't. 32079 32080 -- Macro: DWARF2_FRAME_INFO 32081 Define this macro to a nonzero value if GCC should always output 32082 Dwarf 2 frame information. If `DWARF2_UNWIND_INFO' (*note 32083 Exception Region Output:: is nonzero, GCC will output this 32084 information not matter how you define `DWARF2_FRAME_INFO'. 32085 32086 -- Macro: DWARF2_ASM_LINE_DEBUG_INFO 32087 Define this macro to be a nonzero value if the assembler can 32088 generate Dwarf 2 line debug info sections. This will result in 32089 much more compact line number tables, and hence is desirable if it 32090 works. 32091 32092 -- Macro: ASM_OUTPUT_DWARF_DELTA (STREAM, SIZE, LABEL1, LABEL2) 32093 A C statement to issue assembly directives that create a difference 32094 LAB1 minus LAB2, using an integer of the given SIZE. 32095 32096 -- Macro: ASM_OUTPUT_DWARF_OFFSET (STREAM, SIZE, LABEL, SECTION) 32097 A C statement to issue assembly directives that create a 32098 section-relative reference to the given LABEL, using an integer of 32099 the given SIZE. The label is known to be defined in the given 32100 SECTION. 32101 32102 -- Macro: ASM_OUTPUT_DWARF_PCREL (STREAM, SIZE, LABEL) 32103 A C statement to issue assembly directives that create a 32104 self-relative reference to the given LABEL, using an integer of 32105 the given SIZE. 32106 32107 -- Target Hook: void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *FILE, int 32108 SIZE, rtx X) 32109 If defined, this target hook is a function which outputs a 32110 DTP-relative reference to the given TLS symbol of the specified 32111 size. 32112 32113 -- Macro: PUT_SDB_... 32114 Define these macros to override the assembler syntax for the 32115 special SDB assembler directives. See `sdbout.c' for a list of 32116 these macros and their arguments. If the standard syntax is used, 32117 you need not define them yourself. 32118 32119 -- Macro: SDB_DELIM 32120 Some assemblers do not support a semicolon as a delimiter, even 32121 between SDB assembler directives. In that case, define this macro 32122 to be the delimiter to use (usually `\n'). It is not necessary to 32123 define a new set of `PUT_SDB_OP' macros if this is the only change 32124 required. 32125 32126 -- Macro: SDB_ALLOW_UNKNOWN_REFERENCES 32127 Define this macro to allow references to unknown structure, union, 32128 or enumeration tags to be emitted. Standard COFF does not allow 32129 handling of unknown references, MIPS ECOFF has support for it. 32130 32131 -- Macro: SDB_ALLOW_FORWARD_REFERENCES 32132 Define this macro to allow references to structure, union, or 32133 enumeration tags that have not yet been seen to be handled. Some 32134 assemblers choke if forward tags are used, while some require it. 32135 32136 -- Macro: SDB_OUTPUT_SOURCE_LINE (STREAM, LINE) 32137 A C statement to output SDB debugging information before code for 32138 line number LINE of the current source file to the stdio stream 32139 STREAM. The default is to emit an `.ln' directive. 32140 32141 32142 File: gccint.info, Node: VMS Debug, Prev: SDB and DWARF, Up: Debugging Info 32143 32144 17.22.6 Macros for VMS Debug Format 32145 ----------------------------------- 32146 32147 Here are macros for VMS debug format. 32148 32149 -- Macro: VMS_DEBUGGING_INFO 32150 Define this macro if GCC should produce debugging output for VMS 32151 in response to the `-g' option. The default behavior for VMS is 32152 to generate minimal debug info for a traceback in the absence of 32153 `-g' unless explicitly overridden with `-g0'. This behavior is 32154 controlled by `OPTIMIZATION_OPTIONS' and `OVERRIDE_OPTIONS'. 32155 32156 32157 File: gccint.info, Node: Floating Point, Next: Mode Switching, Prev: Debugging Info, Up: Target Macros 32158 32159 17.23 Cross Compilation and Floating Point 32160 ========================================== 32161 32162 While all modern machines use twos-complement representation for 32163 integers, there are a variety of representations for floating point 32164 numbers. This means that in a cross-compiler the representation of 32165 floating point numbers in the compiled program may be different from 32166 that used in the machine doing the compilation. 32167 32168 Because different representation systems may offer different amounts of 32169 range and precision, all floating point constants must be represented in 32170 the target machine's format. Therefore, the cross compiler cannot 32171 safely use the host machine's floating point arithmetic; it must emulate 32172 the target's arithmetic. To ensure consistency, GCC always uses 32173 emulation to work with floating point values, even when the host and 32174 target floating point formats are identical. 32175 32176 The following macros are provided by `real.h' for the compiler to use. 32177 All parts of the compiler which generate or optimize floating-point 32178 calculations must use these macros. They may evaluate their operands 32179 more than once, so operands must not have side effects. 32180 32181 -- Macro: REAL_VALUE_TYPE 32182 The C data type to be used to hold a floating point value in the 32183 target machine's format. Typically this is a `struct' containing 32184 an array of `HOST_WIDE_INT', but all code should treat it as an 32185 opaque quantity. 32186 32187 -- Macro: int REAL_VALUES_EQUAL (REAL_VALUE_TYPE X, REAL_VALUE_TYPE Y) 32188 Compares for equality the two values, X and Y. If the target 32189 floating point format supports negative zeroes and/or NaNs, 32190 `REAL_VALUES_EQUAL (-0.0, 0.0)' is true, and `REAL_VALUES_EQUAL 32191 (NaN, NaN)' is false. 32192 32193 -- Macro: int REAL_VALUES_LESS (REAL_VALUE_TYPE X, REAL_VALUE_TYPE Y) 32194 Tests whether X is less than Y. 32195 32196 -- Macro: HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE X) 32197 Truncates X to a signed integer, rounding toward zero. 32198 32199 -- Macro: unsigned HOST_WIDE_INT REAL_VALUE_UNSIGNED_FIX 32200 (REAL_VALUE_TYPE X) 32201 Truncates X to an unsigned integer, rounding toward zero. If X is 32202 negative, returns zero. 32203 32204 -- Macro: REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *STRING, enum 32205 machine_mode MODE) 32206 Converts STRING into a floating point number in the target 32207 machine's representation for mode MODE. This routine can handle 32208 both decimal and hexadecimal floating point constants, using the 32209 syntax defined by the C language for both. 32210 32211 -- Macro: int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE X) 32212 Returns 1 if X is negative (including negative zero), 0 otherwise. 32213 32214 -- Macro: int REAL_VALUE_ISINF (REAL_VALUE_TYPE X) 32215 Determines whether X represents infinity (positive or negative). 32216 32217 -- Macro: int REAL_VALUE_ISNAN (REAL_VALUE_TYPE X) 32218 Determines whether X represents a "NaN" (not-a-number). 32219 32220 -- Macro: void REAL_ARITHMETIC (REAL_VALUE_TYPE OUTPUT, enum tree_code 32221 CODE, REAL_VALUE_TYPE X, REAL_VALUE_TYPE Y) 32222 Calculates an arithmetic operation on the two floating point values 32223 X and Y, storing the result in OUTPUT (which must be a variable). 32224 32225 The operation to be performed is specified by CODE. Only the 32226 following codes are supported: `PLUS_EXPR', `MINUS_EXPR', 32227 `MULT_EXPR', `RDIV_EXPR', `MAX_EXPR', `MIN_EXPR'. 32228 32229 If `REAL_ARITHMETIC' is asked to evaluate division by zero and the 32230 target's floating point format cannot represent infinity, it will 32231 call `abort'. Callers should check for this situation first, using 32232 `MODE_HAS_INFINITIES'. *Note Storage Layout::. 32233 32234 -- Macro: REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE X) 32235 Returns the negative of the floating point value X. 32236 32237 -- Macro: REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE X) 32238 Returns the absolute value of X. 32239 32240 -- Macro: REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE MODE, 32241 enum machine_mode X) 32242 Truncates the floating point value X to fit in MODE. The return 32243 value is still a full-size `REAL_VALUE_TYPE', but it has an 32244 appropriate bit pattern to be output as a floating constant whose 32245 precision accords with mode MODE. 32246 32247 -- Macro: void REAL_VALUE_TO_INT (HOST_WIDE_INT LOW, HOST_WIDE_INT 32248 HIGH, REAL_VALUE_TYPE X) 32249 Converts a floating point value X into a double-precision integer 32250 which is then stored into LOW and HIGH. If the value is not 32251 integral, it is truncated. 32252 32253 -- Macro: void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE X, HOST_WIDE_INT 32254 LOW, HOST_WIDE_INT HIGH, enum machine_mode MODE) 32255 Converts a double-precision integer found in LOW and HIGH, into a 32256 floating point value which is then stored into X. The value is 32257 truncated to fit in mode MODE. 32258 32259 32260 File: gccint.info, Node: Mode Switching, Next: Target Attributes, Prev: Floating Point, Up: Target Macros 32261 32262 17.24 Mode Switching Instructions 32263 ================================= 32264 32265 The following macros control mode switching optimizations: 32266 32267 -- Macro: OPTIMIZE_MODE_SWITCHING (ENTITY) 32268 Define this macro if the port needs extra instructions inserted 32269 for mode switching in an optimizing compilation. 32270 32271 For an example, the SH4 can perform both single and double 32272 precision floating point operations, but to perform a single 32273 precision operation, the FPSCR PR bit has to be cleared, while for 32274 a double precision operation, this bit has to be set. Changing 32275 the PR bit requires a general purpose register as a scratch 32276 register, hence these FPSCR sets have to be inserted before 32277 reload, i.e. you can't put this into instruction emitting or 32278 `TARGET_MACHINE_DEPENDENT_REORG'. 32279 32280 You can have multiple entities that are mode-switched, and select 32281 at run time which entities actually need it. 32282 `OPTIMIZE_MODE_SWITCHING' should return nonzero for any ENTITY 32283 that needs mode-switching. If you define this macro, you also 32284 have to define `NUM_MODES_FOR_MODE_SWITCHING', `MODE_NEEDED', 32285 `MODE_PRIORITY_TO_MODE' and `EMIT_MODE_SET'. `MODE_AFTER', 32286 `MODE_ENTRY', and `MODE_EXIT' are optional. 32287 32288 -- Macro: NUM_MODES_FOR_MODE_SWITCHING 32289 If you define `OPTIMIZE_MODE_SWITCHING', you have to define this as 32290 initializer for an array of integers. Each initializer element N 32291 refers to an entity that needs mode switching, and specifies the 32292 number of different modes that might need to be set for this 32293 entity. The position of the initializer in the 32294 initializer--starting counting at zero--determines the integer 32295 that is used to refer to the mode-switched entity in question. In 32296 macros that take mode arguments / yield a mode result, modes are 32297 represented as numbers 0 ... N - 1. N is used to specify that no 32298 mode switch is needed / supplied. 32299 32300 -- Macro: MODE_NEEDED (ENTITY, INSN) 32301 ENTITY is an integer specifying a mode-switched entity. If 32302 `OPTIMIZE_MODE_SWITCHING' is defined, you must define this macro to 32303 return an integer value not larger than the corresponding element 32304 in `NUM_MODES_FOR_MODE_SWITCHING', to denote the mode that ENTITY 32305 must be switched into prior to the execution of INSN. 32306 32307 -- Macro: MODE_AFTER (MODE, INSN) 32308 If this macro is defined, it is evaluated for every INSN during 32309 mode switching. It determines the mode that an insn results in (if 32310 different from the incoming mode). 32311 32312 -- Macro: MODE_ENTRY (ENTITY) 32313 If this macro is defined, it is evaluated for every ENTITY that 32314 needs mode switching. It should evaluate to an integer, which is 32315 a mode that ENTITY is assumed to be switched to at function entry. 32316 If `MODE_ENTRY' is defined then `MODE_EXIT' must be defined. 32317 32318 -- Macro: MODE_EXIT (ENTITY) 32319 If this macro is defined, it is evaluated for every ENTITY that 32320 needs mode switching. It should evaluate to an integer, which is 32321 a mode that ENTITY is assumed to be switched to at function exit. 32322 If `MODE_EXIT' is defined then `MODE_ENTRY' must be defined. 32323 32324 -- Macro: MODE_PRIORITY_TO_MODE (ENTITY, N) 32325 This macro specifies the order in which modes for ENTITY are 32326 processed. 0 is the highest priority, 32327 `NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1' the lowest. The value 32328 of the macro should be an integer designating a mode for ENTITY. 32329 For any fixed ENTITY, `mode_priority_to_mode' (ENTITY, N) shall be 32330 a bijection in 0 ... `num_modes_for_mode_switching[ENTITY] - 1'. 32331 32332 -- Macro: EMIT_MODE_SET (ENTITY, MODE, HARD_REGS_LIVE) 32333 Generate one or more insns to set ENTITY to MODE. HARD_REG_LIVE 32334 is the set of hard registers live at the point where the insn(s) 32335 are to be inserted. 32336 32337 32338 File: gccint.info, Node: Target Attributes, Next: Emulated TLS, Prev: Mode Switching, Up: Target Macros 32339 32340 17.25 Defining target-specific uses of `__attribute__' 32341 ====================================================== 32342 32343 Target-specific attributes may be defined for functions, data and types. 32344 These are described using the following target hooks; they also need to 32345 be documented in `extend.texi'. 32346 32347 -- Target Hook: const struct attribute_spec * TARGET_ATTRIBUTE_TABLE 32348 If defined, this target hook points to an array of `struct 32349 attribute_spec' (defined in `tree.h') specifying the machine 32350 specific attributes for this target and some of the restrictions 32351 on the entities to which these attributes are applied and the 32352 arguments they take. 32353 32354 -- Target Hook: int TARGET_COMP_TYPE_ATTRIBUTES (tree TYPE1, tree 32355 TYPE2) 32356 If defined, this target hook is a function which returns zero if 32357 the attributes on TYPE1 and TYPE2 are incompatible, one if they 32358 are compatible, and two if they are nearly compatible (which 32359 causes a warning to be generated). If this is not defined, 32360 machine-specific attributes are supposed always to be compatible. 32361 32362 -- Target Hook: void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree TYPE) 32363 If defined, this target hook is a function which assigns default 32364 attributes to newly defined TYPE. 32365 32366 -- Target Hook: tree TARGET_MERGE_TYPE_ATTRIBUTES (tree TYPE1, tree 32367 TYPE2) 32368 Define this target hook if the merging of type attributes needs 32369 special handling. If defined, the result is a list of the combined 32370 `TYPE_ATTRIBUTES' of TYPE1 and TYPE2. It is assumed that 32371 `comptypes' has already been called and returned 1. This function 32372 may call `merge_attributes' to handle machine-independent merging. 32373 32374 -- Target Hook: tree TARGET_MERGE_DECL_ATTRIBUTES (tree OLDDECL, tree 32375 NEWDECL) 32376 Define this target hook if the merging of decl attributes needs 32377 special handling. If defined, the result is a list of the combined 32378 `DECL_ATTRIBUTES' of OLDDECL and NEWDECL. NEWDECL is a duplicate 32379 declaration of OLDDECL. Examples of when this is needed are when 32380 one attribute overrides another, or when an attribute is nullified 32381 by a subsequent definition. This function may call 32382 `merge_attributes' to handle machine-independent merging. 32383 32384 If the only target-specific handling you require is `dllimport' 32385 for Microsoft Windows targets, you should define the macro 32386 `TARGET_DLLIMPORT_DECL_ATTRIBUTES' to `1'. The compiler will then 32387 define a function called `merge_dllimport_decl_attributes' which 32388 can then be defined as the expansion of 32389 `TARGET_MERGE_DECL_ATTRIBUTES'. You can also add 32390 `handle_dll_attribute' in the attribute table for your port to 32391 perform initial processing of the `dllimport' and `dllexport' 32392 attributes. This is done in `i386/cygwin.h' and `i386/i386.c', 32393 for example. 32394 32395 -- Target Hook: bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree DECL) 32396 DECL is a variable or function with `__attribute__((dllimport))' 32397 specified. Use this hook if the target needs to add extra 32398 validation checks to `handle_dll_attribute'. 32399 32400 -- Macro: TARGET_DECLSPEC 32401 Define this macro to a nonzero value if you want to treat 32402 `__declspec(X)' as equivalent to `__attribute((X))'. By default, 32403 this behavior is enabled only for targets that define 32404 `TARGET_DLLIMPORT_DECL_ATTRIBUTES'. The current implementation of 32405 `__declspec' is via a built-in macro, but you should not rely on 32406 this implementation detail. 32407 32408 -- Target Hook: void TARGET_INSERT_ATTRIBUTES (tree NODE, tree 32409 *ATTR_PTR) 32410 Define this target hook if you want to be able to add attributes 32411 to a decl when it is being created. This is normally useful for 32412 back ends which wish to implement a pragma by using the attributes 32413 which correspond to the pragma's effect. The NODE argument is the 32414 decl which is being created. The ATTR_PTR argument is a pointer 32415 to the attribute list for this decl. The list itself should not 32416 be modified, since it may be shared with other decls, but 32417 attributes may be chained on the head of the list and `*ATTR_PTR' 32418 modified to point to the new attributes, or a copy of the list may 32419 be made if further changes are needed. 32420 32421 -- Target Hook: bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree 32422 FNDECL) 32423 This target hook returns `true' if it is ok to inline FNDECL into 32424 the current function, despite its having target-specific 32425 attributes, `false' otherwise. By default, if a function has a 32426 target specific attribute attached to it, it will not be inlined. 32427 32428 -- Target Hook: bool TARGET_VALID_OPTION_ATTRIBUTE_P (tree FNDECL, 32429 tree NAME, tree ARGS, int FLAGS) 32430 This hook is called to parse the `attribute(option("..."))', and 32431 it allows the function to set different target machine compile time 32432 options for the current function that might be different than the 32433 options specified on the command line. The hook should return 32434 `true' if the options are valid. 32435 32436 The hook should set the DECL_FUNCTION_SPECIFIC_TARGET field in the 32437 function declaration to hold a pointer to a target specific STRUCT 32438 CL_TARGET_OPTION structure. 32439 32440 -- Target Hook: void TARGET_OPTION_SAVE (struct cl_target_option *PTR) 32441 This hook is called to save any additional target specific 32442 information in the STRUCT CL_TARGET_OPTION structure for function 32443 specific options. *Note Option file format::. 32444 32445 -- Target Hook: void TARGET_OPTION_RESTORE (struct cl_target_option 32446 *PTR) 32447 This hook is called to restore any additional target specific 32448 information in the STRUCT CL_TARGET_OPTION structure for function 32449 specific options. 32450 32451 -- Target Hook: void TARGET_OPTION_PRINT (struct cl_target_option *PTR) 32452 This hook is called to print any additional target specific 32453 information in the STRUCT CL_TARGET_OPTION structure for function 32454 specific options. 32455 32456 -- Target Hook: bool TARGET_OPTION_PRAGMA_PARSE (target ARGS) 32457 This target hook parses the options for `#pragma GCC option' to 32458 set the machine specific options for functions that occur later in 32459 the input stream. The options should be the same as handled by the 32460 `TARGET_VALID_OPTION_ATTRIBUTE_P' hook. 32461 32462 -- Target Hook: bool TARGET_CAN_INLINE_P (tree CALLER, tree CALLEE) 32463 This target hook returns `false' if the CALLER function cannot 32464 inline CALLEE, based on target specific information. By default, 32465 inlining is not allowed if the callee function has function 32466 specific target options and the caller does not use the same 32467 options. 32468 32469 32470 File: gccint.info, Node: Emulated TLS, Next: MIPS Coprocessors, Prev: Target Attributes, Up: Target Macros 32471 32472 17.26 Emulating TLS 32473 =================== 32474 32475 For targets whose psABI does not provide Thread Local Storage via 32476 specific relocations and instruction sequences, an emulation layer is 32477 used. A set of target hooks allows this emulation layer to be 32478 configured for the requirements of a particular target. For instance 32479 the psABI may in fact specify TLS support in terms of an emulation 32480 layer. 32481 32482 The emulation layer works by creating a control object for every TLS 32483 object. To access the TLS object, a lookup function is provided which, 32484 when given the address of the control object, will return the address 32485 of the current thread's instance of the TLS object. 32486 32487 -- Target Hook: const char * TARGET_EMUTLS_GET_ADDRESS 32488 Contains the name of the helper function that uses a TLS control 32489 object to locate a TLS instance. The default causes libgcc's 32490 emulated TLS helper function to be used. 32491 32492 -- Target Hook: const char * TARGET_EMUTLS_REGISTER_COMMON 32493 Contains the name of the helper function that should be used at 32494 program startup to register TLS objects that are implicitly 32495 initialized to zero. If this is `NULL', all TLS objects will have 32496 explicit initializers. The default causes libgcc's emulated TLS 32497 registration function to be used. 32498 32499 -- Target Hook: const char * TARGET_EMUTLS_VAR_SECTION 32500 Contains the name of the section in which TLS control variables 32501 should be placed. The default of `NULL' allows these to be placed 32502 in any section. 32503 32504 -- Target Hook: const char * TARGET_EMUTLS_TMPL_SECTION 32505 Contains the name of the section in which TLS initializers should 32506 be placed. The default of `NULL' allows these to be placed in any 32507 section. 32508 32509 -- Target Hook: const char * TARGET_EMUTLS_VAR_PREFIX 32510 Contains the prefix to be prepended to TLS control variable names. 32511 The default of `NULL' uses a target-specific prefix. 32512 32513 -- Target Hook: const char * TARGET_EMUTLS_TMPL_PREFIX 32514 Contains the prefix to be prepended to TLS initializer objects. 32515 The default of `NULL' uses a target-specific prefix. 32516 32517 -- Target Hook: tree TARGET_EMUTLS_VAR_FIELDS (tree TYPE, tree *NAME) 32518 Specifies a function that generates the FIELD_DECLs for a TLS 32519 control object type. TYPE is the RECORD_TYPE the fields are for 32520 and NAME should be filled with the structure tag, if the default of 32521 `__emutls_object' is unsuitable. The default creates a type 32522 suitable for libgcc's emulated TLS function. 32523 32524 -- Target Hook: tree TARGET_EMUTLS_VAR_INIT (tree VAR, tree DECL, tree 32525 TMPL_ADDR) 32526 Specifies a function that generates the CONSTRUCTOR to initialize a 32527 TLS control object. VAR is the TLS control object, DECL is the 32528 TLS object and TMPL_ADDR is the address of the initializer. The 32529 default initializes libgcc's emulated TLS control object. 32530 32531 -- Target Hook: bool TARGET_EMUTLS_VAR_ALIGN_FIXED 32532 Specifies whether the alignment of TLS control variable objects is 32533 fixed and should not be increased as some backends may do to 32534 optimize single objects. The default is false. 32535 32536 -- Target Hook: bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS 32537 Specifies whether a DWARF `DW_OP_form_tls_address' location 32538 descriptor may be used to describe emulated TLS control objects. 32539 32540 32541 File: gccint.info, Node: MIPS Coprocessors, Next: PCH Target, Prev: Emulated TLS, Up: Target Macros 32542 32543 17.27 Defining coprocessor specifics for MIPS targets. 32544 ====================================================== 32545 32546 The MIPS specification allows MIPS implementations to have as many as 4 32547 coprocessors, each with as many as 32 private registers. GCC supports 32548 accessing these registers and transferring values between the registers 32549 and memory using asm-ized variables. For example: 32550 32551 register unsigned int cp0count asm ("c0r1"); 32552 unsigned int d; 32553 32554 d = cp0count + 3; 32555 32556 ("c0r1" is the default name of register 1 in coprocessor 0; alternate 32557 names may be added as described below, or the default names may be 32558 overridden entirely in `SUBTARGET_CONDITIONAL_REGISTER_USAGE'.) 32559 32560 Coprocessor registers are assumed to be epilogue-used; sets to them 32561 will be preserved even if it does not appear that the register is used 32562 again later in the function. 32563 32564 Another note: according to the MIPS spec, coprocessor 1 (if present) is 32565 the FPU. One accesses COP1 registers through standard mips 32566 floating-point support; they are not included in this mechanism. 32567 32568 There is one macro used in defining the MIPS coprocessor interface 32569 which you may want to override in subtargets; it is described below. 32570 32571 -- Macro: ALL_COP_ADDITIONAL_REGISTER_NAMES 32572 A comma-separated list (with leading comma) of pairs describing the 32573 alternate names of coprocessor registers. The format of each 32574 entry should be 32575 { ALTERNATENAME, REGISTER_NUMBER} 32576 Default: empty. 32577 32578 32579 File: gccint.info, Node: PCH Target, Next: C++ ABI, Prev: MIPS Coprocessors, Up: Target Macros 32580 32581 17.28 Parameters for Precompiled Header Validity Checking 32582 ========================================================= 32583 32584 -- Target Hook: void *TARGET_GET_PCH_VALIDITY (size_t *SZ) 32585 This hook returns the data needed by `TARGET_PCH_VALID_P' and sets 32586 `*SZ' to the size of the data in bytes. 32587 32588 -- Target Hook: const char *TARGET_PCH_VALID_P (const void *DATA, 32589 size_t SZ) 32590 This hook checks whether the options used to create a PCH file are 32591 compatible with the current settings. It returns `NULL' if so and 32592 a suitable error message if not. Error messages will be presented 32593 to the user and must be localized using `_(MSG)'. 32594 32595 DATA is the data that was returned by `TARGET_GET_PCH_VALIDITY' 32596 when the PCH file was created and SZ is the size of that data in 32597 bytes. It's safe to assume that the data was created by the same 32598 version of the compiler, so no format checking is needed. 32599 32600 The default definition of `default_pch_valid_p' should be suitable 32601 for most targets. 32602 32603 -- Target Hook: const char *TARGET_CHECK_PCH_TARGET_FLAGS (int 32604 PCH_FLAGS) 32605 If this hook is nonnull, the default implementation of 32606 `TARGET_PCH_VALID_P' will use it to check for compatible values of 32607 `target_flags'. PCH_FLAGS specifies the value that `target_flags' 32608 had when the PCH file was created. The return value is the same 32609 as for `TARGET_PCH_VALID_P'. 32610 32611 32612 File: gccint.info, Node: C++ ABI, Next: Misc, Prev: PCH Target, Up: Target Macros 32613 32614 17.29 C++ ABI parameters 32615 ======================== 32616 32617 -- Target Hook: tree TARGET_CXX_GUARD_TYPE (void) 32618 Define this hook to override the integer type used for guard 32619 variables. These are used to implement one-time construction of 32620 static objects. The default is long_long_integer_type_node. 32621 32622 -- Target Hook: bool TARGET_CXX_GUARD_MASK_BIT (void) 32623 This hook determines how guard variables are used. It should 32624 return `false' (the default) if first byte should be used. A 32625 return value of `true' indicates the least significant bit should 32626 be used. 32627 32628 -- Target Hook: tree TARGET_CXX_GET_COOKIE_SIZE (tree TYPE) 32629 This hook returns the size of the cookie to use when allocating an 32630 array whose elements have the indicated TYPE. Assumes that it is 32631 already known that a cookie is needed. The default is `max(sizeof 32632 (size_t), alignof(type))', as defined in section 2.7 of the 32633 IA64/Generic C++ ABI. 32634 32635 -- Target Hook: bool TARGET_CXX_COOKIE_HAS_SIZE (void) 32636 This hook should return `true' if the element size should be 32637 stored in array cookies. The default is to return `false'. 32638 32639 -- Target Hook: int TARGET_CXX_IMPORT_EXPORT_CLASS (tree TYPE, int 32640 IMPORT_EXPORT) 32641 If defined by a backend this hook allows the decision made to 32642 export class TYPE to be overruled. Upon entry IMPORT_EXPORT will 32643 contain 1 if the class is going to be exported, -1 if it is going 32644 to be imported and 0 otherwise. This function should return the 32645 modified value and perform any other actions necessary to support 32646 the backend's targeted operating system. 32647 32648 -- Target Hook: bool TARGET_CXX_CDTOR_RETURNS_THIS (void) 32649 This hook should return `true' if constructors and destructors 32650 return the address of the object created/destroyed. The default 32651 is to return `false'. 32652 32653 -- Target Hook: bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void) 32654 This hook returns true if the key method for a class (i.e., the 32655 method which, if defined in the current translation unit, causes 32656 the virtual table to be emitted) may be an inline function. Under 32657 the standard Itanium C++ ABI the key method may be an inline 32658 function so long as the function is not declared inline in the 32659 class definition. Under some variants of the ABI, an inline 32660 function can never be the key method. The default is to return 32661 `true'. 32662 32663 -- Target Hook: void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree 32664 DECL) 32665 DECL is a virtual table, virtual table table, typeinfo object, or 32666 other similar implicit class data object that will be emitted with 32667 external linkage in this translation unit. No ELF visibility has 32668 been explicitly specified. If the target needs to specify a 32669 visibility other than that of the containing class, use this hook 32670 to set `DECL_VISIBILITY' and `DECL_VISIBILITY_SPECIFIED'. 32671 32672 -- Target Hook: bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void) 32673 This hook returns true (the default) if virtual tables and other 32674 similar implicit class data objects are always COMDAT if they have 32675 external linkage. If this hook returns false, then class data for 32676 classes whose virtual table will be emitted in only one translation 32677 unit will not be COMDAT. 32678 32679 -- Target Hook: bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void) 32680 This hook returns true (the default) if the RTTI information for 32681 the basic types which is defined in the C++ runtime should always 32682 be COMDAT, false if it should not be COMDAT. 32683 32684 -- Target Hook: bool TARGET_CXX_USE_AEABI_ATEXIT (void) 32685 This hook returns true if `__aeabi_atexit' (as defined by the ARM 32686 EABI) should be used to register static destructors when 32687 `-fuse-cxa-atexit' is in effect. The default is to return false 32688 to use `__cxa_atexit'. 32689 32690 -- Target Hook: bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void) 32691 This hook returns true if the target `atexit' function can be used 32692 in the same manner as `__cxa_atexit' to register C++ static 32693 destructors. This requires that `atexit'-registered functions in 32694 shared libraries are run in the correct order when the libraries 32695 are unloaded. The default is to return false. 32696 32697 -- Target Hook: void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree TYPE) 32698 TYPE is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has 32699 just been defined. Use this hook to make adjustments to the class 32700 (eg, tweak visibility or perform any other required target 32701 modifications). 32702 32703 32704 File: gccint.info, Node: Misc, Prev: C++ ABI, Up: Target Macros 32705 32706 17.30 Miscellaneous Parameters 32707 ============================== 32708 32709 Here are several miscellaneous parameters. 32710 32711 -- Macro: HAS_LONG_COND_BRANCH 32712 Define this boolean macro to indicate whether or not your 32713 architecture has conditional branches that can span all of memory. 32714 It is used in conjunction with an optimization that partitions 32715 hot and cold basic blocks into separate sections of the 32716 executable. If this macro is set to false, gcc will convert any 32717 conditional branches that attempt to cross between sections into 32718 unconditional branches or indirect jumps. 32719 32720 -- Macro: HAS_LONG_UNCOND_BRANCH 32721 Define this boolean macro to indicate whether or not your 32722 architecture has unconditional branches that can span all of 32723 memory. It is used in conjunction with an optimization that 32724 partitions hot and cold basic blocks into separate sections of the 32725 executable. If this macro is set to false, gcc will convert any 32726 unconditional branches that attempt to cross between sections into 32727 indirect jumps. 32728 32729 -- Macro: CASE_VECTOR_MODE 32730 An alias for a machine mode name. This is the machine mode that 32731 elements of a jump-table should have. 32732 32733 -- Macro: CASE_VECTOR_SHORTEN_MODE (MIN_OFFSET, MAX_OFFSET, BODY) 32734 Optional: return the preferred mode for an `addr_diff_vec' when 32735 the minimum and maximum offset are known. If you define this, it 32736 enables extra code in branch shortening to deal with 32737 `addr_diff_vec'. To make this work, you also have to define 32738 `INSN_ALIGN' and make the alignment for `addr_diff_vec' explicit. 32739 The BODY argument is provided so that the offset_unsigned and scale 32740 flags can be updated. 32741 32742 -- Macro: CASE_VECTOR_PC_RELATIVE 32743 Define this macro to be a C expression to indicate when jump-tables 32744 should contain relative addresses. You need not define this macro 32745 if jump-tables never contain relative addresses, or jump-tables 32746 should contain relative addresses only when `-fPIC' or `-fPIC' is 32747 in effect. 32748 32749 -- Macro: CASE_VALUES_THRESHOLD 32750 Define this to be the smallest number of different values for 32751 which it is best to use a jump-table instead of a tree of 32752 conditional branches. The default is four for machines with a 32753 `casesi' instruction and five otherwise. This is best for most 32754 machines. 32755 32756 -- Macro: CASE_USE_BIT_TESTS 32757 Define this macro to be a C expression to indicate whether C switch 32758 statements may be implemented by a sequence of bit tests. This is 32759 advantageous on processors that can efficiently implement left 32760 shift of 1 by the number of bits held in a register, but 32761 inappropriate on targets that would require a loop. By default, 32762 this macro returns `true' if the target defines an `ashlsi3' 32763 pattern, and `false' otherwise. 32764 32765 -- Macro: WORD_REGISTER_OPERATIONS 32766 Define this macro if operations between registers with integral 32767 mode smaller than a word are always performed on the entire 32768 register. Most RISC machines have this property and most CISC 32769 machines do not. 32770 32771 -- Macro: LOAD_EXTEND_OP (MEM_MODE) 32772 Define this macro to be a C expression indicating when insns that 32773 read memory in MEM_MODE, an integral mode narrower than a word, 32774 set the bits outside of MEM_MODE to be either the sign-extension 32775 or the zero-extension of the data read. Return `SIGN_EXTEND' for 32776 values of MEM_MODE for which the insn sign-extends, `ZERO_EXTEND' 32777 for which it zero-extends, and `UNKNOWN' for other modes. 32778 32779 This macro is not called with MEM_MODE non-integral or with a width 32780 greater than or equal to `BITS_PER_WORD', so you may return any 32781 value in this case. Do not define this macro if it would always 32782 return `UNKNOWN'. On machines where this macro is defined, you 32783 will normally define it as the constant `SIGN_EXTEND' or 32784 `ZERO_EXTEND'. 32785 32786 You may return a non-`UNKNOWN' value even if for some hard 32787 registers the sign extension is not performed, if for the 32788 `REGNO_REG_CLASS' of these hard registers 32789 `CANNOT_CHANGE_MODE_CLASS' returns nonzero when the FROM mode is 32790 MEM_MODE and the TO mode is any integral mode larger than this but 32791 not larger than `word_mode'. 32792 32793 You must return `UNKNOWN' if for some hard registers that allow 32794 this mode, `CANNOT_CHANGE_MODE_CLASS' says that they cannot change 32795 to `word_mode', but that they can change to another integral mode 32796 that is larger then MEM_MODE but still smaller than `word_mode'. 32797 32798 -- Macro: SHORT_IMMEDIATES_SIGN_EXTEND 32799 Define this macro if loading short immediate values into registers 32800 sign extends. 32801 32802 -- Macro: FIXUNS_TRUNC_LIKE_FIX_TRUNC 32803 Define this macro if the same instructions that convert a floating 32804 point number to a signed fixed point number also convert validly 32805 to an unsigned one. 32806 32807 -- Target Hook: int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum 32808 machine_mode MODE) 32809 When `-ffast-math' is in effect, GCC tries to optimize divisions 32810 by the same divisor, by turning them into multiplications by the 32811 reciprocal. This target hook specifies the minimum number of 32812 divisions that should be there for GCC to perform the optimization 32813 for a variable of mode MODE. The default implementation returns 3 32814 if the machine has an instruction for the division, and 2 if it 32815 does not. 32816 32817 -- Macro: MOVE_MAX 32818 The maximum number of bytes that a single instruction can move 32819 quickly between memory and registers or between two memory 32820 locations. 32821 32822 -- Macro: MAX_MOVE_MAX 32823 The maximum number of bytes that a single instruction can move 32824 quickly between memory and registers or between two memory 32825 locations. If this is undefined, the default is `MOVE_MAX'. 32826 Otherwise, it is the constant value that is the largest value that 32827 `MOVE_MAX' can have at run-time. 32828 32829 -- Macro: SHIFT_COUNT_TRUNCATED 32830 A C expression that is nonzero if on this machine the number of 32831 bits actually used for the count of a shift operation is equal to 32832 the number of bits needed to represent the size of the object 32833 being shifted. When this macro is nonzero, the compiler will 32834 assume that it is safe to omit a sign-extend, zero-extend, and 32835 certain bitwise `and' instructions that truncates the count of a 32836 shift operation. On machines that have instructions that act on 32837 bit-fields at variable positions, which may include `bit test' 32838 instructions, a nonzero `SHIFT_COUNT_TRUNCATED' also enables 32839 deletion of truncations of the values that serve as arguments to 32840 bit-field instructions. 32841 32842 If both types of instructions truncate the count (for shifts) and 32843 position (for bit-field operations), or if no variable-position 32844 bit-field instructions exist, you should define this macro. 32845 32846 However, on some machines, such as the 80386 and the 680x0, 32847 truncation only applies to shift operations and not the (real or 32848 pretended) bit-field operations. Define `SHIFT_COUNT_TRUNCATED' 32849 to be zero on such machines. Instead, add patterns to the `md' 32850 file that include the implied truncation of the shift instructions. 32851 32852 You need not define this macro if it would always have the value 32853 of zero. 32854 32855 -- Target Hook: int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode 32856 MODE) 32857 This function describes how the standard shift patterns for MODE 32858 deal with shifts by negative amounts or by more than the width of 32859 the mode. *Note shift patterns::. 32860 32861 On many machines, the shift patterns will apply a mask M to the 32862 shift count, meaning that a fixed-width shift of X by Y is 32863 equivalent to an arbitrary-width shift of X by Y & M. If this is 32864 true for mode MODE, the function should return M, otherwise it 32865 should return 0. A return value of 0 indicates that no particular 32866 behavior is guaranteed. 32867 32868 Note that, unlike `SHIFT_COUNT_TRUNCATED', this function does 32869 _not_ apply to general shift rtxes; it applies only to instructions 32870 that are generated by the named shift patterns. 32871 32872 The default implementation of this function returns 32873 `GET_MODE_BITSIZE (MODE) - 1' if `SHIFT_COUNT_TRUNCATED' and 0 32874 otherwise. This definition is always safe, but if 32875 `SHIFT_COUNT_TRUNCATED' is false, and some shift patterns 32876 nevertheless truncate the shift count, you may get better code by 32877 overriding it. 32878 32879 -- Macro: TRULY_NOOP_TRUNCATION (OUTPREC, INPREC) 32880 A C expression which is nonzero if on this machine it is safe to 32881 "convert" an integer of INPREC bits to one of OUTPREC bits (where 32882 OUTPREC is smaller than INPREC) by merely operating on it as if it 32883 had only OUTPREC bits. 32884 32885 On many machines, this expression can be 1. 32886 32887 When `TRULY_NOOP_TRUNCATION' returns 1 for a pair of sizes for 32888 modes for which `MODES_TIEABLE_P' is 0, suboptimal code can result. 32889 If this is the case, making `TRULY_NOOP_TRUNCATION' return 0 in 32890 such cases may improve things. 32891 32892 -- Target Hook: int TARGET_MODE_REP_EXTENDED (enum machine_mode MODE, 32893 enum machine_mode REP_MODE) 32894 The representation of an integral mode can be such that the values 32895 are always extended to a wider integral mode. Return 32896 `SIGN_EXTEND' if values of MODE are represented in sign-extended 32897 form to REP_MODE. Return `UNKNOWN' otherwise. (Currently, none 32898 of the targets use zero-extended representation this way so unlike 32899 `LOAD_EXTEND_OP', `TARGET_MODE_REP_EXTENDED' is expected to return 32900 either `SIGN_EXTEND' or `UNKNOWN'. Also no target extends MODE to 32901 MODE_REP so that MODE_REP is not the next widest integral mode and 32902 currently we take advantage of this fact.) 32903 32904 Similarly to `LOAD_EXTEND_OP' you may return a non-`UNKNOWN' value 32905 even if the extension is not performed on certain hard registers 32906 as long as for the `REGNO_REG_CLASS' of these hard registers 32907 `CANNOT_CHANGE_MODE_CLASS' returns nonzero. 32908 32909 Note that `TARGET_MODE_REP_EXTENDED' and `LOAD_EXTEND_OP' describe 32910 two related properties. If you define `TARGET_MODE_REP_EXTENDED 32911 (mode, word_mode)' you probably also want to define 32912 `LOAD_EXTEND_OP (mode)' to return the same type of extension. 32913 32914 In order to enforce the representation of `mode', 32915 `TRULY_NOOP_TRUNCATION' should return false when truncating to 32916 `mode'. 32917 32918 -- Macro: STORE_FLAG_VALUE 32919 A C expression describing the value returned by a comparison 32920 operator with an integral mode and stored by a store-flag 32921 instruction (`sCOND') when the condition is true. This 32922 description must apply to _all_ the `sCOND' patterns and all the 32923 comparison operators whose results have a `MODE_INT' mode. 32924 32925 A value of 1 or -1 means that the instruction implementing the 32926 comparison operator returns exactly 1 or -1 when the comparison is 32927 true and 0 when the comparison is false. Otherwise, the value 32928 indicates which bits of the result are guaranteed to be 1 when the 32929 comparison is true. This value is interpreted in the mode of the 32930 comparison operation, which is given by the mode of the first 32931 operand in the `sCOND' pattern. Either the low bit or the sign 32932 bit of `STORE_FLAG_VALUE' be on. Presently, only those bits are 32933 used by the compiler. 32934 32935 If `STORE_FLAG_VALUE' is neither 1 or -1, the compiler will 32936 generate code that depends only on the specified bits. It can also 32937 replace comparison operators with equivalent operations if they 32938 cause the required bits to be set, even if the remaining bits are 32939 undefined. For example, on a machine whose comparison operators 32940 return an `SImode' value and where `STORE_FLAG_VALUE' is defined as 32941 `0x80000000', saying that just the sign bit is relevant, the 32942 expression 32943 32944 (ne:SI (and:SI X (const_int POWER-OF-2)) (const_int 0)) 32945 32946 can be converted to 32947 32948 (ashift:SI X (const_int N)) 32949 32950 where N is the appropriate shift count to move the bit being 32951 tested into the sign bit. 32952 32953 There is no way to describe a machine that always sets the 32954 low-order bit for a true value, but does not guarantee the value 32955 of any other bits, but we do not know of any machine that has such 32956 an instruction. If you are trying to port GCC to such a machine, 32957 include an instruction to perform a logical-and of the result with 32958 1 in the pattern for the comparison operators and let us know at 32959 <gcc (a] gcc.gnu.org>. 32960 32961 Often, a machine will have multiple instructions that obtain a 32962 value from a comparison (or the condition codes). Here are rules 32963 to guide the choice of value for `STORE_FLAG_VALUE', and hence the 32964 instructions to be used: 32965 32966 * Use the shortest sequence that yields a valid definition for 32967 `STORE_FLAG_VALUE'. It is more efficient for the compiler to 32968 "normalize" the value (convert it to, e.g., 1 or 0) than for 32969 the comparison operators to do so because there may be 32970 opportunities to combine the normalization with other 32971 operations. 32972 32973 * For equal-length sequences, use a value of 1 or -1, with -1 32974 being slightly preferred on machines with expensive jumps and 32975 1 preferred on other machines. 32976 32977 * As a second choice, choose a value of `0x80000001' if 32978 instructions exist that set both the sign and low-order bits 32979 but do not define the others. 32980 32981 * Otherwise, use a value of `0x80000000'. 32982 32983 Many machines can produce both the value chosen for 32984 `STORE_FLAG_VALUE' and its negation in the same number of 32985 instructions. On those machines, you should also define a pattern 32986 for those cases, e.g., one matching 32987 32988 (set A (neg:M (ne:M B C))) 32989 32990 Some machines can also perform `and' or `plus' operations on 32991 condition code values with less instructions than the corresponding 32992 `sCOND' insn followed by `and' or `plus'. On those machines, 32993 define the appropriate patterns. Use the names `incscc' and 32994 `decscc', respectively, for the patterns which perform `plus' or 32995 `minus' operations on condition code values. See `rs6000.md' for 32996 some examples. The GNU Superoptizer can be used to find such 32997 instruction sequences on other machines. 32998 32999 If this macro is not defined, the default value, 1, is used. You 33000 need not define `STORE_FLAG_VALUE' if the machine has no store-flag 33001 instructions, or if the value generated by these instructions is 1. 33002 33003 -- Macro: FLOAT_STORE_FLAG_VALUE (MODE) 33004 A C expression that gives a nonzero `REAL_VALUE_TYPE' value that is 33005 returned when comparison operators with floating-point results are 33006 true. Define this macro on machines that have comparison 33007 operations that return floating-point values. If there are no 33008 such operations, do not define this macro. 33009 33010 -- Macro: VECTOR_STORE_FLAG_VALUE (MODE) 33011 A C expression that gives a rtx representing the nonzero true 33012 element for vector comparisons. The returned rtx should be valid 33013 for the inner mode of MODE which is guaranteed to be a vector 33014 mode. Define this macro on machines that have vector comparison 33015 operations that return a vector result. If there are no such 33016 operations, do not define this macro. Typically, this macro is 33017 defined as `const1_rtx' or `constm1_rtx'. This macro may return 33018 `NULL_RTX' to prevent the compiler optimizing such vector 33019 comparison operations for the given mode. 33020 33021 -- Macro: CLZ_DEFINED_VALUE_AT_ZERO (MODE, VALUE) 33022 -- Macro: CTZ_DEFINED_VALUE_AT_ZERO (MODE, VALUE) 33023 A C expression that indicates whether the architecture defines a 33024 value for `clz' or `ctz' with a zero operand. A result of `0' 33025 indicates the value is undefined. If the value is defined for 33026 only the RTL expression, the macro should evaluate to `1'; if the 33027 value applies also to the corresponding optab entry (which is 33028 normally the case if it expands directly into the corresponding 33029 RTL), then the macro should evaluate to `2'. In the cases where 33030 the value is defined, VALUE should be set to this value. 33031 33032 If this macro is not defined, the value of `clz' or `ctz' at zero 33033 is assumed to be undefined. 33034 33035 This macro must be defined if the target's expansion for `ffs' 33036 relies on a particular value to get correct results. Otherwise it 33037 is not necessary, though it may be used to optimize some corner 33038 cases, and to provide a default expansion for the `ffs' optab. 33039 33040 Note that regardless of this macro the "definedness" of `clz' and 33041 `ctz' at zero do _not_ extend to the builtin functions visible to 33042 the user. Thus one may be free to adjust the value at will to 33043 match the target expansion of these operations without fear of 33044 breaking the API. 33045 33046 -- Macro: Pmode 33047 An alias for the machine mode for pointers. On most machines, 33048 define this to be the integer mode corresponding to the width of a 33049 hardware pointer; `SImode' on 32-bit machine or `DImode' on 64-bit 33050 machines. On some machines you must define this to be one of the 33051 partial integer modes, such as `PSImode'. 33052 33053 The width of `Pmode' must be at least as large as the value of 33054 `POINTER_SIZE'. If it is not equal, you must define the macro 33055 `POINTERS_EXTEND_UNSIGNED' to specify how pointers are extended to 33056 `Pmode'. 33057 33058 -- Macro: FUNCTION_MODE 33059 An alias for the machine mode used for memory references to 33060 functions being called, in `call' RTL expressions. On most CISC 33061 machines, where an instruction can begin at any byte address, this 33062 should be `QImode'. On most RISC machines, where all instructions 33063 have fixed size and alignment, this should be a mode with the same 33064 size and alignment as the machine instruction words - typically 33065 `SImode' or `HImode'. 33066 33067 -- Macro: STDC_0_IN_SYSTEM_HEADERS 33068 In normal operation, the preprocessor expands `__STDC__' to the 33069 constant 1, to signify that GCC conforms to ISO Standard C. On 33070 some hosts, like Solaris, the system compiler uses a different 33071 convention, where `__STDC__' is normally 0, but is 1 if the user 33072 specifies strict conformance to the C Standard. 33073 33074 Defining `STDC_0_IN_SYSTEM_HEADERS' makes GNU CPP follows the host 33075 convention when processing system header files, but when 33076 processing user files `__STDC__' will always expand to 1. 33077 33078 -- Macro: NO_IMPLICIT_EXTERN_C 33079 Define this macro if the system header files support C++ as well 33080 as C. This macro inhibits the usual method of using system header 33081 files in C++, which is to pretend that the file's contents are 33082 enclosed in `extern "C" {...}'. 33083 33084 -- Macro: REGISTER_TARGET_PRAGMAS () 33085 Define this macro if you want to implement any target-specific 33086 pragmas. If defined, it is a C expression which makes a series of 33087 calls to `c_register_pragma' or `c_register_pragma_with_expansion' 33088 for each pragma. The macro may also do any setup required for the 33089 pragmas. 33090 33091 The primary reason to define this macro is to provide 33092 compatibility with other compilers for the same target. In 33093 general, we discourage definition of target-specific pragmas for 33094 GCC. 33095 33096 If the pragma can be implemented by attributes then you should 33097 consider defining the target hook `TARGET_INSERT_ATTRIBUTES' as 33098 well. 33099 33100 Preprocessor macros that appear on pragma lines are not expanded. 33101 All `#pragma' directives that do not match any registered pragma 33102 are silently ignored, unless the user specifies 33103 `-Wunknown-pragmas'. 33104 33105 -- Function: void c_register_pragma (const char *SPACE, const char 33106 *NAME, void (*CALLBACK) (struct cpp_reader *)) 33107 -- Function: void c_register_pragma_with_expansion (const char *SPACE, 33108 const char *NAME, void (*CALLBACK) (struct cpp_reader *)) 33109 Each call to `c_register_pragma' or 33110 `c_register_pragma_with_expansion' establishes one pragma. The 33111 CALLBACK routine will be called when the preprocessor encounters a 33112 pragma of the form 33113 33114 #pragma [SPACE] NAME ... 33115 33116 SPACE is the case-sensitive namespace of the pragma, or `NULL' to 33117 put the pragma in the global namespace. The callback routine 33118 receives PFILE as its first argument, which can be passed on to 33119 cpplib's functions if necessary. You can lex tokens after the 33120 NAME by calling `pragma_lex'. Tokens that are not read by the 33121 callback will be silently ignored. The end of the line is 33122 indicated by a token of type `CPP_EOF'. Macro expansion occurs on 33123 the arguments of pragmas registered with 33124 `c_register_pragma_with_expansion' but not on the arguments of 33125 pragmas registered with `c_register_pragma'. 33126 33127 Note that the use of `pragma_lex' is specific to the C and C++ 33128 compilers. It will not work in the Java or Fortran compilers, or 33129 any other language compilers for that matter. Thus if 33130 `pragma_lex' is going to be called from target-specific code, it 33131 must only be done so when building the C and C++ compilers. This 33132 can be done by defining the variables `c_target_objs' and 33133 `cxx_target_objs' in the target entry in the `config.gcc' file. 33134 These variables should name the target-specific, language-specific 33135 object file which contains the code that uses `pragma_lex'. Note 33136 it will also be necessary to add a rule to the makefile fragment 33137 pointed to by `tmake_file' that shows how to build this object 33138 file. 33139 33140 -- Macro: HANDLE_SYSV_PRAGMA 33141 Define this macro (to a value of 1) if you want the System V style 33142 pragmas `#pragma pack(<n>)' and `#pragma weak <name> [=<value>]' 33143 to be supported by gcc. 33144 33145 The pack pragma specifies the maximum alignment (in bytes) of 33146 fields within a structure, in much the same way as the 33147 `__aligned__' and `__packed__' `__attribute__'s do. A pack value 33148 of zero resets the behavior to the default. 33149 33150 A subtlety for Microsoft Visual C/C++ style bit-field packing 33151 (e.g. -mms-bitfields) for targets that support it: When a 33152 bit-field is inserted into a packed record, the whole size of the 33153 underlying type is used by one or more same-size adjacent 33154 bit-fields (that is, if its long:3, 32 bits is used in the record, 33155 and any additional adjacent long bit-fields are packed into the 33156 same chunk of 32 bits. However, if the size changes, a new field 33157 of that size is allocated). 33158 33159 If both MS bit-fields and `__attribute__((packed))' are used, the 33160 latter will take precedence. If `__attribute__((packed))' is used 33161 on a single field when MS bit-fields are in use, it will take 33162 precedence for that field, but the alignment of the rest of the 33163 structure may affect its placement. 33164 33165 The weak pragma only works if `SUPPORTS_WEAK' and 33166 `ASM_WEAKEN_LABEL' are defined. If enabled it allows the creation 33167 of specifically named weak labels, optionally with a value. 33168 33169 -- Macro: HANDLE_PRAGMA_PACK_PUSH_POP 33170 Define this macro (to a value of 1) if you want to support the 33171 Win32 style pragmas `#pragma pack(push[,N])' and `#pragma 33172 pack(pop)'. The `pack(push,[N])' pragma specifies the maximum 33173 alignment (in bytes) of fields within a structure, in much the 33174 same way as the `__aligned__' and `__packed__' `__attribute__'s 33175 do. A pack value of zero resets the behavior to the default. 33176 Successive invocations of this pragma cause the previous values to 33177 be stacked, so that invocations of `#pragma pack(pop)' will return 33178 to the previous value. 33179 33180 -- Macro: HANDLE_PRAGMA_PACK_WITH_EXPANSION 33181 Define this macro, as well as `HANDLE_SYSV_PRAGMA', if macros 33182 should be expanded in the arguments of `#pragma pack'. 33183 33184 -- Macro: TARGET_DEFAULT_PACK_STRUCT 33185 If your target requires a structure packing default other than 0 33186 (meaning the machine default), define this macro to the necessary 33187 value (in bytes). This must be a value that would also be valid 33188 to use with `#pragma pack()' (that is, a small power of two). 33189 33190 -- Macro: DOLLARS_IN_IDENTIFIERS 33191 Define this macro to control use of the character `$' in 33192 identifier names for the C family of languages. 0 means `$' is 33193 not allowed by default; 1 means it is allowed. 1 is the default; 33194 there is no need to define this macro in that case. 33195 33196 -- Macro: NO_DOLLAR_IN_LABEL 33197 Define this macro if the assembler does not accept the character 33198 `$' in label names. By default constructors and destructors in 33199 G++ have `$' in the identifiers. If this macro is defined, `.' is 33200 used instead. 33201 33202 -- Macro: NO_DOT_IN_LABEL 33203 Define this macro if the assembler does not accept the character 33204 `.' in label names. By default constructors and destructors in G++ 33205 have names that use `.'. If this macro is defined, these names 33206 are rewritten to avoid `.'. 33207 33208 -- Macro: INSN_SETS_ARE_DELAYED (INSN) 33209 Define this macro as a C expression that is nonzero if it is safe 33210 for the delay slot scheduler to place instructions in the delay 33211 slot of INSN, even if they appear to use a resource set or 33212 clobbered in INSN. INSN is always a `jump_insn' or an `insn'; GCC 33213 knows that every `call_insn' has this behavior. On machines where 33214 some `insn' or `jump_insn' is really a function call and hence has 33215 this behavior, you should define this macro. 33216 33217 You need not define this macro if it would always return zero. 33218 33219 -- Macro: INSN_REFERENCES_ARE_DELAYED (INSN) 33220 Define this macro as a C expression that is nonzero if it is safe 33221 for the delay slot scheduler to place instructions in the delay 33222 slot of INSN, even if they appear to set or clobber a resource 33223 referenced in INSN. INSN is always a `jump_insn' or an `insn'. 33224 On machines where some `insn' or `jump_insn' is really a function 33225 call and its operands are registers whose use is actually in the 33226 subroutine it calls, you should define this macro. Doing so 33227 allows the delay slot scheduler to move instructions which copy 33228 arguments into the argument registers into the delay slot of INSN. 33229 33230 You need not define this macro if it would always return zero. 33231 33232 -- Macro: MULTIPLE_SYMBOL_SPACES 33233 Define this macro as a C expression that is nonzero if, in some 33234 cases, global symbols from one translation unit may not be bound 33235 to undefined symbols in another translation unit without user 33236 intervention. For instance, under Microsoft Windows symbols must 33237 be explicitly imported from shared libraries (DLLs). 33238 33239 You need not define this macro if it would always evaluate to zero. 33240 33241 -- Target Hook: tree TARGET_MD_ASM_CLOBBERS (tree OUTPUTS, tree 33242 INPUTS, tree CLOBBERS) 33243 This target hook should add to CLOBBERS `STRING_CST' trees for any 33244 hard regs the port wishes to automatically clobber for an asm. It 33245 should return the result of the last `tree_cons' used to add a 33246 clobber. The OUTPUTS, INPUTS and CLOBBER lists are the 33247 corresponding parameters to the asm and may be inspected to avoid 33248 clobbering a register that is an input or output of the asm. You 33249 can use `tree_overlaps_hard_reg_set', declared in `tree.h', to test 33250 for overlap with regards to asm-declared registers. 33251 33252 -- Macro: MATH_LIBRARY 33253 Define this macro as a C string constant for the linker argument 33254 to link in the system math library, or `""' if the target does not 33255 have a separate math library. 33256 33257 You need only define this macro if the default of `"-lm"' is wrong. 33258 33259 -- Macro: LIBRARY_PATH_ENV 33260 Define this macro as a C string constant for the environment 33261 variable that specifies where the linker should look for libraries. 33262 33263 You need only define this macro if the default of `"LIBRARY_PATH"' 33264 is wrong. 33265 33266 -- Macro: TARGET_POSIX_IO 33267 Define this macro if the target supports the following POSIX file 33268 functions, access, mkdir and file locking with fcntl / F_SETLKW. 33269 Defining `TARGET_POSIX_IO' will enable the test coverage code to 33270 use file locking when exiting a program, which avoids race 33271 conditions if the program has forked. It will also create 33272 directories at run-time for cross-profiling. 33273 33274 -- Macro: MAX_CONDITIONAL_EXECUTE 33275 A C expression for the maximum number of instructions to execute 33276 via conditional execution instructions instead of a branch. A 33277 value of `BRANCH_COST'+1 is the default if the machine does not 33278 use cc0, and 1 if it does use cc0. 33279 33280 -- Macro: IFCVT_MODIFY_TESTS (CE_INFO, TRUE_EXPR, FALSE_EXPR) 33281 Used if the target needs to perform machine-dependent 33282 modifications on the conditionals used for turning basic blocks 33283 into conditionally executed code. CE_INFO points to a data 33284 structure, `struct ce_if_block', which contains information about 33285 the currently processed blocks. TRUE_EXPR and FALSE_EXPR are the 33286 tests that are used for converting the then-block and the 33287 else-block, respectively. Set either TRUE_EXPR or FALSE_EXPR to a 33288 null pointer if the tests cannot be converted. 33289 33290 -- Macro: IFCVT_MODIFY_MULTIPLE_TESTS (CE_INFO, BB, TRUE_EXPR, 33291 FALSE_EXPR) 33292 Like `IFCVT_MODIFY_TESTS', but used when converting more 33293 complicated if-statements into conditions combined by `and' and 33294 `or' operations. BB contains the basic block that contains the 33295 test that is currently being processed and about to be turned into 33296 a condition. 33297 33298 -- Macro: IFCVT_MODIFY_INSN (CE_INFO, PATTERN, INSN) 33299 A C expression to modify the PATTERN of an INSN that is to be 33300 converted to conditional execution format. CE_INFO points to a 33301 data structure, `struct ce_if_block', which contains information 33302 about the currently processed blocks. 33303 33304 -- Macro: IFCVT_MODIFY_FINAL (CE_INFO) 33305 A C expression to perform any final machine dependent 33306 modifications in converting code to conditional execution. The 33307 involved basic blocks can be found in the `struct ce_if_block' 33308 structure that is pointed to by CE_INFO. 33309 33310 -- Macro: IFCVT_MODIFY_CANCEL (CE_INFO) 33311 A C expression to cancel any machine dependent modifications in 33312 converting code to conditional execution. The involved basic 33313 blocks can be found in the `struct ce_if_block' structure that is 33314 pointed to by CE_INFO. 33315 33316 -- Macro: IFCVT_INIT_EXTRA_FIELDS (CE_INFO) 33317 A C expression to initialize any extra fields in a `struct 33318 ce_if_block' structure, which are defined by the 33319 `IFCVT_EXTRA_FIELDS' macro. 33320 33321 -- Macro: IFCVT_EXTRA_FIELDS 33322 If defined, it should expand to a set of field declarations that 33323 will be added to the `struct ce_if_block' structure. These should 33324 be initialized by the `IFCVT_INIT_EXTRA_FIELDS' macro. 33325 33326 -- Target Hook: void TARGET_MACHINE_DEPENDENT_REORG () 33327 If non-null, this hook performs a target-specific pass over the 33328 instruction stream. The compiler will run it at all optimization 33329 levels, just before the point at which it normally does 33330 delayed-branch scheduling. 33331 33332 The exact purpose of the hook varies from target to target. Some 33333 use it to do transformations that are necessary for correctness, 33334 such as laying out in-function constant pools or avoiding hardware 33335 hazards. Others use it as an opportunity to do some 33336 machine-dependent optimizations. 33337 33338 You need not implement the hook if it has nothing to do. The 33339 default definition is null. 33340 33341 -- Target Hook: void TARGET_INIT_BUILTINS () 33342 Define this hook if you have any machine-specific built-in 33343 functions that need to be defined. It should be a function that 33344 performs the necessary setup. 33345 33346 Machine specific built-in functions can be useful to expand 33347 special machine instructions that would otherwise not normally be 33348 generated because they have no equivalent in the source language 33349 (for example, SIMD vector instructions or prefetch instructions). 33350 33351 To create a built-in function, call the function 33352 `lang_hooks.builtin_function' which is defined by the language 33353 front end. You can use any type nodes set up by 33354 `build_common_tree_nodes' and `build_common_tree_nodes_2'; only 33355 language front ends that use those two functions will call 33356 `TARGET_INIT_BUILTINS'. 33357 33358 -- Target Hook: rtx TARGET_EXPAND_BUILTIN (tree EXP, rtx TARGET, rtx 33359 SUBTARGET, enum machine_mode MODE, int IGNORE) 33360 Expand a call to a machine specific built-in function that was set 33361 up by `TARGET_INIT_BUILTINS'. EXP is the expression for the 33362 function call; the result should go to TARGET if that is 33363 convenient, and have mode MODE if that is convenient. SUBTARGET 33364 may be used as the target for computing one of EXP's operands. 33365 IGNORE is nonzero if the value is to be ignored. This function 33366 should return the result of the call to the built-in function. 33367 33368 -- Target Hook: tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree FNDECL, 33369 tree ARGLIST) 33370 Select a replacement for a machine specific built-in function that 33371 was set up by `TARGET_INIT_BUILTINS'. This is done _before_ 33372 regular type checking, and so allows the target to implement a 33373 crude form of function overloading. FNDECL is the declaration of 33374 the built-in function. ARGLIST is the list of arguments passed to 33375 the built-in function. The result is a complete expression that 33376 implements the operation, usually another `CALL_EXPR'. 33377 33378 -- Target Hook: tree TARGET_FOLD_BUILTIN (tree FNDECL, tree ARGLIST, 33379 bool IGNORE) 33380 Fold a call to a machine specific built-in function that was set 33381 up by `TARGET_INIT_BUILTINS'. FNDECL is the declaration of the 33382 built-in function. ARGLIST is the list of arguments passed to the 33383 built-in function. The result is another tree containing a 33384 simplified expression for the call's result. If IGNORE is true 33385 the value will be ignored. 33386 33387 -- Target Hook: const char * TARGET_INVALID_WITHIN_DOLOOP (rtx INSN) 33388 Take an instruction in INSN and return NULL if it is valid within a 33389 low-overhead loop, otherwise return a string why doloop could not 33390 be applied. 33391 33392 Many targets use special registers for low-overhead looping. For 33393 any instruction that clobbers these this function should return a 33394 string indicating the reason why the doloop could not be applied. 33395 By default, the RTL loop optimizer does not use a present doloop 33396 pattern for loops containing function calls or branch on table 33397 instructions. 33398 33399 -- Macro: MD_CAN_REDIRECT_BRANCH (BRANCH1, BRANCH2) 33400 Take a branch insn in BRANCH1 and another in BRANCH2. Return true 33401 if redirecting BRANCH1 to the destination of BRANCH2 is possible. 33402 33403 On some targets, branches may have a limited range. Optimizing the 33404 filling of delay slots can result in branches being redirected, 33405 and this may in turn cause a branch offset to overflow. 33406 33407 -- Target Hook: bool TARGET_COMMUTATIVE_P (rtx X, OUTER_CODE) 33408 This target hook returns `true' if X is considered to be 33409 commutative. Usually, this is just COMMUTATIVE_P (X), but the HP 33410 PA doesn't consider PLUS to be commutative inside a MEM. 33411 OUTER_CODE is the rtx code of the enclosing rtl, if known, 33412 otherwise it is UNKNOWN. 