1 This is doc/gccint.info, produced by makeinfo version 4.13 from 2 /tmp/android-build-bb7e003d31d08f72cabc269a652912b7/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 `pass_vrp'. 5639 5640 * Folding built-in functions 5641 5642 This pass simplifies built-in functions, as applicable, with 5643 constant arguments or with inferable string lengths. It is 5644 located in `tree-ssa-ccp.c' and is described by 5645 `pass_fold_builtins'. 5646 5647 * Split critical edges 5648 5649 This pass identifies critical edges and inserts empty basic blocks 5650 such that the edge is no longer critical. The pass is located in 5651 `tree-cfg.c' and is described by `pass_split_crit_edges'. 5652 5653 * Control dependence dead code elimination 5654 5655 This pass is a stronger form of dead code elimination that can 5656 eliminate unnecessary control flow statements. It is located in 5657 `tree-ssa-dce.c' and is described by `pass_cd_dce'. 5658 5659 * Tail call elimination 5660 5661 This pass identifies function calls that may be rewritten into 5662 jumps. No code transformation is actually applied here, but the 5663 data and control flow problem is solved. The code transformation 5664 requires target support, and so is delayed until RTL. In the 5665 meantime `CALL_EXPR_TAILCALL' is set indicating the possibility. 5666 The pass is located in `tree-tailcall.c' and is described by 5667 `pass_tail_calls'. The RTL transformation is handled by 5668 `fixup_tail_calls' in `calls.c'. 5669 5670 * Warn for function return without value 5671 5672 For non-void functions, this pass locates return statements that do 5673 not specify a value and issues a warning. Such a statement may 5674 have been injected by falling off the end of the function. This 5675 pass is run last so that we have as much time as possible to prove 5676 that the statement is not reachable. It is located in 5677 `tree-cfg.c' and is described by `pass_warn_function_return'. 5678 5679 * Mudflap statement annotation 5680 5681 If mudflap is enabled, we rewrite some memory accesses with code to 5682 validate that the memory access is correct. In particular, 5683 expressions involving pointer dereferences (`INDIRECT_REF', 5684 `ARRAY_REF', etc.) are replaced by code that checks the selected 5685 address range against the mudflap runtime's database of valid 5686 regions. This check includes an inline lookup into a 5687 direct-mapped cache, based on shift/mask operations of the pointer 5688 value, with a fallback function call into the runtime. The pass 5689 is located in `tree-mudflap.c' and is described by 5690 `pass_mudflap_2'. 5691 5692 * Leave static single assignment form 5693 5694 This pass rewrites the function such that it is in normal form. At 5695 the same time, we eliminate as many single-use temporaries as 5696 possible, so the intermediate language is no longer GIMPLE, but 5697 GENERIC. The pass is located in `tree-outof-ssa.c' and is 5698 described by `pass_del_ssa'. 5699 5700 * Merge PHI nodes that feed into one another 5701 5702 This is part of the CFG cleanup passes. It attempts to join PHI 5703 nodes from a forwarder CFG block into another block with PHI 5704 nodes. The pass is located in `tree-cfgcleanup.c' and is 5705 described by `pass_merge_phi'. 5706 5707 * Return value optimization 5708 5709 If a function always returns the same local variable, and that 5710 local variable is an aggregate type, then the variable is replaced 5711 with the return value for the function (i.e., the function's 5712 DECL_RESULT). This is equivalent to the C++ named return value 5713 optimization applied to GIMPLE. The pass is located in 5714 `tree-nrv.c' and is described by `pass_nrv'. 5715 5716 * Return slot optimization 5717 5718 If a function returns a memory object and is called as `var = 5719 foo()', this pass tries to change the call so that the address of 5720 `var' is sent to the caller to avoid an extra memory copy. This 5721 pass is located in `tree-nrv.c' and is described by 5722 `pass_return_slot'. 5723 5724 * Optimize calls to `__builtin_object_size' 5725 5726 This is a propagation pass similar to CCP that tries to remove 5727 calls to `__builtin_object_size' when the size of the object can be 5728 computed at compile-time. This pass is located in 5729 `tree-object-size.c' and is described by `pass_object_sizes'. 5730 5731 * Loop invariant motion 5732 5733 This pass removes expensive loop-invariant computations out of 5734 loops. The pass is located in `tree-ssa-loop.c' and described by 5735 `pass_lim'. 5736 5737 * Loop nest optimizations 5738 5739 This is a family of loop transformations that works on loop nests. 5740 It includes loop interchange, scaling, skewing and reversal and 5741 they are all geared to the optimization of data locality in array 5742 traversals and the removal of dependencies that hamper 5743 optimizations such as loop parallelization and vectorization. The 5744 pass is located in `tree-loop-linear.c' and described by 5745 `pass_linear_transform'. 5746 5747 * Removal of empty loops 5748 5749 This pass removes loops with no code in them. The pass is located 5750 in `tree-ssa-loop-ivcanon.c' and described by `pass_empty_loop'. 5751 5752 * Unrolling of small loops 5753 5754 This pass completely unrolls loops with few iterations. The pass 5755 is located in `tree-ssa-loop-ivcanon.c' and described by 5756 `pass_complete_unroll'. 5757 5758 * Predictive commoning 5759 5760 This pass makes the code reuse the computations from the previous 5761 iterations of the loops, especially loads and stores to memory. 5762 It does so by storing the values of these computations to a bank 5763 of temporary variables that are rotated at the end of loop. To 5764 avoid the need for this rotation, the loop is then unrolled and 5765 the copies of the loop body are rewritten to use the appropriate 5766 version of the temporary variable. This pass is located in 5767 `tree-predcom.c' and described by `pass_predcom'. 5768 5769 * Array prefetching 5770 5771 This pass issues prefetch instructions for array references inside 5772 loops. The pass is located in `tree-ssa-loop-prefetch.c' and 5773 described by `pass_loop_prefetch'. 5774 5775 * Reassociation 5776 5777 This pass rewrites arithmetic expressions to enable optimizations 5778 that operate on them, like redundancy elimination and 5779 vectorization. The pass is located in `tree-ssa-reassoc.c' and 5780 described by `pass_reassoc'. 5781 5782 * Optimization of `stdarg' functions 5783 5784 This pass tries to avoid the saving of register arguments into the 5785 stack on entry to `stdarg' functions. If the function doesn't use 5786 any `va_start' macros, no registers need to be saved. If 5787 `va_start' macros are used, the `va_list' variables don't escape 5788 the function, it is only necessary to save registers that will be 5789 used in `va_arg' macros. For instance, if `va_arg' is only used 5790 with integral types in the function, floating point registers 5791 don't need to be saved. This pass is located in `tree-stdarg.c' 5792 and described by `pass_stdarg'. 5793 5794 5795 5796 File: gccint.info, Node: RTL passes, Prev: Tree SSA passes, Up: Passes 5797 5798 8.5 RTL passes 5799 ============== 5800 5801 The following briefly describes the RTL generation and optimization 5802 passes that are run after the Tree optimization passes. 5803 5804 * RTL generation 5805 5806 The source files for RTL generation include `stmt.c', `calls.c', 5807 `expr.c', `explow.c', `expmed.c', `function.c', `optabs.c' and 5808 `emit-rtl.c'. Also, the file `insn-emit.c', generated from the 5809 machine description by the program `genemit', is used in this 5810 pass. The header file `expr.h' is used for communication within 5811 this pass. 5812 5813 The header files `insn-flags.h' and `insn-codes.h', generated from 5814 the machine description by the programs `genflags' and `gencodes', 5815 tell this pass which standard names are available for use and 5816 which patterns correspond to them. 5817 5818 * Generation of exception landing pads 5819 5820 This pass generates the glue that handles communication between the 5821 exception handling library routines and the exception handlers 5822 within the function. Entry points in the function that are 5823 invoked by the exception handling library are called "landing 5824 pads". The code for this pass is located in `except.c'. 5825 5826 * Control flow graph cleanup 5827 5828 This pass removes unreachable code, simplifies jumps to next, 5829 jumps to jump, jumps across jumps, etc. The pass is run multiple 5830 times. For historical reasons, it is occasionally referred to as 5831 the "jump optimization pass". The bulk of the code for this pass 5832 is in `cfgcleanup.c', and there are support routines in `cfgrtl.c' 5833 and `jump.c'. 5834 5835 * Forward propagation of single-def values 5836 5837 This pass attempts to remove redundant computation by substituting 5838 variables that come from a single definition, and seeing if the 5839 result can be simplified. It performs copy propagation and 5840 addressing mode selection. The pass is run twice, with values 5841 being propagated into loops only on the second run. The code is 5842 located in `fwprop.c'. 5843 5844 * Common subexpression elimination 5845 5846 This pass removes redundant computation within basic blocks, and 5847 optimizes addressing modes based on cost. The pass is run twice. 5848 The code for this pass is located in `cse.c'. 5849 5850 * Global common subexpression elimination 5851 5852 This pass performs two different types of GCSE depending on 5853 whether you are optimizing for size or not (LCM based GCSE tends 5854 to increase code size for a gain in speed, while Morel-Renvoise 5855 based GCSE does not). When optimizing for size, GCSE is done 5856 using Morel-Renvoise Partial Redundancy Elimination, with the 5857 exception that it does not try to move invariants out of 5858 loops--that is left to the loop optimization pass. If MR PRE 5859 GCSE is done, code hoisting (aka unification) is also done, as 5860 well as load motion. If you are optimizing for speed, LCM (lazy 5861 code motion) based GCSE is done. LCM is based on the work of 5862 Knoop, Ruthing, and Steffen. LCM based GCSE also does loop 5863 invariant code motion. We also perform load and store motion when 5864 optimizing for speed. Regardless of which type of GCSE is used, 5865 the GCSE pass also performs global constant and copy propagation. 5866 The source file for this pass is `gcse.c', and the LCM routines 5867 are in `lcm.c'. 5868 5869 * Loop optimization 5870 5871 This pass performs several loop related optimizations. The source 5872 files `cfgloopanal.c' and `cfgloopmanip.c' contain generic loop 5873 analysis and manipulation code. Initialization and finalization 5874 of loop structures is handled by `loop-init.c'. A loop invariant 5875 motion pass is implemented in `loop-invariant.c'. Basic block 5876 level optimizations--unrolling, peeling and unswitching loops-- 5877 are implemented in `loop-unswitch.c' and `loop-unroll.c'. 5878 Replacing of the exit condition of loops by special 5879 machine-dependent instructions is handled by `loop-doloop.c'. 5880 5881 * Jump bypassing 5882 5883 This pass is an aggressive form of GCSE that transforms the control 5884 flow graph of a function by propagating constants into conditional 5885 branch instructions. The source file for this pass is `gcse.c'. 5886 5887 * If conversion 5888 5889 This pass attempts to replace conditional branches and surrounding 5890 assignments with arithmetic, boolean value producing comparison 5891 instructions, and conditional move instructions. In the very last 5892 invocation after reload, it will generate predicated instructions 5893 when supported by the target. The code is located in `ifcvt.c'. 5894 5895 * Web construction 5896 5897 This pass splits independent uses of each pseudo-register. This 5898 can improve effect of the other transformation, such as CSE or 5899 register allocation. The code for this pass is located in `web.c'. 5900 5901 * Instruction combination 5902 5903 This pass attempts to combine groups of two or three instructions 5904 that are related by data flow into single instructions. It 5905 combines the RTL expressions for the instructions by substitution, 5906 simplifies the result using algebra, and then attempts to match 5907 the result against the machine description. The code is located 5908 in `combine.c'. 5909 5910 * Register movement 5911 5912 This pass looks for cases where matching constraints would force an 5913 instruction to need a reload, and this reload would be a 5914 register-to-register move. It then attempts to change the 5915 registers used by the instruction to avoid the move instruction. 5916 The code is located in `regmove.c'. 5917 5918 * Mode switching optimization 5919 5920 This pass looks for instructions that require the processor to be 5921 in a specific "mode" and minimizes the number of mode changes 5922 required to satisfy all users. What these modes are, and what 5923 they apply to are completely target-specific. The code for this 5924 pass is located in `mode-switching.c'. 5925 5926 * Modulo scheduling 5927 5928 This pass looks at innermost loops and reorders their instructions 5929 by overlapping different iterations. Modulo scheduling is 5930 performed immediately before instruction scheduling. The code for 5931 this pass is located in `modulo-sched.c'. 5932 5933 * Instruction scheduling 5934 5935 This pass looks for instructions whose output will not be 5936 available by the time that it is used in subsequent instructions. 5937 Memory loads and floating point instructions often have this 5938 behavior on RISC machines. It re-orders instructions within a 5939 basic block to try to separate the definition and use of items 5940 that otherwise would cause pipeline stalls. This pass is 5941 performed twice, before and after register allocation. The code 5942 for this pass is located in `haifa-sched.c', `sched-deps.c', 5943 `sched-ebb.c', `sched-rgn.c' and `sched-vis.c'. 5944 5945 * Register allocation 5946 5947 These passes make sure that all occurrences of pseudo registers are 5948 eliminated, either by allocating them to a hard register, replacing 5949 them by an equivalent expression (e.g. a constant) or by placing 5950 them on the stack. This is done in several subpasses: 5951 5952 * Register move optimizations. This pass makes some simple RTL 5953 code transformations which improve the subsequent register 5954 allocation. The source file is `regmove.c'. 5955 5956 * The integrated register allocator (IRA). It is called 5957 integrated because coalescing, register live range splitting, 5958 and hard register preferencing are done on-the-fly during 5959 coloring. It also has better integration with the reload 5960 pass. Pseudo-registers spilled by the allocator or the 5961 reload have still a chance to get hard-registers if the 5962 reload evicts some pseudo-registers from hard-registers. The 5963 allocator helps to choose better pseudos for spilling based 5964 on their live ranges and to coalesce stack slots allocated 5965 for the spilled pseudo-registers. IRA is a regional register 5966 allocator which is transformed into Chaitin-Briggs allocator 5967 if there is one region. By default, IRA chooses regions using 5968 register pressure but the user can force it to use one region 5969 or regions corresponding to all loops. 5970 5971 Source files of the allocator are `ira.c', `ira-build.c', 5972 `ira-costs.c', `ira-conflicts.c', `ira-color.c', 5973 `ira-emit.c', `ira-lives', plus header files `ira.h' and 5974 `ira-int.h' used for the communication between the allocator 5975 and the rest of the compiler and between the IRA files. 5976 5977 * Reloading. This pass renumbers pseudo registers with the 5978 hardware registers numbers they were allocated. Pseudo 5979 registers that did not get hard registers are replaced with 5980 stack slots. Then it finds instructions that are invalid 5981 because a value has failed to end up in a register, or has 5982 ended up in a register of the wrong kind. It fixes up these 5983 instructions by reloading the problematical values 5984 temporarily into registers. Additional instructions are 5985 generated to do the copying. 5986 5987 The reload pass also optionally eliminates the frame pointer 5988 and inserts instructions to save and restore call-clobbered 5989 registers around calls. 5990 5991 Source files are `reload.c' and `reload1.c', plus the header 5992 `reload.h' used for communication between them. 5993 5994 * Basic block reordering 5995 5996 This pass implements profile guided code positioning. If profile 5997 information is not available, various types of static analysis are 5998 performed to make the predictions normally coming from the profile 5999 feedback (IE execution frequency, branch probability, etc). It is 6000 implemented in the file `bb-reorder.c', and the various prediction 6001 routines are in `predict.c'. 6002 6003 * Variable tracking 6004 6005 This pass computes where the variables are stored at each position 6006 in code and generates notes describing the variable locations to 6007 RTL code. The location lists are then generated according to these 6008 notes to debug information if the debugging information format 6009 supports location lists. The code is located in `var-tracking.c'. 6010 6011 * Delayed branch scheduling 6012 6013 This optional pass attempts to find instructions that can go into 6014 the delay slots of other instructions, usually jumps and calls. 6015 The code for this pass is located in `reorg.c'. 6016 6017 * Branch shortening 6018 6019 On many RISC machines, branch instructions have a limited range. 6020 Thus, longer sequences of instructions must be used for long 6021 branches. In this pass, the compiler figures out what how far 6022 each instruction will be from each other instruction, and 6023 therefore whether the usual instructions, or the longer sequences, 6024 must be used for each branch. The code for this pass is located 6025 in `final.c'. 6026 6027 * Register-to-stack conversion 6028 6029 Conversion from usage of some hard registers to usage of a register 6030 stack may be done at this point. Currently, this is supported only 6031 for the floating-point registers of the Intel 80387 coprocessor. 6032 The code for this pass is located in `reg-stack.c'. 6033 6034 * Final 6035 6036 This pass outputs the assembler code for the function. The source 6037 files are `final.c' plus `insn-output.c'; the latter is generated 6038 automatically from the machine description by the tool `genoutput'. 6039 The header file `conditions.h' is used for communication between 6040 these files. If mudflap is enabled, the queue of deferred 6041 declarations and any addressed constants (e.g., string literals) 6042 is processed by `mudflap_finish_file' into a synthetic constructor 6043 function containing calls into the mudflap runtime. 6044 6045 * Debugging information output 6046 6047 This is run after final because it must output the stack slot 6048 offsets for pseudo registers that did not get hard registers. 6049 Source files are `dbxout.c' for DBX symbol table format, 6050 `sdbout.c' for SDB symbol table format, `dwarfout.c' for DWARF 6051 symbol table format, files `dwarf2out.c' and `dwarf2asm.c' for 6052 DWARF2 symbol table format, and `vmsdbgout.c' for VMS debug symbol 6053 table format. 6054 6055 6056 6057 File: gccint.info, Node: Trees, Next: GENERIC, Prev: Passes, Up: Top 6058 6059 9 Trees: The intermediate representation used by the C and C++ front ends 6060 ************************************************************************* 6061 6062 This chapter documents the internal representation used by GCC to 6063 represent C and C++ source programs. When presented with a C or C++ 6064 source program, GCC parses the program, performs semantic analysis 6065 (including the generation of error messages), and then produces the 6066 internal representation described here. This representation contains a 6067 complete representation for the entire translation unit provided as 6068 input to the front end. This representation is then typically processed 6069 by a code-generator in order to produce machine code, but could also be 6070 used in the creation of source browsers, intelligent editors, automatic 6071 documentation generators, interpreters, and any other programs needing 6072 the ability to process C or C++ code. 6073 6074 This chapter explains the internal representation. In particular, it 6075 documents the internal representation for C and C++ source constructs, 6076 and the macros, functions, and variables that can be used to access 6077 these constructs. The C++ representation is largely a superset of the 6078 representation used in the C front end. There is only one construct 6079 used in C that does not appear in the C++ front end and that is the GNU 6080 "nested function" extension. Many of the macros documented here do not 6081 apply in C because the corresponding language constructs do not appear 6082 in C. 6083 6084 If you are developing a "back end", be it is a code-generator or some 6085 other tool, that uses this representation, you may occasionally find 6086 that you need to ask questions not easily answered by the functions and 6087 macros available here. If that situation occurs, it is quite likely 6088 that GCC already supports the functionality you desire, but that the 6089 interface is simply not documented here. In that case, you should ask 6090 the GCC maintainers (via mail to <gcc (a] gcc.gnu.org>) about documenting 6091 the functionality you require. Similarly, if you find yourself writing 6092 functions that do not deal directly with your back end, but instead 6093 might be useful to other people using the GCC front end, you should 6094 submit your patches for inclusion in GCC. 6095 6096 * Menu: 6097 6098 * Deficiencies:: Topics net yet covered in this document. 6099 * Tree overview:: All about `tree's. 6100 * Types:: Fundamental and aggregate types. 6101 * Scopes:: Namespaces and classes. 6102 * Functions:: Overloading, function bodies, and linkage. 6103 * Declarations:: Type declarations and variables. 6104 * Attributes:: Declaration and type attributes. 6105 * Expression trees:: From `typeid' to `throw'. 6106 6107 6108 File: gccint.info, Node: Deficiencies, Next: Tree overview, Up: Trees 6109 6110 9.1 Deficiencies 6111 ================ 6112 6113 There are many places in which this document is incomplet and incorrekt. 6114 It is, as of yet, only _preliminary_ documentation. 6115 6116 6117 File: gccint.info, Node: Tree overview, Next: Types, Prev: Deficiencies, Up: Trees 6118 6119 9.2 Overview 6120 ============ 6121 6122 The central data structure used by the internal representation is the 6123 `tree'. These nodes, while all of the C type `tree', are of many 6124 varieties. A `tree' is a pointer type, but the object to which it 6125 points may be of a variety of types. From this point forward, we will 6126 refer to trees in ordinary type, rather than in `this font', except 6127 when talking about the actual C type `tree'. 6128 6129 You can tell what kind of node a particular tree is by using the 6130 `TREE_CODE' macro. Many, many macros take trees as input and return 6131 trees as output. However, most macros require a certain kind of tree 6132 node as input. In other words, there is a type-system for trees, but 6133 it is not reflected in the C type-system. 6134 6135 For safety, it is useful to configure GCC with `--enable-checking'. 6136 Although this results in a significant performance penalty (since all 6137 tree types are checked at run-time), and is therefore inappropriate in a 6138 release version, it is extremely helpful during the development process. 6139 6140 Many macros behave as predicates. Many, although not all, of these 6141 predicates end in `_P'. Do not rely on the result type of these macros 6142 being of any particular type. You may, however, rely on the fact that 6143 the type can be compared to `0', so that statements like 6144 if (TEST_P (t) && !TEST_P (y)) 6145 x = 1; 6146 and 6147 int i = (TEST_P (t) != 0); 6148 are legal. Macros that return `int' values now may be changed to 6149 return `tree' values, or other pointers in the future. Even those that 6150 continue to return `int' may return multiple nonzero codes where 6151 previously they returned only zero and one. Therefore, you should not 6152 write code like 6153 if (TEST_P (t) == 1) 6154 as this code is not guaranteed to work correctly in the future. 6155 6156 You should not take the address of values returned by the macros or 6157 functions described here. In particular, no guarantee is given that the 6158 values are lvalues. 6159 6160 In general, the names of macros are all in uppercase, while the names 6161 of functions are entirely in lowercase. There are rare exceptions to 6162 this rule. You should assume that any macro or function whose name is 6163 made up entirely of uppercase letters may evaluate its arguments more 6164 than once. You may assume that a macro or function whose name is made 6165 up entirely of lowercase letters will evaluate its arguments only once. 6166 6167 The `error_mark_node' is a special tree. Its tree code is 6168 `ERROR_MARK', but since there is only ever one node with that code, the 6169 usual practice is to compare the tree against `error_mark_node'. (This 6170 test is just a test for pointer equality.) If an error has occurred 6171 during front-end processing the flag `errorcount' will be set. If the 6172 front end has encountered code it cannot handle, it will issue a 6173 message to the user and set `sorrycount'. When these flags are set, 6174 any macro or function which normally returns a tree of a particular 6175 kind may instead return the `error_mark_node'. Thus, if you intend to 6176 do any processing of erroneous code, you must be prepared to deal with 6177 the `error_mark_node'. 6178 6179 Occasionally, a particular tree slot (like an operand to an expression, 6180 or a particular field in a declaration) will be referred to as 6181 "reserved for the back end". These slots are used to store RTL when 6182 the tree is converted to RTL for use by the GCC back end. However, if 6183 that process is not taking place (e.g., if the front end is being hooked 6184 up to an intelligent editor), then those slots may be used by the back 6185 end presently in use. 6186 6187 If you encounter situations that do not match this documentation, such 6188 as tree nodes of types not mentioned here, or macros documented to 6189 return entities of a particular kind that instead return entities of 6190 some different kind, you have found a bug, either in the front end or in 6191 the documentation. Please report these bugs as you would any other bug. 6192 6193 * Menu: 6194 6195 * Macros and Functions::Macros and functions that can be used with all trees. 6196 * Identifiers:: The names of things. 6197 * Containers:: Lists and vectors. 6198 6199 6200 File: gccint.info, Node: Macros and Functions, Next: Identifiers, Up: Tree overview 6201 6202 9.2.1 Trees 6203 ----------- 6204 6205 This section is not here yet. 6206 6207 6208 File: gccint.info, Node: Identifiers, Next: Containers, Prev: Macros and Functions, Up: Tree overview 6209 6210 9.2.2 Identifiers 6211 ----------------- 6212 6213 An `IDENTIFIER_NODE' represents a slightly more general concept that 6214 the standard C or C++ concept of identifier. In particular, an 6215 `IDENTIFIER_NODE' may contain a `$', or other extraordinary characters. 6216 6217 There are never two distinct `IDENTIFIER_NODE's representing the same 6218 identifier. Therefore, you may use pointer equality to compare 6219 `IDENTIFIER_NODE's, rather than using a routine like `strcmp'. 6220 6221 You can use the following macros to access identifiers: 6222 `IDENTIFIER_POINTER' 6223 The string represented by the identifier, represented as a 6224 `char*'. This string is always `NUL'-terminated, and contains no 6225 embedded `NUL' characters. 6226 6227 `IDENTIFIER_LENGTH' 6228 The length of the string returned by `IDENTIFIER_POINTER', not 6229 including the trailing `NUL'. This value of `IDENTIFIER_LENGTH 6230 (x)' is always the same as `strlen (IDENTIFIER_POINTER (x))'. 6231 6232 `IDENTIFIER_OPNAME_P' 6233 This predicate holds if the identifier represents the name of an 6234 overloaded operator. In this case, you should not depend on the 6235 contents of either the `IDENTIFIER_POINTER' or the 6236 `IDENTIFIER_LENGTH'. 6237 6238 `IDENTIFIER_TYPENAME_P' 6239 This predicate holds if the identifier represents the name of a 6240 user-defined conversion operator. In this case, the `TREE_TYPE' of 6241 the `IDENTIFIER_NODE' holds the type to which the conversion 6242 operator converts. 6243 6244 6245 6246 File: gccint.info, Node: Containers, Prev: Identifiers, Up: Tree overview 6247 6248 9.2.3 Containers 6249 ---------------- 6250 6251 Two common container data structures can be represented directly with 6252 tree nodes. A `TREE_LIST' is a singly linked list containing two trees 6253 per node. These are the `TREE_PURPOSE' and `TREE_VALUE' of each node. 6254 (Often, the `TREE_PURPOSE' contains some kind of tag, or additional 6255 information, while the `TREE_VALUE' contains the majority of the 6256 payload. In other cases, the `TREE_PURPOSE' is simply `NULL_TREE', 6257 while in still others both the `TREE_PURPOSE' and `TREE_VALUE' are of 6258 equal stature.) Given one `TREE_LIST' node, the next node is found by 6259 following the `TREE_CHAIN'. If the `TREE_CHAIN' is `NULL_TREE', then 6260 you have reached the end of the list. 6261 6262 A `TREE_VEC' is a simple vector. The `TREE_VEC_LENGTH' is an integer 6263 (not a tree) giving the number of nodes in the vector. The nodes 6264 themselves are accessed using the `TREE_VEC_ELT' macro, which takes two 6265 arguments. The first is the `TREE_VEC' in question; the second is an 6266 integer indicating which element in the vector is desired. The 6267 elements are indexed from zero. 6268 6269 6270 File: gccint.info, Node: Types, Next: Scopes, Prev: Tree overview, Up: Trees 6271 6272 9.3 Types 6273 ========= 6274 6275 All types have corresponding tree nodes. However, you should not assume 6276 that there is exactly one tree node corresponding to each type. There 6277 are often multiple nodes corresponding to the same type. 6278 6279 For the most part, different kinds of types have different tree codes. 6280 (For example, pointer types use a `POINTER_TYPE' code while arrays use 6281 an `ARRAY_TYPE' code.) However, pointers to member functions use the 6282 `RECORD_TYPE' code. Therefore, when writing a `switch' statement that 6283 depends on the code associated with a particular type, you should take 6284 care to handle pointers to member functions under the `RECORD_TYPE' 6285 case label. 6286 6287 In C++, an array type is not qualified; rather the type of the array 6288 elements is qualified. This situation is reflected in the intermediate 6289 representation. The macros described here will always examine the 6290 qualification of the underlying element type when applied to an array 6291 type. (If the element type is itself an array, then the recursion 6292 continues until a non-array type is found, and the qualification of this 6293 type is examined.) So, for example, `CP_TYPE_CONST_P' will hold of the 6294 type `const int ()[7]', denoting an array of seven `int's. 6295 6296 The following functions and macros deal with cv-qualification of types: 6297 `CP_TYPE_QUALS' 6298 This macro returns the set of type qualifiers applied to this type. 6299 This value is `TYPE_UNQUALIFIED' if no qualifiers have been 6300 applied. The `TYPE_QUAL_CONST' bit is set if the type is 6301 `const'-qualified. The `TYPE_QUAL_VOLATILE' bit is set if the 6302 type is `volatile'-qualified. The `TYPE_QUAL_RESTRICT' bit is set 6303 if the type is `restrict'-qualified. 6304 6305 `CP_TYPE_CONST_P' 6306 This macro holds if the type is `const'-qualified. 6307 6308 `CP_TYPE_VOLATILE_P' 6309 This macro holds if the type is `volatile'-qualified. 6310 6311 `CP_TYPE_RESTRICT_P' 6312 This macro holds if the type is `restrict'-qualified. 6313 6314 `CP_TYPE_CONST_NON_VOLATILE_P' 6315 This predicate holds for a type that is `const'-qualified, but 6316 _not_ `volatile'-qualified; other cv-qualifiers are ignored as 6317 well: only the `const'-ness is tested. 6318 6319 `TYPE_MAIN_VARIANT' 6320 This macro returns the unqualified version of a type. It may be 6321 applied to an unqualified type, but it is not always the identity 6322 function in that case. 6323 6324 A few other macros and functions are usable with all types: 6325 `TYPE_SIZE' 6326 The number of bits required to represent the type, represented as 6327 an `INTEGER_CST'. For an incomplete type, `TYPE_SIZE' will be 6328 `NULL_TREE'. 6329 6330 `TYPE_ALIGN' 6331 The alignment of the type, in bits, represented as an `int'. 6332 6333 `TYPE_NAME' 6334 This macro returns a declaration (in the form of a `TYPE_DECL') for 6335 the type. (Note this macro does _not_ return a `IDENTIFIER_NODE', 6336 as you might expect, given its name!) You can look at the 6337 `DECL_NAME' of the `TYPE_DECL' to obtain the actual name of the 6338 type. The `TYPE_NAME' will be `NULL_TREE' for a type that is not 6339 a built-in type, the result of a typedef, or a named class type. 6340 6341 `CP_INTEGRAL_TYPE' 6342 This predicate holds if the type is an integral type. Notice that 6343 in C++, enumerations are _not_ integral types. 6344 6345 `ARITHMETIC_TYPE_P' 6346 This predicate holds if the type is an integral type (in the C++ 6347 sense) or a floating point type. 6348 6349 `CLASS_TYPE_P' 6350 This predicate holds for a class-type. 6351 6352 `TYPE_BUILT_IN' 6353 This predicate holds for a built-in type. 6354 6355 `TYPE_PTRMEM_P' 6356 This predicate holds if the type is a pointer to data member. 6357 6358 `TYPE_PTR_P' 6359 This predicate holds if the type is a pointer type, and the 6360 pointee is not a data member. 6361 6362 `TYPE_PTRFN_P' 6363 This predicate holds for a pointer to function type. 6364 6365 `TYPE_PTROB_P' 6366 This predicate holds for a pointer to object type. Note however 6367 that it does not hold for the generic pointer to object type `void 6368 *'. You may use `TYPE_PTROBV_P' to test for a pointer to object 6369 type as well as `void *'. 6370 6371 `TYPE_CANONICAL' 6372 This macro returns the "canonical" type for the given type node. 6373 Canonical types are used to improve performance in the C++ and 6374 Objective-C++ front ends by allowing efficient comparison between 6375 two type nodes in `same_type_p': if the `TYPE_CANONICAL' values of 6376 the types are equal, the types are equivalent; otherwise, the types 6377 are not equivalent. The notion of equivalence for canonical types 6378 is the same as the notion of type equivalence in the language 6379 itself. For instance, 6380 6381 When `TYPE_CANONICAL' is `NULL_TREE', there is no canonical type 6382 for the given type node. In this case, comparison between this 6383 type and any other type requires the compiler to perform a deep, 6384 "structural" comparison to see if the two type nodes have the same 6385 form and properties. 6386 6387 The canonical type for a node is always the most fundamental type 6388 in the equivalence class of types. For instance, `int' is its own 6389 canonical type. A typedef `I' of `int' will have `int' as its 6390 canonical type. Similarly, `I*' and a typedef `IP' (defined to 6391 `I*') will has `int*' as their canonical type. When building a new 6392 type node, be sure to set `TYPE_CANONICAL' to the appropriate 6393 canonical type. If the new type is a compound type (built from 6394 other types), and any of those other types require structural 6395 equality, use `SET_TYPE_STRUCTURAL_EQUALITY' to ensure that the 6396 new type also requires structural equality. Finally, if for some 6397 reason you cannot guarantee that `TYPE_CANONICAL' will point to 6398 the canonical type, use `SET_TYPE_STRUCTURAL_EQUALITY' to make 6399 sure that the new type-and any type constructed based on 6400 it-requires structural equality. If you suspect that the canonical 6401 type system is miscomparing types, pass `--param 6402 verify-canonical-types=1' to the compiler or configure with 6403 `--enable-checking' to force the compiler to verify its 6404 canonical-type comparisons against the structural comparisons; the 6405 compiler will then print any warnings if the canonical types 6406 miscompare. 6407 6408 `TYPE_STRUCTURAL_EQUALITY_P' 6409 This predicate holds when the node requires structural equality 6410 checks, e.g., when `TYPE_CANONICAL' is `NULL_TREE'. 6411 6412 `SET_TYPE_STRUCTURAL_EQUALITY' 6413 This macro states that the type node it is given requires 6414 structural equality checks, e.g., it sets `TYPE_CANONICAL' to 6415 `NULL_TREE'. 6416 6417 `same_type_p' 6418 This predicate takes two types as input, and holds if they are the 6419 same type. For example, if one type is a `typedef' for the other, 6420 or both are `typedef's for the same type. This predicate also 6421 holds if the two trees given as input are simply copies of one 6422 another; i.e., there is no difference between them at the source 6423 level, but, for whatever reason, a duplicate has been made in the 6424 representation. You should never use `==' (pointer equality) to 6425 compare types; always use `same_type_p' instead. 6426 6427 Detailed below are the various kinds of types, and the macros that can 6428 be used to access them. Although other kinds of types are used 6429 elsewhere in G++, the types described here are the only ones that you 6430 will encounter while examining the intermediate representation. 6431 6432 `VOID_TYPE' 6433 Used to represent the `void' type. 6434 6435 `INTEGER_TYPE' 6436 Used to represent the various integral types, including `char', 6437 `short', `int', `long', and `long long'. This code is not used 6438 for enumeration types, nor for the `bool' type. The 6439 `TYPE_PRECISION' is the number of bits used in the representation, 6440 represented as an `unsigned int'. (Note that in the general case 6441 this is not the same value as `TYPE_SIZE'; suppose that there were 6442 a 24-bit integer type, but that alignment requirements for the ABI 6443 required 32-bit alignment. Then, `TYPE_SIZE' would be an 6444 `INTEGER_CST' for 32, while `TYPE_PRECISION' would be 24.) The 6445 integer type is unsigned if `TYPE_UNSIGNED' holds; otherwise, it 6446 is signed. 6447 6448 The `TYPE_MIN_VALUE' is an `INTEGER_CST' for the smallest integer 6449 that may be represented by this type. Similarly, the 6450 `TYPE_MAX_VALUE' is an `INTEGER_CST' for the largest integer that 6451 may be represented by this type. 6452 6453 `REAL_TYPE' 6454 Used to represent the `float', `double', and `long double' types. 6455 The number of bits in the floating-point representation is given 6456 by `TYPE_PRECISION', as in the `INTEGER_TYPE' case. 6457 6458 `FIXED_POINT_TYPE' 6459 Used to represent the `short _Fract', `_Fract', `long _Fract', 6460 `long long _Fract', `short _Accum', `_Accum', `long _Accum', and 6461 `long long _Accum' types. The number of bits in the fixed-point 6462 representation is given by `TYPE_PRECISION', as in the 6463 `INTEGER_TYPE' case. There may be padding bits, fractional bits 6464 and integral bits. The number of fractional bits is given by 6465 `TYPE_FBIT', and the number of integral bits is given by 6466 `TYPE_IBIT'. The fixed-point type is unsigned if `TYPE_UNSIGNED' 6467 holds; otherwise, it is signed. The fixed-point type is 6468 saturating if `TYPE_SATURATING' holds; otherwise, it is not 6469 saturating. 6470 6471 `COMPLEX_TYPE' 6472 Used to represent GCC built-in `__complex__' data types. The 6473 `TREE_TYPE' is the type of the real and imaginary parts. 6474 6475 `ENUMERAL_TYPE' 6476 Used to represent an enumeration type. The `TYPE_PRECISION' gives 6477 (as an `int'), the number of bits used to represent the type. If 6478 there are no negative enumeration constants, `TYPE_UNSIGNED' will 6479 hold. The minimum and maximum enumeration constants may be 6480 obtained with `TYPE_MIN_VALUE' and `TYPE_MAX_VALUE', respectively; 6481 each of these macros returns an `INTEGER_CST'. 6482 6483 The actual enumeration constants themselves may be obtained by 6484 looking at the `TYPE_VALUES'. This macro will return a 6485 `TREE_LIST', containing the constants. The `TREE_PURPOSE' of each 6486 node will be an `IDENTIFIER_NODE' giving the name of the constant; 6487 the `TREE_VALUE' will be an `INTEGER_CST' giving the value 6488 assigned to that constant. These constants will appear in the 6489 order in which they were declared. The `TREE_TYPE' of each of 6490 these constants will be the type of enumeration type itself. 6491 6492 `BOOLEAN_TYPE' 6493 Used to represent the `bool' type. 6494 6495 `POINTER_TYPE' 6496 Used to represent pointer types, and pointer to data member types. 6497 The `TREE_TYPE' gives the type to which this type points. If the 6498 type is a pointer to data member type, then `TYPE_PTRMEM_P' will 6499 hold. For a pointer to data member type of the form `T X::*', 6500 `TYPE_PTRMEM_CLASS_TYPE' will be the type `X', while 6501 `TYPE_PTRMEM_POINTED_TO_TYPE' will be the type `T'. 6502 6503 `REFERENCE_TYPE' 6504 Used to represent reference types. The `TREE_TYPE' gives the type 6505 to which this type refers. 6506 6507 `FUNCTION_TYPE' 6508 Used to represent the type of non-member functions and of static 6509 member functions. The `TREE_TYPE' gives the return type of the 6510 function. The `TYPE_ARG_TYPES' are a `TREE_LIST' of the argument 6511 types. The `TREE_VALUE' of each node in this list is the type of 6512 the corresponding argument; the `TREE_PURPOSE' is an expression 6513 for the default argument value, if any. If the last node in the 6514 list is `void_list_node' (a `TREE_LIST' node whose `TREE_VALUE' is 6515 the `void_type_node'), then functions of this type do not take 6516 variable arguments. Otherwise, they do take a variable number of 6517 arguments. 6518 6519 Note that in C (but not in C++) a function declared like `void f()' 6520 is an unprototyped function taking a variable number of arguments; 6521 the `TYPE_ARG_TYPES' of such a function will be `NULL'. 6522 6523 `METHOD_TYPE' 6524 Used to represent the type of a non-static member function. Like a 6525 `FUNCTION_TYPE', the return type is given by the `TREE_TYPE'. The 6526 type of `*this', i.e., the class of which functions of this type 6527 are a member, is given by the `TYPE_METHOD_BASETYPE'. The 6528 `TYPE_ARG_TYPES' is the parameter list, as for a `FUNCTION_TYPE', 6529 and includes the `this' argument. 6530 6531 `ARRAY_TYPE' 6532 Used to represent array types. The `TREE_TYPE' gives the type of 6533 the elements in the array. If the array-bound is present in the 6534 type, the `TYPE_DOMAIN' is an `INTEGER_TYPE' whose 6535 `TYPE_MIN_VALUE' and `TYPE_MAX_VALUE' will be the lower and upper 6536 bounds of the array, respectively. The `TYPE_MIN_VALUE' will 6537 always be an `INTEGER_CST' for zero, while the `TYPE_MAX_VALUE' 6538 will be one less than the number of elements in the array, i.e., 6539 the highest value which may be used to index an element in the 6540 array. 6541 6542 `RECORD_TYPE' 6543 Used to represent `struct' and `class' types, as well as pointers 6544 to member functions and similar constructs in other languages. 6545 `TYPE_FIELDS' contains the items contained in this type, each of 6546 which can be a `FIELD_DECL', `VAR_DECL', `CONST_DECL', or 6547 `TYPE_DECL'. You may not make any assumptions about the ordering 6548 of the fields in the type or whether one or more of them overlap. 6549 If `TYPE_PTRMEMFUNC_P' holds, then this type is a pointer-to-member 6550 type. In that case, the `TYPE_PTRMEMFUNC_FN_TYPE' is a 6551 `POINTER_TYPE' pointing to a `METHOD_TYPE'. The `METHOD_TYPE' is 6552 the type of a function pointed to by the pointer-to-member 6553 function. If `TYPE_PTRMEMFUNC_P' does not hold, this type is a 6554 class type. For more information, see *note Classes::. 6555 6556 `UNION_TYPE' 6557 Used to represent `union' types. Similar to `RECORD_TYPE' except 6558 that all `FIELD_DECL' nodes in `TYPE_FIELD' start at bit position 6559 zero. 6560 6561 `QUAL_UNION_TYPE' 6562 Used to represent part of a variant record in Ada. Similar to 6563 `UNION_TYPE' except that each `FIELD_DECL' has a `DECL_QUALIFIER' 6564 field, which contains a boolean expression that indicates whether 6565 the field is present in the object. The type will only have one 6566 field, so each field's `DECL_QUALIFIER' is only evaluated if none 6567 of the expressions in the previous fields in `TYPE_FIELDS' are 6568 nonzero. Normally these expressions will reference a field in the 6569 outer object using a `PLACEHOLDER_EXPR'. 6570 6571 `UNKNOWN_TYPE' 6572 This node is used to represent a type the knowledge of which is 6573 insufficient for a sound processing. 6574 6575 `OFFSET_TYPE' 6576 This node is used to represent a pointer-to-data member. For a 6577 data member `X::m' the `TYPE_OFFSET_BASETYPE' is `X' and the 6578 `TREE_TYPE' is the type of `m'. 6579 6580 `TYPENAME_TYPE' 6581 Used to represent a construct of the form `typename T::A'. The 6582 `TYPE_CONTEXT' is `T'; the `TYPE_NAME' is an `IDENTIFIER_NODE' for 6583 `A'. If the type is specified via a template-id, then 6584 `TYPENAME_TYPE_FULLNAME' yields a `TEMPLATE_ID_EXPR'. The 6585 `TREE_TYPE' is non-`NULL' if the node is implicitly generated in 6586 support for the implicit typename extension; in which case the 6587 `TREE_TYPE' is a type node for the base-class. 6588 6589 `TYPEOF_TYPE' 6590 Used to represent the `__typeof__' extension. The `TYPE_FIELDS' 6591 is the expression the type of which is being represented. 6592 6593 There are variables whose values represent some of the basic types. 6594 These include: 6595 `void_type_node' 6596 A node for `void'. 6597 6598 `integer_type_node' 6599 A node for `int'. 6600 6601 `unsigned_type_node.' 6602 A node for `unsigned int'. 6603 6604 `char_type_node.' 6605 A node for `char'. 6606 It may sometimes be useful to compare one of these variables with a 6607 type in hand, using `same_type_p'. 6608 6609 6610 File: gccint.info, Node: Scopes, Next: Functions, Prev: Types, Up: Trees 6611 6612 9.4 Scopes 6613 ========== 6614 6615 The root of the entire intermediate representation is the variable 6616 `global_namespace'. This is the namespace specified with `::' in C++ 6617 source code. All other namespaces, types, variables, functions, and so 6618 forth can be found starting with this namespace. 6619 6620 Besides namespaces, the other high-level scoping construct in C++ is 6621 the class. (Throughout this manual the term "class" is used to mean the 6622 types referred to in the ANSI/ISO C++ Standard as classes; these include 6623 types defined with the `class', `struct', and `union' keywords.) 6624 6625 * Menu: 6626 6627 * Namespaces:: Member functions, types, etc. 6628 * Classes:: Members, bases, friends, etc. 6629 6630 6631 File: gccint.info, Node: Namespaces, Next: Classes, Up: Scopes 6632 6633 9.4.1 Namespaces 6634 ---------------- 6635 6636 A namespace is represented by a `NAMESPACE_DECL' node. 6637 6638 However, except for the fact that it is distinguished as the root of 6639 the representation, the global namespace is no different from any other 6640 namespace. Thus, in what follows, we describe namespaces generally, 6641 rather than the global namespace in particular. 6642 6643 The following macros and functions can be used on a `NAMESPACE_DECL': 6644 6645 `DECL_NAME' 6646 This macro is used to obtain the `IDENTIFIER_NODE' corresponding to 6647 the unqualified name of the name of the namespace (*note 6648 Identifiers::). The name of the global namespace is `::', even 6649 though in C++ the global namespace is unnamed. However, you 6650 should use comparison with `global_namespace', rather than 6651 `DECL_NAME' to determine whether or not a namespace is the global 6652 one. An unnamed namespace will have a `DECL_NAME' equal to 6653 `anonymous_namespace_name'. Within a single translation unit, all 6654 unnamed namespaces will have the same name. 6655 6656 `DECL_CONTEXT' 6657 This macro returns the enclosing namespace. The `DECL_CONTEXT' for 6658 the `global_namespace' is `NULL_TREE'. 6659 6660 `DECL_NAMESPACE_ALIAS' 6661 If this declaration is for a namespace alias, then 6662 `DECL_NAMESPACE_ALIAS' is the namespace for which this one is an 6663 alias. 6664 6665 Do not attempt to use `cp_namespace_decls' for a namespace which is 6666 an alias. Instead, follow `DECL_NAMESPACE_ALIAS' links until you 6667 reach an ordinary, non-alias, namespace, and call 6668 `cp_namespace_decls' there. 6669 6670 `DECL_NAMESPACE_STD_P' 6671 This predicate holds if the namespace is the special `::std' 6672 namespace. 6673 6674 `cp_namespace_decls' 6675 This function will return the declarations contained in the 6676 namespace, including types, overloaded functions, other 6677 namespaces, and so forth. If there are no declarations, this 6678 function will return `NULL_TREE'. The declarations are connected 6679 through their `TREE_CHAIN' fields. 6680 6681 Although most entries on this list will be declarations, 6682 `TREE_LIST' nodes may also appear. In this case, the `TREE_VALUE' 6683 will be an `OVERLOAD'. The value of the `TREE_PURPOSE' is 6684 unspecified; back ends should ignore this value. As with the 6685 other kinds of declarations returned by `cp_namespace_decls', the 6686 `TREE_CHAIN' will point to the next declaration in this list. 6687 6688 For more information on the kinds of declarations that can occur 6689 on this list, *Note Declarations::. Some declarations will not 6690 appear on this list. In particular, no `FIELD_DECL', 6691 `LABEL_DECL', or `PARM_DECL' nodes will appear here. 6692 6693 This function cannot be used with namespaces that have 6694 `DECL_NAMESPACE_ALIAS' set. 6695 6696 6697 6698 File: gccint.info, Node: Classes, Prev: Namespaces, Up: Scopes 6699 6700 9.4.2 Classes 6701 ------------- 6702 6703 A class type is represented by either a `RECORD_TYPE' or a 6704 `UNION_TYPE'. A class declared with the `union' tag is represented by 6705 a `UNION_TYPE', while classes declared with either the `struct' or the 6706 `class' tag are represented by `RECORD_TYPE's. You can use the 6707 `CLASSTYPE_DECLARED_CLASS' macro to discern whether or not a particular 6708 type is a `class' as opposed to a `struct'. This macro will be true 6709 only for classes declared with the `class' tag. 6710 6711 Almost all non-function members are available on the `TYPE_FIELDS' 6712 list. Given one member, the next can be found by following the 6713 `TREE_CHAIN'. You should not depend in any way on the order in which 6714 fields appear on this list. All nodes on this list will be `DECL' 6715 nodes. A `FIELD_DECL' is used to represent a non-static data member, a 6716 `VAR_DECL' is used to represent a static data member, and a `TYPE_DECL' 6717 is used to represent a type. Note that the `CONST_DECL' for an 6718 enumeration constant will appear on this list, if the enumeration type 6719 was declared in the class. (Of course, the `TYPE_DECL' for the 6720 enumeration type will appear here as well.) There are no entries for 6721 base classes on this list. In particular, there is no `FIELD_DECL' for 6722 the "base-class portion" of an object. 6723 6724 The `TYPE_VFIELD' is a compiler-generated field used to point to 6725 virtual function tables. It may or may not appear on the `TYPE_FIELDS' 6726 list. However, back ends should handle the `TYPE_VFIELD' just like all 6727 the entries on the `TYPE_FIELDS' list. 6728 6729 The function members are available on the `TYPE_METHODS' list. Again, 6730 subsequent members are found by following the `TREE_CHAIN' field. If a 6731 function is overloaded, each of the overloaded functions appears; no 6732 `OVERLOAD' nodes appear on the `TYPE_METHODS' list. Implicitly 6733 declared functions (including default constructors, copy constructors, 6734 assignment operators, and destructors) will appear on this list as well. 6735 6736 Every class has an associated "binfo", which can be obtained with 6737 `TYPE_BINFO'. Binfos are used to represent base-classes. The binfo 6738 given by `TYPE_BINFO' is the degenerate case, whereby every class is 6739 considered to be its own base-class. The base binfos for a particular 6740 binfo are held in a vector, whose length is obtained with 6741 `BINFO_N_BASE_BINFOS'. The base binfos themselves are obtained with 6742 `BINFO_BASE_BINFO' and `BINFO_BASE_ITERATE'. To add a new binfo, use 6743 `BINFO_BASE_APPEND'. The vector of base binfos can be obtained with 6744 `BINFO_BASE_BINFOS', but normally you do not need to use that. The 6745 class type associated with a binfo is given by `BINFO_TYPE'. It is not 6746 always the case that `BINFO_TYPE (TYPE_BINFO (x))', because of typedefs 6747 and qualified types. Neither is it the case that `TYPE_BINFO 6748 (BINFO_TYPE (y))' is the same binfo as `y'. The reason is that if `y' 6749 is a binfo representing a base-class `B' of a derived class `D', then 6750 `BINFO_TYPE (y)' will be `B', and `TYPE_BINFO (BINFO_TYPE (y))' will be 6751 `B' as its own base-class, rather than as a base-class of `D'. 6752 6753 The access to a base type can be found with `BINFO_BASE_ACCESS'. This 6754 will produce `access_public_node', `access_private_node' or 6755 `access_protected_node'. If bases are always public, 6756 `BINFO_BASE_ACCESSES' may be `NULL'. 6757 6758 `BINFO_VIRTUAL_P' is used to specify whether the binfo is inherited 6759 virtually or not. The other flags, `BINFO_MARKED_P' and `BINFO_FLAG_1' 6760 to `BINFO_FLAG_6' can be used for language specific use. 6761 6762 The following macros can be used on a tree node representing a 6763 class-type. 6764 6765 `LOCAL_CLASS_P' 6766 This predicate holds if the class is local class _i.e._ declared 6767 inside a function body. 6768 6769 `TYPE_POLYMORPHIC_P' 6770 This predicate holds if the class has at least one virtual function 6771 (declared or inherited). 6772 6773 `TYPE_HAS_DEFAULT_CONSTRUCTOR' 6774 This predicate holds whenever its argument represents a class-type 6775 with default constructor. 6776 6777 `CLASSTYPE_HAS_MUTABLE' 6778 `TYPE_HAS_MUTABLE_P' 6779 These predicates hold for a class-type having a mutable data 6780 member. 6781 6782 `CLASSTYPE_NON_POD_P' 6783 This predicate holds only for class-types that are not PODs. 6784 6785 `TYPE_HAS_NEW_OPERATOR' 6786 This predicate holds for a class-type that defines `operator new'. 6787 6788 `TYPE_HAS_ARRAY_NEW_OPERATOR' 6789 This predicate holds for a class-type for which `operator new[]' 6790 is defined. 6791 6792 `TYPE_OVERLOADS_CALL_EXPR' 6793 This predicate holds for class-type for which the function call 6794 `operator()' is overloaded. 6795 6796 `TYPE_OVERLOADS_ARRAY_REF' 6797 This predicate holds for a class-type that overloads `operator[]' 6798 6799 `TYPE_OVERLOADS_ARROW' 6800 This predicate holds for a class-type for which `operator->' is 6801 overloaded. 6802 6803 6804 6805 File: gccint.info, Node: Declarations, Next: Attributes, Prev: Functions, Up: Trees 6806 6807 9.5 Declarations 6808 ================ 6809 6810 This section covers the various kinds of declarations that appear in the 6811 internal representation, except for declarations of functions 6812 (represented by `FUNCTION_DECL' nodes), which are described in *note 6813 Functions::. 6814 6815 * Menu: 6816 6817 * Working with declarations:: Macros and functions that work on 6818 declarations. 6819 * Internal structure:: How declaration nodes are represented. 6820 6821 6822 File: gccint.info, Node: Working with declarations, Next: Internal structure, Up: Declarations 6823 6824 9.5.1 Working with declarations 6825 ------------------------------- 6826 6827 Some macros can be used with any kind of declaration. These include: 6828 `DECL_NAME' 6829 This macro returns an `IDENTIFIER_NODE' giving the name of the 6830 entity. 6831 6832 `TREE_TYPE' 6833 This macro returns the type of the entity declared. 6834 6835 `TREE_FILENAME' 6836 This macro returns the name of the file in which the entity was 6837 declared, as a `char*'. For an entity declared implicitly by the 6838 compiler (like `__builtin_memcpy'), this will be the string 6839 `"<internal>"'. 6840 6841 `TREE_LINENO' 6842 This macro returns the line number at which the entity was 6843 declared, as an `int'. 6844 6845 `DECL_ARTIFICIAL' 6846 This predicate holds if the declaration was implicitly generated 6847 by the compiler. For example, this predicate will hold of an 6848 implicitly declared member function, or of the `TYPE_DECL' 6849 implicitly generated for a class type. Recall that in C++ code 6850 like: 6851 struct S {}; 6852 is roughly equivalent to C code like: 6853 struct S {}; 6854 typedef struct S S; 6855 The implicitly generated `typedef' declaration is represented by a 6856 `TYPE_DECL' for which `DECL_ARTIFICIAL' holds. 6857 6858 `DECL_NAMESPACE_SCOPE_P' 6859 This predicate holds if the entity was declared at a namespace 6860 scope. 6861 6862 `DECL_CLASS_SCOPE_P' 6863 This predicate holds if the entity was declared at a class scope. 6864 6865 `DECL_FUNCTION_SCOPE_P' 6866 This predicate holds if the entity was declared inside a function 6867 body. 6868 6869 6870 The various kinds of declarations include: 6871 `LABEL_DECL' 6872 These nodes are used to represent labels in function bodies. For 6873 more information, see *note Functions::. These nodes only appear 6874 in block scopes. 6875 6876 `CONST_DECL' 6877 These nodes are used to represent enumeration constants. The 6878 value of the constant is given by `DECL_INITIAL' which will be an 6879 `INTEGER_CST' with the same type as the `TREE_TYPE' of the 6880 `CONST_DECL', i.e., an `ENUMERAL_TYPE'. 6881 6882 `RESULT_DECL' 6883 These nodes represent the value returned by a function. When a 6884 value is assigned to a `RESULT_DECL', that indicates that the 6885 value should be returned, via bitwise copy, by the function. You 6886 can use `DECL_SIZE' and `DECL_ALIGN' on a `RESULT_DECL', just as 6887 with a `VAR_DECL'. 6888 6889 `TYPE_DECL' 6890 These nodes represent `typedef' declarations. The `TREE_TYPE' is 6891 the type declared to have the name given by `DECL_NAME'. In some 6892 cases, there is no associated name. 6893 6894 `VAR_DECL' 6895 These nodes represent variables with namespace or block scope, as 6896 well as static data members. The `DECL_SIZE' and `DECL_ALIGN' are 6897 analogous to `TYPE_SIZE' and `TYPE_ALIGN'. For a declaration, you 6898 should always use the `DECL_SIZE' and `DECL_ALIGN' rather than the 6899 `TYPE_SIZE' and `TYPE_ALIGN' given by the `TREE_TYPE', since 6900 special attributes may have been applied to the variable to give 6901 it a particular size and alignment. You may use the predicates 6902 `DECL_THIS_STATIC' or `DECL_THIS_EXTERN' to test whether the 6903 storage class specifiers `static' or `extern' were used to declare 6904 a variable. 6905 6906 If this variable is initialized (but does not require a 6907 constructor), the `DECL_INITIAL' will be an expression for the 6908 initializer. The initializer should be evaluated, and a bitwise 6909 copy into the variable performed. If the `DECL_INITIAL' is the 6910 `error_mark_node', there is an initializer, but it is given by an 6911 explicit statement later in the code; no bitwise copy is required. 6912 6913 GCC provides an extension that allows either automatic variables, 6914 or global variables, to be placed in particular registers. This 6915 extension is being used for a particular `VAR_DECL' if 6916 `DECL_REGISTER' holds for the `VAR_DECL', and if 6917 `DECL_ASSEMBLER_NAME' is not equal to `DECL_NAME'. In that case, 6918 `DECL_ASSEMBLER_NAME' is the name of the register into which the 6919 variable will be placed. 6920 6921 `PARM_DECL' 6922 Used to represent a parameter to a function. Treat these nodes 6923 similarly to `VAR_DECL' nodes. These nodes only appear in the 6924 `DECL_ARGUMENTS' for a `FUNCTION_DECL'. 6925 6926 The `DECL_ARG_TYPE' for a `PARM_DECL' is the type that will 6927 actually be used when a value is passed to this function. It may 6928 be a wider type than the `TREE_TYPE' of the parameter; for 6929 example, the ordinary type might be `short' while the 6930 `DECL_ARG_TYPE' is `int'. 6931 6932 `FIELD_DECL' 6933 These nodes represent non-static data members. The `DECL_SIZE' and 6934 `DECL_ALIGN' behave as for `VAR_DECL' nodes. The position of the 6935 field within the parent record is specified by a combination of 6936 three attributes. `DECL_FIELD_OFFSET' is the position, counting 6937 in bytes, of the `DECL_OFFSET_ALIGN'-bit sized word containing the 6938 bit of the field closest to the beginning of the structure. 6939 `DECL_FIELD_BIT_OFFSET' is the bit offset of the first bit of the 6940 field within this word; this may be nonzero even for fields that 6941 are not bit-fields, since `DECL_OFFSET_ALIGN' may be greater than 6942 the natural alignment of the field's type. 6943 6944 If `DECL_C_BIT_FIELD' holds, this field is a bit-field. In a 6945 bit-field, `DECL_BIT_FIELD_TYPE' also contains the type that was 6946 originally specified for it, while DECL_TYPE may be a modified 6947 type with lesser precision, according to the size of the bit field. 6948 6949 `NAMESPACE_DECL' 6950 *Note Namespaces::. 6951 6952 `TEMPLATE_DECL' 6953 These nodes are used to represent class, function, and variable 6954 (static data member) templates. The 6955 `DECL_TEMPLATE_SPECIALIZATIONS' are a `TREE_LIST'. The 6956 `TREE_VALUE' of each node in the list is a `TEMPLATE_DECL's or 6957 `FUNCTION_DECL's representing specializations (including 6958 instantiations) of this template. Back ends can safely ignore 6959 `TEMPLATE_DECL's, but should examine `FUNCTION_DECL' nodes on the 6960 specializations list just as they would ordinary `FUNCTION_DECL' 6961 nodes. 6962 6963 For a class template, the `DECL_TEMPLATE_INSTANTIATIONS' list 6964 contains the instantiations. The `TREE_VALUE' of each node is an 6965 instantiation of the class. The `DECL_TEMPLATE_SPECIALIZATIONS' 6966 contains partial specializations of the class. 6967 6968 `USING_DECL' 6969 Back ends can safely ignore these nodes. 6970 6971 6972 6973 File: gccint.info, Node: Internal structure, Prev: Working with declarations, Up: Declarations 6974 6975 9.5.2 Internal structure 6976 ------------------------ 6977 6978 `DECL' nodes are represented internally as a hierarchy of structures. 6979 6980 * Menu: 6981 6982 * Current structure hierarchy:: The current DECL node structure 6983 hierarchy. 6984 * Adding new DECL node types:: How to add a new DECL node to a 6985 frontend. 6986 6987 6988 File: gccint.info, Node: Current structure hierarchy, Next: Adding new DECL node types, Up: Internal structure 6989 6990 9.5.2.1 Current structure hierarchy 6991 ................................... 6992 6993 `struct tree_decl_minimal' 6994 This is the minimal structure to inherit from in order for common 6995 `DECL' macros to work. The fields it contains are a unique ID, 6996 source location, context, and name. 6997 6998 `struct tree_decl_common' 6999 This structure inherits from `struct tree_decl_minimal'. It 7000 contains fields that most `DECL' nodes need, such as a field to 7001 store alignment, machine mode, size, and attributes. 7002 7003 `struct tree_field_decl' 7004 This structure inherits from `struct tree_decl_common'. It is 7005 used to represent `FIELD_DECL'. 7006 7007 `struct tree_label_decl' 7008 This structure inherits from `struct tree_decl_common'. It is 7009 used to represent `LABEL_DECL'. 7010 7011 `struct tree_translation_unit_decl' 7012 This structure inherits from `struct tree_decl_common'. It is 7013 used to represent `TRANSLATION_UNIT_DECL'. 7014 7015 `struct tree_decl_with_rtl' 7016 This structure inherits from `struct tree_decl_common'. It 7017 contains a field to store the low-level RTL associated with a 7018 `DECL' node. 7019 7020 `struct tree_result_decl' 7021 This structure inherits from `struct tree_decl_with_rtl'. It is 7022 used to represent `RESULT_DECL'. 7023 7024 `struct tree_const_decl' 7025 This structure inherits from `struct tree_decl_with_rtl'. It is 7026 used to represent `CONST_DECL'. 7027 7028 `struct tree_parm_decl' 7029 This structure inherits from `struct tree_decl_with_rtl'. It is 7030 used to represent `PARM_DECL'. 7031 7032 `struct tree_decl_with_vis' 7033 This structure inherits from `struct tree_decl_with_rtl'. It 7034 contains fields necessary to store visibility information, as well 7035 as a section name and assembler name. 7036 7037 `struct tree_var_decl' 7038 This structure inherits from `struct tree_decl_with_vis'. It is 7039 used to represent `VAR_DECL'. 7040 7041 `struct tree_function_decl' 7042 This structure inherits from `struct tree_decl_with_vis'. It is 7043 used to represent `FUNCTION_DECL'. 7044 7045 7046 7047 File: gccint.info, Node: Adding new DECL node types, Prev: Current structure hierarchy, Up: Internal structure 7048 7049 9.5.2.2 Adding new DECL node types 7050 .................................. 7051 7052 Adding a new `DECL' tree consists of the following steps 7053 7054 Add a new tree code for the `DECL' node 7055 For language specific `DECL' nodes, there is a `.def' file in each 7056 frontend directory where the tree code should be added. For 7057 `DECL' nodes that are part of the middle-end, the code should be 7058 added to `tree.def'. 7059 7060 Create a new structure type for the `DECL' node 7061 These structures should inherit from one of the existing 7062 structures in the language hierarchy by using that structure as 7063 the first member. 7064 7065 struct tree_foo_decl 7066 { 7067 struct tree_decl_with_vis common; 7068 } 7069 7070 Would create a structure name `tree_foo_decl' that inherits from 7071 `struct tree_decl_with_vis'. 7072 7073 For language specific `DECL' nodes, this new structure type should 7074 go in the appropriate `.h' file. For `DECL' nodes that are part 7075 of the middle-end, the structure type should go in `tree.h'. 7076 7077 Add a member to the tree structure enumerator for the node 7078 For garbage collection and dynamic checking purposes, each `DECL' 7079 node structure type is required to have a unique enumerator value 7080 specified with it. For language specific `DECL' nodes, this new 7081 enumerator value should go in the appropriate `.def' file. For 7082 `DECL' nodes that are part of the middle-end, the enumerator 7083 values are specified in `treestruct.def'. 7084 7085 Update `union tree_node' 7086 In order to make your new structure type usable, it must be added 7087 to `union tree_node'. For language specific `DECL' nodes, a new 7088 entry should be added to the appropriate `.h' file of the form 7089 struct tree_foo_decl GTY ((tag ("TS_VAR_DECL"))) foo_decl; 7090 For `DECL' nodes that are part of the middle-end, the additional 7091 member goes directly into `union tree_node' in `tree.h'. 7092 7093 Update dynamic checking info 7094 In order to be able to check whether accessing a named portion of 7095 `union tree_node' is legal, and whether a certain `DECL' node 7096 contains one of the enumerated `DECL' node structures in the 7097 hierarchy, a simple lookup table is used. This lookup table needs 7098 to be kept up to date with the tree structure hierarchy, or else 7099 checking and containment macros will fail inappropriately. 7100 7101 For language specific `DECL' nodes, their is an `init_ts' function 7102 in an appropriate `.c' file, which initializes the lookup table. 7103 Code setting up the table for new `DECL' nodes should be added 7104 there. For each `DECL' tree code and enumerator value 7105 representing a member of the inheritance hierarchy, the table 7106 should contain 1 if that tree code inherits (directly or 7107 indirectly) from that member. Thus, a `FOO_DECL' node derived 7108 from `struct decl_with_rtl', and enumerator value `TS_FOO_DECL', 7109 would be set up as follows 7110 tree_contains_struct[FOO_DECL][TS_FOO_DECL] = 1; 7111 tree_contains_struct[FOO_DECL][TS_DECL_WRTL] = 1; 7112 tree_contains_struct[FOO_DECL][TS_DECL_COMMON] = 1; 7113 tree_contains_struct[FOO_DECL][TS_DECL_MINIMAL] = 1; 7114 7115 For `DECL' nodes that are part of the middle-end, the setup code 7116 goes into `tree.c'. 7117 7118 Add macros to access any new fields and flags 7119 Each added field or flag should have a macro that is used to access 7120 it, that performs appropriate checking to ensure only the right 7121 type of `DECL' nodes access the field. 7122 7123 These macros generally take the following form 7124 #define FOO_DECL_FIELDNAME(NODE) FOO_DECL_CHECK(NODE)->foo_decl.fieldname 7125 However, if the structure is simply a base class for further 7126 structures, something like the following should be used 7127 #define BASE_STRUCT_CHECK(T) CONTAINS_STRUCT_CHECK(T, TS_BASE_STRUCT) 7128 #define BASE_STRUCT_FIELDNAME(NODE) \ 7129 (BASE_STRUCT_CHECK(NODE)->base_struct.fieldname 7130 7131 7132 7133 File: gccint.info, Node: Functions, Next: Declarations, Prev: Scopes, Up: Trees 7134 7135 9.6 Functions 7136 ============= 7137 7138 A function is represented by a `FUNCTION_DECL' node. A set of 7139 overloaded functions is sometimes represented by a `OVERLOAD' node. 7140 7141 An `OVERLOAD' node is not a declaration, so none of the `DECL_' macros 7142 should be used on an `OVERLOAD'. An `OVERLOAD' node is similar to a 7143 `TREE_LIST'. Use `OVL_CURRENT' to get the function associated with an 7144 `OVERLOAD' node; use `OVL_NEXT' to get the next `OVERLOAD' node in the 7145 list of overloaded functions. The macros `OVL_CURRENT' and `OVL_NEXT' 7146 are actually polymorphic; you can use them to work with `FUNCTION_DECL' 7147 nodes as well as with overloads. In the case of a `FUNCTION_DECL', 7148 `OVL_CURRENT' will always return the function itself, and `OVL_NEXT' 7149 will always be `NULL_TREE'. 7150 7151 To determine the scope of a function, you can use the `DECL_CONTEXT' 7152 macro. This macro will return the class (either a `RECORD_TYPE' or a 7153 `UNION_TYPE') or namespace (a `NAMESPACE_DECL') of which the function 7154 is a member. For a virtual function, this macro returns the class in 7155 which the function was actually defined, not the base class in which 7156 the virtual declaration occurred. 7157 7158 If a friend function is defined in a class scope, the 7159 `DECL_FRIEND_CONTEXT' macro can be used to determine the class in which 7160 it was defined. For example, in 7161 class C { friend void f() {} }; 7162 the `DECL_CONTEXT' for `f' will be the `global_namespace', but the 7163 `DECL_FRIEND_CONTEXT' will be the `RECORD_TYPE' for `C'. 7164 7165 In C, the `DECL_CONTEXT' for a function maybe another function. This 7166 representation indicates that the GNU nested function extension is in 7167 use. For details on the semantics of nested functions, see the GCC 7168 Manual. The nested function can refer to local variables in its 7169 containing function. Such references are not explicitly marked in the 7170 tree structure; back ends must look at the `DECL_CONTEXT' for the 7171 referenced `VAR_DECL'. If the `DECL_CONTEXT' for the referenced 7172 `VAR_DECL' is not the same as the function currently being processed, 7173 and neither `DECL_EXTERNAL' nor `TREE_STATIC' hold, then the reference 7174 is to a local variable in a containing function, and the back end must 7175 take appropriate action. 7176 7177 * Menu: 7178 7179 * Function Basics:: Function names, linkage, and so forth. 7180 * Function Bodies:: The statements that make up a function body. 7181 7182 7183 File: gccint.info, Node: Function Basics, Next: Function Bodies, Up: Functions 7184 7185 9.6.1 Function Basics 7186 --------------------- 7187 7188 The following macros and functions can be used on a `FUNCTION_DECL': 7189 `DECL_MAIN_P' 7190 This predicate holds for a function that is the program entry point 7191 `::code'. 7192 7193 `DECL_NAME' 7194 This macro returns the unqualified name of the function, as an 7195 `IDENTIFIER_NODE'. For an instantiation of a function template, 7196 the `DECL_NAME' is the unqualified name of the template, not 7197 something like `f<int>'. The value of `DECL_NAME' is undefined 7198 when used on a constructor, destructor, overloaded operator, or 7199 type-conversion operator, or any function that is implicitly 7200 generated by the compiler. See below for macros that can be used 7201 to distinguish these cases. 7202 7203 `DECL_ASSEMBLER_NAME' 7204 This macro returns the mangled name of the function, also an 7205 `IDENTIFIER_NODE'. This name does not contain leading underscores 7206 on systems that prefix all identifiers with underscores. The 7207 mangled name is computed in the same way on all platforms; if 7208 special processing is required to deal with the object file format 7209 used on a particular platform, it is the responsibility of the 7210 back end to perform those modifications. (Of course, the back end 7211 should not modify `DECL_ASSEMBLER_NAME' itself.) 7212 7213 Using `DECL_ASSEMBLER_NAME' will cause additional memory to be 7214 allocated (for the mangled name of the entity) so it should be used 7215 only when emitting assembly code. It should not be used within the 7216 optimizers to determine whether or not two declarations are the 7217 same, even though some of the existing optimizers do use it in 7218 that way. These uses will be removed over time. 7219 7220 `DECL_EXTERNAL' 7221 This predicate holds if the function is undefined. 7222 7223 `TREE_PUBLIC' 7224 This predicate holds if the function has external linkage. 7225 7226 `DECL_LOCAL_FUNCTION_P' 7227 This predicate holds if the function was declared at block scope, 7228 even though it has a global scope. 7229 7230 `DECL_ANTICIPATED' 7231 This predicate holds if the function is a built-in function but its 7232 prototype is not yet explicitly declared. 7233 7234 `DECL_EXTERN_C_FUNCTION_P' 7235 This predicate holds if the function is declared as an ``extern 7236 "C"'' function. 7237 7238 `DECL_LINKONCE_P' 7239 This macro holds if multiple copies of this function may be 7240 emitted in various translation units. It is the responsibility of 7241 the linker to merge the various copies. Template instantiations 7242 are the most common example of functions for which 7243 `DECL_LINKONCE_P' holds; G++ instantiates needed templates in all 7244 translation units which require them, and then relies on the 7245 linker to remove duplicate instantiations. 7246 7247 FIXME: This macro is not yet implemented. 7248 7249 `DECL_FUNCTION_MEMBER_P' 7250 This macro holds if the function is a member of a class, rather 7251 than a member of a namespace. 7252 7253 `DECL_STATIC_FUNCTION_P' 7254 This predicate holds if the function a static member function. 7255 7256 `DECL_NONSTATIC_MEMBER_FUNCTION_P' 7257 This macro holds for a non-static member function. 7258 7259 `DECL_CONST_MEMFUNC_P' 7260 This predicate holds for a `const'-member function. 7261 7262 `DECL_VOLATILE_MEMFUNC_P' 7263 This predicate holds for a `volatile'-member function. 7264 7265 `DECL_CONSTRUCTOR_P' 7266 This macro holds if the function is a constructor. 7267 7268 `DECL_NONCONVERTING_P' 7269 This predicate holds if the constructor is a non-converting 7270 constructor. 7271 7272 `DECL_COMPLETE_CONSTRUCTOR_P' 7273 This predicate holds for a function which is a constructor for an 7274 object of a complete type. 7275 7276 `DECL_BASE_CONSTRUCTOR_P' 7277 This predicate holds for a function which is a constructor for a 7278 base class sub-object. 7279 7280 `DECL_COPY_CONSTRUCTOR_P' 7281 This predicate holds for a function which is a copy-constructor. 7282 7283 `DECL_DESTRUCTOR_P' 7284 This macro holds if the function is a destructor. 7285 7286 `DECL_COMPLETE_DESTRUCTOR_P' 7287 This predicate holds if the function is the destructor for an 7288 object a complete type. 7289 7290 `DECL_OVERLOADED_OPERATOR_P' 7291 This macro holds if the function is an overloaded operator. 7292 7293 `DECL_CONV_FN_P' 7294 This macro holds if the function is a type-conversion operator. 7295 7296 `DECL_GLOBAL_CTOR_P' 7297 This predicate holds if the function is a file-scope initialization 7298 function. 7299 7300 `DECL_GLOBAL_DTOR_P' 7301 This predicate holds if the function is a file-scope finalization 7302 function. 7303 7304 `DECL_THUNK_P' 7305 This predicate holds if the function is a thunk. 7306 7307 These functions represent stub code that adjusts the `this' pointer 7308 and then jumps to another function. When the jumped-to function 7309 returns, control is transferred directly to the caller, without 7310 returning to the thunk. The first parameter to the thunk is 7311 always the `this' pointer; the thunk should add `THUNK_DELTA' to 7312 this value. (The `THUNK_DELTA' is an `int', not an `INTEGER_CST'.) 7313 7314 Then, if `THUNK_VCALL_OFFSET' (an `INTEGER_CST') is nonzero the 7315 adjusted `this' pointer must be adjusted again. The complete 7316 calculation is given by the following pseudo-code: 7317 7318 this += THUNK_DELTA 7319 if (THUNK_VCALL_OFFSET) 7320 this += (*((ptrdiff_t **) this))[THUNK_VCALL_OFFSET] 7321 7322 Finally, the thunk should jump to the location given by 7323 `DECL_INITIAL'; this will always be an expression for the address 7324 of a function. 7325 7326 `DECL_NON_THUNK_FUNCTION_P' 7327 This predicate holds if the function is _not_ a thunk function. 7328 7329 `GLOBAL_INIT_PRIORITY' 7330 If either `DECL_GLOBAL_CTOR_P' or `DECL_GLOBAL_DTOR_P' holds, then 7331 this gives the initialization priority for the function. The 7332 linker will arrange that all functions for which 7333 `DECL_GLOBAL_CTOR_P' holds are run in increasing order of priority 7334 before `main' is called. When the program exits, all functions for 7335 which `DECL_GLOBAL_DTOR_P' holds are run in the reverse order. 7336 7337 `DECL_ARTIFICIAL' 7338 This macro holds if the function was implicitly generated by the 7339 compiler, rather than explicitly declared. In addition to 7340 implicitly generated class member functions, this macro holds for 7341 the special functions created to implement static initialization 7342 and destruction, to compute run-time type information, and so 7343 forth. 7344 7345 `DECL_ARGUMENTS' 7346 This macro returns the `PARM_DECL' for the first argument to the 7347 function. Subsequent `PARM_DECL' nodes can be obtained by 7348 following the `TREE_CHAIN' links. 7349 7350 `DECL_RESULT' 7351 This macro returns the `RESULT_DECL' for the function. 7352 7353 `TREE_TYPE' 7354 This macro returns the `FUNCTION_TYPE' or `METHOD_TYPE' for the 7355 function. 7356 7357 `TYPE_RAISES_EXCEPTIONS' 7358 This macro returns the list of exceptions that a (member-)function 7359 can raise. The returned list, if non `NULL', is comprised of nodes 7360 whose `TREE_VALUE' represents a type. 7361 7362 `TYPE_NOTHROW_P' 7363 This predicate holds when the exception-specification of its 7364 arguments is of the form ``()''. 7365 7366 `DECL_ARRAY_DELETE_OPERATOR_P' 7367 This predicate holds if the function an overloaded `operator 7368 delete[]'. 7369 7370 `DECL_FUNCTION_SPECIFIC_TARGET' 7371 This macro returns a tree node that holds the target options that 7372 are to be used to compile this particular function or `NULL_TREE' 7373 if the function is to be compiled with the target options 7374 specified on the command line. 7375 7376 `DECL_FUNCTION_SPECIFIC_OPTIMIZATION' 7377 This macro returns a tree node that holds the optimization options 7378 that are to be used to compile this particular function or 7379 `NULL_TREE' if the function is to be compiled with the 7380 optimization options specified on the command line. 7381 7382 7383 File: gccint.info, Node: Function Bodies, Prev: Function Basics, Up: Functions 7384 7385 9.6.2 Function Bodies 7386 --------------------- 7387 7388 A function that has a definition in the current translation unit will 7389 have a non-`NULL' `DECL_INITIAL'. However, back ends should not make 7390 use of the particular value given by `DECL_INITIAL'. 7391 7392 The `DECL_SAVED_TREE' macro will give the complete body of the 7393 function. 7394 7395 9.6.2.1 Statements 7396 .................. 7397 7398 There are tree nodes corresponding to all of the source-level statement 7399 constructs, used within the C and C++ frontends. These are enumerated 7400 here, together with a list of the various macros that can be used to 7401 obtain information about them. There are a few macros that can be used 7402 with all statements: 7403 7404 `STMT_IS_FULL_EXPR_P' 7405 In C++, statements normally constitute "full expressions"; 7406 temporaries created during a statement are destroyed when the 7407 statement is complete. However, G++ sometimes represents 7408 expressions by statements; these statements will not have 7409 `STMT_IS_FULL_EXPR_P' set. Temporaries created during such 7410 statements should be destroyed when the innermost enclosing 7411 statement with `STMT_IS_FULL_EXPR_P' set is exited. 7412 7413 7414 Here is the list of the various statement nodes, and the macros used to 7415 access them. This documentation describes the use of these nodes in 7416 non-template functions (including instantiations of template functions). 7417 In template functions, the same nodes are used, but sometimes in 7418 slightly different ways. 7419 7420 Many of the statements have substatements. For example, a `while' 7421 loop will have a body, which is itself a statement. If the substatement 7422 is `NULL_TREE', it is considered equivalent to a statement consisting 7423 of a single `;', i.e., an expression statement in which the expression 7424 has been omitted. A substatement may in fact be a list of statements, 7425 connected via their `TREE_CHAIN's. So, you should always process the 7426 statement tree by looping over substatements, like this: 7427 void process_stmt (stmt) 7428 tree stmt; 7429 { 7430 while (stmt) 7431 { 7432 switch (TREE_CODE (stmt)) 7433 { 7434 case IF_STMT: 7435 process_stmt (THEN_CLAUSE (stmt)); 7436 /* More processing here. */ 7437 break; 7438 7439 ... 7440 } 7441 7442 stmt = TREE_CHAIN (stmt); 7443 } 7444 } 7445 In other words, while the `then' clause of an `if' statement in C++ 7446 can be only one statement (although that one statement may be a 7447 compound statement), the intermediate representation will sometimes use 7448 several statements chained together. 7449 7450 `ASM_EXPR' 7451 Used to represent an inline assembly statement. For an inline 7452 assembly statement like: 7453 asm ("mov x, y"); 7454 The `ASM_STRING' macro will return a `STRING_CST' node for `"mov 7455 x, y"'. If the original statement made use of the 7456 extended-assembly syntax, then `ASM_OUTPUTS', `ASM_INPUTS', and 7457 `ASM_CLOBBERS' will be the outputs, inputs, and clobbers for the 7458 statement, represented as `STRING_CST' nodes. The 7459 extended-assembly syntax looks like: 7460 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle)); 7461 The first string is the `ASM_STRING', containing the instruction 7462 template. The next two strings are the output and inputs, 7463 respectively; this statement has no clobbers. As this example 7464 indicates, "plain" assembly statements are merely a special case 7465 of extended assembly statements; they have no cv-qualifiers, 7466 outputs, inputs, or clobbers. All of the strings will be 7467 `NUL'-terminated, and will contain no embedded `NUL'-characters. 7468 7469 If the assembly statement is declared `volatile', or if the 7470 statement was not an extended assembly statement, and is therefore 7471 implicitly volatile, then the predicate `ASM_VOLATILE_P' will hold 7472 of the `ASM_EXPR'. 7473 7474 `BREAK_STMT' 7475 Used to represent a `break' statement. There are no additional 7476 fields. 7477 7478 `CASE_LABEL_EXPR' 7479 Use to represent a `case' label, range of `case' labels, or a 7480 `default' label. If `CASE_LOW' is `NULL_TREE', then this is a 7481 `default' label. Otherwise, if `CASE_HIGH' is `NULL_TREE', then 7482 this is an ordinary `case' label. In this case, `CASE_LOW' is an 7483 expression giving the value of the label. Both `CASE_LOW' and 7484 `CASE_HIGH' are `INTEGER_CST' nodes. These values will have the 7485 same type as the condition expression in the switch statement. 7486 7487 Otherwise, if both `CASE_LOW' and `CASE_HIGH' are defined, the 7488 statement is a range of case labels. Such statements originate 7489 with the extension that allows users to write things of the form: 7490 case 2 ... 5: 7491 The first value will be `CASE_LOW', while the second will be 7492 `CASE_HIGH'. 7493 7494 `CLEANUP_STMT' 7495 Used to represent an action that should take place upon exit from 7496 the enclosing scope. Typically, these actions are calls to 7497 destructors for local objects, but back ends cannot rely on this 7498 fact. If these nodes are in fact representing such destructors, 7499 `CLEANUP_DECL' will be the `VAR_DECL' destroyed. Otherwise, 7500 `CLEANUP_DECL' will be `NULL_TREE'. In any case, the 7501 `CLEANUP_EXPR' is the expression to execute. The cleanups 7502 executed on exit from a scope should be run in the reverse order 7503 of the order in which the associated `CLEANUP_STMT's were 7504 encountered. 7505 7506 `CONTINUE_STMT' 7507 Used to represent a `continue' statement. There are no additional 7508 fields. 7509 7510 `CTOR_STMT' 7511 Used to mark the beginning (if `CTOR_BEGIN_P' holds) or end (if 7512 `CTOR_END_P' holds of the main body of a constructor. See also 7513 `SUBOBJECT' for more information on how to use these nodes. 7514 7515 `DECL_STMT' 7516 Used to represent a local declaration. The `DECL_STMT_DECL' macro 7517 can be used to obtain the entity declared. This declaration may 7518 be a `LABEL_DECL', indicating that the label declared is a local 7519 label. (As an extension, GCC allows the declaration of labels 7520 with scope.) In C, this declaration may be a `FUNCTION_DECL', 7521 indicating the use of the GCC nested function extension. For more 7522 information, *note Functions::. 7523 7524 `DO_STMT' 7525 Used to represent a `do' loop. The body of the loop is given by 7526 `DO_BODY' while the termination condition for the loop is given by 7527 `DO_COND'. The condition for a `do'-statement is always an 7528 expression. 7529 7530 `EMPTY_CLASS_EXPR' 7531 Used to represent a temporary object of a class with no data whose 7532 address is never taken. (All such objects are interchangeable.) 7533 The `TREE_TYPE' represents the type of the object. 7534 7535 `EXPR_STMT' 7536 Used to represent an expression statement. Use `EXPR_STMT_EXPR' to 7537 obtain the expression. 7538 7539 `FOR_STMT' 7540 Used to represent a `for' statement. The `FOR_INIT_STMT' is the 7541 initialization statement for the loop. The `FOR_COND' is the 7542 termination condition. The `FOR_EXPR' is the expression executed 7543 right before the `FOR_COND' on each loop iteration; often, this 7544 expression increments a counter. The body of the loop is given by 7545 `FOR_BODY'. Note that `FOR_INIT_STMT' and `FOR_BODY' return 7546 statements, while `FOR_COND' and `FOR_EXPR' return expressions. 7547 7548 `GOTO_EXPR' 7549 Used to represent a `goto' statement. The `GOTO_DESTINATION' will 7550 usually be a `LABEL_DECL'. However, if the "computed goto" 7551 extension has been used, the `GOTO_DESTINATION' will be an 7552 arbitrary expression indicating the destination. This expression 7553 will always have pointer type. 7554 7555 `HANDLER' 7556 Used to represent a C++ `catch' block. The `HANDLER_TYPE' is the 7557 type of exception that will be caught by this handler; it is equal 7558 (by pointer equality) to `NULL' if this handler is for all types. 7559 `HANDLER_PARMS' is the `DECL_STMT' for the catch parameter, and 7560 `HANDLER_BODY' is the code for the block itself. 7561 7562 `IF_STMT' 7563 Used to represent an `if' statement. The `IF_COND' is the 7564 expression. 7565 7566 If the condition is a `TREE_LIST', then the `TREE_PURPOSE' is a 7567 statement (usually a `DECL_STMT'). Each time the condition is 7568 evaluated, the statement should be executed. Then, the 7569 `TREE_VALUE' should be used as the conditional expression itself. 7570 This representation is used to handle C++ code like this: 7571 7572 if (int i = 7) ... 7573 7574 where there is a new local variable (or variables) declared within 7575 the condition. 7576 7577 The `THEN_CLAUSE' represents the statement given by the `then' 7578 condition, while the `ELSE_CLAUSE' represents the statement given 7579 by the `else' condition. 7580 7581 `LABEL_EXPR' 7582 Used to represent a label. The `LABEL_DECL' declared by this 7583 statement can be obtained with the `LABEL_EXPR_LABEL' macro. The 7584 `IDENTIFIER_NODE' giving the name of the label can be obtained from 7585 the `LABEL_DECL' with `DECL_NAME'. 7586 7587 `RETURN_STMT' 7588 Used to represent a `return' statement. The `RETURN_EXPR' is the 7589 expression returned; it will be `NULL_TREE' if the statement was 7590 just 7591 return; 7592 7593 `SUBOBJECT' 7594 In a constructor, these nodes are used to mark the point at which a 7595 subobject of `this' is fully constructed. If, after this point, an 7596 exception is thrown before a `CTOR_STMT' with `CTOR_END_P' set is 7597 encountered, the `SUBOBJECT_CLEANUP' must be executed. The 7598 cleanups must be executed in the reverse order in which they 7599 appear. 7600 7601 `SWITCH_STMT' 7602 Used to represent a `switch' statement. The `SWITCH_STMT_COND' is 7603 the expression on which the switch is occurring. See the 7604 documentation for an `IF_STMT' for more information on the 7605 representation used for the condition. The `SWITCH_STMT_BODY' is 7606 the body of the switch statement. The `SWITCH_STMT_TYPE' is the 7607 original type of switch expression as given in the source, before 7608 any compiler conversions. 7609 7610 `TRY_BLOCK' 7611 Used to represent a `try' block. The body of the try block is 7612 given by `TRY_STMTS'. Each of the catch blocks is a `HANDLER' 7613 node. The first handler is given by `TRY_HANDLERS'. Subsequent 7614 handlers are obtained by following the `TREE_CHAIN' link from one 7615 handler to the next. The body of the handler is given by 7616 `HANDLER_BODY'. 7617 7618 If `CLEANUP_P' holds of the `TRY_BLOCK', then the `TRY_HANDLERS' 7619 will not be a `HANDLER' node. Instead, it will be an expression 7620 that should be executed if an exception is thrown in the try 7621 block. It must rethrow the exception after executing that code. 7622 And, if an exception is thrown while the expression is executing, 7623 `terminate' must be called. 7624 7625 `USING_STMT' 7626 Used to represent a `using' directive. The namespace is given by 7627 `USING_STMT_NAMESPACE', which will be a NAMESPACE_DECL. This node 7628 is needed inside template functions, to implement using directives 7629 during instantiation. 7630 7631 `WHILE_STMT' 7632 Used to represent a `while' loop. The `WHILE_COND' is the 7633 termination condition for the loop. See the documentation for an 7634 `IF_STMT' for more information on the representation used for the 7635 condition. 7636 7637 The `WHILE_BODY' is the body of the loop. 7638 7639 7640 7641 File: gccint.info, Node: Attributes, Next: Expression trees, Prev: Declarations, Up: Trees 7642 7643 9.7 Attributes in trees 7644 ======================= 7645 7646 Attributes, as specified using the `__attribute__' keyword, are 7647 represented internally as a `TREE_LIST'. The `TREE_PURPOSE' is the 7648 name of the attribute, as an `IDENTIFIER_NODE'. The `TREE_VALUE' is a 7649 `TREE_LIST' of the arguments of the attribute, if any, or `NULL_TREE' 7650 if there are no arguments; the arguments are stored as the `TREE_VALUE' 7651 of successive entries in the list, and may be identifiers or 7652 expressions. The `TREE_CHAIN' of the attribute is the next attribute 7653 in a list of attributes applying to the same declaration or type, or 7654 `NULL_TREE' if there are no further attributes in the list. 7655 7656 Attributes may be attached to declarations and to types; these 7657 attributes may be accessed with the following macros. All attributes 7658 are stored in this way, and many also cause other changes to the 7659 declaration or type or to other internal compiler data structures. 7660 7661 -- Tree Macro: tree DECL_ATTRIBUTES (tree DECL) 7662 This macro returns the attributes on the declaration DECL. 7663 7664 -- Tree Macro: tree TYPE_ATTRIBUTES (tree TYPE) 7665 This macro returns the attributes on the type TYPE. 7666 7667 7668 File: gccint.info, Node: Expression trees, Prev: Attributes, Up: Trees 7669 7670 9.8 Expressions 7671 =============== 7672 7673 The internal representation for expressions is for the most part quite 7674 straightforward. However, there are a few facts that one must bear in 7675 mind. In particular, the expression "tree" is actually a directed 7676 acyclic graph. (For example there may be many references to the integer 7677 constant zero throughout the source program; many of these will be 7678 represented by the same expression node.) You should not rely on 7679 certain kinds of node being shared, nor should you rely on certain 7680 kinds of nodes being unshared. 7681 7682 The following macros can be used with all expression nodes: 7683 7684 `TREE_TYPE' 7685 Returns the type of the expression. This value may not be 7686 precisely the same type that would be given the expression in the 7687 original program. 7688 7689 In what follows, some nodes that one might expect to always have type 7690 `bool' are documented to have either integral or boolean type. At some 7691 point in the future, the C front end may also make use of this same 7692 intermediate representation, and at this point these nodes will 7693 certainly have integral type. The previous sentence is not meant to 7694 imply that the C++ front end does not or will not give these nodes 7695 integral type. 7696 7697 Below, we list the various kinds of expression nodes. Except where 7698 noted otherwise, the operands to an expression are accessed using the 7699 `TREE_OPERAND' macro. For example, to access the first operand to a 7700 binary plus expression `expr', use: 7701 7702 TREE_OPERAND (expr, 0) 7703 As this example indicates, the operands are zero-indexed. 7704 7705 All the expressions starting with `OMP_' represent directives and 7706 clauses used by the OpenMP API `http://www.openmp.org/'. 7707 7708 The table below begins with constants, moves on to unary expressions, 7709 then proceeds to binary expressions, and concludes with various other 7710 kinds of expressions: 7711 7712 `INTEGER_CST' 7713 These nodes represent integer constants. Note that the type of 7714 these constants is obtained with `TREE_TYPE'; they are not always 7715 of type `int'. In particular, `char' constants are represented 7716 with `INTEGER_CST' nodes. The value of the integer constant `e' is 7717 given by 7718 ((TREE_INT_CST_HIGH (e) << HOST_BITS_PER_WIDE_INT) 7719 + TREE_INST_CST_LOW (e)) 7720 HOST_BITS_PER_WIDE_INT is at least thirty-two on all platforms. 7721 Both `TREE_INT_CST_HIGH' and `TREE_INT_CST_LOW' return a 7722 `HOST_WIDE_INT'. The value of an `INTEGER_CST' is interpreted as 7723 a signed or unsigned quantity depending on the type of the 7724 constant. In general, the expression given above will overflow, 7725 so it should not be used to calculate the value of the constant. 7726 7727 The variable `integer_zero_node' is an integer constant with value 7728 zero. Similarly, `integer_one_node' is an integer constant with 7729 value one. The `size_zero_node' and `size_one_node' variables are 7730 analogous, but have type `size_t' rather than `int'. 7731 7732 The function `tree_int_cst_lt' is a predicate which holds if its 7733 first argument is less than its second. Both constants are 7734 assumed to have the same signedness (i.e., either both should be 7735 signed or both should be unsigned.) The full width of the 7736 constant is used when doing the comparison; the usual rules about 7737 promotions and conversions are ignored. Similarly, 7738 `tree_int_cst_equal' holds if the two constants are equal. The 7739 `tree_int_cst_sgn' function returns the sign of a constant. The 7740 value is `1', `0', or `-1' according on whether the constant is 7741 greater than, equal to, or less than zero. Again, the signedness 7742 of the constant's type is taken into account; an unsigned constant 7743 is never less than zero, no matter what its bit-pattern. 7744 7745 `REAL_CST' 7746 FIXME: Talk about how to obtain representations of this constant, 7747 do comparisons, and so forth. 7748 7749 `FIXED_CST' 7750 These nodes represent fixed-point constants. The type of these 7751 constants is obtained with `TREE_TYPE'. `TREE_FIXED_CST_PTR' 7752 points to to struct fixed_value; `TREE_FIXED_CST' returns the 7753 structure itself. Struct fixed_value contains `data' with the 7754 size of two HOST_BITS_PER_WIDE_INT and `mode' as the associated 7755 fixed-point machine mode for `data'. 7756 7757 `COMPLEX_CST' 7758 These nodes are used to represent complex number constants, that 7759 is a `__complex__' whose parts are constant nodes. The 7760 `TREE_REALPART' and `TREE_IMAGPART' return the real and the 7761 imaginary parts respectively. 7762 7763 `VECTOR_CST' 7764 These nodes are used to represent vector constants, whose parts are 7765 constant nodes. Each individual constant node is either an 7766 integer or a double constant node. The first operand is a 7767 `TREE_LIST' of the constant nodes and is accessed through 7768 `TREE_VECTOR_CST_ELTS'. 7769 7770 `STRING_CST' 7771 These nodes represent string-constants. The `TREE_STRING_LENGTH' 7772 returns the length of the string, as an `int'. The 7773 `TREE_STRING_POINTER' is a `char*' containing the string itself. 7774 The string may not be `NUL'-terminated, and it may contain 7775 embedded `NUL' characters. Therefore, the `TREE_STRING_LENGTH' 7776 includes the trailing `NUL' if it is present. 7777 7778 For wide string constants, the `TREE_STRING_LENGTH' is the number 7779 of bytes in the string, and the `TREE_STRING_POINTER' points to an 7780 array of the bytes of the string, as represented on the target 7781 system (that is, as integers in the target endianness). Wide and 7782 non-wide string constants are distinguished only by the `TREE_TYPE' 7783 of the `STRING_CST'. 7784 7785 FIXME: The formats of string constants are not well-defined when 7786 the target system bytes are not the same width as host system 7787 bytes. 7788 7789 `PTRMEM_CST' 7790 These nodes are used to represent pointer-to-member constants. The 7791 `PTRMEM_CST_CLASS' is the class type (either a `RECORD_TYPE' or 7792 `UNION_TYPE' within which the pointer points), and the 7793 `PTRMEM_CST_MEMBER' is the declaration for the pointed to object. 7794 Note that the `DECL_CONTEXT' for the `PTRMEM_CST_MEMBER' is in 7795 general different from the `PTRMEM_CST_CLASS'. For example, given: 7796 struct B { int i; }; 7797 struct D : public B {}; 7798 int D::*dp = &D::i; 7799 The `PTRMEM_CST_CLASS' for `&D::i' is `D', even though the 7800 `DECL_CONTEXT' for the `PTRMEM_CST_MEMBER' is `B', since `B::i' is 7801 a member of `B', not `D'. 7802 7803 `VAR_DECL' 7804 These nodes represent variables, including static data members. 7805 For more information, *note Declarations::. 7806 7807 `NEGATE_EXPR' 7808 These nodes represent unary negation of the single operand, for 7809 both integer and floating-point types. The type of negation can be 7810 determined by looking at the type of the expression. 7811 7812 The behavior of this operation on signed arithmetic overflow is 7813 controlled by the `flag_wrapv' and `flag_trapv' variables. 7814 7815 `ABS_EXPR' 7816 These nodes represent the absolute value of the single operand, for 7817 both integer and floating-point types. This is typically used to 7818 implement the `abs', `labs' and `llabs' builtins for integer 7819 types, and the `fabs', `fabsf' and `fabsl' builtins for floating 7820 point types. The type of abs operation can be determined by 7821 looking at the type of the expression. 7822 7823 This node is not used for complex types. To represent the modulus 7824 or complex abs of a complex value, use the `BUILT_IN_CABS', 7825 `BUILT_IN_CABSF' or `BUILT_IN_CABSL' builtins, as used to 7826 implement the C99 `cabs', `cabsf' and `cabsl' built-in functions. 7827 7828 `BIT_NOT_EXPR' 7829 These nodes represent bitwise complement, and will always have 7830 integral type. The only operand is the value to be complemented. 7831 7832 `TRUTH_NOT_EXPR' 7833 These nodes represent logical negation, and will always have 7834 integral (or boolean) type. The operand is the value being 7835 negated. The type of the operand and that of the result are 7836 always of `BOOLEAN_TYPE' or `INTEGER_TYPE'. 