33413 33414 -- Target Hook: rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx HARD_REG) 33415 When the initial value of a hard register has been copied in a 33416 pseudo register, it is often not necessary to actually allocate 33417 another register to this pseudo register, because the original 33418 hard register or a stack slot it has been saved into can be used. 33419 `TARGET_ALLOCATE_INITIAL_VALUE' is called at the start of register 33420 allocation once for each hard register that had its initial value 33421 copied by using `get_func_hard_reg_initial_val' or 33422 `get_hard_reg_initial_val'. Possible values are `NULL_RTX', if 33423 you don't want to do any special allocation, a `REG' rtx--that 33424 would typically be the hard register itself, if it is known not to 33425 be clobbered--or a `MEM'. If you are returning a `MEM', this is 33426 only a hint for the allocator; it might decide to use another 33427 register anyways. You may use `current_function_leaf_function' in 33428 the hook, functions that use `REG_N_SETS', to determine if the hard 33429 register in question will not be clobbered. The default value of 33430 this hook is `NULL', which disables any special allocation. 33431 33432 -- Target Hook: int TARGET_UNSPEC_MAY_TRAP_P (const_rtx X, unsigned 33433 FLAGS) 33434 This target hook returns nonzero if X, an `unspec' or 33435 `unspec_volatile' operation, might cause a trap. Targets can use 33436 this hook to enhance precision of analysis for `unspec' and 33437 `unspec_volatile' operations. You may call `may_trap_p_1' to 33438 analyze inner elements of X in which case FLAGS should be passed 33439 along. 33440 33441 -- Target Hook: void TARGET_SET_CURRENT_FUNCTION (tree DECL) 33442 The compiler invokes this hook whenever it changes its current 33443 function context (`cfun'). You can define this function if the 33444 back end needs to perform any initialization or reset actions on a 33445 per-function basis. For example, it may be used to implement 33446 function attributes that affect register usage or code generation 33447 patterns. The argument DECL is the declaration for the new 33448 function context, and may be null to indicate that the compiler 33449 has left a function context and is returning to processing at the 33450 top level. The default hook function does nothing. 33451 33452 GCC sets `cfun' to a dummy function context during initialization 33453 of some parts of the back end. The hook function is not invoked 33454 in this situation; you need not worry about the hook being invoked 33455 recursively, or when the back end is in a partially-initialized 33456 state. 33457 33458 -- Macro: TARGET_OBJECT_SUFFIX 33459 Define this macro to be a C string representing the suffix for 33460 object files on your target machine. If you do not define this 33461 macro, GCC will use `.o' as the suffix for object files. 33462 33463 -- Macro: TARGET_EXECUTABLE_SUFFIX 33464 Define this macro to be a C string representing the suffix to be 33465 automatically added to executable files on your target machine. 33466 If you do not define this macro, GCC will use the null string as 33467 the suffix for executable files. 33468 33469 -- Macro: COLLECT_EXPORT_LIST 33470 If defined, `collect2' will scan the individual object files 33471 specified on its command line and create an export list for the 33472 linker. Define this macro for systems like AIX, where the linker 33473 discards object files that are not referenced from `main' and uses 33474 export lists. 33475 33476 -- Macro: MODIFY_JNI_METHOD_CALL (MDECL) 33477 Define this macro to a C expression representing a variant of the 33478 method call MDECL, if Java Native Interface (JNI) methods must be 33479 invoked differently from other methods on your target. For 33480 example, on 32-bit Microsoft Windows, JNI methods must be invoked 33481 using the `stdcall' calling convention and this macro is then 33482 defined as this expression: 33483 33484 build_type_attribute_variant (MDECL, 33485 build_tree_list 33486 (get_identifier ("stdcall"), 33487 NULL)) 33488 33489 -- Target Hook: bool TARGET_CANNOT_MODIFY_JUMPS_P (void) 33490 This target hook returns `true' past the point in which new jump 33491 instructions could be created. On machines that require a 33492 register for every jump such as the SHmedia ISA of SH5, this point 33493 would typically be reload, so this target hook should be defined 33494 to a function such as: 33495 33496 static bool 33497 cannot_modify_jumps_past_reload_p () 33498 { 33499 return (reload_completed || reload_in_progress); 33500 } 33501 33502 -- Target Hook: int TARGET_BRANCH_TARGET_REGISTER_CLASS (void) 33503 This target hook returns a register class for which branch target 33504 register optimizations should be applied. All registers in this 33505 class should be usable interchangeably. After reload, registers 33506 in this class will be re-allocated and loads will be hoisted out 33507 of loops and be subjected to inter-block scheduling. 33508 33509 -- Target Hook: bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool 33510 AFTER_PROLOGUE_EPILOGUE_GEN) 33511 Branch target register optimization will by default exclude 33512 callee-saved registers that are not already live during the 33513 current function; if this target hook returns true, they will be 33514 included. The target code must than make sure that all target 33515 registers in the class returned by 33516 `TARGET_BRANCH_TARGET_REGISTER_CLASS' that might need saving are 33517 saved. AFTER_PROLOGUE_EPILOGUE_GEN indicates if prologues and 33518 epilogues have already been generated. Note, even if you only 33519 return true when AFTER_PROLOGUE_EPILOGUE_GEN is false, you still 33520 are likely to have to make special provisions in 33521 `INITIAL_ELIMINATION_OFFSET' to reserve space for caller-saved 33522 target registers. 33523 33524 -- Target Hook: bool TARGET_HAVE_CONDITIONAL_EXECUTION (void) 33525 This target hook returns true if the target supports conditional 33526 execution. This target hook is required only when the target has 33527 several different modes and they have different conditional 33528 execution capability, such as ARM. 33529 33530 -- Macro: POWI_MAX_MULTS 33531 If defined, this macro is interpreted as a signed integer C 33532 expression that specifies the maximum number of floating point 33533 multiplications that should be emitted when expanding 33534 exponentiation by an integer constant inline. When this value is 33535 defined, exponentiation requiring more than this number of 33536 multiplications is implemented by calling the system library's 33537 `pow', `powf' or `powl' routines. The default value places no 33538 upper bound on the multiplication count. 33539 33540 -- Macro: void TARGET_EXTRA_INCLUDES (const char *SYSROOT, const char 33541 *IPREFIX, int STDINC) 33542 This target hook should register any extra include files for the 33543 target. The parameter STDINC indicates if normal include files 33544 are present. The parameter SYSROOT is the system root directory. 33545 The parameter IPREFIX is the prefix for the gcc directory. 33546 33547 -- Macro: void TARGET_EXTRA_PRE_INCLUDES (const char *SYSROOT, const 33548 char *IPREFIX, int STDINC) 33549 This target hook should register any extra include files for the 33550 target before any standard headers. The parameter STDINC 33551 indicates if normal include files are present. The parameter 33552 SYSROOT is the system root directory. The parameter IPREFIX is 33553 the prefix for the gcc directory. 33554 33555 -- Macro: void TARGET_OPTF (char *PATH) 33556 This target hook should register special include paths for the 33557 target. The parameter PATH is the include to register. On Darwin 33558 systems, this is used for Framework includes, which have semantics 33559 that are different from `-I'. 33560 33561 -- Target Hook: bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree FNDECL) 33562 This target hook returns `true' if it is safe to use a local alias 33563 for a virtual function FNDECL when constructing thunks, `false' 33564 otherwise. By default, the hook returns `true' for all functions, 33565 if a target supports aliases (i.e. defines `ASM_OUTPUT_DEF'), 33566 `false' otherwise, 33567 33568 -- Macro: TARGET_FORMAT_TYPES 33569 If defined, this macro is the name of a global variable containing 33570 target-specific format checking information for the `-Wformat' 33571 option. The default is to have no target-specific format checks. 33572 33573 -- Macro: TARGET_N_FORMAT_TYPES 33574 If defined, this macro is the number of entries in 33575 `TARGET_FORMAT_TYPES'. 33576 33577 -- Macro: TARGET_OVERRIDES_FORMAT_ATTRIBUTES 33578 If defined, this macro is the name of a global variable containing 33579 target-specific format overrides for the `-Wformat' option. The 33580 default is to have no target-specific format overrides. If defined, 33581 `TARGET_FORMAT_TYPES' must be defined, too. 33582 33583 -- Macro: TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT 33584 If defined, this macro specifies the number of entries in 33585 `TARGET_OVERRIDES_FORMAT_ATTRIBUTES'. 33586 33587 -- Macro: TARGET_OVERRIDES_FORMAT_INIT 33588 If defined, this macro specifies the optional initialization 33589 routine for target specific customizations of the system printf 33590 and scanf formatter settings. 33591 33592 -- Target Hook: bool TARGET_RELAXED_ORDERING 33593 If set to `true', means that the target's memory model does not 33594 guarantee that loads which do not depend on one another will access 33595 main memory in the order of the instruction stream; if ordering is 33596 important, an explicit memory barrier must be used. This is true 33597 of many recent processors which implement a policy of "relaxed," 33598 "weak," or "release" memory consistency, such as Alpha, PowerPC, 33599 and ia64. The default is `false'. 33600 33601 -- Target Hook: const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN 33602 (tree TYPELIST, tree FUNCDECL, tree VAL) 33603 If defined, this macro returns the diagnostic message when it is 33604 illegal to pass argument VAL to function FUNCDECL with prototype 33605 TYPELIST. 33606 33607 -- Target Hook: const char * TARGET_INVALID_CONVERSION (tree FROMTYPE, 33608 tree TOTYPE) 33609 If defined, this macro returns the diagnostic message when it is 33610 invalid to convert from FROMTYPE to TOTYPE, or `NULL' if validity 33611 should be determined by the front end. 33612 33613 -- Target Hook: const char * TARGET_INVALID_UNARY_OP (int OP, tree 33614 TYPE) 33615 If defined, this macro returns the diagnostic message when it is 33616 invalid to apply operation OP (where unary plus is denoted by 33617 `CONVERT_EXPR') to an operand of type TYPE, or `NULL' if validity 33618 should be determined by the front end. 33619 33620 -- Target Hook: const char * TARGET_INVALID_BINARY_OP (int OP, tree 33621 TYPE1, tree TYPE2) 33622 If defined, this macro returns the diagnostic message when it is 33623 invalid to apply operation OP to operands of types TYPE1 and 33624 TYPE2, or `NULL' if validity should be determined by the front end. 33625 33626 -- Macro: TARGET_USE_JCR_SECTION 33627 This macro determines whether to use the JCR section to register 33628 Java classes. By default, TARGET_USE_JCR_SECTION is defined to 1 33629 if both SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, 33630 else 0. 33631 33632 -- Macro: OBJC_JBLEN 33633 This macro determines the size of the objective C jump buffer for 33634 the NeXT runtime. By default, OBJC_JBLEN is defined to an 33635 innocuous value. 33636 33637 -- Macro: LIBGCC2_UNWIND_ATTRIBUTE 33638 Define this macro if any target-specific attributes need to be 33639 attached to the functions in `libgcc' that provide low-level 33640 support for call stack unwinding. It is used in declarations in 33641 `unwind-generic.h' and the associated definitions of those 33642 functions. 33643 33644 -- Target Hook: void TARGET_UPDATE_STACK_BOUNDARY (void) 33645 Define this macro to update the current function stack boundary if 33646 necessary. 33647 33648 -- Target Hook: rtx TARGET_GET_DRAP_RTX (void) 33649 Define this macro to an rtx for Dynamic Realign Argument Pointer 33650 if a different argument pointer register is needed to access the 33651 function's argument list when stack is aligned. 33652 33653 -- Target Hook: bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void) 33654 When optimization is disabled, this hook indicates whether or not 33655 arguments should be allocated to stack slots. Normally, GCC 33656 allocates stacks slots for arguments when not optimizing in order 33657 to make debugging easier. However, when a function is declared 33658 with `__attribute__((naked))', there is no stack frame, and the 33659 compiler cannot safely move arguments from the registers in which 33660 they are passed to the stack. Therefore, this hook should return 33661 true in general, but false for naked functions. The default 33662 implementation always returns true. 33663 33664 -- Target Hook: rtx TARGET_GET_PIC_REG (void) 33665 Return the pic_reg pseudo register which holds the base address of 33666 GOT. It is only required by the simplify-got optimization. 33667 33668 -- Target Hook: void TARGET_CLEAR_PIC_REG (void) 33669 After successful simplify-got optimization, the pic_reg is 33670 useless. So a target can use this hook to clear pic_reg. 33671 33672 -- Target Hook: rtx TARGET_LOADED_GLOBAL_VAR (rtx INSN, rtx * 33673 OFFSET_REG, rtx * OFFSET_INSN) 33674 This hook is used to detect if the given INSN loads a global 33675 variable's address from GOT with the form of 33676 33677 (set ADDRESS_REG (mem (plus pic_reg OFFSET_REG))) 33678 33679 If so return the global variable whose address will be loaded and 33680 fill in OFFSET_INSN and OFFSET_REG. OFFSET_REG is set at 33681 OFFSET_INSN to hold the offset from GOT base to the GOT entry of 33682 the global variable. Otherwise return `NULL_RTX'. 33683 33684 -- Target Hook: bool TARGET_CAN_SIMPLIFY_GOT_ACCESS (int N_SYMBOL, int 33685 N_ACCESS) 33686 This hook determines if it satisfy the target dependent conditions 33687 to do simplify-got when given the number of global variable 33688 accessing and the number of accessed symbols. If the returned 33689 value is false the GOT access insns will not be rewritten. 33690 Otherwise we will rewrite these insns. 33691 33692 -- Target Hook: void TARGET_LOAD_GLOBAL_ADDRESS (rtx SYMBOL, rtx 33693 OFFSET_REG, rtx ADDRESS_REG, rtx LOAD_INSN, rtx OFFSET_INSN) 33694 This hook does the actual rewriting of GOT access insn LOAD_INSN. 33695 The global variable is SYMBOL. The global address should be loaded 33696 into ADDRESS_REG. The register OFFSET_REG was previously set in 33697 insn OFFSET_INSN to hold the offset from GOT base to the GOT entry 33698 of the global variable. Now it can be used as a scratch register. 33699 33700 33701 File: gccint.info, Node: Host Config, Next: Fragments, Prev: Target Macros, Up: Top 33702 33703 18 Host Configuration 33704 ********************* 33705 33706 Most details about the machine and system on which the compiler is 33707 actually running are detected by the `configure' script. Some things 33708 are impossible for `configure' to detect; these are described in two 33709 ways, either by macros defined in a file named `xm-MACHINE.h' or by 33710 hook functions in the file specified by the OUT_HOST_HOOK_OBJ variable 33711 in `config.gcc'. (The intention is that very few hosts will need a 33712 header file but nearly every fully supported host will need to override 33713 some hooks.) 33714 33715 If you need to define only a few macros, and they have simple 33716 definitions, consider using the `xm_defines' variable in your 33717 `config.gcc' entry instead of creating a host configuration header. 33718 *Note System Config::. 33719 33720 * Menu: 33721 33722 * Host Common:: Things every host probably needs implemented. 33723 * Filesystem:: Your host can't have the letter `a' in filenames? 33724 * Host Misc:: Rare configuration options for hosts. 33725 33726 33727 File: gccint.info, Node: Host Common, Next: Filesystem, Up: Host Config 33728 33729 18.1 Host Common 33730 ================ 33731 33732 Some things are just not portable, even between similar operating 33733 systems, and are too difficult for autoconf to detect. They get 33734 implemented using hook functions in the file specified by the 33735 HOST_HOOK_OBJ variable in `config.gcc'. 33736 33737 -- Host Hook: void HOST_HOOKS_EXTRA_SIGNALS (void) 33738 This host hook is used to set up handling for extra signals. The 33739 most common thing to do in this hook is to detect stack overflow. 33740 33741 -- Host Hook: void * HOST_HOOKS_GT_PCH_GET_ADDRESS (size_t SIZE, int 33742 FD) 33743 This host hook returns the address of some space that is likely to 33744 be free in some subsequent invocation of the compiler. We intend 33745 to load the PCH data at this address such that the data need not 33746 be relocated. The area should be able to hold SIZE bytes. If the 33747 host uses `mmap', FD is an open file descriptor that can be used 33748 for probing. 33749 33750 -- Host Hook: int HOST_HOOKS_GT_PCH_USE_ADDRESS (void * ADDRESS, 33751 size_t SIZE, int FD, size_t OFFSET) 33752 This host hook is called when a PCH file is about to be loaded. 33753 We want to load SIZE bytes from FD at OFFSET into memory at 33754 ADDRESS. The given address will be the result of a previous 33755 invocation of `HOST_HOOKS_GT_PCH_GET_ADDRESS'. Return -1 if we 33756 couldn't allocate SIZE bytes at ADDRESS. Return 0 if the memory 33757 is allocated but the data is not loaded. Return 1 if the hook has 33758 performed everything. 33759 33760 If the implementation uses reserved address space, free any 33761 reserved space beyond SIZE, regardless of the return value. If no 33762 PCH will be loaded, this hook may be called with SIZE zero, in 33763 which case all reserved address space should be freed. 33764 33765 Do not try to handle values of ADDRESS that could not have been 33766 returned by this executable; just return -1. Such values usually 33767 indicate an out-of-date PCH file (built by some other GCC 33768 executable), and such a PCH file won't work. 33769 33770 -- Host Hook: size_t HOST_HOOKS_GT_PCH_ALLOC_GRANULARITY (void); 33771 This host hook returns the alignment required for allocating 33772 virtual memory. Usually this is the same as getpagesize, but on 33773 some hosts the alignment for reserving memory differs from the 33774 pagesize for committing memory. 33775 33776 33777 File: gccint.info, Node: Filesystem, Next: Host Misc, Prev: Host Common, Up: Host Config 33778 33779 18.2 Host Filesystem 33780 ==================== 33781 33782 GCC needs to know a number of things about the semantics of the host 33783 machine's filesystem. Filesystems with Unix and MS-DOS semantics are 33784 automatically detected. For other systems, you can define the 33785 following macros in `xm-MACHINE.h'. 33786 33787 `HAVE_DOS_BASED_FILE_SYSTEM' 33788 This macro is automatically defined by `system.h' if the host file 33789 system obeys the semantics defined by MS-DOS instead of Unix. DOS 33790 file systems are case insensitive, file specifications may begin 33791 with a drive letter, and both forward slash and backslash (`/' and 33792 `\') are directory separators. 33793 33794 `DIR_SEPARATOR' 33795 `DIR_SEPARATOR_2' 33796 If defined, these macros expand to character constants specifying 33797 separators for directory names within a file specification. 33798 `system.h' will automatically give them appropriate values on Unix 33799 and MS-DOS file systems. If your file system is neither of these, 33800 define one or both appropriately in `xm-MACHINE.h'. 33801 33802 However, operating systems like VMS, where constructing a pathname 33803 is more complicated than just stringing together directory names 33804 separated by a special character, should not define either of these 33805 macros. 33806 33807 `PATH_SEPARATOR' 33808 If defined, this macro should expand to a character constant 33809 specifying the separator for elements of search paths. The default 33810 value is a colon (`:'). DOS-based systems usually, but not 33811 always, use semicolon (`;'). 33812 33813 `VMS' 33814 Define this macro if the host system is VMS. 33815 33816 `HOST_OBJECT_SUFFIX' 33817 Define this macro to be a C string representing the suffix for 33818 object files on your host machine. If you do not define this 33819 macro, GCC will use `.o' as the suffix for object files. 33820 33821 `HOST_EXECUTABLE_SUFFIX' 33822 Define this macro to be a C string representing the suffix for 33823 executable files on your host machine. If you do not define this 33824 macro, GCC will use the null string as the suffix for executable 33825 files. 33826 33827 `HOST_BIT_BUCKET' 33828 A pathname defined by the host operating system, which can be 33829 opened as a file and written to, but all the information written 33830 is discarded. This is commonly known as a "bit bucket" or "null 33831 device". If you do not define this macro, GCC will use 33832 `/dev/null' as the bit bucket. If the host does not support a bit 33833 bucket, define this macro to an invalid filename. 33834 33835 `UPDATE_PATH_HOST_CANONICALIZE (PATH)' 33836 If defined, a C statement (sans semicolon) that performs 33837 host-dependent canonicalization when a path used in a compilation 33838 driver or preprocessor is canonicalized. PATH is a malloc-ed path 33839 to be canonicalized. If the C statement does canonicalize PATH 33840 into a different buffer, the old path should be freed and the new 33841 buffer should have been allocated with malloc. 33842 33843 `DUMPFILE_FORMAT' 33844 Define this macro to be a C string representing the format to use 33845 for constructing the index part of debugging dump file names. The 33846 resultant string must fit in fifteen bytes. The full filename 33847 will be the concatenation of: the prefix of the assembler file 33848 name, the string resulting from applying this format to an index 33849 number, and a string unique to each dump file kind, e.g. `rtl'. 33850 33851 If you do not define this macro, GCC will use `.%02d.'. You should 33852 define this macro if using the default will create an invalid file 33853 name. 33854 33855 `DELETE_IF_ORDINARY' 33856 Define this macro to be a C statement (sans semicolon) that 33857 performs host-dependent removal of ordinary temp files in the 33858 compilation driver. 33859 33860 If you do not define this macro, GCC will use the default version. 33861 You should define this macro if the default version does not 33862 reliably remove the temp file as, for example, on VMS which allows 33863 multiple versions of a file. 33864 33865 `HOST_LACKS_INODE_NUMBERS' 33866 Define this macro if the host filesystem does not report 33867 meaningful inode numbers in struct stat. 33868 33869 33870 File: gccint.info, Node: Host Misc, Prev: Filesystem, Up: Host Config 33871 33872 18.3 Host Misc 33873 ============== 33874 33875 `FATAL_EXIT_CODE' 33876 A C expression for the status code to be returned when the compiler 33877 exits after serious errors. The default is the system-provided 33878 macro `EXIT_FAILURE', or `1' if the system doesn't define that 33879 macro. Define this macro only if these defaults are incorrect. 33880 33881 `SUCCESS_EXIT_CODE' 33882 A C expression for the status code to be returned when the compiler 33883 exits without serious errors. (Warnings are not serious errors.) 33884 The default is the system-provided macro `EXIT_SUCCESS', or `0' if 33885 the system doesn't define that macro. Define this macro only if 33886 these defaults are incorrect. 33887 33888 `USE_C_ALLOCA' 33889 Define this macro if GCC should use the C implementation of 33890 `alloca' provided by `libiberty.a'. This only affects how some 33891 parts of the compiler itself allocate memory. It does not change 33892 code generation. 33893 33894 When GCC is built with a compiler other than itself, the C `alloca' 33895 is always used. This is because most other implementations have 33896 serious bugs. You should define this macro only on a system where 33897 no stack-based `alloca' can possibly work. For instance, if a 33898 system has a small limit on the size of the stack, GCC's builtin 33899 `alloca' will not work reliably. 33900 33901 `COLLECT2_HOST_INITIALIZATION' 33902 If defined, a C statement (sans semicolon) that performs 33903 host-dependent initialization when `collect2' is being initialized. 33904 33905 `GCC_DRIVER_HOST_INITIALIZATION' 33906 If defined, a C statement (sans semicolon) that performs 33907 host-dependent initialization when a compilation driver is being 33908 initialized. 33909 33910 `HOST_LONG_LONG_FORMAT' 33911 If defined, the string used to indicate an argument of type `long 33912 long' to functions like `printf'. The default value is `"ll"'. 33913 33914 In addition, if `configure' generates an incorrect definition of any 33915 of the macros in `auto-host.h', you can override that definition in a 33916 host configuration header. If you need to do this, first see if it is 33917 possible to fix `configure'. 33918 33919 33920 File: gccint.info, Node: Fragments, Next: Collect2, Prev: Host Config, Up: Top 33921 33922 19 Makefile Fragments 33923 ********************* 33924 33925 When you configure GCC using the `configure' script, it will construct 33926 the file `Makefile' from the template file `Makefile.in'. When it does 33927 this, it can incorporate makefile fragments from the `config' 33928 directory. These are used to set Makefile parameters that are not 33929 amenable to being calculated by autoconf. The list of fragments to 33930 incorporate is set by `config.gcc' (and occasionally `config.build' and 33931 `config.host'); *Note System Config::. 33932 33933 Fragments are named either `t-TARGET' or `x-HOST', depending on 33934 whether they are relevant to configuring GCC to produce code for a 33935 particular target, or to configuring GCC to run on a particular host. 33936 Here TARGET and HOST are mnemonics which usually have some relationship 33937 to the canonical system name, but no formal connection. 33938 33939 If these files do not exist, it means nothing needs to be added for a 33940 given target or host. Most targets need a few `t-TARGET' fragments, 33941 but needing `x-HOST' fragments is rare. 33942 33943 * Menu: 33944 33945 * Target Fragment:: Writing `t-TARGET' files. 33946 * Host Fragment:: Writing `x-HOST' files. 33947 33948 33949 File: gccint.info, Node: Target Fragment, Next: Host Fragment, Up: Fragments 33950 33951 19.1 Target Makefile Fragments 33952 ============================== 33953 33954 Target makefile fragments can set these Makefile variables. 33955 33956 `LIBGCC2_CFLAGS' 33957 Compiler flags to use when compiling `libgcc2.c'. 33958 33959 `LIB2FUNCS_EXTRA' 33960 A list of source file names to be compiled or assembled and 33961 inserted into `libgcc.a'. 33962 33963 `Floating Point Emulation' 33964 To have GCC include software floating point libraries in `libgcc.a' 33965 define `FPBIT' and `DPBIT' along with a few rules as follows: 33966 # We want fine grained libraries, so use the new code 33967 # to build the floating point emulation libraries. 33968 FPBIT = fp-bit.c 33969 DPBIT = dp-bit.c 33970 33971 33972 fp-bit.c: $(srcdir)/config/fp-bit.c 33973 echo '#define FLOAT' > fp-bit.c 33974 cat $(srcdir)/config/fp-bit.c >> fp-bit.c 33975 33976 dp-bit.c: $(srcdir)/config/fp-bit.c 33977 cat $(srcdir)/config/fp-bit.c > dp-bit.c 33978 33979 You may need to provide additional #defines at the beginning of 33980 `fp-bit.c' and `dp-bit.c' to control target endianness and other 33981 options. 33982 33983 `CRTSTUFF_T_CFLAGS' 33984 Special flags used when compiling `crtstuff.c'. *Note 33985 Initialization::. 33986 33987 `CRTSTUFF_T_CFLAGS_S' 33988 Special flags used when compiling `crtstuff.c' for shared linking. 33989 Used if you use `crtbeginS.o' and `crtendS.o' in `EXTRA-PARTS'. 33990 *Note Initialization::. 33991 33992 `MULTILIB_OPTIONS' 33993 For some targets, invoking GCC in different ways produces objects 33994 that can not be linked together. For example, for some targets GCC 33995 produces both big and little endian code. For these targets, you 33996 must arrange for multiple versions of `libgcc.a' to be compiled, 33997 one for each set of incompatible options. When GCC invokes the 33998 linker, it arranges to link in the right version of `libgcc.a', 33999 based on the command line options used. 34000 34001 The `MULTILIB_OPTIONS' macro lists the set of options for which 34002 special versions of `libgcc.a' must be built. Write options that 34003 are mutually incompatible side by side, separated by a slash. 34004 Write options that may be used together separated by a space. The 34005 build procedure will build all combinations of compatible options. 34006 34007 For example, if you set `MULTILIB_OPTIONS' to `m68000/m68020 34008 msoft-float', `Makefile' will build special versions of `libgcc.a' 34009 using the following sets of options: `-m68000', `-m68020', 34010 `-msoft-float', `-m68000 -msoft-float', and `-m68020 -msoft-float'. 34011 34012 `MULTILIB_DIRNAMES' 34013 If `MULTILIB_OPTIONS' is used, this variable specifies the 34014 directory names that should be used to hold the various libraries. 34015 Write one element in `MULTILIB_DIRNAMES' for each element in 34016 `MULTILIB_OPTIONS'. If `MULTILIB_DIRNAMES' is not used, the 34017 default value will be `MULTILIB_OPTIONS', with all slashes treated 34018 as spaces. 34019 34020 For example, if `MULTILIB_OPTIONS' is set to `m68000/m68020 34021 msoft-float', then the default value of `MULTILIB_DIRNAMES' is 34022 `m68000 m68020 msoft-float'. You may specify a different value if 34023 you desire a different set of directory names. 34024 34025 `MULTILIB_MATCHES' 34026 Sometimes the same option may be written in two different ways. 34027 If an option is listed in `MULTILIB_OPTIONS', GCC needs to know 34028 about any synonyms. In that case, set `MULTILIB_MATCHES' to a 34029 list of items of the form `option=option' to describe all relevant 34030 synonyms. For example, `m68000=mc68000 m68020=mc68020'. 34031 34032 `MULTILIB_EXCEPTIONS' 34033 Sometimes when there are multiple sets of `MULTILIB_OPTIONS' being 34034 specified, there are combinations that should not be built. In 34035 that case, set `MULTILIB_EXCEPTIONS' to be all of the switch 34036 exceptions in shell case syntax that should not be built. 34037 34038 For example the ARM processor cannot execute both hardware floating 34039 point instructions and the reduced size THUMB instructions at the 34040 same time, so there is no need to build libraries with both of 34041 these options enabled. Therefore `MULTILIB_EXCEPTIONS' is set to: 34042 *mthumb/*mhard-float* 34043 34044 `MULTILIB_EXTRA_OPTS' 34045 Sometimes it is desirable that when building multiple versions of 34046 `libgcc.a' certain options should always be passed on to the 34047 compiler. In that case, set `MULTILIB_EXTRA_OPTS' to be the list 34048 of options to be used for all builds. If you set this, you should 34049 probably set `CRTSTUFF_T_CFLAGS' to a dash followed by it. 34050 34051 `NATIVE_SYSTEM_HEADER_DIR' 34052 If the default location for system headers is not `/usr/include', 34053 you must set this to the directory containing the headers. This 34054 value should match the value of the `SYSTEM_INCLUDE_DIR' macro. 34055 34056 `SPECS' 34057 Unfortunately, setting `MULTILIB_EXTRA_OPTS' is not enough, since 34058 it does not affect the build of target libraries, at least not the 34059 build of the default multilib. One possible work-around is to use 34060 `DRIVER_SELF_SPECS' to bring options from the `specs' file as if 34061 they had been passed in the compiler driver command line. 34062 However, you don't want to be adding these options after the 34063 toolchain is installed, so you can instead tweak the `specs' file 34064 that will be used during the toolchain build, while you still 34065 install the original, built-in `specs'. The trick is to set 34066 `SPECS' to some other filename (say `specs.install'), that will 34067 then be created out of the built-in specs, and introduce a 34068 `Makefile' rule to generate the `specs' file that's going to be 34069 used at build time out of your `specs.install'. 34070 34071 `T_CFLAGS' 34072 These are extra flags to pass to the C compiler. They are used 34073 both when building GCC, and when compiling things with the 34074 just-built GCC. This variable is deprecated and should not be 34075 used. 34076 34077 34078 File: gccint.info, Node: Host Fragment, Prev: Target Fragment, Up: Fragments 34079 34080 19.2 Host Makefile Fragments 34081 ============================ 34082 34083 The use of `x-HOST' fragments is discouraged. You should only use it 34084 for makefile dependencies. 34085 34086 34087 File: gccint.info, Node: Collect2, Next: Header Dirs, Prev: Fragments, Up: Top 34088 34089 20 `collect2' 34090 ************* 34091 34092 GCC uses a utility called `collect2' on nearly all systems to arrange 34093 to call various initialization functions at start time. 34094 34095 The program `collect2' works by linking the program once and looking 34096 through the linker output file for symbols with particular names 34097 indicating they are constructor functions. If it finds any, it creates 34098 a new temporary `.c' file containing a table of them, compiles it, and 34099 links the program a second time including that file. 34100 34101 The actual calls to the constructors are carried out by a subroutine 34102 called `__main', which is called (automatically) at the beginning of 34103 the body of `main' (provided `main' was compiled with GNU CC). Calling 34104 `__main' is necessary, even when compiling C code, to allow linking C 34105 and C++ object code together. (If you use `-nostdlib', you get an 34106 unresolved reference to `__main', since it's defined in the standard 34107 GCC library. Include `-lgcc' at the end of your compiler command line 34108 to resolve this reference.) 34109 34110 The program `collect2' is installed as `ld' in the directory where the 34111 passes of the compiler are installed. When `collect2' needs to find 34112 the _real_ `ld', it tries the following file names: 34113 34114 * `real-ld' in the directories listed in the compiler's search 34115 directories. 34116 34117 * `real-ld' in the directories listed in the environment variable 34118 `PATH'. 34119 34120 * The file specified in the `REAL_LD_FILE_NAME' configuration macro, 34121 if specified. 34122 34123 * `ld' in the compiler's search directories, except that `collect2' 34124 will not execute itself recursively. 34125 34126 * `ld' in `PATH'. 34127 34128 "The compiler's search directories" means all the directories where 34129 `gcc' searches for passes of the compiler. This includes directories 34130 that you specify with `-B'. 34131 34132 Cross-compilers search a little differently: 34133 34134 * `real-ld' in the compiler's search directories. 34135 34136 * `TARGET-real-ld' in `PATH'. 34137 34138 * The file specified in the `REAL_LD_FILE_NAME' configuration macro, 34139 if specified. 34140 34141 * `ld' in the compiler's search directories. 34142 34143 * `TARGET-ld' in `PATH'. 34144 34145 `collect2' explicitly avoids running `ld' using the file name under 34146 which `collect2' itself was invoked. In fact, it remembers up a list 34147 of such names--in case one copy of `collect2' finds another copy (or 34148 version) of `collect2' installed as `ld' in a second place in the 34149 search path. 34150 34151 `collect2' searches for the utilities `nm' and `strip' using the same 34152 algorithm as above for `ld'. 34153 34154 34155 File: gccint.info, Node: Header Dirs, Next: Type Information, Prev: Collect2, Up: Top 34156 34157 21 Standard Header File Directories 34158 *********************************** 34159 34160 `GCC_INCLUDE_DIR' means the same thing for native and cross. It is 34161 where GCC stores its private include files, and also where GCC stores 34162 the fixed include files. A cross compiled GCC runs `fixincludes' on 34163 the header files in `$(tooldir)/include'. (If the cross compilation 34164 header files need to be fixed, they must be installed before GCC is 34165 built. If the cross compilation header files are already suitable for 34166 GCC, nothing special need be done). 34167 34168 `GPLUSPLUS_INCLUDE_DIR' means the same thing for native and cross. It 34169 is where `g++' looks first for header files. The C++ library installs 34170 only target independent header files in that directory. 34171 34172 `LOCAL_INCLUDE_DIR' is used only by native compilers. GCC doesn't 34173 install anything there. It is normally `/usr/local/include'. This is 34174 where local additions to a packaged system should place header files. 34175 34176 `CROSS_INCLUDE_DIR' is used only by cross compilers. GCC doesn't 34177 install anything there. 34178 34179 `TOOL_INCLUDE_DIR' is used for both native and cross compilers. It is 34180 the place for other packages to install header files that GCC will use. 34181 For a cross-compiler, this is the equivalent of `/usr/include'. When 34182 you build a cross-compiler, `fixincludes' processes any header files in 34183 this directory. 34184 34185 34186 File: gccint.info, Node: Type Information, Next: Plugins, Prev: Header Dirs, Up: Top 34187 34188 22 Memory Management and Type Information 34189 ***************************************** 34190 34191 GCC uses some fairly sophisticated memory management techniques, which 34192 involve determining information about GCC's data structures from GCC's 34193 source code and using this information to perform garbage collection and 34194 implement precompiled headers. 34195 34196 A full C parser would be too complicated for this task, so a limited 34197 subset of C is interpreted and special markers are used to determine 34198 what parts of the source to look at. All `struct' and `union' 34199 declarations that define data structures that are allocated under 34200 control of the garbage collector must be marked. All global variables 34201 that hold pointers to garbage-collected memory must also be marked. 34202 Finally, all global variables that need to be saved and restored by a 34203 precompiled header must be marked. (The precompiled header mechanism 34204 can only save static variables if they're scalar. Complex data 34205 structures must be allocated in garbage-collected memory to be saved in 34206 a precompiled header.) 34207 34208 The full format of a marker is 34209 GTY (([OPTION] [(PARAM)], [OPTION] [(PARAM)] ...)) 34210 but in most cases no options are needed. The outer double parentheses 34211 are still necessary, though: `GTY(())'. Markers can appear: 34212 34213 * In a structure definition, before the open brace; 34214 34215 * In a global variable declaration, after the keyword `static' or 34216 `extern'; and 34217 34218 * In a structure field definition, before the name of the field. 34219 34220 Here are some examples of marking simple data structures and globals. 34221 34222 struct TAG GTY(()) 34223 { 34224 FIELDS... 34225 }; 34226 34227 typedef struct TAG GTY(()) 34228 { 34229 FIELDS... 34230 } *TYPENAME; 34231 34232 static GTY(()) struct TAG *LIST; /* points to GC memory */ 34233 static GTY(()) int COUNTER; /* save counter in a PCH */ 34234 34235 The parser understands simple typedefs such as `typedef struct TAG 34236 *NAME;' and `typedef int NAME;'. These don't need to be marked. 34237 34238 * Menu: 34239 34240 * GTY Options:: What goes inside a `GTY(())'. 34241 * GGC Roots:: Making global variables GGC roots. 34242 * Files:: How the generated files work. 34243 * Invoking the garbage collector:: How to invoke the garbage collector. 34244 34245 34246 File: gccint.info, Node: GTY Options, Next: GGC Roots, Up: Type Information 34247 34248 22.1 The Inside of a `GTY(())' 34249 ============================== 34250 34251 Sometimes the C code is not enough to fully describe the type 34252 structure. Extra information can be provided with `GTY' options and 34253 additional markers. Some options take a parameter, which may be either 34254 a string or a type name, depending on the parameter. If an option 34255 takes no parameter, it is acceptable either to omit the parameter 34256 entirely, or to provide an empty string as a parameter. For example, 34257 `GTY ((skip))' and `GTY ((skip ("")))' are equivalent. 34258 34259 When the parameter is a string, often it is a fragment of C code. Four 34260 special escapes may be used in these strings, to refer to pieces of the 34261 data structure being marked: 34262 34263 `%h' 34264 The current structure. 34265 34266 `%1' 34267 The structure that immediately contains the current structure. 34268 34269 `%0' 34270 The outermost structure that contains the current structure. 34271 34272 `%a' 34273 A partial expression of the form `[i1][i2]...' that indexes the 34274 array item currently being marked. 34275 34276 For instance, suppose that you have a structure of the form 34277 struct A { 34278 ... 34279 }; 34280 struct B { 34281 struct A foo[12]; 34282 }; 34283 and `b' is a variable of type `struct B'. When marking `b.foo[11]', 34284 `%h' would expand to `b.foo[11]', `%0' and `%1' would both expand to 34285 `b', and `%a' would expand to `[11]'. 34286 34287 As in ordinary C, adjacent strings will be concatenated; this is 34288 helpful when you have a complicated expression. 34289 GTY ((chain_next ("TREE_CODE (&%h.generic) == INTEGER_TYPE" 34290 " ? TYPE_NEXT_VARIANT (&%h.generic)" 34291 " : TREE_CHAIN (&%h.generic)"))) 34292 34293 The available options are: 34294 34295 `length ("EXPRESSION")' 34296 There are two places the type machinery will need to be explicitly 34297 told the length of an array. The first case is when a structure 34298 ends in a variable-length array, like this: 34299 struct rtvec_def GTY(()) { 34300 int num_elem; /* number of elements */ 34301 rtx GTY ((length ("%h.num_elem"))) elem[1]; 34302 }; 34303 34304 In this case, the `length' option is used to override the specified 34305 array length (which should usually be `1'). The parameter of the 34306 option is a fragment of C code that calculates the length. 34307 34308 The second case is when a structure or a global variable contains a 34309 pointer to an array, like this: 34310 tree * 34311 GTY ((length ("%h.regno_pointer_align_length"))) regno_decl; 34312 In this case, `regno_decl' has been allocated by writing something 34313 like 34314 x->regno_decl = 34315 ggc_alloc (x->regno_pointer_align_length * sizeof (tree)); 34316 and the `length' provides the length of the field. 34317 34318 This second use of `length' also works on global variables, like: 34319 static GTY((length ("reg_base_value_size"))) 34320 rtx *reg_base_value; 34321 34322 `skip' 34323 If `skip' is applied to a field, the type machinery will ignore it. 34324 This is somewhat dangerous; the only safe use is in a union when 34325 one field really isn't ever used. 34326 34327 `desc ("EXPRESSION")' 34328 `tag ("CONSTANT")' 34329 `default' 34330 The type machinery needs to be told which field of a `union' is 34331 currently active. This is done by giving each field a constant 34332 `tag' value, and then specifying a discriminator using `desc'. 34333 The value of the expression given by `desc' is compared against 34334 each `tag' value, each of which should be different. If no `tag' 34335 is matched, the field marked with `default' is used if there is 34336 one, otherwise no field in the union will be marked. 34337 34338 In the `desc' option, the "current structure" is the union that it 34339 discriminates. Use `%1' to mean the structure containing it. 34340 There are no escapes available to the `tag' option, since it is a 34341 constant. 34342 34343 For example, 34344 struct tree_binding GTY(()) 34345 { 34346 struct tree_common common; 34347 union tree_binding_u { 34348 tree GTY ((tag ("0"))) scope; 34349 struct cp_binding_level * GTY ((tag ("1"))) level; 34350 } GTY ((desc ("BINDING_HAS_LEVEL_P ((tree)&%0)"))) xscope; 34351 tree value; 34352 }; 34353 34354 In this example, the value of BINDING_HAS_LEVEL_P when applied to a 34355 `struct tree_binding *' is presumed to be 0 or 1. If 1, the type 34356 mechanism will treat the field `level' as being present and if 0, 34357 will treat the field `scope' as being present. 34358 34359 `param_is (TYPE)' 34360 `use_param' 34361 Sometimes it's convenient to define some data structure to work on 34362 generic pointers (that is, `PTR') and then use it with a specific 34363 type. `param_is' specifies the real type pointed to, and 34364 `use_param' says where in the generic data structure that type 34365 should be put. 34366 34367 For instance, to have a `htab_t' that points to trees, one would 34368 write the definition of `htab_t' like this: 34369 typedef struct GTY(()) { 34370 ... 34371 void ** GTY ((use_param, ...)) entries; 34372 ... 34373 } htab_t; 34374 and then declare variables like this: 34375 static htab_t GTY ((param_is (union tree_node))) ict; 34376 34377 `paramN_is (TYPE)' 34378 `use_paramN' 34379 In more complicated cases, the data structure might need to work on 34380 several different types, which might not necessarily all be 34381 pointers. For this, `param1_is' through `param9_is' may be used to 34382 specify the real type of a field identified by `use_param1' through 34383 `use_param9'. 34384 34385 `use_params' 34386 When a structure contains another structure that is parameterized, 34387 there's no need to do anything special, the inner structure 34388 inherits the parameters of the outer one. When a structure 34389 contains a pointer to a parameterized structure, the type 34390 machinery won't automatically detect this (it could, it just 34391 doesn't yet), so it's necessary to tell it that the pointed-to 34392 structure should use the same parameters as the outer structure. 34393 This is done by marking the pointer with the `use_params' option. 34394 34395 `deletable' 34396 `deletable', when applied to a global variable, indicates that when 34397 garbage collection runs, there's no need to mark anything pointed 34398 to by this variable, it can just be set to `NULL' instead. This 34399 is used to keep a list of free structures around for re-use. 34400 34401 `if_marked ("EXPRESSION")' 34402 Suppose you want some kinds of object to be unique, and so you put 34403 them in a hash table. If garbage collection marks the hash table, 34404 these objects will never be freed, even if the last other 34405 reference to them goes away. GGC has special handling to deal 34406 with this: if you use the `if_marked' option on a global hash 34407 table, GGC will call the routine whose name is the parameter to 34408 the option on each hash table entry. If the routine returns 34409 nonzero, the hash table entry will be marked as usual. If the 34410 routine returns zero, the hash table entry will be deleted. 34411 34412 The routine `ggc_marked_p' can be used to determine if an element 34413 has been marked already; in fact, the usual case is to use 34414 `if_marked ("ggc_marked_p")'. 34415 34416 `mark_hook ("HOOK-ROUTINE-NAME")' 34417 If provided for a structure or union type, the given 34418 HOOK-ROUTINE-NAME (between double-quotes) is the name of a routine 34419 called when the garbage collector has just marked the data as 34420 reachable. This routine should not change the data, or call any ggc 34421 routine. Its only argument is a pointer to the just marked (const) 34422 structure or union. 34423 34424 `maybe_undef' 34425 When applied to a field, `maybe_undef' indicates that it's OK if 34426 the structure that this fields points to is never defined, so long 34427 as this field is always `NULL'. This is used to avoid requiring 34428 backends to define certain optional structures. It doesn't work 34429 with language frontends. 34430 34431 `nested_ptr (TYPE, "TO EXPRESSION", "FROM EXPRESSION")' 34432 The type machinery expects all pointers to point to the start of an 34433 object. Sometimes for abstraction purposes it's convenient to have 34434 a pointer which points inside an object. So long as it's possible 34435 to convert the original object to and from the pointer, such 34436 pointers can still be used. TYPE is the type of the original 34437 object, the TO EXPRESSION returns the pointer given the original 34438 object, and the FROM EXPRESSION returns the original object given 34439 the pointer. The pointer will be available using the `%h' escape. 34440 34441 `chain_next ("EXPRESSION")' 34442 `chain_prev ("EXPRESSION")' 34443 `chain_circular ("EXPRESSION")' 34444 It's helpful for the type machinery to know if objects are often 34445 chained together in long lists; this lets it generate code that 34446 uses less stack space by iterating along the list instead of 34447 recursing down it. `chain_next' is an expression for the next 34448 item in the list, `chain_prev' is an expression for the previous 34449 item. For singly linked lists, use only `chain_next'; for doubly 34450 linked lists, use both. The machinery requires that taking the 34451 next item of the previous item gives the original item. 34452 `chain_circular' is similar to `chain_next', but can be used for 34453 circular single linked lists. 34454 34455 `reorder ("FUNCTION NAME")' 34456 Some data structures depend on the relative ordering of pointers. 34457 If the precompiled header machinery needs to change that ordering, 34458 it will call the function referenced by the `reorder' option, 34459 before changing the pointers in the object that's pointed to by 34460 the field the option applies to. The function must take four 34461 arguments, with the signature 34462 `void *, void *, gt_pointer_operator, void *'. The first 34463 parameter is a pointer to the structure that contains the object 34464 being updated, or the object itself if there is no containing 34465 structure. The second parameter is a cookie that should be 34466 ignored. The third parameter is a routine that, given a pointer, 34467 will update it to its correct new value. The fourth parameter is 34468 a cookie that must be passed to the second parameter. 34469 34470 PCH cannot handle data structures that depend on the absolute 34471 values of pointers. `reorder' functions can be expensive. When 34472 possible, it is better to depend on properties of the data, like 34473 an ID number or the hash of a string instead. 34474 34475 `special ("NAME")' 34476 The `special' option is used to mark types that have to be dealt 34477 with by special case machinery. The parameter is the name of the 34478 special case. See `gengtype.c' for further details. Avoid adding 34479 new special cases unless there is no other alternative. 34480 34481 34482 File: gccint.info, Node: GGC Roots, Next: Files, Prev: GTY Options, Up: Type Information 34483 34484 22.2 Marking Roots for the Garbage Collector 34485 ============================================ 34486 34487 In addition to keeping track of types, the type machinery also locates 34488 the global variables ("roots") that the garbage collector starts at. 34489 Roots must be declared using one of the following syntaxes: 34490 34491 * `extern GTY(([OPTIONS])) TYPE NAME;' 34492 34493 * `static GTY(([OPTIONS])) TYPE NAME;' 34494 The syntax 34495 * `GTY(([OPTIONS])) TYPE NAME;' 34496 is _not_ accepted. There should be an `extern' declaration of such a 34497 variable in a header somewhere--mark that, not the definition. Or, if 34498 the variable is only used in one file, make it `static'. 34499 34500 34501 File: gccint.info, Node: Files, Next: Invoking the garbage collector, Prev: GGC Roots, Up: Type Information 34502 34503 22.3 Source Files Containing Type Information 34504 ============================================= 34505 34506 Whenever you add `GTY' markers to a source file that previously had 34507 none, or create a new source file containing `GTY' markers, there are 34508 three things you need to do: 34509 34510 1. You need to add the file to the list of source files the type 34511 machinery scans. There are four cases: 34512 34513 a. For a back-end file, this is usually done automatically; if 34514 not, you should add it to `target_gtfiles' in the appropriate 34515 port's entries in `config.gcc'. 34516 34517 b. For files shared by all front ends, add the filename to the 34518 `GTFILES' variable in `Makefile.in'. 34519 34520 c. For files that are part of one front end, add the filename to 34521 the `gtfiles' variable defined in the appropriate 34522 `config-lang.in'. For C, the file is `c-config-lang.in'. 34523 Headers should appear before non-headers in this list. 34524 34525 d. For files that are part of some but not all front ends, add 34526 the filename to the `gtfiles' variable of _all_ the front ends 34527 that use it. 34528 34529 2. If the file was a header file, you'll need to check that it's 34530 included in the right place to be visible to the generated files. 34531 For a back-end header file, this should be done automatically. 34532 For a front-end header file, it needs to be included by the same 34533 file that includes `gtype-LANG.h'. For other header files, it 34534 needs to be included in `gtype-desc.c', which is a generated file, 34535 so add it to `ifiles' in `open_base_file' in `gengtype.c'. 34536 34537 For source files that aren't header files, the machinery will 34538 generate a header file that should be included in the source file 34539 you just changed. The file will be called `gt-PATH.h' where PATH 34540 is the pathname relative to the `gcc' directory with slashes 34541 replaced by -, so for example the header file to be included in 34542 `cp/parser.c' is called `gt-cp-parser.c'. The generated header 34543 file should be included after everything else in the source file. 34544 Don't forget to mention this file as a dependency in the 34545 `Makefile'! 34546 34547 34548 For language frontends, there is another file that needs to be included 34549 somewhere. It will be called `gtype-LANG.h', where LANG is the name of 34550 the subdirectory the language is contained in. 34551 34552 Plugins can add additional root tables. Run the `gengtype' utility in 34553 plugin mode as `gengtype -p SOURCE-DIR FILE-LIST PLUGIN*.C' with your 34554 plugin files PLUGIN*.C using `GTY' to generate the corresponding 34555 GT-PLUGIN*.H files. The GCC build tree is needed to be present in that 34556 mode. 34557 34558 34559 File: gccint.info, Node: Invoking the garbage collector, Prev: Files, Up: Type Information 34560 34561 22.4 How to invoke the garbage collector 34562 ======================================== 34563 34564 The GCC garbage collector GGC is only invoked explicitly. In contrast 34565 with many other garbage collectors, it is not implicitly invoked by 34566 allocation routines when a lot of memory has been consumed. So the only 34567 way to have GGC reclaim storage it to call the `ggc_collect' function 34568 explicitly. This call is an expensive operation, as it may have to scan 34569 the entire heap. Beware that local variables (on the GCC call stack) 34570 are not followed by such an invocation (as many other garbage 34571 collectors do): you should reference all your data from static or 34572 external `GTY'-ed variables, and it is advised to call `ggc_collect' 34573 with a shallow call stack. The GGC is an exact mark and sweep garbage 34574 collector (so it does not scan the call stack for pointers). In 34575 practice GCC passes don't often call `ggc_collect' themselves, because 34576 it is called by the pass manager between passes. 34577 34578 34579 File: gccint.info, Node: Plugins, Next: Funding, Prev: Type Information, Up: Top 34580 34581 23 Plugins 34582 ********** 34583 34584 23.1 Loading Plugins 34585 ==================== 34586 34587 Plugins are supported on platforms that support `-ldl -rdynamic'. They 34588 are loaded by the compiler using `dlopen' and invoked at pre-determined 34589 locations in the compilation process. 34590 34591 Plugins are loaded with 34592 34593 `-fplugin=/path/to/NAME.so' `-fplugin-arg-NAME-<key1>[=<value1>]' 34594 34595 The plugin arguments are parsed by GCC and passed to respective 34596 plugins as key-value pairs. Multiple plugins can be invoked by 34597 specifying multiple `-fplugin' arguments. 34598 34599 23.2 Plugin API 34600 =============== 34601 34602 Plugins are activated by the compiler at specific events as defined in 34603 `gcc-plugin.h'. For each event of interest, the plugin should call 34604 `register_callback' specifying the name of the event and address of the 34605 callback function that will handle that event. 34606 34607 The header `gcc-plugin.h' must be the first gcc header to be included. 34608 34609 23.2.1 Plugin initialization 34610 ---------------------------- 34611 34612 Every plugin should export a function called `plugin_init' that is 34613 called right after the plugin is loaded. This function is responsible 34614 for registering all the callbacks required by the plugin and do any 34615 other required initialization. 34616 34617 This function is called from `compile_file' right before invoking the 34618 parser. The arguments to `plugin_init' are: 34619 34620 * `plugin_info': Plugin invocation information. 34621 34622 * `version': GCC version. 34623 34624 The `plugin_info' struct is defined as follows: 34625 34626 struct plugin_name_args 34627 { 34628 char *base_name; /* Short name of the plugin 34629 (filename without .so suffix). */ 34630 const char *full_name; /* Path to the plugin as specified with 34631 -fplugin=. */ 34632 int argc; /* Number of arguments specified with 34633 -fplugin-arg-.... */ 34634 struct plugin_argument *argv; /* Array of ARGC key-value pairs. */ 34635 const char *version; /* Version string provided by plugin. */ 34636 const char *help; /* Help string provided by plugin. */ 34637 } 34638 34639 If initialization fails, `plugin_init' must return a non-zero value. 34640 Otherwise, it should return 0. 34641 34642 The version of the GCC compiler loading the plugin is described by the 34643 following structure: 34644 34645 struct plugin_gcc_version 34646 { 34647 const char *basever; 34648 const char *datestamp; 34649 const char *devphase; 34650 const char *revision; 34651 const char *configuration_arguments; 34652 }; 34653 34654 The function `plugin_default_version_check' takes two pointers to such 34655 structure and compare them field by field. It can be used by the 34656 plugin's `plugin_init' function. 34657 34658 23.2.2 Plugin callbacks 34659 ----------------------- 34660 34661 Callback functions have the following prototype: 34662 34663 /* The prototype for a plugin callback function. 34664 gcc_data - event-specific data provided by GCC 34665 user_data - plugin-specific data provided by the plug-in. */ 34666 typedef void (*plugin_callback_func)(void *gcc_data, void *user_data); 34667 34668 Callbacks can be invoked at the following pre-determined events: 34669 34670 enum plugin_event 34671 { 34672 PLUGIN_PASS_MANAGER_SETUP, /* To hook into pass manager. */ 34673 PLUGIN_FINISH_TYPE, /* After finishing parsing a type. */ 34674 PLUGIN_FINISH_UNIT, /* Useful for summary processing. */ 34675 PLUGIN_CXX_CP_PRE_GENERICIZE, /* Allows to see low level AST in C++ FE. */ 34676 PLUGIN_FINISH, /* Called before GCC exits. */ 34677 PLUGIN_INFO, /* Information about the plugin. */ 34678 PLUGIN_GGC_START, /* Called at start of GCC Garbage Collection. */ 34679 PLUGIN_GGC_MARKING, /* Extend the GGC marking. */ 34680 PLUGIN_GGC_END, /* Called at end of GGC. */ 34681 PLUGIN_REGISTER_GGC_ROOTS, /* Register an extra GGC root table. */ 34682 PLUGIN_ATTRIBUTES, /* Called during attribute registration */ 34683 PLUGIN_START_UNIT, /* Called before processing a translation unit. */ 34684 PLUGIN_EVENT_LAST /* Dummy event used for indexing callback 34685 array. */ 34686 }; 34687 34688 To register a callback, the plugin calls `register_callback' with the 34689 arguments: 34690 34691 * `char *name': Plugin name. 34692 34693 * `enum plugin_event event': The event code. 34694 34695 * `plugin_callback_func callback': The function that handles `event'. 34696 34697 * `void *user_data': Pointer to plugin-specific data. 34698 34699 For the PLUGIN_PASS_MANAGER_SETUP, PLUGIN_INFO, and 34700 PLUGIN_REGISTER_GGC_ROOTS pseudo-events the `callback' should be null, 34701 and the `user_data' is specific. 34702 34703 23.3 Interacting with the pass manager 34704 ====================================== 34705 34706 There needs to be a way to add/reorder/remove passes dynamically. This 34707 is useful for both analysis plugins (plugging in after a certain pass 34708 such as CFG or an IPA pass) and optimization plugins. 34709 34710 Basic support for inserting new passes or replacing existing passes is 34711 provided. A plugin registers a new pass with GCC by calling 34712 `register_callback' with the `PLUGIN_PASS_MANAGER_SETUP' event and a 34713 pointer to a `struct plugin_pass' object defined as follows 34714 34715 enum pass_positioning_ops 34716 { 34717 PASS_POS_INSERT_AFTER, // Insert after the reference pass. 34718 PASS_POS_INSERT_BEFORE, // Insert before the reference pass. 34719 PASS_POS_REPLACE // Replace the reference pass. 34720 }; 34721 34722 struct plugin_pass 34723 { 34724 struct opt_pass *pass; /* New pass provided by the plugin. */ 34725 const char *reference_pass_name; /* Name of the reference pass for hooking 34726 up the new pass. */ 34727 int ref_pass_instance_number; /* Insert the pass at the specified 34728 instance number of the reference pass. */ 34729 /* Do it for every instance if it is 0. */ 34730 enum pass_positioning_ops pos_op; /* how to insert the new pass. */ 34731 }; 34732 34733 34734 /* Sample plugin code that registers a new pass. */ 34735 int 34736 plugin_init (struct plugin_name_args *plugin_info, 34737 struct plugin_gcc_version *version) 34738 { 34739 struct plugin_pass pass_info; 34740 34741 ... 34742 34743 /* Code to fill in the pass_info object with new pass information. */ 34744 34745 ... 34746 34747 /* Register the new pass. */ 34748 register_callback (plugin_info->base_name, PLUGIN_PASS_MANAGER_SETUP, NULL, &pass_info); 34749 34750 ... 34751 } 34752 34753 23.4 Interacting with the GCC Garbage Collector 34754 =============================================== 34755 34756 Some plugins may want to be informed when GGC (the GCC Garbage 34757 Collector) is running. They can register callbacks for the 34758 `PLUGIN_GGC_START' and `PLUGIN_GGC_END' events (for which the callback 34759 is called with a null `gcc_data') to be notified of the start or end of 34760 the GCC garbage collection. 34761 34762 Some plugins may need to have GGC mark additional data. This can be 34763 done by registering a callback (called with a null `gcc_data') for the 34764 `PLUGIN_GGC_MARKING' event. Such callbacks can call the `ggc_set_mark' 34765 routine, preferably thru the `ggc_mark' macro (and conversely, these 34766 routines should usually not be used in plugins outside of the 34767 `PLUGIN_GGC_MARKING' event). 34768 34769 Some plugins may need to add extra GGC root tables, e.g. to handle 34770 their own `GTY'-ed data. This can be done with the 34771 `PLUGIN_REGISTER_GGC_ROOTS' pseudo-event with a null callback and the 34772 extra root table as `user_data'. Running the `gengtype -p SOURCE-DIR 34773 FILE-LIST PLUGIN*.C ...' utility generates this extra root table. 34774 34775 You should understand the details of memory management inside GCC 34776 before using `PLUGIN_GGC_MARKING' or `PLUGIN_REGISTER_GGC_ROOTS'. 34777 34778 23.5 Giving information about a plugin 34779 ====================================== 34780 34781 A plugin should give some information to the user about itself. This 34782 uses the following structure: 34783 34784 struct plugin_info 34785 { 34786 const char *version; 34787 const char *help; 34788 }; 34789 34790 Such a structure is passed as the `user_data' by the plugin's init 34791 routine using `register_callback' with the `PLUGIN_INFO' pseudo-event 34792 and a null callback. 34793 34794 23.6 Registering custom attributes 34795 ================================== 34796 34797 For analysis purposes it is useful to be able to add custom attributes. 34798 34799 The `PLUGIN_ATTRIBUTES' callback is called during attribute 34800 registration. Use the `register_attribute' function to register custom 34801 attributes. 34802 34803 /* Attribute handler callback */ 34804 static tree 34805 handle_user_attribute (tree *node, tree name, tree args, 34806 int flags, bool *no_add_attrs) 34807 { 34808 return NULL_TREE; 34809 } 34810 34811 /* Attribute definition */ 34812 static struct attribute_spec user_attr = 34813 { "user", 1, 1, false, false, false, handle_user_attribute }; 34814 34815 /* Plugin callback called during attribute registration. 34816 Registered with register_callback (plugin_name, PLUGIN_ATTRIBUTES, register_attributes, NULL) 34817 */ 34818 static void 34819 register_attributes (void *event_data, void *data) 34820 { 34821 warning (0, G_("Callback to register attributes")); 34822 register_attribute (&user_attr); 34823 } 34824 34825 23.7 Building GCC plugins 34826 ========================= 34827 34828 If plugins are enabled, GCC installs the headers needed to build a 34829 plugin (somehwere in the installation tree, e.g. under `/usr/local'). 34830 In particular a `plugin/include' directory is installed, containing all 34831 the header files needed to build plugins. 34832 34833 On most systems, you can query this `plugin' directory by invoking 34834 `gcc -print-file-name=plugin' (replace if needed `gcc' with the 34835 appropriate program path). 34836 34837 The following GNU Makefile excerpt shows how to build a simple plugin: 34838 34839 GCC=gcc 34840 PLUGIN_SOURCE_FILES= plugin1.c plugin2.c 34841 PLUGIN_OBJECT_FILES= $(patsubst %.c,%.o,$(PLUGIN_SOURCE_FILES)) 34842 GCCPLUGINS_DIR:= $(shell $(GCC) -print-file-name=plugin) 34843 CFLAGS+= -I$(GCCPLUGINS_DIR)/include -fPIC -O2 34844 34845 plugin.so: $(PLUGIN_OBJECT_FILES) 34846 $(GCC) -shared $^ -o $ 34847 34848 A single source file plugin may be built with `gcc -I`gcc 34849 -print-file-name=plugin`/include -fPIC -shared -O2 plugin.c -o 34850 plugin.so', using backquote shell syntax to query the `plugin' 34851 directory. 34852 34853 Plugins needing to use `gengtype' require a GCC build directory for 34854 the same version of GCC that they will be linked against. 34855 34856 34857 File: gccint.info, Node: Funding, Next: GNU Project, Prev: Plugins, Up: Top 34858 34859 Funding Free Software 34860 ********************* 34861 34862 If you want to have more free software a few years from now, it makes 34863 sense for you to help encourage people to contribute funds for its 34864 development. The most effective approach known is to encourage 34865 commercial redistributors to donate. 34866 34867 Users of free software systems can boost the pace of development by 34868 encouraging for-a-fee distributors to donate part of their selling price 34869 to free software developers--the Free Software Foundation, and others. 34870 34871 The way to convince distributors to do this is to demand it and expect 34872 it from them. So when you compare distributors, judge them partly by 34873 how much they give to free software development. Show distributors 34874 they must compete to be the one who gives the most. 34875 34876 To make this approach work, you must insist on numbers that you can 34877 compare, such as, "We will donate ten dollars to the Frobnitz project 34878 for each disk sold." Don't be satisfied with a vague promise, such as 34879 "A portion of the profits are donated," since it doesn't give a basis 34880 for comparison. 34881 34882 Even a precise fraction "of the profits from this disk" is not very 34883 meaningful, since creative accounting and unrelated business decisions 34884 can greatly alter what fraction of the sales price counts as profit. 34885 If the price you pay is $50, ten percent of the profit is probably less 34886 than a dollar; it might be a few cents, or nothing at all. 34887 34888 Some redistributors do development work themselves. This is useful 34889 too; but to keep everyone honest, you need to inquire how much they do, 34890 and what kind. Some kinds of development make much more long-term 34891 difference than others. For example, maintaining a separate version of 34892 a program contributes very little; maintaining the standard version of a 34893 program for the whole community contributes much. Easy new ports 34894 contribute little, since someone else would surely do them; difficult 34895 ports such as adding a new CPU to the GNU Compiler Collection 34896 contribute more; major new features or packages contribute the most. 34897 34898 By establishing the idea that supporting further development is "the 34899 proper thing to do" when distributing free software for a fee, we can 34900 assure a steady flow of resources into making more free software. 