7837 7838 `PREDECREMENT_EXPR' 7839 `PREINCREMENT_EXPR' 7840 `POSTDECREMENT_EXPR' 7841 `POSTINCREMENT_EXPR' 7842 These nodes represent increment and decrement expressions. The 7843 value of the single operand is computed, and the operand 7844 incremented or decremented. In the case of `PREDECREMENT_EXPR' and 7845 `PREINCREMENT_EXPR', the value of the expression is the value 7846 resulting after the increment or decrement; in the case of 7847 `POSTDECREMENT_EXPR' and `POSTINCREMENT_EXPR' is the value before 7848 the increment or decrement occurs. The type of the operand, like 7849 that of the result, will be either integral, boolean, or 7850 floating-point. 7851 7852 `ADDR_EXPR' 7853 These nodes are used to represent the address of an object. (These 7854 expressions will always have pointer or reference type.) The 7855 operand may be another expression, or it may be a declaration. 7856 7857 As an extension, GCC allows users to take the address of a label. 7858 In this case, the operand of the `ADDR_EXPR' will be a 7859 `LABEL_DECL'. The type of such an expression is `void*'. 7860 7861 If the object addressed is not an lvalue, a temporary is created, 7862 and the address of the temporary is used. 7863 7864 `INDIRECT_REF' 7865 These nodes are used to represent the object pointed to by a 7866 pointer. The operand is the pointer being dereferenced; it will 7867 always have pointer or reference type. 7868 7869 `FIX_TRUNC_EXPR' 7870 These nodes represent conversion of a floating-point value to an 7871 integer. The single operand will have a floating-point type, while 7872 the complete expression will have an integral (or boolean) type. 7873 The operand is rounded towards zero. 7874 7875 `FLOAT_EXPR' 7876 These nodes represent conversion of an integral (or boolean) value 7877 to a floating-point value. The single operand will have integral 7878 type, while the complete expression will have a floating-point 7879 type. 7880 7881 FIXME: How is the operand supposed to be rounded? Is this 7882 dependent on `-mieee'? 7883 7884 `COMPLEX_EXPR' 7885 These nodes are used to represent complex numbers constructed from 7886 two expressions of the same (integer or real) type. The first 7887 operand is the real part and the second operand is the imaginary 7888 part. 7889 7890 `CONJ_EXPR' 7891 These nodes represent the conjugate of their operand. 7892 7893 `REALPART_EXPR' 7894 `IMAGPART_EXPR' 7895 These nodes represent respectively the real and the imaginary parts 7896 of complex numbers (their sole argument). 7897 7898 `NON_LVALUE_EXPR' 7899 These nodes indicate that their one and only operand is not an 7900 lvalue. A back end can treat these identically to the single 7901 operand. 7902 7903 `NOP_EXPR' 7904 These nodes are used to represent conversions that do not require 7905 any code-generation. For example, conversion of a `char*' to an 7906 `int*' does not require any code be generated; such a conversion is 7907 represented by a `NOP_EXPR'. The single operand is the expression 7908 to be converted. The conversion from a pointer to a reference is 7909 also represented with a `NOP_EXPR'. 7910 7911 `CONVERT_EXPR' 7912 These nodes are similar to `NOP_EXPR's, but are used in those 7913 situations where code may need to be generated. For example, if an 7914 `int*' is converted to an `int' code may need to be generated on 7915 some platforms. These nodes are never used for C++-specific 7916 conversions, like conversions between pointers to different 7917 classes in an inheritance hierarchy. Any adjustments that need to 7918 be made in such cases are always indicated explicitly. Similarly, 7919 a user-defined conversion is never represented by a 7920 `CONVERT_EXPR'; instead, the function calls are made explicit. 7921 7922 `FIXED_CONVERT_EXPR' 7923 These nodes are used to represent conversions that involve 7924 fixed-point values. For example, from a fixed-point value to 7925 another fixed-point value, from an integer to a fixed-point value, 7926 from a fixed-point value to an integer, from a floating-point 7927 value to a fixed-point value, or from a fixed-point value to a 7928 floating-point value. 7929 7930 `THROW_EXPR' 7931 These nodes represent `throw' expressions. The single operand is 7932 an expression for the code that should be executed to throw the 7933 exception. However, there is one implicit action not represented 7934 in that expression; namely the call to `__throw'. This function 7935 takes no arguments. If `setjmp'/`longjmp' exceptions are used, the 7936 function `__sjthrow' is called instead. The normal GCC back end 7937 uses the function `emit_throw' to generate this code; you can 7938 examine this function to see what needs to be done. 7939 7940 `LSHIFT_EXPR' 7941 `RSHIFT_EXPR' 7942 These nodes represent left and right shifts, respectively. The 7943 first operand is the value to shift; it will always be of integral 7944 type. The second operand is an expression for the number of bits 7945 by which to shift. Right shift should be treated as arithmetic, 7946 i.e., the high-order bits should be zero-filled when the 7947 expression has unsigned type and filled with the sign bit when the 7948 expression has signed type. Note that the result is undefined if 7949 the second operand is larger than or equal to the first operand's 7950 type size. 7951 7952 `BIT_IOR_EXPR' 7953 `BIT_XOR_EXPR' 7954 `BIT_AND_EXPR' 7955 These nodes represent bitwise inclusive or, bitwise exclusive or, 7956 and bitwise and, respectively. Both operands will always have 7957 integral type. 7958 7959 `TRUTH_ANDIF_EXPR' 7960 `TRUTH_ORIF_EXPR' 7961 These nodes represent logical "and" and logical "or", respectively. 7962 These operators are not strict; i.e., the second operand is 7963 evaluated only if the value of the expression is not determined by 7964 evaluation of the first operand. The type of the operands and 7965 that of the result are always of `BOOLEAN_TYPE' or `INTEGER_TYPE'. 7966 7967 `TRUTH_AND_EXPR' 7968 `TRUTH_OR_EXPR' 7969 `TRUTH_XOR_EXPR' 7970 These nodes represent logical and, logical or, and logical 7971 exclusive or. They are strict; both arguments are always 7972 evaluated. There are no corresponding operators in C or C++, but 7973 the front end will sometimes generate these expressions anyhow, if 7974 it can tell that strictness does not matter. The type of the 7975 operands and that of the result are always of `BOOLEAN_TYPE' or 7976 `INTEGER_TYPE'. 7977 7978 `POINTER_PLUS_EXPR' 7979 This node represents pointer arithmetic. The first operand is 7980 always a pointer/reference type. The second operand is always an 7981 unsigned integer type compatible with sizetype. This is the only 7982 binary arithmetic operand that can operate on pointer types. 7983 7984 `PLUS_EXPR' 7985 `MINUS_EXPR' 7986 `MULT_EXPR' 7987 These nodes represent various binary arithmetic operations. 7988 Respectively, these operations are addition, subtraction (of the 7989 second operand from the first) and multiplication. Their operands 7990 may have either integral or floating type, but there will never be 7991 case in which one operand is of floating type and the other is of 7992 integral type. 7993 7994 The behavior of these operations on signed arithmetic overflow is 7995 controlled by the `flag_wrapv' and `flag_trapv' variables. 7996 7997 `RDIV_EXPR' 7998 This node represents a floating point division operation. 7999 8000 `TRUNC_DIV_EXPR' 8001 `FLOOR_DIV_EXPR' 8002 `CEIL_DIV_EXPR' 8003 `ROUND_DIV_EXPR' 8004 These nodes represent integer division operations that return an 8005 integer result. `TRUNC_DIV_EXPR' rounds towards zero, 8006 `FLOOR_DIV_EXPR' rounds towards negative infinity, `CEIL_DIV_EXPR' 8007 rounds towards positive infinity and `ROUND_DIV_EXPR' rounds to 8008 the closest integer. Integer division in C and C++ is truncating, 8009 i.e. `TRUNC_DIV_EXPR'. 8010 8011 The behavior of these operations on signed arithmetic overflow, 8012 when dividing the minimum signed integer by minus one, is 8013 controlled by the `flag_wrapv' and `flag_trapv' variables. 8014 8015 `TRUNC_MOD_EXPR' 8016 `FLOOR_MOD_EXPR' 8017 `CEIL_MOD_EXPR' 8018 `ROUND_MOD_EXPR' 8019 These nodes represent the integer remainder or modulus operation. 8020 The integer modulus of two operands `a' and `b' is defined as `a - 8021 (a/b)*b' where the division calculated using the corresponding 8022 division operator. Hence for `TRUNC_MOD_EXPR' this definition 8023 assumes division using truncation towards zero, i.e. 8024 `TRUNC_DIV_EXPR'. Integer remainder in C and C++ uses truncating 8025 division, i.e. `TRUNC_MOD_EXPR'. 8026 8027 `EXACT_DIV_EXPR' 8028 The `EXACT_DIV_EXPR' code is used to represent integer divisions 8029 where the numerator is known to be an exact multiple of the 8030 denominator. This allows the backend to choose between the faster 8031 of `TRUNC_DIV_EXPR', `CEIL_DIV_EXPR' and `FLOOR_DIV_EXPR' for the 8032 current target. 8033 8034 `ARRAY_REF' 8035 These nodes represent array accesses. The first operand is the 8036 array; the second is the index. To calculate the address of the 8037 memory accessed, you must scale the index by the size of the type 8038 of the array elements. The type of these expressions must be the 8039 type of a component of the array. The third and fourth operands 8040 are used after gimplification to represent the lower bound and 8041 component size but should not be used directly; call 8042 `array_ref_low_bound' and `array_ref_element_size' instead. 8043 8044 `ARRAY_RANGE_REF' 8045 These nodes represent access to a range (or "slice") of an array. 8046 The operands are the same as that for `ARRAY_REF' and have the same 8047 meanings. The type of these expressions must be an array whose 8048 component type is the same as that of the first operand. The 8049 range of that array type determines the amount of data these 8050 expressions access. 8051 8052 `TARGET_MEM_REF' 8053 These nodes represent memory accesses whose address directly map to 8054 an addressing mode of the target architecture. The first argument 8055 is `TMR_SYMBOL' and must be a `VAR_DECL' of an object with a fixed 8056 address. The second argument is `TMR_BASE' and the third one is 8057 `TMR_INDEX'. The fourth argument is `TMR_STEP' and must be an 8058 `INTEGER_CST'. The fifth argument is `TMR_OFFSET' and must be an 8059 `INTEGER_CST'. Any of the arguments may be NULL if the 8060 appropriate component does not appear in the address. Address of 8061 the `TARGET_MEM_REF' is determined in the following way. 8062 8063 &TMR_SYMBOL + TMR_BASE + TMR_INDEX * TMR_STEP + TMR_OFFSET 8064 8065 The sixth argument is the reference to the original memory access, 8066 which is preserved for the purposes of the RTL alias analysis. 8067 The seventh argument is a tag representing the results of tree 8068 level alias analysis. 8069 8070 `LT_EXPR' 8071 `LE_EXPR' 8072 `GT_EXPR' 8073 `GE_EXPR' 8074 `EQ_EXPR' 8075 `NE_EXPR' 8076 These nodes represent the less than, less than or equal to, greater 8077 than, greater than or equal to, equal, and not equal comparison 8078 operators. The first and second operand with either be both of 8079 integral type or both of floating type. The result type of these 8080 expressions will always be of integral or boolean type. These 8081 operations return the result type's zero value for false, and the 8082 result type's one value for true. 8083 8084 For floating point comparisons, if we honor IEEE NaNs and either 8085 operand is NaN, then `NE_EXPR' always returns true and the 8086 remaining operators always return false. On some targets, 8087 comparisons against an IEEE NaN, other than equality and 8088 inequality, may generate a floating point exception. 8089 8090 `ORDERED_EXPR' 8091 `UNORDERED_EXPR' 8092 These nodes represent non-trapping ordered and unordered comparison 8093 operators. These operations take two floating point operands and 8094 determine whether they are ordered or unordered relative to each 8095 other. If either operand is an IEEE NaN, their comparison is 8096 defined to be unordered, otherwise the comparison is defined to be 8097 ordered. The result type of these expressions will always be of 8098 integral or boolean type. These operations return the result 8099 type's zero value for false, and the result type's one value for 8100 true. 8101 8102 `UNLT_EXPR' 8103 `UNLE_EXPR' 8104 `UNGT_EXPR' 8105 `UNGE_EXPR' 8106 `UNEQ_EXPR' 8107 `LTGT_EXPR' 8108 These nodes represent the unordered comparison operators. These 8109 operations take two floating point operands and determine whether 8110 the operands are unordered or are less than, less than or equal to, 8111 greater than, greater than or equal to, or equal respectively. For 8112 example, `UNLT_EXPR' returns true if either operand is an IEEE NaN 8113 or the first operand is less than the second. With the possible 8114 exception of `LTGT_EXPR', all of these operations are guaranteed 8115 not to generate a floating point exception. The result type of 8116 these expressions will always be of integral or boolean type. 8117 These operations return the result type's zero value for false, 8118 and the result type's one value for true. 8119 8120 `MODIFY_EXPR' 8121 These nodes represent assignment. The left-hand side is the first 8122 operand; the right-hand side is the second operand. The left-hand 8123 side will be a `VAR_DECL', `INDIRECT_REF', `COMPONENT_REF', or 8124 other lvalue. 8125 8126 These nodes are used to represent not only assignment with `=' but 8127 also compound assignments (like `+='), by reduction to `=' 8128 assignment. In other words, the representation for `i += 3' looks 8129 just like that for `i = i + 3'. 8130 8131 `INIT_EXPR' 8132 These nodes are just like `MODIFY_EXPR', but are used only when a 8133 variable is initialized, rather than assigned to subsequently. 8134 This means that we can assume that the target of the 8135 initialization is not used in computing its own value; any 8136 reference to the lhs in computing the rhs is undefined. 8137 8138 `COMPONENT_REF' 8139 These nodes represent non-static data member accesses. The first 8140 operand is the object (rather than a pointer to it); the second 8141 operand is the `FIELD_DECL' for the data member. The third 8142 operand represents the byte offset of the field, but should not be 8143 used directly; call `component_ref_field_offset' instead. 8144 8145 `COMPOUND_EXPR' 8146 These nodes represent comma-expressions. The first operand is an 8147 expression whose value is computed and thrown away prior to the 8148 evaluation of the second operand. The value of the entire 8149 expression is the value of the second operand. 8150 8151 `COND_EXPR' 8152 These nodes represent `?:' expressions. The first operand is of 8153 boolean or integral type. If it evaluates to a nonzero value, the 8154 second operand should be evaluated, and returned as the value of 8155 the expression. Otherwise, the third operand is evaluated, and 8156 returned as the value of the expression. 8157 8158 The second operand must have the same type as the entire 8159 expression, unless it unconditionally throws an exception or calls 8160 a noreturn function, in which case it should have void type. The 8161 same constraints apply to the third operand. This allows array 8162 bounds checks to be represented conveniently as `(i >= 0 && i < 8163 10) ? i : abort()'. 8164 8165 As a GNU extension, the C language front-ends allow the second 8166 operand of the `?:' operator may be omitted in the source. For 8167 example, `x ? : 3' is equivalent to `x ? x : 3', assuming that `x' 8168 is an expression without side-effects. In the tree 8169 representation, however, the second operand is always present, 8170 possibly protected by `SAVE_EXPR' if the first argument does cause 8171 side-effects. 8172 8173 `CALL_EXPR' 8174 These nodes are used to represent calls to functions, including 8175 non-static member functions. `CALL_EXPR's are implemented as 8176 expression nodes with a variable number of operands. Rather than 8177 using `TREE_OPERAND' to extract them, it is preferable to use the 8178 specialized accessor macros and functions that operate 8179 specifically on `CALL_EXPR' nodes. 8180 8181 `CALL_EXPR_FN' returns a pointer to the function to call; it is 8182 always an expression whose type is a `POINTER_TYPE'. 8183 8184 The number of arguments to the call is returned by 8185 `call_expr_nargs', while the arguments themselves can be accessed 8186 with the `CALL_EXPR_ARG' macro. The arguments are zero-indexed 8187 and numbered left-to-right. You can iterate over the arguments 8188 using `FOR_EACH_CALL_EXPR_ARG', as in: 8189 8190 tree call, arg; 8191 call_expr_arg_iterator iter; 8192 FOR_EACH_CALL_EXPR_ARG (arg, iter, call) 8193 /* arg is bound to successive arguments of call. */ 8194 ...; 8195 8196 For non-static member functions, there will be an operand 8197 corresponding to the `this' pointer. There will always be 8198 expressions corresponding to all of the arguments, even if the 8199 function is declared with default arguments and some arguments are 8200 not explicitly provided at the call sites. 8201 8202 `CALL_EXPR's also have a `CALL_EXPR_STATIC_CHAIN' operand that is 8203 used to implement nested functions. This operand is otherwise 8204 null. 8205 8206 `STMT_EXPR' 8207 These nodes are used to represent GCC's statement-expression 8208 extension. The statement-expression extension allows code like 8209 this: 8210 int f() { return ({ int j; j = 3; j + 7; }); } 8211 In other words, an sequence of statements may occur where a single 8212 expression would normally appear. The `STMT_EXPR' node represents 8213 such an expression. The `STMT_EXPR_STMT' gives the statement 8214 contained in the expression. The value of the expression is the 8215 value of the last sub-statement in the body. More precisely, the 8216 value is the value computed by the last statement nested inside 8217 `BIND_EXPR', `TRY_FINALLY_EXPR', or `TRY_CATCH_EXPR'. For 8218 example, in: 8219 ({ 3; }) 8220 the value is `3' while in: 8221 ({ if (x) { 3; } }) 8222 there is no value. If the `STMT_EXPR' does not yield a value, 8223 it's type will be `void'. 8224 8225 `BIND_EXPR' 8226 These nodes represent local blocks. The first operand is a list of 8227 variables, connected via their `TREE_CHAIN' field. These will 8228 never require cleanups. The scope of these variables is just the 8229 body of the `BIND_EXPR'. The body of the `BIND_EXPR' is the 8230 second operand. 8231 8232 `LOOP_EXPR' 8233 These nodes represent "infinite" loops. The `LOOP_EXPR_BODY' 8234 represents the body of the loop. It should be executed forever, 8235 unless an `EXIT_EXPR' is encountered. 8236 8237 `EXIT_EXPR' 8238 These nodes represent conditional exits from the nearest enclosing 8239 `LOOP_EXPR'. The single operand is the condition; if it is 8240 nonzero, then the loop should be exited. An `EXIT_EXPR' will only 8241 appear within a `LOOP_EXPR'. 8242 8243 `CLEANUP_POINT_EXPR' 8244 These nodes represent full-expressions. The single operand is an 8245 expression to evaluate. Any destructor calls engendered by the 8246 creation of temporaries during the evaluation of that expression 8247 should be performed immediately after the expression is evaluated. 8248 8249 `CONSTRUCTOR' 8250 These nodes represent the brace-enclosed initializers for a 8251 structure or array. The first operand is reserved for use by the 8252 back end. The second operand is a `TREE_LIST'. If the 8253 `TREE_TYPE' of the `CONSTRUCTOR' is a `RECORD_TYPE' or 8254 `UNION_TYPE', then the `TREE_PURPOSE' of each node in the 8255 `TREE_LIST' will be a `FIELD_DECL' and the `TREE_VALUE' of each 8256 node will be the expression used to initialize that field. 8257 8258 If the `TREE_TYPE' of the `CONSTRUCTOR' is an `ARRAY_TYPE', then 8259 the `TREE_PURPOSE' of each element in the `TREE_LIST' will be an 8260 `INTEGER_CST' or a `RANGE_EXPR' of two `INTEGER_CST's. A single 8261 `INTEGER_CST' indicates which element of the array (indexed from 8262 zero) is being assigned to. A `RANGE_EXPR' indicates an inclusive 8263 range of elements to initialize. In both cases the `TREE_VALUE' 8264 is the corresponding initializer. It is re-evaluated for each 8265 element of a `RANGE_EXPR'. If the `TREE_PURPOSE' is `NULL_TREE', 8266 then the initializer is for the next available array element. 8267 8268 In the front end, you should not depend on the fields appearing in 8269 any particular order. However, in the middle end, fields must 8270 appear in declaration order. You should not assume that all 8271 fields will be represented. Unrepresented fields will be set to 8272 zero. 8273 8274 `COMPOUND_LITERAL_EXPR' 8275 These nodes represent ISO C99 compound literals. The 8276 `COMPOUND_LITERAL_EXPR_DECL_STMT' is a `DECL_STMT' containing an 8277 anonymous `VAR_DECL' for the unnamed object represented by the 8278 compound literal; the `DECL_INITIAL' of that `VAR_DECL' is a 8279 `CONSTRUCTOR' representing the brace-enclosed list of initializers 8280 in the compound literal. That anonymous `VAR_DECL' can also be 8281 accessed directly by the `COMPOUND_LITERAL_EXPR_DECL' macro. 8282 8283 `SAVE_EXPR' 8284 A `SAVE_EXPR' represents an expression (possibly involving 8285 side-effects) that is used more than once. The side-effects should 8286 occur only the first time the expression is evaluated. Subsequent 8287 uses should just reuse the computed value. The first operand to 8288 the `SAVE_EXPR' is the expression to evaluate. The side-effects 8289 should be executed where the `SAVE_EXPR' is first encountered in a 8290 depth-first preorder traversal of the expression tree. 8291 8292 `TARGET_EXPR' 8293 A `TARGET_EXPR' represents a temporary object. The first operand 8294 is a `VAR_DECL' for the temporary variable. The second operand is 8295 the initializer for the temporary. The initializer is evaluated 8296 and, if non-void, copied (bitwise) into the temporary. If the 8297 initializer is void, that means that it will perform the 8298 initialization itself. 8299 8300 Often, a `TARGET_EXPR' occurs on the right-hand side of an 8301 assignment, or as the second operand to a comma-expression which is 8302 itself the right-hand side of an assignment, etc. In this case, 8303 we say that the `TARGET_EXPR' is "normal"; otherwise, we say it is 8304 "orphaned". For a normal `TARGET_EXPR' the temporary variable 8305 should be treated as an alias for the left-hand side of the 8306 assignment, rather than as a new temporary variable. 8307 8308 The third operand to the `TARGET_EXPR', if present, is a 8309 cleanup-expression (i.e., destructor call) for the temporary. If 8310 this expression is orphaned, then this expression must be executed 8311 when the statement containing this expression is complete. These 8312 cleanups must always be executed in the order opposite to that in 8313 which they were encountered. Note that if a temporary is created 8314 on one branch of a conditional operator (i.e., in the second or 8315 third operand to a `COND_EXPR'), the cleanup must be run only if 8316 that branch is actually executed. 8317 8318 See `STMT_IS_FULL_EXPR_P' for more information about running these 8319 cleanups. 8320 8321 `AGGR_INIT_EXPR' 8322 An `AGGR_INIT_EXPR' represents the initialization as the return 8323 value of a function call, or as the result of a constructor. An 8324 `AGGR_INIT_EXPR' will only appear as a full-expression, or as the 8325 second operand of a `TARGET_EXPR'. `AGGR_INIT_EXPR's have a 8326 representation similar to that of `CALL_EXPR's. You can use the 8327 `AGGR_INIT_EXPR_FN' and `AGGR_INIT_EXPR_ARG' macros to access the 8328 function to call and the arguments to pass. 8329 8330 If `AGGR_INIT_VIA_CTOR_P' holds of the `AGGR_INIT_EXPR', then the 8331 initialization is via a constructor call. The address of the 8332 `AGGR_INIT_EXPR_SLOT' operand, which is always a `VAR_DECL', is 8333 taken, and this value replaces the first argument in the argument 8334 list. 8335 8336 In either case, the expression is void. 8337 8338 `VA_ARG_EXPR' 8339 This node is used to implement support for the C/C++ variable 8340 argument-list mechanism. It represents expressions like `va_arg 8341 (ap, type)'. Its `TREE_TYPE' yields the tree representation for 8342 `type' and its sole argument yields the representation for `ap'. 8343 8344 `CHANGE_DYNAMIC_TYPE_EXPR' 8345 Indicates the special aliasing required by C++ placement new. It 8346 has two operands: a type and a location. It means that the 8347 dynamic type of the location is changing to be the specified type. 8348 The alias analysis code takes this into account when doing type 8349 based alias analysis. 8350 8351 `OMP_PARALLEL' 8352 Represents `#pragma omp parallel [clause1 ... clauseN]'. It has 8353 four operands: 8354 8355 Operand `OMP_PARALLEL_BODY' is valid while in GENERIC and High 8356 GIMPLE forms. It contains the body of code to be executed by all 8357 the threads. During GIMPLE lowering, this operand becomes `NULL' 8358 and the body is emitted linearly after `OMP_PARALLEL'. 8359 8360 Operand `OMP_PARALLEL_CLAUSES' is the list of clauses associated 8361 with the directive. 8362 8363 Operand `OMP_PARALLEL_FN' is created by `pass_lower_omp', it 8364 contains the `FUNCTION_DECL' for the function that will contain 8365 the body of the parallel region. 8366 8367 Operand `OMP_PARALLEL_DATA_ARG' is also created by 8368 `pass_lower_omp'. If there are shared variables to be communicated 8369 to the children threads, this operand will contain the `VAR_DECL' 8370 that contains all the shared values and variables. 8371 8372 `OMP_FOR' 8373 Represents `#pragma omp for [clause1 ... clauseN]'. It has 5 8374 operands: 8375 8376 Operand `OMP_FOR_BODY' contains the loop body. 8377 8378 Operand `OMP_FOR_CLAUSES' is the list of clauses associated with 8379 the directive. 8380 8381 Operand `OMP_FOR_INIT' is the loop initialization code of the form 8382 `VAR = N1'. 8383 8384 Operand `OMP_FOR_COND' is the loop conditional expression of the 8385 form `VAR {<,>,<=,>=} N2'. 8386 8387 Operand `OMP_FOR_INCR' is the loop index increment of the form 8388 `VAR {+=,-=} INCR'. 8389 8390 Operand `OMP_FOR_PRE_BODY' contains side-effect code from operands 8391 `OMP_FOR_INIT', `OMP_FOR_COND' and `OMP_FOR_INC'. These 8392 side-effects are part of the `OMP_FOR' block but must be evaluated 8393 before the start of loop body. 8394 8395 The loop index variable `VAR' must be a signed integer variable, 8396 which is implicitly private to each thread. Bounds `N1' and `N2' 8397 and the increment expression `INCR' are required to be loop 8398 invariant integer expressions that are evaluated without any 8399 synchronization. The evaluation order, frequency of evaluation and 8400 side-effects are unspecified by the standard. 8401 8402 `OMP_SECTIONS' 8403 Represents `#pragma omp sections [clause1 ... clauseN]'. 8404 8405 Operand `OMP_SECTIONS_BODY' contains the sections body, which in 8406 turn contains a set of `OMP_SECTION' nodes for each of the 8407 concurrent sections delimited by `#pragma omp section'. 8408 8409 Operand `OMP_SECTIONS_CLAUSES' is the list of clauses associated 8410 with the directive. 8411 8412 `OMP_SECTION' 8413 Section delimiter for `OMP_SECTIONS'. 8414 8415 `OMP_SINGLE' 8416 Represents `#pragma omp single'. 8417 8418 Operand `OMP_SINGLE_BODY' contains the body of code to be executed 8419 by a single thread. 8420 8421 Operand `OMP_SINGLE_CLAUSES' is the list of clauses associated 8422 with the directive. 8423 8424 `OMP_MASTER' 8425 Represents `#pragma omp master'. 8426 8427 Operand `OMP_MASTER_BODY' contains the body of code to be executed 8428 by the master thread. 8429 8430 `OMP_ORDERED' 8431 Represents `#pragma omp ordered'. 8432 8433 Operand `OMP_ORDERED_BODY' contains the body of code to be 8434 executed in the sequential order dictated by the loop index 8435 variable. 8436 8437 `OMP_CRITICAL' 8438 Represents `#pragma omp critical [name]'. 8439 8440 Operand `OMP_CRITICAL_BODY' is the critical section. 8441 8442 Operand `OMP_CRITICAL_NAME' is an optional identifier to label the 8443 critical section. 8444 8445 `OMP_RETURN' 8446 This does not represent any OpenMP directive, it is an artificial 8447 marker to indicate the end of the body of an OpenMP. It is used by 8448 the flow graph (`tree-cfg.c') and OpenMP region building code 8449 (`omp-low.c'). 8450 8451 `OMP_CONTINUE' 8452 Similarly, this instruction does not represent an OpenMP 8453 directive, it is used by `OMP_FOR' and `OMP_SECTIONS' to mark the 8454 place where the code needs to loop to the next iteration (in the 8455 case of `OMP_FOR') or the next section (in the case of 8456 `OMP_SECTIONS'). 8457 8458 In some cases, `OMP_CONTINUE' is placed right before `OMP_RETURN'. 8459 But if there are cleanups that need to occur right after the 8460 looping body, it will be emitted between `OMP_CONTINUE' and 8461 `OMP_RETURN'. 8462 8463 `OMP_ATOMIC' 8464 Represents `#pragma omp atomic'. 8465 8466 Operand 0 is the address at which the atomic operation is to be 8467 performed. 8468 8469 Operand 1 is the expression to evaluate. The gimplifier tries 8470 three alternative code generation strategies. Whenever possible, 8471 an atomic update built-in is used. If that fails, a 8472 compare-and-swap loop is attempted. If that also fails, a regular 8473 critical section around the expression is used. 8474 8475 `OMP_CLAUSE' 8476 Represents clauses associated with one of the `OMP_' directives. 8477 Clauses are represented by separate sub-codes defined in `tree.h'. 8478 Clauses codes can be one of: `OMP_CLAUSE_PRIVATE', 8479 `OMP_CLAUSE_SHARED', `OMP_CLAUSE_FIRSTPRIVATE', 8480 `OMP_CLAUSE_LASTPRIVATE', `OMP_CLAUSE_COPYIN', 8481 `OMP_CLAUSE_COPYPRIVATE', `OMP_CLAUSE_IF', 8482 `OMP_CLAUSE_NUM_THREADS', `OMP_CLAUSE_SCHEDULE', 8483 `OMP_CLAUSE_NOWAIT', `OMP_CLAUSE_ORDERED', `OMP_CLAUSE_DEFAULT', 8484 and `OMP_CLAUSE_REDUCTION'. Each code represents the 8485 corresponding OpenMP clause. 8486 8487 Clauses associated with the same directive are chained together 8488 via `OMP_CLAUSE_CHAIN'. Those clauses that accept a list of 8489 variables are restricted to exactly one, accessed with 8490 `OMP_CLAUSE_VAR'. Therefore, multiple variables under the same 8491 clause `C' need to be represented as multiple `C' clauses chained 8492 together. This facilitates adding new clauses during compilation. 8493 8494 `VEC_LSHIFT_EXPR' 8495 8496 `VEC_RSHIFT_EXPR' 8497 These nodes represent whole vector left and right shifts, 8498 respectively. The first operand is the vector to shift; it will 8499 always be of vector type. The second operand is an expression for 8500 the number of bits by which to shift. Note that the result is 8501 undefined if the second operand is larger than or equal to the 8502 first operand's type size. 8503 8504 `VEC_WIDEN_MULT_HI_EXPR' 8505 8506 `VEC_WIDEN_MULT_LO_EXPR' 8507 These nodes represent widening vector multiplication of the high 8508 and low parts of the two input vectors, respectively. Their 8509 operands are vectors that contain the same number of elements 8510 (`N') of the same integral type. The result is a vector that 8511 contains half as many elements, of an integral type whose size is 8512 twice as wide. In the case of `VEC_WIDEN_MULT_HI_EXPR' the high 8513 `N/2' elements of the two vector are multiplied to produce the 8514 vector of `N/2' products. In the case of `VEC_WIDEN_MULT_LO_EXPR' 8515 the low `N/2' elements of the two vector are multiplied to produce 8516 the vector of `N/2' products. 8517 8518 `VEC_UNPACK_HI_EXPR' 8519 8520 `VEC_UNPACK_LO_EXPR' 8521 These nodes represent unpacking of the high and low parts of the 8522 input vector, respectively. The single operand is a vector that 8523 contains `N' elements of the same integral or floating point type. 8524 The result is a vector that contains half as many elements, of an 8525 integral or floating point type whose size is twice as wide. In 8526 the case of `VEC_UNPACK_HI_EXPR' the high `N/2' elements of the 8527 vector are extracted and widened (promoted). In the case of 8528 `VEC_UNPACK_LO_EXPR' the low `N/2' elements of the vector are 8529 extracted and widened (promoted). 8530 8531 `VEC_UNPACK_FLOAT_HI_EXPR' 8532 8533 `VEC_UNPACK_FLOAT_LO_EXPR' 8534 These nodes represent unpacking of the high and low parts of the 8535 input vector, where the values are converted from fixed point to 8536 floating point. The single operand is a vector that contains `N' 8537 elements of the same integral type. The result is a vector that 8538 contains half as many elements of a floating point type whose size 8539 is twice as wide. In the case of `VEC_UNPACK_HI_EXPR' the high 8540 `N/2' elements of the vector are extracted, converted and widened. 8541 In the case of `VEC_UNPACK_LO_EXPR' the low `N/2' elements of the 8542 vector are extracted, converted and widened. 8543 8544 `VEC_PACK_TRUNC_EXPR' 8545 This node represents packing of truncated elements of the two 8546 input vectors into the output vector. Input operands are vectors 8547 that contain the same number of elements of the same integral or 8548 floating point type. The result is a vector that contains twice 8549 as many elements of an integral or floating point type whose size 8550 is half as wide. The elements of the two vectors are demoted and 8551 merged (concatenated) to form the output vector. 8552 8553 `VEC_PACK_SAT_EXPR' 8554 This node represents packing of elements of the two input vectors 8555 into the output vector using saturation. Input operands are 8556 vectors that contain the same number of elements of the same 8557 integral type. The result is a vector that contains twice as many 8558 elements of an integral type whose size is half as wide. The 8559 elements of the two vectors are demoted and merged (concatenated) 8560 to form the output vector. 8561 8562 `VEC_PACK_FIX_TRUNC_EXPR' 8563 This node represents packing of elements of the two input vectors 8564 into the output vector, where the values are converted from 8565 floating point to fixed point. Input operands are vectors that 8566 contain the same number of elements of a floating point type. The 8567 result is a vector that contains twice as many elements of an 8568 integral type whose size is half as wide. The elements of the two 8569 vectors are merged (concatenated) to form the output vector. 8570 8571 `VEC_EXTRACT_EVEN_EXPR' 8572 8573 `VEC_EXTRACT_ODD_EXPR' 8574 These nodes represent extracting of the even/odd elements of the 8575 two input vectors, respectively. Their operands and result are 8576 vectors that contain the same number of elements of the same type. 8577 8578 `VEC_INTERLEAVE_HIGH_EXPR' 8579 8580 `VEC_INTERLEAVE_LOW_EXPR' 8581 These nodes represent merging and interleaving of the high/low 8582 elements of the two input vectors, respectively. The operands and 8583 the result are vectors that contain the same number of elements 8584 (`N') of the same type. In the case of 8585 `VEC_INTERLEAVE_HIGH_EXPR', the high `N/2' elements of the first 8586 input vector are interleaved with the high `N/2' elements of the 8587 second input vector. In the case of `VEC_INTERLEAVE_LOW_EXPR', the 8588 low `N/2' elements of the first input vector are interleaved with 8589 the low `N/2' elements of the second input vector. 8590 8591 8592 8593 File: gccint.info, Node: RTL, Next: Control Flow, Prev: Tree SSA, Up: Top 8594 8595 10 RTL Representation 8596 ********************* 8597 8598 The last part of the compiler work is done on a low-level intermediate 8599 representation called Register Transfer Language. In this language, the 8600 instructions to be output are described, pretty much one by one, in an 8601 algebraic form that describes what the instruction does. 8602 8603 RTL is inspired by Lisp lists. It has both an internal form, made up 8604 of structures that point at other structures, and a textual form that 8605 is used in the machine description and in printed debugging dumps. The 8606 textual form uses nested parentheses to indicate the pointers in the 8607 internal form. 8608 8609 * Menu: 8610 8611 * RTL Objects:: Expressions vs vectors vs strings vs integers. 8612 * RTL Classes:: Categories of RTL expression objects, and their structure. 8613 * Accessors:: Macros to access expression operands or vector elts. 8614 * Special Accessors:: Macros to access specific annotations on RTL. 8615 * Flags:: Other flags in an RTL expression. 8616 * Machine Modes:: Describing the size and format of a datum. 8617 * Constants:: Expressions with constant values. 8618 * Regs and Memory:: Expressions representing register contents or memory. 8619 * Arithmetic:: Expressions representing arithmetic on other expressions. 8620 * Comparisons:: Expressions representing comparison of expressions. 8621 * Bit-Fields:: Expressions representing bit-fields in memory or reg. 8622 * Vector Operations:: Expressions involving vector datatypes. 8623 * Conversions:: Extending, truncating, floating or fixing. 8624 * RTL Declarations:: Declaring volatility, constancy, etc. 8625 * Side Effects:: Expressions for storing in registers, etc. 8626 * Incdec:: Embedded side-effects for autoincrement addressing. 8627 * Assembler:: Representing `asm' with operands. 8628 * Insns:: Expression types for entire insns. 8629 * Calls:: RTL representation of function call insns. 8630 * Sharing:: Some expressions are unique; others *must* be copied. 8631 * Reading RTL:: Reading textual RTL from a file. 8632 8633 8634 File: gccint.info, Node: RTL Objects, Next: RTL Classes, Up: RTL 8635 8636 10.1 RTL Object Types 8637 ===================== 8638 8639 RTL uses five kinds of objects: expressions, integers, wide integers, 8640 strings and vectors. Expressions are the most important ones. An RTL 8641 expression ("RTX", for short) is a C structure, but it is usually 8642 referred to with a pointer; a type that is given the typedef name `rtx'. 8643 8644 An integer is simply an `int'; their written form uses decimal digits. 8645 A wide integer is an integral object whose type is `HOST_WIDE_INT'; 8646 their written form uses decimal digits. 8647 8648 A string is a sequence of characters. In core it is represented as a 8649 `char *' in usual C fashion, and it is written in C syntax as well. 8650 However, strings in RTL may never be null. If you write an empty 8651 string in a machine description, it is represented in core as a null 8652 pointer rather than as a pointer to a null character. In certain 8653 contexts, these null pointers instead of strings are valid. Within RTL 8654 code, strings are most commonly found inside `symbol_ref' expressions, 8655 but they appear in other contexts in the RTL expressions that make up 8656 machine descriptions. 8657 8658 In a machine description, strings are normally written with double 8659 quotes, as you would in C. However, strings in machine descriptions may 8660 extend over many lines, which is invalid C, and adjacent string 8661 constants are not concatenated as they are in C. Any string constant 8662 may be surrounded with a single set of parentheses. Sometimes this 8663 makes the machine description easier to read. 8664 8665 There is also a special syntax for strings, which can be useful when C 8666 code is embedded in a machine description. Wherever a string can 8667 appear, it is also valid to write a C-style brace block. The entire 8668 brace block, including the outermost pair of braces, is considered to be 8669 the string constant. Double quote characters inside the braces are not 8670 special. Therefore, if you write string constants in the C code, you 8671 need not escape each quote character with a backslash. 8672 8673 A vector contains an arbitrary number of pointers to expressions. The 8674 number of elements in the vector is explicitly present in the vector. 8675 The written form of a vector consists of square brackets (`[...]') 8676 surrounding the elements, in sequence and with whitespace separating 8677 them. Vectors of length zero are not created; null pointers are used 8678 instead. 8679 8680 Expressions are classified by "expression codes" (also called RTX 8681 codes). The expression code is a name defined in `rtl.def', which is 8682 also (in uppercase) a C enumeration constant. The possible expression 8683 codes and their meanings are machine-independent. The code of an RTX 8684 can be extracted with the macro `GET_CODE (X)' and altered with 8685 `PUT_CODE (X, NEWCODE)'. 8686 8687 The expression code determines how many operands the expression 8688 contains, and what kinds of objects they are. In RTL, unlike Lisp, you 8689 cannot tell by looking at an operand what kind of object it is. 8690 Instead, you must know from its context--from the expression code of 8691 the containing expression. For example, in an expression of code 8692 `subreg', the first operand is to be regarded as an expression and the 8693 second operand as an integer. In an expression of code `plus', there 8694 are two operands, both of which are to be regarded as expressions. In 8695 a `symbol_ref' expression, there is one operand, which is to be 8696 regarded as a string. 8697 8698 Expressions are written as parentheses containing the name of the 8699 expression type, its flags and machine mode if any, and then the 8700 operands of the expression (separated by spaces). 8701 8702 Expression code names in the `md' file are written in lowercase, but 8703 when they appear in C code they are written in uppercase. In this 8704 manual, they are shown as follows: `const_int'. 8705 8706 In a few contexts a null pointer is valid where an expression is 8707 normally wanted. The written form of this is `(nil)'. 8708 8709 8710 File: gccint.info, Node: RTL Classes, Next: Accessors, Prev: RTL Objects, Up: RTL 8711 8712 10.2 RTL Classes and Formats 8713 ============================ 8714 8715 The various expression codes are divided into several "classes", which 8716 are represented by single characters. You can determine the class of 8717 an RTX code with the macro `GET_RTX_CLASS (CODE)'. Currently, 8718 `rtl.def' defines these classes: 8719 8720 `RTX_OBJ' 8721 An RTX code that represents an actual object, such as a register 8722 (`REG') or a memory location (`MEM', `SYMBOL_REF'). `LO_SUM') is 8723 also included; instead, `SUBREG' and `STRICT_LOW_PART' are not in 8724 this class, but in class `x'. 8725 8726 `RTX_CONST_OBJ' 8727 An RTX code that represents a constant object. `HIGH' is also 8728 included in this class. 8729 8730 `RTX_COMPARE' 8731 An RTX code for a non-symmetric comparison, such as `GEU' or `LT'. 8732 8733 `RTX_COMM_COMPARE' 8734 An RTX code for a symmetric (commutative) comparison, such as `EQ' 8735 or `ORDERED'. 8736 8737 `RTX_UNARY' 8738 An RTX code for a unary arithmetic operation, such as `NEG', 8739 `NOT', or `ABS'. This category also includes value extension 8740 (sign or zero) and conversions between integer and floating point. 8741 8742 `RTX_COMM_ARITH' 8743 An RTX code for a commutative binary operation, such as `PLUS' or 8744 `AND'. `NE' and `EQ' are comparisons, so they have class `<'. 8745 8746 `RTX_BIN_ARITH' 8747 An RTX code for a non-commutative binary operation, such as 8748 `MINUS', `DIV', or `ASHIFTRT'. 8749 8750 `RTX_BITFIELD_OPS' 8751 An RTX code for a bit-field operation. Currently only 8752 `ZERO_EXTRACT' and `SIGN_EXTRACT'. These have three inputs and 8753 are lvalues (so they can be used for insertion as well). *Note 8754 Bit-Fields::. 8755 8756 `RTX_TERNARY' 8757 An RTX code for other three input operations. Currently only 8758 `IF_THEN_ELSE' and `VEC_MERGE'. 8759 8760 `RTX_INSN' 8761 An RTX code for an entire instruction: `INSN', `JUMP_INSN', and 8762 `CALL_INSN'. *Note Insns::. 8763 8764 `RTX_MATCH' 8765 An RTX code for something that matches in insns, such as 8766 `MATCH_DUP'. These only occur in machine descriptions. 8767 8768 `RTX_AUTOINC' 8769 An RTX code for an auto-increment addressing mode, such as 8770 `POST_INC'. 8771 8772 `RTX_EXTRA' 8773 All other RTX codes. This category includes the remaining codes 8774 used only in machine descriptions (`DEFINE_*', etc.). It also 8775 includes all the codes describing side effects (`SET', `USE', 8776 `CLOBBER', etc.) and the non-insns that may appear on an insn 8777 chain, such as `NOTE', `BARRIER', and `CODE_LABEL'. `SUBREG' is 8778 also part of this class. 8779 8780 For each expression code, `rtl.def' specifies the number of contained 8781 objects and their kinds using a sequence of characters called the 8782 "format" of the expression code. For example, the format of `subreg' 8783 is `ei'. 8784 8785 These are the most commonly used format characters: 8786 8787 `e' 8788 An expression (actually a pointer to an expression). 8789 8790 `i' 8791 An integer. 8792 8793 `w' 8794 A wide integer. 8795 8796 `s' 8797 A string. 8798 8799 `E' 8800 A vector of expressions. 8801 8802 A few other format characters are used occasionally: 8803 8804 `u' 8805 `u' is equivalent to `e' except that it is printed differently in 8806 debugging dumps. It is used for pointers to insns. 8807 8808 `n' 8809 `n' is equivalent to `i' except that it is printed differently in 8810 debugging dumps. It is used for the line number or code number of 8811 a `note' insn. 8812 8813 `S' 8814 `S' indicates a string which is optional. In the RTL objects in 8815 core, `S' is equivalent to `s', but when the object is read, from 8816 an `md' file, the string value of this operand may be omitted. An 8817 omitted string is taken to be the null string. 8818 8819 `V' 8820 `V' indicates a vector which is optional. In the RTL objects in 8821 core, `V' is equivalent to `E', but when the object is read from 8822 an `md' file, the vector value of this operand may be omitted. An 8823 omitted vector is effectively the same as a vector of no elements. 8824 8825 `B' 8826 `B' indicates a pointer to basic block structure. 8827 8828 `0' 8829 `0' means a slot whose contents do not fit any normal category. 8830 `0' slots are not printed at all in dumps, and are often used in 8831 special ways by small parts of the compiler. 8832 8833 There are macros to get the number of operands and the format of an 8834 expression code: 8835 8836 `GET_RTX_LENGTH (CODE)' 8837 Number of operands of an RTX of code CODE. 8838 8839 `GET_RTX_FORMAT (CODE)' 8840 The format of an RTX of code CODE, as a C string. 8841 8842 Some classes of RTX codes always have the same format. For example, it 8843 is safe to assume that all comparison operations have format `ee'. 8844 8845 `1' 8846 All codes of this class have format `e'. 8847 8848 `<' 8849 `c' 8850 `2' 8851 All codes of these classes have format `ee'. 8852 8853 `b' 8854 `3' 8855 All codes of these classes have format `eee'. 8856 8857 `i' 8858 All codes of this class have formats that begin with `iuueiee'. 8859 *Note Insns::. Note that not all RTL objects linked onto an insn 8860 chain are of class `i'. 8861 8862 `o' 8863 `m' 8864 `x' 8865 You can make no assumptions about the format of these codes. 8866 8867 8868 File: gccint.info, Node: Accessors, Next: Special Accessors, Prev: RTL Classes, Up: RTL 8869 8870 10.3 Access to Operands 8871 ======================= 8872 8873 Operands of expressions are accessed using the macros `XEXP', `XINT', 8874 `XWINT' and `XSTR'. Each of these macros takes two arguments: an 8875 expression-pointer (RTX) and an operand number (counting from zero). 8876 Thus, 8877 8878 XEXP (X, 2) 8879 8880 accesses operand 2 of expression X, as an expression. 8881 8882 XINT (X, 2) 8883 8884 accesses the same operand as an integer. `XSTR', used in the same 8885 fashion, would access it as a string. 8886 8887 Any operand can be accessed as an integer, as an expression or as a 8888 string. You must choose the correct method of access for the kind of 8889 value actually stored in the operand. You would do this based on the 8890 expression code of the containing expression. That is also how you 8891 would know how many operands there are. 8892 8893 For example, if X is a `subreg' expression, you know that it has two 8894 operands which can be correctly accessed as `XEXP (X, 0)' and `XINT (X, 8895 1)'. If you did `XINT (X, 0)', you would get the address of the 8896 expression operand but cast as an integer; that might occasionally be 8897 useful, but it would be cleaner to write `(int) XEXP (X, 0)'. `XEXP 8898 (X, 1)' would also compile without error, and would return the second, 8899 integer operand cast as an expression pointer, which would probably 8900 result in a crash when accessed. Nothing stops you from writing `XEXP 8901 (X, 28)' either, but this will access memory past the end of the 8902 expression with unpredictable results. 8903 8904 Access to operands which are vectors is more complicated. You can use 8905 the macro `XVEC' to get the vector-pointer itself, or the macros 8906 `XVECEXP' and `XVECLEN' to access the elements and length of a vector. 8907 8908 `XVEC (EXP, IDX)' 8909 Access the vector-pointer which is operand number IDX in EXP. 8910 8911 `XVECLEN (EXP, IDX)' 8912 Access the length (number of elements) in the vector which is in 8913 operand number IDX in EXP. This value is an `int'. 8914 8915 `XVECEXP (EXP, IDX, ELTNUM)' 8916 Access element number ELTNUM in the vector which is in operand 8917 number IDX in EXP. This value is an RTX. 8918 8919 It is up to you to make sure that ELTNUM is not negative and is 8920 less than `XVECLEN (EXP, IDX)'. 8921 8922 All the macros defined in this section expand into lvalues and 8923 therefore can be used to assign the operands, lengths and vector 8924 elements as well as to access them. 8925 8926 8927 File: gccint.info, Node: Special Accessors, Next: Flags, Prev: Accessors, Up: RTL 8928 8929 10.4 Access to Special Operands 8930 =============================== 8931 8932 Some RTL nodes have special annotations associated with them. 8933 8934 `MEM' 8935 8936 `MEM_ALIAS_SET (X)' 8937 If 0, X is not in any alias set, and may alias anything. 8938 Otherwise, X can only alias `MEM's in a conflicting alias 8939 set. This value is set in a language-dependent manner in the 8940 front-end, and should not be altered in the back-end. In 8941 some front-ends, these numbers may correspond in some way to 8942 types, or other language-level entities, but they need not, 8943 and the back-end makes no such assumptions. These set 8944 numbers are tested with `alias_sets_conflict_p'. 8945 8946 `MEM_EXPR (X)' 8947 If this register is known to hold the value of some user-level 8948 declaration, this is that tree node. It may also be a 8949 `COMPONENT_REF', in which case this is some field reference, 8950 and `TREE_OPERAND (X, 0)' contains the declaration, or 8951 another `COMPONENT_REF', or null if there is no compile-time 8952 object associated with the reference. 8953 8954 `MEM_OFFSET (X)' 8955 The offset from the start of `MEM_EXPR' as a `CONST_INT' rtx. 8956 8957 `MEM_SIZE (X)' 8958 The size in bytes of the memory reference as a `CONST_INT' 8959 rtx. This is mostly relevant for `BLKmode' references as 8960 otherwise the size is implied by the mode. 8961 8962 `MEM_ALIGN (X)' 8963 The known alignment in bits of the memory reference. 8964 8965 `REG' 8966 8967 `ORIGINAL_REGNO (X)' 8968 This field holds the number the register "originally" had; 8969 for a pseudo register turned into a hard reg this will hold 8970 the old pseudo register number. 8971 8972 `REG_EXPR (X)' 8973 If this register is known to hold the value of some user-level 8974 declaration, this is that tree node. 8975 8976 `REG_OFFSET (X)' 8977 If this register is known to hold the value of some user-level 8978 declaration, this is the offset into that logical storage. 8979 8980 `SYMBOL_REF' 8981 8982 `SYMBOL_REF_DECL (X)' 8983 If the `symbol_ref' X was created for a `VAR_DECL' or a 8984 `FUNCTION_DECL', that tree is recorded here. If this value is 8985 null, then X was created by back end code generation routines, 8986 and there is no associated front end symbol table entry. 8987 8988 `SYMBOL_REF_DECL' may also point to a tree of class `'c'', 8989 that is, some sort of constant. In this case, the 8990 `symbol_ref' is an entry in the per-file constant pool; 8991 again, there is no associated front end symbol table entry. 8992 8993 `SYMBOL_REF_CONSTANT (X)' 8994 If `CONSTANT_POOL_ADDRESS_P (X)' is true, this is the constant 8995 pool entry for X. It is null otherwise. 8996 8997 `SYMBOL_REF_DATA (X)' 8998 A field of opaque type used to store `SYMBOL_REF_DECL' or 8999 `SYMBOL_REF_CONSTANT'. 9000 9001 `SYMBOL_REF_FLAGS (X)' 9002 In a `symbol_ref', this is used to communicate various 9003 predicates about the symbol. Some of these are common enough 9004 to be computed by common code, some are specific to the 9005 target. The common bits are: 9006 9007 `SYMBOL_FLAG_FUNCTION' 9008 Set if the symbol refers to a function. 9009 9010 `SYMBOL_FLAG_LOCAL' 9011 Set if the symbol is local to this "module". See 9012 `TARGET_BINDS_LOCAL_P'. 9013 9014 `SYMBOL_FLAG_EXTERNAL' 9015 Set if this symbol is not defined in this translation 9016 unit. Note that this is not the inverse of 9017 `SYMBOL_FLAG_LOCAL'. 9018 9019 `SYMBOL_FLAG_SMALL' 9020 Set if the symbol is located in the small data section. 9021 See `TARGET_IN_SMALL_DATA_P'. 9022 9023 `SYMBOL_REF_TLS_MODEL (X)' 9024 This is a multi-bit field accessor that returns the 9025 `tls_model' to be used for a thread-local storage 9026 symbol. It returns zero for non-thread-local symbols. 9027 9028 `SYMBOL_FLAG_HAS_BLOCK_INFO' 9029 Set if the symbol has `SYMBOL_REF_BLOCK' and 9030 `SYMBOL_REF_BLOCK_OFFSET' fields. 9031 9032 `SYMBOL_FLAG_ANCHOR' 9033 Set if the symbol is used as a section anchor. "Section 9034 anchors" are symbols that have a known position within 9035 an `object_block' and that can be used to access nearby 9036 members of that block. They are used to implement 9037 `-fsection-anchors'. 9038 9039 If this flag is set, then `SYMBOL_FLAG_HAS_BLOCK_INFO' 9040 will be too. 9041 9042 Bits beginning with `SYMBOL_FLAG_MACH_DEP' are available for 9043 the target's use. 9044 9045 `SYMBOL_REF_BLOCK (X)' 9046 If `SYMBOL_REF_HAS_BLOCK_INFO_P (X)', this is the `object_block' 9047 structure to which the symbol belongs, or `NULL' if it has not 9048 been assigned a block. 9049 9050 `SYMBOL_REF_BLOCK_OFFSET (X)' 9051 If `SYMBOL_REF_HAS_BLOCK_INFO_P (X)', this is the offset of X from 9052 the first object in `SYMBOL_REF_BLOCK (X)'. The value is negative 9053 if X has not yet been assigned to a block, or it has not been 9054 given an offset within that block. 9055 9056 9057 File: gccint.info, Node: Flags, Next: Machine Modes, Prev: Special Accessors, Up: RTL 9058 9059 10.5 Flags in an RTL Expression 9060 =============================== 9061 9062 RTL expressions contain several flags (one-bit bit-fields) that are 9063 used in certain types of expression. Most often they are accessed with 9064 the following macros, which expand into lvalues. 9065 9066 `CONSTANT_POOL_ADDRESS_P (X)' 9067 Nonzero in a `symbol_ref' if it refers to part of the current 9068 function's constant pool. For most targets these addresses are in 9069 a `.rodata' section entirely separate from the function, but for 9070 some targets the addresses are close to the beginning of the 9071 function. In either case GCC assumes these addresses can be 9072 addressed directly, perhaps with the help of base registers. 9073 Stored in the `unchanging' field and printed as `/u'. 9074 9075 `RTL_CONST_CALL_P (X)' 9076 In a `call_insn' indicates that the insn represents a call to a 9077 const function. Stored in the `unchanging' field and printed as 9078 `/u'. 9079 9080 `RTL_PURE_CALL_P (X)' 9081 In a `call_insn' indicates that the insn represents a call to a 9082 pure function. Stored in the `return_val' field and printed as 9083 `/i'. 9084 9085 `RTL_CONST_OR_PURE_CALL_P (X)' 9086 In a `call_insn', true if `RTL_CONST_CALL_P' or `RTL_PURE_CALL_P' 9087 is true. 9088 9089 `RTL_LOOPING_CONST_OR_PURE_CALL_P (X)' 9090 In a `call_insn' indicates that the insn represents a possibly 9091 infinite looping call to a const or pure function. Stored in the 9092 `call' field and printed as `/c'. Only true if one of 9093 `RTL_CONST_CALL_P' or `RTL_PURE_CALL_P' is true. 9094 9095 `INSN_ANNULLED_BRANCH_P (X)' 9096 In a `jump_insn', `call_insn', or `insn' indicates that the branch 9097 is an annulling one. See the discussion under `sequence' below. 9098 Stored in the `unchanging' field and printed as `/u'. 9099 9100 `INSN_DELETED_P (X)' 9101 In an `insn', `call_insn', `jump_insn', `code_label', `barrier', 9102 or `note', nonzero if the insn has been deleted. Stored in the 9103 `volatil' field and printed as `/v'. 9104 9105 `INSN_FROM_TARGET_P (X)' 9106 In an `insn' or `jump_insn' or `call_insn' in a delay slot of a 9107 branch, indicates that the insn is from the target of the branch. 9108 If the branch insn has `INSN_ANNULLED_BRANCH_P' set, this insn 9109 will only be executed if the branch is taken. For annulled 9110 branches with `INSN_FROM_TARGET_P' clear, the insn will be 9111 executed only if the branch is not taken. When 9112 `INSN_ANNULLED_BRANCH_P' is not set, this insn will always be 9113 executed. Stored in the `in_struct' field and printed as `/s'. 9114 9115 `LABEL_PRESERVE_P (X)' 9116 In a `code_label' or `note', indicates that the label is 9117 referenced by code or data not visible to the RTL of a given 9118 function. Labels referenced by a non-local goto will have this 9119 bit set. Stored in the `in_struct' field and printed as `/s'. 9120 9121 `LABEL_REF_NONLOCAL_P (X)' 9122 In `label_ref' and `reg_label' expressions, nonzero if this is a 9123 reference to a non-local label. Stored in the `volatil' field and 9124 printed as `/v'. 9125 9126 `MEM_IN_STRUCT_P (X)' 9127 In `mem' expressions, nonzero for reference to an entire structure, 9128 union or array, or to a component of one. Zero for references to a 9129 scalar variable or through a pointer to a scalar. If both this 9130 flag and `MEM_SCALAR_P' are clear, then we don't know whether this 9131 `mem' is in a structure or not. Both flags should never be 9132 simultaneously set. Stored in the `in_struct' field and printed 9133 as `/s'. 9134 9135 `MEM_KEEP_ALIAS_SET_P (X)' 9136 In `mem' expressions, 1 if we should keep the alias set for this 9137 mem unchanged when we access a component. Set to 1, for example, 9138 when we are already in a non-addressable component of an aggregate. 9139 Stored in the `jump' field and printed as `/j'. 9140 9141 `MEM_SCALAR_P (X)' 9142 In `mem' expressions, nonzero for reference to a scalar known not 9143 to be a member of a structure, union, or array. Zero for such 9144 references and for indirections through pointers, even pointers 9145 pointing to scalar types. If both this flag and `MEM_IN_STRUCT_P' 9146 are clear, then we don't know whether this `mem' is in a structure 9147 or not. Both flags should never be simultaneously set. Stored in 9148 the `return_val' field and printed as `/i'. 9149 9150 `MEM_VOLATILE_P (X)' 9151 In `mem', `asm_operands', and `asm_input' expressions, nonzero for 9152 volatile memory references. Stored in the `volatil' field and 9153 printed as `/v'. 9154 9155 `MEM_NOTRAP_P (X)' 9156 In `mem', nonzero for memory references that will not trap. 9157 Stored in the `call' field and printed as `/c'. 9158 9159 `MEM_POINTER (X)' 9160 Nonzero in a `mem' if the memory reference holds a pointer. 9161 Stored in the `frame_related' field and printed as `/f'. 9162 9163 `REG_FUNCTION_VALUE_P (X)' 9164 Nonzero in a `reg' if it is the place in which this function's 9165 value is going to be returned. (This happens only in a hard 9166 register.) Stored in the `return_val' field and printed as `/i'. 9167 9168 `REG_POINTER (X)' 9169 Nonzero in a `reg' if the register holds a pointer. Stored in the 9170 `frame_related' field and printed as `/f'. 9171 9172 `REG_USERVAR_P (X)' 9173 In a `reg', nonzero if it corresponds to a variable present in the 9174 user's source code. Zero for temporaries generated internally by 9175 the compiler. Stored in the `volatil' field and printed as `/v'. 9176 9177 The same hard register may be used also for collecting the values 9178 of functions called by this one, but `REG_FUNCTION_VALUE_P' is zero 9179 in this kind of use. 9180 9181 `RTX_FRAME_RELATED_P (X)' 9182 Nonzero in an `insn', `call_insn', `jump_insn', `barrier', or 9183 `set' which is part of a function prologue and sets the stack 9184 pointer, sets the frame pointer, or saves a register. This flag 9185 should also be set on an instruction that sets up a temporary 9186 register to use in place of the frame pointer. Stored in the 9187 `frame_related' field and printed as `/f'. 9188 9189 In particular, on RISC targets where there are limits on the sizes 9190 of immediate constants, it is sometimes impossible to reach the 9191 register save area directly from the stack pointer. In that case, 9192 a temporary register is used that is near enough to the register 9193 save area, and the Canonical Frame Address, i.e., DWARF2's logical 9194 frame pointer, register must (temporarily) be changed to be this 9195 temporary register. So, the instruction that sets this temporary 9196 register must be marked as `RTX_FRAME_RELATED_P'. 9197 9198 If the marked instruction is overly complex (defined in terms of 9199 what `dwarf2out_frame_debug_expr' can handle), you will also have 9200 to create a `REG_FRAME_RELATED_EXPR' note and attach it to the 9201 instruction. This note should contain a simple expression of the 9202 computation performed by this instruction, i.e., one that 9203 `dwarf2out_frame_debug_expr' can handle. 9204 9205 This flag is required for exception handling support on targets 9206 with RTL prologues. 9207 9208 `MEM_READONLY_P (X)' 9209 Nonzero in a `mem', if the memory is statically allocated and 9210 read-only. 9211 9212 Read-only in this context means never modified during the lifetime 9213 of the program, not necessarily in ROM or in write-disabled pages. 9214 A common example of the later is a shared library's global offset 9215 table. This table is initialized by the runtime loader, so the 9216 memory is technically writable, but after control is transfered 9217 from the runtime loader to the application, this memory will never 9218 be subsequently modified. 9219 9220 Stored in the `unchanging' field and printed as `/u'. 9221 9222 `SCHED_GROUP_P (X)' 9223 During instruction scheduling, in an `insn', `call_insn' or 9224 `jump_insn', indicates that the previous insn must be scheduled 9225 together with this insn. This is used to ensure that certain 9226 groups of instructions will not be split up by the instruction 9227 scheduling pass, for example, `use' insns before a `call_insn' may 9228 not be separated from the `call_insn'. Stored in the `in_struct' 9229 field and printed as `/s'. 9230 9231 `SET_IS_RETURN_P (X)' 9232 For a `set', nonzero if it is for a return. Stored in the `jump' 9233 field and printed as `/j'. 9234 9235 `SIBLING_CALL_P (X)' 9236 For a `call_insn', nonzero if the insn is a sibling call. Stored 9237 in the `jump' field and printed as `/j'. 9238 9239 `STRING_POOL_ADDRESS_P (X)' 9240 For a `symbol_ref' expression, nonzero if it addresses this 9241 function's string constant pool. Stored in the `frame_related' 9242 field and printed as `/f'. 9243 9244 `SUBREG_PROMOTED_UNSIGNED_P (X)' 9245 Returns a value greater then zero for a `subreg' that has 9246 `SUBREG_PROMOTED_VAR_P' nonzero if the object being referenced is 9247 kept zero-extended, zero if it is kept sign-extended, and less 9248 then zero if it is extended some other way via the `ptr_extend' 9249 instruction. Stored in the `unchanging' field and `volatil' 9250 field, printed as `/u' and `/v'. This macro may only be used to 9251 get the value it may not be used to change the value. Use 9252 `SUBREG_PROMOTED_UNSIGNED_SET' to change the value. 9253 9254 `SUBREG_PROMOTED_UNSIGNED_SET (X)' 9255 Set the `unchanging' and `volatil' fields in a `subreg' to reflect 9256 zero, sign, or other extension. If `volatil' is zero, then 9257 `unchanging' as nonzero means zero extension and as zero means 9258 sign extension. If `volatil' is nonzero then some other type of 9259 extension was done via the `ptr_extend' instruction. 9260 9261 `SUBREG_PROMOTED_VAR_P (X)' 9262 Nonzero in a `subreg' if it was made when accessing an object that 9263 was promoted to a wider mode in accord with the `PROMOTED_MODE' 9264 machine description macro (*note Storage Layout::). In this case, 9265 the mode of the `subreg' is the declared mode of the object and 9266 the mode of `SUBREG_REG' is the mode of the register that holds 9267 the object. Promoted variables are always either sign- or 9268 zero-extended to the wider mode on every assignment. Stored in 9269 the `in_struct' field and printed as `/s'. 9270 9271 `SYMBOL_REF_USED (X)' 9272 In a `symbol_ref', indicates that X has been used. This is 9273 normally only used to ensure that X is only declared external 9274 once. Stored in the `used' field. 9275 9276 `SYMBOL_REF_WEAK (X)' 9277 In a `symbol_ref', indicates that X has been declared weak. 9278 Stored in the `return_val' field and printed as `/i'. 9279 9280 `SYMBOL_REF_FLAG (X)' 9281 In a `symbol_ref', this is used as a flag for machine-specific 9282 purposes. Stored in the `volatil' field and printed as `/v'. 9283 9284 Most uses of `SYMBOL_REF_FLAG' are historic and may be subsumed by 9285 `SYMBOL_REF_FLAGS'. Certainly use of `SYMBOL_REF_FLAGS' is 9286 mandatory if the target requires more than one bit of storage. 9287 9288 These are the fields to which the above macros refer: 9289 9290 `call' 9291 In a `mem', 1 means that the memory reference will not trap. 9292 9293 In a `call', 1 means that this pure or const call may possibly 9294 infinite loop. 9295 9296 In an RTL dump, this flag is represented as `/c'. 9297 9298 `frame_related' 9299 In an `insn' or `set' expression, 1 means that it is part of a 9300 function prologue and sets the stack pointer, sets the frame 9301 pointer, saves a register, or sets up a temporary register to use 9302 in place of the frame pointer. 9303 9304 In `reg' expressions, 1 means that the register holds a pointer. 9305 9306 In `mem' expressions, 1 means that the memory reference holds a 9307 pointer. 9308 9309 In `symbol_ref' expressions, 1 means that the reference addresses 9310 this function's string constant pool. 9311 9312 In an RTL dump, this flag is represented as `/f'. 9313 9314 `in_struct' 9315 In `mem' expressions, it is 1 if the memory datum referred to is 9316 all or part of a structure or array; 0 if it is (or might be) a 9317 scalar variable. A reference through a C pointer has 0 because 9318 the pointer might point to a scalar variable. This information 9319 allows the compiler to determine something about possible cases of 9320 aliasing. 9321 9322 In `reg' expressions, it is 1 if the register has its entire life 9323 contained within the test expression of some loop. 9324 9325 In `subreg' expressions, 1 means that the `subreg' is accessing an 9326 object that has had its mode promoted from a wider mode. 9327 9328 In `label_ref' expressions, 1 means that the referenced label is 9329 outside the innermost loop containing the insn in which the 9330 `label_ref' was found. 9331 9332 In `code_label' expressions, it is 1 if the label may never be 9333 deleted. This is used for labels which are the target of 9334 non-local gotos. Such a label that would have been deleted is 9335 replaced with a `note' of type `NOTE_INSN_DELETED_LABEL'. 9336 9337 In an `insn' during dead-code elimination, 1 means that the insn is 9338 dead code. 9339 9340 In an `insn' or `jump_insn' during reorg for an insn in the delay 9341 slot of a branch, 1 means that this insn is from the target of the 9342 branch. 9343 9344 In an `insn' during instruction scheduling, 1 means that this insn 9345 must be scheduled as part of a group together with the previous 9346 insn. 9347 9348 In an RTL dump, this flag is represented as `/s'. 9349 9350 `return_val' 9351 In `reg' expressions, 1 means the register contains the value to 9352 be returned by the current function. On machines that pass 9353 parameters in registers, the same register number may be used for 9354 parameters as well, but this flag is not set on such uses. 9355 9356 In `mem' expressions, 1 means the memory reference is to a scalar 9357 known not to be a member of a structure, union, or array. 9358 9359 In `symbol_ref' expressions, 1 means the referenced symbol is weak. 9360 9361 In `call' expressions, 1 means the call is pure. 9362 9363 In an RTL dump, this flag is represented as `/i'. 9364 9365 `jump' 9366 In a `mem' expression, 1 means we should keep the alias set for 9367 this mem unchanged when we access a component. 9368 9369 In a `set', 1 means it is for a return. 9370 9371 In a `call_insn', 1 means it is a sibling call. 9372 9373 In an RTL dump, this flag is represented as `/j'. 9374 9375 `unchanging' 9376 In `reg' and `mem' expressions, 1 means that the value of the 9377 expression never changes. 9378 9379 In `subreg' expressions, it is 1 if the `subreg' references an 9380 unsigned object whose mode has been promoted to a wider mode. 9381 9382 In an `insn' or `jump_insn' in the delay slot of a branch 9383 instruction, 1 means an annulling branch should be used. 9384 9385 In a `symbol_ref' expression, 1 means that this symbol addresses 9386 something in the per-function constant pool. 9387 9388 In a `call_insn' 1 means that this instruction is a call to a const 9389 function. 9390 9391 In an RTL dump, this flag is represented as `/u'. 9392 9393 `used' 9394 This flag is used directly (without an access macro) at the end of 9395 RTL generation for a function, to count the number of times an 9396 expression appears in insns. Expressions that appear more than 9397 once are copied, according to the rules for shared structure 9398 (*note Sharing::). 9399 9400 For a `reg', it is used directly (without an access macro) by the 9401 leaf register renumbering code to ensure that each register is only 9402 renumbered once. 9403 9404 In a `symbol_ref', it indicates that an external declaration for 9405 the symbol has already been written. 9406 9407 `volatil' 9408 In a `mem', `asm_operands', or `asm_input' expression, it is 1 if 9409 the memory reference is volatile. Volatile memory references may 9410 not be deleted, reordered or combined. 9411 9412 In a `symbol_ref' expression, it is used for machine-specific 9413 purposes. 9414 9415 In a `reg' expression, it is 1 if the value is a user-level 9416 variable. 0 indicates an internal compiler temporary. 9417 9418 In an `insn', 1 means the insn has been deleted. 9419 9420 In `label_ref' and `reg_label' expressions, 1 means a reference to 9421 a non-local label. 9422 9423 In an RTL dump, this flag is represented as `/v'. 9424 9425 9426 File: gccint.info, Node: Machine Modes, Next: Constants, Prev: Flags, Up: RTL 9427 9428 10.6 Machine Modes 9429 ================== 9430 9431 A machine mode describes a size of data object and the representation 9432 used for it. In the C code, machine modes are represented by an 9433 enumeration type, `enum machine_mode', defined in `machmode.def'. Each 9434 RTL expression has room for a machine mode and so do certain kinds of 9435 tree expressions (declarations and types, to be precise). 9436 9437 In debugging dumps and machine descriptions, the machine mode of an RTL 9438 expression is written after the expression code with a colon to separate 9439 them. The letters `mode' which appear at the end of each machine mode 9440 name are omitted. For example, `(reg:SI 38)' is a `reg' expression 9441 with machine mode `SImode'. If the mode is `VOIDmode', it is not 9442 written at all. 9443 9444 Here is a table of machine modes. The term "byte" below refers to an 9445 object of `BITS_PER_UNIT' bits (*note Storage Layout::). 9446 9447 `BImode' 9448 "Bit" mode represents a single bit, for predicate registers. 9449 9450 `QImode' 9451 "Quarter-Integer" mode represents a single byte treated as an 9452 integer. 9453 9454 `HImode' 9455 "Half-Integer" mode represents a two-byte integer. 9456 9457 `PSImode' 9458 "Partial Single Integer" mode represents an integer which occupies 9459 four bytes but which doesn't really use all four. On some 9460 machines, this is the right mode to use for pointers. 9461 9462 `SImode' 9463 "Single Integer" mode represents a four-byte integer. 9464 9465 `PDImode' 9466 "Partial Double Integer" mode represents an integer which occupies 9467 eight bytes but which doesn't really use all eight. On some 9468 machines, this is the right mode to use for certain pointers. 9469 9470 `DImode' 9471 "Double Integer" mode represents an eight-byte integer. 9472 9473 `TImode' 9474 "Tetra Integer" (?) mode represents a sixteen-byte integer. 9475 9476 `OImode' 9477 "Octa Integer" (?) mode represents a thirty-two-byte integer. 9478 9479 `QFmode' 9480 "Quarter-Floating" mode represents a quarter-precision (single 9481 byte) floating point number. 9482 9483 `HFmode' 9484 "Half-Floating" mode represents a half-precision (two byte) 9485 floating point number. 9486 9487 `TQFmode' 9488 "Three-Quarter-Floating" (?) mode represents a 9489 three-quarter-precision (three byte) floating point number. 9490 9491 `SFmode' 9492 "Single Floating" mode represents a four byte floating point 9493 number. In the common case, of a processor with IEEE arithmetic 9494 and 8-bit bytes, this is a single-precision IEEE floating point 9495 number; it can also be used for double-precision (on processors 9496 with 16-bit bytes) and single-precision VAX and IBM types. 9497 9498 `DFmode' 9499 "Double Floating" mode represents an eight byte floating point 9500 number. In the common case, of a processor with IEEE arithmetic 9501 and 8-bit bytes, this is a double-precision IEEE floating point 9502 number. 9503 9504 `XFmode' 9505 "Extended Floating" mode represents an IEEE extended floating point 9506 number. This mode only has 80 meaningful bits (ten bytes). Some 9507 processors require such numbers to be padded to twelve bytes, 9508 others to sixteen; this mode is used for either. 9509 9510 `SDmode' 9511 "Single Decimal Floating" mode represents a four byte decimal 9512 floating point number (as distinct from conventional binary 9513 floating point). 9514 9515 `DDmode' 9516 "Double Decimal Floating" mode represents an eight byte decimal 9517 floating point number. 9518 9519 `TDmode' 9520 "Tetra Decimal Floating" mode represents a sixteen byte decimal 9521 floating point number all 128 of whose bits are meaningful. 9522 9523 `TFmode' 9524 "Tetra Floating" mode represents a sixteen byte floating point 9525 number all 128 of whose bits are meaningful. One common use is the 9526 IEEE quad-precision format. 9527 9528 `QQmode' 9529 "Quarter-Fractional" mode represents a single byte treated as a 9530 signed fractional number. The default format is "s.7". 9531 9532 `HQmode' 9533 "Half-Fractional" mode represents a two-byte signed fractional 9534 number. The default format is "s.15". 9535 9536 `SQmode' 9537 "Single Fractional" mode represents a four-byte signed fractional 9538 number. The default format is "s.31". 9539 9540 `DQmode' 9541 "Double Fractional" mode represents an eight-byte signed 9542 fractional number. The default format is "s.63". 9543 9544 `TQmode' 9545 "Tetra Fractional" mode represents a sixteen-byte signed 9546 fractional number. The default format is "s.127". 9547 9548 `UQQmode' 9549 "Unsigned Quarter-Fractional" mode represents a single byte 9550 treated as an unsigned fractional number. The default format is 9551 ".8". 9552 9553 `UHQmode' 9554 "Unsigned Half-Fractional" mode represents a two-byte unsigned 9555 fractional number. The default format is ".16". 9556 9557 `USQmode' 9558 "Unsigned Single Fractional" mode represents a four-byte unsigned 9559 fractional number. The default format is ".32". 9560 9561 `UDQmode' 9562 "Unsigned Double Fractional" mode represents an eight-byte unsigned 9563 fractional number. The default format is ".64". 9564 9565 `UTQmode' 9566 "Unsigned Tetra Fractional" mode represents a sixteen-byte unsigned 9567 fractional number. The default format is ".128". 9568 9569 `HAmode' 9570 "Half-Accumulator" mode represents a two-byte signed accumulator. 9571 The default format is "s8.7". 9572 9573 `SAmode' 9574 "Single Accumulator" mode represents a four-byte signed 9575 accumulator. The default format is "s16.15". 9576 9577 `DAmode' 9578 "Double Accumulator" mode represents an eight-byte signed 9579 accumulator. The default format is "s32.31". 9580 9581 `TAmode' 9582 "Tetra Accumulator" mode represents a sixteen-byte signed 9583 accumulator. The default format is "s64.63". 9584 9585 `UHAmode' 9586 "Unsigned Half-Accumulator" mode represents a two-byte unsigned 9587 accumulator. The default format is "8.8". 9588 9589 `USAmode' 9590 "Unsigned Single Accumulator" mode represents a four-byte unsigned 9591 accumulator. The default format is "16.16". 9592 9593 `UDAmode' 9594 "Unsigned Double Accumulator" mode represents an eight-byte 9595 unsigned accumulator. The default format is "32.32". 9596 9597 `UTAmode' 9598 "Unsigned Tetra Accumulator" mode represents a sixteen-byte 9599 unsigned accumulator. The default format is "64.64". 9600 9601 `CCmode' 9602 "Condition Code" mode represents the value of a condition code, 9603 which is a machine-specific set of bits used to represent the 9604 result of a comparison operation. Other machine-specific modes 9605 may also be used for the condition code. These modes are not used 9606 on machines that use `cc0' (see *note Condition Code::). 9607 9608 `BLKmode' 9609 "Block" mode represents values that are aggregates to which none of 9610 the other modes apply. In RTL, only memory references can have 9611 this mode, and only if they appear in string-move or vector 9612 instructions. On machines which have no such instructions, 9613 `BLKmode' will not appear in RTL. 9614 9615 `VOIDmode' 9616 Void mode means the absence of a mode or an unspecified mode. For 9617 example, RTL expressions of code `const_int' have mode `VOIDmode' 9618 because they can be taken to have whatever mode the context 9619 requires. In debugging dumps of RTL, `VOIDmode' is expressed by 9620 the absence of any mode. 9621 9622 `QCmode, HCmode, SCmode, DCmode, XCmode, TCmode' 9623 These modes stand for a complex number represented as a pair of 9624 floating point values. The floating point values are in `QFmode', 9625 `HFmode', `SFmode', `DFmode', `XFmode', and `TFmode', respectively. 9626 9627 `CQImode, CHImode, CSImode, CDImode, CTImode, COImode' 9628 These modes stand for a complex number represented as a pair of 9629 integer values. The integer values are in `QImode', `HImode', 9630 `SImode', `DImode', `TImode', and `OImode', respectively. 9631 9632 The machine description defines `Pmode' as a C macro which expands 9633 into the machine mode used for addresses. Normally this is the mode 9634 whose size is `BITS_PER_WORD', `SImode' on 32-bit machines. 9635 9636 The only modes which a machine description must support are `QImode', 9637 and the modes corresponding to `BITS_PER_WORD', `FLOAT_TYPE_SIZE' and 9638 `DOUBLE_TYPE_SIZE'. The compiler will attempt to use `DImode' for 9639 8-byte structures and unions, but this can be prevented by overriding 9640 the definition of `MAX_FIXED_MODE_SIZE'. Alternatively, you can have 9641 the compiler use `TImode' for 16-byte structures and unions. Likewise, 9642 you can arrange for the C type `short int' to avoid using `HImode'. 9643 9644 Very few explicit references to machine modes remain in the compiler 9645 and these few references will soon be removed. Instead, the machine 9646 modes are divided into mode classes. These are represented by the 9647 enumeration type `enum mode_class' defined in `machmode.h'. The 9648 possible mode classes are: 9649 9650 `MODE_INT' 9651 Integer modes. By default these are `BImode', `QImode', `HImode', 9652 `SImode', `DImode', `TImode', and `OImode'. 9653 9654 `MODE_PARTIAL_INT' 9655 The "partial integer" modes, `PQImode', `PHImode', `PSImode' and 9656 `PDImode'. 9657 9658 `MODE_FLOAT' 9659 Floating point modes. By default these are `QFmode', `HFmode', 9660 `TQFmode', `SFmode', `DFmode', `XFmode' and `TFmode'. 9661 9662 `MODE_DECIMAL_FLOAT' 9663 Decimal floating point modes. By default these are `SDmode', 9664 `DDmode' and `TDmode'. 9665 9666 `MODE_FRACT' 9667 Signed fractional modes. By default these are `QQmode', `HQmode', 9668 `SQmode', `DQmode' and `TQmode'. 9669 9670 `MODE_UFRACT' 9671 Unsigned fractional modes. By default these are `UQQmode', 9672 `UHQmode', `USQmode', `UDQmode' and `UTQmode'. 9673 9674 `MODE_ACCUM' 9675 Signed accumulator modes. By default these are `HAmode', 9676 `SAmode', `DAmode' and `TAmode'. 9677 9678 `MODE_UACCUM' 9679 Unsigned accumulator modes. By default these are `UHAmode', 9680 `USAmode', `UDAmode' and `UTAmode'. 9681 9682 `MODE_COMPLEX_INT' 9683 Complex integer modes. (These are not currently implemented). 9684 9685 `MODE_COMPLEX_FLOAT' 9686 Complex floating point modes. By default these are `QCmode', 9687 `HCmode', `SCmode', `DCmode', `XCmode', and `TCmode'. 9688 9689 `MODE_FUNCTION' 9690 Algol or Pascal function variables including a static chain. 9691 (These are not currently implemented). 9692 9693 `MODE_CC' 9694 Modes representing condition code values. These are `CCmode' plus 9695 any `CC_MODE' modes listed in the `MACHINE-modes.def'. *Note Jump 9696 Patterns::, also see *note Condition Code::. 9697 9698 `MODE_RANDOM' 9699 This is a catchall mode class for modes which don't fit into the 9700 above classes. Currently `VOIDmode' and `BLKmode' are in 9701 `MODE_RANDOM'. 9702 9703 Here are some C macros that relate to machine modes: 9704 9705 `GET_MODE (X)' 9706 Returns the machine mode of the RTX X. 9707 9708 `PUT_MODE (X, NEWMODE)' 9709 Alters the machine mode of the RTX X to be NEWMODE. 9710 9711 `NUM_MACHINE_MODES' 9712 Stands for the number of machine modes available on the target 9713 machine. This is one greater than the largest numeric value of any 9714 machine mode. 9715 9716 `GET_MODE_NAME (M)' 9717 Returns the name of mode M as a string. 9718 9719 `GET_MODE_CLASS (M)' 9720 Returns the mode class of mode M. 9721 9722 `GET_MODE_WIDER_MODE (M)' 9723 Returns the next wider natural mode. For example, the expression 9724 `GET_MODE_WIDER_MODE (QImode)' returns `HImode'. 9725 9726 `GET_MODE_SIZE (M)' 9727 Returns the size in bytes of a datum of mode M. 9728 9729 `GET_MODE_BITSIZE (M)' 9730 Returns the size in bits of a datum of mode M. 9731 9732 `GET_MODE_IBIT (M)' 9733 Returns the number of integral bits of a datum of fixed-point mode 9734 M. 9735 9736 `GET_MODE_FBIT (M)' 9737 Returns the number of fractional bits of a datum of fixed-point 9738 mode M. 9739 9740 `GET_MODE_MASK (M)' 9741 Returns a bitmask containing 1 for all bits in a word that fit 9742 within mode M. This macro can only be used for modes whose 9743 bitsize is less than or equal to `HOST_BITS_PER_INT'. 9744 9745 `GET_MODE_ALIGNMENT (M)' 9746 Return the required alignment, in bits, for an object of mode M. 9747 9748 `GET_MODE_UNIT_SIZE (M)' 9749 Returns the size in bytes of the subunits of a datum of mode M. 9750 This is the same as `GET_MODE_SIZE' except in the case of complex 9751 modes. For them, the unit size is the size of the real or 9752 imaginary part. 9753 9754 `GET_MODE_NUNITS (M)' 9755 Returns the number of units contained in a mode, i.e., 9756 `GET_MODE_SIZE' divided by `GET_MODE_UNIT_SIZE'. 9757 9758 `GET_CLASS_NARROWEST_MODE (C)' 9759 Returns the narrowest mode in mode class C. 9760 9761 The global variables `byte_mode' and `word_mode' contain modes whose 9762 classes are `MODE_INT' and whose bitsizes are either `BITS_PER_UNIT' or 9763 `BITS_PER_WORD', respectively. On 32-bit machines, these are `QImode' 9764 and `SImode', respectively. 9765 9766 9767 File: gccint.info, Node: Constants, Next: Regs and Memory, Prev: Machine Modes, Up: RTL 9768 9769 10.7 Constant Expression Types 9770 ============================== 9771 9772 The simplest RTL expressions are those that represent constant values. 9773 9774 `(const_int I)' 9775 This type of expression represents the integer value I. I is 9776 customarily accessed with the macro `INTVAL' as in `INTVAL (EXP)', 9777 which is equivalent to `XWINT (EXP, 0)'. 9778 9779 Constants generated for modes with fewer bits than `HOST_WIDE_INT' 9780 must be sign extended to full width (e.g., with `gen_int_mode'). 9781 9782 There is only one expression object for the integer value zero; it 9783 is the value of the variable `const0_rtx'. Likewise, the only 9784 expression for integer value one is found in `const1_rtx', the only 9785 expression for integer value two is found in `const2_rtx', and the 9786 only expression for integer value negative one is found in 9787 `constm1_rtx'. Any attempt to create an expression of code 9788 `const_int' and value zero, one, two or negative one will return 9789 `const0_rtx', `const1_rtx', `const2_rtx' or `constm1_rtx' as 9790 appropriate. 9791 9792 Similarly, there is only one object for the integer whose value is 9793 `STORE_FLAG_VALUE'. It is found in `const_true_rtx'. If 9794 `STORE_FLAG_VALUE' is one, `const_true_rtx' and `const1_rtx' will 9795 point to the same object. If `STORE_FLAG_VALUE' is -1, 9796 `const_true_rtx' and `constm1_rtx' will point to the same object. 9797 9798 `(const_double:M I0 I1 ...)' 9799 Represents either a floating-point constant of mode M or an 9800 integer constant too large to fit into `HOST_BITS_PER_WIDE_INT' 9801 bits but small enough to fit within twice that number of bits (GCC 9802 does not provide a mechanism to represent even larger constants). 9803 In the latter case, M will be `VOIDmode'. 9804 9805 If M is `VOIDmode', the bits of the value are stored in I0 and I1. 9806 I0 is customarily accessed with the macro `CONST_DOUBLE_LOW' and 9807 I1 with `CONST_DOUBLE_HIGH'. 9808 9809 If the constant is floating point (regardless of its precision), 9810 then the number of integers used to store the value depends on the 9811 size of `REAL_VALUE_TYPE' (*note Floating Point::). The integers 9812 represent a floating point number, but not precisely in the target 9813 machine's or host machine's floating point format. To convert 9814 them to the precise bit pattern used by the target machine, use 9815 the macro `REAL_VALUE_TO_TARGET_DOUBLE' and friends (*note Data 9816 Output::). 9817 9818 `(const_fixed:M ...)' 9819 Represents a fixed-point constant of mode M. The operand is a 9820 data structure of type `struct fixed_value' and is accessed with 9821 the macro `CONST_FIXED_VALUE'. The high part of data is accessed 9822 with `CONST_FIXED_VALUE_HIGH'; the low part is accessed with 9823 `CONST_FIXED_VALUE_LOW'. 9824 9825 `(const_vector:M [X0 X1 ...])' 9826 Represents a vector constant. The square brackets stand for the 9827 vector containing the constant elements. X0, X1 and so on are the 9828 `const_int', `const_double' or `const_fixed' elements. 9829 9830 The number of units in a `const_vector' is obtained with the macro 9831 `CONST_VECTOR_NUNITS' as in `CONST_VECTOR_NUNITS (V)'. 9832 9833 Individual elements in a vector constant are accessed with the 9834 macro `CONST_VECTOR_ELT' as in `CONST_VECTOR_ELT (V, N)' where V 9835 is the vector constant and N is the element desired. 9836 9837 `(const_string STR)' 9838 Represents a constant string with value STR. Currently this is 9839 used only for insn attributes (*note Insn Attributes::) since 9840 constant strings in C are placed in memory. 9841 9842 `(symbol_ref:MODE SYMBOL)' 9843 Represents the value of an assembler label for data. SYMBOL is a 9844 string that describes the name of the assembler label. If it 9845 starts with a `*', the label is the rest of SYMBOL not including 9846 the `*'. Otherwise, the label is SYMBOL, usually prefixed with 9847 `_'. 9848 9849 The `symbol_ref' contains a mode, which is usually `Pmode'. 9850 Usually that is the only mode for which a symbol is directly valid. 9851 9852 `(label_ref:MODE LABEL)' 9853 Represents the value of an assembler label for code. It contains 9854 one operand, an expression, which must be a `code_label' or a 9855 `note' of type `NOTE_INSN_DELETED_LABEL' that appears in the 9856 instruction sequence to identify the place where the label should 9857 go. 9858 9859 The reason for using a distinct expression type for code label 9860 references is so that jump optimization can distinguish them. 9861 9862 The `label_ref' contains a mode, which is usually `Pmode'. 9863 Usually that is the only mode for which a label is directly valid. 9864 9865 `(const:M EXP)' 9866 Represents a constant that is the result of an assembly-time 9867 arithmetic computation. The operand, EXP, is an expression that 9868 contains only constants (`const_int', `symbol_ref' and `label_ref' 9869 expressions) combined with `plus' and `minus'. However, not all 9870 combinations are valid, since the assembler cannot do arbitrary 9871 arithmetic on relocatable symbols. 9872 9873 M should be `Pmode'. 9874 9875 `(high:M EXP)' 9876 Represents the high-order bits of EXP, usually a `symbol_ref'. 9877 The number of bits is machine-dependent and is normally the number 9878 of bits specified in an instruction that initializes the high 9879 order bits of a register. It is used with `lo_sum' to represent 9880 the typical two-instruction sequence used in RISC machines to 9881 reference a global memory location. 9882 9883 M should be `Pmode'. 9884 9885 The macro `CONST0_RTX (MODE)' refers to an expression with value 0 in 9886 mode MODE. If mode MODE is of mode class `MODE_INT', it returns 9887 `const0_rtx'. If mode MODE is of mode class `MODE_FLOAT', it returns a 9888 `CONST_DOUBLE' expression in mode MODE. Otherwise, it returns a 9889 `CONST_VECTOR' expression in mode MODE. Similarly, the macro 9890 `CONST1_RTX (MODE)' refers to an expression with value 1 in mode MODE 9891 and similarly for `CONST2_RTX'. The `CONST1_RTX' and `CONST2_RTX' 9892 macros are undefined for vector modes. 9893 9894 9895 File: gccint.info, Node: Regs and Memory, Next: Arithmetic, Prev: Constants, Up: RTL 9896 9897 10.8 Registers and Memory 9898 ========================= 9899 9900 Here are the RTL expression types for describing access to machine 9901 registers and to main memory. 9902 9903 `(reg:M N)' 9904 For small values of the integer N (those that are less than 9905 `FIRST_PSEUDO_REGISTER'), this stands for a reference to machine 9906 register number N: a "hard register". For larger values of N, it 9907 stands for a temporary value or "pseudo register". The compiler's 9908 strategy is to generate code assuming an unlimited number of such 9909 pseudo registers, and later convert them into hard registers or 9910 into memory references. 9911 9912 M is the machine mode of the reference. It is necessary because 9913 machines can generally refer to each register in more than one 9914 mode. For example, a register may contain a full word but there 9915 may be instructions to refer to it as a half word or as a single 9916 byte, as well as instructions to refer to it as a floating point 9917 number of various precisions. 9918 9919 Even for a register that the machine can access in only one mode, 9920 the mode must always be specified. 9921 9922 The symbol `FIRST_PSEUDO_REGISTER' is defined by the machine 9923 description, since the number of hard registers on the machine is 9924 an invariant characteristic of the machine. Note, however, that 9925 not all of the machine registers must be general registers. All 9926 the machine registers that can be used for storage of data are 9927 given hard register numbers, even those that can be used only in 9928 certain instructions or can hold only certain types of data. 9929 9930 A hard register may be accessed in various modes throughout one 9931 function, but each pseudo register is given a natural mode and is 9932 accessed only in that mode. When it is necessary to describe an 9933 access to a pseudo register using a nonnatural mode, a `subreg' 9934 expression is used. 9935 9936 A `reg' expression with a machine mode that specifies more than 9937 one word of data may actually stand for several consecutive 9938 registers. If in addition the register number specifies a 9939 hardware register, then it actually represents several consecutive 9940 hardware registers starting with the specified one. 9941 9942 Each pseudo register number used in a function's RTL code is 9943 represented by a unique `reg' expression. 9944 9945 Some pseudo register numbers, those within the range of 9946 `FIRST_VIRTUAL_REGISTER' to `LAST_VIRTUAL_REGISTER' only appear 9947 during the RTL generation phase and are eliminated before the 9948 optimization phases. These represent locations in the stack frame 9949 that cannot be determined until RTL generation for the function 9950 has been completed. The following virtual register numbers are 9951 defined: 9952 9953 `VIRTUAL_INCOMING_ARGS_REGNUM' 9954 This points to the first word of the incoming arguments 9955 passed on the stack. Normally these arguments are placed 9956 there by the caller, but the callee may have pushed some 9957 arguments that were previously passed in registers. 9958 9959 When RTL generation is complete, this virtual register is 9960 replaced by the sum of the register given by 9961 `ARG_POINTER_REGNUM' and the value of `FIRST_PARM_OFFSET'. 9962 9963 `VIRTUAL_STACK_VARS_REGNUM' 9964 If `FRAME_GROWS_DOWNWARD' is defined to a nonzero value, this 9965 points to immediately above the first variable on the stack. 9966 Otherwise, it points to the first variable on the stack. 9967 9968 `VIRTUAL_STACK_VARS_REGNUM' is replaced with the sum of the 9969 register given by `FRAME_POINTER_REGNUM' and the value 9970 `STARTING_FRAME_OFFSET'. 9971 9972 `VIRTUAL_STACK_DYNAMIC_REGNUM' 9973 This points to the location of dynamically allocated memory 9974 on the stack immediately after the stack pointer has been 9975 adjusted by the amount of memory desired. 9976 9977 This virtual register is replaced by the sum of the register 9978 given by `STACK_POINTER_REGNUM' and the value 9979 `STACK_DYNAMIC_OFFSET'. 9980 9981 `VIRTUAL_OUTGOING_ARGS_REGNUM' 9982 This points to the location in the stack at which outgoing 9983 arguments should be written when the stack is pre-pushed 9984 (arguments pushed using push insns should always use 9985 `STACK_POINTER_REGNUM'). 9986 9987 This virtual register is replaced by the sum of the register 9988 given by `STACK_POINTER_REGNUM' and the value 9989 `STACK_POINTER_OFFSET'. 9990 9991 `(subreg:M1 REG:M2 BYTENUM)' 9992 `subreg' expressions are used to refer to a register in a machine 9993 mode other than its natural one, or to refer to one register of a 9994 multi-part `reg' that actually refers to several registers. 9995 9996 Each pseudo register has a natural mode. If it is necessary to 9997 operate on it in a different mode, the register must be enclosed 9998 in a `subreg'. 9999 10000 There are currently three supported types for the first operand of 10001 a `subreg': 10002 * pseudo registers This is the most common case. Most 10003 `subreg's have pseudo `reg's as their first operand. 10004 10005 * mem `subreg's of `mem' were common in earlier versions of GCC 10006 and are still supported. During the reload pass these are 10007 replaced by plain `mem's. On machines that do not do 10008 instruction scheduling, use of `subreg's of `mem' are still 10009 used, but this is no longer recommended. Such `subreg's are 10010 considered to be `register_operand's rather than 10011 `memory_operand's before and during reload. Because of this, 10012 the scheduling passes cannot properly schedule instructions 10013 with `subreg's of `mem', so for machines that do scheduling, 10014 `subreg's of `mem' should never be used. To support this, 10015 the combine and recog passes have explicit code to inhibit 10016 the creation of `subreg's of `mem' when `INSN_SCHEDULING' is 10017 defined. 10018 10019 The use of `subreg's of `mem' after the reload pass is an area 10020 that is not well understood and should be avoided. There is 10021 still some code in the compiler to support this, but this 10022 code has possibly rotted. This use of `subreg's is 10023 discouraged and will most likely not be supported in the 10024 future. 10025 10026 * hard registers It is seldom necessary to wrap hard registers 10027 in `subreg's; such registers would normally reduce to a 10028 single `reg' rtx. This use of `subreg's is discouraged and 10029 may not be supported in the future. 10030 10031 10032 `subreg's of `subreg's are not supported. Using 10033 `simplify_gen_subreg' is the recommended way to avoid this problem. 10034 10035 `subreg's come in two distinct flavors, each having its own usage 10036 and rules: 10037 10038 Paradoxical subregs 10039 When M1 is strictly wider than M2, the `subreg' expression is 10040 called "paradoxical". The canonical test for this class of 10041 `subreg' is: 10042 10043 GET_MODE_SIZE (M1) > GET_MODE_SIZE (M2) 10044 10045 Paradoxical `subreg's can be used as both lvalues and rvalues. 10046 When used as an lvalue, the low-order bits of the source value 10047 are stored in REG and the high-order bits are discarded. 10048 When used as an rvalue, the low-order bits of the `subreg' are 10049 taken from REG while the high-order bits may or may not be 10050 defined. 10051 10052 The high-order bits of rvalues are in the following 10053 circumstances: 10054 10055 * `subreg's of `mem' When M2 is smaller than a word, the 10056 macro `LOAD_EXTEND_OP', can control how the high-order 10057 bits are defined. 10058 10059 * `subreg' of `reg's The upper bits are defined when 10060 `SUBREG_PROMOTED_VAR_P' is true. 10061 `SUBREG_PROMOTED_UNSIGNED_P' describes what the upper 10062 bits hold. Such subregs usually represent local 10063 variables, register variables and parameter pseudo 10064 variables that have been promoted to a wider mode. 10065 10066 10067 BYTENUM is always zero for a paradoxical `subreg', even on 10068 big-endian targets. 10069 10070 For example, the paradoxical `subreg': 10071 10072 (set (subreg:SI (reg:HI X) 0) Y) 10073 10074 stores the lower 2 bytes of Y in X and discards the upper 2 10075 bytes. A subsequent: 10076 10077 (set Z (subreg:SI (reg:HI X) 0)) 10078 10079 would set the lower two bytes of Z to Y and set the upper two 10080 bytes to an unknown value assuming `SUBREG_PROMOTED_VAR_P' is 10081 false. 10082 10083 Normal subregs 10084 When M1 is at least as narrow as M2 the `subreg' expression 10085 is called "normal". 10086 10087 Normal `subreg's restrict consideration to certain bits of 10088 REG. There are two cases. If M1 is smaller than a word, the 10089 `subreg' refers to the least-significant part (or "lowpart") 10090 of one word of REG. If M1 is word-sized or greater, the 10091 `subreg' refers to one or more complete words. 10092 10093 When used as an lvalue, `subreg' is a word-based accessor. 10094 Storing to a `subreg' modifies all the words of REG that 10095 overlap the `subreg', but it leaves the other words of REG 10096 alone. 10097 10098 When storing to a normal `subreg' that is smaller than a word, 10099 the other bits of the referenced word are usually left in an 10100 undefined state. This laxity makes it easier to generate 10101 efficient code for such instructions. To represent an 10102 instruction that preserves all the bits outside of those in 10103 the `subreg', use `strict_low_part' or `zero_extract' around 10104 the `subreg'. 10105 10106 BYTENUM must identify the offset of the first byte of the 10107 `subreg' from the start of REG, assuming that REG is laid out 10108 in memory order. The memory order of bytes is defined by two 10109 target macros, `WORDS_BIG_ENDIAN' and `BYTES_BIG_ENDIAN': 10110 10111 * `WORDS_BIG_ENDIAN', if set to 1, says that byte number 10112 zero is part of the most significant word; otherwise, it 10113 is part of the least significant word. 10114 10115 * `BYTES_BIG_ENDIAN', if set to 1, says that byte number 10116 zero is the most significant byte within a word; 10117 otherwise, it is the least significant byte within a 10118 word. 10119 10120 On a few targets, `FLOAT_WORDS_BIG_ENDIAN' disagrees with 10121 `WORDS_BIG_ENDIAN'. However, most parts of the compiler treat 10122 floating point values as if they had the same endianness as 10123 integer values. This works because they handle them solely 10124 as a collection of integer values, with no particular 10125 numerical value. Only real.c and the runtime libraries care 10126 about `FLOAT_WORDS_BIG_ENDIAN'. 