34901 34902 Copyright (C) 1994 Free Software Foundation, Inc. 34903 Verbatim copying and redistribution of this section is permitted 34904 without royalty; alteration is not permitted. 34905 34906 34907 File: gccint.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top 34908 34909 The GNU Project and GNU/Linux 34910 ***************************** 34911 34912 The GNU Project was launched in 1984 to develop a complete Unix-like 34913 operating system which is free software: the GNU system. (GNU is a 34914 recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".) 34915 Variants of the GNU operating system, which use the kernel Linux, are 34916 now widely used; though these systems are often referred to as "Linux", 34917 they are more accurately called GNU/Linux systems. 34918 34919 For more information, see: 34920 `http://www.gnu.org/' 34921 `http://www.gnu.org/gnu/linux-and-gnu.html' 34922 34923 34924 File: gccint.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top 34925 34926 GNU General Public License 34927 ************************** 34928 34929 Version 3, 29 June 2007 34930 34931 Copyright (C) 2007 Free Software Foundation, Inc. `http://fsf.org/' 34932 34933 Everyone is permitted to copy and distribute verbatim copies of this 34934 license document, but changing it is not allowed. 34935 34936 Preamble 34937 ======== 34938 34939 The GNU General Public License is a free, copyleft license for software 34940 and other kinds of works. 34941 34942 The licenses for most software and other practical works are designed 34943 to take away your freedom to share and change the works. By contrast, 34944 the GNU General Public License is intended to guarantee your freedom to 34945 share and change all versions of a program-to make sure it remains free 34946 software for all its users. We, the Free Software Foundation, use the 34947 GNU General Public License for most of our software; it applies also to 34948 any other work released this way by its authors. You can apply it to 34949 your programs, too. 34950 34951 When we speak of free software, we are referring to freedom, not 34952 price. Our General Public Licenses are designed to make sure that you 34953 have the freedom to distribute copies of free software (and charge for 34954 them if you wish), that you receive source code or can get it if you 34955 want it, that you can change the software or use pieces of it in new 34956 free programs, and that you know you can do these things. 34957 34958 To protect your rights, we need to prevent others from denying you 34959 these rights or asking you to surrender the rights. Therefore, you 34960 have certain responsibilities if you distribute copies of the software, 34961 or if you modify it: responsibilities to respect the freedom of others. 34962 34963 For example, if you distribute copies of such a program, whether 34964 gratis or for a fee, you must pass on to the recipients the same 34965 freedoms that you received. You must make sure that they, too, receive 34966 or can get the source code. And you must show them these terms so they 34967 know their rights. 34968 34969 Developers that use the GNU GPL protect your rights with two steps: 34970 (1) assert copyright on the software, and (2) offer you this License 34971 giving you legal permission to copy, distribute and/or modify it. 34972 34973 For the developers' and authors' protection, the GPL clearly explains 34974 that there is no warranty for this free software. For both users' and 34975 authors' sake, the GPL requires that modified versions be marked as 34976 changed, so that their problems will not be attributed erroneously to 34977 authors of previous versions. 34978 34979 Some devices are designed to deny users access to install or run 34980 modified versions of the software inside them, although the 34981 manufacturer can do so. This is fundamentally incompatible with the 34982 aim of protecting users' freedom to change the software. The 34983 systematic pattern of such abuse occurs in the area of products for 34984 individuals to use, which is precisely where it is most unacceptable. 34985 Therefore, we have designed this version of the GPL to prohibit the 34986 practice for those products. If such problems arise substantially in 34987 other domains, we stand ready to extend this provision to those domains 34988 in future versions of the GPL, as needed to protect the freedom of 34989 users. 34990 34991 Finally, every program is threatened constantly by software patents. 34992 States should not allow patents to restrict development and use of 34993 software on general-purpose computers, but in those that do, we wish to 34994 avoid the special danger that patents applied to a free program could 34995 make it effectively proprietary. To prevent this, the GPL assures that 34996 patents cannot be used to render the program non-free. 34997 34998 The precise terms and conditions for copying, distribution and 34999 modification follow. 35000 35001 TERMS AND CONDITIONS 35002 ==================== 35003 35004 0. Definitions. 35005 35006 "This License" refers to version 3 of the GNU General Public 35007 License. 35008 35009 "Copyright" also means copyright-like laws that apply to other 35010 kinds of works, such as semiconductor masks. 35011 35012 "The Program" refers to any copyrightable work licensed under this 35013 License. Each licensee is addressed as "you". "Licensees" and 35014 "recipients" may be individuals or organizations. 35015 35016 To "modify" a work means to copy from or adapt all or part of the 35017 work in a fashion requiring copyright permission, other than the 35018 making of an exact copy. The resulting work is called a "modified 35019 version" of the earlier work or a work "based on" the earlier work. 35020 35021 A "covered work" means either the unmodified Program or a work 35022 based on the Program. 35023 35024 To "propagate" a work means to do anything with it that, without 35025 permission, would make you directly or secondarily liable for 35026 infringement under applicable copyright law, except executing it 35027 on a computer or modifying a private copy. Propagation includes 35028 copying, distribution (with or without modification), making 35029 available to the public, and in some countries other activities as 35030 well. 35031 35032 To "convey" a work means any kind of propagation that enables other 35033 parties to make or receive copies. Mere interaction with a user 35034 through a computer network, with no transfer of a copy, is not 35035 conveying. 35036 35037 An interactive user interface displays "Appropriate Legal Notices" 35038 to the extent that it includes a convenient and prominently visible 35039 feature that (1) displays an appropriate copyright notice, and (2) 35040 tells the user that there is no warranty for the work (except to 35041 the extent that warranties are provided), that licensees may 35042 convey the work under this License, and how to view a copy of this 35043 License. If the interface presents a list of user commands or 35044 options, such as a menu, a prominent item in the list meets this 35045 criterion. 35046 35047 1. Source Code. 35048 35049 The "source code" for a work means the preferred form of the work 35050 for making modifications to it. "Object code" means any 35051 non-source form of a work. 35052 35053 A "Standard Interface" means an interface that either is an 35054 official standard defined by a recognized standards body, or, in 35055 the case of interfaces specified for a particular programming 35056 language, one that is widely used among developers working in that 35057 language. 35058 35059 The "System Libraries" of an executable work include anything, 35060 other than the work as a whole, that (a) is included in the normal 35061 form of packaging a Major Component, but which is not part of that 35062 Major Component, and (b) serves only to enable use of the work 35063 with that Major Component, or to implement a Standard Interface 35064 for which an implementation is available to the public in source 35065 code form. A "Major Component", in this context, means a major 35066 essential component (kernel, window system, and so on) of the 35067 specific operating system (if any) on which the executable work 35068 runs, or a compiler used to produce the work, or an object code 35069 interpreter used to run it. 35070 35071 The "Corresponding Source" for a work in object code form means all 35072 the source code needed to generate, install, and (for an executable 35073 work) run the object code and to modify the work, including 35074 scripts to control those activities. However, it does not include 35075 the work's System Libraries, or general-purpose tools or generally 35076 available free programs which are used unmodified in performing 35077 those activities but which are not part of the work. For example, 35078 Corresponding Source includes interface definition files 35079 associated with source files for the work, and the source code for 35080 shared libraries and dynamically linked subprograms that the work 35081 is specifically designed to require, such as by intimate data 35082 communication or control flow between those subprograms and other 35083 parts of the work. 35084 35085 The Corresponding Source need not include anything that users can 35086 regenerate automatically from other parts of the Corresponding 35087 Source. 35088 35089 The Corresponding Source for a work in source code form is that 35090 same work. 35091 35092 2. Basic Permissions. 35093 35094 All rights granted under this License are granted for the term of 35095 copyright on the Program, and are irrevocable provided the stated 35096 conditions are met. This License explicitly affirms your unlimited 35097 permission to run the unmodified Program. The output from running 35098 a covered work is covered by this License only if the output, 35099 given its content, constitutes a covered work. This License 35100 acknowledges your rights of fair use or other equivalent, as 35101 provided by copyright law. 35102 35103 You may make, run and propagate covered works that you do not 35104 convey, without conditions so long as your license otherwise 35105 remains in force. You may convey covered works to others for the 35106 sole purpose of having them make modifications exclusively for 35107 you, or provide you with facilities for running those works, 35108 provided that you comply with the terms of this License in 35109 conveying all material for which you do not control copyright. 35110 Those thus making or running the covered works for you must do so 35111 exclusively on your behalf, under your direction and control, on 35112 terms that prohibit them from making any copies of your 35113 copyrighted material outside their relationship with you. 35114 35115 Conveying under any other circumstances is permitted solely under 35116 the conditions stated below. Sublicensing is not allowed; section 35117 10 makes it unnecessary. 35118 35119 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 35120 35121 No covered work shall be deemed part of an effective technological 35122 measure under any applicable law fulfilling obligations under 35123 article 11 of the WIPO copyright treaty adopted on 20 December 35124 1996, or similar laws prohibiting or restricting circumvention of 35125 such measures. 35126 35127 When you convey a covered work, you waive any legal power to forbid 35128 circumvention of technological measures to the extent such 35129 circumvention is effected by exercising rights under this License 35130 with respect to the covered work, and you disclaim any intention 35131 to limit operation or modification of the work as a means of 35132 enforcing, against the work's users, your or third parties' legal 35133 rights to forbid circumvention of technological measures. 35134 35135 4. Conveying Verbatim Copies. 35136 35137 You may convey verbatim copies of the Program's source code as you 35138 receive it, in any medium, provided that you conspicuously and 35139 appropriately publish on each copy an appropriate copyright notice; 35140 keep intact all notices stating that this License and any 35141 non-permissive terms added in accord with section 7 apply to the 35142 code; keep intact all notices of the absence of any warranty; and 35143 give all recipients a copy of this License along with the Program. 35144 35145 You may charge any price or no price for each copy that you convey, 35146 and you may offer support or warranty protection for a fee. 35147 35148 5. Conveying Modified Source Versions. 35149 35150 You may convey a work based on the Program, or the modifications to 35151 produce it from the Program, in the form of source code under the 35152 terms of section 4, provided that you also meet all of these 35153 conditions: 35154 35155 a. The work must carry prominent notices stating that you 35156 modified it, and giving a relevant date. 35157 35158 b. The work must carry prominent notices stating that it is 35159 released under this License and any conditions added under 35160 section 7. This requirement modifies the requirement in 35161 section 4 to "keep intact all notices". 35162 35163 c. You must license the entire work, as a whole, under this 35164 License to anyone who comes into possession of a copy. This 35165 License will therefore apply, along with any applicable 35166 section 7 additional terms, to the whole of the work, and all 35167 its parts, regardless of how they are packaged. This License 35168 gives no permission to license the work in any other way, but 35169 it does not invalidate such permission if you have separately 35170 received it. 35171 35172 d. If the work has interactive user interfaces, each must display 35173 Appropriate Legal Notices; however, if the Program has 35174 interactive interfaces that do not display Appropriate Legal 35175 Notices, your work need not make them do so. 35176 35177 A compilation of a covered work with other separate and independent 35178 works, which are not by their nature extensions of the covered 35179 work, and which are not combined with it such as to form a larger 35180 program, in or on a volume of a storage or distribution medium, is 35181 called an "aggregate" if the compilation and its resulting 35182 copyright are not used to limit the access or legal rights of the 35183 compilation's users beyond what the individual works permit. 35184 Inclusion of a covered work in an aggregate does not cause this 35185 License to apply to the other parts of the aggregate. 35186 35187 6. Conveying Non-Source Forms. 35188 35189 You may convey a covered work in object code form under the terms 35190 of sections 4 and 5, provided that you also convey the 35191 machine-readable Corresponding Source under the terms of this 35192 License, in one of these ways: 35193 35194 a. Convey the object code in, or embodied in, a physical product 35195 (including a physical distribution medium), accompanied by the 35196 Corresponding Source fixed on a durable physical medium 35197 customarily used for software interchange. 35198 35199 b. Convey the object code in, or embodied in, a physical product 35200 (including a physical distribution medium), accompanied by a 35201 written offer, valid for at least three years and valid for 35202 as long as you offer spare parts or customer support for that 35203 product model, to give anyone who possesses the object code 35204 either (1) a copy of the Corresponding Source for all the 35205 software in the product that is covered by this License, on a 35206 durable physical medium customarily used for software 35207 interchange, for a price no more than your reasonable cost of 35208 physically performing this conveying of source, or (2) access 35209 to copy the Corresponding Source from a network server at no 35210 charge. 35211 35212 c. Convey individual copies of the object code with a copy of 35213 the written offer to provide the Corresponding Source. This 35214 alternative is allowed only occasionally and noncommercially, 35215 and only if you received the object code with such an offer, 35216 in accord with subsection 6b. 35217 35218 d. Convey the object code by offering access from a designated 35219 place (gratis or for a charge), and offer equivalent access 35220 to the Corresponding Source in the same way through the same 35221 place at no further charge. You need not require recipients 35222 to copy the Corresponding Source along with the object code. 35223 If the place to copy the object code is a network server, the 35224 Corresponding Source may be on a different server (operated 35225 by you or a third party) that supports equivalent copying 35226 facilities, provided you maintain clear directions next to 35227 the object code saying where to find the Corresponding Source. 35228 Regardless of what server hosts the Corresponding Source, you 35229 remain obligated to ensure that it is available for as long 35230 as needed to satisfy these requirements. 35231 35232 e. Convey the object code using peer-to-peer transmission, 35233 provided you inform other peers where the object code and 35234 Corresponding Source of the work are being offered to the 35235 general public at no charge under subsection 6d. 35236 35237 35238 A separable portion of the object code, whose source code is 35239 excluded from the Corresponding Source as a System Library, need 35240 not be included in conveying the object code work. 35241 35242 A "User Product" is either (1) a "consumer product", which means 35243 any tangible personal property which is normally used for personal, 35244 family, or household purposes, or (2) anything designed or sold for 35245 incorporation into a dwelling. In determining whether a product 35246 is a consumer product, doubtful cases shall be resolved in favor of 35247 coverage. For a particular product received by a particular user, 35248 "normally used" refers to a typical or common use of that class of 35249 product, regardless of the status of the particular user or of the 35250 way in which the particular user actually uses, or expects or is 35251 expected to use, the product. A product is a consumer product 35252 regardless of whether the product has substantial commercial, 35253 industrial or non-consumer uses, unless such uses represent the 35254 only significant mode of use of the product. 35255 35256 "Installation Information" for a User Product means any methods, 35257 procedures, authorization keys, or other information required to 35258 install and execute modified versions of a covered work in that 35259 User Product from a modified version of its Corresponding Source. 35260 The information must suffice to ensure that the continued 35261 functioning of the modified object code is in no case prevented or 35262 interfered with solely because modification has been made. 35263 35264 If you convey an object code work under this section in, or with, 35265 or specifically for use in, a User Product, and the conveying 35266 occurs as part of a transaction in which the right of possession 35267 and use of the User Product is transferred to the recipient in 35268 perpetuity or for a fixed term (regardless of how the transaction 35269 is characterized), the Corresponding Source conveyed under this 35270 section must be accompanied by the Installation Information. But 35271 this requirement does not apply if neither you nor any third party 35272 retains the ability to install modified object code on the User 35273 Product (for example, the work has been installed in ROM). 35274 35275 The requirement to provide Installation Information does not 35276 include a requirement to continue to provide support service, 35277 warranty, or updates for a work that has been modified or 35278 installed by the recipient, or for the User Product in which it 35279 has been modified or installed. Access to a network may be denied 35280 when the modification itself materially and adversely affects the 35281 operation of the network or violates the rules and protocols for 35282 communication across the network. 35283 35284 Corresponding Source conveyed, and Installation Information 35285 provided, in accord with this section must be in a format that is 35286 publicly documented (and with an implementation available to the 35287 public in source code form), and must require no special password 35288 or key for unpacking, reading or copying. 35289 35290 7. Additional Terms. 35291 35292 "Additional permissions" are terms that supplement the terms of 35293 this License by making exceptions from one or more of its 35294 conditions. Additional permissions that are applicable to the 35295 entire Program shall be treated as though they were included in 35296 this License, to the extent that they are valid under applicable 35297 law. If additional permissions apply only to part of the Program, 35298 that part may be used separately under those permissions, but the 35299 entire Program remains governed by this License without regard to 35300 the additional permissions. 35301 35302 When you convey a copy of a covered work, you may at your option 35303 remove any additional permissions from that copy, or from any part 35304 of it. (Additional permissions may be written to require their own 35305 removal in certain cases when you modify the work.) You may place 35306 additional permissions on material, added by you to a covered work, 35307 for which you have or can give appropriate copyright permission. 35308 35309 Notwithstanding any other provision of this License, for material 35310 you add to a covered work, you may (if authorized by the copyright 35311 holders of that material) supplement the terms of this License 35312 with terms: 35313 35314 a. Disclaiming warranty or limiting liability differently from 35315 the terms of sections 15 and 16 of this License; or 35316 35317 b. Requiring preservation of specified reasonable legal notices 35318 or author attributions in that material or in the Appropriate 35319 Legal Notices displayed by works containing it; or 35320 35321 c. Prohibiting misrepresentation of the origin of that material, 35322 or requiring that modified versions of such material be 35323 marked in reasonable ways as different from the original 35324 version; or 35325 35326 d. Limiting the use for publicity purposes of names of licensors 35327 or authors of the material; or 35328 35329 e. Declining to grant rights under trademark law for use of some 35330 trade names, trademarks, or service marks; or 35331 35332 f. Requiring indemnification of licensors and authors of that 35333 material by anyone who conveys the material (or modified 35334 versions of it) with contractual assumptions of liability to 35335 the recipient, for any liability that these contractual 35336 assumptions directly impose on those licensors and authors. 35337 35338 All other non-permissive additional terms are considered "further 35339 restrictions" within the meaning of section 10. If the Program as 35340 you received it, or any part of it, contains a notice stating that 35341 it is governed by this License along with a term that is a further 35342 restriction, you may remove that term. If a license document 35343 contains a further restriction but permits relicensing or 35344 conveying under this License, you may add to a covered work 35345 material governed by the terms of that license document, provided 35346 that the further restriction does not survive such relicensing or 35347 conveying. 35348 35349 If you add terms to a covered work in accord with this section, you 35350 must place, in the relevant source files, a statement of the 35351 additional terms that apply to those files, or a notice indicating 35352 where to find the applicable terms. 35353 35354 Additional terms, permissive or non-permissive, may be stated in 35355 the form of a separately written license, or stated as exceptions; 35356 the above requirements apply either way. 35357 35358 8. Termination. 35359 35360 You may not propagate or modify a covered work except as expressly 35361 provided under this License. Any attempt otherwise to propagate or 35362 modify it is void, and will automatically terminate your rights 35363 under this License (including any patent licenses granted under 35364 the third paragraph of section 11). 35365 35366 However, if you cease all violation of this License, then your 35367 license from a particular copyright holder is reinstated (a) 35368 provisionally, unless and until the copyright holder explicitly 35369 and finally terminates your license, and (b) permanently, if the 35370 copyright holder fails to notify you of the violation by some 35371 reasonable means prior to 60 days after the cessation. 35372 35373 Moreover, your license from a particular copyright holder is 35374 reinstated permanently if the copyright holder notifies you of the 35375 violation by some reasonable means, this is the first time you have 35376 received notice of violation of this License (for any work) from 35377 that copyright holder, and you cure the violation prior to 30 days 35378 after your receipt of the notice. 35379 35380 Termination of your rights under this section does not terminate 35381 the licenses of parties who have received copies or rights from 35382 you under this License. If your rights have been terminated and 35383 not permanently reinstated, you do not qualify to receive new 35384 licenses for the same material under section 10. 35385 35386 9. Acceptance Not Required for Having Copies. 35387 35388 You are not required to accept this License in order to receive or 35389 run a copy of the Program. Ancillary propagation of a covered work 35390 occurring solely as a consequence of using peer-to-peer 35391 transmission to receive a copy likewise does not require 35392 acceptance. However, nothing other than this License grants you 35393 permission to propagate or modify any covered work. These actions 35394 infringe copyright if you do not accept this License. Therefore, 35395 by modifying or propagating a covered work, you indicate your 35396 acceptance of this License to do so. 35397 35398 10. Automatic Licensing of Downstream Recipients. 35399 35400 Each time you convey a covered work, the recipient automatically 35401 receives a license from the original licensors, to run, modify and 35402 propagate that work, subject to this License. You are not 35403 responsible for enforcing compliance by third parties with this 35404 License. 35405 35406 An "entity transaction" is a transaction transferring control of an 35407 organization, or substantially all assets of one, or subdividing an 35408 organization, or merging organizations. If propagation of a 35409 covered work results from an entity transaction, each party to that 35410 transaction who receives a copy of the work also receives whatever 35411 licenses to the work the party's predecessor in interest had or 35412 could give under the previous paragraph, plus a right to 35413 possession of the Corresponding Source of the work from the 35414 predecessor in interest, if the predecessor has it or can get it 35415 with reasonable efforts. 35416 35417 You may not impose any further restrictions on the exercise of the 35418 rights granted or affirmed under this License. For example, you 35419 may not impose a license fee, royalty, or other charge for 35420 exercise of rights granted under this License, and you may not 35421 initiate litigation (including a cross-claim or counterclaim in a 35422 lawsuit) alleging that any patent claim is infringed by making, 35423 using, selling, offering for sale, or importing the Program or any 35424 portion of it. 35425 35426 11. Patents. 35427 35428 A "contributor" is a copyright holder who authorizes use under this 35429 License of the Program or a work on which the Program is based. 35430 The work thus licensed is called the contributor's "contributor 35431 version". 35432 35433 A contributor's "essential patent claims" are all patent claims 35434 owned or controlled by the contributor, whether already acquired or 35435 hereafter acquired, that would be infringed by some manner, 35436 permitted by this License, of making, using, or selling its 35437 contributor version, but do not include claims that would be 35438 infringed only as a consequence of further modification of the 35439 contributor version. For purposes of this definition, "control" 35440 includes the right to grant patent sublicenses in a manner 35441 consistent with the requirements of this License. 35442 35443 Each contributor grants you a non-exclusive, worldwide, 35444 royalty-free patent license under the contributor's essential 35445 patent claims, to make, use, sell, offer for sale, import and 35446 otherwise run, modify and propagate the contents of its 35447 contributor version. 35448 35449 In the following three paragraphs, a "patent license" is any 35450 express agreement or commitment, however denominated, not to 35451 enforce a patent (such as an express permission to practice a 35452 patent or covenant not to sue for patent infringement). To 35453 "grant" such a patent license to a party means to make such an 35454 agreement or commitment not to enforce a patent against the party. 35455 35456 If you convey a covered work, knowingly relying on a patent 35457 license, and the Corresponding Source of the work is not available 35458 for anyone to copy, free of charge and under the terms of this 35459 License, through a publicly available network server or other 35460 readily accessible means, then you must either (1) cause the 35461 Corresponding Source to be so available, or (2) arrange to deprive 35462 yourself of the benefit of the patent license for this particular 35463 work, or (3) arrange, in a manner consistent with the requirements 35464 of this License, to extend the patent license to downstream 35465 recipients. "Knowingly relying" means you have actual knowledge 35466 that, but for the patent license, your conveying the covered work 35467 in a country, or your recipient's use of the covered work in a 35468 country, would infringe one or more identifiable patents in that 35469 country that you have reason to believe are valid. 35470 35471 If, pursuant to or in connection with a single transaction or 35472 arrangement, you convey, or propagate by procuring conveyance of, a 35473 covered work, and grant a patent license to some of the parties 35474 receiving the covered work authorizing them to use, propagate, 35475 modify or convey a specific copy of the covered work, then the 35476 patent license you grant is automatically extended to all 35477 recipients of the covered work and works based on it. 35478 35479 A patent license is "discriminatory" if it does not include within 35480 the scope of its coverage, prohibits the exercise of, or is 35481 conditioned on the non-exercise of one or more of the rights that 35482 are specifically granted under this License. You may not convey a 35483 covered work if you are a party to an arrangement with a third 35484 party that is in the business of distributing software, under 35485 which you make payment to the third party based on the extent of 35486 your activity of conveying the work, and under which the third 35487 party grants, to any of the parties who would receive the covered 35488 work from you, a discriminatory patent license (a) in connection 35489 with copies of the covered work conveyed by you (or copies made 35490 from those copies), or (b) primarily for and in connection with 35491 specific products or compilations that contain the covered work, 35492 unless you entered into that arrangement, or that patent license 35493 was granted, prior to 28 March 2007. 35494 35495 Nothing in this License shall be construed as excluding or limiting 35496 any implied license or other defenses to infringement that may 35497 otherwise be available to you under applicable patent law. 35498 35499 12. No Surrender of Others' Freedom. 35500 35501 If conditions are imposed on you (whether by court order, 35502 agreement or otherwise) that contradict the conditions of this 35503 License, they do not excuse you from the conditions of this 35504 License. If you cannot convey a covered work so as to satisfy 35505 simultaneously your obligations under this License and any other 35506 pertinent obligations, then as a consequence you may not convey it 35507 at all. For example, if you agree to terms that obligate you to 35508 collect a royalty for further conveying from those to whom you 35509 convey the Program, the only way you could satisfy both those 35510 terms and this License would be to refrain entirely from conveying 35511 the Program. 35512 35513 13. Use with the GNU Affero General Public License. 35514 35515 Notwithstanding any other provision of this License, you have 35516 permission to link or combine any covered work with a work licensed 35517 under version 3 of the GNU Affero General Public License into a 35518 single combined work, and to convey the resulting work. The terms 35519 of this License will continue to apply to the part which is the 35520 covered work, but the special requirements of the GNU Affero 35521 General Public License, section 13, concerning interaction through 35522 a network will apply to the combination as such. 35523 35524 14. Revised Versions of this License. 35525 35526 The Free Software Foundation may publish revised and/or new 35527 versions of the GNU General Public License from time to time. 35528 Such new versions will be similar in spirit to the present 35529 version, but may differ in detail to address new problems or 35530 concerns. 35531 35532 Each version is given a distinguishing version number. If the 35533 Program specifies that a certain numbered version of the GNU 35534 General Public License "or any later version" applies to it, you 35535 have the option of following the terms and conditions either of 35536 that numbered version or of any later version published by the 35537 Free Software Foundation. If the Program does not specify a 35538 version number of the GNU General Public License, you may choose 35539 any version ever published by the Free Software Foundation. 35540 35541 If the Program specifies that a proxy can decide which future 35542 versions of the GNU General Public License can be used, that 35543 proxy's public statement of acceptance of a version permanently 35544 authorizes you to choose that version for the Program. 35545 35546 Later license versions may give you additional or different 35547 permissions. However, no additional obligations are imposed on any 35548 author or copyright holder as a result of your choosing to follow a 35549 later version. 35550 35551 15. Disclaimer of Warranty. 35552 35553 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 35554 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE 35555 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" 35556 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, 35557 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 35558 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE 35559 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. 35560 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL 35561 NECESSARY SERVICING, REPAIR OR CORRECTION. 35562 35563 16. Limitation of Liability. 35564 35565 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN 35566 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES 35567 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU 35568 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR 35569 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE 35570 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA 35571 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 35572 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 35573 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF 35574 THE POSSIBILITY OF SUCH DAMAGES. 35575 35576 17. Interpretation of Sections 15 and 16. 35577 35578 If the disclaimer of warranty and limitation of liability provided 35579 above cannot be given local legal effect according to their terms, 35580 reviewing courts shall apply local law that most closely 35581 approximates an absolute waiver of all civil liability in 35582 connection with the Program, unless a warranty or assumption of 35583 liability accompanies a copy of the Program in return for a fee. 35584 35585 35586 END OF TERMS AND CONDITIONS 35587 =========================== 35588 35589 How to Apply These Terms to Your New Programs 35590 ============================================= 35591 35592 If you develop a new program, and you want it to be of the greatest 35593 possible use to the public, the best way to achieve this is to make it 35594 free software which everyone can redistribute and change under these 35595 terms. 35596 35597 To do so, attach the following notices to the program. It is safest 35598 to attach them to the start of each source file to most effectively 35599 state the exclusion of warranty; and each file should have at least the 35600 "copyright" line and a pointer to where the full notice is found. 35601 35602 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 35603 Copyright (C) YEAR NAME OF AUTHOR 35604 35605 This program is free software: you can redistribute it and/or modify 35606 it under the terms of the GNU General Public License as published by 35607 the Free Software Foundation, either version 3 of the License, or (at 35608 your option) any later version. 35609 35610 This program is distributed in the hope that it will be useful, but 35611 WITHOUT ANY WARRANTY; without even the implied warranty of 35612 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 35613 General Public License for more details. 35614 35615 You should have received a copy of the GNU General Public License 35616 along with this program. If not, see `http://www.gnu.org/licenses/'. 35617 35618 Also add information on how to contact you by electronic and paper 35619 mail. 35620 35621 If the program does terminal interaction, make it output a short 35622 notice like this when it starts in an interactive mode: 35623 35624 PROGRAM Copyright (C) YEAR NAME OF AUTHOR 35625 This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 35626 This is free software, and you are welcome to redistribute it 35627 under certain conditions; type `show c' for details. 35628 35629 The hypothetical commands `show w' and `show c' should show the 35630 appropriate parts of the General Public License. Of course, your 35631 program's commands might be different; for a GUI interface, you would 35632 use an "about box". 35633 35634 You should also get your employer (if you work as a programmer) or 35635 school, if any, to sign a "copyright disclaimer" for the program, if 35636 necessary. For more information on this, and how to apply and follow 35637 the GNU GPL, see `http://www.gnu.org/licenses/'. 35638 35639 The GNU General Public License does not permit incorporating your 35640 program into proprietary programs. If your program is a subroutine 35641 library, you may consider it more useful to permit linking proprietary 35642 applications with the library. If this is what you want to do, use the 35643 GNU Lesser General Public License instead of this License. But first, 35644 please read `http://www.gnu.org/philosophy/why-not-lgpl.html'. 35645 35646 35647 File: gccint.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top 35648 35649 GNU Free Documentation License 35650 ****************************** 35651 35652 Version 1.2, November 2002 35653 35654 Copyright (C) 2000,2001,2002 Free Software Foundation, Inc. 35655 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA 35656 35657 Everyone is permitted to copy and distribute verbatim copies 35658 of this license document, but changing it is not allowed. 35659 35660 0. PREAMBLE 35661 35662 The purpose of this License is to make a manual, textbook, or other 35663 functional and useful document "free" in the sense of freedom: to 35664 assure everyone the effective freedom to copy and redistribute it, 35665 with or without modifying it, either commercially or 35666 noncommercially. Secondarily, this License preserves for the 35667 author and publisher a way to get credit for their work, while not 35668 being considered responsible for modifications made by others. 35669 35670 This License is a kind of "copyleft", which means that derivative 35671 works of the document must themselves be free in the same sense. 35672 It complements the GNU General Public License, which is a copyleft 35673 license designed for free software. 35674 35675 We have designed this License in order to use it for manuals for 35676 free software, because free software needs free documentation: a 35677 free program should come with manuals providing the same freedoms 35678 that the software does. But this License is not limited to 35679 software manuals; it can be used for any textual work, regardless 35680 of subject matter or whether it is published as a printed book. 35681 We recommend this License principally for works whose purpose is 35682 instruction or reference. 35683 35684 1. APPLICABILITY AND DEFINITIONS 35685 35686 This License applies to any manual or other work, in any medium, 35687 that contains a notice placed by the copyright holder saying it 35688 can be distributed under the terms of this License. Such a notice 35689 grants a world-wide, royalty-free license, unlimited in duration, 35690 to use that work under the conditions stated herein. The 35691 "Document", below, refers to any such manual or work. Any member 35692 of the public is a licensee, and is addressed as "you". You 35693 accept the license if you copy, modify or distribute the work in a 35694 way requiring permission under copyright law. 35695 35696 A "Modified Version" of the Document means any work containing the 35697 Document or a portion of it, either copied verbatim, or with 35698 modifications and/or translated into another language. 35699 35700 A "Secondary Section" is a named appendix or a front-matter section 35701 of the Document that deals exclusively with the relationship of the 35702 publishers or authors of the Document to the Document's overall 35703 subject (or to related matters) and contains nothing that could 35704 fall directly within that overall subject. (Thus, if the Document 35705 is in part a textbook of mathematics, a Secondary Section may not 35706 explain any mathematics.) The relationship could be a matter of 35707 historical connection with the subject or with related matters, or 35708 of legal, commercial, philosophical, ethical or political position 35709 regarding them. 35710 35711 The "Invariant Sections" are certain Secondary Sections whose 35712 titles are designated, as being those of Invariant Sections, in 35713 the notice that says that the Document is released under this 35714 License. If a section does not fit the above definition of 35715 Secondary then it is not allowed to be designated as Invariant. 35716 The Document may contain zero Invariant Sections. If the Document 35717 does not identify any Invariant Sections then there are none. 35718 35719 The "Cover Texts" are certain short passages of text that are 35720 listed, as Front-Cover Texts or Back-Cover Texts, in the notice 35721 that says that the Document is released under this License. A 35722 Front-Cover Text may be at most 5 words, and a Back-Cover Text may 35723 be at most 25 words. 35724 35725 A "Transparent" copy of the Document means a machine-readable copy, 35726 represented in a format whose specification is available to the 35727 general public, that is suitable for revising the document 35728 straightforwardly with generic text editors or (for images 35729 composed of pixels) generic paint programs or (for drawings) some 35730 widely available drawing editor, and that is suitable for input to 35731 text formatters or for automatic translation to a variety of 35732 formats suitable for input to text formatters. A copy made in an 35733 otherwise Transparent file format whose markup, or absence of 35734 markup, has been arranged to thwart or discourage subsequent 35735 modification by readers is not Transparent. An image format is 35736 not Transparent if used for any substantial amount of text. A 35737 copy that is not "Transparent" is called "Opaque". 35738 35739 Examples of suitable formats for Transparent copies include plain 35740 ASCII without markup, Texinfo input format, LaTeX input format, 35741 SGML or XML using a publicly available DTD, and 35742 standard-conforming simple HTML, PostScript or PDF designed for 35743 human modification. Examples of transparent image formats include 35744 PNG, XCF and JPG. Opaque formats include proprietary formats that 35745 can be read and edited only by proprietary word processors, SGML or 35746 XML for which the DTD and/or processing tools are not generally 35747 available, and the machine-generated HTML, PostScript or PDF 35748 produced by some word processors for output purposes only. 35749 35750 The "Title Page" means, for a printed book, the title page itself, 35751 plus such following pages as are needed to hold, legibly, the 35752 material this License requires to appear in the title page. For 35753 works in formats which do not have any title page as such, "Title 35754 Page" means the text near the most prominent appearance of the 35755 work's title, preceding the beginning of the body of the text. 35756 35757 A section "Entitled XYZ" means a named subunit of the Document 35758 whose title either is precisely XYZ or contains XYZ in parentheses 35759 following text that translates XYZ in another language. (Here XYZ 35760 stands for a specific section name mentioned below, such as 35761 "Acknowledgements", "Dedications", "Endorsements", or "History".) 35762 To "Preserve the Title" of such a section when you modify the 35763 Document means that it remains a section "Entitled XYZ" according 35764 to this definition. 35765 35766 The Document may include Warranty Disclaimers next to the notice 35767 which states that this License applies to the Document. These 35768 Warranty Disclaimers are considered to be included by reference in 35769 this License, but only as regards disclaiming warranties: any other 35770 implication that these Warranty Disclaimers may have is void and 35771 has no effect on the meaning of this License. 35772 35773 2. VERBATIM COPYING 35774 35775 You may copy and distribute the Document in any medium, either 35776 commercially or noncommercially, provided that this License, the 35777 copyright notices, and the license notice saying this License 35778 applies to the Document are reproduced in all copies, and that you 35779 add no other conditions whatsoever to those of this License. You 35780 may not use technical measures to obstruct or control the reading 35781 or further copying of the copies you make or distribute. However, 35782 you may accept compensation in exchange for copies. If you 35783 distribute a large enough number of copies you must also follow 35784 the conditions in section 3. 35785 35786 You may also lend copies, under the same conditions stated above, 35787 and you may publicly display copies. 35788 35789 3. COPYING IN QUANTITY 35790 35791 If you publish printed copies (or copies in media that commonly 35792 have printed covers) of the Document, numbering more than 100, and 35793 the Document's license notice requires Cover Texts, you must 35794 enclose the copies in covers that carry, clearly and legibly, all 35795 these Cover Texts: Front-Cover Texts on the front cover, and 35796 Back-Cover Texts on the back cover. Both covers must also clearly 35797 and legibly identify you as the publisher of these copies. The 35798 front cover must present the full title with all words of the 35799 title equally prominent and visible. You may add other material 35800 on the covers in addition. Copying with changes limited to the 35801 covers, as long as they preserve the title of the Document and 35802 satisfy these conditions, can be treated as verbatim copying in 35803 other respects. 35804 35805 If the required texts for either cover are too voluminous to fit 35806 legibly, you should put the first ones listed (as many as fit 35807 reasonably) on the actual cover, and continue the rest onto 35808 adjacent pages. 35809 35810 If you publish or distribute Opaque copies of the Document 35811 numbering more than 100, you must either include a 35812 machine-readable Transparent copy along with each Opaque copy, or 35813 state in or with each Opaque copy a computer-network location from 35814 which the general network-using public has access to download 35815 using public-standard network protocols a complete Transparent 35816 copy of the Document, free of added material. If you use the 35817 latter option, you must take reasonably prudent steps, when you 35818 begin distribution of Opaque copies in quantity, to ensure that 35819 this Transparent copy will remain thus accessible at the stated 35820 location until at least one year after the last time you 35821 distribute an Opaque copy (directly or through your agents or 35822 retailers) of that edition to the public. 35823 35824 It is requested, but not required, that you contact the authors of 35825 the Document well before redistributing any large number of 35826 copies, to give them a chance to provide you with an updated 35827 version of the Document. 35828 35829 4. MODIFICATIONS 35830 35831 You may copy and distribute a Modified Version of the Document 35832 under the conditions of sections 2 and 3 above, provided that you 35833 release the Modified Version under precisely this License, with 35834 the Modified Version filling the role of the Document, thus 35835 licensing distribution and modification of the Modified Version to 35836 whoever possesses a copy of it. In addition, you must do these 35837 things in the Modified Version: 35838 35839 A. Use in the Title Page (and on the covers, if any) a title 35840 distinct from that of the Document, and from those of 35841 previous versions (which should, if there were any, be listed 35842 in the History section of the Document). You may use the 35843 same title as a previous version if the original publisher of 35844 that version gives permission. 35845 35846 B. List on the Title Page, as authors, one or more persons or 35847 entities responsible for authorship of the modifications in 35848 the Modified Version, together with at least five of the 35849 principal authors of the Document (all of its principal 35850 authors, if it has fewer than five), unless they release you 35851 from this requirement. 35852 35853 C. State on the Title page the name of the publisher of the 35854 Modified Version, as the publisher. 35855 35856 D. Preserve all the copyright notices of the Document. 35857 35858 E. Add an appropriate copyright notice for your modifications 35859 adjacent to the other copyright notices. 35860 35861 F. Include, immediately after the copyright notices, a license 35862 notice giving the public permission to use the Modified 35863 Version under the terms of this License, in the form shown in 35864 the Addendum below. 35865 35866 G. Preserve in that license notice the full lists of Invariant 35867 Sections and required Cover Texts given in the Document's 35868 license notice. 35869 35870 H. Include an unaltered copy of this License. 35871 35872 I. Preserve the section Entitled "History", Preserve its Title, 35873 and add to it an item stating at least the title, year, new 35874 authors, and publisher of the Modified Version as given on 35875 the Title Page. If there is no section Entitled "History" in 35876 the Document, create one stating the title, year, authors, 35877 and publisher of the Document as given on its Title Page, 35878 then add an item describing the Modified Version as stated in 35879 the previous sentence. 35880 35881 J. Preserve the network location, if any, given in the Document 35882 for public access to a Transparent copy of the Document, and 35883 likewise the network locations given in the Document for 35884 previous versions it was based on. These may be placed in 35885 the "History" section. You may omit a network location for a 35886 work that was published at least four years before the 35887 Document itself, or if the original publisher of the version 35888 it refers to gives permission. 35889 35890 K. For any section Entitled "Acknowledgements" or "Dedications", 35891 Preserve the Title of the section, and preserve in the 35892 section all the substance and tone of each of the contributor 35893 acknowledgements and/or dedications given therein. 35894 35895 L. Preserve all the Invariant Sections of the Document, 35896 unaltered in their text and in their titles. Section numbers 35897 or the equivalent are not considered part of the section 35898 titles. 35899 35900 M. Delete any section Entitled "Endorsements". Such a section 35901 may not be included in the Modified Version. 35902 35903 N. Do not retitle any existing section to be Entitled 35904 "Endorsements" or to conflict in title with any Invariant 35905 Section. 35906 35907 O. Preserve any Warranty Disclaimers. 35908 35909 If the Modified Version includes new front-matter sections or 35910 appendices that qualify as Secondary Sections and contain no 35911 material copied from the Document, you may at your option 35912 designate some or all of these sections as invariant. To do this, 35913 add their titles to the list of Invariant Sections in the Modified 35914 Version's license notice. These titles must be distinct from any 35915 other section titles. 35916 35917 You may add a section Entitled "Endorsements", provided it contains 35918 nothing but endorsements of your Modified Version by various 35919 parties--for example, statements of peer review or that the text 35920 has been approved by an organization as the authoritative 35921 definition of a standard. 35922 35923 You may add a passage of up to five words as a Front-Cover Text, 35924 and a passage of up to 25 words as a Back-Cover Text, to the end 35925 of the list of Cover Texts in the Modified Version. Only one 35926 passage of Front-Cover Text and one of Back-Cover Text may be 35927 added by (or through arrangements made by) any one entity. If the 35928 Document already includes a cover text for the same cover, 35929 previously added by you or by arrangement made by the same entity 35930 you are acting on behalf of, you may not add another; but you may 35931 replace the old one, on explicit permission from the previous 35932 publisher that added the old one. 35933 35934 The author(s) and publisher(s) of the Document do not by this 35935 License give permission to use their names for publicity for or to 35936 assert or imply endorsement of any Modified Version. 35937 35938 5. COMBINING DOCUMENTS 35939 35940 You may combine the Document with other documents released under 35941 this License, under the terms defined in section 4 above for 35942 modified versions, provided that you include in the combination 35943 all of the Invariant Sections of all of the original documents, 35944 unmodified, and list them all as Invariant Sections of your 35945 combined work in its license notice, and that you preserve all 35946 their Warranty Disclaimers. 35947 35948 The combined work need only contain one copy of this License, and 35949 multiple identical Invariant Sections may be replaced with a single 35950 copy. If there are multiple Invariant Sections with the same name 35951 but different contents, make the title of each such section unique 35952 by adding at the end of it, in parentheses, the name of the 35953 original author or publisher of that section if known, or else a 35954 unique number. Make the same adjustment to the section titles in 35955 the list of Invariant Sections in the license notice of the 35956 combined work. 35957 35958 In the combination, you must combine any sections Entitled 35959 "History" in the various original documents, forming one section 35960 Entitled "History"; likewise combine any sections Entitled 35961 "Acknowledgements", and any sections Entitled "Dedications". You 35962 must delete all sections Entitled "Endorsements." 35963 35964 6. COLLECTIONS OF DOCUMENTS 35965 35966 You may make a collection consisting of the Document and other 35967 documents released under this License, and replace the individual 35968 copies of this License in the various documents with a single copy 35969 that is included in the collection, provided that you follow the 35970 rules of this License for verbatim copying of each of the 35971 documents in all other respects. 35972 35973 You may extract a single document from such a collection, and 35974 distribute it individually under this License, provided you insert 35975 a copy of this License into the extracted document, and follow 35976 this License in all other respects regarding verbatim copying of 35977 that document. 35978 35979 7. AGGREGATION WITH INDEPENDENT WORKS 35980 35981 A compilation of the Document or its derivatives with other 35982 separate and independent documents or works, in or on a volume of 35983 a storage or distribution medium, is called an "aggregate" if the 35984 copyright resulting from the compilation is not used to limit the 35985 legal rights of the compilation's users beyond what the individual 35986 works permit. When the Document is included in an aggregate, this 35987 License does not apply to the other works in the aggregate which 35988 are not themselves derivative works of the Document. 35989 35990 If the Cover Text requirement of section 3 is applicable to these 35991 copies of the Document, then if the Document is less than one half 35992 of the entire aggregate, the Document's Cover Texts may be placed 35993 on covers that bracket the Document within the aggregate, or the 35994 electronic equivalent of covers if the Document is in electronic 35995 form. Otherwise they must appear on printed covers that bracket 35996 the whole aggregate. 35997 35998 8. TRANSLATION 35999 36000 Translation is considered a kind of modification, so you may 36001 distribute translations of the Document under the terms of section 36002 4. Replacing Invariant Sections with translations requires special 36003 permission from their copyright holders, but you may include 36004 translations of some or all Invariant Sections in addition to the 36005 original versions of these Invariant Sections. You may include a 36006 translation of this License, and all the license notices in the 36007 Document, and any Warranty Disclaimers, provided that you also 36008 include the original English version of this License and the 36009 original versions of those notices and disclaimers. In case of a 36010 disagreement between the translation and the original version of 36011 this License or a notice or disclaimer, the original version will 36012 prevail. 36013 36014 If a section in the Document is Entitled "Acknowledgements", 36015 "Dedications", or "History", the requirement (section 4) to 36016 Preserve its Title (section 1) will typically require changing the 36017 actual title. 36018 36019 9. TERMINATION 36020 36021 You may not copy, modify, sublicense, or distribute the Document 36022 except as expressly provided for under this License. Any other 36023 attempt to copy, modify, sublicense or distribute the Document is 36024 void, and will automatically terminate your rights under this 36025 License. However, parties who have received copies, or rights, 36026 from you under this License will not have their licenses 36027 terminated so long as such parties remain in full compliance. 36028 36029 10. FUTURE REVISIONS OF THIS LICENSE 36030 36031 The Free Software Foundation may publish new, revised versions of 36032 the GNU Free Documentation License from time to time. Such new 36033 versions will be similar in spirit to the present version, but may 36034 differ in detail to address new problems or concerns. See 36035 `http://www.gnu.org/copyleft/'. 36036 36037 Each version of the License is given a distinguishing version 36038 number. If the Document specifies that a particular numbered 36039 version of this License "or any later version" applies to it, you 36040 have the option of following the terms and conditions either of 36041 that specified version or of any later version that has been 36042 published (not as a draft) by the Free Software Foundation. If 36043 the Document does not specify a version number of this License, 36044 you may choose any version ever published (not as a draft) by the 36045 Free Software Foundation. 36046 36047 ADDENDUM: How to use this License for your documents 36048 ==================================================== 36049 36050 To use this License in a document you have written, include a copy of 36051 the License in the document and put the following copyright and license 36052 notices just after the title page: 36053 36054 Copyright (C) YEAR YOUR NAME. 36055 Permission is granted to copy, distribute and/or modify this document 36056 under the terms of the GNU Free Documentation License, Version 1.2 36057 or any later version published by the Free Software Foundation; 36058 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 36059 Texts. A copy of the license is included in the section entitled ``GNU 36060 Free Documentation License''. 36061 36062 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, 36063 replace the "with...Texts." line with this: 36064 36065 with the Invariant Sections being LIST THEIR TITLES, with 36066 the Front-Cover Texts being LIST, and with the Back-Cover Texts 36067 being LIST. 36068 36069 If you have Invariant Sections without Cover Texts, or some other 36070 combination of the three, merge those two alternatives to suit the 36071 situation. 36072 36073 If your document contains nontrivial examples of program code, we 36074 recommend releasing these examples in parallel under your choice of 36075 free software license, such as the GNU General Public License, to 36076 permit their use in free software. 36077 36078 36079 File: gccint.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top 36080 36081 Contributors to GCC 36082 ******************* 36083 36084 The GCC project would like to thank its many contributors. Without 36085 them the project would not have been nearly as successful as it has 36086 been. Any omissions in this list are accidental. Feel free to contact 36087 <law (a] redhat.com> or <gerald (a] pfeifer.com> if you have been left out or 36088 some of your contributions are not listed. Please keep this list in 36089 alphabetical order. 36090 36091 * Analog Devices helped implement the support for complex data types 36092 and iterators. 36093 36094 * John David Anglin for threading-related fixes and improvements to 36095 libstdc++-v3, and the HP-UX port. 36096 36097 * James van Artsdalen wrote the code that makes efficient use of the 36098 Intel 80387 register stack. 36099 36100 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta 36101 Series port. 36102 36103 * Alasdair Baird for various bug fixes. 36104 36105 * Giovanni Bajo for analyzing lots of complicated C++ problem 36106 reports. 36107 36108 * Peter Barada for his work to improve code generation for new 36109 ColdFire cores. 36110 36111 * Gerald Baumgartner added the signature extension to the C++ front 36112 end. 36113 36114 * Godmar Back for his Java improvements and encouragement. 36115 36116 * Scott Bambrough for help porting the Java compiler. 36117 36118 * Wolfgang Bangerth for processing tons of bug reports. 36119 36120 * Jon Beniston for his Microsoft Windows port of Java. 36121 36122 * Daniel Berlin for better DWARF2 support, faster/better 36123 optimizations, improved alias analysis, plus migrating GCC to 36124 Bugzilla. 36125 36126 * Geoff Berry for his Java object serialization work and various 36127 patches. 36128 36129 * Uros Bizjak for the implementation of x87 math built-in functions 36130 and for various middle end and i386 back end improvements and bug 36131 fixes. 36132 36133 * Eric Blake for helping to make GCJ and libgcj conform to the 36134 specifications. 36135 36136 * Janne Blomqvist for contributions to GNU Fortran. 36137 36138 * Segher Boessenkool for various fixes. 36139 36140 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and 36141 other Java work. 36142 36143 * Neil Booth for work on cpplib, lang hooks, debug hooks and other 36144 miscellaneous clean-ups. 36145 36146 * Steven Bosscher for integrating the GNU Fortran front end into GCC 36147 and for contributing to the tree-ssa branch. 36148 36149 * Eric Botcazou for fixing middle- and backend bugs left and right. 36150 36151 * Per Bothner for his direction via the steering committee and 36152 various improvements to the infrastructure for supporting new 36153 languages. Chill front end implementation. Initial 36154 implementations of cpplib, fix-header, config.guess, libio, and 36155 past C++ library (libg++) maintainer. Dreaming up, designing and 36156 implementing much of GCJ. 36157 36158 * Devon Bowen helped port GCC to the Tahoe. 36159 36160 * Don Bowman for mips-vxworks contributions. 36161 36162 * Dave Brolley for work on cpplib and Chill. 36163 36164 * Paul Brook for work on the ARM architecture and maintaining GNU 36165 Fortran. 36166 36167 * Robert Brown implemented the support for Encore 32000 systems. 36168 36169 * Christian Bruel for improvements to local store elimination. 36170 36171 * Herman A.J. ten Brugge for various fixes. 36172 36173 * Joerg Brunsmann for Java compiler hacking and help with the GCJ 36174 FAQ. 36175 36176 * Joe Buck for his direction via the steering committee. 36177 36178 * Craig Burley for leadership of the G77 Fortran effort. 36179 36180 * Stephan Buys for contributing Doxygen notes for libstdc++. 36181 36182 * Paolo Carlini for libstdc++ work: lots of efficiency improvements 36183 to the C++ strings, streambufs and formatted I/O, hard detective 36184 work on the frustrating localization issues, and keeping up with 36185 the problem reports. 36186 36187 * John Carr for his alias work, SPARC hacking, infrastructure 36188 improvements, previous contributions to the steering committee, 36189 loop optimizations, etc. 36190 36191 * Stephane Carrez for 68HC11 and 68HC12 ports. 36192 36193 * Steve Chamberlain for support for the Renesas SH and H8 processors 36194 and the PicoJava processor, and for GCJ config fixes. 36195 36196 * Glenn Chambers for help with the GCJ FAQ. 36197 36198 * John-Marc Chandonia for various libgcj patches. 36199 36200 * Scott Christley for his Objective-C contributions. 36201 36202 * Eric Christopher for his Java porting help and clean-ups. 36203 36204 * Branko Cibej for more warning contributions. 36205 36206 * The GNU Classpath project for all of their merged runtime code. 36207 36208 * Nick Clifton for arm, mcore, fr30, v850, m32r work, `--help', and 36209 other random hacking. 36210 36211 * Michael Cook for libstdc++ cleanup patches to reduce warnings. 36212 36213 * R. Kelley Cook for making GCC buildable from a read-only directory 36214 as well as other miscellaneous build process and documentation 36215 clean-ups. 36216 36217 * Ralf Corsepius for SH testing and minor bug fixing. 36218 36219 * Stan Cox for care and feeding of the x86 port and lots of behind 36220 the scenes hacking. 36221 36222 * Alex Crain provided changes for the 3b1. 36223 36224 * Ian Dall for major improvements to the NS32k port. 36225 36226 * Paul Dale for his work to add uClinux platform support to the m68k 36227 backend. 36228 36229 * Dario Dariol contributed the four varieties of sample programs 36230 that print a copy of their source. 36231 36232 * Russell Davidson for fstream and stringstream fixes in libstdc++. 36233 36234 * Bud Davis for work on the G77 and GNU Fortran compilers. 36235 36236 * Mo DeJong for GCJ and libgcj bug fixes. 36237 36238 * DJ Delorie for the DJGPP port, build and libiberty maintenance, 36239 various bug fixes, and the M32C port. 36240 36241 * Arnaud Desitter for helping to debug GNU Fortran. 36242 36243 * Gabriel Dos Reis for contributions to G++, contributions and 36244 maintenance of GCC diagnostics infrastructure, libstdc++-v3, 36245 including `valarray<>', `complex<>', maintaining the numerics 36246 library (including that pesky `<limits>' :-) and keeping 36247 up-to-date anything to do with numbers. 