10127 10128 Thus, 10129 10130 (subreg:HI (reg:SI X) 2) 10131 10132 on a `BYTES_BIG_ENDIAN', `UNITS_PER_WORD == 4' target is the 10133 same as 10134 10135 (subreg:HI (reg:SI X) 0) 10136 10137 on a little-endian, `UNITS_PER_WORD == 4' target. Both 10138 `subreg's access the lower two bytes of register X. 10139 10140 10141 A `MODE_PARTIAL_INT' mode behaves as if it were as wide as the 10142 corresponding `MODE_INT' mode, except that it has an unknown 10143 number of undefined bits. For example: 10144 10145 (subreg:PSI (reg:SI 0) 0) 10146 10147 accesses the whole of `(reg:SI 0)', but the exact relationship 10148 between the `PSImode' value and the `SImode' value is not defined. 10149 If we assume `UNITS_PER_WORD <= 4', then the following two 10150 `subreg's: 10151 10152 (subreg:PSI (reg:DI 0) 0) 10153 (subreg:PSI (reg:DI 0) 4) 10154 10155 represent independent 4-byte accesses to the two halves of 10156 `(reg:DI 0)'. Both `subreg's have an unknown number of undefined 10157 bits. 10158 10159 If `UNITS_PER_WORD <= 2' then these two `subreg's: 10160 10161 (subreg:HI (reg:PSI 0) 0) 10162 (subreg:HI (reg:PSI 0) 2) 10163 10164 represent independent 2-byte accesses that together span the whole 10165 of `(reg:PSI 0)'. Storing to the first `subreg' does not affect 10166 the value of the second, and vice versa. `(reg:PSI 0)' has an 10167 unknown number of undefined bits, so the assignment: 10168 10169 (set (subreg:HI (reg:PSI 0) 0) (reg:HI 4)) 10170 10171 does not guarantee that `(subreg:HI (reg:PSI 0) 0)' has the value 10172 `(reg:HI 4)'. 10173 10174 The rules above apply to both pseudo REGs and hard REGs. If the 10175 semantics are not correct for particular combinations of M1, M2 10176 and hard REG, the target-specific code must ensure that those 10177 combinations are never used. For example: 10178 10179 CANNOT_CHANGE_MODE_CLASS (M2, M1, CLASS) 10180 10181 must be true for every class CLASS that includes REG. 10182 10183 The first operand of a `subreg' expression is customarily accessed 10184 with the `SUBREG_REG' macro and the second operand is customarily 10185 accessed with the `SUBREG_BYTE' macro. 10186 10187 It has been several years since a platform in which 10188 `BYTES_BIG_ENDIAN' not equal to `WORDS_BIG_ENDIAN' has been 10189 tested. Anyone wishing to support such a platform in the future 10190 may be confronted with code rot. 10191 10192 `(scratch:M)' 10193 This represents a scratch register that will be required for the 10194 execution of a single instruction and not used subsequently. It is 10195 converted into a `reg' by either the local register allocator or 10196 the reload pass. 10197 10198 `scratch' is usually present inside a `clobber' operation (*note 10199 Side Effects::). 10200 10201 `(cc0)' 10202 This refers to the machine's condition code register. It has no 10203 operands and may not have a machine mode. There are two ways to 10204 use it: 10205 10206 * To stand for a complete set of condition code flags. This is 10207 best on most machines, where each comparison sets the entire 10208 series of flags. 10209 10210 With this technique, `(cc0)' may be validly used in only two 10211 contexts: as the destination of an assignment (in test and 10212 compare instructions) and in comparison operators comparing 10213 against zero (`const_int' with value zero; that is to say, 10214 `const0_rtx'). 10215 10216 * To stand for a single flag that is the result of a single 10217 condition. This is useful on machines that have only a 10218 single flag bit, and in which comparison instructions must 10219 specify the condition to test. 10220 10221 With this technique, `(cc0)' may be validly used in only two 10222 contexts: as the destination of an assignment (in test and 10223 compare instructions) where the source is a comparison 10224 operator, and as the first operand of `if_then_else' (in a 10225 conditional branch). 10226 10227 There is only one expression object of code `cc0'; it is the value 10228 of the variable `cc0_rtx'. Any attempt to create an expression of 10229 code `cc0' will return `cc0_rtx'. 10230 10231 Instructions can set the condition code implicitly. On many 10232 machines, nearly all instructions set the condition code based on 10233 the value that they compute or store. It is not necessary to 10234 record these actions explicitly in the RTL because the machine 10235 description includes a prescription for recognizing the 10236 instructions that do so (by means of the macro 10237 `NOTICE_UPDATE_CC'). *Note Condition Code::. Only instructions 10238 whose sole purpose is to set the condition code, and instructions 10239 that use the condition code, need mention `(cc0)'. 10240 10241 On some machines, the condition code register is given a register 10242 number and a `reg' is used instead of `(cc0)'. This is usually the 10243 preferable approach if only a small subset of instructions modify 10244 the condition code. Other machines store condition codes in 10245 general registers; in such cases a pseudo register should be used. 10246 10247 Some machines, such as the SPARC and RS/6000, have two sets of 10248 arithmetic instructions, one that sets and one that does not set 10249 the condition code. This is best handled by normally generating 10250 the instruction that does not set the condition code, and making a 10251 pattern that both performs the arithmetic and sets the condition 10252 code register (which would not be `(cc0)' in this case). For 10253 examples, search for `addcc' and `andcc' in `sparc.md'. 10254 10255 `(pc)' 10256 This represents the machine's program counter. It has no operands 10257 and may not have a machine mode. `(pc)' may be validly used only 10258 in certain specific contexts in jump instructions. 10259 10260 There is only one expression object of code `pc'; it is the value 10261 of the variable `pc_rtx'. Any attempt to create an expression of 10262 code `pc' will return `pc_rtx'. 10263 10264 All instructions that do not jump alter the program counter 10265 implicitly by incrementing it, but there is no need to mention 10266 this in the RTL. 10267 10268 `(mem:M ADDR ALIAS)' 10269 This RTX represents a reference to main memory at an address 10270 represented by the expression ADDR. M specifies how large a unit 10271 of memory is accessed. ALIAS specifies an alias set for the 10272 reference. In general two items are in different alias sets if 10273 they cannot reference the same memory address. 10274 10275 The construct `(mem:BLK (scratch))' is considered to alias all 10276 other memories. Thus it may be used as a memory barrier in 10277 epilogue stack deallocation patterns. 10278 10279 `(concatM RTX RTX)' 10280 This RTX represents the concatenation of two other RTXs. This is 10281 used for complex values. It should only appear in the RTL 10282 attached to declarations and during RTL generation. It should not 10283 appear in the ordinary insn chain. 10284 10285 `(concatnM [RTX ...])' 10286 This RTX represents the concatenation of all the RTX to make a 10287 single value. Like `concat', this should only appear in 10288 declarations, and not in the insn chain. 10289 10290 10291 File: gccint.info, Node: Arithmetic, Next: Comparisons, Prev: Regs and Memory, Up: RTL 10292 10293 10.9 RTL Expressions for Arithmetic 10294 =================================== 10295 10296 Unless otherwise specified, all the operands of arithmetic expressions 10297 must be valid for mode M. An operand is valid for mode M if it has 10298 mode M, or if it is a `const_int' or `const_double' and M is a mode of 10299 class `MODE_INT'. 10300 10301 For commutative binary operations, constants should be placed in the 10302 second operand. 10303 10304 `(plus:M X Y)' 10305 `(ss_plus:M X Y)' 10306 `(us_plus:M X Y)' 10307 These three expressions all represent the sum of the values 10308 represented by X and Y carried out in machine mode M. They differ 10309 in their behavior on overflow of integer modes. `plus' wraps 10310 round modulo the width of M; `ss_plus' saturates at the maximum 10311 signed value representable in M; `us_plus' saturates at the 10312 maximum unsigned value. 10313 10314 `(lo_sum:M X Y)' 10315 This expression represents the sum of X and the low-order bits of 10316 Y. It is used with `high' (*note Constants::) to represent the 10317 typical two-instruction sequence used in RISC machines to 10318 reference a global memory location. 10319 10320 The number of low order bits is machine-dependent but is normally 10321 the number of bits in a `Pmode' item minus the number of bits set 10322 by `high'. 10323 10324 M should be `Pmode'. 10325 10326 `(minus:M X Y)' 10327 `(ss_minus:M X Y)' 10328 `(us_minus:M X Y)' 10329 These three expressions represent the result of subtracting Y from 10330 X, carried out in mode M. Behavior on overflow is the same as for 10331 the three variants of `plus' (see above). 10332 10333 `(compare:M X Y)' 10334 Represents the result of subtracting Y from X for purposes of 10335 comparison. The result is computed without overflow, as if with 10336 infinite precision. 10337 10338 Of course, machines can't really subtract with infinite precision. 10339 However, they can pretend to do so when only the sign of the 10340 result will be used, which is the case when the result is stored 10341 in the condition code. And that is the _only_ way this kind of 10342 expression may validly be used: as a value to be stored in the 10343 condition codes, either `(cc0)' or a register. *Note 10344 Comparisons::. 10345 10346 The mode M is not related to the modes of X and Y, but instead is 10347 the mode of the condition code value. If `(cc0)' is used, it is 10348 `VOIDmode'. Otherwise it is some mode in class `MODE_CC', often 10349 `CCmode'. *Note Condition Code::. If M is `VOIDmode' or 10350 `CCmode', the operation returns sufficient information (in an 10351 unspecified format) so that any comparison operator can be applied 10352 to the result of the `COMPARE' operation. For other modes in 10353 class `MODE_CC', the operation only returns a subset of this 10354 information. 10355 10356 Normally, X and Y must have the same mode. Otherwise, `compare' 10357 is valid only if the mode of X is in class `MODE_INT' and Y is a 10358 `const_int' or `const_double' with mode `VOIDmode'. The mode of X 10359 determines what mode the comparison is to be done in; thus it must 10360 not be `VOIDmode'. 10361 10362 If one of the operands is a constant, it should be placed in the 10363 second operand and the comparison code adjusted as appropriate. 10364 10365 A `compare' specifying two `VOIDmode' constants is not valid since 10366 there is no way to know in what mode the comparison is to be 10367 performed; the comparison must either be folded during the 10368 compilation or the first operand must be loaded into a register 10369 while its mode is still known. 10370 10371 `(neg:M X)' 10372 `(ss_neg:M X)' 10373 `(us_neg:M X)' 10374 These two expressions represent the negation (subtraction from 10375 zero) of the value represented by X, carried out in mode M. They 10376 differ in the behavior on overflow of integer modes. In the case 10377 of `neg', the negation of the operand may be a number not 10378 representable in mode M, in which case it is truncated to M. 10379 `ss_neg' and `us_neg' ensure that an out-of-bounds result 10380 saturates to the maximum or minimum signed or unsigned value. 10381 10382 `(mult:M X Y)' 10383 `(ss_mult:M X Y)' 10384 `(us_mult:M X Y)' 10385 Represents the signed product of the values represented by X and Y 10386 carried out in machine mode M. `ss_mult' and `us_mult' ensure 10387 that an out-of-bounds result saturates to the maximum or minimum 10388 signed or unsigned value. 10389 10390 Some machines support a multiplication that generates a product 10391 wider than the operands. Write the pattern for this as 10392 10393 (mult:M (sign_extend:M X) (sign_extend:M Y)) 10394 10395 where M is wider than the modes of X and Y, which need not be the 10396 same. 10397 10398 For unsigned widening multiplication, use the same idiom, but with 10399 `zero_extend' instead of `sign_extend'. 10400 10401 `(div:M X Y)' 10402 `(ss_div:M X Y)' 10403 Represents the quotient in signed division of X by Y, carried out 10404 in machine mode M. If M is a floating point mode, it represents 10405 the exact quotient; otherwise, the integerized quotient. `ss_div' 10406 ensures that an out-of-bounds result saturates to the maximum or 10407 minimum signed value. 10408 10409 Some machines have division instructions in which the operands and 10410 quotient widths are not all the same; you should represent such 10411 instructions using `truncate' and `sign_extend' as in, 10412 10413 (truncate:M1 (div:M2 X (sign_extend:M2 Y))) 10414 10415 `(udiv:M X Y)' 10416 `(us_div:M X Y)' 10417 Like `div' but represents unsigned division. `us_div' ensures 10418 that an out-of-bounds result saturates to the maximum or minimum 10419 unsigned value. 10420 10421 `(mod:M X Y)' 10422 `(umod:M X Y)' 10423 Like `div' and `udiv' but represent the remainder instead of the 10424 quotient. 10425 10426 `(smin:M X Y)' 10427 `(smax:M X Y)' 10428 Represents the smaller (for `smin') or larger (for `smax') of X 10429 and Y, interpreted as signed values in mode M. When used with 10430 floating point, if both operands are zeros, or if either operand 10431 is `NaN', then it is unspecified which of the two operands is 10432 returned as the result. 10433 10434 `(umin:M X Y)' 10435 `(umax:M X Y)' 10436 Like `smin' and `smax', but the values are interpreted as unsigned 10437 integers. 10438 10439 `(not:M X)' 10440 Represents the bitwise complement of the value represented by X, 10441 carried out in mode M, which must be a fixed-point machine mode. 10442 10443 `(and:M X Y)' 10444 Represents the bitwise logical-and of the values represented by X 10445 and Y, carried out in machine mode M, which must be a fixed-point 10446 machine mode. 10447 10448 `(ior:M X Y)' 10449 Represents the bitwise inclusive-or of the values represented by X 10450 and Y, carried out in machine mode M, which must be a fixed-point 10451 mode. 10452 10453 `(xor:M X Y)' 10454 Represents the bitwise exclusive-or of the values represented by X 10455 and Y, carried out in machine mode M, which must be a fixed-point 10456 mode. 10457 10458 `(ashift:M X C)' 10459 `(ss_ashift:M X C)' 10460 `(us_ashift:M X C)' 10461 These three expressions represent the result of arithmetically 10462 shifting X left by C places. They differ in their behavior on 10463 overflow of integer modes. An `ashift' operation is a plain shift 10464 with no special behavior in case of a change in the sign bit; 10465 `ss_ashift' and `us_ashift' saturates to the minimum or maximum 10466 representable value if any of the bits shifted out differs from 10467 the final sign bit. 10468 10469 X have mode M, a fixed-point machine mode. C be a fixed-point 10470 mode or be a constant with mode `VOIDmode'; which mode is 10471 determined by the mode called for in the machine description entry 10472 for the left-shift instruction. For example, on the VAX, the mode 10473 of C is `QImode' regardless of M. 10474 10475 `(lshiftrt:M X C)' 10476 `(ashiftrt:M X C)' 10477 Like `ashift' but for right shift. Unlike the case for left shift, 10478 these two operations are distinct. 10479 10480 `(rotate:M X C)' 10481 `(rotatert:M X C)' 10482 Similar but represent left and right rotate. If C is a constant, 10483 use `rotate'. 10484 10485 `(abs:M X)' 10486 Represents the absolute value of X, computed in mode M. 10487 10488 `(sqrt:M X)' 10489 Represents the square root of X, computed in mode M. Most often M 10490 will be a floating point mode. 10491 10492 `(ffs:M X)' 10493 Represents one plus the index of the least significant 1-bit in X, 10494 represented as an integer of mode M. (The value is zero if X is 10495 zero.) The mode of X need not be M; depending on the target 10496 machine, various mode combinations may be valid. 10497 10498 `(clz:M X)' 10499 Represents the number of leading 0-bits in X, represented as an 10500 integer of mode M, starting at the most significant bit position. 10501 If X is zero, the value is determined by 10502 `CLZ_DEFINED_VALUE_AT_ZERO' (*note Misc::). Note that this is one 10503 of the few expressions that is not invariant under widening. The 10504 mode of X will usually be an integer mode. 10505 10506 `(ctz:M X)' 10507 Represents the number of trailing 0-bits in X, represented as an 10508 integer of mode M, starting at the least significant bit position. 10509 If X is zero, the value is determined by 10510 `CTZ_DEFINED_VALUE_AT_ZERO' (*note Misc::). Except for this case, 10511 `ctz(x)' is equivalent to `ffs(X) - 1'. The mode of X will 10512 usually be an integer mode. 10513 10514 `(popcount:M X)' 10515 Represents the number of 1-bits in X, represented as an integer of 10516 mode M. The mode of X will usually be an integer mode. 10517 10518 `(parity:M X)' 10519 Represents the number of 1-bits modulo 2 in X, represented as an 10520 integer of mode M. The mode of X will usually be an integer mode. 10521 10522 `(bswap:M X)' 10523 Represents the value X with the order of bytes reversed, carried 10524 out in mode M, which must be a fixed-point machine mode. 10525 10526 10527 File: gccint.info, Node: Comparisons, Next: Bit-Fields, Prev: Arithmetic, Up: RTL 10528 10529 10.10 Comparison Operations 10530 =========================== 10531 10532 Comparison operators test a relation on two operands and are considered 10533 to represent a machine-dependent nonzero value described by, but not 10534 necessarily equal to, `STORE_FLAG_VALUE' (*note Misc::) if the relation 10535 holds, or zero if it does not, for comparison operators whose results 10536 have a `MODE_INT' mode, `FLOAT_STORE_FLAG_VALUE' (*note Misc::) if the 10537 relation holds, or zero if it does not, for comparison operators that 10538 return floating-point values, and a vector of either 10539 `VECTOR_STORE_FLAG_VALUE' (*note Misc::) if the relation holds, or of 10540 zeros if it does not, for comparison operators that return vector 10541 results. The mode of the comparison operation is independent of the 10542 mode of the data being compared. If the comparison operation is being 10543 tested (e.g., the first operand of an `if_then_else'), the mode must be 10544 `VOIDmode'. 10545 10546 There are two ways that comparison operations may be used. The 10547 comparison operators may be used to compare the condition codes `(cc0)' 10548 against zero, as in `(eq (cc0) (const_int 0))'. Such a construct 10549 actually refers to the result of the preceding instruction in which the 10550 condition codes were set. The instruction setting the condition code 10551 must be adjacent to the instruction using the condition code; only 10552 `note' insns may separate them. 10553 10554 Alternatively, a comparison operation may directly compare two data 10555 objects. The mode of the comparison is determined by the operands; they 10556 must both be valid for a common machine mode. A comparison with both 10557 operands constant would be invalid as the machine mode could not be 10558 deduced from it, but such a comparison should never exist in RTL due to 10559 constant folding. 10560 10561 In the example above, if `(cc0)' were last set to `(compare X Y)', the 10562 comparison operation is identical to `(eq X Y)'. Usually only one style 10563 of comparisons is supported on a particular machine, but the combine 10564 pass will try to merge the operations to produce the `eq' shown in case 10565 it exists in the context of the particular insn involved. 10566 10567 Inequality comparisons come in two flavors, signed and unsigned. Thus, 10568 there are distinct expression codes `gt' and `gtu' for signed and 10569 unsigned greater-than. These can produce different results for the same 10570 pair of integer values: for example, 1 is signed greater-than -1 but not 10571 unsigned greater-than, because -1 when regarded as unsigned is actually 10572 `0xffffffff' which is greater than 1. 10573 10574 The signed comparisons are also used for floating point values. 10575 Floating point comparisons are distinguished by the machine modes of 10576 the operands. 10577 10578 `(eq:M X Y)' 10579 `STORE_FLAG_VALUE' if the values represented by X and Y are equal, 10580 otherwise 0. 10581 10582 `(ne:M X Y)' 10583 `STORE_FLAG_VALUE' if the values represented by X and Y are not 10584 equal, otherwise 0. 10585 10586 `(gt:M X Y)' 10587 `STORE_FLAG_VALUE' if the X is greater than Y. If they are 10588 fixed-point, the comparison is done in a signed sense. 10589 10590 `(gtu:M X Y)' 10591 Like `gt' but does unsigned comparison, on fixed-point numbers 10592 only. 10593 10594 `(lt:M X Y)' 10595 `(ltu:M X Y)' 10596 Like `gt' and `gtu' but test for "less than". 10597 10598 `(ge:M X Y)' 10599 `(geu:M X Y)' 10600 Like `gt' and `gtu' but test for "greater than or equal". 10601 10602 `(le:M X Y)' 10603 `(leu:M X Y)' 10604 Like `gt' and `gtu' but test for "less than or equal". 10605 10606 `(if_then_else COND THEN ELSE)' 10607 This is not a comparison operation but is listed here because it is 10608 always used in conjunction with a comparison operation. To be 10609 precise, COND is a comparison expression. This expression 10610 represents a choice, according to COND, between the value 10611 represented by THEN and the one represented by ELSE. 10612 10613 On most machines, `if_then_else' expressions are valid only to 10614 express conditional jumps. 10615 10616 `(cond [TEST1 VALUE1 TEST2 VALUE2 ...] DEFAULT)' 10617 Similar to `if_then_else', but more general. Each of TEST1, 10618 TEST2, ... is performed in turn. The result of this expression is 10619 the VALUE corresponding to the first nonzero test, or DEFAULT if 10620 none of the tests are nonzero expressions. 10621 10622 This is currently not valid for instruction patterns and is 10623 supported only for insn attributes. *Note Insn Attributes::. 10624 10625 10626 File: gccint.info, Node: Bit-Fields, Next: Vector Operations, Prev: Comparisons, Up: RTL 10627 10628 10.11 Bit-Fields 10629 ================ 10630 10631 Special expression codes exist to represent bit-field instructions. 10632 10633 `(sign_extract:M LOC SIZE POS)' 10634 This represents a reference to a sign-extended bit-field contained 10635 or starting in LOC (a memory or register reference). The bit-field 10636 is SIZE bits wide and starts at bit POS. The compilation option 10637 `BITS_BIG_ENDIAN' says which end of the memory unit POS counts 10638 from. 10639 10640 If LOC is in memory, its mode must be a single-byte integer mode. 10641 If LOC is in a register, the mode to use is specified by the 10642 operand of the `insv' or `extv' pattern (*note Standard Names::) 10643 and is usually a full-word integer mode, which is the default if 10644 none is specified. 10645 10646 The mode of POS is machine-specific and is also specified in the 10647 `insv' or `extv' pattern. 10648 10649 The mode M is the same as the mode that would be used for LOC if 10650 it were a register. 10651 10652 A `sign_extract' can not appear as an lvalue, or part thereof, in 10653 RTL. 10654 10655 `(zero_extract:M LOC SIZE POS)' 10656 Like `sign_extract' but refers to an unsigned or zero-extended 10657 bit-field. The same sequence of bits are extracted, but they are 10658 filled to an entire word with zeros instead of by sign-extension. 10659 10660 Unlike `sign_extract', this type of expressions can be lvalues in 10661 RTL; they may appear on the left side of an assignment, indicating 10662 insertion of a value into the specified bit-field. 10663 10664 10665 File: gccint.info, Node: Vector Operations, Next: Conversions, Prev: Bit-Fields, Up: RTL 10666 10667 10.12 Vector Operations 10668 ======================= 10669 10670 All normal RTL expressions can be used with vector modes; they are 10671 interpreted as operating on each part of the vector independently. 10672 Additionally, there are a few new expressions to describe specific 10673 vector operations. 10674 10675 `(vec_merge:M VEC1 VEC2 ITEMS)' 10676 This describes a merge operation between two vectors. The result 10677 is a vector of mode M; its elements are selected from either VEC1 10678 or VEC2. Which elements are selected is described by ITEMS, which 10679 is a bit mask represented by a `const_int'; a zero bit indicates 10680 the corresponding element in the result vector is taken from VEC2 10681 while a set bit indicates it is taken from VEC1. 10682 10683 `(vec_select:M VEC1 SELECTION)' 10684 This describes an operation that selects parts of a vector. VEC1 10685 is the source vector, SELECTION is a `parallel' that contains a 10686 `const_int' for each of the subparts of the result vector, giving 10687 the number of the source subpart that should be stored into it. 10688 10689 `(vec_concat:M VEC1 VEC2)' 10690 Describes a vector concat operation. The result is a 10691 concatenation of the vectors VEC1 and VEC2; its length is the sum 10692 of the lengths of the two inputs. 10693 10694 `(vec_duplicate:M VEC)' 10695 This operation converts a small vector into a larger one by 10696 duplicating the input values. The output vector mode must have 10697 the same submodes as the input vector mode, and the number of 10698 output parts must be an integer multiple of the number of input 10699 parts. 10700 10701 10702 10703 File: gccint.info, Node: Conversions, Next: RTL Declarations, Prev: Vector Operations, Up: RTL 10704 10705 10.13 Conversions 10706 ================= 10707 10708 All conversions between machine modes must be represented by explicit 10709 conversion operations. For example, an expression which is the sum of 10710 a byte and a full word cannot be written as `(plus:SI (reg:QI 34) 10711 (reg:SI 80))' because the `plus' operation requires two operands of the 10712 same machine mode. Therefore, the byte-sized operand is enclosed in a 10713 conversion operation, as in 10714 10715 (plus:SI (sign_extend:SI (reg:QI 34)) (reg:SI 80)) 10716 10717 The conversion operation is not a mere placeholder, because there may 10718 be more than one way of converting from a given starting mode to the 10719 desired final mode. The conversion operation code says how to do it. 10720 10721 For all conversion operations, X must not be `VOIDmode' because the 10722 mode in which to do the conversion would not be known. The conversion 10723 must either be done at compile-time or X must be placed into a register. 10724 10725 `(sign_extend:M X)' 10726 Represents the result of sign-extending the value X to machine 10727 mode M. M must be a fixed-point mode and X a fixed-point value of 10728 a mode narrower than M. 10729 10730 `(zero_extend:M X)' 10731 Represents the result of zero-extending the value X to machine 10732 mode M. M must be a fixed-point mode and X a fixed-point value of 10733 a mode narrower than M. 10734 10735 `(float_extend:M X)' 10736 Represents the result of extending the value X to machine mode M. 10737 M must be a floating point mode and X a floating point value of a 10738 mode narrower than M. 10739 10740 `(truncate:M X)' 10741 Represents the result of truncating the value X to machine mode M. 10742 M must be a fixed-point mode and X a fixed-point value of a mode 10743 wider than M. 10744 10745 `(ss_truncate:M X)' 10746 Represents the result of truncating the value X to machine mode M, 10747 using signed saturation in the case of overflow. Both M and the 10748 mode of X must be fixed-point modes. 10749 10750 `(us_truncate:M X)' 10751 Represents the result of truncating the value X to machine mode M, 10752 using unsigned saturation in the case of overflow. Both M and the 10753 mode of X must be fixed-point modes. 10754 10755 `(float_truncate:M X)' 10756 Represents the result of truncating the value X to machine mode M. 10757 M must be a floating point mode and X a floating point value of a 10758 mode wider than M. 10759 10760 `(float:M X)' 10761 Represents the result of converting fixed point value X, regarded 10762 as signed, to floating point mode M. 10763 10764 `(unsigned_float:M X)' 10765 Represents the result of converting fixed point value X, regarded 10766 as unsigned, to floating point mode M. 10767 10768 `(fix:M X)' 10769 When M is a floating-point mode, represents the result of 10770 converting floating point value X (valid for mode M) to an 10771 integer, still represented in floating point mode M, by rounding 10772 towards zero. 10773 10774 When M is a fixed-point mode, represents the result of converting 10775 floating point value X to mode M, regarded as signed. How 10776 rounding is done is not specified, so this operation may be used 10777 validly in compiling C code only for integer-valued operands. 10778 10779 `(unsigned_fix:M X)' 10780 Represents the result of converting floating point value X to 10781 fixed point mode M, regarded as unsigned. How rounding is done is 10782 not specified. 10783 10784 `(fract_convert:M X)' 10785 Represents the result of converting fixed-point value X to 10786 fixed-point mode M, signed integer value X to fixed-point mode M, 10787 floating-point value X to fixed-point mode M, fixed-point value X 10788 to integer mode M regarded as signed, or fixed-point value X to 10789 floating-point mode M. When overflows or underflows happen, the 10790 results are undefined. 10791 10792 `(sat_fract:M X)' 10793 Represents the result of converting fixed-point value X to 10794 fixed-point mode M, signed integer value X to fixed-point mode M, 10795 or floating-point value X to fixed-point mode M. When overflows 10796 or underflows happen, the results are saturated to the maximum or 10797 the minimum. 10798 10799 `(unsigned_fract_convert:M X)' 10800 Represents the result of converting fixed-point value X to integer 10801 mode M regarded as unsigned, or unsigned integer value X to 10802 fixed-point mode M. When overflows or underflows happen, the 10803 results are undefined. 10804 10805 `(unsigned_sat_fract:M X)' 10806 Represents the result of converting unsigned integer value X to 10807 fixed-point mode M. When overflows or underflows happen, the 10808 results are saturated to the maximum or the minimum. 10809 10810 10811 File: gccint.info, Node: RTL Declarations, Next: Side Effects, Prev: Conversions, Up: RTL 10812 10813 10.14 Declarations 10814 ================== 10815 10816 Declaration expression codes do not represent arithmetic operations but 10817 rather state assertions about their operands. 10818 10819 `(strict_low_part (subreg:M (reg:N R) 0))' 10820 This expression code is used in only one context: as the 10821 destination operand of a `set' expression. In addition, the 10822 operand of this expression must be a non-paradoxical `subreg' 10823 expression. 10824 10825 The presence of `strict_low_part' says that the part of the 10826 register which is meaningful in mode N, but is not part of mode M, 10827 is not to be altered. Normally, an assignment to such a subreg is 10828 allowed to have undefined effects on the rest of the register when 10829 M is less than a word. 10830 10831 10832 File: gccint.info, Node: Side Effects, Next: Incdec, Prev: RTL Declarations, Up: RTL 10833 10834 10.15 Side Effect Expressions 10835 ============================= 10836 10837 The expression codes described so far represent values, not actions. 10838 But machine instructions never produce values; they are meaningful only 10839 for their side effects on the state of the machine. Special expression 10840 codes are used to represent side effects. 10841 10842 The body of an instruction is always one of these side effect codes; 10843 the codes described above, which represent values, appear only as the 10844 operands of these. 10845 10846 `(set LVAL X)' 10847 Represents the action of storing the value of X into the place 10848 represented by LVAL. LVAL must be an expression representing a 10849 place that can be stored in: `reg' (or `subreg', `strict_low_part' 10850 or `zero_extract'), `mem', `pc', `parallel', or `cc0'. 10851 10852 If LVAL is a `reg', `subreg' or `mem', it has a machine mode; then 10853 X must be valid for that mode. 10854 10855 If LVAL is a `reg' whose machine mode is less than the full width 10856 of the register, then it means that the part of the register 10857 specified by the machine mode is given the specified value and the 10858 rest of the register receives an undefined value. Likewise, if 10859 LVAL is a `subreg' whose machine mode is narrower than the mode of 10860 the register, the rest of the register can be changed in an 10861 undefined way. 10862 10863 If LVAL is a `strict_low_part' of a subreg, then the part of the 10864 register specified by the machine mode of the `subreg' is given 10865 the value X and the rest of the register is not changed. 10866 10867 If LVAL is a `zero_extract', then the referenced part of the 10868 bit-field (a memory or register reference) specified by the 10869 `zero_extract' is given the value X and the rest of the bit-field 10870 is not changed. Note that `sign_extract' can not appear in LVAL. 10871 10872 If LVAL is `(cc0)', it has no machine mode, and X may be either a 10873 `compare' expression or a value that may have any mode. The 10874 latter case represents a "test" instruction. The expression `(set 10875 (cc0) (reg:M N))' is equivalent to `(set (cc0) (compare (reg:M N) 10876 (const_int 0)))'. Use the former expression to save space during 10877 the compilation. 10878 10879 If LVAL is a `parallel', it is used to represent the case of a 10880 function returning a structure in multiple registers. Each element 10881 of the `parallel' is an `expr_list' whose first operand is a `reg' 10882 and whose second operand is a `const_int' representing the offset 10883 (in bytes) into the structure at which the data in that register 10884 corresponds. The first element may be null to indicate that the 10885 structure is also passed partly in memory. 10886 10887 If LVAL is `(pc)', we have a jump instruction, and the 10888 possibilities for X are very limited. It may be a `label_ref' 10889 expression (unconditional jump). It may be an `if_then_else' 10890 (conditional jump), in which case either the second or the third 10891 operand must be `(pc)' (for the case which does not jump) and the 10892 other of the two must be a `label_ref' (for the case which does 10893 jump). X may also be a `mem' or `(plus:SI (pc) Y)', where Y may 10894 be a `reg' or a `mem'; these unusual patterns are used to 10895 represent jumps through branch tables. 10896 10897 If LVAL is neither `(cc0)' nor `(pc)', the mode of LVAL must not 10898 be `VOIDmode' and the mode of X must be valid for the mode of LVAL. 10899 10900 LVAL is customarily accessed with the `SET_DEST' macro and X with 10901 the `SET_SRC' macro. 10902 10903 `(return)' 10904 As the sole expression in a pattern, represents a return from the 10905 current function, on machines where this can be done with one 10906 instruction, such as VAXen. On machines where a multi-instruction 10907 "epilogue" must be executed in order to return from the function, 10908 returning is done by jumping to a label which precedes the 10909 epilogue, and the `return' expression code is never used. 10910 10911 Inside an `if_then_else' expression, represents the value to be 10912 placed in `pc' to return to the caller. 10913 10914 Note that an insn pattern of `(return)' is logically equivalent to 10915 `(set (pc) (return))', but the latter form is never used. 10916 10917 `(call FUNCTION NARGS)' 10918 Represents a function call. FUNCTION is a `mem' expression whose 10919 address is the address of the function to be called. NARGS is an 10920 expression which can be used for two purposes: on some machines it 10921 represents the number of bytes of stack argument; on others, it 10922 represents the number of argument registers. 10923 10924 Each machine has a standard machine mode which FUNCTION must have. 10925 The machine description defines macro `FUNCTION_MODE' to expand 10926 into the requisite mode name. The purpose of this mode is to 10927 specify what kind of addressing is allowed, on machines where the 10928 allowed kinds of addressing depend on the machine mode being 10929 addressed. 10930 10931 `(clobber X)' 10932 Represents the storing or possible storing of an unpredictable, 10933 undescribed value into X, which must be a `reg', `scratch', 10934 `parallel' or `mem' expression. 10935 10936 One place this is used is in string instructions that store 10937 standard values into particular hard registers. It may not be 10938 worth the trouble to describe the values that are stored, but it 10939 is essential to inform the compiler that the registers will be 10940 altered, lest it attempt to keep data in them across the string 10941 instruction. 10942 10943 If X is `(mem:BLK (const_int 0))' or `(mem:BLK (scratch))', it 10944 means that all memory locations must be presumed clobbered. If X 10945 is a `parallel', it has the same meaning as a `parallel' in a 10946 `set' expression. 10947 10948 Note that the machine description classifies certain hard 10949 registers as "call-clobbered". All function call instructions are 10950 assumed by default to clobber these registers, so there is no need 10951 to use `clobber' expressions to indicate this fact. Also, each 10952 function call is assumed to have the potential to alter any memory 10953 location, unless the function is declared `const'. 10954 10955 If the last group of expressions in a `parallel' are each a 10956 `clobber' expression whose arguments are `reg' or `match_scratch' 10957 (*note RTL Template::) expressions, the combiner phase can add the 10958 appropriate `clobber' expressions to an insn it has constructed 10959 when doing so will cause a pattern to be matched. 10960 10961 This feature can be used, for example, on a machine that whose 10962 multiply and add instructions don't use an MQ register but which 10963 has an add-accumulate instruction that does clobber the MQ 10964 register. Similarly, a combined instruction might require a 10965 temporary register while the constituent instructions might not. 10966 10967 When a `clobber' expression for a register appears inside a 10968 `parallel' with other side effects, the register allocator 10969 guarantees that the register is unoccupied both before and after 10970 that insn if it is a hard register clobber. For pseudo-register 10971 clobber, the register allocator and the reload pass do not assign 10972 the same hard register to the clobber and the input operands if 10973 there is an insn alternative containing the `&' constraint (*note 10974 Modifiers::) for the clobber and the hard register is in register 10975 classes of the clobber in the alternative. You can clobber either 10976 a specific hard register, a pseudo register, or a `scratch' 10977 expression; in the latter two cases, GCC will allocate a hard 10978 register that is available there for use as a temporary. 10979 10980 For instructions that require a temporary register, you should use 10981 `scratch' instead of a pseudo-register because this will allow the 10982 combiner phase to add the `clobber' when required. You do this by 10983 coding (`clobber' (`match_scratch' ...)). If you do clobber a 10984 pseudo register, use one which appears nowhere else--generate a 10985 new one each time. Otherwise, you may confuse CSE. 10986 10987 There is one other known use for clobbering a pseudo register in a 10988 `parallel': when one of the input operands of the insn is also 10989 clobbered by the insn. In this case, using the same pseudo 10990 register in the clobber and elsewhere in the insn produces the 10991 expected results. 10992 10993 `(use X)' 10994 Represents the use of the value of X. It indicates that the value 10995 in X at this point in the program is needed, even though it may 10996 not be apparent why this is so. Therefore, the compiler will not 10997 attempt to delete previous instructions whose only effect is to 10998 store a value in X. X must be a `reg' expression. 10999 11000 In some situations, it may be tempting to add a `use' of a 11001 register in a `parallel' to describe a situation where the value 11002 of a special register will modify the behavior of the instruction. 11003 An hypothetical example might be a pattern for an addition that can 11004 either wrap around or use saturating addition depending on the 11005 value of a special control register: 11006 11007 (parallel [(set (reg:SI 2) (unspec:SI [(reg:SI 3) 11008 (reg:SI 4)] 0)) 11009 (use (reg:SI 1))]) 11010 11011 This will not work, several of the optimizers only look at 11012 expressions locally; it is very likely that if you have multiple 11013 insns with identical inputs to the `unspec', they will be 11014 optimized away even if register 1 changes in between. 11015 11016 This means that `use' can _only_ be used to describe that the 11017 register is live. You should think twice before adding `use' 11018 statements, more often you will want to use `unspec' instead. The 11019 `use' RTX is most commonly useful to describe that a fixed 11020 register is implicitly used in an insn. It is also safe to use in 11021 patterns where the compiler knows for other reasons that the result 11022 of the whole pattern is variable, such as `movmemM' or `call' 11023 patterns. 11024 11025 During the reload phase, an insn that has a `use' as pattern can 11026 carry a reg_equal note. These `use' insns will be deleted before 11027 the reload phase exits. 11028 11029 During the delayed branch scheduling phase, X may be an insn. 11030 This indicates that X previously was located at this place in the 11031 code and its data dependencies need to be taken into account. 11032 These `use' insns will be deleted before the delayed branch 11033 scheduling phase exits. 11034 11035 `(parallel [X0 X1 ...])' 11036 Represents several side effects performed in parallel. The square 11037 brackets stand for a vector; the operand of `parallel' is a vector 11038 of expressions. X0, X1 and so on are individual side effect 11039 expressions--expressions of code `set', `call', `return', 11040 `clobber' or `use'. 11041 11042 "In parallel" means that first all the values used in the 11043 individual side-effects are computed, and second all the actual 11044 side-effects are performed. For example, 11045 11046 (parallel [(set (reg:SI 1) (mem:SI (reg:SI 1))) 11047 (set (mem:SI (reg:SI 1)) (reg:SI 1))]) 11048 11049 says unambiguously that the values of hard register 1 and the 11050 memory location addressed by it are interchanged. In both places 11051 where `(reg:SI 1)' appears as a memory address it refers to the 11052 value in register 1 _before_ the execution of the insn. 11053 11054 It follows that it is _incorrect_ to use `parallel' and expect the 11055 result of one `set' to be available for the next one. For 11056 example, people sometimes attempt to represent a jump-if-zero 11057 instruction this way: 11058 11059 (parallel [(set (cc0) (reg:SI 34)) 11060 (set (pc) (if_then_else 11061 (eq (cc0) (const_int 0)) 11062 (label_ref ...) 11063 (pc)))]) 11064 11065 But this is incorrect, because it says that the jump condition 11066 depends on the condition code value _before_ this instruction, not 11067 on the new value that is set by this instruction. 11068 11069 Peephole optimization, which takes place together with final 11070 assembly code output, can produce insns whose patterns consist of 11071 a `parallel' whose elements are the operands needed to output the 11072 resulting assembler code--often `reg', `mem' or constant 11073 expressions. This would not be well-formed RTL at any other stage 11074 in compilation, but it is ok then because no further optimization 11075 remains to be done. However, the definition of the macro 11076 `NOTICE_UPDATE_CC', if any, must deal with such insns if you 11077 define any peephole optimizations. 11078 11079 `(cond_exec [COND EXPR])' 11080 Represents a conditionally executed expression. The EXPR is 11081 executed only if the COND is nonzero. The COND expression must 11082 not have side-effects, but the EXPR may very well have 11083 side-effects. 11084 11085 `(sequence [INSNS ...])' 11086 Represents a sequence of insns. Each of the INSNS that appears in 11087 the vector is suitable for appearing in the chain of insns, so it 11088 must be an `insn', `jump_insn', `call_insn', `code_label', 11089 `barrier' or `note'. 11090 11091 A `sequence' RTX is never placed in an actual insn during RTL 11092 generation. It represents the sequence of insns that result from a 11093 `define_expand' _before_ those insns are passed to `emit_insn' to 11094 insert them in the chain of insns. When actually inserted, the 11095 individual sub-insns are separated out and the `sequence' is 11096 forgotten. 11097 11098 After delay-slot scheduling is completed, an insn and all the 11099 insns that reside in its delay slots are grouped together into a 11100 `sequence'. The insn requiring the delay slot is the first insn 11101 in the vector; subsequent insns are to be placed in the delay slot. 11102 11103 `INSN_ANNULLED_BRANCH_P' is set on an insn in a delay slot to 11104 indicate that a branch insn should be used that will conditionally 11105 annul the effect of the insns in the delay slots. In such a case, 11106 `INSN_FROM_TARGET_P' indicates that the insn is from the target of 11107 the branch and should be executed only if the branch is taken; 11108 otherwise the insn should be executed only if the branch is not 11109 taken. *Note Delay Slots::. 11110 11111 These expression codes appear in place of a side effect, as the body of 11112 an insn, though strictly speaking they do not always describe side 11113 effects as such: 11114 11115 `(asm_input S)' 11116 Represents literal assembler code as described by the string S. 11117 11118 `(unspec [OPERANDS ...] INDEX)' 11119 `(unspec_volatile [OPERANDS ...] INDEX)' 11120 Represents a machine-specific operation on OPERANDS. INDEX 11121 selects between multiple machine-specific operations. 11122 `unspec_volatile' is used for volatile operations and operations 11123 that may trap; `unspec' is used for other operations. 11124 11125 These codes may appear inside a `pattern' of an insn, inside a 11126 `parallel', or inside an expression. 11127 11128 `(addr_vec:M [LR0 LR1 ...])' 11129 Represents a table of jump addresses. The vector elements LR0, 11130 etc., are `label_ref' expressions. The mode M specifies how much 11131 space is given to each address; normally M would be `Pmode'. 11132 11133 `(addr_diff_vec:M BASE [LR0 LR1 ...] MIN MAX FLAGS)' 11134 Represents a table of jump addresses expressed as offsets from 11135 BASE. The vector elements LR0, etc., are `label_ref' expressions 11136 and so is BASE. The mode M specifies how much space is given to 11137 each address-difference. MIN and MAX are set up by branch 11138 shortening and hold a label with a minimum and a maximum address, 11139 respectively. FLAGS indicates the relative position of BASE, MIN 11140 and MAX to the containing insn and of MIN and MAX to BASE. See 11141 rtl.def for details. 11142 11143 `(prefetch:M ADDR RW LOCALITY)' 11144 Represents prefetch of memory at address ADDR. Operand RW is 1 if 11145 the prefetch is for data to be written, 0 otherwise; targets that 11146 do not support write prefetches should treat this as a normal 11147 prefetch. Operand LOCALITY specifies the amount of temporal 11148 locality; 0 if there is none or 1, 2, or 3 for increasing levels 11149 of temporal locality; targets that do not support locality hints 11150 should ignore this. 11151 11152 This insn is used to minimize cache-miss latency by moving data 11153 into a cache before it is accessed. It should use only 11154 non-faulting data prefetch instructions. 11155 11156 11157 File: gccint.info, Node: Incdec, Next: Assembler, Prev: Side Effects, Up: RTL 11158 11159 10.16 Embedded Side-Effects on Addresses 11160 ======================================== 11161 11162 Six special side-effect expression codes appear as memory addresses. 11163 11164 `(pre_dec:M X)' 11165 Represents the side effect of decrementing X by a standard amount 11166 and represents also the value that X has after being decremented. 11167 X must be a `reg' or `mem', but most machines allow only a `reg'. 11168 M must be the machine mode for pointers on the machine in use. 11169 The amount X is decremented by is the length in bytes of the 11170 machine mode of the containing memory reference of which this 11171 expression serves as the address. Here is an example of its use: 11172 11173 (mem:DF (pre_dec:SI (reg:SI 39))) 11174 11175 This says to decrement pseudo register 39 by the length of a 11176 `DFmode' value and use the result to address a `DFmode' value. 11177 11178 `(pre_inc:M X)' 11179 Similar, but specifies incrementing X instead of decrementing it. 11180 11181 `(post_dec:M X)' 11182 Represents the same side effect as `pre_dec' but a different 11183 value. The value represented here is the value X has before being 11184 decremented. 11185 11186 `(post_inc:M X)' 11187 Similar, but specifies incrementing X instead of decrementing it. 11188 11189 `(post_modify:M X Y)' 11190 Represents the side effect of setting X to Y and represents X 11191 before X is modified. X must be a `reg' or `mem', but most 11192 machines allow only a `reg'. M must be the machine mode for 11193 pointers on the machine in use. 11194 11195 The expression Y must be one of three forms: `(plus:M X Z)', 11196 `(minus:M X Z)', or `(plus:M X I)', where Z is an index register 11197 and I is a constant. 11198 11199 Here is an example of its use: 11200 11201 (mem:SF (post_modify:SI (reg:SI 42) (plus (reg:SI 42) 11202 (reg:SI 48)))) 11203 11204 This says to modify pseudo register 42 by adding the contents of 11205 pseudo register 48 to it, after the use of what ever 42 points to. 11206 11207 `(pre_modify:M X EXPR)' 11208 Similar except side effects happen before the use. 11209 11210 These embedded side effect expressions must be used with care. 11211 Instruction patterns may not use them. Until the `flow' pass of the 11212 compiler, they may occur only to represent pushes onto the stack. The 11213 `flow' pass finds cases where registers are incremented or decremented 11214 in one instruction and used as an address shortly before or after; 11215 these cases are then transformed to use pre- or post-increment or 11216 -decrement. 11217 11218 If a register used as the operand of these expressions is used in 11219 another address in an insn, the original value of the register is used. 11220 Uses of the register outside of an address are not permitted within the 11221 same insn as a use in an embedded side effect expression because such 11222 insns behave differently on different machines and hence must be treated 11223 as ambiguous and disallowed. 11224 11225 An instruction that can be represented with an embedded side effect 11226 could also be represented using `parallel' containing an additional 11227 `set' to describe how the address register is altered. This is not 11228 done because machines that allow these operations at all typically 11229 allow them wherever a memory address is called for. Describing them as 11230 additional parallel stores would require doubling the number of entries 11231 in the machine description. 11232 11233 11234 File: gccint.info, Node: Assembler, Next: Insns, Prev: Incdec, Up: RTL 11235 11236 10.17 Assembler Instructions as Expressions 11237 =========================================== 11238 11239 The RTX code `asm_operands' represents a value produced by a 11240 user-specified assembler instruction. It is used to represent an `asm' 11241 statement with arguments. An `asm' statement with a single output 11242 operand, like this: 11243 11244 asm ("foo %1,%2,%0" : "=a" (outputvar) : "g" (x + y), "di" (*z)); 11245 11246 is represented using a single `asm_operands' RTX which represents the 11247 value that is stored in `outputvar': 11248 11249 (set RTX-FOR-OUTPUTVAR 11250 (asm_operands "foo %1,%2,%0" "a" 0 11251 [RTX-FOR-ADDITION-RESULT RTX-FOR-*Z] 11252 [(asm_input:M1 "g") 11253 (asm_input:M2 "di")])) 11254 11255 Here the operands of the `asm_operands' RTX are the assembler template 11256 string, the output-operand's constraint, the index-number of the output 11257 operand among the output operands specified, a vector of input operand 11258 RTX's, and a vector of input-operand modes and constraints. The mode 11259 M1 is the mode of the sum `x+y'; M2 is that of `*z'. 11260 11261 When an `asm' statement has multiple output values, its insn has 11262 several such `set' RTX's inside of a `parallel'. Each `set' contains a 11263 `asm_operands'; all of these share the same assembler template and 11264 vectors, but each contains the constraint for the respective output 11265 operand. They are also distinguished by the output-operand index 11266 number, which is 0, 1, ... for successive output operands. 11267 11268 11269 File: gccint.info, Node: Insns, Next: Calls, Prev: Assembler, Up: RTL 11270 11271 10.18 Insns 11272 =========== 11273 11274 The RTL representation of the code for a function is a doubly-linked 11275 chain of objects called "insns". Insns are expressions with special 11276 codes that are used for no other purpose. Some insns are actual 11277 instructions; others represent dispatch tables for `switch' statements; 11278 others represent labels to jump to or various sorts of declarative 11279 information. 11280 11281 In addition to its own specific data, each insn must have a unique 11282 id-number that distinguishes it from all other insns in the current 11283 function (after delayed branch scheduling, copies of an insn with the 11284 same id-number may be present in multiple places in a function, but 11285 these copies will always be identical and will only appear inside a 11286 `sequence'), and chain pointers to the preceding and following insns. 11287 These three fields occupy the same position in every insn, independent 11288 of the expression code of the insn. They could be accessed with `XEXP' 11289 and `XINT', but instead three special macros are always used: 11290 11291 `INSN_UID (I)' 11292 Accesses the unique id of insn I. 11293 11294 `PREV_INSN (I)' 11295 Accesses the chain pointer to the insn preceding I. If I is the 11296 first insn, this is a null pointer. 11297 11298 `NEXT_INSN (I)' 11299 Accesses the chain pointer to the insn following I. If I is the 11300 last insn, this is a null pointer. 11301 11302 The first insn in the chain is obtained by calling `get_insns'; the 11303 last insn is the result of calling `get_last_insn'. Within the chain 11304 delimited by these insns, the `NEXT_INSN' and `PREV_INSN' pointers must 11305 always correspond: if INSN is not the first insn, 11306 11307 NEXT_INSN (PREV_INSN (INSN)) == INSN 11308 11309 is always true and if INSN is not the last insn, 11310 11311 PREV_INSN (NEXT_INSN (INSN)) == INSN 11312 11313 is always true. 11314 11315 After delay slot scheduling, some of the insns in the chain might be 11316 `sequence' expressions, which contain a vector of insns. The value of 11317 `NEXT_INSN' in all but the last of these insns is the next insn in the 11318 vector; the value of `NEXT_INSN' of the last insn in the vector is the 11319 same as the value of `NEXT_INSN' for the `sequence' in which it is 11320 contained. Similar rules apply for `PREV_INSN'. 11321 11322 This means that the above invariants are not necessarily true for insns 11323 inside `sequence' expressions. Specifically, if INSN is the first insn 11324 in a `sequence', `NEXT_INSN (PREV_INSN (INSN))' is the insn containing 11325 the `sequence' expression, as is the value of `PREV_INSN (NEXT_INSN 11326 (INSN))' if INSN is the last insn in the `sequence' expression. You 11327 can use these expressions to find the containing `sequence' expression. 11328 11329 Every insn has one of the following six expression codes: 11330 11331 `insn' 11332 The expression code `insn' is used for instructions that do not 11333 jump and do not do function calls. `sequence' expressions are 11334 always contained in insns with code `insn' even if one of those 11335 insns should jump or do function calls. 11336 11337 Insns with code `insn' have four additional fields beyond the three 11338 mandatory ones listed above. These four are described in a table 11339 below. 11340 11341 `jump_insn' 11342 The expression code `jump_insn' is used for instructions that may 11343 jump (or, more generally, may contain `label_ref' expressions to 11344 which `pc' can be set in that instruction). If there is an 11345 instruction to return from the current function, it is recorded as 11346 a `jump_insn'. 11347 11348 `jump_insn' insns have the same extra fields as `insn' insns, 11349 accessed in the same way and in addition contain a field 11350 `JUMP_LABEL' which is defined once jump optimization has completed. 11351 11352 For simple conditional and unconditional jumps, this field contains 11353 the `code_label' to which this insn will (possibly conditionally) 11354 branch. In a more complex jump, `JUMP_LABEL' records one of the 11355 labels that the insn refers to; other jump target labels are 11356 recorded as `REG_LABEL_TARGET' notes. The exception is `addr_vec' 11357 and `addr_diff_vec', where `JUMP_LABEL' is `NULL_RTX' and the only 11358 way to find the labels is to scan the entire body of the insn. 11359 11360 Return insns count as jumps, but since they do not refer to any 11361 labels, their `JUMP_LABEL' is `NULL_RTX'. 11362 11363 `call_insn' 11364 The expression code `call_insn' is used for instructions that may 11365 do function calls. It is important to distinguish these 11366 instructions because they imply that certain registers and memory 11367 locations may be altered unpredictably. 11368 11369 `call_insn' insns have the same extra fields as `insn' insns, 11370 accessed in the same way and in addition contain a field 11371 `CALL_INSN_FUNCTION_USAGE', which contains a list (chain of 11372 `expr_list' expressions) containing `use' and `clobber' 11373 expressions that denote hard registers and `MEM's used or 11374 clobbered by the called function. 11375 11376 A `MEM' generally points to a stack slots in which arguments passed 11377 to the libcall by reference (*note TARGET_PASS_BY_REFERENCE: 11378 Register Arguments.) are stored. If the argument is caller-copied 11379 (*note TARGET_CALLEE_COPIES: Register Arguments.), the stack slot 11380 will be mentioned in `CLOBBER' and `USE' entries; if it's 11381 callee-copied, only a `USE' will appear, and the `MEM' may point 11382 to addresses that are not stack slots. 11383 11384 `CLOBBER'ed registers in this list augment registers specified in 11385 `CALL_USED_REGISTERS' (*note Register Basics::). 11386 11387 `code_label' 11388 A `code_label' insn represents a label that a jump insn can jump 11389 to. It contains two special fields of data in addition to the 11390 three standard ones. `CODE_LABEL_NUMBER' is used to hold the 11391 "label number", a number that identifies this label uniquely among 11392 all the labels in the compilation (not just in the current 11393 function). Ultimately, the label is represented in the assembler 11394 output as an assembler label, usually of the form `LN' where N is 11395 the label number. 11396 11397 When a `code_label' appears in an RTL expression, it normally 11398 appears within a `label_ref' which represents the address of the 11399 label, as a number. 11400 11401 Besides as a `code_label', a label can also be represented as a 11402 `note' of type `NOTE_INSN_DELETED_LABEL'. 11403 11404 The field `LABEL_NUSES' is only defined once the jump optimization 11405 phase is completed. It contains the number of times this label is 11406 referenced in the current function. 11407 11408 The field `LABEL_KIND' differentiates four different types of 11409 labels: `LABEL_NORMAL', `LABEL_STATIC_ENTRY', 11410 `LABEL_GLOBAL_ENTRY', and `LABEL_WEAK_ENTRY'. The only labels 11411 that do not have type `LABEL_NORMAL' are "alternate entry points" 11412 to the current function. These may be static (visible only in the 11413 containing translation unit), global (exposed to all translation 11414 units), or weak (global, but can be overridden by another symbol 11415 with the same name). 11416 11417 Much of the compiler treats all four kinds of label identically. 11418 Some of it needs to know whether or not a label is an alternate 11419 entry point; for this purpose, the macro `LABEL_ALT_ENTRY_P' is 11420 provided. It is equivalent to testing whether `LABEL_KIND (label) 11421 == LABEL_NORMAL'. The only place that cares about the distinction 11422 between static, global, and weak alternate entry points, besides 11423 the front-end code that creates them, is the function 11424 `output_alternate_entry_point', in `final.c'. 11425 11426 To set the kind of a label, use the `SET_LABEL_KIND' macro. 11427 11428 `barrier' 11429 Barriers are placed in the instruction stream when control cannot 11430 flow past them. They are placed after unconditional jump 11431 instructions to indicate that the jumps are unconditional and 11432 after calls to `volatile' functions, which do not return (e.g., 11433 `exit'). They contain no information beyond the three standard 11434 fields. 11435 11436 `note' 11437 `note' insns are used to represent additional debugging and 11438 declarative information. They contain two nonstandard fields, an 11439 integer which is accessed with the macro `NOTE_LINE_NUMBER' and a 11440 string accessed with `NOTE_SOURCE_FILE'. 11441 11442 If `NOTE_LINE_NUMBER' is positive, the note represents the 11443 position of a source line and `NOTE_SOURCE_FILE' is the source 11444 file name that the line came from. These notes control generation 11445 of line number data in the assembler output. 11446 11447 Otherwise, `NOTE_LINE_NUMBER' is not really a line number but a 11448 code with one of the following values (and `NOTE_SOURCE_FILE' must 11449 contain a null pointer): 11450 11451 `NOTE_INSN_DELETED' 11452 Such a note is completely ignorable. Some passes of the 11453 compiler delete insns by altering them into notes of this 11454 kind. 11455 11456 `NOTE_INSN_DELETED_LABEL' 11457 This marks what used to be a `code_label', but was not used 11458 for other purposes than taking its address and was 11459 transformed to mark that no code jumps to it. 11460 11461 `NOTE_INSN_BLOCK_BEG' 11462 `NOTE_INSN_BLOCK_END' 11463 These types of notes indicate the position of the beginning 11464 and end of a level of scoping of variable names. They 11465 control the output of debugging information. 11466 11467 `NOTE_INSN_EH_REGION_BEG' 11468 `NOTE_INSN_EH_REGION_END' 11469 These types of notes indicate the position of the beginning 11470 and end of a level of scoping for exception handling. 11471 `NOTE_BLOCK_NUMBER' identifies which `CODE_LABEL' or `note' 11472 of type `NOTE_INSN_DELETED_LABEL' is associated with the 11473 given region. 11474 11475 `NOTE_INSN_LOOP_BEG' 11476 `NOTE_INSN_LOOP_END' 11477 These types of notes indicate the position of the beginning 11478 and end of a `while' or `for' loop. They enable the loop 11479 optimizer to find loops quickly. 11480 11481 `NOTE_INSN_LOOP_CONT' 11482 Appears at the place in a loop that `continue' statements 11483 jump to. 11484 11485 `NOTE_INSN_LOOP_VTOP' 11486 This note indicates the place in a loop where the exit test 11487 begins for those loops in which the exit test has been 11488 duplicated. This position becomes another virtual start of 11489 the loop when considering loop invariants. 11490 11491 `NOTE_INSN_FUNCTION_BEG' 11492 Appears at the start of the function body, after the function 11493 prologue. 11494 11495 11496 These codes are printed symbolically when they appear in debugging 11497 dumps. 11498 11499 The machine mode of an insn is normally `VOIDmode', but some phases 11500 use the mode for various purposes. 11501 11502 The common subexpression elimination pass sets the mode of an insn to 11503 `QImode' when it is the first insn in a block that has already been 11504 processed. 11505 11506 The second Haifa scheduling pass, for targets that can multiple issue, 11507 sets the mode of an insn to `TImode' when it is believed that the 11508 instruction begins an issue group. That is, when the instruction 11509 cannot issue simultaneously with the previous. This may be relied on 11510 by later passes, in particular machine-dependent reorg. 11511 11512 Here is a table of the extra fields of `insn', `jump_insn' and 11513 `call_insn' insns: 11514 11515 `PATTERN (I)' 11516 An expression for the side effect performed by this insn. This 11517 must be one of the following codes: `set', `call', `use', 11518 `clobber', `return', `asm_input', `asm_output', `addr_vec', 11519 `addr_diff_vec', `trap_if', `unspec', `unspec_volatile', 11520 `parallel', `cond_exec', or `sequence'. If it is a `parallel', 11521 each element of the `parallel' must be one these codes, except that 11522 `parallel' expressions cannot be nested and `addr_vec' and 11523 `addr_diff_vec' are not permitted inside a `parallel' expression. 11524 11525 `INSN_CODE (I)' 11526 An integer that says which pattern in the machine description 11527 matches this insn, or -1 if the matching has not yet been 11528 attempted. 11529 11530 Such matching is never attempted and this field remains -1 on an 11531 insn whose pattern consists of a single `use', `clobber', 11532 `asm_input', `addr_vec' or `addr_diff_vec' expression. 11533 11534 Matching is also never attempted on insns that result from an `asm' 11535 statement. These contain at least one `asm_operands' expression. 11536 The function `asm_noperands' returns a non-negative value for such 11537 insns. 11538 11539 In the debugging output, this field is printed as a number 11540 followed by a symbolic representation that locates the pattern in 11541 the `md' file as some small positive or negative offset from a 11542 named pattern. 11543 11544 `LOG_LINKS (I)' 11545 A list (chain of `insn_list' expressions) giving information about 11546 dependencies between instructions within a basic block. Neither a 11547 jump nor a label may come between the related insns. These are 11548 only used by the schedulers and by combine. This is a deprecated 11549 data structure. Def-use and use-def chains are now preferred. 11550 11551 `REG_NOTES (I)' 11552 A list (chain of `expr_list' and `insn_list' expressions) giving 11553 miscellaneous information about the insn. It is often information 11554 pertaining to the registers used in this insn. 11555 11556 The `LOG_LINKS' field of an insn is a chain of `insn_list' 11557 expressions. Each of these has two operands: the first is an insn, and 11558 the second is another `insn_list' expression (the next one in the 11559 chain). The last `insn_list' in the chain has a null pointer as second 11560 operand. The significant thing about the chain is which insns appear 11561 in it (as first operands of `insn_list' expressions). Their order is 11562 not significant. 11563 11564 This list is originally set up by the flow analysis pass; it is a null 11565 pointer until then. Flow only adds links for those data dependencies 11566 which can be used for instruction combination. For each insn, the flow 11567 analysis pass adds a link to insns which store into registers values 11568 that are used for the first time in this insn. 11569 11570 The `REG_NOTES' field of an insn is a chain similar to the `LOG_LINKS' 11571 field but it includes `expr_list' expressions in addition to 11572 `insn_list' expressions. There are several kinds of register notes, 11573 which are distinguished by the machine mode, which in a register note 11574 is really understood as being an `enum reg_note'. The first operand OP 11575 of the note is data whose meaning depends on the kind of note. 11576 11577 The macro `REG_NOTE_KIND (X)' returns the kind of register note. Its 11578 counterpart, the macro `PUT_REG_NOTE_KIND (X, NEWKIND)' sets the 11579 register note type of X to be NEWKIND. 11580 11581 Register notes are of three classes: They may say something about an 11582 input to an insn, they may say something about an output of an insn, or 11583 they may create a linkage between two insns. There are also a set of 11584 values that are only used in `LOG_LINKS'. 11585 11586 These register notes annotate inputs to an insn: 11587 11588 `REG_DEAD' 11589 The value in OP dies in this insn; that is to say, altering the 11590 value immediately after this insn would not affect the future 11591 behavior of the program. 11592 11593 It does not follow that the register OP has no useful value after 11594 this insn since OP is not necessarily modified by this insn. 11595 Rather, no subsequent instruction uses the contents of OP. 11596 11597 `REG_UNUSED' 11598 The register OP being set by this insn will not be used in a 11599 subsequent insn. This differs from a `REG_DEAD' note, which 11600 indicates that the value in an input will not be used subsequently. 11601 These two notes are independent; both may be present for the same 11602 register. 11603 11604 `REG_INC' 11605 The register OP is incremented (or decremented; at this level 11606 there is no distinction) by an embedded side effect inside this 11607 insn. This means it appears in a `post_inc', `pre_inc', 11608 `post_dec' or `pre_dec' expression. 11609 11610 `REG_NONNEG' 11611 The register OP is known to have a nonnegative value when this 11612 insn is reached. This is used so that decrement and branch until 11613 zero instructions, such as the m68k dbra, can be matched. 11614 11615 The `REG_NONNEG' note is added to insns only if the machine 11616 description has a `decrement_and_branch_until_zero' pattern. 11617 11618 `REG_LABEL_OPERAND' 11619 This insn uses OP, a `code_label' or a `note' of type 11620 `NOTE_INSN_DELETED_LABEL', but is not a `jump_insn', or it is a 11621 `jump_insn' that refers to the operand as an ordinary operand. 11622 The label may still eventually be a jump target, but if so in an 11623 indirect jump in a subsequent insn. The presence of this note 11624 allows jump optimization to be aware that OP is, in fact, being 11625 used, and flow optimization to build an accurate flow graph. 11626 11627 `REG_LABEL_TARGET' 11628 This insn is a `jump_insn' but not a `addr_vec' or 11629 `addr_diff_vec'. It uses OP, a `code_label' as a direct or 11630 indirect jump target. Its purpose is similar to that of 11631 `REG_LABEL_OPERAND'. This note is only present if the insn has 11632 multiple targets; the last label in the insn (in the highest 11633 numbered insn-field) goes into the `JUMP_LABEL' field and does not 11634 have a `REG_LABEL_TARGET' note. *Note JUMP_LABEL: Insns. 11635 11636 `REG_CROSSING_JUMP' 11637 This insn is an branching instruction (either an unconditional 11638 jump or an indirect jump) which crosses between hot and cold 11639 sections, which could potentially be very far apart in the 11640 executable. The presence of this note indicates to other 11641 optimizations that this branching instruction should not be 11642 "collapsed" into a simpler branching construct. It is used when 11643 the optimization to partition basic blocks into hot and cold 11644 sections is turned on. 11645 11646 `REG_SETJMP' 11647 Appears attached to each `CALL_INSN' to `setjmp' or a related 11648 function. 11649 11650 The following notes describe attributes of outputs of an insn: 11651 11652 `REG_EQUIV' 11653 `REG_EQUAL' 11654 This note is only valid on an insn that sets only one register and 11655 indicates that that register will be equal to OP at run time; the 11656 scope of this equivalence differs between the two types of notes. 11657 The value which the insn explicitly copies into the register may 11658 look different from OP, but they will be equal at run time. If the 11659 output of the single `set' is a `strict_low_part' expression, the 11660 note refers to the register that is contained in `SUBREG_REG' of 11661 the `subreg' expression. 11662 11663 For `REG_EQUIV', the register is equivalent to OP throughout the 11664 entire function, and could validly be replaced in all its 11665 occurrences by OP. ("Validly" here refers to the data flow of the 11666 program; simple replacement may make some insns invalid.) For 11667 example, when a constant is loaded into a register that is never 11668 assigned any other value, this kind of note is used. 11669 11670 When a parameter is copied into a pseudo-register at entry to a 11671 function, a note of this kind records that the register is 11672 equivalent to the stack slot where the parameter was passed. 11673 Although in this case the register may be set by other insns, it 11674 is still valid to replace the register by the stack slot 11675 throughout the function. 11676 11677 A `REG_EQUIV' note is also used on an instruction which copies a 11678 register parameter into a pseudo-register at entry to a function, 11679 if there is a stack slot where that parameter could be stored. 11680 Although other insns may set the pseudo-register, it is valid for 11681 the compiler to replace the pseudo-register by stack slot 11682 throughout the function, provided the compiler ensures that the 11683 stack slot is properly initialized by making the replacement in 11684 the initial copy instruction as well. This is used on machines 11685 for which the calling convention allocates stack space for 11686 register parameters. See `REG_PARM_STACK_SPACE' in *note Stack 11687 Arguments::. 11688 11689 In the case of `REG_EQUAL', the register that is set by this insn 11690 will be equal to OP at run time at the end of this insn but not 11691 necessarily elsewhere in the function. In this case, OP is 11692 typically an arithmetic expression. For example, when a sequence 11693 of insns such as a library call is used to perform an arithmetic 11694 operation, this kind of note is attached to the insn that produces 11695 or copies the final value. 11696 11697 These two notes are used in different ways by the compiler passes. 11698 `REG_EQUAL' is used by passes prior to register allocation (such as 11699 common subexpression elimination and loop optimization) to tell 11700 them how to think of that value. `REG_EQUIV' notes are used by 11701 register allocation to indicate that there is an available 11702 substitute expression (either a constant or a `mem' expression for 11703 the location of a parameter on the stack) that may be used in 11704 place of a register if insufficient registers are available. 11705 11706 Except for stack homes for parameters, which are indicated by a 11707 `REG_EQUIV' note and are not useful to the early optimization 11708 passes and pseudo registers that are equivalent to a memory 11709 location throughout their entire life, which is not detected until 11710 later in the compilation, all equivalences are initially indicated 11711 by an attached `REG_EQUAL' note. In the early stages of register 11712 allocation, a `REG_EQUAL' note is changed into a `REG_EQUIV' note 11713 if OP is a constant and the insn represents the only set of its 11714 destination register. 11715 11716 Thus, compiler passes prior to register allocation need only check 11717 for `REG_EQUAL' notes and passes subsequent to register allocation 11718 need only check for `REG_EQUIV' notes. 11719 11720 These notes describe linkages between insns. They occur in pairs: one 11721 insn has one of a pair of notes that points to a second insn, which has 11722 the inverse note pointing back to the first insn. 11723 11724 `REG_CC_SETTER' 11725 `REG_CC_USER' 11726 On machines that use `cc0', the insns which set and use `cc0' set 11727 and use `cc0' are adjacent. However, when branch delay slot 11728 filling is done, this may no longer be true. In this case a 11729 `REG_CC_USER' note will be placed on the insn setting `cc0' to 11730 point to the insn using `cc0' and a `REG_CC_SETTER' note will be 11731 placed on the insn using `cc0' to point to the insn setting `cc0'. 11732 11733 These values are only used in the `LOG_LINKS' field, and indicate the 11734 type of dependency that each link represents. Links which indicate a 11735 data dependence (a read after write dependence) do not use any code, 11736 they simply have mode `VOIDmode', and are printed without any 11737 descriptive text. 11738 11739 `REG_DEP_TRUE' 11740 This indicates a true dependence (a read after write dependence). 11741 11742 `REG_DEP_OUTPUT' 11743 This indicates an output dependence (a write after write 11744 dependence). 11745 11746 `REG_DEP_ANTI' 11747 This indicates an anti dependence (a write after read dependence). 11748 11749 11750 These notes describe information gathered from gcov profile data. They 11751 are stored in the `REG_NOTES' field of an insn as an `expr_list'. 11752 11753 `REG_BR_PROB' 11754 This is used to specify the ratio of branches to non-branches of a 11755 branch insn according to the profile data. The value is stored as 11756 a value between 0 and REG_BR_PROB_BASE; larger values indicate a 11757 higher probability that the branch will be taken. 11758 11759 `REG_BR_PRED' 11760 These notes are found in JUMP insns after delayed branch scheduling 11761 has taken place. They indicate both the direction and the 11762 likelihood of the JUMP. The format is a bitmask of ATTR_FLAG_* 11763 values. 11764 11765 `REG_FRAME_RELATED_EXPR' 11766 This is used on an RTX_FRAME_RELATED_P insn wherein the attached 11767 expression is used in place of the actual insn pattern. This is 11768 done in cases where the pattern is either complex or misleading. 11769 11770 For convenience, the machine mode in an `insn_list' or `expr_list' is 11771 printed using these symbolic codes in debugging dumps. 11772 11773 The only difference between the expression codes `insn_list' and 11774 `expr_list' is that the first operand of an `insn_list' is assumed to 11775 be an insn and is printed in debugging dumps as the insn's unique id; 11776 the first operand of an `expr_list' is printed in the ordinary way as 11777 an expression. 11778 11779 11780 File: gccint.info, Node: Calls, Next: Sharing, Prev: Insns, Up: RTL 11781 11782 10.19 RTL Representation of Function-Call Insns 11783 =============================================== 11784 11785 Insns that call subroutines have the RTL expression code `call_insn'. 11786 These insns must satisfy special rules, and their bodies must use a 11787 special RTL expression code, `call'. 11788 11789 A `call' expression has two operands, as follows: 11790 11791 (call (mem:FM ADDR) NBYTES) 11792 11793 Here NBYTES is an operand that represents the number of bytes of 11794 argument data being passed to the subroutine, FM is a machine mode 11795 (which must equal as the definition of the `FUNCTION_MODE' macro in the 11796 machine description) and ADDR represents the address of the subroutine. 11797 11798 For a subroutine that returns no value, the `call' expression as shown 11799 above is the entire body of the insn, except that the insn might also 11800 contain `use' or `clobber' expressions. 11801 11802 For a subroutine that returns a value whose mode is not `BLKmode', the 11803 value is returned in a hard register. If this register's number is R, 11804 then the body of the call insn looks like this: 11805 11806 (set (reg:M R) 11807 (call (mem:FM ADDR) NBYTES)) 11808 11809 This RTL expression makes it clear (to the optimizer passes) that the 11810 appropriate register receives a useful value in this insn. 11811 11812 When a subroutine returns a `BLKmode' value, it is handled by passing 11813 to the subroutine the address of a place to store the value. So the 11814 call insn itself does not "return" any value, and it has the same RTL 11815 form as a call that returns nothing. 11816 11817 On some machines, the call instruction itself clobbers some register, 11818 for example to contain the return address. `call_insn' insns on these 11819 machines should have a body which is a `parallel' that contains both 11820 the `call' expression and `clobber' expressions that indicate which 11821 registers are destroyed. Similarly, if the call instruction requires 11822 some register other than the stack pointer that is not explicitly 11823 mentioned in its RTL, a `use' subexpression should mention that 11824 register. 11825 11826 Functions that are called are assumed to modify all registers listed in 11827 the configuration macro `CALL_USED_REGISTERS' (*note Register Basics::) 11828 and, with the exception of `const' functions and library calls, to 11829 modify all of memory. 11830 11831 Insns containing just `use' expressions directly precede the 11832 `call_insn' insn to indicate which registers contain inputs to the 11833 function. Similarly, if registers other than those in 11834 `CALL_USED_REGISTERS' are clobbered by the called function, insns 11835 containing a single `clobber' follow immediately after the call to 11836 indicate which registers. 11837 11838 11839 File: gccint.info, Node: Sharing, Next: Reading RTL, Prev: Calls, Up: RTL 11840 11841 10.20 Structure Sharing Assumptions 11842 =================================== 11843 11844 The compiler assumes that certain kinds of RTL expressions are unique; 11845 there do not exist two distinct objects representing the same value. 11846 In other cases, it makes an opposite assumption: that no RTL expression 11847 object of a certain kind appears in more than one place in the 11848 containing structure. 11849 11850 These assumptions refer to a single function; except for the RTL 11851 objects that describe global variables and external functions, and a 11852 few standard objects such as small integer constants, no RTL objects 11853 are common to two functions. 11854 11855 * Each pseudo-register has only a single `reg' object to represent 11856 it, and therefore only a single machine mode. 11857 11858 * For any symbolic label, there is only one `symbol_ref' object 11859 referring to it. 11860 11861 * All `const_int' expressions with equal values are shared. 11862 11863 * There is only one `pc' expression. 11864 11865 * There is only one `cc0' expression. 11866 11867 * There is only one `const_double' expression with value 0 for each 11868 floating point mode. Likewise for values 1 and 2. 11869 11870 * There is only one `const_vector' expression with value 0 for each 11871 vector mode, be it an integer or a double constant vector. 11872 11873 * No `label_ref' or `scratch' appears in more than one place in the 11874 RTL structure; in other words, it is safe to do a tree-walk of all 11875 the insns in the function and assume that each time a `label_ref' 11876 or `scratch' is seen it is distinct from all others that are seen. 11877 11878 * Only one `mem' object is normally created for each static variable 11879 or stack slot, so these objects are frequently shared in all the 11880 places they appear. However, separate but equal objects for these 11881 variables are occasionally made. 11882 11883 * When a single `asm' statement has multiple output operands, a 11884 distinct `asm_operands' expression is made for each output operand. 11885 However, these all share the vector which contains the sequence of 11886 input operands. This sharing is used later on to test whether two 11887 `asm_operands' expressions come from the same statement, so all 11888 optimizations must carefully preserve the sharing if they copy the 11889 vector at all. 11890 11891 * No RTL object appears in more than one place in the RTL structure 11892 except as described above. Many passes of the compiler rely on 11893 this by assuming that they can modify RTL objects in place without 11894 unwanted side-effects on other insns. 11895 11896 * During initial RTL generation, shared structure is freely 11897 introduced. After all the RTL for a function has been generated, 11898 all shared structure is copied by `unshare_all_rtl' in 11899 `emit-rtl.c', after which the above rules are guaranteed to be 11900 followed. 11901 11902 * During the combiner pass, shared structure within an insn can exist 11903 temporarily. However, the shared structure is copied before the 11904 combiner is finished with the insn. This is done by calling 11905 `copy_rtx_if_shared', which is a subroutine of `unshare_all_rtl'. 11906 11907 11908 File: gccint.info, Node: Reading RTL, Prev: Sharing, Up: RTL 11909 11910 10.21 Reading RTL 11911 ================= 11912 11913 To read an RTL object from a file, call `read_rtx'. It takes one 11914 argument, a stdio stream, and returns a single RTL object. This routine 11915 is defined in `read-rtl.c'. It is not available in the compiler 11916 itself, only the various programs that generate the compiler back end 11917 from the machine description. 11918 11919 People frequently have the idea of using RTL stored as text in a file 11920 as an interface between a language front end and the bulk of GCC. This 11921 idea is not feasible. 11922 11923 GCC was designed to use RTL internally only. Correct RTL for a given 11924 program is very dependent on the particular target machine. And the RTL 11925 does not contain all the information about the program. 11926 11927 The proper way to interface GCC to a new language front end is with 11928 the "tree" data structure, described in the files `tree.h' and 11929 `tree.def'. The documentation for this structure (*note Trees::) is 11930 incomplete. 11931 11932 11933 File: gccint.info, Node: GENERIC, Next: GIMPLE, Prev: Trees, Up: Top 11934 11935 11 GENERIC 11936 ********** 11937 11938 The purpose of GENERIC is simply to provide a language-independent way 11939 of representing an entire function in trees. To this end, it was 11940 necessary to add a few new tree codes to the back end, but most 11941 everything was already there. If you can express it with the codes in 11942 `gcc/tree.def', it's GENERIC. 11943 11944 Early on, there was a great deal of debate about how to think about 11945 statements in a tree IL. In GENERIC, a statement is defined as any 11946 expression whose value, if any, is ignored. A statement will always 11947 have `TREE_SIDE_EFFECTS' set (or it will be discarded), but a 11948 non-statement expression may also have side effects. A `CALL_EXPR', 11949 for instance. 11950 11951 It would be possible for some local optimizations to work on the 11952 GENERIC form of a function; indeed, the adapted tree inliner works fine 11953 on GENERIC, but the current compiler performs inlining after lowering 11954 to GIMPLE (a restricted form described in the next section). Indeed, 11955 currently the frontends perform this lowering before handing off to 11956 `tree_rest_of_compilation', but this seems inelegant. 11957 11958 If necessary, a front end can use some language-dependent tree codes 11959 in its GENERIC representation, so long as it provides a hook for 11960 converting them to GIMPLE and doesn't expect them to work with any 11961 (hypothetical) optimizers that run before the conversion to GIMPLE. The 11962 intermediate representation used while parsing C and C++ looks very 11963 little like GENERIC, but the C and C++ gimplifier hooks are perfectly 11964 happy to take it as input and spit out GIMPLE. 11965 11966 * Menu: 11967 11968 * Statements:: 11969 11970 11971 File: gccint.info, Node: Statements, Up: GENERIC 11972 11973 11.1 Statements 11974 =============== 11975 11976 Most statements in GIMPLE are assignment statements, represented by 11977 `GIMPLE_ASSIGN'. No other C expressions can appear at statement level; 11978 a reference to a volatile object is converted into a `GIMPLE_ASSIGN'. 11979 11980 There are also several varieties of complex statements. 11981 11982 * Menu: 11983 11984 * Blocks:: 11985 * Statement Sequences:: 11986 * Empty Statements:: 11987 * Jumps:: 11988 * Cleanups:: 11989 11990 11991 File: gccint.info, Node: Blocks, Next: Statement Sequences, Up: Statements 11992 11993 11.1.1 Blocks 11994 ------------- 11995 11996 Block scopes and the variables they declare in GENERIC are expressed 11997 using the `BIND_EXPR' code, which in previous versions of GCC was 11998 primarily used for the C statement-expression extension. 11999 12000 Variables in a block are collected into `BIND_EXPR_VARS' in 12001 declaration order. Any runtime initialization is moved out of 12002 `DECL_INITIAL' and into a statement in the controlled block. When 12003 gimplifying from C or C++, this initialization replaces the `DECL_STMT'. 12004 12005 Variable-length arrays (VLAs) complicate this process, as their size 12006 often refers to variables initialized earlier in the block. To handle 12007 this, we currently split the block at that point, and move the VLA into 12008 a new, inner `BIND_EXPR'. This strategy may change in the future. 12009 12010 A C++ program will usually contain more `BIND_EXPR's than there are 12011 syntactic blocks in the source code, since several C++ constructs have 12012 implicit scopes associated with them. On the other hand, although the 12013 C++ front end uses pseudo-scopes to handle cleanups for objects with 12014 destructors, these don't translate into the GIMPLE form; multiple 12015 declarations at the same level use the same `BIND_EXPR'. 12016 12017 12018 File: gccint.info, Node: Statement Sequences, Next: Empty Statements, Prev: Blocks, Up: Statements 12019 12020 11.1.2 Statement Sequences 12021 -------------------------- 12022 12023 Multiple statements at the same nesting level are collected into a 12024 `STATEMENT_LIST'. Statement lists are modified and traversed using the 12025 interface in `tree-iterator.h'. 12026 12027 12028 File: gccint.info, Node: Empty Statements, Next: Jumps, Prev: Statement Sequences, Up: Statements 12029 12030 11.1.3 Empty Statements 12031 ----------------------- 12032 12033 Whenever possible, statements with no effect are discarded. But if 12034 they are nested within another construct which cannot be discarded for 12035 some reason, they are instead replaced with an empty statement, 12036 generated by `build_empty_stmt'. Initially, all empty statements were 12037 shared, after the pattern of the Java front end, but this caused a lot 12038 of trouble in practice. 12039 12040 An empty statement is represented as `(void)0'. 12041 12042 12043 File: gccint.info, Node: Jumps, Next: Cleanups, Prev: Empty Statements, Up: Statements 12044 12045 11.1.4 Jumps 12046 ------------ 12047 12048 Other jumps are expressed by either `GOTO_EXPR' or `RETURN_EXPR'. 12049 12050 The operand of a `GOTO_EXPR' must be either a label or a variable 12051 containing the address to jump to. 12052 12053 The operand of a `RETURN_EXPR' is either `NULL_TREE', `RESULT_DECL', 12054 or a `MODIFY_EXPR' which sets the return value. It would be nice to 12055 move the `MODIFY_EXPR' into a separate statement, but the special 12056 return semantics in `expand_return' make that difficult. It may still 12057 happen in the future, perhaps by moving most of that logic into 12058 `expand_assignment'. 12059 12060 12061 File: gccint.info, Node: Cleanups, Prev: Jumps, Up: Statements 12062 12063 11.1.5 Cleanups 12064 --------------- 12065 12066 Destructors for local C++ objects and similar dynamic cleanups are 12067 represented in GIMPLE by a `TRY_FINALLY_EXPR'. `TRY_FINALLY_EXPR' has 12068 two operands, both of which are a sequence of statements to execute. 12069 The first sequence is executed. When it completes the second sequence 12070 is executed. 12071 12072 The first sequence may complete in the following ways: 12073 12074 1. Execute the last statement in the sequence and fall off the end. 12075 12076 2. Execute a goto statement (`GOTO_EXPR') to an ordinary label 12077 outside the sequence. 12078 12079 3. Execute a return statement (`RETURN_EXPR'). 12080 12081 4. Throw an exception. This is currently not explicitly represented 12082 in GIMPLE. 12083 12084 12085 The second sequence is not executed if the first sequence completes by 12086 calling `setjmp' or `exit' or any other function that does not return. 12087 The second sequence is also not executed if the first sequence 12088 completes via a non-local goto or a computed goto (in general the 12089 compiler does not know whether such a goto statement exits the first 12090 sequence or not, so we assume that it doesn't). 12091 12092 After the second sequence is executed, if it completes normally by 12093 falling off the end, execution continues wherever the first sequence 12094 would have continued, by falling off the end, or doing a goto, etc. 12095 12096 `TRY_FINALLY_EXPR' complicates the flow graph, since the cleanup needs 12097 to appear on every edge out of the controlled block; this reduces the 12098 freedom to move code across these edges. Therefore, the EH lowering 12099 pass which runs before most of the optimization passes eliminates these 12100 expressions by explicitly adding the cleanup to each edge. Rethrowing 12101 the exception is represented using `RESX_EXPR'. 12102 12103 12104 File: gccint.info, Node: GIMPLE, Next: Tree SSA, Prev: GENERIC, Up: Top 12105 12106 12 GIMPLE 12107 ********* 12108 12109 GIMPLE is a three-address representation derived from GENERIC by 12110 breaking down GENERIC expressions into tuples of no more than 3 12111 operands (with some exceptions like function calls). GIMPLE was 12112 heavily influenced by the SIMPLE IL used by the McCAT compiler project 12113 at McGill University, though we have made some different choices. For 12114 one thing, SIMPLE doesn't support `goto'. 12115 12116 Temporaries are introduced to hold intermediate values needed to 12117 compute complex expressions. Additionally, all the control structures 12118 used in GENERIC are lowered into conditional jumps, lexical scopes are 12119 removed and exception regions are converted into an on the side 12120 exception region tree. 12121 12122 The compiler pass which converts GENERIC into GIMPLE is referred to as 12123 the `gimplifier'. The gimplifier works recursively, generating GIMPLE 12124 tuples out of the original GENERIC expressions. 12125 12126 One of the early implementation strategies used for the GIMPLE 12127 representation was to use the same internal data structures used by 12128 front ends to represent parse trees. This simplified implementation 12129 because we could leverage existing functionality and interfaces. 12130 However, GIMPLE is a much more restrictive representation than abstract 12131 syntax trees (AST), therefore it does not require the full structural 12132 complexity provided by the main tree data structure. 12133 12134 The GENERIC representation of a function is stored in the 12135 `DECL_SAVED_TREE' field of the associated `FUNCTION_DECL' tree node. 12136 It is converted to GIMPLE by a call to `gimplify_function_tree'. 12137 12138 If a front end wants to include language-specific tree codes in the 12139 tree representation which it provides to the back end, it must provide a 12140 definition of `LANG_HOOKS_GIMPLIFY_EXPR' which knows how to convert the 12141 front end trees to GIMPLE. Usually such a hook will involve much of 12142 the same code for expanding front end trees to RTL. This function can 12143 return fully lowered GIMPLE, or it can return GENERIC trees and let the 12144 main gimplifier lower them the rest of the way; this is often simpler. 12145 GIMPLE that is not fully lowered is known as "High GIMPLE" and consists 12146 of the IL before the pass `pass_lower_cf'. High GIMPLE contains some 12147 container statements like lexical scopes (represented by `GIMPLE_BIND') 12148 and nested expressions (e.g., `GIMPLE_TRY'), while "Low GIMPLE" exposes 12149 all of the implicit jumps for control and exception expressions 12150 directly in the IL and EH region trees. 12151 12152 The C and C++ front ends currently convert directly from front end 12153 trees to GIMPLE, and hand that off to the back end rather than first 12154 converting to GENERIC. Their gimplifier hooks know about all the 12155 `_STMT' nodes and how to convert them to GENERIC forms. There was some 12156 work done on a genericization pass which would run first, but the 12157 existence of `STMT_EXPR' meant that in order to convert all of the C 12158 statements into GENERIC equivalents would involve walking the entire 12159 tree anyway, so it was simpler to lower all the way. This might change 12160 in the future if someone writes an optimization pass which would work 12161 better with higher-level trees, but currently the optimizers all expect 12162 GIMPLE. 12163 12164 You can request to dump a C-like representation of the GIMPLE form 12165 with the flag `-fdump-tree-gimple'. 12166 12167 * Menu: 12168 12169 * Tuple representation:: 12170 * GIMPLE instruction set:: 12171 * GIMPLE Exception Handling:: 12172 * Temporaries:: 12173 * Operands:: 12174 * Manipulating GIMPLE statements:: 12175 * Tuple specific accessors:: 12176 * GIMPLE sequences:: 12177 * Sequence iterators:: 12178 * Adding a new GIMPLE statement code:: 12179 * Statement and operand traversals:: 12180 12181 12182 File: gccint.info, Node: Tuple representation, Next: GIMPLE instruction set, Up: GIMPLE 12183 12184 12.1 Tuple representation 12185 ========================= 12186 12187 GIMPLE instructions are tuples of variable size divided in two groups: 12188 a header describing the instruction and its locations, and a variable 12189 length body with all the operands. Tuples are organized into a 12190 hierarchy with 3 main classes of tuples. 12191 12192 12.1.1 `gimple_statement_base' (gsbase) 12193 --------------------------------------- 12194 12195 This is the root of the hierarchy, it holds basic information needed by 12196 most GIMPLE statements. There are some fields that may not be relevant 12197 to every GIMPLE statement, but those were moved into the base structure 12198 to take advantage of holes left by other fields (thus making the 12199 structure more compact). The structure takes 4 words (32 bytes) on 64 12200 bit hosts: 12201 12202 Field Size (bits) 12203 `code' 8 12204 `subcode' 16 12205 `no_warning' 1 12206 `visited' 1 12207 `nontemporal_move' 1 12208 `plf' 2 12209 `modified' 1 12210 `has_volatile_ops' 1 12211 `references_memory_p' 1 12212 `uid' 32 12213 `location' 32 12214 `num_ops' 32 12215 `bb' 64 12216 `block' 63 12217 Total size 32 bytes 12218 12219 * `code' Main identifier for a GIMPLE instruction. 12220 12221 * `subcode' Used to distinguish different variants of the same basic 12222 instruction or provide flags applicable to a given code. The 12223 `subcode' flags field has different uses depending on the code of 12224 the instruction, but mostly it distinguishes instructions of the 12225 same family. The most prominent use of this field is in 12226 assignments, where subcode indicates the operation done on the RHS 12227 of the assignment. For example, a = b + c is encoded as 12228 `GIMPLE_ASSIGN <PLUS_EXPR, a, b, c>'. 12229 12230 * `no_warning' Bitflag to indicate whether a warning has already 12231 been issued on this statement. 12232 12233 * `visited' General purpose "visited" marker. Set and cleared by 12234 each pass when needed. 12235 12236 * `nontemporal_move' Bitflag used in assignments that represent 12237 non-temporal moves. Although this bitflag is only used in 12238 assignments, it was moved into the base to take advantage of the 12239 bit holes left by the previous fields. 12240 12241 * `plf' Pass Local Flags. This 2-bit mask can be used as general 12242 purpose markers by any pass. Passes are responsible for clearing 12243 and setting these two flags accordingly. 12244 12245 * `modified' Bitflag to indicate whether the statement has been 12246 modified. Used mainly by the operand scanner to determine when to 12247 re-scan a statement for operands. 12248 12249 * `has_volatile_ops' Bitflag to indicate whether this statement 12250 contains operands that have been marked volatile. 12251 12252 * `references_memory_p' Bitflag to indicate whether this statement 12253 contains memory references (i.e., its operands are either global 12254 variables, or pointer dereferences or anything that must reside in 12255 memory). 12256 12257 * `uid' This is an unsigned integer used by passes that want to 12258 assign IDs to every statement. These IDs must be assigned and used 12259 by each pass. 12260 12261 * `location' This is a `location_t' identifier to specify source code 12262 location for this statement. It is inherited from the front end. 12263 12264 * `num_ops' Number of operands that this statement has. This 12265 specifies the size of the operand vector embedded in the tuple. 12266 Only used in some tuples, but it is declared in the base tuple to 12267 take advantage of the 32-bit hole left by the previous fields. 12268 12269 * `bb' Basic block holding the instruction. 12270 12271 * `block' Lexical block holding this statement. Also used for debug 12272 information generation. 