36248 36249 * Ulrich Drepper for his work on glibc, testing of GCC using glibc, 36250 ISO C99 support, CFG dumping support, etc., plus support of the 36251 C++ runtime libraries including for all kinds of C interface 36252 issues, contributing and maintaining `complex<>', sanity checking 36253 and disbursement, configuration architecture, libio maintenance, 36254 and early math work. 36255 36256 * Zdenek Dvorak for a new loop unroller and various fixes. 36257 36258 * Richard Earnshaw for his ongoing work with the ARM. 36259 36260 * David Edelsohn for his direction via the steering committee, 36261 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa 36262 loop changes, doing the entire AIX port of libstdc++ with his bare 36263 hands, and for ensuring GCC properly keeps working on AIX. 36264 36265 * Kevin Ediger for the floating point formatting of num_put::do_put 36266 in libstdc++. 36267 36268 * Phil Edwards for libstdc++ work including configuration hackery, 36269 documentation maintainer, chief breaker of the web pages, the 36270 occasional iostream bug fix, and work on shared library symbol 36271 versioning. 36272 36273 * Paul Eggert for random hacking all over GCC. 36274 36275 * Mark Elbrecht for various DJGPP improvements, and for libstdc++ 36276 configuration support for locales and fstream-related fixes. 36277 36278 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and 36279 iostreams. 36280 36281 * Christian Ehrhardt for dealing with bug reports. 36282 36283 * Ben Elliston for his work to move the Objective-C runtime into its 36284 own subdirectory and for his work on autoconf. 36285 36286 * Revital Eres for work on the PowerPC 750CL port. 36287 36288 * Marc Espie for OpenBSD support. 36289 36290 * Doug Evans for much of the global optimization framework, arc, 36291 m32r, and SPARC work. 36292 36293 * Christopher Faylor for his work on the Cygwin port and for caring 36294 and feeding the gcc.gnu.org box and saving its users tons of spam. 36295 36296 * Fred Fish for BeOS support and Ada fixes. 36297 36298 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ. 36299 36300 * Peter Gerwinski for various bug fixes and the Pascal front end. 36301 36302 * Kaveh R. Ghazi for his direction via the steering committee, 36303 amazing work to make `-W -Wall -W* -Werror' useful, and 36304 continuously testing GCC on a plethora of platforms. Kaveh 36305 extends his gratitude to the CAIP Center at Rutgers University for 36306 providing him with computing resources to work on Free Software 36307 since the late 1980s. 36308 36309 * John Gilmore for a donation to the FSF earmarked improving GNU 36310 Java. 36311 36312 * Judy Goldberg for c++ contributions. 36313 36314 * Torbjorn Granlund for various fixes and the c-torture testsuite, 36315 multiply- and divide-by-constant optimization, improved long long 36316 support, improved leaf function register allocation, and his 36317 direction via the steering committee. 36318 36319 * Anthony Green for his `-Os' contributions and Java front end work. 36320 36321 * Stu Grossman for gdb hacking, allowing GCJ developers to debug 36322 Java code. 36323 36324 * Michael K. Gschwind contributed the port to the PDP-11. 36325 36326 * Ron Guilmette implemented the `protoize' and `unprotoize' tools, 36327 the support for Dwarf symbolic debugging information, and much of 36328 the support for System V Release 4. He has also worked heavily on 36329 the Intel 386 and 860 support. 36330 36331 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload 36332 GCSE. 36333 36334 * Bruno Haible for improvements in the runtime overhead for EH, new 36335 warnings and assorted bug fixes. 36336 36337 * Andrew Haley for his amazing Java compiler and library efforts. 36338 36339 * Chris Hanson assisted in making GCC work on HP-UX for the 9000 36340 series 300. 36341 36342 * Michael Hayes for various thankless work he's done trying to get 36343 the c30/c40 ports functional. Lots of loop and unroll 36344 improvements and fixes. 36345 36346 * Dara Hazeghi for wading through myriads of target-specific bug 36347 reports. 36348 36349 * Kate Hedstrom for staking the G77 folks with an initial testsuite. 36350 36351 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64 36352 work, loop opts, and generally fixing lots of old problems we've 36353 ignored for years, flow rewrite and lots of further stuff, 36354 including reviewing tons of patches. 36355 36356 * Aldy Hernandez for working on the PowerPC port, SIMD support, and 36357 various fixes. 36358 36359 * Nobuyuki Hikichi of Software Research Associates, Tokyo, 36360 contributed the support for the Sony NEWS machine. 36361 36362 * Kazu Hirata for caring and feeding the Renesas H8/300 port and 36363 various fixes. 36364 36365 * Katherine Holcomb for work on GNU Fortran. 36366 36367 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots 36368 of testing and bug fixing, particularly of GCC configury code. 36369 36370 * Steve Holmgren for MachTen patches. 36371 36372 * Jan Hubicka for his x86 port improvements. 36373 36374 * Falk Hueffner for working on C and optimization bug reports. 36375 36376 * Bernardo Innocenti for his m68k work, including merging of 36377 ColdFire improvements and uClinux support. 36378 36379 * Christian Iseli for various bug fixes. 36380 36381 * Kamil Iskra for general m68k hacking. 36382 36383 * Lee Iverson for random fixes and MIPS testing. 36384 36385 * Andreas Jaeger for testing and benchmarking of GCC and various bug 36386 fixes. 36387 36388 * Jakub Jelinek for his SPARC work and sibling call optimizations as 36389 well as lots of bug fixes and test cases, and for improving the 36390 Java build system. 36391 36392 * Janis Johnson for ia64 testing and fixes, her quality improvement 36393 sidetracks, and web page maintenance. 36394 36395 * Kean Johnston for SCO OpenServer support and various fixes. 36396 36397 * Tim Josling for the sample language treelang based originally on 36398 Richard Kenner's "toy" language. 36399 36400 * Nicolai Josuttis for additional libstdc++ documentation. 36401 36402 * Klaus Kaempf for his ongoing work to make alpha-vms a viable 36403 target. 36404 36405 * Steven G. Kargl for work on GNU Fortran. 36406 36407 * David Kashtan of SRI adapted GCC to VMS. 36408 36409 * Ryszard Kabatek for many, many libstdc++ bug fixes and 36410 optimizations of strings, especially member functions, and for 36411 auto_ptr fixes. 36412 36413 * Geoffrey Keating for his ongoing work to make the PPC work for 36414 GNU/Linux and his automatic regression tester. 36415 36416 * Brendan Kehoe for his ongoing work with G++ and for a lot of early 36417 work in just about every part of libstdc++. 36418 36419 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the 36420 MIL-STD-1750A. 36421 36422 * Richard Kenner of the New York University Ultracomputer Research 36423 Laboratory wrote the machine descriptions for the AMD 29000, the 36424 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the 36425 support for instruction attributes. He also made changes to 36426 better support RISC processors including changes to common 36427 subexpression elimination, strength reduction, function calling 36428 sequence handling, and condition code support, in addition to 36429 generalizing the code for frame pointer elimination and delay slot 36430 scheduling. Richard Kenner was also the head maintainer of GCC 36431 for several years. 36432 36433 * Mumit Khan for various contributions to the Cygwin and Mingw32 36434 ports and maintaining binary releases for Microsoft Windows hosts, 36435 and for massive libstdc++ porting work to Cygwin/Mingw32. 36436 36437 * Robin Kirkham for cpu32 support. 36438 36439 * Mark Klein for PA improvements. 36440 36441 * Thomas Koenig for various bug fixes. 36442 36443 * Bruce Korb for the new and improved fixincludes code. 36444 36445 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3 36446 effort. 36447 36448 * Charles LaBrec contributed the support for the Integrated Solutions 36449 68020 system. 36450 36451 * Asher Langton and Mike Kumbera for contributing Cray pointer 36452 support to GNU Fortran, and for other GNU Fortran improvements. 36453 36454 * Jeff Law for his direction via the steering committee, 36455 coordinating the entire egcs project and GCC 2.95, rolling out 36456 snapshots and releases, handling merges from GCC2, reviewing tons 36457 of patches that might have fallen through the cracks else, and 36458 random but extensive hacking. 36459 36460 * Marc Lehmann for his direction via the steering committee and 36461 helping with analysis and improvements of x86 performance. 36462 36463 * Victor Leikehman for work on GNU Fortran. 36464 36465 * Ted Lemon wrote parts of the RTL reader and printer. 36466 36467 * Kriang Lerdsuwanakij for C++ improvements including template as 36468 template parameter support, and many C++ fixes. 36469 36470 * Warren Levy for tremendous work on libgcj (Java Runtime Library) 36471 and random work on the Java front end. 36472 36473 * Alain Lichnewsky ported GCC to the MIPS CPU. 36474 36475 * Oskar Liljeblad for hacking on AWT and his many Java bug reports 36476 and patches. 36477 36478 * Robert Lipe for OpenServer support, new testsuites, testing, etc. 36479 36480 * Chen Liqin for various S+core related fixes/improvement, and for 36481 maintaining the S+core port. 36482 36483 * Weiwen Liu for testing and various bug fixes. 36484 36485 * Manuel Lo'pez-Iba'n~ez for improving `-Wconversion' and many other 36486 diagnostics fixes and improvements. 36487 36488 * Dave Love for his ongoing work with the Fortran front end and 36489 runtime libraries. 36490 36491 * Martin von Lo"wis for internal consistency checking infrastructure, 36492 various C++ improvements including namespace support, and tons of 36493 assistance with libstdc++/compiler merges. 36494 36495 * H.J. Lu for his previous contributions to the steering committee, 36496 many x86 bug reports, prototype patches, and keeping the GNU/Linux 36497 ports working. 36498 36499 * Greg McGary for random fixes and (someday) bounded pointers. 36500 36501 * Andrew MacLeod for his ongoing work in building a real EH system, 36502 various code generation improvements, work on the global 36503 optimizer, etc. 36504 36505 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC 36506 hacking improvements to compile-time performance, overall 36507 knowledge and direction in the area of instruction scheduling, and 36508 design and implementation of the automaton based instruction 36509 scheduler. 36510 36511 * Bob Manson for his behind the scenes work on dejagnu. 36512 36513 * Philip Martin for lots of libstdc++ string and vector iterator 36514 fixes and improvements, and string clean up and testsuites. 36515 36516 * All of the Mauve project contributors, for Java test code. 36517 36518 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements. 36519 36520 * Adam Megacz for his work on the Microsoft Windows port of GCJ. 36521 36522 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS, 36523 powerpc, haifa, ECOFF debug support, and other assorted hacking. 36524 36525 * Jason Merrill for his direction via the steering committee and 36526 leading the G++ effort. 36527 36528 * Martin Michlmayr for testing GCC on several architectures using the 36529 entire Debian archive. 36530 36531 * David Miller for his direction via the steering committee, lots of 36532 SPARC work, improvements in jump.c and interfacing with the Linux 36533 kernel developers. 36534 36535 * Gary Miller ported GCC to Charles River Data Systems machines. 36536 36537 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning 36538 the entire libstdc++ testsuite namespace-compatible. 36539 36540 * Mark Mitchell for his direction via the steering committee, 36541 mountains of C++ work, load/store hoisting out of loops, alias 36542 analysis improvements, ISO C `restrict' support, and serving as 36543 release manager for GCC 3.x. 36544 36545 * Alan Modra for various GNU/Linux bits and testing. 36546 36547 * Toon Moene for his direction via the steering committee, Fortran 36548 maintenance, and his ongoing work to make us make Fortran run fast. 36549 36550 * Jason Molenda for major help in the care and feeding of all the 36551 services on the gcc.gnu.org (formerly egcs.cygnus.com) 36552 machine--mail, web services, ftp services, etc etc. Doing all 36553 this work on scrap paper and the backs of envelopes would have 36554 been... difficult. 36555 36556 * Catherine Moore for fixing various ugly problems we have sent her 36557 way, including the haifa bug which was killing the Alpha & PowerPC 36558 Linux kernels. 36559 36560 * Mike Moreton for his various Java patches. 36561 36562 * David Mosberger-Tang for various Alpha improvements, and for the 36563 initial IA-64 port. 36564 36565 * Stephen Moshier contributed the floating point emulator that 36566 assists in cross-compilation and permits support for floating 36567 point numbers wider than 64 bits and for ISO C99 support. 36568 36569 * Bill Moyer for his behind the scenes work on various issues. 36570 36571 * Philippe De Muyter for his work on the m68k port. 36572 36573 * Joseph S. Myers for his work on the PDP-11 port, format checking 36574 and ISO C99 support, and continuous emphasis on (and contributions 36575 to) documentation. 36576 36577 * Nathan Myers for his work on libstdc++-v3: architecture and 36578 authorship through the first three snapshots, including 36579 implementation of locale infrastructure, string, shadow C headers, 36580 and the initial project documentation (DESIGN, CHECKLIST, and so 36581 forth). Later, more work on MT-safe string and shadow headers. 36582 36583 * Felix Natter for documentation on porting libstdc++. 36584 36585 * Nathanael Nerode for cleaning up the configuration/build process. 36586 36587 * NeXT, Inc. donated the front end that supports the Objective-C 36588 language. 36589 36590 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to 36591 the search engine setup, various documentation fixes and other 36592 small fixes. 36593 36594 * Geoff Noer for his work on getting cygwin native builds working. 36595 36596 * Diego Novillo for his work on Tree SSA, OpenMP, SPEC performance 36597 tracking web pages, GIMPLE tuples, and assorted fixes. 36598 36599 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64, 36600 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and 36601 related infrastructure improvements. 36602 36603 * Alexandre Oliva for various build infrastructure improvements, 36604 scripts and amazing testing work, including keeping libtool issues 36605 sane and happy. 36606 36607 * Stefan Olsson for work on mt_alloc. 36608 36609 * Melissa O'Neill for various NeXT fixes. 36610 36611 * Rainer Orth for random MIPS work, including improvements to GCC's 36612 o32 ABI support, improvements to dejagnu's MIPS support, Java 36613 configuration clean-ups and porting work, etc. 36614 36615 * Hartmut Penner for work on the s390 port. 36616 36617 * Paul Petersen wrote the machine description for the Alliant FX/8. 36618 36619 * Alexandre Petit-Bianco for implementing much of the Java compiler 36620 and continued Java maintainership. 36621 36622 * Matthias Pfaller for major improvements to the NS32k port. 36623 36624 * Gerald Pfeifer for his direction via the steering committee, 36625 pointing out lots of problems we need to solve, maintenance of the 36626 web pages, and taking care of documentation maintenance in general. 36627 36628 * Andrew Pinski for processing bug reports by the dozen. 36629 36630 * Ovidiu Predescu for his work on the Objective-C front end and 36631 runtime libraries. 36632 36633 * Jerry Quinn for major performance improvements in C++ formatted 36634 I/O. 36635 36636 * Ken Raeburn for various improvements to checker, MIPS ports and 36637 various cleanups in the compiler. 36638 36639 * Rolf W. Rasmussen for hacking on AWT. 36640 36641 * David Reese of Sun Microsystems contributed to the Solaris on 36642 PowerPC port. 36643 36644 * Volker Reichelt for keeping up with the problem reports. 36645 36646 * Joern Rennecke for maintaining the sh port, loop, regmove & reload 36647 hacking. 36648 36649 * Loren J. Rittle for improvements to libstdc++-v3 including the 36650 FreeBSD port, threading fixes, thread-related configury changes, 36651 critical threading documentation, and solutions to really tricky 36652 I/O problems, as well as keeping GCC properly working on FreeBSD 36653 and continuous testing. 36654 36655 * Craig Rodrigues for processing tons of bug reports. 36656 36657 * Ola Ro"nnerup for work on mt_alloc. 36658 36659 * Gavin Romig-Koch for lots of behind the scenes MIPS work. 36660 36661 * David Ronis inspired and encouraged Craig to rewrite the G77 36662 documentation in texinfo format by contributing a first pass at a 36663 translation of the old `g77-0.5.16/f/DOC' file. 36664 36665 * Ken Rose for fixes to GCC's delay slot filling code. 36666 36667 * Paul Rubin wrote most of the preprocessor. 36668 36669 * Pe'tur Runo'lfsson for major performance improvements in C++ 36670 formatted I/O and large file support in C++ filebuf. 36671 36672 * Chip Salzenberg for libstdc++ patches and improvements to locales, 36673 traits, Makefiles, libio, libtool hackery, and "long long" support. 36674 36675 * Juha Sarlin for improvements to the H8 code generator. 36676 36677 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series 36678 300. 36679 36680 * Roger Sayle for improvements to constant folding and GCC's RTL 36681 optimizers as well as for fixing numerous bugs. 36682 36683 * Bradley Schatz for his work on the GCJ FAQ. 36684 36685 * Peter Schauer wrote the code to allow debugging to work on the 36686 Alpha. 36687 36688 * William Schelter did most of the work on the Intel 80386 support. 36689 36690 * Tobias Schlu"ter for work on GNU Fortran. 36691 36692 * Bernd Schmidt for various code generation improvements and major 36693 work in the reload pass as well a serving as release manager for 36694 GCC 2.95.3. 36695 36696 * Peter Schmid for constant testing of libstdc++--especially 36697 application testing, going above and beyond what was requested for 36698 the release criteria--and libstdc++ header file tweaks. 36699 36700 * Jason Schroeder for jcf-dump patches. 36701 36702 * Andreas Schwab for his work on the m68k port. 36703 36704 * Lars Segerlund for work on GNU Fortran. 36705 36706 * Joel Sherrill for his direction via the steering committee, RTEMS 36707 contributions and RTEMS testing. 36708 36709 * Nathan Sidwell for many C++ fixes/improvements. 36710 36711 * Jeffrey Siegal for helping RMS with the original design of GCC, 36712 some code which handles the parse tree and RTL data structures, 36713 constant folding and help with the original VAX & m68k ports. 36714 36715 * Kenny Simpson for prompting libstdc++ fixes due to defect reports 36716 from the LWG (thereby keeping GCC in line with updates from the 36717 ISO). 36718 36719 * Franz Sirl for his ongoing work with making the PPC port stable 36720 for GNU/Linux. 36721 36722 * Andrey Slepuhin for assorted AIX hacking. 36723 36724 * Trevor Smigiel for contributing the SPU port. 36725 36726 * Christopher Smith did the port for Convex machines. 36727 36728 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports. 36729 36730 * Randy Smith finished the Sun FPA support. 36731 36732 * Scott Snyder for queue, iterator, istream, and string fixes and 36733 libstdc++ testsuite entries. Also for providing the patch to G77 36734 to add rudimentary support for `INTEGER*1', `INTEGER*2', and 36735 `LOGICAL*1'. 36736 36737 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique. 36738 36739 * Richard Stallman, for writing the original GCC and launching the 36740 GNU project. 36741 36742 * Jan Stein of the Chalmers Computer Society provided support for 36743 Genix, as well as part of the 32000 machine description. 36744 36745 * Nigel Stephens for various mips16 related fixes/improvements. 36746 36747 * Jonathan Stone wrote the machine description for the Pyramid 36748 computer. 36749 36750 * Graham Stott for various infrastructure improvements. 36751 36752 * John Stracke for his Java HTTP protocol fixes. 36753 36754 * Mike Stump for his Elxsi port, G++ contributions over the years 36755 and more recently his vxworks contributions 36756 36757 * Jeff Sturm for Java porting help, bug fixes, and encouragement. 36758 36759 * Shigeya Suzuki for this fixes for the bsdi platforms. 36760 36761 * Ian Lance Taylor for his mips16 work, general configury hacking, 36762 fixincludes, etc. 36763 36764 * Holger Teutsch provided the support for the Clipper CPU. 36765 36766 * Gary Thomas for his ongoing work to make the PPC work for 36767 GNU/Linux. 36768 36769 * Philipp Thomas for random bug fixes throughout the compiler 36770 36771 * Jason Thorpe for thread support in libstdc++ on NetBSD. 36772 36773 * Kresten Krab Thorup wrote the run time support for the Objective-C 36774 language and the fantastic Java bytecode interpreter. 36775 36776 * Michael Tiemann for random bug fixes, the first instruction 36777 scheduler, initial C++ support, function integration, NS32k, SPARC 36778 and M88k machine description work, delay slot scheduling. 36779 36780 * Andreas Tobler for his work porting libgcj to Darwin. 36781 36782 * Teemu Torma for thread safe exception handling support. 36783 36784 * Leonard Tower wrote parts of the parser, RTL generator, and RTL 36785 definitions, and of the VAX machine description. 36786 36787 * Daniel Towner and Hariharan Sandanagobalane contributed and 36788 maintain the picoChip port. 36789 36790 * Tom Tromey for internationalization support and for his many Java 36791 contributions and libgcj maintainership. 36792 36793 * Lassi Tuura for improvements to config.guess to determine HP 36794 processor types. 36795 36796 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes. 36797 36798 * Andy Vaught for the design and initial implementation of the GNU 36799 Fortran front end. 36800 36801 * Brent Verner for work with the libstdc++ cshadow files and their 36802 associated configure steps. 36803 36804 * Todd Vierling for contributions for NetBSD ports. 36805 36806 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML 36807 guidance. 36808 36809 * Dean Wakerley for converting the install documentation from HTML 36810 to texinfo in time for GCC 3.0. 36811 36812 * Krister Walfridsson for random bug fixes. 36813 36814 * Feng Wang for contributions to GNU Fortran. 36815 36816 * Stephen M. Webb for time and effort on making libstdc++ shadow 36817 files work with the tricky Solaris 8+ headers, and for pushing the 36818 build-time header tree. 36819 36820 * John Wehle for various improvements for the x86 code generator, 36821 related infrastructure improvements to help x86 code generation, 36822 value range propagation and other work, WE32k port. 36823 36824 * Ulrich Weigand for work on the s390 port. 36825 36826 * Zack Weinberg for major work on cpplib and various other bug fixes. 36827 36828 * Matt Welsh for help with Linux Threads support in GCJ. 36829 36830 * Urban Widmark for help fixing java.io. 36831 36832 * Mark Wielaard for new Java library code and his work integrating 36833 with Classpath. 36834 36835 * Dale Wiles helped port GCC to the Tahoe. 36836 36837 * Bob Wilson from Tensilica, Inc. for the Xtensa port. 36838 36839 * Jim Wilson for his direction via the steering committee, tackling 36840 hard problems in various places that nobody else wanted to work 36841 on, strength reduction and other loop optimizations. 36842 36843 * Paul Woegerer and Tal Agmon for the CRX port. 36844 36845 * Carlo Wood for various fixes. 36846 36847 * Tom Wood for work on the m88k port. 36848 36849 * Canqun Yang for work on GNU Fortran. 36850 36851 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine 36852 description for the Tron architecture (specifically, the Gmicro). 36853 36854 * Kevin Zachmann helped port GCC to the Tahoe. 36855 36856 * Ayal Zaks for Swing Modulo Scheduling (SMS). 36857 36858 * Xiaoqiang Zhang for work on GNU Fortran. 36859 36860 * Gilles Zunino for help porting Java to Irix. 36861 36862 36863 The following people are recognized for their contributions to GNAT, 36864 the Ada front end of GCC: 36865 * Bernard Banner 36866 36867 * Romain Berrendonner 36868 36869 * Geert Bosch 36870 36871 * Emmanuel Briot 36872 36873 * Joel Brobecker 36874 36875 * Ben Brosgol 36876 36877 * Vincent Celier 36878 36879 * Arnaud Charlet 36880 36881 * Chien Chieng 36882 36883 * Cyrille Comar 36884 36885 * Cyrille Crozes 36886 36887 * Robert Dewar 36888 36889 * Gary Dismukes 36890 36891 * Robert Duff 36892 36893 * Ed Falis 36894 36895 * Ramon Fernandez 36896 36897 * Sam Figueroa 36898 36899 * Vasiliy Fofanov 36900 36901 * Michael Friess 36902 36903 * Franco Gasperoni 36904 36905 * Ted Giering 36906 36907 * Matthew Gingell 36908 36909 * Laurent Guerby 36910 36911 * Jerome Guitton 36912 36913 * Olivier Hainque 36914 36915 * Jerome Hugues 36916 36917 * Hristian Kirtchev 36918 36919 * Jerome Lambourg 36920 36921 * Bruno Leclerc 36922 36923 * Albert Lee 36924 36925 * Sean McNeil 36926 36927 * Javier Miranda 36928 36929 * Laurent Nana 36930 36931 * Pascal Obry 36932 36933 * Dong-Ik Oh 36934 36935 * Laurent Pautet 36936 36937 * Brett Porter 36938 36939 * Thomas Quinot 36940 36941 * Nicolas Roche 36942 36943 * Pat Rogers 36944 36945 * Jose Ruiz 36946 36947 * Douglas Rupp 36948 36949 * Sergey Rybin 36950 36951 * Gail Schenker 36952 36953 * Ed Schonberg 36954 36955 * Nicolas Setton 36956 36957 * Samuel Tardieu 36958 36959 36960 The following people are recognized for their contributions of new 36961 features, bug reports, testing and integration of classpath/libgcj for 36962 GCC version 4.1: 36963 * Lillian Angel for `JTree' implementation and lots Free Swing 36964 additions and bug fixes. 36965 36966 * Wolfgang Baer for `GapContent' bug fixes. 36967 36968 * Anthony Balkissoon for `JList', Free Swing 1.5 updates and mouse 36969 event fixes, lots of Free Swing work including `JTable' editing. 36970 36971 * Stuart Ballard for RMI constant fixes. 36972 36973 * Goffredo Baroncelli for `HTTPURLConnection' fixes. 36974 36975 * Gary Benson for `MessageFormat' fixes. 36976 36977 * Daniel Bonniot for `Serialization' fixes. 36978 36979 * Chris Burdess for lots of gnu.xml and http protocol fixes, `StAX' 36980 and `DOM xml:id' support. 36981 36982 * Ka-Hing Cheung for `TreePath' and `TreeSelection' fixes. 36983 36984 * Archie Cobbs for build fixes, VM interface updates, 36985 `URLClassLoader' updates. 36986 36987 * Kelley Cook for build fixes. 36988 36989 * Martin Cordova for Suggestions for better `SocketTimeoutException'. 36990 36991 * David Daney for `BitSet' bug fixes, `HttpURLConnection' rewrite 36992 and improvements. 36993 36994 * Thomas Fitzsimmons for lots of upgrades to the gtk+ AWT and Cairo 36995 2D support. Lots of imageio framework additions, lots of AWT and 36996 Free Swing bug fixes. 36997 36998 * Jeroen Frijters for `ClassLoader' and nio cleanups, serialization 36999 fixes, better `Proxy' support, bug fixes and IKVM integration. 37000 37001 * Santiago Gala for `AccessControlContext' fixes. 37002 37003 * Nicolas Geoffray for `VMClassLoader' and `AccessController' 37004 improvements. 37005 37006 * David Gilbert for `basic' and `metal' icon and plaf support and 37007 lots of documenting, Lots of Free Swing and metal theme additions. 37008 `MetalIconFactory' implementation. 37009 37010 * Anthony Green for `MIDI' framework, `ALSA' and `DSSI' providers. 37011 37012 * Andrew Haley for `Serialization' and `URLClassLoader' fixes, gcj 37013 build speedups. 37014 37015 * Kim Ho for `JFileChooser' implementation. 37016 37017 * Andrew John Hughes for `Locale' and net fixes, URI RFC2986 37018 updates, `Serialization' fixes, `Properties' XML support and 37019 generic branch work, VMIntegration guide update. 37020 37021 * Bastiaan Huisman for `TimeZone' bug fixing. 37022 37023 * Andreas Jaeger for mprec updates. 37024 37025 * Paul Jenner for better `-Werror' support. 37026 37027 * Ito Kazumitsu for `NetworkInterface' implementation and updates. 37028 37029 * Roman Kennke for `BoxLayout', `GrayFilter' and `SplitPane', plus 37030 bug fixes all over. Lots of Free Swing work including styled text. 37031 37032 * Simon Kitching for `String' cleanups and optimization suggestions. 37033 37034 * Michael Koch for configuration fixes, `Locale' updates, bug and 37035 build fixes. 37036 37037 * Guilhem Lavaux for configuration, thread and channel fixes and 37038 Kaffe integration. JCL native `Pointer' updates. Logger bug fixes. 37039 37040 * David Lichteblau for JCL support library global/local reference 37041 cleanups. 37042 37043 * Aaron Luchko for JDWP updates and documentation fixes. 37044 37045 * Ziga Mahkovec for `Graphics2D' upgraded to Cairo 0.5 and new regex 37046 features. 37047 37048 * Sven de Marothy for BMP imageio support, CSS and `TextLayout' 37049 fixes. `GtkImage' rewrite, 2D, awt, free swing and date/time fixes 37050 and implementing the Qt4 peers. 37051 37052 * Casey Marshall for crypto algorithm fixes, `FileChannel' lock, 37053 `SystemLogger' and `FileHandler' rotate implementations, NIO 37054 `FileChannel.map' support, security and policy updates. 37055 37056 * Bryce McKinlay for RMI work. 37057 37058 * Audrius Meskauskas for lots of Free Corba, RMI and HTML work plus 37059 testing and documenting. 37060 37061 * Kalle Olavi Niemitalo for build fixes. 37062 37063 * Rainer Orth for build fixes. 37064 37065 * Andrew Overholt for `File' locking fixes. 37066 37067 * Ingo Proetel for `Image', `Logger' and `URLClassLoader' updates. 37068 37069 * Olga Rodimina for `MenuSelectionManager' implementation. 37070 37071 * Jan Roehrich for `BasicTreeUI' and `JTree' fixes. 37072 37073 * Julian Scheid for documentation updates and gjdoc support. 37074 37075 * Christian Schlichtherle for zip fixes and cleanups. 37076 37077 * Robert Schuster for documentation updates and beans fixes, 37078 `TreeNode' enumerations and `ActionCommand' and various fixes, XML 37079 and URL, AWT and Free Swing bug fixes. 37080 37081 * Keith Seitz for lots of JDWP work. 37082 37083 * Christian Thalinger for 64-bit cleanups, Configuration and VM 37084 interface fixes and `CACAO' integration, `fdlibm' updates. 37085 37086 * Gael Thomas for `VMClassLoader' boot packages support suggestions. 37087 37088 * Andreas Tobler for Darwin and Solaris testing and fixing, `Qt4' 37089 support for Darwin/OS X, `Graphics2D' support, `gtk+' updates. 37090 37091 * Dalibor Topic for better `DEBUG' support, build cleanups and Kaffe 37092 integration. `Qt4' build infrastructure, `SHA1PRNG' and 37093 `GdkPixbugDecoder' updates. 37094 37095 * Tom Tromey for Eclipse integration, generics work, lots of bug 37096 fixes and gcj integration including coordinating The Big Merge. 37097 37098 * Mark Wielaard for bug fixes, packaging and release management, 37099 `Clipboard' implementation, system call interrupts and network 37100 timeouts and `GdkPixpufDecoder' fixes. 37101 37102 37103 In addition to the above, all of which also contributed time and 37104 energy in testing GCC, we would like to thank the following for their 37105 contributions to testing: 37106 37107 * Michael Abd-El-Malek 37108 37109 * Thomas Arend 37110 37111 * Bonzo Armstrong 37112 37113 * Steven Ashe 37114 37115 * Chris Baldwin 37116 37117 * David Billinghurst 37118 37119 * Jim Blandy 37120 37121 * Stephane Bortzmeyer 37122 37123 * Horst von Brand 37124 37125 * Frank Braun 37126 37127 * Rodney Brown 37128 37129 * Sidney Cadot 37130 37131 * Bradford Castalia 37132 37133 * Robert Clark 37134 37135 * Jonathan Corbet 37136 37137 * Ralph Doncaster 37138 37139 * Richard Emberson 37140 37141 * Levente Farkas 37142 37143 * Graham Fawcett 37144 37145 * Mark Fernyhough 37146 37147 * Robert A. French 37148 37149 * Jo"rgen Freyh 37150 37151 * Mark K. Gardner 37152 37153 * Charles-Antoine Gauthier 37154 37155 * Yung Shing Gene 37156 37157 * David Gilbert 37158 37159 * Simon Gornall 37160 37161 * Fred Gray 37162 37163 * John Griffin 37164 37165 * Patrik Hagglund 37166 37167 * Phil Hargett 37168 37169 * Amancio Hasty 37170 37171 * Takafumi Hayashi 37172 37173 * Bryan W. Headley 37174 37175 * Kevin B. Hendricks 37176 37177 * Joep Jansen 37178 37179 * Christian Joensson 37180 37181 * Michel Kern 37182 37183 * David Kidd 37184 37185 * Tobias Kuipers 37186 37187 * Anand Krishnaswamy 37188 37189 * A. O. V. Le Blanc 37190 37191 * llewelly 37192 37193 * Damon Love 37194 37195 * Brad Lucier 37196 37197 * Matthias Klose 37198 37199 * Martin Knoblauch 37200 37201 * Rick Lutowski 37202 37203 * Jesse Macnish 37204 37205 * Stefan Morrell 37206 37207 * Anon A. Mous 37208 37209 * Matthias Mueller 37210 37211 * Pekka Nikander 37212 37213 * Rick Niles 37214 37215 * Jon Olson 37216 37217 * Magnus Persson 37218 37219 * Chris Pollard 37220 37221 * Richard Polton 37222 37223 * Derk Reefman 37224 37225 * David Rees 37226 37227 * Paul Reilly 37228 37229 * Tom Reilly 37230 37231 * Torsten Rueger 37232 37233 * Danny Sadinoff 37234 37235 * Marc Schifer 37236 37237 * Erik Schnetter 37238 37239 * Wayne K. Schroll 37240 37241 * David Schuler 37242 37243 * Vin Shelton 37244 37245 * Tim Souder 37246 37247 * Adam Sulmicki 37248 37249 * Bill Thorson 37250 37251 * George Talbot 37252 37253 * Pedro A. M. Vazquez 37254 37255 * Gregory Warnes 37256 37257 * Ian Watson 37258 37259 * David E. Young 37260 37261 * And many others 37262 37263 And finally we'd like to thank everyone who uses the compiler, provides 37264 feedback and generally reminds us why we're doing this work in the first 37265 place. 37266 37267 37268 File: gccint.info, Node: Option Index, Next: Concept Index, Prev: Contributors, Up: Top 37269 37270 Option Index 37271 ************ 37272 37273 GCC's command line options are indexed here without any initial `-' or 37274 `--'. Where an option has both positive and negative forms (such as 37275 `-fOPTION' and `-fno-OPTION'), relevant entries in the manual are 37276 indexed under the most appropriate form; it may sometimes be useful to 37277 look up both forms. 37278 37279 [index] 37280 * Menu: 37281 37282 * msoft-float: Soft float library routines. 37283 (line 6) 37284 37285 37286 File: gccint.info, Node: Concept Index, Prev: Option Index, Up: Top 37287 37288 Concept Index 37289 ************* 37290 37291 [index] 37292 * Menu: 37293 37294 * ! in constraint: Multi-Alternative. (line 47) 37295 * # in constraint: Modifiers. (line 67) 37296 * # in template: Output Template. (line 66) 37297 * #pragma: Misc. (line 381) 37298 * % in constraint: Modifiers. (line 45) 37299 * % in GTY option: GTY Options. (line 18) 37300 * % in template: Output Template. (line 6) 37301 * & in constraint: Modifiers. (line 25) 37302 * ( <1>: GIMPLE_ASM. (line 27) 37303 * ( <2>: GIMPLE_CALL. (line 63) 37304 * ( <3>: Logical Operators. (line 124) 37305 * ( <4>: GIMPLE_ASM. (line 24) 37306 * ( <5>: Logical Operators. (line 128) 37307 * ( <6>: Sections. (line 160) 37308 * (: Logical Operators. (line 131) 37309 * (nil): RTL Objects. (line 73) 37310 * * <1>: Scheduling. (line 268) 37311 * *: Host Common. (line 17) 37312 * * in constraint: Modifiers. (line 72) 37313 * * in template: Output Statement. (line 29) 37314 * *gimple_assign_lhs_ptr: GIMPLE_ASSIGN. (line 54) 37315 * *gimple_assign_rhs1_ptr: GIMPLE_ASSIGN. (line 60) 37316 * *gimple_assign_rhs2_ptr: GIMPLE_ASSIGN. (line 67) 37317 * *gimple_call_arg_ptr: GIMPLE_CALL. (line 71) 37318 * *gimple_call_lhs_ptr: GIMPLE_CALL. (line 32) 37319 * *gimple_catch_types_ptr: GIMPLE_CATCH. (line 16) 37320 * *gimple_cdt_location_ptr: GIMPLE_CHANGE_DYNAMIC_TYPE. 37321 (line 28) 37322 * *gimple_cdt_new_type_ptr: GIMPLE_CHANGE_DYNAMIC_TYPE. 37323 (line 15) 37324 * *gimple_eh_filter_types_ptr: GIMPLE_EH_FILTER. (line 15) 37325 * *gimple_omp_critical_name_ptr: GIMPLE_OMP_CRITICAL. 37326 (line 16) 37327 * *gimple_omp_for_clauses_ptr: GIMPLE_OMP_FOR. (line 23) 37328 * *gimple_omp_for_final_ptr: GIMPLE_OMP_FOR. (line 54) 37329 * *gimple_omp_for_incr_ptr: GIMPLE_OMP_FOR. (line 64) 37330 * *gimple_omp_for_index_ptr: GIMPLE_OMP_FOR. (line 34) 37331 * *gimple_omp_for_initial_ptr: GIMPLE_OMP_FOR. (line 44) 37332 * *gimple_omp_parallel_child_fn_ptr: GIMPLE_OMP_PARALLEL. 37333 (line 46) 37334 * *gimple_omp_parallel_clauses_ptr: GIMPLE_OMP_PARALLEL. 37335 (line 34) 37336 * *gimple_omp_parallel_data_arg_ptr: GIMPLE_OMP_PARALLEL. 37337 (line 58) 37338 * *gimple_omp_sections_clauses_ptr: GIMPLE_OMP_SECTIONS. 37339 (line 33) 37340 * *gimple_omp_sections_control_ptr: GIMPLE_OMP_SECTIONS. 37341 (line 21) 37342 * *gimple_omp_single_clauses_ptr: GIMPLE_OMP_SINGLE. (line 17) 37343 * *gimple_op_ptr: Manipulating GIMPLE statements. 37344 (line 84) 37345 * *gimple_ops <1>: Manipulating GIMPLE statements. 37346 (line 78) 37347 * *gimple_ops: Logical Operators. (line 82) 37348 * *gimple_phi_result_ptr: GIMPLE_PHI. (line 22) 37349 * *gsi_stmt_ptr: Sequence iterators. (line 80) 37350 * *TARGET_GET_PCH_VALIDITY: PCH Target. (line 7) 37351 * + in constraint: Modifiers. (line 12) 37352 * -fsection-anchors <1>: Special Accessors. (line 106) 37353 * -fsection-anchors: Anchored Addresses. (line 6) 37354 * /c in RTL dump: Flags. (line 234) 37355 * /f in RTL dump: Flags. (line 242) 37356 * /i in RTL dump: Flags. (line 294) 37357 * /j in RTL dump: Flags. (line 309) 37358 * /s in RTL dump: Flags. (line 258) 37359 * /u in RTL dump: Flags. (line 319) 37360 * /v in RTL dump: Flags. (line 351) 37361 * 0 in constraint: Simple Constraints. (line 120) 37362 * < in constraint: Simple Constraints. (line 48) 37363 * = in constraint: Modifiers. (line 8) 37364 * > in constraint: Simple Constraints. (line 52) 37365 * ? in constraint: Multi-Alternative. (line 41) 37366 * \: Output Template. (line 46) 37367 * __absvdi2: Integer library routines. 37368 (line 107) 37369 * __absvsi2: Integer library routines. 37370 (line 106) 37371 * __addda3: Fixed-point fractional library routines. 37372 (line 45) 37373 * __adddf3: Soft float library routines. 37374 (line 23) 37375 * __adddq3: Fixed-point fractional library routines. 37376 (line 33) 37377 * __addha3: Fixed-point fractional library routines. 37378 (line 43) 37379 * __addhq3: Fixed-point fractional library routines. 37380 (line 30) 37381 * __addqq3: Fixed-point fractional library routines. 37382 (line 29) 37383 * __addsa3: Fixed-point fractional library routines. 37384 (line 44) 37385 * __addsf3: Soft float library routines. 37386 (line 22) 37387 * __addsq3: Fixed-point fractional library routines. 37388 (line 31) 37389 * __addta3: Fixed-point fractional library routines. 37390 (line 47) 37391 * __addtf3: Soft float library routines. 37392 (line 25) 37393 * __adduda3: Fixed-point fractional library routines. 37394 (line 53) 37395 * __addudq3: Fixed-point fractional library routines. 37396 (line 41) 37397 * __adduha3: Fixed-point fractional library routines. 37398 (line 49) 37399 * __adduhq3: Fixed-point fractional library routines. 37400 (line 37) 37401 * __adduqq3: Fixed-point fractional library routines. 37402 (line 35) 37403 * __addusa3: Fixed-point fractional library routines. 37404 (line 51) 37405 * __addusq3: Fixed-point fractional library routines. 37406 (line 39) 37407 * __adduta3: Fixed-point fractional library routines. 37408 (line 55) 37409 * __addvdi3: Integer library routines. 37410 (line 111) 37411 * __addvsi3: Integer library routines. 37412 (line 110) 37413 * __addxf3: Soft float library routines. 37414 (line 27) 37415 * __ashlda3: Fixed-point fractional library routines. 37416 (line 351) 37417 * __ashldi3: Integer library routines. 37418 (line 14) 37419 * __ashldq3: Fixed-point fractional library routines. 37420 (line 340) 37421 * __ashlha3: Fixed-point fractional library routines. 37422 (line 349) 37423 * __ashlhq3: Fixed-point fractional library routines. 37424 (line 337) 37425 * __ashlqq3: Fixed-point fractional library routines. 37426 (line 336) 37427 * __ashlsa3: Fixed-point fractional library routines. 37428 (line 350) 37429 * __ashlsi3: Integer library routines. 37430 (line 13) 37431 * __ashlsq3: Fixed-point fractional library routines. 37432 (line 338) 37433 * __ashlta3: Fixed-point fractional library routines. 37434 (line 353) 37435 * __ashlti3: Integer library routines. 37436 (line 15) 37437 * __ashluda3: Fixed-point fractional library routines. 37438 (line 359) 37439 * __ashludq3: Fixed-point fractional library routines. 37440 (line 348) 37441 * __ashluha3: Fixed-point fractional library routines. 37442 (line 355) 37443 * __ashluhq3: Fixed-point fractional library routines. 37444 (line 344) 37445 * __ashluqq3: Fixed-point fractional library routines. 37446 (line 342) 37447 * __ashlusa3: Fixed-point fractional library routines. 37448 (line 357) 37449 * __ashlusq3: Fixed-point fractional library routines. 37450 (line 346) 37451 * __ashluta3: Fixed-point fractional library routines. 37452 (line 361) 37453 * __ashrda3: Fixed-point fractional library routines. 37454 (line 371) 37455 * __ashrdi3: Integer library routines. 37456 (line 19) 37457 * __ashrdq3: Fixed-point fractional library routines. 37458 (line 368) 37459 * __ashrha3: Fixed-point fractional library routines. 37460 (line 369) 37461 * __ashrhq3: Fixed-point fractional library routines. 37462 (line 365) 37463 * __ashrqq3: Fixed-point fractional library routines. 37464 (line 364) 37465 * __ashrsa3: Fixed-point fractional library routines. 37466 (line 370) 37467 * __ashrsi3: Integer library routines. 37468 (line 18) 37469 * __ashrsq3: Fixed-point fractional library routines. 37470 (line 366) 37471 * __ashrta3: Fixed-point fractional library routines. 37472 (line 373) 37473 * __ashrti3: Integer library routines. 37474 (line 20) 37475 * __bid_adddd3: Decimal float library routines. 37476 (line 25) 37477 * __bid_addsd3: Decimal float library routines. 37478 (line 21) 37479 * __bid_addtd3: Decimal float library routines. 37480 (line 29) 37481 * __bid_divdd3: Decimal float library routines. 37482 (line 68) 37483 * __bid_divsd3: Decimal float library routines. 37484 (line 64) 37485 * __bid_divtd3: Decimal float library routines. 37486 (line 72) 37487 * __bid_eqdd2: Decimal float library routines. 37488 (line 259) 37489 * __bid_eqsd2: Decimal float library routines. 37490 (line 257) 37491 * __bid_eqtd2: Decimal float library routines. 37492 (line 261) 37493 * __bid_extendddtd2: Decimal float library routines. 37494 (line 92) 37495 * __bid_extendddtf: Decimal float library routines. 37496 (line 140) 37497 * __bid_extendddxf: Decimal float library routines. 37498 (line 134) 37499 * __bid_extenddfdd: Decimal float library routines. 37500 (line 147) 37501 * __bid_extenddftd: Decimal float library routines. 37502 (line 107) 37503 * __bid_extendsddd2: Decimal float library routines. 37504 (line 88) 37505 * __bid_extendsddf: Decimal float library routines. 37506 (line 128) 37507 * __bid_extendsdtd2: Decimal float library routines. 37508 (line 90) 37509 * __bid_extendsdtf: Decimal float library routines. 37510 (line 138) 37511 * __bid_extendsdxf: Decimal float library routines. 37512 (line 132) 37513 * __bid_extendsfdd: Decimal float library routines. 37514 (line 103) 37515 * __bid_extendsfsd: Decimal float library routines. 37516 (line 145) 37517 * __bid_extendsftd: Decimal float library routines. 37518 (line 105) 37519 * __bid_extendtftd: Decimal float library routines. 37520 (line 149) 37521 * __bid_extendxftd: Decimal float library routines. 37522 (line 109) 37523 * __bid_fixdddi: Decimal float library routines. 37524 (line 170) 37525 * __bid_fixddsi: Decimal float library routines. 37526 (line 162) 37527 * __bid_fixsddi: Decimal float library routines. 37528 (line 168) 37529 * __bid_fixsdsi: Decimal float library routines. 37530 (line 160) 37531 * __bid_fixtddi: Decimal float library routines. 37532 (line 172) 37533 * __bid_fixtdsi: Decimal float library routines. 37534 (line 164) 37535 * __bid_fixunsdddi: Decimal float library routines. 37536 (line 187) 37537 * __bid_fixunsddsi: Decimal float library routines. 37538 (line 178) 37539 * __bid_fixunssddi: Decimal float library routines. 37540 (line 185) 37541 * __bid_fixunssdsi: Decimal float library routines. 37542 (line 176) 37543 * __bid_fixunstddi: Decimal float library routines. 37544 (line 189) 37545 * __bid_fixunstdsi: Decimal float library routines. 37546 (line 180) 37547 * __bid_floatdidd: Decimal float library routines. 37548 (line 205) 37549 * __bid_floatdisd: Decimal float library routines. 37550 (line 203) 37551 * __bid_floatditd: Decimal float library routines. 37552 (line 207) 37553 * __bid_floatsidd: Decimal float library routines. 37554 (line 196) 37555 * __bid_floatsisd: Decimal float library routines. 37556 (line 194) 37557 * __bid_floatsitd: Decimal float library routines. 37558 (line 198) 37559 * __bid_floatunsdidd: Decimal float library routines. 37560 (line 223) 37561 * __bid_floatunsdisd: Decimal float library routines. 37562 (line 221) 37563 * __bid_floatunsditd: Decimal float library routines. 37564 (line 225) 37565 * __bid_floatunssidd: Decimal float library routines. 37566 (line 214) 37567 * __bid_floatunssisd: Decimal float library routines. 37568 (line 212) 37569 * __bid_floatunssitd: Decimal float library routines. 37570 (line 216) 37571 * __bid_gedd2: Decimal float library routines. 37572 (line 277) 37573 * __bid_gesd2: Decimal float library routines. 37574 (line 275) 37575 * __bid_getd2: Decimal float library routines. 37576 (line 279) 37577 * __bid_gtdd2: Decimal float library routines. 37578 (line 304) 37579 * __bid_gtsd2: Decimal float library routines. 37580 (line 302) 37581 * __bid_gttd2: Decimal float library routines. 37582 (line 306) 37583 * __bid_ledd2: Decimal float library routines. 37584 (line 295) 37585 * __bid_lesd2: Decimal float library routines. 37586 (line 293) 37587 * __bid_letd2: Decimal float library routines. 37588 (line 297) 37589 * __bid_ltdd2: Decimal float library routines. 37590 (line 286) 37591 * __bid_ltsd2: Decimal float library routines. 37592 (line 284) 37593 * __bid_lttd2: Decimal float library routines. 37594 (line 288) 37595 * __bid_muldd3: Decimal float library routines. 37596 (line 54) 37597 * __bid_mulsd3: Decimal float library routines. 37598 (line 50) 37599 * __bid_multd3: Decimal float library routines. 37600 (line 58) 37601 * __bid_nedd2: Decimal float library routines. 37602 (line 268) 37603 * __bid_negdd2: Decimal float library routines. 37604 (line 78) 37605 * __bid_negsd2: Decimal float library routines. 37606 (line 76) 37607 * __bid_negtd2: Decimal float library routines. 37608 (line 80) 37609 * __bid_nesd2: Decimal float library routines. 37610 (line 266) 37611 * __bid_netd2: Decimal float library routines. 37612 (line 270) 37613 * __bid_subdd3: Decimal float library routines. 37614 (line 39) 37615 * __bid_subsd3: Decimal float library routines. 37616 (line 35) 37617 * __bid_subtd3: Decimal float library routines. 37618 (line 43) 37619 * __bid_truncdddf: Decimal float library routines. 37620 (line 153) 37621 * __bid_truncddsd2: Decimal float library routines. 37622 (line 94) 37623 * __bid_truncddsf: Decimal float library routines. 37624 (line 124) 37625 * __bid_truncdfsd: Decimal float library routines. 37626 (line 111) 37627 * __bid_truncsdsf: Decimal float library routines. 37628 (line 151) 37629 * __bid_trunctddd2: Decimal float library routines. 37630 (line 98) 37631 * __bid_trunctddf: Decimal float library routines. 37632 (line 130) 37633 * __bid_trunctdsd2: Decimal float library routines. 37634 (line 96) 37635 * __bid_trunctdsf: Decimal float library routines. 37636 (line 126) 37637 * __bid_trunctdtf: Decimal float library routines. 37638 (line 155) 37639 * __bid_trunctdxf: Decimal float library routines. 37640 (line 136) 37641 * __bid_trunctfdd: Decimal float library routines. 37642 (line 119) 37643 * __bid_trunctfsd: Decimal float library routines. 37644 (line 115) 37645 * __bid_truncxfdd: Decimal float library routines. 37646 (line 117) 37647 * __bid_truncxfsd: Decimal float library routines. 37648 (line 113) 37649 * __bid_unorddd2: Decimal float library routines. 37650 (line 235) 37651 * __bid_unordsd2: Decimal float library routines. 37652 (line 233) 37653 * __bid_unordtd2: Decimal float library routines. 37654 (line 237) 37655 * __bswapdi2: Integer library routines. 37656 (line 162) 37657 * __bswapsi2: Integer library routines. 37658 (line 161) 37659 * __builtin_args_info: Varargs. (line 42) 37660 * __builtin_classify_type: Varargs. (line 76) 37661 * __builtin_next_arg: Varargs. (line 66) 37662 * __builtin_saveregs: Varargs. (line 24) 37663 * __clear_cache: Miscellaneous routines. 37664 (line 10) 37665 * __clzdi2: Integer library routines. 37666 (line 131) 37667 * __clzsi2: Integer library routines. 37668 (line 130) 37669 * __clzti2: Integer library routines. 37670 (line 132) 37671 * __cmpda2: Fixed-point fractional library routines. 37672 (line 451) 37673 * __cmpdf2: Soft float library routines. 37674 (line 164) 37675 * __cmpdi2: Integer library routines. 37676 (line 87) 37677 * __cmpdq2: Fixed-point fractional library routines. 37678 (line 441) 37679 * __cmpha2: Fixed-point fractional library routines. 37680 (line 449) 37681 * __cmphq2: Fixed-point fractional library routines. 37682 (line 438) 37683 * __cmpqq2: Fixed-point fractional library routines. 37684 (line 437) 37685 * __cmpsa2: Fixed-point fractional library routines. 37686 (line 450) 37687 * __cmpsf2: Soft float library routines. 37688 (line 163) 37689 * __cmpsq2: Fixed-point fractional library routines. 37690 (line 439) 37691 * __cmpta2: Fixed-point fractional library routines. 37692 (line 453) 37693 * __cmptf2: Soft float library routines. 37694 (line 165) 37695 * __cmpti2: Integer library routines. 37696 (line 88) 37697 * __cmpuda2: Fixed-point fractional library routines. 37698 (line 458) 37699 * __cmpudq2: Fixed-point fractional library routines. 37700 (line 448) 37701 * __cmpuha2: Fixed-point fractional library routines. 37702 (line 455) 37703 * __cmpuhq2: Fixed-point fractional library routines. 37704 (line 444) 37705 * __cmpuqq2: Fixed-point fractional library routines. 37706 (line 443) 37707 * __cmpusa2: Fixed-point fractional library routines. 37708 (line 456) 37709 * __cmpusq2: Fixed-point fractional library routines. 37710 (line 446) 37711 * __cmputa2: Fixed-point fractional library routines. 37712 (line 460) 37713 * __CTOR_LIST__: Initialization. (line 25) 37714 * __ctzdi2: Integer library routines. 37715 (line 138) 37716 * __ctzsi2: Integer library routines. 37717 (line 137) 37718 * __ctzti2: Integer library routines. 37719 (line 139) 37720 * __divda3: Fixed-point fractional library routines. 37721 (line 227) 37722 * __divdc3: Soft float library routines. 37723 (line 252) 37724 * __divdf3: Soft float library routines. 37725 (line 48) 37726 * __divdi3: Integer library routines. 37727 (line 25) 37728 * __divdq3: Fixed-point fractional library routines. 37729 (line 223) 37730 * __divha3: Fixed-point fractional library routines. 37731 (line 225) 37732 * __divhq3: Fixed-point fractional library routines. 37733 (line 220) 37734 * __divqq3: Fixed-point fractional library routines. 37735 (line 219) 37736 * __divsa3: Fixed-point fractional library routines. 37737 (line 226) 37738 * __divsc3: Soft float library routines. 37739 (line 250) 37740 * __divsf3: Soft float library routines. 37741 (line 47) 37742 * __divsi3: Integer library routines. 37743 (line 24) 37744 * __divsq3: Fixed-point fractional library routines. 37745 (line 221) 37746 * __divta3: Fixed-point fractional library routines. 37747 (line 229) 37748 * __divtc3: Soft float library routines. 37749 (line 254) 37750 * __divtf3: Soft float library routines. 37751 (line 50) 37752 * __divti3: Integer library routines. 37753 (line 26) 37754 * __divxc3: Soft float library routines. 37755 (line 256) 37756 * __divxf3: Soft float library routines. 37757 (line 52) 37758 * __dpd_adddd3: Decimal float library routines. 37759 (line 23) 37760 * __dpd_addsd3: Decimal float library routines. 37761 (line 19) 37762 * __dpd_addtd3: Decimal float library routines. 37763 (line 27) 37764 * __dpd_divdd3: Decimal float library routines. 37765 (line 66) 37766 * __dpd_divsd3: Decimal float library routines. 37767 (line 62) 37768 * __dpd_divtd3: Decimal float library routines. 37769 (line 70) 37770 * __dpd_eqdd2: Decimal float library routines. 37771 (line 258) 37772 * __dpd_eqsd2: Decimal float library routines. 37773 (line 256) 37774 * __dpd_eqtd2: Decimal float library routines. 37775 (line 260) 37776 * __dpd_extendddtd2: Decimal float library routines. 37777 (line 91) 37778 * __dpd_extendddtf: Decimal float library routines. 37779 (line 139) 37780 * __dpd_extendddxf: Decimal float library routines. 37781 (line 133) 37782 * __dpd_extenddfdd: Decimal float library routines. 37783 (line 146) 37784 * __dpd_extenddftd: Decimal float library routines. 37785 (line 106) 37786 * __dpd_extendsddd2: Decimal float library routines. 37787 (line 87) 37788 * __dpd_extendsddf: Decimal float library routines. 37789 (line 127) 37790 * __dpd_extendsdtd2: Decimal float library routines. 37791 (line 89) 37792 * __dpd_extendsdtf: Decimal float library routines. 37793 (line 137) 37794 * __dpd_extendsdxf: Decimal float library routines. 37795 (line 131) 37796 * __dpd_extendsfdd: Decimal float library routines. 37797 (line 102) 37798 * __dpd_extendsfsd: Decimal float library routines. 37799 (line 144) 37800 * __dpd_extendsftd: Decimal float library routines. 37801 (line 104) 37802 * __dpd_extendtftd: Decimal float library routines. 37803 (line 148) 37804 * __dpd_extendxftd: Decimal float library routines. 37805 (line 108) 37806 * __dpd_fixdddi: Decimal float library routines. 37807 (line 169) 37808 * __dpd_fixddsi: Decimal float library routines. 37809 (line 161) 37810 * __dpd_fixsddi: Decimal float library routines. 37811 (line 167) 37812 * __dpd_fixsdsi: Decimal float library routines. 37813 (line 159) 37814 * __dpd_fixtddi: Decimal float library routines. 37815 (line 171) 37816 * __dpd_fixtdsi: Decimal float library routines. 37817 (line 163) 37818 * __dpd_fixunsdddi: Decimal float library routines. 37819 (line 186) 37820 * __dpd_fixunsddsi: Decimal float library routines. 37821 (line 177) 37822 * __dpd_fixunssddi: Decimal float library routines. 37823 (line 184) 37824 * __dpd_fixunssdsi: Decimal float library routines. 37825 (line 175) 37826 * __dpd_fixunstddi: Decimal float library routines. 37827 (line 188) 37828 * __dpd_fixunstdsi: Decimal float library routines. 37829 (line 179) 37830 * __dpd_floatdidd: Decimal float library routines. 37831 (line 204) 37832 * __dpd_floatdisd: Decimal float library routines. 37833 (line 202) 37834 * __dpd_floatditd: Decimal float library routines. 37835 (line 206) 37836 * __dpd_floatsidd: Decimal float library routines. 37837 (line 195) 37838 * __dpd_floatsisd: Decimal float library routines. 37839 (line 193) 37840 * __dpd_floatsitd: Decimal float library routines. 37841 (line 197) 37842 * __dpd_floatunsdidd: Decimal float library routines. 37843 (line 222) 37844 * __dpd_floatunsdisd: Decimal float library routines. 37845 (line 220) 37846 * __dpd_floatunsditd: Decimal float library routines. 37847 (line 224) 37848 * __dpd_floatunssidd: Decimal float library routines. 37849 (line 213) 37850 * __dpd_floatunssisd: Decimal float library routines. 37851 (line 211) 37852 * __dpd_floatunssitd: Decimal float library routines. 37853 (line 215) 37854 * __dpd_gedd2: Decimal float library routines. 37855 (line 276) 37856 * __dpd_gesd2: Decimal float library routines. 37857 (line 274) 37858 * __dpd_getd2: Decimal float library routines. 37859 (line 278) 37860 * __dpd_gtdd2: Decimal float library routines. 37861 (line 303) 37862 * __dpd_gtsd2: Decimal float library routines. 37863 (line 301) 37864 * __dpd_gttd2: Decimal float library routines. 37865 (line 305) 37866 * __dpd_ledd2: Decimal float library routines. 37867 (line 294) 37868 * __dpd_lesd2: Decimal float library routines. 37869 (line 292) 37870 * __dpd_letd2: Decimal float library routines. 37871 (line 296) 37872 * __dpd_ltdd2: Decimal float library routines. 37873 (line 285) 37874 * __dpd_ltsd2: Decimal float library routines. 37875 (line 283) 37876 * __dpd_lttd2: Decimal float library routines. 37877 (line 287) 37878 * __dpd_muldd3: Decimal float library routines. 37879 (line 52) 37880 * __dpd_mulsd3: Decimal float library routines. 37881 (line 48) 37882 * __dpd_multd3: Decimal float library routines. 37883 (line 56) 37884 * __dpd_nedd2: Decimal float library routines. 37885 (line 267) 37886 * __dpd_negdd2: Decimal float library routines. 37887 (line 77) 37888 * __dpd_negsd2: Decimal float library routines. 37889 (line 75) 37890 * __dpd_negtd2: Decimal float library routines. 37891 (line 79) 37892 * __dpd_nesd2: Decimal float library routines. 37893 (line 265) 37894 * __dpd_netd2: Decimal float library routines. 37895 (line 269) 37896 * __dpd_subdd3: Decimal float library routines. 37897 (line 37) 37898 * __dpd_subsd3: Decimal float library routines. 37899 (line 33) 37900 * __dpd_subtd3: Decimal float library routines. 37901 (line 41) 37902 * __dpd_truncdddf: Decimal float library routines. 37903 (line 152) 37904 * __dpd_truncddsd2: Decimal float library routines. 37905 (line 93) 37906 * __dpd_truncddsf: Decimal float library routines. 37907 (line 123) 37908 * __dpd_truncdfsd: Decimal float library routines. 37909 (line 110) 37910 * __dpd_truncsdsf: Decimal float library routines. 37911 (line 150) 37912 * __dpd_trunctddd2: Decimal float library routines. 37913 (line 97) 37914 * __dpd_trunctddf: Decimal float library routines. 37915 (line 129) 37916 * __dpd_trunctdsd2: Decimal float library routines. 37917 (line 95) 37918 * __dpd_trunctdsf: Decimal float library routines. 37919 (line 125) 37920 * __dpd_trunctdtf: Decimal float library routines. 37921 (line 154) 37922 * __dpd_trunctdxf: Decimal float library routines. 37923 (line 135) 37924 * __dpd_trunctfdd: Decimal float library routines. 37925 (line 118) 37926 * __dpd_trunctfsd: Decimal float library routines. 37927 (line 114) 37928 * __dpd_truncxfdd: Decimal float library routines. 37929 (line 116) 37930 * __dpd_truncxfsd: Decimal float library routines. 37931 (line 112) 37932 * __dpd_unorddd2: Decimal float library routines. 37933 (line 234) 37934 * __dpd_unordsd2: Decimal float library routines. 37935 (line 232) 37936 * __dpd_unordtd2: Decimal float library routines. 37937 (line 236) 37938 * __DTOR_LIST__: Initialization. (line 25) 37939 * __eqdf2: Soft float library routines. 37940 (line 194) 37941 * __eqsf2: Soft float library routines. 37942 (line 193) 37943 * __eqtf2: Soft float library routines. 37944 (line 195) 37945 * __extenddftf2: Soft float library routines. 37946 (line 68) 37947 * __extenddfxf2: Soft float library routines. 37948 (line 69) 37949 * __extendsfdf2: Soft float library routines. 37950 (line 65) 37951 * __extendsftf2: Soft float library routines. 37952 (line 66) 37953 * __extendsfxf2: Soft float library routines. 37954 (line 67) 37955 * __ffsdi2: Integer library routines. 37956 (line 144) 37957 * __ffsti2: Integer library routines. 37958 (line 145) 37959 * __fixdfdi: Soft float library routines. 37960 (line 88) 37961 * __fixdfsi: Soft float library routines. 37962 (line 81) 37963 * __fixdfti: Soft float library routines. 37964 (line 94) 37965 * __fixsfdi: Soft float library routines. 37966 (line 87) 37967 * __fixsfsi: Soft float library routines. 37968 (line 80) 37969 * __fixsfti: Soft float library routines. 37970 (line 93) 37971 * __fixtfdi: Soft float library routines. 37972 (line 89) 37973 * __fixtfsi: Soft float library routines. 37974 (line 82) 37975 * __fixtfti: Soft float library routines. 37976 (line 95) 37977 * __fixunsdfdi: Soft float library routines. 37978 (line 108) 37979 * __fixunsdfsi: Soft float library routines. 37980 (line 101) 37981 * __fixunsdfti: Soft float library routines. 37982 (line 115) 37983 * __fixunssfdi: Soft float library routines. 37984 (line 107) 37985 * __fixunssfsi: Soft float library routines. 37986 (line 100) 37987 * __fixunssfti: Soft float library routines. 37988 (line 114) 37989 * __fixunstfdi: Soft float library routines. 37990 (line 109) 37991 * __fixunstfsi: Soft float library routines. 37992 (line 102) 37993 * __fixunstfti: Soft float library routines. 37994 (line 116) 37995 * __fixunsxfdi: Soft float library routines. 37996 (line 110) 37997 * __fixunsxfsi: Soft float library routines. 37998 (line 103) 37999 * __fixunsxfti: Soft float library routines. 38000 (line 117) 38001 * __fixxfdi: Soft float library routines. 38002 (line 90) 38003 * __fixxfsi: Soft float library routines. 38004 (line 83) 38005 * __fixxfti: Soft float library routines. 38006 (line 96) 38007 * __floatdidf: Soft float library routines. 38008 (line 128) 38009 * __floatdisf: Soft float library routines. 38010 (line 127) 38011 * __floatditf: Soft float library routines. 38012 (line 129) 38013 * __floatdixf: Soft float library routines. 38014 (line 130) 38015 * __floatsidf: Soft float library routines. 38016 (line 122) 38017 * __floatsisf: Soft float library routines. 38018 (line 121) 38019 * __floatsitf: Soft float library routines. 38020 (line 123) 38021 * __floatsixf: Soft float library routines. 38022 (line 124) 38023 * __floattidf: Soft float library routines. 38024 (line 134) 38025 * __floattisf: Soft float library routines. 38026 (line 133) 38027 * __floattitf: Soft float library routines. 38028 (line 135) 38029 * __floattixf: Soft float library routines. 38030 (line 136) 38031 * __floatundidf: Soft float library routines. 38032 (line 146) 38033 * __floatundisf: Soft float library routines. 38034 (line 145) 38035 * __floatunditf: Soft float library routines. 38036 (line 147) 38037 * __floatundixf: Soft float library routines. 38038 (line 148) 38039 * __floatunsidf: Soft float library routines. 38040 (line 140) 38041 * __floatunsisf: Soft float library routines. 38042 (line 139) 38043 * __floatunsitf: Soft float library routines. 38044 (line 141) 38045 * __floatunsixf: Soft float library routines. 38046 (line 142) 38047 * __floatuntidf: Soft float library routines. 38048 (line 152) 38049 * __floatuntisf: Soft float library routines. 38050 (line 151) 38051 * __floatuntitf: Soft float library routines. 38052 (line 153) 38053 * __floatuntixf: Soft float library routines. 38054 (line 154) 38055 * __fractdadf: Fixed-point fractional library routines. 38056 (line 636) 38057 * __fractdadi: Fixed-point fractional library routines. 38058 (line 633) 38059 * __fractdadq: Fixed-point fractional library routines. 38060 (line 616) 38061 * __fractdaha2: Fixed-point fractional library routines. 38062 (line 617) 38063 * __fractdahi: Fixed-point fractional library routines. 38064 (line 631) 38065 * __fractdahq: Fixed-point fractional library routines. 38066 (line 614) 38067 * __fractdaqi: Fixed-point fractional library routines. 38068 (line 630) 38069 * __fractdaqq: Fixed-point fractional library routines. 38070 (line 613) 38071 * __fractdasa2: Fixed-point fractional library routines. 38072 (line 618) 38073 * __fractdasf: Fixed-point fractional library routines. 38074 (line 635) 38075 * __fractdasi: Fixed-point fractional library routines. 38076 (line 632) 38077 * __fractdasq: Fixed-point fractional library routines. 38078 (line 615) 38079 * __fractdata2: Fixed-point fractional library routines. 38080 (line 619) 38081 * __fractdati: Fixed-point fractional library routines. 38082 (line 634) 38083 * __fractdauda: Fixed-point fractional library routines. 38084 (line 627) 38085 * __fractdaudq: Fixed-point fractional library routines. 38086 (line 624) 38087 * __fractdauha: Fixed-point fractional library routines. 38088 (line 625) 38089 * __fractdauhq: Fixed-point fractional library routines. 38090 (line 621) 38091 * __fractdauqq: Fixed-point fractional library routines. 38092 (line 620) 38093 * __fractdausa: Fixed-point fractional library routines. 38094 (line 626) 38095 * __fractdausq: Fixed-point fractional library routines. 38096 (line 622) 38097 * __fractdauta: Fixed-point fractional library routines. 38098 (line 629) 38099 * __fractdfda: Fixed-point fractional library routines. 38100 (line 1025) 38101 * __fractdfdq: Fixed-point fractional library routines. 38102 (line 1022) 38103 * __fractdfha: Fixed-point fractional library routines. 38104 (line 1023) 38105 * __fractdfhq: Fixed-point fractional library routines. 38106 (line 1020) 38107 * __fractdfqq: Fixed-point fractional library routines. 38108 (line 1019) 38109 * __fractdfsa: Fixed-point fractional library routines. 38110 (line 1024) 38111 * __fractdfsq: Fixed-point fractional library routines. 38112 (line 1021) 38113 * __fractdfta: Fixed-point fractional library routines. 38114 (line 1026) 38115 * __fractdfuda: Fixed-point fractional library routines. 38116 (line 1033) 38117 * __fractdfudq: Fixed-point fractional library routines. 38118 (line 1030) 38119 * __fractdfuha: Fixed-point fractional library routines. 38120 (line 1031) 38121 * __fractdfuhq: Fixed-point fractional library routines. 38122 (line 1028) 38123 * __fractdfuqq: Fixed-point fractional library routines. 38124 (line 1027) 38125 * __fractdfusa: Fixed-point fractional library routines. 38126 (line 1032) 38127 * __fractdfusq: Fixed-point fractional library routines. 38128 (line 1029) 38129 * __fractdfuta: Fixed-point fractional library routines. 38130 (line 1034) 38131 * __fractdida: Fixed-point fractional library routines. 38132 (line 975) 38133 * __fractdidq: Fixed-point fractional library routines. 38134 (line 972) 38135 * __fractdiha: Fixed-point fractional library routines. 38136 (line 973) 38137 * __fractdihq: Fixed-point fractional library routines. 38138 (line 970) 38139 * __fractdiqq: Fixed-point fractional library routines. 38140 (line 969) 38141 * __fractdisa: Fixed-point fractional library routines. 38142 (line 974) 38143 * __fractdisq: Fixed-point fractional library routines. 38144 (line 971) 38145 * __fractdita: Fixed-point fractional library routines. 38146 (line 976) 38147 * __fractdiuda: Fixed-point fractional library routines. 38148 (line 983) 38149 * __fractdiudq: Fixed-point fractional library routines. 38150 (line 980) 38151 * __fractdiuha: Fixed-point fractional library routines. 38152 (line 981) 38153 * __fractdiuhq: Fixed-point fractional library routines. 38154 (line 978) 38155 * __fractdiuqq: Fixed-point fractional library routines. 38156 (line 977) 38157 * __fractdiusa: Fixed-point fractional library routines. 38158 (line 982) 38159 * __fractdiusq: Fixed-point fractional library routines. 38160 (line 979) 38161 * __fractdiuta: Fixed-point fractional library routines. 38162 (line 984) 38163 * __fractdqda: Fixed-point fractional library routines. 38164 (line 544) 38165 * __fractdqdf: Fixed-point fractional library routines. 38166 (line 566) 38167 * __fractdqdi: Fixed-point fractional library routines. 38168 (line 563) 38169 * __fractdqha: Fixed-point fractional library routines. 38170 (line 542) 38171 * __fractdqhi: Fixed-point fractional library routines. 38172 (line 561) 38173 * __fractdqhq2: Fixed-point fractional library routines. 38174 (line 540) 38175 * __fractdqqi: Fixed-point fractional library routines. 38176 (line 560) 38177 * __fractdqqq2: Fixed-point fractional library routines. 38178 (line 539) 38179 * __fractdqsa: Fixed-point fractional library routines. 38180 (line 543) 38181 * __fractdqsf: Fixed-point fractional library routines. 38182 (line 565) 38183 * __fractdqsi: Fixed-point fractional library routines. 38184 (line 562) 38185 * __fractdqsq2: Fixed-point fractional library routines. 38186 (line 541) 38187 * __fractdqta: Fixed-point fractional library routines. 38188 (line 545) 38189 * __fractdqti: Fixed-point fractional library routines. 38190 (line 564) 38191 * __fractdquda: Fixed-point fractional library routines. 38192 (line 557) 38193 * __fractdqudq: Fixed-point fractional library routines. 38194 (line 552) 38195 * __fractdquha: Fixed-point fractional library routines. 38196 (line 554) 38197 * __fractdquhq: Fixed-point fractional library routines. 38198 (line 548) 38199 * __fractdquqq: Fixed-point fractional library routines. 38200 (line 547) 38201 * __fractdqusa: Fixed-point fractional library routines. 38202 (line 555) 38203 * __fractdqusq: Fixed-point fractional library routines. 38204 (line 550) 38205 * __fractdquta: Fixed-point fractional library routines. 38206 (line 559) 38207 * __fracthada2: Fixed-point fractional library routines. 38208 (line 572) 38209 * __fracthadf: Fixed-point fractional library routines. 38210 (line 590) 38211 * __fracthadi: Fixed-point fractional library routines. 38212 (line 587) 38213 * __fracthadq: Fixed-point fractional library routines. 38214 (line 570) 38215 * __fracthahi: Fixed-point fractional library routines. 38216 (line 585) 38217 * __fracthahq: Fixed-point fractional library routines. 38218 (line 568) 38219 * __fracthaqi: Fixed-point fractional library routines. 38220 (line 584) 38221 * __fracthaqq: Fixed-point fractional library routines. 38222 (line 567) 38223 * __fracthasa2: Fixed-point fractional library routines. 38224 (line 571) 38225 * __fracthasf: Fixed-point fractional library routines. 38226 (line 589) 38227 * __fracthasi: Fixed-point fractional library routines. 38228 (line 586) 38229 * __fracthasq: Fixed-point fractional library routines. 38230 (line 569) 38231 * __fracthata2: Fixed-point fractional library routines. 38232 (line 573) 38233 * __fracthati: Fixed-point fractional library routines. 38234 (line 588) 38235 * __fracthauda: Fixed-point fractional library routines. 38236 (line 581) 38237 * __fracthaudq: Fixed-point fractional library routines. 38238 (line 578) 38239 * __fracthauha: Fixed-point fractional library routines. 38240 (line 579) 38241 * __fracthauhq: Fixed-point fractional library routines. 38242 (line 575) 38243 * __fracthauqq: Fixed-point fractional library routines. 38244 (line 574) 38245 * __fracthausa: Fixed-point fractional library routines. 38246 (line 580) 38247 * __fracthausq: Fixed-point fractional library routines. 38248 (line 576) 38249 * __fracthauta: Fixed-point fractional library routines. 38250 (line 583) 38251 * __fracthida: Fixed-point fractional library routines. 38252 (line 943) 38253 * __fracthidq: Fixed-point fractional library routines. 38254 (line 940) 38255 * __fracthiha: Fixed-point fractional library routines. 38256 (line 941) 38257 * __fracthihq: Fixed-point fractional library routines. 38258 (line 938) 38259 * __fracthiqq: Fixed-point fractional library routines. 38260 (line 937) 38261 * __fracthisa: Fixed-point fractional library routines. 38262 (line 942) 38263 * __fracthisq: Fixed-point fractional library routines. 38264 (line 939) 38265 * __fracthita: Fixed-point fractional library routines. 38266 (line 944) 38267 * __fracthiuda: Fixed-point fractional library routines. 38268 (line 951) 38269 * __fracthiudq: Fixed-point fractional library routines. 38270 (line 948) 38271 * __fracthiuha: Fixed-point fractional library routines. 38272 (line 949) 38273 * __fracthiuhq: Fixed-point fractional library routines. 38274 (line 946) 38275 * __fracthiuqq: Fixed-point fractional library routines. 38276 (line 945) 38277 * __fracthiusa: Fixed-point fractional library routines. 38278 (line 950) 38279 * __fracthiusq: Fixed-point fractional library routines. 38280 (line 947) 38281 * __fracthiuta: Fixed-point fractional library routines. 38282 (line 952) 38283 * __fracthqda: Fixed-point fractional library routines. 38284 (line 498) 38285 * __fracthqdf: Fixed-point fractional library routines. 38286 (line 514) 38287 * __fracthqdi: Fixed-point fractional library routines. 38288 (line 511) 38289 * __fracthqdq2: Fixed-point fractional library routines. 38290 (line 495) 38291 * __fracthqha: Fixed-point fractional library routines. 38292 (line 496) 38293 * __fracthqhi: Fixed-point fractional library routines. 38294 (line 509) 38295 * __fracthqqi: Fixed-point fractional library routines. 38296 (line 508) 38297 * __fracthqqq2: Fixed-point fractional library routines. 38298 (line 493) 38299 * __fracthqsa: Fixed-point fractional library routines. 38300 (line 497) 38301 * __fracthqsf: Fixed-point fractional library routines. 38302 (line 513) 38303 * __fracthqsi: Fixed-point fractional library routines. 38304 (line 510) 38305 * __fracthqsq2: Fixed-point fractional library routines. 38306 (line 494) 38307 * __fracthqta: Fixed-point fractional library routines. 38308 (line 499) 38309 * __fracthqti: Fixed-point fractional library routines. 38310 (line 512) 38311 * __fracthquda: Fixed-point fractional library routines. 38312 (line 506) 38313 * __fracthqudq: Fixed-point fractional library routines. 38314 (line 503) 38315 * __fracthquha: Fixed-point fractional library routines. 38316 (line 504) 38317 * __fracthquhq: Fixed-point fractional library routines. 38318 (line 501) 38319 * __fracthquqq: Fixed-point fractional library routines. 38320 (line 500) 38321 * __fracthqusa: Fixed-point fractional library routines. 38322 (line 505) 38323 * __fracthqusq: Fixed-point fractional library routines. 38324 (line 502) 38325 * __fracthquta: Fixed-point fractional library routines. 38326 (line 507) 38327 * __fractqida: Fixed-point fractional library routines. 38328 (line 925) 38329 * __fractqidq: Fixed-point fractional library routines. 38330 (line 922) 38331 * __fractqiha: Fixed-point fractional library routines. 38332 (line 923) 38333 * __fractqihq: Fixed-point fractional library routines. 38334 (line 920) 38335 * __fractqiqq: Fixed-point fractional library routines. 38336 (line 919) 38337 * __fractqisa: Fixed-point fractional library routines. 38338 (line 924) 38339 * __fractqisq: Fixed-point fractional library routines. 38340 (line 921) 38341 * __fractqita: Fixed-point fractional library routines. 38342 (line 926) 38343 * __fractqiuda: Fixed-point fractional library routines. 38344 (line 934) 38345 * __fractqiudq: Fixed-point fractional library routines. 38346 (line 931) 38347 * __fractqiuha: Fixed-point fractional library routines. 38348 (line 932) 38349 * __fractqiuhq: Fixed-point fractional library routines. 38350 (line 928) 38351 * __fractqiuqq: Fixed-point fractional library routines. 38352 (line 927) 38353 * __fractqiusa: Fixed-point fractional library routines. 38354 (line 933) 38355 * __fractqiusq: Fixed-point fractional library routines. 38356 (line 929) 38357 * __fractqiuta: Fixed-point fractional library routines. 38358 (line 936) 38359 * __fractqqda: Fixed-point fractional library routines. 38360 (line 474) 38361 * __fractqqdf: Fixed-point fractional library routines. 38362 (line 492) 38363 * __fractqqdi: Fixed-point fractional library routines. 38364 (line 489) 38365 * __fractqqdq2: Fixed-point fractional library routines. 38366 (line 471) 38367 * __fractqqha: Fixed-point fractional library routines. 38368 (line 472) 38369 * __fractqqhi: Fixed-point fractional library routines. 38370 (line 487) 38371 * __fractqqhq2: Fixed-point fractional library routines. 38372 (line 469) 38373 * __fractqqqi: Fixed-point fractional library routines. 38374 (line 486) 38375 * __fractqqsa: Fixed-point fractional library routines. 38376 (line 473) 38377 * __fractqqsf: Fixed-point fractional library routines. 38378 (line 491) 38379 * __fractqqsi: Fixed-point fractional library routines. 38380 (line 488) 38381 * __fractqqsq2: Fixed-point fractional library routines. 38382 (line 470) 38383 * __fractqqta: Fixed-point fractional library routines. 38384 (line 475) 38385 * __fractqqti: Fixed-point fractional library routines. 38386 (line 490) 38387 * __fractqquda: Fixed-point fractional library routines. 38388 (line 483) 38389 * __fractqqudq: Fixed-point fractional library routines. 38390 (line 480) 38391 * __fractqquha: Fixed-point fractional library routines. 38392 (line 481) 38393 * __fractqquhq: Fixed-point fractional library routines. 38394 (line 477) 38395 * __fractqquqq: Fixed-point fractional library routines. 38396 (line 476) 38397 * __fractqqusa: Fixed-point fractional library routines. 38398 (line 482) 38399 * __fractqqusq: Fixed-point fractional library routines. 38400 (line 478) 38401 * __fractqquta: Fixed-point fractional library routines. 38402 (line 485) 38403 * __fractsada2: Fixed-point fractional library routines. 38404 (line 596) 38405 * __fractsadf: Fixed-point fractional library routines. 38406 (line 612) 38407 * __fractsadi: Fixed-point fractional library routines. 38408 (line 609) 38409 * __fractsadq: Fixed-point fractional library routines. 38410 (line 594) 38411 * __fractsaha2: Fixed-point fractional library routines. 38412 (line 595) 38413 * __fractsahi: Fixed-point fractional library routines. 38414 (line 607) 38415 * __fractsahq: Fixed-point fractional library routines. 38416 (line 592) 38417 * __fractsaqi: Fixed-point fractional library routines. 38418 (line 606) 38419 * __fractsaqq: Fixed-point fractional library routines. 38420 (line 591) 38421 * __fractsasf: Fixed-point fractional library routines. 38422 (line 611) 38423 * __fractsasi: Fixed-point fractional library routines. 38424 (line 608) 38425 * __fractsasq: Fixed-point fractional library routines. 38426 (line 593) 38427 * __fractsata2: Fixed-point fractional library routines. 38428 (line 597) 38429 * __fractsati: Fixed-point fractional library routines. 38430 (line 610) 38431 * __fractsauda: Fixed-point fractional library routines. 38432 (line 604) 38433 * __fractsaudq: Fixed-point fractional library routines. 38434 (line 601) 38435 * __fractsauha: Fixed-point fractional library routines. 38436 (line 602) 38437 * __fractsauhq: Fixed-point fractional library routines. 38438 (line 599) 38439 * __fractsauqq: Fixed-point fractional library routines. 38440 (line 598) 38441 * __fractsausa: Fixed-point fractional library routines. 38442 (line 603) 38443 * __fractsausq: Fixed-point fractional library routines. 38444 (line 600) 38445 * __fractsauta: Fixed-point fractional library routines. 38446 (line 605) 38447 * __fractsfda: Fixed-point fractional library routines. 38448 (line 1009) 38449 * __fractsfdq: Fixed-point fractional library routines. 38450 (line 1006) 38451 * __fractsfha: Fixed-point fractional library routines. 38452 (line 1007) 38453 * __fractsfhq: Fixed-point fractional library routines. 38454 (line 1004) 38455 * __fractsfqq: Fixed-point fractional library routines. 38456 (line 1003) 38457 * __fractsfsa: Fixed-point fractional library routines. 38458 (line 1008) 38459 * __fractsfsq: Fixed-point fractional library routines. 38460 (line 1005) 38461 * __fractsfta: Fixed-point fractional library routines. 38462 (line 1010) 38463 * __fractsfuda: Fixed-point fractional library routines. 38464 (line 1017) 38465 * __fractsfudq: Fixed-point fractional library routines. 38466 (line 1014) 38467 * __fractsfuha: Fixed-point fractional library routines. 38468 (line 1015) 38469 * __fractsfuhq: Fixed-point fractional library routines. 38470 (line 1012) 38471 * __fractsfuqq: Fixed-point fractional library routines. 38472 (line 1011) 38473 * __fractsfusa: Fixed-point fractional library routines. 38474 (line 1016) 38475 * __fractsfusq: Fixed-point fractional library routines. 38476 (line 1013) 38477 * __fractsfuta: Fixed-point fractional library routines. 38478 (line 1018) 38479 * __fractsida: Fixed-point fractional library routines. 38480 (line 959) 38481 * __fractsidq: Fixed-point fractional library routines. 38482 (line 956) 38483 * __fractsiha: Fixed-point fractional library routines. 38484 (line 957) 38485 * __fractsihq: Fixed-point fractional library routines. 38486 (line 954) 38487 * __fractsiqq: Fixed-point fractional library routines. 38488 (line 953) 38489 * __fractsisa: Fixed-point fractional library routines. 38490 (line 958) 38491 * __fractsisq: Fixed-point fractional library routines. 38492 (line 955) 38493 * __fractsita: Fixed-point fractional library routines. 38494 (line 960) 38495 * __fractsiuda: Fixed-point fractional library routines. 38496 (line 967) 38497 * __fractsiudq: Fixed-point fractional library routines. 38498 (line 964) 38499 * __fractsiuha: Fixed-point fractional library routines. 38500 (line 965) 38501 * __fractsiuhq: Fixed-point fractional library routines. 38502 (line 962) 38503 * __fractsiuqq: Fixed-point fractional library routines. 38504 (line 961) 38505 * __fractsiusa: Fixed-point fractional library routines. 38506 (line 966) 38507 * __fractsiusq: Fixed-point fractional library routines. 38508 (line 963) 38509 * __fractsiuta: Fixed-point fractional library routines. 38510 (line 968) 38511 * __fractsqda: Fixed-point fractional library routines. 38512 (line 520) 38513 * __fractsqdf: Fixed-point fractional library routines. 38514 (line 538) 38515 * __fractsqdi: Fixed-point fractional library routines. 38516 (line 535) 38517 * __fractsqdq2: Fixed-point fractional library routines. 38518 (line 517) 38519 * __fractsqha: Fixed-point fractional library routines. 38520 (line 518) 38521 * __fractsqhi: Fixed-point fractional library routines. 38522 (line 533) 38523 * __fractsqhq2: Fixed-point fractional library routines. 38524 (line 516) 38525 * __fractsqqi: Fixed-point fractional library routines. 38526 (line 532) 38527 * __fractsqqq2: Fixed-point fractional library routines. 38528 (line 515) 38529 * __fractsqsa: Fixed-point fractional library routines. 38530 (line 519) 38531 * __fractsqsf: Fixed-point fractional library routines. 38532 (line 537) 38533 * __fractsqsi: Fixed-point fractional library routines. 38534 (line 534) 38535 * __fractsqta: Fixed-point fractional library routines. 38536 (line 521) 38537 * __fractsqti: Fixed-point fractional library routines. 38538 (line 536) 38539 * __fractsquda: Fixed-point fractional library routines. 38540 (line 529) 38541 * __fractsqudq: Fixed-point fractional library routines. 38542 (line 526) 38543 * __fractsquha: Fixed-point fractional library routines. 38544 (line 527) 38545 * __fractsquhq: Fixed-point fractional library routines. 38546 (line 523) 38547 * __fractsquqq: Fixed-point fractional library routines. 38548 (line 522) 38549 * __fractsqusa: Fixed-point fractional library routines. 38550 (line 528) 38551 * __fractsqusq: Fixed-point fractional library routines. 38552 (line 524) 38553 * __fractsquta: Fixed-point fractional library routines. 38554 (line 531) 38555 * __fracttada2: Fixed-point fractional library routines. 38556 (line 643) 38557 * __fracttadf: Fixed-point fractional library routines. 38558 (line 664) 38559 * __fracttadi: Fixed-point fractional library routines. 38560 (line 661) 38561 * __fracttadq: Fixed-point fractional library routines. 38562 (line 640) 38563 * __fracttaha2: Fixed-point fractional library routines. 38564 (line 641) 38565 * __fracttahi: Fixed-point fractional library routines. 38566 (line 659) 38567 * __fracttahq: Fixed-point fractional library routines. 38568 (line 638) 38569 * __fracttaqi: Fixed-point fractional library routines. 38570 (line 658) 38571 * __fracttaqq: Fixed-point fractional library routines. 38572 (line 637) 38573 * __fracttasa2: Fixed-point fractional library routines. 38574 (line 642) 38575 * __fracttasf: Fixed-point fractional library routines. 38576 (line 663) 38577 * __fracttasi: Fixed-point fractional library routines. 38578 (line 660) 38579 * __fracttasq: Fixed-point fractional library routines. 38580 (line 639) 38581 * __fracttati: Fixed-point fractional library routines. 38582 (line 662) 38583 * __fracttauda: Fixed-point fractional library routines. 38584 (line 655) 38585 * __fracttaudq: Fixed-point fractional library routines. 38586 (line 650) 38587 * __fracttauha: Fixed-point fractional library routines. 38588 (line 652) 38589 * __fracttauhq: Fixed-point fractional library routines. 38590 (line 646) 38591 * __fracttauqq: Fixed-point fractional library routines. 38592 (line 645) 38593 * __fracttausa: Fixed-point fractional library routines. 38594 (line 653) 38595 * __fracttausq: Fixed-point fractional library routines. 38596 (line 648) 38597 * __fracttauta: Fixed-point fractional library routines. 38598 (line 657) 38599 * __fracttida: Fixed-point fractional library routines. 38600 (line 991) 38601 * __fracttidq: Fixed-point fractional library routines. 38602 (line 988) 38603 * __fracttiha: Fixed-point fractional library routines. 38604 (line 989) 38605 * __fracttihq: Fixed-point fractional library routines. 38606 (line 986) 38607 * __fracttiqq: Fixed-point fractional library routines. 38608 (line 985) 38609 * __fracttisa: Fixed-point fractional library routines. 38610 (line 990) 38611 * __fracttisq: Fixed-point fractional library routines. 38612 (line 987) 38613 * __fracttita: Fixed-point fractional library routines. 38614 (line 992) 38615 * __fracttiuda: Fixed-point fractional library routines. 38616 (line 1000) 38617 * __fracttiudq: Fixed-point fractional library routines. 38618 (line 997) 38619 * __fracttiuha: Fixed-point fractional library routines. 38620 (line 998) 38621 * __fracttiuhq: Fixed-point fractional library routines. 38622 (line 994) 38623 * __fracttiuqq: Fixed-point fractional library routines. 38624 (line 993) 38625 * __fracttiusa: Fixed-point fractional library routines. 38626 (line 999) 38627 * __fracttiusq: Fixed-point fractional library routines. 38628 (line 995) 38629 * __fracttiuta: Fixed-point fractional library routines. 38630 (line 1002) 38631 * __fractudada: Fixed-point fractional library routines. 38632 (line 858) 38633 * __fractudadf: Fixed-point fractional library routines. 38634 (line 881) 38635 * __fractudadi: Fixed-point fractional library routines. 38636 (line 878) 38637 * __fractudadq: Fixed-point fractional library routines. 38638 (line 855) 38639 * __fractudaha: Fixed-point fractional library routines. 38640 (line 856) 38641 * __fractudahi: Fixed-point fractional library routines. 38642 (line 876) 38643 * __fractudahq: Fixed-point fractional library routines. 38644 (line 852) 38645 * __fractudaqi: Fixed-point fractional library routines. 38646 (line 875) 38647 * __fractudaqq: Fixed-point fractional library routines. 38648 (line 851) 38649 * __fractudasa: Fixed-point fractional library routines. 38650 (line 857) 38651 * __fractudasf: Fixed-point fractional library routines. 38652 (line 880) 38653 * __fractudasi: Fixed-point fractional library routines. 38654 (line 877) 38655 * __fractudasq: Fixed-point fractional library routines. 38656 (line 853) 38657 * __fractudata: Fixed-point fractional library routines. 38658 (line 860) 38659 * __fractudati: Fixed-point fractional library routines. 38660 (line 879) 38661 * __fractudaudq: Fixed-point fractional library routines. 38662 (line 868) 38663 * __fractudauha2: Fixed-point fractional library routines. 38664 (line 870) 38665 * __fractudauhq: Fixed-point fractional library routines. 38666 (line 864) 38667 * __fractudauqq: Fixed-point fractional library routines. 38668 (line 862) 38669 * __fractudausa2: Fixed-point fractional library routines. 38670 (line 872) 38671 * __fractudausq: Fixed-point fractional library routines. 38672 (line 866) 38673 * __fractudauta2: Fixed-point fractional library routines. 38674 (line 874) 38675 * __fractudqda: Fixed-point fractional library routines. 38676 (line 766) 38677 * __fractudqdf: Fixed-point fractional library routines. 38678 (line 791) 38679 * __fractudqdi: Fixed-point fractional library routines. 38680 (line 787) 38681 * __fractudqdq: Fixed-point fractional library routines. 38682 (line 761) 38683 * __fractudqha: Fixed-point fractional library routines. 38684 (line 763) 38685 * __fractudqhi: Fixed-point fractional library routines. 38686 (line 785) 38687 * __fractudqhq: Fixed-point fractional library routines. 38688 (line 757) 38689 * __fractudqqi: Fixed-point fractional library routines. 38690 (line 784) 38691 * __fractudqqq: Fixed-point fractional library routines. 38692 (line 756) 38693 * __fractudqsa: Fixed-point fractional library routines. 38694 (line 764) 38695 * __fractudqsf: Fixed-point fractional library routines. 38696 (line 790) 38697 * __fractudqsi: Fixed-point fractional library routines. 38698 (line 786) 38699 * __fractudqsq: Fixed-point fractional library routines. 38700 (line 759) 38701 * __fractudqta: Fixed-point fractional library routines. 38702 (line 768) 38703 * __fractudqti: Fixed-point fractional library routines. 38704 (line 789) 38705 * __fractudquda: Fixed-point fractional library routines. 38706 (line 780) 38707 * __fractudquha: Fixed-point fractional library routines. 38708 (line 776) 38709 * __fractudquhq2: Fixed-point fractional library routines. 38710 (line 772) 38711 * __fractudquqq2: Fixed-point fractional library routines. 38712 (line 770) 38713 * __fractudqusa: Fixed-point fractional library routines. 38714 (line 778) 38715 * __fractudqusq2: Fixed-point fractional library routines. 38716 (line 774) 38717 * __fractudquta: Fixed-point fractional library routines. 38718 (line 782) 38719 * __fractuhada: Fixed-point fractional library routines. 38720 (line 799) 38721 * __fractuhadf: Fixed-point fractional library routines. 38722 (line 822) 38723 * __fractuhadi: Fixed-point fractional library routines. 38724 (line 819) 38725 * __fractuhadq: Fixed-point fractional library routines. 38726 (line 796) 38727 * __fractuhaha: Fixed-point fractional library routines. 38728 (line 797) 38729 * __fractuhahi: Fixed-point fractional library routines. 38730 (line 817) 38731 * __fractuhahq: Fixed-point fractional library routines. 38732 (line 793) 38733 * __fractuhaqi: Fixed-point fractional library routines. 38734 (line 816) 38735 * __fractuhaqq: Fixed-point fractional library routines. 38736 (line 792) 38737 * __fractuhasa: Fixed-point fractional library routines. 38738 (line 798) 38739 * __fractuhasf: Fixed-point fractional library routines. 38740 (line 821) 38741 * __fractuhasi: Fixed-point fractional library routines. 38742 (line 818) 38743 * __fractuhasq: Fixed-point fractional library routines. 38744 (line 794) 38745 * __fractuhata: Fixed-point fractional library routines. 38746 (line 801) 38747 * __fractuhati: Fixed-point fractional library routines. 38748 (line 820) 38749 * __fractuhauda2: Fixed-point fractional library routines. 38750 (line 813) 38751 * __fractuhaudq: Fixed-point fractional library routines. 38752 (line 809) 38753 * __fractuhauhq: Fixed-point fractional library routines. 38754 (line 805) 38755 * __fractuhauqq: Fixed-point fractional library routines. 38756 (line 803) 38757 * __fractuhausa2: Fixed-point fractional library routines. 38758 (line 811) 38759 * __fractuhausq: Fixed-point fractional library routines. 38760 (line 807) 38761 * __fractuhauta2: Fixed-point fractional library routines. 38762 (line 815) 38763 * __fractuhqda: Fixed-point fractional library routines. 38764 (line 702) 38765 * __fractuhqdf: Fixed-point fractional library routines. 38766 (line 723) 38767 * __fractuhqdi: Fixed-point fractional library routines. 38768 (line 720) 38769 * __fractuhqdq: Fixed-point fractional library routines. 38770 (line 699) 38771 * __fractuhqha: Fixed-point fractional library routines. 38772 (line 700) 38773 * __fractuhqhi: Fixed-point fractional library routines. 38774 (line 718) 38775 * __fractuhqhq: Fixed-point fractional library routines. 38776 (line 697) 38777 * __fractuhqqi: Fixed-point fractional library routines. 38778 (line 717) 38779 * __fractuhqqq: Fixed-point fractional library routines. 38780 (line 696) 38781 * __fractuhqsa: Fixed-point fractional library routines. 38782 (line 701) 38783 * __fractuhqsf: Fixed-point fractional library routines. 38784 (line 722) 38785 * __fractuhqsi: Fixed-point fractional library routines. 38786 (line 719) 38787 * __fractuhqsq: Fixed-point fractional library routines. 38788 (line 698) 38789 * __fractuhqta: Fixed-point fractional library routines. 38790 (line 703) 38791 * __fractuhqti: Fixed-point fractional library routines. 38792 (line 721) 38793 * __fractuhquda: Fixed-point fractional library routines. 38794 (line 714) 38795 * __fractuhqudq2: Fixed-point fractional library routines. 38796 (line 709) 38797 * __fractuhquha: Fixed-point fractional library routines. 38798 (line 711) 38799 * __fractuhquqq2: Fixed-point fractional library routines. 38800 (line 705) 38801 * __fractuhqusa: Fixed-point fractional library routines. 38802 (line 712) 38803 * __fractuhqusq2: Fixed-point fractional library routines. 38804 (line 707) 38805 * __fractuhquta: Fixed-point fractional library routines. 38806 (line 716) 38807 * __fractunsdadi: Fixed-point fractional library routines. 38808 (line 1555) 38809 * __fractunsdahi: Fixed-point fractional library routines. 38810 (line 1553) 38811 * __fractunsdaqi: Fixed-point fractional library routines. 38812 (line 1552) 38813 * __fractunsdasi: Fixed-point fractional library routines. 38814 (line 1554) 38815 * __fractunsdati: Fixed-point fractional library routines. 38816 (line 1556) 38817 * __fractunsdida: Fixed-point fractional library routines. 38818 (line 1707) 38819 * __fractunsdidq: Fixed-point fractional library routines. 38820 (line 1704) 38821 * __fractunsdiha: Fixed-point fractional library routines. 38822 (line 1705) 38823 * __fractunsdihq: Fixed-point fractional library routines. 38824 (line 1702) 38825 * __fractunsdiqq: Fixed-point fractional library routines. 38826 (line 1701) 38827 * __fractunsdisa: Fixed-point fractional library routines. 38828 (line 1706) 38829 * __fractunsdisq: Fixed-point fractional library routines. 38830 (line 1703) 38831 * __fractunsdita: Fixed-point fractional library routines. 38832 (line 1708) 38833 * __fractunsdiuda: Fixed-point fractional library routines. 38834 (line 1720) 38835 * __fractunsdiudq: Fixed-point fractional library routines. 38836 (line 1715) 38837 * __fractunsdiuha: Fixed-point fractional library routines. 38838 (line 1717) 38839 * __fractunsdiuhq: Fixed-point fractional library routines. 38840 (line 1711) 38841 * __fractunsdiuqq: Fixed-point fractional library routines. 38842 (line 1710) 38843 * __fractunsdiusa: Fixed-point fractional library routines. 38844 (line 1718) 38845 * __fractunsdiusq: Fixed-point fractional library routines. 38846 (line 1713) 38847 * __fractunsdiuta: Fixed-point fractional library routines. 38848 (line 1722) 38849 * __fractunsdqdi: Fixed-point fractional library routines. 38850 (line 1539) 38851 * __fractunsdqhi: Fixed-point fractional library routines. 38852 (line 1537) 38853 * __fractunsdqqi: Fixed-point fractional library routines. 38854 (line 1536) 38855 * __fractunsdqsi: Fixed-point fractional library routines. 38856 (line 1538) 38857 * __fractunsdqti: Fixed-point fractional library routines. 38858 (line 1541) 38859 * __fractunshadi: Fixed-point fractional library routines. 38860 (line 1545) 38861 * __fractunshahi: Fixed-point fractional library routines. 38862 (line 1543) 38863 * __fractunshaqi: Fixed-point fractional library routines. 38864 (line 1542) 38865 * __fractunshasi: Fixed-point fractional library routines. 38866 (line 1544) 38867 * __fractunshati: Fixed-point fractional library routines. 38868 (line 1546) 38869 * __fractunshida: Fixed-point fractional library routines. 38870 (line 1663) 38871 * __fractunshidq: Fixed-point fractional library routines. 38872 (line 1660) 38873 * __fractunshiha: Fixed-point fractional library routines. 38874 (line 1661) 38875 * __fractunshihq: Fixed-point fractional library routines. 38876 (line 1658) 38877 * __fractunshiqq: Fixed-point fractional library routines. 38878 (line 1657) 38879 * __fractunshisa: Fixed-point fractional library routines. 38880 (line 1662) 38881 * __fractunshisq: Fixed-point fractional library routines. 38882 (line 1659) 38883 * __fractunshita: Fixed-point fractional library routines. 38884 (line 1664) 38885 * __fractunshiuda: Fixed-point fractional library routines. 38886 (line 1676) 38887 * __fractunshiudq: Fixed-point fractional library routines. 38888 (line 1671) 38889 * __fractunshiuha: Fixed-point fractional library routines. 38890 (line 1673) 38891 * __fractunshiuhq: Fixed-point fractional library routines. 38892 (line 1667) 38893 * __fractunshiuqq: Fixed-point fractional library routines. 38894 (line 1666) 38895 * __fractunshiusa: Fixed-point fractional library routines. 38896 (line 1674) 38897 * __fractunshiusq: Fixed-point fractional library routines. 38898 (line 1669) 38899 * __fractunshiuta: Fixed-point fractional library routines. 38900 (line 1678) 38901 * __fractunshqdi: Fixed-point fractional library routines. 38902 (line 1529) 38903 * __fractunshqhi: Fixed-point fractional library routines. 38904 (line 1527) 38905 * __fractunshqqi: Fixed-point fractional library routines. 38906 (line 1526) 38907 * __fractunshqsi: Fixed-point fractional library routines. 38908 (line 1528) 38909 * __fractunshqti: Fixed-point fractional library routines. 38910 (line 1530) 38911 * __fractunsqida: Fixed-point fractional library routines. 38912 (line 1641) 38913 * __fractunsqidq: Fixed-point fractional library routines. 38914 (line 1638) 38915 * __fractunsqiha: Fixed-point fractional library routines. 38916 (line 1639) 38917 * __fractunsqihq: Fixed-point fractional library routines. 38918 (line 1636) 38919 * __fractunsqiqq: Fixed-point fractional library routines. 38920 (line 1635) 38921 * __fractunsqisa: Fixed-point fractional library routines. 38922 (line 1640) 38923 * __fractunsqisq: Fixed-point fractional library routines. 38924 (line 1637) 38925 * __fractunsqita: Fixed-point fractional library routines. 38926 (line 1642) 38927 * __fractunsqiuda: Fixed-point fractional library routines. 38928 (line 1654) 38929 * __fractunsqiudq: Fixed-point fractional library routines. 38930 (line 1649) 38931 * __fractunsqiuha: Fixed-point fractional library routines. 38932 (line 1651) 38933 * __fractunsqiuhq: Fixed-point fractional library routines. 38934 (line 1645) 38935 * __fractunsqiuqq: Fixed-point fractional library routines. 38936 (line 1644) 38937 * __fractunsqiusa: Fixed-point fractional library routines. 38938 (line 1652) 38939 * __fractunsqiusq: Fixed-point fractional library routines. 38940 (line 1647) 38941 * __fractunsqiuta: Fixed-point fractional library routines. 38942 (line 1656) 38943 * __fractunsqqdi: Fixed-point fractional library routines. 38944 (line 1524) 38945 * __fractunsqqhi: Fixed-point fractional library routines. 38946 (line 1522) 38947 * __fractunsqqqi: Fixed-point fractional library routines. 38948 (line 1521) 38949 * __fractunsqqsi: Fixed-point fractional library routines. 38950 (line 1523) 38951 * __fractunsqqti: Fixed-point fractional library routines. 38952 (line 1525) 38953 * __fractunssadi: Fixed-point fractional library routines. 38954 (line 1550) 38955 * __fractunssahi: Fixed-point fractional library routines. 38956 (line 1548) 38957 * __fractunssaqi: Fixed-point fractional library routines. 38958 (line 1547) 38959 * __fractunssasi: Fixed-point fractional library routines. 38960 (line 1549) 38961 * __fractunssati: Fixed-point fractional library routines. 38962 (line 1551) 38963 * __fractunssida: Fixed-point fractional library routines. 38964 (line 1685) 38965 * __fractunssidq: Fixed-point fractional library routines. 38966 (line 1682) 38967 * __fractunssiha: Fixed-point fractional library routines. 38968 (line 1683) 38969 * __fractunssihq: Fixed-point fractional library routines. 38970 (line 1680) 38971 * __fractunssiqq: Fixed-point fractional library routines. 38972 (line 1679) 38973 * __fractunssisa: Fixed-point fractional library routines. 38974 (line 1684) 38975 * __fractunssisq: Fixed-point fractional library routines. 38976 (line 1681) 38977 * __fractunssita: Fixed-point fractional library routines. 38978 (line 1686) 38979 * __fractunssiuda: Fixed-point fractional library routines. 38980 (line 1698) 38981 * __fractunssiudq: Fixed-point fractional library routines. 38982 (line 1693) 38983 * __fractunssiuha: Fixed-point fractional library routines. 38984 (line 1695) 38985 * __fractunssiuhq: Fixed-point fractional library routines. 38986 (line 1689) 38987 * __fractunssiuqq: Fixed-point fractional library routines. 38988 (line 1688) 38989 * __fractunssiusa: Fixed-point fractional library routines. 38990 (line 1696) 38991 * __fractunssiusq: Fixed-point fractional library routines. 38992 (line 1691) 38993 * __fractunssiuta: Fixed-point fractional library routines. 38994 (line 1700) 38995 * __fractunssqdi: Fixed-point fractional library routines. 38996 (line 1534) 38997 * __fractunssqhi: Fixed-point fractional library routines. 38998 (line 1532) 38999 * __fractunssqqi: Fixed-point fractional library routines. 39000 (line 1531) 39001 * __fractunssqsi: Fixed-point fractional library routines. 39002 (line 1533) 39003 * __fractunssqti: Fixed-point fractional library routines. 39004 (line 1535) 39005 * __fractunstadi: Fixed-point fractional library routines. 39006 (line 1560) 39007 * __fractunstahi: Fixed-point fractional library routines. 39008 (line 1558) 39009 * __fractunstaqi: Fixed-point fractional library routines. 39010 (line 1557) 39011 * __fractunstasi: Fixed-point fractional library routines. 39012 (line 1559) 39013 * __fractunstati: Fixed-point fractional library routines. 39014 (line 1562) 39015 * __fractunstida: Fixed-point fractional library routines. 39016 (line 1730) 39017 * __fractunstidq: Fixed-point fractional library routines. 39018 (line 1727) 39019 * __fractunstiha: Fixed-point fractional library routines. 39020 (line 1728) 39021 * __fractunstihq: Fixed-point fractional library routines. 39022 (line 1724) 39023 * __fractunstiqq: Fixed-point fractional library routines. 39024 (line 1723) 39025 * __fractunstisa: Fixed-point fractional library routines. 39026 (line 1729) 39027 * __fractunstisq: Fixed-point fractional library routines. 39028 (line 1725) 39029 * __fractunstita: Fixed-point fractional library routines. 39030 (line 1732) 39031 * __fractunstiuda: Fixed-point fractional library routines. 39032 (line 1746) 39033 * __fractunstiudq: Fixed-point fractional library routines. 39034 (line 1740) 39035 * __fractunstiuha: Fixed-point fractional library routines. 39036 (line 1742) 39037 * __fractunstiuhq: Fixed-point fractional library routines. 39038 (line 1736) 39039 * __fractunstiuqq: Fixed-point fractional library routines. 39040 (line 1734) 39041 * __fractunstiusa: Fixed-point fractional library routines. 39042 (line 1744) 39043 * __fractunstiusq: Fixed-point fractional library routines. 39044 (line 1738) 39045 * __fractunstiuta: Fixed-point fractional library routines. 39046 (line 1748) 39047 * __fractunsudadi: Fixed-point fractional library routines. 39048 (line 1622) 39049 * __fractunsudahi: Fixed-point fractional library routines. 39050 (line 1618) 39051 * __fractunsudaqi: Fixed-point fractional library routines. 39052 (line 1616) 39053 * __fractunsudasi: Fixed-point fractional library routines. 39054 (line 1620) 39055 * __fractunsudati: Fixed-point fractional library routines. 39056 (line 1624) 39057 * __fractunsudqdi: Fixed-point fractional library routines. 39058 (line 1596) 39059 * __fractunsudqhi: Fixed-point fractional library routines. 39060 (line 1592) 39061 * __fractunsudqqi: Fixed-point fractional library routines. 39062 (line 1590) 39063 * __fractunsudqsi: Fixed-point fractional library routines. 39064 (line 1594) 39065 * __fractunsudqti: Fixed-point fractional library routines. 39066 (line 1598) 39067 * __fractunsuhadi: Fixed-point fractional library routines. 39068 (line 1606) 39069 * __fractunsuhahi: Fixed-point fractional library routines. 39070 (line 1602) 39071 * __fractunsuhaqi: Fixed-point fractional library routines. 39072 (line 1600) 39073 * __fractunsuhasi: Fixed-point fractional library routines. 39074 (line 1604) 39075 * __fractunsuhati: Fixed-point fractional library routines. 39076 (line 1608) 39077 * __fractunsuhqdi: Fixed-point fractional library routines. 39078 (line 1576) 39079 * __fractunsuhqhi: Fixed-point fractional library routines. 39080 (line 1574) 39081 * __fractunsuhqqi: Fixed-point fractional library routines. 39082 (line 1573) 39083 * __fractunsuhqsi: Fixed-point fractional library routines. 39084 (line 1575) 39085 * __fractunsuhqti: Fixed-point fractional library routines. 39086 (line 1578) 39087 * __fractunsuqqdi: Fixed-point fractional library routines. 39088 (line 1570) 39089 * __fractunsuqqhi: Fixed-point fractional library routines. 39090 (line 1566) 39091 * __fractunsuqqqi: Fixed-point fractional library routines. 39092 (line 1564) 39093 * __fractunsuqqsi: Fixed-point fractional library routines. 39094 (line 1568) 39095 * __fractunsuqqti: Fixed-point fractional library routines. 39096 (line 1572) 39097 * __fractunsusadi: Fixed-point fractional library routines. 39098 (line 1612) 39099 * __fractunsusahi: Fixed-point fractional library routines. 39100 (line 1610) 39101 * __fractunsusaqi: Fixed-point fractional library routines. 39102 (line 1609) 39103 * __fractunsusasi: Fixed-point fractional library routines. 39104 (line 1611) 39105 * __fractunsusati: Fixed-point fractional library routines. 39106 (line 1614) 39107 * __fractunsusqdi: Fixed-point fractional library routines. 39108 (line 1586) 39109 * __fractunsusqhi: Fixed-point fractional library routines. 39110 (line 1582) 39111 * __fractunsusqqi: Fixed-point fractional library routines. 39112 (line 1580) 39113 * __fractunsusqsi: Fixed-point fractional library routines. 39114 (line 1584) 39115 * __fractunsusqti: Fixed-point fractional library routines. 39116 (line 1588) 39117 * __fractunsutadi: Fixed-point fractional library routines. 39118 (line 1632) 39119 * __fractunsutahi: Fixed-point fractional library routines. 39120 (line 1628) 39121 * __fractunsutaqi: Fixed-point fractional library routines. 39122 (line 1626) 39123 * __fractunsutasi: Fixed-point fractional library routines. 39124 (line 1630) 39125 * __fractunsutati: Fixed-point fractional library routines. 39126 (line 1634) 39127 * __fractuqqda: Fixed-point fractional library routines. 39128 (line 672) 39129 * __fractuqqdf: Fixed-point fractional library routines. 39130 (line 695) 39131 * __fractuqqdi: Fixed-point fractional library routines. 39132 (line 692) 39133 * __fractuqqdq: Fixed-point fractional library routines. 39134 (line 669) 39135 * __fractuqqha: Fixed-point fractional library routines. 39136 (line 670) 39137 * __fractuqqhi: Fixed-point fractional library routines. 39138 (line 690) 39139 * __fractuqqhq: Fixed-point fractional library routines. 39140 (line 666) 39141 * __fractuqqqi: Fixed-point fractional library routines. 39142 (line 689) 39143 * __fractuqqqq: Fixed-point fractional library routines. 39144 (line 665) 39145 * __fractuqqsa: Fixed-point fractional library routines. 39146 (line 671) 39147 * __fractuqqsf: Fixed-point fractional library routines. 39148 (line 694) 39149 * __fractuqqsi: Fixed-point fractional library routines. 39150 (line 691) 39151 * __fractuqqsq: Fixed-point fractional library routines. 39152 (line 667) 39153 * __fractuqqta: Fixed-point fractional library routines. 39154 (line 674) 39155 * __fractuqqti: Fixed-point fractional library routines. 39156 (line 693) 39157 * __fractuqquda: Fixed-point fractional library routines. 39158 (line 686) 39159 * __fractuqqudq2: Fixed-point fractional library routines. 39160 (line 680) 39161 * __fractuqquha: Fixed-point fractional library routines. 39162 (line 682) 39163 * __fractuqquhq2: Fixed-point fractional library routines. 39164 (line 676) 39165 * __fractuqqusa: Fixed-point fractional library routines. 39166 (line 684) 39167 * __fractuqqusq2: Fixed-point fractional library routines. 39168 (line 678) 39169 * __fractuqquta: Fixed-point fractional library routines. 39170 (line 688) 39171 * __fractusada: Fixed-point fractional library routines. 39172 (line 829) 39173 * __fractusadf: Fixed-point fractional library routines. 39174 (line 850) 39175 * __fractusadi: Fixed-point fractional library routines. 39176 (line 847) 39177 * __fractusadq: Fixed-point fractional library routines. 39178 (line 826) 39179 * __fractusaha: Fixed-point fractional library routines. 39180 (line 827) 39181 * __fractusahi: Fixed-point fractional library routines. 39182 (line 845) 39183 * __fractusahq: Fixed-point fractional library routines. 39184 (line 824) 39185 * __fractusaqi: Fixed-point fractional library routines. 39186 (line 844) 39187 * __fractusaqq: Fixed-point fractional library routines. 39188 (line 823) 39189 * __fractusasa: Fixed-point fractional library routines. 39190 (line 828) 39191 * __fractusasf: Fixed-point fractional library routines. 39192 (line 849) 39193 * __fractusasi: Fixed-point fractional library routines. 39194 (line 846) 39195 * __fractusasq: Fixed-point fractional library routines. 39196 (line 825) 39197 * __fractusata: Fixed-point fractional library routines. 39198 (line 830) 39199 * __fractusati: Fixed-point fractional library routines. 39200 (line 848) 39201 * __fractusauda2: Fixed-point fractional library routines. 39202 (line 841) 39203 * __fractusaudq: Fixed-point fractional library routines. 39204 (line 837) 39205 * __fractusauha2: Fixed-point fractional library routines. 39206 (line 839) 39207 * __fractusauhq: Fixed-point fractional library routines. 39208 (line 833) 39209 * __fractusauqq: Fixed-point fractional library routines. 39210 (line 832) 39211 * __fractusausq: Fixed-point fractional library routines. 39212 (line 835) 39213 * __fractusauta2: Fixed-point fractional library routines. 39214 (line 843) 39215 * __fractusqda: Fixed-point fractional library routines. 39216 (line 731) 39217 * __fractusqdf: Fixed-point fractional library routines. 39218 (line 754) 39219 * __fractusqdi: Fixed-point fractional library routines. 39220 (line 751) 39221 * __fractusqdq: Fixed-point fractional library routines. 39222 (line 728) 39223 * __fractusqha: Fixed-point fractional library routines. 39224 (line 729) 39225 * __fractusqhi: Fixed-point fractional library routines. 39226 (line 749) 39227 * __fractusqhq: Fixed-point fractional library routines. 39228 (line 725) 39229 * __fractusqqi: Fixed-point fractional library routines. 39230 (line 748) 39231 * __fractusqqq: Fixed-point fractional library routines. 39232 (line 724) 39233 * __fractusqsa: Fixed-point fractional library routines. 39234 (line 730) 39235 * __fractusqsf: Fixed-point fractional library routines. 39236 (line 753) 39237 * __fractusqsi: Fixed-point fractional library routines. 39238 (line 750) 39239 * __fractusqsq: Fixed-point fractional library routines. 39240 (line 726) 39241 * __fractusqta: Fixed-point fractional library routines. 39242 (line 733) 39243 * __fractusqti: Fixed-point fractional library routines. 39244 (line 752) 39245 * __fractusquda: Fixed-point fractional library routines. 39246 (line 745) 39247 * __fractusqudq2: Fixed-point fractional library routines. 39248 (line 739) 39249 * __fractusquha: Fixed-point fractional library routines. 39250 (line 741) 39251 * __fractusquhq2: Fixed-point fractional library routines. 39252 (line 737) 39253 * __fractusquqq2: Fixed-point fractional library routines. 39254 (line 735) 39255 * __fractusqusa: Fixed-point fractional library routines. 39256 (line 743) 39257 * __fractusquta: Fixed-point fractional library routines. 39258 (line 747) 39259 * __fractutada: Fixed-point fractional library routines. 39260 (line 893) 39261 * __fractutadf: Fixed-point fractional library routines. 39262 (line 918) 39263 * __fractutadi: Fixed-point fractional library routines. 39264 (line 914) 39265 * __fractutadq: Fixed-point fractional library routines. 39266 (line 888) 39267 * __fractutaha: Fixed-point fractional library routines. 39268 (line 890) 39269 * __fractutahi: Fixed-point fractional library routines. 39270 (line 912) 39271 * __fractutahq: Fixed-point fractional library routines. 39272 (line 884) 39273 * __fractutaqi: Fixed-point fractional library routines. 39274 (line 911) 39275 * __fractutaqq: Fixed-point fractional library routines. 39276 (line 883) 39277 * __fractutasa: Fixed-point fractional library routines. 39278 (line 891) 39279 * __fractutasf: Fixed-point fractional library routines. 39280 (line 917) 39281 * __fractutasi: Fixed-point fractional library routines. 39282 (line 913) 39283 * __fractutasq: Fixed-point fractional library routines. 39284 (line 886) 39285 * __fractutata: Fixed-point fractional library routines. 39286 (line 895) 39287 * __fractutati: Fixed-point fractional library routines. 39288 (line 916) 39289 * __fractutauda2: Fixed-point fractional library routines. 39290 (line 909) 39291 * __fractutaudq: Fixed-point fractional library routines. 39292 (line 903) 39293 * __fractutauha2: Fixed-point fractional library routines. 39294 (line 905) 39295 * __fractutauhq: Fixed-point fractional library routines. 39296 (line 899) 39297 * __fractutauqq: Fixed-point fractional library routines. 39298 (line 897) 39299 * __fractutausa2: Fixed-point fractional library routines. 39300 (line 907) 39301 * __fractutausq: Fixed-point fractional library routines. 39302 (line 901) 39303 * __gedf2: Soft float library routines. 39304 (line 206) 39305 * __gesf2: Soft float library routines. 39306 (line 205) 39307 * __getf2: Soft float library routines. 39308 (line 207) 39309 * __gtdf2: Soft float library routines. 39310 (line 224) 39311 * __gtsf2: Soft float library routines. 39312 (line 223) 39313 * __gttf2: Soft float library routines. 39314 (line 225) 39315 * __ledf2: Soft float library routines. 39316 (line 218) 39317 * __lesf2: Soft float library routines. 39318 (line 217) 39319 * __letf2: Soft float library routines. 39320 (line 219) 39321 * __lshrdi3: Integer library routines. 39322 (line 31) 39323 * __lshrsi3: Integer library routines. 39324 (line 30) 39325 * __lshrti3: Integer library routines. 39326 (line 32) 39327 * __lshruda3: Fixed-point fractional library routines. 39328 (line 390) 39329 * __lshrudq3: Fixed-point fractional library routines. 39330 (line 384) 39331 * __lshruha3: Fixed-point fractional library routines. 39332 (line 386) 39333 * __lshruhq3: Fixed-point fractional library routines. 39334 (line 380) 39335 * __lshruqq3: Fixed-point fractional library routines. 39336 (line 378) 39337 * __lshrusa3: Fixed-point fractional library routines. 39338 (line 388) 39339 * __lshrusq3: Fixed-point fractional library routines. 39340 (line 382) 39341 * __lshruta3: Fixed-point fractional library routines. 39342 (line 392) 39343 * __ltdf2: Soft float library routines. 39344 (line 212) 39345 * __ltsf2: Soft float library routines. 39346 (line 211) 39347 * __lttf2: Soft float library routines. 39348 (line 213) 39349 * __main: Collect2. (line 15) 39350 * __moddi3: Integer library routines. 39351 (line 37) 39352 * __modsi3: Integer library routines. 39353 (line 36) 39354 * __modti3: Integer library routines. 39355 (line 38) 39356 * __mulda3: Fixed-point fractional library routines. 39357 (line 171) 39358 * __muldc3: Soft float library routines. 39359 (line 241) 39360 * __muldf3: Soft float library routines. 39361 (line 40) 39362 * __muldi3: Integer library routines. 39363 (line 43) 39364 * __muldq3: Fixed-point fractional library routines. 39365 (line 159) 39366 * __mulha3: Fixed-point fractional library routines. 39367 (line 169) 39368 * __mulhq3: Fixed-point fractional library routines. 39369 (line 156) 39370 * __mulqq3: Fixed-point fractional library routines. 39371 (line 155) 39372 * __mulsa3: Fixed-point fractional library routines. 39373 (line 170) 39374 * __mulsc3: Soft float library routines. 39375 (line 239) 39376 * __mulsf3: Soft float library routines. 39377 (line 39) 39378 * __mulsi3: Integer library routines. 39379 (line 42) 39380 * __mulsq3: Fixed-point fractional library routines. 39381 (line 157) 39382 * __multa3: Fixed-point fractional library routines. 39383 (line 173) 39384 * __multc3: Soft float library routines. 39385 (line 243) 39386 * __multf3: Soft float library routines. 39387 (line 42) 39388 * __multi3: Integer library routines. 39389 (line 44) 39390 * __muluda3: Fixed-point fractional library routines. 39391 (line 179) 39392 * __muludq3: Fixed-point fractional library routines. 39393 (line 167) 39394 * __muluha3: Fixed-point fractional library routines. 39395 (line 175) 39396 * __muluhq3: Fixed-point fractional library routines. 39397 (line 163) 39398 * __muluqq3: Fixed-point fractional library routines. 39399 (line 161) 39400 * __mulusa3: Fixed-point fractional library routines. 39401 (line 177) 39402 * __mulusq3: Fixed-point fractional library routines. 39403 (line 165) 39404 * __muluta3: Fixed-point fractional library routines. 39405 (line 181) 39406 * __mulvdi3: Integer library routines. 39407 (line 115) 39408 * __mulvsi3: Integer library routines. 39409 (line 114) 39410 * __mulxc3: Soft float library routines. 39411 (line 245) 39412 * __mulxf3: Soft float library routines. 39413 (line 44) 39414 * __nedf2: Soft float library routines. 39415 (line 200) 39416 * __negda2: Fixed-point fractional library routines. 39417 (line 299) 39418 * __negdf2: Soft float library routines. 39419 (line 56) 39420 * __negdi2: Integer library routines. 39421 (line 47) 39422 * __negdq2: Fixed-point fractional library routines. 39423 (line 289) 39424 * __negha2: Fixed-point fractional library routines. 39425 (line 297) 39426 * __neghq2: Fixed-point fractional library routines. 39427 (line 287) 39428 * __negqq2: Fixed-point fractional library routines. 39429 (line 286) 39430 * __negsa2: Fixed-point fractional library routines. 39431 (line 298) 39432 * __negsf2: Soft float library routines. 39433 (line 55) 39434 * __negsq2: Fixed-point fractional library routines. 39435 (line 288) 39436 * __negta2: Fixed-point fractional library routines. 39437 (line 300) 39438 * __negtf2: Soft float library routines. 39439 (line 57) 39440 * __negti2: Integer library routines. 39441 (line 48) 39442 * __neguda2: Fixed-point fractional library routines. 39443 (line 305) 39444 * __negudq2: Fixed-point fractional library routines. 39445 (line 296) 39446 * __neguha2: Fixed-point fractional library routines. 39447 (line 302) 39448 * __neguhq2: Fixed-point fractional library routines. 39449 (line 292) 39450 * __neguqq2: Fixed-point fractional library routines. 39451 (line 291) 39452 * __negusa2: Fixed-point fractional library routines. 39453 (line 303) 39454 * __negusq2: Fixed-point fractional library routines. 39455 (line 294) 39456 * __neguta2: Fixed-point fractional library routines. 39457 (line 307) 39458 * __negvdi2: Integer library routines. 39459 (line 119) 39460 * __negvsi2: Integer library routines. 39461 (line 118) 39462 * __negxf2: Soft float library routines. 39463 (line 58) 39464 * __nesf2: Soft float library routines. 39465 (line 199) 39466 * __netf2: Soft float library routines. 39467 (line 201) 39468 * __paritydi2: Integer library routines. 39469 (line 151) 39470 * __paritysi2: Integer library routines. 39471 (line 150) 39472 * __parityti2: Integer library routines. 39473 (line 152) 39474 * __popcountdi2: Integer library routines. 39475 (line 157) 39476 * __popcountsi2: Integer library routines. 39477 (line 156) 39478 * __popcountti2: Integer library routines. 39479 (line 158) 39480 * __powidf2: Soft float library routines. 39481 (line 233) 39482 * __powisf2: Soft float library routines. 39483 (line 232) 39484 * __powitf2: Soft float library routines. 39485 (line 234) 39486 * __powixf2: Soft float library routines. 39487 (line 235) 39488 * __satfractdadq: Fixed-point fractional library routines. 39489 (line 1153) 39490 * __satfractdaha2: Fixed-point fractional library routines. 39491 (line 1154) 39492 * __satfractdahq: Fixed-point fractional library routines. 39493 (line 1151) 39494 * __satfractdaqq: Fixed-point fractional library routines. 39495 (line 1150) 39496 * __satfractdasa2: Fixed-point fractional library routines. 39497 (line 1155) 39498 * __satfractdasq: Fixed-point fractional library routines. 39499 (line 1152) 39500 * __satfractdata2: Fixed-point fractional library routines. 39501 (line 1156) 39502 * __satfractdauda: Fixed-point fractional library routines. 39503 (line 1166) 39504 * __satfractdaudq: Fixed-point fractional library routines. 39505 (line 1162) 39506 * __satfractdauha: Fixed-point fractional library routines. 39507 (line 1164) 39508 * __satfractdauhq: Fixed-point fractional library routines. 39509 (line 1159) 39510 * __satfractdauqq: Fixed-point fractional library routines. 39511 (line 1158) 39512 * __satfractdausa: Fixed-point fractional library routines. 39513 (line 1165) 39514 * __satfractdausq: Fixed-point fractional library routines. 39515 (line 1160) 39516 * __satfractdauta: Fixed-point fractional library routines. 39517 (line 1168) 39518 * __satfractdfda: Fixed-point fractional library routines. 39519 (line 1506) 39520 * __satfractdfdq: Fixed-point fractional library routines. 39521 (line 1503) 39522 * __satfractdfha: Fixed-point fractional library routines. 39523 (line 1504) 39524 * __satfractdfhq: Fixed-point fractional library routines. 39525 (line 1501) 39526 * __satfractdfqq: Fixed-point fractional library routines. 39527 (line 1500) 39528 * __satfractdfsa: Fixed-point fractional library routines. 39529 (line 1505) 39530 * __satfractdfsq: Fixed-point fractional library routines. 39531 (line 1502) 39532 * __satfractdfta: Fixed-point fractional library routines. 39533 (line 1507) 39534 * __satfractdfuda: Fixed-point fractional library routines. 39535 (line 1515) 39536 * __satfractdfudq: Fixed-point fractional library routines. 39537 (line 1512) 39538 * __satfractdfuha: Fixed-point fractional library routines. 39539 (line 1513) 39540 * __satfractdfuhq: Fixed-point fractional library routines. 39541 (line 1509) 39542 * __satfractdfuqq: Fixed-point fractional library routines. 39543 (line 1508) 39544 * __satfractdfusa: Fixed-point fractional library routines. 39545 (line 1514) 39546 * __satfractdfusq: Fixed-point fractional library routines. 39547 (line 1510) 39548 * __satfractdfuta: Fixed-point fractional library routines. 39549 (line 1517) 39550 * __satfractdida: Fixed-point fractional library routines. 39551 (line 1456) 39552 * __satfractdidq: Fixed-point fractional library routines. 39553 (line 1453) 39554 * __satfractdiha: Fixed-point fractional library routines. 39555 (line 1454) 39556 * __satfractdihq: Fixed-point fractional library routines. 39557 (line 1451) 39558 * __satfractdiqq: Fixed-point fractional library routines. 39559 (line 1450) 39560 * __satfractdisa: Fixed-point fractional library routines. 39561 (line 1455) 39562 * __satfractdisq: Fixed-point fractional library routines. 39563 (line 1452) 39564 * __satfractdita: Fixed-point fractional library routines. 39565 (line 1457) 39566 * __satfractdiuda: Fixed-point fractional library routines. 39567 (line 1464) 39568 * __satfractdiudq: Fixed-point fractional library routines. 39569 (line 1461) 39570 * __satfractdiuha: Fixed-point fractional library routines. 39571 (line 1462) 39572 * __satfractdiuhq: Fixed-point fractional library routines. 39573 (line 1459) 39574 * __satfractdiuqq: Fixed-point fractional library routines. 39575 (line 1458) 39576 * __satfractdiusa: Fixed-point fractional library routines. 39577 (line 1463) 39578 * __satfractdiusq: Fixed-point fractional library routines. 39579 (line 1460) 39580 * __satfractdiuta: Fixed-point fractional library routines. 39581 (line 1465) 39582 * __satfractdqda: Fixed-point fractional library routines. 39583 (line 1098) 39584 * __satfractdqha: Fixed-point fractional library routines. 39585 (line 1096) 39586 * __satfractdqhq2: Fixed-point fractional library routines. 39587 (line 1094) 39588 * __satfractdqqq2: Fixed-point fractional library routines. 39589 (line 1093) 39590 * __satfractdqsa: Fixed-point fractional library routines. 39591 (line 1097) 39592 * __satfractdqsq2: Fixed-point fractional library routines. 39593 (line 1095) 39594 * __satfractdqta: Fixed-point fractional library routines. 39595 (line 1099) 39596 * __satfractdquda: Fixed-point fractional library routines. 39597 (line 1111) 39598 * __satfractdqudq: Fixed-point fractional library routines. 39599 (line 1106) 39600 * __satfractdquha: Fixed-point fractional library routines. 39601 (line 1108) 39602 * __satfractdquhq: Fixed-point fractional library routines. 39603 (line 1102) 39604 * __satfractdquqq: Fixed-point fractional library routines. 39605 (line 1101) 39606 * __satfractdqusa: Fixed-point fractional library routines. 39607 (line 1109) 39608 * __satfractdqusq: Fixed-point fractional library routines. 39609 (line 1104) 39610 * __satfractdquta: Fixed-point fractional library routines. 39611 (line 1113) 39612 * __satfracthada2: Fixed-point fractional library routines. 39613 (line 1119) 39614 * __satfracthadq: Fixed-point fractional library routines. 39615 (line 1117) 39616 * __satfracthahq: Fixed-point fractional library routines. 39617 (line 1115) 39618 * __satfracthaqq: Fixed-point fractional library routines. 39619 (line 1114) 39620 * __satfracthasa2: Fixed-point fractional library routines. 39621 (line 1118) 39622 * __satfracthasq: Fixed-point fractional library routines. 39623 (line 1116) 39624 * __satfracthata2: Fixed-point fractional library routines. 39625 (line 1120) 39626 * __satfracthauda: Fixed-point fractional library routines. 39627 (line 1132) 39628 * __satfracthaudq: Fixed-point fractional library routines. 39629 (line 1127) 39630 * __satfracthauha: Fixed-point fractional library routines. 39631 (line 1129) 39632 * __satfracthauhq: Fixed-point fractional library routines. 39633 (line 1123) 39634 * __satfracthauqq: Fixed-point fractional library routines. 39635 (line 1122) 39636 * __satfracthausa: Fixed-point fractional library routines. 39637 (line 1130) 39638 * __satfracthausq: Fixed-point fractional library routines. 39639 (line 1125) 39640 * __satfracthauta: Fixed-point fractional library routines. 39641 (line 1134) 39642 * __satfracthida: Fixed-point fractional library routines. 39643 (line 1424) 39644 * __satfracthidq: Fixed-point fractional library routines. 39645 (line 1421) 39646 * __satfracthiha: Fixed-point fractional library routines. 39647 (line 1422) 39648 * __satfracthihq: Fixed-point fractional library routines. 39649 (line 1419) 39650 * __satfracthiqq: Fixed-point fractional library routines. 39651 (line 1418) 39652 * __satfracthisa: Fixed-point fractional library routines. 39653 (line 1423) 39654 * __satfracthisq: Fixed-point fractional library routines. 39655 (line 1420) 39656 * __satfracthita: Fixed-point fractional library routines. 39657 (line 1425) 39658 * __satfracthiuda: Fixed-point fractional library routines. 39659 (line 1432) 39660 * __satfracthiudq: Fixed-point fractional library routines. 39661 (line 1429) 39662 * __satfracthiuha: Fixed-point fractional library routines. 39663 (line 1430) 39664 * __satfracthiuhq: Fixed-point fractional library routines. 39665 (line 1427) 39666 * __satfracthiuqq: Fixed-point fractional library routines. 39667 (line 1426) 39668 * __satfracthiusa: Fixed-point fractional library routines. 39669 (line 1431) 39670 * __satfracthiusq: Fixed-point fractional library routines. 39671 (line 1428) 39672 * __satfracthiuta: Fixed-point fractional library routines. 39673 (line 1433) 39674 * __satfracthqda: Fixed-point fractional library routines. 39675 (line 1064) 39676 * __satfracthqdq2: Fixed-point fractional library routines. 39677 (line 1061) 39678 * __satfracthqha: Fixed-point fractional library routines. 39679 (line 1062) 39680 * __satfracthqqq2: Fixed-point fractional library routines. 39681 (line 1059) 39682 * __satfracthqsa: Fixed-point fractional library routines. 39683 (line 1063) 39684 * __satfracthqsq2: Fixed-point fractional library routines. 39685 (line 1060) 39686 * __satfracthqta: Fixed-point fractional library routines. 39687 (line 1065) 39688 * __satfracthquda: Fixed-point fractional library routines. 39689 (line 1072) 39690 * __satfracthqudq: Fixed-point fractional library routines. 39691 (line 1069) 39692 * __satfracthquha: Fixed-point fractional library routines. 39693 (line 1070) 39694 * __satfracthquhq: Fixed-point fractional library routines. 39695 (line 1067) 39696 * __satfracthquqq: Fixed-point fractional library routines. 39697 (line 1066) 39698 * __satfracthqusa: Fixed-point fractional library routines. 39699 (line 1071) 39700 * __satfracthqusq: Fixed-point fractional library routines. 39701 (line 1068) 39702 * __satfracthquta: Fixed-point fractional library routines. 39703 (line 1073) 39704 * __satfractqida: Fixed-point fractional library routines. 39705 (line 1402) 39706 * __satfractqidq: Fixed-point fractional library routines. 39707 (line 1399) 39708 * __satfractqiha: Fixed-point fractional library routines. 39709 (line 1400) 39710 * __satfractqihq: Fixed-point fractional library routines. 39711 (line 1397) 39712 * __satfractqiqq: Fixed-point fractional library routines. 39713 (line 1396) 39714 * __satfractqisa: Fixed-point fractional library routines. 39715 (line 1401) 39716 * __satfractqisq: Fixed-point fractional library routines. 39717 (line 1398) 39718 * __satfractqita: Fixed-point fractional library routines. 39719 (line 1403) 39720 * __satfractqiuda: Fixed-point fractional library routines. 39721 (line 1415) 39722 * __satfractqiudq: Fixed-point fractional library routines. 39723 (line 1410) 39724 * __satfractqiuha: Fixed-point fractional library routines. 39725 (line 1412) 39726 * __satfractqiuhq: Fixed-point fractional library routines. 39727 (line 1406) 39728 * __satfractqiuqq: Fixed-point fractional library routines. 39729 (line 1405) 39730 * __satfractqiusa: Fixed-point fractional library routines. 39731 (line 1413) 39732 * __satfractqiusq: Fixed-point fractional library routines. 39733 (line 1408) 39734 * __satfractqiuta: Fixed-point fractional library routines. 39735 (line 1417) 39736 * __satfractqqda: Fixed-point fractional library routines. 39737 (line 1043) 39738 * __satfractqqdq2: Fixed-point fractional library routines. 39739 (line 1040) 39740 * __satfractqqha: Fixed-point fractional library routines. 39741 (line 1041) 39742 * __satfractqqhq2: Fixed-point fractional library routines. 39743 (line 1038) 39744 * __satfractqqsa: Fixed-point fractional library routines. 39745 (line 1042) 39746 * __satfractqqsq2: Fixed-point fractional library routines. 39747 (line 1039) 39748 * __satfractqqta: Fixed-point fractional library routines. 39749 (line 1044) 39750 * __satfractqquda: Fixed-point fractional library routines. 39751 (line 1056) 39752 * __satfractqqudq: Fixed-point fractional library routines. 39753 (line 1051) 39754 * __satfractqquha: Fixed-point fractional library routines. 39755 (line 1053) 39756 * __satfractqquhq: Fixed-point fractional library routines. 39757 (line 1047) 39758 * __satfractqquqq: Fixed-point fractional library routines. 39759 (line 1046) 39760 * __satfractqqusa: Fixed-point fractional library routines. 39761 (line 1054) 39762 * __satfractqqusq: Fixed-point fractional library routines. 39763 (line 1049) 39764 * __satfractqquta: Fixed-point fractional library routines. 39765 (line 1058) 39766 * __satfractsada2: Fixed-point fractional library routines. 39767 (line 1140) 39768 * __satfractsadq: Fixed-point fractional library routines. 39769 (line 1138) 39770 * __satfractsaha2: Fixed-point fractional library routines. 39771 (line 1139) 39772 * __satfractsahq: Fixed-point fractional library routines. 39773 (line 1136) 39774 * __satfractsaqq: Fixed-point fractional library routines. 39775 (line 1135) 39776 * __satfractsasq: Fixed-point fractional library routines. 39777 (line 1137) 39778 * __satfractsata2: Fixed-point fractional library routines. 39779 (line 1141) 39780 * __satfractsauda: Fixed-point fractional library routines. 39781 (line 1148) 39782 * __satfractsaudq: Fixed-point fractional library routines. 39783 (line 1145) 39784 * __satfractsauha: Fixed-point fractional library routines. 39785 (line 1146) 39786 * __satfractsauhq: Fixed-point fractional library routines. 39787 (line 1143) 39788 * __satfractsauqq: Fixed-point fractional library routines. 39789 (line 1142) 39790 * __satfractsausa: Fixed-point fractional library routines. 39791 (line 1147) 39792 * __satfractsausq: Fixed-point fractional library routines. 39793 (line 1144) 39794 * __satfractsauta: Fixed-point fractional library routines. 39795 (line 1149) 39796 * __satfractsfda: Fixed-point fractional library routines. 39797 (line 1490) 39798 * __satfractsfdq: Fixed-point fractional library routines. 39799 (line 1487) 39800 * __satfractsfha: Fixed-point fractional library routines. 39801 (line 1488) 39802 * __satfractsfhq: Fixed-point fractional library routines. 39803 (line 1485) 39804 * __satfractsfqq: Fixed-point fractional library routines. 39805 (line 1484) 39806 * __satfractsfsa: Fixed-point fractional library routines. 39807 (line 1489) 39808 * __satfractsfsq: Fixed-point fractional library routines. 39809 (line 1486) 39810 * __satfractsfta: Fixed-point fractional library routines. 39811 (line 1491) 39812 * __satfractsfuda: Fixed-point fractional library routines. 39813 (line 1498) 39814 * __satfractsfudq: Fixed-point fractional library routines. 39815 (line 1495) 39816 * __satfractsfuha: Fixed-point fractional library routines. 39817 (line 1496) 39818 * __satfractsfuhq: Fixed-point fractional library routines. 39819 (line 1493) 39820 * __satfractsfuqq: Fixed-point fractional library routines. 39821 (line 1492) 39822 * __satfractsfusa: Fixed-point fractional library routines. 39823 (line 1497) 39824 * __satfractsfusq: Fixed-point fractional library routines. 39825 (line 1494) 39826 * __satfractsfuta: Fixed-point fractional library routines. 39827 (line 1499) 39828 * __satfractsida: Fixed-point fractional library routines. 39829 (line 1440) 39830 * __satfractsidq: Fixed-point fractional library routines. 39831 (line 1437) 39832 * __satfractsiha: Fixed-point fractional library routines. 39833 (line 1438) 39834 * __satfractsihq: Fixed-point fractional library routines. 39835 (line 1435) 39836 * __satfractsiqq: Fixed-point fractional library routines. 39837 (line 1434) 39838 * __satfractsisa: Fixed-point fractional library routines. 39839 (line 1439) 39840 * __satfractsisq: Fixed-point fractional library routines. 39841 (line 1436) 39842 * __satfractsita: Fixed-point fractional library routines. 39843 (line 1441) 39844 * __satfractsiuda: Fixed-point fractional library routines. 39845 (line 1448) 39846 * __satfractsiudq: Fixed-point fractional library routines. 39847 (line 1445) 39848 * __satfractsiuha: Fixed-point fractional library routines. 39849 (line 1446) 39850 * __satfractsiuhq: Fixed-point fractional library routines. 39851 (line 1443) 39852 * __satfractsiuqq: Fixed-point fractional library routines. 39853 (line 1442) 39854 * __satfractsiusa: Fixed-point fractional library routines. 39855 (line 1447) 39856 * __satfractsiusq: Fixed-point fractional library routines. 39857 (line 1444) 39858 * __satfractsiuta: Fixed-point fractional library routines. 39859 (line 1449) 39860 * __satfractsqda: Fixed-point fractional library routines. 39861 (line 1079) 39862 * __satfractsqdq2: Fixed-point fractional library routines. 39863 (line 1076) 39864 * __satfractsqha: Fixed-point fractional library routines. 39865 (line 1077) 39866 * __satfractsqhq2: Fixed-point fractional library routines. 39867 (line 1075) 39868 * __satfractsqqq2: Fixed-point fractional library routines. 39869 (line 1074) 39870 * __satfractsqsa: Fixed-point fractional library routines. 39871 (line 1078) 39872 * __satfractsqta: Fixed-point fractional library routines. 39873 (line 1080) 39874 * __satfractsquda: Fixed-point fractional library routines. 39875 (line 1090) 39876 * __satfractsqudq: Fixed-point fractional library routines. 39877 (line 1086) 39878 * __satfractsquha: Fixed-point fractional library routines. 39879 (line 1088) 39880 * __satfractsquhq: Fixed-point fractional library routines. 39881 (line 1083) 39882 * __satfractsquqq: Fixed-point fractional library routines. 39883 (line 1082) 39884 * __satfractsqusa: Fixed-point fractional library routines. 39885 (line 1089) 39886 * __satfractsqusq: Fixed-point fractional library routines. 39887 (line 1084) 39888 * __satfractsquta: Fixed-point fractional library routines. 39889 (line 1092) 39890 * __satfracttada2: Fixed-point fractional library routines. 39891 (line 1175) 39892 * __satfracttadq: Fixed-point fractional library routines. 39893 (line 1172) 39894 * __satfracttaha2: Fixed-point fractional library routines. 39895 (line 1173) 39896 * __satfracttahq: Fixed-point fractional library routines. 39897 (line 1170) 39898 * __satfracttaqq: Fixed-point fractional library routines. 39899 (line 1169) 39900 * __satfracttasa2: Fixed-point fractional library routines. 39901 (line 1174) 39902 * __satfracttasq: Fixed-point fractional library routines. 39903 (line 1171) 39904 * __satfracttauda: Fixed-point fractional library routines. 39905 (line 1187) 39906 * __satfracttaudq: Fixed-point fractional library routines. 39907 (line 1182) 39908 * __satfracttauha: Fixed-point fractional library routines. 39909 (line 1184) 39910 * __satfracttauhq: Fixed-point fractional library routines. 39911 (line 1178) 39912 * __satfracttauqq: Fixed-point fractional library routines. 39913 (line 1177) 39914 * __satfracttausa: Fixed-point fractional library routines. 39915 (line 1185) 39916 * __satfracttausq: Fixed-point fractional library routines. 39917 (line 1180) 39918 * __satfracttauta: Fixed-point fractional library routines. 39919 (line 1189) 39920 * __satfracttida: Fixed-point fractional library routines. 39921 (line 1472) 39922 * __satfracttidq: Fixed-point fractional library routines. 39923 (line 1469) 39924 * __satfracttiha: Fixed-point fractional library routines. 39925 (line 1470) 39926 * __satfracttihq: Fixed-point fractional library routines. 39927 (line 1467) 39928 * __satfracttiqq: Fixed-point fractional library routines. 39929 (line 1466) 39930 * __satfracttisa: Fixed-point fractional library routines. 