12273 12274 12.1.2 `gimple_statement_with_ops' 12275 ---------------------------------- 12276 12277 This tuple is actually split in two: `gimple_statement_with_ops_base' 12278 and `gimple_statement_with_ops'. This is needed to accommodate the way 12279 the operand vector is allocated. The operand vector is defined to be an 12280 array of 1 element. So, to allocate a dynamic number of operands, the 12281 memory allocator (`gimple_alloc') simply allocates enough memory to 12282 hold the structure itself plus `N - 1' operands which run "off the end" 12283 of the structure. For example, to allocate space for a tuple with 3 12284 operands, `gimple_alloc' reserves `sizeof (struct 12285 gimple_statement_with_ops) + 2 * sizeof (tree)' bytes. 12286 12287 On the other hand, several fields in this tuple need to be shared with 12288 the `gimple_statement_with_memory_ops' tuple. So, these common fields 12289 are placed in `gimple_statement_with_ops_base' which is then inherited 12290 from the other two tuples. 12291 12292 `gsbase' 256 12293 `addresses_taken' 64 12294 `def_ops' 64 12295 `use_ops' 64 12296 `op' `num_ops' * 64 12297 Total size 56 + 8 * `num_ops' bytes 12298 12299 * `gsbase' Inherited from `struct gimple_statement_base'. 12300 12301 * `addresses_taken' Bitmap holding the UIDs of all the `VAR_DECL's 12302 whose addresses are taken by this statement. For example, a 12303 statement of the form `p = &b' will have the UID for symbol `b' in 12304 this set. 12305 12306 * `def_ops' Array of pointers into the operand array indicating all 12307 the slots that contain a variable written-to by the statement. 12308 This array is also used for immediate use chaining. Note that it 12309 would be possible to not rely on this array, but the changes 12310 required to implement this are pretty invasive. 12311 12312 * `use_ops' Similar to `def_ops' but for variables read by the 12313 statement. 12314 12315 * `op' Array of trees with `num_ops' slots. 12316 12317 12.1.3 `gimple_statement_with_memory_ops' 12318 ----------------------------------------- 12319 12320 This tuple is essentially identical to `gimple_statement_with_ops', 12321 except that it contains 4 additional fields to hold vectors related 12322 memory stores and loads. Similar to the previous case, the structure 12323 is split in two to accommodate for the operand vector 12324 (`gimple_statement_with_memory_ops_base' and 12325 `gimple_statement_with_memory_ops'). 12326 12327 Field Size (bits) 12328 `gsbase' 256 12329 `addresses_taken' 64 12330 `def_ops' 64 12331 `use_ops' 64 12332 `vdef_ops' 64 12333 `vuse_ops' 64 12334 `stores' 64 12335 `loads' 64 12336 `op' `num_ops' * 64 12337 Total size 88 + 8 * `num_ops' bytes 12338 12339 * `vdef_ops' Similar to `def_ops' but for `VDEF' operators. There is 12340 one entry per memory symbol written by this statement. This is 12341 used to maintain the memory SSA use-def and def-def chains. 12342 12343 * `vuse_ops' Similar to `use_ops' but for `VUSE' operators. There is 12344 one entry per memory symbol loaded by this statement. This is used 12345 to maintain the memory SSA use-def chains. 12346 12347 * `stores' Bitset with all the UIDs for the symbols written-to by the 12348 statement. This is different than `vdef_ops' in that all the 12349 affected symbols are mentioned in this set. If memory 12350 partitioning is enabled, the `vdef_ops' vector will refer to memory 12351 partitions. Furthermore, no SSA information is stored in this set. 12352 12353 * `loads' Similar to `stores', but for memory loads. (Note that there 12354 is some amount of redundancy here, it should be possible to reduce 12355 memory utilization further by removing these sets). 12356 12357 All the other tuples are defined in terms of these three basic ones. 12358 Each tuple will add some fields. The main gimple type is defined to be 12359 the union of all these structures (`GTY' markers elided for clarity): 12360 12361 union gimple_statement_d 12362 { 12363 struct gimple_statement_base gsbase; 12364 struct gimple_statement_with_ops gsops; 12365 struct gimple_statement_with_memory_ops gsmem; 12366 struct gimple_statement_omp omp; 12367 struct gimple_statement_bind gimple_bind; 12368 struct gimple_statement_catch gimple_catch; 12369 struct gimple_statement_eh_filter gimple_eh_filter; 12370 struct gimple_statement_phi gimple_phi; 12371 struct gimple_statement_resx gimple_resx; 12372 struct gimple_statement_try gimple_try; 12373 struct gimple_statement_wce gimple_wce; 12374 struct gimple_statement_asm gimple_asm; 12375 struct gimple_statement_omp_critical gimple_omp_critical; 12376 struct gimple_statement_omp_for gimple_omp_for; 12377 struct gimple_statement_omp_parallel gimple_omp_parallel; 12378 struct gimple_statement_omp_task gimple_omp_task; 12379 struct gimple_statement_omp_sections gimple_omp_sections; 12380 struct gimple_statement_omp_single gimple_omp_single; 12381 struct gimple_statement_omp_continue gimple_omp_continue; 12382 struct gimple_statement_omp_atomic_load gimple_omp_atomic_load; 12383 struct gimple_statement_omp_atomic_store gimple_omp_atomic_store; 12384 }; 12385 12386 12387 File: gccint.info, Node: GIMPLE instruction set, Next: GIMPLE Exception Handling, Prev: Tuple representation, Up: GIMPLE 12388 12389 12.2 GIMPLE instruction set 12390 =========================== 12391 12392 The following table briefly describes the GIMPLE instruction set. 12393 12394 Instruction High GIMPLE Low GIMPLE 12395 `GIMPLE_ASM' x x 12396 `GIMPLE_ASSIGN' x x 12397 `GIMPLE_BIND' x 12398 `GIMPLE_CALL' x x 12399 `GIMPLE_CATCH' x 12400 `GIMPLE_CHANGE_DYNAMIC_TYPE' x x 12401 `GIMPLE_COND' x x 12402 `GIMPLE_EH_FILTER' x 12403 `GIMPLE_GOTO' x x 12404 `GIMPLE_LABEL' x x 12405 `GIMPLE_NOP' x x 12406 `GIMPLE_OMP_ATOMIC_LOAD' x x 12407 `GIMPLE_OMP_ATOMIC_STORE' x x 12408 `GIMPLE_OMP_CONTINUE' x x 12409 `GIMPLE_OMP_CRITICAL' x x 12410 `GIMPLE_OMP_FOR' x x 12411 `GIMPLE_OMP_MASTER' x x 12412 `GIMPLE_OMP_ORDERED' x x 12413 `GIMPLE_OMP_PARALLEL' x x 12414 `GIMPLE_OMP_RETURN' x x 12415 `GIMPLE_OMP_SECTION' x x 12416 `GIMPLE_OMP_SECTIONS' x x 12417 `GIMPLE_OMP_SECTIONS_SWITCH' x x 12418 `GIMPLE_OMP_SINGLE' x x 12419 `GIMPLE_PHI' x 12420 `GIMPLE_RESX' x 12421 `GIMPLE_RETURN' x x 12422 `GIMPLE_SWITCH' x x 12423 `GIMPLE_TRY' x 12424 12425 12426 File: gccint.info, Node: GIMPLE Exception Handling, Next: Temporaries, Prev: GIMPLE instruction set, Up: GIMPLE 12427 12428 12.3 Exception Handling 12429 ======================= 12430 12431 Other exception handling constructs are represented using 12432 `GIMPLE_TRY_CATCH'. `GIMPLE_TRY_CATCH' has two operands. The first 12433 operand is a sequence of statements to execute. If executing these 12434 statements does not throw an exception, then the second operand is 12435 ignored. Otherwise, if an exception is thrown, then the second operand 12436 of the `GIMPLE_TRY_CATCH' is checked. The second operand may have the 12437 following forms: 12438 12439 1. A sequence of statements to execute. When an exception occurs, 12440 these statements are executed, and then the exception is rethrown. 12441 12442 2. A sequence of `GIMPLE_CATCH' statements. Each `GIMPLE_CATCH' has 12443 a list of applicable exception types and handler code. If the 12444 thrown exception matches one of the caught types, the associated 12445 handler code is executed. If the handler code falls off the 12446 bottom, execution continues after the original `GIMPLE_TRY_CATCH'. 12447 12448 3. An `GIMPLE_EH_FILTER' statement. This has a list of permitted 12449 exception types, and code to handle a match failure. If the 12450 thrown exception does not match one of the allowed types, the 12451 associated match failure code is executed. If the thrown exception 12452 does match, it continues unwinding the stack looking for the next 12453 handler. 12454 12455 12456 Currently throwing an exception is not directly represented in GIMPLE, 12457 since it is implemented by calling a function. At some point in the 12458 future we will want to add some way to express that the call will throw 12459 an exception of a known type. 12460 12461 Just before running the optimizers, the compiler lowers the high-level 12462 EH constructs above into a set of `goto's, magic labels, and EH 12463 regions. Continuing to unwind at the end of a cleanup is represented 12464 with a `GIMPLE_RESX'. 12465 12466 12467 File: gccint.info, Node: Temporaries, Next: Operands, Prev: GIMPLE Exception Handling, Up: GIMPLE 12468 12469 12.4 Temporaries 12470 ================ 12471 12472 When gimplification encounters a subexpression that is too complex, it 12473 creates a new temporary variable to hold the value of the 12474 subexpression, and adds a new statement to initialize it before the 12475 current statement. These special temporaries are known as `expression 12476 temporaries', and are allocated using `get_formal_tmp_var'. The 12477 compiler tries to always evaluate identical expressions into the same 12478 temporary, to simplify elimination of redundant calculations. 12479 12480 We can only use expression temporaries when we know that it will not 12481 be reevaluated before its value is used, and that it will not be 12482 otherwise modified(1). Other temporaries can be allocated using 12483 `get_initialized_tmp_var' or `create_tmp_var'. 12484 12485 Currently, an expression like `a = b + 5' is not reduced any further. 12486 We tried converting it to something like 12487 T1 = b + 5; 12488 a = T1; 12489 but this bloated the representation for minimal benefit. However, a 12490 variable which must live in memory cannot appear in an expression; its 12491 value is explicitly loaded into a temporary first. Similarly, storing 12492 the value of an expression to a memory variable goes through a 12493 temporary. 12494 12495 ---------- Footnotes ---------- 12496 12497 (1) These restrictions are derived from those in Morgan 4.8. 12498 12499 12500 File: gccint.info, Node: Operands, Next: Manipulating GIMPLE statements, Prev: Temporaries, Up: GIMPLE 12501 12502 12.5 Operands 12503 ============= 12504 12505 In general, expressions in GIMPLE consist of an operation and the 12506 appropriate number of simple operands; these operands must either be a 12507 GIMPLE rvalue (`is_gimple_val'), i.e. a constant or a register 12508 variable. More complex operands are factored out into temporaries, so 12509 that 12510 a = b + c + d 12511 becomes 12512 T1 = b + c; 12513 a = T1 + d; 12514 12515 The same rule holds for arguments to a `GIMPLE_CALL'. 12516 12517 The target of an assignment is usually a variable, but can also be an 12518 `INDIRECT_REF' or a compound lvalue as described below. 12519 12520 * Menu: 12521 12522 * Compound Expressions:: 12523 * Compound Lvalues:: 12524 * Conditional Expressions:: 12525 * Logical Operators:: 12526 12527 12528 File: gccint.info, Node: Compound Expressions, Next: Compound Lvalues, Up: Operands 12529 12530 12.5.1 Compound Expressions 12531 --------------------------- 12532 12533 The left-hand side of a C comma expression is simply moved into a 12534 separate statement. 12535 12536 12537 File: gccint.info, Node: Compound Lvalues, Next: Conditional Expressions, Prev: Compound Expressions, Up: Operands 12538 12539 12.5.2 Compound Lvalues 12540 ----------------------- 12541 12542 Currently compound lvalues involving array and structure field 12543 references are not broken down; an expression like `a.b[2] = 42' is not 12544 reduced any further (though complex array subscripts are). This 12545 restriction is a workaround for limitations in later optimizers; if we 12546 were to convert this to 12547 12548 T1 = &a.b; 12549 T1[2] = 42; 12550 12551 alias analysis would not remember that the reference to `T1[2]' came 12552 by way of `a.b', so it would think that the assignment could alias 12553 another member of `a'; this broke `struct-alias-1.c'. Future optimizer 12554 improvements may make this limitation unnecessary. 12555 12556 12557 File: gccint.info, Node: Conditional Expressions, Next: Logical Operators, Prev: Compound Lvalues, Up: Operands 12558 12559 12.5.3 Conditional Expressions 12560 ------------------------------ 12561 12562 A C `?:' expression is converted into an `if' statement with each 12563 branch assigning to the same temporary. So, 12564 12565 a = b ? c : d; 12566 becomes 12567 if (b == 1) 12568 T1 = c; 12569 else 12570 T1 = d; 12571 a = T1; 12572 12573 The GIMPLE level if-conversion pass re-introduces `?:' expression, if 12574 appropriate. It is used to vectorize loops with conditions using vector 12575 conditional operations. 12576 12577 Note that in GIMPLE, `if' statements are represented using 12578 `GIMPLE_COND', as described below. 12579 12580 12581 File: gccint.info, Node: Logical Operators, Prev: Conditional Expressions, Up: Operands 12582 12583 12.5.4 Logical Operators 12584 ------------------------ 12585 12586 Except when they appear in the condition operand of a `GIMPLE_COND', 12587 logical `and' and `or' operators are simplified as follows: `a = b && 12588 c' becomes 12589 12590 T1 = (bool)b; 12591 if (T1 == true) 12592 T1 = (bool)c; 12593 a = T1; 12594 12595 Note that `T1' in this example cannot be an expression temporary, 12596 because it has two different assignments. 12597 12598 12.5.5 Manipulating operands 12599 ---------------------------- 12600 12601 All gimple operands are of type `tree'. But only certain types of 12602 trees are allowed to be used as operand tuples. Basic validation is 12603 controlled by the function `get_gimple_rhs_class', which given a tree 12604 code, returns an `enum' with the following values of type `enum 12605 gimple_rhs_class' 12606 12607 * `GIMPLE_INVALID_RHS' The tree cannot be used as a GIMPLE operand. 12608 12609 * `GIMPLE_BINARY_RHS' The tree is a valid GIMPLE binary operation. 12610 12611 * `GIMPLE_UNARY_RHS' The tree is a valid GIMPLE unary operation. 12612 12613 * `GIMPLE_SINGLE_RHS' The tree is a single object, that cannot be 12614 split into simpler operands (for instance, `SSA_NAME', `VAR_DECL', 12615 `COMPONENT_REF', etc). 12616 12617 This operand class also acts as an escape hatch for tree nodes 12618 that may be flattened out into the operand vector, but would need 12619 more than two slots on the RHS. For instance, a `COND_EXPR' 12620 expression of the form `(a op b) ? x : y' could be flattened out 12621 on the operand vector using 4 slots, but it would also require 12622 additional processing to distinguish `c = a op b' from `c = a op b 12623 ? x : y'. Something similar occurs with `ASSERT_EXPR'. In time, 12624 these special case tree expressions should be flattened into the 12625 operand vector. 12626 12627 For tree nodes in the categories `GIMPLE_BINARY_RHS' and 12628 `GIMPLE_UNARY_RHS', they cannot be stored inside tuples directly. They 12629 first need to be flattened and separated into individual components. 12630 For instance, given the GENERIC expression 12631 12632 a = b + c 12633 12634 its tree representation is: 12635 12636 MODIFY_EXPR <VAR_DECL <a>, PLUS_EXPR <VAR_DECL <b>, VAR_DECL <c>>> 12637 12638 In this case, the GIMPLE form for this statement is logically 12639 identical to its GENERIC form but in GIMPLE, the `PLUS_EXPR' on the RHS 12640 of the assignment is not represented as a tree, instead the two 12641 operands are taken out of the `PLUS_EXPR' sub-tree and flattened into 12642 the GIMPLE tuple as follows: 12643 12644 GIMPLE_ASSIGN <PLUS_EXPR, VAR_DECL <a>, VAR_DECL <b>, VAR_DECL <c>> 12645 12646 12.5.6 Operand vector allocation 12647 -------------------------------- 12648 12649 The operand vector is stored at the bottom of the three tuple 12650 structures that accept operands. This means, that depending on the code 12651 of a given statement, its operand vector will be at different offsets 12652 from the base of the structure. To access tuple operands use the 12653 following accessors 12654 12655 -- GIMPLE function: unsigned gimple_num_ops (gimple g) 12656 Returns the number of operands in statement G. 12657 12658 -- GIMPLE function: tree gimple_op (gimple g, unsigned i) 12659 Returns operand `I' from statement `G'. 12660 12661 -- GIMPLE function: tree *gimple_ops (gimple g) 12662 Returns a pointer into the operand vector for statement `G'. This 12663 is computed using an internal table called `gimple_ops_offset_'[]. 12664 This table is indexed by the gimple code of `G'. 12665 12666 When the compiler is built, this table is filled-in using the 12667 sizes of the structures used by each statement code defined in 12668 gimple.def. Since the operand vector is at the bottom of the 12669 structure, for a gimple code `C' the offset is computed as sizeof 12670 (struct-of `C') - sizeof (tree). 12671 12672 This mechanism adds one memory indirection to every access when 12673 using `gimple_op'(), if this becomes a bottleneck, a pass can 12674 choose to memoize the result from `gimple_ops'() and use that to 12675 access the operands. 12676 12677 12.5.7 Operand validation 12678 ------------------------- 12679 12680 When adding a new operand to a gimple statement, the operand will be 12681 validated according to what each tuple accepts in its operand vector. 12682 These predicates are called by the `gimple_<name>_set_...()'. Each 12683 tuple will use one of the following predicates (Note, this list is not 12684 exhaustive): 12685 12686 -- GIMPLE function: is_gimple_operand (tree t) 12687 This is the most permissive of the predicates. It essentially 12688 checks whether t has a `gimple_rhs_class' of `GIMPLE_SINGLE_RHS'. 12689 12690 -- GIMPLE function: is_gimple_val (tree t) 12691 Returns true if t is a "GIMPLE value", which are all the 12692 non-addressable stack variables (variables for which 12693 `is_gimple_reg' returns true) and constants (expressions for which 12694 `is_gimple_min_invariant' returns true). 12695 12696 -- GIMPLE function: is_gimple_addressable (tree t) 12697 Returns true if t is a symbol or memory reference whose address 12698 can be taken. 12699 12700 -- GIMPLE function: is_gimple_asm_val (tree t) 12701 Similar to `is_gimple_val' but it also accepts hard registers. 12702 12703 -- GIMPLE function: is_gimple_call_addr (tree t) 12704 Return true if t is a valid expression to use as the function 12705 called by a `GIMPLE_CALL'. 12706 12707 -- GIMPLE function: is_gimple_constant (tree t) 12708 Return true if t is a valid gimple constant. 12709 12710 -- GIMPLE function: is_gimple_min_invariant (tree t) 12711 Return true if t is a valid minimal invariant. This is different 12712 from constants, in that the specific value of t may not be known 12713 at compile time, but it is known that it doesn't change (e.g., the 12714 address of a function local variable). 12715 12716 -- GIMPLE function: is_gimple_min_invariant_address (tree t) 12717 Return true if t is an `ADDR_EXPR' that does not change once the 12718 program is running. 12719 12720 12.5.8 Statement validation 12721 --------------------------- 12722 12723 -- GIMPLE function: is_gimple_assign (gimple g) 12724 Return true if the code of g is `GIMPLE_ASSIGN'. 12725 12726 -- GIMPLE function: is_gimple_call (gimple g) 12727 Return true if the code of g is `GIMPLE_CALL' 12728 12729 -- GIMPLE function: gimple_assign_cast_p (gimple g) 12730 Return true if g is a `GIMPLE_ASSIGN' that performs a type cast 12731 operation 12732 12733 12734 File: gccint.info, Node: Manipulating GIMPLE statements, Next: Tuple specific accessors, Prev: Operands, Up: GIMPLE 12735 12736 12.6 Manipulating GIMPLE statements 12737 =================================== 12738 12739 This section documents all the functions available to handle each of 12740 the GIMPLE instructions. 12741 12742 12.6.1 Common accessors 12743 ----------------------- 12744 12745 The following are common accessors for gimple statements. 12746 12747 -- GIMPLE function: enum gimple_code gimple_code (gimple g) 12748 Return the code for statement `G'. 12749 12750 -- GIMPLE function: basic_block gimple_bb (gimple g) 12751 Return the basic block to which statement `G' belongs to. 12752 12753 -- GIMPLE function: tree gimple_block (gimple g) 12754 Return the lexical scope block holding statement `G'. 12755 12756 -- GIMPLE function: tree gimple_expr_type (gimple stmt) 12757 Return the type of the main expression computed by `STMT'. Return 12758 `void_type_node' if `STMT' computes nothing. This will only return 12759 something meaningful for `GIMPLE_ASSIGN', `GIMPLE_COND' and 12760 `GIMPLE_CALL'. For all other tuple codes, it will return 12761 `void_type_node'. 12762 12763 -- GIMPLE function: enum tree_code gimple_expr_code (gimple stmt) 12764 Return the tree code for the expression computed by `STMT'. This 12765 is only meaningful for `GIMPLE_CALL', `GIMPLE_ASSIGN' and 12766 `GIMPLE_COND'. If `STMT' is `GIMPLE_CALL', it will return 12767 `CALL_EXPR'. For `GIMPLE_COND', it returns the code of the 12768 comparison predicate. For `GIMPLE_ASSIGN' it returns the code of 12769 the operation performed by the `RHS' of the assignment. 12770 12771 -- GIMPLE function: void gimple_set_block (gimple g, tree block) 12772 Set the lexical scope block of `G' to `BLOCK'. 12773 12774 -- GIMPLE function: location_t gimple_locus (gimple g) 12775 Return locus information for statement `G'. 12776 12777 -- GIMPLE function: void gimple_set_locus (gimple g, location_t locus) 12778 Set locus information for statement `G'. 12779 12780 -- GIMPLE function: bool gimple_locus_empty_p (gimple g) 12781 Return true if `G' does not have locus information. 12782 12783 -- GIMPLE function: bool gimple_no_warning_p (gimple stmt) 12784 Return true if no warnings should be emitted for statement `STMT'. 12785 12786 -- GIMPLE function: void gimple_set_visited (gimple stmt, bool 12787 visited_p) 12788 Set the visited status on statement `STMT' to `VISITED_P'. 12789 12790 -- GIMPLE function: bool gimple_visited_p (gimple stmt) 12791 Return the visited status on statement `STMT'. 12792 12793 -- GIMPLE function: void gimple_set_plf (gimple stmt, enum plf_mask 12794 plf, bool val_p) 12795 Set pass local flag `PLF' on statement `STMT' to `VAL_P'. 12796 12797 -- GIMPLE function: unsigned int gimple_plf (gimple stmt, enum 12798 plf_mask plf) 12799 Return the value of pass local flag `PLF' on statement `STMT'. 12800 12801 -- GIMPLE function: bool gimple_has_ops (gimple g) 12802 Return true if statement `G' has register or memory operands. 12803 12804 -- GIMPLE function: bool gimple_has_mem_ops (gimple g) 12805 Return true if statement `G' has memory operands. 12806 12807 -- GIMPLE function: unsigned gimple_num_ops (gimple g) 12808 Return the number of operands for statement `G'. 12809 12810 -- GIMPLE function: tree *gimple_ops (gimple g) 12811 Return the array of operands for statement `G'. 12812 12813 -- GIMPLE function: tree gimple_op (gimple g, unsigned i) 12814 Return operand `I' for statement `G'. 12815 12816 -- GIMPLE function: tree *gimple_op_ptr (gimple g, unsigned i) 12817 Return a pointer to operand `I' for statement `G'. 12818 12819 -- GIMPLE function: void gimple_set_op (gimple g, unsigned i, tree op) 12820 Set operand `I' of statement `G' to `OP'. 12821 12822 -- GIMPLE function: bitmap gimple_addresses_taken (gimple stmt) 12823 Return the set of symbols that have had their address taken by 12824 `STMT'. 12825 12826 -- GIMPLE function: struct def_optype_d *gimple_def_ops (gimple g) 12827 Return the set of `DEF' operands for statement `G'. 12828 12829 -- GIMPLE function: void gimple_set_def_ops (gimple g, struct 12830 def_optype_d *def) 12831 Set `DEF' to be the set of `DEF' operands for statement `G'. 12832 12833 -- GIMPLE function: struct use_optype_d *gimple_use_ops (gimple g) 12834 Return the set of `USE' operands for statement `G'. 12835 12836 -- GIMPLE function: void gimple_set_use_ops (gimple g, struct 12837 use_optype_d *use) 12838 Set `USE' to be the set of `USE' operands for statement `G'. 12839 12840 -- GIMPLE function: struct voptype_d *gimple_vuse_ops (gimple g) 12841 Return the set of `VUSE' operands for statement `G'. 12842 12843 -- GIMPLE function: void gimple_set_vuse_ops (gimple g, struct 12844 voptype_d *ops) 12845 Set `OPS' to be the set of `VUSE' operands for statement `G'. 12846 12847 -- GIMPLE function: struct voptype_d *gimple_vdef_ops (gimple g) 12848 Return the set of `VDEF' operands for statement `G'. 12849 12850 -- GIMPLE function: void gimple_set_vdef_ops (gimple g, struct 12851 voptype_d *ops) 12852 Set `OPS' to be the set of `VDEF' operands for statement `G'. 12853 12854 -- GIMPLE function: bitmap gimple_loaded_syms (gimple g) 12855 Return the set of symbols loaded by statement `G'. Each element of 12856 the set is the `DECL_UID' of the corresponding symbol. 12857 12858 -- GIMPLE function: bitmap gimple_stored_syms (gimple g) 12859 Return the set of symbols stored by statement `G'. Each element of 12860 the set is the `DECL_UID' of the corresponding symbol. 12861 12862 -- GIMPLE function: bool gimple_modified_p (gimple g) 12863 Return true if statement `G' has operands and the modified field 12864 has been set. 12865 12866 -- GIMPLE function: bool gimple_has_volatile_ops (gimple stmt) 12867 Return true if statement `STMT' contains volatile operands. 12868 12869 -- GIMPLE function: void gimple_set_has_volatile_ops (gimple stmt, 12870 bool volatilep) 12871 Return true if statement `STMT' contains volatile operands. 12872 12873 -- GIMPLE function: void update_stmt (gimple s) 12874 Mark statement `S' as modified, and update it. 12875 12876 -- GIMPLE function: void update_stmt_if_modified (gimple s) 12877 Update statement `S' if it has been marked modified. 12878 12879 -- GIMPLE function: gimple gimple_copy (gimple stmt) 12880 Return a deep copy of statement `STMT'. 12881 12882 12883 File: gccint.info, Node: Tuple specific accessors, Next: GIMPLE sequences, Prev: Manipulating GIMPLE statements, Up: GIMPLE 12884 12885 12.7 Tuple specific accessors 12886 ============================= 12887 12888 * Menu: 12889 12890 * `GIMPLE_ASM':: 12891 * `GIMPLE_ASSIGN':: 12892 * `GIMPLE_BIND':: 12893 * `GIMPLE_CALL':: 12894 * `GIMPLE_CATCH':: 12895 * `GIMPLE_CHANGE_DYNAMIC_TYPE':: 12896 * `GIMPLE_COND':: 12897 * `GIMPLE_EH_FILTER':: 12898 * `GIMPLE_LABEL':: 12899 * `GIMPLE_NOP':: 12900 * `GIMPLE_OMP_ATOMIC_LOAD':: 12901 * `GIMPLE_OMP_ATOMIC_STORE':: 12902 * `GIMPLE_OMP_CONTINUE':: 12903 * `GIMPLE_OMP_CRITICAL':: 12904 * `GIMPLE_OMP_FOR':: 12905 * `GIMPLE_OMP_MASTER':: 12906 * `GIMPLE_OMP_ORDERED':: 12907 * `GIMPLE_OMP_PARALLEL':: 12908 * `GIMPLE_OMP_RETURN':: 12909 * `GIMPLE_OMP_SECTION':: 12910 * `GIMPLE_OMP_SECTIONS':: 12911 * `GIMPLE_OMP_SINGLE':: 12912 * `GIMPLE_PHI':: 12913 * `GIMPLE_RESX':: 12914 * `GIMPLE_RETURN':: 12915 * `GIMPLE_SWITCH':: 12916 * `GIMPLE_TRY':: 12917 * `GIMPLE_WITH_CLEANUP_EXPR':: 12918 12919 12920 File: gccint.info, Node: `GIMPLE_ASM', Next: `GIMPLE_ASSIGN', Up: Tuple specific accessors 12921 12922 12.7.1 `GIMPLE_ASM' 12923 ------------------- 12924 12925 -- GIMPLE function: gimple gimple_build_asm (const char *string, 12926 ninputs, noutputs, nclobbers, ...) 12927 Build a `GIMPLE_ASM' statement. This statement is used for 12928 building in-line assembly constructs. `STRING' is the assembly 12929 code. `NINPUT' is the number of register inputs. `NOUTPUT' is the 12930 number of register outputs. `NCLOBBERS' is the number of clobbered 12931 registers. The rest of the arguments trees for each input, 12932 output, and clobbered registers. 12933 12934 -- GIMPLE function: gimple gimple_build_asm_vec (const char *, 12935 VEC(tree,gc) *, VEC(tree,gc) *, VEC(tree,gc) *) 12936 Identical to gimple_build_asm, but the arguments are passed in 12937 VECs. 12938 12939 -- GIMPLE function: gimple_asm_ninputs (gimple g) 12940 Return the number of input operands for `GIMPLE_ASM' `G'. 12941 12942 -- GIMPLE function: gimple_asm_noutputs (gimple g) 12943 Return the number of output operands for `GIMPLE_ASM' `G'. 12944 12945 -- GIMPLE function: gimple_asm_nclobbers (gimple g) 12946 Return the number of clobber operands for `GIMPLE_ASM' `G'. 12947 12948 -- GIMPLE function: tree gimple_asm_input_op (gimple g, unsigned index) 12949 Return input operand `INDEX' of `GIMPLE_ASM' `G'. 12950 12951 -- GIMPLE function: void gimple_asm_set_input_op (gimple g, unsigned 12952 index, tree in_op) 12953 Set `IN_OP' to be input operand `INDEX' in `GIMPLE_ASM' `G'. 12954 12955 -- GIMPLE function: tree gimple_asm_output_op (gimple g, unsigned 12956 index) 12957 Return output operand `INDEX' of `GIMPLE_ASM' `G'. 12958 12959 -- GIMPLE function: void gimple_asm_set_output_op (gimple g, unsigned 12960 index, tree out_op) 12961 Set `OUT_OP' to be output operand `INDEX' in `GIMPLE_ASM' `G'. 12962 12963 -- GIMPLE function: tree gimple_asm_clobber_op (gimple g, unsigned 12964 index) 12965 Return clobber operand `INDEX' of `GIMPLE_ASM' `G'. 12966 12967 -- GIMPLE function: void gimple_asm_set_clobber_op (gimple g, unsigned 12968 index, tree clobber_op) 12969 Set `CLOBBER_OP' to be clobber operand `INDEX' in `GIMPLE_ASM' `G'. 12970 12971 -- GIMPLE function: const char *gimple_asm_string (gimple g) 12972 Return the string representing the assembly instruction in 12973 `GIMPLE_ASM' `G'. 12974 12975 -- GIMPLE function: bool gimple_asm_volatile_p (gimple g) 12976 Return true if `G' is an asm statement marked volatile. 12977 12978 -- GIMPLE function: void gimple_asm_set_volatile (gimple g) 12979 Mark asm statement `G' as volatile. 12980 12981 -- GIMPLE function: void gimple_asm_clear_volatile (gimple g) 12982 Remove volatile marker from asm statement `G'. 12983 12984 12985 File: gccint.info, Node: `GIMPLE_ASSIGN', Next: `GIMPLE_BIND', Prev: `GIMPLE_ASM', Up: Tuple specific accessors 12986 12987 12.7.2 `GIMPLE_ASSIGN' 12988 ---------------------- 12989 12990 -- GIMPLE function: gimple gimple_build_assign (tree lhs, tree rhs) 12991 Build a `GIMPLE_ASSIGN' statement. The left-hand side is an lvalue 12992 passed in lhs. The right-hand side can be either a unary or 12993 binary tree expression. The expression tree rhs will be flattened 12994 and its operands assigned to the corresponding operand slots in 12995 the new statement. This function is useful when you already have 12996 a tree expression that you want to convert into a tuple. However, 12997 try to avoid building expression trees for the sole purpose of 12998 calling this function. If you already have the operands in 12999 separate trees, it is better to use `gimple_build_assign_with_ops'. 13000 13001 -- GIMPLE function: gimple gimplify_assign (tree dst, tree src, 13002 gimple_seq *seq_p) 13003 Build a new `GIMPLE_ASSIGN' tuple and append it to the end of 13004 `*SEQ_P'. 13005 13006 `DST'/`SRC' are the destination and source respectively. You can pass 13007 ungimplified trees in `DST' or `SRC', in which case they will be 13008 converted to a gimple operand if necessary. 13009 13010 This function returns the newly created `GIMPLE_ASSIGN' tuple. 13011 13012 -- GIMPLE function: gimple gimple_build_assign_with_ops (enum 13013 tree_code subcode, tree lhs, tree op1, tree op2) 13014 This function is similar to `gimple_build_assign', but is used to 13015 build a `GIMPLE_ASSIGN' statement when the operands of the 13016 right-hand side of the assignment are already split into different 13017 operands. 13018 13019 The left-hand side is an lvalue passed in lhs. Subcode is the 13020 `tree_code' for the right-hand side of the assignment. Op1 and op2 13021 are the operands. If op2 is null, subcode must be a `tree_code' 13022 for a unary expression. 13023 13024 -- GIMPLE function: enum tree_code gimple_assign_rhs_code (gimple g) 13025 Return the code of the expression computed on the `RHS' of 13026 assignment statement `G'. 13027 13028 -- GIMPLE function: enum gimple_rhs_class gimple_assign_rhs_class 13029 (gimple g) 13030 Return the gimple rhs class of the code for the expression 13031 computed on the rhs of assignment statement `G'. This will never 13032 return `GIMPLE_INVALID_RHS'. 13033 13034 -- GIMPLE function: tree gimple_assign_lhs (gimple g) 13035 Return the `LHS' of assignment statement `G'. 13036 13037 -- GIMPLE function: tree *gimple_assign_lhs_ptr (gimple g) 13038 Return a pointer to the `LHS' of assignment statement `G'. 13039 13040 -- GIMPLE function: tree gimple_assign_rhs1 (gimple g) 13041 Return the first operand on the `RHS' of assignment statement `G'. 13042 13043 -- GIMPLE function: tree *gimple_assign_rhs1_ptr (gimple g) 13044 Return the address of the first operand on the `RHS' of assignment 13045 statement `G'. 13046 13047 -- GIMPLE function: tree gimple_assign_rhs2 (gimple g) 13048 Return the second operand on the `RHS' of assignment statement `G'. 13049 13050 -- GIMPLE function: tree *gimple_assign_rhs2_ptr (gimple g) 13051 Return the address of the second operand on the `RHS' of assignment 13052 statement `G'. 13053 13054 -- GIMPLE function: void gimple_assign_set_lhs (gimple g, tree lhs) 13055 Set `LHS' to be the `LHS' operand of assignment statement `G'. 13056 13057 -- GIMPLE function: void gimple_assign_set_rhs1 (gimple g, tree rhs) 13058 Set `RHS' to be the first operand on the `RHS' of assignment 13059 statement `G'. 13060 13061 -- GIMPLE function: tree gimple_assign_rhs2 (gimple g) 13062 Return the second operand on the `RHS' of assignment statement `G'. 13063 13064 -- GIMPLE function: tree *gimple_assign_rhs2_ptr (gimple g) 13065 Return a pointer to the second operand on the `RHS' of assignment 13066 statement `G'. 13067 13068 -- GIMPLE function: void gimple_assign_set_rhs2 (gimple g, tree rhs) 13069 Set `RHS' to be the second operand on the `RHS' of assignment 13070 statement `G'. 13071 13072 -- GIMPLE function: bool gimple_assign_cast_p (gimple s) 13073 Return true if `S' is an type-cast assignment. 13074 13075 13076 File: gccint.info, Node: `GIMPLE_BIND', Next: `GIMPLE_CALL', Prev: `GIMPLE_ASSIGN', Up: Tuple specific accessors 13077 13078 12.7.3 `GIMPLE_BIND' 13079 -------------------- 13080 13081 -- GIMPLE function: gimple gimple_build_bind (tree vars, gimple_seq 13082 body) 13083 Build a `GIMPLE_BIND' statement with a list of variables in `VARS' 13084 and a body of statements in sequence `BODY'. 13085 13086 -- GIMPLE function: tree gimple_bind_vars (gimple g) 13087 Return the variables declared in the `GIMPLE_BIND' statement `G'. 13088 13089 -- GIMPLE function: void gimple_bind_set_vars (gimple g, tree vars) 13090 Set `VARS' to be the set of variables declared in the `GIMPLE_BIND' 13091 statement `G'. 13092 13093 -- GIMPLE function: void gimple_bind_append_vars (gimple g, tree vars) 13094 Append `VARS' to the set of variables declared in the `GIMPLE_BIND' 13095 statement `G'. 13096 13097 -- GIMPLE function: gimple_seq gimple_bind_body (gimple g) 13098 Return the GIMPLE sequence contained in the `GIMPLE_BIND' statement 13099 `G'. 13100 13101 -- GIMPLE function: void gimple_bind_set_body (gimple g, gimple_seq 13102 seq) 13103 Set `SEQ' to be sequence contained in the `GIMPLE_BIND' statement 13104 `G'. 13105 13106 -- GIMPLE function: void gimple_bind_add_stmt (gimple gs, gimple stmt) 13107 Append a statement to the end of a `GIMPLE_BIND''s body. 13108 13109 -- GIMPLE function: void gimple_bind_add_seq (gimple gs, gimple_seq 13110 seq) 13111 Append a sequence of statements to the end of a `GIMPLE_BIND''s 13112 body. 13113 13114 -- GIMPLE function: tree gimple_bind_block (gimple g) 13115 Return the `TREE_BLOCK' node associated with `GIMPLE_BIND' 13116 statement `G'. This is analogous to the `BIND_EXPR_BLOCK' field in 13117 trees. 13118 13119 -- GIMPLE function: void gimple_bind_set_block (gimple g, tree block) 13120 Set `BLOCK' to be the `TREE_BLOCK' node associated with 13121 `GIMPLE_BIND' statement `G'. 13122 13123 13124 File: gccint.info, Node: `GIMPLE_CALL', Next: `GIMPLE_CATCH', Prev: `GIMPLE_BIND', Up: Tuple specific accessors 13125 13126 12.7.4 `GIMPLE_CALL' 13127 -------------------- 13128 13129 -- GIMPLE function: gimple gimple_build_call (tree fn, unsigned nargs, 13130 ...) 13131 Build a `GIMPLE_CALL' statement to function `FN'. The argument 13132 `FN' must be either a `FUNCTION_DECL' or a gimple call address as 13133 determined by `is_gimple_call_addr'. `NARGS' are the number of 13134 arguments. The rest of the arguments follow the argument `NARGS', 13135 and must be trees that are valid as rvalues in gimple (i.e., each 13136 operand is validated with `is_gimple_operand'). 13137 13138 -- GIMPLE function: gimple gimple_build_call_from_tree (tree call_expr) 13139 Build a `GIMPLE_CALL' from a `CALL_EXPR' node. The arguments and 13140 the function are taken from the expression directly. This routine 13141 assumes that `call_expr' is already in GIMPLE form. That is, its 13142 operands are GIMPLE values and the function call needs no further 13143 simplification. All the call flags in `call_expr' are copied over 13144 to the new `GIMPLE_CALL'. 13145 13146 -- GIMPLE function: gimple gimple_build_call_vec (tree fn, `VEC'(tree, 13147 heap) *args) 13148 Identical to `gimple_build_call' but the arguments are stored in a 13149 `VEC'(). 13150 13151 -- GIMPLE function: tree gimple_call_lhs (gimple g) 13152 Return the `LHS' of call statement `G'. 13153 13154 -- GIMPLE function: tree *gimple_call_lhs_ptr (gimple g) 13155 Return a pointer to the `LHS' of call statement `G'. 13156 13157 -- GIMPLE function: void gimple_call_set_lhs (gimple g, tree lhs) 13158 Set `LHS' to be the `LHS' operand of call statement `G'. 13159 13160 -- GIMPLE function: tree gimple_call_fn (gimple g) 13161 Return the tree node representing the function called by call 13162 statement `G'. 13163 13164 -- GIMPLE function: void gimple_call_set_fn (gimple g, tree fn) 13165 Set `FN' to be the function called by call statement `G'. This has 13166 to be a gimple value specifying the address of the called function. 13167 13168 -- GIMPLE function: tree gimple_call_fndecl (gimple g) 13169 If a given `GIMPLE_CALL''s callee is a `FUNCTION_DECL', return it. 13170 Otherwise return `NULL'. This function is analogous to 13171 `get_callee_fndecl' in `GENERIC'. 13172 13173 -- GIMPLE function: tree gimple_call_set_fndecl (gimple g, tree fndecl) 13174 Set the called function to `FNDECL'. 13175 13176 -- GIMPLE function: tree gimple_call_return_type (gimple g) 13177 Return the type returned by call statement `G'. 13178 13179 -- GIMPLE function: tree gimple_call_chain (gimple g) 13180 Return the static chain for call statement `G'. 13181 13182 -- GIMPLE function: void gimple_call_set_chain (gimple g, tree chain) 13183 Set `CHAIN' to be the static chain for call statement `G'. 13184 13185 -- GIMPLE function: gimple_call_num_args (gimple g) 13186 Return the number of arguments used by call statement `G'. 13187 13188 -- GIMPLE function: tree gimple_call_arg (gimple g, unsigned index) 13189 Return the argument at position `INDEX' for call statement `G'. 13190 The first argument is 0. 13191 13192 -- GIMPLE function: tree *gimple_call_arg_ptr (gimple g, unsigned 13193 index) 13194 Return a pointer to the argument at position `INDEX' for call 13195 statement `G'. 13196 13197 -- GIMPLE function: void gimple_call_set_arg (gimple g, unsigned 13198 index, tree arg) 13199 Set `ARG' to be the argument at position `INDEX' for call statement 13200 `G'. 13201 13202 -- GIMPLE function: void gimple_call_set_tail (gimple s) 13203 Mark call statement `S' as being a tail call (i.e., a call just 13204 before the exit of a function). These calls are candidate for tail 13205 call optimization. 13206 13207 -- GIMPLE function: bool gimple_call_tail_p (gimple s) 13208 Return true if `GIMPLE_CALL' `S' is marked as a tail call. 13209 13210 -- GIMPLE function: void gimple_call_mark_uninlinable (gimple s) 13211 Mark `GIMPLE_CALL' `S' as being uninlinable. 13212 13213 -- GIMPLE function: bool gimple_call_cannot_inline_p (gimple s) 13214 Return true if `GIMPLE_CALL' `S' cannot be inlined. 13215 13216 -- GIMPLE function: bool gimple_call_noreturn_p (gimple s) 13217 Return true if `S' is a noreturn call. 13218 13219 -- GIMPLE function: gimple gimple_call_copy_skip_args (gimple stmt, 13220 bitmap args_to_skip) 13221 Build a `GIMPLE_CALL' identical to `STMT' but skipping the 13222 arguments in the positions marked by the set `ARGS_TO_SKIP'. 13223 13224 13225 File: gccint.info, Node: `GIMPLE_CATCH', Next: `GIMPLE_CHANGE_DYNAMIC_TYPE', Prev: `GIMPLE_CALL', Up: Tuple specific accessors 13226 13227 12.7.5 `GIMPLE_CATCH' 13228 --------------------- 13229 13230 -- GIMPLE function: gimple gimple_build_catch (tree types, gimple_seq 13231 handler) 13232 Build a `GIMPLE_CATCH' statement. `TYPES' are the tree types this 13233 catch handles. `HANDLER' is a sequence of statements with the code 13234 for the handler. 13235 13236 -- GIMPLE function: tree gimple_catch_types (gimple g) 13237 Return the types handled by `GIMPLE_CATCH' statement `G'. 13238 13239 -- GIMPLE function: tree *gimple_catch_types_ptr (gimple g) 13240 Return a pointer to the types handled by `GIMPLE_CATCH' statement 13241 `G'. 13242 13243 -- GIMPLE function: gimple_seq gimple_catch_handler (gimple g) 13244 Return the GIMPLE sequence representing the body of the handler of 13245 `GIMPLE_CATCH' statement `G'. 13246 13247 -- GIMPLE function: void gimple_catch_set_types (gimple g, tree t) 13248 Set `T' to be the set of types handled by `GIMPLE_CATCH' `G'. 13249 13250 -- GIMPLE function: void gimple_catch_set_handler (gimple g, 13251 gimple_seq handler) 13252 Set `HANDLER' to be the body of `GIMPLE_CATCH' `G'. 13253 13254 13255 File: gccint.info, Node: `GIMPLE_CHANGE_DYNAMIC_TYPE', Next: `GIMPLE_COND', Prev: `GIMPLE_CATCH', Up: Tuple specific accessors 13256 13257 12.7.6 `GIMPLE_CHANGE_DYNAMIC_TYPE' 13258 ----------------------------------- 13259 13260 -- GIMPLE function: gimple gimple_build_cdt (tree type, tree ptr) 13261 Build a `GIMPLE_CHANGE_DYNAMIC_TYPE' statement. `TYPE' is the new 13262 type for the location `PTR'. 13263 13264 -- GIMPLE function: tree gimple_cdt_new_type (gimple g) 13265 Return the new type set by `GIMPLE_CHANGE_DYNAMIC_TYPE' statement 13266 `G'. 13267 13268 -- GIMPLE function: tree *gimple_cdt_new_type_ptr (gimple g) 13269 Return a pointer to the new type set by 13270 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13271 13272 -- GIMPLE function: void gimple_cdt_set_new_type (gimple g, tree 13273 new_type) 13274 Set `NEW_TYPE' to be the type returned by 13275 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13276 13277 -- GIMPLE function: tree gimple_cdt_location (gimple g) 13278 Return the location affected by `GIMPLE_CHANGE_DYNAMIC_TYPE' 13279 statement `G'. 13280 13281 -- GIMPLE function: tree *gimple_cdt_location_ptr (gimple g) 13282 Return a pointer to the location affected by 13283 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13284 13285 -- GIMPLE function: void gimple_cdt_set_location (gimple g, tree ptr) 13286 Set `PTR' to be the location affected by 13287 `GIMPLE_CHANGE_DYNAMIC_TYPE' statement `G'. 13288 13289 13290 File: gccint.info, Node: `GIMPLE_COND', Next: `GIMPLE_EH_FILTER', Prev: `GIMPLE_CHANGE_DYNAMIC_TYPE', Up: Tuple specific accessors 13291 13292 12.7.7 `GIMPLE_COND' 13293 -------------------- 13294 13295 -- GIMPLE function: gimple gimple_build_cond (enum tree_code 13296 pred_code, tree lhs, tree rhs, tree t_label, tree f_label) 13297 Build a `GIMPLE_COND' statement. `A' `GIMPLE_COND' statement 13298 compares `LHS' and `RHS' and if the condition in `PRED_CODE' is 13299 true, jump to the label in `t_label', otherwise jump to the label 13300 in `f_label'. `PRED_CODE' are relational operator tree codes like 13301 `EQ_EXPR', `LT_EXPR', `LE_EXPR', `NE_EXPR', etc. 13302 13303 -- GIMPLE function: gimple gimple_build_cond_from_tree (tree cond, 13304 tree t_label, tree f_label) 13305 Build a `GIMPLE_COND' statement from the conditional expression 13306 tree `COND'. `T_LABEL' and `F_LABEL' are as in 13307 `gimple_build_cond'. 13308 13309 -- GIMPLE function: enum tree_code gimple_cond_code (gimple g) 13310 Return the code of the predicate computed by conditional statement 13311 `G'. 13312 13313 -- GIMPLE function: void gimple_cond_set_code (gimple g, enum 13314 tree_code code) 13315 Set `CODE' to be the predicate code for the conditional statement 13316 `G'. 13317 13318 -- GIMPLE function: tree gimple_cond_lhs (gimple g) 13319 Return the `LHS' of the predicate computed by conditional statement 13320 `G'. 13321 13322 -- GIMPLE function: void gimple_cond_set_lhs (gimple g, tree lhs) 13323 Set `LHS' to be the `LHS' operand of the predicate computed by 13324 conditional statement `G'. 13325 13326 -- GIMPLE function: tree gimple_cond_rhs (gimple g) 13327 Return the `RHS' operand of the predicate computed by conditional 13328 `G'. 13329 13330 -- GIMPLE function: void gimple_cond_set_rhs (gimple g, tree rhs) 13331 Set `RHS' to be the `RHS' operand of the predicate computed by 13332 conditional statement `G'. 13333 13334 -- GIMPLE function: tree gimple_cond_true_label (gimple g) 13335 Return the label used by conditional statement `G' when its 13336 predicate evaluates to true. 13337 13338 -- GIMPLE function: void gimple_cond_set_true_label (gimple g, tree 13339 label) 13340 Set `LABEL' to be the label used by conditional statement `G' when 13341 its predicate evaluates to true. 13342 13343 -- GIMPLE function: void gimple_cond_set_false_label (gimple g, tree 13344 label) 13345 Set `LABEL' to be the label used by conditional statement `G' when 13346 its predicate evaluates to false. 13347 13348 -- GIMPLE function: tree gimple_cond_false_label (gimple g) 13349 Return the label used by conditional statement `G' when its 13350 predicate evaluates to false. 13351 13352 -- GIMPLE function: void gimple_cond_make_false (gimple g) 13353 Set the conditional `COND_STMT' to be of the form 'if (1 == 0)'. 13354 13355 -- GIMPLE function: void gimple_cond_make_true (gimple g) 13356 Set the conditional `COND_STMT' to be of the form 'if (1 == 1)'. 13357 13358 13359 File: gccint.info, Node: `GIMPLE_EH_FILTER', Next: `GIMPLE_LABEL', Prev: `GIMPLE_COND', Up: Tuple specific accessors 13360 13361 12.7.8 `GIMPLE_EH_FILTER' 13362 ------------------------- 13363 13364 -- GIMPLE function: gimple gimple_build_eh_filter (tree types, 13365 gimple_seq failure) 13366 Build a `GIMPLE_EH_FILTER' statement. `TYPES' are the filter's 13367 types. `FAILURE' is a sequence with the filter's failure action. 13368 13369 -- GIMPLE function: tree gimple_eh_filter_types (gimple g) 13370 Return the types handled by `GIMPLE_EH_FILTER' statement `G'. 13371 13372 -- GIMPLE function: tree *gimple_eh_filter_types_ptr (gimple g) 13373 Return a pointer to the types handled by `GIMPLE_EH_FILTER' 13374 statement `G'. 13375 13376 -- GIMPLE function: gimple_seq gimple_eh_filter_failure (gimple g) 13377 Return the sequence of statement to execute when `GIMPLE_EH_FILTER' 13378 statement fails. 13379 13380 -- GIMPLE function: void gimple_eh_filter_set_types (gimple g, tree 13381 types) 13382 Set `TYPES' to be the set of types handled by `GIMPLE_EH_FILTER' 13383 `G'. 13384 13385 -- GIMPLE function: void gimple_eh_filter_set_failure (gimple g, 13386 gimple_seq failure) 13387 Set `FAILURE' to be the sequence of statements to execute on 13388 failure for `GIMPLE_EH_FILTER' `G'. 13389 13390 -- GIMPLE function: bool gimple_eh_filter_must_not_throw (gimple g) 13391 Return the `EH_FILTER_MUST_NOT_THROW' flag. 13392 13393 -- GIMPLE function: void gimple_eh_filter_set_must_not_throw (gimple 13394 g, bool mntp) 13395 Set the `EH_FILTER_MUST_NOT_THROW' flag. 13396 13397 13398 File: gccint.info, Node: `GIMPLE_LABEL', Next: `GIMPLE_NOP', Prev: `GIMPLE_EH_FILTER', Up: Tuple specific accessors 13399 13400 12.7.9 `GIMPLE_LABEL' 13401 --------------------- 13402 13403 -- GIMPLE function: gimple gimple_build_label (tree label) 13404 Build a `GIMPLE_LABEL' statement with corresponding to the tree 13405 label, `LABEL'. 13406 13407 -- GIMPLE function: tree gimple_label_label (gimple g) 13408 Return the `LABEL_DECL' node used by `GIMPLE_LABEL' statement `G'. 13409 13410 -- GIMPLE function: void gimple_label_set_label (gimple g, tree label) 13411 Set `LABEL' to be the `LABEL_DECL' node used by `GIMPLE_LABEL' 13412 statement `G'. 13413 13414 -- GIMPLE function: gimple gimple_build_goto (tree dest) 13415 Build a `GIMPLE_GOTO' statement to label `DEST'. 13416 13417 -- GIMPLE function: tree gimple_goto_dest (gimple g) 13418 Return the destination of the unconditional jump `G'. 13419 13420 -- GIMPLE function: void gimple_goto_set_dest (gimple g, tree dest) 13421 Set `DEST' to be the destination of the unconditional jump `G'. 13422 13423 13424 File: gccint.info, Node: `GIMPLE_NOP', Next: `GIMPLE_OMP_ATOMIC_LOAD', Prev: `GIMPLE_LABEL', Up: Tuple specific accessors 13425 13426 12.7.10 `GIMPLE_NOP' 13427 -------------------- 13428 13429 -- GIMPLE function: gimple gimple_build_nop (void) 13430 Build a `GIMPLE_NOP' statement. 13431 13432 -- GIMPLE function: bool gimple_nop_p (gimple g) 13433 Returns `TRUE' if statement `G' is a `GIMPLE_NOP'. 13434 13435 13436 File: gccint.info, Node: `GIMPLE_OMP_ATOMIC_LOAD', Next: `GIMPLE_OMP_ATOMIC_STORE', Prev: `GIMPLE_NOP', Up: Tuple specific accessors 13437 13438 12.7.11 `GIMPLE_OMP_ATOMIC_LOAD' 13439 -------------------------------- 13440 13441 -- GIMPLE function: gimple gimple_build_omp_atomic_load (tree lhs, 13442 tree rhs) 13443 Build a `GIMPLE_OMP_ATOMIC_LOAD' statement. `LHS' is the left-hand 13444 side of the assignment. `RHS' is the right-hand side of the 13445 assignment. 13446 13447 -- GIMPLE function: void gimple_omp_atomic_load_set_lhs (gimple g, 13448 tree lhs) 13449 Set the `LHS' of an atomic load. 13450 13451 -- GIMPLE function: tree gimple_omp_atomic_load_lhs (gimple g) 13452 Get the `LHS' of an atomic load. 13453 13454 -- GIMPLE function: void gimple_omp_atomic_load_set_rhs (gimple g, 13455 tree rhs) 13456 Set the `RHS' of an atomic set. 13457 13458 -- GIMPLE function: tree gimple_omp_atomic_load_rhs (gimple g) 13459 Get the `RHS' of an atomic set. 13460 13461 13462 File: gccint.info, Node: `GIMPLE_OMP_ATOMIC_STORE', Next: `GIMPLE_OMP_CONTINUE', Prev: `GIMPLE_OMP_ATOMIC_LOAD', Up: Tuple specific accessors 13463 13464 12.7.12 `GIMPLE_OMP_ATOMIC_STORE' 13465 --------------------------------- 13466 13467 -- GIMPLE function: gimple gimple_build_omp_atomic_store (tree val) 13468 Build a `GIMPLE_OMP_ATOMIC_STORE' statement. `VAL' is the value to 13469 be stored. 13470 13471 -- GIMPLE function: void gimple_omp_atomic_store_set_val (gimple g, 13472 tree val) 13473 Set the value being stored in an atomic store. 13474 13475 -- GIMPLE function: tree gimple_omp_atomic_store_val (gimple g) 13476 Return the value being stored in an atomic store. 13477 13478 13479 File: gccint.info, Node: `GIMPLE_OMP_CONTINUE', Next: `GIMPLE_OMP_CRITICAL', Prev: `GIMPLE_OMP_ATOMIC_STORE', Up: Tuple specific accessors 13480 13481 12.7.13 `GIMPLE_OMP_CONTINUE' 13482 ----------------------------- 13483 13484 -- GIMPLE function: gimple gimple_build_omp_continue (tree 13485 control_def, tree control_use) 13486 Build a `GIMPLE_OMP_CONTINUE' statement. `CONTROL_DEF' is the 13487 definition of the control variable. `CONTROL_USE' is the use of 13488 the control variable. 13489 13490 -- GIMPLE function: tree gimple_omp_continue_control_def (gimple s) 13491 Return the definition of the control variable on a 13492 `GIMPLE_OMP_CONTINUE' in `S'. 13493 13494 -- GIMPLE function: tree gimple_omp_continue_control_def_ptr (gimple s) 13495 Same as above, but return the pointer. 13496 13497 -- GIMPLE function: tree gimple_omp_continue_set_control_def (gimple s) 13498 Set the control variable definition for a `GIMPLE_OMP_CONTINUE' 13499 statement in `S'. 13500 13501 -- GIMPLE function: tree gimple_omp_continue_control_use (gimple s) 13502 Return the use of the control variable on a `GIMPLE_OMP_CONTINUE' 13503 in `S'. 13504 13505 -- GIMPLE function: tree gimple_omp_continue_control_use_ptr (gimple s) 13506 Same as above, but return the pointer. 13507 13508 -- GIMPLE function: tree gimple_omp_continue_set_control_use (gimple s) 13509 Set the control variable use for a `GIMPLE_OMP_CONTINUE' statement 13510 in `S'. 13511 13512 13513 File: gccint.info, Node: `GIMPLE_OMP_CRITICAL', Next: `GIMPLE_OMP_FOR', Prev: `GIMPLE_OMP_CONTINUE', Up: Tuple specific accessors 13514 13515 12.7.14 `GIMPLE_OMP_CRITICAL' 13516 ----------------------------- 13517 13518 -- GIMPLE function: gimple gimple_build_omp_critical (gimple_seq body, 13519 tree name) 13520 Build a `GIMPLE_OMP_CRITICAL' statement. `BODY' is the sequence of 13521 statements for which only one thread can execute. `NAME' is an 13522 optional identifier for this critical block. 13523 13524 -- GIMPLE function: tree gimple_omp_critical_name (gimple g) 13525 Return the name associated with `OMP_CRITICAL' statement `G'. 13526 13527 -- GIMPLE function: tree *gimple_omp_critical_name_ptr (gimple g) 13528 Return a pointer to the name associated with `OMP' critical 13529 statement `G'. 13530 13531 -- GIMPLE function: void gimple_omp_critical_set_name (gimple g, tree 13532 name) 13533 Set `NAME' to be the name associated with `OMP' critical statement 13534 `G'. 13535 13536 13537 File: gccint.info, Node: `GIMPLE_OMP_FOR', Next: `GIMPLE_OMP_MASTER', Prev: `GIMPLE_OMP_CRITICAL', Up: Tuple specific accessors 13538 13539 12.7.15 `GIMPLE_OMP_FOR' 13540 ------------------------ 13541 13542 -- GIMPLE function: gimple gimple_build_omp_for (gimple_seq body, tree 13543 clauses, tree index, tree initial, tree final, tree incr, 13544 gimple_seq pre_body, enum tree_code omp_for_cond) 13545 Build a `GIMPLE_OMP_FOR' statement. `BODY' is sequence of 13546 statements inside the for loop. `CLAUSES', are any of the `OMP' 13547 loop construct's clauses: private, firstprivate, lastprivate, 13548 reductions, ordered, schedule, and nowait. `PRE_BODY' is the 13549 sequence of statements that are loop invariant. `INDEX' is the 13550 index variable. `INITIAL' is the initial value of `INDEX'. 13551 `FINAL' is final value of `INDEX'. OMP_FOR_COND is the predicate 13552 used to compare `INDEX' and `FINAL'. `INCR' is the increment 13553 expression. 13554 13555 -- GIMPLE function: tree gimple_omp_for_clauses (gimple g) 13556 Return the clauses associated with `OMP_FOR' `G'. 13557 13558 -- GIMPLE function: tree *gimple_omp_for_clauses_ptr (gimple g) 13559 Return a pointer to the `OMP_FOR' `G'. 13560 13561 -- GIMPLE function: void gimple_omp_for_set_clauses (gimple g, tree 13562 clauses) 13563 Set `CLAUSES' to be the list of clauses associated with `OMP_FOR' 13564 `G'. 13565 13566 -- GIMPLE function: tree gimple_omp_for_index (gimple g) 13567 Return the index variable for `OMP_FOR' `G'. 13568 13569 -- GIMPLE function: tree *gimple_omp_for_index_ptr (gimple g) 13570 Return a pointer to the index variable for `OMP_FOR' `G'. 13571 13572 -- GIMPLE function: void gimple_omp_for_set_index (gimple g, tree 13573 index) 13574 Set `INDEX' to be the index variable for `OMP_FOR' `G'. 13575 13576 -- GIMPLE function: tree gimple_omp_for_initial (gimple g) 13577 Return the initial value for `OMP_FOR' `G'. 13578 13579 -- GIMPLE function: tree *gimple_omp_for_initial_ptr (gimple g) 13580 Return a pointer to the initial value for `OMP_FOR' `G'. 13581 13582 -- GIMPLE function: void gimple_omp_for_set_initial (gimple g, tree 13583 initial) 13584 Set `INITIAL' to be the initial value for `OMP_FOR' `G'. 13585 13586 -- GIMPLE function: tree gimple_omp_for_final (gimple g) 13587 Return the final value for `OMP_FOR' `G'. 13588 13589 -- GIMPLE function: tree *gimple_omp_for_final_ptr (gimple g) 13590 turn a pointer to the final value for `OMP_FOR' `G'. 13591 13592 -- GIMPLE function: void gimple_omp_for_set_final (gimple g, tree 13593 final) 13594 Set `FINAL' to be the final value for `OMP_FOR' `G'. 13595 13596 -- GIMPLE function: tree gimple_omp_for_incr (gimple g) 13597 Return the increment value for `OMP_FOR' `G'. 13598 13599 -- GIMPLE function: tree *gimple_omp_for_incr_ptr (gimple g) 13600 Return a pointer to the increment value for `OMP_FOR' `G'. 13601 13602 -- GIMPLE function: void gimple_omp_for_set_incr (gimple g, tree incr) 13603 Set `INCR' to be the increment value for `OMP_FOR' `G'. 13604 13605 -- GIMPLE function: gimple_seq gimple_omp_for_pre_body (gimple g) 13606 Return the sequence of statements to execute before the `OMP_FOR' 13607 statement `G' starts. 13608 13609 -- GIMPLE function: void gimple_omp_for_set_pre_body (gimple g, 13610 gimple_seq pre_body) 13611 Set `PRE_BODY' to be the sequence of statements to execute before 13612 the `OMP_FOR' statement `G' starts. 13613 13614 -- GIMPLE function: void gimple_omp_for_set_cond (gimple g, enum 13615 tree_code cond) 13616 Set `COND' to be the condition code for `OMP_FOR' `G'. 13617 13618 -- GIMPLE function: enum tree_code gimple_omp_for_cond (gimple g) 13619 Return the condition code associated with `OMP_FOR' `G'. 13620 13621 13622 File: gccint.info, Node: `GIMPLE_OMP_MASTER', Next: `GIMPLE_OMP_ORDERED', Prev: `GIMPLE_OMP_FOR', Up: Tuple specific accessors 13623 13624 12.7.16 `GIMPLE_OMP_MASTER' 13625 --------------------------- 13626 13627 -- GIMPLE function: gimple gimple_build_omp_master (gimple_seq body) 13628 Build a `GIMPLE_OMP_MASTER' statement. `BODY' is the sequence of 13629 statements to be executed by just the master. 13630 13631 13632 File: gccint.info, Node: `GIMPLE_OMP_ORDERED', Next: `GIMPLE_OMP_PARALLEL', Prev: `GIMPLE_OMP_MASTER', Up: Tuple specific accessors 13633 13634 12.7.17 `GIMPLE_OMP_ORDERED' 13635 ---------------------------- 13636 13637 -- GIMPLE function: gimple gimple_build_omp_ordered (gimple_seq body) 13638 Build a `GIMPLE_OMP_ORDERED' statement. 13639 13640 `BODY' is the sequence of statements inside a loop that will executed 13641 in sequence. 13642 13643 13644 File: gccint.info, Node: `GIMPLE_OMP_PARALLEL', Next: `GIMPLE_OMP_RETURN', Prev: `GIMPLE_OMP_ORDERED', Up: Tuple specific accessors 13645 13646 12.7.18 `GIMPLE_OMP_PARALLEL' 13647 ----------------------------- 13648 13649 -- GIMPLE function: gimple gimple_build_omp_parallel (gimple_seq body, 13650 tree clauses, tree child_fn, tree data_arg) 13651 Build a `GIMPLE_OMP_PARALLEL' statement. 13652 13653 `BODY' is sequence of statements which are executed in parallel. 13654 `CLAUSES', are the `OMP' parallel construct's clauses. `CHILD_FN' is 13655 the function created for the parallel threads to execute. `DATA_ARG' 13656 are the shared data argument(s). 13657 13658 -- GIMPLE function: bool gimple_omp_parallel_combined_p (gimple g) 13659 Return true if `OMP' parallel statement `G' has the 13660 `GF_OMP_PARALLEL_COMBINED' flag set. 13661 13662 -- GIMPLE function: void gimple_omp_parallel_set_combined_p (gimple g) 13663 Set the `GF_OMP_PARALLEL_COMBINED' field in `OMP' parallel 13664 statement `G'. 13665 13666 -- GIMPLE function: gimple_seq gimple_omp_body (gimple g) 13667 Return the body for the `OMP' statement `G'. 13668 13669 -- GIMPLE function: void gimple_omp_set_body (gimple g, gimple_seq 13670 body) 13671 Set `BODY' to be the body for the `OMP' statement `G'. 13672 13673 -- GIMPLE function: tree gimple_omp_parallel_clauses (gimple g) 13674 Return the clauses associated with `OMP_PARALLEL' `G'. 13675 13676 -- GIMPLE function: tree *gimple_omp_parallel_clauses_ptr (gimple g) 13677 Return a pointer to the clauses associated with `OMP_PARALLEL' `G'. 13678 13679 -- GIMPLE function: void gimple_omp_parallel_set_clauses (gimple g, 13680 tree clauses) 13681 Set `CLAUSES' to be the list of clauses associated with 13682 `OMP_PARALLEL' `G'. 13683 13684 -- GIMPLE function: tree gimple_omp_parallel_child_fn (gimple g) 13685 Return the child function used to hold the body of `OMP_PARALLEL' 13686 `G'. 13687 13688 -- GIMPLE function: tree *gimple_omp_parallel_child_fn_ptr (gimple g) 13689 Return a pointer to the child function used to hold the body of 13690 `OMP_PARALLEL' `G'. 13691 13692 -- GIMPLE function: void gimple_omp_parallel_set_child_fn (gimple g, 13693 tree child_fn) 13694 Set `CHILD_FN' to be the child function for `OMP_PARALLEL' `G'. 13695 13696 -- GIMPLE function: tree gimple_omp_parallel_data_arg (gimple g) 13697 Return the artificial argument used to send variables and values 13698 from the parent to the children threads in `OMP_PARALLEL' `G'. 13699 13700 -- GIMPLE function: tree *gimple_omp_parallel_data_arg_ptr (gimple g) 13701 Return a pointer to the data argument for `OMP_PARALLEL' `G'. 13702 13703 -- GIMPLE function: void gimple_omp_parallel_set_data_arg (gimple g, 13704 tree data_arg) 13705 Set `DATA_ARG' to be the data argument for `OMP_PARALLEL' `G'. 13706 13707 -- GIMPLE function: bool is_gimple_omp (gimple stmt) 13708 Returns true when the gimple statement `STMT' is any of the OpenMP 13709 types. 13710 13711 13712 File: gccint.info, Node: `GIMPLE_OMP_RETURN', Next: `GIMPLE_OMP_SECTION', Prev: `GIMPLE_OMP_PARALLEL', Up: Tuple specific accessors 13713 13714 12.7.19 `GIMPLE_OMP_RETURN' 13715 --------------------------- 13716 13717 -- GIMPLE function: gimple gimple_build_omp_return (bool wait_p) 13718 Build a `GIMPLE_OMP_RETURN' statement. `WAIT_P' is true if this is 13719 a non-waiting return. 13720 13721 -- GIMPLE function: void gimple_omp_return_set_nowait (gimple s) 13722 Set the nowait flag on `GIMPLE_OMP_RETURN' statement `S'. 13723 13724 -- GIMPLE function: bool gimple_omp_return_nowait_p (gimple g) 13725 Return true if `OMP' return statement `G' has the 13726 `GF_OMP_RETURN_NOWAIT' flag set. 13727 13728 13729 File: gccint.info, Node: `GIMPLE_OMP_SECTION', Next: `GIMPLE_OMP_SECTIONS', Prev: `GIMPLE_OMP_RETURN', Up: Tuple specific accessors 13730 13731 12.7.20 `GIMPLE_OMP_SECTION' 13732 ---------------------------- 13733 13734 -- GIMPLE function: gimple gimple_build_omp_section (gimple_seq body) 13735 Build a `GIMPLE_OMP_SECTION' statement for a sections statement. 13736 13737 `BODY' is the sequence of statements in the section. 13738 13739 -- GIMPLE function: bool gimple_omp_section_last_p (gimple g) 13740 Return true if `OMP' section statement `G' has the 13741 `GF_OMP_SECTION_LAST' flag set. 13742 13743 -- GIMPLE function: void gimple_omp_section_set_last (gimple g) 13744 Set the `GF_OMP_SECTION_LAST' flag on `G'. 13745 13746 13747 File: gccint.info, Node: `GIMPLE_OMP_SECTIONS', Next: `GIMPLE_OMP_SINGLE', Prev: `GIMPLE_OMP_SECTION', Up: Tuple specific accessors 13748 13749 12.7.21 `GIMPLE_OMP_SECTIONS' 13750 ----------------------------- 13751 13752 -- GIMPLE function: gimple gimple_build_omp_sections (gimple_seq body, 13753 tree clauses) 13754 Build a `GIMPLE_OMP_SECTIONS' statement. `BODY' is a sequence of 13755 section statements. `CLAUSES' are any of the `OMP' sections 13756 construct's clauses: private, firstprivate, lastprivate, 13757 reduction, and nowait. 13758 13759 -- GIMPLE function: gimple gimple_build_omp_sections_switch (void) 13760 Build a `GIMPLE_OMP_SECTIONS_SWITCH' statement. 13761 13762 -- GIMPLE function: tree gimple_omp_sections_control (gimple g) 13763 Return the control variable associated with the 13764 `GIMPLE_OMP_SECTIONS' in `G'. 13765 13766 -- GIMPLE function: tree *gimple_omp_sections_control_ptr (gimple g) 13767 Return a pointer to the clauses associated with the 13768 `GIMPLE_OMP_SECTIONS' in `G'. 13769 13770 -- GIMPLE function: void gimple_omp_sections_set_control (gimple g, 13771 tree control) 13772 Set `CONTROL' to be the set of clauses associated with the 13773 `GIMPLE_OMP_SECTIONS' in `G'. 13774 13775 -- GIMPLE function: tree gimple_omp_sections_clauses (gimple g) 13776 Return the clauses associated with `OMP_SECTIONS' `G'. 13777 13778 -- GIMPLE function: tree *gimple_omp_sections_clauses_ptr (gimple g) 13779 Return a pointer to the clauses associated with `OMP_SECTIONS' `G'. 13780 13781 -- GIMPLE function: void gimple_omp_sections_set_clauses (gimple g, 13782 tree clauses) 13783 Set `CLAUSES' to be the set of clauses associated with 13784 `OMP_SECTIONS' `G'. 13785 13786 13787 File: gccint.info, Node: `GIMPLE_OMP_SINGLE', Next: `GIMPLE_PHI', Prev: `GIMPLE_OMP_SECTIONS', Up: Tuple specific accessors 13788 13789 12.7.22 `GIMPLE_OMP_SINGLE' 13790 --------------------------- 13791 13792 -- GIMPLE function: gimple gimple_build_omp_single (gimple_seq body, 13793 tree clauses) 13794 Build a `GIMPLE_OMP_SINGLE' statement. `BODY' is the sequence of 13795 statements that will be executed once. `CLAUSES' are any of the 13796 `OMP' single construct's clauses: private, firstprivate, 13797 copyprivate, nowait. 13798 13799 -- GIMPLE function: tree gimple_omp_single_clauses (gimple g) 13800 Return the clauses associated with `OMP_SINGLE' `G'. 13801 13802 -- GIMPLE function: tree *gimple_omp_single_clauses_ptr (gimple g) 13803 Return a pointer to the clauses associated with `OMP_SINGLE' `G'. 13804 13805 -- GIMPLE function: void gimple_omp_single_set_clauses (gimple g, tree 13806 clauses) 13807 Set `CLAUSES' to be the clauses associated with `OMP_SINGLE' `G'. 13808 13809 13810 File: gccint.info, Node: `GIMPLE_PHI', Next: `GIMPLE_RESX', Prev: `GIMPLE_OMP_SINGLE', Up: Tuple specific accessors 13811 13812 12.7.23 `GIMPLE_PHI' 13813 -------------------- 13814 13815 -- GIMPLE function: gimple make_phi_node (tree var, int len) 13816 Build a `PHI' node with len argument slots for variable var. 13817 13818 -- GIMPLE function: unsigned gimple_phi_capacity (gimple g) 13819 Return the maximum number of arguments supported by `GIMPLE_PHI' 13820 `G'. 13821 13822 -- GIMPLE function: unsigned gimple_phi_num_args (gimple g) 13823 Return the number of arguments in `GIMPLE_PHI' `G'. This must 13824 always be exactly the number of incoming edges for the basic block 13825 holding `G'. 13826 13827 -- GIMPLE function: tree gimple_phi_result (gimple g) 13828 Return the `SSA' name created by `GIMPLE_PHI' `G'. 13829 13830 -- GIMPLE function: tree *gimple_phi_result_ptr (gimple g) 13831 Return a pointer to the `SSA' name created by `GIMPLE_PHI' `G'. 13832 13833 -- GIMPLE function: void gimple_phi_set_result (gimple g, tree result) 13834 Set `RESULT' to be the `SSA' name created by `GIMPLE_PHI' `G'. 13835 13836 -- GIMPLE function: struct phi_arg_d *gimple_phi_arg (gimple g, index) 13837 Return the `PHI' argument corresponding to incoming edge `INDEX' 13838 for `GIMPLE_PHI' `G'. 13839 13840 -- GIMPLE function: void gimple_phi_set_arg (gimple g, index, struct 13841 phi_arg_d * phiarg) 13842 Set `PHIARG' to be the argument corresponding to incoming edge 13843 `INDEX' for `GIMPLE_PHI' `G'. 13844 13845 13846 File: gccint.info, Node: `GIMPLE_RESX', Next: `GIMPLE_RETURN', Prev: `GIMPLE_PHI', Up: Tuple specific accessors 13847 13848 12.7.24 `GIMPLE_RESX' 13849 --------------------- 13850 13851 -- GIMPLE function: gimple gimple_build_resx (int region) 13852 Build a `GIMPLE_RESX' statement which is a statement. This 13853 statement is a placeholder for _Unwind_Resume before we know if a 13854 function call or a branch is needed. `REGION' is the exception 13855 region from which control is flowing. 13856 13857 -- GIMPLE function: int gimple_resx_region (gimple g) 13858 Return the region number for `GIMPLE_RESX' `G'. 13859 13860 -- GIMPLE function: void gimple_resx_set_region (gimple g, int region) 13861 Set `REGION' to be the region number for `GIMPLE_RESX' `G'. 13862 13863 13864 File: gccint.info, Node: `GIMPLE_RETURN', Next: `GIMPLE_SWITCH', Prev: `GIMPLE_RESX', Up: Tuple specific accessors 13865 13866 12.7.25 `GIMPLE_RETURN' 13867 ----------------------- 13868 13869 -- GIMPLE function: gimple gimple_build_return (tree retval) 13870 Build a `GIMPLE_RETURN' statement whose return value is retval. 13871 13872 -- GIMPLE function: tree gimple_return_retval (gimple g) 13873 Return the return value for `GIMPLE_RETURN' `G'. 13874 13875 -- GIMPLE function: void gimple_return_set_retval (gimple g, tree 13876 retval) 13877 Set `RETVAL' to be the return value for `GIMPLE_RETURN' `G'. 13878 13879 13880 File: gccint.info, Node: `GIMPLE_SWITCH', Next: `GIMPLE_TRY', Prev: `GIMPLE_RETURN', Up: Tuple specific accessors 13881 13882 12.7.26 `GIMPLE_SWITCH' 13883 ----------------------- 13884 13885 -- GIMPLE function: gimple gimple_build_switch ( nlabels, tree index, 13886 tree default_label, ...) 13887 Build a `GIMPLE_SWITCH' statement. `NLABELS' are the number of 13888 labels excluding the default label. The default label is passed 13889 in `DEFAULT_LABEL'. The rest of the arguments are trees 13890 representing the labels. Each label is a tree of code 13891 `CASE_LABEL_EXPR'. 13892 13893 -- GIMPLE function: gimple gimple_build_switch_vec (tree index, tree 13894 default_label, `VEC'(tree,heap) *args) 13895 This function is an alternate way of building `GIMPLE_SWITCH' 13896 statements. `INDEX' and `DEFAULT_LABEL' are as in 13897 gimple_build_switch. `ARGS' is a vector of `CASE_LABEL_EXPR' trees 13898 that contain the labels. 13899 13900 -- GIMPLE function: unsigned gimple_switch_num_labels (gimple g) 13901 Return the number of labels associated with the switch statement 13902 `G'. 13903 13904 -- GIMPLE function: void gimple_switch_set_num_labels (gimple g, 13905 unsigned nlabels) 13906 Set `NLABELS' to be the number of labels for the switch statement 13907 `G'. 13908 13909 -- GIMPLE function: tree gimple_switch_index (gimple g) 13910 Return the index variable used by the switch statement `G'. 13911 13912 -- GIMPLE function: void gimple_switch_set_index (gimple g, tree index) 13913 Set `INDEX' to be the index variable for switch statement `G'. 13914 13915 -- GIMPLE function: tree gimple_switch_label (gimple g, unsigned index) 13916 Return the label numbered `INDEX'. The default label is 0, followed 13917 by any labels in a switch statement. 13918 13919 -- GIMPLE function: void gimple_switch_set_label (gimple g, unsigned 13920 index, tree label) 13921 Set the label number `INDEX' to `LABEL'. 0 is always the default 13922 label. 13923 13924 -- GIMPLE function: tree gimple_switch_default_label (gimple g) 13925 Return the default label for a switch statement. 13926 13927 -- GIMPLE function: void gimple_switch_set_default_label (gimple g, 13928 tree label) 13929 Set the default label for a switch statement. 13930 13931 13932 File: gccint.info, Node: `GIMPLE_TRY', Next: `GIMPLE_WITH_CLEANUP_EXPR', Prev: `GIMPLE_SWITCH', Up: Tuple specific accessors 13933 13934 12.7.27 `GIMPLE_TRY' 13935 -------------------- 13936 13937 -- GIMPLE function: gimple gimple_build_try (gimple_seq eval, 13938 gimple_seq cleanup, unsigned int kind) 13939 Build a `GIMPLE_TRY' statement. `EVAL' is a sequence with the 13940 expression to evaluate. `CLEANUP' is a sequence of statements to 13941 run at clean-up time. `KIND' is the enumeration value 13942 `GIMPLE_TRY_CATCH' if this statement denotes a try/catch construct 13943 or `GIMPLE_TRY_FINALLY' if this statement denotes a try/finally 13944 construct. 13945 13946 -- GIMPLE function: enum gimple_try_flags gimple_try_kind (gimple g) 13947 Return the kind of try block represented by `GIMPLE_TRY' `G'. This 13948 is either `GIMPLE_TRY_CATCH' or `GIMPLE_TRY_FINALLY'. 13949 13950 -- GIMPLE function: bool gimple_try_catch_is_cleanup (gimple g) 13951 Return the `GIMPLE_TRY_CATCH_IS_CLEANUP' flag. 13952 13953 -- GIMPLE function: gimple_seq gimple_try_eval (gimple g) 13954 Return the sequence of statements used as the body for `GIMPLE_TRY' 13955 `G'. 13956 13957 -- GIMPLE function: gimple_seq gimple_try_cleanup (gimple g) 13958 Return the sequence of statements used as the cleanup body for 13959 `GIMPLE_TRY' `G'. 13960 13961 -- GIMPLE function: void gimple_try_set_catch_is_cleanup (gimple g, 13962 bool catch_is_cleanup) 13963 Set the `GIMPLE_TRY_CATCH_IS_CLEANUP' flag. 13964 13965 -- GIMPLE function: void gimple_try_set_eval (gimple g, gimple_seq 13966 eval) 13967 Set `EVAL' to be the sequence of statements to use as the body for 13968 `GIMPLE_TRY' `G'. 13969 13970 -- GIMPLE function: void gimple_try_set_cleanup (gimple g, gimple_seq 13971 cleanup) 13972 Set `CLEANUP' to be the sequence of statements to use as the 13973 cleanup body for `GIMPLE_TRY' `G'. 13974 13975 13976 File: gccint.info, Node: `GIMPLE_WITH_CLEANUP_EXPR', Prev: `GIMPLE_TRY', Up: Tuple specific accessors 13977 13978 12.7.28 `GIMPLE_WITH_CLEANUP_EXPR' 13979 ---------------------------------- 13980 13981 -- GIMPLE function: gimple gimple_build_wce (gimple_seq cleanup) 13982 Build a `GIMPLE_WITH_CLEANUP_EXPR' statement. `CLEANUP' is the 13983 clean-up expression. 13984 13985 -- GIMPLE function: gimple_seq gimple_wce_cleanup (gimple g) 13986 Return the cleanup sequence for cleanup statement `G'. 13987 13988 -- GIMPLE function: void gimple_wce_set_cleanup (gimple g, gimple_seq 13989 cleanup) 13990 Set `CLEANUP' to be the cleanup sequence for `G'. 13991 13992 -- GIMPLE function: bool gimple_wce_cleanup_eh_only (gimple g) 13993 Return the `CLEANUP_EH_ONLY' flag for a `WCE' tuple. 13994 13995 -- GIMPLE function: void gimple_wce_set_cleanup_eh_only (gimple g, 13996 bool eh_only_p) 13997 Set the `CLEANUP_EH_ONLY' flag for a `WCE' tuple. 13998 13999 14000 File: gccint.info, Node: GIMPLE sequences, Next: Sequence iterators, Prev: Tuple specific accessors, Up: GIMPLE 14001 14002 12.8 GIMPLE sequences 14003 ===================== 14004 14005 GIMPLE sequences are the tuple equivalent of `STATEMENT_LIST''s used in 14006 `GENERIC'. They are used to chain statements together, and when used 14007 in conjunction with sequence iterators, provide a framework for 14008 iterating through statements. 14009 14010 GIMPLE sequences are of type struct `gimple_sequence', but are more 14011 commonly passed by reference to functions dealing with sequences. The 14012 type for a sequence pointer is `gimple_seq' which is the same as struct 14013 `gimple_sequence' *. When declaring a local sequence, you can define a 14014 local variable of type struct `gimple_sequence'. When declaring a 14015 sequence allocated on the garbage collected heap, use the function 14016 `gimple_seq_alloc' documented below. 14017 14018 There are convenience functions for iterating through sequences in the 14019 section entitled Sequence Iterators. 14020 14021 Below is a list of functions to manipulate and query sequences. 14022 14023 -- GIMPLE function: void gimple_seq_add_stmt (gimple_seq *seq, gimple 14024 g) 14025 Link a gimple statement to the end of the sequence *`SEQ' if `G' is 14026 not `NULL'. If *`SEQ' is `NULL', allocate a sequence before 14027 linking. 14028 14029 -- GIMPLE function: void gimple_seq_add_seq (gimple_seq *dest, 14030 gimple_seq src) 14031 Append sequence `SRC' to the end of sequence *`DEST' if `SRC' is 14032 not `NULL'. If *`DEST' is `NULL', allocate a new sequence before 14033 appending. 14034 14035 -- GIMPLE function: gimple_seq gimple_seq_deep_copy (gimple_seq src) 14036 Perform a deep copy of sequence `SRC' and return the result. 14037 14038 -- GIMPLE function: gimple_seq gimple_seq_reverse (gimple_seq seq) 14039 Reverse the order of the statements in the sequence `SEQ'. Return 14040 `SEQ'. 14041 14042 -- GIMPLE function: gimple gimple_seq_first (gimple_seq s) 14043 Return the first statement in sequence `S'. 14044 14045 -- GIMPLE function: gimple gimple_seq_last (gimple_seq s) 14046 Return the last statement in sequence `S'. 14047 14048 -- GIMPLE function: void gimple_seq_set_last (gimple_seq s, gimple 14049 last) 14050 Set the last statement in sequence `S' to the statement in `LAST'. 14051 14052 -- GIMPLE function: void gimple_seq_set_first (gimple_seq s, gimple 14053 first) 14054 Set the first statement in sequence `S' to the statement in 14055 `FIRST'. 14056 14057 -- GIMPLE function: void gimple_seq_init (gimple_seq s) 14058 Initialize sequence `S' to an empty sequence. 14059 14060 -- GIMPLE function: gimple_seq gimple_seq_alloc (void) 14061 Allocate a new sequence in the garbage collected store and return 14062 it. 14063 14064 -- GIMPLE function: void gimple_seq_copy (gimple_seq dest, gimple_seq 14065 src) 14066 Copy the sequence `SRC' into the sequence `DEST'. 14067 14068 -- GIMPLE function: bool gimple_seq_empty_p (gimple_seq s) 14069 Return true if the sequence `S' is empty. 14070 14071 -- GIMPLE function: gimple_seq bb_seq (basic_block bb) 14072 Returns the sequence of statements in `BB'. 14073 14074 -- GIMPLE function: void set_bb_seq (basic_block bb, gimple_seq seq) 14075 Sets the sequence of statements in `BB' to `SEQ'. 14076 14077 -- GIMPLE function: bool gimple_seq_singleton_p (gimple_seq seq) 14078 Determine whether `SEQ' contains exactly one statement. 14079 14080 14081 File: gccint.info, Node: Sequence iterators, Next: Adding a new GIMPLE statement code, Prev: GIMPLE sequences, Up: GIMPLE 14082 14083 12.9 Sequence iterators 14084 ======================= 14085 14086 Sequence iterators are convenience constructs for iterating through 14087 statements in a sequence. Given a sequence `SEQ', here is a typical 14088 use of gimple sequence iterators: 14089 14090 gimple_stmt_iterator gsi; 14091 14092 for (gsi = gsi_start (seq); !gsi_end_p (gsi); gsi_next (&gsi)) 14093 { 14094 gimple g = gsi_stmt (gsi); 14095 /* Do something with gimple statement `G'. */ 14096 } 14097 14098 Backward iterations are possible: 14099 14100 for (gsi = gsi_last (seq); !gsi_end_p (gsi); gsi_prev (&gsi)) 14101 14102 Forward and backward iterations on basic blocks are possible with 14103 `gsi_start_bb' and `gsi_last_bb'. 14104 14105 In the documentation below we sometimes refer to enum 14106 `gsi_iterator_update'. The valid options for this enumeration are: 14107 14108 * `GSI_NEW_STMT' Only valid when a single statement is added. Move 14109 the iterator to it. 14110 14111 * `GSI_SAME_STMT' Leave the iterator at the same statement. 14112 14113 * `GSI_CONTINUE_LINKING' Move iterator to whatever position is 14114 suitable for linking other statements in the same direction. 14115 14116 Below is a list of the functions used to manipulate and use statement 14117 iterators. 14118 14119 -- GIMPLE function: gimple_stmt_iterator gsi_start (gimple_seq seq) 14120 Return a new iterator pointing to the sequence `SEQ''s first 14121 statement. If `SEQ' is empty, the iterator's basic block is 14122 `NULL'. Use `gsi_start_bb' instead when the iterator needs to 14123 always have the correct basic block set. 14124 14125 -- GIMPLE function: gimple_stmt_iterator gsi_start_bb (basic_block bb) 14126 Return a new iterator pointing to the first statement in basic 14127 block `BB'. 14128 14129 -- GIMPLE function: gimple_stmt_iterator gsi_last (gimple_seq seq) 14130 Return a new iterator initially pointing to the last statement of 14131 sequence `SEQ'. If `SEQ' is empty, the iterator's basic block is 14132 `NULL'. Use `gsi_last_bb' instead when the iterator needs to 14133 always have the correct basic block set. 14134 14135 -- GIMPLE function: gimple_stmt_iterator gsi_last_bb (basic_block bb) 14136 Return a new iterator pointing to the last statement in basic 14137 block `BB'. 14138 14139 -- GIMPLE function: bool gsi_end_p (gimple_stmt_iterator i) 14140 Return `TRUE' if at the end of `I'. 14141 14142 -- GIMPLE function: bool gsi_one_before_end_p (gimple_stmt_iterator i) 14143 Return `TRUE' if we're one statement before the end of `I'. 14144 14145 -- GIMPLE function: void gsi_next (gimple_stmt_iterator *i) 14146 Advance the iterator to the next gimple statement. 14147 14148 -- GIMPLE function: void gsi_prev (gimple_stmt_iterator *i) 14149 Advance the iterator to the previous gimple statement. 14150 14151 -- GIMPLE function: gimple gsi_stmt (gimple_stmt_iterator i) 14152 Return the current stmt. 14153 14154 -- GIMPLE function: gimple_stmt_iterator gsi_after_labels (basic_block 14155 bb) 14156 Return a block statement iterator that points to the first 14157 non-label statement in block `BB'. 14158 14159 -- GIMPLE function: gimple *gsi_stmt_ptr (gimple_stmt_iterator *i) 14160 Return a pointer to the current stmt. 14161 14162 -- GIMPLE function: basic_block gsi_bb (gimple_stmt_iterator i) 14163 Return the basic block associated with this iterator. 14164 14165 -- GIMPLE function: gimple_seq gsi_seq (gimple_stmt_iterator i) 14166 Return the sequence associated with this iterator. 14167 14168 -- GIMPLE function: void gsi_remove (gimple_stmt_iterator *i, bool 14169 remove_eh_info) 14170 Remove the current stmt from the sequence. The iterator is 14171 updated to point to the next statement. When `REMOVE_EH_INFO' is 14172 true we remove the statement pointed to by iterator `I' from the 14173 `EH' tables. Otherwise we do not modify the `EH' tables. 14174 Generally, `REMOVE_EH_INFO' should be true when the statement is 14175 going to be removed from the `IL' and not reinserted elsewhere. 14176 14177 -- GIMPLE function: void gsi_link_seq_before (gimple_stmt_iterator *i, 14178 gimple_seq seq, enum gsi_iterator_update mode) 14179 Links the sequence of statements `SEQ' before the statement pointed 14180 by iterator `I'. `MODE' indicates what to do with the iterator 14181 after insertion (see `enum gsi_iterator_update' above). 14182 14183 -- GIMPLE function: void gsi_link_before (gimple_stmt_iterator *i, 14184 gimple g, enum gsi_iterator_update mode) 14185 Links statement `G' before the statement pointed-to by iterator 14186 `I'. Updates iterator `I' according to `MODE'. 14187 14188 -- GIMPLE function: void gsi_link_seq_after (gimple_stmt_iterator *i, 14189 gimple_seq seq, enum gsi_iterator_update mode) 14190 Links sequence `SEQ' after the statement pointed-to by iterator 14191 `I'. `MODE' is as in `gsi_insert_after'. 14192 14193 -- GIMPLE function: void gsi_link_after (gimple_stmt_iterator *i, 14194 gimple g, enum gsi_iterator_update mode) 14195 Links statement `G' after the statement pointed-to by iterator `I'. 14196 `MODE' is as in `gsi_insert_after'. 14197 14198 -- GIMPLE function: gimple_seq gsi_split_seq_after 14199 (gimple_stmt_iterator i) 14200 Move all statements in the sequence after `I' to a new sequence. 14201 Return this new sequence. 14202 14203 -- GIMPLE function: gimple_seq gsi_split_seq_before 14204 (gimple_stmt_iterator *i) 14205 Move all statements in the sequence before `I' to a new sequence. 14206 Return this new sequence. 14207 14208 -- GIMPLE function: void gsi_replace (gimple_stmt_iterator *i, gimple 14209 stmt, bool update_eh_info) 14210 Replace the statement pointed-to by `I' to `STMT'. If 14211 `UPDATE_EH_INFO' is true, the exception handling information of 14212 the original statement is moved to the new statement. 14213 14214 -- GIMPLE function: void gsi_insert_before (gimple_stmt_iterator *i, 14215 gimple stmt, enum gsi_iterator_update mode) 14216 Insert statement `STMT' before the statement pointed-to by iterator 14217 `I', update `STMT''s basic block and scan it for new operands. 14218 `MODE' specifies how to update iterator `I' after insertion (see 14219 enum `gsi_iterator_update'). 14220 14221 -- GIMPLE function: void gsi_insert_seq_before (gimple_stmt_iterator 14222 *i, gimple_seq seq, enum gsi_iterator_update mode) 14223 Like `gsi_insert_before', but for all the statements in `SEQ'. 14224 14225 -- GIMPLE function: void gsi_insert_after (gimple_stmt_iterator *i, 14226 gimple stmt, enum gsi_iterator_update mode) 14227 Insert statement `STMT' after the statement pointed-to by iterator 14228 `I', update `STMT''s basic block and scan it for new operands. 14229 `MODE' specifies how to update iterator `I' after insertion (see 14230 enum `gsi_iterator_update'). 14231 14232 -- GIMPLE function: void gsi_insert_seq_after (gimple_stmt_iterator 14233 *i, gimple_seq seq, enum gsi_iterator_update mode) 14234 Like `gsi_insert_after', but for all the statements in `SEQ'. 14235 14236 -- GIMPLE function: gimple_stmt_iterator gsi_for_stmt (gimple stmt) 14237 Finds iterator for `STMT'. 14238 14239 -- GIMPLE function: void gsi_move_after (gimple_stmt_iterator *from, 14240 gimple_stmt_iterator *to) 14241 Move the statement at `FROM' so it comes right after the statement 14242 at `TO'. 14243 14244 -- GIMPLE function: void gsi_move_before (gimple_stmt_iterator *from, 14245 gimple_stmt_iterator *to) 14246 Move the statement at `FROM' so it comes right before the statement 14247 at `TO'. 14248 14249 -- GIMPLE function: void gsi_move_to_bb_end (gimple_stmt_iterator 14250 *from, basic_block bb) 14251 Move the statement at `FROM' to the end of basic block `BB'. 14252 14253 -- GIMPLE function: void gsi_insert_on_edge (edge e, gimple stmt) 14254 Add `STMT' to the pending list of edge `E'. No actual insertion is 14255 made until a call to `gsi_commit_edge_inserts'() is made. 14256 14257 -- GIMPLE function: void gsi_insert_seq_on_edge (edge e, gimple_seq 14258 seq) 14259 Add the sequence of statements in `SEQ' to the pending list of edge 14260 `E'. No actual insertion is made until a call to 14261 `gsi_commit_edge_inserts'() is made. 14262 14263 -- GIMPLE function: basic_block gsi_insert_on_edge_immediate (edge e, 14264 gimple stmt) 14265 Similar to `gsi_insert_on_edge'+`gsi_commit_edge_inserts'. If a 14266 new block has to be created, it is returned. 14267 14268 -- GIMPLE function: void gsi_commit_one_edge_insert (edge e, 14269 basic_block *new_bb) 14270 Commit insertions pending at edge `E'. If a new block is created, 14271 set `NEW_BB' to this block, otherwise set it to `NULL'. 14272 14273 -- GIMPLE function: void gsi_commit_edge_inserts (void) 14274 This routine will commit all pending edge insertions, creating any 14275 new basic blocks which are necessary. 14276 14277 14278 File: gccint.info, Node: Adding a new GIMPLE statement code, Next: Statement and operand traversals, Prev: Sequence iterators, Up: GIMPLE 14279 14280 12.10 Adding a new GIMPLE statement code 14281 ======================================== 14282 14283 The first step in adding a new GIMPLE statement code, is modifying the 14284 file `gimple.def', which contains all the GIMPLE codes. Then you must 14285 add a corresponding structure, and an entry in `union 14286 gimple_statement_d', both of which are located in `gimple.h'. This in 14287 turn, will require you to add a corresponding `GTY' tag in 14288 `gsstruct.def', and code to handle this tag in `gss_for_code' which is 14289 located in `gimple.c'. 14290 14291 In order for the garbage collector to know the size of the structure 14292 you created in `gimple.h', you need to add a case to handle your new 14293 GIMPLE statement in `gimple_size' which is located in `gimple.c'. 14294 14295 You will probably want to create a function to build the new gimple 14296 statement in `gimple.c'. The function should be called 14297 `gimple_build_<`NEW_TUPLE_NAME'>', and should return the new tuple of 14298 type gimple. 14299 14300 If your new statement requires accessors for any members or operands 14301 it may have, put simple inline accessors in `gimple.h' and any 14302 non-trivial accessors in `gimple.c' with a corresponding prototype in 14303 `gimple.h'. 14304 14305 14306 File: gccint.info, Node: Statement and operand traversals, Prev: Adding a new GIMPLE statement code, Up: GIMPLE 14307 14308 12.11 Statement and operand traversals 14309 ====================================== 14310 14311 There are two functions available for walking statements and sequences: 14312 `walk_gimple_stmt' and `walk_gimple_seq', accordingly, and a third 14313 function for walking the operands in a statement: `walk_gimple_op'. 14314 14315 -- GIMPLE function: tree walk_gimple_stmt (gimple_stmt_iterator *gsi, 14316 walk_stmt_fn callback_stmt, walk_tree_fn callback_op, struct 14317 walk_stmt_info *wi) 14318 This function is used to walk the current statement in `GSI', 14319 optionally using traversal state stored in `WI'. If `WI' is 14320 `NULL', no state is kept during the traversal. 14321 14322 The callback `CALLBACK_STMT' is called. If `CALLBACK_STMT' returns 14323 true, it means that the callback function has handled all the 14324 operands of the statement and it is not necessary to walk its 14325 operands. 14326 14327 If `CALLBACK_STMT' is `NULL' or it returns false, `CALLBACK_OP' is 14328 called on each operand of the statement via `walk_gimple_op'. If 14329 `walk_gimple_op' returns non-`NULL' for any operand, the remaining 14330 operands are not scanned. 14331 14332 The return value is that returned by the last call to 14333 `walk_gimple_op', or `NULL_TREE' if no `CALLBACK_OP' is specified. 14334 14335 -- GIMPLE function: tree walk_gimple_op (gimple stmt, walk_tree_fn 14336 callback_op, struct walk_stmt_info *wi) 14337 Use this function to walk the operands of statement `STMT'. Every 14338 operand is walked via `walk_tree' with optional state information 14339 in `WI'. 14340 14341 `CALLBACK_OP' is called on each operand of `STMT' via `walk_tree'. 14342 Additional parameters to `walk_tree' must be stored in `WI'. For 14343 each operand `OP', `walk_tree' is called as: 14344 14345 walk_tree (&`OP', `CALLBACK_OP', `WI', `WI'- `PSET') 14346 14347 If `CALLBACK_OP' returns non-`NULL' for an operand, the remaining 14348 operands are not scanned. The return value is that returned by 14349 the last call to `walk_tree', or `NULL_TREE' if no `CALLBACK_OP' is 14350 specified. 14351 14352 -- GIMPLE function: tree walk_gimple_seq (gimple_seq seq, walk_stmt_fn 14353 callback_stmt, walk_tree_fn callback_op, struct 14354 walk_stmt_info *wi) 14355 This function walks all the statements in the sequence `SEQ' 14356 calling `walk_gimple_stmt' on each one. `WI' is as in 14357 `walk_gimple_stmt'. If `walk_gimple_stmt' returns non-`NULL', the 14358 walk is stopped and the value returned. Otherwise, all the 14359 statements are walked and `NULL_TREE' returned. 14360 14361 14362 File: gccint.info, Node: Tree SSA, Next: RTL, Prev: GIMPLE, Up: Top 14363 14364 13 Analysis and Optimization of GIMPLE tuples 14365 ********************************************* 14366 14367 GCC uses three main intermediate languages to represent the program 14368 during compilation: GENERIC, GIMPLE and RTL. GENERIC is a 14369 language-independent representation generated by each front end. It is 14370 used to serve as an interface between the parser and optimizer. 14371 GENERIC is a common representation that is able to represent programs 14372 written in all the languages supported by GCC. 14373 14374 GIMPLE and RTL are used to optimize the program. GIMPLE is used for 14375 target and language independent optimizations (e.g., inlining, constant 14376 propagation, tail call elimination, redundancy elimination, etc). Much 14377 like GENERIC, GIMPLE is a language independent, tree based 14378 representation. However, it differs from GENERIC in that the GIMPLE 14379 grammar is more restrictive: expressions contain no more than 3 14380 operands (except function calls), it has no control flow structures and 14381 expressions with side-effects are only allowed on the right hand side 14382 of assignments. See the chapter describing GENERIC and GIMPLE for more 14383 details. 14384 14385 This chapter describes the data structures and functions used in the 14386 GIMPLE optimizers (also known as "tree optimizers" or "middle end"). 14387 In particular, it focuses on all the macros, data structures, functions 14388 and programming constructs needed to implement optimization passes for 14389 GIMPLE. 14390 14391 * Menu: 14392 14393 * Annotations:: Attributes for variables. 14394 * SSA Operands:: SSA names referenced by GIMPLE statements. 14395 * SSA:: Static Single Assignment representation. 14396 * Alias analysis:: Representing aliased loads and stores. 14397 14398 14399 File: gccint.info, Node: Annotations, Next: SSA Operands, Up: Tree SSA 14400 14401 13.1 Annotations 14402 ================ 14403 14404 The optimizers need to associate attributes with variables during the 14405 optimization process. For instance, we need to know whether a variable 14406 has aliases. All these attributes are stored in data structures called 14407 annotations which are then linked to the field `ann' in `struct 14408 tree_common'. 14409 14410 Presently, we define annotations for variables (`var_ann_t'). 14411 Annotations are defined and documented in `tree-flow.h'. 14412 14413 14414 File: gccint.info, Node: SSA Operands, Next: SSA, Prev: Annotations, Up: Tree SSA 14415 14416 13.2 SSA Operands 14417 ================= 14418 14419 Almost every GIMPLE statement will contain a reference to a variable or 14420 memory location. Since statements come in different shapes and sizes, 14421 their operands are going to be located at various spots inside the 14422 statement's tree. To facilitate access to the statement's operands, 14423 they are organized into lists associated inside each statement's 14424 annotation. Each element in an operand list is a pointer to a 14425 `VAR_DECL', `PARM_DECL' or `SSA_NAME' tree node. This provides a very 14426 convenient way of examining and replacing operands. 14427 14428 Data flow analysis and optimization is done on all tree nodes 14429 representing variables. Any node for which `SSA_VAR_P' returns nonzero 14430 is considered when scanning statement operands. However, not all 14431 `SSA_VAR_P' variables are processed in the same way. For the purposes 14432 of optimization, we need to distinguish between references to local 14433 scalar variables and references to globals, statics, structures, 14434 arrays, aliased variables, etc. The reason is simple, the compiler can 14435 gather complete data flow information for a local scalar. On the other 14436 hand, a global variable may be modified by a function call, it may not 14437 be possible to keep track of all the elements of an array or the fields 14438 of a structure, etc. 14439 14440 The operand scanner gathers two kinds of operands: "real" and 14441 "virtual". An operand for which `is_gimple_reg' returns true is 14442 considered real, otherwise it is a virtual operand. We also 14443 distinguish between uses and definitions. An operand is used if its 14444 value is loaded by the statement (e.g., the operand at the RHS of an 14445 assignment). If the statement assigns a new value to the operand, the 14446 operand is considered a definition (e.g., the operand at the LHS of an 14447 assignment). 14448 14449 Virtual and real operands also have very different data flow 14450 properties. Real operands are unambiguous references to the full 14451 object that they represent. For instance, given 14452 14453 { 14454 int a, b; 14455 a = b 14456 } 14457 14458 Since `a' and `b' are non-aliased locals, the statement `a = b' will 14459 have one real definition and one real use because variable `b' is 14460 completely modified with the contents of variable `a'. Real definition 14461 are also known as "killing definitions". Similarly, the use of `a' 14462 reads all its bits. 14463 14464 In contrast, virtual operands are used with variables that can have a 14465 partial or ambiguous reference. This includes structures, arrays, 14466 globals, and aliased variables. In these cases, we have two types of 14467 definitions. For globals, structures, and arrays, we can determine from 14468 a statement whether a variable of these types has a killing definition. 14469 If the variable does, then the statement is marked as having a "must 14470 definition" of that variable. However, if a statement is only defining 14471 a part of the variable (i.e. a field in a structure), or if we know 14472 that a statement might define the variable but we cannot say for sure, 14473 then we mark that statement as having a "may definition". For 14474 instance, given 14475 14476 { 14477 int a, b, *p; 14478 14479 if (...) 14480 p = &a; 14481 else 14482 p = &b; 14483 *p = 5; 14484 return *p; 14485 } 14486 14487 The assignment `*p = 5' may be a definition of `a' or `b'. If we 14488 cannot determine statically where `p' is pointing to at the time of the 14489 store operation, we create virtual definitions to mark that statement 14490 as a potential definition site for `a' and `b'. Memory loads are 14491 similarly marked with virtual use operands. Virtual operands are shown 14492 in tree dumps right before the statement that contains them. To 14493 request a tree dump with virtual operands, use the `-vops' option to 14494 `-fdump-tree': 14495 14496 { 14497 int a, b, *p; 14498 14499 if (...) 14500 p = &a; 14501 else 14502 p = &b; 14503 # a = VDEF <a> 14504 # b = VDEF <b> 14505 *p = 5; 14506 14507 # VUSE <a> 14508 # VUSE <b> 14509 return *p; 14510 } 14511 14512 Notice that `VDEF' operands have two copies of the referenced 14513 variable. This indicates that this is not a killing definition of that 14514 variable. In this case we refer to it as a "may definition" or 14515 "aliased store". The presence of the second copy of the variable in 14516 the `VDEF' operand will become important when the function is converted 14517 into SSA form. This will be used to link all the non-killing 14518 definitions to prevent optimizations from making incorrect assumptions 14519 about them. 14520 14521 Operands are updated as soon as the statement is finished via a call 14522 to `update_stmt'. If statement elements are changed via `SET_USE' or 14523 `SET_DEF', then no further action is required (i.e., those macros take 14524 care of updating the statement). If changes are made by manipulating 14525 the statement's tree directly, then a call must be made to 14526 `update_stmt' when complete. Calling one of the `bsi_insert' routines 14527 or `bsi_replace' performs an implicit call to `update_stmt'. 14528 14529 13.2.1 Operand Iterators And Access Routines 14530 -------------------------------------------- 14531 14532 Operands are collected by `tree-ssa-operands.c'. They are stored 14533 inside each statement's annotation and can be accessed through either 14534 the operand iterators or an access routine. 14535 14536 The following access routines are available for examining operands: 14537 14538 1. `SINGLE_SSA_{USE,DEF,TREE}_OPERAND': These accessors will return 14539 NULL unless there is exactly one operand matching the specified 14540 flags. If there is exactly one operand, the operand is returned 14541 as either a `tree', `def_operand_p', or `use_operand_p'. 14542 14543 tree t = SINGLE_SSA_TREE_OPERAND (stmt, flags); 14544 use_operand_p u = SINGLE_SSA_USE_OPERAND (stmt, SSA_ALL_VIRTUAL_USES); 14545 def_operand_p d = SINGLE_SSA_DEF_OPERAND (stmt, SSA_OP_ALL_DEFS); 14546 14547 2. `ZERO_SSA_OPERANDS': This macro returns true if there are no 14548 operands matching the specified flags. 14549 14550 if (ZERO_SSA_OPERANDS (stmt, SSA_OP_ALL_VIRTUALS)) 14551 return; 14552 14553 3. `NUM_SSA_OPERANDS': This macro Returns the number of operands 14554 matching 'flags'. This actually executes a loop to perform the 14555 count, so only use this if it is really needed. 14556 14557 int count = NUM_SSA_OPERANDS (stmt, flags) 14558 14559 If you wish to iterate over some or all operands, use the 14560 `FOR_EACH_SSA_{USE,DEF,TREE}_OPERAND' iterator. For example, to print 14561 all the operands for a statement: 14562 14563 void 14564 print_ops (tree stmt) 14565 { 14566 ssa_op_iter; 14567 tree var; 14568 14569 FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_ALL_OPERANDS) 14570 print_generic_expr (stderr, var, TDF_SLIM); 14571 } 14572 14573 How to choose the appropriate iterator: 14574 14575 1. Determine whether you are need to see the operand pointers, or 14576 just the trees, and choose the appropriate macro: 14577 14578 Need Macro: 14579 ---- ------- 14580 use_operand_p FOR_EACH_SSA_USE_OPERAND 14581 def_operand_p FOR_EACH_SSA_DEF_OPERAND 14582 tree FOR_EACH_SSA_TREE_OPERAND 14583 14584 2. You need to declare a variable of the type you are interested in, 14585 and an ssa_op_iter structure which serves as the loop controlling 14586 variable. 14587 14588 3. Determine which operands you wish to use, and specify the flags of 14589 those you are interested in. They are documented in 14590 `tree-ssa-operands.h': 14591 14592 #define SSA_OP_USE 0x01 /* Real USE operands. */ 14593 #define SSA_OP_DEF 0x02 /* Real DEF operands. */ 14594 #define SSA_OP_VUSE 0x04 /* VUSE operands. */ 14595 #define SSA_OP_VMAYUSE 0x08 /* USE portion of VDEFS. */ 14596 #define SSA_OP_VDEF 0x10 /* DEF portion of VDEFS. */ 14597 14598 /* These are commonly grouped operand flags. */ 14599 #define SSA_OP_VIRTUAL_USES (SSA_OP_VUSE | SSA_OP_VMAYUSE) 14600 #define SSA_OP_VIRTUAL_DEFS (SSA_OP_VDEF) 14601 #define SSA_OP_ALL_USES (SSA_OP_VIRTUAL_USES | SSA_OP_USE) 14602 #define SSA_OP_ALL_DEFS (SSA_OP_VIRTUAL_DEFS | SSA_OP_DEF) 14603 #define SSA_OP_ALL_OPERANDS (SSA_OP_ALL_USES | SSA_OP_ALL_DEFS) 14604 14605 So if you want to look at the use pointers for all the `USE' and 14606 `VUSE' operands, you would do something like: 14607 14608 use_operand_p use_p; 14609 ssa_op_iter iter; 14610 14611 FOR_EACH_SSA_USE_OPERAND (use_p, stmt, iter, (SSA_OP_USE | SSA_OP_VUSE)) 14612 { 14613 process_use_ptr (use_p); 14614 } 14615 14616 The `TREE' macro is basically the same as the `USE' and `DEF' macros, 14617 only with the use or def dereferenced via `USE_FROM_PTR (use_p)' and 14618 `DEF_FROM_PTR (def_p)'. Since we aren't using operand pointers, use 14619 and defs flags can be mixed. 