39931 (line 1471) 39932 * __satfracttisq: Fixed-point fractional library routines. 39933 (line 1468) 39934 * __satfracttita: Fixed-point fractional library routines. 39935 (line 1473) 39936 * __satfracttiuda: Fixed-point fractional library routines. 39937 (line 1481) 39938 * __satfracttiudq: Fixed-point fractional library routines. 39939 (line 1478) 39940 * __satfracttiuha: Fixed-point fractional library routines. 39941 (line 1479) 39942 * __satfracttiuhq: Fixed-point fractional library routines. 39943 (line 1475) 39944 * __satfracttiuqq: Fixed-point fractional library routines. 39945 (line 1474) 39946 * __satfracttiusa: Fixed-point fractional library routines. 39947 (line 1480) 39948 * __satfracttiusq: Fixed-point fractional library routines. 39949 (line 1476) 39950 * __satfracttiuta: Fixed-point fractional library routines. 39951 (line 1483) 39952 * __satfractudada: Fixed-point fractional library routines. 39953 (line 1351) 39954 * __satfractudadq: Fixed-point fractional library routines. 39955 (line 1347) 39956 * __satfractudaha: Fixed-point fractional library routines. 39957 (line 1349) 39958 * __satfractudahq: Fixed-point fractional library routines. 39959 (line 1344) 39960 * __satfractudaqq: Fixed-point fractional library routines. 39961 (line 1343) 39962 * __satfractudasa: Fixed-point fractional library routines. 39963 (line 1350) 39964 * __satfractudasq: Fixed-point fractional library routines. 39965 (line 1345) 39966 * __satfractudata: Fixed-point fractional library routines. 39967 (line 1353) 39968 * __satfractudaudq: Fixed-point fractional library routines. 39969 (line 1361) 39970 * __satfractudauha2: Fixed-point fractional library routines. 39971 (line 1363) 39972 * __satfractudauhq: Fixed-point fractional library routines. 39973 (line 1357) 39974 * __satfractudauqq: Fixed-point fractional library routines. 39975 (line 1355) 39976 * __satfractudausa2: Fixed-point fractional library routines. 39977 (line 1365) 39978 * __satfractudausq: Fixed-point fractional library routines. 39979 (line 1359) 39980 * __satfractudauta2: Fixed-point fractional library routines. 39981 (line 1367) 39982 * __satfractudqda: Fixed-point fractional library routines. 39983 (line 1276) 39984 * __satfractudqdq: Fixed-point fractional library routines. 39985 (line 1271) 39986 * __satfractudqha: Fixed-point fractional library routines. 39987 (line 1273) 39988 * __satfractudqhq: Fixed-point fractional library routines. 39989 (line 1267) 39990 * __satfractudqqq: Fixed-point fractional library routines. 39991 (line 1266) 39992 * __satfractudqsa: Fixed-point fractional library routines. 39993 (line 1274) 39994 * __satfractudqsq: Fixed-point fractional library routines. 39995 (line 1269) 39996 * __satfractudqta: Fixed-point fractional library routines. 39997 (line 1278) 39998 * __satfractudquda: Fixed-point fractional library routines. 39999 (line 1290) 40000 * __satfractudquha: Fixed-point fractional library routines. 40001 (line 1286) 40002 * __satfractudquhq2: Fixed-point fractional library routines. 40003 (line 1282) 40004 * __satfractudquqq2: Fixed-point fractional library routines. 40005 (line 1280) 40006 * __satfractudqusa: Fixed-point fractional library routines. 40007 (line 1288) 40008 * __satfractudqusq2: Fixed-point fractional library routines. 40009 (line 1284) 40010 * __satfractudquta: Fixed-point fractional library routines. 40011 (line 1292) 40012 * __satfractuhada: Fixed-point fractional library routines. 40013 (line 1304) 40014 * __satfractuhadq: Fixed-point fractional library routines. 40015 (line 1299) 40016 * __satfractuhaha: Fixed-point fractional library routines. 40017 (line 1301) 40018 * __satfractuhahq: Fixed-point fractional library routines. 40019 (line 1295) 40020 * __satfractuhaqq: Fixed-point fractional library routines. 40021 (line 1294) 40022 * __satfractuhasa: Fixed-point fractional library routines. 40023 (line 1302) 40024 * __satfractuhasq: Fixed-point fractional library routines. 40025 (line 1297) 40026 * __satfractuhata: Fixed-point fractional library routines. 40027 (line 1306) 40028 * __satfractuhauda2: Fixed-point fractional library routines. 40029 (line 1318) 40030 * __satfractuhaudq: Fixed-point fractional library routines. 40031 (line 1314) 40032 * __satfractuhauhq: Fixed-point fractional library routines. 40033 (line 1310) 40034 * __satfractuhauqq: Fixed-point fractional library routines. 40035 (line 1308) 40036 * __satfractuhausa2: Fixed-point fractional library routines. 40037 (line 1316) 40038 * __satfractuhausq: Fixed-point fractional library routines. 40039 (line 1312) 40040 * __satfractuhauta2: Fixed-point fractional library routines. 40041 (line 1320) 40042 * __satfractuhqda: Fixed-point fractional library routines. 40043 (line 1224) 40044 * __satfractuhqdq: Fixed-point fractional library routines. 40045 (line 1221) 40046 * __satfractuhqha: Fixed-point fractional library routines. 40047 (line 1222) 40048 * __satfractuhqhq: Fixed-point fractional library routines. 40049 (line 1219) 40050 * __satfractuhqqq: Fixed-point fractional library routines. 40051 (line 1218) 40052 * __satfractuhqsa: Fixed-point fractional library routines. 40053 (line 1223) 40054 * __satfractuhqsq: Fixed-point fractional library routines. 40055 (line 1220) 40056 * __satfractuhqta: Fixed-point fractional library routines. 40057 (line 1225) 40058 * __satfractuhquda: Fixed-point fractional library routines. 40059 (line 1236) 40060 * __satfractuhqudq2: Fixed-point fractional library routines. 40061 (line 1231) 40062 * __satfractuhquha: Fixed-point fractional library routines. 40063 (line 1233) 40064 * __satfractuhquqq2: Fixed-point fractional library routines. 40065 (line 1227) 40066 * __satfractuhqusa: Fixed-point fractional library routines. 40067 (line 1234) 40068 * __satfractuhqusq2: Fixed-point fractional library routines. 40069 (line 1229) 40070 * __satfractuhquta: Fixed-point fractional library routines. 40071 (line 1238) 40072 * __satfractunsdida: Fixed-point fractional library routines. 40073 (line 1834) 40074 * __satfractunsdidq: Fixed-point fractional library routines. 40075 (line 1831) 40076 * __satfractunsdiha: Fixed-point fractional library routines. 40077 (line 1832) 40078 * __satfractunsdihq: Fixed-point fractional library routines. 40079 (line 1828) 40080 * __satfractunsdiqq: Fixed-point fractional library routines. 40081 (line 1827) 40082 * __satfractunsdisa: Fixed-point fractional library routines. 40083 (line 1833) 40084 * __satfractunsdisq: Fixed-point fractional library routines. 40085 (line 1829) 40086 * __satfractunsdita: Fixed-point fractional library routines. 40087 (line 1836) 40088 * __satfractunsdiuda: Fixed-point fractional library routines. 40089 (line 1850) 40090 * __satfractunsdiudq: Fixed-point fractional library routines. 40091 (line 1844) 40092 * __satfractunsdiuha: Fixed-point fractional library routines. 40093 (line 1846) 40094 * __satfractunsdiuhq: Fixed-point fractional library routines. 40095 (line 1840) 40096 * __satfractunsdiuqq: Fixed-point fractional library routines. 40097 (line 1838) 40098 * __satfractunsdiusa: Fixed-point fractional library routines. 40099 (line 1848) 40100 * __satfractunsdiusq: Fixed-point fractional library routines. 40101 (line 1842) 40102 * __satfractunsdiuta: Fixed-point fractional library routines. 40103 (line 1852) 40104 * __satfractunshida: Fixed-point fractional library routines. 40105 (line 1786) 40106 * __satfractunshidq: Fixed-point fractional library routines. 40107 (line 1783) 40108 * __satfractunshiha: Fixed-point fractional library routines. 40109 (line 1784) 40110 * __satfractunshihq: Fixed-point fractional library routines. 40111 (line 1780) 40112 * __satfractunshiqq: Fixed-point fractional library routines. 40113 (line 1779) 40114 * __satfractunshisa: Fixed-point fractional library routines. 40115 (line 1785) 40116 * __satfractunshisq: Fixed-point fractional library routines. 40117 (line 1781) 40118 * __satfractunshita: Fixed-point fractional library routines. 40119 (line 1788) 40120 * __satfractunshiuda: Fixed-point fractional library routines. 40121 (line 1802) 40122 * __satfractunshiudq: Fixed-point fractional library routines. 40123 (line 1796) 40124 * __satfractunshiuha: Fixed-point fractional library routines. 40125 (line 1798) 40126 * __satfractunshiuhq: Fixed-point fractional library routines. 40127 (line 1792) 40128 * __satfractunshiuqq: Fixed-point fractional library routines. 40129 (line 1790) 40130 * __satfractunshiusa: Fixed-point fractional library routines. 40131 (line 1800) 40132 * __satfractunshiusq: Fixed-point fractional library routines. 40133 (line 1794) 40134 * __satfractunshiuta: Fixed-point fractional library routines. 40135 (line 1804) 40136 * __satfractunsqida: Fixed-point fractional library routines. 40137 (line 1760) 40138 * __satfractunsqidq: Fixed-point fractional library routines. 40139 (line 1757) 40140 * __satfractunsqiha: Fixed-point fractional library routines. 40141 (line 1758) 40142 * __satfractunsqihq: Fixed-point fractional library routines. 40143 (line 1754) 40144 * __satfractunsqiqq: Fixed-point fractional library routines. 40145 (line 1753) 40146 * __satfractunsqisa: Fixed-point fractional library routines. 40147 (line 1759) 40148 * __satfractunsqisq: Fixed-point fractional library routines. 40149 (line 1755) 40150 * __satfractunsqita: Fixed-point fractional library routines. 40151 (line 1762) 40152 * __satfractunsqiuda: Fixed-point fractional library routines. 40153 (line 1776) 40154 * __satfractunsqiudq: Fixed-point fractional library routines. 40155 (line 1770) 40156 * __satfractunsqiuha: Fixed-point fractional library routines. 40157 (line 1772) 40158 * __satfractunsqiuhq: Fixed-point fractional library routines. 40159 (line 1766) 40160 * __satfractunsqiuqq: Fixed-point fractional library routines. 40161 (line 1764) 40162 * __satfractunsqiusa: Fixed-point fractional library routines. 40163 (line 1774) 40164 * __satfractunsqiusq: Fixed-point fractional library routines. 40165 (line 1768) 40166 * __satfractunsqiuta: Fixed-point fractional library routines. 40167 (line 1778) 40168 * __satfractunssida: Fixed-point fractional library routines. 40169 (line 1811) 40170 * __satfractunssidq: Fixed-point fractional library routines. 40171 (line 1808) 40172 * __satfractunssiha: Fixed-point fractional library routines. 40173 (line 1809) 40174 * __satfractunssihq: Fixed-point fractional library routines. 40175 (line 1806) 40176 * __satfractunssiqq: Fixed-point fractional library routines. 40177 (line 1805) 40178 * __satfractunssisa: Fixed-point fractional library routines. 40179 (line 1810) 40180 * __satfractunssisq: Fixed-point fractional library routines. 40181 (line 1807) 40182 * __satfractunssita: Fixed-point fractional library routines. 40183 (line 1812) 40184 * __satfractunssiuda: Fixed-point fractional library routines. 40185 (line 1824) 40186 * __satfractunssiudq: Fixed-point fractional library routines. 40187 (line 1819) 40188 * __satfractunssiuha: Fixed-point fractional library routines. 40189 (line 1821) 40190 * __satfractunssiuhq: Fixed-point fractional library routines. 40191 (line 1815) 40192 * __satfractunssiuqq: Fixed-point fractional library routines. 40193 (line 1814) 40194 * __satfractunssiusa: Fixed-point fractional library routines. 40195 (line 1822) 40196 * __satfractunssiusq: Fixed-point fractional library routines. 40197 (line 1817) 40198 * __satfractunssiuta: Fixed-point fractional library routines. 40199 (line 1826) 40200 * __satfractunstida: Fixed-point fractional library routines. 40201 (line 1864) 40202 * __satfractunstidq: Fixed-point fractional library routines. 40203 (line 1859) 40204 * __satfractunstiha: Fixed-point fractional library routines. 40205 (line 1861) 40206 * __satfractunstihq: Fixed-point fractional library routines. 40207 (line 1855) 40208 * __satfractunstiqq: Fixed-point fractional library routines. 40209 (line 1854) 40210 * __satfractunstisa: Fixed-point fractional library routines. 40211 (line 1862) 40212 * __satfractunstisq: Fixed-point fractional library routines. 40213 (line 1857) 40214 * __satfractunstita: Fixed-point fractional library routines. 40215 (line 1866) 40216 * __satfractunstiuda: Fixed-point fractional library routines. 40217 (line 1880) 40218 * __satfractunstiudq: Fixed-point fractional library routines. 40219 (line 1874) 40220 * __satfractunstiuha: Fixed-point fractional library routines. 40221 (line 1876) 40222 * __satfractunstiuhq: Fixed-point fractional library routines. 40223 (line 1870) 40224 * __satfractunstiuqq: Fixed-point fractional library routines. 40225 (line 1868) 40226 * __satfractunstiusa: Fixed-point fractional library routines. 40227 (line 1878) 40228 * __satfractunstiusq: Fixed-point fractional library routines. 40229 (line 1872) 40230 * __satfractunstiuta: Fixed-point fractional library routines. 40231 (line 1882) 40232 * __satfractuqqda: Fixed-point fractional library routines. 40233 (line 1201) 40234 * __satfractuqqdq: Fixed-point fractional library routines. 40235 (line 1196) 40236 * __satfractuqqha: Fixed-point fractional library routines. 40237 (line 1198) 40238 * __satfractuqqhq: Fixed-point fractional library routines. 40239 (line 1192) 40240 * __satfractuqqqq: Fixed-point fractional library routines. 40241 (line 1191) 40242 * __satfractuqqsa: Fixed-point fractional library routines. 40243 (line 1199) 40244 * __satfractuqqsq: Fixed-point fractional library routines. 40245 (line 1194) 40246 * __satfractuqqta: Fixed-point fractional library routines. 40247 (line 1203) 40248 * __satfractuqquda: Fixed-point fractional library routines. 40249 (line 1215) 40250 * __satfractuqqudq2: Fixed-point fractional library routines. 40251 (line 1209) 40252 * __satfractuqquha: Fixed-point fractional library routines. 40253 (line 1211) 40254 * __satfractuqquhq2: Fixed-point fractional library routines. 40255 (line 1205) 40256 * __satfractuqqusa: Fixed-point fractional library routines. 40257 (line 1213) 40258 * __satfractuqqusq2: Fixed-point fractional library routines. 40259 (line 1207) 40260 * __satfractuqquta: Fixed-point fractional library routines. 40261 (line 1217) 40262 * __satfractusada: Fixed-point fractional library routines. 40263 (line 1327) 40264 * __satfractusadq: Fixed-point fractional library routines. 40265 (line 1324) 40266 * __satfractusaha: Fixed-point fractional library routines. 40267 (line 1325) 40268 * __satfractusahq: Fixed-point fractional library routines. 40269 (line 1322) 40270 * __satfractusaqq: Fixed-point fractional library routines. 40271 (line 1321) 40272 * __satfractusasa: Fixed-point fractional library routines. 40273 (line 1326) 40274 * __satfractusasq: Fixed-point fractional library routines. 40275 (line 1323) 40276 * __satfractusata: Fixed-point fractional library routines. 40277 (line 1328) 40278 * __satfractusauda2: Fixed-point fractional library routines. 40279 (line 1339) 40280 * __satfractusaudq: Fixed-point fractional library routines. 40281 (line 1335) 40282 * __satfractusauha2: Fixed-point fractional library routines. 40283 (line 1337) 40284 * __satfractusauhq: Fixed-point fractional library routines. 40285 (line 1331) 40286 * __satfractusauqq: Fixed-point fractional library routines. 40287 (line 1330) 40288 * __satfractusausq: Fixed-point fractional library routines. 40289 (line 1333) 40290 * __satfractusauta2: Fixed-point fractional library routines. 40291 (line 1341) 40292 * __satfractusqda: Fixed-point fractional library routines. 40293 (line 1248) 40294 * __satfractusqdq: Fixed-point fractional library routines. 40295 (line 1244) 40296 * __satfractusqha: Fixed-point fractional library routines. 40297 (line 1246) 40298 * __satfractusqhq: Fixed-point fractional library routines. 40299 (line 1241) 40300 * __satfractusqqq: Fixed-point fractional library routines. 40301 (line 1240) 40302 * __satfractusqsa: Fixed-point fractional library routines. 40303 (line 1247) 40304 * __satfractusqsq: Fixed-point fractional library routines. 40305 (line 1242) 40306 * __satfractusqta: Fixed-point fractional library routines. 40307 (line 1250) 40308 * __satfractusquda: Fixed-point fractional library routines. 40309 (line 1262) 40310 * __satfractusqudq2: Fixed-point fractional library routines. 40311 (line 1256) 40312 * __satfractusquha: Fixed-point fractional library routines. 40313 (line 1258) 40314 * __satfractusquhq2: Fixed-point fractional library routines. 40315 (line 1254) 40316 * __satfractusquqq2: Fixed-point fractional library routines. 40317 (line 1252) 40318 * __satfractusqusa: Fixed-point fractional library routines. 40319 (line 1260) 40320 * __satfractusquta: Fixed-point fractional library routines. 40321 (line 1264) 40322 * __satfractutada: Fixed-point fractional library routines. 40323 (line 1379) 40324 * __satfractutadq: Fixed-point fractional library routines. 40325 (line 1374) 40326 * __satfractutaha: Fixed-point fractional library routines. 40327 (line 1376) 40328 * __satfractutahq: Fixed-point fractional library routines. 40329 (line 1370) 40330 * __satfractutaqq: Fixed-point fractional library routines. 40331 (line 1369) 40332 * __satfractutasa: Fixed-point fractional library routines. 40333 (line 1377) 40334 * __satfractutasq: Fixed-point fractional library routines. 40335 (line 1372) 40336 * __satfractutata: Fixed-point fractional library routines. 40337 (line 1381) 40338 * __satfractutauda2: Fixed-point fractional library routines. 40339 (line 1395) 40340 * __satfractutaudq: Fixed-point fractional library routines. 40341 (line 1389) 40342 * __satfractutauha2: Fixed-point fractional library routines. 40343 (line 1391) 40344 * __satfractutauhq: Fixed-point fractional library routines. 40345 (line 1385) 40346 * __satfractutauqq: Fixed-point fractional library routines. 40347 (line 1383) 40348 * __satfractutausa2: Fixed-point fractional library routines. 40349 (line 1393) 40350 * __satfractutausq: Fixed-point fractional library routines. 40351 (line 1387) 40352 * __ssaddda3: Fixed-point fractional library routines. 40353 (line 67) 40354 * __ssadddq3: Fixed-point fractional library routines. 40355 (line 63) 40356 * __ssaddha3: Fixed-point fractional library routines. 40357 (line 65) 40358 * __ssaddhq3: Fixed-point fractional library routines. 40359 (line 60) 40360 * __ssaddqq3: Fixed-point fractional library routines. 40361 (line 59) 40362 * __ssaddsa3: Fixed-point fractional library routines. 40363 (line 66) 40364 * __ssaddsq3: Fixed-point fractional library routines. 40365 (line 61) 40366 * __ssaddta3: Fixed-point fractional library routines. 40367 (line 69) 40368 * __ssashlda3: Fixed-point fractional library routines. 40369 (line 402) 40370 * __ssashldq3: Fixed-point fractional library routines. 40371 (line 399) 40372 * __ssashlha3: Fixed-point fractional library routines. 40373 (line 400) 40374 * __ssashlhq3: Fixed-point fractional library routines. 40375 (line 396) 40376 * __ssashlsa3: Fixed-point fractional library routines. 40377 (line 401) 40378 * __ssashlsq3: Fixed-point fractional library routines. 40379 (line 397) 40380 * __ssashlta3: Fixed-point fractional library routines. 40381 (line 404) 40382 * __ssdivda3: Fixed-point fractional library routines. 40383 (line 261) 40384 * __ssdivdq3: Fixed-point fractional library routines. 40385 (line 257) 40386 * __ssdivha3: Fixed-point fractional library routines. 40387 (line 259) 40388 * __ssdivhq3: Fixed-point fractional library routines. 40389 (line 254) 40390 * __ssdivqq3: Fixed-point fractional library routines. 40391 (line 253) 40392 * __ssdivsa3: Fixed-point fractional library routines. 40393 (line 260) 40394 * __ssdivsq3: Fixed-point fractional library routines. 40395 (line 255) 40396 * __ssdivta3: Fixed-point fractional library routines. 40397 (line 263) 40398 * __ssmulda3: Fixed-point fractional library routines. 40399 (line 193) 40400 * __ssmuldq3: Fixed-point fractional library routines. 40401 (line 189) 40402 * __ssmulha3: Fixed-point fractional library routines. 40403 (line 191) 40404 * __ssmulhq3: Fixed-point fractional library routines. 40405 (line 186) 40406 * __ssmulqq3: Fixed-point fractional library routines. 40407 (line 185) 40408 * __ssmulsa3: Fixed-point fractional library routines. 40409 (line 192) 40410 * __ssmulsq3: Fixed-point fractional library routines. 40411 (line 187) 40412 * __ssmulta3: Fixed-point fractional library routines. 40413 (line 195) 40414 * __ssnegda2: Fixed-point fractional library routines. 40415 (line 316) 40416 * __ssnegdq2: Fixed-point fractional library routines. 40417 (line 313) 40418 * __ssnegha2: Fixed-point fractional library routines. 40419 (line 314) 40420 * __ssneghq2: Fixed-point fractional library routines. 40421 (line 311) 40422 * __ssnegqq2: Fixed-point fractional library routines. 40423 (line 310) 40424 * __ssnegsa2: Fixed-point fractional library routines. 40425 (line 315) 40426 * __ssnegsq2: Fixed-point fractional library routines. 40427 (line 312) 40428 * __ssnegta2: Fixed-point fractional library routines. 40429 (line 317) 40430 * __sssubda3: Fixed-point fractional library routines. 40431 (line 129) 40432 * __sssubdq3: Fixed-point fractional library routines. 40433 (line 125) 40434 * __sssubha3: Fixed-point fractional library routines. 40435 (line 127) 40436 * __sssubhq3: Fixed-point fractional library routines. 40437 (line 122) 40438 * __sssubqq3: Fixed-point fractional library routines. 40439 (line 121) 40440 * __sssubsa3: Fixed-point fractional library routines. 40441 (line 128) 40442 * __sssubsq3: Fixed-point fractional library routines. 40443 (line 123) 40444 * __sssubta3: Fixed-point fractional library routines. 40445 (line 131) 40446 * __subda3: Fixed-point fractional library routines. 40447 (line 107) 40448 * __subdf3: Soft float library routines. 40449 (line 31) 40450 * __subdq3: Fixed-point fractional library routines. 40451 (line 95) 40452 * __subha3: Fixed-point fractional library routines. 40453 (line 105) 40454 * __subhq3: Fixed-point fractional library routines. 40455 (line 92) 40456 * __subqq3: Fixed-point fractional library routines. 40457 (line 91) 40458 * __subsa3: Fixed-point fractional library routines. 40459 (line 106) 40460 * __subsf3: Soft float library routines. 40461 (line 30) 40462 * __subsq3: Fixed-point fractional library routines. 40463 (line 93) 40464 * __subta3: Fixed-point fractional library routines. 40465 (line 109) 40466 * __subtf3: Soft float library routines. 40467 (line 33) 40468 * __subuda3: Fixed-point fractional library routines. 40469 (line 115) 40470 * __subudq3: Fixed-point fractional library routines. 40471 (line 103) 40472 * __subuha3: Fixed-point fractional library routines. 40473 (line 111) 40474 * __subuhq3: Fixed-point fractional library routines. 40475 (line 99) 40476 * __subuqq3: Fixed-point fractional library routines. 40477 (line 97) 40478 * __subusa3: Fixed-point fractional library routines. 40479 (line 113) 40480 * __subusq3: Fixed-point fractional library routines. 40481 (line 101) 40482 * __subuta3: Fixed-point fractional library routines. 40483 (line 117) 40484 * __subvdi3: Integer library routines. 40485 (line 123) 40486 * __subvsi3: Integer library routines. 40487 (line 122) 40488 * __subxf3: Soft float library routines. 40489 (line 35) 40490 * __truncdfsf2: Soft float library routines. 40491 (line 76) 40492 * __trunctfdf2: Soft float library routines. 40493 (line 73) 40494 * __trunctfsf2: Soft float library routines. 40495 (line 75) 40496 * __truncxfdf2: Soft float library routines. 40497 (line 72) 40498 * __truncxfsf2: Soft float library routines. 40499 (line 74) 40500 * __ucmpdi2: Integer library routines. 40501 (line 93) 40502 * __ucmpti2: Integer library routines. 40503 (line 95) 40504 * __udivdi3: Integer library routines. 40505 (line 54) 40506 * __udivmoddi3: Integer library routines. 40507 (line 61) 40508 * __udivsi3: Integer library routines. 40509 (line 52) 40510 * __udivti3: Integer library routines. 40511 (line 63) 40512 * __udivuda3: Fixed-point fractional library routines. 40513 (line 246) 40514 * __udivudq3: Fixed-point fractional library routines. 40515 (line 240) 40516 * __udivuha3: Fixed-point fractional library routines. 40517 (line 242) 40518 * __udivuhq3: Fixed-point fractional library routines. 40519 (line 236) 40520 * __udivuqq3: Fixed-point fractional library routines. 40521 (line 234) 40522 * __udivusa3: Fixed-point fractional library routines. 40523 (line 244) 40524 * __udivusq3: Fixed-point fractional library routines. 40525 (line 238) 40526 * __udivuta3: Fixed-point fractional library routines. 40527 (line 248) 40528 * __umoddi3: Integer library routines. 40529 (line 71) 40530 * __umodsi3: Integer library routines. 40531 (line 69) 40532 * __umodti3: Integer library routines. 40533 (line 73) 40534 * __unorddf2: Soft float library routines. 40535 (line 173) 40536 * __unordsf2: Soft float library routines. 40537 (line 172) 40538 * __unordtf2: Soft float library routines. 40539 (line 174) 40540 * __usadduda3: Fixed-point fractional library routines. 40541 (line 85) 40542 * __usaddudq3: Fixed-point fractional library routines. 40543 (line 79) 40544 * __usadduha3: Fixed-point fractional library routines. 40545 (line 81) 40546 * __usadduhq3: Fixed-point fractional library routines. 40547 (line 75) 40548 * __usadduqq3: Fixed-point fractional library routines. 40549 (line 73) 40550 * __usaddusa3: Fixed-point fractional library routines. 40551 (line 83) 40552 * __usaddusq3: Fixed-point fractional library routines. 40553 (line 77) 40554 * __usadduta3: Fixed-point fractional library routines. 40555 (line 87) 40556 * __usashluda3: Fixed-point fractional library routines. 40557 (line 421) 40558 * __usashludq3: Fixed-point fractional library routines. 40559 (line 415) 40560 * __usashluha3: Fixed-point fractional library routines. 40561 (line 417) 40562 * __usashluhq3: Fixed-point fractional library routines. 40563 (line 411) 40564 * __usashluqq3: Fixed-point fractional library routines. 40565 (line 409) 40566 * __usashlusa3: Fixed-point fractional library routines. 40567 (line 419) 40568 * __usashlusq3: Fixed-point fractional library routines. 40569 (line 413) 40570 * __usashluta3: Fixed-point fractional library routines. 40571 (line 423) 40572 * __usdivuda3: Fixed-point fractional library routines. 40573 (line 280) 40574 * __usdivudq3: Fixed-point fractional library routines. 40575 (line 274) 40576 * __usdivuha3: Fixed-point fractional library routines. 40577 (line 276) 40578 * __usdivuhq3: Fixed-point fractional library routines. 40579 (line 270) 40580 * __usdivuqq3: Fixed-point fractional library routines. 40581 (line 268) 40582 * __usdivusa3: Fixed-point fractional library routines. 40583 (line 278) 40584 * __usdivusq3: Fixed-point fractional library routines. 40585 (line 272) 40586 * __usdivuta3: Fixed-point fractional library routines. 40587 (line 282) 40588 * __usmuluda3: Fixed-point fractional library routines. 40589 (line 212) 40590 * __usmuludq3: Fixed-point fractional library routines. 40591 (line 206) 40592 * __usmuluha3: Fixed-point fractional library routines. 40593 (line 208) 40594 * __usmuluhq3: Fixed-point fractional library routines. 40595 (line 202) 40596 * __usmuluqq3: Fixed-point fractional library routines. 40597 (line 200) 40598 * __usmulusa3: Fixed-point fractional library routines. 40599 (line 210) 40600 * __usmulusq3: Fixed-point fractional library routines. 40601 (line 204) 40602 * __usmuluta3: Fixed-point fractional library routines. 40603 (line 214) 40604 * __usneguda2: Fixed-point fractional library routines. 40605 (line 331) 40606 * __usnegudq2: Fixed-point fractional library routines. 40607 (line 326) 40608 * __usneguha2: Fixed-point fractional library routines. 40609 (line 328) 40610 * __usneguhq2: Fixed-point fractional library routines. 40611 (line 322) 40612 * __usneguqq2: Fixed-point fractional library routines. 40613 (line 321) 40614 * __usnegusa2: Fixed-point fractional library routines. 40615 (line 329) 40616 * __usnegusq2: Fixed-point fractional library routines. 40617 (line 324) 40618 * __usneguta2: Fixed-point fractional library routines. 40619 (line 333) 40620 * __ussubuda3: Fixed-point fractional library routines. 40621 (line 148) 40622 * __ussubudq3: Fixed-point fractional library routines. 40623 (line 142) 40624 * __ussubuha3: Fixed-point fractional library routines. 40625 (line 144) 40626 * __ussubuhq3: Fixed-point fractional library routines. 40627 (line 138) 40628 * __ussubuqq3: Fixed-point fractional library routines. 40629 (line 136) 40630 * __ussubusa3: Fixed-point fractional library routines. 40631 (line 146) 40632 * __ussubusq3: Fixed-point fractional library routines. 40633 (line 140) 40634 * __ussubuta3: Fixed-point fractional library routines. 40635 (line 150) 40636 * abort: Portability. (line 21) 40637 * abs: Arithmetic. (line 195) 40638 * abs and attributes: Expressions. (line 64) 40639 * ABS_EXPR: Expression trees. (line 6) 40640 * absence_set: Processor pipeline description. 40641 (line 220) 40642 * absM2 instruction pattern: Standard Names. (line 452) 40643 * absolute value: Arithmetic. (line 195) 40644 * access to operands: Accessors. (line 6) 40645 * access to special operands: Special Accessors. (line 6) 40646 * accessors: Accessors. (line 6) 40647 * ACCUM_TYPE_SIZE: Type Layout. (line 88) 40648 * ACCUMULATE_OUTGOING_ARGS: Stack Arguments. (line 46) 40649 * ACCUMULATE_OUTGOING_ARGS and stack frames: Function Entry. (line 135) 40650 * ADA_LONG_TYPE_SIZE: Type Layout. (line 26) 40651 * Adding a new GIMPLE statement code: Adding a new GIMPLE statement code. 40652 (line 6) 40653 * ADDITIONAL_REGISTER_NAMES: Instruction Output. (line 15) 40654 * addM3 instruction pattern: Standard Names. (line 216) 40655 * addMODEcc instruction pattern: Standard Names. (line 904) 40656 * addr_diff_vec: Side Effects. (line 302) 40657 * addr_diff_vec, length of: Insn Lengths. (line 26) 40658 * ADDR_EXPR: Expression trees. (line 6) 40659 * addr_vec: Side Effects. (line 297) 40660 * addr_vec, length of: Insn Lengths. (line 26) 40661 * address constraints: Simple Constraints. (line 154) 40662 * address_operand <1>: Machine-Independent Predicates. 40663 (line 63) 40664 * address_operand: Simple Constraints. (line 158) 40665 * addressing modes: Addressing Modes. (line 6) 40666 * ADJUST_FIELD_ALIGN: Storage Layout. (line 201) 40667 * ADJUST_INSN_LENGTH: Insn Lengths. (line 35) 40668 * ADJUST_REG_ALLOC_ORDER: Allocation Order. (line 23) 40669 * AGGR_INIT_EXPR: Expression trees. (line 6) 40670 * aggregates as return values: Aggregate Return. (line 6) 40671 * alias: Alias analysis. (line 6) 40672 * ALL_COP_ADDITIONAL_REGISTER_NAMES: MIPS Coprocessors. (line 32) 40673 * ALL_REGS: Register Classes. (line 17) 40674 * allocate_stack instruction pattern: Standard Names. (line 1227) 40675 * alternate entry points: Insns. (line 140) 40676 * anchored addresses: Anchored Addresses. (line 6) 40677 * and: Arithmetic. (line 153) 40678 * and and attributes: Expressions. (line 50) 40679 * and, canonicalization of: Insn Canonicalizations. 40680 (line 57) 40681 * andM3 instruction pattern: Standard Names. (line 222) 40682 * annotations: Annotations. (line 6) 40683 * APPLY_RESULT_SIZE: Scalar Return. (line 107) 40684 * ARG_POINTER_CFA_OFFSET: Frame Layout. (line 194) 40685 * ARG_POINTER_REGNUM: Frame Registers. (line 41) 40686 * ARG_POINTER_REGNUM and virtual registers: Regs and Memory. (line 65) 40687 * arg_pointer_rtx: Frame Registers. (line 85) 40688 * ARGS_GROW_DOWNWARD: Frame Layout. (line 35) 40689 * argument passing: Interface. (line 36) 40690 * arguments in registers: Register Arguments. (line 6) 40691 * arguments on stack: Stack Arguments. (line 6) 40692 * arithmetic library: Soft float library routines. 40693 (line 6) 40694 * arithmetic shift: Arithmetic. (line 168) 40695 * arithmetic shift with signed saturation: Arithmetic. (line 168) 40696 * arithmetic shift with unsigned saturation: Arithmetic. (line 168) 40697 * arithmetic, in RTL: Arithmetic. (line 6) 40698 * ARITHMETIC_TYPE_P: Types. (line 76) 40699 * array: Types. (line 6) 40700 * ARRAY_RANGE_REF: Expression trees. (line 6) 40701 * ARRAY_REF: Expression trees. (line 6) 40702 * ARRAY_TYPE: Types. (line 6) 40703 * AS_NEEDS_DASH_FOR_PIPED_INPUT: Driver. (line 151) 40704 * ashift: Arithmetic. (line 168) 40705 * ashift and attributes: Expressions. (line 64) 40706 * ashiftrt: Arithmetic. (line 185) 40707 * ashiftrt and attributes: Expressions. (line 64) 40708 * ashlM3 instruction pattern: Standard Names. (line 431) 40709 * ashrM3 instruction pattern: Standard Names. (line 441) 40710 * ASM_APP_OFF: File Framework. (line 61) 40711 * ASM_APP_ON: File Framework. (line 54) 40712 * ASM_COMMENT_START: File Framework. (line 49) 40713 * ASM_DECLARE_CLASS_REFERENCE: Label Output. (line 436) 40714 * ASM_DECLARE_CONSTANT_NAME: Label Output. (line 128) 40715 * ASM_DECLARE_FUNCTION_NAME: Label Output. (line 87) 40716 * ASM_DECLARE_FUNCTION_SIZE: Label Output. (line 101) 40717 * ASM_DECLARE_OBJECT_NAME: Label Output. (line 114) 40718 * ASM_DECLARE_REGISTER_GLOBAL: Label Output. (line 143) 40719 * ASM_DECLARE_UNRESOLVED_REFERENCE: Label Output. (line 442) 40720 * ASM_FINAL_SPEC: Driver. (line 144) 40721 * ASM_FINISH_DECLARE_OBJECT: Label Output. (line 151) 40722 * ASM_FORMAT_PRIVATE_NAME: Label Output. (line 354) 40723 * asm_fprintf: Instruction Output. (line 123) 40724 * ASM_FPRINTF_EXTENSIONS: Instruction Output. (line 134) 40725 * ASM_GENERATE_INTERNAL_LABEL: Label Output. (line 338) 40726 * asm_input: Side Effects. (line 284) 40727 * asm_input and /v: Flags. (line 94) 40728 * ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX: Exception Handling. (line 82) 40729 * ASM_NO_SKIP_IN_TEXT: Alignment Output. (line 72) 40730 * asm_noperands: Insns. (line 266) 40731 * asm_operands and /v: Flags. (line 94) 40732 * asm_operands, RTL sharing: Sharing. (line 45) 40733 * asm_operands, usage: Assembler. (line 6) 40734 * ASM_OUTPUT_ADDR_DIFF_ELT: Dispatch Tables. (line 9) 40735 * ASM_OUTPUT_ADDR_VEC_ELT: Dispatch Tables. (line 26) 40736 * ASM_OUTPUT_ALIGN: Alignment Output. (line 79) 40737 * ASM_OUTPUT_ALIGN_WITH_NOP: Alignment Output. (line 84) 40738 * ASM_OUTPUT_ALIGNED_BSS: Uninitialized Data. (line 64) 40739 * ASM_OUTPUT_ALIGNED_COMMON: Uninitialized Data. (line 23) 40740 * ASM_OUTPUT_ALIGNED_DECL_COMMON: Uninitialized Data. (line 31) 40741 * ASM_OUTPUT_ALIGNED_DECL_LOCAL: Uninitialized Data. (line 95) 40742 * ASM_OUTPUT_ALIGNED_LOCAL: Uninitialized Data. (line 87) 40743 * ASM_OUTPUT_ASCII: Data Output. (line 50) 40744 * ASM_OUTPUT_BSS: Uninitialized Data. (line 39) 40745 * ASM_OUTPUT_CASE_END: Dispatch Tables. (line 51) 40746 * ASM_OUTPUT_CASE_LABEL: Dispatch Tables. (line 38) 40747 * ASM_OUTPUT_COMMON: Uninitialized Data. (line 10) 40748 * ASM_OUTPUT_DEBUG_LABEL: Label Output. (line 326) 40749 * ASM_OUTPUT_DEF: Label Output. (line 375) 40750 * ASM_OUTPUT_DEF_FROM_DECLS: Label Output. (line 383) 40751 * ASM_OUTPUT_DWARF_DELTA: SDB and DWARF. (line 42) 40752 * ASM_OUTPUT_DWARF_OFFSET: SDB and DWARF. (line 46) 40753 * ASM_OUTPUT_DWARF_PCREL: SDB and DWARF. (line 52) 40754 * ASM_OUTPUT_EXTERNAL: Label Output. (line 264) 40755 * ASM_OUTPUT_FDESC: Data Output. (line 59) 40756 * ASM_OUTPUT_IDENT: File Framework. (line 83) 40757 * ASM_OUTPUT_INTERNAL_LABEL: Label Output. (line 17) 40758 * ASM_OUTPUT_LABEL: Label Output. (line 9) 40759 * ASM_OUTPUT_LABEL_REF: Label Output. (line 299) 40760 * ASM_OUTPUT_LABELREF: Label Output. (line 285) 40761 * ASM_OUTPUT_LOCAL: Uninitialized Data. (line 74) 40762 * ASM_OUTPUT_MAX_SKIP_ALIGN: Alignment Output. (line 88) 40763 * ASM_OUTPUT_MEASURED_SIZE: Label Output. (line 41) 40764 * ASM_OUTPUT_OPCODE: Instruction Output. (line 21) 40765 * ASM_OUTPUT_POOL_EPILOGUE: Data Output. (line 109) 40766 * ASM_OUTPUT_POOL_PROLOGUE: Data Output. (line 72) 40767 * ASM_OUTPUT_REG_POP: Instruction Output. (line 178) 40768 * ASM_OUTPUT_REG_PUSH: Instruction Output. (line 173) 40769 * ASM_OUTPUT_SIZE_DIRECTIVE: Label Output. (line 35) 40770 * ASM_OUTPUT_SKIP: Alignment Output. (line 66) 40771 * ASM_OUTPUT_SOURCE_FILENAME: File Framework. (line 68) 40772 * ASM_OUTPUT_SPECIAL_POOL_ENTRY: Data Output. (line 84) 40773 * ASM_OUTPUT_SYMBOL_REF: Label Output. (line 292) 40774 * ASM_OUTPUT_TYPE_DIRECTIVE: Label Output. (line 77) 40775 * ASM_OUTPUT_WEAK_ALIAS: Label Output. (line 401) 40776 * ASM_OUTPUT_WEAKREF: Label Output. (line 203) 40777 * ASM_PREFERRED_EH_DATA_FORMAT: Exception Handling. (line 67) 40778 * ASM_SPEC: Driver. (line 136) 40779 * ASM_STABD_OP: DBX Options. (line 36) 40780 * ASM_STABN_OP: DBX Options. (line 43) 40781 * ASM_STABS_OP: DBX Options. (line 29) 40782 * ASM_WEAKEN_DECL: Label Output. (line 195) 40783 * ASM_WEAKEN_LABEL: Label Output. (line 182) 40784 * assemble_name: Label Output. (line 8) 40785 * assemble_name_raw: Label Output. (line 16) 40786 * assembler format: File Framework. (line 6) 40787 * assembler instructions in RTL: Assembler. (line 6) 40788 * ASSEMBLER_DIALECT: Instruction Output. (line 146) 40789 * assigning attribute values to insns: Tagging Insns. (line 6) 40790 * assignment operator: Function Basics. (line 6) 40791 * asterisk in template: Output Statement. (line 29) 40792 * atan2M3 instruction pattern: Standard Names. (line 522) 40793 * attr <1>: Tagging Insns. (line 54) 40794 * attr: Expressions. (line 154) 40795 * attr_flag: Expressions. (line 119) 40796 * attribute expressions: Expressions. (line 6) 40797 * attribute specifications: Attr Example. (line 6) 40798 * attribute specifications example: Attr Example. (line 6) 40799 * ATTRIBUTE_ALIGNED_VALUE: Storage Layout. (line 183) 40800 * attributes: Attributes. (line 6) 40801 * attributes, defining: Defining Attributes. 40802 (line 6) 40803 * attributes, target-specific: Target Attributes. (line 6) 40804 * autoincrement addressing, availability: Portability. (line 21) 40805 * autoincrement/decrement addressing: Simple Constraints. (line 30) 40806 * automata_option: Processor pipeline description. 40807 (line 301) 40808 * automaton based pipeline description: Processor pipeline description. 40809 (line 6) 40810 * automaton based scheduler: Processor pipeline description. 40811 (line 6) 40812 * AVOID_CCMODE_COPIES: Values in Registers. 40813 (line 153) 40814 * backslash: Output Template. (line 46) 40815 * barrier: Insns. (line 160) 40816 * barrier and /f: Flags. (line 125) 40817 * barrier and /v: Flags. (line 44) 40818 * BASE_REG_CLASS: Register Classes. (line 107) 40819 * basic block: Basic Blocks. (line 6) 40820 * basic-block.h: Control Flow. (line 6) 40821 * BASIC_BLOCK: Basic Blocks. (line 19) 40822 * basic_block: Basic Blocks. (line 6) 40823 * BB_HEAD, BB_END: Maintaining the CFG. 40824 (line 88) 40825 * bb_seq: GIMPLE sequences. (line 73) 40826 * bCOND instruction pattern: Standard Names. (line 941) 40827 * BIGGEST_ALIGNMENT: Storage Layout. (line 173) 40828 * BIGGEST_FIELD_ALIGNMENT: Storage Layout. (line 194) 40829 * BImode: Machine Modes. (line 22) 40830 * BIND_EXPR: Expression trees. (line 6) 40831 * BINFO_TYPE: Classes. (line 6) 40832 * bit-fields: Bit-Fields. (line 6) 40833 * BIT_AND_EXPR: Expression trees. (line 6) 40834 * BIT_IOR_EXPR: Expression trees. (line 6) 40835 * BIT_NOT_EXPR: Expression trees. (line 6) 40836 * BIT_XOR_EXPR: Expression trees. (line 6) 40837 * BITFIELD_NBYTES_LIMITED: Storage Layout. (line 382) 40838 * BITS_BIG_ENDIAN: Storage Layout. (line 12) 40839 * BITS_BIG_ENDIAN, effect on sign_extract: Bit-Fields. (line 8) 40840 * BITS_PER_UNIT: Storage Layout. (line 52) 40841 * BITS_PER_WORD: Storage Layout. (line 57) 40842 * bitwise complement: Arithmetic. (line 149) 40843 * bitwise exclusive-or: Arithmetic. (line 163) 40844 * bitwise inclusive-or: Arithmetic. (line 158) 40845 * bitwise logical-and: Arithmetic. (line 153) 40846 * BLKmode: Machine Modes. (line 183) 40847 * BLKmode, and function return values: Calls. (line 23) 40848 * block statement iterators <1>: Maintaining the CFG. 40849 (line 45) 40850 * block statement iterators: Basic Blocks. (line 68) 40851 * BLOCK_FOR_INSN, bb_for_stmt: Maintaining the CFG. 40852 (line 40) 40853 * BLOCK_REG_PADDING: Register Arguments. (line 229) 40854 * blockage instruction pattern: Standard Names. (line 1408) 40855 * Blocks: Blocks. (line 6) 40856 * bool <1>: Sections. (line 280) 40857 * bool <2>: Exception Region Output. 40858 (line 60) 40859 * bool: Sections. (line 293) 40860 * BOOL_TYPE_SIZE: Type Layout. (line 44) 40861 * BOOLEAN_TYPE: Types. (line 6) 40862 * branch prediction: Profile information. 40863 (line 24) 40864 * BRANCH_COST: Costs. (line 52) 40865 * break_out_memory_refs: Addressing Modes. (line 130) 40866 * BREAK_STMT: Function Bodies. (line 6) 40867 * bsi_commit_edge_inserts: Maintaining the CFG. 40868 (line 118) 40869 * bsi_end_p: Maintaining the CFG. 40870 (line 60) 40871 * bsi_insert_after: Maintaining the CFG. 40872 (line 72) 40873 * bsi_insert_before: Maintaining the CFG. 40874 (line 78) 40875 * bsi_insert_on_edge: Maintaining the CFG. 40876 (line 118) 40877 * bsi_last: Maintaining the CFG. 40878 (line 56) 40879 * bsi_next: Maintaining the CFG. 40880 (line 64) 40881 * bsi_prev: Maintaining the CFG. 40882 (line 68) 40883 * bsi_remove: Maintaining the CFG. 40884 (line 84) 40885 * bsi_start: Maintaining the CFG. 40886 (line 52) 40887 * BSS_SECTION_ASM_OP: Sections. (line 68) 40888 * bswap: Arithmetic. (line 232) 40889 * btruncM2 instruction pattern: Standard Names. (line 540) 40890 * builtin_longjmp instruction pattern: Standard Names. (line 1313) 40891 * builtin_setjmp_receiver instruction pattern: Standard Names. 40892 (line 1303) 40893 * builtin_setjmp_setup instruction pattern: Standard Names. (line 1292) 40894 * byte_mode: Machine Modes. (line 336) 40895 * BYTES_BIG_ENDIAN: Storage Layout. (line 24) 40896 * BYTES_BIG_ENDIAN, effect on subreg: Regs and Memory. (line 221) 40897 * C statements for assembler output: Output Statement. (line 6) 40898 * C/C++ Internal Representation: Trees. (line 6) 40899 * C99 math functions, implicit usage: Library Calls. (line 76) 40900 * C_COMMON_OVERRIDE_OPTIONS: Run-time Target. (line 114) 40901 * c_register_pragma: Misc. (line 404) 40902 * c_register_pragma_with_expansion: Misc. (line 406) 40903 * call <1>: Flags. (line 234) 40904 * call: Side Effects. (line 86) 40905 * call instruction pattern: Standard Names. (line 974) 40906 * call usage: Calls. (line 10) 40907 * call, in call_insn: Flags. (line 33) 40908 * call, in mem: Flags. (line 99) 40909 * call-clobbered register: Register Basics. (line 46) 40910 * call-saved register: Register Basics. (line 53) 40911 * call-used register: Register Basics. (line 35) 40912 * CALL_EXPR: Expression trees. (line 6) 40913 * call_insn: Insns. (line 95) 40914 * call_insn and /c: Flags. (line 33) 40915 * call_insn and /f: Flags. (line 125) 40916 * call_insn and /i: Flags. (line 24) 40917 * call_insn and /j: Flags. (line 179) 40918 * call_insn and /s: Flags. (line 166) 40919 * call_insn and /u: Flags. (line 19) 40920 * call_insn and /u or /i: Flags. (line 29) 40921 * call_insn and /v: Flags. (line 44) 40922 * CALL_INSN_FUNCTION_USAGE: Insns. (line 101) 40923 * call_pop instruction pattern: Standard Names. (line 1002) 40924 * CALL_POPS_ARGS: Stack Arguments. (line 130) 40925 * CALL_REALLY_USED_REGISTERS: Register Basics. (line 46) 40926 * CALL_USED_REGISTERS: Register Basics. (line 35) 40927 * call_used_regs: Register Basics. (line 59) 40928 * call_value instruction pattern: Standard Names. (line 994) 40929 * call_value_pop instruction pattern: Standard Names. (line 1002) 40930 * CALLER_SAVE_PROFITABLE: Caller Saves. (line 11) 40931 * calling conventions: Stack and Calling. (line 6) 40932 * calling functions in RTL: Calls. (line 6) 40933 * can_create_pseudo_p: Standard Names. (line 75) 40934 * CAN_DEBUG_WITHOUT_FP: Run-time Target. (line 146) 40935 * CAN_ELIMINATE: Elimination. (line 71) 40936 * can_fallthru: Basic Blocks. (line 57) 40937 * canadian: Configure Terms. (line 6) 40938 * CANNOT_CHANGE_MODE_CLASS: Register Classes. (line 481) 40939 * CANNOT_CHANGE_MODE_CLASS and subreg semantics: Regs and Memory. 40940 (line 280) 40941 * canonicalization of instructions: Insn Canonicalizations. 40942 (line 6) 40943 * CANONICALIZE_COMPARISON: Condition Code. (line 84) 40944 * canonicalize_funcptr_for_compare instruction pattern: Standard Names. 40945 (line 1158) 40946 * CASE_USE_BIT_TESTS: Misc. (line 54) 40947 * CASE_VALUES_THRESHOLD: Misc. (line 47) 40948 * CASE_VECTOR_MODE: Misc. (line 27) 40949 * CASE_VECTOR_PC_RELATIVE: Misc. (line 40) 40950 * CASE_VECTOR_SHORTEN_MODE: Misc. (line 31) 40951 * casesi instruction pattern: Standard Names. (line 1082) 40952 * cbranchMODE4 instruction pattern: Standard Names. (line 963) 40953 * cc0: Regs and Memory. (line 307) 40954 * cc0, RTL sharing: Sharing. (line 27) 40955 * cc0_rtx: Regs and Memory. (line 333) 40956 * CC1_SPEC: Driver. (line 118) 40957 * CC1PLUS_SPEC: Driver. (line 126) 40958 * cc_status: Condition Code. (line 8) 40959 * CC_STATUS_MDEP: Condition Code. (line 19) 40960 * CC_STATUS_MDEP_INIT: Condition Code. (line 25) 40961 * CCmode: Machine Modes. (line 176) 40962 * CDImode: Machine Modes. (line 202) 40963 * CEIL_DIV_EXPR: Expression trees. (line 6) 40964 * CEIL_MOD_EXPR: Expression trees. (line 6) 40965 * ceilM2 instruction pattern: Standard Names. (line 556) 40966 * CFA_FRAME_BASE_OFFSET: Frame Layout. (line 226) 40967 * CFG, Control Flow Graph: Control Flow. (line 6) 40968 * cfghooks.h: Maintaining the CFG. 40969 (line 6) 40970 * cgraph_finalize_function: Parsing pass. (line 52) 40971 * chain_circular: GTY Options. (line 196) 40972 * chain_next: GTY Options. (line 196) 40973 * chain_prev: GTY Options. (line 196) 40974 * change_address: Standard Names. (line 47) 40975 * CHANGE_DYNAMIC_TYPE_EXPR: Expression trees. (line 6) 40976 * char <1>: GIMPLE_ASM. (line 53) 40977 * char <2>: PCH Target. (line 12) 40978 * char <3>: Misc. (line 685) 40979 * char: Sections. (line 272) 40980 * CHAR_TYPE_SIZE: Type Layout. (line 39) 40981 * check_stack instruction pattern: Standard Names. (line 1245) 40982 * CHImode: Machine Modes. (line 202) 40983 * class: Classes. (line 6) 40984 * class definitions, register: Register Classes. (line 6) 40985 * class preference constraints: Class Preferences. (line 6) 40986 * CLASS_LIKELY_SPILLED_P: Register Classes. (line 452) 40987 * CLASS_MAX_NREGS: Register Classes. (line 469) 40988 * CLASS_TYPE_P: Types. (line 80) 40989 * classes of RTX codes: RTL Classes. (line 6) 40990 * CLASSTYPE_DECLARED_CLASS: Classes. (line 6) 40991 * CLASSTYPE_HAS_MUTABLE: Classes. (line 80) 40992 * CLASSTYPE_NON_POD_P: Classes. (line 85) 40993 * CLEANUP_DECL: Function Bodies. (line 6) 40994 * CLEANUP_EXPR: Function Bodies. (line 6) 40995 * CLEANUP_POINT_EXPR: Expression trees. (line 6) 40996 * CLEANUP_STMT: Function Bodies. (line 6) 40997 * Cleanups: Cleanups. (line 6) 40998 * CLEAR_BY_PIECES_P: Costs. (line 130) 40999 * clear_cache instruction pattern: Standard Names. (line 1553) 41000 * CLEAR_INSN_CACHE: Trampolines. (line 100) 41001 * CLEAR_RATIO: Costs. (line 121) 41002 * clobber: Side Effects. (line 100) 41003 * clz: Arithmetic. (line 208) 41004 * CLZ_DEFINED_VALUE_AT_ZERO: Misc. (line 319) 41005 * clzM2 instruction pattern: Standard Names. (line 621) 41006 * cmpM instruction pattern: Standard Names. (line 654) 41007 * cmpmemM instruction pattern: Standard Names. (line 769) 41008 * cmpstrM instruction pattern: Standard Names. (line 750) 41009 * cmpstrnM instruction pattern: Standard Names. (line 738) 41010 * code generation RTL sequences: Expander Definitions. 41011 (line 6) 41012 * code iterators in .md files: Code Iterators. (line 6) 41013 * code_label: Insns. (line 119) 41014 * code_label and /i: Flags. (line 59) 41015 * code_label and /v: Flags. (line 44) 41016 * CODE_LABEL_NUMBER: Insns. (line 119) 41017 * codes, RTL expression: RTL Objects. (line 47) 41018 * COImode: Machine Modes. (line 202) 41019 * COLLECT2_HOST_INITIALIZATION: Host Misc. (line 32) 41020 * COLLECT_EXPORT_LIST: Misc. (line 767) 41021 * COLLECT_SHARED_FINI_FUNC: Macros for Initialization. 41022 (line 44) 41023 * COLLECT_SHARED_INIT_FUNC: Macros for Initialization. 41024 (line 33) 41025 * commit_edge_insertions: Maintaining the CFG. 41026 (line 118) 41027 * compare: Arithmetic. (line 43) 41028 * compare, canonicalization of: Insn Canonicalizations. 41029 (line 37) 41030 * comparison_operator: Machine-Independent Predicates. 41031 (line 111) 41032 * compiler passes and files: Passes. (line 6) 41033 * complement, bitwise: Arithmetic. (line 149) 41034 * COMPLEX_CST: Expression trees. (line 6) 41035 * COMPLEX_EXPR: Expression trees. (line 6) 41036 * COMPLEX_TYPE: Types. (line 6) 41037 * COMPONENT_REF: Expression trees. (line 6) 41038 * Compound Expressions: Compound Expressions. 41039 (line 6) 41040 * Compound Lvalues: Compound Lvalues. (line 6) 41041 * COMPOUND_EXPR: Expression trees. (line 6) 41042 * COMPOUND_LITERAL_EXPR: Expression trees. (line 6) 41043 * COMPOUND_LITERAL_EXPR_DECL: Expression trees. (line 608) 41044 * COMPOUND_LITERAL_EXPR_DECL_STMT: Expression trees. (line 608) 41045 * computed jump: Edges. (line 128) 41046 * computing the length of an insn: Insn Lengths. (line 6) 41047 * concat: Regs and Memory. (line 385) 41048 * concatn: Regs and Memory. (line 391) 41049 * cond: Comparisons. (line 90) 41050 * cond and attributes: Expressions. (line 37) 41051 * cond_exec: Side Effects. (line 248) 41052 * COND_EXPR: Expression trees. (line 6) 41053 * condition code register: Regs and Memory. (line 307) 41054 * condition code status: Condition Code. (line 6) 41055 * condition codes: Comparisons. (line 20) 41056 * conditional execution: Conditional Execution. 41057 (line 6) 41058 * Conditional Expressions: Conditional Expressions. 41059 (line 6) 41060 * CONDITIONAL_REGISTER_USAGE: Register Basics. (line 60) 41061 * conditional_trap instruction pattern: Standard Names. (line 1379) 41062 * conditions, in patterns: Patterns. (line 43) 41063 * configuration file <1>: Host Misc. (line 6) 41064 * configuration file: Filesystem. (line 6) 41065 * configure terms: Configure Terms. (line 6) 41066 * CONJ_EXPR: Expression trees. (line 6) 41067 * const: Constants. (line 99) 41068 * const0_rtx: Constants. (line 16) 41069 * CONST0_RTX: Constants. (line 119) 41070 * CONST1_RTX: Constants. (line 119) 41071 * const1_rtx: Constants. (line 16) 41072 * const2_rtx: Constants. (line 16) 41073 * CONST2_RTX: Constants. (line 119) 41074 * CONST_DECL: Declarations. (line 6) 41075 * const_double: Constants. (line 32) 41076 * const_double, RTL sharing: Sharing. (line 29) 41077 * CONST_DOUBLE_LOW: Constants. (line 39) 41078 * CONST_DOUBLE_OK_FOR_CONSTRAINT_P: Old Constraints. (line 69) 41079 * CONST_DOUBLE_OK_FOR_LETTER_P: Old Constraints. (line 54) 41080 * const_double_operand: Machine-Independent Predicates. 41081 (line 21) 41082 * const_fixed: Constants. (line 52) 41083 * const_int: Constants. (line 8) 41084 * const_int and attribute tests: Expressions. (line 47) 41085 * const_int and attributes: Expressions. (line 10) 41086 * const_int, RTL sharing: Sharing. (line 23) 41087 * const_int_operand: Machine-Independent Predicates. 41088 (line 16) 41089 * CONST_OK_FOR_CONSTRAINT_P: Old Constraints. (line 49) 41090 * CONST_OK_FOR_LETTER_P: Old Constraints. (line 40) 41091 * const_string: Constants. (line 71) 41092 * const_string and attributes: Expressions. (line 20) 41093 * const_true_rtx: Constants. (line 26) 41094 * const_vector: Constants. (line 59) 41095 * const_vector, RTL sharing: Sharing. (line 32) 41096 * constant attributes: Constant Attributes. 41097 (line 6) 41098 * constant definitions: Constant Definitions. 41099 (line 6) 41100 * CONSTANT_ADDRESS_P: Addressing Modes. (line 29) 41101 * CONSTANT_ALIGNMENT: Storage Layout. (line 241) 41102 * CONSTANT_P: Addressing Modes. (line 35) 41103 * CONSTANT_POOL_ADDRESS_P: Flags. (line 10) 41104 * CONSTANT_POOL_BEFORE_FUNCTION: Data Output. (line 64) 41105 * constants in constraints: Simple Constraints. (line 60) 41106 * constm1_rtx: Constants. (line 16) 41107 * constraint modifier characters: Modifiers. (line 6) 41108 * constraint, matching: Simple Constraints. (line 132) 41109 * CONSTRAINT_LEN: Old Constraints. (line 12) 41110 * constraint_num: C Constraint Interface. 41111 (line 38) 41112 * constraint_satisfied_p: C Constraint Interface. 41113 (line 54) 41114 * constraints: Constraints. (line 6) 41115 * constraints, defining: Define Constraints. (line 6) 41116 * constraints, defining, obsolete method: Old Constraints. (line 6) 41117 * constraints, machine specific: Machine Constraints. 41118 (line 6) 41119 * constraints, testing: C Constraint Interface. 41120 (line 6) 41121 * CONSTRUCTOR: Expression trees. (line 6) 41122 * constructor: Function Basics. (line 6) 41123 * constructors, automatic calls: Collect2. (line 15) 41124 * constructors, output of: Initialization. (line 6) 41125 * container: Containers. (line 6) 41126 * CONTINUE_STMT: Function Bodies. (line 6) 41127 * contributors: Contributors. (line 6) 41128 * controlling register usage: Register Basics. (line 76) 41129 * controlling the compilation driver: Driver. (line 6) 41130 * conventions, run-time: Interface. (line 6) 41131 * conversions: Conversions. (line 6) 41132 * CONVERT_EXPR: Expression trees. (line 6) 41133 * copy constructor: Function Basics. (line 6) 41134 * copy_rtx: Addressing Modes. (line 182) 41135 * copy_rtx_if_shared: Sharing. (line 64) 41136 * copysignM3 instruction pattern: Standard Names. (line 602) 41137 * cosM2 instruction pattern: Standard Names. (line 481) 41138 * costs of instructions: Costs. (line 6) 41139 * CP_INTEGRAL_TYPE: Types. (line 72) 41140 * cp_namespace_decls: Namespaces. (line 44) 41141 * CP_TYPE_CONST_NON_VOLATILE_P: Types. (line 45) 41142 * CP_TYPE_CONST_P: Types. (line 36) 41143 * CP_TYPE_QUALS: Types. (line 6) 41144 * CP_TYPE_RESTRICT_P: Types. (line 42) 41145 * CP_TYPE_VOLATILE_P: Types. (line 39) 41146 * CPLUSPLUS_CPP_SPEC: Driver. (line 113) 41147 * CPP_SPEC: Driver. (line 106) 41148 * CQImode: Machine Modes. (line 202) 41149 * cross compilation and floating point: Floating Point. (line 6) 41150 * CRT_CALL_STATIC_FUNCTION: Sections. (line 112) 41151 * CRTSTUFF_T_CFLAGS: Target Fragment. (line 35) 41152 * CRTSTUFF_T_CFLAGS_S: Target Fragment. (line 39) 41153 * CSImode: Machine Modes. (line 202) 41154 * CTImode: Machine Modes. (line 202) 41155 * ctz: Arithmetic. (line 216) 41156 * CTZ_DEFINED_VALUE_AT_ZERO: Misc. (line 320) 41157 * ctzM2 instruction pattern: Standard Names. (line 630) 41158 * CUMULATIVE_ARGS: Register Arguments. (line 127) 41159 * current_function_epilogue_delay_list: Function Entry. (line 181) 41160 * current_function_is_leaf: Leaf Functions. (line 51) 41161 * current_function_outgoing_args_size: Stack Arguments. (line 45) 41162 * current_function_pops_args: Function Entry. (line 106) 41163 * current_function_pretend_args_size: Function Entry. (line 112) 41164 * current_function_uses_only_leaf_regs: Leaf Functions. (line 51) 41165 * current_insn_predicate: Conditional Execution. 41166 (line 26) 41167 * DAmode: Machine Modes. (line 152) 41168 * data bypass: Processor pipeline description. 41169 (line 106) 41170 * data dependence delays: Processor pipeline description. 41171 (line 6) 41172 * Data Dependency Analysis: Dependency analysis. 41173 (line 6) 41174 * data structures: Per-Function Data. (line 6) 41175 * DATA_ALIGNMENT: Storage Layout. (line 228) 41176 * DATA_SECTION_ASM_OP: Sections. (line 53) 41177 * DBR_OUTPUT_SEQEND: Instruction Output. (line 107) 41178 * dbr_sequence_length: Instruction Output. (line 106) 41179 * DBX_BLOCKS_FUNCTION_RELATIVE: DBX Options. (line 103) 41180 * DBX_CONTIN_CHAR: DBX Options. (line 66) 41181 * DBX_CONTIN_LENGTH: DBX Options. (line 56) 41182 * DBX_DEBUGGING_INFO: DBX Options. (line 9) 41183 * DBX_FUNCTION_FIRST: DBX Options. (line 97) 41184 * DBX_LINES_FUNCTION_RELATIVE: DBX Options. (line 109) 41185 * DBX_NO_XREFS: DBX Options. (line 50) 41186 * DBX_OUTPUT_LBRAC: DBX Hooks. (line 9) 41187 * DBX_OUTPUT_MAIN_SOURCE_FILE_END: File Names and DBX. (line 34) 41188 * DBX_OUTPUT_MAIN_SOURCE_FILENAME: File Names and DBX. (line 9) 41189 * DBX_OUTPUT_NFUN: DBX Hooks. (line 18) 41190 * DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END: File Names and DBX. 41191 (line 42) 41192 * DBX_OUTPUT_RBRAC: DBX Hooks. (line 15) 41193 * DBX_OUTPUT_SOURCE_LINE: DBX Hooks. (line 22) 41194 * DBX_REGISTER_NUMBER: All Debuggers. (line 9) 41195 * DBX_REGPARM_STABS_CODE: DBX Options. (line 87) 41196 * DBX_REGPARM_STABS_LETTER: DBX Options. (line 92) 41197 * DBX_STATIC_CONST_VAR_CODE: DBX Options. (line 82) 41198 * DBX_STATIC_STAB_DATA_SECTION: DBX Options. (line 73) 41199 * DBX_TYPE_DECL_STABS_CODE: DBX Options. (line 78) 41200 * DBX_USE_BINCL: DBX Options. (line 115) 41201 * DCmode: Machine Modes. (line 197) 41202 * DDmode: Machine Modes. (line 90) 41203 * De Morgan's law: Insn Canonicalizations. 41204 (line 57) 41205 * dead_or_set_p: define_peephole. (line 65) 41206 * DEBUG_SYMS_TEXT: DBX Options. (line 25) 41207 * DEBUGGER_ARG_OFFSET: All Debuggers. (line 37) 41208 * DEBUGGER_AUTO_OFFSET: All Debuggers. (line 28) 41209 * decimal float library: Decimal float library routines. 41210 (line 6) 41211 * DECL_ALIGN: Declarations. (line 6) 41212 * DECL_ANTICIPATED: Function Basics. (line 48) 41213 * DECL_ARGUMENTS: Function Basics. (line 163) 41214 * DECL_ARRAY_DELETE_OPERATOR_P: Function Basics. (line 184) 41215 * DECL_ARTIFICIAL <1>: Working with declarations. 41216 (line 24) 41217 * DECL_ARTIFICIAL: Function Basics. (line 155) 41218 * DECL_ASSEMBLER_NAME: Function Basics. (line 21) 41219 * DECL_ATTRIBUTES: Attributes. (line 22) 41220 * DECL_BASE_CONSTRUCTOR_P: Function Basics. (line 94) 41221 * DECL_CLASS_SCOPE_P: Working with declarations. 41222 (line 41) 41223 * DECL_COMPLETE_CONSTRUCTOR_P: Function Basics. (line 90) 41224 * DECL_COMPLETE_DESTRUCTOR_P: Function Basics. (line 104) 41225 * DECL_CONST_MEMFUNC_P: Function Basics. (line 77) 41226 * DECL_CONSTRUCTOR_P: Function Basics. (line 6) 41227 * DECL_CONTEXT: Namespaces. (line 26) 41228 * DECL_CONV_FN_P: Function Basics. (line 6) 41229 * DECL_COPY_CONSTRUCTOR_P: Function Basics. (line 98) 41230 * DECL_DESTRUCTOR_P: Function Basics. (line 6) 41231 * DECL_EXTERN_C_FUNCTION_P: Function Basics. (line 52) 41232 * DECL_EXTERNAL <1>: Declarations. (line 6) 41233 * DECL_EXTERNAL: Function Basics. (line 38) 41234 * DECL_FUNCTION_MEMBER_P: Function Basics. (line 6) 41235 * DECL_FUNCTION_SCOPE_P: Working with declarations. 41236 (line 44) 41237 * DECL_FUNCTION_SPECIFIC_OPTIMIZATION: Function Basics. (line 194) 41238 * DECL_FUNCTION_SPECIFIC_TARGET: Function Basics. (line 6) 41239 * DECL_GLOBAL_CTOR_P: Function Basics. (line 114) 41240 * DECL_GLOBAL_DTOR_P: Function Basics. (line 6) 41241 * DECL_INITIAL: Declarations. (line 6) 41242 * DECL_LINKONCE_P: Function Basics. (line 6) 41243 * DECL_LOCAL_FUNCTION_P: Function Basics. (line 44) 41244 * DECL_MAIN_P: Function Basics. (line 7) 41245 * DECL_NAME <1>: Namespaces. (line 15) 41246 * DECL_NAME <2>: Function Basics. (line 6) 41247 * DECL_NAME <3>: Working with declarations. 41248 (line 7) 41249 * DECL_NAME: Function Basics. (line 11) 41250 * DECL_NAMESPACE_ALIAS: Namespaces. (line 30) 41251 * DECL_NAMESPACE_SCOPE_P: Working with declarations. 41252 (line 37) 41253 * DECL_NAMESPACE_STD_P: Namespaces. (line 40) 41254 * DECL_NON_THUNK_FUNCTION_P: Function Basics. (line 144) 41255 * DECL_NONCONVERTING_P: Function Basics. (line 86) 41256 * DECL_NONSTATIC_MEMBER_FUNCTION_P: Function Basics. (line 74) 41257 * DECL_OVERLOADED_OPERATOR_P: Function Basics. (line 6) 41258 * DECL_RESULT: Function Basics. (line 168) 41259 * DECL_SIZE: Declarations. (line 6) 41260 * DECL_STATIC_FUNCTION_P: Function Basics. (line 71) 41261 * DECL_STMT: Function Bodies. (line 6) 41262 * DECL_STMT_DECL: Function Bodies. (line 6) 41263 * DECL_THUNK_P: Function Basics. (line 122) 41264 * DECL_VOLATILE_MEMFUNC_P: Function Basics. (line 80) 41265 * declaration: Declarations. (line 6) 41266 * declarations, RTL: RTL Declarations. (line 6) 41267 * DECLARE_LIBRARY_RENAMES: Library Calls. (line 9) 41268 * decrement_and_branch_until_zero instruction pattern: Standard Names. 41269 (line 1120) 41270 * def_optype_d: Manipulating GIMPLE statements. 41271 (line 94) 41272 * default: GTY Options. (line 82) 41273 * default_file_start: File Framework. (line 9) 41274 * DEFAULT_GDB_EXTENSIONS: DBX Options. (line 18) 41275 * DEFAULT_PCC_STRUCT_RETURN: Aggregate Return. (line 34) 41276 * DEFAULT_SIGNED_CHAR: Type Layout. (line 154) 41277 * define_address_constraint: Define Constraints. (line 107) 41278 * define_asm_attributes: Tagging Insns. (line 73) 41279 * define_attr: Defining Attributes. 41280 (line 6) 41281 * define_automaton: Processor pipeline description. 41282 (line 53) 41283 * define_bypass: Processor pipeline description. 41284 (line 197) 41285 * define_code_attr: Code Iterators. (line 6) 41286 * define_code_iterator: Code Iterators. (line 6) 41287 * define_cond_exec: Conditional Execution. 41288 (line 13) 41289 * define_constants: Constant Definitions. 41290 (line 6) 41291 * define_constraint: Define Constraints. (line 48) 41292 * define_cpu_unit: Processor pipeline description. 41293 (line 68) 41294 * define_delay: Delay Slots. (line 25) 41295 * define_expand: Expander Definitions. 41296 (line 11) 41297 * define_insn: Patterns. (line 6) 41298 * define_insn example: Example. (line 6) 41299 * define_insn_and_split: Insn Splitting. (line 170) 41300 * define_insn_reservation: Processor pipeline description. 41301 (line 106) 41302 * define_memory_constraint: Define Constraints. (line 88) 41303 * define_mode_attr: Substitutions. (line 6) 41304 * define_mode_iterator: Defining Mode Iterators. 41305 (line 6) 41306 * define_peephole: define_peephole. (line 6) 41307 * define_peephole2: define_peephole2. (line 6) 41308 * define_predicate: Defining Predicates. 41309 (line 6) 41310 * define_query_cpu_unit: Processor pipeline description. 41311 (line 90) 41312 * define_register_constraint: Define Constraints. (line 28) 41313 * define_reservation: Processor pipeline description. 41314 (line 186) 41315 * define_special_predicate: Defining Predicates. 41316 (line 6) 41317 * define_split: Insn Splitting. (line 32) 41318 * defining attributes and their values: Defining Attributes. 41319 (line 6) 41320 * defining constraints: Define Constraints. (line 6) 41321 * defining constraints, obsolete method: Old Constraints. (line 6) 41322 * defining jump instruction patterns: Jump Patterns. (line 6) 41323 * defining looping instruction patterns: Looping Patterns. (line 6) 41324 * defining peephole optimizers: Peephole Definitions. 41325 (line 6) 41326 * defining predicates: Defining Predicates. 41327 (line 6) 41328 * defining RTL sequences for code generation: Expander Definitions. 41329 (line 6) 41330 * delay slots, defining: Delay Slots. (line 6) 41331 * DELAY_SLOTS_FOR_EPILOGUE: Function Entry. (line 163) 41332 * deletable: GTY Options. (line 150) 41333 * DELETE_IF_ORDINARY: Filesystem. (line 79) 41334 * Dependent Patterns: Dependent Patterns. (line 6) 41335 * desc: GTY Options. (line 82) 41336 * destructor: Function Basics. (line 6) 41337 * destructors, output of: Initialization. (line 6) 41338 * deterministic finite state automaton: Processor pipeline description. 41339 (line 301) 41340 * DF_SIZE: Type Layout. (line 130) 41341 * DFmode: Machine Modes. (line 73) 41342 * digits in constraint: Simple Constraints. (line 120) 41343 * DImode: Machine Modes. (line 45) 41344 * DIR_SEPARATOR: Filesystem. (line 18) 41345 * DIR_SEPARATOR_2: Filesystem. (line 19) 41346 * directory options .md: Including Patterns. (line 44) 41347 * disabling certain registers: Register Basics. (line 76) 41348 * dispatch table: Dispatch Tables. (line 8) 41349 * div: Arithmetic. (line 111) 41350 * div and attributes: Expressions. (line 64) 41351 * division: Arithmetic. (line 131) 41352 * divM3 instruction pattern: Standard Names. (line 222) 41353 * divmodM4 instruction pattern: Standard Names. (line 411) 41354 * DO_BODY: Function Bodies. (line 6) 41355 * DO_COND: Function Bodies. (line 6) 41356 * DO_STMT: Function Bodies. (line 6) 41357 * DOLLARS_IN_IDENTIFIERS: Misc. (line 488) 41358 * doloop_begin instruction pattern: Standard Names. (line 1151) 41359 * doloop_end instruction pattern: Standard Names. (line 1130) 41360 * DONE: Expander Definitions. 41361 (line 74) 41362 * DONT_USE_BUILTIN_SETJMP: Exception Region Output. 41363 (line 70) 41364 * DOUBLE_TYPE_SIZE: Type Layout. (line 53) 41365 * DQmode: Machine Modes. (line 115) 41366 * driver: Driver. (line 6) 41367 * DRIVER_SELF_SPECS: Driver. (line 71) 41368 * DUMPFILE_FORMAT: Filesystem. (line 67) 41369 * DWARF2_ASM_LINE_DEBUG_INFO: SDB and DWARF. (line 36) 41370 * DWARF2_DEBUGGING_INFO: SDB and DWARF. (line 13) 41371 * DWARF2_FRAME_INFO: SDB and DWARF. (line 30) 41372 * DWARF2_FRAME_REG_OUT: Frame Registers. (line 133) 41373 * DWARF2_UNWIND_INFO: Exception Region Output. 41374 (line 40) 41375 * DWARF_ALT_FRAME_RETURN_COLUMN: Frame Layout. (line 152) 41376 * DWARF_CIE_DATA_ALIGNMENT: Exception Region Output. 41377 (line 75) 41378 * DWARF_FRAME_REGISTERS: Frame Registers. (line 93) 41379 * DWARF_FRAME_REGNUM: Frame Registers. (line 125) 41380 * DWARF_REG_TO_UNWIND_COLUMN: Frame Registers. (line 117) 41381 * DWARF_ZERO_REG: Frame Layout. (line 163) 41382 * DYNAMIC_CHAIN_ADDRESS: Frame Layout. (line 92) 41383 * E in constraint: Simple Constraints. (line 79) 41384 * earlyclobber operand: Modifiers. (line 25) 41385 * edge: Edges. (line 6) 41386 * edge in the flow graph: Edges. (line 6) 41387 * edge iterators: Edges. (line 15) 41388 * edge splitting: Maintaining the CFG. 41389 (line 118) 41390 * EDGE_ABNORMAL: Edges. (line 128) 41391 * EDGE_ABNORMAL, EDGE_ABNORMAL_CALL: Edges. (line 171) 41392 * EDGE_ABNORMAL, EDGE_EH: Edges. (line 96) 41393 * EDGE_ABNORMAL, EDGE_SIBCALL: Edges. (line 122) 41394 * EDGE_FALLTHRU, force_nonfallthru: Edges. (line 86) 41395 * EDOM, implicit usage: Library Calls. (line 58) 41396 * EH_FRAME_IN_DATA_SECTION: Exception Region Output. 41397 (line 20) 41398 * EH_FRAME_SECTION_NAME: Exception Region Output. 41399 (line 10) 41400 * eh_return instruction pattern: Standard Names. (line 1319) 41401 * EH_RETURN_DATA_REGNO: Exception Handling. (line 7) 41402 * EH_RETURN_HANDLER_RTX: Exception Handling. (line 39) 41403 * EH_RETURN_STACKADJ_RTX: Exception Handling. (line 22) 41404 * EH_TABLES_CAN_BE_READ_ONLY: Exception Region Output. 41405 (line 29) 41406 * EH_USES: Function Entry. (line 158) 41407 * ei_edge: Edges. (line 43) 41408 * ei_end_p: Edges. (line 27) 41409 * ei_last: Edges. (line 23) 41410 * ei_next: Edges. (line 35) 41411 * ei_one_before_end_p: Edges. (line 31) 41412 * ei_prev: Edges. (line 39) 41413 * ei_safe_safe: Edges. (line 47) 41414 * ei_start: Edges. (line 19) 41415 * ELIGIBLE_FOR_EPILOGUE_DELAY: Function Entry. (line 169) 41416 * ELIMINABLE_REGS: Elimination. (line 44) 41417 * ELSE_CLAUSE: Function Bodies. (line 6) 41418 * Embedded C: Fixed-point fractional library routines. 41419 (line 6) 41420 * EMIT_MODE_SET: Mode Switching. (line 74) 41421 * Empty Statements: Empty Statements. (line 6) 41422 * EMPTY_CLASS_EXPR: Function Bodies. (line 6) 41423 * EMPTY_FIELD_BOUNDARY: Storage Layout. (line 295) 41424 * Emulated TLS: Emulated TLS. (line 6) 41425 * ENABLE_EXECUTE_STACK: Trampolines. (line 110) 41426 * enabled: Disable Insn Alternatives. 41427 (line 6) 41428 * ENDFILE_SPEC: Driver. (line 218) 41429 * endianness: Portability. (line 21) 41430 * ENTRY_BLOCK_PTR, EXIT_BLOCK_PTR: Basic Blocks. (line 28) 41431 * enum machine_mode: Machine Modes. (line 6) 41432 * enum reg_class: Register Classes. (line 65) 41433 * ENUMERAL_TYPE: Types. (line 6) 41434 * epilogue: Function Entry. (line 6) 41435 * epilogue instruction pattern: Standard Names. (line 1351) 41436 * EPILOGUE_USES: Function Entry. (line 152) 41437 * eq: Comparisons. (line 52) 41438 * eq and attributes: Expressions. (line 64) 41439 * eq_attr: Expressions. (line 85) 41440 * EQ_EXPR: Expression trees. (line 6) 41441 * equal: Comparisons. (line 52) 41442 * errno, implicit usage: Library Calls. (line 70) 41443 * EXACT_DIV_EXPR: Expression trees. (line 6) 41444 * examining SSA_NAMEs: SSA. (line 218) 41445 * exception handling <1>: Exception Handling. (line 6) 41446 * exception handling: Edges. (line 96) 41447 * exception_receiver instruction pattern: Standard Names. (line 1283) 41448 * exclamation point: Multi-Alternative. (line 47) 41449 * exclusion_set: Processor pipeline description. 41450 (line 220) 41451 * exclusive-or, bitwise: Arithmetic. (line 163) 41452 * EXIT_EXPR: Expression trees. (line 6) 41453 * EXIT_IGNORE_STACK: Function Entry. (line 140) 41454 * expander definitions: Expander Definitions. 41455 (line 6) 41456 * expM2 instruction pattern: Standard Names. (line 497) 41457 * expr_list: Insns. (line 505) 41458 * EXPR_STMT: Function Bodies. (line 6) 41459 * EXPR_STMT_EXPR: Function Bodies. (line 6) 41460 * expression: Expression trees. (line 6) 41461 * expression codes: RTL Objects. (line 47) 41462 * extendMN2 instruction pattern: Standard Names. (line 826) 41463 * extensible constraints: Simple Constraints. (line 163) 41464 * EXTRA_ADDRESS_CONSTRAINT: Old Constraints. (line 123) 41465 * EXTRA_CONSTRAINT: Old Constraints. (line 74) 41466 * EXTRA_CONSTRAINT_STR: Old Constraints. (line 95) 41467 * EXTRA_MEMORY_CONSTRAINT: Old Constraints. (line 100) 41468 * EXTRA_SPECS: Driver. (line 245) 41469 * extv instruction pattern: Standard Names. (line 862) 41470 * extzv instruction pattern: Standard Names. (line 877) 41471 * F in constraint: Simple Constraints. (line 84) 41472 * FAIL: Expander Definitions. 41473 (line 80) 41474 * fall-thru: Edges. (line 69) 41475 * FATAL_EXIT_CODE: Host Misc. (line 6) 41476 * FDL, GNU Free Documentation License: GNU Free Documentation License. 41477 (line 6) 41478 * features, optional, in system conventions: Run-time Target. 41479 (line 59) 41480 * ffs: Arithmetic. (line 202) 41481 * ffsM2 instruction pattern: Standard Names. (line 611) 41482 * FIELD_DECL: Declarations. (line 6) 41483 * file_end_indicate_exec_stack: File Framework. (line 41) 41484 * files and passes of the compiler: Passes. (line 6) 41485 * files, generated: Files. (line 6) 41486 * final_absence_set: Processor pipeline description. 41487 (line 220) 41488 * FINAL_PRESCAN_INSN: Instruction Output. (line 46) 41489 * final_presence_set: Processor pipeline description. 41490 (line 220) 41491 * final_scan_insn: Function Entry. (line 181) 41492 * final_sequence: Instruction Output. (line 117) 41493 * FIND_BASE_TERM: Addressing Modes. (line 110) 41494 * FINI_ARRAY_SECTION_ASM_OP: Sections. (line 105) 41495 * FINI_SECTION_ASM_OP: Sections. (line 90) 41496 * finite state automaton minimization: Processor pipeline description. 41497 (line 301) 41498 * FIRST_PARM_OFFSET: Frame Layout. (line 67) 41499 * FIRST_PARM_OFFSET and virtual registers: Regs and Memory. (line 65) 41500 * FIRST_PSEUDO_REGISTER: Register Basics. (line 9) 41501 * FIRST_STACK_REG: Stack Registers. (line 23) 41502 * FIRST_VIRTUAL_REGISTER: Regs and Memory. (line 51) 41503 * fix: Conversions. (line 66) 41504 * FIX_TRUNC_EXPR: Expression trees. (line 6) 41505 * fix_truncMN2 instruction pattern: Standard Names. (line 813) 41506 * fixed register: Register Basics. (line 15) 41507 * fixed-point fractional library: Fixed-point fractional library routines. 41508 (line 6) 41509 * FIXED_CONVERT_EXPR: Expression trees. (line 6) 41510 * FIXED_CST: Expression trees. (line 6) 41511 * FIXED_POINT_TYPE: Types. (line 6) 41512 * FIXED_REGISTERS: Register Basics. (line 15) 41513 * fixed_regs: Register Basics. (line 59) 41514 * fixMN2 instruction pattern: Standard Names. (line 793) 41515 * FIXUNS_TRUNC_LIKE_FIX_TRUNC: Misc. (line 100) 41516 * fixuns_truncMN2 instruction pattern: Standard Names. (line 817) 41517 * fixunsMN2 instruction pattern: Standard Names. (line 802) 41518 * flags in RTL expression: Flags. (line 6) 41519 * float: Conversions. (line 58) 41520 * FLOAT_EXPR: Expression trees. (line 6) 41521 * float_extend: Conversions. (line 33) 41522 * FLOAT_LIB_COMPARE_RETURNS_BOOL: Library Calls. (line 25) 41523 * FLOAT_STORE_FLAG_VALUE: Misc. (line 301) 41524 * float_truncate: Conversions. (line 53) 41525 * FLOAT_TYPE_SIZE: Type Layout. (line 49) 41526 * FLOAT_WORDS_BIG_ENDIAN: Storage Layout. (line 43) 41527 * FLOAT_WORDS_BIG_ENDIAN, (lack of) effect on subreg: Regs and Memory. 41528 (line 226) 41529 * floating point and cross compilation: Floating Point. (line 6) 41530 * Floating Point Emulation: Target Fragment. (line 15) 41531 * floating point emulation library, US Software GOFAST: Library Calls. 41532 (line 44) 41533 * floatMN2 instruction pattern: Standard Names. (line 785) 41534 * floatunsMN2 instruction pattern: Standard Names. (line 789) 41535 * FLOOR_DIV_EXPR: Expression trees. (line 6) 41536 * FLOOR_MOD_EXPR: Expression trees. (line 6) 41537 * floorM2 instruction pattern: Standard Names. (line 532) 41538 * flow-insensitive alias analysis: Alias analysis. (line 6) 41539 * flow-sensitive alias analysis: Alias analysis. (line 6) 41540 * fmodM3 instruction pattern: Standard Names. (line 463) 41541 * FOR_BODY: Function Bodies. (line 6) 41542 * FOR_COND: Function Bodies. (line 6) 41543 * FOR_EXPR: Function Bodies. (line 6) 41544 * FOR_INIT_STMT: Function Bodies. (line 6) 41545 * FOR_STMT: Function Bodies. (line 6) 41546 * FORCE_CODE_SECTION_ALIGN: Sections. (line 136) 41547 * force_reg: Standard Names. (line 36) 41548 * fract_convert: Conversions. (line 82) 41549 * FRACT_TYPE_SIZE: Type Layout. (line 68) 41550 * fractional types: Fixed-point fractional library routines. 41551 (line 6) 41552 * fractMN2 instruction pattern: Standard Names. (line 835) 41553 * fractunsMN2 instruction pattern: Standard Names. (line 850) 41554 * frame layout: Frame Layout. (line 6) 41555 * FRAME_ADDR_RTX: Frame Layout. (line 116) 41556 * FRAME_GROWS_DOWNWARD: Frame Layout. (line 31) 41557 * FRAME_GROWS_DOWNWARD and virtual registers: Regs and Memory. 41558 (line 69) 41559 * FRAME_POINTER_CFA_OFFSET: Frame Layout. (line 212) 41560 * frame_pointer_needed: Function Entry. (line 34) 41561 * FRAME_POINTER_REGNUM: Frame Registers. (line 14) 41562 * FRAME_POINTER_REGNUM and virtual registers: Regs and Memory. 41563 (line 74) 41564 * FRAME_POINTER_REQUIRED: Elimination. (line 9) 41565 * frame_pointer_rtx: Frame Registers. (line 85) 41566 * frame_related: Flags. (line 242) 41567 * frame_related, in insn, call_insn, jump_insn, barrier, and set: Flags. 41568 (line 125) 41569 * frame_related, in mem: Flags. (line 103) 41570 * frame_related, in reg: Flags. (line 112) 41571 * frame_related, in symbol_ref: Flags. (line 183) 41572 * frequency, count, BB_FREQ_BASE: Profile information. 41573 (line 30) 41574 * ftruncM2 instruction pattern: Standard Names. (line 808) 41575 * function: Functions. (line 6) 41576 * function body: Function Bodies. (line 6) 41577 * function call conventions: Interface. (line 6) 41578 * function entry and exit: Function Entry. (line 6) 41579 * function entry point, alternate function entry point: Edges. 41580 (line 180) 41581 * function-call insns: Calls. (line 6) 41582 * FUNCTION_ARG: Register Arguments. (line 11) 41583 * FUNCTION_ARG_ADVANCE: Register Arguments. (line 186) 41584 * FUNCTION_ARG_BOUNDARY: Register Arguments. (line 239) 41585 * FUNCTION_ARG_OFFSET: Register Arguments. (line 197) 41586 * FUNCTION_ARG_PADDING: Register Arguments. (line 204) 41587 * FUNCTION_ARG_REGNO_P: Register Arguments. (line 244) 41588 * FUNCTION_BOUNDARY: Storage Layout. (line 170) 41589 * FUNCTION_DECL: Functions. (line 6) 41590 * FUNCTION_INCOMING_ARG: Register Arguments. (line 68) 41591 * FUNCTION_MODE: Misc. (line 356) 41592 * FUNCTION_OUTGOING_VALUE: Scalar Return. (line 56) 41593 * FUNCTION_PROFILER: Profiling. (line 9) 41594 * FUNCTION_TYPE: Types. (line 6) 41595 * FUNCTION_VALUE: Scalar Return. (line 52) 41596 * FUNCTION_VALUE_REGNO_P: Scalar Return. (line 81) 41597 * functions, leaf: Leaf Functions. (line 6) 41598 * fundamental type: Types. (line 6) 41599 * G in constraint: Simple Constraints. (line 88) 41600 * g in constraint: Simple Constraints. (line 110) 41601 * garbage collector, invocation: Invoking the garbage collector. 41602 (line 6) 41603 * GCC and portability: Portability. (line 6) 41604 * GCC_DRIVER_HOST_INITIALIZATION: Host Misc. (line 36) 41605 * gcov_type: Profile information. 41606 (line 41) 41607 * ge: Comparisons. (line 72) 41608 * ge and attributes: Expressions. (line 64) 41609 * GE_EXPR: Expression trees. (line 6) 41610 * GEN_ERRNO_RTX: Library Calls. (line 71) 41611 * gencodes: RTL passes. (line 18) 41612 * general_operand: Machine-Independent Predicates. 41613 (line 105) 41614 * GENERAL_REGS: Register Classes. (line 23) 41615 * generated files: Files. (line 6) 41616 * generating assembler output: Output Statement. (line 6) 41617 * generating insns: RTL Template. (line 6) 41618 * GENERIC <1>: Parsing pass. (line 6) 41619 * GENERIC <2>: GENERIC. (line 6) 41620 * GENERIC: Gimplification pass. 41621 (line 12) 41622 * generic predicates: Machine-Independent Predicates. 41623 (line 6) 41624 * genflags: RTL passes. (line 18) 41625 * get_attr: Expressions. (line 80) 41626 * get_attr_length: Insn Lengths. (line 46) 41627 * GET_CLASS_NARROWEST_MODE: Machine Modes. (line 333) 41628 * GET_CODE: RTL Objects. (line 47) 41629 * get_frame_size: Elimination. (line 31) 41630 * get_insns: Insns. (line 34) 41631 * get_last_insn: Insns. (line 34) 41632 * GET_MODE: Machine Modes. (line 280) 41633 * GET_MODE_ALIGNMENT: Machine Modes. (line 320) 41634 * GET_MODE_BITSIZE: Machine Modes. (line 304) 41635 * GET_MODE_CLASS: Machine Modes. (line 294) 41636 * GET_MODE_FBIT: Machine Modes. (line 311) 41637 * GET_MODE_IBIT: Machine Modes. (line 307) 41638 * GET_MODE_MASK: Machine Modes. (line 315) 41639 * GET_MODE_NAME: Machine Modes. (line 291) 41640 * GET_MODE_NUNITS: Machine Modes. (line 329) 41641 * GET_MODE_SIZE: Machine Modes. (line 301) 41642 * GET_MODE_UNIT_SIZE: Machine Modes. (line 323) 41643 * GET_MODE_WIDER_MODE: Machine Modes. (line 297) 41644 * GET_RTX_CLASS: RTL Classes. (line 6) 41645 * GET_RTX_FORMAT: RTL Classes. (line 130) 41646 * GET_RTX_LENGTH: RTL Classes. (line 127) 41647 * geu: Comparisons. (line 72) 41648 * geu and attributes: Expressions. (line 64) 41649 * GGC: Type Information. (line 6) 41650 * ggc_collect: Invoking the garbage collector. 41651 (line 6) 41652 * GIMPLE <1>: GIMPLE. (line 6) 41653 * GIMPLE <2>: Gimplification pass. 41654 (line 6) 41655 * GIMPLE: Parsing pass. (line 14) 41656 * GIMPLE Exception Handling: GIMPLE Exception Handling. 41657 (line 6) 41658 * GIMPLE instruction set: GIMPLE instruction set. 41659 (line 6) 41660 * GIMPLE sequences: GIMPLE sequences. (line 6) 41661 * gimple_addresses_taken: Manipulating GIMPLE statements. 41662 (line 90) 41663 * GIMPLE_ASM: GIMPLE_ASM. (line 6) 41664 * gimple_asm_clear_volatile: GIMPLE_ASM. (line 63) 41665 * gimple_asm_clobber_op: GIMPLE_ASM. (line 46) 41666 * gimple_asm_input_op: GIMPLE_ASM. (line 30) 41667 * gimple_asm_output_op: GIMPLE_ASM. (line 38) 41668 * gimple_asm_set_clobber_op: GIMPLE_ASM. (line 50) 41669 * gimple_asm_set_input_op: GIMPLE_ASM. (line 34) 41670 * gimple_asm_set_output_op: GIMPLE_ASM. (line 42) 41671 * gimple_asm_set_volatile: GIMPLE_ASM. (line 60) 41672 * gimple_asm_volatile_p: GIMPLE_ASM. (line 57) 41673 * GIMPLE_ASSIGN: GIMPLE_ASSIGN. (line 6) 41674 * gimple_assign_cast_p: GIMPLE_ASSIGN. (line 89) 41675 * gimple_assign_lhs: GIMPLE_ASSIGN. (line 51) 41676 * gimple_assign_rhs1: GIMPLE_ASSIGN. (line 57) 41677 * gimple_assign_rhs2: GIMPLE_ASSIGN. (line 64) 41678 * gimple_assign_set_lhs: GIMPLE_ASSIGN. (line 71) 41679 * gimple_assign_set_rhs1: GIMPLE_ASSIGN. (line 74) 41680 * gimple_assign_set_rhs2: GIMPLE_ASSIGN. (line 85) 41681 * gimple_bb: Manipulating GIMPLE statements. 41682 (line 18) 41683 * GIMPLE_BIND: GIMPLE_BIND. (line 6) 41684 * gimple_bind_add_seq: GIMPLE_BIND. (line 36) 41685 * gimple_bind_add_stmt: GIMPLE_BIND. (line 32) 41686 * gimple_bind_append_vars: GIMPLE_BIND. (line 19) 41687 * gimple_bind_block: GIMPLE_BIND. (line 40) 41688 * gimple_bind_body: GIMPLE_BIND. (line 23) 41689 * gimple_bind_set_block: GIMPLE_BIND. (line 45) 41690 * gimple_bind_set_body: GIMPLE_BIND. (line 28) 41691 * gimple_bind_set_vars: GIMPLE_BIND. (line 15) 41692 * gimple_bind_vars: GIMPLE_BIND. (line 12) 41693 * gimple_block: Manipulating GIMPLE statements. 41694 (line 21) 41695 * gimple_build_asm: GIMPLE_ASM. (line 8) 41696 * gimple_build_asm_vec: GIMPLE_ASM. (line 17) 41697 * gimple_build_assign: GIMPLE_ASSIGN. (line 7) 41698 * gimple_build_assign_with_ops: GIMPLE_ASSIGN. (line 30) 41699 * gimple_build_bind: GIMPLE_BIND. (line 8) 41700 * gimple_build_call: GIMPLE_CALL. (line 8) 41701 * gimple_build_call_from_tree: GIMPLE_CALL. (line 16) 41702 * gimple_build_call_vec: GIMPLE_CALL. (line 25) 41703 * gimple_build_catch: GIMPLE_CATCH. (line 8) 41704 * gimple_build_cdt: GIMPLE_CHANGE_DYNAMIC_TYPE. 41705 (line 7) 41706 * gimple_build_cond: GIMPLE_COND. (line 8) 41707 * gimple_build_cond_from_tree: GIMPLE_COND. (line 16) 41708 * gimple_build_eh_filter: GIMPLE_EH_FILTER. (line 8) 41709 * gimple_build_goto: GIMPLE_LABEL. (line 18) 41710 * gimple_build_label: GIMPLE_LABEL. (line 7) 41711 * gimple_build_nop: GIMPLE_NOP. (line 7) 41712 * gimple_build_omp_atomic_load: GIMPLE_OMP_ATOMIC_LOAD. 41713 (line 8) 41714 * gimple_build_omp_atomic_store: GIMPLE_OMP_ATOMIC_STORE. 41715 (line 7) 41716 * gimple_build_omp_continue: GIMPLE_OMP_CONTINUE. 41717 (line 8) 41718 * gimple_build_omp_critical: GIMPLE_OMP_CRITICAL. 41719 (line 8) 41720 * gimple_build_omp_for: GIMPLE_OMP_FOR. (line 9) 41721 * gimple_build_omp_master: GIMPLE_OMP_MASTER. (line 7) 41722 * gimple_build_omp_ordered: GIMPLE_OMP_ORDERED. (line 7) 41723 * gimple_build_omp_parallel: GIMPLE_OMP_PARALLEL. 41724 (line 8) 41725 * gimple_build_omp_return: GIMPLE_OMP_RETURN. (line 7) 41726 * gimple_build_omp_section: GIMPLE_OMP_SECTION. (line 7) 41727 * gimple_build_omp_sections: GIMPLE_OMP_SECTIONS. 41728 (line 8) 41729 * gimple_build_omp_sections_switch: GIMPLE_OMP_SECTIONS. 41730 (line 14) 41731 * gimple_build_omp_single: GIMPLE_OMP_SINGLE. (line 8) 41732 * gimple_build_resx: GIMPLE_RESX. (line 7) 41733 * gimple_build_return: GIMPLE_RETURN. (line 7) 41734 * gimple_build_switch: GIMPLE_SWITCH. (line 8) 41735 * gimple_build_switch_vec: GIMPLE_SWITCH. (line 16) 41736 * gimple_build_try: GIMPLE_TRY. (line 8) 41737 * gimple_build_wce: GIMPLE_WITH_CLEANUP_EXPR. 41738 (line 7) 41739 * GIMPLE_CALL: GIMPLE_CALL. (line 6) 41740 * gimple_call_arg: GIMPLE_CALL. (line 66) 41741 * gimple_call_cannot_inline_p: GIMPLE_CALL. (line 91) 41742 * gimple_call_chain: GIMPLE_CALL. (line 57) 41743 * gimple_call_copy_skip_args: GIMPLE_CALL. (line 98) 41744 * gimple_call_fn: GIMPLE_CALL. (line 38) 41745 * gimple_call_fndecl: GIMPLE_CALL. (line 46) 41746 * gimple_call_lhs: GIMPLE_CALL. (line 29) 41747 * gimple_call_mark_uninlinable: GIMPLE_CALL. (line 88) 41748 * gimple_call_noreturn_p: GIMPLE_CALL. (line 94) 41749 * gimple_call_return_type: GIMPLE_CALL. (line 54) 41750 * gimple_call_set_arg: GIMPLE_CALL. (line 76) 41751 * gimple_call_set_chain: GIMPLE_CALL. (line 60) 41752 * gimple_call_set_fn: GIMPLE_CALL. (line 42) 41753 * gimple_call_set_fndecl: GIMPLE_CALL. (line 51) 41754 * gimple_call_set_lhs: GIMPLE_CALL. (line 35) 41755 * gimple_call_set_tail: GIMPLE_CALL. (line 80) 41756 * gimple_call_tail_p: GIMPLE_CALL. (line 85) 41757 * GIMPLE_CATCH: GIMPLE_CATCH. (line 6) 41758 * gimple_catch_handler: GIMPLE_CATCH. (line 20) 41759 * gimple_catch_set_handler: GIMPLE_CATCH. (line 28) 41760 * gimple_catch_set_types: GIMPLE_CATCH. (line 24) 41761 * gimple_catch_types: GIMPLE_CATCH. (line 13) 41762 * gimple_cdt_location: GIMPLE_CHANGE_DYNAMIC_TYPE. 41763 (line 24) 41764 * gimple_cdt_new_type: GIMPLE_CHANGE_DYNAMIC_TYPE. 41765 (line 11) 41766 * gimple_cdt_set_location: GIMPLE_CHANGE_DYNAMIC_TYPE. 41767 (line 32) 41768 * gimple_cdt_set_new_type: GIMPLE_CHANGE_DYNAMIC_TYPE. 41769 (line 20) 41770 * GIMPLE_CHANGE_DYNAMIC_TYPE: GIMPLE_CHANGE_DYNAMIC_TYPE. 41771 (line 6) 41772 * gimple_code: Manipulating GIMPLE statements. 41773 (line 15) 41774 * GIMPLE_COND: GIMPLE_COND. (line 6) 41775 * gimple_cond_false_label: GIMPLE_COND. (line 60) 41776 * gimple_cond_lhs: GIMPLE_COND. (line 30) 41777 * gimple_cond_make_false: GIMPLE_COND. (line 64) 41778 * gimple_cond_make_true: GIMPLE_COND. (line 67) 41779 * gimple_cond_rhs: GIMPLE_COND. (line 38) 41780 * gimple_cond_set_code: GIMPLE_COND. (line 26) 41781 * gimple_cond_set_false_label: GIMPLE_COND. (line 56) 41782 * gimple_cond_set_lhs: GIMPLE_COND. (line 34) 41783 * gimple_cond_set_rhs: GIMPLE_COND. (line 42) 41784 * gimple_cond_set_true_label: GIMPLE_COND. (line 51) 41785 * gimple_cond_true_label: GIMPLE_COND. (line 46) 41786 * gimple_copy: Manipulating GIMPLE statements. 41787 (line 147) 41788 * GIMPLE_EH_FILTER: GIMPLE_EH_FILTER. (line 6) 41789 * gimple_eh_filter_failure: GIMPLE_EH_FILTER. (line 19) 41790 * gimple_eh_filter_must_not_throw: GIMPLE_EH_FILTER. (line 33) 41791 * gimple_eh_filter_set_failure: GIMPLE_EH_FILTER. (line 29) 41792 * gimple_eh_filter_set_must_not_throw: GIMPLE_EH_FILTER. (line 37) 41793 * gimple_eh_filter_set_types: GIMPLE_EH_FILTER. (line 24) 41794 * gimple_eh_filter_types: GIMPLE_EH_FILTER. (line 12) 41795 * gimple_expr_type: Manipulating GIMPLE statements. 41796 (line 24) 41797 * gimple_goto_dest: GIMPLE_LABEL. (line 21) 41798 * gimple_goto_set_dest: GIMPLE_LABEL. (line 24) 41799 * gimple_has_mem_ops: Manipulating GIMPLE statements. 41800 (line 72) 41801 * gimple_has_ops: Manipulating GIMPLE statements. 41802 (line 69) 41803 * gimple_has_volatile_ops: Manipulating GIMPLE statements. 41804 (line 134) 41805 * GIMPLE_LABEL: GIMPLE_LABEL. (line 6) 41806 * gimple_label_label: GIMPLE_LABEL. (line 11) 41807 * gimple_label_set_label: GIMPLE_LABEL. (line 14) 41808 * gimple_loaded_syms: Manipulating GIMPLE statements. 41809 (line 122) 41810 * gimple_locus: Manipulating GIMPLE statements. 41811 (line 42) 41812 * gimple_locus_empty_p: Manipulating GIMPLE statements. 41813 (line 48) 41814 * gimple_modified_p: Manipulating GIMPLE statements. 41815 (line 130) 41816 * gimple_no_warning_p: Manipulating GIMPLE statements. 41817 (line 51) 41818 * GIMPLE_NOP: GIMPLE_NOP. (line 6) 41819 * gimple_nop_p: GIMPLE_NOP. (line 10) 41820 * gimple_num_ops <1>: Logical Operators. (line 76) 41821 * gimple_num_ops: Manipulating GIMPLE statements. 41822 (line 75) 41823 * GIMPLE_OMP_ATOMIC_LOAD: GIMPLE_OMP_ATOMIC_LOAD. 41824 (line 6) 41825 * gimple_omp_atomic_load_lhs: GIMPLE_OMP_ATOMIC_LOAD. 41826 (line 17) 41827 * gimple_omp_atomic_load_rhs: GIMPLE_OMP_ATOMIC_LOAD. 41828 (line 24) 41829 * gimple_omp_atomic_load_set_lhs: GIMPLE_OMP_ATOMIC_LOAD. 41830 (line 14) 41831 * gimple_omp_atomic_load_set_rhs: GIMPLE_OMP_ATOMIC_LOAD. 41832 (line 21) 41833 * GIMPLE_OMP_ATOMIC_STORE: GIMPLE_OMP_ATOMIC_STORE. 41834 (line 6) 41835 * gimple_omp_atomic_store_set_val: GIMPLE_OMP_ATOMIC_STORE. 41836 (line 12) 41837 * gimple_omp_atomic_store_val: GIMPLE_OMP_ATOMIC_STORE. 41838 (line 15) 41839 * gimple_omp_body: GIMPLE_OMP_PARALLEL. 41840 (line 24) 41841 * GIMPLE_OMP_CONTINUE: GIMPLE_OMP_CONTINUE. 41842 (line 6) 41843 * gimple_omp_continue_control_def: GIMPLE_OMP_CONTINUE. 41844 (line 13) 41845 * gimple_omp_continue_control_def_ptr: GIMPLE_OMP_CONTINUE. 41846 (line 17) 41847 * gimple_omp_continue_control_use: GIMPLE_OMP_CONTINUE. 41848 (line 24) 41849 * gimple_omp_continue_control_use_ptr: GIMPLE_OMP_CONTINUE. 41850 (line 28) 41851 * gimple_omp_continue_set_control_def: GIMPLE_OMP_CONTINUE. 41852 (line 20) 41853 * gimple_omp_continue_set_control_use: GIMPLE_OMP_CONTINUE. 41854 (line 31) 41855 * GIMPLE_OMP_CRITICAL: GIMPLE_OMP_CRITICAL. 41856 (line 6) 41857 * gimple_omp_critical_name: GIMPLE_OMP_CRITICAL. 41858 (line 13) 41859 * gimple_omp_critical_set_name: GIMPLE_OMP_CRITICAL. 41860 (line 21) 41861 * GIMPLE_OMP_FOR: GIMPLE_OMP_FOR. (line 6) 41862 * gimple_omp_for_clauses: GIMPLE_OMP_FOR. (line 20) 41863 * gimple_omp_for_final: GIMPLE_OMP_FOR. (line 51) 41864 * gimple_omp_for_incr: GIMPLE_OMP_FOR. (line 61) 41865 * gimple_omp_for_index: GIMPLE_OMP_FOR. (line 31) 41866 * gimple_omp_for_initial: GIMPLE_OMP_FOR. (line 41) 41867 * gimple_omp_for_pre_body: GIMPLE_OMP_FOR. (line 70) 41868 * gimple_omp_for_set_clauses: GIMPLE_OMP_FOR. (line 27) 41869 * gimple_omp_for_set_cond: GIMPLE_OMP_FOR. (line 80) 41870 * gimple_omp_for_set_final: GIMPLE_OMP_FOR. (line 58) 41871 * gimple_omp_for_set_incr: GIMPLE_OMP_FOR. (line 67) 41872 * gimple_omp_for_set_index: GIMPLE_OMP_FOR. (line 38) 41873 * gimple_omp_for_set_initial: GIMPLE_OMP_FOR. (line 48) 41874 * gimple_omp_for_set_pre_body: GIMPLE_OMP_FOR. (line 75) 41875 * GIMPLE_OMP_MASTER: GIMPLE_OMP_MASTER. (line 6) 41876 * GIMPLE_OMP_ORDERED: GIMPLE_OMP_ORDERED. (line 6) 41877 * GIMPLE_OMP_PARALLEL: GIMPLE_OMP_PARALLEL. 41878 (line 6) 41879 * gimple_omp_parallel_child_fn: GIMPLE_OMP_PARALLEL. 41880 (line 42) 41881 * gimple_omp_parallel_clauses: GIMPLE_OMP_PARALLEL. 41882 (line 31) 41883 * gimple_omp_parallel_combined_p: GIMPLE_OMP_PARALLEL. 41884 (line 16) 41885 * gimple_omp_parallel_data_arg: GIMPLE_OMP_PARALLEL. 41886 (line 54) 41887 * gimple_omp_parallel_set_child_fn: GIMPLE_OMP_PARALLEL. 41888 (line 51) 41889 * gimple_omp_parallel_set_clauses: GIMPLE_OMP_PARALLEL. 41890 (line 38) 41891 * gimple_omp_parallel_set_combined_p: GIMPLE_OMP_PARALLEL. 41892 (line 20) 41893 * gimple_omp_parallel_set_data_arg: GIMPLE_OMP_PARALLEL. 41894 (line 62) 41895 * GIMPLE_OMP_RETURN: GIMPLE_OMP_RETURN. (line 6) 41896 * gimple_omp_return_nowait_p: GIMPLE_OMP_RETURN. (line 14) 41897 * gimple_omp_return_set_nowait: GIMPLE_OMP_RETURN. (line 11) 41898 * GIMPLE_OMP_SECTION: GIMPLE_OMP_SECTION. (line 6) 41899 * gimple_omp_section_last_p: GIMPLE_OMP_SECTION. (line 12) 41900 * gimple_omp_section_set_last: GIMPLE_OMP_SECTION. (line 16) 41901 * GIMPLE_OMP_SECTIONS: GIMPLE_OMP_SECTIONS. 41902 (line 6) 41903 * gimple_omp_sections_clauses: GIMPLE_OMP_SECTIONS. 41904 (line 30) 41905 * gimple_omp_sections_control: GIMPLE_OMP_SECTIONS. 41906 (line 17) 41907 * gimple_omp_sections_set_clauses: GIMPLE_OMP_SECTIONS. 41908 (line 37) 41909 * gimple_omp_sections_set_control: GIMPLE_OMP_SECTIONS. 41910 (line 26) 41911 * gimple_omp_set_body: GIMPLE_OMP_PARALLEL. 41912 (line 28) 41913 * GIMPLE_OMP_SINGLE: GIMPLE_OMP_SINGLE. (line 6) 41914 * gimple_omp_single_clauses: GIMPLE_OMP_SINGLE. (line 14) 41915 * gimple_omp_single_set_clauses: GIMPLE_OMP_SINGLE. (line 21) 41916 * gimple_op <1>: Logical Operators. (line 79) 41917 * gimple_op: Manipulating GIMPLE statements. 41918 (line 81) 41919 * GIMPLE_PHI: GIMPLE_PHI. (line 6) 41920 * gimple_phi_capacity: GIMPLE_PHI. (line 10) 41921 * gimple_phi_num_args: GIMPLE_PHI. (line 14) 41922 * gimple_phi_result: GIMPLE_PHI. (line 19) 41923 * gimple_phi_set_arg: GIMPLE_PHI. (line 33) 41924 * gimple_phi_set_result: GIMPLE_PHI. (line 25) 41925 * GIMPLE_RESX: GIMPLE_RESX. (line 6) 41926 * gimple_resx_region: GIMPLE_RESX. (line 13) 41927 * gimple_resx_set_region: GIMPLE_RESX. (line 16) 41928 * GIMPLE_RETURN: GIMPLE_RETURN. (line 6) 41929 * gimple_return_retval: GIMPLE_RETURN. (line 10) 41930 * gimple_return_set_retval: GIMPLE_RETURN. (line 14) 41931 * gimple_rhs_class: GIMPLE_ASSIGN. (line 46) 41932 * gimple_seq_add_seq: GIMPLE sequences. (line 32) 41933 * gimple_seq_add_stmt: GIMPLE sequences. (line 26) 41934 * gimple_seq_alloc: GIMPLE sequences. (line 62) 41935 * gimple_seq_copy: GIMPLE sequences. (line 67) 41936 * gimple_seq_deep_copy: GIMPLE sequences. (line 37) 41937 * gimple_seq_empty_p: GIMPLE sequences. (line 70) 41938 * gimple_seq_first: GIMPLE sequences. (line 44) 41939 * gimple_seq_init: GIMPLE sequences. (line 59) 41940 * gimple_seq_last: GIMPLE sequences. (line 47) 41941 * gimple_seq_reverse: GIMPLE sequences. (line 40) 41942 * gimple_seq_set_first: GIMPLE sequences. (line 55) 41943 * gimple_seq_set_last: GIMPLE sequences. (line 51) 41944 * gimple_seq_singleton_p: GIMPLE sequences. (line 79) 41945 * gimple_set_block: Manipulating GIMPLE statements. 41946 (line 39) 41947 * gimple_set_def_ops: Manipulating GIMPLE statements. 41948 (line 98) 41949 * gimple_set_has_volatile_ops: Manipulating GIMPLE statements. 41950 (line 138) 41951 * gimple_set_locus: Manipulating GIMPLE statements. 41952 (line 45) 41953 * gimple_set_op: Manipulating GIMPLE statements. 41954 (line 87) 41955 * gimple_set_plf: Manipulating GIMPLE statements. 41956 (line 62) 41957 * gimple_set_use_ops: Manipulating GIMPLE statements. 41958 (line 105) 41959 * gimple_set_vdef_ops: Manipulating GIMPLE statements. 41960 (line 119) 41961 * gimple_set_visited: Manipulating GIMPLE statements. 41962 (line 55) 41963 * gimple_set_vuse_ops: Manipulating GIMPLE statements. 41964 (line 112) 41965 * gimple_statement_base: Tuple representation. 41966 (line 14) 41967 * gimple_statement_with_ops: Tuple representation. 41968 (line 96) 41969 * gimple_stored_syms: Manipulating GIMPLE statements. 41970 (line 126) 41971 * GIMPLE_SWITCH: GIMPLE_SWITCH. (line 6) 41972 * gimple_switch_default_label: GIMPLE_SWITCH. (line 46) 41973 * gimple_switch_index: GIMPLE_SWITCH. (line 31) 41974 * gimple_switch_label: GIMPLE_SWITCH. (line 37) 41975 * gimple_switch_num_labels: GIMPLE_SWITCH. (line 22) 41976 * gimple_switch_set_default_label: GIMPLE_SWITCH. (line 50) 41977 * gimple_switch_set_index: GIMPLE_SWITCH. (line 34) 41978 * gimple_switch_set_label: GIMPLE_SWITCH. (line 42) 41979 * gimple_switch_set_num_labels: GIMPLE_SWITCH. (line 27) 41980 * GIMPLE_TRY: GIMPLE_TRY. (line 6) 41981 * gimple_try_catch_is_cleanup: GIMPLE_TRY. (line 20) 41982 * gimple_try_cleanup: GIMPLE_TRY. (line 27) 41983 * gimple_try_eval: GIMPLE_TRY. (line 23) 41984 * gimple_try_flags: GIMPLE_TRY. (line 16) 41985 * gimple_try_set_catch_is_cleanup: GIMPLE_TRY. (line 32) 41986 * gimple_try_set_cleanup: GIMPLE_TRY. (line 41) 41987 * gimple_try_set_eval: GIMPLE_TRY. (line 36) 41988 * gimple_visited_p: Manipulating GIMPLE statements. 41989 (line 58) 41990 * gimple_wce_cleanup: GIMPLE_WITH_CLEANUP_EXPR. 41991 (line 11) 41992 * gimple_wce_cleanup_eh_only: GIMPLE_WITH_CLEANUP_EXPR. 41993 (line 18) 41994 * gimple_wce_set_cleanup: GIMPLE_WITH_CLEANUP_EXPR. 41995 (line 15) 41996 * gimple_wce_set_cleanup_eh_only: GIMPLE_WITH_CLEANUP_EXPR. 41997 (line 22) 41998 * GIMPLE_WITH_CLEANUP_EXPR: GIMPLE_WITH_CLEANUP_EXPR. 41999 (line 6) 42000 * gimplification <1>: Parsing pass. (line 14) 42001 * gimplification: Gimplification pass. 42002 (line 6) 42003 * gimplifier: Parsing pass. (line 14) 42004 * gimplify_assign: GIMPLE_ASSIGN. (line 19) 42005 * gimplify_expr: Gimplification pass. 42006 (line 18) 42007 * gimplify_function_tree: Gimplification pass. 42008 (line 18) 42009 * GLOBAL_INIT_PRIORITY: Function Basics. (line 6) 42010 * global_regs: Register Basics. (line 59) 42011 * GO_IF_LEGITIMATE_ADDRESS: Addressing Modes. (line 48) 42012 * GO_IF_MODE_DEPENDENT_ADDRESS: Addressing Modes. (line 190) 42013 * GOFAST, floating point emulation library: Library Calls. (line 44) 42014 * gofast_maybe_init_libfuncs: Library Calls. (line 44) 42015 * greater than: Comparisons. (line 64) 42016 * gsi_after_labels: Sequence iterators. (line 76) 42017 * gsi_bb: Sequence iterators. (line 83) 42018 * gsi_commit_edge_inserts: Sequence iterators. (line 194) 42019 * gsi_commit_one_edge_insert: Sequence iterators. (line 190) 42020 * gsi_end_p: Sequence iterators. (line 60) 42021 * gsi_for_stmt: Sequence iterators. (line 157) 42022 * gsi_insert_after: Sequence iterators. (line 147) 42023 * gsi_insert_before: Sequence iterators. (line 136) 42024 * gsi_insert_on_edge: Sequence iterators. (line 174) 42025 * gsi_insert_on_edge_immediate: Sequence iterators. (line 185) 42026 * gsi_insert_seq_after: Sequence iterators. (line 154) 42027 * gsi_insert_seq_before: Sequence iterators. (line 143) 42028 * gsi_insert_seq_on_edge: Sequence iterators. (line 179) 42029 * gsi_last: Sequence iterators. (line 50) 42030 * gsi_last_bb: Sequence iterators. (line 56) 42031 * gsi_link_after: Sequence iterators. (line 115) 42032 * gsi_link_before: Sequence iterators. (line 105) 42033 * gsi_link_seq_after: Sequence iterators. (line 110) 42034 * gsi_link_seq_before: Sequence iterators. (line 99) 42035 * gsi_move_after: Sequence iterators. (line 161) 42036 * gsi_move_before: Sequence iterators. (line 166) 42037 * gsi_move_to_bb_end: Sequence iterators. (line 171) 42038 * gsi_next: Sequence iterators. (line 66) 42039 * gsi_one_before_end_p: Sequence iterators. (line 63) 42040 * gsi_prev: Sequence iterators. (line 69) 42041 * gsi_remove: Sequence iterators. (line 90) 42042 * gsi_replace: Sequence iterators. (line 130) 42043 * gsi_seq: Sequence iterators. (line 86) 42044 * gsi_split_seq_after: Sequence iterators. (line 120) 42045 * gsi_split_seq_before: Sequence iterators. (line 125) 42046 * gsi_start: Sequence iterators. (line 40) 42047 * gsi_start_bb: Sequence iterators. (line 46) 42048 * gsi_stmt: Sequence iterators. (line 72) 42049 * gt: Comparisons. (line 60) 42050 * gt and attributes: Expressions. (line 64) 42051 * GT_EXPR: Expression trees. (line 6) 42052 * gtu: Comparisons. (line 64) 42053 * gtu and attributes: Expressions. (line 64) 42054 * GTY: Type Information. (line 6) 42055 * H in constraint: Simple Constraints. (line 88) 42056 * HAmode: Machine Modes. (line 144) 42057 * HANDLE_PRAGMA_PACK_PUSH_POP: Misc. (line 467) 42058 * HANDLE_PRAGMA_PACK_WITH_EXPANSION: Misc. (line 478) 42059 * HANDLE_SYSV_PRAGMA: Misc. (line 438) 42060 * HANDLER: Function Bodies. (line 6) 42061 * HANDLER_BODY: Function Bodies. (line 6) 42062 * HANDLER_PARMS: Function Bodies. (line 6) 42063 * hard registers: Regs and Memory. (line 9) 42064 * HARD_FRAME_POINTER_REGNUM: Frame Registers. (line 20) 42065 * HARD_REGNO_CALL_PART_CLOBBERED: Register Basics. (line 53) 42066 * HARD_REGNO_CALLER_SAVE_MODE: Caller Saves. (line 20) 42067 * HARD_REGNO_MODE_OK: Values in Registers. 42068 (line 58) 42069 * HARD_REGNO_NREGS: Values in Registers. 42070 (line 11) 42071 * HARD_REGNO_NREGS_HAS_PADDING: Values in Registers. 42072 (line 25) 42073 * HARD_REGNO_NREGS_WITH_PADDING: Values in Registers. 42074 (line 43) 42075 * HARD_REGNO_RENAME_OK: Values in Registers. 42076 (line 119) 42077 * HAS_INIT_SECTION: Macros for Initialization. 42078 (line 19) 42079 * HAS_LONG_COND_BRANCH: Misc. (line 9) 42080 * HAS_LONG_UNCOND_BRANCH: Misc. (line 18) 42081 * HAVE_DOS_BASED_FILE_SYSTEM: Filesystem. (line 11) 42082 * HAVE_POST_DECREMENT: Addressing Modes. (line 12) 42083 * HAVE_POST_INCREMENT: Addressing Modes. (line 11) 42084 * HAVE_POST_MODIFY_DISP: Addressing Modes. (line 18) 42085 * HAVE_POST_MODIFY_REG: Addressing Modes. (line 24) 42086 * HAVE_PRE_DECREMENT: Addressing Modes. (line 10) 42087 * HAVE_PRE_INCREMENT: Addressing Modes. (line 9) 42088 * HAVE_PRE_MODIFY_DISP: Addressing Modes. (line 17) 42089 * HAVE_PRE_MODIFY_REG: Addressing Modes. (line 23) 42090 * HCmode: Machine Modes. (line 197) 42091 * HFmode: Machine Modes. (line 58) 42092 * high: Constants. (line 109) 42093 * HImode: Machine Modes. (line 29) 42094 * HImode, in insn: Insns. (line 231) 42095 * HONOR_REG_ALLOC_ORDER: Allocation Order. (line 37) 42096 * host configuration: Host Config. (line 6) 42097 * host functions: Host Common. (line 6) 42098 * host hooks: Host Common. (line 6) 42099 * host makefile fragment: Host Fragment. (line 6) 42100 * HOST_BIT_BUCKET: Filesystem. (line 51) 42101 * HOST_EXECUTABLE_SUFFIX: Filesystem. (line 45) 42102 * HOST_HOOKS_EXTRA_SIGNALS: Host Common. (line 12) 42103 * HOST_HOOKS_GT_PCH_ALLOC_GRANULARITY: Host Common. (line 45) 42104 * HOST_HOOKS_GT_PCH_USE_ADDRESS: Host Common. (line 26) 42105 * HOST_LACKS_INODE_NUMBERS: Filesystem. (line 89) 42106 * HOST_LONG_LONG_FORMAT: Host Misc. (line 41) 42107 * HOST_OBJECT_SUFFIX: Filesystem. (line 40) 42108 * HOST_WIDE_INT: Anchored Addresses. (line 33) 42109 * HOT_TEXT_SECTION_NAME: Sections. (line 43) 42110 * HQmode: Machine Modes. (line 107) 42111 * i in constraint: Simple Constraints. (line 60) 42112 * I in constraint: Simple Constraints. (line 71) 42113 * identifier: Identifiers. (line 6) 42114 * IDENTIFIER_LENGTH: Identifiers. (line 20) 42115 * IDENTIFIER_NODE: Identifiers. (line 6) 42116 * IDENTIFIER_OPNAME_P: Identifiers. (line 25) 42117 * IDENTIFIER_POINTER: Identifiers. (line 15) 42118 * IDENTIFIER_TYPENAME_P: Identifiers. (line 31) 42119 * IEEE 754-2008: Decimal float library routines. 42120 (line 6) 42121 * IF_COND: Function Bodies. (line 6) 42122 * if_marked: GTY Options. (line 156) 42123 * IF_STMT: Function Bodies. (line 6) 42124 * if_then_else: Comparisons. (line 80) 42125 * if_then_else and attributes: Expressions. (line 32) 42126 * if_then_else usage: Side Effects. (line 56) 42127 * IFCVT_EXTRA_FIELDS: Misc. (line 619) 42128 * IFCVT_INIT_EXTRA_FIELDS: Misc. (line 614) 42129 * IFCVT_MODIFY_CANCEL: Misc. (line 608) 42130 * IFCVT_MODIFY_FINAL: Misc. (line 602) 42131 * IFCVT_MODIFY_INSN: Misc. (line 596) 42132 * IFCVT_MODIFY_MULTIPLE_TESTS: Misc. (line 589) 42133 * IFCVT_MODIFY_TESTS: Misc. (line 578) 42134 * IMAGPART_EXPR: Expression trees. (line 6) 42135 * Immediate Uses: SSA Operands. (line 274) 42136 * immediate_operand: Machine-Independent Predicates. 42137 (line 11) 42138 * IMMEDIATE_PREFIX: Instruction Output. (line 127) 42139 * in_struct: Flags. (line 258) 42140 * in_struct, in code_label and note: Flags. (line 59) 42141 * in_struct, in insn and jump_insn and call_insn: Flags. (line 49) 42142 * in_struct, in insn, jump_insn and call_insn: Flags. (line 166) 42143 * in_struct, in mem: Flags. (line 70) 42144 * in_struct, in subreg: Flags. (line 205) 42145 * include: Including Patterns. (line 6) 42146 * INCLUDE_DEFAULTS: Driver. (line 430) 42147 * inclusive-or, bitwise: Arithmetic. (line 158) 42148 * INCOMING_FRAME_SP_OFFSET: Frame Layout. (line 183) 42149 * INCOMING_REGNO: Register Basics. (line 91) 42150 * INCOMING_RETURN_ADDR_RTX: Frame Layout. (line 139) 42151 * INCOMING_STACK_BOUNDARY: Storage Layout. (line 165) 42152 * INDEX_REG_CLASS: Register Classes. (line 134) 42153 * indirect_jump instruction pattern: Standard Names. (line 1078) 42154 * indirect_operand: Machine-Independent Predicates. 42155 (line 71) 42156 * INDIRECT_REF: Expression trees. (line 6) 42157 * INIT_ARRAY_SECTION_ASM_OP: Sections. (line 98) 42158 * INIT_CUMULATIVE_ARGS: Register Arguments. (line 149) 42159 * INIT_CUMULATIVE_INCOMING_ARGS: Register Arguments. (line 177) 42160 * INIT_CUMULATIVE_LIBCALL_ARGS: Register Arguments. (line 170) 42161 * INIT_ENVIRONMENT: Driver. (line 369) 42162 * INIT_EXPANDERS: Per-Function Data. (line 39) 42163 * INIT_EXPR: Expression trees. (line 6) 42164 * init_machine_status: Per-Function Data. (line 45) 42165 * init_one_libfunc: Library Calls. (line 15) 42166 * INIT_SECTION_ASM_OP <1>: Sections. (line 82) 42167 * INIT_SECTION_ASM_OP: Macros for Initialization. 42168 (line 10) 42169 * INITIAL_ELIMINATION_OFFSET: Elimination. (line 79) 42170 * INITIAL_FRAME_ADDRESS_RTX: Frame Layout. (line 83) 42171 * INITIAL_FRAME_POINTER_OFFSET: Elimination. (line 32) 42172 * initialization routines: Initialization. (line 6) 42173 * INITIALIZE_TRAMPOLINE: Trampolines. (line 55) 42174 * inlining: Target Attributes. (line 86) 42175 * insert_insn_on_edge: Maintaining the CFG. 42176 (line 118) 42177 * insn: Insns. (line 63) 42178 * insn and /f: Flags. (line 125) 42179 * insn and /j: Flags. (line 175) 42180 * insn and /s: Flags. (line 166) 42181 * insn and /u: Flags. (line 39) 42182 * insn and /v: Flags. (line 44) 42183 * insn attributes: Insn Attributes. (line 6) 42184 * insn canonicalization: Insn Canonicalizations. 42185 (line 6) 42186 * insn includes: Including Patterns. (line 6) 42187 * insn lengths, computing: Insn Lengths. (line 6) 42188 * insn splitting: Insn Splitting. (line 6) 42189 * insn-attr.h: Defining Attributes. 42190 (line 24) 42191 * INSN_ANNULLED_BRANCH_P: Flags. (line 39) 42192 * INSN_CODE: Insns. (line 257) 42193 * INSN_DELETED_P: Flags. (line 44) 42194 * INSN_FROM_TARGET_P: Flags. (line 49) 42195 * insn_list: Insns. (line 505) 42196 * INSN_REFERENCES_ARE_DELAYED: Misc. (line 517) 42197 * INSN_SETS_ARE_DELAYED: Misc. (line 506) 42198 * INSN_UID: Insns. (line 23) 42199 * insns: Insns. (line 6) 42200 * insns, generating: RTL Template. (line 6) 42201 * insns, recognizing: RTL Template. (line 6) 42202 * instruction attributes: Insn Attributes. (line 6) 42203 * instruction latency time: Processor pipeline description. 42204 (line 106) 42205 * instruction patterns: Patterns. (line 6) 42206 * instruction splitting: Insn Splitting. (line 6) 42207 * insv instruction pattern: Standard Names. (line 880) 42208 * int <1>: Manipulating GIMPLE statements. 42209 (line 66) 42210 * int: Run-time Target. (line 56) 42211 * INT_TYPE_SIZE: Type Layout. (line 12) 42212 * INTEGER_CST: Expression trees. (line 6) 42213 * INTEGER_TYPE: Types. (line 6) 42214 * Interdependence of Patterns: Dependent Patterns. (line 6) 42215 * interfacing to GCC output: Interface. (line 6) 42216 * interlock delays: Processor pipeline description. 42217 (line 6) 42218 * intermediate representation lowering: Parsing pass. (line 14) 42219 * INTMAX_TYPE: Type Layout. (line 213) 42220 * introduction: Top. (line 6) 42221 * INVOKE__main: Macros for Initialization. 42222 (line 51) 42223 * ior: Arithmetic. (line 158) 42224 * ior and attributes: Expressions. (line 50) 42225 * ior, canonicalization of: Insn Canonicalizations. 42226 (line 57) 42227 * iorM3 instruction pattern: Standard Names. (line 222) 42228 * IRA_COVER_CLASSES: Register Classes. (line 516) 42229 * IRA_HARD_REGNO_ADD_COST_MULTIPLIER: Allocation Order. (line 45) 42230 * IS_ASM_LOGICAL_LINE_SEPARATOR: Data Output. (line 120) 42231 * is_gimple_omp: GIMPLE_OMP_PARALLEL. 42232 (line 65) 42233 * iterators in .md files: Iterators. (line 6) 42234 * IV analysis on GIMPLE: Scalar evolutions. (line 6) 42235 * IV analysis on RTL: loop-iv. (line 6) 42236 * jump: Flags. (line 309) 42237 * jump instruction pattern: Standard Names. (line 969) 42238 * jump instruction patterns: Jump Patterns. (line 6) 42239 * jump instructions and set: Side Effects. (line 56) 42240 * jump, in call_insn: Flags. (line 179) 42241 * jump, in insn: Flags. (line 175) 42242 * jump, in mem: Flags. (line 79) 42243 * JUMP_ALIGN: Alignment Output. (line 9) 42244 * jump_insn: Insns. (line 73) 42245 * jump_insn and /f: Flags. (line 125) 42246 * jump_insn and /s: Flags. (line 49) 42247 * jump_insn and /u: Flags. (line 39) 42248 * jump_insn and /v: Flags. (line 44) 42249 * JUMP_LABEL: Insns. (line 80) 42250 * JUMP_TABLES_IN_TEXT_SECTION: Sections. (line 142) 42251 * Jumps: Jumps. (line 6) 42252 * LABEL_ALIGN: Alignment Output. (line 52) 42253 * LABEL_ALIGN_AFTER_BARRIER: Alignment Output. (line 22) 42254 * LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP: Alignment Output. (line 30) 42255 * LABEL_ALIGN_MAX_SKIP: Alignment Output. (line 62) 42256 * LABEL_ALT_ENTRY_P: Insns. (line 140) 42257 * LABEL_ALTERNATE_NAME: Edges. (line 180) 42258 * LABEL_DECL: Declarations. (line 6) 42259 * LABEL_KIND: Insns. (line 140) 42260 * LABEL_NUSES: Insns. (line 136) 42261 * LABEL_PRESERVE_P: Flags. (line 59) 42262 * label_ref: Constants. (line 86) 42263 * label_ref and /v: Flags. (line 65) 42264 * label_ref, RTL sharing: Sharing. (line 35) 42265 * LABEL_REF_NONLOCAL_P: Flags. (line 65) 42266 * lang_hooks.gimplify_expr: Gimplification pass. 42267 (line 18) 42268 * lang_hooks.parse_file: Parsing pass. (line 6) 42269 * language-independent intermediate representation: Parsing pass. 42270 (line 14) 42271 * large return values: Aggregate Return. (line 6) 42272 * LARGEST_EXPONENT_IS_NORMAL: Storage Layout. (line 469) 42273 * LAST_STACK_REG: Stack Registers. (line 27) 42274 * LAST_VIRTUAL_REGISTER: Regs and Memory. (line 51) 42275 * lceilMN2: Standard Names. (line 597) 42276 * LCSSA: LCSSA. (line 6) 42277 * LD_FINI_SWITCH: Macros for Initialization. 42278 (line 29) 42279 * LD_INIT_SWITCH: Macros for Initialization. 42280 (line 25) 42281 * LDD_SUFFIX: Macros for Initialization. 42282 (line 116) 42283 * le: Comparisons. (line 76) 42284 * le and attributes: Expressions. (line 64) 42285 * LE_EXPR: Expression trees. (line 6) 42286 * leaf functions: Leaf Functions. (line 6) 42287 * leaf_function_p: Standard Names. (line 1040) 42288 * LEAF_REG_REMAP: Leaf Functions. (line 39) 42289 * LEAF_REGISTERS: Leaf Functions. (line 25) 42290 * left rotate: Arithmetic. (line 190) 42291 * left shift: Arithmetic. (line 168) 42292 * LEGITIMATE_CONSTANT_P: Addressing Modes. (line 205) 42293 * LEGITIMATE_PIC_OPERAND_P: PIC. (line 31) 42294 * LEGITIMIZE_ADDRESS: Addressing Modes. (line 122) 42295 * LEGITIMIZE_RELOAD_ADDRESS: Addressing Modes. (line 145) 42296 * length: GTY Options. (line 50) 42297 * less than: Comparisons. (line 68) 42298 * less than or equal: Comparisons. (line 76) 42299 * leu: Comparisons. (line 76) 42300 * leu and attributes: Expressions. (line 64) 42301 * lfloorMN2: Standard Names. (line 592) 42302 * LIB2FUNCS_EXTRA: Target Fragment. (line 11) 42303 * LIB_SPEC: Driver. (line 170) 42304 * LIBCALL_VALUE: Scalar Return. (line 60) 42305 * libgcc.a: Library Calls. (line 6) 42306 * LIBGCC2_CFLAGS: Target Fragment. (line 8) 42307 * LIBGCC2_HAS_DF_MODE: Type Layout. (line 109) 42308 * LIBGCC2_HAS_TF_MODE: Type Layout. (line 123) 42309 * LIBGCC2_HAS_XF_MODE: Type Layout. (line 117) 42310 * LIBGCC2_LONG_DOUBLE_TYPE_SIZE: Type Layout. (line 103) 42311 * LIBGCC2_UNWIND_ATTRIBUTE: Misc. (line 935) 42312 * LIBGCC2_WORDS_BIG_ENDIAN: Storage Layout. (line 36) 42313 * LIBGCC_SPEC: Driver. (line 178) 42314 * library subroutine names: Library Calls. (line 6) 42315 * LIBRARY_PATH_ENV: Misc. (line 557) 42316 * LIMIT_RELOAD_CLASS: Register Classes. (line 239) 42317 * Linear loop transformations framework: Lambda. (line 6) 42318 * LINK_COMMAND_SPEC: Driver. (line 299) 42319 * LINK_EH_SPEC: Driver. (line 205) 42320 * LINK_ELIMINATE_DUPLICATE_LDIRECTORIES: Driver. (line 309) 42321 * LINK_GCC_C_SEQUENCE_SPEC: Driver. (line 295) 42322 * LINK_LIBGCC_SPECIAL_1: Driver. (line 290) 42323 * LINK_SPEC: Driver. (line 163) 42324 * linkage: Function Basics. (line 6) 42325 * list: Containers. (line 6) 42326 * Liveness representation: Liveness information. 42327 (line 6) 42328 * lo_sum: Arithmetic. (line 24) 42329 * load address instruction: Simple Constraints. (line 154) 42330 * LOAD_EXTEND_OP: Misc. (line 69) 42331 * load_multiple instruction pattern: Standard Names. (line 137) 42332 * LOCAL_ALIGNMENT: Storage Layout. (line 254) 42333 * LOCAL_CLASS_P: Classes. (line 68) 42334 * LOCAL_DECL_ALIGNMENT: Storage Layout. (line 278) 42335 * LOCAL_INCLUDE_DIR: Driver. (line 376) 42336 * LOCAL_LABEL_PREFIX: Instruction Output. (line 125) 42337 * LOCAL_REGNO: Register Basics. (line 105) 42338 * LOG_LINKS: Insns. (line 276) 42339 * Logical Operators: Logical Operators. (line 6) 42340 * logical-and, bitwise: Arithmetic. (line 153) 42341 * logM2 instruction pattern: Standard Names. (line 505) 42342 * LONG_ACCUM_TYPE_SIZE: Type Layout. (line 93) 42343 * LONG_DOUBLE_TYPE_SIZE: Type Layout. (line 58) 42344 * LONG_FRACT_TYPE_SIZE: Type Layout. (line 73) 42345 * LONG_LONG_ACCUM_TYPE_SIZE: Type Layout. (line 98) 42346 * LONG_LONG_FRACT_TYPE_SIZE: Type Layout. (line 78) 42347 * LONG_LONG_TYPE_SIZE: Type Layout. (line 33) 42348 * LONG_TYPE_SIZE: Type Layout. (line 22) 42349 * longjmp and automatic variables: Interface. (line 52) 42350 * Loop analysis: Loop representation. 42351 (line 6) 42352 * Loop manipulation: Loop manipulation. (line 6) 42353 * Loop querying: Loop querying. (line 6) 42354 * Loop representation: Loop representation. 42355 (line 6) 42356 * Loop-closed SSA form: LCSSA. (line 6) 42357 * LOOP_ALIGN: Alignment Output. (line 35) 42358 * LOOP_ALIGN_MAX_SKIP: Alignment Output. (line 48) 42359 * LOOP_EXPR: Expression trees. (line 6) 42360 * looping instruction patterns: Looping Patterns. (line 6) 42361 * lowering, language-dependent intermediate representation: Parsing pass. 42362 (line 14) 42363 * lrintMN2: Standard Names. (line 582) 42364 * lroundMN2: Standard Names. (line 587) 42365 * LSHIFT_EXPR: Expression trees. (line 6) 42366 * lshiftrt: Arithmetic. (line 185) 42367 * lshiftrt and attributes: Expressions. (line 64) 42368 * lshrM3 instruction pattern: Standard Names. (line 441) 42369 * lt: Comparisons. (line 68) 42370 * lt and attributes: Expressions. (line 64) 42371 * LT_EXPR: Expression trees. (line 6) 42372 * LTGT_EXPR: Expression trees. (line 6) 42373 * ltu: Comparisons. (line 68) 42374 * m in constraint: Simple Constraints. (line 17) 42375 * machine attributes: Target Attributes. (line 6) 42376 * machine description macros: Target Macros. (line 6) 42377 * machine descriptions: Machine Desc. (line 6) 42378 * machine mode conversions: Conversions. (line 6) 42379 * machine modes: Machine Modes. (line 6) 42380 * machine specific constraints: Machine Constraints. 42381 (line 6) 42382 * machine-independent predicates: Machine-Independent Predicates. 42383 (line 6) 42384 * machine_mode: Condition Code. (line 157) 42385 * macros, target description: Target Macros. (line 6) 42386 * maddMN4 instruction pattern: Standard Names. (line 364) 42387 * MAKE_DECL_ONE_ONLY: Label Output. (line 218) 42388 * make_phi_node: GIMPLE_PHI. (line 7) 42389 * make_safe_from: Expander Definitions. 42390 (line 148) 42391 * makefile fragment: Fragments. (line 6) 42392 * makefile targets: Makefile. (line 6) 42393 * MALLOC_ABI_ALIGNMENT: Storage Layout. (line 179) 42394 * Manipulating GIMPLE statements: Manipulating GIMPLE statements. 42395 (line 6) 42396 * mark_hook: GTY Options. (line 171) 42397 * marking roots: GGC Roots. (line 6) 42398 * MASK_RETURN_ADDR: Exception Region Output. 42399 (line 35) 42400 * match_dup <1>: RTL Template. (line 73) 42401 * match_dup: define_peephole2. (line 28) 42402 * match_dup and attributes: Insn Lengths. (line 16) 42403 * match_op_dup: RTL Template. (line 163) 42404 * match_operand: RTL Template. (line 16) 42405 * match_operand and attributes: Expressions. (line 55) 42406 * match_operator: RTL Template. (line 95) 42407 * match_par_dup: RTL Template. (line 219) 42408 * match_parallel: RTL Template. (line 172) 42409 * match_scratch <1>: RTL Template. (line 58) 42410 * match_scratch: define_peephole2. (line 28) 42411 * matching constraint: Simple Constraints. (line 132) 42412 * matching operands: Output Template. (line 49) 42413 * math library: Soft float library routines. 42414 (line 6) 42415 * math, in RTL: Arithmetic. (line 6) 42416 * MATH_LIBRARY: Misc. (line 550) 42417 * matherr: Library Calls. (line 58) 42418 * MAX_BITS_PER_WORD: Storage Layout. (line 61) 42419 * MAX_CONDITIONAL_EXECUTE: Misc. (line 572) 42420 * MAX_FIXED_MODE_SIZE: Storage Layout. (line 420) 42421 * MAX_MOVE_MAX: Misc. (line 120) 42422 * MAX_OFILE_ALIGNMENT: Storage Layout. (line 216) 42423 * MAX_REGS_PER_ADDRESS: Addressing Modes. (line 42) 42424 * MAX_STACK_ALIGNMENT: Storage Layout. (line 209) 42425 * maxM3 instruction pattern: Standard Names. (line 234) 42426 * may_trap_p, tree_could_trap_p: Edges. (line 115) 42427 * maybe_undef: GTY Options. (line 179) 42428 * mcount: Profiling. (line 12) 42429 * MD_CAN_REDIRECT_BRANCH: Misc. (line 697) 42430 * MD_EXEC_PREFIX: Driver. (line 330) 42431 * MD_FALLBACK_FRAME_STATE_FOR: Exception Handling. (line 98) 42432 * MD_HANDLE_UNWABI: Exception Handling. (line 118) 42433 * MD_STARTFILE_PREFIX: Driver. (line 358) 42434 * MD_STARTFILE_PREFIX_1: Driver. (line 364) 42435 * MD_UNWIND_SUPPORT: Exception Handling. (line 94) 42436 * mem: Regs and Memory. (line 374) 42437 * mem and /c: Flags. (line 99) 42438 * mem and /f: Flags. (line 103) 42439 * mem and /i: Flags. (line 85) 42440 * mem and /j: Flags. (line 79) 42441 * mem and /s: Flags. (line 70) 42442 * mem and /u: Flags. (line 152) 42443 * mem and /v: Flags. (line 94) 42444 * mem, RTL sharing: Sharing. (line 40) 42445 * MEM_ALIAS_SET: Special Accessors. (line 9) 42446 * MEM_ALIGN: Special Accessors. (line 36) 42447 * MEM_EXPR: Special Accessors. (line 20) 42448 * MEM_IN_STRUCT_P: Flags. (line 70) 42449 * MEM_KEEP_ALIAS_SET_P: Flags. (line 79) 42450 * MEM_NOTRAP_P: Flags. (line 99) 42451 * MEM_OFFSET: Special Accessors. (line 28) 42452 * MEM_POINTER: Flags. (line 103) 42453 * MEM_READONLY_P: Flags. (line 152) 42454 * MEM_SCALAR_P: Flags. (line 85) 42455 * MEM_SIZE: Special Accessors. (line 31) 42456 * MEM_VOLATILE_P: Flags. (line 94) 42457 * MEMBER_TYPE_FORCES_BLK: Storage Layout. (line 400) 42458 * memory reference, nonoffsettable: Simple Constraints. (line 246) 42459 * memory references in constraints: Simple Constraints. (line 17) 42460 * memory_barrier instruction pattern: Standard Names. (line 1413) 42461 * MEMORY_MOVE_COST: Costs. (line 29) 42462 * memory_operand: Machine-Independent Predicates. 42463 (line 58) 42464 * METHOD_TYPE: Types. (line 6) 42465 * MIN_UNITS_PER_WORD: Storage Layout. (line 71) 42466 * MINIMUM_ALIGNMENT: Storage Layout. (line 288) 42467 * MINIMUM_ATOMIC_ALIGNMENT: Storage Layout. (line 187) 42468 * minM3 instruction pattern: Standard Names. (line 234) 42469 * minus: Arithmetic. (line 36) 42470 * minus and attributes: Expressions. (line 64) 42471 * minus, canonicalization of: Insn Canonicalizations. 42472 (line 27) 42473 * MINUS_EXPR: Expression trees. (line 6) 42474 * MIPS coprocessor-definition macros: MIPS Coprocessors. (line 6) 42475 * mod: Arithmetic. (line 131) 42476 * mod and attributes: Expressions. (line 64) 42477 * mode classes: Machine Modes. (line 219) 42478 * mode iterators in .md files: Mode Iterators. (line 6) 42479 * mode switching: Mode Switching. (line 6) 42480 * MODE_ACCUM: Machine Modes. (line 249) 42481 * MODE_AFTER: Mode Switching. (line 49) 42482 * MODE_BASE_REG_CLASS: Register Classes. (line 112) 42483 * MODE_BASE_REG_REG_CLASS: Register Classes. (line 118) 42484 * MODE_CC: Machine Modes. (line 268) 42485 * MODE_CODE_BASE_REG_CLASS: Register Classes. (line 125) 42486 * MODE_COMPLEX_FLOAT: Machine Modes. (line 260) 42487 * MODE_COMPLEX_INT: Machine Modes. (line 257) 42488 * MODE_DECIMAL_FLOAT: Machine Modes. (line 237) 42489 * MODE_ENTRY: Mode Switching. (line 54) 42490 * MODE_EXIT: Mode Switching. (line 60) 42491 * MODE_FLOAT: Machine Modes. (line 233) 42492 * MODE_FRACT: Machine Modes. (line 241) 42493 * MODE_FUNCTION: Machine Modes. (line 264) 42494 * MODE_INT: Machine Modes. (line 225) 42495 * MODE_NEEDED: Mode Switching. (line 42) 42496 * MODE_PARTIAL_INT: Machine Modes. (line 229) 42497 * MODE_PRIORITY_TO_MODE: Mode Switching. (line 66) 42498 * MODE_RANDOM: Machine Modes. (line 273) 42499 * MODE_UACCUM: Machine Modes. (line 253) 42500 * MODE_UFRACT: Machine Modes. (line 245) 42501 * MODES_TIEABLE_P: Values in Registers. 42502 (line 129) 42503 * modifiers in constraints: Modifiers. (line 6) 42504 * MODIFY_EXPR: Expression trees. (line 6) 42505 * MODIFY_JNI_METHOD_CALL: Misc. (line 774) 42506 * MODIFY_TARGET_NAME: Driver. (line 385) 42507 * modM3 instruction pattern: Standard Names. (line 222) 42508 * modulo scheduling: RTL passes. (line 131) 42509 * MOVE_BY_PIECES_P: Costs. (line 110) 42510 * MOVE_MAX: Misc. (line 115) 42511 * MOVE_MAX_PIECES: Costs. (line 116) 42512 * MOVE_RATIO: Costs. (line 97) 42513 * movM instruction pattern: Standard Names. (line 11) 42514 * movmemM instruction pattern: Standard Names. (line 672) 42515 * movmisalignM instruction pattern: Standard Names. (line 126) 42516 * movMODEcc instruction pattern: Standard Names. (line 891) 42517 * movstr instruction pattern: Standard Names. (line 707) 42518 * movstrictM instruction pattern: Standard Names. (line 120) 42519 * msubMN4 instruction pattern: Standard Names. (line 387) 42520 * mulhisi3 instruction pattern: Standard Names. (line 340) 42521 * mulM3 instruction pattern: Standard Names. (line 222) 42522 * mulqihi3 instruction pattern: Standard Names. (line 344) 42523 * mulsidi3 instruction pattern: Standard Names. (line 344) 42524 * mult: Arithmetic. (line 92) 42525 * mult and attributes: Expressions. (line 64) 42526 * mult, canonicalization of: Insn Canonicalizations. 42527 (line 27) 42528 * MULT_EXPR: Expression trees. (line 6) 42529 * MULTILIB_DEFAULTS: Driver. (line 315) 42530 * MULTILIB_DIRNAMES: Target Fragment. (line 64) 42531 * MULTILIB_EXCEPTIONS: Target Fragment. (line 84) 42532 * MULTILIB_EXTRA_OPTS: Target Fragment. (line 96) 42533 * MULTILIB_MATCHES: Target Fragment. (line 77) 42534 * MULTILIB_OPTIONS: Target Fragment. (line 44) 42535 * multiple alternative constraints: Multi-Alternative. (line 6) 42536 * MULTIPLE_SYMBOL_SPACES: Misc. (line 530) 42537 * multiplication: Arithmetic. (line 92) 42538 * multiplication with signed saturation: Arithmetic. (line 92) 42539 * multiplication with unsigned saturation: Arithmetic. (line 92) 42540 * MUST_USE_SJLJ_EXCEPTIONS: Exception Region Output. 42541 (line 64) 42542 * n in constraint: Simple Constraints. (line 65) 42543 * N_REG_CLASSES: Register Classes. (line 76) 42544 * name: Identifiers. (line 6) 42545 * named patterns and conditions: Patterns. (line 47) 42546 * names, pattern: Standard Names. (line 6) 42547 * namespace: Namespaces. (line 6) 42548 * namespace, class, scope: Scopes. (line 6) 42549 * NAMESPACE_DECL <1>: Namespaces. (line 6) 42550 * NAMESPACE_DECL: Declarations. (line 6) 42551 * NATIVE_SYSTEM_HEADER_DIR: Target Fragment. (line 103) 42552 * ne: Comparisons. (line 56) 42553 * ne and attributes: Expressions. (line 64) 42554 * NE_EXPR: Expression trees. (line 6) 42555 * nearbyintM2 instruction pattern: Standard Names. (line 564) 42556 * neg: Arithmetic. (line 81) 42557 * neg and attributes: Expressions. (line 64) 42558 * neg, canonicalization of: Insn Canonicalizations. 42559 (line 27) 42560 * NEGATE_EXPR: Expression trees. (line 6) 42561 * negation: Arithmetic. (line 81) 42562 * negation with signed saturation: Arithmetic. (line 81) 42563 * negation with unsigned saturation: Arithmetic. (line 81) 42564 * negM2 instruction pattern: Standard Names. (line 449) 42565 * nested functions, trampolines for: Trampolines. (line 6) 42566 * nested_ptr: GTY Options. (line 186) 42567 * next_bb, prev_bb, FOR_EACH_BB: Basic Blocks. (line 10) 42568 * next_cc0_user: Jump Patterns. (line 64) 42569 * NEXT_INSN: Insns. (line 30) 42570 * NEXT_OBJC_RUNTIME: Library Calls. (line 94) 42571 * nil: RTL Objects. (line 73) 42572 * NO_DBX_BNSYM_ENSYM: DBX Hooks. (line 39) 42573 * NO_DBX_FUNCTION_END: DBX Hooks. (line 33) 42574 * NO_DBX_GCC_MARKER: File Names and DBX. (line 28) 42575 * NO_DBX_MAIN_SOURCE_DIRECTORY: File Names and DBX. (line 23) 42576 * NO_DOLLAR_IN_LABEL: Misc. (line 494) 42577 * NO_DOT_IN_LABEL: Misc. (line 500) 42578 * NO_FUNCTION_CSE: Costs. (line 200) 42579 * NO_IMPLICIT_EXTERN_C: Misc. (line 376) 42580 * NO_PROFILE_COUNTERS: Profiling. (line 28) 42581 * NO_REGS: Register Classes. (line 17) 42582 * NON_LVALUE_EXPR: Expression trees. (line 6) 42583 * nondeterministic finite state automaton: Processor pipeline description. 42584 (line 301) 42585 * nonimmediate_operand: Machine-Independent Predicates. 42586 (line 101) 42587 * nonlocal goto handler: Edges. (line 171) 42588 * nonlocal_goto instruction pattern: Standard Names. (line 1255) 42589 * nonlocal_goto_receiver instruction pattern: Standard Names. 42590 (line 1272) 42591 * nonmemory_operand: Machine-Independent Predicates. 42592 (line 97) 42593 * nonoffsettable memory reference: Simple Constraints. (line 246) 42594 * nop instruction pattern: Standard Names. (line 1073) 42595 * NOP_EXPR: Expression trees. (line 6) 42596 * normal predicates: Predicates. (line 31) 42597 * not: Arithmetic. (line 149) 42598 * not and attributes: Expressions. (line 50) 42599 * not equal: Comparisons. (line 56) 42600 * not, canonicalization of: Insn Canonicalizations. 42601 (line 27) 42602 * note: Insns. (line 168) 42603 * note and /i: Flags. (line 59) 42604 * note and /v: Flags. (line 44) 42605 * NOTE_INSN_BASIC_BLOCK, CODE_LABEL, notes: Basic Blocks. (line 41) 42606 * NOTE_INSN_BLOCK_BEG: Insns. (line 193) 42607 * NOTE_INSN_BLOCK_END: Insns. (line 193) 42608 * NOTE_INSN_DELETED: Insns. (line 183) 42609 * NOTE_INSN_DELETED_LABEL: Insns. (line 188) 42610 * NOTE_INSN_EH_REGION_BEG: Insns. (line 199) 42611 * NOTE_INSN_EH_REGION_END: Insns. (line 199) 42612 * NOTE_INSN_FUNCTION_BEG: Insns. (line 223) 42613 * NOTE_INSN_LOOP_BEG: Insns. (line 207) 42614 * NOTE_INSN_LOOP_CONT: Insns. (line 213) 42615 * NOTE_INSN_LOOP_END: Insns. (line 207) 42616 * NOTE_INSN_LOOP_VTOP: Insns. (line 217) 42617 * NOTE_LINE_NUMBER: Insns. (line 168) 42618 * NOTE_SOURCE_FILE: Insns. (line 168) 42619 * NOTICE_UPDATE_CC: Condition Code. (line 33) 42620 * NUM_MACHINE_MODES: Machine Modes. (line 286) 42621 * NUM_MODES_FOR_MODE_SWITCHING: Mode Switching. (line 30) 42622 * Number of iterations analysis: Number of iterations. 42623 (line 6) 42624 * o in constraint: Simple Constraints. (line 23) 42625 * OBJC_GEN_METHOD_LABEL: Label Output. (line 411) 42626 * OBJC_JBLEN: Misc. (line 930) 42627 * OBJECT_FORMAT_COFF: Macros for Initialization. 42628 (line 97) 42629 * OFFSET_TYPE: Types. (line 6) 42630 * offsettable address: Simple Constraints. (line 23) 42631 * OImode: Machine Modes. (line 51) 42632 * Omega a solver for linear programming problems: Omega. (line 6) 42633 * OMP_ATOMIC: Expression trees. (line 6) 42634 * OMP_CLAUSE: Expression trees. (line 6) 42635 * OMP_CONTINUE: Expression trees. (line 6) 42636 * OMP_CRITICAL: Expression trees. (line 6) 42637 * OMP_FOR: Expression trees. (line 6) 42638 * OMP_MASTER: Expression trees. (line 6) 42639 * OMP_ORDERED: Expression trees. (line 6) 42640 * OMP_PARALLEL: Expression trees. (line 6) 42641 * OMP_RETURN: Expression trees. (line 6) 42642 * OMP_SECTION: Expression trees. (line 6) 42643 * OMP_SECTIONS: Expression trees. (line 6) 42644 * OMP_SINGLE: Expression trees. (line 6) 42645 * one_cmplM2 instruction pattern: Standard Names. (line 651) 42646 * operand access: Accessors. (line 6) 42647 * Operand Access Routines: SSA Operands. (line 119) 42648 * operand constraints: Constraints. (line 6) 42649 * Operand Iterators: SSA Operands. (line 119) 42650 * operand predicates: Predicates. (line 6) 42651 * operand substitution: Output Template. (line 6) 42652 * Operands: Operands. (line 6) 42653 * operands <1>: Patterns. (line 53) 42654 * operands: SSA Operands. (line 6) 42655 * operator predicates: Predicates. (line 6) 42656 * optc-gen.awk: Options. (line 6) 42657 * Optimization infrastructure for GIMPLE: Tree SSA. (line 6) 42658 * OPTIMIZATION_OPTIONS: Run-time Target. (line 120) 42659 * OPTIMIZE_MODE_SWITCHING: Mode Switching. (line 9) 42660 * option specification files: Options. (line 6) 42661 * OPTION_DEFAULT_SPECS: Driver. (line 88) 42662 * optional hardware or system features: Run-time Target. (line 59) 42663 * options, directory search: Including Patterns. (line 44) 42664 * order of register allocation: Allocation Order. (line 6) 42665 * ORDERED_EXPR: Expression trees. (line 6) 42666 * Ordering of Patterns: Pattern Ordering. (line 6) 42667 * ORIGINAL_REGNO: Special Accessors. (line 40) 42668 * other register constraints: Simple Constraints. (line 163) 42669 * OUTGOING_REG_PARM_STACK_SPACE: Stack Arguments. (line 71) 42670 * OUTGOING_REGNO: Register Basics. (line 98) 42671 * output of assembler code: File Framework. (line 6) 42672 * output statements: Output Statement. (line 6) 42673 * output templates: Output Template. (line 6) 42674 * OUTPUT_ADDR_CONST_EXTRA: Data Output. (line 39) 42675 * output_asm_insn: Output Statement. (line 53) 42676 * OUTPUT_QUOTED_STRING: File Framework. (line 76) 42677 * OVERLOAD: Functions. (line 6) 42678 * OVERRIDE_ABI_FORMAT: Register Arguments. (line 140) 42679 * OVERRIDE_OPTIONS: Run-time Target. (line 104) 42680 * OVL_CURRENT: Functions. (line 6) 42681 * OVL_NEXT: Functions. (line 6) 42682 * p in constraint: Simple Constraints. (line 154) 42683 * PAD_VARARGS_DOWN: Register Arguments. (line 221) 42684 * parallel: Side Effects. (line 204) 42685 * param_is: GTY Options. (line 114) 42686 * parameters, c++ abi: C++ ABI. (line 6) 42687 * parameters, miscellaneous: Misc. (line 6) 42688 * parameters, precompiled headers: PCH Target. (line 6) 42689 * paramN_is: GTY Options. (line 132) 42690 * parity: Arithmetic. (line 228) 42691 * parityM2 instruction pattern: Standard Names. (line 645) 42692 * PARM_BOUNDARY: Storage Layout. (line 144) 42693 * PARM_DECL: Declarations. (line 6) 42694 * PARSE_LDD_OUTPUT: Macros for Initialization. 42695 (line 121) 42696 * passes and files of the compiler: Passes. (line 6) 42697 * passing arguments: Interface. (line 36) 42698 * PATH_SEPARATOR: Filesystem. (line 31) 42699 * PATTERN: Insns. (line 247) 42700 * pattern conditions: Patterns. (line 43) 42701 * pattern names: Standard Names. (line 6) 42702 * Pattern Ordering: Pattern Ordering. (line 6) 42703 * patterns: Patterns. (line 6) 42704 * pc: Regs and Memory. (line 361) 42705 * pc and attributes: Insn Lengths. (line 20) 42706 * pc, RTL sharing: Sharing. (line 25) 42707 * PC_REGNUM: Register Basics. (line 112) 42708 * pc_rtx: Regs and Memory. (line 366) 42709 * PCC_BITFIELD_TYPE_MATTERS: Storage Layout. (line 314) 42710 * PCC_STATIC_STRUCT_RETURN: Aggregate Return. (line 64) 42711 * PDImode: Machine Modes. (line 40) 42712 * peephole optimization, RTL representation: Side Effects. (line 238) 42713 * peephole optimizer definitions: Peephole Definitions. 42714 (line 6) 42715 * per-function data: Per-Function Data. (line 6) 42716 * percent sign: Output Template. (line 6) 42717 * PHI nodes: SSA. (line 31) 42718 * phi_arg_d: GIMPLE_PHI. (line 28) 42719 * PHI_ARG_DEF: SSA. (line 71) 42720 * PHI_ARG_EDGE: SSA. (line 68) 42721 * PHI_ARG_ELT: SSA. (line 63) 42722 * PHI_NUM_ARGS: SSA. (line 59) 42723 * PHI_RESULT: SSA. (line 56) 42724 * PIC: PIC. (line 6) 42725 * PIC_OFFSET_TABLE_REG_CALL_CLOBBERED: PIC. (line 26) 42726 * PIC_OFFSET_TABLE_REGNUM: PIC. (line 16) 42727 * pipeline hazard recognizer: Processor pipeline description. 42728 (line 6) 42729 * Plugins: Plugins. (line 6) 42730 * plus: Arithmetic. (line 14) 42731 * plus and attributes: Expressions. (line 64) 42732 * plus, canonicalization of: Insn Canonicalizations. 42733 (line 27) 42734 * PLUS_EXPR: Expression trees. (line 6) 42735 * Pmode: Misc. (line 344) 42736 * pmode_register_operand: Machine-Independent Predicates. 42737 (line 35) 42738 * pointer: Types. (line 6) 42739 * POINTER_PLUS_EXPR: Expression trees. (line 6) 42740 * POINTER_SIZE: Storage Layout. (line 83) 42741 * POINTER_TYPE: Types. (line 6) 42742 * POINTERS_EXTEND_UNSIGNED: Storage Layout. (line 89) 42743 * pop_operand: Machine-Independent Predicates. 42744 (line 88) 42745 * popcount: Arithmetic. (line 224) 42746 * popcountM2 instruction pattern: Standard Names. (line 639) 42747 * portability: Portability. (line 6) 42748 * position independent code: PIC. (line 6) 42749 * post_dec: Incdec. (line 25) 42750 * post_inc: Incdec. (line 30) 42751 * post_modify: Incdec. (line 33) 42752 * POSTDECREMENT_EXPR: Expression trees. (line 6) 42753 * POSTINCREMENT_EXPR: Expression trees. (line 6) 42754 * POWI_MAX_MULTS: Misc. (line 828) 42755 * powM3 instruction pattern: Standard Names. (line 513) 42756 * pragma: Misc. (line 437) 42757 * pre_dec: Incdec. (line 8) 42758 * PRE_GCC3_DWARF_FRAME_REGISTERS: Frame Registers. (line 110) 42759 * pre_inc: Incdec. (line 22) 42760 * pre_modify: Incdec. (line 51) 42761 * PREDECREMENT_EXPR: Expression trees. (line 6) 42762 * predefined macros: Run-time Target. (line 6) 42763 * predicates: Predicates. (line 6) 42764 * predicates and machine modes: Predicates. (line 31) 42765 * predication: Conditional Execution. 42766 (line 6) 42767 * predict.def: Profile information. 42768 (line 24) 42769 * PREFERRED_DEBUGGING_TYPE: All Debuggers. (line 42) 42770 * PREFERRED_OUTPUT_RELOAD_CLASS: Register Classes. (line 231) 42771 * PREFERRED_RELOAD_CLASS: Register Classes. (line 196) 42772 * PREFERRED_STACK_BOUNDARY: Storage Layout. (line 158) 42773 * prefetch: Side Effects. (line 312) 42774 * prefetch instruction pattern: Standard Names. (line 1392) 42775 * PREINCREMENT_EXPR: Expression trees. (line 6) 42776 * presence_set: Processor pipeline description. 42777 (line 220) 42778 * preserving SSA form: SSA. (line 76) 42779 * preserving virtual SSA form: SSA. (line 186) 42780 * prev_active_insn: define_peephole. (line 60) 42781 * prev_cc0_setter: Jump Patterns. (line 64) 42782 * PREV_INSN: Insns. (line 26) 42783 * PRINT_OPERAND: Instruction Output. (line 68) 42784 * PRINT_OPERAND_ADDRESS: Instruction Output. (line 96) 42785 * PRINT_OPERAND_PUNCT_VALID_P: Instruction Output. (line 89) 42786 * processor functional units: Processor pipeline description. 42787 (line 68) 42788 * processor pipeline description: Processor pipeline description. 42789 (line 6) 42790 * product: Arithmetic. (line 92) 42791 * profile feedback: Profile information. 42792 (line 14) 42793 * profile representation: Profile information. 42794 (line 6) 42795 * PROFILE_BEFORE_PROLOGUE: Profiling. (line 35) 42796 * PROFILE_HOOK: Profiling. (line 23) 42797 * profiling, code generation: Profiling. (line 6) 42798 * program counter: Regs and Memory. (line 362) 42799 * prologue: Function Entry. (line 6) 42800 * prologue instruction pattern: Standard Names. (line 1338) 42801 * PROMOTE_FUNCTION_MODE: Storage Layout. (line 123) 42802 * PROMOTE_MODE: Storage Layout. (line 100) 42803 * pseudo registers: Regs and Memory. (line 9) 42804 * PSImode: Machine Modes. (line 32) 42805 * PTRDIFF_TYPE: Type Layout. (line 184) 42806 * PTRMEM_CST: Expression trees. (line 6) 42807 * PTRMEM_CST_CLASS: Expression trees. (line 6) 42808 * PTRMEM_CST_MEMBER: Expression trees. (line 6) 42809 * purge_dead_edges <1>: Maintaining the CFG. 42810 (line 93) 42811 * purge_dead_edges: Edges. (line 104) 42812 * push address instruction: Simple Constraints. (line 154) 42813 * PUSH_ARGS: Stack Arguments. (line 18) 42814 * PUSH_ARGS_REVERSED: Stack Arguments. (line 26) 42815 * push_operand: Machine-Independent Predicates. 42816 (line 81) 42817 * push_reload: Addressing Modes. (line 169) 42818 * PUSH_ROUNDING: Stack Arguments. (line 32) 42819 * pushM1 instruction pattern: Standard Names. (line 209) 42820 * PUT_CODE: RTL Objects. (line 47) 42821 * PUT_MODE: Machine Modes. (line 283) 42822 * PUT_REG_NOTE_KIND: Insns. (line 309) 42823 * PUT_SDB_: SDB and DWARF. (line 63) 42824 * QCmode: Machine Modes. (line 197) 42825 * QFmode: Machine Modes. (line 54) 42826 * QImode: Machine Modes. (line 25) 42827 * QImode, in insn: Insns. (line 231) 42828 * QQmode: Machine Modes. (line 103) 42829 * qualified type: Types. (line 6) 42830 * querying function unit reservations: Processor pipeline description. 42831 (line 90) 42832 * question mark: Multi-Alternative. (line 41) 42833 * quotient: Arithmetic. (line 111) 42834 * r in constraint: Simple Constraints. (line 56) 42835 * RANGE_TEST_NON_SHORT_CIRCUIT: Costs. (line 204) 42836 * RDIV_EXPR: Expression trees. (line 6) 42837 * READONLY_DATA_SECTION_ASM_OP: Sections. (line 63) 42838 * real operands: SSA Operands. (line 6) 42839 * REAL_ARITHMETIC: Floating Point. (line 66) 42840 * REAL_CST: Expression trees. (line 6) 42841 * REAL_LIBGCC_SPEC: Driver. (line 187) 42842 * REAL_NM_FILE_NAME: Macros for Initialization. 42843 (line 106) 42844 * REAL_TYPE: Types. (line 6) 42845 * REAL_VALUE_ABS: Floating Point. (line 82) 42846 * REAL_VALUE_ATOF: Floating Point. (line 50) 42847 * REAL_VALUE_FIX: Floating Point. (line 41) 42848 * REAL_VALUE_FROM_INT: Floating Point. (line 99) 42849 * REAL_VALUE_ISINF: Floating Point. (line 59) 42850 * REAL_VALUE_ISNAN: Floating Point. (line 62) 42851 * REAL_VALUE_NEGATE: Floating Point. (line 79) 42852 * REAL_VALUE_NEGATIVE: Floating Point. (line 56) 42853 * REAL_VALUE_TO_INT: Floating Point. (line 93) 42854 * REAL_VALUE_TO_TARGET_DECIMAL128: Data Output. (line 144) 42855 * REAL_VALUE_TO_TARGET_DECIMAL32: Data Output. (line 142) 42856 * REAL_VALUE_TO_TARGET_DECIMAL64: Data Output. (line 143) 42857 * REAL_VALUE_TO_TARGET_DOUBLE: Data Output. (line 140) 42858 * REAL_VALUE_TO_TARGET_LONG_DOUBLE: Data Output. (line 141) 42859 * REAL_VALUE_TO_TARGET_SINGLE: Data Output. (line 139) 42860 * REAL_VALUE_TRUNCATE: Floating Point. (line 86) 42861 * REAL_VALUE_TYPE: Floating Point. (line 26) 42862 * REAL_VALUE_UNSIGNED_FIX: Floating Point. (line 45) 42863 * REAL_VALUES_EQUAL: Floating Point. (line 32) 42864 * REAL_VALUES_LESS: Floating Point. (line 38) 42865 * REALPART_EXPR: Expression trees. (line 6) 42866 * recog_data.operand: Instruction Output. (line 39) 42867 * recognizing insns: RTL Template. (line 6) 42868 * RECORD_TYPE <1>: Types. (line 6) 42869 * RECORD_TYPE: Classes. (line 6) 42870 * redirect_edge_and_branch: Profile information. 42871 (line 71) 42872 * redirect_edge_and_branch, redirect_jump: Maintaining the CFG. 42873 (line 103) 42874 * reduc_smax_M instruction pattern: Standard Names. (line 240) 42875 * reduc_smin_M instruction pattern: Standard Names. (line 240) 42876 * reduc_splus_M instruction pattern: Standard Names. (line 252) 42877 * reduc_umax_M instruction pattern: Standard Names. (line 246) 42878 * reduc_umin_M instruction pattern: Standard Names. (line 246) 42879 * reduc_uplus_M instruction pattern: Standard Names. (line 258) 42880 * reference: Types. (line 6) 42881 * REFERENCE_TYPE: Types. (line 6) 42882 * reg: Regs and Memory. (line 9) 42883 * reg and /f: Flags. (line 112) 42884 * reg and /i: Flags. (line 107) 42885 * reg and /v: Flags. (line 116) 42886 * reg, RTL sharing: Sharing. (line 17) 42887 * REG_ALLOC_ORDER: Allocation Order. (line 9) 42888 * REG_BR_PRED: Insns. (line 491) 42889 * REG_BR_PROB: Insns. (line 485) 42890 * REG_BR_PROB_BASE, BB_FREQ_BASE, count: Profile information. 42891 (line 82) 42892 * REG_BR_PROB_BASE, EDGE_FREQUENCY: Profile information. 