14620 14621 tree var; 14622 ssa_op_iter iter; 14623 14624 FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_VUSE) 14625 { 14626 print_generic_expr (stderr, var, TDF_SLIM); 14627 } 14628 14629 `VDEF's are broken into two flags, one for the `DEF' portion 14630 (`SSA_OP_VDEF') and one for the USE portion (`SSA_OP_VMAYUSE'). If all 14631 you want to look at are the `VDEF's together, there is a fourth 14632 iterator macro for this, which returns both a def_operand_p and a 14633 use_operand_p for each `VDEF' in the statement. Note that you don't 14634 need any flags for this one. 14635 14636 use_operand_p use_p; 14637 def_operand_p def_p; 14638 ssa_op_iter iter; 14639 14640 FOR_EACH_SSA_MAYDEF_OPERAND (def_p, use_p, stmt, iter) 14641 { 14642 my_code; 14643 } 14644 14645 There are many examples in the code as well, as well as the 14646 documentation in `tree-ssa-operands.h'. 14647 14648 There are also a couple of variants on the stmt iterators regarding PHI 14649 nodes. 14650 14651 `FOR_EACH_PHI_ARG' Works exactly like `FOR_EACH_SSA_USE_OPERAND', 14652 except it works over `PHI' arguments instead of statement operands. 14653 14654 /* Look at every virtual PHI use. */ 14655 FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_VIRTUAL_USES) 14656 { 14657 my_code; 14658 } 14659 14660 /* Look at every real PHI use. */ 14661 FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_USES) 14662 my_code; 14663 14664 /* Look at every PHI use. */ 14665 FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_ALL_USES) 14666 my_code; 14667 14668 `FOR_EACH_PHI_OR_STMT_{USE,DEF}' works exactly like 14669 `FOR_EACH_SSA_{USE,DEF}_OPERAND', except it will function on either a 14670 statement or a `PHI' node. These should be used when it is appropriate 14671 but they are not quite as efficient as the individual `FOR_EACH_PHI' 14672 and `FOR_EACH_SSA' routines. 14673 14674 FOR_EACH_PHI_OR_STMT_USE (use_operand_p, stmt, iter, flags) 14675 { 14676 my_code; 14677 } 14678 14679 FOR_EACH_PHI_OR_STMT_DEF (def_operand_p, phi, iter, flags) 14680 { 14681 my_code; 14682 } 14683 14684 13.2.2 Immediate Uses 14685 --------------------- 14686 14687 Immediate use information is now always available. Using the immediate 14688 use iterators, you may examine every use of any `SSA_NAME'. For 14689 instance, to change each use of `ssa_var' to `ssa_var2' and call 14690 fold_stmt on each stmt after that is done: 14691 14692 use_operand_p imm_use_p; 14693 imm_use_iterator iterator; 14694 tree ssa_var, stmt; 14695 14696 14697 FOR_EACH_IMM_USE_STMT (stmt, iterator, ssa_var) 14698 { 14699 FOR_EACH_IMM_USE_ON_STMT (imm_use_p, iterator) 14700 SET_USE (imm_use_p, ssa_var_2); 14701 fold_stmt (stmt); 14702 } 14703 14704 There are 2 iterators which can be used. `FOR_EACH_IMM_USE_FAST' is 14705 used when the immediate uses are not changed, i.e., you are looking at 14706 the uses, but not setting them. 14707 14708 If they do get changed, then care must be taken that things are not 14709 changed under the iterators, so use the `FOR_EACH_IMM_USE_STMT' and 14710 `FOR_EACH_IMM_USE_ON_STMT' iterators. They attempt to preserve the 14711 sanity of the use list by moving all the uses for a statement into a 14712 controlled position, and then iterating over those uses. Then the 14713 optimization can manipulate the stmt when all the uses have been 14714 processed. This is a little slower than the FAST version since it adds 14715 a placeholder element and must sort through the list a bit for each 14716 statement. This placeholder element must be also be removed if the 14717 loop is terminated early. The macro `BREAK_FROM_IMM_USE_SAFE' is 14718 provided to do this : 14719 14720 FOR_EACH_IMM_USE_STMT (stmt, iterator, ssa_var) 14721 { 14722 if (stmt == last_stmt) 14723 BREAK_FROM_SAFE_IMM_USE (iter); 14724 14725 FOR_EACH_IMM_USE_ON_STMT (imm_use_p, iterator) 14726 SET_USE (imm_use_p, ssa_var_2); 14727 fold_stmt (stmt); 14728 } 14729 14730 There are checks in `verify_ssa' which verify that the immediate use 14731 list is up to date, as well as checking that an optimization didn't 14732 break from the loop without using this macro. It is safe to simply 14733 'break'; from a `FOR_EACH_IMM_USE_FAST' traverse. 14734 14735 Some useful functions and macros: 14736 1. `has_zero_uses (ssa_var)' : Returns true if there are no uses of 14737 `ssa_var'. 14738 14739 2. `has_single_use (ssa_var)' : Returns true if there is only a 14740 single use of `ssa_var'. 14741 14742 3. `single_imm_use (ssa_var, use_operand_p *ptr, tree *stmt)' : 14743 Returns true if there is only a single use of `ssa_var', and also 14744 returns the use pointer and statement it occurs in, in the second 14745 and third parameters. 14746 14747 4. `num_imm_uses (ssa_var)' : Returns the number of immediate uses of 14748 `ssa_var'. It is better not to use this if possible since it simply 14749 utilizes a loop to count the uses. 14750 14751 5. `PHI_ARG_INDEX_FROM_USE (use_p)' : Given a use within a `PHI' 14752 node, return the index number for the use. An assert is triggered 14753 if the use isn't located in a `PHI' node. 14754 14755 6. `USE_STMT (use_p)' : Return the statement a use occurs in. 14756 14757 Note that uses are not put into an immediate use list until their 14758 statement is actually inserted into the instruction stream via a 14759 `bsi_*' routine. 14760 14761 It is also still possible to utilize lazy updating of statements, but 14762 this should be used only when absolutely required. Both alias analysis 14763 and the dominator optimizations currently do this. 14764 14765 When lazy updating is being used, the immediate use information is out 14766 of date and cannot be used reliably. Lazy updating is achieved by 14767 simply marking statements modified via calls to `mark_stmt_modified' 14768 instead of `update_stmt'. When lazy updating is no longer required, 14769 all the modified statements must have `update_stmt' called in order to 14770 bring them up to date. This must be done before the optimization is 14771 finished, or `verify_ssa' will trigger an abort. 14772 14773 This is done with a simple loop over the instruction stream: 14774 block_stmt_iterator bsi; 14775 basic_block bb; 14776 FOR_EACH_BB (bb) 14777 { 14778 for (bsi = bsi_start (bb); !bsi_end_p (bsi); bsi_next (&bsi)) 14779 update_stmt_if_modified (bsi_stmt (bsi)); 14780 } 14781 14782 14783 File: gccint.info, Node: SSA, Next: Alias analysis, Prev: SSA Operands, Up: Tree SSA 14784 14785 13.3 Static Single Assignment 14786 ============================= 14787 14788 Most of the tree optimizers rely on the data flow information provided 14789 by the Static Single Assignment (SSA) form. We implement the SSA form 14790 as described in `R. Cytron, J. Ferrante, B. Rosen, M. Wegman, and K. 14791 Zadeck. Efficiently Computing Static Single Assignment Form and the 14792 Control Dependence Graph. ACM Transactions on Programming Languages 14793 and Systems, 13(4):451-490, October 1991'. 14794 14795 The SSA form is based on the premise that program variables are 14796 assigned in exactly one location in the program. Multiple assignments 14797 to the same variable create new versions of that variable. Naturally, 14798 actual programs are seldom in SSA form initially because variables tend 14799 to be assigned multiple times. The compiler modifies the program 14800 representation so that every time a variable is assigned in the code, a 14801 new version of the variable is created. Different versions of the same 14802 variable are distinguished by subscripting the variable name with its 14803 version number. Variables used in the right-hand side of expressions 14804 are renamed so that their version number matches that of the most 14805 recent assignment. 14806 14807 We represent variable versions using `SSA_NAME' nodes. The renaming 14808 process in `tree-ssa.c' wraps every real and virtual operand with an 14809 `SSA_NAME' node which contains the version number and the statement 14810 that created the `SSA_NAME'. Only definitions and virtual definitions 14811 may create new `SSA_NAME' nodes. 14812 14813 Sometimes, flow of control makes it impossible to determine the most 14814 recent version of a variable. In these cases, the compiler inserts an 14815 artificial definition for that variable called "PHI function" or "PHI 14816 node". This new definition merges all the incoming versions of the 14817 variable to create a new name for it. For instance, 14818 14819 if (...) 14820 a_1 = 5; 14821 else if (...) 14822 a_2 = 2; 14823 else 14824 a_3 = 13; 14825 14826 # a_4 = PHI <a_1, a_2, a_3> 14827 return a_4; 14828 14829 Since it is not possible to determine which of the three branches will 14830 be taken at runtime, we don't know which of `a_1', `a_2' or `a_3' to 14831 use at the return statement. So, the SSA renamer creates a new version 14832 `a_4' which is assigned the result of "merging" `a_1', `a_2' and `a_3'. 14833 Hence, PHI nodes mean "one of these operands. I don't know which". 14834 14835 The following macros can be used to examine PHI nodes 14836 14837 -- Macro: PHI_RESULT (PHI) 14838 Returns the `SSA_NAME' created by PHI node PHI (i.e., PHI's LHS). 14839 14840 -- Macro: PHI_NUM_ARGS (PHI) 14841 Returns the number of arguments in PHI. This number is exactly 14842 the number of incoming edges to the basic block holding PHI. 14843 14844 -- Macro: PHI_ARG_ELT (PHI, I) 14845 Returns a tuple representing the Ith argument of PHI. Each 14846 element of this tuple contains an `SSA_NAME' VAR and the incoming 14847 edge through which VAR flows. 14848 14849 -- Macro: PHI_ARG_EDGE (PHI, I) 14850 Returns the incoming edge for the Ith argument of PHI. 14851 14852 -- Macro: PHI_ARG_DEF (PHI, I) 14853 Returns the `SSA_NAME' for the Ith argument of PHI. 14854 14855 13.3.1 Preserving the SSA form 14856 ------------------------------ 14857 14858 Some optimization passes make changes to the function that invalidate 14859 the SSA property. This can happen when a pass has added new symbols or 14860 changed the program so that variables that were previously aliased 14861 aren't anymore. Whenever something like this happens, the affected 14862 symbols must be renamed into SSA form again. Transformations that emit 14863 new code or replicate existing statements will also need to update the 14864 SSA form. 14865 14866 Since GCC implements two different SSA forms for register and virtual 14867 variables, keeping the SSA form up to date depends on whether you are 14868 updating register or virtual names. In both cases, the general idea 14869 behind incremental SSA updates is similar: when new SSA names are 14870 created, they typically are meant to replace other existing names in 14871 the program. 14872 14873 For instance, given the following code: 14874 14875 1 L0: 14876 2 x_1 = PHI (0, x_5) 14877 3 if (x_1 < 10) 14878 4 if (x_1 > 7) 14879 5 y_2 = 0 14880 6 else 14881 7 y_3 = x_1 + x_7 14882 8 endif 14883 9 x_5 = x_1 + 1 14884 10 goto L0; 14885 11 endif 14886 14887 Suppose that we insert new names `x_10' and `x_11' (lines `4' and `8'). 14888 14889 1 L0: 14890 2 x_1 = PHI (0, x_5) 14891 3 if (x_1 < 10) 14892 4 x_10 = ... 14893 5 if (x_1 > 7) 14894 6 y_2 = 0 14895 7 else 14896 8 x_11 = ... 14897 9 y_3 = x_1 + x_7 14898 10 endif 14899 11 x_5 = x_1 + 1 14900 12 goto L0; 14901 13 endif 14902 14903 We want to replace all the uses of `x_1' with the new definitions of 14904 `x_10' and `x_11'. Note that the only uses that should be replaced are 14905 those at lines `5', `9' and `11'. Also, the use of `x_7' at line `9' 14906 should _not_ be replaced (this is why we cannot just mark symbol `x' for 14907 renaming). 14908 14909 Additionally, we may need to insert a PHI node at line `11' because 14910 that is a merge point for `x_10' and `x_11'. So the use of `x_1' at 14911 line `11' will be replaced with the new PHI node. The insertion of PHI 14912 nodes is optional. They are not strictly necessary to preserve the SSA 14913 form, and depending on what the caller inserted, they may not even be 14914 useful for the optimizers. 14915 14916 Updating the SSA form is a two step process. First, the pass has to 14917 identify which names need to be updated and/or which symbols need to be 14918 renamed into SSA form for the first time. When new names are 14919 introduced to replace existing names in the program, the mapping 14920 between the old and the new names are registered by calling 14921 `register_new_name_mapping' (note that if your pass creates new code by 14922 duplicating basic blocks, the call to `tree_duplicate_bb' will set up 14923 the necessary mappings automatically). On the other hand, if your pass 14924 exposes a new symbol that should be put in SSA form for the first time, 14925 the new symbol should be registered with `mark_sym_for_renaming'. 14926 14927 After the replacement mappings have been registered and new symbols 14928 marked for renaming, a call to `update_ssa' makes the registered 14929 changes. This can be done with an explicit call or by creating `TODO' 14930 flags in the `tree_opt_pass' structure for your pass. There are 14931 several `TODO' flags that control the behavior of `update_ssa': 14932 14933 * `TODO_update_ssa'. Update the SSA form inserting PHI nodes for 14934 newly exposed symbols and virtual names marked for updating. When 14935 updating real names, only insert PHI nodes for a real name `O_j' 14936 in blocks reached by all the new and old definitions for `O_j'. 14937 If the iterated dominance frontier for `O_j' is not pruned, we may 14938 end up inserting PHI nodes in blocks that have one or more edges 14939 with no incoming definition for `O_j'. This would lead to 14940 uninitialized warnings for `O_j''s symbol. 14941 14942 * `TODO_update_ssa_no_phi'. Update the SSA form without inserting 14943 any new PHI nodes at all. This is used by passes that have either 14944 inserted all the PHI nodes themselves or passes that need only to 14945 patch use-def and def-def chains for virtuals (e.g., DCE). 14946 14947 * `TODO_update_ssa_full_phi'. Insert PHI nodes everywhere they are 14948 needed. No pruning of the IDF is done. This is used by passes 14949 that need the PHI nodes for `O_j' even if it means that some 14950 arguments will come from the default definition of `O_j''s symbol 14951 (e.g., `pass_linear_transform'). 14952 14953 WARNING: If you need to use this flag, chances are that your pass 14954 may be doing something wrong. Inserting PHI nodes for an old name 14955 where not all edges carry a new replacement may lead to silent 14956 codegen errors or spurious uninitialized warnings. 14957 14958 * `TODO_update_ssa_only_virtuals'. Passes that update the SSA form 14959 on their own may want to delegate the updating of virtual names to 14960 the generic updater. Since FUD chains are easier to maintain, 14961 this simplifies the work they need to do. NOTE: If this flag is 14962 used, any OLD->NEW mappings for real names are explicitly 14963 destroyed and only the symbols marked for renaming are processed. 14964 14965 13.3.2 Preserving the virtual SSA form 14966 -------------------------------------- 14967 14968 The virtual SSA form is harder to preserve than the non-virtual SSA form 14969 mainly because the set of virtual operands for a statement may change at 14970 what some would consider unexpected times. In general, statement 14971 modifications should be bracketed between calls to `push_stmt_changes' 14972 and `pop_stmt_changes'. For example, 14973 14974 munge_stmt (tree stmt) 14975 { 14976 push_stmt_changes (&stmt); 14977 ... rewrite STMT ... 14978 pop_stmt_changes (&stmt); 14979 } 14980 14981 The call to `push_stmt_changes' saves the current state of the 14982 statement operands and the call to `pop_stmt_changes' compares the 14983 saved state with the current one and does the appropriate symbol 14984 marking for the SSA renamer. 14985 14986 It is possible to modify several statements at a time, provided that 14987 `push_stmt_changes' and `pop_stmt_changes' are called in LIFO order, as 14988 when processing a stack of statements. 14989 14990 Additionally, if the pass discovers that it did not need to make 14991 changes to the statement after calling `push_stmt_changes', it can 14992 simply discard the topmost change buffer by calling 14993 `discard_stmt_changes'. This will avoid the expensive operand re-scan 14994 operation and the buffer comparison that determines if symbols need to 14995 be marked for renaming. 14996 14997 13.3.3 Examining `SSA_NAME' nodes 14998 --------------------------------- 14999 15000 The following macros can be used to examine `SSA_NAME' nodes 15001 15002 -- Macro: SSA_NAME_DEF_STMT (VAR) 15003 Returns the statement S that creates the `SSA_NAME' VAR. If S is 15004 an empty statement (i.e., `IS_EMPTY_STMT (S)' returns `true'), it 15005 means that the first reference to this variable is a USE or a VUSE. 15006 15007 -- Macro: SSA_NAME_VERSION (VAR) 15008 Returns the version number of the `SSA_NAME' object VAR. 15009 15010 13.3.4 Walking use-def chains 15011 ----------------------------- 15012 15013 -- Tree SSA function: void walk_use_def_chains (VAR, FN, DATA) 15014 Walks use-def chains starting at the `SSA_NAME' node VAR. Calls 15015 function FN at each reaching definition found. Function FN takes 15016 three arguments: VAR, its defining statement (DEF_STMT) and a 15017 generic pointer to whatever state information that FN may want to 15018 maintain (DATA). Function FN is able to stop the walk by 15019 returning `true', otherwise in order to continue the walk, FN 15020 should return `false'. 15021 15022 Note, that if DEF_STMT is a `PHI' node, the semantics are slightly 15023 different. For each argument ARG of the PHI node, this function 15024 will: 15025 15026 1. Walk the use-def chains for ARG. 15027 15028 2. Call `FN (ARG, PHI, DATA)'. 15029 15030 Note how the first argument to FN is no longer the original 15031 variable VAR, but the PHI argument currently being examined. If 15032 FN wants to get at VAR, it should call `PHI_RESULT' (PHI). 15033 15034 13.3.5 Walking the dominator tree 15035 --------------------------------- 15036 15037 -- Tree SSA function: void walk_dominator_tree (WALK_DATA, BB) 15038 This function walks the dominator tree for the current CFG calling 15039 a set of callback functions defined in STRUCT DOM_WALK_DATA in 15040 `domwalk.h'. The call back functions you need to define give you 15041 hooks to execute custom code at various points during traversal: 15042 15043 1. Once to initialize any local data needed while processing BB 15044 and its children. This local data is pushed into an internal 15045 stack which is automatically pushed and popped as the walker 15046 traverses the dominator tree. 15047 15048 2. Once before traversing all the statements in the BB. 15049 15050 3. Once for every statement inside BB. 15051 15052 4. Once after traversing all the statements and before recursing 15053 into BB's dominator children. 15054 15055 5. It then recurses into all the dominator children of BB. 15056 15057 6. After recursing into all the dominator children of BB it can, 15058 optionally, traverse every statement in BB again (i.e., 15059 repeating steps 2 and 3). 15060 15061 7. Once after walking the statements in BB and BB's dominator 15062 children. At this stage, the block local data stack is 15063 popped. 15064 15065 15066 File: gccint.info, Node: Alias analysis, Prev: SSA, Up: Tree SSA 15067 15068 13.4 Alias analysis 15069 =================== 15070 15071 Alias analysis proceeds in 4 main phases: 15072 15073 1. Structural alias analysis. 15074 15075 This phase walks the types for structure variables, and determines 15076 which of the fields can overlap using offset and size of each 15077 field. For each field, a "subvariable" called a "Structure field 15078 tag" (SFT) is created, which represents that field as a separate 15079 variable. All accesses that could possibly overlap with a given 15080 field will have virtual operands for the SFT of that field. 15081 15082 struct foo 15083 { 15084 int a; 15085 int b; 15086 } 15087 struct foo temp; 15088 int bar (void) 15089 { 15090 int tmp1, tmp2, tmp3; 15091 SFT.0_2 = VDEF <SFT.0_1> 15092 temp.a = 5; 15093 SFT.1_4 = VDEF <SFT.1_3> 15094 temp.b = 6; 15095 15096 VUSE <SFT.1_4> 15097 tmp1_5 = temp.b; 15098 VUSE <SFT.0_2> 15099 tmp2_6 = temp.a; 15100 15101 tmp3_7 = tmp1_5 + tmp2_6; 15102 return tmp3_7; 15103 } 15104 15105 If you copy the symbol tag for a variable for some reason, you 15106 probably also want to copy the subvariables for that variable. 15107 15108 2. Points-to and escape analysis. 15109 15110 This phase walks the use-def chains in the SSA web looking for 15111 three things: 15112 15113 * Assignments of the form `P_i = &VAR' 15114 15115 * Assignments of the form P_i = malloc() 15116 15117 * Pointers and ADDR_EXPR that escape the current function. 15118 15119 The concept of `escaping' is the same one used in the Java world. 15120 When a pointer or an ADDR_EXPR escapes, it means that it has been 15121 exposed outside of the current function. So, assignment to global 15122 variables, function arguments and returning a pointer are all 15123 escape sites. 15124 15125 This is where we are currently limited. Since not everything is 15126 renamed into SSA, we lose track of escape properties when a 15127 pointer is stashed inside a field in a structure, for instance. 15128 In those cases, we are assuming that the pointer does escape. 15129 15130 We use escape analysis to determine whether a variable is 15131 call-clobbered. Simply put, if an ADDR_EXPR escapes, then the 15132 variable is call-clobbered. If a pointer P_i escapes, then all 15133 the variables pointed-to by P_i (and its memory tag) also escape. 15134 15135 3. Compute flow-sensitive aliases 15136 15137 We have two classes of memory tags. Memory tags associated with 15138 the pointed-to data type of the pointers in the program. These 15139 tags are called "symbol memory tag" (SMT). The other class are 15140 those associated with SSA_NAMEs, called "name memory tag" (NMT). 15141 The basic idea is that when adding operands for an INDIRECT_REF 15142 *P_i, we will first check whether P_i has a name tag, if it does 15143 we use it, because that will have more precise aliasing 15144 information. Otherwise, we use the standard symbol tag. 15145 15146 In this phase, we go through all the pointers we found in 15147 points-to analysis and create alias sets for the name memory tags 15148 associated with each pointer P_i. If P_i escapes, we mark 15149 call-clobbered the variables it points to and its tag. 15150 15151 4. Compute flow-insensitive aliases 15152 15153 This pass will compare the alias set of every symbol memory tag and 15154 every addressable variable found in the program. Given a symbol 15155 memory tag SMT and an addressable variable V. If the alias sets 15156 of SMT and V conflict (as computed by may_alias_p), then V is 15157 marked as an alias tag and added to the alias set of SMT. 15158 15159 Every language that wishes to perform language-specific alias 15160 analysis should define a function that computes, given a `tree' 15161 node, an alias set for the node. Nodes in different alias sets 15162 are not allowed to alias. For an example, see the C front-end 15163 function `c_get_alias_set'. 15164 15165 For instance, consider the following function: 15166 15167 foo (int i) 15168 { 15169 int *p, *q, a, b; 15170 15171 if (i > 10) 15172 p = &a; 15173 else 15174 q = &b; 15175 15176 *p = 3; 15177 *q = 5; 15178 a = b + 2; 15179 return *p; 15180 } 15181 15182 After aliasing analysis has finished, the symbol memory tag for 15183 pointer `p' will have two aliases, namely variables `a' and `b'. Every 15184 time pointer `p' is dereferenced, we want to mark the operation as a 15185 potential reference to `a' and `b'. 15186 15187 foo (int i) 15188 { 15189 int *p, a, b; 15190 15191 if (i_2 > 10) 15192 p_4 = &a; 15193 else 15194 p_6 = &b; 15195 # p_1 = PHI <p_4(1), p_6(2)>; 15196 15197 # a_7 = VDEF <a_3>; 15198 # b_8 = VDEF <b_5>; 15199 *p_1 = 3; 15200 15201 # a_9 = VDEF <a_7> 15202 # VUSE <b_8> 15203 a_9 = b_8 + 2; 15204 15205 # VUSE <a_9>; 15206 # VUSE <b_8>; 15207 return *p_1; 15208 } 15209 15210 In certain cases, the list of may aliases for a pointer may grow too 15211 large. This may cause an explosion in the number of virtual operands 15212 inserted in the code. Resulting in increased memory consumption and 15213 compilation time. 15214 15215 When the number of virtual operands needed to represent aliased loads 15216 and stores grows too large (configurable with `--param 15217 max-aliased-vops'), alias sets are grouped to avoid severe compile-time 15218 slow downs and memory consumption. The alias grouping heuristic 15219 proceeds as follows: 15220 15221 1. Sort the list of pointers in decreasing number of contributed 15222 virtual operands. 15223 15224 2. Take the first pointer from the list and reverse the role of the 15225 memory tag and its aliases. Usually, whenever an aliased variable 15226 Vi is found to alias with a memory tag T, we add Vi to the 15227 may-aliases set for T. Meaning that after alias analysis, we will 15228 have: 15229 15230 may-aliases(T) = { V1, V2, V3, ..., Vn } 15231 15232 This means that every statement that references T, will get `n' 15233 virtual operands for each of the Vi tags. But, when alias 15234 grouping is enabled, we make T an alias tag and add it to the 15235 alias set of all the Vi variables: 15236 15237 may-aliases(V1) = { T } 15238 may-aliases(V2) = { T } 15239 ... 15240 may-aliases(Vn) = { T } 15241 15242 This has two effects: (a) statements referencing T will only get a 15243 single virtual operand, and, (b) all the variables Vi will now 15244 appear to alias each other. So, we lose alias precision to 15245 improve compile time. But, in theory, a program with such a high 15246 level of aliasing should not be very optimizable in the first 15247 place. 15248 15249 3. Since variables may be in the alias set of more than one memory 15250 tag, the grouping done in step (2) needs to be extended to all the 15251 memory tags that have a non-empty intersection with the 15252 may-aliases set of tag T. For instance, if we originally had 15253 these may-aliases sets: 15254 15255 may-aliases(T) = { V1, V2, V3 } 15256 may-aliases(R) = { V2, V4 } 15257 15258 In step (2) we would have reverted the aliases for T as: 15259 15260 may-aliases(V1) = { T } 15261 may-aliases(V2) = { T } 15262 may-aliases(V3) = { T } 15263 15264 But note that now V2 is no longer aliased with R. We could add R 15265 to may-aliases(V2), but we are in the process of grouping aliases 15266 to reduce virtual operands so what we do is add V4 to the grouping 15267 to obtain: 15268 15269 may-aliases(V1) = { T } 15270 may-aliases(V2) = { T } 15271 may-aliases(V3) = { T } 15272 may-aliases(V4) = { T } 15273 15274 4. If the total number of virtual operands due to aliasing is still 15275 above the threshold set by max-alias-vops, go back to (2). 15276 15277 15278 File: gccint.info, Node: Loop Analysis and Representation, Next: Machine Desc, Prev: Control Flow, Up: Top 15279 15280 14 Analysis and Representation of Loops 15281 *************************************** 15282 15283 GCC provides extensive infrastructure for work with natural loops, i.e., 15284 strongly connected components of CFG with only one entry block. This 15285 chapter describes representation of loops in GCC, both on GIMPLE and in 15286 RTL, as well as the interfaces to loop-related analyses (induction 15287 variable analysis and number of iterations analysis). 15288 15289 * Menu: 15290 15291 * Loop representation:: Representation and analysis of loops. 15292 * Loop querying:: Getting information about loops. 15293 * Loop manipulation:: Loop manipulation functions. 15294 * LCSSA:: Loop-closed SSA form. 15295 * Scalar evolutions:: Induction variables on GIMPLE. 15296 * loop-iv:: Induction variables on RTL. 15297 * Number of iterations:: Number of iterations analysis. 15298 * Dependency analysis:: Data dependency analysis. 15299 * Lambda:: Linear loop transformations framework. 15300 * Omega:: A solver for linear programming problems. 15301 15302 15303 File: gccint.info, Node: Loop representation, Next: Loop querying, Up: Loop Analysis and Representation 15304 15305 14.1 Loop representation 15306 ======================== 15307 15308 This chapter describes the representation of loops in GCC, and functions 15309 that can be used to build, modify and analyze this representation. Most 15310 of the interfaces and data structures are declared in `cfgloop.h'. At 15311 the moment, loop structures are analyzed and this information is 15312 updated only by the optimization passes that deal with loops, but some 15313 efforts are being made to make it available throughout most of the 15314 optimization passes. 15315 15316 In general, a natural loop has one entry block (header) and possibly 15317 several back edges (latches) leading to the header from the inside of 15318 the loop. Loops with several latches may appear if several loops share 15319 a single header, or if there is a branching in the middle of the loop. 15320 The representation of loops in GCC however allows only loops with a 15321 single latch. During loop analysis, headers of such loops are split and 15322 forwarder blocks are created in order to disambiguate their structures. 15323 Heuristic based on profile information and structure of the induction 15324 variables in the loops is used to determine whether the latches 15325 correspond to sub-loops or to control flow in a single loop. This means 15326 that the analysis sometimes changes the CFG, and if you run it in the 15327 middle of an optimization pass, you must be able to deal with the new 15328 blocks. You may avoid CFG changes by passing 15329 `LOOPS_MAY_HAVE_MULTIPLE_LATCHES' flag to the loop discovery, note 15330 however that most other loop manipulation functions will not work 15331 correctly for loops with multiple latch edges (the functions that only 15332 query membership of blocks to loops and subloop relationships, or 15333 enumerate and test loop exits, can be expected to work). 15334 15335 Body of the loop is the set of blocks that are dominated by its header, 15336 and reachable from its latch against the direction of edges in CFG. The 15337 loops are organized in a containment hierarchy (tree) such that all the 15338 loops immediately contained inside loop L are the children of L in the 15339 tree. This tree is represented by the `struct loops' structure. The 15340 root of this tree is a fake loop that contains all blocks in the 15341 function. Each of the loops is represented in a `struct loop' 15342 structure. Each loop is assigned an index (`num' field of the `struct 15343 loop' structure), and the pointer to the loop is stored in the 15344 corresponding field of the `larray' vector in the loops structure. The 15345 indices do not have to be continuous, there may be empty (`NULL') 15346 entries in the `larray' created by deleting loops. Also, there is no 15347 guarantee on the relative order of a loop and its subloops in the 15348 numbering. The index of a loop never changes. 15349 15350 The entries of the `larray' field should not be accessed directly. 15351 The function `get_loop' returns the loop description for a loop with 15352 the given index. `number_of_loops' function returns number of loops in 15353 the function. To traverse all loops, use `FOR_EACH_LOOP' macro. The 15354 `flags' argument of the macro is used to determine the direction of 15355 traversal and the set of loops visited. Each loop is guaranteed to be 15356 visited exactly once, regardless of the changes to the loop tree, and 15357 the loops may be removed during the traversal. The newly created loops 15358 are never traversed, if they need to be visited, this must be done 15359 separately after their creation. The `FOR_EACH_LOOP' macro allocates 15360 temporary variables. If the `FOR_EACH_LOOP' loop were ended using 15361 break or goto, they would not be released; `FOR_EACH_LOOP_BREAK' macro 15362 must be used instead. 15363 15364 Each basic block contains the reference to the innermost loop it 15365 belongs to (`loop_father'). For this reason, it is only possible to 15366 have one `struct loops' structure initialized at the same time for each 15367 CFG. The global variable `current_loops' contains the `struct loops' 15368 structure. Many of the loop manipulation functions assume that 15369 dominance information is up-to-date. 15370 15371 The loops are analyzed through `loop_optimizer_init' function. The 15372 argument of this function is a set of flags represented in an integer 15373 bitmask. These flags specify what other properties of the loop 15374 structures should be calculated/enforced and preserved later: 15375 15376 * `LOOPS_MAY_HAVE_MULTIPLE_LATCHES': If this flag is set, no changes 15377 to CFG will be performed in the loop analysis, in particular, 15378 loops with multiple latch edges will not be disambiguated. If a 15379 loop has multiple latches, its latch block is set to NULL. Most of 15380 the loop manipulation functions will not work for loops in this 15381 shape. No other flags that require CFG changes can be passed to 15382 loop_optimizer_init. 15383 15384 * `LOOPS_HAVE_PREHEADERS': Forwarder blocks are created in such a 15385 way that each loop has only one entry edge, and additionally, the 15386 source block of this entry edge has only one successor. This 15387 creates a natural place where the code can be moved out of the 15388 loop, and ensures that the entry edge of the loop leads from its 15389 immediate super-loop. 15390 15391 * `LOOPS_HAVE_SIMPLE_LATCHES': Forwarder blocks are created to force 15392 the latch block of each loop to have only one successor. This 15393 ensures that the latch of the loop does not belong to any of its 15394 sub-loops, and makes manipulation with the loops significantly 15395 easier. Most of the loop manipulation functions assume that the 15396 loops are in this shape. Note that with this flag, the "normal" 15397 loop without any control flow inside and with one exit consists of 15398 two basic blocks. 15399 15400 * `LOOPS_HAVE_MARKED_IRREDUCIBLE_REGIONS': Basic blocks and edges in 15401 the strongly connected components that are not natural loops (have 15402 more than one entry block) are marked with `BB_IRREDUCIBLE_LOOP' 15403 and `EDGE_IRREDUCIBLE_LOOP' flags. The flag is not set for blocks 15404 and edges that belong to natural loops that are in such an 15405 irreducible region (but it is set for the entry and exit edges of 15406 such a loop, if they lead to/from this region). 15407 15408 * `LOOPS_HAVE_RECORDED_EXITS': The lists of exits are recorded and 15409 updated for each loop. This makes some functions (e.g., 15410 `get_loop_exit_edges') more efficient. Some functions (e.g., 15411 `single_exit') can be used only if the lists of exits are recorded. 15412 15413 These properties may also be computed/enforced later, using functions 15414 `create_preheaders', `force_single_succ_latches', 15415 `mark_irreducible_loops' and `record_loop_exits'. 15416 15417 The memory occupied by the loops structures should be freed with 15418 `loop_optimizer_finalize' function. 15419 15420 The CFG manipulation functions in general do not update loop 15421 structures. Specialized versions that additionally do so are provided 15422 for the most common tasks. On GIMPLE, `cleanup_tree_cfg_loop' function 15423 can be used to cleanup CFG while updating the loops structures if 15424 `current_loops' is set. 15425 15426 15427 File: gccint.info, Node: Loop querying, Next: Loop manipulation, Prev: Loop representation, Up: Loop Analysis and Representation 15428 15429 14.2 Loop querying 15430 ================== 15431 15432 The functions to query the information about loops are declared in 15433 `cfgloop.h'. Some of the information can be taken directly from the 15434 structures. `loop_father' field of each basic block contains the 15435 innermost loop to that the block belongs. The most useful fields of 15436 loop structure (that are kept up-to-date at all times) are: 15437 15438 * `header', `latch': Header and latch basic blocks of the loop. 15439 15440 * `num_nodes': Number of basic blocks in the loop (including the 15441 basic blocks of the sub-loops). 15442 15443 * `depth': The depth of the loop in the loops tree, i.e., the number 15444 of super-loops of the loop. 15445 15446 * `outer', `inner', `next': The super-loop, the first sub-loop, and 15447 the sibling of the loop in the loops tree. 15448 15449 There are other fields in the loop structures, many of them used only 15450 by some of the passes, or not updated during CFG changes; in general, 15451 they should not be accessed directly. 15452 15453 The most important functions to query loop structures are: 15454 15455 * `flow_loops_dump': Dumps the information about loops to a file. 15456 15457 * `verify_loop_structure': Checks consistency of the loop structures. 15458 15459 * `loop_latch_edge': Returns the latch edge of a loop. 15460 15461 * `loop_preheader_edge': If loops have preheaders, returns the 15462 preheader edge of a loop. 15463 15464 * `flow_loop_nested_p': Tests whether loop is a sub-loop of another 15465 loop. 15466 15467 * `flow_bb_inside_loop_p': Tests whether a basic block belongs to a 15468 loop (including its sub-loops). 15469 15470 * `find_common_loop': Finds the common super-loop of two loops. 15471 15472 * `superloop_at_depth': Returns the super-loop of a loop with the 15473 given depth. 15474 15475 * `tree_num_loop_insns', `num_loop_insns': Estimates the number of 15476 insns in the loop, on GIMPLE and on RTL. 15477 15478 * `loop_exit_edge_p': Tests whether edge is an exit from a loop. 15479 15480 * `mark_loop_exit_edges': Marks all exit edges of all loops with 15481 `EDGE_LOOP_EXIT' flag. 15482 15483 * `get_loop_body', `get_loop_body_in_dom_order', 15484 `get_loop_body_in_bfs_order': Enumerates the basic blocks in the 15485 loop in depth-first search order in reversed CFG, ordered by 15486 dominance relation, and breath-first search order, respectively. 15487 15488 * `single_exit': Returns the single exit edge of the loop, or `NULL' 15489 if the loop has more than one exit. You can only use this 15490 function if LOOPS_HAVE_MARKED_SINGLE_EXITS property is used. 15491 15492 * `get_loop_exit_edges': Enumerates the exit edges of a loop. 15493 15494 * `just_once_each_iteration_p': Returns true if the basic block is 15495 executed exactly once during each iteration of a loop (that is, it 15496 does not belong to a sub-loop, and it dominates the latch of the 15497 loop). 15498 15499 15500 File: gccint.info, Node: Loop manipulation, Next: LCSSA, Prev: Loop querying, Up: Loop Analysis and Representation 15501 15502 14.3 Loop manipulation 15503 ====================== 15504 15505 The loops tree can be manipulated using the following functions: 15506 15507 * `flow_loop_tree_node_add': Adds a node to the tree. 15508 15509 * `flow_loop_tree_node_remove': Removes a node from the tree. 15510 15511 * `add_bb_to_loop': Adds a basic block to a loop. 15512 15513 * `remove_bb_from_loops': Removes a basic block from loops. 15514 15515 Most low-level CFG functions update loops automatically. The following 15516 functions handle some more complicated cases of CFG manipulations: 15517 15518 * `remove_path': Removes an edge and all blocks it dominates. 15519 15520 * `split_loop_exit_edge': Splits exit edge of the loop, ensuring 15521 that PHI node arguments remain in the loop (this ensures that 15522 loop-closed SSA form is preserved). Only useful on GIMPLE. 15523 15524 Finally, there are some higher-level loop transformations implemented. 15525 While some of them are written so that they should work on non-innermost 15526 loops, they are mostly untested in that case, and at the moment, they 15527 are only reliable for the innermost loops: 15528 15529 * `create_iv': Creates a new induction variable. Only works on 15530 GIMPLE. `standard_iv_increment_position' can be used to find a 15531 suitable place for the iv increment. 15532 15533 * `duplicate_loop_to_header_edge', 15534 `tree_duplicate_loop_to_header_edge': These functions (on RTL and 15535 on GIMPLE) duplicate the body of the loop prescribed number of 15536 times on one of the edges entering loop header, thus performing 15537 either loop unrolling or loop peeling. `can_duplicate_loop_p' 15538 (`can_unroll_loop_p' on GIMPLE) must be true for the duplicated 15539 loop. 15540 15541 * `loop_version', `tree_ssa_loop_version': These function create a 15542 copy of a loop, and a branch before them that selects one of them 15543 depending on the prescribed condition. This is useful for 15544 optimizations that need to verify some assumptions in runtime (one 15545 of the copies of the loop is usually left unchanged, while the 15546 other one is transformed in some way). 15547 15548 * `tree_unroll_loop': Unrolls the loop, including peeling the extra 15549 iterations to make the number of iterations divisible by unroll 15550 factor, updating the exit condition, and removing the exits that 15551 now cannot be taken. Works only on GIMPLE. 15552 15553 15554 File: gccint.info, Node: LCSSA, Next: Scalar evolutions, Prev: Loop manipulation, Up: Loop Analysis and Representation 15555 15556 14.4 Loop-closed SSA form 15557 ========================= 15558 15559 Throughout the loop optimizations on tree level, one extra condition is 15560 enforced on the SSA form: No SSA name is used outside of the loop in 15561 that it is defined. The SSA form satisfying this condition is called 15562 "loop-closed SSA form" - LCSSA. To enforce LCSSA, PHI nodes must be 15563 created at the exits of the loops for the SSA names that are used 15564 outside of them. Only the real operands (not virtual SSA names) are 15565 held in LCSSA, in order to save memory. 15566 15567 There are various benefits of LCSSA: 15568 15569 * Many optimizations (value range analysis, final value replacement) 15570 are interested in the values that are defined in the loop and used 15571 outside of it, i.e., exactly those for that we create new PHI 15572 nodes. 15573 15574 * In induction variable analysis, it is not necessary to specify the 15575 loop in that the analysis should be performed - the scalar 15576 evolution analysis always returns the results with respect to the 15577 loop in that the SSA name is defined. 15578 15579 * It makes updating of SSA form during loop transformations simpler. 15580 Without LCSSA, operations like loop unrolling may force creation 15581 of PHI nodes arbitrarily far from the loop, while in LCSSA, the 15582 SSA form can be updated locally. However, since we only keep real 15583 operands in LCSSA, we cannot use this advantage (we could have 15584 local updating of real operands, but it is not much more efficient 15585 than to use generic SSA form updating for it as well; the amount 15586 of changes to SSA is the same). 15587 15588 However, it also means LCSSA must be updated. This is usually 15589 straightforward, unless you create a new value in loop and use it 15590 outside, or unless you manipulate loop exit edges (functions are 15591 provided to make these manipulations simple). 15592 `rewrite_into_loop_closed_ssa' is used to rewrite SSA form to LCSSA, 15593 and `verify_loop_closed_ssa' to check that the invariant of LCSSA is 15594 preserved. 15595 15596 15597 File: gccint.info, Node: Scalar evolutions, Next: loop-iv, Prev: LCSSA, Up: Loop Analysis and Representation 15598 15599 14.5 Scalar evolutions 15600 ====================== 15601 15602 Scalar evolutions (SCEV) are used to represent results of induction 15603 variable analysis on GIMPLE. They enable us to represent variables with 15604 complicated behavior in a simple and consistent way (we only use it to 15605 express values of polynomial induction variables, but it is possible to 15606 extend it). The interfaces to SCEV analysis are declared in 15607 `tree-scalar-evolution.h'. To use scalar evolutions analysis, 15608 `scev_initialize' must be used. To stop using SCEV, `scev_finalize' 15609 should be used. SCEV analysis caches results in order to save time and 15610 memory. This cache however is made invalid by most of the loop 15611 transformations, including removal of code. If such a transformation 15612 is performed, `scev_reset' must be called to clean the caches. 15613 15614 Given an SSA name, its behavior in loops can be analyzed using the 15615 `analyze_scalar_evolution' function. The returned SCEV however does 15616 not have to be fully analyzed and it may contain references to other 15617 SSA names defined in the loop. To resolve these (potentially 15618 recursive) references, `instantiate_parameters' or `resolve_mixers' 15619 functions must be used. `instantiate_parameters' is useful when you 15620 use the results of SCEV only for some analysis, and when you work with 15621 whole nest of loops at once. It will try replacing all SSA names by 15622 their SCEV in all loops, including the super-loops of the current loop, 15623 thus providing a complete information about the behavior of the 15624 variable in the loop nest. `resolve_mixers' is useful if you work with 15625 only one loop at a time, and if you possibly need to create code based 15626 on the value of the induction variable. It will only resolve the SSA 15627 names defined in the current loop, leaving the SSA names defined 15628 outside unchanged, even if their evolution in the outer loops is known. 15629 15630 The SCEV is a normal tree expression, except for the fact that it may 15631 contain several special tree nodes. One of them is `SCEV_NOT_KNOWN', 15632 used for SSA names whose value cannot be expressed. The other one is 15633 `POLYNOMIAL_CHREC'. Polynomial chrec has three arguments - base, step 15634 and loop (both base and step may contain further polynomial chrecs). 15635 Type of the expression and of base and step must be the same. A 15636 variable has evolution `POLYNOMIAL_CHREC(base, step, loop)' if it is 15637 (in the specified loop) equivalent to `x_1' in the following example 15638 15639 while (...) 15640 { 15641 x_1 = phi (base, x_2); 15642 x_2 = x_1 + step; 15643 } 15644 15645 Note that this includes the language restrictions on the operations. 15646 For example, if we compile C code and `x' has signed type, then the 15647 overflow in addition would cause undefined behavior, and we may assume 15648 that this does not happen. Hence, the value with this SCEV cannot 15649 overflow (which restricts the number of iterations of such a loop). 15650 15651 In many cases, one wants to restrict the attention just to affine 15652 induction variables. In this case, the extra expressive power of SCEV 15653 is not useful, and may complicate the optimizations. In this case, 15654 `simple_iv' function may be used to analyze a value - the result is a 15655 loop-invariant base and step. 15656 15657 15658 File: gccint.info, Node: loop-iv, Next: Number of iterations, Prev: Scalar evolutions, Up: Loop Analysis and Representation 15659 15660 14.6 IV analysis on RTL 15661 ======================= 15662 15663 The induction variable on RTL is simple and only allows analysis of 15664 affine induction variables, and only in one loop at once. The interface 15665 is declared in `cfgloop.h'. Before analyzing induction variables in a 15666 loop L, `iv_analysis_loop_init' function must be called on L. After 15667 the analysis (possibly calling `iv_analysis_loop_init' for several 15668 loops) is finished, `iv_analysis_done' should be called. The following 15669 functions can be used to access the results of the analysis: 15670 15671 * `iv_analyze': Analyzes a single register used in the given insn. 15672 If no use of the register in this insn is found, the following 15673 insns are scanned, so that this function can be called on the insn 15674 returned by get_condition. 15675 15676 * `iv_analyze_result': Analyzes result of the assignment in the 15677 given insn. 15678 15679 * `iv_analyze_expr': Analyzes a more complicated expression. All 15680 its operands are analyzed by `iv_analyze', and hence they must be 15681 used in the specified insn or one of the following insns. 15682 15683 The description of the induction variable is provided in `struct 15684 rtx_iv'. In order to handle subregs, the representation is a bit 15685 complicated; if the value of the `extend' field is not `UNKNOWN', the 15686 value of the induction variable in the i-th iteration is 15687 15688 delta + mult * extend_{extend_mode} (subreg_{mode} (base + i * step)), 15689 15690 with the following exception: if `first_special' is true, then the 15691 value in the first iteration (when `i' is zero) is `delta + mult * 15692 base'. However, if `extend' is equal to `UNKNOWN', then 15693 `first_special' must be false, `delta' 0, `mult' 1 and the value in the 15694 i-th iteration is 15695 15696 subreg_{mode} (base + i * step) 15697 15698 The function `get_iv_value' can be used to perform these calculations. 15699 15700 15701 File: gccint.info, Node: Number of iterations, Next: Dependency analysis, Prev: loop-iv, Up: Loop Analysis and Representation 15702 15703 14.7 Number of iterations analysis 15704 ================================== 15705 15706 Both on GIMPLE and on RTL, there are functions available to determine 15707 the number of iterations of a loop, with a similar interface. The 15708 number of iterations of a loop in GCC is defined as the number of 15709 executions of the loop latch. In many cases, it is not possible to 15710 determine the number of iterations unconditionally - the determined 15711 number is correct only if some assumptions are satisfied. The analysis 15712 tries to verify these conditions using the information contained in the 15713 program; if it fails, the conditions are returned together with the 15714 result. The following information and conditions are provided by the 15715 analysis: 15716 15717 * `assumptions': If this condition is false, the rest of the 15718 information is invalid. 15719 15720 * `noloop_assumptions' on RTL, `may_be_zero' on GIMPLE: If this 15721 condition is true, the loop exits in the first iteration. 15722 15723 * `infinite': If this condition is true, the loop is infinite. This 15724 condition is only available on RTL. On GIMPLE, conditions for 15725 finiteness of the loop are included in `assumptions'. 15726 15727 * `niter_expr' on RTL, `niter' on GIMPLE: The expression that gives 15728 number of iterations. The number of iterations is defined as the 15729 number of executions of the loop latch. 15730 15731 Both on GIMPLE and on RTL, it necessary for the induction variable 15732 analysis framework to be initialized (SCEV on GIMPLE, loop-iv on RTL). 15733 On GIMPLE, the results are stored to `struct tree_niter_desc' 15734 structure. Number of iterations before the loop is exited through a 15735 given exit can be determined using `number_of_iterations_exit' 15736 function. On RTL, the results are returned in `struct niter_desc' 15737 structure. The corresponding function is named `check_simple_exit'. 15738 There are also functions that pass through all the exits of a loop and 15739 try to find one with easy to determine number of iterations - 15740 `find_loop_niter' on GIMPLE and `find_simple_exit' on RTL. Finally, 15741 there are functions that provide the same information, but additionally 15742 cache it, so that repeated calls to number of iterations are not so 15743 costly - `number_of_latch_executions' on GIMPLE and 15744 `get_simple_loop_desc' on RTL. 15745 15746 Note that some of these functions may behave slightly differently than 15747 others - some of them return only the expression for the number of 15748 iterations, and fail if there are some assumptions. The function 15749 `number_of_latch_executions' works only for single-exit loops. The 15750 function `number_of_cond_exit_executions' can be used to determine 15751 number of executions of the exit condition of a single-exit loop (i.e., 15752 the `number_of_latch_executions' increased by one). 15753 15754 15755 File: gccint.info, Node: Dependency analysis, Next: Lambda, Prev: Number of iterations, Up: Loop Analysis and Representation 15756 15757 14.8 Data Dependency Analysis 15758 ============================= 15759 15760 The code for the data dependence analysis can be found in 15761 `tree-data-ref.c' and its interface and data structures are described 15762 in `tree-data-ref.h'. The function that computes the data dependences 15763 for all the array and pointer references for a given loop is 15764 `compute_data_dependences_for_loop'. This function is currently used 15765 by the linear loop transform and the vectorization passes. Before 15766 calling this function, one has to allocate two vectors: a first vector 15767 will contain the set of data references that are contained in the 15768 analyzed loop body, and the second vector will contain the dependence 15769 relations between the data references. Thus if the vector of data 15770 references is of size `n', the vector containing the dependence 15771 relations will contain `n*n' elements. However if the analyzed loop 15772 contains side effects, such as calls that potentially can interfere 15773 with the data references in the current analyzed loop, the analysis 15774 stops while scanning the loop body for data references, and inserts a 15775 single `chrec_dont_know' in the dependence relation array. 15776 15777 The data references are discovered in a particular order during the 15778 scanning of the loop body: the loop body is analyzed in execution order, 15779 and the data references of each statement are pushed at the end of the 15780 data reference array. Two data references syntactically occur in the 15781 program in the same order as in the array of data references. This 15782 syntactic order is important in some classical data dependence tests, 15783 and mapping this order to the elements of this array avoids costly 15784 queries to the loop body representation. 15785 15786 Three types of data references are currently handled: ARRAY_REF, 15787 INDIRECT_REF and COMPONENT_REF. The data structure for the data 15788 reference is `data_reference', where `data_reference_p' is a name of a 15789 pointer to the data reference structure. The structure contains the 15790 following elements: 15791 15792 * `base_object_info': Provides information about the base object of 15793 the data reference and its access functions. These access functions 15794 represent the evolution of the data reference in the loop relative 15795 to its base, in keeping with the classical meaning of the data 15796 reference access function for the support of arrays. For example, 15797 for a reference `a.b[i][j]', the base object is `a.b' and the 15798 access functions, one for each array subscript, are: `{i_init, + 15799 i_step}_1, {j_init, +, j_step}_2'. 15800 15801 * `first_location_in_loop': Provides information about the first 15802 location accessed by the data reference in the loop and about the 15803 access function used to represent evolution relative to this 15804 location. This data is used to support pointers, and is not used 15805 for arrays (for which we have base objects). Pointer accesses are 15806 represented as a one-dimensional access that starts from the first 15807 location accessed in the loop. For example: 15808 15809 for1 i 15810 for2 j 15811 *((int *)p + i + j) = a[i][j]; 15812 15813 The access function of the pointer access is `{0, + 4B}_for2' 15814 relative to `p + i'. The access functions of the array are 15815 `{i_init, + i_step}_for1' and `{j_init, +, j_step}_for2' relative 15816 to `a'. 15817 15818 Usually, the object the pointer refers to is either unknown, or we 15819 can't prove that the access is confined to the boundaries of a 15820 certain object. 15821 15822 Two data references can be compared only if at least one of these 15823 two representations has all its fields filled for both data 15824 references. 15825 15826 The current strategy for data dependence tests is as follows: If 15827 both `a' and `b' are represented as arrays, compare 15828 `a.base_object' and `b.base_object'; if they are equal, apply 15829 dependence tests (use access functions based on base_objects). 15830 Else if both `a' and `b' are represented as pointers, compare 15831 `a.first_location' and `b.first_location'; if they are equal, 15832 apply dependence tests (use access functions based on first 15833 location). However, if `a' and `b' are represented differently, 15834 only try to prove that the bases are definitely different. 15835 15836 * Aliasing information. 15837 15838 * Alignment information. 15839 15840 The structure describing the relation between two data references is 15841 `data_dependence_relation' and the shorter name for a pointer to such a 15842 structure is `ddr_p'. This structure contains: 15843 15844 * a pointer to each data reference, 15845 15846 * a tree node `are_dependent' that is set to `chrec_known' if the 15847 analysis has proved that there is no dependence between these two 15848 data references, `chrec_dont_know' if the analysis was not able to 15849 determine any useful result and potentially there could exist a 15850 dependence between these data references, and `are_dependent' is 15851 set to `NULL_TREE' if there exist a dependence relation between the 15852 data references, and the description of this dependence relation is 15853 given in the `subscripts', `dir_vects', and `dist_vects' arrays, 15854 15855 * a boolean that determines whether the dependence relation can be 15856 represented by a classical distance vector, 15857 15858 * an array `subscripts' that contains a description of each 15859 subscript of the data references. Given two array accesses a 15860 subscript is the tuple composed of the access functions for a given 15861 dimension. For example, given `A[f1][f2][f3]' and 15862 `B[g1][g2][g3]', there are three subscripts: `(f1, g1), (f2, g2), 15863 (f3, g3)'. 15864 15865 * two arrays `dir_vects' and `dist_vects' that contain classical 15866 representations of the data dependences under the form of 15867 direction and distance dependence vectors, 15868 15869 * an array of loops `loop_nest' that contains the loops to which the 15870 distance and direction vectors refer to. 15871 15872 Several functions for pretty printing the information extracted by the 15873 data dependence analysis are available: `dump_ddrs' prints with a 15874 maximum verbosity the details of a data dependence relations array, 15875 `dump_dist_dir_vectors' prints only the classical distance and 15876 direction vectors for a data dependence relations array, and 15877 `dump_data_references' prints the details of the data references 15878 contained in a data reference array. 15879 15880 15881 File: gccint.info, Node: Lambda, Next: Omega, Prev: Dependency analysis, Up: Loop Analysis and Representation 15882 15883 14.9 Linear loop transformations framework 15884 ========================================== 15885 15886 Lambda is a framework that allows transformations of loops using 15887 non-singular matrix based transformations of the iteration space and 15888 loop bounds. This allows compositions of skewing, scaling, interchange, 15889 and reversal transformations. These transformations are often used to 15890 improve cache behavior or remove inner loop dependencies to allow 15891 parallelization and vectorization to take place. 15892 15893 To perform these transformations, Lambda requires that the loopnest be 15894 converted into a internal form that can be matrix transformed easily. 15895 To do this conversion, the function `gcc_loopnest_to_lambda_loopnest' 15896 is provided. If the loop cannot be transformed using lambda, this 15897 function will return NULL. 15898 15899 Once a `lambda_loopnest' is obtained from the conversion function, it 15900 can be transformed by using `lambda_loopnest_transform', which takes a 15901 transformation matrix to apply. Note that it is up to the caller to 15902 verify that the transformation matrix is legal to apply to the loop 15903 (dependence respecting, etc). Lambda simply applies whatever matrix it 15904 is told to provide. It can be extended to make legal matrices out of 15905 any non-singular matrix, but this is not currently implemented. 15906 Legality of a matrix for a given loopnest can be verified using 15907 `lambda_transform_legal_p'. 15908 15909 Given a transformed loopnest, conversion back into gcc IR is done by 15910 `lambda_loopnest_to_gcc_loopnest'. This function will modify the loops 15911 so that they match the transformed loopnest. 15912 15913 15914 File: gccint.info, Node: Omega, Prev: Lambda, Up: Loop Analysis and Representation 15915 15916 14.10 Omega a solver for linear programming problems 15917 ==================================================== 15918 15919 The data dependence analysis contains several solvers triggered 15920 sequentially from the less complex ones to the more sophisticated. For 15921 ensuring the consistency of the results of these solvers, a data 15922 dependence check pass has been implemented based on two different 15923 solvers. The second method that has been integrated to GCC is based on 15924 the Omega dependence solver, written in the 1990's by William Pugh and 15925 David Wonnacott. Data dependence tests can be formulated using a 15926 subset of the Presburger arithmetics that can be translated to linear 15927 constraint systems. These linear constraint systems can then be solved 15928 using the Omega solver. 15929 15930 The Omega solver is using Fourier-Motzkin's algorithm for variable 15931 elimination: a linear constraint system containing `n' variables is 15932 reduced to a linear constraint system with `n-1' variables. The Omega 15933 solver can also be used for solving other problems that can be 15934 expressed under the form of a system of linear equalities and 15935 inequalities. The Omega solver is known to have an exponential worst 15936 case, also known under the name of "omega nightmare" in the literature, 15937 but in practice, the omega test is known to be efficient for the common 15938 data dependence tests. 15939 15940 The interface used by the Omega solver for describing the linear 15941 programming problems is described in `omega.h', and the solver is 15942 `omega_solve_problem'. 15943 15944 15945 File: gccint.info, Node: Control Flow, Next: Loop Analysis and Representation, Prev: RTL, Up: Top 15946 15947 15 Control Flow Graph 15948 ********************* 15949 15950 A control flow graph (CFG) is a data structure built on top of the 15951 intermediate code representation (the RTL or `tree' instruction stream) 15952 abstracting the control flow behavior of a function that is being 15953 compiled. The CFG is a directed graph where the vertices represent 15954 basic blocks and edges represent possible transfer of control flow from 15955 one basic block to another. The data structures used to represent the 15956 control flow graph are defined in `basic-block.h'. 15957 15958 * Menu: 15959 15960 * Basic Blocks:: The definition and representation of basic blocks. 15961 * Edges:: Types of edges and their representation. 15962 * Profile information:: Representation of frequencies and probabilities. 15963 * Maintaining the CFG:: Keeping the control flow graph and up to date. 15964 * Liveness information:: Using and maintaining liveness information. 15965 15966 15967 File: gccint.info, Node: Basic Blocks, Next: Edges, Up: Control Flow 15968 15969 15.1 Basic Blocks 15970 ================= 15971 15972 A basic block is a straight-line sequence of code with only one entry 15973 point and only one exit. In GCC, basic blocks are represented using 15974 the `basic_block' data type. 15975 15976 Two pointer members of the `basic_block' structure are the pointers 15977 `next_bb' and `prev_bb'. These are used to keep doubly linked chain of 15978 basic blocks in the same order as the underlying instruction stream. 15979 The chain of basic blocks is updated transparently by the provided API 15980 for manipulating the CFG. The macro `FOR_EACH_BB' can be used to visit 15981 all the basic blocks in lexicographical order. Dominator traversals 15982 are also possible using `walk_dominator_tree'. Given two basic blocks 15983 A and B, block A dominates block B if A is _always_ executed before B. 15984 15985 The `BASIC_BLOCK' array contains all basic blocks in an unspecified 15986 order. Each `basic_block' structure has a field that holds a unique 15987 integer identifier `index' that is the index of the block in the 15988 `BASIC_BLOCK' array. The total number of basic blocks in the function 15989 is `n_basic_blocks'. Both the basic block indices and the total number 15990 of basic blocks may vary during the compilation process, as passes 15991 reorder, create, duplicate, and destroy basic blocks. The index for 15992 any block should never be greater than `last_basic_block'. 15993 15994 Special basic blocks represent possible entry and exit points of a 15995 function. These blocks are called `ENTRY_BLOCK_PTR' and 15996 `EXIT_BLOCK_PTR'. These blocks do not contain any code, and are not 15997 elements of the `BASIC_BLOCK' array. Therefore they have been assigned 15998 unique, negative index numbers. 15999 16000 Each `basic_block' also contains pointers to the first instruction 16001 (the "head") and the last instruction (the "tail") or "end" of the 16002 instruction stream contained in a basic block. In fact, since the 16003 `basic_block' data type is used to represent blocks in both major 16004 intermediate representations of GCC (`tree' and RTL), there are 16005 pointers to the head and end of a basic block for both representations. 16006 16007 For RTL, these pointers are `rtx head, end'. In the RTL function 16008 representation, the head pointer always points either to a 16009 `NOTE_INSN_BASIC_BLOCK' or to a `CODE_LABEL', if present. In the RTL 16010 representation of a function, the instruction stream contains not only 16011 the "real" instructions, but also "notes". Any function that moves or 16012 duplicates the basic blocks needs to take care of updating of these 16013 notes. Many of these notes expect that the instruction stream consists 16014 of linear regions, making such updates difficult. The 16015 `NOTE_INSN_BASIC_BLOCK' note is the only kind of note that may appear 16016 in the instruction stream contained in a basic block. The instruction 16017 stream of a basic block always follows a `NOTE_INSN_BASIC_BLOCK', but 16018 zero or more `CODE_LABEL' nodes can precede the block note. A basic 16019 block ends by control flow instruction or last instruction before 16020 following `CODE_LABEL' or `NOTE_INSN_BASIC_BLOCK'. A `CODE_LABEL' 16021 cannot appear in the instruction stream of a basic block. 16022 16023 In addition to notes, the jump table vectors are also represented as 16024 "pseudo-instructions" inside the insn stream. These vectors never 16025 appear in the basic block and should always be placed just after the 16026 table jump instructions referencing them. After removing the 16027 table-jump it is often difficult to eliminate the code computing the 16028 address and referencing the vector, so cleaning up these vectors is 16029 postponed until after liveness analysis. Thus the jump table vectors 16030 may appear in the insn stream unreferenced and without any purpose. 16031 Before any edge is made "fall-thru", the existence of such construct in 16032 the way needs to be checked by calling `can_fallthru' function. 16033 16034 For the `tree' representation, the head and end of the basic block are 16035 being pointed to by the `stmt_list' field, but this special `tree' 16036 should never be referenced directly. Instead, at the tree level 16037 abstract containers and iterators are used to access statements and 16038 expressions in basic blocks. These iterators are called "block 16039 statement iterators" (BSIs). Grep for `^bsi' in the various `tree-*' 16040 files. The following snippet will pretty-print all the statements of 16041 the program in the GIMPLE representation. 16042 16043 FOR_EACH_BB (bb) 16044 { 16045 block_stmt_iterator si; 16046 16047 for (si = bsi_start (bb); !bsi_end_p (si); bsi_next (&si)) 16048 { 16049 tree stmt = bsi_stmt (si); 16050 print_generic_stmt (stderr, stmt, 0); 16051 } 16052 } 16053 16054 16055 File: gccint.info, Node: Edges, Next: Profile information, Prev: Basic Blocks, Up: Control Flow 16056 16057 15.2 Edges 16058 ========== 16059 16060 Edges represent possible control flow transfers from the end of some 16061 basic block A to the head of another basic block B. We say that A is a 16062 predecessor of B, and B is a successor of A. Edges are represented in 16063 GCC with the `edge' data type. Each `edge' acts as a link between two 16064 basic blocks: the `src' member of an edge points to the predecessor 16065 basic block of the `dest' basic block. The members `preds' and `succs' 16066 of the `basic_block' data type point to type-safe vectors of edges to 16067 the predecessors and successors of the block. 16068 16069 When walking the edges in an edge vector, "edge iterators" should be 16070 used. Edge iterators are constructed using the `edge_iterator' data 16071 structure and several methods are available to operate on them: 16072 16073 `ei_start' 16074 This function initializes an `edge_iterator' that points to the 16075 first edge in a vector of edges. 16076 16077 `ei_last' 16078 This function initializes an `edge_iterator' that points to the 16079 last edge in a vector of edges. 16080 16081 `ei_end_p' 16082 This predicate is `true' if an `edge_iterator' represents the last 16083 edge in an edge vector. 16084 16085 `ei_one_before_end_p' 16086 This predicate is `true' if an `edge_iterator' represents the 16087 second last edge in an edge vector. 16088 16089 `ei_next' 16090 This function takes a pointer to an `edge_iterator' and makes it 16091 point to the next edge in the sequence. 16092 16093 `ei_prev' 16094 This function takes a pointer to an `edge_iterator' and makes it 16095 point to the previous edge in the sequence. 16096 16097 `ei_edge' 16098 This function returns the `edge' currently pointed to by an 16099 `edge_iterator'. 16100 16101 `ei_safe_safe' 16102 This function returns the `edge' currently pointed to by an 16103 `edge_iterator', but returns `NULL' if the iterator is pointing at 16104 the end of the sequence. This function has been provided for 16105 existing code makes the assumption that a `NULL' edge indicates 16106 the end of the sequence. 16107 16108 16109 The convenience macro `FOR_EACH_EDGE' can be used to visit all of the 16110 edges in a sequence of predecessor or successor edges. It must not be 16111 used when an element might be removed during the traversal, otherwise 16112 elements will be missed. Here is an example of how to use the macro: 16113 16114 edge e; 16115 edge_iterator ei; 16116 16117 FOR_EACH_EDGE (e, ei, bb->succs) 16118 { 16119 if (e->flags & EDGE_FALLTHRU) 16120 break; 16121 } 16122 16123 There are various reasons why control flow may transfer from one block 16124 to another. One possibility is that some instruction, for example a 16125 `CODE_LABEL', in a linearized instruction stream just always starts a 16126 new basic block. In this case a "fall-thru" edge links the basic block 16127 to the first following basic block. But there are several other 16128 reasons why edges may be created. The `flags' field of the `edge' data 16129 type is used to store information about the type of edge we are dealing 16130 with. Each edge is of one of the following types: 16131 16132 _jump_ 16133 No type flags are set for edges corresponding to jump instructions. 16134 These edges are used for unconditional or conditional jumps and in 16135 RTL also for table jumps. They are the easiest to manipulate as 16136 they may be freely redirected when the flow graph is not in SSA 16137 form. 16138 16139 _fall-thru_ 16140 Fall-thru edges are present in case where the basic block may 16141 continue execution to the following one without branching. These 16142 edges have the `EDGE_FALLTHRU' flag set. Unlike other types of 16143 edges, these edges must come into the basic block immediately 16144 following in the instruction stream. The function 16145 `force_nonfallthru' is available to insert an unconditional jump 16146 in the case that redirection is needed. Note that this may 16147 require creation of a new basic block. 16148 16149 _exception handling_ 16150 Exception handling edges represent possible control transfers from 16151 a trapping instruction to an exception handler. The definition of 16152 "trapping" varies. In C++, only function calls can throw, but for 16153 Java, exceptions like division by zero or segmentation fault are 16154 defined and thus each instruction possibly throwing this kind of 16155 exception needs to be handled as control flow instruction. 16156 Exception edges have the `EDGE_ABNORMAL' and `EDGE_EH' flags set. 16157 16158 When updating the instruction stream it is easy to change possibly 16159 trapping instruction to non-trapping, by simply removing the 16160 exception edge. The opposite conversion is difficult, but should 16161 not happen anyway. The edges can be eliminated via 16162 `purge_dead_edges' call. 16163 16164 In the RTL representation, the destination of an exception edge is 16165 specified by `REG_EH_REGION' note attached to the insn. In case 16166 of a trapping call the `EDGE_ABNORMAL_CALL' flag is set too. In 16167 the `tree' representation, this extra flag is not set. 16168 16169 In the RTL representation, the predicate `may_trap_p' may be used 16170 to check whether instruction still may trap or not. For the tree 16171 representation, the `tree_could_trap_p' predicate is available, 16172 but this predicate only checks for possible memory traps, as in 16173 dereferencing an invalid pointer location. 16174 16175 _sibling calls_ 16176 Sibling calls or tail calls terminate the function in a 16177 non-standard way and thus an edge to the exit must be present. 16178 `EDGE_SIBCALL' and `EDGE_ABNORMAL' are set in such case. These 16179 edges only exist in the RTL representation. 16180 16181 _computed jumps_ 16182 Computed jumps contain edges to all labels in the function 16183 referenced from the code. All those edges have `EDGE_ABNORMAL' 16184 flag set. The edges used to represent computed jumps often cause 16185 compile time performance problems, since functions consisting of 16186 many taken labels and many computed jumps may have _very_ dense 16187 flow graphs, so these edges need to be handled with special care. 16188 During the earlier stages of the compilation process, GCC tries to 16189 avoid such dense flow graphs by factoring computed jumps. For 16190 example, given the following series of jumps, 16191 16192 goto *x; 16193 [ ... ] 16194 16195 goto *x; 16196 [ ... ] 16197 16198 goto *x; 16199 [ ... ] 16200 16201 factoring the computed jumps results in the following code sequence 16202 which has a much simpler flow graph: 16203 16204 goto y; 16205 [ ... ] 16206 16207 goto y; 16208 [ ... ] 16209 16210 goto y; 16211 [ ... ] 16212 16213 y: 16214 goto *x; 16215 16216 However, the classic problem with this transformation is that it 16217 has a runtime cost in there resulting code: An extra jump. 16218 Therefore, the computed jumps are un-factored in the later passes 16219 of the compiler. Be aware of that when you work on passes in that 16220 area. There have been numerous examples already where the compile 16221 time for code with unfactored computed jumps caused some serious 16222 headaches. 16223 16224 _nonlocal goto handlers_ 16225 GCC allows nested functions to return into caller using a `goto' 16226 to a label passed to as an argument to the callee. The labels 16227 passed to nested functions contain special code to cleanup after 16228 function call. Such sections of code are referred to as "nonlocal 16229 goto receivers". If a function contains such nonlocal goto 16230 receivers, an edge from the call to the label is created with the 16231 `EDGE_ABNORMAL' and `EDGE_ABNORMAL_CALL' flags set. 16232 16233 _function entry points_ 16234 By definition, execution of function starts at basic block 0, so 16235 there is always an edge from the `ENTRY_BLOCK_PTR' to basic block 16236 0. There is no `tree' representation for alternate entry points at 16237 this moment. In RTL, alternate entry points are specified by 16238 `CODE_LABEL' with `LABEL_ALTERNATE_NAME' defined. This feature is 16239 currently used for multiple entry point prologues and is limited 16240 to post-reload passes only. This can be used by back-ends to emit 16241 alternate prologues for functions called from different contexts. 16242 In future full support for multiple entry functions defined by 16243 Fortran 90 needs to be implemented. 16244 16245 _function exits_ 16246 In the pre-reload representation a function terminates after the 16247 last instruction in the insn chain and no explicit return 16248 instructions are used. This corresponds to the fall-thru edge 16249 into exit block. After reload, optimal RTL epilogues are used 16250 that use explicit (conditional) return instructions that are 16251 represented by edges with no flags set. 16252 16253 16254 16255 File: gccint.info, Node: Profile information, Next: Maintaining the CFG, Prev: Edges, Up: Control Flow 16256 16257 15.3 Profile information 16258 ======================== 16259 16260 In many cases a compiler must make a choice whether to trade speed in 16261 one part of code for speed in another, or to trade code size for code 16262 speed. In such cases it is useful to know information about how often 16263 some given block will be executed. That is the purpose for maintaining 16264 profile within the flow graph. GCC can handle profile information 16265 obtained through "profile feedback", but it can also estimate branch 16266 probabilities based on statics and heuristics. 16267 16268 The feedback based profile is produced by compiling the program with 16269 instrumentation, executing it on a train run and reading the numbers of 16270 executions of basic blocks and edges back to the compiler while 16271 re-compiling the program to produce the final executable. This method 16272 provides very accurate information about where a program spends most of 16273 its time on the train run. Whether it matches the average run of 16274 course depends on the choice of train data set, but several studies 16275 have shown that the behavior of a program usually changes just 16276 marginally over different data sets. 16277 16278 When profile feedback is not available, the compiler may be asked to 16279 attempt to predict the behavior of each branch in the program using a 16280 set of heuristics (see `predict.def' for details) and compute estimated 16281 frequencies of each basic block by propagating the probabilities over 16282 the graph. 16283 16284 Each `basic_block' contains two integer fields to represent profile 16285 information: `frequency' and `count'. The `frequency' is an estimation 16286 how often is basic block executed within a function. It is represented 16287 as an integer scaled in the range from 0 to `BB_FREQ_BASE'. The most 16288 frequently executed basic block in function is initially set to 16289 `BB_FREQ_BASE' and the rest of frequencies are scaled accordingly. 16290 During optimization, the frequency of the most frequent basic block can 16291 both decrease (for instance by loop unrolling) or grow (for instance by 16292 cross-jumping optimization), so scaling sometimes has to be performed 16293 multiple times. 16294 16295 The `count' contains hard-counted numbers of execution measured during 16296 training runs and is nonzero only when profile feedback is available. 16297 This value is represented as the host's widest integer (typically a 64 16298 bit integer) of the special type `gcov_type'. 16299 16300 Most optimization passes can use only the frequency information of a 16301 basic block, but a few passes may want to know hard execution counts. 16302 The frequencies should always match the counts after scaling, however 16303 during updating of the profile information numerical error may 16304 accumulate into quite large errors. 16305 16306 Each edge also contains a branch probability field: an integer in the 16307 range from 0 to `REG_BR_PROB_BASE'. It represents probability of 16308 passing control from the end of the `src' basic block to the `dest' 16309 basic block, i.e. the probability that control will flow along this 16310 edge. The `EDGE_FREQUENCY' macro is available to compute how 16311 frequently a given edge is taken. There is a `count' field for each 16312 edge as well, representing same information as for a basic block. 16313 16314 The basic block frequencies are not represented in the instruction 16315 stream, but in the RTL representation the edge frequencies are 16316 represented for conditional jumps (via the `REG_BR_PROB' macro) since 16317 they are used when instructions are output to the assembly file and the 16318 flow graph is no longer maintained. 16319 16320 The probability that control flow arrives via a given edge to its 16321 destination basic block is called "reverse probability" and is not 16322 directly represented, but it may be easily computed from frequencies of 16323 basic blocks. 16324 16325 Updating profile information is a delicate task that can unfortunately 16326 not be easily integrated with the CFG manipulation API. Many of the 16327 functions and hooks to modify the CFG, such as 16328 `redirect_edge_and_branch', do not have enough information to easily 16329 update the profile, so updating it is in the majority of cases left up 16330 to the caller. It is difficult to uncover bugs in the profile updating 16331 code, because they manifest themselves only by producing worse code, 16332 and checking profile consistency is not possible because of numeric 16333 error accumulation. Hence special attention needs to be given to this 16334 issue in each pass that modifies the CFG. 16335 16336 It is important to point out that `REG_BR_PROB_BASE' and 16337 `BB_FREQ_BASE' are both set low enough to be possible to compute second 16338 power of any frequency or probability in the flow graph, it is not 16339 possible to even square the `count' field, as modern CPUs are fast 16340 enough to execute $2^32$ operations quickly. 16341 16342 16343 File: gccint.info, Node: Maintaining the CFG, Next: Liveness information, Prev: Profile information, Up: Control Flow 16344 16345 15.4 Maintaining the CFG 16346 ======================== 16347 16348 An important task of each compiler pass is to keep both the control 16349 flow graph and all profile information up-to-date. Reconstruction of 16350 the control flow graph after each pass is not an option, since it may be 16351 very expensive and lost profile information cannot be reconstructed at 16352 all. 16353 16354 GCC has two major intermediate representations, and both use the 16355 `basic_block' and `edge' data types to represent control flow. Both 16356 representations share as much of the CFG maintenance code as possible. 16357 For each representation, a set of "hooks" is defined so that each 16358 representation can provide its own implementation of CFG manipulation 16359 routines when necessary. These hooks are defined in `cfghooks.h'. 16360 There are hooks for almost all common CFG manipulations, including 16361 block splitting and merging, edge redirection and creating and deleting 16362 basic blocks. These hooks should provide everything you need to 16363 maintain and manipulate the CFG in both the RTL and `tree' 16364 representation. 16365 16366 At the moment, the basic block boundaries are maintained transparently 16367 when modifying instructions, so there rarely is a need to move them 16368 manually (such as in case someone wants to output instruction outside 16369 basic block explicitly). Often the CFG may be better viewed as 16370 integral part of instruction chain, than structure built on the top of 16371 it. However, in principle the control flow graph for the `tree' 16372 representation is _not_ an integral part of the representation, in that 16373 a function tree may be expanded without first building a flow graph 16374 for the `tree' representation at all. This happens when compiling 16375 without any `tree' optimization enabled. When the `tree' optimizations 16376 are enabled and the instruction stream is rewritten in SSA form, the 16377 CFG is very tightly coupled with the instruction stream. In 16378 particular, statement insertion and removal has to be done with care. 16379 In fact, the whole `tree' representation can not be easily used or 16380 maintained without proper maintenance of the CFG simultaneously. 16381 16382 In the RTL representation, each instruction has a `BLOCK_FOR_INSN' 16383 value that represents pointer to the basic block that contains the 16384 instruction. In the `tree' representation, the function `bb_for_stmt' 16385 returns a pointer to the basic block containing the queried statement. 16386 16387 When changes need to be applied to a function in its `tree' 16388 representation, "block statement iterators" should be used. These 16389 iterators provide an integrated abstraction of the flow graph and the 16390 instruction stream. Block statement iterators are constructed using 16391 the `block_stmt_iterator' data structure and several modifier are 16392 available, including the following: 16393 16394 `bsi_start' 16395 This function initializes a `block_stmt_iterator' that points to 16396 the first non-empty statement in a basic block. 16397 16398 `bsi_last' 16399 This function initializes a `block_stmt_iterator' that points to 16400 the last statement in a basic block. 16401 16402 `bsi_end_p' 16403 This predicate is `true' if a `block_stmt_iterator' represents the 16404 end of a basic block. 16405 16406 `bsi_next' 16407 This function takes a `block_stmt_iterator' and makes it point to 16408 its successor. 16409 16410 `bsi_prev' 16411 This function takes a `block_stmt_iterator' and makes it point to 16412 its predecessor. 16413 16414 `bsi_insert_after' 16415 This function inserts a statement after the `block_stmt_iterator' 16416 passed in. The final parameter determines whether the statement 16417 iterator is updated to point to the newly inserted statement, or 16418 left pointing to the original statement. 16419 16420 `bsi_insert_before' 16421 This function inserts a statement before the `block_stmt_iterator' 16422 passed in. The final parameter determines whether the statement 16423 iterator is updated to point to the newly inserted statement, or 16424 left pointing to the original statement. 16425 16426 `bsi_remove' 16427 This function removes the `block_stmt_iterator' passed in and 16428 rechains the remaining statements in a basic block, if any. 16429 16430 In the RTL representation, the macros `BB_HEAD' and `BB_END' may be 16431 used to get the head and end `rtx' of a basic block. No abstract 16432 iterators are defined for traversing the insn chain, but you can just 16433 use `NEXT_INSN' and `PREV_INSN' instead. See *Note Insns::. 16434 16435 Usually a code manipulating pass simplifies the instruction stream and 16436 the flow of control, possibly eliminating some edges. This may for 16437 example happen when a conditional jump is replaced with an 16438 unconditional jump, but also when simplifying possibly trapping 16439 instruction to non-trapping while compiling Java. Updating of edges is 16440 not transparent and each optimization pass is required to do so 16441 manually. However only few cases occur in practice. The pass may call 16442 `purge_dead_edges' on a given basic block to remove superfluous edges, 16443 if any. 16444 16445 Another common scenario is redirection of branch instructions, but 16446 this is best modeled as redirection of edges in the control flow graph 16447 and thus use of `redirect_edge_and_branch' is preferred over more low 16448 level functions, such as `redirect_jump' that operate on RTL chain 16449 only. The CFG hooks defined in `cfghooks.h' should provide the 16450 complete API required for manipulating and maintaining the CFG. 16451 16452 It is also possible that a pass has to insert control flow instruction 16453 into the middle of a basic block, thus creating an entry point in the 16454 middle of the basic block, which is impossible by definition: The block 16455 must be split to make sure it only has one entry point, i.e. the head 16456 of the basic block. The CFG hook `split_block' may be used when an 16457 instruction in the middle of a basic block has to become the target of 16458 a jump or branch instruction. 16459 16460 For a global optimizer, a common operation is to split edges in the 16461 flow graph and insert instructions on them. In the RTL representation, 16462 this can be easily done using the `insert_insn_on_edge' function that 16463 emits an instruction "on the edge", caching it for a later 16464 `commit_edge_insertions' call that will take care of moving the 16465 inserted instructions off the edge into the instruction stream 16466 contained in a basic block. This includes the creation of new basic 16467 blocks where needed. In the `tree' representation, the equivalent 16468 functions are `bsi_insert_on_edge' which inserts a block statement 16469 iterator on an edge, and `bsi_commit_edge_inserts' which flushes the 16470 instruction to actual instruction stream. 16471 16472 While debugging the optimization pass, an `verify_flow_info' function 16473 may be useful to find bugs in the control flow graph updating code. 16474 16475 Note that at present, the representation of control flow in the `tree' 16476 representation is discarded before expanding to RTL. Long term the CFG 16477 should be maintained and "expanded" to the RTL representation along 16478 with the function `tree' itself. 16479 16480 16481 File: gccint.info, Node: Liveness information, Prev: Maintaining the CFG, Up: Control Flow 16482 16483 15.5 Liveness information 16484 ========================= 16485 16486 Liveness information is useful to determine whether some register is 16487 "live" at given point of program, i.e. that it contains a value that 16488 may be used at a later point in the program. This information is used, 16489 for instance, during register allocation, as the pseudo registers only 16490 need to be assigned to a unique hard register or to a stack slot if 16491 they are live. The hard registers and stack slots may be freely reused 16492 for other values when a register is dead. 16493 16494 Liveness information is available in the back end starting with 16495 `pass_df_initialize' and ending with `pass_df_finish'. Three flavors 16496 of live analysis are available: With `LR', it is possible to determine 16497 at any point `P' in the function if the register may be used on some 16498 path from `P' to the end of the function. With `UR', it is possible to 16499 determine if there is a path from the beginning of the function to `P' 16500 that defines the variable. `LIVE' is the intersection of the `LR' and 16501 `UR' and a variable is live at `P' if there is both an assignment that 16502 reaches it from the beginning of the function and a uses that can be 16503 reached on some path from `P' to the end of the function. 16504 16505 In general `LIVE' is the most useful of the three. The macros 16506 `DF_[LR,UR,LIVE]_[IN,OUT]' can be used to access this information. The 16507 macros take a basic block number and return a bitmap that is indexed by 16508 the register number. This information is only guaranteed to be up to 16509 date after calls are made to `df_analyze'. See the file `df-core.c' 16510 for details on using the dataflow. 16511 16512 The liveness information is stored partly in the RTL instruction stream 16513 and partly in the flow graph. Local information is stored in the 16514 instruction stream: Each instruction may contain `REG_DEAD' notes 16515 representing that the value of a given register is no longer needed, or 16516 `REG_UNUSED' notes representing that the value computed by the 16517 instruction is never used. The second is useful for instructions 16518 computing multiple values at once. 16519 16520 16521 File: gccint.info, Node: Machine Desc, Next: Target Macros, Prev: Loop Analysis and Representation, Up: Top 16522 16523 16 Machine Descriptions 16524 *********************** 16525 16526 A machine description has two parts: a file of instruction patterns 16527 (`.md' file) and a C header file of macro definitions. 16528 16529 The `.md' file for a target machine contains a pattern for each 16530 instruction that the target machine supports (or at least each 16531 instruction that is worth telling the compiler about). It may also 16532 contain comments. A semicolon causes the rest of the line to be a 16533 comment, unless the semicolon is inside a quoted string. 16534 16535 See the next chapter for information on the C header file. 16536 16537 * Menu: 16538 16539 * Overview:: How the machine description is used. 16540 * Patterns:: How to write instruction patterns. 16541 * Example:: An explained example of a `define_insn' pattern. 16542 * RTL Template:: The RTL template defines what insns match a pattern. 16543 * Output Template:: The output template says how to make assembler code 16544 from such an insn. 16545 * Output Statement:: For more generality, write C code to output 16546 the assembler code. 16547 * Predicates:: Controlling what kinds of operands can be used 16548 for an insn. 16549 * Constraints:: Fine-tuning operand selection. 16550 * Standard Names:: Names mark patterns to use for code generation. 16551 * Pattern Ordering:: When the order of patterns makes a difference. 16552 * Dependent Patterns:: Having one pattern may make you need another. 16553 * Jump Patterns:: Special considerations for patterns for jump insns. 16554 * Looping Patterns:: How to define patterns for special looping insns. 16555 * Insn Canonicalizations::Canonicalization of Instructions 16556 * Expander Definitions::Generating a sequence of several RTL insns 16557 for a standard operation. 16558 * Insn Splitting:: Splitting Instructions into Multiple Instructions. 16559 * Including Patterns:: Including Patterns in Machine Descriptions. 16560 * Peephole Definitions::Defining machine-specific peephole optimizations. 16561 * Insn Attributes:: Specifying the value of attributes for generated insns. 16562 * Conditional Execution::Generating `define_insn' patterns for 16563 predication. 16564 * Constant Definitions::Defining symbolic constants that can be used in the 16565 md file. 16566 * Iterators:: Using iterators to generate patterns from a template. 16567 16568 16569 File: gccint.info, Node: Overview, Next: Patterns, Up: Machine Desc 16570 16571 16.1 Overview of How the Machine Description is Used 16572 ==================================================== 16573 16574 There are three main conversions that happen in the compiler: 16575 16576 1. The front end reads the source code and builds a parse tree. 16577 16578 2. The parse tree is used to generate an RTL insn list based on named 16579 instruction patterns. 16580 16581 3. The insn list is matched against the RTL templates to produce 16582 assembler code. 16583 16584 16585 For the generate pass, only the names of the insns matter, from either 16586 a named `define_insn' or a `define_expand'. The compiler will choose 16587 the pattern with the right name and apply the operands according to the 16588 documentation later in this chapter, without regard for the RTL 16589 template or operand constraints. Note that the names the compiler looks 16590 for are hard-coded in the compiler--it will ignore unnamed patterns and 16591 patterns with names it doesn't know about, but if you don't provide a 16592 named pattern it needs, it will abort. 16593 16594 If a `define_insn' is used, the template given is inserted into the 16595 insn list. If a `define_expand' is used, one of three things happens, 16596 based on the condition logic. The condition logic may manually create 16597 new insns for the insn list, say via `emit_insn()', and invoke `DONE'. 16598 For certain named patterns, it may invoke `FAIL' to tell the compiler 16599 to use an alternate way of performing that task. If it invokes neither 16600 `DONE' nor `FAIL', the template given in the pattern is inserted, as if 16601 the `define_expand' were a `define_insn'. 16602 16603 Once the insn list is generated, various optimization passes convert, 16604 replace, and rearrange the insns in the insn list. This is where the 16605 `define_split' and `define_peephole' patterns get used, for example. 16606 16607 Finally, the insn list's RTL is matched up with the RTL templates in 16608 the `define_insn' patterns, and those patterns are used to emit the 16609 final assembly code. For this purpose, each named `define_insn' acts 16610 like it's unnamed, since the names are ignored. 16611 16612 16613 File: gccint.info, Node: Patterns, Next: Example, Prev: Overview, Up: Machine Desc 16614 16615 16.2 Everything about Instruction Patterns 16616 ========================================== 16617 16618 Each instruction pattern contains an incomplete RTL expression, with 16619 pieces to be filled in later, operand constraints that restrict how the 16620 pieces can be filled in, and an output pattern or C code to generate 16621 the assembler output, all wrapped up in a `define_insn' expression. 16622 16623 A `define_insn' is an RTL expression containing four or five operands: 16624 16625 1. An optional name. The presence of a name indicate that this 16626 instruction pattern can perform a certain standard job for the 16627 RTL-generation pass of the compiler. This pass knows certain 16628 names and will use the instruction patterns with those names, if 16629 the names are defined in the machine description. 16630 16631 The absence of a name is indicated by writing an empty string 16632 where the name should go. Nameless instruction patterns are never 16633 used for generating RTL code, but they may permit several simpler 16634 insns to be combined later on. 16635 16636 Names that are not thus known and used in RTL-generation have no 16637 effect; they are equivalent to no name at all. 16638 16639 For the purpose of debugging the compiler, you may also specify a 16640 name beginning with the `*' character. Such a name is used only 16641 for identifying the instruction in RTL dumps; it is entirely 16642 equivalent to having a nameless pattern for all other purposes. 16643 16644 2. The "RTL template" (*note RTL Template::) is a vector of incomplete 16645 RTL expressions which show what the instruction should look like. 16646 It is incomplete because it may contain `match_operand', 16647 `match_operator', and `match_dup' expressions that stand for 16648 operands of the instruction. 16649 16650 If the vector has only one element, that element is the template 16651 for the instruction pattern. If the vector has multiple elements, 16652 then the instruction pattern is a `parallel' expression containing 16653 the elements described. 16654 16655 3. A condition. This is a string which contains a C expression that 16656 is the final test to decide whether an insn body matches this 16657 pattern. 16658 16659 For a named pattern, the condition (if present) may not depend on 16660 the data in the insn being matched, but only the 16661 target-machine-type flags. The compiler needs to test these 16662 conditions during initialization in order to learn exactly which 16663 named instructions are available in a particular run. 16664 16665 For nameless patterns, the condition is applied only when matching 16666 an individual insn, and only after the insn has matched the 16667 pattern's recognition template. The insn's operands may be found 16668 in the vector `operands'. For an insn where the condition has 16669 once matched, it can't be used to control register allocation, for 16670 example by excluding certain hard registers or hard register 16671 combinations. 16672 16673 4. The "output template": a string that says how to output matching 16674 insns as assembler code. `%' in this string specifies where to 16675 substitute the value of an operand. *Note Output Template::. 16676 16677 When simple substitution isn't general enough, you can specify a 16678 piece of C code to compute the output. *Note Output Statement::. 16679 16680 5. Optionally, a vector containing the values of attributes for insns 16681 matching this pattern. *Note Insn Attributes::. 16682 16683 16684 File: gccint.info, Node: Example, Next: RTL Template, Prev: Patterns, Up: Machine Desc 16685 16686 16.3 Example of `define_insn' 16687 ============================= 16688 16689 Here is an actual example of an instruction pattern, for the 16690 68000/68020. 