42893 (line 52) 42894 * REG_CC_SETTER: Insns. (line 456) 42895 * REG_CC_USER: Insns. (line 456) 42896 * reg_class_contents: Register Basics. (line 59) 42897 * REG_CLASS_CONTENTS: Register Classes. (line 86) 42898 * REG_CLASS_FROM_CONSTRAINT: Old Constraints. (line 35) 42899 * REG_CLASS_FROM_LETTER: Old Constraints. (line 27) 42900 * REG_CLASS_NAMES: Register Classes. (line 81) 42901 * REG_CROSSING_JUMP: Insns. (line 368) 42902 * REG_DEAD: Insns. (line 320) 42903 * REG_DEAD, REG_UNUSED: Liveness information. 42904 (line 32) 42905 * REG_DEP_ANTI: Insns. (line 478) 42906 * REG_DEP_OUTPUT: Insns. (line 474) 42907 * REG_DEP_TRUE: Insns. (line 471) 42908 * REG_EH_REGION, EDGE_ABNORMAL_CALL: Edges. (line 110) 42909 * REG_EQUAL: Insns. (line 384) 42910 * REG_EQUIV: Insns. (line 384) 42911 * REG_EXPR: Special Accessors. (line 46) 42912 * REG_FRAME_RELATED_EXPR: Insns. (line 497) 42913 * REG_FUNCTION_VALUE_P: Flags. (line 107) 42914 * REG_INC: Insns. (line 336) 42915 * reg_label and /v: Flags. (line 65) 42916 * REG_LABEL_OPERAND: Insns. (line 350) 42917 * REG_LABEL_TARGET: Insns. (line 359) 42918 * reg_names <1>: Register Basics. (line 59) 42919 * reg_names: Instruction Output. (line 80) 42920 * REG_NONNEG: Insns. (line 342) 42921 * REG_NOTE_KIND: Insns. (line 309) 42922 * REG_NOTES: Insns. (line 283) 42923 * REG_OFFSET: Special Accessors. (line 50) 42924 * REG_OK_STRICT: Addressing Modes. (line 67) 42925 * REG_PARM_STACK_SPACE: Stack Arguments. (line 56) 42926 * REG_PARM_STACK_SPACE, and FUNCTION_ARG: Register Arguments. 42927 (line 52) 42928 * REG_POINTER: Flags. (line 112) 42929 * REG_SETJMP: Insns. (line 378) 42930 * REG_UNUSED: Insns. (line 329) 42931 * REG_USERVAR_P: Flags. (line 116) 42932 * regclass_for_constraint: C Constraint Interface. 42933 (line 60) 42934 * register allocation order: Allocation Order. (line 6) 42935 * register class definitions: Register Classes. (line 6) 42936 * register class preference constraints: Class Preferences. (line 6) 42937 * register pairs: Values in Registers. 42938 (line 69) 42939 * Register Transfer Language (RTL): RTL. (line 6) 42940 * register usage: Registers. (line 6) 42941 * REGISTER_MOVE_COST: Costs. (line 10) 42942 * REGISTER_NAMES: Instruction Output. (line 9) 42943 * register_operand: Machine-Independent Predicates. 42944 (line 30) 42945 * REGISTER_PREFIX: Instruction Output. (line 124) 42946 * REGISTER_TARGET_PRAGMAS: Misc. (line 382) 42947 * registers arguments: Register Arguments. (line 6) 42948 * registers in constraints: Simple Constraints. (line 56) 42949 * REGMODE_NATURAL_SIZE: Values in Registers. 42950 (line 50) 42951 * REGNO_MODE_CODE_OK_FOR_BASE_P: Register Classes. (line 170) 42952 * REGNO_MODE_OK_FOR_BASE_P: Register Classes. (line 146) 42953 * REGNO_MODE_OK_FOR_REG_BASE_P: Register Classes. (line 157) 42954 * REGNO_OK_FOR_BASE_P: Register Classes. (line 140) 42955 * REGNO_OK_FOR_INDEX_P: Register Classes. (line 181) 42956 * REGNO_REG_CLASS: Register Classes. (line 101) 42957 * regs_ever_live: Function Entry. (line 21) 42958 * regular expressions: Processor pipeline description. 42959 (line 6) 42960 * relative costs: Costs. (line 6) 42961 * RELATIVE_PREFIX_NOT_LINKDIR: Driver. (line 325) 42962 * reload_completed: Standard Names. (line 1040) 42963 * reload_in instruction pattern: Standard Names. (line 99) 42964 * reload_in_progress: Standard Names. (line 57) 42965 * reload_out instruction pattern: Standard Names. (line 99) 42966 * reloading: RTL passes. (line 182) 42967 * remainder: Arithmetic. (line 131) 42968 * remainderM3 instruction pattern: Standard Names. (line 472) 42969 * reorder: GTY Options. (line 210) 42970 * representation of RTL: RTL. (line 6) 42971 * reservation delays: Processor pipeline description. 42972 (line 6) 42973 * rest_of_decl_compilation: Parsing pass. (line 52) 42974 * rest_of_type_compilation: Parsing pass. (line 52) 42975 * restore_stack_block instruction pattern: Standard Names. (line 1174) 42976 * restore_stack_function instruction pattern: Standard Names. 42977 (line 1174) 42978 * restore_stack_nonlocal instruction pattern: Standard Names. 42979 (line 1174) 42980 * RESULT_DECL: Declarations. (line 6) 42981 * return: Side Effects. (line 72) 42982 * return instruction pattern: Standard Names. (line 1027) 42983 * return values in registers: Scalar Return. (line 6) 42984 * RETURN_ADDR_IN_PREVIOUS_FRAME: Frame Layout. (line 135) 42985 * RETURN_ADDR_OFFSET: Exception Handling. (line 60) 42986 * RETURN_ADDR_RTX: Frame Layout. (line 124) 42987 * RETURN_ADDRESS_POINTER_REGNUM: Frame Registers. (line 51) 42988 * RETURN_EXPR: Function Bodies. (line 6) 42989 * RETURN_POPS_ARGS: Stack Arguments. (line 90) 42990 * RETURN_STMT: Function Bodies. (line 6) 42991 * return_val: Flags. (line 294) 42992 * return_val, in call_insn: Flags. (line 24) 42993 * return_val, in mem: Flags. (line 85) 42994 * return_val, in reg: Flags. (line 107) 42995 * return_val, in symbol_ref: Flags. (line 220) 42996 * returning aggregate values: Aggregate Return. (line 6) 42997 * returning structures and unions: Interface. (line 10) 42998 * reverse probability: Profile information. 42999 (line 66) 43000 * REVERSE_CONDEXEC_PREDICATES_P: Condition Code. (line 129) 43001 * REVERSE_CONDITION: Condition Code. (line 116) 43002 * REVERSIBLE_CC_MODE: Condition Code. (line 102) 43003 * right rotate: Arithmetic. (line 190) 43004 * right shift: Arithmetic. (line 185) 43005 * rintM2 instruction pattern: Standard Names. (line 572) 43006 * RISC: Processor pipeline description. 43007 (line 220) 43008 * roots, marking: GGC Roots. (line 6) 43009 * rotate: Arithmetic. (line 190) 43010 * rotatert: Arithmetic. (line 190) 43011 * rotlM3 instruction pattern: Standard Names. (line 441) 43012 * rotrM3 instruction pattern: Standard Names. (line 441) 43013 * ROUND_DIV_EXPR: Expression trees. (line 6) 43014 * ROUND_MOD_EXPR: Expression trees. (line 6) 43015 * ROUND_TOWARDS_ZERO: Storage Layout. (line 460) 43016 * ROUND_TYPE_ALIGN: Storage Layout. (line 411) 43017 * roundM2 instruction pattern: Standard Names. (line 548) 43018 * RSHIFT_EXPR: Expression trees. (line 6) 43019 * RTL addition: Arithmetic. (line 14) 43020 * RTL addition with signed saturation: Arithmetic. (line 14) 43021 * RTL addition with unsigned saturation: Arithmetic. (line 14) 43022 * RTL classes: RTL Classes. (line 6) 43023 * RTL comparison: Arithmetic. (line 43) 43024 * RTL comparison operations: Comparisons. (line 6) 43025 * RTL constant expression types: Constants. (line 6) 43026 * RTL constants: Constants. (line 6) 43027 * RTL declarations: RTL Declarations. (line 6) 43028 * RTL difference: Arithmetic. (line 36) 43029 * RTL expression: RTL Objects. (line 6) 43030 * RTL expressions for arithmetic: Arithmetic. (line 6) 43031 * RTL format: RTL Classes. (line 71) 43032 * RTL format characters: RTL Classes. (line 76) 43033 * RTL function-call insns: Calls. (line 6) 43034 * RTL insn template: RTL Template. (line 6) 43035 * RTL integers: RTL Objects. (line 6) 43036 * RTL memory expressions: Regs and Memory. (line 6) 43037 * RTL object types: RTL Objects. (line 6) 43038 * RTL postdecrement: Incdec. (line 6) 43039 * RTL postincrement: Incdec. (line 6) 43040 * RTL predecrement: Incdec. (line 6) 43041 * RTL preincrement: Incdec. (line 6) 43042 * RTL register expressions: Regs and Memory. (line 6) 43043 * RTL representation: RTL. (line 6) 43044 * RTL side effect expressions: Side Effects. (line 6) 43045 * RTL strings: RTL Objects. (line 6) 43046 * RTL structure sharing assumptions: Sharing. (line 6) 43047 * RTL subtraction: Arithmetic. (line 36) 43048 * RTL subtraction with signed saturation: Arithmetic. (line 36) 43049 * RTL subtraction with unsigned saturation: Arithmetic. (line 36) 43050 * RTL sum: Arithmetic. (line 14) 43051 * RTL vectors: RTL Objects. (line 6) 43052 * RTL_CONST_CALL_P: Flags. (line 19) 43053 * RTL_CONST_OR_PURE_CALL_P: Flags. (line 29) 43054 * RTL_LOOPING_CONST_OR_PURE_CALL_P: Flags. (line 33) 43055 * RTL_PURE_CALL_P: Flags. (line 24) 43056 * RTX (See RTL): RTL Objects. (line 6) 43057 * RTX codes, classes of: RTL Classes. (line 6) 43058 * RTX_FRAME_RELATED_P: Flags. (line 125) 43059 * run-time conventions: Interface. (line 6) 43060 * run-time target specification: Run-time Target. (line 6) 43061 * s in constraint: Simple Constraints. (line 92) 43062 * same_type_p: Types. (line 148) 43063 * SAmode: Machine Modes. (line 148) 43064 * sat_fract: Conversions. (line 90) 43065 * satfractMN2 instruction pattern: Standard Names. (line 843) 43066 * satfractunsMN2 instruction pattern: Standard Names. (line 856) 43067 * satisfies_constraint_: C Constraint Interface. 43068 (line 47) 43069 * SAVE_EXPR: Expression trees. (line 6) 43070 * save_stack_block instruction pattern: Standard Names. (line 1174) 43071 * save_stack_function instruction pattern: Standard Names. (line 1174) 43072 * save_stack_nonlocal instruction pattern: Standard Names. (line 1174) 43073 * SBSS_SECTION_ASM_OP: Sections. (line 77) 43074 * Scalar evolutions: Scalar evolutions. (line 6) 43075 * scalars, returned as values: Scalar Return. (line 6) 43076 * SCHED_GROUP_P: Flags. (line 166) 43077 * SCmode: Machine Modes. (line 197) 43078 * sCOND instruction pattern: Standard Names. (line 911) 43079 * scratch: Regs and Memory. (line 298) 43080 * scratch operands: Regs and Memory. (line 298) 43081 * scratch, RTL sharing: Sharing. (line 35) 43082 * scratch_operand: Machine-Independent Predicates. 43083 (line 50) 43084 * SDATA_SECTION_ASM_OP: Sections. (line 58) 43085 * SDB_ALLOW_FORWARD_REFERENCES: SDB and DWARF. (line 81) 43086 * SDB_ALLOW_UNKNOWN_REFERENCES: SDB and DWARF. (line 76) 43087 * SDB_DEBUGGING_INFO: SDB and DWARF. (line 9) 43088 * SDB_DELIM: SDB and DWARF. (line 69) 43089 * SDB_OUTPUT_SOURCE_LINE: SDB and DWARF. (line 86) 43090 * SDmode: Machine Modes. (line 85) 43091 * sdot_prodM instruction pattern: Standard Names. (line 264) 43092 * search options: Including Patterns. (line 44) 43093 * SECONDARY_INPUT_RELOAD_CLASS: Register Classes. (line 335) 43094 * SECONDARY_MEMORY_NEEDED: Register Classes. (line 391) 43095 * SECONDARY_MEMORY_NEEDED_MODE: Register Classes. (line 410) 43096 * SECONDARY_MEMORY_NEEDED_RTX: Register Classes. (line 401) 43097 * SECONDARY_OUTPUT_RELOAD_CLASS: Register Classes. (line 336) 43098 * SECONDARY_RELOAD_CLASS: Register Classes. (line 334) 43099 * SELECT_CC_MODE: Condition Code. (line 68) 43100 * sequence: Side Effects. (line 254) 43101 * Sequence iterators: Sequence iterators. (line 6) 43102 * set: Side Effects. (line 15) 43103 * set and /f: Flags. (line 125) 43104 * SET_ASM_OP: Label Output. (line 389) 43105 * set_attr: Tagging Insns. (line 31) 43106 * set_attr_alternative: Tagging Insns. (line 49) 43107 * set_bb_seq: GIMPLE sequences. (line 76) 43108 * SET_BY_PIECES_P: Costs. (line 145) 43109 * SET_DEST: Side Effects. (line 69) 43110 * SET_IS_RETURN_P: Flags. (line 175) 43111 * SET_LABEL_KIND: Insns. (line 140) 43112 * set_optab_libfunc: Library Calls. (line 15) 43113 * SET_RATIO: Costs. (line 136) 43114 * SET_SRC: Side Effects. (line 69) 43115 * SET_TYPE_STRUCTURAL_EQUALITY: Types. (line 6) 43116 * setmemM instruction pattern: Standard Names. (line 715) 43117 * SETUP_FRAME_ADDRESSES: Frame Layout. (line 102) 43118 * SF_SIZE: Type Layout. (line 129) 43119 * SFmode: Machine Modes. (line 66) 43120 * sharing of RTL components: Sharing. (line 6) 43121 * shift: Arithmetic. (line 168) 43122 * SHIFT_COUNT_TRUNCATED: Misc. (line 127) 43123 * SHLIB_SUFFIX: Macros for Initialization. 43124 (line 129) 43125 * SHORT_ACCUM_TYPE_SIZE: Type Layout. (line 83) 43126 * SHORT_FRACT_TYPE_SIZE: Type Layout. (line 63) 43127 * SHORT_IMMEDIATES_SIGN_EXTEND: Misc. (line 96) 43128 * SHORT_TYPE_SIZE: Type Layout. (line 16) 43129 * sibcall_epilogue instruction pattern: Standard Names. (line 1364) 43130 * sibling call: Edges. (line 122) 43131 * SIBLING_CALL_P: Flags. (line 179) 43132 * sign_extend: Conversions. (line 23) 43133 * sign_extract: Bit-Fields. (line 8) 43134 * sign_extract, canonicalization of: Insn Canonicalizations. 43135 (line 96) 43136 * signed division: Arithmetic. (line 111) 43137 * signed division with signed saturation: Arithmetic. (line 111) 43138 * signed maximum: Arithmetic. (line 136) 43139 * signed minimum: Arithmetic. (line 136) 43140 * SImode: Machine Modes. (line 37) 43141 * simple constraints: Simple Constraints. (line 6) 43142 * sincos math function, implicit usage: Library Calls. (line 84) 43143 * sinM2 instruction pattern: Standard Names. (line 489) 43144 * SIZE_ASM_OP: Label Output. (line 23) 43145 * SIZE_TYPE: Type Layout. (line 168) 43146 * skip: GTY Options. (line 77) 43147 * SLOW_BYTE_ACCESS: Costs. (line 66) 43148 * SLOW_UNALIGNED_ACCESS: Costs. (line 81) 43149 * SMALL_REGISTER_CLASSES: Register Classes. (line 433) 43150 * smax: Arithmetic. (line 136) 43151 * smin: Arithmetic. (line 136) 43152 * sms, swing, software pipelining: RTL passes. (line 131) 43153 * smulM3_highpart instruction pattern: Standard Names. (line 356) 43154 * soft float library: Soft float library routines. 43155 (line 6) 43156 * special: GTY Options. (line 230) 43157 * special predicates: Predicates. (line 31) 43158 * SPECS: Target Fragment. (line 108) 43159 * speed of instructions: Costs. (line 6) 43160 * split_block: Maintaining the CFG. 43161 (line 110) 43162 * splitting instructions: Insn Splitting. (line 6) 43163 * SQmode: Machine Modes. (line 111) 43164 * sqrt: Arithmetic. (line 198) 43165 * sqrtM2 instruction pattern: Standard Names. (line 455) 43166 * square root: Arithmetic. (line 198) 43167 * ss_ashift: Arithmetic. (line 168) 43168 * ss_div: Arithmetic. (line 111) 43169 * ss_minus: Arithmetic. (line 36) 43170 * ss_mult: Arithmetic. (line 92) 43171 * ss_neg: Arithmetic. (line 81) 43172 * ss_plus: Arithmetic. (line 14) 43173 * ss_truncate: Conversions. (line 43) 43174 * SSA: SSA. (line 6) 43175 * SSA_NAME_DEF_STMT: SSA. (line 221) 43176 * SSA_NAME_VERSION: SSA. (line 226) 43177 * ssaddM3 instruction pattern: Standard Names. (line 222) 43178 * ssashlM3 instruction pattern: Standard Names. (line 431) 43179 * ssdivM3 instruction pattern: Standard Names. (line 222) 43180 * ssmaddMN4 instruction pattern: Standard Names. (line 379) 43181 * ssmsubMN4 instruction pattern: Standard Names. (line 403) 43182 * ssmulM3 instruction pattern: Standard Names. (line 222) 43183 * ssnegM2 instruction pattern: Standard Names. (line 449) 43184 * sssubM3 instruction pattern: Standard Names. (line 222) 43185 * ssum_widenM3 instruction pattern: Standard Names. (line 274) 43186 * stack arguments: Stack Arguments. (line 6) 43187 * stack frame layout: Frame Layout. (line 6) 43188 * stack smashing protection: Stack Smashing Protection. 43189 (line 6) 43190 * STACK_ALIGNMENT_NEEDED: Frame Layout. (line 48) 43191 * STACK_BOUNDARY: Storage Layout. (line 150) 43192 * STACK_CHECK_BUILTIN: Stack Checking. (line 32) 43193 * STACK_CHECK_FIXED_FRAME_SIZE: Stack Checking. (line 77) 43194 * STACK_CHECK_MAX_FRAME_SIZE: Stack Checking. (line 68) 43195 * STACK_CHECK_MAX_VAR_SIZE: Stack Checking. (line 84) 43196 * STACK_CHECK_PROBE_INTERVAL: Stack Checking. (line 46) 43197 * STACK_CHECK_PROBE_LOAD: Stack Checking. (line 53) 43198 * STACK_CHECK_PROTECT: Stack Checking. (line 59) 43199 * STACK_CHECK_STATIC_BUILTIN: Stack Checking. (line 39) 43200 * STACK_DYNAMIC_OFFSET: Frame Layout. (line 75) 43201 * STACK_DYNAMIC_OFFSET and virtual registers: Regs and Memory. 43202 (line 83) 43203 * STACK_GROWS_DOWNWARD: Frame Layout. (line 9) 43204 * STACK_PARMS_IN_REG_PARM_AREA: Stack Arguments. (line 81) 43205 * STACK_POINTER_OFFSET: Frame Layout. (line 58) 43206 * STACK_POINTER_OFFSET and virtual registers: Regs and Memory. 43207 (line 93) 43208 * STACK_POINTER_REGNUM: Frame Registers. (line 9) 43209 * STACK_POINTER_REGNUM and virtual registers: Regs and Memory. 43210 (line 83) 43211 * stack_pointer_rtx: Frame Registers. (line 85) 43212 * stack_protect_set instruction pattern: Standard Names. (line 1534) 43213 * stack_protect_test instruction pattern: Standard Names. (line 1544) 43214 * STACK_PUSH_CODE: Frame Layout. (line 17) 43215 * STACK_REGS: Stack Registers. (line 20) 43216 * STACK_SAVEAREA_MODE: Storage Layout. (line 427) 43217 * STACK_SIZE_MODE: Storage Layout. (line 439) 43218 * STACK_SLOT_ALIGNMENT: Storage Layout. (line 265) 43219 * standard pattern names: Standard Names. (line 6) 43220 * STANDARD_INCLUDE_COMPONENT: Driver. (line 425) 43221 * STANDARD_INCLUDE_DIR: Driver. (line 417) 43222 * STANDARD_STARTFILE_PREFIX: Driver. (line 337) 43223 * STANDARD_STARTFILE_PREFIX_1: Driver. (line 344) 43224 * STANDARD_STARTFILE_PREFIX_2: Driver. (line 351) 43225 * STARTFILE_SPEC: Driver. (line 210) 43226 * STARTING_FRAME_OFFSET: Frame Layout. (line 39) 43227 * STARTING_FRAME_OFFSET and virtual registers: Regs and Memory. 43228 (line 74) 43229 * Statement and operand traversals: Statement and operand traversals. 43230 (line 6) 43231 * Statement Sequences: Statement Sequences. 43232 (line 6) 43233 * Statements: Statements. (line 6) 43234 * statements: Function Bodies. (line 6) 43235 * Static profile estimation: Profile information. 43236 (line 24) 43237 * static single assignment: SSA. (line 6) 43238 * STATIC_CHAIN: Frame Registers. (line 77) 43239 * STATIC_CHAIN_INCOMING: Frame Registers. (line 78) 43240 * STATIC_CHAIN_INCOMING_REGNUM: Frame Registers. (line 64) 43241 * STATIC_CHAIN_REGNUM: Frame Registers. (line 63) 43242 * stdarg.h and register arguments: Register Arguments. (line 47) 43243 * STDC_0_IN_SYSTEM_HEADERS: Misc. (line 365) 43244 * STMT_EXPR: Expression trees. (line 6) 43245 * STMT_IS_FULL_EXPR_P: Function Bodies. (line 22) 43246 * storage layout: Storage Layout. (line 6) 43247 * STORE_BY_PIECES_P: Costs. (line 152) 43248 * STORE_FLAG_VALUE: Misc. (line 216) 43249 * store_multiple instruction pattern: Standard Names. (line 160) 43250 * strcpy: Storage Layout. (line 235) 43251 * STRICT_ALIGNMENT: Storage Layout. (line 309) 43252 * strict_low_part: RTL Declarations. (line 9) 43253 * strict_memory_address_p: Addressing Modes. (line 179) 43254 * STRING_CST: Expression trees. (line 6) 43255 * STRING_POOL_ADDRESS_P: Flags. (line 183) 43256 * strlenM instruction pattern: Standard Names. (line 778) 43257 * structure value address: Aggregate Return. (line 6) 43258 * STRUCTURE_SIZE_BOUNDARY: Storage Layout. (line 301) 43259 * structures, returning: Interface. (line 10) 43260 * subM3 instruction pattern: Standard Names. (line 222) 43261 * SUBOBJECT: Function Bodies. (line 6) 43262 * SUBOBJECT_CLEANUP: Function Bodies. (line 6) 43263 * subreg: Regs and Memory. (line 97) 43264 * subreg and /s: Flags. (line 205) 43265 * subreg and /u: Flags. (line 198) 43266 * subreg and /u and /v: Flags. (line 188) 43267 * subreg, in strict_low_part: RTL Declarations. (line 9) 43268 * SUBREG_BYTE: Regs and Memory. (line 289) 43269 * SUBREG_PROMOTED_UNSIGNED_P: Flags. (line 188) 43270 * SUBREG_PROMOTED_UNSIGNED_SET: Flags. (line 198) 43271 * SUBREG_PROMOTED_VAR_P: Flags. (line 205) 43272 * SUBREG_REG: Regs and Memory. (line 289) 43273 * SUCCESS_EXIT_CODE: Host Misc. (line 12) 43274 * SUPPORTS_INIT_PRIORITY: Macros for Initialization. 43275 (line 58) 43276 * SUPPORTS_ONE_ONLY: Label Output. (line 227) 43277 * SUPPORTS_WEAK: Label Output. (line 208) 43278 * SWITCH_BODY: Function Bodies. (line 6) 43279 * SWITCH_COND: Function Bodies. (line 6) 43280 * SWITCH_CURTAILS_COMPILATION: Driver. (line 33) 43281 * SWITCH_STMT: Function Bodies. (line 6) 43282 * SWITCH_TAKES_ARG: Driver. (line 9) 43283 * SWITCHES_NEED_SPACES: Driver. (line 47) 43284 * SYMBOL_FLAG_ANCHOR: Special Accessors. (line 106) 43285 * SYMBOL_FLAG_EXTERNAL: Special Accessors. (line 88) 43286 * SYMBOL_FLAG_FUNCTION: Special Accessors. (line 81) 43287 * SYMBOL_FLAG_HAS_BLOCK_INFO: Special Accessors. (line 102) 43288 * SYMBOL_FLAG_LOCAL: Special Accessors. (line 84) 43289 * SYMBOL_FLAG_SMALL: Special Accessors. (line 93) 43290 * SYMBOL_FLAG_TLS_SHIFT: Special Accessors. (line 97) 43291 * symbol_ref: Constants. (line 76) 43292 * symbol_ref and /f: Flags. (line 183) 43293 * symbol_ref and /i: Flags. (line 220) 43294 * symbol_ref and /u: Flags. (line 10) 43295 * symbol_ref and /v: Flags. (line 224) 43296 * symbol_ref, RTL sharing: Sharing. (line 20) 43297 * SYMBOL_REF_ANCHOR_P: Special Accessors. (line 106) 43298 * SYMBOL_REF_BLOCK: Special Accessors. (line 119) 43299 * SYMBOL_REF_BLOCK_OFFSET: Special Accessors. (line 124) 43300 * SYMBOL_REF_CONSTANT: Special Accessors. (line 67) 43301 * SYMBOL_REF_DATA: Special Accessors. (line 71) 43302 * SYMBOL_REF_DECL: Special Accessors. (line 55) 43303 * SYMBOL_REF_EXTERNAL_P: Special Accessors. (line 88) 43304 * SYMBOL_REF_FLAG: Flags. (line 224) 43305 * SYMBOL_REF_FLAG, in TARGET_ENCODE_SECTION_INFO: Sections. (line 259) 43306 * SYMBOL_REF_FLAGS: Special Accessors. (line 75) 43307 * SYMBOL_REF_FUNCTION_P: Special Accessors. (line 81) 43308 * SYMBOL_REF_HAS_BLOCK_INFO_P: Special Accessors. (line 102) 43309 * SYMBOL_REF_LOCAL_P: Special Accessors. (line 84) 43310 * SYMBOL_REF_SMALL_P: Special Accessors. (line 93) 43311 * SYMBOL_REF_TLS_MODEL: Special Accessors. (line 97) 43312 * SYMBOL_REF_USED: Flags. (line 215) 43313 * SYMBOL_REF_WEAK: Flags. (line 220) 43314 * symbolic label: Sharing. (line 20) 43315 * sync_addMODE instruction pattern: Standard Names. (line 1450) 43316 * sync_andMODE instruction pattern: Standard Names. (line 1450) 43317 * sync_compare_and_swap_ccMODE instruction pattern: Standard Names. 43318 (line 1437) 43319 * sync_compare_and_swapMODE instruction pattern: Standard Names. 43320 (line 1419) 43321 * sync_iorMODE instruction pattern: Standard Names. (line 1450) 43322 * sync_lock_releaseMODE instruction pattern: Standard Names. (line 1515) 43323 * sync_lock_test_and_setMODE instruction pattern: Standard Names. 43324 (line 1489) 43325 * sync_nandMODE instruction pattern: Standard Names. (line 1450) 43326 * sync_new_addMODE instruction pattern: Standard Names. (line 1482) 43327 * sync_new_andMODE instruction pattern: Standard Names. (line 1482) 43328 * sync_new_iorMODE instruction pattern: Standard Names. (line 1482) 43329 * sync_new_nandMODE instruction pattern: Standard Names. (line 1482) 43330 * sync_new_subMODE instruction pattern: Standard Names. (line 1482) 43331 * sync_new_xorMODE instruction pattern: Standard Names. (line 1482) 43332 * sync_old_addMODE instruction pattern: Standard Names. (line 1465) 43333 * sync_old_andMODE instruction pattern: Standard Names. (line 1465) 43334 * sync_old_iorMODE instruction pattern: Standard Names. (line 1465) 43335 * sync_old_nandMODE instruction pattern: Standard Names. (line 1465) 43336 * sync_old_subMODE instruction pattern: Standard Names. (line 1465) 43337 * sync_old_xorMODE instruction pattern: Standard Names. (line 1465) 43338 * sync_subMODE instruction pattern: Standard Names. (line 1450) 43339 * sync_xorMODE instruction pattern: Standard Names. (line 1450) 43340 * SYSROOT_HEADERS_SUFFIX_SPEC: Driver. (line 239) 43341 * SYSROOT_SUFFIX_SPEC: Driver. (line 234) 43342 * SYSTEM_INCLUDE_DIR: Driver. (line 408) 43343 * t-TARGET: Target Fragment. (line 6) 43344 * table jump: Basic Blocks. (line 57) 43345 * tablejump instruction pattern: Standard Names. (line 1102) 43346 * tag: GTY Options. (line 82) 43347 * tagging insns: Tagging Insns. (line 6) 43348 * tail calls: Tail Calls. (line 6) 43349 * TAmode: Machine Modes. (line 156) 43350 * target attributes: Target Attributes. (line 6) 43351 * target description macros: Target Macros. (line 6) 43352 * target functions: Target Structure. (line 6) 43353 * target hooks: Target Structure. (line 6) 43354 * target makefile fragment: Target Fragment. (line 6) 43355 * target specifications: Run-time Target. (line 6) 43356 * TARGET_ADDRESS_COST: Costs. (line 236) 43357 * TARGET_ALIGN_ANON_BITFIELD: Storage Layout. (line 386) 43358 * TARGET_ALLOCATE_INITIAL_VALUE: Misc. (line 712) 43359 * TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS: Misc. (line 951) 43360 * TARGET_ARG_PARTIAL_BYTES: Register Arguments. (line 83) 43361 * TARGET_ARM_EABI_UNWINDER: Exception Region Output. 43362 (line 113) 43363 * TARGET_ASM_ALIGNED_DI_OP: Data Output. (line 10) 43364 * TARGET_ASM_ALIGNED_HI_OP: Data Output. (line 8) 43365 * TARGET_ASM_ALIGNED_SI_OP: Data Output. (line 9) 43366 * TARGET_ASM_ALIGNED_TI_OP: Data Output. (line 11) 43367 * TARGET_ASM_ASSEMBLE_VISIBILITY: Label Output. (line 239) 43368 * TARGET_ASM_BYTE_OP: Data Output. (line 7) 43369 * TARGET_ASM_CAN_OUTPUT_MI_THUNK: Function Entry. (line 237) 43370 * TARGET_ASM_CLOSE_PAREN: Data Output. (line 130) 43371 * TARGET_ASM_CONSTRUCTOR: Macros for Initialization. 43372 (line 69) 43373 * TARGET_ASM_DESTRUCTOR: Macros for Initialization. 43374 (line 83) 43375 * TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL: Dispatch Tables. (line 74) 43376 * TARGET_ASM_EMIT_UNWIND_LABEL: Dispatch Tables. (line 63) 43377 * TARGET_ASM_EXTERNAL_LIBCALL: Label Output. (line 274) 43378 * TARGET_ASM_FILE_END: File Framework. (line 37) 43379 * TARGET_ASM_FILE_START: File Framework. (line 9) 43380 * TARGET_ASM_FILE_START_APP_OFF: File Framework. (line 17) 43381 * TARGET_ASM_FILE_START_FILE_DIRECTIVE: File Framework. (line 31) 43382 * TARGET_ASM_FUNCTION_BEGIN_EPILOGUE: Function Entry. (line 61) 43383 * TARGET_ASM_FUNCTION_END_PROLOGUE: Function Entry. (line 55) 43384 * TARGET_ASM_FUNCTION_EPILOGUE: Function Entry. (line 68) 43385 * TARGET_ASM_FUNCTION_EPILOGUE and trampolines: Trampolines. (line 70) 43386 * TARGET_ASM_FUNCTION_PROLOGUE: Function Entry. (line 11) 43387 * TARGET_ASM_FUNCTION_PROLOGUE and trampolines: Trampolines. (line 70) 43388 * TARGET_ASM_FUNCTION_RODATA_SECTION: Sections. (line 206) 43389 * TARGET_ASM_GLOBALIZE_DECL_NAME: Label Output. (line 174) 43390 * TARGET_ASM_GLOBALIZE_LABEL: Label Output. (line 165) 43391 * TARGET_ASM_INIT_SECTIONS: Sections. (line 151) 43392 * TARGET_ASM_INTEGER: Data Output. (line 27) 43393 * TARGET_ASM_INTERNAL_LABEL: Label Output. (line 309) 43394 * TARGET_ASM_MARK_DECL_PRESERVED: Label Output. (line 280) 43395 * TARGET_ASM_NAMED_SECTION: File Framework. (line 89) 43396 * TARGET_ASM_OPEN_PAREN: Data Output. (line 129) 43397 * TARGET_ASM_OUTPUT_ANCHOR: Anchored Addresses. (line 44) 43398 * TARGET_ASM_OUTPUT_DWARF_DTPREL: SDB and DWARF. (line 58) 43399 * TARGET_ASM_OUTPUT_MI_THUNK: Function Entry. (line 195) 43400 * TARGET_ASM_RECORD_GCC_SWITCHES: File Framework. (line 122) 43401 * TARGET_ASM_RECORD_GCC_SWITCHES_SECTION: File Framework. (line 166) 43402 * TARGET_ASM_SELECT_RTX_SECTION: Sections. (line 214) 43403 * TARGET_ASM_SELECT_SECTION: Sections. (line 172) 43404 * TARGET_ASM_TTYPE: Exception Region Output. 43405 (line 107) 43406 * TARGET_ASM_UNALIGNED_DI_OP: Data Output. (line 14) 43407 * TARGET_ASM_UNALIGNED_HI_OP: Data Output. (line 12) 43408 * TARGET_ASM_UNALIGNED_SI_OP: Data Output. (line 13) 43409 * TARGET_ASM_UNALIGNED_TI_OP: Data Output. (line 15) 43410 * TARGET_ASM_UNIQUE_SECTION: Sections. (line 193) 43411 * TARGET_ATTRIBUTE_TABLE: Target Attributes. (line 11) 43412 * TARGET_BINDS_LOCAL_P: Sections. (line 284) 43413 * TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED: Misc. (line 808) 43414 * TARGET_BRANCH_TARGET_REGISTER_CLASS: Misc. (line 800) 43415 * TARGET_BUILD_BUILTIN_VA_LIST: Register Arguments. (line 264) 43416 * TARGET_BUILTIN_RECIPROCAL: Addressing Modes. (line 240) 43417 * TARGET_BUILTIN_SETJMP_FRAME_VALUE: Frame Layout. (line 109) 43418 * TARGET_C99_FUNCTIONS: Library Calls. (line 77) 43419 * TARGET_CALLEE_COPIES: Register Arguments. (line 115) 43420 * TARGET_CAN_INLINE_P: Target Attributes. (line 126) 43421 * TARGET_CAN_SIMPLIFY_GOT_ACCESS: Misc. (line 983) 43422 * TARGET_CANNOT_FORCE_CONST_MEM: Addressing Modes. (line 221) 43423 * TARGET_CANNOT_MODIFY_JUMPS_P: Misc. (line 787) 43424 * TARGET_CANONICAL_VA_LIST_TYPE: Register Arguments. (line 273) 43425 * TARGET_CLEAR_PIC_REG: Misc. (line 966) 43426 * TARGET_COMMUTATIVE_P: Misc. (line 705) 43427 * TARGET_COMP_TYPE_ATTRIBUTES: Target Attributes. (line 19) 43428 * TARGET_CPU_CPP_BUILTINS: Run-time Target. (line 9) 43429 * TARGET_CXX_ADJUST_CLASS_AT_DEFINITION: C++ ABI. (line 87) 43430 * TARGET_CXX_CDTOR_RETURNS_THIS: C++ ABI. (line 38) 43431 * TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT: C++ ABI. (line 62) 43432 * TARGET_CXX_COOKIE_HAS_SIZE: C++ ABI. (line 25) 43433 * TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY: C++ ABI. (line 54) 43434 * TARGET_CXX_GET_COOKIE_SIZE: C++ ABI. (line 18) 43435 * TARGET_CXX_GUARD_MASK_BIT: C++ ABI. (line 12) 43436 * TARGET_CXX_GUARD_TYPE: C++ ABI. (line 7) 43437 * TARGET_CXX_IMPORT_EXPORT_CLASS: C++ ABI. (line 30) 43438 * TARGET_CXX_KEY_METHOD_MAY_BE_INLINE: C++ ABI. (line 43) 43439 * TARGET_CXX_LIBRARY_RTTI_COMDAT: C++ ABI. (line 69) 43440 * TARGET_CXX_USE_AEABI_ATEXIT: C++ ABI. (line 74) 43441 * TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT: C++ ABI. (line 80) 43442 * TARGET_DECIMAL_FLOAT_SUPPORTED_P: Storage Layout. (line 513) 43443 * TARGET_DECLSPEC: Target Attributes. (line 64) 43444 * TARGET_DEFAULT_PACK_STRUCT: Misc. (line 482) 43445 * TARGET_DEFAULT_SHORT_ENUMS: Type Layout. (line 160) 43446 * TARGET_DEFERRED_OUTPUT_DEFS: Label Output. (line 393) 43447 * TARGET_DELEGITIMIZE_ADDRESS: Addressing Modes. (line 212) 43448 * TARGET_DLLIMPORT_DECL_ATTRIBUTES: Target Attributes. (line 47) 43449 * TARGET_DWARF_CALLING_CONVENTION: SDB and DWARF. (line 18) 43450 * TARGET_DWARF_HANDLE_FRAME_UNSPEC: Frame Layout. (line 172) 43451 * TARGET_DWARF_REGISTER_SPAN: Exception Region Output. 43452 (line 90) 43453 * TARGET_EDOM: Library Calls. (line 59) 43454 * TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS: Emulated TLS. (line 68) 43455 * TARGET_EMUTLS_GET_ADDRESS: Emulated TLS. (line 19) 43456 * TARGET_EMUTLS_REGISTER_COMMON: Emulated TLS. (line 24) 43457 * TARGET_EMUTLS_TMPL_PREFIX: Emulated TLS. (line 45) 43458 * TARGET_EMUTLS_TMPL_SECTION: Emulated TLS. (line 36) 43459 * TARGET_EMUTLS_VAR_ALIGN_FIXED: Emulated TLS. (line 63) 43460 * TARGET_EMUTLS_VAR_FIELDS: Emulated TLS. (line 49) 43461 * TARGET_EMUTLS_VAR_INIT: Emulated TLS. (line 57) 43462 * TARGET_EMUTLS_VAR_PREFIX: Emulated TLS. (line 41) 43463 * TARGET_EMUTLS_VAR_SECTION: Emulated TLS. (line 31) 43464 * TARGET_ENCODE_SECTION_INFO: Sections. (line 235) 43465 * TARGET_ENCODE_SECTION_INFO and address validation: Addressing Modes. 43466 (line 91) 43467 * TARGET_ENCODE_SECTION_INFO usage: Instruction Output. (line 100) 43468 * TARGET_ENUM_VA_LIST: Scalar Return. (line 96) 43469 * TARGET_EXECUTABLE_SUFFIX: Misc. (line 761) 43470 * TARGET_EXPAND_BUILTIN: Misc. (line 657) 43471 * TARGET_EXPAND_BUILTIN_SAVEREGS: Varargs. (line 92) 43472 * TARGET_EXPAND_TO_RTL_HOOK: Storage Layout. (line 519) 43473 * TARGET_EXPR: Expression trees. (line 6) 43474 * TARGET_EXTRA_INCLUDES: Misc. (line 839) 43475 * TARGET_EXTRA_LIVE_ON_ENTRY: Tail Calls. (line 21) 43476 * TARGET_EXTRA_PRE_INCLUDES: Misc. (line 846) 43477 * TARGET_FIXED_CONDITION_CODE_REGS: Condition Code. (line 142) 43478 * TARGET_FIXED_POINT_SUPPORTED_P: Storage Layout. (line 516) 43479 * target_flags: Run-time Target. (line 52) 43480 * TARGET_FLT_EVAL_METHOD: Type Layout. (line 141) 43481 * TARGET_FN_ABI_VA_LIST: Register Arguments. (line 268) 43482 * TARGET_FOLD_BUILTIN: Misc. (line 677) 43483 * TARGET_FORMAT_TYPES: Misc. (line 866) 43484 * TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P: Target Attributes. (line 86) 43485 * TARGET_FUNCTION_OK_FOR_SIBCALL: Tail Calls. (line 8) 43486 * TARGET_FUNCTION_VALUE: Scalar Return. (line 11) 43487 * TARGET_GET_DRAP_RTX: Misc. (line 946) 43488 * TARGET_GET_PIC_REG: Misc. (line 962) 43489 * TARGET_GIMPLIFY_VA_ARG_EXPR: Register Arguments. (line 279) 43490 * TARGET_HANDLE_C_OPTION: Run-time Target. (line 78) 43491 * TARGET_HANDLE_OPTION: Run-time Target. (line 61) 43492 * TARGET_HARD_REGNO_SCRATCH_OK: Values in Registers. 43493 (line 144) 43494 * TARGET_HAS_SINCOS: Library Calls. (line 85) 43495 * TARGET_HAVE_CONDITIONAL_EXECUTION: Misc. (line 822) 43496 * TARGET_HAVE_CTORS_DTORS: Macros for Initialization. 43497 (line 64) 43498 * TARGET_HAVE_NAMED_SECTIONS: File Framework. (line 99) 43499 * TARGET_HAVE_SWITCHABLE_BSS_SECTIONS: File Framework. (line 103) 43500 * TARGET_HELP: Run-time Target. (line 140) 43501 * TARGET_IN_SMALL_DATA_P: Sections. (line 276) 43502 * TARGET_INIT_BUILTINS: Misc. (line 639) 43503 * TARGET_INIT_DWARF_REG_SIZES_EXTRA: Exception Region Output. 43504 (line 99) 43505 * TARGET_INIT_LIBFUNCS: Library Calls. (line 16) 43506 * TARGET_INSERT_ATTRIBUTES: Target Attributes. (line 73) 43507 * TARGET_INSTANTIATE_DECLS: Storage Layout. (line 527) 43508 * TARGET_INVALID_BINARY_OP: Misc. (line 919) 43509 * TARGET_INVALID_CONVERSION: Misc. (line 906) 43510 * TARGET_INVALID_UNARY_OP: Misc. (line 912) 43511 * TARGET_IRA_COVER_CLASSES: Register Classes. (line 496) 43512 * TARGET_LIB_INT_CMP_BIASED: Library Calls. (line 35) 43513 * TARGET_LIBCALL_VALUE: Scalar Return. (line 69) 43514 * TARGET_LIBGCC_CMP_RETURN_MODE: Storage Layout. (line 448) 43515 * TARGET_LIBGCC_SDATA_SECTION: Sections. (line 123) 43516 * TARGET_LIBGCC_SHIFT_COUNT_MODE: Storage Layout. (line 454) 43517 * TARGET_LOAD_GLOBAL_ADDRESS: Misc. (line 991) 43518 * TARGET_LOADED_GLOBAL_VAR: Misc. (line 971) 43519 * TARGET_MACHINE_DEPENDENT_REORG: Misc. (line 624) 43520 * TARGET_MANGLE_DECL_ASSEMBLER_NAME: Sections. (line 225) 43521 * TARGET_MANGLE_TYPE: Storage Layout. (line 531) 43522 * TARGET_MD_ASM_CLOBBERS: Misc. (line 540) 43523 * TARGET_MEM_CONSTRAINT: Addressing Modes. (line 100) 43524 * TARGET_MEM_REF: Expression trees. (line 6) 43525 * TARGET_MERGE_DECL_ATTRIBUTES: Target Attributes. (line 39) 43526 * TARGET_MERGE_TYPE_ATTRIBUTES: Target Attributes. (line 31) 43527 * TARGET_MIN_DIVISIONS_FOR_RECIP_MUL: Misc. (line 106) 43528 * TARGET_MODE_REP_EXTENDED: Misc. (line 191) 43529 * TARGET_MS_BITFIELD_LAYOUT_P: Storage Layout. (line 486) 43530 * TARGET_MUST_PASS_IN_STACK: Register Arguments. (line 62) 43531 * TARGET_MUST_PASS_IN_STACK, and FUNCTION_ARG: Register Arguments. 43532 (line 52) 43533 * TARGET_N_FORMAT_TYPES: Misc. (line 871) 43534 * TARGET_NARROW_VOLATILE_BITFIELD: Storage Layout. (line 392) 43535 * TARGET_OBJECT_SUFFIX: Misc. (line 756) 43536 * TARGET_OBJFMT_CPP_BUILTINS: Run-time Target. (line 46) 43537 * TARGET_OPTF: Misc. (line 853) 43538 * TARGET_OPTION_PRAGMA_PARSE: Target Attributes. (line 120) 43539 * TARGET_OPTION_PRINT: Target Attributes. (line 115) 43540 * TARGET_OPTION_RESTORE: Target Attributes. (line 110) 43541 * TARGET_OPTION_SAVE: Target Attributes. (line 104) 43542 * TARGET_OPTION_TRANSLATE_TABLE: Driver. (line 53) 43543 * TARGET_OS_CPP_BUILTINS: Run-time Target. (line 42) 43544 * TARGET_OVERRIDES_FORMAT_ATTRIBUTES: Misc. (line 875) 43545 * TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT: Misc. (line 881) 43546 * TARGET_OVERRIDES_FORMAT_INIT: Misc. (line 885) 43547 * TARGET_PASS_BY_REFERENCE: Register Arguments. (line 103) 43548 * TARGET_POSIX_IO: Misc. (line 564) 43549 * TARGET_PRETEND_OUTGOING_VARARGS_NAMED: Varargs. (line 152) 43550 * TARGET_PROMOTE_FUNCTION_ARGS: Storage Layout. (line 131) 43551 * TARGET_PROMOTE_FUNCTION_RETURN: Storage Layout. (line 136) 43552 * TARGET_PROMOTE_PROTOTYPES: Stack Arguments. (line 11) 43553 * TARGET_PTRMEMFUNC_VBIT_LOCATION: Type Layout. (line 235) 43554 * TARGET_RELAXED_ORDERING: Misc. (line 890) 43555 * TARGET_RESOLVE_OVERLOADED_BUILTIN: Misc. (line 667) 43556 * TARGET_RETURN_IN_MEMORY: Aggregate Return. (line 16) 43557 * TARGET_RETURN_IN_MSB: Scalar Return. (line 112) 43558 * TARGET_RTX_COSTS: Costs. (line 210) 43559 * TARGET_SCALAR_MODE_SUPPORTED_P: Register Arguments. (line 291) 43560 * TARGET_SCHED_ADJUST_COST: Scheduling. (line 37) 43561 * TARGET_SCHED_ADJUST_PRIORITY: Scheduling. (line 52) 43562 * TARGET_SCHED_CLEAR_SCHED_CONTEXT: Scheduling. (line 283) 43563 * TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK: Scheduling. (line 89) 43564 * TARGET_SCHED_DFA_NEW_CYCLE: Scheduling. (line 205) 43565 * TARGET_SCHED_DFA_POST_CYCLE_ADVANCE: Scheduling. (line 160) 43566 * TARGET_SCHED_DFA_POST_CYCLE_INSN: Scheduling. (line 144) 43567 * TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE: Scheduling. (line 153) 43568 * TARGET_SCHED_DFA_PRE_CYCLE_INSN: Scheduling. (line 132) 43569 * TARGET_SCHED_FINISH: Scheduling. (line 109) 43570 * TARGET_SCHED_FINISH_GLOBAL: Scheduling. (line 126) 43571 * TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD: Scheduling. 43572 (line 168) 43573 * TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD: Scheduling. 43574 (line 196) 43575 * TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC: Scheduling. 43576 (line 321) 43577 * TARGET_SCHED_FREE_SCHED_CONTEXT: Scheduling. (line 265) 43578 * TARGET_SCHED_GEN_CHECK: Scheduling. (line 309) 43579 * TARGET_SCHED_H_I_D_EXTENDED: Scheduling. (line 241) 43580 * TARGET_SCHED_INIT: Scheduling. (line 99) 43581 * TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN: Scheduling. (line 149) 43582 * TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN: Scheduling. (line 141) 43583 * TARGET_SCHED_INIT_GLOBAL: Scheduling. (line 118) 43584 * TARGET_SCHED_INIT_SCHED_CONTEXT: Scheduling. (line 251) 43585 * TARGET_SCHED_IS_COSTLY_DEPENDENCE: Scheduling. (line 219) 43586 * TARGET_SCHED_ISSUE_RATE: Scheduling. (line 12) 43587 * TARGET_SCHED_NEEDS_BLOCK_P: Scheduling. (line 302) 43588 * TARGET_SCHED_REORDER: Scheduling. (line 60) 43589 * TARGET_SCHED_REORDER2: Scheduling. (line 77) 43590 * TARGET_SCHED_SET_SCHED_CONTEXT: Scheduling. (line 279) 43591 * TARGET_SCHED_SET_SCHED_FLAGS: Scheduling. (line 332) 43592 * TARGET_SCHED_SMS_RES_MII: Scheduling. (line 343) 43593 * TARGET_SCHED_SPECULATE_INSN: Scheduling. (line 291) 43594 * TARGET_SCHED_VARIABLE_ISSUE: Scheduling. (line 24) 43595 * TARGET_SECONDARY_RELOAD: Register Classes. (line 257) 43596 * TARGET_SECTION_TYPE_FLAGS: File Framework. (line 109) 43597 * TARGET_SET_CURRENT_FUNCTION: Misc. (line 739) 43598 * TARGET_SET_DEFAULT_TYPE_ATTRIBUTES: Target Attributes. (line 26) 43599 * TARGET_SETUP_INCOMING_VARARGS: Varargs. (line 101) 43600 * TARGET_SHIFT_TRUNCATION_MASK: Misc. (line 154) 43601 * TARGET_SPLIT_COMPLEX_ARG: Register Arguments. (line 252) 43602 * TARGET_STACK_PROTECT_FAIL: Stack Smashing Protection. 43603 (line 17) 43604 * TARGET_STACK_PROTECT_GUARD: Stack Smashing Protection. 43605 (line 7) 43606 * TARGET_STRICT_ARGUMENT_NAMING: Varargs. (line 137) 43607 * TARGET_STRUCT_VALUE_RTX: Aggregate Return. (line 44) 43608 * TARGET_UNSPEC_MAY_TRAP_P: Misc. (line 731) 43609 * TARGET_UNWIND_EMIT: Dispatch Tables. (line 81) 43610 * TARGET_UNWIND_INFO: Exception Region Output. 43611 (line 56) 43612 * TARGET_UPDATE_STACK_BOUNDARY: Misc. (line 942) 43613 * TARGET_USE_ANCHORS_FOR_SYMBOL_P: Anchored Addresses. (line 55) 43614 * TARGET_USE_BLOCKS_FOR_CONSTANT_P: Addressing Modes. (line 233) 43615 * TARGET_USE_JCR_SECTION: Misc. (line 924) 43616 * TARGET_USE_LOCAL_THUNK_ALIAS_P: Misc. (line 859) 43617 * TARGET_USES_WEAK_UNWIND_INFO: Exception Handling. (line 129) 43618 * TARGET_VALID_DLLIMPORT_ATTRIBUTE_P: Target Attributes. (line 59) 43619 * TARGET_VALID_OPTION_ATTRIBUTE_P: Target Attributes. (line 93) 43620 * TARGET_VALID_POINTER_MODE: Register Arguments. (line 285) 43621 * TARGET_VECTOR_MODE_SUPPORTED_P: Register Arguments. (line 303) 43622 * TARGET_VECTOR_OPAQUE_P: Storage Layout. (line 479) 43623 * TARGET_VECTORIZE_BUILTIN_CONVERSION: Addressing Modes. (line 300) 43624 * TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD: Addressing Modes. (line 249) 43625 * TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN: Addressing Modes. (line 275) 43626 * TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD: Addressing Modes. (line 287) 43627 * TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION: Addressing Modes. 43628 (line 315) 43629 * TARGET_VERSION: Run-time Target. (line 91) 43630 * TARGET_VTABLE_DATA_ENTRY_DISTANCE: Type Layout. (line 288) 43631 * TARGET_VTABLE_ENTRY_ALIGN: Type Layout. (line 282) 43632 * TARGET_VTABLE_USES_DESCRIPTORS: Type Layout. (line 271) 43633 * TARGET_WEAK_NOT_IN_ARCHIVE_TOC: Label Output. (line 245) 43634 * targetm: Target Structure. (line 7) 43635 * targets, makefile: Makefile. (line 6) 43636 * TCmode: Machine Modes. (line 197) 43637 * TDmode: Machine Modes. (line 94) 43638 * TEMPLATE_DECL: Declarations. (line 6) 43639 * Temporaries: Temporaries. (line 6) 43640 * termination routines: Initialization. (line 6) 43641 * testing constraints: C Constraint Interface. 43642 (line 6) 43643 * TEXT_SECTION_ASM_OP: Sections. (line 38) 43644 * TF_SIZE: Type Layout. (line 132) 43645 * TFmode: Machine Modes. (line 98) 43646 * THEN_CLAUSE: Function Bodies. (line 6) 43647 * THREAD_MODEL_SPEC: Driver. (line 225) 43648 * THROW_EXPR: Expression trees. (line 6) 43649 * THUNK_DECL: Declarations. (line 6) 43650 * THUNK_DELTA: Declarations. (line 6) 43651 * TImode: Machine Modes. (line 48) 43652 * TImode, in insn: Insns. (line 231) 43653 * tm.h macros: Target Macros. (line 6) 43654 * TQFmode: Machine Modes. (line 62) 43655 * TQmode: Machine Modes. (line 119) 43656 * TRAMPOLINE_ADJUST_ADDRESS: Trampolines. (line 62) 43657 * TRAMPOLINE_ALIGNMENT: Trampolines. (line 49) 43658 * TRAMPOLINE_SECTION: Trampolines. (line 40) 43659 * TRAMPOLINE_SIZE: Trampolines. (line 45) 43660 * TRAMPOLINE_TEMPLATE: Trampolines. (line 29) 43661 * trampolines for nested functions: Trampolines. (line 6) 43662 * TRANSFER_FROM_TRAMPOLINE: Trampolines. (line 124) 43663 * trap instruction pattern: Standard Names. (line 1374) 43664 * tree <1>: Tree overview. (line 6) 43665 * tree: Macros and Functions. 43666 (line 6) 43667 * Tree SSA: Tree SSA. (line 6) 43668 * tree_code <1>: GIMPLE_OMP_FOR. (line 83) 43669 * tree_code <2>: GIMPLE_ASSIGN. (line 41) 43670 * tree_code <3>: GIMPLE_COND. (line 21) 43671 * tree_code: Manipulating GIMPLE statements. 43672 (line 31) 43673 * TREE_CODE: Tree overview. (line 6) 43674 * TREE_FILENAME: Working with declarations. 43675 (line 14) 43676 * tree_int_cst_equal: Expression trees. (line 6) 43677 * TREE_INT_CST_HIGH: Expression trees. (line 6) 43678 * TREE_INT_CST_LOW: Expression trees. (line 6) 43679 * tree_int_cst_lt: Expression trees. (line 6) 43680 * TREE_LINENO: Working with declarations. 43681 (line 20) 43682 * TREE_LIST: Containers. (line 6) 43683 * TREE_OPERAND: Expression trees. (line 6) 43684 * TREE_PUBLIC: Function Basics. (line 41) 43685 * TREE_PURPOSE: Containers. (line 6) 43686 * TREE_STRING_LENGTH: Expression trees. (line 6) 43687 * TREE_STRING_POINTER: Expression trees. (line 6) 43688 * TREE_TYPE <1>: Expression trees. (line 6) 43689 * TREE_TYPE <2>: Types. (line 6) 43690 * TREE_TYPE <3>: Working with declarations. 43691 (line 11) 43692 * TREE_TYPE <4>: Expression trees. (line 17) 43693 * TREE_TYPE: Function Basics. (line 171) 43694 * TREE_VALUE: Containers. (line 6) 43695 * TREE_VEC: Containers. (line 6) 43696 * TREE_VEC_ELT: Containers. (line 6) 43697 * TREE_VEC_LENGTH: Containers. (line 6) 43698 * Trees: Trees. (line 6) 43699 * TRULY_NOOP_TRUNCATION: Misc. (line 177) 43700 * TRUNC_DIV_EXPR: Expression trees. (line 6) 43701 * TRUNC_MOD_EXPR: Expression trees. (line 6) 43702 * truncate: Conversions. (line 38) 43703 * truncMN2 instruction pattern: Standard Names. (line 821) 43704 * TRUTH_AND_EXPR: Expression trees. (line 6) 43705 * TRUTH_ANDIF_EXPR: Expression trees. (line 6) 43706 * TRUTH_NOT_EXPR: Expression trees. (line 6) 43707 * TRUTH_OR_EXPR: Expression trees. (line 6) 43708 * TRUTH_ORIF_EXPR: Expression trees. (line 6) 43709 * TRUTH_XOR_EXPR: Expression trees. (line 6) 43710 * TRY_BLOCK: Function Bodies. (line 6) 43711 * TRY_HANDLERS: Function Bodies. (line 6) 43712 * TRY_STMTS: Function Bodies. (line 6) 43713 * tstM instruction pattern: Standard Names. (line 661) 43714 * Tuple specific accessors: Tuple specific accessors. 43715 (line 6) 43716 * tuples: Tuple representation. 43717 (line 6) 43718 * type: Types. (line 6) 43719 * type declaration: Declarations. (line 6) 43720 * TYPE_ALIGN: Types. (line 61) 43721 * TYPE_ARG_TYPES: Types. (line 6) 43722 * TYPE_ASM_OP: Label Output. (line 55) 43723 * TYPE_ATTRIBUTES: Attributes. (line 25) 43724 * TYPE_BINFO: Classes. (line 6) 43725 * TYPE_BUILT_IN: Types. (line 83) 43726 * TYPE_CANONICAL: Types. (line 6) 43727 * TYPE_CONTEXT: Types. (line 6) 43728 * TYPE_DECL: Declarations. (line 6) 43729 * TYPE_FIELDS <1>: Classes. (line 6) 43730 * TYPE_FIELDS: Types. (line 6) 43731 * TYPE_HAS_ARRAY_NEW_OPERATOR: Classes. (line 91) 43732 * TYPE_HAS_DEFAULT_CONSTRUCTOR: Classes. (line 76) 43733 * TYPE_HAS_MUTABLE_P: Classes. (line 81) 43734 * TYPE_HAS_NEW_OPERATOR: Classes. (line 88) 43735 * TYPE_MAIN_VARIANT: Types. (line 50) 43736 * TYPE_MAX_VALUE: Types. (line 6) 43737 * TYPE_METHOD_BASETYPE: Types. (line 6) 43738 * TYPE_METHODS: Classes. (line 6) 43739 * TYPE_MIN_VALUE: Types. (line 6) 43740 * TYPE_NAME: Types. (line 6) 43741 * TYPE_NOTHROW_P: Function Basics. (line 180) 43742 * TYPE_OFFSET_BASETYPE: Types. (line 6) 43743 * TYPE_OPERAND_FMT: Label Output. (line 66) 43744 * TYPE_OVERLOADS_ARRAY_REF: Classes. (line 99) 43745 * TYPE_OVERLOADS_ARROW: Classes. (line 102) 43746 * TYPE_OVERLOADS_CALL_EXPR: Classes. (line 95) 43747 * TYPE_POLYMORPHIC_P: Classes. (line 72) 43748 * TYPE_PRECISION: Types. (line 6) 43749 * TYPE_PTR_P: Types. (line 89) 43750 * TYPE_PTRFN_P: Types. (line 93) 43751 * TYPE_PTRMEM_P: Types. (line 6) 43752 * TYPE_PTROB_P: Types. (line 96) 43753 * TYPE_PTROBV_P: Types. (line 6) 43754 * TYPE_QUAL_CONST: Types. (line 6) 43755 * TYPE_QUAL_RESTRICT: Types. (line 6) 43756 * TYPE_QUAL_VOLATILE: Types. (line 6) 43757 * TYPE_RAISES_EXCEPTIONS: Function Basics. (line 175) 43758 * TYPE_SIZE: Types. (line 6) 43759 * TYPE_STRUCTURAL_EQUALITY_P: Types. (line 6) 43760 * TYPE_UNQUALIFIED: Types. (line 6) 43761 * TYPE_VFIELD: Classes. (line 6) 43762 * TYPENAME_TYPE: Types. (line 6) 43763 * TYPENAME_TYPE_FULLNAME: Types. (line 6) 43764 * TYPEOF_TYPE: Types. (line 6) 43765 * UDAmode: Machine Modes. (line 168) 43766 * udiv: Arithmetic. (line 125) 43767 * udivM3 instruction pattern: Standard Names. (line 222) 43768 * udivmodM4 instruction pattern: Standard Names. (line 428) 43769 * udot_prodM instruction pattern: Standard Names. (line 265) 43770 * UDQmode: Machine Modes. (line 136) 43771 * UHAmode: Machine Modes. (line 160) 43772 * UHQmode: Machine Modes. (line 128) 43773 * UINTMAX_TYPE: Type Layout. (line 224) 43774 * umaddMN4 instruction pattern: Standard Names. (line 375) 43775 * umax: Arithmetic. (line 144) 43776 * umaxM3 instruction pattern: Standard Names. (line 222) 43777 * umin: Arithmetic. (line 144) 43778 * uminM3 instruction pattern: Standard Names. (line 222) 43779 * umod: Arithmetic. (line 131) 43780 * umodM3 instruction pattern: Standard Names. (line 222) 43781 * umsubMN4 instruction pattern: Standard Names. (line 399) 43782 * umulhisi3 instruction pattern: Standard Names. (line 347) 43783 * umulM3_highpart instruction pattern: Standard Names. (line 361) 43784 * umulqihi3 instruction pattern: Standard Names. (line 347) 43785 * umulsidi3 instruction pattern: Standard Names. (line 347) 43786 * unchanging: Flags. (line 319) 43787 * unchanging, in call_insn: Flags. (line 19) 43788 * unchanging, in jump_insn, call_insn and insn: Flags. (line 39) 43789 * unchanging, in mem: Flags. (line 152) 43790 * unchanging, in subreg: Flags. (line 188) 43791 * unchanging, in symbol_ref: Flags. (line 10) 43792 * UNEQ_EXPR: Expression trees. (line 6) 43793 * UNGE_EXPR: Expression trees. (line 6) 43794 * UNGT_EXPR: Expression trees. (line 6) 43795 * UNION_TYPE <1>: Types. (line 6) 43796 * UNION_TYPE: Classes. (line 6) 43797 * unions, returning: Interface. (line 10) 43798 * UNITS_PER_SIMD_WORD: Storage Layout. (line 77) 43799 * UNITS_PER_WORD: Storage Layout. (line 67) 43800 * UNKNOWN_TYPE: Types. (line 6) 43801 * UNLE_EXPR: Expression trees. (line 6) 43802 * UNLIKELY_EXECUTED_TEXT_SECTION_NAME: Sections. (line 49) 43803 * UNLT_EXPR: Expression trees. (line 6) 43804 * UNORDERED_EXPR: Expression trees. (line 6) 43805 * unshare_all_rtl: Sharing. (line 58) 43806 * unsigned division: Arithmetic. (line 125) 43807 * unsigned division with unsigned saturation: Arithmetic. (line 125) 43808 * unsigned greater than: Comparisons. (line 64) 43809 * unsigned less than: Comparisons. (line 68) 43810 * unsigned minimum and maximum: Arithmetic. (line 144) 43811 * unsigned_fix: Conversions. (line 77) 43812 * unsigned_float: Conversions. (line 62) 43813 * unsigned_fract_convert: Conversions. (line 97) 43814 * unsigned_sat_fract: Conversions. (line 103) 43815 * unspec: Side Effects. (line 287) 43816 * unspec_volatile: Side Effects. (line 287) 43817 * untyped_call instruction pattern: Standard Names. (line 1012) 43818 * untyped_return instruction pattern: Standard Names. (line 1062) 43819 * UPDATE_PATH_HOST_CANONICALIZE (PATH): Filesystem. (line 59) 43820 * update_ssa: SSA. (line 76) 43821 * update_stmt <1>: Manipulating GIMPLE statements. 43822 (line 141) 43823 * update_stmt: SSA Operands. (line 6) 43824 * update_stmt_if_modified: Manipulating GIMPLE statements. 43825 (line 144) 43826 * UQQmode: Machine Modes. (line 123) 43827 * US Software GOFAST, floating point emulation library: Library Calls. 43828 (line 44) 43829 * us_ashift: Arithmetic. (line 168) 43830 * us_minus: Arithmetic. (line 36) 43831 * us_mult: Arithmetic. (line 92) 43832 * us_neg: Arithmetic. (line 81) 43833 * us_plus: Arithmetic. (line 14) 43834 * US_SOFTWARE_GOFAST: Library Calls. (line 45) 43835 * us_truncate: Conversions. (line 48) 43836 * usaddM3 instruction pattern: Standard Names. (line 222) 43837 * USAmode: Machine Modes. (line 164) 43838 * usashlM3 instruction pattern: Standard Names. (line 431) 43839 * usdivM3 instruction pattern: Standard Names. (line 222) 43840 * use: Side Effects. (line 162) 43841 * USE_C_ALLOCA: Host Misc. (line 19) 43842 * USE_LD_AS_NEEDED: Driver. (line 198) 43843 * USE_LOAD_POST_DECREMENT: Costs. (line 165) 43844 * USE_LOAD_POST_INCREMENT: Costs. (line 160) 43845 * USE_LOAD_PRE_DECREMENT: Costs. (line 175) 43846 * USE_LOAD_PRE_INCREMENT: Costs. (line 170) 43847 * use_optype_d: Manipulating GIMPLE statements. 43848 (line 101) 43849 * use_param: GTY Options. (line 114) 43850 * use_paramN: GTY Options. (line 132) 43851 * use_params: GTY Options. (line 140) 43852 * USE_SELECT_SECTION_FOR_FUNCTIONS: Sections. (line 185) 43853 * USE_STORE_POST_DECREMENT: Costs. (line 185) 43854 * USE_STORE_POST_INCREMENT: Costs. (line 180) 43855 * USE_STORE_PRE_DECREMENT: Costs. (line 195) 43856 * USE_STORE_PRE_INCREMENT: Costs. (line 190) 43857 * used: Flags. (line 337) 43858 * used, in symbol_ref: Flags. (line 215) 43859 * USER_LABEL_PREFIX: Instruction Output. (line 126) 43860 * USING_DECL: Declarations. (line 6) 43861 * USING_STMT: Function Bodies. (line 6) 43862 * usmaddMN4 instruction pattern: Standard Names. (line 383) 43863 * usmsubMN4 instruction pattern: Standard Names. (line 407) 43864 * usmulhisi3 instruction pattern: Standard Names. (line 351) 43865 * usmulM3 instruction pattern: Standard Names. (line 222) 43866 * usmulqihi3 instruction pattern: Standard Names. (line 351) 43867 * usmulsidi3 instruction pattern: Standard Names. (line 351) 43868 * usnegM2 instruction pattern: Standard Names. (line 449) 43869 * USQmode: Machine Modes. (line 132) 43870 * ussubM3 instruction pattern: Standard Names. (line 222) 43871 * usum_widenM3 instruction pattern: Standard Names. (line 275) 43872 * UTAmode: Machine Modes. (line 172) 43873 * UTQmode: Machine Modes. (line 140) 43874 * V in constraint: Simple Constraints. (line 43) 43875 * VA_ARG_EXPR: Expression trees. (line 6) 43876 * values, returned by functions: Scalar Return. (line 6) 43877 * VAR_DECL <1>: Declarations. (line 6) 43878 * VAR_DECL: Expression trees. (line 6) 43879 * varargs implementation: Varargs. (line 6) 43880 * variable: Declarations. (line 6) 43881 * vashlM3 instruction pattern: Standard Names. (line 445) 43882 * vashrM3 instruction pattern: Standard Names. (line 445) 43883 * vec_concat: Vector Operations. (line 25) 43884 * vec_duplicate: Vector Operations. (line 30) 43885 * VEC_EXTRACT_EVEN_EXPR: Expression trees. (line 6) 43886 * vec_extract_evenM instruction pattern: Standard Names. (line 176) 43887 * VEC_EXTRACT_ODD_EXPR: Expression trees. (line 6) 43888 * vec_extract_oddM instruction pattern: Standard Names. (line 183) 43889 * vec_extractM instruction pattern: Standard Names. (line 171) 43890 * vec_initM instruction pattern: Standard Names. (line 204) 43891 * VEC_INTERLEAVE_HIGH_EXPR: Expression trees. (line 6) 43892 * vec_interleave_highM instruction pattern: Standard Names. (line 190) 43893 * VEC_INTERLEAVE_LOW_EXPR: Expression trees. (line 6) 43894 * vec_interleave_lowM instruction pattern: Standard Names. (line 197) 43895 * VEC_LSHIFT_EXPR: Expression trees. (line 6) 43896 * vec_merge: Vector Operations. (line 11) 43897 * VEC_PACK_FIX_TRUNC_EXPR: Expression trees. (line 6) 43898 * VEC_PACK_SAT_EXPR: Expression trees. (line 6) 43899 * vec_pack_sfix_trunc_M instruction pattern: Standard Names. (line 302) 43900 * vec_pack_ssat_M instruction pattern: Standard Names. (line 295) 43901 * VEC_PACK_TRUNC_EXPR: Expression trees. (line 6) 43902 * vec_pack_trunc_M instruction pattern: Standard Names. (line 288) 43903 * vec_pack_ufix_trunc_M instruction pattern: Standard Names. (line 302) 43904 * vec_pack_usat_M instruction pattern: Standard Names. (line 295) 43905 * VEC_RSHIFT_EXPR: Expression trees. (line 6) 43906 * vec_select: Vector Operations. (line 19) 43907 * vec_setM instruction pattern: Standard Names. (line 166) 43908 * vec_shl_M instruction pattern: Standard Names. (line 282) 43909 * vec_shr_M instruction pattern: Standard Names. (line 282) 43910 * VEC_UNPACK_FLOAT_HI_EXPR: Expression trees. (line 6) 43911 * VEC_UNPACK_FLOAT_LO_EXPR: Expression trees. (line 6) 43912 * VEC_UNPACK_HI_EXPR: Expression trees. (line 6) 43913 * VEC_UNPACK_LO_EXPR: Expression trees. (line 6) 43914 * vec_unpacks_float_hi_M instruction pattern: Standard Names. 43915 (line 324) 43916 * vec_unpacks_float_lo_M instruction pattern: Standard Names. 43917 (line 324) 43918 * vec_unpacks_hi_M instruction pattern: Standard Names. (line 309) 43919 * vec_unpacks_lo_M instruction pattern: Standard Names. (line 309) 43920 * vec_unpacku_float_hi_M instruction pattern: Standard Names. 43921 (line 324) 43922 * vec_unpacku_float_lo_M instruction pattern: Standard Names. 43923 (line 324) 43924 * vec_unpacku_hi_M instruction pattern: Standard Names. (line 317) 43925 * vec_unpacku_lo_M instruction pattern: Standard Names. (line 317) 43926 * VEC_WIDEN_MULT_HI_EXPR: Expression trees. (line 6) 43927 * VEC_WIDEN_MULT_LO_EXPR: Expression trees. (line 6) 43928 * vec_widen_smult_hi_M instruction pattern: Standard Names. (line 333) 43929 * vec_widen_smult_lo_M instruction pattern: Standard Names. (line 333) 43930 * vec_widen_umult_hi_M instruction pattern: Standard Names. (line 333) 43931 * vec_widen_umult_lo__M instruction pattern: Standard Names. (line 333) 43932 * vector: Containers. (line 6) 43933 * vector operations: Vector Operations. (line 6) 43934 * VECTOR_CST: Expression trees. (line 6) 43935 * VECTOR_STORE_FLAG_VALUE: Misc. (line 308) 43936 * virtual operands: SSA Operands. (line 6) 43937 * VIRTUAL_INCOMING_ARGS_REGNUM: Regs and Memory. (line 59) 43938 * VIRTUAL_OUTGOING_ARGS_REGNUM: Regs and Memory. (line 87) 43939 * VIRTUAL_STACK_DYNAMIC_REGNUM: Regs and Memory. (line 78) 43940 * VIRTUAL_STACK_VARS_REGNUM: Regs and Memory. (line 69) 43941 * VLIW: Processor pipeline description. 43942 (line 6) 43943 * vlshrM3 instruction pattern: Standard Names. (line 445) 43944 * VMS: Filesystem. (line 37) 43945 * VMS_DEBUGGING_INFO: VMS Debug. (line 9) 43946 * VOID_TYPE: Types. (line 6) 43947 * VOIDmode: Machine Modes. (line 190) 43948 * volatil: Flags. (line 351) 43949 * volatil, in insn, call_insn, jump_insn, code_label, barrier, and note: Flags. 43950 (line 44) 43951 * volatil, in label_ref and reg_label: Flags. (line 65) 43952 * volatil, in mem, asm_operands, and asm_input: Flags. (line 94) 43953 * volatil, in reg: Flags. (line 116) 43954 * volatil, in subreg: Flags. (line 188) 43955 * volatil, in symbol_ref: Flags. (line 224) 43956 * volatile memory references: Flags. (line 352) 43957 * voptype_d: Manipulating GIMPLE statements. 43958 (line 115) 43959 * voting between constraint alternatives: Class Preferences. (line 6) 43960 * vrotlM3 instruction pattern: Standard Names. (line 445) 43961 * vrotrM3 instruction pattern: Standard Names. (line 445) 43962 * walk_dominator_tree: SSA. (line 256) 43963 * walk_gimple_op: Statement and operand traversals. 43964 (line 32) 43965 * walk_gimple_seq: Statement and operand traversals. 43966 (line 50) 43967 * walk_gimple_stmt: Statement and operand traversals. 43968 (line 13) 43969 * walk_use_def_chains: SSA. (line 232) 43970 * WCHAR_TYPE: Type Layout. (line 192) 43971 * WCHAR_TYPE_SIZE: Type Layout. (line 200) 43972 * which_alternative: Output Statement. (line 59) 43973 * WHILE_BODY: Function Bodies. (line 6) 43974 * WHILE_COND: Function Bodies. (line 6) 43975 * WHILE_STMT: Function Bodies. (line 6) 43976 * WIDEST_HARDWARE_FP_SIZE: Type Layout. (line 147) 43977 * WINT_TYPE: Type Layout. (line 205) 43978 * word_mode: Machine Modes. (line 336) 43979 * WORD_REGISTER_OPERATIONS: Misc. (line 63) 43980 * WORD_SWITCH_TAKES_ARG: Driver. (line 20) 43981 * WORDS_BIG_ENDIAN: Storage Layout. (line 29) 43982 * WORDS_BIG_ENDIAN, effect on subreg: Regs and Memory. (line 217) 43983 * X in constraint: Simple Constraints. (line 114) 43984 * x-HOST: Host Fragment. (line 6) 43985 * XCmode: Machine Modes. (line 197) 43986 * XCOFF_DEBUGGING_INFO: DBX Options. (line 13) 43987 * XEXP: Accessors. (line 6) 43988 * XF_SIZE: Type Layout. (line 131) 43989 * XFmode: Machine Modes. (line 79) 43990 * XINT: Accessors. (line 6) 43991 * xm-MACHINE.h <1>: Filesystem. (line 6) 43992 * xm-MACHINE.h: Host Misc. (line 6) 43993 * xor: Arithmetic. (line 163) 43994 * xor, canonicalization of: Insn Canonicalizations. 43995 (line 84) 43996 * xorM3 instruction pattern: Standard Names. (line 222) 43997 * XSTR: Accessors. (line 6) 43998 * XVEC: Accessors. (line 41) 43999 * XVECEXP: Accessors. (line 48) 44000 * XVECLEN: Accessors. (line 44) 44001 * XWINT: Accessors. (line 6) 44002 * zero_extend: Conversions. (line 28) 44003 * zero_extendMN2 instruction pattern: Standard Names. (line 831) 44004 * zero_extract: Bit-Fields. (line 30) 44005 * zero_extract, canonicalization of: Insn Canonicalizations. 44006 (line 96) 44007 44008 44009 44010 Tag Table: 44011 Node: Top2090 44012 Node: Contributing5172 44013 Node: Portability5913 44014 Node: Interface7701 44015 Node: Libgcc10741 44016 Node: Integer library routines12582 44017 Node: Soft float library routines19421 44018 Node: Decimal float library routines31358 44019 Node: Fixed-point fractional library routines47115 44020 Node: Exception handling routines147513 44021 Node: Miscellaneous routines148620 44022 Node: Languages149003 44023 Node: Source Tree150550 44024 Node: Configure Terms151169 44025 Node: Top Level154127 44026 Node: gcc Directory156897 44027 Node: Subdirectories157866 44028 Node: Configuration159716 44029 Node: Config Fragments160436 44030 Node: System Config161665 44031 Node: Configuration Files162601 44032 Node: Build165176 44033 Node: Makefile165588 44034 Ref: Makefile-Footnote-1172306 44035 Ref: Makefile-Footnote-2172451 44036 Node: Library Files172523 44037 Node: Headers173085 44038 Node: Documentation175168 44039 Node: Texinfo Manuals176027 44040 Node: Man Page Generation178365 44041 Node: Miscellaneous Docs180280 44042 Node: Front End181579 44043 Node: Front End Directory185280 44044 Node: Front End Config190470 44045 Node: Back End193384 44046 Node: Testsuites197061 44047 Node: Test Idioms197925 44048 Node: Test Directives201326 44049 Node: Ada Tests213390 44050 Node: C Tests214682 44051 Node: libgcj Tests219037 44052 Node: gcov Testing220169 44053 Node: profopt Testing223153 44054 Node: compat Testing224596 44055 Node: Torture Tests228840 44056 Node: Options230472 44057 Node: Option file format230913 44058 Node: Option properties233662 44059 Node: Passes239718 44060 Node: Parsing pass240460 44061 Node: Gimplification pass243988 44062 Node: Pass manager245815 44063 Node: Tree SSA passes247298 44064 Node: RTL passes269129 44065 Node: Trees281472 44066 Node: Deficiencies284202 44067 Node: Tree overview284439 44068 Node: Macros and Functions288562 44069 Node: Identifiers288708 44070 Node: Containers290233 44071 Node: Types291388 44072 Node: Scopes307091 44073 Node: Namespaces307853 44074 Node: Classes310665 44075 Node: Declarations315422 44076 Node: Working with declarations315917 44077 Node: Internal structure322374 44078 Node: Current structure hierarchy322756 44079 Node: Adding new DECL node types324848 44080 Node: Functions328919 44081 Node: Function Basics331322 44082 Node: Function Bodies339052 44083 Node: Attributes350294 44084 Node: Expression trees351535 44085 Node: RTL394144 44086 Node: RTL Objects396262 44087 Node: RTL Classes400136 44088 Node: Accessors405088 44089 Node: Special Accessors407482 44090 Node: Flags412700 44091 Node: Machine Modes428568 44092 Node: Constants440884 44093 Node: Regs and Memory446913 44094 Node: Arithmetic464814 44095 Node: Comparisons474334 44096 Node: Bit-Fields478626 44097 Node: Vector Operations480178 44098 Node: Conversions481804 44099 Node: RTL Declarations486302 44100 Node: Side Effects487123 44101 Node: Incdec503446 44102 Node: Assembler506781 44103 Node: Insns508313 44104 Node: Calls532202 44105 Node: Sharing534795 44106 Node: Reading RTL537905 44107 Node: GENERIC538895 44108 Node: Statements540534 44109 Node: Blocks540979 44110 Node: Statement Sequences542232 44111 Node: Empty Statements542565 44112 Node: Jumps543139 44113 Node: Cleanups543792 44114 Node: GIMPLE545545 44115 Node: Tuple representation549166 44116 Node: GIMPLE instruction set557821 44117 Node: GIMPLE Exception Handling559489 44118 Node: Temporaries561404 44119 Ref: Temporaries-Footnote-1562723 44120 Node: Operands562786 44121 Node: Compound Expressions563560 44122 Node: Compound Lvalues563794 44123 Node: Conditional Expressions564560 44124 Node: Logical Operators565230 44125 Node: Manipulating GIMPLE statements571321 44126 Node: Tuple specific accessors577249 44127 Node: `GIMPLE_ASM'578082 44128 Node: `GIMPLE_ASSIGN'580687 44129 Node: `GIMPLE_BIND'584633 44130 Node: `GIMPLE_CALL'586440 44131 Node: `GIMPLE_CATCH'590699 44132 Node: `GIMPLE_CHANGE_DYNAMIC_TYPE'591857 44133 Node: `GIMPLE_COND'593190 44134 Node: `GIMPLE_EH_FILTER'595996 44135 Node: `GIMPLE_LABEL'597482 44136 Node: `GIMPLE_NOP'598457 44137 Node: `GIMPLE_OMP_ATOMIC_LOAD'598826 44138 Node: `GIMPLE_OMP_ATOMIC_STORE'599736 44139 Node: `GIMPLE_OMP_CONTINUE'600375 44140 Node: `GIMPLE_OMP_CRITICAL'601725 44141 Node: `GIMPLE_OMP_FOR'602661 44142 Node: `GIMPLE_OMP_MASTER'606171 44143 Node: `GIMPLE_OMP_ORDERED'606554 44144 Node: `GIMPLE_OMP_PARALLEL'606954 44145 Node: `GIMPLE_OMP_RETURN'609723 44146 Node: `GIMPLE_OMP_SECTION'610373 44147 Node: `GIMPLE_OMP_SECTIONS'611039 44148 Node: `GIMPLE_OMP_SINGLE'612643 44149 Node: `GIMPLE_PHI'613579 44150 Node: `GIMPLE_RESX'614992 44151 Node: `GIMPLE_RETURN'615711 44152 Node: `GIMPLE_SWITCH'616279 44153 Node: `GIMPLE_TRY'618409 44154 Node: `GIMPLE_WITH_CLEANUP_EXPR'620199 44155 Node: GIMPLE sequences621082 44156 Node: Sequence iterators624288 44157 Node: Adding a new GIMPLE statement code632743 44158 Node: Statement and operand traversals634023 44159 Node: Tree SSA636633 44160 Node: Annotations638353 44161 Node: SSA Operands638879 44162 Node: SSA653410 44163 Node: Alias analysis665701 44164 Node: Loop Analysis and Representation673157 44165 Node: Loop representation674338 44166 Node: Loop querying681258 44167 Node: Loop manipulation684091 44168 Node: LCSSA686459 44169 Node: Scalar evolutions688531 44170 Node: loop-iv691775 44171 Node: Number of iterations693701 44172 Node: Dependency analysis696510 44173 Node: Lambda702878 44174 Node: Omega704548 44175 Node: Control Flow706113 44176 Node: Basic Blocks707108 44177 Node: Edges711676 44178 Node: Profile information720238 44179 Node: Maintaining the CFG724924 44180 Node: Liveness information731806 44181 Node: Machine Desc733933 44182 Node: Overview736401 44183 Node: Patterns738442 44184 Node: Example741880 44185 Node: RTL Template743315 44186 Node: Output Template753970 44187 Node: Output Statement757936 44188 Node: Predicates761898 44189 Node: Machine-Independent Predicates764816 44190 Node: Defining Predicates769448 44191 Node: Constraints775413 44192 Node: Simple Constraints776661 44193 Node: Multi-Alternative788867 44194 Node: Class Preferences791708 44195 Node: Modifiers792600 44196 Node: Machine Constraints796732 44197 Node: Disable Insn Alternatives829455 44198 Node: Define Constraints832348 44199 Node: C Constraint Interface839128 44200 Node: Standard Names842769 44201 Ref: shift patterns861697 44202 Ref: prologue instruction pattern902715 44203 Ref: epilogue instruction pattern903208 44204 Node: Pattern Ordering912707 44205 Node: Dependent Patterns913943 44206 Node: Jump Patterns916757 44207 Node: Looping Patterns922453 44208 Node: Insn Canonicalizations927181 44209 Node: Expander Definitions931565 44210 Node: Insn Splitting939683 44211 Node: Including Patterns949286 44212 Node: Peephole Definitions951066 44213 Node: define_peephole952319 44214 Node: define_peephole2958650 44215 Node: Insn Attributes961717 44216 Node: Defining Attributes962823 44217 Node: Expressions965343 44218 Node: Tagging Insns971945 44219 Node: Attr Example976298 44220 Node: Insn Lengths978672 44221 Node: Constant Attributes981731 44222 Node: Delay Slots982900 44223 Node: Processor pipeline description986124 44224 Ref: Processor pipeline description-Footnote-11003742 44225 Node: Conditional Execution1004064 44226 Node: Constant Definitions1006917 44227 Node: Iterators1008512 44228 Node: Mode Iterators1008959 44229 Node: Defining Mode Iterators1009937 44230 Node: Substitutions1011431 44231 Node: Examples1013672 44232 Node: Code Iterators1015120 44233 Node: Target Macros1017377 44234 Node: Target Structure1020400 44235 Node: Driver1021669 44236 Node: Run-time Target1045350 44237 Node: Per-Function Data1052474 44238 Node: Storage Layout1055237 44239 Node: Type Layout1080651 44240 Node: Registers1093608 44241 Node: Register Basics1094582 44242 Node: Allocation Order1100149 44243 Node: Values in Registers1102595 44244 Node: Leaf Functions1110084 44245 Node: Stack Registers1112942 44246 Node: Register Classes1114058 44247 Node: Old Constraints1140770 44248 Node: Stack and Calling1147921 44249 Node: Frame Layout1148455 44250 Node: Exception Handling1159301 44251 Node: Stack Checking1165679 44252 Node: Frame Registers1170066 44253 Node: Elimination1176672 44254 Node: Stack Arguments1180703 44255 Node: Register Arguments1187506 44256 Node: Scalar Return1202959 44257 Node: Aggregate Return1209032 44258 Node: Caller Saves1212691 44259 Node: Function Entry1213869 44260 Node: Profiling1226484 44261 Node: Tail Calls1228183 44262 Node: Stack Smashing Protection1229550 44263 Node: Varargs1230662 44264 Node: Trampolines1238622 44265 Node: Library Calls1245288 44266 Node: Addressing Modes1250138 44267 Node: Anchored Addresses1266056 44268 Node: Condition Code1268717 44269 Node: Costs1277006 44270 Node: Scheduling1290105 44271 Node: Sections1308666 44272 Node: PIC1323316 44273 Node: Assembler Format1325306 44274 Node: File Framework1326444 44275 Ref: TARGET_HAVE_SWITCHABLE_BSS_SECTIONS1331350 44276 Node: Data Output1334616 44277 Node: Uninitialized Data1342375 44278 Node: Label Output1347446 44279 Node: Initialization1369113 44280 Node: Macros for Initialization1375075 44281 Node: Instruction Output1381527 44282 Node: Dispatch Tables1390521 44283 Node: Exception Region Output1394316 44284 Node: Alignment Output1400076 44285 Node: Debugging Info1404239 44286 Node: All Debuggers1404909 44287 Node: DBX Options1407764 44288 Node: DBX Hooks1413213 44289 Node: File Names and DBX1415139 44290 Node: SDB and DWARF1417250 44291 Node: VMS Debug1421242 44292 Node: Floating Point1421812 44293 Node: Mode Switching1426635 44294 Node: Target Attributes1430561 44295 Node: Emulated TLS1437325 44296 Node: MIPS Coprocessors1440715 44297 Node: PCH Target1442284 44298 Node: C++ ABI1443805 44299 Node: Misc1448424 44300 Ref: TARGET_SHIFT_TRUNCATION_MASK1455795 44301 Node: Host Config1498459 44302 Node: Host Common1499527 44303 Node: Filesystem1501906 44304 Node: Host Misc1506021 44305 Node: Fragments1508160 44306 Node: Target Fragment1509355 44307 Node: Host Fragment1515245 44308 Node: Collect21515485 44309 Node: Header Dirs1518028 44310 Node: Type Information1519451 44311 Node: GTY Options1521742 44312 Node: GGC Roots1532422 44313 Node: Files1533142 44314 Node: Invoking the garbage collector1535892 44315 Node: Plugins1536945 44316 Node: Funding1547310 44317 Node: GNU Project1549797 44318 Node: Copying1550446 44319 Node: GNU Free Documentation License1587977 44320 Node: Contributors1610386 44321 Node: Option Index1646716 44322 Node: Concept Index1647301 44323 44324 End Tag Table 44325