16691 16692 (define_insn "tstsi" 16693 [(set (cc0) 16694 (match_operand:SI 0 "general_operand" "rm"))] 16695 "" 16696 "* 16697 { 16698 if (TARGET_68020 || ! ADDRESS_REG_P (operands[0])) 16699 return \"tstl %0\"; 16700 return \"cmpl #0,%0\"; 16701 }") 16702 16703 This can also be written using braced strings: 16704 16705 (define_insn "tstsi" 16706 [(set (cc0) 16707 (match_operand:SI 0 "general_operand" "rm"))] 16708 "" 16709 { 16710 if (TARGET_68020 || ! ADDRESS_REG_P (operands[0])) 16711 return "tstl %0"; 16712 return "cmpl #0,%0"; 16713 }) 16714 16715 This is an instruction that sets the condition codes based on the 16716 value of a general operand. It has no condition, so any insn whose RTL 16717 description has the form shown may be handled according to this 16718 pattern. The name `tstsi' means "test a `SImode' value" and tells the 16719 RTL generation pass that, when it is necessary to test such a value, an 16720 insn to do so can be constructed using this pattern. 16721 16722 The output control string is a piece of C code which chooses which 16723 output template to return based on the kind of operand and the specific 16724 type of CPU for which code is being generated. 16725 16726 `"rm"' is an operand constraint. Its meaning is explained below. 16727 16728 16729 File: gccint.info, Node: RTL Template, Next: Output Template, Prev: Example, Up: Machine Desc 16730 16731 16.4 RTL Template 16732 ================= 16733 16734 The RTL template is used to define which insns match the particular 16735 pattern and how to find their operands. For named patterns, the RTL 16736 template also says how to construct an insn from specified operands. 16737 16738 Construction involves substituting specified operands into a copy of 16739 the template. Matching involves determining the values that serve as 16740 the operands in the insn being matched. Both of these activities are 16741 controlled by special expression types that direct matching and 16742 substitution of the operands. 16743 16744 `(match_operand:M N PREDICATE CONSTRAINT)' 16745 This expression is a placeholder for operand number N of the insn. 16746 When constructing an insn, operand number N will be substituted at 16747 this point. When matching an insn, whatever appears at this 16748 position in the insn will be taken as operand number N; but it 16749 must satisfy PREDICATE or this instruction pattern will not match 16750 at all. 16751 16752 Operand numbers must be chosen consecutively counting from zero in 16753 each instruction pattern. There may be only one `match_operand' 16754 expression in the pattern for each operand number. Usually 16755 operands are numbered in the order of appearance in `match_operand' 16756 expressions. In the case of a `define_expand', any operand numbers 16757 used only in `match_dup' expressions have higher values than all 16758 other operand numbers. 16759 16760 PREDICATE is a string that is the name of a function that accepts 16761 two arguments, an expression and a machine mode. *Note 16762 Predicates::. During matching, the function will be called with 16763 the putative operand as the expression and M as the mode argument 16764 (if M is not specified, `VOIDmode' will be used, which normally 16765 causes PREDICATE to accept any mode). If it returns zero, this 16766 instruction pattern fails to match. PREDICATE may be an empty 16767 string; then it means no test is to be done on the operand, so 16768 anything which occurs in this position is valid. 16769 16770 Most of the time, PREDICATE will reject modes other than M--but 16771 not always. For example, the predicate `address_operand' uses M 16772 as the mode of memory ref that the address should be valid for. 16773 Many predicates accept `const_int' nodes even though their mode is 16774 `VOIDmode'. 16775 16776 CONSTRAINT controls reloading and the choice of the best register 16777 class to use for a value, as explained later (*note Constraints::). 16778 If the constraint would be an empty string, it can be omitted. 16779 16780 People are often unclear on the difference between the constraint 16781 and the predicate. The predicate helps decide whether a given 16782 insn matches the pattern. The constraint plays no role in this 16783 decision; instead, it controls various decisions in the case of an 16784 insn which does match. 16785 16786 `(match_scratch:M N CONSTRAINT)' 16787 This expression is also a placeholder for operand number N and 16788 indicates that operand must be a `scratch' or `reg' expression. 16789 16790 When matching patterns, this is equivalent to 16791 16792 (match_operand:M N "scratch_operand" PRED) 16793 16794 but, when generating RTL, it produces a (`scratch':M) expression. 16795 16796 If the last few expressions in a `parallel' are `clobber' 16797 expressions whose operands are either a hard register or 16798 `match_scratch', the combiner can add or delete them when 16799 necessary. *Note Side Effects::. 16800 16801 `(match_dup N)' 16802 This expression is also a placeholder for operand number N. It is 16803 used when the operand needs to appear more than once in the insn. 16804 16805 In construction, `match_dup' acts just like `match_operand': the 16806 operand is substituted into the insn being constructed. But in 16807 matching, `match_dup' behaves differently. It assumes that operand 16808 number N has already been determined by a `match_operand' 16809 appearing earlier in the recognition template, and it matches only 16810 an identical-looking expression. 16811 16812 Note that `match_dup' should not be used to tell the compiler that 16813 a particular register is being used for two operands (example: 16814 `add' that adds one register to another; the second register is 16815 both an input operand and the output operand). Use a matching 16816 constraint (*note Simple Constraints::) for those. `match_dup' is 16817 for the cases where one operand is used in two places in the 16818 template, such as an instruction that computes both a quotient and 16819 a remainder, where the opcode takes two input operands but the RTL 16820 template has to refer to each of those twice; once for the 16821 quotient pattern and once for the remainder pattern. 16822 16823 `(match_operator:M N PREDICATE [OPERANDS...])' 16824 This pattern is a kind of placeholder for a variable RTL expression 16825 code. 16826 16827 When constructing an insn, it stands for an RTL expression whose 16828 expression code is taken from that of operand N, and whose 16829 operands are constructed from the patterns OPERANDS. 16830 16831 When matching an expression, it matches an expression if the 16832 function PREDICATE returns nonzero on that expression _and_ the 16833 patterns OPERANDS match the operands of the expression. 16834 16835 Suppose that the function `commutative_operator' is defined as 16836 follows, to match any expression whose operator is one of the 16837 commutative arithmetic operators of RTL and whose mode is MODE: 16838 16839 int 16840 commutative_integer_operator (x, mode) 16841 rtx x; 16842 enum machine_mode mode; 16843 { 16844 enum rtx_code code = GET_CODE (x); 16845 if (GET_MODE (x) != mode) 16846 return 0; 16847 return (GET_RTX_CLASS (code) == RTX_COMM_ARITH 16848 || code == EQ || code == NE); 16849 } 16850 16851 Then the following pattern will match any RTL expression consisting 16852 of a commutative operator applied to two general operands: 16853 16854 (match_operator:SI 3 "commutative_operator" 16855 [(match_operand:SI 1 "general_operand" "g") 16856 (match_operand:SI 2 "general_operand" "g")]) 16857 16858 Here the vector `[OPERANDS...]' contains two patterns because the 16859 expressions to be matched all contain two operands. 16860 16861 When this pattern does match, the two operands of the commutative 16862 operator are recorded as operands 1 and 2 of the insn. (This is 16863 done by the two instances of `match_operand'.) Operand 3 of the 16864 insn will be the entire commutative expression: use `GET_CODE 16865 (operands[3])' to see which commutative operator was used. 16866 16867 The machine mode M of `match_operator' works like that of 16868 `match_operand': it is passed as the second argument to the 16869 predicate function, and that function is solely responsible for 16870 deciding whether the expression to be matched "has" that mode. 16871 16872 When constructing an insn, argument 3 of the gen-function will 16873 specify the operation (i.e. the expression code) for the 16874 expression to be made. It should be an RTL expression, whose 16875 expression code is copied into a new expression whose operands are 16876 arguments 1 and 2 of the gen-function. The subexpressions of 16877 argument 3 are not used; only its expression code matters. 16878 16879 When `match_operator' is used in a pattern for matching an insn, 16880 it usually best if the operand number of the `match_operator' is 16881 higher than that of the actual operands of the insn. This improves 16882 register allocation because the register allocator often looks at 16883 operands 1 and 2 of insns to see if it can do register tying. 16884 16885 There is no way to specify constraints in `match_operator'. The 16886 operand of the insn which corresponds to the `match_operator' 16887 never has any constraints because it is never reloaded as a whole. 16888 However, if parts of its OPERANDS are matched by `match_operand' 16889 patterns, those parts may have constraints of their own. 16890 16891 `(match_op_dup:M N[OPERANDS...])' 16892 Like `match_dup', except that it applies to operators instead of 16893 operands. When constructing an insn, operand number N will be 16894 substituted at this point. But in matching, `match_op_dup' behaves 16895 differently. It assumes that operand number N has already been 16896 determined by a `match_operator' appearing earlier in the 16897 recognition template, and it matches only an identical-looking 16898 expression. 16899 16900 `(match_parallel N PREDICATE [SUBPAT...])' 16901 This pattern is a placeholder for an insn that consists of a 16902 `parallel' expression with a variable number of elements. This 16903 expression should only appear at the top level of an insn pattern. 16904 16905 When constructing an insn, operand number N will be substituted at 16906 this point. When matching an insn, it matches if the body of the 16907 insn is a `parallel' expression with at least as many elements as 16908 the vector of SUBPAT expressions in the `match_parallel', if each 16909 SUBPAT matches the corresponding element of the `parallel', _and_ 16910 the function PREDICATE returns nonzero on the `parallel' that is 16911 the body of the insn. It is the responsibility of the predicate 16912 to validate elements of the `parallel' beyond those listed in the 16913 `match_parallel'. 16914 16915 A typical use of `match_parallel' is to match load and store 16916 multiple expressions, which can contain a variable number of 16917 elements in a `parallel'. For example, 16918 16919 (define_insn "" 16920 [(match_parallel 0 "load_multiple_operation" 16921 [(set (match_operand:SI 1 "gpc_reg_operand" "=r") 16922 (match_operand:SI 2 "memory_operand" "m")) 16923 (use (reg:SI 179)) 16924 (clobber (reg:SI 179))])] 16925 "" 16926 "loadm 0,0,%1,%2") 16927 16928 This example comes from `a29k.md'. The function 16929 `load_multiple_operation' is defined in `a29k.c' and checks that 16930 subsequent elements in the `parallel' are the same as the `set' in 16931 the pattern, except that they are referencing subsequent registers 16932 and memory locations. 16933 16934 An insn that matches this pattern might look like: 16935 16936 (parallel 16937 [(set (reg:SI 20) (mem:SI (reg:SI 100))) 16938 (use (reg:SI 179)) 16939 (clobber (reg:SI 179)) 16940 (set (reg:SI 21) 16941 (mem:SI (plus:SI (reg:SI 100) 16942 (const_int 4)))) 16943 (set (reg:SI 22) 16944 (mem:SI (plus:SI (reg:SI 100) 16945 (const_int 8))))]) 16946 16947 `(match_par_dup N [SUBPAT...])' 16948 Like `match_op_dup', but for `match_parallel' instead of 16949 `match_operator'. 16950 16951 16952 16953 File: gccint.info, Node: Output Template, Next: Output Statement, Prev: RTL Template, Up: Machine Desc 16954 16955 16.5 Output Templates and Operand Substitution 16956 ============================================== 16957 16958 The "output template" is a string which specifies how to output the 16959 assembler code for an instruction pattern. Most of the template is a 16960 fixed string which is output literally. The character `%' is used to 16961 specify where to substitute an operand; it can also be used to identify 16962 places where different variants of the assembler require different 16963 syntax. 16964 16965 In the simplest case, a `%' followed by a digit N says to output 16966 operand N at that point in the string. 16967 16968 `%' followed by a letter and a digit says to output an operand in an 16969 alternate fashion. Four letters have standard, built-in meanings 16970 described below. The machine description macro `PRINT_OPERAND' can 16971 define additional letters with nonstandard meanings. 16972 16973 `%cDIGIT' can be used to substitute an operand that is a constant 16974 value without the syntax that normally indicates an immediate operand. 16975 16976 `%nDIGIT' is like `%cDIGIT' except that the value of the constant is 16977 negated before printing. 16978 16979 `%aDIGIT' can be used to substitute an operand as if it were a memory 16980 reference, with the actual operand treated as the address. This may be 16981 useful when outputting a "load address" instruction, because often the 16982 assembler syntax for such an instruction requires you to write the 16983 operand as if it were a memory reference. 16984 16985 `%lDIGIT' is used to substitute a `label_ref' into a jump instruction. 16986 16987 `%=' outputs a number which is unique to each instruction in the 16988 entire compilation. This is useful for making local labels to be 16989 referred to more than once in a single template that generates multiple 16990 assembler instructions. 16991 16992 `%' followed by a punctuation character specifies a substitution that 16993 does not use an operand. Only one case is standard: `%%' outputs a `%' 16994 into the assembler code. Other nonstandard cases can be defined in the 16995 `PRINT_OPERAND' macro. You must also define which punctuation 16996 characters are valid with the `PRINT_OPERAND_PUNCT_VALID_P' macro. 16997 16998 The template may generate multiple assembler instructions. Write the 16999 text for the instructions, with `\;' between them. 17000 17001 When the RTL contains two operands which are required by constraint to 17002 match each other, the output template must refer only to the 17003 lower-numbered operand. Matching operands are not always identical, 17004 and the rest of the compiler arranges to put the proper RTL expression 17005 for printing into the lower-numbered operand. 17006 17007 One use of nonstandard letters or punctuation following `%' is to 17008 distinguish between different assembler languages for the same machine; 17009 for example, Motorola syntax versus MIT syntax for the 68000. Motorola 17010 syntax requires periods in most opcode names, while MIT syntax does 17011 not. For example, the opcode `movel' in MIT syntax is `move.l' in 17012 Motorola syntax. The same file of patterns is used for both kinds of 17013 output syntax, but the character sequence `%.' is used in each place 17014 where Motorola syntax wants a period. The `PRINT_OPERAND' macro for 17015 Motorola syntax defines the sequence to output a period; the macro for 17016 MIT syntax defines it to do nothing. 17017 17018 As a special case, a template consisting of the single character `#' 17019 instructs the compiler to first split the insn, and then output the 17020 resulting instructions separately. This helps eliminate redundancy in 17021 the output templates. If you have a `define_insn' that needs to emit 17022 multiple assembler instructions, and there is an matching `define_split' 17023 already defined, then you can simply use `#' as the output template 17024 instead of writing an output template that emits the multiple assembler 17025 instructions. 17026 17027 If the macro `ASSEMBLER_DIALECT' is defined, you can use construct of 17028 the form `{option0|option1|option2}' in the templates. These describe 17029 multiple variants of assembler language syntax. *Note Instruction 17030 Output::. 17031 17032 17033 File: gccint.info, Node: Output Statement, Next: Predicates, Prev: Output Template, Up: Machine Desc 17034 17035 16.6 C Statements for Assembler Output 17036 ====================================== 17037 17038 Often a single fixed template string cannot produce correct and 17039 efficient assembler code for all the cases that are recognized by a 17040 single instruction pattern. For example, the opcodes may depend on the 17041 kinds of operands; or some unfortunate combinations of operands may 17042 require extra machine instructions. 17043 17044 If the output control string starts with a `@', then it is actually a 17045 series of templates, each on a separate line. (Blank lines and leading 17046 spaces and tabs are ignored.) The templates correspond to the 17047 pattern's constraint alternatives (*note Multi-Alternative::). For 17048 example, if a target machine has a two-address add instruction `addr' 17049 to add into a register and another `addm' to add a register to memory, 17050 you might write this pattern: 17051 17052 (define_insn "addsi3" 17053 [(set (match_operand:SI 0 "general_operand" "=r,m") 17054 (plus:SI (match_operand:SI 1 "general_operand" "0,0") 17055 (match_operand:SI 2 "general_operand" "g,r")))] 17056 "" 17057 "@ 17058 addr %2,%0 17059 addm %2,%0") 17060 17061 If the output control string starts with a `*', then it is not an 17062 output template but rather a piece of C program that should compute a 17063 template. It should execute a `return' statement to return the 17064 template-string you want. Most such templates use C string literals, 17065 which require doublequote characters to delimit them. To include these 17066 doublequote characters in the string, prefix each one with `\'. 17067 17068 If the output control string is written as a brace block instead of a 17069 double-quoted string, it is automatically assumed to be C code. In that 17070 case, it is not necessary to put in a leading asterisk, or to escape the 17071 doublequotes surrounding C string literals. 17072 17073 The operands may be found in the array `operands', whose C data type 17074 is `rtx []'. 17075 17076 It is very common to select different ways of generating assembler code 17077 based on whether an immediate operand is within a certain range. Be 17078 careful when doing this, because the result of `INTVAL' is an integer 17079 on the host machine. If the host machine has more bits in an `int' 17080 than the target machine has in the mode in which the constant will be 17081 used, then some of the bits you get from `INTVAL' will be superfluous. 17082 For proper results, you must carefully disregard the values of those 17083 bits. 17084 17085 It is possible to output an assembler instruction and then go on to 17086 output or compute more of them, using the subroutine `output_asm_insn'. 17087 This receives two arguments: a template-string and a vector of 17088 operands. The vector may be `operands', or it may be another array of 17089 `rtx' that you declare locally and initialize yourself. 17090 17091 When an insn pattern has multiple alternatives in its constraints, 17092 often the appearance of the assembler code is determined mostly by 17093 which alternative was matched. When this is so, the C code can test 17094 the variable `which_alternative', which is the ordinal number of the 17095 alternative that was actually satisfied (0 for the first, 1 for the 17096 second alternative, etc.). 17097 17098 For example, suppose there are two opcodes for storing zero, `clrreg' 17099 for registers and `clrmem' for memory locations. Here is how a pattern 17100 could use `which_alternative' to choose between them: 17101 17102 (define_insn "" 17103 [(set (match_operand:SI 0 "general_operand" "=r,m") 17104 (const_int 0))] 17105 "" 17106 { 17107 return (which_alternative == 0 17108 ? "clrreg %0" : "clrmem %0"); 17109 }) 17110 17111 The example above, where the assembler code to generate was _solely_ 17112 determined by the alternative, could also have been specified as 17113 follows, having the output control string start with a `@': 17114 17115 (define_insn "" 17116 [(set (match_operand:SI 0 "general_operand" "=r,m") 17117 (const_int 0))] 17118 "" 17119 "@ 17120 clrreg %0 17121 clrmem %0") 17122 17123 17124 File: gccint.info, Node: Predicates, Next: Constraints, Prev: Output Statement, Up: Machine Desc 17125 17126 16.7 Predicates 17127 =============== 17128 17129 A predicate determines whether a `match_operand' or `match_operator' 17130 expression matches, and therefore whether the surrounding instruction 17131 pattern will be used for that combination of operands. GCC has a 17132 number of machine-independent predicates, and you can define 17133 machine-specific predicates as needed. By convention, predicates used 17134 with `match_operand' have names that end in `_operand', and those used 17135 with `match_operator' have names that end in `_operator'. 17136 17137 All predicates are Boolean functions (in the mathematical sense) of 17138 two arguments: the RTL expression that is being considered at that 17139 position in the instruction pattern, and the machine mode that the 17140 `match_operand' or `match_operator' specifies. In this section, the 17141 first argument is called OP and the second argument MODE. Predicates 17142 can be called from C as ordinary two-argument functions; this can be 17143 useful in output templates or other machine-specific code. 17144 17145 Operand predicates can allow operands that are not actually acceptable 17146 to the hardware, as long as the constraints give reload the ability to 17147 fix them up (*note Constraints::). However, GCC will usually generate 17148 better code if the predicates specify the requirements of the machine 17149 instructions as closely as possible. Reload cannot fix up operands 17150 that must be constants ("immediate operands"); you must use a predicate 17151 that allows only constants, or else enforce the requirement in the 17152 extra condition. 17153 17154 Most predicates handle their MODE argument in a uniform manner. If 17155 MODE is `VOIDmode' (unspecified), then OP can have any mode. If MODE 17156 is anything else, then OP must have the same mode, unless OP is a 17157 `CONST_INT' or integer `CONST_DOUBLE'. These RTL expressions always 17158 have `VOIDmode', so it would be counterproductive to check that their 17159 mode matches. Instead, predicates that accept `CONST_INT' and/or 17160 integer `CONST_DOUBLE' check that the value stored in the constant will 17161 fit in the requested mode. 17162 17163 Predicates with this behavior are called "normal". `genrecog' can 17164 optimize the instruction recognizer based on knowledge of how normal 17165 predicates treat modes. It can also diagnose certain kinds of common 17166 errors in the use of normal predicates; for instance, it is almost 17167 always an error to use a normal predicate without specifying a mode. 17168 17169 Predicates that do something different with their MODE argument are 17170 called "special". The generic predicates `address_operand' and 17171 `pmode_register_operand' are special predicates. `genrecog' does not 17172 do any optimizations or diagnosis when special predicates are used. 17173 17174 * Menu: 17175 17176 * Machine-Independent Predicates:: Predicates available to all back ends. 17177 * Defining Predicates:: How to write machine-specific predicate 17178 functions. 17179 17180 17181 File: gccint.info, Node: Machine-Independent Predicates, Next: Defining Predicates, Up: Predicates 17182 17183 16.7.1 Machine-Independent Predicates 17184 ------------------------------------- 17185 17186 These are the generic predicates available to all back ends. They are 17187 defined in `recog.c'. The first category of predicates allow only 17188 constant, or "immediate", operands. 17189 17190 -- Function: immediate_operand 17191 This predicate allows any sort of constant that fits in MODE. It 17192 is an appropriate choice for instructions that take operands that 17193 must be constant. 17194 17195 -- Function: const_int_operand 17196 This predicate allows any `CONST_INT' expression that fits in 17197 MODE. It is an appropriate choice for an immediate operand that 17198 does not allow a symbol or label. 17199 17200 -- Function: const_double_operand 17201 This predicate accepts any `CONST_DOUBLE' expression that has 17202 exactly MODE. If MODE is `VOIDmode', it will also accept 17203 `CONST_INT'. It is intended for immediate floating point 17204 constants. 17205 17206 The second category of predicates allow only some kind of machine 17207 register. 17208 17209 -- Function: register_operand 17210 This predicate allows any `REG' or `SUBREG' expression that is 17211 valid for MODE. It is often suitable for arithmetic instruction 17212 operands on a RISC machine. 17213 17214 -- Function: pmode_register_operand 17215 This is a slight variant on `register_operand' which works around 17216 a limitation in the machine-description reader. 17217 17218 (match_operand N "pmode_register_operand" CONSTRAINT) 17219 17220 means exactly what 17221 17222 (match_operand:P N "register_operand" CONSTRAINT) 17223 17224 would mean, if the machine-description reader accepted `:P' mode 17225 suffixes. Unfortunately, it cannot, because `Pmode' is an alias 17226 for some other mode, and might vary with machine-specific options. 17227 *Note Misc::. 17228 17229 -- Function: scratch_operand 17230 This predicate allows hard registers and `SCRATCH' expressions, 17231 but not pseudo-registers. It is used internally by 17232 `match_scratch'; it should not be used directly. 17233 17234 The third category of predicates allow only some kind of memory 17235 reference. 17236 17237 -- Function: memory_operand 17238 This predicate allows any valid reference to a quantity of mode 17239 MODE in memory, as determined by the weak form of 17240 `GO_IF_LEGITIMATE_ADDRESS' (*note Addressing Modes::). 17241 17242 -- Function: address_operand 17243 This predicate is a little unusual; it allows any operand that is a 17244 valid expression for the _address_ of a quantity of mode MODE, 17245 again determined by the weak form of `GO_IF_LEGITIMATE_ADDRESS'. 17246 To first order, if `(mem:MODE (EXP))' is acceptable to 17247 `memory_operand', then EXP is acceptable to `address_operand'. 17248 Note that EXP does not necessarily have the mode MODE. 17249 17250 -- Function: indirect_operand 17251 This is a stricter form of `memory_operand' which allows only 17252 memory references with a `general_operand' as the address 17253 expression. New uses of this predicate are discouraged, because 17254 `general_operand' is very permissive, so it's hard to tell what an 17255 `indirect_operand' does or does not allow. If a target has 17256 different requirements for memory operands for different 17257 instructions, it is better to define target-specific predicates 17258 which enforce the hardware's requirements explicitly. 17259 17260 -- Function: push_operand 17261 This predicate allows a memory reference suitable for pushing a 17262 value onto the stack. This will be a `MEM' which refers to 17263 `stack_pointer_rtx', with a side-effect in its address expression 17264 (*note Incdec::); which one is determined by the `STACK_PUSH_CODE' 17265 macro (*note Frame Layout::). 17266 17267 -- Function: pop_operand 17268 This predicate allows a memory reference suitable for popping a 17269 value off the stack. Again, this will be a `MEM' referring to 17270 `stack_pointer_rtx', with a side-effect in its address expression. 17271 However, this time `STACK_POP_CODE' is expected. 17272 17273 The fourth category of predicates allow some combination of the above 17274 operands. 17275 17276 -- Function: nonmemory_operand 17277 This predicate allows any immediate or register operand valid for 17278 MODE. 17279 17280 -- Function: nonimmediate_operand 17281 This predicate allows any register or memory operand valid for 17282 MODE. 17283 17284 -- Function: general_operand 17285 This predicate allows any immediate, register, or memory operand 17286 valid for MODE. 17287 17288 Finally, there is one generic operator predicate. 17289 17290 -- Function: comparison_operator 17291 This predicate matches any expression which performs an arithmetic 17292 comparison in MODE; that is, `COMPARISON_P' is true for the 17293 expression code. 17294 17295 17296 File: gccint.info, Node: Defining Predicates, Prev: Machine-Independent Predicates, Up: Predicates 17297 17298 16.7.2 Defining Machine-Specific Predicates 17299 ------------------------------------------- 17300 17301 Many machines have requirements for their operands that cannot be 17302 expressed precisely using the generic predicates. You can define 17303 additional predicates using `define_predicate' and 17304 `define_special_predicate' expressions. These expressions have three 17305 operands: 17306 17307 * The name of the predicate, as it will be referred to in 17308 `match_operand' or `match_operator' expressions. 17309 17310 * An RTL expression which evaluates to true if the predicate allows 17311 the operand OP, false if it does not. This expression can only use 17312 the following RTL codes: 17313 17314 `MATCH_OPERAND' 17315 When written inside a predicate expression, a `MATCH_OPERAND' 17316 expression evaluates to true if the predicate it names would 17317 allow OP. The operand number and constraint are ignored. 17318 Due to limitations in `genrecog', you can only refer to 17319 generic predicates and predicates that have already been 17320 defined. 17321 17322 `MATCH_CODE' 17323 This expression evaluates to true if OP or a specified 17324 subexpression of OP has one of a given list of RTX codes. 17325 17326 The first operand of this expression is a string constant 17327 containing a comma-separated list of RTX code names (in lower 17328 case). These are the codes for which the `MATCH_CODE' will 17329 be true. 17330 17331 The second operand is a string constant which indicates what 17332 subexpression of OP to examine. If it is absent or the empty 17333 string, OP itself is examined. Otherwise, the string constant 17334 must be a sequence of digits and/or lowercase letters. Each 17335 character indicates a subexpression to extract from the 17336 current expression; for the first character this is OP, for 17337 the second and subsequent characters it is the result of the 17338 previous character. A digit N extracts `XEXP (E, N)'; a 17339 letter L extracts `XVECEXP (E, 0, N)' where N is the 17340 alphabetic ordinal of L (0 for `a', 1 for 'b', and so on). 17341 The `MATCH_CODE' then examines the RTX code of the 17342 subexpression extracted by the complete string. It is not 17343 possible to extract components of an `rtvec' that is not at 17344 position 0 within its RTX object. 17345 17346 `MATCH_TEST' 17347 This expression has one operand, a string constant containing 17348 a C expression. The predicate's arguments, OP and MODE, are 17349 available with those names in the C expression. The 17350 `MATCH_TEST' evaluates to true if the C expression evaluates 17351 to a nonzero value. `MATCH_TEST' expressions must not have 17352 side effects. 17353 17354 `AND' 17355 `IOR' 17356 `NOT' 17357 `IF_THEN_ELSE' 17358 The basic `MATCH_' expressions can be combined using these 17359 logical operators, which have the semantics of the C operators 17360 `&&', `||', `!', and `? :' respectively. As in Common Lisp, 17361 you may give an `AND' or `IOR' expression an arbitrary number 17362 of arguments; this has exactly the same effect as writing a 17363 chain of two-argument `AND' or `IOR' expressions. 17364 17365 * An optional block of C code, which should execute `return true' if 17366 the predicate is found to match and `return false' if it does not. 17367 It must not have any side effects. The predicate arguments, OP 17368 and MODE, are available with those names. 17369 17370 If a code block is present in a predicate definition, then the RTL 17371 expression must evaluate to true _and_ the code block must execute 17372 `return true' for the predicate to allow the operand. The RTL 17373 expression is evaluated first; do not re-check anything in the 17374 code block that was checked in the RTL expression. 17375 17376 The program `genrecog' scans `define_predicate' and 17377 `define_special_predicate' expressions to determine which RTX codes are 17378 possibly allowed. You should always make this explicit in the RTL 17379 predicate expression, using `MATCH_OPERAND' and `MATCH_CODE'. 17380 17381 Here is an example of a simple predicate definition, from the IA64 17382 machine description: 17383 17384 ;; True if OP is a `SYMBOL_REF' which refers to the sdata section. 17385 (define_predicate "small_addr_symbolic_operand" 17386 (and (match_code "symbol_ref") 17387 (match_test "SYMBOL_REF_SMALL_ADDR_P (op)"))) 17388 17389 And here is another, showing the use of the C block. 17390 17391 ;; True if OP is a register operand that is (or could be) a GR reg. 17392 (define_predicate "gr_register_operand" 17393 (match_operand 0 "register_operand") 17394 { 17395 unsigned int regno; 17396 if (GET_CODE (op) == SUBREG) 17397 op = SUBREG_REG (op); 17398 17399 regno = REGNO (op); 17400 return (regno >= FIRST_PSEUDO_REGISTER || GENERAL_REGNO_P (regno)); 17401 }) 17402 17403 Predicates written with `define_predicate' automatically include a 17404 test that MODE is `VOIDmode', or OP has the same mode as MODE, or OP is 17405 a `CONST_INT' or `CONST_DOUBLE'. They do _not_ check specifically for 17406 integer `CONST_DOUBLE', nor do they test that the value of either kind 17407 of constant fits in the requested mode. This is because 17408 target-specific predicates that take constants usually have to do more 17409 stringent value checks anyway. If you need the exact same treatment of 17410 `CONST_INT' or `CONST_DOUBLE' that the generic predicates provide, use 17411 a `MATCH_OPERAND' subexpression to call `const_int_operand', 17412 `const_double_operand', or `immediate_operand'. 17413 17414 Predicates written with `define_special_predicate' do not get any 17415 automatic mode checks, and are treated as having special mode handling 17416 by `genrecog'. 17417 17418 The program `genpreds' is responsible for generating code to test 17419 predicates. It also writes a header file containing function 17420 declarations for all machine-specific predicates. It is not necessary 17421 to declare these predicates in `CPU-protos.h'. 17422 17423 17424 File: gccint.info, Node: Constraints, Next: Standard Names, Prev: Predicates, Up: Machine Desc 17425 17426 16.8 Operand Constraints 17427 ======================== 17428 17429 Each `match_operand' in an instruction pattern can specify constraints 17430 for the operands allowed. The constraints allow you to fine-tune 17431 matching within the set of operands allowed by the predicate. 17432 17433 Constraints can say whether an operand may be in a register, and which 17434 kinds of register; whether the operand can be a memory reference, and 17435 which kinds of address; whether the operand may be an immediate 17436 constant, and which possible values it may have. Constraints can also 17437 require two operands to match. 17438 17439 * Menu: 17440 17441 * Simple Constraints:: Basic use of constraints. 17442 * Multi-Alternative:: When an insn has two alternative constraint-patterns. 17443 * Class Preferences:: Constraints guide which hard register to put things in. 17444 * Modifiers:: More precise control over effects of constraints. 17445 * Disable Insn Alternatives:: Disable insn alternatives using the `enabled' attribute. 17446 * Machine Constraints:: Existing constraints for some particular machines. 17447 * Define Constraints:: How to define machine-specific constraints. 17448 * C Constraint Interface:: How to test constraints from C code. 17449 17450 17451 File: gccint.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints 17452 17453 16.8.1 Simple Constraints 17454 ------------------------- 17455 17456 The simplest kind of constraint is a string full of letters, each of 17457 which describes one kind of operand that is permitted. Here are the 17458 letters that are allowed: 17459 17460 whitespace 17461 Whitespace characters are ignored and can be inserted at any 17462 position except the first. This enables each alternative for 17463 different operands to be visually aligned in the machine 17464 description even if they have different number of constraints and 17465 modifiers. 17466 17467 `m' 17468 A memory operand is allowed, with any kind of address that the 17469 machine supports in general. Note that the letter used for the 17470 general memory constraint can be re-defined by a back end using 17471 the `TARGET_MEM_CONSTRAINT' macro. 17472 17473 `o' 17474 A memory operand is allowed, but only if the address is 17475 "offsettable". This means that adding a small integer (actually, 17476 the width in bytes of the operand, as determined by its machine 17477 mode) may be added to the address and the result is also a valid 17478 memory address. 17479 17480 For example, an address which is constant is offsettable; so is an 17481 address that is the sum of a register and a constant (as long as a 17482 slightly larger constant is also within the range of 17483 address-offsets supported by the machine); but an autoincrement or 17484 autodecrement address is not offsettable. More complicated 17485 indirect/indexed addresses may or may not be offsettable depending 17486 on the other addressing modes that the machine supports. 17487 17488 Note that in an output operand which can be matched by another 17489 operand, the constraint letter `o' is valid only when accompanied 17490 by both `<' (if the target machine has predecrement addressing) 17491 and `>' (if the target machine has preincrement addressing). 17492 17493 `V' 17494 A memory operand that is not offsettable. In other words, 17495 anything that would fit the `m' constraint but not the `o' 17496 constraint. 17497 17498 `<' 17499 A memory operand with autodecrement addressing (either 17500 predecrement or postdecrement) is allowed. 17501 17502 `>' 17503 A memory operand with autoincrement addressing (either 17504 preincrement or postincrement) is allowed. 17505 17506 `r' 17507 A register operand is allowed provided that it is in a general 17508 register. 17509 17510 `i' 17511 An immediate integer operand (one with constant value) is allowed. 17512 This includes symbolic constants whose values will be known only at 17513 assembly time or later. 17514 17515 `n' 17516 An immediate integer operand with a known numeric value is allowed. 17517 Many systems cannot support assembly-time constants for operands 17518 less than a word wide. Constraints for these operands should use 17519 `n' rather than `i'. 17520 17521 `I', `J', `K', ... `P' 17522 Other letters in the range `I' through `P' may be defined in a 17523 machine-dependent fashion to permit immediate integer operands with 17524 explicit integer values in specified ranges. For example, on the 17525 68000, `I' is defined to stand for the range of values 1 to 8. 17526 This is the range permitted as a shift count in the shift 17527 instructions. 17528 17529 `E' 17530 An immediate floating operand (expression code `const_double') is 17531 allowed, but only if the target floating point format is the same 17532 as that of the host machine (on which the compiler is running). 17533 17534 `F' 17535 An immediate floating operand (expression code `const_double' or 17536 `const_vector') is allowed. 17537 17538 `G', `H' 17539 `G' and `H' may be defined in a machine-dependent fashion to 17540 permit immediate floating operands in particular ranges of values. 17541 17542 `s' 17543 An immediate integer operand whose value is not an explicit 17544 integer is allowed. 17545 17546 This might appear strange; if an insn allows a constant operand 17547 with a value not known at compile time, it certainly must allow 17548 any known value. So why use `s' instead of `i'? Sometimes it 17549 allows better code to be generated. 17550 17551 For example, on the 68000 in a fullword instruction it is possible 17552 to use an immediate operand; but if the immediate value is between 17553 -128 and 127, better code results from loading the value into a 17554 register and using the register. This is because the load into 17555 the register can be done with a `moveq' instruction. We arrange 17556 for this to happen by defining the letter `K' to mean "any integer 17557 outside the range -128 to 127", and then specifying `Ks' in the 17558 operand constraints. 17559 17560 `g' 17561 Any register, memory or immediate integer operand is allowed, 17562 except for registers that are not general registers. 17563 17564 `X' 17565 Any operand whatsoever is allowed, even if it does not satisfy 17566 `general_operand'. This is normally used in the constraint of a 17567 `match_scratch' when certain alternatives will not actually 17568 require a scratch register. 17569 17570 `0', `1', `2', ... `9' 17571 An operand that matches the specified operand number is allowed. 17572 If a digit is used together with letters within the same 17573 alternative, the digit should come last. 17574 17575 This number is allowed to be more than a single digit. If multiple 17576 digits are encountered consecutively, they are interpreted as a 17577 single decimal integer. There is scant chance for ambiguity, 17578 since to-date it has never been desirable that `10' be interpreted 17579 as matching either operand 1 _or_ operand 0. Should this be 17580 desired, one can use multiple alternatives instead. 17581 17582 This is called a "matching constraint" and what it really means is 17583 that the assembler has only a single operand that fills two roles 17584 considered separate in the RTL insn. For example, an add insn has 17585 two input operands and one output operand in the RTL, but on most 17586 CISC machines an add instruction really has only two operands, one 17587 of them an input-output operand: 17588 17589 addl #35,r12 17590 17591 Matching constraints are used in these circumstances. More 17592 precisely, the two operands that match must include one input-only 17593 operand and one output-only operand. Moreover, the digit must be a 17594 smaller number than the number of the operand that uses it in the 17595 constraint. 17596 17597 For operands to match in a particular case usually means that they 17598 are identical-looking RTL expressions. But in a few special cases 17599 specific kinds of dissimilarity are allowed. For example, `*x' as 17600 an input operand will match `*x++' as an output operand. For 17601 proper results in such cases, the output template should always 17602 use the output-operand's number when printing the operand. 17603 17604 `p' 17605 An operand that is a valid memory address is allowed. This is for 17606 "load address" and "push address" instructions. 17607 17608 `p' in the constraint must be accompanied by `address_operand' as 17609 the predicate in the `match_operand'. This predicate interprets 17610 the mode specified in the `match_operand' as the mode of the memory 17611 reference for which the address would be valid. 17612 17613 OTHER-LETTERS 17614 Other letters can be defined in machine-dependent fashion to stand 17615 for particular classes of registers or other arbitrary operand 17616 types. `d', `a' and `f' are defined on the 68000/68020 to stand 17617 for data, address and floating point registers. 17618 17619 In order to have valid assembler code, each operand must satisfy its 17620 constraint. But a failure to do so does not prevent the pattern from 17621 applying to an insn. Instead, it directs the compiler to modify the 17622 code so that the constraint will be satisfied. Usually this is done by 17623 copying an operand into a register. 17624 17625 Contrast, therefore, the two instruction patterns that follow: 17626 17627 (define_insn "" 17628 [(set (match_operand:SI 0 "general_operand" "=r") 17629 (plus:SI (match_dup 0) 17630 (match_operand:SI 1 "general_operand" "r")))] 17631 "" 17632 "...") 17633 17634 which has two operands, one of which must appear in two places, and 17635 17636 (define_insn "" 17637 [(set (match_operand:SI 0 "general_operand" "=r") 17638 (plus:SI (match_operand:SI 1 "general_operand" "0") 17639 (match_operand:SI 2 "general_operand" "r")))] 17640 "" 17641 "...") 17642 17643 which has three operands, two of which are required by a constraint to 17644 be identical. If we are considering an insn of the form 17645 17646 (insn N PREV NEXT 17647 (set (reg:SI 3) 17648 (plus:SI (reg:SI 6) (reg:SI 109))) 17649 ...) 17650 17651 the first pattern would not apply at all, because this insn does not 17652 contain two identical subexpressions in the right place. The pattern 17653 would say, "That does not look like an add instruction; try other 17654 patterns". The second pattern would say, "Yes, that's an add 17655 instruction, but there is something wrong with it". It would direct 17656 the reload pass of the compiler to generate additional insns to make 17657 the constraint true. The results might look like this: 17658 17659 (insn N2 PREV N 17660 (set (reg:SI 3) (reg:SI 6)) 17661 ...) 17662 17663 (insn N N2 NEXT 17664 (set (reg:SI 3) 17665 (plus:SI (reg:SI 3) (reg:SI 109))) 17666 ...) 17667 17668 It is up to you to make sure that each operand, in each pattern, has 17669 constraints that can handle any RTL expression that could be present for 17670 that operand. (When multiple alternatives are in use, each pattern 17671 must, for each possible combination of operand expressions, have at 17672 least one alternative which can handle that combination of operands.) 17673 The constraints don't need to _allow_ any possible operand--when this is 17674 the case, they do not constrain--but they must at least point the way to 17675 reloading any possible operand so that it will fit. 17676 17677 * If the constraint accepts whatever operands the predicate permits, 17678 there is no problem: reloading is never necessary for this operand. 17679 17680 For example, an operand whose constraints permit everything except 17681 registers is safe provided its predicate rejects registers. 17682 17683 An operand whose predicate accepts only constant values is safe 17684 provided its constraints include the letter `i'. If any possible 17685 constant value is accepted, then nothing less than `i' will do; if 17686 the predicate is more selective, then the constraints may also be 17687 more selective. 17688 17689 * Any operand expression can be reloaded by copying it into a 17690 register. So if an operand's constraints allow some kind of 17691 register, it is certain to be safe. It need not permit all 17692 classes of registers; the compiler knows how to copy a register 17693 into another register of the proper class in order to make an 17694 instruction valid. 17695 17696 * A nonoffsettable memory reference can be reloaded by copying the 17697 address into a register. So if the constraint uses the letter 17698 `o', all memory references are taken care of. 17699 17700 * A constant operand can be reloaded by allocating space in memory to 17701 hold it as preinitialized data. Then the memory reference can be 17702 used in place of the constant. So if the constraint uses the 17703 letters `o' or `m', constant operands are not a problem. 17704 17705 * If the constraint permits a constant and a pseudo register used in 17706 an insn was not allocated to a hard register and is equivalent to 17707 a constant, the register will be replaced with the constant. If 17708 the predicate does not permit a constant and the insn is 17709 re-recognized for some reason, the compiler will crash. Thus the 17710 predicate must always recognize any objects allowed by the 17711 constraint. 17712 17713 If the operand's predicate can recognize registers, but the constraint 17714 does not permit them, it can make the compiler crash. When this 17715 operand happens to be a register, the reload pass will be stymied, 17716 because it does not know how to copy a register temporarily into memory. 17717 17718 If the predicate accepts a unary operator, the constraint applies to 17719 the operand. For example, the MIPS processor at ISA level 3 supports an 17720 instruction which adds two registers in `SImode' to produce a `DImode' 17721 result, but only if the registers are correctly sign extended. This 17722 predicate for the input operands accepts a `sign_extend' of an `SImode' 17723 register. Write the constraint to indicate the type of register that 17724 is required for the operand of the `sign_extend'. 17725 17726 17727 File: gccint.info, Node: Multi-Alternative, Next: Class Preferences, Prev: Simple Constraints, Up: Constraints 17728 17729 16.8.2 Multiple Alternative Constraints 17730 --------------------------------------- 17731 17732 Sometimes a single instruction has multiple alternative sets of possible 17733 operands. For example, on the 68000, a logical-or instruction can 17734 combine register or an immediate value into memory, or it can combine 17735 any kind of operand into a register; but it cannot combine one memory 17736 location into another. 17737 17738 These constraints are represented as multiple alternatives. An 17739 alternative can be described by a series of letters for each operand. 17740 The overall constraint for an operand is made from the letters for this 17741 operand from the first alternative, a comma, the letters for this 17742 operand from the second alternative, a comma, and so on until the last 17743 alternative. Here is how it is done for fullword logical-or on the 17744 68000: 17745 17746 (define_insn "iorsi3" 17747 [(set (match_operand:SI 0 "general_operand" "=m,d") 17748 (ior:SI (match_operand:SI 1 "general_operand" "%0,0") 17749 (match_operand:SI 2 "general_operand" "dKs,dmKs")))] 17750 ...) 17751 17752 The first alternative has `m' (memory) for operand 0, `0' for operand 17753 1 (meaning it must match operand 0), and `dKs' for operand 2. The 17754 second alternative has `d' (data register) for operand 0, `0' for 17755 operand 1, and `dmKs' for operand 2. The `=' and `%' in the 17756 constraints apply to all the alternatives; their meaning is explained 17757 in the next section (*note Class Preferences::). 17758 17759 If all the operands fit any one alternative, the instruction is valid. 17760 Otherwise, for each alternative, the compiler counts how many 17761 instructions must be added to copy the operands so that that 17762 alternative applies. The alternative requiring the least copying is 17763 chosen. If two alternatives need the same amount of copying, the one 17764 that comes first is chosen. These choices can be altered with the `?' 17765 and `!' characters: 17766 17767 `?' 17768 Disparage slightly the alternative that the `?' appears in, as a 17769 choice when no alternative applies exactly. The compiler regards 17770 this alternative as one unit more costly for each `?' that appears 17771 in it. 17772 17773 `!' 17774 Disparage severely the alternative that the `!' appears in. This 17775 alternative can still be used if it fits without reloading, but if 17776 reloading is needed, some other alternative will be used. 17777 17778 When an insn pattern has multiple alternatives in its constraints, 17779 often the appearance of the assembler code is determined mostly by which 17780 alternative was matched. When this is so, the C code for writing the 17781 assembler code can use the variable `which_alternative', which is the 17782 ordinal number of the alternative that was actually satisfied (0 for 17783 the first, 1 for the second alternative, etc.). *Note Output 17784 Statement::. 17785 17786 17787 File: gccint.info, Node: Class Preferences, Next: Modifiers, Prev: Multi-Alternative, Up: Constraints 17788 17789 16.8.3 Register Class Preferences 17790 --------------------------------- 17791 17792 The operand constraints have another function: they enable the compiler 17793 to decide which kind of hardware register a pseudo register is best 17794 allocated to. The compiler examines the constraints that apply to the 17795 insns that use the pseudo register, looking for the machine-dependent 17796 letters such as `d' and `a' that specify classes of registers. The 17797 pseudo register is put in whichever class gets the most "votes". The 17798 constraint letters `g' and `r' also vote: they vote in favor of a 17799 general register. The machine description says which registers are 17800 considered general. 17801 17802 Of course, on some machines all registers are equivalent, and no 17803 register classes are defined. Then none of this complexity is relevant. 17804 17805 17806 File: gccint.info, Node: Modifiers, Next: Disable Insn Alternatives, Prev: Class Preferences, Up: Constraints 17807 17808 16.8.4 Constraint Modifier Characters 17809 ------------------------------------- 17810 17811 Here are constraint modifier characters. 17812 17813 `=' 17814 Means that this operand is write-only for this instruction: the 17815 previous value is discarded and replaced by output data. 17816 17817 `+' 17818 Means that this operand is both read and written by the 17819 instruction. 17820 17821 When the compiler fixes up the operands to satisfy the constraints, 17822 it needs to know which operands are inputs to the instruction and 17823 which are outputs from it. `=' identifies an output; `+' 17824 identifies an operand that is both input and output; all other 17825 operands are assumed to be input only. 17826 17827 If you specify `=' or `+' in a constraint, you put it in the first 17828 character of the constraint string. 17829 17830 `&' 17831 Means (in a particular alternative) that this operand is an 17832 "earlyclobber" operand, which is modified before the instruction is 17833 finished using the input operands. Therefore, this operand may 17834 not lie in a register that is used as an input operand or as part 17835 of any memory address. 17836 17837 `&' applies only to the alternative in which it is written. In 17838 constraints with multiple alternatives, sometimes one alternative 17839 requires `&' while others do not. See, for example, the `movdf' 17840 insn of the 68000. 17841 17842 An input operand can be tied to an earlyclobber operand if its only 17843 use as an input occurs before the early result is written. Adding 17844 alternatives of this form often allows GCC to produce better code 17845 when only some of the inputs can be affected by the earlyclobber. 17846 See, for example, the `mulsi3' insn of the ARM. 17847 17848 `&' does not obviate the need to write `='. 17849 17850 `%' 17851 Declares the instruction to be commutative for this operand and the 17852 following operand. This means that the compiler may interchange 17853 the two operands if that is the cheapest way to make all operands 17854 fit the constraints. This is often used in patterns for addition 17855 instructions that really have only two operands: the result must 17856 go in one of the arguments. Here for example, is how the 68000 17857 halfword-add instruction is defined: 17858 17859 (define_insn "addhi3" 17860 [(set (match_operand:HI 0 "general_operand" "=m,r") 17861 (plus:HI (match_operand:HI 1 "general_operand" "%0,0") 17862 (match_operand:HI 2 "general_operand" "di,g")))] 17863 ...) 17864 GCC can only handle one commutative pair in an asm; if you use 17865 more, the compiler may fail. Note that you need not use the 17866 modifier if the two alternatives are strictly identical; this 17867 would only waste time in the reload pass. The modifier is not 17868 operational after register allocation, so the result of 17869 `define_peephole2' and `define_split's performed after reload 17870 cannot rely on `%' to make the intended insn match. 17871 17872 `#' 17873 Says that all following characters, up to the next comma, are to be 17874 ignored as a constraint. They are significant only for choosing 17875 register preferences. 17876 17877 `*' 17878 Says that the following character should be ignored when choosing 17879 register preferences. `*' has no effect on the meaning of the 17880 constraint as a constraint, and no effect on reloading. 17881 17882 Here is an example: the 68000 has an instruction to sign-extend a 17883 halfword in a data register, and can also sign-extend a value by 17884 copying it into an address register. While either kind of 17885 register is acceptable, the constraints on an address-register 17886 destination are less strict, so it is best if register allocation 17887 makes an address register its goal. Therefore, `*' is used so 17888 that the `d' constraint letter (for data register) is ignored when 17889 computing register preferences. 17890 17891 (define_insn "extendhisi2" 17892 [(set (match_operand:SI 0 "general_operand" "=*d,a") 17893 (sign_extend:SI 17894 (match_operand:HI 1 "general_operand" "0,g")))] 17895 ...) 17896 17897 17898 File: gccint.info, Node: Machine Constraints, Next: Define Constraints, Prev: Disable Insn Alternatives, Up: Constraints 17899 17900 16.8.5 Constraints for Particular Machines 17901 ------------------------------------------ 17902 17903 Whenever possible, you should use the general-purpose constraint letters 17904 in `asm' arguments, since they will convey meaning more readily to 17905 people reading your code. Failing that, use the constraint letters 17906 that usually have very similar meanings across architectures. The most 17907 commonly used constraints are `m' and `r' (for memory and 17908 general-purpose registers respectively; *note Simple Constraints::), and 17909 `I', usually the letter indicating the most common immediate-constant 17910 format. 17911 17912 Each architecture defines additional constraints. These constraints 17913 are used by the compiler itself for instruction generation, as well as 17914 for `asm' statements; therefore, some of the constraints are not 17915 particularly useful for `asm'. Here is a summary of some of the 17916 machine-dependent constraints available on some particular machines; it 17917 includes both constraints that are useful for `asm' and constraints 17918 that aren't. The compiler source file mentioned in the table heading 17919 for each architecture is the definitive reference for the meanings of 17920 that architecture's constraints. 17921 17922 _ARM family--`config/arm/arm.h'_ 17923 17924 `f' 17925 Floating-point register 17926 17927 `w' 17928 VFP floating-point register 17929 17930 `F' 17931 One of the floating-point constants 0.0, 0.5, 1.0, 2.0, 3.0, 17932 4.0, 5.0 or 10.0 17933 17934 `G' 17935 Floating-point constant that would satisfy the constraint `F' 17936 if it were negated 17937 17938 `I' 17939 Integer that is valid as an immediate operand in a data 17940 processing instruction. That is, an integer in the range 0 17941 to 255 rotated by a multiple of 2 17942 17943 `J' 17944 Integer in the range -4095 to 4095 17945 17946 `K' 17947 Integer that satisfies constraint `I' when inverted (ones 17948 complement) 17949 17950 `L' 17951 Integer that satisfies constraint `I' when negated (twos 17952 complement) 17953 17954 `M' 17955 Integer in the range 0 to 32 17956 17957 `Q' 17958 A memory reference where the exact address is in a single 17959 register (``m'' is preferable for `asm' statements) 17960 17961 `R' 17962 An item in the constant pool 17963 17964 `S' 17965 A symbol in the text segment of the current file 17966 17967 `Uv' 17968 A memory reference suitable for VFP load/store insns 17969 (reg+constant offset) 17970 17971 `Uy' 17972 A memory reference suitable for iWMMXt load/store 17973 instructions. 17974 17975 `Uq' 17976 A memory reference suitable for the ARMv4 ldrsb instruction. 17977 17978 _AVR family--`config/avr/constraints.md'_ 17979 17980 `l' 17981 Registers from r0 to r15 17982 17983 `a' 17984 Registers from r16 to r23 17985 17986 `d' 17987 Registers from r16 to r31 17988 17989 `w' 17990 Registers from r24 to r31. These registers can be used in 17991 `adiw' command 17992 17993 `e' 17994 Pointer register (r26-r31) 17995 17996 `b' 17997 Base pointer register (r28-r31) 17998 17999 `q' 18000 Stack pointer register (SPH:SPL) 18001 18002 `t' 18003 Temporary register r0 18004 18005 `x' 18006 Register pair X (r27:r26) 18007 18008 `y' 18009 Register pair Y (r29:r28) 18010 18011 `z' 18012 Register pair Z (r31:r30) 18013 18014 `I' 18015 Constant greater than -1, less than 64 18016 18017 `J' 18018 Constant greater than -64, less than 1 18019 18020 `K' 18021 Constant integer 2 18022 18023 `L' 18024 Constant integer 0 18025 18026 `M' 18027 Constant that fits in 8 bits 18028 18029 `N' 18030 Constant integer -1 18031 18032 `O' 18033 Constant integer 8, 16, or 24 18034 18035 `P' 18036 Constant integer 1 18037 18038 `G' 18039 A floating point constant 0.0 18040 18041 `R' 18042 Integer constant in the range -6 ... 5. 18043 18044 `Q' 18045 A memory address based on Y or Z pointer with displacement. 18046 18047 _CRX Architecture--`config/crx/crx.h'_ 18048 18049 `b' 18050 Registers from r0 to r14 (registers without stack pointer) 18051 18052 `l' 18053 Register r16 (64-bit accumulator lo register) 18054 18055 `h' 18056 Register r17 (64-bit accumulator hi register) 18057 18058 `k' 18059 Register pair r16-r17. (64-bit accumulator lo-hi pair) 18060 18061 `I' 18062 Constant that fits in 3 bits 18063 18064 `J' 18065 Constant that fits in 4 bits 18066 18067 `K' 18068 Constant that fits in 5 bits 18069 18070 `L' 18071 Constant that is one of -1, 4, -4, 7, 8, 12, 16, 20, 32, 48 18072 18073 `G' 18074 Floating point constant that is legal for store immediate 18075 18076 _Hewlett-Packard PA-RISC--`config/pa/pa.h'_ 18077 18078 `a' 18079 General register 1 18080 18081 `f' 18082 Floating point register 18083 18084 `q' 18085 Shift amount register 18086 18087 `x' 18088 Floating point register (deprecated) 18089 18090 `y' 18091 Upper floating point register (32-bit), floating point 18092 register (64-bit) 18093 18094 `Z' 18095 Any register 18096 18097 `I' 18098 Signed 11-bit integer constant 18099 18100 `J' 18101 Signed 14-bit integer constant 18102 18103 `K' 18104 Integer constant that can be deposited with a `zdepi' 18105 instruction 18106 18107 `L' 18108 Signed 5-bit integer constant 18109 18110 `M' 18111 Integer constant 0 18112 18113 `N' 18114 Integer constant that can be loaded with a `ldil' instruction 18115 18116 `O' 18117 Integer constant whose value plus one is a power of 2 18118 18119 `P' 18120 Integer constant that can be used for `and' operations in 18121 `depi' and `extru' instructions 18122 18123 `S' 18124 Integer constant 31 18125 18126 `U' 18127 Integer constant 63 18128 18129 `G' 18130 Floating-point constant 0.0 18131 18132 `A' 18133 A `lo_sum' data-linkage-table memory operand 18134 18135 `Q' 18136 A memory operand that can be used as the destination operand 18137 of an integer store instruction 18138 18139 `R' 18140 A scaled or unscaled indexed memory operand 18141 18142 `T' 18143 A memory operand for floating-point loads and stores 18144 18145 `W' 18146 A register indirect memory operand 18147 18148 _picoChip family--`picochip.h'_ 18149 18150 `k' 18151 Stack register. 18152 18153 `f' 18154 Pointer register. A register which can be used to access 18155 memory without supplying an offset. Any other register can 18156 be used to access memory, but will need a constant offset. 18157 In the case of the offset being zero, it is more efficient to 18158 use a pointer register, since this reduces code size. 18159 18160 `t' 18161 A twin register. A register which may be paired with an 18162 adjacent register to create a 32-bit register. 18163 18164 `a' 18165 Any absolute memory address (e.g., symbolic constant, symbolic 18166 constant + offset). 18167 18168 `I' 18169 4-bit signed integer. 18170 18171 `J' 18172 4-bit unsigned integer. 18173 18174 `K' 18175 8-bit signed integer. 18176 18177 `M' 18178 Any constant whose absolute value is no greater than 4-bits. 18179 18180 `N' 18181 10-bit signed integer 18182 18183 `O' 18184 16-bit signed integer. 18185 18186 18187 _PowerPC and IBM RS6000--`config/rs6000/rs6000.h'_ 18188 18189 `b' 18190 Address base register 18191 18192 `f' 18193 Floating point register 18194 18195 `v' 18196 Vector register 18197 18198 `h' 18199 `MQ', `CTR', or `LINK' register 18200 18201 `q' 18202 `MQ' register 18203 18204 `c' 18205 `CTR' register 18206 18207 `l' 18208 `LINK' register 18209 18210 `x' 18211 `CR' register (condition register) number 0 18212 18213 `y' 18214 `CR' register (condition register) 18215 18216 `z' 18217 `FPMEM' stack memory for FPR-GPR transfers 18218 18219 `I' 18220 Signed 16-bit constant 18221 18222 `J' 18223 Unsigned 16-bit constant shifted left 16 bits (use `L' 18224 instead for `SImode' constants) 18225 18226 `K' 18227 Unsigned 16-bit constant 18228 18229 `L' 18230 Signed 16-bit constant shifted left 16 bits 18231 18232 `M' 18233 Constant larger than 31 18234 18235 `N' 18236 Exact power of 2 18237 18238 `O' 18239 Zero 18240 18241 `P' 18242 Constant whose negation is a signed 16-bit constant 18243 18244 `G' 18245 Floating point constant that can be loaded into a register 18246 with one instruction per word 18247 18248 `H' 18249 Integer/Floating point constant that can be loaded into a 18250 register using three instructions 18251 18252 `Q' 18253 Memory operand that is an offset from a register (`m' is 18254 preferable for `asm' statements) 18255 18256 `Z' 18257 Memory operand that is an indexed or indirect from a register 18258 (`m' is preferable for `asm' statements) 18259 18260 `R' 18261 AIX TOC entry 18262 18263 `a' 18264 Address operand that is an indexed or indirect from a 18265 register (`p' is preferable for `asm' statements) 18266 18267 `S' 18268 Constant suitable as a 64-bit mask operand 18269 18270 `T' 18271 Constant suitable as a 32-bit mask operand 18272 18273 `U' 18274 System V Release 4 small data area reference 18275 18276 `t' 18277 AND masks that can be performed by two rldic{l, r} 18278 instructions 18279 18280 `W' 18281 Vector constant that does not require memory 18282 18283 18284 _Intel 386--`config/i386/constraints.md'_ 18285 18286 `R' 18287 Legacy register--the eight integer registers available on all 18288 i386 processors (`a', `b', `c', `d', `si', `di', `bp', `sp'). 18289 18290 `q' 18291 Any register accessible as `Rl'. In 32-bit mode, `a', `b', 18292 `c', and `d'; in 64-bit mode, any integer register. 18293 18294 `Q' 18295 Any register accessible as `Rh': `a', `b', `c', and `d'. 18296 18297 `l' 18298 Any register that can be used as the index in a base+index 18299 memory access: that is, any general register except the stack 18300 pointer. 18301 18302 `a' 18303 The `a' register. 18304 18305 `b' 18306 The `b' register. 18307 18308 `c' 18309 The `c' register. 18310 18311 `d' 18312 The `d' register. 18313 18314 `S' 18315 The `si' register. 18316 18317 `D' 18318 The `di' register. 18319 18320 `A' 18321 The `a' and `d' registers, as a pair (for instructions that 18322 return half the result in one and half in the other). 18323 18324 `f' 18325 Any 80387 floating-point (stack) register. 18326 18327 `t' 18328 Top of 80387 floating-point stack (`%st(0)'). 18329 18330 `u' 18331 Second from top of 80387 floating-point stack (`%st(1)'). 18332 18333 `y' 18334 Any MMX register. 18335 18336 `x' 18337 Any SSE register. 18338 18339 `Yz' 18340 First SSE register (`%xmm0'). 18341 18342 `Y2' 18343 Any SSE register, when SSE2 is enabled. 18344 18345 `Yi' 18346 Any SSE register, when SSE2 and inter-unit moves are enabled. 18347 18348 `Ym' 18349 Any MMX register, when inter-unit moves are enabled. 18350 18351 `I' 18352 Integer constant in the range 0 ... 31, for 32-bit shifts. 18353 18354 `J' 18355 Integer constant in the range 0 ... 63, for 64-bit shifts. 18356 18357 `K' 18358 Signed 8-bit integer constant. 18359 18360 `L' 18361 `0xFF' or `0xFFFF', for andsi as a zero-extending move. 18362 18363 `M' 18364 0, 1, 2, or 3 (shifts for the `lea' instruction). 18365 18366 `N' 18367 Unsigned 8-bit integer constant (for `in' and `out' 18368 instructions). 18369 18370 `O' 18371 Integer constant in the range 0 ... 127, for 128-bit shifts. 18372 18373 `G' 18374 Standard 80387 floating point constant. 18375 18376 `C' 18377 Standard SSE floating point constant. 18378 18379 `e' 18380 32-bit signed integer constant, or a symbolic reference known 18381 to fit that range (for immediate operands in sign-extending 18382 x86-64 instructions). 18383 18384 `Z' 18385 32-bit unsigned integer constant, or a symbolic reference 18386 known to fit that range (for immediate operands in 18387 zero-extending x86-64 instructions). 18388 18389 18390 _Intel IA-64--`config/ia64/ia64.h'_ 18391 18392 `a' 18393 General register `r0' to `r3' for `addl' instruction 18394 18395 `b' 18396 Branch register 18397 18398 `c' 18399 Predicate register (`c' as in "conditional") 18400 18401 `d' 18402 Application register residing in M-unit 18403 18404 `e' 18405 Application register residing in I-unit 18406 18407 `f' 18408 Floating-point register 18409 18410 `m' 18411 Memory operand. Remember that `m' allows postincrement and 18412 postdecrement which require printing with `%Pn' on IA-64. 18413 Use `S' to disallow postincrement and postdecrement. 18414 18415 `G' 18416 Floating-point constant 0.0 or 1.0 18417 18418 `I' 18419 14-bit signed integer constant 18420 18421 `J' 18422 22-bit signed integer constant 18423 18424 `K' 18425 8-bit signed integer constant for logical instructions 18426 18427 `L' 18428 8-bit adjusted signed integer constant for compare pseudo-ops 18429 18430 `M' 18431 6-bit unsigned integer constant for shift counts 18432 18433 `N' 18434 9-bit signed integer constant for load and store 18435 postincrements 18436 18437 `O' 18438 The constant zero 18439 18440 `P' 18441 0 or -1 for `dep' instruction 18442 18443 `Q' 18444 Non-volatile memory for floating-point loads and stores 18445 18446 `R' 18447 Integer constant in the range 1 to 4 for `shladd' instruction 18448 18449 `S' 18450 Memory operand except postincrement and postdecrement 18451 18452 _FRV--`config/frv/frv.h'_ 18453 18454 `a' 18455 Register in the class `ACC_REGS' (`acc0' to `acc7'). 18456 18457 `b' 18458 Register in the class `EVEN_ACC_REGS' (`acc0' to `acc7'). 18459 18460 `c' 18461 Register in the class `CC_REGS' (`fcc0' to `fcc3' and `icc0' 18462 to `icc3'). 18463 18464 `d' 18465 Register in the class `GPR_REGS' (`gr0' to `gr63'). 18466 18467 `e' 18468 Register in the class `EVEN_REGS' (`gr0' to `gr63'). Odd 18469 registers are excluded not in the class but through the use 18470 of a machine mode larger than 4 bytes. 18471 18472 `f' 18473 Register in the class `FPR_REGS' (`fr0' to `fr63'). 18474 18475 `h' 18476 Register in the class `FEVEN_REGS' (`fr0' to `fr63'). Odd 18477 registers are excluded not in the class but through the use 18478 of a machine mode larger than 4 bytes. 18479 18480 `l' 18481 Register in the class `LR_REG' (the `lr' register). 18482 18483 `q' 18484 Register in the class `QUAD_REGS' (`gr2' to `gr63'). 18485 Register numbers not divisible by 4 are excluded not in the 18486 class but through the use of a machine mode larger than 8 18487 bytes. 18488 18489 `t' 18490 Register in the class `ICC_REGS' (`icc0' to `icc3'). 18491 18492 `u' 18493 Register in the class `FCC_REGS' (`fcc0' to `fcc3'). 18494 18495 `v' 18496 Register in the class `ICR_REGS' (`cc4' to `cc7'). 18497 18498 `w' 18499 Register in the class `FCR_REGS' (`cc0' to `cc3'). 18500 18501 `x' 18502 Register in the class `QUAD_FPR_REGS' (`fr0' to `fr63'). 18503 Register numbers not divisible by 4 are excluded not in the 18504 class but through the use of a machine mode larger than 8 18505 bytes. 18506 18507 `z' 18508 Register in the class `SPR_REGS' (`lcr' and `lr'). 18509 18510 `A' 18511 Register in the class `QUAD_ACC_REGS' (`acc0' to `acc7'). 18512 18513 `B' 18514 Register in the class `ACCG_REGS' (`accg0' to `accg7'). 18515 18516 `C' 18517 Register in the class `CR_REGS' (`cc0' to `cc7'). 18518 18519 `G' 18520 Floating point constant zero 18521 18522 `I' 18523 6-bit signed integer constant 18524 18525 `J' 18526 10-bit signed integer constant 18527 18528 `L' 18529 16-bit signed integer constant 18530 18531 `M' 18532 16-bit unsigned integer constant 18533 18534 `N' 18535 12-bit signed integer constant that is negative--i.e. in the 18536 range of -2048 to -1 18537 18538 `O' 18539 Constant zero 18540 18541 `P' 18542 12-bit signed integer constant that is greater than 18543 zero--i.e. in the range of 1 to 2047. 18544 18545 18546 _Blackfin family--`config/bfin/constraints.md'_ 18547 18548 `a' 18549 P register 18550 18551 `d' 18552 D register 18553 18554 `z' 18555 A call clobbered P register. 18556 18557 `qN' 18558 A single register. If N is in the range 0 to 7, the 18559 corresponding D register. If it is `A', then the register P0. 18560 18561 `D' 18562 Even-numbered D register 18563 18564 `W' 18565 Odd-numbered D register 18566 18567 `e' 18568 Accumulator register. 18569 18570 `A' 18571 Even-numbered accumulator register. 18572 18573 `B' 18574 Odd-numbered accumulator register. 18575 18576 `b' 18577 I register 18578 18579 `v' 18580 B register 18581 18582 `f' 18583 M register 18584 18585 `c' 18586 Registers used for circular buffering, i.e. I, B, or L 18587 registers. 18588 18589 `C' 18590 The CC register. 18591 18592 `t' 18593 LT0 or LT1. 18594 18595 `k' 18596 LC0 or LC1. 18597 18598 `u' 18599 LB0 or LB1. 18600 18601 `x' 18602 Any D, P, B, M, I or L register. 18603 18604 `y' 18605 Additional registers typically used only in prologues and 18606 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and 18607 USP. 18608 18609 `w' 18610 Any register except accumulators or CC. 18611 18612 `Ksh' 18613 Signed 16 bit integer (in the range -32768 to 32767) 18614 18615 `Kuh' 18616 Unsigned 16 bit integer (in the range 0 to 65535) 18617 18618 `Ks7' 18619 Signed 7 bit integer (in the range -64 to 63) 18620 18621 `Ku7' 18622 Unsigned 7 bit integer (in the range 0 to 127) 18623 18624 `Ku5' 18625 Unsigned 5 bit integer (in the range 0 to 31) 18626 18627 `Ks4' 18628 Signed 4 bit integer (in the range -8 to 7) 18629 18630 `Ks3' 18631 Signed 3 bit integer (in the range -3 to 4) 18632 18633 `Ku3' 18634 Unsigned 3 bit integer (in the range 0 to 7) 18635 18636 `PN' 18637 Constant N, where N is a single-digit constant in the range 0 18638 to 4. 18639 18640 `PA' 18641 An integer equal to one of the MACFLAG_XXX constants that is 18642 suitable for use with either accumulator. 18643 18644 `PB' 18645 An integer equal to one of the MACFLAG_XXX constants that is 18646 suitable for use only with accumulator A1. 18647 18648 `M1' 18649 Constant 255. 18650 18651 `M2' 18652 Constant 65535. 18653 18654 `J' 18655 An integer constant with exactly a single bit set. 18656 18657 `L' 18658 An integer constant with all bits set except exactly one. 18659 18660 `H' 18661 18662 `Q' 18663 Any SYMBOL_REF. 18664 18665 _M32C--`config/m32c/m32c.c'_ 18666 18667 `Rsp' 18668 `Rfb' 18669 `Rsb' 18670 `$sp', `$fb', `$sb'. 18671 18672 `Rcr' 18673 Any control register, when they're 16 bits wide (nothing if 18674 control registers are 24 bits wide) 18675 18676 `Rcl' 18677 Any control register, when they're 24 bits wide. 18678 18679 `R0w' 18680 `R1w' 18681 `R2w' 18682 `R3w' 18683 $r0, $r1, $r2, $r3. 18684 18685 `R02' 18686 $r0 or $r2, or $r2r0 for 32 bit values. 18687 18688 `R13' 18689 $r1 or $r3, or $r3r1 for 32 bit values. 18690 18691 `Rdi' 18692 A register that can hold a 64 bit value. 18693 18694 `Rhl' 18695 $r0 or $r1 (registers with addressable high/low bytes) 18696 18697 `R23' 18698 $r2 or $r3 18699 18700 `Raa' 18701 Address registers 18702 18703 `Raw' 18704 Address registers when they're 16 bits wide. 18705 18706 `Ral' 18707 Address registers when they're 24 bits wide. 18708 18709 `Rqi' 18710 Registers that can hold QI values. 18711 18712 `Rad' 18713 Registers that can be used with displacements ($a0, $a1, $sb). 18714 18715 `Rsi' 18716 Registers that can hold 32 bit values. 18717 18718 `Rhi' 18719 Registers that can hold 16 bit values. 18720 18721 `Rhc' 18722 Registers chat can hold 16 bit values, including all control 18723 registers. 18724 18725 `Rra' 18726 $r0 through R1, plus $a0 and $a1. 18727 18728 `Rfl' 18729 The flags register. 18730 18731 `Rmm' 18732 The memory-based pseudo-registers $mem0 through $mem15. 18733 18734 `Rpi' 18735 Registers that can hold pointers (16 bit registers for r8c, 18736 m16c; 24 bit registers for m32cm, m32c). 18737 18738 `Rpa' 18739 Matches multiple registers in a PARALLEL to form a larger 18740 register. Used to match function return values. 18741 18742 `Is3' 18743 -8 ... 7 18744 18745 `IS1' 18746 -128 ... 127 18747 18748 `IS2' 18749 -32768 ... 32767 18750 18751 `IU2' 18752 0 ... 65535 18753 18754 `In4' 18755 -8 ... -1 or 1 ... 8 18756 18757 `In5' 18758 -16 ... -1 or 1 ... 16 18759 18760 `In6' 18761 -32 ... -1 or 1 ... 32 18762 18763 `IM2' 18764 -65536 ... -1 18765 18766 `Ilb' 18767 An 8 bit value with exactly one bit set. 18768 18769 `Ilw' 18770 A 16 bit value with exactly one bit set. 18771 18772 `Sd' 18773 The common src/dest memory addressing modes. 18774 18775 `Sa' 18776 Memory addressed using $a0 or $a1. 18777 18778 `Si' 18779 Memory addressed with immediate addresses. 18780 18781 `Ss' 18782 Memory addressed using the stack pointer ($sp). 18783 18784 `Sf' 18785 Memory addressed using the frame base register ($fb). 18786 18787 `Ss' 18788 Memory addressed using the small base register ($sb). 18789 18790 `S1' 18791 $r1h 18792 18793 _MIPS--`config/mips/constraints.md'_ 18794 18795 `d' 18796 An address register. This is equivalent to `r' unless 18797 generating MIPS16 code. 18798 18799 `f' 18800 A floating-point register (if available). 18801 18802 `h' 18803 Formerly the `hi' register. This constraint is no longer 18804 supported. 18805 18806 `l' 18807 The `lo' register. Use this register to store values that are 18808 no bigger than a word. 18809 18810 `x' 18811 The concatenated `hi' and `lo' registers. Use this register 18812 to store doubleword values. 18813 18814 `c' 18815 A register suitable for use in an indirect jump. This will 18816 always be `$25' for `-mabicalls'. 18817 18818 `v' 18819 Register `$3'. Do not use this constraint in new code; it is 18820 retained only for compatibility with glibc. 18821 18822 `y' 18823 Equivalent to `r'; retained for backwards compatibility. 18824 18825 `z' 18826 A floating-point condition code register. 18827 18828 `I' 18829 A signed 16-bit constant (for arithmetic instructions). 18830 18831 `J' 18832 Integer zero. 18833 18834 `K' 18835 An unsigned 16-bit constant (for logic instructions). 18836 18837 `L' 18838 A signed 32-bit constant in which the lower 16 bits are zero. 18839 Such constants can be loaded using `lui'. 18840 18841 `M' 18842 A constant that cannot be loaded using `lui', `addiu' or 18843 `ori'. 18844 18845 `N' 18846 A constant in the range -65535 to -1 (inclusive). 18847 18848 `O' 18849 A signed 15-bit constant. 18850 18851 `P' 18852 A constant in the range 1 to 65535 (inclusive). 18853 18854 `G' 18855 Floating-point zero. 18856 18857 `R' 18858 An address that can be used in a non-macro load or store. 18859 18860 _Motorola 680x0--`config/m68k/constraints.md'_ 18861 18862 `a' 18863 Address register 18864 18865 `d' 18866 Data register 18867 18868 `f' 18869 68881 floating-point register, if available 18870 18871 `I' 18872 Integer in the range 1 to 8 18873 18874 `J' 18875 16-bit signed number 18876 18877 `K' 18878 Signed number whose magnitude is greater than 0x80 18879 18880 `L' 18881 Integer in the range -8 to -1 18882 18883 `M' 18884 Signed number whose magnitude is greater than 0x100 18885 18886 `N' 18887 Range 24 to 31, rotatert:SI 8 to 1 expressed as rotate 18888 18889 `O' 18890 16 (for rotate using swap) 18891 18892 `P' 18893 Range 8 to 15, rotatert:HI 8 to 1 expressed as rotate 18894 18895 `R' 18896 Numbers that mov3q can handle 18897 18898 `G' 18899 Floating point constant that is not a 68881 constant 18900 18901 `S' 18902 Operands that satisfy 'm' when -mpcrel is in effect 18903 18904 `T' 18905 Operands that satisfy 's' when -mpcrel is not in effect 18906 18907 `Q' 18908 Address register indirect addressing mode 18909 18910 `U' 18911 Register offset addressing 18912 18913 `W' 18914 const_call_operand 18915 18916 `Cs' 18917 symbol_ref or const 18918 18919 `Ci' 18920 const_int 18921 18922 `C0' 18923 const_int 0 18924 18925 `Cj' 18926 Range of signed numbers that don't fit in 16 bits 18927 18928 `Cmvq' 18929 Integers valid for mvq 18930 18931 `Capsw' 18932 Integers valid for a moveq followed by a swap 18933 18934 `Cmvz' 18935 Integers valid for mvz 18936 18937 `Cmvs' 18938 Integers valid for mvs 18939 18940 `Ap' 18941 push_operand 18942 18943 `Ac' 18944 Non-register operands allowed in clr 18945 18946 18947 _Motorola 68HC11 & 68HC12 families--`config/m68hc11/m68hc11.h'_ 18948 18949 `a' 18950 Register `a' 18951 18952 `b' 18953 Register `b' 18954 18955 `d' 18956 Register `d' 18957 18958 `q' 18959 An 8-bit register 18960 18961 `t' 18962 Temporary soft register _.tmp 18963 18964 `u' 18965 A soft register _.d1 to _.d31 18966 18967 `w' 18968 Stack pointer register 18969 18970 `x' 18971 Register `x' 18972 18973 `y' 18974 Register `y' 18975 18976 `z' 18977 Pseudo register `z' (replaced by `x' or `y' at the end) 18978 18979 `A' 18980 An address register: x, y or z 18981 18982 `B' 18983 An address register: x or y 18984 18985 `D' 18986 Register pair (x:d) to form a 32-bit value 18987 18988 `L' 18989 Constants in the range -65536 to 65535 18990 18991 `M' 18992 Constants whose 16-bit low part is zero 18993 18994 `N' 18995 Constant integer 1 or -1 18996 18997 `O' 18998 Constant integer 16 18999 19000 `P' 19001 Constants in the range -8 to 2 19002 19003 19004 _SPARC--`config/sparc/sparc.h'_ 19005 19006 `f' 19007 Floating-point register on the SPARC-V8 architecture and 19008 lower floating-point register on the SPARC-V9 architecture. 19009 19010 `e' 19011 Floating-point register. It is equivalent to `f' on the 19012 SPARC-V8 architecture and contains both lower and upper 19013 floating-point registers on the SPARC-V9 architecture. 19014 19015 `c' 19016 Floating-point condition code register. 19017 19018 `d' 19019 Lower floating-point register. It is only valid on the 19020 SPARC-V9 architecture when the Visual Instruction Set is 19021 available. 19022 19023 `b' 19024 Floating-point register. It is only valid on the SPARC-V9 19025 architecture when the Visual Instruction Set is available. 19026 19027 `h' 19028 64-bit global or out register for the SPARC-V8+ architecture. 19029 19030 `D' 19031 A vector constant 19032 19033 `I' 19034 Signed 13-bit constant 19035 19036 `J' 19037 Zero 19038 19039 `K' 19040 32-bit constant with the low 12 bits clear (a constant that 19041 can be loaded with the `sethi' instruction) 19042 19043 `L' 19044 A constant in the range supported by `movcc' instructions 19045 19046 `M' 19047 A constant in the range supported by `movrcc' instructions 19048 19049 `N' 19050 Same as `K', except that it verifies that bits that are not 19051 in the lower 32-bit range are all zero. Must be used instead 19052 of `K' for modes wider than `SImode' 19053 19054 `O' 19055 The constant 4096 19056 19057 `G' 19058 Floating-point zero 19059 19060 `H' 19061 Signed 13-bit constant, sign-extended to 32 or 64 bits 19062 19063 `Q' 19064 Floating-point constant whose integral representation can be 19065 moved into an integer register using a single sethi 19066 instruction 19067 19068 `R' 19069 Floating-point constant whose integral representation can be 19070 moved into an integer register using a single mov instruction 19071 19072 `S' 19073 Floating-point constant whose integral representation can be 19074 moved into an integer register using a high/lo_sum 19075 instruction sequence 19076 19077 `T' 19078 Memory address aligned to an 8-byte boundary 19079 19080 `U' 19081 Even register 19082 19083 `W' 19084 Memory address for `e' constraint registers 19085 19086 `Y' 19087 Vector zero 19088 19089 19090 _SPU--`config/spu/spu.h'_ 19091 19092 `a' 19093 An immediate which can be loaded with the il/ila/ilh/ilhu 19094 instructions. const_int is treated as a 64 bit value. 19095 19096 `c' 19097 An immediate for and/xor/or instructions. const_int is 19098 treated as a 64 bit value. 19099 19100 `d' 19101 An immediate for the `iohl' instruction. const_int is 19102 treated as a 64 bit value. 19103 19104 `f' 19105 An immediate which can be loaded with `fsmbi'. 19106 19107 `A' 19108 An immediate which can be loaded with the il/ila/ilh/ilhu 19109 instructions. const_int is treated as a 32 bit value. 19110 19111 `B' 19112 An immediate for most arithmetic instructions. const_int is 19113 treated as a 32 bit value. 19114 19115 `C' 19116 An immediate for and/xor/or instructions. const_int is 19117 treated as a 32 bit value. 19118 19119 `D' 19120 An immediate for the `iohl' instruction. const_int is 19121 treated as a 32 bit value. 19122 19123 `I' 19124 A constant in the range [-64, 63] for shift/rotate 19125 instructions. 19126 19127 `J' 19128 An unsigned 7-bit constant for conversion/nop/channel 19129 instructions. 19130 19131 `K' 19132 A signed 10-bit constant for most arithmetic instructions. 19133 19134 `M' 19135 A signed 16 bit immediate for `stop'. 19136 19137 `N' 19138 An unsigned 16-bit constant for `iohl' and `fsmbi'. 19139 19140 `O' 19141 An unsigned 7-bit constant whose 3 least significant bits are 19142 0. 19143 19144 `P' 19145 An unsigned 3-bit constant for 16-byte rotates and shifts 19146 19147 `R' 19148 Call operand, reg, for indirect calls 19149 19150 `S' 19151 Call operand, symbol, for relative calls. 19152 19153 `T' 19154 Call operand, const_int, for absolute calls. 19155 19156 `U' 19157 An immediate which can be loaded with the il/ila/ilh/ilhu 19158 instructions. const_int is sign extended to 128 bit. 19159 19160 `W' 19161 An immediate for shift and rotate instructions. const_int is 19162 treated as a 32 bit value. 19163 19164 `Y' 19165 An immediate for and/xor/or instructions. const_int is sign 19166 extended as a 128 bit. 19167 19168 `Z' 19169 An immediate for the `iohl' instruction. const_int is sign 19170 extended to 128 bit. 19171 19172 19173 _S/390 and zSeries--`config/s390/s390.h'_ 19174 19175 `a' 19176 Address register (general purpose register except r0) 19177 19178 `c' 19179 Condition code register 19180 19181 `d' 19182 Data register (arbitrary general purpose register) 19183 19184 `f' 19185 Floating-point register 19186 19187 `I' 19188 Unsigned 8-bit constant (0-255) 19189 19190 `J' 19191 Unsigned 12-bit constant (0-4095) 19192 19193 `K' 19194 Signed 16-bit constant (-32768-32767) 19195 19196 `L' 19197 Value appropriate as displacement. 19198 `(0..4095)' 19199 for short displacement 19200 19201 `(-524288..524287)' 19202 for long displacement 19203 19204 `M' 19205 Constant integer with a value of 0x7fffffff. 19206 19207 `N' 19208 Multiple letter constraint followed by 4 parameter letters. 19209 `0..9:' 19210 number of the part counting from most to least 19211 significant 19212 19213 `H,Q:' 19214 mode of the part 19215 19216 `D,S,H:' 19217 mode of the containing operand 19218 19219 `0,F:' 19220 value of the other parts (F--all bits set) 19221 The constraint matches if the specified part of a constant 19222 has a value different from its other parts. 19223 19224 `Q' 19225 Memory reference without index register and with short 19226 displacement. 19227 19228 `R' 19229 Memory reference with index register and short displacement. 19230 19231 `S' 19232 Memory reference without index register but with long 19233 displacement. 19234 19235 `T' 19236 Memory reference with index register and long displacement. 19237 19238 `U' 19239 Pointer with short displacement. 19240 19241 `W' 19242 Pointer with long displacement. 19243 19244 `Y' 19245 Shift count operand. 19246 19247 19248 _Score family--`config/score/score.h'_ 19249 19250 `d' 19251 Registers from r0 to r32. 19252 19253 `e' 19254 Registers from r0 to r16. 19255 19256 `t' 19257 r8--r11 or r22--r27 registers. 19258 19259 `h' 19260 hi register. 19261 19262 `l' 19263 lo register. 19264 19265 `x' 19266 hi + lo register. 19267 19268 `q' 19269 cnt register. 19270 19271 `y' 19272 lcb register. 19273 19274 `z' 19275 scb register. 19276 19277 `a' 19278 cnt + lcb + scb register. 19279 19280 `c' 19281 cr0--cr15 register. 19282 19283 `b' 19284 cp1 registers. 19285 19286 `f' 19287 cp2 registers. 19288 19289 `i' 19290 cp3 registers. 19291 19292 `j' 19293 cp1 + cp2 + cp3 registers. 19294 19295 `I' 19296 High 16-bit constant (32-bit constant with 16 LSBs zero). 19297 19298 `J' 19299 Unsigned 5 bit integer (in the range 0 to 31). 19300 19301 `K' 19302 Unsigned 16 bit integer (in the range 0 to 65535). 19303 19304 `L' 19305 Signed 16 bit integer (in the range -32768 to 32767). 19306 19307 `M' 19308 Unsigned 14 bit integer (in the range 0 to 16383). 19309 19310 `N' 19311 Signed 14 bit integer (in the range -8192 to 8191). 19312 19313 `Z' 19314 Any SYMBOL_REF. 19315 19316 _Xstormy16--`config/stormy16/stormy16.h'_ 19317 19318 `a' 19319 Register r0. 19320 19321 `b' 19322 Register r1. 19323 19324 `c' 19325 Register r2. 19326 19327 `d' 19328 Register r8. 19329 19330 `e' 19331 Registers r0 through r7. 19332 19333 `t' 19334 Registers r0 and r1. 19335 19336 `y' 19337 The carry register. 19338 19339 `z' 19340 Registers r8 and r9. 19341 19342 `I' 19343 A constant between 0 and 3 inclusive. 19344 19345 `J' 19346 A constant that has exactly one bit set. 19347 19348 `K' 19349 A constant that has exactly one bit clear. 19350 19351 `L' 19352 A constant between 0 and 255 inclusive. 19353 19354 `M' 19355 A constant between -255 and 0 inclusive. 19356 19357 `N' 19358 A constant between -3 and 0 inclusive. 19359 19360 `O' 19361 A constant between 1 and 4 inclusive. 19362 19363 `P' 19364 A constant between -4 and -1 inclusive. 19365 19366 `Q' 19367 A memory reference that is a stack push. 19368 19369 `R' 19370 A memory reference that is a stack pop. 19371 19372 `S' 19373 A memory reference that refers to a constant address of known 19374 value. 19375 19376 `T' 19377 The register indicated by Rx (not implemented yet). 19378 19379 `U' 19380 A constant that is not between 2 and 15 inclusive. 19381 19382 `Z' 19383 The constant 0. 19384 19385 19386 _Xtensa--`config/xtensa/constraints.md'_ 19387 19388 `a' 19389 General-purpose 32-bit register 19390 19391 `b' 19392 One-bit boolean register 19393 19394 `A' 19395 MAC16 40-bit accumulator register 19396 19397 `I' 19398 Signed 12-bit integer constant, for use in MOVI instructions 19399 19400 `J' 19401 Signed 8-bit integer constant, for use in ADDI instructions 19402 19403 `K' 19404 Integer constant valid for BccI instructions 19405 19406 `L' 19407 Unsigned constant valid for BccUI instructions 19408 19409 19410 19411 19412 File: gccint.info, Node: Disable Insn Alternatives, Next: Machine Constraints, Prev: Modifiers, Up: Constraints 19413 19414 16.8.6 Disable insn alternatives using the `enabled' attribute 19415 -------------------------------------------------------------- 19416 19417 The `enabled' insn attribute may be used to disable certain insn 19418 alternatives for machine-specific reasons. This is useful when adding 19419 new instructions to an existing pattern which are only available for 19420 certain cpu architecture levels as specified with the `-march=' option. 19421 19422 If an insn alternative is disabled, then it will never be used. The 19423 compiler treats the constraints for the disabled alternative as 19424 unsatisfiable. 19425 19426 In order to make use of the `enabled' attribute a back end has to add 19427 in the machine description files: 19428 19429 1. A definition of the `enabled' insn attribute. The attribute is 19430 defined as usual using the `define_attr' command. This definition 19431 should be based on other insn attributes and/or target flags. The 19432 `enabled' attribute is a numeric attribute and should evaluate to 19433 `(const_int 1)' for an enabled alternative and to `(const_int 0)' 19434 otherwise. 19435 19436 2. A definition of another insn attribute used to describe for what 19437 reason an insn alternative might be available or not. E.g. 19438 `cpu_facility' as in the example below. 19439 19440 3. An assignment for the second attribute to each insn definition 19441 combining instructions which are not all available under the same 19442 circumstances. (Note: It obviously only makes sense for 19443 definitions with more than one alternative. Otherwise the insn 19444 pattern should be disabled or enabled using the insn condition.) 19445 19446 E.g. the following two patterns could easily be merged using the 19447 `enabled' attribute: 19448 19449 19450 (define_insn "*movdi_old" 19451 [(set (match_operand:DI 0 "register_operand" "=d") 19452 (match_operand:DI 1 "register_operand" " d"))] 19453 "!TARGET_NEW" 19454 "lgr %0,%1") 19455 19456 (define_insn "*movdi_new" 19457 [(set (match_operand:DI 0 "register_operand" "=d,f,d") 19458 (match_operand:DI 1 "register_operand" " d,d,f"))] 19459 "TARGET_NEW" 19460 "@ 19461 lgr %0,%1 19462 ldgr %0,%1 19463 lgdr %0,%1") 19464 19465 to: 19466 19467 19468 (define_insn "*movdi_combined" 19469 [(set (match_operand:DI 0 "register_operand" "=d,f,d") 19470 (match_operand:DI 1 "register_operand" " d,d,f"))] 19471 "" 19472 "@ 19473 lgr %0,%1 19474 ldgr %0,%1 19475 lgdr %0,%1" 19476 [(set_attr "cpu_facility" "*,new,new")]) 19477 19478 with the `enabled' attribute defined like this: 19479 19480 19481 (define_attr "cpu_facility" "standard,new" (const_string "standard")) 19482 19483 (define_attr "enabled" "" 19484 (cond [(eq_attr "cpu_facility" "standard") (const_int 1) 19485 (and (eq_attr "cpu_facility" "new") 19486 (ne (symbol_ref "TARGET_NEW") (const_int 0))) 19487 (const_int 1)] 19488 (const_int 0))) 19489 19490 19491 File: gccint.info, Node: Define Constraints, Next: C Constraint Interface, Prev: Machine Constraints, Up: Constraints 19492 19493 16.8.7 Defining Machine-Specific Constraints 19494 -------------------------------------------- 19495 19496 Machine-specific constraints fall into two categories: register and 19497 non-register constraints. Within the latter category, constraints 19498 which allow subsets of all possible memory or address operands should 19499 be specially marked, to give `reload' more information. 19500 19501 Machine-specific constraints can be given names of arbitrary length, 19502 but they must be entirely composed of letters, digits, underscores 19503 (`_'), and angle brackets (`< >'). Like C identifiers, they must begin 19504 with a letter or underscore. 19505 19506 In order to avoid ambiguity in operand constraint strings, no 19507 constraint can have a name that begins with any other constraint's 19508 name. For example, if `x' is defined as a constraint name, `xy' may 19509 not be, and vice versa. As a consequence of this rule, no constraint 19510 may begin with one of the generic constraint letters: `E F V X g i m n 19511 o p r s'. 19512 19513 Register constraints correspond directly to register classes. *Note 19514 Register Classes::. There is thus not much flexibility in their 19515 definitions. 19516 19517 -- MD Expression: define_register_constraint name regclass docstring 19518 All three arguments are string constants. NAME is the name of the 19519 constraint, as it will appear in `match_operand' expressions. If 19520 NAME is a multi-letter constraint its length shall be the same for 19521 all constraints starting with the same letter. REGCLASS can be 19522 either the name of the corresponding register class (*note 19523 Register Classes::), or a C expression which evaluates to the 19524 appropriate register class. If it is an expression, it must have 19525 no side effects, and it cannot look at the operand. The usual use 19526 of expressions is to map some register constraints to `NO_REGS' 19527 when the register class is not available on a given 19528 subarchitecture. 19529 19530 DOCSTRING is a sentence documenting the meaning of the constraint. 19531 Docstrings are explained further below. 19532 19533 Non-register constraints are more like predicates: the constraint 19534 definition gives a Boolean expression which indicates whether the 19535 constraint matches. 19536 19537 -- MD Expression: define_constraint name docstring exp 19538 The NAME and DOCSTRING arguments are the same as for 19539 `define_register_constraint', but note that the docstring comes 19540 immediately after the name for these expressions. EXP is an RTL 19541 expression, obeying the same rules as the RTL expressions in 19542 predicate definitions. *Note Defining Predicates::, for details. 19543 If it evaluates true, the constraint matches; if it evaluates 19544 false, it doesn't. Constraint expressions should indicate which 19545 RTL codes they might match, just like predicate expressions. 19546 19547 `match_test' C expressions have access to the following variables: 19548 19549 OP 19550 The RTL object defining the operand. 19551 19552 MODE 19553 The machine mode of OP. 19554 19555 IVAL 19556 `INTVAL (OP)', if OP is a `const_int'. 19557 19558 HVAL 19559 `CONST_DOUBLE_HIGH (OP)', if OP is an integer `const_double'. 19560 19561 LVAL 19562 `CONST_DOUBLE_LOW (OP)', if OP is an integer `const_double'. 19563 19564 RVAL 19565 `CONST_DOUBLE_REAL_VALUE (OP)', if OP is a floating-point 19566 `const_double'. 19567 19568 The *VAL variables should only be used once another piece of the 19569 expression has verified that OP is the appropriate kind of RTL 19570 object. 19571 19572 Most non-register constraints should be defined with 19573 `define_constraint'. The remaining two definition expressions are only 19574 appropriate for constraints that should be handled specially by 19575 `reload' if they fail to match. 19576 19577 -- MD Expression: define_memory_constraint name docstring exp 19578 Use this expression for constraints that match a subset of all 19579 memory operands: that is, `reload' can make them match by 19580 converting the operand to the form `(mem (reg X))', where X is a 19581 base register (from the register class specified by 19582 `BASE_REG_CLASS', *note Register Classes::). 19583 19584 For example, on the S/390, some instructions do not accept 19585 arbitrary memory references, but only those that do not make use 19586 of an index register. The constraint letter `Q' is defined to 19587 represent a memory address of this type. If `Q' is defined with 19588 `define_memory_constraint', a `Q' constraint can handle any memory 19589 operand, because `reload' knows it can simply copy the memory 19590 address into a base register if required. This is analogous to 19591 the way a `o' constraint can handle any memory operand. 19592 19593 The syntax and semantics are otherwise identical to 19594 `define_constraint'. 19595 19596 -- MD Expression: define_address_constraint name docstring exp 19597 Use this expression for constraints that match a subset of all 19598 address operands: that is, `reload' can make the constraint match 19599 by converting the operand to the form `(reg X)', again with X a 19600 base register. 19601 19602 Constraints defined with `define_address_constraint' can only be 19603 used with the `address_operand' predicate, or machine-specific 19604 predicates that work the same way. They are treated analogously to 19605 the generic `p' constraint. 19606 19607 The syntax and semantics are otherwise identical to 19608 `define_constraint'. 19609 19610 For historical reasons, names beginning with the letters `G H' are 19611 reserved for constraints that match only `const_double's, and names 19612 beginning with the letters `I J K L M N O P' are reserved for 19613 constraints that match only `const_int's. This may change in the 19614 future. For the time being, constraints with these names must be 19615 written in a stylized form, so that `genpreds' can tell you did it 19616 correctly: 19617 19618 (define_constraint "[GHIJKLMNOP]..." 19619 "DOC..." 19620 (and (match_code "const_int") ; `const_double' for G/H 19621 CONDITION...)) ; usually a `match_test' 19622 19623 It is fine to use names beginning with other letters for constraints 19624 that match `const_double's or `const_int's. 19625 19626 Each docstring in a constraint definition should be one or more 19627 complete sentences, marked up in Texinfo format. _They are currently 19628 unused._ In the future they will be copied into the GCC manual, in 19629 *note Machine Constraints::, replacing the hand-maintained tables 19630 currently found in that section. Also, in the future the compiler may 19631 use this to give more helpful diagnostics when poor choice of `asm' 19632 constraints causes a reload failure. 19633 19634 If you put the pseudo-Texinfo directive `@internal' at the beginning 19635 of a docstring, then (in the future) it will appear only in the 19636 internals manual's version of the machine-specific constraint tables. 19637 Use this for constraints that should not appear in `asm' statements. 19638 19639 19640 File: gccint.info, Node: C Constraint Interface, Prev: Define Constraints, Up: Constraints 19641 19642 16.8.8 Testing constraints from C 19643 --------------------------------- 19644 19645 It is occasionally useful to test a constraint from C code rather than 19646 implicitly via the constraint string in a `match_operand'. The 19647 generated file `tm_p.h' declares a few interfaces for working with 19648 machine-specific constraints. None of these interfaces work with the 19649 generic constraints described in *note Simple Constraints::. This may 19650 change in the future. 19651 19652 *Warning:* `tm_p.h' may declare other functions that operate on 19653 constraints, besides the ones documented here. Do not use those 19654 functions from machine-dependent code. They exist to implement the old 19655 constraint interface that machine-independent components of the 19656 compiler still expect. They will change or disappear in the future. 19657 19658 Some valid constraint names are not valid C identifiers, so there is a 19659 mangling scheme for referring to them from C. Constraint names that do 19660 not contain angle brackets or underscores are left unchanged. 19661 Underscores are doubled, each `<' is replaced with `_l', and each `>' 19662 with `_g'. Here are some examples: 19663 19664 *Original* *Mangled* 19665 `x' `x' 19666 `P42x' `P42x' 19667 `P4_x' `P4__x' 19668 `P4>x' `P4_gx' 19669 `P4>>' `P4_g_g' 19670 `P4_g>' `P4__g_g' 19671 19672 Throughout this section, the variable C is either a constraint in the 19673 abstract sense, or a constant from `enum constraint_num'; the variable 19674 M is a mangled constraint name (usually as part of a larger identifier). 19675 19676 -- Enum: constraint_num 19677 For each machine-specific constraint, there is a corresponding 19678 enumeration constant: `CONSTRAINT_' plus the mangled name of the 19679 constraint. Functions that take an `enum constraint_num' as an 19680 argument expect one of these constants. 19681 19682 Machine-independent constraints do not have associated constants. 19683 This may change in the future. 19684 19685 -- Function: inline bool satisfies_constraint_M (rtx EXP) 19686 For each machine-specific, non-register constraint M, there is one 19687 of these functions; it returns `true' if EXP satisfies the 19688 constraint. These functions are only visible if `rtl.h' was 19689 included before `tm_p.h'. 19690 19691 -- Function: bool constraint_satisfied_p (rtx EXP, enum constraint_num 19692 C) 19693 Like the `satisfies_constraint_M' functions, but the constraint to 19694 test is given as an argument, C. If C specifies a register 19695 constraint, this function will always return `false'. 19696 19697 -- Function: enum reg_class regclass_for_constraint (enum 19698 constraint_num C) 19699 Returns the register class associated with C. If C is not a 19700 register constraint, or those registers are not available for the 19701 currently selected subtarget, returns `NO_REGS'. 19702 19703 Here is an example use of `satisfies_constraint_M'. In peephole 19704 optimizations (*note Peephole Definitions::), operand constraint 19705 strings are ignored, so if there are relevant constraints, they must be 19706 tested in the C condition. In the example, the optimization is applied 19707 if operand 2 does _not_ satisfy the `K' constraint. (This is a 19708 simplified version of a peephole definition from the i386 machine 19709 description.) 19710 19711 (define_peephole2 19712 [(match_scratch:SI 3 "r") 19713 (set (match_operand:SI 0 "register_operand" "") 19714 (mult:SI (match_operand:SI 1 "memory_operand" "") 19715 (match_operand:SI 2 "immediate_operand" "")))] 19716 19717 "!satisfies_constraint_K (operands[2])" 19718 19719 [(set (match_dup 3) (match_dup 1)) 19720 (set (match_dup 0) (mult:SI (match_dup 3) (match_dup 2)))] 19721 19722 "") 19723 19724 19725 File: gccint.info, Node: Standard Names, Next: Pattern Ordering, Prev: Constraints, Up: Machine Desc 19726 19727 16.9 Standard Pattern Names For Generation 19728 ========================================== 19729 19730 Here is a table of the instruction names that are meaningful in the RTL 19731 generation pass of the compiler. Giving one of these names to an 19732 instruction pattern tells the RTL generation pass that it can use the 19733 pattern to accomplish a certain task. 19734 19735 `movM' 19736 Here M stands for a two-letter machine mode name, in lowercase. 19737 This instruction pattern moves data with that machine mode from 19738 operand 1 to operand 0. For example, `movsi' moves full-word data. 19739 19740 If operand 0 is a `subreg' with mode M of a register whose own 19741 mode is wider than M, the effect of this instruction is to store 19742 the specified value in the part of the register that corresponds 19743 to mode M. Bits outside of M, but which are within the same 19744 target word as the `subreg' are undefined. Bits which are outside 19745 the target word are left unchanged. 19746 19747 This class of patterns is special in several ways. First of all, 19748 each of these names up to and including full word size _must_ be 19749 defined, because there is no other way to copy a datum from one 19750 place to another. If there are patterns accepting operands in 19751 larger modes, `movM' must be defined for integer modes of those 19752 sizes. 19753 19754 Second, these patterns are not used solely in the RTL generation 19755 pass. Even the reload pass can generate move insns to copy values 19756 from stack slots into temporary registers. When it does so, one 19757 of the operands is a hard register and the other is an operand 19758 that can need to be reloaded into a register. 19759 19760 Therefore, when given such a pair of operands, the pattern must 19761 generate RTL which needs no reloading and needs no temporary 19762 registers--no registers other than the operands. For example, if 19763 you support the pattern with a `define_expand', then in such a 19764 case the `define_expand' mustn't call `force_reg' or any other such 19765 function which might generate new pseudo registers. 19766 19767 This requirement exists even for subword modes on a RISC machine 19768 where fetching those modes from memory normally requires several 19769 insns and some temporary registers. 19770 19771 During reload a memory reference with an invalid address may be 19772 passed as an operand. Such an address will be replaced with a 19773 valid address later in the reload pass. In this case, nothing may 19774 be done with the address except to use it as it stands. If it is 19775 copied, it will not be replaced with a valid address. No attempt 19776 should be made to make such an address into a valid address and no 19777 routine (such as `change_address') that will do so may be called. 19778 Note that `general_operand' will fail when applied to such an 19779 address. 19780 19781 The global variable `reload_in_progress' (which must be explicitly 19782 declared if required) can be used to determine whether such special 19783 handling is required. 19784 19785 The variety of operands that have reloads depends on the rest of 19786 the machine description, but typically on a RISC machine these can 19787 only be pseudo registers that did not get hard registers, while on 19788 other machines explicit memory references will get optional 19789 reloads. 19790 19791 If a scratch register is required to move an object to or from 19792 memory, it can be allocated using `gen_reg_rtx' prior to life 19793 analysis. 19794 19795 If there are cases which need scratch registers during or after 19796 reload, you must provide an appropriate secondary_reload target 19797 hook. 19798 19799 The macro `can_create_pseudo_p' can be used to determine if it is 19800 unsafe to create new pseudo registers. If this variable is 19801 nonzero, then it is unsafe to call `gen_reg_rtx' to allocate a new 19802 pseudo. 19803 19804 The constraints on a `movM' must permit moving any hard register 19805 to any other hard register provided that `HARD_REGNO_MODE_OK' 19806 permits mode M in both registers and `REGISTER_MOVE_COST' applied 19807 to their classes returns a value of 2. 19808 19809 It is obligatory to support floating point `movM' instructions 19810 into and out of any registers that can hold fixed point values, 19811 because unions and structures (which have modes `SImode' or 19812 `DImode') can be in those registers and they may have floating 19813 point members. 19814 19815 There may also be a need to support fixed point `movM' 19816 instructions in and out of floating point registers. 19817 Unfortunately, I have forgotten why this was so, and I don't know 19818 whether it is still true. If `HARD_REGNO_MODE_OK' rejects fixed 19819 point values in floating point registers, then the constraints of 19820 the fixed point `movM' instructions must be designed to avoid ever 19821 trying to reload into a floating point register. 19822 19823 `reload_inM' 19824 `reload_outM' 19825 These named patterns have been obsoleted by the target hook 19826 `secondary_reload'. 19827 19828 Like `movM', but used when a scratch register is required to move 19829 between operand 0 and operand 1. Operand 2 describes the scratch 19830 register. See the discussion of the `SECONDARY_RELOAD_CLASS' 19831 macro in *note Register Classes::. 19832 19833 There are special restrictions on the form of the `match_operand's 19834 used in these patterns. First, only the predicate for the reload 19835 operand is examined, i.e., `reload_in' examines operand 1, but not 19836 the predicates for operand 0 or 2. Second, there may be only one 19837 alternative in the constraints. Third, only a single register 19838 class letter may be used for the constraint; subsequent constraint 19839 letters are ignored. As a special exception, an empty constraint 19840 string matches the `ALL_REGS' register class. This may relieve 19841 ports of the burden of defining an `ALL_REGS' constraint letter 19842 just for these patterns. 19843 19844 `movstrictM' 19845 Like `movM' except that if operand 0 is a `subreg' with mode M of 19846 a register whose natural mode is wider, the `movstrictM' 19847 instruction is guaranteed not to alter any of the register except 19848 the part which belongs to mode M. 19849 19850 `movmisalignM' 19851 This variant of a move pattern is designed to load or store a value 19852 from a memory address that is not naturally aligned for its mode. 19853 For a store, the memory will be in operand 0; for a load, the 19854 memory will be in operand 1. The other operand is guaranteed not 19855 to be a memory, so that it's easy to tell whether this is a load 19856 or store. 19857 19858 This pattern is used by the autovectorizer, and when expanding a 19859 `MISALIGNED_INDIRECT_REF' expression. 19860 19861 `load_multiple' 19862 Load several consecutive memory locations into consecutive 19863 registers. Operand 0 is the first of the consecutive registers, 19864 operand 1 is the first memory location, and operand 2 is a 19865 constant: the number of consecutive registers. 19866 19867 Define this only if the target machine really has such an 19868 instruction; do not define this if the most efficient way of 19869 loading consecutive registers from memory is to do them one at a 19870 time. 19871 19872 On some machines, there are restrictions as to which consecutive 19873 registers can be stored into memory, such as particular starting or 19874 ending register numbers or only a range of valid counts. For those 19875 machines, use a `define_expand' (*note Expander Definitions::) and 19876 make the pattern fail if the restrictions are not met. 19877 19878 Write the generated insn as a `parallel' with elements being a 19879 `set' of one register from the appropriate memory location (you may 19880 also need `use' or `clobber' elements). Use a `match_parallel' 19881 (*note RTL Template::) to recognize the insn. See `rs6000.md' for 19882 examples of the use of this insn pattern. 19883 19884 `store_multiple' 19885 Similar to `load_multiple', but store several consecutive registers 19886 into consecutive memory locations. Operand 0 is the first of the 19887 consecutive memory locations, operand 1 is the first register, and 19888 operand 2 is a constant: the number of consecutive registers. 19889 19890 `vec_setM' 19891 Set given field in the vector value. Operand 0 is the vector to 19892 modify, operand 1 is new value of field and operand 2 specify the 19893 field index. 19894 19895 `vec_extractM' 19896 Extract given field from the vector value. Operand 1 is the 19897 vector, operand 2 specify field index and operand 0 place to store 19898 value into. 19899 19900 `vec_extract_evenM' 19901 Extract even elements from the input vectors (operand 1 and 19902 operand 2). The even elements of operand 2 are concatenated to 19903 the even elements of operand 1 in their original order. The result 19904 is stored in operand 0. The output and input vectors should have 19905 the same modes. 19906 19907 `vec_extract_oddM' 19908 Extract odd elements from the input vectors (operand 1 and operand 19909 2). The odd elements of operand 2 are concatenated to the odd 19910 elements of operand 1 in their original order. The result is 19911 stored in operand 0. The output and input vectors should have the 19912 same modes. 19913 19914 `vec_interleave_highM' 19915 Merge high elements of the two input vectors into the output 19916 vector. The output and input vectors should have the same modes 19917 (`N' elements). The high `N/2' elements of the first input vector 19918 are interleaved with the high `N/2' elements of the second input 19919 vector. 19920 19921 `vec_interleave_lowM' 19922 Merge low elements of the two input vectors into the output 19923 vector. The output and input vectors should have the same modes 19924 (`N' elements). The low `N/2' elements of the first input vector 19925 are interleaved with the low `N/2' elements of the second input 19926 vector. 19927 19928 `vec_initM' 19929 Initialize the vector to given values. Operand 0 is the vector to 19930 initialize and operand 1 is parallel containing values for 19931 individual fields. 19932 19933 `pushM1' 19934 Output a push instruction. Operand 0 is value to push. Used only 19935 when `PUSH_ROUNDING' is defined. For historical reason, this 19936 pattern may be missing and in such case an `mov' expander is used 19937 instead, with a `MEM' expression forming the push operation. The 19938 `mov' expander method is deprecated. 19939 19940 `addM3' 19941 Add operand 2 and operand 1, storing the result in operand 0. All 19942 operands must have mode M. This can be used even on two-address 19943 machines, by means of constraints requiring operands 1 and 0 to be 19944 the same location. 19945 19946 `ssaddM3', `usaddM3' 19947 19948 `subM3', `sssubM3', `ussubM3' 19949 19950 `mulM3', `ssmulM3', `usmulM3' 19951 `divM3', `ssdivM3' 19952 `udivM3', `usdivM3' 19953 `modM3', `umodM3' 19954 `uminM3', `umaxM3' 19955 `andM3', `iorM3', `xorM3' 19956 Similar, for other arithmetic operations. 19957 19958 `sminM3', `smaxM3' 19959 Signed minimum and maximum operations. When used with floating 19960 point, if both operands are zeros, or if either operand is `NaN', 19961 then it is unspecified which of the two operands is returned as 19962 the result. 19963 19964 `reduc_smin_M', `reduc_smax_M' 19965 Find the signed minimum/maximum of the elements of a vector. The 19966 vector is operand 1, and the scalar result is stored in the least 19967 significant bits of operand 0 (also a vector). The output and 19968 input vector should have the same modes. 19969 19970 `reduc_umin_M', `reduc_umax_M' 19971 Find the unsigned minimum/maximum of the elements of a vector. The 19972 vector is operand 1, and the scalar result is stored in the least 19973 significant bits of operand 0 (also a vector). The output and 19974 input vector should have the same modes. 19975 19976 `reduc_splus_M' 19977 Compute the sum of the signed elements of a vector. The vector is 19978 operand 1, and the scalar result is stored in the least 19979 significant bits of operand 0 (also a vector). The output and 19980 input vector should have the same modes. 19981 19982 `reduc_uplus_M' 19983 Compute the sum of the unsigned elements of a vector. The vector 19984 is operand 1, and the scalar result is stored in the least 19985 significant bits of operand 0 (also a vector). The output and 19986 input vector should have the same modes. 19987 19988 `sdot_prodM' 19989 19990 `udot_prodM' 19991 Compute the sum of the products of two signed/unsigned elements. 19992 Operand 1 and operand 2 are of the same mode. Their product, which 19993 is of a wider mode, is computed and added to operand 3. Operand 3 19994 is of a mode equal or wider than the mode of the product. The 19995 result is placed in operand 0, which is of the same mode as 19996 operand 3. 19997 19998 `ssum_widenM3' 19999 20000 `usum_widenM3' 20001 Operands 0 and 2 are of the same mode, which is wider than the 20002 mode of operand 1. Add operand 1 to operand 2 and place the 20003 widened result in operand 0. (This is used express accumulation of 20004 elements into an accumulator of a wider mode.) 20005 20006 `vec_shl_M', `vec_shr_M' 20007 Whole vector left/right shift in bits. Operand 1 is a vector to 20008 be shifted. Operand 2 is an integer shift amount in bits. 20009 Operand 0 is where the resulting shifted vector is stored. The 20010 output and input vectors should have the same modes. 20011 20012 `vec_pack_trunc_M' 20013 Narrow (demote) and merge the elements of two vectors. Operands 1 20014 and 2 are vectors of the same mode having N integral or floating 20015 point elements of size S. Operand 0 is the resulting vector in 20016 which 2*N elements of size N/2 are concatenated after narrowing 20017 them down using truncation. 20018 20019 `vec_pack_ssat_M', `vec_pack_usat_M' 20020 Narrow (demote) and merge the elements of two vectors. Operands 1 20021 and 2 are vectors of the same mode having N integral elements of 20022 size S. Operand 0 is the resulting vector in which the elements 20023 of the two input vectors are concatenated after narrowing them 20024 down using signed/unsigned saturating arithmetic. 20025 20026 `vec_pack_sfix_trunc_M', `vec_pack_ufix_trunc_M' 20027 Narrow, convert to signed/unsigned integral type and merge the 20028 elements of two vectors. Operands 1 and 2 are vectors of the same 20029 mode having N floating point elements of size S. Operand 0 is the 20030 resulting vector in which 2*N elements of size N/2 are 20031 concatenated. 20032 20033 `vec_unpacks_hi_M', `vec_unpacks_lo_M' 20034 Extract and widen (promote) the high/low part of a vector of signed 20035 integral or floating point elements. The input vector (operand 1) 20036 has N elements of size S. Widen (promote) the high/low elements 20037 of the vector using signed or floating point extension and place 20038 the resulting N/2 values of size 2*S in the output vector (operand 20039 0). 20040 20041 `vec_unpacku_hi_M', `vec_unpacku_lo_M' 20042 Extract and widen (promote) the high/low part of a vector of 20043 unsigned integral elements. The input vector (operand 1) has N 20044 elements of size S. Widen (promote) the high/low elements of the 20045 vector using zero extension and place the resulting N/2 values of 20046 size 2*S in the output vector (operand 0). 20047 20048 `vec_unpacks_float_hi_M', `vec_unpacks_float_lo_M' 20049 `vec_unpacku_float_hi_M', `vec_unpacku_float_lo_M' 20050 Extract, convert to floating point type and widen the high/low 20051 part of a vector of signed/unsigned integral elements. The input 20052 vector (operand 1) has N elements of size S. Convert the high/low 20053 elements of the vector using floating point conversion and place 20054 the resulting N/2 values of size 2*S in the output vector (operand 20055 0). 20056 20057 `vec_widen_umult_hi_M', `vec_widen_umult_lo_M' 20058 `vec_widen_smult_hi_M', `vec_widen_smult_lo_M' 20059 Signed/Unsigned widening multiplication. The two inputs (operands 20060 1 and 2) are vectors with N signed/unsigned elements of size S. 20061 Multiply the high/low elements of the two vectors, and put the N/2 20062 products of size 2*S in the output vector (operand 0). 20063 20064 `mulhisi3' 20065 Multiply operands 1 and 2, which have mode `HImode', and store a 20066 `SImode' product in operand 0. 20067 20068 `mulqihi3', `mulsidi3' 20069 Similar widening-multiplication instructions of other widths. 20070 20071 `umulqihi3', `umulhisi3', `umulsidi3' 20072 Similar widening-multiplication instructions that do unsigned 20073 multiplication. 20074 20075 `usmulqihi3', `usmulhisi3', `usmulsidi3' 20076 Similar widening-multiplication instructions that interpret the 20077 first operand as unsigned and the second operand as signed, then 20078 do a signed multiplication. 20079 20080 `smulM3_highpart' 20081 Perform a signed multiplication of operands 1 and 2, which have 20082 mode M, and store the most significant half of the product in 20083 operand 0. The least significant half of the product is discarded. 20084 20085 `umulM3_highpart' 20086 Similar, but the multiplication is unsigned. 20087 20088 `maddMN4' 20089 Multiply operands 1 and 2, sign-extend them to mode N, add operand 20090 3, and store the result in operand 0. Operands 1 and 2 have mode 20091 M and operands 0 and 3 have mode N. Both modes must be integer or 20092 fixed-point modes and N must be twice the size of M. 20093 20094 In other words, `maddMN4' is like `mulMN3' except that it also 20095 adds operand 3. 20096 20097 These instructions are not allowed to `FAIL'. 20098 20099 `umaddMN4' 20100 Like `maddMN4', but zero-extend the multiplication operands 20101 instead of sign-extending them. 20102 20103 `ssmaddMN4' 20104 Like `maddMN4', but all involved operations must be 20105 signed-saturating. 20106 20107 `usmaddMN4' 20108 Like `umaddMN4', but all involved operations must be 20109 unsigned-saturating. 20110 20111 `msubMN4' 20112 Multiply operands 1 and 2, sign-extend them to mode N, subtract the 20113 result from operand 3, and store the result in operand 0. 20114 Operands 1 and 2 have mode M and operands 0 and 3 have mode N. 20115 Both modes must be integer or fixed-point modes and N must be twice 20116 the size of M. 20117 20118 In other words, `msubMN4' is like `mulMN3' except that it also 20119 subtracts the result from operand 3. 20120 20121 These instructions are not allowed to `FAIL'. 20122 20123 `umsubMN4' 20124 Like `msubMN4', but zero-extend the multiplication operands 20125 instead of sign-extending them. 20126 20127 `ssmsubMN4' 20128 Like `msubMN4', but all involved operations must be 20129 signed-saturating. 20130 20131 `usmsubMN4' 20132 Like `umsubMN4', but all involved operations must be 20133 unsigned-saturating. 20134 20135 `divmodM4' 20136 Signed division that produces both a quotient and a remainder. 20137 Operand 1 is divided by operand 2 to produce a quotient stored in 20138 operand 0 and a remainder stored in operand 3. 20139 20140 For machines with an instruction that produces both a quotient and 20141 a remainder, provide a pattern for `divmodM4' but do not provide 20142 patterns for `divM3' and `modM3'. This allows optimization in the 20143 relatively common case when both the quotient and remainder are 20144 computed. 20145 20146 If an instruction that just produces a quotient or just a remainder 20147 exists and is more efficient than the instruction that produces 20148 both, write the output routine of `divmodM4' to call 20149 `find_reg_note' and look for a `REG_UNUSED' note on the quotient 20150 or remainder and generate the appropriate instruction. 20151 20152 `udivmodM4' 20153 Similar, but does unsigned division. 20154 20155 `ashlM3', `ssashlM3', `usashlM3' 20156 Arithmetic-shift operand 1 left by a number of bits specified by 20157 operand 2, and store the result in operand 0. Here M is the mode 20158 of operand 0 and operand 1; operand 2's mode is specified by the 20159 instruction pattern, and the compiler will convert the operand to 20160 that mode before generating the instruction. The meaning of 20161 out-of-range shift counts can optionally be specified by 20162 `TARGET_SHIFT_TRUNCATION_MASK'. *Note 20163 TARGET_SHIFT_TRUNCATION_MASK::. Operand 2 is always a scalar type. 20164 20165 `ashrM3', `lshrM3', `rotlM3', `rotrM3' 20166 Other shift and rotate instructions, analogous to the `ashlM3' 20167 instructions. Operand 2 is always a scalar type. 20168 20169 `vashlM3', `vashrM3', `vlshrM3', `vrotlM3', `vrotrM3' 20170 Vector shift and rotate instructions that take vectors as operand 2 20171 instead of a scalar type. 20172 20173 `negM2', `ssnegM2', `usnegM2' 20174 Negate operand 1 and store the result in operand 0. 20175 20176 `absM2' 20177 Store the absolute value of operand 1 into operand 0. 20178 20179 `sqrtM2' 20180 Store the square root of operand 1 into operand 0. 20181 20182 The `sqrt' built-in function of C always uses the mode which 20183 corresponds to the C data type `double' and the `sqrtf' built-in 20184 function uses the mode which corresponds to the C data type 20185 `float'. 20186 20187 `fmodM3' 20188 Store the remainder of dividing operand 1 by operand 2 into 20189 operand 0, rounded towards zero to an integer. 20190 20191 The `fmod' built-in function of C always uses the mode which 20192 corresponds to the C data type `double' and the `fmodf' built-in 20193 function uses the mode which corresponds to the C data type 20194 `float'. 20195 20196 `remainderM3' 20197 Store the remainder of dividing operand 1 by operand 2 into 20198 operand 0, rounded to the nearest integer. 20199 20200 The `remainder' built-in function of C always uses the mode which 20201 corresponds to the C data type `double' and the `remainderf' 20202 built-in function uses the mode which corresponds to the C data 20203 type `float'. 20204 20205 `cosM2' 20206 Store the cosine of operand 1 into operand 0. 20207 20208 The `cos' built-in function of C always uses the mode which 20209 corresponds to the C data type `double' and the `cosf' built-in 20210 function uses the mode which corresponds to the C data type 20211 `float'. 20212 20213 `sinM2' 20214 Store the sine of operand 1 into operand 0. 20215 20216 The `sin' built-in function of C always uses the mode which 20217 corresponds to the C data type `double' and the `sinf' built-in 20218 function uses the mode which corresponds to the C data type 20219 `float'. 20220 20221 `expM2' 20222 Store the exponential of operand 1 into operand 0. 20223 20224 The `exp' built-in function of C always uses the mode which 20225 corresponds to the C data type `double' and the `expf' built-in 20226 function uses the mode which corresponds to the C data type 20227 `float'. 20228 20229 `logM2' 20230 Store the natural logarithm of operand 1 into operand 0. 20231 20232 The `log' built-in function of C always uses the mode which 20233 corresponds to the C data type `double' and the `logf' built-in 20234 function uses the mode which corresponds to the C data type 20235 `float'. 20236 20237 `powM3' 20238 Store the value of operand 1 raised to the exponent operand 2 into 20239 operand 0. 20240 20241 The `pow' built-in function of C always uses the mode which 20242 corresponds to the C data type `double' and the `powf' built-in 20243 function uses the mode which corresponds to the C data type 20244 `float'. 20245 20246 `atan2M3' 20247 Store the arc tangent (inverse tangent) of operand 1 divided by 20248 operand 2 into operand 0, using the signs of both arguments to 20249 determine the quadrant of the result. 20250 20251 The `atan2' built-in function of C always uses the mode which 20252 corresponds to the C data type `double' and the `atan2f' built-in 20253 function uses the mode which corresponds to the C data type 20254 `float'. 20255 20256 `floorM2' 20257 Store the largest integral value not greater than argument. 20258 20259 The `floor' built-in function of C always uses the mode which 20260 corresponds to the C data type `double' and the `floorf' built-in 20261 function uses the mode which corresponds to the C data type 20262 `float'. 20263 20264 `btruncM2' 20265 Store the argument rounded to integer towards zero. 20266 20267 The `trunc' built-in function of C always uses the mode which 20268 corresponds to the C data type `double' and the `truncf' built-in 20269 function uses the mode which corresponds to the C data type 20270 `float'. 20271 20272 `roundM2' 20273 Store the argument rounded to integer away from zero. 20274 20275 The `round' built-in function of C always uses the mode which 20276 corresponds to the C data type `double' and the `roundf' built-in 20277 function uses the mode which corresponds to the C data type 20278 `float'. 20279 20280 `ceilM2' 20281 Store the argument rounded to integer away from zero. 20282 20283 The `ceil' built-in function of C always uses the mode which 20284 corresponds to the C data type `double' and the `ceilf' built-in 20285 function uses the mode which corresponds to the C data type 20286 `float'. 20287 20288 `nearbyintM2' 20289 Store the argument rounded according to the default rounding mode 20290 20291 The `nearbyint' built-in function of C always uses the mode which 20292 corresponds to the C data type `double' and the `nearbyintf' 20293 built-in function uses the mode which corresponds to the C data 20294 type `float'. 20295 20296 `rintM2' 20297 Store the argument rounded according to the default rounding mode 20298 and raise the inexact exception when the result differs in value 20299 from the argument 20300 20301 The `rint' built-in function of C always uses the mode which 20302 corresponds to the C data type `double' and the `rintf' built-in 20303 function uses the mode which corresponds to the C data type 20304 `float'. 20305 20306 `lrintMN2' 20307 Convert operand 1 (valid for floating point mode M) to fixed point 20308 mode N as a signed number according to the current rounding mode 20309 and store in operand 0 (which has mode N). 20310 20311 `lroundM2' 20312 Convert operand 1 (valid for floating point mode M) to fixed point 20313 mode N as a signed number rounding to nearest and away from zero 20314 and store in operand 0 (which has mode N). 20315 20316 `lfloorM2' 20317 Convert operand 1 (valid for floating point mode M) to fixed point 20318 mode N as a signed number rounding down and store in operand 0 20319 (which has mode N). 20320 20321 `lceilM2' 20322 Convert operand 1 (valid for floating point mode M) to fixed point 20323 mode N as a signed number rounding up and store in operand 0 20324 (which has mode N). 20325 20326 `copysignM3' 20327 Store a value with the magnitude of operand 1 and the sign of 20328 operand 2 into operand 0. 20329 20330 The `copysign' built-in function of C always uses the mode which 20331 corresponds to the C data type `double' and the `copysignf' 20332 built-in function uses the mode which corresponds to the C data 20333 type `float'. 20334 20335 `ffsM2' 20336 Store into operand 0 one plus the index of the least significant 20337 1-bit of operand 1. If operand 1 is zero, store zero. M is the 20338 mode of operand 0; operand 1's mode is specified by the instruction 20339 pattern, and the compiler will convert the operand to that mode 20340 before generating the instruction. 20341 20342 The `ffs' built-in function of C always uses the mode which 20343 corresponds to the C data type `int'. 20344 20345 `clzM2' 20346 Store into operand 0 the number of leading 0-bits in X, starting 20347 at the most significant bit position. If X is 0, the 20348 `CLZ_DEFINED_VALUE_AT_ZERO' (*note Misc::) macro defines if the 20349 result is undefined or has a useful value. M is the mode of 20350 operand 0; operand 1's mode is specified by the instruction 20351 pattern, and the compiler will convert the operand to that mode 20352 before generating the instruction. 20353 20354 `ctzM2' 20355 Store into operand 0 the number of trailing 0-bits in X, starting 20356 at the least significant bit position. If X is 0, the 20357 `CTZ_DEFINED_VALUE_AT_ZERO' (*note Misc::) macro defines if the 20358 result is undefined or has a useful value. M is the mode of 20359 operand 0; operand 1's mode is specified by the instruction 20360 pattern, and the compiler will convert the operand to that mode 20361 before generating the instruction. 20362 20363 `popcountM2' 20364 Store into operand 0 the number of 1-bits in X. M is the mode of 20365 operand 0; operand 1's mode is specified by the instruction 20366 pattern, and the compiler will convert the operand to that mode 20367 before generating the instruction. 20368 20369 `parityM2' 20370 Store into operand 0 the parity of X, i.e. the number of 1-bits in 20371 X modulo 2. M is the mode of operand 0; operand 1's mode is 20372 specified by the instruction pattern, and the compiler will convert 20373 the operand to that mode before generating the instruction. 20374 20375 `one_cmplM2' 20376 Store the bitwise-complement of operand 1 into operand 0. 20377 20378 `cmpM' 20379 Compare operand 0 and operand 1, and set the condition codes. The 20380 RTL pattern should look like this: 20381 20382 (set (cc0) (compare (match_operand:M 0 ...) 20383 (match_operand:M 1 ...))) 20384 20385 `tstM' 20386 Compare operand 0 against zero, and set the condition codes. The 20387 RTL pattern should look like this: 20388 20389 (set (cc0) (match_operand:M 0 ...)) 20390 20391 `tstM' patterns should not be defined for machines that do not use 20392 `(cc0)'. Doing so would confuse the optimizer since it would no 20393 longer be clear which `set' operations were comparisons. The 20394 `cmpM' patterns should be used instead. 20395 20396 `movmemM' 20397 Block move instruction. The destination and source blocks of 20398 memory are the first two operands, and both are `mem:BLK's with an 20399 address in mode `Pmode'. 20400 20401 The number of bytes to move is the third operand, in mode M. 20402 Usually, you specify `word_mode' for M. However, if you can 20403 generate better code knowing the range of valid lengths is smaller 20404 than those representable in a full word, you should provide a 20405 pattern with a mode corresponding to the range of values you can 20406 handle efficiently (e.g., `QImode' for values in the range 0-127; 20407 note we avoid numbers that appear negative) and also a pattern 20408 with `word_mode'. 20409 20410 The fourth operand is the known shared alignment of the source and 20411 destination, in the form of a `const_int' rtx. Thus, if the 20412 compiler knows that both source and destination are word-aligned, 20413 it may provide the value 4 for this operand. 20414 20415 Optional operands 5 and 6 specify expected alignment and size of 20416 block respectively. The expected alignment differs from alignment 20417 in operand 4 in a way that the blocks are not required to be 20418 aligned according to it in all cases. This expected alignment is 20419 also in bytes, just like operand 4. Expected size, when unknown, 20420 is set to `(const_int -1)'. 20421 20422 Descriptions of multiple `movmemM' patterns can only be beneficial 20423 if the patterns for smaller modes have fewer restrictions on their 20424 first, second and fourth operands. Note that the mode M in 20425 `movmemM' does not impose any restriction on the mode of 20426 individually moved data units in the block. 20427 20428 These patterns need not give special consideration to the 20429 possibility that the source and destination strings might overlap. 20430 20431 `movstr' 20432 String copy instruction, with `stpcpy' semantics. Operand 0 is an 20433 output operand in mode `Pmode'. The addresses of the destination 20434 and source strings are operands 1 and 2, and both are `mem:BLK's 20435 with addresses in mode `Pmode'. The execution of the expansion of 20436 this pattern should store in operand 0 the address in which the 20437 `NUL' terminator was stored in the destination string. 20438 20439 `setmemM' 20440 Block set instruction. The destination string is the first 20441 operand, given as a `mem:BLK' whose address is in mode `Pmode'. 20442 The number of bytes to set is the second operand, in mode M. The 20443 value to initialize the memory with is the third operand. Targets 20444 that only support the clearing of memory should reject any value 20445 that is not the constant 0. See `movmemM' for a discussion of the 20446 choice of mode. 20447 20448 The fourth operand is the known alignment of the destination, in 20449 the form of a `const_int' rtx. Thus, if the compiler knows that 20450 the destination is word-aligned, it may provide the value 4 for 20451 this operand. 20452 20453 Optional operands 5 and 6 specify expected alignment and size of 20454 block respectively. The expected alignment differs from alignment 20455 in operand 4 in a way that the blocks are not required to be 20456 aligned according to it in all cases. This expected alignment is 20457 also in bytes, just like operand 4. Expected size, when unknown, 20458 is set to `(const_int -1)'. 20459 20460 The use for multiple `setmemM' is as for `movmemM'. 20461 20462 `cmpstrnM' 20463 String compare instruction, with five operands. Operand 0 is the 20464 output; it has mode M. The remaining four operands are like the 20465 operands of `movmemM'. The two memory blocks specified are 20466 compared byte by byte in lexicographic order starting at the 20467 beginning of each string. The instruction is not allowed to 20468 prefetch more than one byte at a time since either string may end 20469 in the first byte and reading past that may access an invalid page 20470 or segment and cause a fault. The effect of the instruction is to 20471 store a value in operand 0 whose sign indicates the result of the 20472 comparison. 20473 20474 `cmpstrM' 20475 String compare instruction, without known maximum length. Operand 20476 0 is the output; it has mode M. The second and third operand are 20477 the blocks of memory to be compared; both are `mem:BLK' with an 20478 address in mode `Pmode'. 20479 20480 The fourth operand is the known shared alignment of the source and 20481 destination, in the form of a `const_int' rtx. Thus, if the 20482 compiler knows that both source and destination are word-aligned, 20483 it may provide the value 4 for this operand. 20484 20485 The two memory blocks specified are compared byte by byte in 20486 lexicographic order starting at the beginning of each string. The 20487 instruction is not allowed to prefetch more than one byte at a 20488 time since either string may end in the first byte and reading 20489 past that may access an invalid page or segment and cause a fault. 20490 The effect of the instruction is to store a value in operand 0 20491 whose sign indicates the result of the comparison. 20492 20493 `cmpmemM' 20494 Block compare instruction, with five operands like the operands of 20495 `cmpstrM'. The two memory blocks specified are compared byte by 20496 byte in lexicographic order starting at the beginning of each 20497 block. Unlike `cmpstrM' the instruction can prefetch any bytes in 20498 the two memory blocks. The effect of the instruction is to store 20499 a value in operand 0 whose sign indicates the result of the 20500 comparison. 20501 20502 `strlenM' 20503 Compute the length of a string, with three operands. Operand 0 is 20504 the result (of mode M), operand 1 is a `mem' referring to the 20505 first character of the string, operand 2 is the character to 20506 search for (normally zero), and operand 3 is a constant describing 20507 the known alignment of the beginning of the string. 20508 20509 `floatMN2' 20510 Convert signed integer operand 1 (valid for fixed point mode M) to 20511 floating point mode N and store in operand 0 (which has mode N). 20512 20513 `floatunsMN2' 20514 Convert unsigned integer operand 1 (valid for fixed point mode M) 20515 to floating point mode N and store in operand 0 (which has mode N). 20516 20517 `fixMN2' 20518 Convert operand 1 (valid for floating point mode M) to fixed point 20519 mode N as a signed number and store in operand 0 (which has mode 20520 N). This instruction's result is defined only when the value of 20521 operand 1 is an integer. 20522 20523 If the machine description defines this pattern, it also needs to 20524 define the `ftrunc' pattern. 20525 20526 `fixunsMN2' 20527 Convert operand 1 (valid for floating point mode M) to fixed point 20528 mode N as an unsigned number and store in operand 0 (which has 20529 mode N). This instruction's result is defined only when the value 20530 of operand 1 is an integer. 20531 20532 `ftruncM2' 20533 Convert operand 1 (valid for floating point mode M) to an integer 20534 value, still represented in floating point mode M, and store it in 20535 operand 0 (valid for floating point mode M). 20536 20537 `fix_truncMN2' 20538 Like `fixMN2' but works for any floating point value of mode M by 20539 converting the value to an integer. 20540 20541 `fixuns_truncMN2' 20542 Like `fixunsMN2' but works for any floating point value of mode M 20543 by converting the value to an integer. 20544 20545 `truncMN2' 20546 Truncate operand 1 (valid for mode M) to mode N and store in 20547 operand 0 (which has mode N). Both modes must be fixed point or 20548 both floating point. 20549 20550 `extendMN2' 20551 Sign-extend operand 1 (valid for mode M) to mode N and store in 20552 operand 0 (which has mode N). Both modes must be fixed point or 20553 both floating point. 20554 20555 `zero_extendMN2' 20556 Zero-extend operand 1 (valid for mode M) to mode N and store in 20557 operand 0 (which has mode N). Both modes must be fixed point. 20558 20559 `fractMN2' 20560 Convert operand 1 of mode M to mode N and store in operand 0 20561 (which has mode N). Mode M and mode N could be fixed-point to 20562 fixed-point, signed integer to fixed-point, fixed-point to signed 20563 integer, floating-point to fixed-point, or fixed-point to 20564 floating-point. When overflows or underflows happen, the results 20565 are undefined. 20566 20567 `satfractMN2' 20568 Convert operand 1 of mode M to mode N and store in operand 0 20569 (which has mode N). Mode M and mode N could be fixed-point to 20570 fixed-point, signed integer to fixed-point, or floating-point to 20571 fixed-point. When overflows or underflows happen, the instruction 20572 saturates the results to the maximum or the minimum. 20573 20574 `fractunsMN2' 20575 Convert operand 1 of mode M to mode N and store in operand 0 20576 (which has mode N). Mode M and mode N could be unsigned integer 20577 to fixed-point, or fixed-point to unsigned integer. When 20578 overflows or underflows happen, the results are undefined. 20579 20580 `satfractunsMN2' 20581 Convert unsigned integer operand 1 of mode M to fixed-point mode N 20582 and store in operand 0 (which has mode N). When overflows or 20583 underflows happen, the instruction saturates the results to the 20584 maximum or the minimum. 20585 20586 `extv' 20587 Extract a bit-field from operand 1 (a register or memory operand), 20588 where operand 2 specifies the width in bits and operand 3 the 20589 starting bit, and store it in operand 0. Operand 0 must have mode 20590 `word_mode'. Operand 1 may have mode `byte_mode' or `word_mode'; 20591 often `word_mode' is allowed only for registers. Operands 2 and 3 20592 must be valid for `word_mode'. 20593 20594 The RTL generation pass generates this instruction only with 20595 constants for operands 2 and 3 and the constant is never zero for 20596 operand 2. 20597 20598 The bit-field value is sign-extended to a full word integer before 20599 it is stored in operand 0. 20600 20601 `extzv' 20602 Like `extv' except that the bit-field value is zero-extended. 20603 20604 `insv' 20605 Store operand 3 (which must be valid for `word_mode') into a 20606 bit-field in operand 0, where operand 1 specifies the width in 20607 bits and operand 2 the starting bit. Operand 0 may have mode 20608 `byte_mode' or `word_mode'; often `word_mode' is allowed only for 20609 registers. Operands 1 and 2 must be valid for `word_mode'. 20610 20611 The RTL generation pass generates this instruction only with 20612 constants for operands 1 and 2 and the constant is never zero for 20613 operand 1. 20614 20615 `movMODEcc' 20616 Conditionally move operand 2 or operand 3 into operand 0 according 20617 to the comparison in operand 1. If the comparison is true, 20618 operand 2 is moved into operand 0, otherwise operand 3 is moved. 20619 20620 The mode of the operands being compared need not be the same as 20621 the operands being moved. Some machines, sparc64 for example, 20622 have instructions that conditionally move an integer value based 20623 on the floating point condition codes and vice versa. 20624 20625 If the machine does not have conditional move instructions, do not 20626 define these patterns. 20627 20628 `addMODEcc' 20629 Similar to `movMODEcc' but for conditional addition. Conditionally 20630 move operand 2 or (operands 2 + operand 3) into operand 0 20631 according to the comparison in operand 1. If the comparison is 20632 true, operand 2 is moved into operand 0, otherwise (operand 2 + 20633 operand 3) is moved. 20634 20635 `sCOND' 20636 Store zero or nonzero in the operand according to the condition 20637 codes. Value stored is nonzero iff the condition COND is true. 20638 COND is the name of a comparison operation expression code, such 20639 as `eq', `lt' or `leu'. 20640 20641 You specify the mode that the operand must have when you write the 20642 `match_operand' expression. The compiler automatically sees which 20643 mode you have used and supplies an operand of that mode. 20644 20645 The value stored for a true condition must have 1 as its low bit, 20646 or else must be negative. Otherwise the instruction is not 20647 suitable and you should omit it from the machine description. You 20648 describe to the compiler exactly which value is stored by defining 20649 the macro `STORE_FLAG_VALUE' (*note Misc::). If a description 20650 cannot be found that can be used for all the `sCOND' patterns, you 20651 should omit those operations from the machine description. 20652 20653 These operations may fail, but should do so only in relatively 20654 uncommon cases; if they would fail for common cases involving 20655 integer comparisons, it is best to omit these patterns. 20656 20657 If these operations are omitted, the compiler will usually 20658 generate code that copies the constant one to the target and 20659 branches around an assignment of zero to the target. If this code 20660 is more efficient than the potential instructions used for the 20661 `sCOND' pattern followed by those required to convert the result 20662 into a 1 or a zero in `SImode', you should omit the `sCOND' 20663 operations from the machine description. 20664 20665 `bCOND' 20666 Conditional branch instruction. Operand 0 is a `label_ref' that 20667 refers to the label to jump to. Jump if the condition codes meet 20668 condition COND. 20669 20670 Some machines do not follow the model assumed here where a 20671 comparison instruction is followed by a conditional branch 20672 instruction. In that case, the `cmpM' (and `tstM') patterns should 20673 simply store the operands away and generate all the required insns 20674 in a `define_expand' (*note Expander Definitions::) for the 20675 conditional branch operations. All calls to expand `bCOND' 20676 patterns are immediately preceded by calls to expand either a 20677 `cmpM' pattern or a `tstM' pattern. 20678 20679 Machines that use a pseudo register for the condition code value, 20680 or where the mode used for the comparison depends on the condition 20681 being tested, should also use the above mechanism. *Note Jump 20682 Patterns::. 20683 20684 The above discussion also applies to the `movMODEcc' and `sCOND' 20685 patterns. 20686 20687 `cbranchMODE4' 20688 Conditional branch instruction combined with a compare instruction. 20689 Operand 0 is a comparison operator. Operand 1 and operand 2 are 20690 the first and second operands of the comparison, respectively. 20691 Operand 3 is a `label_ref' that refers to the label to jump to. 20692 20693 `jump' 20694 A jump inside a function; an unconditional branch. Operand 0 is 20695 the `label_ref' of the label to jump to. This pattern name is 20696 mandatory on all machines. 20697 20698 `call' 20699 Subroutine call instruction returning no value. Operand 0 is the 20700 function to call; operand 1 is the number of bytes of arguments 20701 pushed as a `const_int'; operand 2 is the number of registers used 20702 as operands. 20703 20704 On most machines, operand 2 is not actually stored into the RTL 20705 pattern. It is supplied for the sake of some RISC machines which 20706 need to put this information into the assembler code; they can put 20707 it in the RTL instead of operand 1. 20708 20709 Operand 0 should be a `mem' RTX whose address is the address of the 20710 function. Note, however, that this address can be a `symbol_ref' 20711 expression even if it would not be a legitimate memory address on 20712 the target machine. If it is also not a valid argument for a call 20713 instruction, the pattern for this operation should be a 20714 `define_expand' (*note Expander Definitions::) that places the 20715 address into a register and uses that register in the call 20716 instruction. 20717 20718 `call_value' 20719 Subroutine call instruction returning a value. Operand 0 is the 20720 hard register in which the value is returned. There are three more 20721 operands, the same as the three operands of the `call' instruction 20722 (but with numbers increased by one). 20723 20724 Subroutines that return `BLKmode' objects use the `call' insn. 20725 20726 `call_pop', `call_value_pop' 20727 Similar to `call' and `call_value', except used if defined and if 20728 `RETURN_POPS_ARGS' is nonzero. They should emit a `parallel' that 20729 contains both the function call and a `set' to indicate the 20730 adjustment made to the frame pointer. 20731 20732 For machines where `RETURN_POPS_ARGS' can be nonzero, the use of 20733 these patterns increases the number of functions for which the 20734 frame pointer can be eliminated, if desired. 20735 20736 `untyped_call' 20737 Subroutine call instruction returning a value of any type. 20738 Operand 0 is the function to call; operand 1 is a memory location 20739 where the result of calling the function is to be stored; operand 20740 2 is a `parallel' expression where each element is a `set' 20741 expression that indicates the saving of a function return value 20742 into the result block. 20743 20744 This instruction pattern should be defined to support 20745 `__builtin_apply' on machines where special instructions are needed 20746 to call a subroutine with arbitrary arguments or to save the value 20747 returned. This instruction pattern is required on machines that 20748 have multiple registers that can hold a return value (i.e. 20749 `FUNCTION_VALUE_REGNO_P' is true for more than one register). 20750 20751 `return' 20752 Subroutine return instruction. This instruction pattern name 20753 should be defined only if a single instruction can do all the work 20754 of returning from a function. 20755 20756 Like the `movM' patterns, this pattern is also used after the RTL 20757 generation phase. In this case it is to support machines where 20758 multiple instructions are usually needed to return from a 20759 function, but some class of functions only requires one 20760 instruction to implement a return. Normally, the applicable 20761 functions are those which do not need to save any registers or 20762 allocate stack space. 20763 20764 For such machines, the condition specified in this pattern should 20765 only be true when `reload_completed' is nonzero and the function's 20766 epilogue would only be a single instruction. For machines with 20767 register windows, the routine `leaf_function_p' may be used to 20768 determine if a register window push is required. 20769 20770 Machines that have conditional return instructions should define 20771 patterns such as 20772 20773 (define_insn "" 20774 [(set (pc) 20775 (if_then_else (match_operator 20776 0 "comparison_operator" 20777 [(cc0) (const_int 0)]) 20778 (return) 20779 (pc)))] 20780 "CONDITION" 20781 "...") 20782 20783 where CONDITION would normally be the same condition specified on 20784 the named `return' pattern. 20785 20786 `untyped_return' 20787 Untyped subroutine return instruction. This instruction pattern 20788 should be defined to support `__builtin_return' on machines where 20789 special instructions are needed to return a value of any type. 20790 20791 Operand 0 is a memory location where the result of calling a 20792 function with `__builtin_apply' is stored; operand 1 is a 20793 `parallel' expression where each element is a `set' expression 20794 that indicates the restoring of a function return value from the 20795 result block. 20796 20797 `nop' 20798 No-op instruction. This instruction pattern name should always be 20799 defined to output a no-op in assembler code. `(const_int 0)' will 20800 do as an RTL pattern. 20801 20802 `indirect_jump' 20803 An instruction to jump to an address which is operand zero. This 20804 pattern name is mandatory on all machines. 20805 20806 `casesi' 20807 Instruction to jump through a dispatch table, including bounds 20808 checking. This instruction takes five operands: 20809 20810 1. The index to dispatch on, which has mode `SImode'. 20811 20812 2. The lower bound for indices in the table, an integer constant. 20813 20814 3. The total range of indices in the table--the largest index 20815 minus the smallest one (both inclusive). 20816 20817 4. A label that precedes the table itself. 20818 20819 5. A label to jump to if the index has a value outside the 20820 bounds. 20821 20822 The table is a `addr_vec' or `addr_diff_vec' inside of a 20823 `jump_insn'. The number of elements in the table is one plus the 20824 difference between the upper bound and the lower bound. 20825 20826 `tablejump' 20827 Instruction to jump to a variable address. This is a low-level 20828 capability which can be used to implement a dispatch table when 20829 there is no `casesi' pattern. 20830 20831 This pattern requires two operands: the address or offset, and a 20832 label which should immediately precede the jump table. If the 20833 macro `CASE_VECTOR_PC_RELATIVE' evaluates to a nonzero value then 20834 the first operand is an offset which counts from the address of 20835 the table; otherwise, it is an absolute address to jump to. In 20836 either case, the first operand has mode `Pmode'. 20837 20838 The `tablejump' insn is always the last insn before the jump table 20839 it uses. Its assembler code normally has no need to use the 20840 second operand, but you should incorporate it in the RTL pattern so 20841 that the jump optimizer will not delete the table as unreachable 20842 code. 20843 20844 `decrement_and_branch_until_zero' 20845 Conditional branch instruction that decrements a register and 20846 jumps if the register is nonzero. Operand 0 is the register to 20847 decrement and test; operand 1 is the label to jump to if the 20848 register is nonzero. *Note Looping Patterns::. 20849 20850 This optional instruction pattern is only used by the combiner, 20851 typically for loops reversed by the loop optimizer when strength 20852 reduction is enabled. 20853 20854 `doloop_end' 20855 Conditional branch instruction that decrements a register and 20856 jumps if the register is nonzero. This instruction takes five 20857 operands: Operand 0 is the register to decrement and test; operand 20858 1 is the number of loop iterations as a `const_int' or 20859 `const0_rtx' if this cannot be determined until run-time; operand 20860 2 is the actual or estimated maximum number of iterations as a 20861 `const_int'; operand 3 is the number of enclosed loops as a 20862 `const_int' (an innermost loop has a value of 1); operand 4 is the 20863 label to jump to if the register is nonzero. *Note Looping 20864 Patterns::. 20865 20866 This optional instruction pattern should be defined for machines 20867 with low-overhead looping instructions as the loop optimizer will 20868 try to modify suitable loops to utilize it. If nested 20869 low-overhead looping is not supported, use a `define_expand' 20870 (*note Expander Definitions::) and make the pattern fail if 20871 operand 3 is not `const1_rtx'. Similarly, if the actual or 20872 estimated maximum number of iterations is too large for this 20873 instruction, make it fail. 20874 20875 `doloop_begin' 20876 Companion instruction to `doloop_end' required for machines that 20877 need to perform some initialization, such as loading special 20878 registers used by a low-overhead looping instruction. If 20879 initialization insns do not always need to be emitted, use a 20880 `define_expand' (*note Expander Definitions::) and make it fail. 20881 20882 `canonicalize_funcptr_for_compare' 20883 Canonicalize the function pointer in operand 1 and store the result 20884 into operand 0. 20885 20886 Operand 0 is always a `reg' and has mode `Pmode'; operand 1 may be 20887 a `reg', `mem', `symbol_ref', `const_int', etc and also has mode 20888 `Pmode'. 20889 20890 Canonicalization of a function pointer usually involves computing 20891 the address of the function which would be called if the function 20892 pointer were used in an indirect call. 20893 20894 Only define this pattern if function pointers on the target machine 20895 can have different values but still call the same function when 20896 used in an indirect call. 20897 20898 `save_stack_block' 20899 `save_stack_function' 20900 `save_stack_nonlocal' 20901 `restore_stack_block' 20902 `restore_stack_function' 20903 `restore_stack_nonlocal' 20904 Most machines save and restore the stack pointer by copying it to 20905 or from an object of mode `Pmode'. Do not define these patterns on 20906 such machines. 20907 20908 Some machines require special handling for stack pointer saves and 20909 restores. On those machines, define the patterns corresponding to 20910 the non-standard cases by using a `define_expand' (*note Expander 20911 Definitions::) that produces the required insns. The three types 20912 of saves and restores are: 20913 20914 1. `save_stack_block' saves the stack pointer at the start of a 20915 block that allocates a variable-sized object, and 20916 `restore_stack_block' restores the stack pointer when the 20917 block is exited. 20918 20919 2. `save_stack_function' and `restore_stack_function' do a 20920 similar job for the outermost block of a function and are 20921 used when the function allocates variable-sized objects or 20922 calls `alloca'. Only the epilogue uses the restored stack 20923 pointer, allowing a simpler save or restore sequence on some 20924 machines. 20925 20926 3. `save_stack_nonlocal' is used in functions that contain labels 20927 branched to by nested functions. It saves the stack pointer 20928 in such a way that the inner function can use 20929 `restore_stack_nonlocal' to restore the stack pointer. The 20930 compiler generates code to restore the frame and argument 20931 pointer registers, but some machines require saving and 20932 restoring additional data such as register window information 20933 or stack backchains. Place insns in these patterns to save 20934 and restore any such required data. 20935 20936 When saving the stack pointer, operand 0 is the save area and 20937 operand 1 is the stack pointer. The mode used to allocate the 20938 save area defaults to `Pmode' but you can override that choice by 20939 defining the `STACK_SAVEAREA_MODE' macro (*note Storage Layout::). 20940 You must specify an integral mode, or `VOIDmode' if no save area 20941 is needed for a particular type of save (either because no save is 20942 needed or because a machine-specific save area can be used). 20943 Operand 0 is the stack pointer and operand 1 is the save area for 20944 restore operations. If `save_stack_block' is defined, operand 0 20945 must not be `VOIDmode' since these saves can be arbitrarily nested. 20946 20947 A save area is a `mem' that is at a constant offset from 20948 `virtual_stack_vars_rtx' when the stack pointer is saved for use by 20949 nonlocal gotos and a `reg' in the other two cases. 20950 20951 `allocate_stack' 20952 Subtract (or add if `STACK_GROWS_DOWNWARD' is undefined) operand 1 20953 from the stack pointer to create space for dynamically allocated 20954 data. 20955 20956 Store the resultant pointer to this space into operand 0. If you 20957 are allocating space from the main stack, do this by emitting a 20958 move insn to copy `virtual_stack_dynamic_rtx' to operand 0. If 20959 you are allocating the space elsewhere, generate code to copy the 20960 location of the space to operand 0. In the latter case, you must 20961 ensure this space gets freed when the corresponding space on the 20962 main stack is free. 20963 20964 Do not define this pattern if all that must be done is the 20965 subtraction. Some machines require other operations such as stack 20966 probes or maintaining the back chain. Define this pattern to emit 20967 those operations in addition to updating the stack pointer. 20968 20969 `check_stack' 20970 If stack checking cannot be done on your system by probing the 20971 stack with a load or store instruction (*note Stack Checking::), 20972 define this pattern to perform the needed check and signaling an 20973 error if the stack has overflowed. The single operand is the 20974 location in the stack furthest from the current stack pointer that 20975 you need to validate. Normally, on machines where this pattern is 20976 needed, you would obtain the stack limit from a global or 20977 thread-specific variable or register. 20978 20979 `nonlocal_goto' 20980 Emit code to generate a non-local goto, e.g., a jump from one 20981 function to a label in an outer function. This pattern has four 20982 arguments, each representing a value to be used in the jump. The 20983 first argument is to be loaded into the frame pointer, the second 20984 is the address to branch to (code to dispatch to the actual label), 20985 the third is the address of a location where the stack is saved, 20986 and the last is the address of the label, to be placed in the 20987 location for the incoming static chain. 20988 20989 On most machines you need not define this pattern, since GCC will 20990 already generate the correct code, which is to load the frame 20991 pointer and static chain, restore the stack (using the 20992 `restore_stack_nonlocal' pattern, if defined), and jump indirectly 20993 to the dispatcher. You need only define this pattern if this code 20994 will not work on your machine. 20995 20996 `nonlocal_goto_receiver' 20997 This pattern, if defined, contains code needed at the target of a 20998 nonlocal goto after the code already generated by GCC. You will 20999 not normally need to define this pattern. A typical reason why 21000 you might need this pattern is if some value, such as a pointer to 21001 a global table, must be restored when the frame pointer is 21002 restored. Note that a nonlocal goto only occurs within a 21003 unit-of-translation, so a global table pointer that is shared by 21004 all functions of a given module need not be restored. There are 21005 no arguments. 21006 21007 `exception_receiver' 21008 This pattern, if defined, contains code needed at the site of an 21009 exception handler that isn't needed at the site of a nonlocal 21010 goto. You will not normally need to define this pattern. A 21011 typical reason why you might need this pattern is if some value, 21012 such as a pointer to a global table, must be restored after 21013 control flow is branched to the handler of an exception. There 21014 are no arguments. 21015 21016 `builtin_setjmp_setup' 21017 This pattern, if defined, contains additional code needed to 21018 initialize the `jmp_buf'. You will not normally need to define 21019 this pattern. A typical reason why you might need this pattern is 21020 if some value, such as a pointer to a global table, must be 21021 restored. Though it is preferred that the pointer value be 21022 recalculated if possible (given the address of a label for 21023 instance). The single argument is a pointer to the `jmp_buf'. 21024 Note that the buffer is five words long and that the first three 21025 are normally used by the generic mechanism. 21026 21027 `builtin_setjmp_receiver' 21028 This pattern, if defined, contains code needed at the site of an 21029 built-in setjmp that isn't needed at the site of a nonlocal goto. 21030 You will not normally need to define this pattern. A typical 21031 reason why you might need this pattern is if some value, such as a 21032 pointer to a global table, must be restored. It takes one 21033 argument, which is the label to which builtin_longjmp transfered 21034 control; this pattern may be emitted at a small offset from that 21035 label. 21036 21037 `builtin_longjmp' 21038 This pattern, if defined, performs the entire action of the 21039 longjmp. You will not normally need to define this pattern unless 21040 you also define `builtin_setjmp_setup'. The single argument is a 21041 pointer to the `jmp_buf'. 21042 21043 `eh_return' 21044 This pattern, if defined, affects the way `__builtin_eh_return', 21045 and thence the call frame exception handling library routines, are 21046 built. It is intended to handle non-trivial actions needed along 21047 the abnormal return path. 21048 21049 The address of the exception handler to which the function should 21050 return is passed as operand to this pattern. It will normally 21051 need to copied by the pattern to some special register or memory 21052 location. If the pattern needs to determine the location of the 21053 target call frame in order to do so, it may use 21054 `EH_RETURN_STACKADJ_RTX', if defined; it will have already been 21055 assigned. 21056 21057 If this pattern is not defined, the default action will be to 21058 simply copy the return address to `EH_RETURN_HANDLER_RTX'. Either 21059 that macro or this pattern needs to be defined if call frame 21060 exception handling is to be used. 21061 21062 `prologue' 21063 This pattern, if defined, emits RTL for entry to a function. The 21064 function entry is responsible for setting up the stack frame, 21065 initializing the frame pointer register, saving callee saved 21066 registers, etc. 21067 21068 Using a prologue pattern is generally preferred over defining 21069 `TARGET_ASM_FUNCTION_PROLOGUE' to emit assembly code for the 21070 prologue. 21071 21072 The `prologue' pattern is particularly useful for targets which 21073 perform instruction scheduling. 21074 21075 `epilogue' 21076 This pattern emits RTL for exit from a function. The function 21077 exit is responsible for deallocating the stack frame, restoring 21078 callee saved registers and emitting the return instruction. 21079 21080 Using an epilogue pattern is generally preferred over defining 21081 `TARGET_ASM_FUNCTION_EPILOGUE' to emit assembly code for the 21082 epilogue. 21083 21084 The `epilogue' pattern is particularly useful for targets which 21085 perform instruction scheduling or which have delay slots for their 21086 return instruction. 21087 21088 `sibcall_epilogue' 21089 This pattern, if defined, emits RTL for exit from a function 21090 without the final branch back to the calling function. This 21091 pattern will be emitted before any sibling call (aka tail call) 21092 sites. 21093 21094 The `sibcall_epilogue' pattern must not clobber any arguments used 21095 for parameter passing or any stack slots for arguments passed to 21096 the current function. 21097 21098 `trap' 21099 This pattern, if defined, signals an error, typically by causing 21100 some kind of signal to be raised. Among other places, it is used 21101 by the Java front end to signal `invalid array index' exceptions. 21102 21103 `conditional_trap' 21104 Conditional trap instruction. Operand 0 is a piece of RTL which 21105 performs a comparison. Operand 1 is the trap code, an integer. 21106 21107 A typical `conditional_trap' pattern looks like 21108 21109 (define_insn "conditional_trap" 21110 [(trap_if (match_operator 0 "trap_operator" 21111 [(cc0) (const_int 0)]) 21112 (match_operand 1 "const_int_operand" "i"))] 21113 "" 21114 "...") 21115 21116 `prefetch' 21117 This pattern, if defined, emits code for a non-faulting data 21118 prefetch instruction. Operand 0 is the address of the memory to 21119 prefetch. Operand 1 is a constant 1 if the prefetch is preparing 21120 for a write to the memory address, or a constant 0 otherwise. 21121 Operand 2 is the expected degree of temporal locality of the data 21122 and is a value between 0 and 3, inclusive; 0 means that the data 21123 has no temporal locality, so it need not be left in the cache 21124 after the access; 3 means that the data has a high degree of 21125 temporal locality and should be left in all levels of cache 21126 possible; 1 and 2 mean, respectively, a low or moderate degree of 21127 temporal locality. 21128 21129 Targets that do not support write prefetches or locality hints can 21130 ignore the values of operands 1 and 2. 21131 21132 `blockage' 21133 This pattern defines a pseudo insn that prevents the instruction 21134 scheduler from moving instructions across the boundary defined by 21135 the blockage insn. Normally an UNSPEC_VOLATILE pattern. 21136 21137 `memory_barrier' 21138 If the target memory model is not fully synchronous, then this 21139 pattern should be defined to an instruction that orders both loads 21140 and stores before the instruction with respect to loads and stores 21141 after the instruction. This pattern has no operands. 21142 21143 `sync_compare_and_swapMODE' 21144 This pattern, if defined, emits code for an atomic compare-and-swap 21145 operation. Operand 1 is the memory on which the atomic operation 21146 is performed. Operand 2 is the "old" value to be compared against 21147 the current contents of the memory location. Operand 3 is the 21148 "new" value to store in the memory if the compare succeeds. 21149 Operand 0 is the result of the operation; it should contain the 21150 contents of the memory before the operation. If the compare 21151 succeeds, this should obviously be a copy of operand 2. 21152 21153 This pattern must show that both operand 0 and operand 1 are 21154 modified. 21155 21156 This pattern must issue any memory barrier instructions such that 21157 all memory operations before the atomic operation occur before the 21158 atomic operation and all memory operations after the atomic 21159 operation occur after the atomic operation. 21160 21161 `sync_compare_and_swap_ccMODE' 21162 This pattern is just like `sync_compare_and_swapMODE', except it 21163 should act as if compare part of the compare-and-swap were issued 21164 via `cmpM'. This comparison will only be used with `EQ' and `NE' 21165 branches and `setcc' operations. 21166 21167 Some targets do expose the success or failure of the 21168 compare-and-swap operation via the status flags. Ideally we 21169 wouldn't need a separate named pattern in order to take advantage 21170 of this, but the combine pass does not handle patterns with 21171 multiple sets, which is required by definition for 21172 `sync_compare_and_swapMODE'. 21173 21174 `sync_addMODE', `sync_subMODE' 21175 `sync_iorMODE', `sync_andMODE' 21176 `sync_xorMODE', `sync_nandMODE' 21177 These patterns emit code for an atomic operation on memory. 21178 Operand 0 is the memory on which the atomic operation is performed. 21179 Operand 1 is the second operand to the binary operator. 21180 21181 This pattern must issue any memory barrier instructions such that 21182 all memory operations before the atomic operation occur before the 21183 atomic operation and all memory operations after the atomic 21184 operation occur after the atomic operation. 21185 21186 If these patterns are not defined, the operation will be 21187 constructed from a compare-and-swap operation, if defined. 21188 21189 `sync_old_addMODE', `sync_old_subMODE' 21190 `sync_old_iorMODE', `sync_old_andMODE' 21191 `sync_old_xorMODE', `sync_old_nandMODE' 21192 These patterns are emit code for an atomic operation on memory, 21193 and return the value that the memory contained before the 21194 operation. Operand 0 is the result value, operand 1 is the memory 21195 on which the atomic operation is performed, and operand 2 is the 21196 second operand to the binary operator. 21197 21198 This pattern must issue any memory barrier instructions such that 21199 all memory operations before the atomic operation occur before the 21200 atomic operation and all memory operations after the atomic 21201 operation occur after the atomic operation. 21202 21203 If these patterns are not defined, the operation will be 21204 constructed from a compare-and-swap operation, if defined. 21205 21206 `sync_new_addMODE', `sync_new_subMODE' 21207 `sync_new_iorMODE', `sync_new_andMODE' 21208 `sync_new_xorMODE', `sync_new_nandMODE' 21209 These patterns are like their `sync_old_OP' counterparts, except 21210 that they return the value that exists in the memory location 21211 after the operation, rather than before the operation. 21212 21213 `sync_lock_test_and_setMODE' 21214 This pattern takes two forms, based on the capabilities of the 21215 target. In either case, operand 0 is the result of the operand, 21216 operand 1 is the memory on which the atomic operation is 21217 performed, and operand 2 is the value to set in the lock. 21218 21219 In the ideal case, this operation is an atomic exchange operation, 21220 in which the previous value in memory operand is copied into the 21221 result operand, and the value operand is stored in the memory 21222 operand. 21223 21224 For less capable targets, any value operand that is not the 21225 constant 1 should be rejected with `FAIL'. In this case the 21226 target may use an atomic test-and-set bit operation. The result 21227 operand should contain 1 if the bit was previously set and 0 if 21228 the bit was previously clear. The true contents of the memory 21229 operand are implementation defined. 21230 21231 This pattern must issue any memory barrier instructions such that 21232 the pattern as a whole acts as an acquire barrier, that is all 21233 memory operations after the pattern do not occur until the lock is 21234 acquired. 21235 21236 If this pattern is not defined, the operation will be constructed 21237 from a compare-and-swap operation, if defined. 21238 21239 `sync_lock_releaseMODE' 21240 This pattern, if defined, releases a lock set by 21241 `sync_lock_test_and_setMODE'. Operand 0 is the memory that 21242 contains the lock; operand 1 is the value to store in the lock. 21243 21244 If the target doesn't implement full semantics for 21245 `sync_lock_test_and_setMODE', any value operand which is not the 21246 constant 0 should be rejected with `FAIL', and the true contents 21247 of the memory operand are implementation defined. 21248 21249 This pattern must issue any memory barrier instructions such that 21250 the pattern as a whole acts as a release barrier, that is the lock 21251 is released only after all previous memory operations have 21252 completed. 21253 21254 If this pattern is not defined, then a `memory_barrier' pattern 21255 will be emitted, followed by a store of the value to the memory 21256 operand. 21257 21258 `stack_protect_set' 21259 This pattern, if defined, moves a `Pmode' value from the memory in 21260 operand 1 to the memory in operand 0 without leaving the value in 21261 a register afterward. This is to avoid leaking the value some 21262 place that an attacker might use to rewrite the stack guard slot 21263 after having clobbered it. 21264 21265 If this pattern is not defined, then a plain move pattern is 21266 generated. 21267 21268 `stack_protect_test' 21269 This pattern, if defined, compares a `Pmode' value from the memory 21270 in operand 1 with the memory in operand 0 without leaving the 21271 value in a register afterward and branches to operand 2 if the 21272 values weren't equal. 21273 21274 If this pattern is not defined, then a plain compare pattern and 21275 conditional branch pattern is used. 21276 21277 `clear_cache' 21278 This pattern, if defined, flushes the instruction cache for a 21279 region of memory. The region is bounded to by the Pmode pointers 21280 in operand 0 inclusive and operand 1 exclusive. 21281 21282 If this pattern is not defined, a call to the library function 21283 `__clear_cache' is used. 21284 21285 21286 21287 File: gccint.info, Node: Pattern Ordering, Next: Dependent Patterns, Prev: Standard Names, Up: Machine Desc 21288 21289 16.10 When the Order of Patterns Matters 21290 ======================================== 21291 21292 Sometimes an insn can match more than one instruction pattern. Then the 21293 pattern that appears first in the machine description is the one used. 21294 Therefore, more specific patterns (patterns that will match fewer 21295 things) and faster instructions (those that will produce better code 21296 when they do match) should usually go first in the description. 21297 21298 In some cases the effect of ordering the patterns can be used to hide 21299 a pattern when it is not valid. For example, the 68000 has an 21300 instruction for converting a fullword to floating point and another for 21301 converting a byte to floating point. An instruction converting an 21302 integer to floating point could match either one. We put the pattern 21303 to convert the fullword first to make sure that one will be used rather 21304 than the other. (Otherwise a large integer might be generated as a 21305 single-byte immediate quantity, which would not work.) Instead of 21306 using this pattern ordering it would be possible to make the pattern 21307 for convert-a-byte smart enough to deal properly with any constant 21308 value. 21309 21310 21311 File: gccint.info, Node: Dependent Patterns, Next: Jump Patterns, Prev: Pattern Ordering, Up: Machine Desc 21312 21313 16.11 Interdependence of Patterns 21314 ================================= 21315 21316 Every machine description must have a named pattern for each of the 21317 conditional branch names `bCOND'. The recognition template must always 21318 have the form 21319 21320 (set (pc) 21321 (if_then_else (COND (cc0) (const_int 0)) 21322 (label_ref (match_operand 0 "" "")) 21323 (pc))) 21324 21325 In addition, every machine description must have an anonymous pattern 21326 for each of the possible reverse-conditional branches. Their templates 21327 look like 21328 21329 (set (pc) 21330 (if_then_else (COND (cc0) (const_int 0)) 21331 (pc) 21332 (label_ref (match_operand 0 "" "")))) 21333 21334 They are necessary because jump optimization can turn direct-conditional 21335 branches into reverse-conditional branches. 21336 21337 It is often convenient to use the `match_operator' construct to reduce 21338 the number of patterns that must be specified for branches. For 21339 example, 21340 21341 (define_insn "" 21342 [(set (pc) 21343 (if_then_else (match_operator 0 "comparison_operator" 21344 [(cc0) (const_int 0)]) 21345 (pc) 21346 (label_ref (match_operand 1 "" ""))))] 21347 "CONDITION" 21348 "...") 21349 21350 In some cases machines support instructions identical except for the 21351 machine mode of one or more operands. For example, there may be 21352 "sign-extend halfword" and "sign-extend byte" instructions whose 21353 patterns are 21354 21355 (set (match_operand:SI 0 ...) 21356 (extend:SI (match_operand:HI 1 ...))) 21357 21358 (set (match_operand:SI 0 ...) 21359 (extend:SI (match_operand:QI 1 ...))) 21360 21361 Constant integers do not specify a machine mode, so an instruction to 21362 extend a constant value could match either pattern. The pattern it 21363 actually will match is the one that appears first in the file. For 21364 correct results, this must be the one for the widest possible mode 21365 (`HImode', here). If the pattern matches the `QImode' instruction, the 21366 results will be incorrect if the constant value does not actually fit 21367 that mode. 21368 21369 Such instructions to extend constants are rarely generated because 21370 they are optimized away, but they do occasionally happen in nonoptimized 21371 compilations. 21372 21373 If a constraint in a pattern allows a constant, the reload pass may 21374 replace a register with a constant permitted by the constraint in some 21375 cases. Similarly for memory references. Because of this substitution, 21376 you should not provide separate patterns for increment and decrement 21377 instructions. Instead, they should be generated from the same pattern 21378 that supports register-register add insns by examining the operands and 21379 generating the appropriate machine instruction. 21380 21381 21382 File: gccint.info, Node: Jump Patterns, Next: Looping Patterns, Prev: Dependent Patterns, Up: Machine Desc 21383 21384 16.12 Defining Jump Instruction Patterns 21385 ======================================== 21386 21387 For most machines, GCC assumes that the machine has a condition code. 21388 A comparison insn sets the condition code, recording the results of both 21389 signed and unsigned comparison of the given operands. A separate branch 21390 insn tests the condition code and branches or not according its value. 21391 The branch insns come in distinct signed and unsigned flavors. Many 21392 common machines, such as the VAX, the 68000 and the 32000, work this 21393 way. 21394 21395 Some machines have distinct signed and unsigned compare instructions, 21396 and only one set of conditional branch instructions. The easiest way 21397 to handle these machines is to treat them just like the others until 21398 the final stage where assembly code is written. At this time, when 21399 outputting code for the compare instruction, peek ahead at the 21400 following branch using `next_cc0_user (insn)'. (The variable `insn' 21401 refers to the insn being output, in the output-writing code in an 21402 instruction pattern.) If the RTL says that is an unsigned branch, 21403 output an unsigned compare; otherwise output a signed compare. When 21404 the branch itself is output, you can treat signed and unsigned branches 21405 identically. 21406 21407 The reason you can do this is that GCC always generates a pair of 21408 consecutive RTL insns, possibly separated by `note' insns, one to set 21409 the condition code and one to test it, and keeps the pair inviolate 21410 until the end. 21411 21412 To go with this technique, you must define the machine-description 21413 macro `NOTICE_UPDATE_CC' to do `CC_STATUS_INIT'; in other words, no 21414 compare instruction is superfluous. 21415 21416 Some machines have compare-and-branch instructions and no condition 21417 code. A similar technique works for them. When it is time to "output" 21418 a compare instruction, record its operands in two static variables. 21419 When outputting the branch-on-condition-code instruction that follows, 21420 actually output a compare-and-branch instruction that uses the 21421 remembered operands. 21422 21423 It also works to define patterns for compare-and-branch instructions. 21424 In optimizing compilation, the pair of compare and branch instructions 21425 will be combined according to these patterns. But this does not happen 21426 if optimization is not requested. So you must use one of the solutions 21427 above in addition to any special patterns you define. 21428 21429 In many RISC machines, most instructions do not affect the condition 21430 code and there may not even be a separate condition code register. On 21431 these machines, the restriction that the definition and use of the 21432 condition code be adjacent insns is not necessary and can prevent 21433 important optimizations. For example, on the IBM RS/6000, there is a 21434 delay for taken branches unless the condition code register is set three 21435 instructions earlier than the conditional branch. The instruction 21436 scheduler cannot perform this optimization if it is not permitted to 21437 separate the definition and use of the condition code register. 21438 21439 On these machines, do not use `(cc0)', but instead use a register to 21440 represent the condition code. If there is a specific condition code 21441 register in the machine, use a hard register. If the condition code or 21442 comparison result can be placed in any general register, or if there are 21443 multiple condition registers, use a pseudo register. 21444 21445 On some machines, the type of branch instruction generated may depend 21446 on the way the condition code was produced; for example, on the 68k and 21447 SPARC, setting the condition code directly from an add or subtract 21448 instruction does not clear the overflow bit the way that a test 21449 instruction does, so a different branch instruction must be used for 21450 some conditional branches. For machines that use `(cc0)', the set and 21451 use of the condition code must be adjacent (separated only by `note' 21452 insns) allowing flags in `cc_status' to be used. (*Note Condition 21453 Code::.) Also, the comparison and branch insns can be located from 21454 each other by using the functions `prev_cc0_setter' and `next_cc0_user'. 21455 21456 However, this is not true on machines that do not use `(cc0)'. On 21457 those machines, no assumptions can be made about the adjacency of the 21458 compare and branch insns and the above methods cannot be used. Instead, 21459 we use the machine mode of the condition code register to record 21460 different formats of the condition code register. 21461 21462 Registers used to store the condition code value should have a mode 21463 that is in class `MODE_CC'. Normally, it will be `CCmode'. If 21464 additional modes are required (as for the add example mentioned above in 21465 the SPARC), define them in `MACHINE-modes.def' (*note Condition 21466 Code::). Also define `SELECT_CC_MODE' to choose a mode given an 21467 operand of a compare. 21468 21469 If it is known during RTL generation that a different mode will be 21470 required (for example, if the machine has separate compare instructions 21471 for signed and unsigned quantities, like most IBM processors), they can 21472 be specified at that time. 21473 21474 If the cases that require different modes would be made by instruction 21475 combination, the macro `SELECT_CC_MODE' determines which machine mode 21476 should be used for the comparison result. The patterns should be 21477 written using that mode. To support the case of the add on the SPARC 21478 discussed above, we have the pattern 21479 21480 (define_insn "" 21481 [(set (reg:CC_NOOV 0) 21482 (compare:CC_NOOV 21483 (plus:SI (match_operand:SI 0 "register_operand" "%r") 21484 (match_operand:SI 1 "arith_operand" "rI")) 21485 (const_int 0)))] 21486 "" 21487 "...") 21488 21489 The `SELECT_CC_MODE' macro on the SPARC returns `CC_NOOVmode' for 21490 comparisons whose argument is a `plus'. 21491 21492 21493 File: gccint.info, Node: Looping Patterns, Next: Insn Canonicalizations, Prev: Jump Patterns, Up: Machine Desc 21494 21495 16.13 Defining Looping Instruction Patterns 21496 =========================================== 21497 21498 Some machines have special jump instructions that can be utilized to 21499 make loops more efficient. A common example is the 68000 `dbra' 21500 instruction which performs a decrement of a register and a branch if the 21501 result was greater than zero. Other machines, in particular digital 21502 signal processors (DSPs), have special block repeat instructions to 21503 provide low-overhead loop support. For example, the TI TMS320C3x/C4x 21504 DSPs have a block repeat instruction that loads special registers to 21505 mark the top and end of a loop and to count the number of loop 21506 iterations. This avoids the need for fetching and executing a 21507 `dbra'-like instruction and avoids pipeline stalls associated with the 21508 jump. 21509 21510 GCC has three special named patterns to support low overhead looping. 21511 They are `decrement_and_branch_until_zero', `doloop_begin', and 21512 `doloop_end'. The first pattern, `decrement_and_branch_until_zero', is 21513 not emitted during RTL generation but may be emitted during the 21514 instruction combination phase. This requires the assistance of the 21515 loop optimizer, using information collected during strength reduction, 21516 to reverse a loop to count down to zero. Some targets also require the 21517 loop optimizer to add a `REG_NONNEG' note to indicate that the 21518 iteration count is always positive. This is needed if the target 21519 performs a signed loop termination test. For example, the 68000 uses a 21520 pattern similar to the following for its `dbra' instruction: 21521 21522 (define_insn "decrement_and_branch_until_zero" 21523 [(set (pc) 21524 (if_then_else 21525 (ge (plus:SI (match_operand:SI 0 "general_operand" "+d*am") 21526 (const_int -1)) 21527 (const_int 0)) 21528 (label_ref (match_operand 1 "" "")) 21529 (pc))) 21530 (set (match_dup 0) 21531 (plus:SI (match_dup 0) 21532 (const_int -1)))] 21533 "find_reg_note (insn, REG_NONNEG, 0)" 21534 "...") 21535 21536 Note that since the insn is both a jump insn and has an output, it must 21537 deal with its own reloads, hence the `m' constraints. Also note that 21538 since this insn is generated by the instruction combination phase 21539 combining two sequential insns together into an implicit parallel insn, 21540 the iteration counter needs to be biased by the same amount as the 21541 decrement operation, in this case -1. Note that the following similar 21542 pattern will not be matched by the combiner. 21543 21544 (define_insn "decrement_and_branch_until_zero" 21545 [(set (pc) 21546 (if_then_else 21547 (ge (match_operand:SI 0 "general_operand" "+d*am") 21548 (const_int 1)) 21549 (label_ref (match_operand 1 "" "")) 21550 (pc))) 21551 (set (match_dup 0) 21552 (plus:SI (match_dup 0) 21553 (const_int -1)))] 21554 "find_reg_note (insn, REG_NONNEG, 0)" 21555 "...") 21556 21557 The other two special looping patterns, `doloop_begin' and 21558 `doloop_end', are emitted by the loop optimizer for certain 21559 well-behaved loops with a finite number of loop iterations using 21560 information collected during strength reduction. 21561 21562 The `doloop_end' pattern describes the actual looping instruction (or 21563 the implicit looping operation) and the `doloop_begin' pattern is an 21564 optional companion pattern that can be used for initialization needed 21565 for some low-overhead looping instructions. 21566 21567 Note that some machines require the actual looping instruction to be 21568 emitted at the top of the loop (e.g., the TMS320C3x/C4x DSPs). Emitting 21569 the true RTL for a looping instruction at the top of the loop can cause 21570 problems with flow analysis. So instead, a dummy `doloop' insn is 21571 emitted at the end of the loop. The machine dependent reorg pass checks 21572 for the presence of this `doloop' insn and then searches back to the 21573 top of the loop, where it inserts the true looping insn (provided there 21574 are no instructions in the loop which would cause problems). Any 21575 additional labels can be emitted at this point. In addition, if the 21576 desired special iteration counter register was not allocated, this 21577 machine dependent reorg pass could emit a traditional compare and jump 21578 instruction pair. 21579 21580 The essential difference between the `decrement_and_branch_until_zero' 21581 and the `doloop_end' patterns is that the loop optimizer allocates an 21582 additional pseudo register for the latter as an iteration counter. 21583 This pseudo register cannot be used within the loop (i.e., general 21584 induction variables cannot be derived from it), however, in many cases 21585 the loop induction variable may become redundant and removed by the 21586 flow pass. 21587 21588 21589 File: gccint.info, Node: Insn Canonicalizations, Next: Expander Definitions, Prev: Looping Patterns, Up: Machine Desc 21590 21591 16.14 Canonicalization of Instructions 21592 ====================================== 21593 21594 There are often cases where multiple RTL expressions could represent an 21595 operation performed by a single machine instruction. This situation is 21596 most commonly encountered with logical, branch, and multiply-accumulate 21597 instructions. In such cases, the compiler attempts to convert these 21598 multiple RTL expressions into a single canonical form to reduce the 21599 number of insn patterns required. 21600 21601 In addition to algebraic simplifications, following canonicalizations 21602 are performed: 21603 21604 * For commutative and comparison operators, a constant is always 21605 made the second operand. If a machine only supports a constant as 21606 the second operand, only patterns that match a constant in the 21607 second operand need be supplied. 21608 21609 * For associative operators, a sequence of operators will always 21610 chain to the left; for instance, only the left operand of an 21611 integer `plus' can itself be a `plus'. `and', `ior', `xor', 21612 `plus', `mult', `smin', `smax', `umin', and `umax' are associative 21613 when applied to integers, and sometimes to floating-point. 21614 21615 * For these operators, if only one operand is a `neg', `not', 21616 `mult', `plus', or `minus' expression, it will be the first 21617 operand. 21618 21619 * In combinations of `neg', `mult', `plus', and `minus', the `neg' 21620 operations (if any) will be moved inside the operations as far as 21621 possible. For instance, `(neg (mult A B))' is canonicalized as 21622 `(mult (neg A) B)', but `(plus (mult (neg A) B) C)' is 21623 canonicalized as `(minus A (mult B C))'. 21624 21625 * For the `compare' operator, a constant is always the second operand 21626 on machines where `cc0' is used (*note Jump Patterns::). On other 21627 machines, there are rare cases where the compiler might want to 21628 construct a `compare' with a constant as the first operand. 21629 However, these cases are not common enough for it to be worthwhile 21630 to provide a pattern matching a constant as the first operand 21631 unless the machine actually has such an instruction. 21632 21633 An operand of `neg', `not', `mult', `plus', or `minus' is made the 21634 first operand under the same conditions as above. 21635 21636 * `(ltu (plus A B) B)' is converted to `(ltu (plus A B) A)'. 21637 Likewise with `geu' instead of `ltu'. 21638 21639 * `(minus X (const_int N))' is converted to `(plus X (const_int 21640 -N))'. 21641 21642 * Within address computations (i.e., inside `mem'), a left shift is 21643 converted into the appropriate multiplication by a power of two. 21644 21645 * De Morgan's Law is used to move bitwise negation inside a bitwise 21646 logical-and or logical-or operation. If this results in only one 21647 operand being a `not' expression, it will be the first one. 21648 21649 A machine that has an instruction that performs a bitwise 21650 logical-and of one operand with the bitwise negation of the other 21651 should specify the pattern for that instruction as 21652 21653 (define_insn "" 21654 [(set (match_operand:M 0 ...) 21655 (and:M (not:M (match_operand:M 1 ...)) 21656 (match_operand:M 2 ...)))] 21657 "..." 21658 "...") 21659 21660 Similarly, a pattern for a "NAND" instruction should be written 21661 21662 (define_insn "" 21663 [(set (match_operand:M 0 ...) 21664 (ior:M (not:M (match_operand:M 1 ...)) 21665 (not:M (match_operand:M 2 ...))))] 21666 "..." 21667 "...") 21668 21669 In both cases, it is not necessary to include patterns for the many 21670 logically equivalent RTL expressions. 21671 21672 * The only possible RTL expressions involving both bitwise 21673 exclusive-or and bitwise negation are `(xor:M X Y)' and `(not:M 21674 (xor:M X Y))'. 21675 21676 * The sum of three items, one of which is a constant, will only 21677 appear in the form 21678 21679 (plus:M (plus:M X Y) CONSTANT) 21680 21681 * On machines that do not use `cc0', `(compare X (const_int 0))' 21682 will be converted to X. 21683 21684 * Equality comparisons of a group of bits (usually a single bit) 21685 with zero will be written using `zero_extract' rather than the 21686 equivalent `and' or `sign_extract' operations. 21687 21688 21689 Further canonicalization rules are defined in the function 21690 `commutative_operand_precedence' in `gcc/rtlanal.c'. 21691 21692 21693 File: gccint.info, Node: Expander Definitions, Next: Insn Splitting, Prev: Insn Canonicalizations, Up: Machine Desc 21694 21695 16.15 Defining RTL Sequences for Code Generation 21696 ================================================ 21697 21698 On some target machines, some standard pattern names for RTL generation 21699 cannot be handled with single insn, but a sequence of RTL insns can 21700 represent them. For these target machines, you can write a 21701 `define_expand' to specify how to generate the sequence of RTL. 21702 21703 A `define_expand' is an RTL expression that looks almost like a 21704 `define_insn'; but, unlike the latter, a `define_expand' is used only 21705 for RTL generation and it can produce more than one RTL insn. 21706 21707 A `define_expand' RTX has four operands: 21708 21709 * The name. Each `define_expand' must have a name, since the only 21710 use for it is to refer to it by name. 21711 21712 * The RTL template. This is a vector of RTL expressions representing 21713 a sequence of separate instructions. Unlike `define_insn', there 21714 is no implicit surrounding `PARALLEL'. 21715 21716 * The condition, a string containing a C expression. This 21717 expression is used to express how the availability of this pattern 21718 depends on subclasses of target machine, selected by command-line 21719 options when GCC is run. This is just like the condition of a 21720 `define_insn' that has a standard name. Therefore, the condition 21721 (if present) may not depend on the data in the insn being matched, 21722 but only the target-machine-type flags. The compiler needs to 21723 test these conditions during initialization in order to learn 21724 exactly which named instructions are available in a particular run. 21725 21726 * The preparation statements, a string containing zero or more C 21727 statements which are to be executed before RTL code is generated 21728 from the RTL template. 21729 21730 Usually these statements prepare temporary registers for use as 21731 internal operands in the RTL template, but they can also generate 21732 RTL insns directly by calling routines such as `emit_insn', etc. 21733 Any such insns precede the ones that come from the RTL template. 21734 21735 Every RTL insn emitted by a `define_expand' must match some 21736 `define_insn' in the machine description. Otherwise, the compiler will 21737 crash when trying to generate code for the insn or trying to optimize 21738 it. 21739 21740 The RTL template, in addition to controlling generation of RTL insns, 21741 also describes the operands that need to be specified when this pattern 21742 is used. In particular, it gives a predicate for each operand. 21743 21744 A true operand, which needs to be specified in order to generate RTL 21745 from the pattern, should be described with a `match_operand' in its 21746 first occurrence in the RTL template. This enters information on the 21747 operand's predicate into the tables that record such things. GCC uses 21748 the information to preload the operand into a register if that is 21749 required for valid RTL code. If the operand is referred to more than 21750 once, subsequent references should use `match_dup'. 21751 21752 The RTL template may also refer to internal "operands" which are 21753 temporary registers or labels used only within the sequence made by the 21754 `define_expand'. Internal operands are substituted into the RTL 21755 template with `match_dup', never with `match_operand'. The values of 21756 the internal operands are not passed in as arguments by the compiler 21757 when it requests use of this pattern. Instead, they are computed 21758 within the pattern, in the preparation statements. These statements 21759 compute the values and store them into the appropriate elements of 21760 `operands' so that `match_dup' can find them. 21761 21762 There are two special macros defined for use in the preparation 21763 statements: `DONE' and `FAIL'. Use them with a following semicolon, as 21764 a statement. 21765 21766 `DONE' 21767 Use the `DONE' macro to end RTL generation for the pattern. The 21768 only RTL insns resulting from the pattern on this occasion will be 21769 those already emitted by explicit calls to `emit_insn' within the 21770 preparation statements; the RTL template will not be generated. 21771 21772 `FAIL' 21773 Make the pattern fail on this occasion. When a pattern fails, it 21774 means that the pattern was not truly available. The calling 21775 routines in the compiler will try other strategies for code 21776 generation using other patterns. 21777 21778 Failure is currently supported only for binary (addition, 21779 multiplication, shifting, etc.) and bit-field (`extv', `extzv', 21780 and `insv') operations. 21781 21782 If the preparation falls through (invokes neither `DONE' nor `FAIL'), 21783 then the `define_expand' acts like a `define_insn' in that the RTL 21784 template is used to generate the insn. 21785 21786 The RTL template is not used for matching, only for generating the 21787 initial insn list. If the preparation statement always invokes `DONE' 21788 or `FAIL', the RTL template may be reduced to a simple list of 21789 operands, such as this example: 21790 21791 (define_expand "addsi3" 21792 [(match_operand:SI 0 "register_operand" "") 21793 (match_operand:SI 1 "register_operand" "") 21794 (match_operand:SI 2 "register_operand" "")] 21795 "" 21796 " 21797 { 21798 handle_add (operands[0], operands[1], operands[2]); 21799 DONE; 21800 }") 21801 21802 Here is an example, the definition of left-shift for the SPUR chip: 21803 21804 (define_expand "ashlsi3" 21805 [(set (match_operand:SI 0 "register_operand" "") 21806 (ashift:SI 21807 (match_operand:SI 1 "register_operand" "") 21808 (match_operand:SI 2 "nonmemory_operand" "")))] 21809 "" 21810 " 21811 21812 { 21813 if (GET_CODE (operands[2]) != CONST_INT 21814 || (unsigned) INTVAL (operands[2]) > 3) 21815 FAIL; 21816 }") 21817 21818 This example uses `define_expand' so that it can generate an RTL insn 21819 for shifting when the shift-count is in the supported range of 0 to 3 21820 but fail in other cases where machine insns aren't available. When it 21821 fails, the compiler tries another strategy using different patterns 21822 (such as, a library call). 21823 21824 If the compiler were able to handle nontrivial condition-strings in 21825 patterns with names, then it would be possible to use a `define_insn' 21826 in that case. Here is another case (zero-extension on the 68000) which 21827 makes more use of the power of `define_expand': 21828 21829 (define_expand "zero_extendhisi2" 21830 [(set (match_operand:SI 0 "general_operand" "") 21831 (const_int 0)) 21832 (set (strict_low_part 21833 (subreg:HI 21834 (match_dup 0) 21835 0)) 21836 (match_operand:HI 1 "general_operand" ""))] 21837 "" 21838 "operands[1] = make_safe_from (operands[1], operands[0]);") 21839 21840 Here two RTL insns are generated, one to clear the entire output operand 21841 and the other to copy the input operand into its low half. This 21842 sequence is incorrect if the input operand refers to [the old value of] 21843 the output operand, so the preparation statement makes sure this isn't 21844 so. The function `make_safe_from' copies the `operands[1]' into a 21845 temporary register if it refers to `operands[0]'. It does this by 21846 emitting another RTL insn. 21847 21848 Finally, a third example shows the use of an internal operand. 21849 Zero-extension on the SPUR chip is done by `and'-ing the result against 21850 a halfword mask. But this mask cannot be represented by a `const_int' 21851 because the constant value is too large to be legitimate on this 21852 machine. So it must be copied into a register with `force_reg' and 21853 then the register used in the `and'. 21854 21855 (define_expand "zero_extendhisi2" 21856 [(set (match_operand:SI 0 "register_operand" "") 21857 (and:SI (subreg:SI 21858 (match_operand:HI 1 "register_operand" "") 21859 0) 21860 (match_dup 2)))] 21861 "" 21862 "operands[2] 21863 = force_reg (SImode, GEN_INT (65535)); ") 21864 21865 _Note:_ If the `define_expand' is used to serve a standard binary or 21866 unary arithmetic operation or a bit-field operation, then the last insn 21867 it generates must not be a `code_label', `barrier' or `note'. It must 21868 be an `insn', `jump_insn' or `call_insn'. If you don't need a real insn 21869 at the end, emit an insn to copy the result of the operation into 21870 itself. Such an insn will generate no code, but it can avoid problems 21871 in the compiler. 21872 21873 21874 File: gccint.info, Node: Insn Splitting, Next: Including Patterns, Prev: Expander Definitions, Up: Machine Desc 21875 21876 16.16 Defining How to Split Instructions 21877 ======================================== 21878 21879 There are two cases where you should specify how to split a pattern 21880 into multiple insns. On machines that have instructions requiring 21881 delay slots (*note Delay Slots::) or that have instructions whose 21882 output is not available for multiple cycles (*note Processor pipeline 21883 description::), the compiler phases that optimize these cases need to 21884 be able to move insns into one-instruction delay slots. However, some 21885 insns may generate more than one machine instruction. These insns 21886 cannot be placed into a delay slot. 21887 21888 Often you can rewrite the single insn as a list of individual insns, 21889 each corresponding to one machine instruction. The disadvantage of 21890 doing so is that it will cause the compilation to be slower and require 21891 more space. If the resulting insns are too complex, it may also 21892 suppress some optimizations. The compiler splits the insn if there is a 21893 reason to believe that it might improve instruction or delay slot 21894 scheduling. 21895 21896 The insn combiner phase also splits putative insns. If three insns are 21897 merged into one insn with a complex expression that cannot be matched by 21898 some `define_insn' pattern, the combiner phase attempts to split the 21899 complex pattern into two insns that are recognized. Usually it can 21900 break the complex pattern into two patterns by splitting out some 21901 subexpression. However, in some other cases, such as performing an 21902 addition of a large constant in two insns on a RISC machine, the way to 21903 split the addition into two insns is machine-dependent. 21904 21905 The `define_split' definition tells the compiler how to split a 21906 complex insn into several simpler insns. It looks like this: 21907 21908 (define_split 21909 [INSN-PATTERN] 21910 "CONDITION" 21911 [NEW-INSN-PATTERN-1 21912 NEW-INSN-PATTERN-2 21913 ...] 21914 "PREPARATION-STATEMENTS") 21915 21916 INSN-PATTERN is a pattern that needs to be split and CONDITION is the 21917 final condition to be tested, as in a `define_insn'. When an insn 21918 matching INSN-PATTERN and satisfying CONDITION is found, it is replaced 21919 in the insn list with the insns given by NEW-INSN-PATTERN-1, 21920 NEW-INSN-PATTERN-2, etc. 21921 21922 The PREPARATION-STATEMENTS are similar to those statements that are 21923 specified for `define_expand' (*note Expander Definitions::) and are 21924 executed before the new RTL is generated to prepare for the generated 21925 code or emit some insns whose pattern is not fixed. Unlike those in 21926 `define_expand', however, these statements must not generate any new 21927 pseudo-registers. Once reload has completed, they also must not 21928 allocate any space in the stack frame. 21929 21930 Patterns are matched against INSN-PATTERN in two different 21931 circumstances. If an insn needs to be split for delay slot scheduling 21932 or insn scheduling, the insn is already known to be valid, which means 21933 that it must have been matched by some `define_insn' and, if 21934 `reload_completed' is nonzero, is known to satisfy the constraints of 21935 that `define_insn'. In that case, the new insn patterns must also be 21936 insns that are matched by some `define_insn' and, if `reload_completed' 21937 is nonzero, must also satisfy the constraints of those definitions. 21938 21939 As an example of this usage of `define_split', consider the following 21940 example from `a29k.md', which splits a `sign_extend' from `HImode' to 21941 `SImode' into a pair of shift insns: 21942 21943 (define_split 21944 [(set (match_operand:SI 0 "gen_reg_operand" "") 21945 (sign_extend:SI (match_operand:HI 1 "gen_reg_operand" "")))] 21946 "" 21947 [(set (match_dup 0) 21948 (ashift:SI (match_dup 1) 21949 (const_int 16))) 21950 (set (match_dup 0) 21951 (ashiftrt:SI (match_dup 0) 21952 (const_int 16)))] 21953 " 21954 { operands[1] = gen_lowpart (SImode, operands[1]); }") 21955 21956 When the combiner phase tries to split an insn pattern, it is always 21957 the case that the pattern is _not_ matched by any `define_insn'. The 21958 combiner pass first tries to split a single `set' expression and then 21959 the same `set' expression inside a `parallel', but followed by a 21960 `clobber' of a pseudo-reg to use as a scratch register. In these 21961 cases, the combiner expects exactly two new insn patterns to be 21962 generated. It will verify that these patterns match some `define_insn' 21963 definitions, so you need not do this test in the `define_split' (of 21964 course, there is no point in writing a `define_split' that will never 21965 produce insns that match). 21966 21967 Here is an example of this use of `define_split', taken from 21968 `rs6000.md': 21969 21970 (define_split 21971 [(set (match_operand:SI 0 "gen_reg_operand" "") 21972 (plus:SI (match_operand:SI 1 "gen_reg_operand" "") 21973 (match_operand:SI 2 "non_add_cint_operand" "")))] 21974 "" 21975 [(set (match_dup 0) (plus:SI (match_dup 1) (match_dup 3))) 21976 (set (match_dup 0) (plus:SI (match_dup 0) (match_dup 4)))] 21977 " 21978 { 21979 int low = INTVAL (operands[2]) & 0xffff; 21980 int high = (unsigned) INTVAL (operands[2]) >> 16; 21981 21982 if (low & 0x8000) 21983 high++, low |= 0xffff0000; 21984 21985 operands[3] = GEN_INT (high << 16); 21986 operands[4] = GEN_INT (low); 21987 }") 21988 21989 Here the predicate `non_add_cint_operand' matches any `const_int' that 21990 is _not_ a valid operand of a single add insn. The add with the 21991 smaller displacement is written so that it can be substituted into the 21992 address of a subsequent operation. 21993 21994 An example that uses a scratch register, from the same file, generates 21995 an equality comparison of a register and a large constant: 21996 21997 (define_split 21998 [(set (match_operand:CC 0 "cc_reg_operand" "") 21999 (compare:CC (match_operand:SI 1 "gen_reg_operand" "") 22000 (match_operand:SI 2 "non_short_cint_operand" ""))) 22001 (clobber (match_operand:SI 3 "gen_reg_operand" ""))] 22002 "find_single_use (operands[0], insn, 0) 22003 && (GET_CODE (*find_single_use (operands[0], insn, 0)) == EQ 22004 || GET_CODE (*find_single_use (operands[0], insn, 0)) == NE)" 22005 [(set (match_dup 3) (xor:SI (match_dup 1) (match_dup 4))) 22006 (set (match_dup 0) (compare:CC (match_dup 3) (match_dup 5)))] 22007 " 22008 { 22009 /* Get the constant we are comparing against, C, and see what it 22010 looks like sign-extended to 16 bits. Then see what constant 22011 could be XOR'ed with C to get the sign-extended value. */ 22012 22013 int c = INTVAL (operands[2]); 22014 int sextc = (c << 16) >> 16; 22015 int xorv = c ^ sextc; 22016 22017 operands[4] = GEN_INT (xorv); 22018 operands[5] = GEN_INT (sextc); 22019 }") 22020 22021 To avoid confusion, don't write a single `define_split' that accepts 22022 some insns that match some `define_insn' as well as some insns that 22023 don't. Instead, write two separate `define_split' definitions, one for 22024 the insns that are valid and one for the insns that are not valid. 22025 22026 The splitter is allowed to split jump instructions into sequence of 22027 jumps or create new jumps in while splitting non-jump instructions. As 22028 the central flowgraph and branch prediction information needs to be 22029 updated, several restriction apply. 22030 22031 Splitting of jump instruction into sequence that over by another jump 22032 instruction is always valid, as compiler expect identical behavior of 22033 new jump. When new sequence contains multiple jump instructions or new 22034 labels, more assistance is needed. Splitter is required to create only 22035 unconditional jumps, or simple conditional jump instructions. 22036 Additionally it must attach a `REG_BR_PROB' note to each conditional 22037 jump. A global variable `split_branch_probability' holds the 22038 probability of the original branch in case it was an simple conditional 22039 jump, -1 otherwise. To simplify recomputing of edge frequencies, the 22040 new sequence is required to have only forward jumps to the newly 22041 created labels. 22042 22043 For the common case where the pattern of a define_split exactly 22044 matches the pattern of a define_insn, use `define_insn_and_split'. It 22045 looks like this: 22046 22047 (define_insn_and_split 22048 [INSN-PATTERN] 22049 "CONDITION" 22050 "OUTPUT-TEMPLATE" 22051 "SPLIT-CONDITION" 22052 [NEW-INSN-PATTERN-1 22053 NEW-INSN-PATTERN-2 22054 ...] 22055 "PREPARATION-STATEMENTS" 22056 [INSN-ATTRIBUTES]) 22057 22058 INSN-PATTERN, CONDITION, OUTPUT-TEMPLATE, and INSN-ATTRIBUTES are used 22059 as in `define_insn'. The NEW-INSN-PATTERN vector and the 22060 PREPARATION-STATEMENTS are used as in a `define_split'. The 22061 SPLIT-CONDITION is also used as in `define_split', with the additional 22062 behavior that if the condition starts with `&&', the condition used for 22063 the split will be the constructed as a logical "and" of the split 22064 condition with the insn condition. For example, from i386.md: 22065 22066 (define_insn_and_split "zero_extendhisi2_and" 22067 [(set (match_operand:SI 0 "register_operand" "=r") 22068 (zero_extend:SI (match_operand:HI 1 "register_operand" "0"))) 22069 (clobber (reg:CC 17))] 22070 "TARGET_ZERO_EXTEND_WITH_AND && !optimize_size" 22071 "#" 22072 "&& reload_completed" 22073 [(parallel [(set (match_dup 0) 22074 (and:SI (match_dup 0) (const_int 65535))) 22075 (clobber (reg:CC 17))])] 22076 "" 22077 [(set_attr "type" "alu1")]) 22078 22079 In this case, the actual split condition will be 22080 `TARGET_ZERO_EXTEND_WITH_AND && !optimize_size && reload_completed'. 22081 22082 The `define_insn_and_split' construction provides exactly the same 22083 functionality as two separate `define_insn' and `define_split' 22084 patterns. It exists for compactness, and as a maintenance tool to 22085 prevent having to ensure the two patterns' templates match. 22086 22087 22088 File: gccint.info, Node: Including Patterns, Next: Peephole Definitions, Prev: Insn Splitting, Up: Machine Desc 22089 22090 16.17 Including Patterns in Machine Descriptions. 22091 ================================================= 22092 22093 The `include' pattern tells the compiler tools where to look for 22094 patterns that are in files other than in the file `.md'. This is used 22095 only at build time and there is no preprocessing allowed. 22096 22097 It looks like: 22098 22099 22100 (include 22101 PATHNAME) 22102 22103 For example: 22104 22105 22106 (include "filestuff") 22107 22108 Where PATHNAME is a string that specifies the location of the file, 22109 specifies the include file to be in `gcc/config/target/filestuff'. The 22110 directory `gcc/config/target' is regarded as the default directory. 22111 22112 Machine descriptions may be split up into smaller more manageable 22113 subsections and placed into subdirectories. 22114 22115 By specifying: 22116 22117 22118 (include "BOGUS/filestuff") 22119 22120 the include file is specified to be in 22121 `gcc/config/TARGET/BOGUS/filestuff'. 22122 22123 Specifying an absolute path for the include file such as; 22124 22125 (include "/u2/BOGUS/filestuff") 22126 is permitted but is not encouraged. 22127 22128 16.17.1 RTL Generation Tool Options for Directory Search 22129 -------------------------------------------------------- 22130 22131 The `-IDIR' option specifies directories to search for machine 22132 descriptions. For example: 22133 22134 22135 genrecog -I/p1/abc/proc1 -I/p2/abcd/pro2 target.md 22136 22137 Add the directory DIR to the head of the list of directories to be 22138 searched for header files. This can be used to override a system 22139 machine definition file, substituting your own version, since these 22140 directories are searched before the default machine description file 22141 directories. If you use more than one `-I' option, the directories are 22142 scanned in left-to-right order; the standard default directory come 22143 after. 22144 22145 22146 File: gccint.info, Node: Peephole Definitions, Next: Insn Attributes, Prev: Including Patterns, Up: Machine Desc 22147 22148 16.18 Machine-Specific Peephole Optimizers 22149 ========================================== 22150 22151 In addition to instruction patterns the `md' file may contain 22152 definitions of machine-specific peephole optimizations. 22153 22154 The combiner does not notice certain peephole optimizations when the 22155 data flow in the program does not suggest that it should try them. For 22156 example, sometimes two consecutive insns related in purpose can be 22157 combined even though the second one does not appear to use a register 22158 computed in the first one. A machine-specific peephole optimizer can 22159 detect such opportunities. 22160 22161 There are two forms of peephole definitions that may be used. The 22162 original `define_peephole' is run at assembly output time to match 22163 insns and substitute assembly text. Use of `define_peephole' is 22164 deprecated. 22165 22166 A newer `define_peephole2' matches insns and substitutes new insns. 22167 The `peephole2' pass is run after register allocation but before 22168 scheduling, which may result in much better code for targets that do 22169 scheduling. 22170 22171 * Menu: 22172 22173 * define_peephole:: RTL to Text Peephole Optimizers 22174 * define_peephole2:: RTL to RTL Peephole Optimizers 22175 22176 22177 File: gccint.info, Node: define_peephole, Next: define_peephole2, Up: Peephole Definitions 22178 22179 16.18.1 RTL to Text Peephole Optimizers 22180 --------------------------------------- 22181 22182 A definition looks like this: 22183 22184 (define_peephole 22185 [INSN-PATTERN-1 22186 INSN-PATTERN-2 22187 ...] 22188 "CONDITION" 22189 "TEMPLATE" 22190 "OPTIONAL-INSN-ATTRIBUTES") 22191 22192 The last string operand may be omitted if you are not using any 22193 machine-specific information in this machine description. If present, 22194 it must obey the same rules as in a `define_insn'. 22195 22196 In this skeleton, INSN-PATTERN-1 and so on are patterns to match 22197 consecutive insns. The optimization applies to a sequence of insns when 22198 INSN-PATTERN-1 matches the first one, INSN-PATTERN-2 matches the next, 22199 and so on. 22200 22201 Each of the insns matched by a peephole must also match a 22202 `define_insn'. Peepholes are checked only at the last stage just 22203 before code generation, and only optionally. Therefore, any insn which 22204 would match a peephole but no `define_insn' will cause a crash in code 22205 generation in an unoptimized compilation, or at various optimization 22206 stages. 22207 22208 The operands of the insns are matched with `match_operands', 22209 `match_operator', and `match_dup', as usual. What is not usual is that 22210 the operand numbers apply to all the insn patterns in the definition. 22211 So, you can check for identical operands in two insns by using 22212 `match_operand' in one insn and `match_dup' in the other. 22213 22214 The operand constraints used in `match_operand' patterns do not have 22215 any direct effect on the applicability of the peephole, but they will 22216 be validated afterward, so make sure your constraints are general enough 22217 to apply whenever the peephole matches. If the peephole matches but 22218 the constraints are not satisfied, the compiler will crash. 22219 22220 It is safe to omit constraints in all the operands of the peephole; or 22221 you can write constraints which serve as a double-check on the criteria 22222 previously tested. 22223 22224 Once a sequence of insns matches the patterns, the CONDITION is 22225 checked. This is a C expression which makes the final decision whether 22226 to perform the optimization (we do so if the expression is nonzero). If 22227 CONDITION is omitted (in other words, the string is empty) then the 22228 optimization is applied to every sequence of insns that matches the 22229 patterns. 22230 22231 The defined peephole optimizations are applied after register 22232 allocation is complete. Therefore, the peephole definition can check 22233 which operands have ended up in which kinds of registers, just by 22234 looking at the operands. 22235 22236 The way to refer to the operands in CONDITION is to write 22237 `operands[I]' for operand number I (as matched by `(match_operand I 22238 ...)'). Use the variable `insn' to refer to the last of the insns 22239 being matched; use `prev_active_insn' to find the preceding insns. 22240 22241 When optimizing computations with intermediate results, you can use 22242 CONDITION to match only when the intermediate results are not used 22243 elsewhere. Use the C expression `dead_or_set_p (INSN, OP)', where INSN 22244 is the insn in which you expect the value to be used for the last time 22245 (from the value of `insn', together with use of `prev_nonnote_insn'), 22246 and OP is the intermediate value (from `operands[I]'). 22247 22248 Applying the optimization means replacing the sequence of insns with 22249 one new insn. The TEMPLATE controls ultimate output of assembler code 22250 for this combined insn. It works exactly like the template of a 22251 `define_insn'. Operand numbers in this template are the same ones used 22252 in matching the original sequence of insns. 22253 22254 The result of a defined peephole optimizer does not need to match any 22255 of the insn patterns in the machine description; it does not even have 22256 an opportunity to match them. The peephole optimizer definition itself 22257 serves as the insn pattern to control how the insn is output. 22258 22259 Defined peephole optimizers are run as assembler code is being output, 22260 so the insns they produce are never combined or rearranged in any way. 22261 22262 Here is an example, taken from the 68000 machine description: 22263 22264 (define_peephole 22265 [(set (reg:SI 15) (plus:SI (reg:SI 15) (const_int 4))) 22266 (set (match_operand:DF 0 "register_operand" "=f") 22267 (match_operand:DF 1 "register_operand" "ad"))] 22268 "FP_REG_P (operands[0]) && ! FP_REG_P (operands[1])" 22269 { 22270 rtx xoperands[2]; 22271 xoperands[1] = gen_rtx_REG (SImode, REGNO (operands[1]) + 1); 22272 #ifdef MOTOROLA 22273 output_asm_insn ("move.l %1,(sp)", xoperands); 22274 output_asm_insn ("move.l %1,-(sp)", operands); 22275 return "fmove.d (sp)+,%0"; 22276 #else 22277 output_asm_insn ("movel %1,sp@", xoperands); 22278 output_asm_insn ("movel %1,sp@-", operands); 22279 return "fmoved sp@+,%0"; 22280 #endif 22281 }) 22282 22283 The effect of this optimization is to change 22284 22285 jbsr _foobar 22286 addql #4,sp 22287 movel d1,sp@- 22288 movel d0,sp@- 22289 fmoved sp@+,fp0 22290 22291 into 22292 22293 jbsr _foobar 22294 movel d1,sp@ 22295 movel d0,sp@- 22296 fmoved sp@+,fp0 22297 22298 INSN-PATTERN-1 and so on look _almost_ like the second operand of 22299 `define_insn'. There is one important difference: the second operand 22300 of `define_insn' consists of one or more RTX's enclosed in square 22301 brackets. Usually, there is only one: then the same action can be 22302 written as an element of a `define_peephole'. But when there are 22303 multiple actions in a `define_insn', they are implicitly enclosed in a 22304 `parallel'. Then you must explicitly write the `parallel', and the 22305 square brackets within it, in the `define_peephole'. Thus, if an insn 22306 pattern looks like this, 22307 22308 (define_insn "divmodsi4" 22309 [(set (match_operand:SI 0 "general_operand" "=d") 22310 (div:SI (match_operand:SI 1 "general_operand" "0") 22311 (match_operand:SI 2 "general_operand" "dmsK"))) 22312 (set (match_operand:SI 3 "general_operand" "=d") 22313 (mod:SI (match_dup 1) (match_dup 2)))] 22314 "TARGET_68020" 22315 "divsl%.l %2,%3:%0") 22316 22317 then the way to mention this insn in a peephole is as follows: 22318 22319 (define_peephole 22320 [... 22321 (parallel 22322 [(set (match_operand:SI 0 "general_operand" "=d") 22323 (div:SI (match_operand:SI 1 "general_operand" "0") 22324 (match_operand:SI 2 "general_operand" "dmsK"))) 22325 (set (match_operand:SI 3 "general_operand" "=d") 22326 (mod:SI (match_dup 1) (match_dup 2)))]) 22327 ...] 22328 ...) 22329 22330 22331 File: gccint.info, Node: define_peephole2, Prev: define_peephole, Up: Peephole Definitions 22332 22333 16.18.2 RTL to RTL Peephole Optimizers 22334 -------------------------------------- 22335 22336 The `define_peephole2' definition tells the compiler how to substitute 22337 one sequence of instructions for another sequence, what additional 22338 scratch registers may be needed and what their lifetimes must be. 22339 22340 (define_peephole2 22341 [INSN-PATTERN-1 22342 INSN-PATTERN-2 22343 ...] 22344 "CONDITION" 22345 [NEW-INSN-PATTERN-1 22346 NEW-INSN-PATTERN-2 22347 ...] 22348 "PREPARATION-STATEMENTS") 22349 22350 The definition is almost identical to `define_split' (*note Insn 22351 Splitting::) except that the pattern to match is not a single 22352 instruction, but a sequence of instructions. 22353 22354 It is possible to request additional scratch registers for use in the 22355 output template. If appropriate registers are not free, the pattern 22356 will simply not match. 22357 22358 Scratch registers are requested with a `match_scratch' pattern at the 22359 top level of the input pattern. The allocated register (initially) will 22360 be dead at the point requested within the original sequence. If the 22361 scratch is used at more than a single point, a `match_dup' pattern at 22362 the top level of the input pattern marks the last position in the input 22363 sequence at which the register must be available. 22364 22365 Here is an example from the IA-32 machine description: 22366 22367 (define_peephole2 22368 [(match_scratch:SI 2 "r") 22369 (parallel [(set (match_operand:SI 0 "register_operand" "") 22370 (match_operator:SI 3 "arith_or_logical_operator" 22371 [(match_dup 0) 22372 (match_operand:SI 1 "memory_operand" "")])) 22373 (clobber (reg:CC 17))])] 22374 "! optimize_size && ! TARGET_READ_MODIFY" 22375 [(set (match_dup 2) (match_dup 1)) 22376 (parallel [(set (match_dup 0) 22377 (match_op_dup 3 [(match_dup 0) (match_dup 2)])) 22378 (clobber (reg:CC 17))])] 22379 "") 22380 22381 This pattern tries to split a load from its use in the hopes that we'll 22382 be able to schedule around the memory load latency. It allocates a 22383 single `SImode' register of class `GENERAL_REGS' (`"r"') that needs to 22384 be live only at the point just before the arithmetic. 22385 22386 A real example requiring extended scratch lifetimes is harder to come 22387 by, so here's a silly made-up example: 22388 22389 (define_peephole2 22390 [(match_scratch:SI 4 "r") 22391 (set (match_operand:SI 0 "" "") (match_operand:SI 1 "" "")) 22392 (set (match_operand:SI 2 "" "") (match_dup 1)) 22393 (match_dup 4) 22394 (set (match_operand:SI 3 "" "") (match_dup 1))] 22395 "/* determine 1 does not overlap 0 and 2 */" 22396 [(set (match_dup 4) (match_dup 1)) 22397 (set (match_dup 0) (match_dup 4)) 22398 (set (match_dup 2) (match_dup 4))] 22399 (set (match_dup 3) (match_dup 4))] 22400 "") 22401 22402 If we had not added the `(match_dup 4)' in the middle of the input 22403 sequence, it might have been the case that the register we chose at the 22404 beginning of the sequence is killed by the first or second `set'. 22405 22406 22407 File: gccint.info, Node: Insn Attributes, Next: Conditional Execution, Prev: Peephole Definitions, Up: Machine Desc 22408 22409 16.19 Instruction Attributes 22410 ============================ 22411 22412 In addition to describing the instruction supported by the target 22413 machine, the `md' file also defines a group of "attributes" and a set of 22414 values for each. Every generated insn is assigned a value for each 22415 attribute. One possible attribute would be the effect that the insn 22416 has on the machine's condition code. This attribute can then be used 22417 by `NOTICE_UPDATE_CC' to track the condition codes. 22418 22419 * Menu: 22420 22421 * Defining Attributes:: Specifying attributes and their values. 22422 * Expressions:: Valid expressions for attribute values. 22423 * Tagging Insns:: Assigning attribute values to insns. 22424 * Attr Example:: An example of assigning attributes. 22425 * Insn Lengths:: Computing the length of insns. 22426 * Constant Attributes:: Defining attributes that are constant. 22427 * Delay Slots:: Defining delay slots required for a machine. 22428 * Processor pipeline description:: Specifying information for insn scheduling. 22429 22430 22431 File: gccint.info, Node: Defining Attributes, Next: Expressions, Up: Insn Attributes 22432 22433 16.19.1 Defining Attributes and their Values 22434 -------------------------------------------- 22435 22436 The `define_attr' expression is used to define each attribute required 22437 by the target machine. It looks like: 22438 22439 (define_attr NAME LIST-OF-VALUES DEFAULT) 22440 22441 NAME is a string specifying the name of the attribute being defined. 22442 22443 LIST-OF-VALUES is either a string that specifies a comma-separated 22444 list of values that can be assigned to the attribute, or a null string 22445 to indicate that the attribute takes numeric values. 22446 22447 DEFAULT is an attribute expression that gives the value of this 22448 attribute for insns that match patterns whose definition does not 22449 include an explicit value for this attribute. *Note Attr Example::, 22450 for more information on the handling of defaults. *Note Constant 22451 Attributes::, for information on attributes that do not depend on any 22452 particular insn. 22453 22454 For each defined attribute, a number of definitions are written to the 22455 `insn-attr.h' file. For cases where an explicit set of values is 22456 specified for an attribute, the following are defined: 22457 22458 * A `#define' is written for the symbol `HAVE_ATTR_NAME'. 22459 22460 * An enumerated class is defined for `attr_NAME' with elements of 22461 the form `UPPER-NAME_UPPER-VALUE' where the attribute name and 22462 value are first converted to uppercase. 22463 22464 * A function `get_attr_NAME' is defined that is passed an insn and 22465 returns the attribute value for that insn. 22466 22467 For example, if the following is present in the `md' file: 22468 22469 (define_attr "type" "branch,fp,load,store,arith" ...) 22470 22471 the following lines will be written to the file `insn-attr.h'. 22472 22473 #define HAVE_ATTR_type 22474 enum attr_type {TYPE_BRANCH, TYPE_FP, TYPE_LOAD, 22475 TYPE_STORE, TYPE_ARITH}; 22476 extern enum attr_type get_attr_type (); 22477 22478 If the attribute takes numeric values, no `enum' type will be defined 22479 and the function to obtain the attribute's value will return `int'. 22480 22481 There are attributes which are tied to a specific meaning. These 22482 attributes are not free to use for other purposes: 22483 22484 `length' 22485 The `length' attribute is used to calculate the length of emitted 22486 code chunks. This is especially important when verifying branch 22487 distances. *Note Insn Lengths::. 22488 22489 `enabled' 22490 The `enabled' attribute can be defined to prevent certain 22491 alternatives of an insn definition from being used during code 22492 generation. *Note Disable Insn Alternatives::. 22493 22494 22495 22496 File: gccint.info, Node: Expressions, Next: Tagging Insns, Prev: Defining Attributes, Up: Insn Attributes 22497 22498 16.19.2 Attribute Expressions 22499 ----------------------------- 22500 22501 RTL expressions used to define attributes use the codes described above 22502 plus a few specific to attribute definitions, to be discussed below. 22503 Attribute value expressions must have one of the following forms: 22504 22505 `(const_int I)' 22506 The integer I specifies the value of a numeric attribute. I must 22507 be non-negative. 22508 22509 The value of a numeric attribute can be specified either with a 22510 `const_int', or as an integer represented as a string in 22511 `const_string', `eq_attr' (see below), `attr', `symbol_ref', 22512 simple arithmetic expressions, and `set_attr' overrides on 22513 specific instructions (*note Tagging Insns::). 22514 22515 `(const_string VALUE)' 22516 The string VALUE specifies a constant attribute value. If VALUE 22517 is specified as `"*"', it means that the default value of the 22518 attribute is to be used for the insn containing this expression. 22519 `"*"' obviously cannot be used in the DEFAULT expression of a 22520 `define_attr'. 22521 22522 If the attribute whose value is being specified is numeric, VALUE 22523 must be a string containing a non-negative integer (normally 22524 `const_int' would be used in this case). Otherwise, it must 22525 contain one of the valid values for the attribute. 22526 22527 `(if_then_else TEST TRUE-VALUE FALSE-VALUE)' 22528 TEST specifies an attribute test, whose format is defined below. 22529 The value of this expression is TRUE-VALUE if TEST is true, 22530 otherwise it is FALSE-VALUE. 22531 22532 `(cond [TEST1 VALUE1 ...] DEFAULT)' 22533 The first operand of this expression is a vector containing an even 22534 number of expressions and consisting of pairs of TEST and VALUE 22535 expressions. The value of the `cond' expression is that of the 22536 VALUE corresponding to the first true TEST expression. If none of 22537 the TEST expressions are true, the value of the `cond' expression 22538 is that of the DEFAULT expression. 22539 22540 TEST expressions can have one of the following forms: 22541 22542 `(const_int I)' 22543 This test is true if I is nonzero and false otherwise. 22544 22545 `(not TEST)' 22546 `(ior TEST1 TEST2)' 22547 `(and TEST1 TEST2)' 22548 These tests are true if the indicated logical function is true. 22549 22550 `(match_operand:M N PRED CONSTRAINTS)' 22551 This test is true if operand N of the insn whose attribute value 22552 is being determined has mode M (this part of the test is ignored 22553 if M is `VOIDmode') and the function specified by the string PRED 22554 returns a nonzero value when passed operand N and mode M (this 22555 part of the test is ignored if PRED is the null string). 22556 22557 The CONSTRAINTS operand is ignored and should be the null string. 22558 22559 `(le ARITH1 ARITH2)' 22560 `(leu ARITH1 ARITH2)' 22561 `(lt ARITH1 ARITH2)' 22562 `(ltu ARITH1 ARITH2)' 22563 `(gt ARITH1 ARITH2)' 22564 `(gtu ARITH1 ARITH2)' 22565 `(ge ARITH1 ARITH2)' 22566 `(geu ARITH1 ARITH2)' 22567 `(ne ARITH1 ARITH2)' 22568 `(eq ARITH1 ARITH2)' 22569 These tests are true if the indicated comparison of the two 22570 arithmetic expressions is true. Arithmetic expressions are formed 22571 with `plus', `minus', `mult', `div', `mod', `abs', `neg', `and', 22572 `ior', `xor', `not', `ashift', `lshiftrt', and `ashiftrt' 22573 expressions. 22574 22575 `const_int' and `symbol_ref' are always valid terms (*note Insn 22576 Lengths::,for additional forms). `symbol_ref' is a string 22577 denoting a C expression that yields an `int' when evaluated by the 22578 `get_attr_...' routine. It should normally be a global variable. 22579 22580 `(eq_attr NAME VALUE)' 22581 NAME is a string specifying the name of an attribute. 22582 22583 VALUE is a string that is either a valid value for attribute NAME, 22584 a comma-separated list of values, or `!' followed by a value or 22585 list. If VALUE does not begin with a `!', this test is true if 22586 the value of the NAME attribute of the current insn is in the list 22587 specified by VALUE. If VALUE begins with a `!', this test is true 22588 if the attribute's value is _not_ in the specified list. 22589 22590 For example, 22591 22592 (eq_attr "type" "load,store") 22593 22594 is equivalent to 22595 22596 (ior (eq_attr "type" "load") (eq_attr "type" "store")) 22597 22598 If NAME specifies an attribute of `alternative', it refers to the 22599 value of the compiler variable `which_alternative' (*note Output 22600 Statement::) and the values must be small integers. For example, 22601 22602 (eq_attr "alternative" "2,3") 22603 22604 is equivalent to 22605 22606 (ior (eq (symbol_ref "which_alternative") (const_int 2)) 22607 (eq (symbol_ref "which_alternative") (const_int 3))) 22608 22609 Note that, for most attributes, an `eq_attr' test is simplified in 22610 cases where the value of the attribute being tested is known for 22611 all insns matching a particular pattern. This is by far the most 22612 common case. 22613 22614 `(attr_flag NAME)' 22615 The value of an `attr_flag' expression is true if the flag 22616 specified by NAME is true for the `insn' currently being scheduled. 22617 22618 NAME is a string specifying one of a fixed set of flags to test. 22619 Test the flags `forward' and `backward' to determine the direction 22620 of a conditional branch. Test the flags `very_likely', `likely', 22621 `very_unlikely', and `unlikely' to determine if a conditional 22622 branch is expected to be taken. 22623 22624 If the `very_likely' flag is true, then the `likely' flag is also 22625 true. Likewise for the `very_unlikely' and `unlikely' flags. 22626 22627 This example describes a conditional branch delay slot which can 22628 be nullified for forward branches that are taken (annul-true) or 22629 for backward branches which are not taken (annul-false). 22630 22631 (define_delay (eq_attr "type" "cbranch") 22632 [(eq_attr "in_branch_delay" "true") 22633 (and (eq_attr "in_branch_delay" "true") 22634 (attr_flag "forward")) 22635 (and (eq_attr "in_branch_delay" "true") 22636 (attr_flag "backward"))]) 22637 22638 The `forward' and `backward' flags are false if the current `insn' 22639 being scheduled is not a conditional branch. 22640 22641 The `very_likely' and `likely' flags are true if the `insn' being 22642 scheduled is not a conditional branch. The `very_unlikely' and 22643 `unlikely' flags are false if the `insn' being scheduled is not a 22644 conditional branch. 22645 22646 `attr_flag' is only used during delay slot scheduling and has no 22647 meaning to other passes of the compiler. 22648 22649 `(attr NAME)' 22650 The value of another attribute is returned. This is most useful 22651 for numeric attributes, as `eq_attr' and `attr_flag' produce more 22652 efficient code for non-numeric attributes. 22653 22654 22655 File: gccint.info, Node: Tagging Insns, Next: Attr Example, Prev: Expressions, Up: Insn Attributes 22656 22657 16.19.3 Assigning Attribute Values to Insns 22658 ------------------------------------------- 22659 22660 The value assigned to an attribute of an insn is primarily determined by 22661 which pattern is matched by that insn (or which `define_peephole' 22662 generated it). Every `define_insn' and `define_peephole' can have an 22663 optional last argument to specify the values of attributes for matching 22664 insns. The value of any attribute not specified in a particular insn 22665 is set to the default value for that attribute, as specified in its 22666 `define_attr'. Extensive use of default values for attributes permits 22667 the specification of the values for only one or two attributes in the 22668 definition of most insn patterns, as seen in the example in the next 22669 section. 22670 22671 The optional last argument of `define_insn' and `define_peephole' is a 22672 vector of expressions, each of which defines the value for a single 22673 attribute. The most general way of assigning an attribute's value is 22674 to use a `set' expression whose first operand is an `attr' expression 22675 giving the name of the attribute being set. The second operand of the 22676 `set' is an attribute expression (*note Expressions::) giving the value 22677 of the attribute. 22678 22679 When the attribute value depends on the `alternative' attribute (i.e., 22680 which is the applicable alternative in the constraint of the insn), the 22681 `set_attr_alternative' expression can be used. It allows the 22682 specification of a vector of attribute expressions, one for each 22683 alternative. 22684 22685 When the generality of arbitrary attribute expressions is not required, 22686 the simpler `set_attr' expression can be used, which allows specifying 22687 a string giving either a single attribute value or a list of attribute 22688 values, one for each alternative. 22689 22690 The form of each of the above specifications is shown below. In each 22691 case, NAME is a string specifying the attribute to be set. 22692 22693 `(set_attr NAME VALUE-STRING)' 22694 VALUE-STRING is either a string giving the desired attribute value, 22695 or a string containing a comma-separated list giving the values for 22696 succeeding alternatives. The number of elements must match the 22697 number of alternatives in the constraint of the insn pattern. 22698 22699 Note that it may be useful to specify `*' for some alternative, in 22700 which case the attribute will assume its default value for insns 22701 matching that alternative. 22702 22703 `(set_attr_alternative NAME [VALUE1 VALUE2 ...])' 22704 Depending on the alternative of the insn, the value will be one of 22705 the specified values. This is a shorthand for using a `cond' with 22706 tests on the `alternative' attribute. 22707 22708 `(set (attr NAME) VALUE)' 22709 The first operand of this `set' must be the special RTL expression 22710 `attr', whose sole operand is a string giving the name of the 22711 attribute being set. VALUE is the value of the attribute. 22712 22713 The following shows three different ways of representing the same 22714 attribute value specification: 22715 22716 (set_attr "type" "load,store,arith") 22717 22718 (set_attr_alternative "type" 22719 [(const_string "load") (const_string "store") 22720 (const_string "arith")]) 22721 22722 (set (attr "type") 22723 (cond [(eq_attr "alternative" "1") (const_string "load") 22724 (eq_attr "alternative" "2") (const_string "store")] 22725 (const_string "arith"))) 22726 22727 The `define_asm_attributes' expression provides a mechanism to specify 22728 the attributes assigned to insns produced from an `asm' statement. It 22729 has the form: 22730 22731 (define_asm_attributes [ATTR-SETS]) 22732 22733 where ATTR-SETS is specified the same as for both the `define_insn' and 22734 the `define_peephole' expressions. 22735 22736 These values will typically be the "worst case" attribute values. For 22737 example, they might indicate that the condition code will be clobbered. 22738 22739 A specification for a `length' attribute is handled specially. The 22740 way to compute the length of an `asm' insn is to multiply the length 22741 specified in the expression `define_asm_attributes' by the number of 22742 machine instructions specified in the `asm' statement, determined by 22743 counting the number of semicolons and newlines in the string. 22744 Therefore, the value of the `length' attribute specified in a 22745 `define_asm_attributes' should be the maximum possible length of a 22746 single machine instruction. 22747 22748 22749 File: gccint.info, Node: Attr Example, Next: Insn Lengths, Prev: Tagging Insns, Up: Insn Attributes 22750 22751 16.19.4 Example of Attribute Specifications 22752 ------------------------------------------- 22753 22754 The judicious use of defaulting is important in the efficient use of 22755 insn attributes. Typically, insns are divided into "types" and an 22756 attribute, customarily called `type', is used to represent this value. 22757 This attribute is normally used only to define the default value for 22758 other attributes. An example will clarify this usage. 22759 22760 Assume we have a RISC machine with a condition code and in which only 22761 full-word operations are performed in registers. Let us assume that we 22762 can divide all insns into loads, stores, (integer) arithmetic 22763 operations, floating point operations, and branches. 22764 22765 Here we will concern ourselves with determining the effect of an insn 22766 on the condition code and will limit ourselves to the following possible 22767 effects: The condition code can be set unpredictably (clobbered), not 22768 be changed, be set to agree with the results of the operation, or only 22769 changed if the item previously set into the condition code has been 22770 modified. 22771 22772 Here is part of a sample `md' file for such a machine: 22773 22774 (define_attr "type" "load,store,arith,fp,branch" (const_string "arith")) 22775 22776 (define_attr "cc" "clobber,unchanged,set,change0" 22777 (cond [(eq_attr "type" "load") 22778 (const_string "change0") 22779 (eq_attr "type" "store,branch") 22780 (const_string "unchanged") 22781 (eq_attr "type" "arith") 22782 (if_then_else (match_operand:SI 0 "" "") 22783 (const_string "set") 22784 (const_string "clobber"))] 22785 (const_string "clobber"))) 22786 22787 (define_insn "" 22788 [(set (match_operand:SI 0 "general_operand" "=r,r,m") 22789 (match_operand:SI 1 "general_operand" "r,m,r"))] 22790 "" 22791 "@ 22792 move %0,%1 22793 load %0,%1 22794 store %0,%1" 22795 [(set_attr "type" "arith,load,store")]) 22796 22797 Note that we assume in the above example that arithmetic operations 22798 performed on quantities smaller than a machine word clobber the 22799 condition code since they will set the condition code to a value 22800 corresponding to the full-word result. 22801 22802 22803 File: gccint.info, Node: Insn Lengths, Next: Constant Attributes, Prev: Attr Example, Up: Insn Attributes 22804 22805 16.19.5 Computing the Length of an Insn 22806 --------------------------------------- 22807 22808 For many machines, multiple types of branch instructions are provided, 22809 each for different length branch displacements. In most cases, the 22810 assembler will choose the correct instruction to use. However, when 22811 the assembler cannot do so, GCC can when a special attribute, the 22812 `length' attribute, is defined. This attribute must be defined to have 22813 numeric values by specifying a null string in its `define_attr'. 22814 22815 In the case of the `length' attribute, two additional forms of 22816 arithmetic terms are allowed in test expressions: 22817 22818 `(match_dup N)' 22819 This refers to the address of operand N of the current insn, which 22820 must be a `label_ref'. 22821 22822 `(pc)' 22823 This refers to the address of the _current_ insn. It might have 22824 been more consistent with other usage to make this the address of 22825 the _next_ insn but this would be confusing because the length of 22826 the current insn is to be computed. 22827 22828 For normal insns, the length will be determined by value of the 22829 `length' attribute. In the case of `addr_vec' and `addr_diff_vec' insn 22830 patterns, the length is computed as the number of vectors multiplied by 22831 the size of each vector. 22832 22833 Lengths are measured in addressable storage units (bytes). 22834 22835 The following macros can be used to refine the length computation: 22836 22837 `ADJUST_INSN_LENGTH (INSN, LENGTH)' 22838 If defined, modifies the length assigned to instruction INSN as a 22839 function of the context in which it is used. LENGTH is an lvalue 22840 that contains the initially computed length of the insn and should 22841 be updated with the correct length of the insn. 22842 22843 This macro will normally not be required. A case in which it is 22844 required is the ROMP. On this machine, the size of an `addr_vec' 22845 insn must be increased by two to compensate for the fact that 22846 alignment may be required. 22847 22848 The routine that returns `get_attr_length' (the value of the `length' 22849 attribute) can be used by the output routine to determine the form of 22850 the branch instruction to be written, as the example below illustrates. 22851 22852 As an example of the specification of variable-length branches, 22853 consider the IBM 360. If we adopt the convention that a register will 22854 be set to the starting address of a function, we can jump to labels 22855 within 4k of the start using a four-byte instruction. Otherwise, we 22856 need a six-byte sequence to load the address from memory and then 22857 branch to it. 22858 22859 On such a machine, a pattern for a branch instruction might be 22860 specified as follows: 22861 22862 (define_insn "jump" 22863 [(set (pc) 22864 (label_ref (match_operand 0 "" "")))] 22865 "" 22866 { 22867 return (get_attr_length (insn) == 4 22868 ? "b %l0" : "l r15,=a(%l0); br r15"); 22869 } 22870 [(set (attr "length") 22871 (if_then_else (lt (match_dup 0) (const_int 4096)) 22872 (const_int 4) 22873 (const_int 6)))]) 22874 22875 22876 File: gccint.info, Node: Constant Attributes, Next: Delay Slots, Prev: Insn Lengths, Up: Insn Attributes 22877 22878 16.19.6 Constant Attributes 22879 --------------------------- 22880 22881 A special form of `define_attr', where the expression for the default 22882 value is a `const' expression, indicates an attribute that is constant 22883 for a given run of the compiler. Constant attributes may be used to 22884 specify which variety of processor is used. For example, 22885 22886 (define_attr "cpu" "m88100,m88110,m88000" 22887 (const 22888 (cond [(symbol_ref "TARGET_88100") (const_string "m88100") 22889 (symbol_ref "TARGET_88110") (const_string "m88110")] 22890 (const_string "m88000")))) 22891 22892 (define_attr "memory" "fast,slow" 22893 (const 22894 (if_then_else (symbol_ref "TARGET_FAST_MEM") 22895 (const_string "fast") 22896 (const_string "slow")))) 22897 22898 The routine generated for constant attributes has no parameters as it 22899 does not depend on any particular insn. RTL expressions used to define 22900 the value of a constant attribute may use the `symbol_ref' form, but 22901 may not use either the `match_operand' form or `eq_attr' forms 22902 involving insn attributes. 22903 22904 22905 File: gccint.info, Node: Delay Slots, Next: Processor pipeline description, Prev: Constant Attributes, Up: Insn Attributes 22906 22907 16.19.7 Delay Slot Scheduling 22908 ----------------------------- 22909 22910 The insn attribute mechanism can be used to specify the requirements for 22911 delay slots, if any, on a target machine. An instruction is said to 22912 require a "delay slot" if some instructions that are physically after 22913 the instruction are executed as if they were located before it. 22914 Classic examples are branch and call instructions, which often execute 22915 the following instruction before the branch or call is performed. 22916 22917 On some machines, conditional branch instructions can optionally 22918 "annul" instructions in the delay slot. This means that the 22919 instruction will not be executed for certain branch outcomes. Both 22920 instructions that annul if the branch is true and instructions that 22921 annul if the branch is false are supported. 22922 22923 Delay slot scheduling differs from instruction scheduling in that 22924 determining whether an instruction needs a delay slot is dependent only 22925 on the type of instruction being generated, not on data flow between the 22926 instructions. See the next section for a discussion of data-dependent 22927 instruction scheduling. 22928 22929 The requirement of an insn needing one or more delay slots is indicated 22930 via the `define_delay' expression. It has the following form: 22931 22932 (define_delay TEST 22933 [DELAY-1 ANNUL-TRUE-1 ANNUL-FALSE-1 22934 DELAY-2 ANNUL-TRUE-2 ANNUL-FALSE-2 22935 ...]) 22936 22937 TEST is an attribute test that indicates whether this `define_delay' 22938 applies to a particular insn. If so, the number of required delay 22939 slots is determined by the length of the vector specified as the second 22940 argument. An insn placed in delay slot N must satisfy attribute test 22941 DELAY-N. ANNUL-TRUE-N is an attribute test that specifies which insns 22942 may be annulled if the branch is true. Similarly, ANNUL-FALSE-N 22943 specifies which insns in the delay slot may be annulled if the branch 22944 is false. If annulling is not supported for that delay slot, `(nil)' 22945 should be coded. 22946 22947 For example, in the common case where branch and call insns require a 22948 single delay slot, which may contain any insn other than a branch or 22949 call, the following would be placed in the `md' file: 22950 22951 (define_delay (eq_attr "type" "branch,call") 22952 [(eq_attr "type" "!branch,call") (nil) (nil)]) 22953 22954 Multiple `define_delay' expressions may be specified. In this case, 22955 each such expression specifies different delay slot requirements and 22956 there must be no insn for which tests in two `define_delay' expressions 22957 are both true. 22958 22959 For example, if we have a machine that requires one delay slot for 22960 branches but two for calls, no delay slot can contain a branch or call 22961 insn, and any valid insn in the delay slot for the branch can be 22962 annulled if the branch is true, we might represent this as follows: 22963 22964 (define_delay (eq_attr "type" "branch") 22965 [(eq_attr "type" "!branch,call") 22966 (eq_attr "type" "!branch,call") 22967 (nil)]) 22968 22969 (define_delay (eq_attr "type" "call") 22970 [(eq_attr "type" "!branch,call") (nil) (nil) 22971 (eq_attr "type" "!branch,call") (nil) (nil)]) 22972 22973 22974 File: gccint.info, Node: Processor pipeline description, Prev: Delay Slots, Up: Insn Attributes 22975 22976 16.19.8 Specifying processor pipeline description 22977 ------------------------------------------------- 22978 22979 To achieve better performance, most modern processors (super-pipelined, 22980 superscalar RISC, and VLIW processors) have many "functional units" on 22981 which several instructions can be executed simultaneously. An 22982 instruction starts execution if its issue conditions are satisfied. If 22983 not, the instruction is stalled until its conditions are satisfied. 22984 Such "interlock (pipeline) delay" causes interruption of the fetching 22985 of successor instructions (or demands nop instructions, e.g. for some 22986 MIPS processors). 22987 22988 There are two major kinds of interlock delays in modern processors. 22989 The first one is a data dependence delay determining "instruction 22990 latency time". The instruction execution is not started until all 22991 source data have been evaluated by prior instructions (there are more 22992 complex cases when the instruction execution starts even when the data 22993 are not available but will be ready in given time after the instruction 22994 execution start). Taking the data dependence delays into account is 22995 simple. The data dependence (true, output, and anti-dependence) delay 22996 between two instructions is given by a constant. In most cases this 22997 approach is adequate. The second kind of interlock delays is a 22998 reservation delay. The reservation delay means that two instructions 22999 under execution will be in need of shared processors resources, i.e. 23000 buses, internal registers, and/or functional units, which are reserved 23001 for some time. Taking this kind of delay into account is complex 23002 especially for modern RISC processors. 23003 23004 The task of exploiting more processor parallelism is solved by an 23005 instruction scheduler. For a better solution to this problem, the 23006 instruction scheduler has to have an adequate description of the 23007 processor parallelism (or "pipeline description"). GCC machine 23008 descriptions describe processor parallelism and functional unit 23009 reservations for groups of instructions with the aid of "regular 23010 expressions". 23011 23012 The GCC instruction scheduler uses a "pipeline hazard recognizer" to 23013 figure out the possibility of the instruction issue by the processor on 23014 a given simulated processor cycle. The pipeline hazard recognizer is 23015 automatically generated from the processor pipeline description. The 23016 pipeline hazard recognizer generated from the machine description is 23017 based on a deterministic finite state automaton (DFA): the instruction 23018 issue is possible if there is a transition from one automaton state to 23019 another one. This algorithm is very fast, and furthermore, its speed 23020 is not dependent on processor complexity(1). 23021 23022 The rest of this section describes the directives that constitute an 23023 automaton-based processor pipeline description. The order of these 23024 constructions within the machine description file is not important. 23025 23026 The following optional construction describes names of automata 23027 generated and used for the pipeline hazards recognition. Sometimes the 23028 generated finite state automaton used by the pipeline hazard recognizer 23029 is large. If we use more than one automaton and bind functional units 23030 to the automata, the total size of the automata is usually less than 23031 the size of the single automaton. If there is no one such 23032 construction, only one finite state automaton is generated. 23033 23034 (define_automaton AUTOMATA-NAMES) 23035 23036 AUTOMATA-NAMES is a string giving names of the automata. The names 23037 are separated by commas. All the automata should have unique names. 23038 The automaton name is used in the constructions `define_cpu_unit' and 23039 `define_query_cpu_unit'. 23040 23041 Each processor functional unit used in the description of instruction 23042 reservations should be described by the following construction. 23043 23044 (define_cpu_unit UNIT-NAMES [AUTOMATON-NAME]) 23045 23046 UNIT-NAMES is a string giving the names of the functional units 23047 separated by commas. Don't use name `nothing', it is reserved for 23048 other goals. 23049 23050 AUTOMATON-NAME is a string giving the name of the automaton with which 23051 the unit is bound. The automaton should be described in construction 23052 `define_automaton'. You should give "automaton-name", if there is a 23053 defined automaton. 23054 23055 The assignment of units to automata are constrained by the uses of the 23056 units in insn reservations. The most important constraint is: if a 23057 unit reservation is present on a particular cycle of an alternative for 23058 an insn reservation, then some unit from the same automaton must be 23059 present on the same cycle for the other alternatives of the insn 23060 reservation. The rest of the constraints are mentioned in the 23061 description of the subsequent constructions. 23062 23063 The following construction describes CPU functional units analogously 23064 to `define_cpu_unit'. The reservation of such units can be queried for 23065 an automaton state. The instruction scheduler never queries 23066 reservation of functional units for given automaton state. So as a 23067 rule, you don't need this construction. This construction could be 23068 used for future code generation goals (e.g. to generate VLIW insn 23069 templates). 23070 23071 (define_query_cpu_unit UNIT-NAMES [AUTOMATON-NAME]) 23072 23073 UNIT-NAMES is a string giving names of the functional units separated 23074 by commas. 23075 23076 AUTOMATON-NAME is a string giving the name of the automaton with which 23077 the unit is bound. 23078 23079 The following construction is the major one to describe pipeline 23080 characteristics of an instruction. 23081 23082 (define_insn_reservation INSN-NAME DEFAULT_LATENCY 23083 CONDITION REGEXP) 23084 23085 DEFAULT_LATENCY is a number giving latency time of the instruction. 23086 There is an important difference between the old description and the 23087 automaton based pipeline description. The latency time is used for all 23088 dependencies when we use the old description. In the automaton based 23089 pipeline description, the given latency time is only used for true 23090 dependencies. The cost of anti-dependencies is always zero and the 23091 cost of output dependencies is the difference between latency times of 23092 the producing and consuming insns (if the difference is negative, the 23093 cost is considered to be zero). You can always change the default 23094 costs for any description by using the target hook 23095 `TARGET_SCHED_ADJUST_COST' (*note Scheduling::). 23096 23097 INSN-NAME is a string giving the internal name of the insn. The 23098 internal names are used in constructions `define_bypass' and in the 23099 automaton description file generated for debugging. The internal name 23100 has nothing in common with the names in `define_insn'. It is a good 23101 practice to use insn classes described in the processor manual. 23102 23103 CONDITION defines what RTL insns are described by this construction. 23104 You should remember that you will be in trouble if CONDITION for two or 23105 more different `define_insn_reservation' constructions is TRUE for an 23106 insn. In this case what reservation will be used for the insn is not 23107 defined. Such cases are not checked during generation of the pipeline 23108 hazards recognizer because in general recognizing that two conditions 23109 may have the same value is quite difficult (especially if the conditions 23110 contain `symbol_ref'). It is also not checked during the pipeline 23111 hazard recognizer work because it would slow down the recognizer 23112 considerably. 23113 23114 REGEXP is a string describing the reservation of the cpu's functional 23115 units by the instruction. The reservations are described by a regular 23116 expression according to the following syntax: 23117 23118 regexp = regexp "," oneof 23119 | oneof 23120 23121 oneof = oneof "|" allof 23122 | allof 23123 23124 allof = allof "+" repeat 23125 | repeat 23126 23127 repeat = element "*" number 23128 | element 23129 23130 element = cpu_function_unit_name 23131 | reservation_name 23132 | result_name 23133 | "nothing" 23134 | "(" regexp ")" 23135 23136 * `,' is used for describing the start of the next cycle in the 23137 reservation. 23138 23139 * `|' is used for describing a reservation described by the first 23140 regular expression *or* a reservation described by the second 23141 regular expression *or* etc. 23142 23143 * `+' is used for describing a reservation described by the first 23144 regular expression *and* a reservation described by the second 23145 regular expression *and* etc. 23146 23147 * `*' is used for convenience and simply means a sequence in which 23148 the regular expression are repeated NUMBER times with cycle 23149 advancing (see `,'). 23150 23151 * `cpu_function_unit_name' denotes reservation of the named 23152 functional unit. 23153 23154 * `reservation_name' -- see description of construction 23155 `define_reservation'. 23156 23157 * `nothing' denotes no unit reservations. 23158 23159 Sometimes unit reservations for different insns contain common parts. 23160 In such case, you can simplify the pipeline description by describing 23161 the common part by the following construction 23162 23163 (define_reservation RESERVATION-NAME REGEXP) 23164 23165 RESERVATION-NAME is a string giving name of REGEXP. Functional unit 23166 names and reservation names are in the same name space. So the 23167 reservation names should be different from the functional unit names 23168 and can not be the reserved name `nothing'. 23169 23170 The following construction is used to describe exceptions in the 23171 latency time for given instruction pair. This is so called bypasses. 23172 23173 (define_bypass NUMBER OUT_INSN_NAMES IN_INSN_NAMES 23174 [GUARD]) 23175 23176 NUMBER defines when the result generated by the instructions given in 23177 string OUT_INSN_NAMES will be ready for the instructions given in 23178 string IN_INSN_NAMES. The instructions in the string are separated by 23179 commas. 23180 23181 GUARD is an optional string giving the name of a C function which 23182 defines an additional guard for the bypass. The function will get the 23183 two insns as parameters. If the function returns zero the bypass will 23184 be ignored for this case. The additional guard is necessary to 23185 recognize complicated bypasses, e.g. when the consumer is only an 23186 address of insn `store' (not a stored value). 23187 23188 If there are more one bypass with the same output and input insns, the 23189 chosen bypass is the first bypass with a guard in description whose 23190 guard function returns nonzero. If there is no such bypass, then 23191 bypass without the guard function is chosen. 23192 23193 The following five constructions are usually used to describe VLIW 23194 processors, or more precisely, to describe a placement of small 23195 instructions into VLIW instruction slots. They can be used for RISC 23196 processors, too. 23197 23198 (exclusion_set UNIT-NAMES UNIT-NAMES) 23199 (presence_set UNIT-NAMES PATTERNS) 23200 (final_presence_set UNIT-NAMES PATTERNS) 23201 (absence_set UNIT-NAMES PATTERNS) 23202 (final_absence_set UNIT-NAMES PATTERNS) 23203 23204 UNIT-NAMES is a string giving names of functional units separated by 23205 commas. 23206 23207 PATTERNS is a string giving patterns of functional units separated by 23208 comma. Currently pattern is one unit or units separated by 23209 white-spaces. 23210 23211 The first construction (`exclusion_set') means that each functional 23212 unit in the first string can not be reserved simultaneously with a unit 23213 whose name is in the second string and vice versa. For example, the 23214 construction is useful for describing processors (e.g. some SPARC 23215 processors) with a fully pipelined floating point functional unit which 23216 can execute simultaneously only single floating point insns or only 23217 double floating point insns. 23218 23219 The second construction (`presence_set') means that each functional 23220 unit in the first string can not be reserved unless at least one of 23221 pattern of units whose names are in the second string is reserved. 23222 This is an asymmetric relation. For example, it is useful for 23223 description that VLIW `slot1' is reserved after `slot0' reservation. 23224 We could describe it by the following construction 23225 23226 (presence_set "slot1" "slot0") 23227 23228 Or `slot1' is reserved only after `slot0' and unit `b0' reservation. 23229 In this case we could write 23230 23231 (presence_set "slot1" "slot0 b0") 23232 23233 The third construction (`final_presence_set') is analogous to 23234 `presence_set'. The difference between them is when checking is done. 23235 When an instruction is issued in given automaton state reflecting all 23236 current and planned unit reservations, the automaton state is changed. 23237 The first state is a source state, the second one is a result state. 23238 Checking for `presence_set' is done on the source state reservation, 23239 checking for `final_presence_set' is done on the result reservation. 23240 This construction is useful to describe a reservation which is actually 23241 two subsequent reservations. For example, if we use 23242 23243 (presence_set "slot1" "slot0") 23244 23245 the following insn will be never issued (because `slot1' requires 23246 `slot0' which is absent in the source state). 23247 23248 (define_reservation "insn_and_nop" "slot0 + slot1") 23249 23250 but it can be issued if we use analogous `final_presence_set'. 23251 23252 The forth construction (`absence_set') means that each functional unit 23253 in the first string can be reserved only if each pattern of units whose 23254 names are in the second string is not reserved. This is an asymmetric 23255 relation (actually `exclusion_set' is analogous to this one but it is 23256 symmetric). For example it might be useful in a VLIW description to 23257 say that `slot0' cannot be reserved after either `slot1' or `slot2' 23258 have been reserved. This can be described as: 23259 23260 (absence_set "slot0" "slot1, slot2") 23261 23262 Or `slot2' can not be reserved if `slot0' and unit `b0' are reserved 23263 or `slot1' and unit `b1' are reserved. In this case we could write 23264 23265 (absence_set "slot2" "slot0 b0, slot1 b1") 23266 23267 All functional units mentioned in a set should belong to the same 23268 automaton. 23269 23270 The last construction (`final_absence_set') is analogous to 23271 `absence_set' but checking is done on the result (state) reservation. 23272 See comments for `final_presence_set'. 23273 23274 You can control the generator of the pipeline hazard recognizer with 23275 the following construction. 23276 23277 (automata_option OPTIONS) 23278 23279 OPTIONS is a string giving options which affect the generated code. 23280 Currently there are the following options: 23281 23282 * "no-minimization" makes no minimization of the automaton. This is 23283 only worth to do when we are debugging the description and need to 23284 look more accurately at reservations of states. 23285 23286 * "time" means printing time statistics about the generation of 23287 automata. 23288 23289 * "stats" means printing statistics about the generated automata 23290 such as the number of DFA states, NDFA states and arcs. 23291 23292 * "v" means a generation of the file describing the result automata. 23293 The file has suffix `.dfa' and can be used for the description 23294 verification and debugging. 23295 23296 * "w" means a generation of warning instead of error for 23297 non-critical errors. 23298 23299 * "ndfa" makes nondeterministic finite state automata. This affects 23300 the treatment of operator `|' in the regular expressions. The 23301 usual treatment of the operator is to try the first alternative 23302 and, if the reservation is not possible, the second alternative. 23303 The nondeterministic treatment means trying all alternatives, some 23304 of them may be rejected by reservations in the subsequent insns. 23305 23306 * "progress" means output of a progress bar showing how many states 23307 were generated so far for automaton being processed. This is 23308 useful during debugging a DFA description. If you see too many 23309 generated states, you could interrupt the generator of the pipeline 23310 hazard recognizer and try to figure out a reason for generation of 23311 the huge automaton. 23312 23313 As an example, consider a superscalar RISC machine which can issue 23314 three insns (two integer insns and one floating point insn) on the 23315 cycle but can finish only two insns. To describe this, we define the 23316 following functional units. 23317 23318 (define_cpu_unit "i0_pipeline, i1_pipeline, f_pipeline") 23319 (define_cpu_unit "port0, port1") 23320 23321 All simple integer insns can be executed in any integer pipeline and 23322 their result is ready in two cycles. The simple integer insns are 23323 issued into the first pipeline unless it is reserved, otherwise they 23324 are issued into the second pipeline. Integer division and 23325 multiplication insns can be executed only in the second integer 23326 pipeline and their results are ready correspondingly in 8 and 4 cycles. 23327 The integer division is not pipelined, i.e. the subsequent integer 23328 division insn can not be issued until the current division insn 23329 finished. Floating point insns are fully pipelined and their results 23330 are ready in 3 cycles. Where the result of a floating point insn is 23331 used by an integer insn, an additional delay of one cycle is incurred. 23332 To describe all of this we could specify 23333 23334 (define_cpu_unit "div") 23335 23336 (define_insn_reservation "simple" 2 (eq_attr "type" "int") 23337 "(i0_pipeline | i1_pipeline), (port0 | port1)") 23338 23339 (define_insn_reservation "mult" 4 (eq_attr "type" "mult") 23340 "i1_pipeline, nothing*2, (port0 | port1)") 23341 23342 (define_insn_reservation "div" 8 (eq_attr "type" "div") 23343 "i1_pipeline, div*7, div + (port0 | port1)") 23344 23345 (define_insn_reservation "float" 3 (eq_attr "type" "float") 23346 "f_pipeline, nothing, (port0 | port1)) 23347 23348 (define_bypass 4 "float" "simple,mult,div") 23349 23350 To simplify the description we could describe the following reservation 23351 23352 (define_reservation "finish" "port0|port1") 23353 23354 and use it in all `define_insn_reservation' as in the following 23355 construction 23356 23357 (define_insn_reservation "simple" 2 (eq_attr "type" "int") 23358 "(i0_pipeline | i1_pipeline), finish") 23359 23360 ---------- Footnotes ---------- 23361 23362 (1) However, the size of the automaton depends on processor 23363 complexity. To limit this effect, machine descriptions can split 23364 orthogonal parts of the machine description among several automata: but 23365 then, since each of these must be stepped independently, this does 23366 cause a small decrease in the algorithm's performance. 23367 23368 23369 File: gccint.info, Node: Conditional Execution, Next: Constant Definitions, Prev: Insn Attributes, Up: Machine Desc 23370 23371 16.20 Conditional Execution 23372 =========================== 23373 23374 A number of architectures provide for some form of conditional 23375 execution, or predication. The hallmark of this feature is the ability 23376 to nullify most of the instructions in the instruction set. When the 23377 instruction set is large and not entirely symmetric, it can be quite 23378 tedious to describe these forms directly in the `.md' file. An 23379 alternative is the `define_cond_exec' template. 23380 23381 (define_cond_exec 23382 [PREDICATE-PATTERN] 23383 "CONDITION" 23384 "OUTPUT-TEMPLATE") 23385 23386 PREDICATE-PATTERN is the condition that must be true for the insn to 23387 be executed at runtime and should match a relational operator. One can 23388 use `match_operator' to match several relational operators at once. 23389 Any `match_operand' operands must have no more than one alternative. 23390 23391 CONDITION is a C expression that must be true for the generated 23392 pattern to match. 23393 23394 OUTPUT-TEMPLATE is a string similar to the `define_insn' output 23395 template (*note Output Template::), except that the `*' and `@' special 23396 cases do not apply. This is only useful if the assembly text for the 23397 predicate is a simple prefix to the main insn. In order to handle the 23398 general case, there is a global variable `current_insn_predicate' that 23399 will contain the entire predicate if the current insn is predicated, 23400 and will otherwise be `NULL'. 23401 23402 When `define_cond_exec' is used, an implicit reference to the 23403 `predicable' instruction attribute is made. *Note Insn Attributes::. 23404 This attribute must be boolean (i.e. have exactly two elements in its 23405 LIST-OF-VALUES). Further, it must not be used with complex 23406 expressions. That is, the default and all uses in the insns must be a 23407 simple constant, not dependent on the alternative or anything else. 23408 23409 For each `define_insn' for which the `predicable' attribute is true, a 23410 new `define_insn' pattern will be generated that matches a predicated 23411 version of the instruction. For example, 23412 23413 (define_insn "addsi" 23414 [(set (match_operand:SI 0 "register_operand" "r") 23415 (plus:SI (match_operand:SI 1 "register_operand" "r") 23416 (match_operand:SI 2 "register_operand" "r")))] 23417 "TEST1" 23418 "add %2,%1,%0") 23419 23420 (define_cond_exec 23421 [(ne (match_operand:CC 0 "register_operand" "c") 23422 (const_int 0))] 23423 "TEST2" 23424 "(%0)") 23425 23426 generates a new pattern 23427 23428 (define_insn "" 23429 [(cond_exec 23430 (ne (match_operand:CC 3 "register_operand" "c") (const_int 0)) 23431 (set (match_operand:SI 0 "register_operand" "r") 23432 (plus:SI (match_operand:SI 1 "register_operand" "r") 23433 (match_operand:SI 2 "register_operand" "r"))))] 23434 "(TEST2) && (TEST1)" 23435 "(%3) add %2,%1,%0") 23436 23437 23438 File: gccint.info, Node: Constant Definitions, Next: Iterators, Prev: Conditional Execution, Up: Machine Desc 23439 23440 16.21 Constant Definitions 23441 ========================== 23442 23443 Using literal constants inside instruction patterns reduces legibility 23444 and can be a maintenance problem. 23445 23446 To overcome this problem, you may use the `define_constants' 23447 expression. It contains a vector of name-value pairs. From that point 23448 on, wherever any of the names appears in the MD file, it is as if the 23449 corresponding value had been written instead. You may use 23450 `define_constants' multiple times; each appearance adds more constants 23451 to the table. It is an error to redefine a constant with a different 23452 value. 23453 23454 To come back to the a29k load multiple example, instead of 23455 23456 (define_insn "" 23457 [(match_parallel 0 "load_multiple_operation" 23458 [(set (match_operand:SI 1 "gpc_reg_operand" "=r") 23459 (match_operand:SI 2 "memory_operand" "m")) 23460 (use (reg:SI 179)) 23461 (clobber (reg:SI 179))])] 23462 "" 23463 "loadm 0,0,%1,%2") 23464 23465 You could write: 23466 23467 (define_constants [ 23468 (R_BP 177) 23469 (R_FC 178) 23470 (R_CR 179) 23471 (R_Q 180) 23472 ]) 23473 23474 (define_insn "" 23475 [(match_parallel 0 "load_multiple_operation" 23476 [(set (match_operand:SI 1 "gpc_reg_operand" "=r") 23477 (match_operand:SI 2 "memory_operand" "m")) 23478 (use (reg:SI R_CR)) 23479 (clobber (reg:SI R_CR))])] 23480 "" 23481 "loadm 0,0,%1,%2") 23482 23483 The constants that are defined with a define_constant are also output 23484 in the insn-codes.h header file as #defines. 23485 23486 23487 File: gccint.info, Node: Iterators, Prev: Constant Definitions, Up: Machine Desc 23488 23489 16.22 Iterators 23490 =============== 23491 23492 Ports often need to define similar patterns for more than one machine 23493 mode or for more than one rtx code. GCC provides some simple iterator 23494 facilities to make this process easier. 23495 23496 * Menu: 23497 23498 * Mode Iterators:: Generating variations of patterns for different modes. 23499 * Code Iterators:: Doing the same for codes. 23500 23501 23502 File: gccint.info, Node: Mode Iterators, Next: Code Iterators, Up: Iterators 23503 23504 16.22.1 Mode Iterators 23505 ---------------------- 23506 23507 Ports often need to define similar patterns for two or more different 23508 modes. For example: 23509 23510 * If a processor has hardware support for both single and double 23511 floating-point arithmetic, the `SFmode' patterns tend to be very 23512 similar to the `DFmode' ones. 23513 23514 * If a port uses `SImode' pointers in one configuration and `DImode' 23515 pointers in another, it will usually have very similar `SImode' 23516 and `DImode' patterns for manipulating pointers. 23517 23518 Mode iterators allow several patterns to be instantiated from one 23519 `.md' file template. They can be used with any type of rtx-based 23520 construct, such as a `define_insn', `define_split', or 23521 `define_peephole2'. 23522 23523 * Menu: 23524 23525 * Defining Mode Iterators:: Defining a new mode iterator. 23526 * Substitutions:: Combining mode iterators with substitutions 23527 * Examples:: Examples 23528 23529 23530 File: gccint.info, Node: Defining Mode Iterators, Next: Substitutions, Up: Mode Iterators 23531 23532 16.22.1.1 Defining Mode Iterators 23533 ................................. 23534 23535 The syntax for defining a mode iterator is: 23536 23537 (define_mode_iterator NAME [(MODE1 "COND1") ... (MODEN "CONDN")]) 23538 23539 This allows subsequent `.md' file constructs to use the mode suffix 23540 `:NAME'. Every construct that does so will be expanded N times, once 23541 with every use of `:NAME' replaced by `:MODE1', once with every use 23542 replaced by `:MODE2', and so on. In the expansion for a particular 23543 MODEI, every C condition will also require that CONDI be true. 23544 23545 For example: 23546 23547 (define_mode_iterator P [(SI "Pmode == SImode") (DI "Pmode == DImode")]) 23548 23549 defines a new mode suffix `:P'. Every construct that uses `:P' will 23550 be expanded twice, once with every `:P' replaced by `:SI' and once with 23551 every `:P' replaced by `:DI'. The `:SI' version will only apply if 23552 `Pmode == SImode' and the `:DI' version will only apply if `Pmode == 23553 DImode'. 23554 23555 As with other `.md' conditions, an empty string is treated as "always 23556 true". `(MODE "")' can also be abbreviated to `MODE'. For example: 23557 23558 (define_mode_iterator GPR [SI (DI "TARGET_64BIT")]) 23559 23560 means that the `:DI' expansion only applies if `TARGET_64BIT' but that 23561 the `:SI' expansion has no such constraint. 23562 23563 Iterators are applied in the order they are defined. This can be 23564 significant if two iterators are used in a construct that requires 23565 substitutions. *Note Substitutions::. 23566 23567 23568 File: gccint.info, Node: Substitutions, Next: Examples, Prev: Defining Mode Iterators, Up: Mode Iterators 23569 23570 16.22.1.2 Substitution in Mode Iterators 23571 ........................................ 23572 23573 If an `.md' file construct uses mode iterators, each version of the 23574 construct will often need slightly different strings or modes. For 23575 example: 23576 23577 * When a `define_expand' defines several `addM3' patterns (*note 23578 Standard Names::), each expander will need to use the appropriate 23579 mode name for M. 23580 23581 * When a `define_insn' defines several instruction patterns, each 23582 instruction will often use a different assembler mnemonic. 23583 23584 * When a `define_insn' requires operands with different modes, using 23585 an iterator for one of the operand modes usually requires a 23586 specific mode for the other operand(s). 23587 23588 GCC supports such variations through a system of "mode attributes". 23589 There are two standard attributes: `mode', which is the name of the 23590 mode in lower case, and `MODE', which is the same thing in upper case. 23591 You can define other attributes using: 23592 23593 (define_mode_attr NAME [(MODE1 "VALUE1") ... (MODEN "VALUEN")]) 23594 23595 where NAME is the name of the attribute and VALUEI is the value 23596 associated with MODEI. 23597 23598 When GCC replaces some :ITERATOR with :MODE, it will scan each string 23599 and mode in the pattern for sequences of the form `<ITERATOR:ATTR>', 23600 where ATTR is the name of a mode attribute. If the attribute is 23601 defined for MODE, the whole `<...>' sequence will be replaced by the 23602 appropriate attribute value. 23603 23604 For example, suppose an `.md' file has: 23605 23606 (define_mode_iterator P [(SI "Pmode == SImode") (DI "Pmode == DImode")]) 23607 (define_mode_attr load [(SI "lw") (DI "ld")]) 23608 23609 If one of the patterns that uses `:P' contains the string 23610 `"<P:load>\t%0,%1"', the `SI' version of that pattern will use 23611 `"lw\t%0,%1"' and the `DI' version will use `"ld\t%0,%1"'. 23612 23613 Here is an example of using an attribute for a mode: 23614 23615 (define_mode_iterator LONG [SI DI]) 23616 (define_mode_attr SHORT [(SI "HI") (DI "SI")]) 23617 (define_insn ... 23618 (sign_extend:LONG (match_operand:<LONG:SHORT> ...)) ...) 23619 23620 The `ITERATOR:' prefix may be omitted, in which case the substitution 23621 will be attempted for every iterator expansion. 23622 23623 23624 File: gccint.info, Node: Examples, Prev: Substitutions, Up: Mode Iterators 23625 23626 16.22.1.3 Mode Iterator Examples 23627 ................................ 23628 23629 Here is an example from the MIPS port. It defines the following modes 23630 and attributes (among others): 23631 23632 (define_mode_iterator GPR [SI (DI "TARGET_64BIT")]) 23633 (define_mode_attr d [(SI "") (DI "d")]) 23634 23635 and uses the following template to define both `subsi3' and `subdi3': 23636 23637 (define_insn "sub<mode>3" 23638 [(set (match_operand:GPR 0 "register_operand" "=d") 23639 (minus:GPR (match_operand:GPR 1 "register_operand" "d") 23640 (match_operand:GPR 2 "register_operand" "d")))] 23641 "" 23642 "<d>subu\t%0,%1,%2" 23643 [(set_attr "type" "arith") 23644 (set_attr "mode" "<MODE>")]) 23645 23646 This is exactly equivalent to: 23647 23648 (define_insn "subsi3" 23649 [(set (match_operand:SI 0 "register_operand" "=d") 23650 (minus:SI (match_operand:SI 1 "register_operand" "d") 23651 (match_operand:SI 2 "register_operand" "d")))] 23652 "" 23653 "subu\t%0,%1,%2" 23654 [(set_attr "type" "arith") 23655 (set_attr "mode" "SI")]) 23656 23657 (define_insn "subdi3" 23658 [(set (match_operand:DI 0 "register_operand" "=d") 23659 (minus:DI (match_operand:DI 1 "register_operand" "d") 23660 (match_operand:DI 2 "register_operand" "d")))] 23661 "" 23662 "dsubu\t%0,%1,%2" 23663 [(set_attr "type" "arith") 23664 (set_attr "mode" "DI")]) 23665 23666 23667 File: gccint.info, Node: Code Iterators, Prev: Mode Iterators, Up: Iterators 23668 23669 16.22.2 Code Iterators 23670 ---------------------- 23671 23672 Code iterators operate in a similar way to mode iterators. *Note Mode 23673 Iterators::. 23674 23675 The construct: 23676 23677 (define_code_iterator NAME [(CODE1 "COND1") ... (CODEN "CONDN")]) 23678 23679 defines a pseudo rtx code NAME that can be instantiated as CODEI if 23680 condition CONDI is true. Each CODEI must have the same rtx format. 23681 *Note RTL Classes::. 23682 23683 As with mode iterators, each pattern that uses NAME will be expanded N 23684 times, once with all uses of NAME replaced by CODE1, once with all uses 23685 replaced by CODE2, and so on. *Note Defining Mode Iterators::. 23686 23687 It is possible to define attributes for codes as well as for modes. 23688 There are two standard code attributes: `code', the name of the code in 23689 lower case, and `CODE', the name of the code in upper case. Other 23690 attributes are defined using: 23691 23692 (define_code_attr NAME [(CODE1 "VALUE1") ... (CODEN "VALUEN")]) 23693 23694 Here's an example of code iterators in action, taken from the MIPS 23695 port: 23696 23697 (define_code_iterator any_cond [unordered ordered unlt unge uneq ltgt unle ungt 23698 eq ne gt ge lt le gtu geu ltu leu]) 23699 23700 (define_expand "b<code>" 23701 [(set (pc) 23702 (if_then_else (any_cond:CC (cc0) 23703 (const_int 0)) 23704 (label_ref (match_operand 0 "")) 23705 (pc)))] 23706 "" 23707 { 23708 gen_conditional_branch (operands, <CODE>); 23709 DONE; 23710 }) 23711 23712 This is equivalent to: 23713 23714 (define_expand "bunordered" 23715 [(set (pc) 23716 (if_then_else (unordered:CC (cc0) 23717 (const_int 0)) 23718 (label_ref (match_operand 0 "")) 23719 (pc)))] 23720 "" 23721 { 23722 gen_conditional_branch (operands, UNORDERED); 23723 DONE; 23724 }) 23725 23726 (define_expand "bordered" 23727 [(set (pc) 23728 (if_then_else (ordered:CC (cc0) 23729 (const_int 0)) 23730 (label_ref (match_operand 0 "")) 23731 (pc)))] 23732 "" 23733 { 23734 gen_conditional_branch (operands, ORDERED); 23735 DONE; 23736 }) 23737 23738 ... 23739 23740 23741 File: gccint.info, Node: Target Macros, Next: Host Config, Prev: Machine Desc, Up: Top 23742 23743 17 Target Description Macros and Functions 23744 ****************************************** 23745 23746 In addition to the file `MACHINE.md', a machine description includes a 23747 C header file conventionally given the name `MACHINE.h' and a C source 23748 file named `MACHINE.c'. The header file defines numerous macros that 23749 convey the information about the target machine that does not fit into 23750 the scheme of the `.md' file. The file `tm.h' should be a link to 23751 `MACHINE.h'. The header file `config.h' includes `tm.h' and most 23752 compiler source files include `config.h'. The source file defines a 23753 variable `targetm', which is a structure containing pointers to 23754 functions and data relating to the target machine. `MACHINE.c' should 23755 also contain their definitions, if they are not defined elsewhere in 23756 GCC, and other functions called through the macros defined in the `.h' 23757 file. 23758 23759 * Menu: 23760 23761 * Target Structure:: The `targetm' variable. 23762 * Driver:: Controlling how the driver runs the compilation passes. 23763 * Run-time Target:: Defining `-m' options like `-m68000' and `-m68020'. 23764 * Per-Function Data:: Defining data structures for per-function information. 23765 * Storage Layout:: Defining sizes and alignments of data. 23766 * Type Layout:: Defining sizes and properties of basic user data types. 23767 * Registers:: Naming and describing the hardware registers. 23768 * Register Classes:: Defining the classes of hardware registers. 23769 * Old Constraints:: The old way to define machine-specific constraints. 23770 * Stack and Calling:: Defining which way the stack grows and by how much. 23771 * Varargs:: Defining the varargs macros. 23772 * Trampolines:: Code set up at run time to enter a nested function. 23773 * Library Calls:: Controlling how library routines are implicitly called. 23774 * Addressing Modes:: Defining addressing modes valid for memory operands. 23775 * Anchored Addresses:: Defining how `-fsection-anchors' should work. 23776 * Condition Code:: Defining how insns update the condition code. 23777 * Costs:: Defining relative costs of different operations. 23778 * Scheduling:: Adjusting the behavior of the instruction scheduler. 23779 * Sections:: Dividing storage into text, data, and other sections. 23780 * PIC:: Macros for position independent code. 23781 * Assembler Format:: Defining how to write insns and pseudo-ops to output. 23782 * Debugging Info:: Defining the format of debugging output. 23783 * Floating Point:: Handling floating point for cross-compilers. 23784 * Mode Switching:: Insertion of mode-switching instructions. 23785 * Target Attributes:: Defining target-specific uses of `__attribute__'. 23786 * Emulated TLS:: Emulated TLS support. 23787 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it. 23788 * PCH Target:: Validity checking for precompiled headers. 23789 * C++ ABI:: Controlling C++ ABI changes. 23790 * Misc:: Everything else. 23791 23792 23793 File: gccint.info, Node: Target Structure, Next: Driver, Up: Target Macros 23794 23795 17.1 The Global `targetm' Variable 23796 ================================== 23797 23798 -- Variable: struct gcc_target targetm 23799 The target `.c' file must define the global `targetm' variable 23800 which contains pointers to functions and data relating to the 23801 target machine. The variable is declared in `target.h'; 23802 `target-def.h' defines the macro `TARGET_INITIALIZER' which is 23803 used to initialize the variable, and macros for the default 23804 initializers for elements of the structure. The `.c' file should 23805 override those macros for which the default definition is 23806 inappropriate. For example: 23807 #include "target.h" 23808 #include "target-def.h" 23809 23810 /* Initialize the GCC target structure. */ 23811 23812 #undef TARGET_COMP_TYPE_ATTRIBUTES 23813 #define TARGET_COMP_TYPE_ATTRIBUTES MACHINE_comp_type_attributes 23814 23815 struct gcc_target targetm = TARGET_INITIALIZER; 23816 23817 Where a macro should be defined in the `.c' file in this manner to form 23818 part of the `targetm' structure, it is documented below as a "Target 23819 Hook" with a prototype. Many macros will change in future from being 23820 defined in the `.h' file to being part of the `targetm' structure. 23821 23822 23823 File: gccint.info, Node: Driver, Next: Run-time Target, Prev: Target Structure, Up: Target Macros 23824 23825 17.2 Controlling the Compilation Driver, `gcc' 23826 ============================================== 23827 23828 You can control the compilation driver. 23829 23830 -- Macro: SWITCH_TAKES_ARG (CHAR) 23831 A C expression which determines whether the option `-CHAR' takes 23832 arguments. The value should be the number of arguments that 23833 option takes-zero, for many options. 23834 23835 By default, this macro is defined as `DEFAULT_SWITCH_TAKES_ARG', 23836 which handles the standard options properly. You need not define 23837 `SWITCH_TAKES_ARG' unless you wish to add additional options which 23838 take arguments. Any redefinition should call 23839 `DEFAULT_SWITCH_TAKES_ARG' and then check for additional options. 23840 23841 -- Macro: WORD_SWITCH_TAKES_ARG (NAME) 23842 A C expression which determines whether the option `-NAME' takes 23843 arguments. The value should be the number of arguments that 23844 option takes-zero, for many options. This macro rather than 23845 `SWITCH_TAKES_ARG' is used for multi-character option names. 23846 23847 By default, this macro is defined as 23848 `DEFAULT_WORD_SWITCH_TAKES_ARG', which handles the standard options 23849 properly. You need not define `WORD_SWITCH_TAKES_ARG' unless you 23850 wish to add additional options which take arguments. Any 23851 redefinition should call `DEFAULT_WORD_SWITCH_TAKES_ARG' and then 23852 check for additional options. 23853 23854 -- Macro: SWITCH_CURTAILS_COMPILATION (CHAR) 23855 A C expression which determines whether the option `-CHAR' stops 23856 compilation before the generation of an executable. The value is 23857 boolean, nonzero if the option does stop an executable from being 23858 generated, zero otherwise. 23859 23860 By default, this macro is defined as 23861 `DEFAULT_SWITCH_CURTAILS_COMPILATION', which handles the standard 23862 options properly. You need not define 23863 `SWITCH_CURTAILS_COMPILATION' unless you wish to add additional 23864 options which affect the generation of an executable. Any 23865 redefinition should call `DEFAULT_SWITCH_CURTAILS_COMPILATION' and 23866 then check for additional options. 23867 23868 -- Macro: SWITCHES_NEED_SPACES 23869 A string-valued C expression which enumerates the options for which 23870 the linker needs a space between the option and its argument. 23871 23872 If this macro is not defined, the default value is `""'. 23873 23874 -- Macro: TARGET_OPTION_TRANSLATE_TABLE 23875 If defined, a list of pairs of strings, the first of which is a 23876 potential command line target to the `gcc' driver program, and the 23877 second of which is a space-separated (tabs and other whitespace 23878 are not supported) list of options with which to replace the first 23879 option. The target defining this list is responsible for assuring 23880 that the results are valid. Replacement options may not be the 23881 `--opt' style, they must be the `-opt' style. It is the intention 23882 of this macro to provide a mechanism for substitution that affects 23883 the multilibs chosen, such as one option that enables many 23884 options, some of which select multilibs. Example nonsensical 23885 definition, where `-malt-abi', `-EB', and `-mspoo' cause different 23886 multilibs to be chosen: 23887 23888 #define TARGET_OPTION_TRANSLATE_TABLE \ 23889 { "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" }, \ 23890 { "-compat", "-EB -malign=4 -mspoo" } 23891 23892 -- Macro: DRIVER_SELF_SPECS 23893 A list of specs for the driver itself. It should be a suitable 23894 initializer for an array of strings, with no surrounding braces. 23895 23896 The driver applies these specs to its own command line between 23897 loading default `specs' files (but not command-line specified 23898 ones) and choosing the multilib directory or running any 23899 subcommands. It applies them in the order given, so each spec can 23900 depend on the options added by earlier ones. It is also possible 23901 to remove options using `%<OPTION' in the usual way. 23902 23903 This macro can be useful when a port has several interdependent 23904 target options. It provides a way of standardizing the command 23905 line so that the other specs are easier to write. 23906 23907 Do not define this macro if it does not need to do anything. 23908 23909 -- Macro: OPTION_DEFAULT_SPECS 23910 A list of specs used to support configure-time default options 23911 (i.e. `--with' options) in the driver. It should be a suitable 23912 initializer for an array of structures, each containing two 23913 strings, without the outermost pair of surrounding braces. 23914 23915 The first item in the pair is the name of the default. This must 23916 match the code in `config.gcc' for the target. The second item is 23917 a spec to apply if a default with this name was specified. The 23918 string `%(VALUE)' in the spec will be replaced by the value of the 23919 default everywhere it occurs. 23920 23921 The driver will apply these specs to its own command line between 23922 loading default `specs' files and processing `DRIVER_SELF_SPECS', 23923 using the same mechanism as `DRIVER_SELF_SPECS'. 23924 23925 Do not define this macro if it does not need to do anything. 23926 23927 -- Macro: CPP_SPEC 23928 A C string constant that tells the GCC driver program options to 23929 pass to CPP. It can also specify how to translate options you 23930 give to GCC into options for GCC to pass to the CPP. 23931 23932 Do not define this macro if it does not need to do anything. 23933 23934 -- Macro: CPLUSPLUS_CPP_SPEC 23935 This macro is just like `CPP_SPEC', but is used for C++, rather 23936 than C. If you do not define this macro, then the value of 23937 `CPP_SPEC' (if any) will be used instead. 23938 23939 -- Macro: CC1_SPEC 23940 A C string constant that tells the GCC driver program options to 23941 pass to `cc1', `cc1plus', `f771', and the other language front 23942 ends. It can also specify how to translate options you give to 23943 GCC into options for GCC to pass to front ends. 23944 23945 Do not define this macro if it does not need to do anything. 23946 23947 -- Macro: CC1PLUS_SPEC 23948 A C string constant that tells the GCC driver program options to 23949 pass to `cc1plus'. It can also specify how to translate options 23950 you give to GCC into options for GCC to pass to the `cc1plus'. 23951 23952 Do not define this macro if it does not need to do anything. Note 23953 that everything defined in CC1_SPEC is already passed to `cc1plus' 23954 so there is no need to duplicate the contents of CC1_SPEC in 23955 CC1PLUS_SPEC. 23956 23957 -- Macro: ASM_SPEC 23958 A C string constant that tells the GCC driver program options to 23959 pass to the assembler. It can also specify how to translate 23960 options you give to GCC into options for GCC to pass to the 23961 assembler. See the file `sun3.h' for an example of this. 23962 23963 Do not define this macro if it does not need to do anything. 23964 23965 -- Macro: ASM_FINAL_SPEC 23966 A C string constant that tells the GCC driver program how to run 23967 any programs which cleanup after the normal assembler. Normally, 23968 this is not needed. See the file `mips.h' for an example of this. 23969 23970 Do not define this macro if it does not need to do anything. 23971 23972 -- Macro: AS_NEEDS_DASH_FOR_PIPED_INPUT 23973 Define this macro, with no value, if the driver should give the 23974 assembler an argument consisting of a single dash, `-', to 23975 instruct it to read from its standard input (which will be a pipe 23976 connected to the output of the compiler proper). This argument is 23977 given after any `-o' option specifying the name of the output file. 23978 23979 If you do not define this macro, the assembler is assumed to read 23980 its standard input if given no non-option arguments. If your 23981 assembler cannot read standard input at all, use a `%{pipe:%e}' 23982 construct; see `mips.h' for instance. 23983 23984 -- Macro: LINK_SPEC 23985 A C string constant that tells the GCC driver program options to 23986 pass to the linker. It can also specify how to translate options 23987 you give to GCC into options for GCC to pass to the linker. 23988 23989 Do not define this macro if it does not need to do anything. 23990 23991 -- Macro: LIB_SPEC 23992 Another C string constant used much like `LINK_SPEC'. The 23993 difference between the two is that `LIB_SPEC' is used at the end 23994 of the command given to the linker. 23995 23996 If this macro is not defined, a default is provided that loads the 23997 standard C library from the usual place. See `gcc.c'. 23998 23999 -- Macro: LIBGCC_SPEC 24000 Another C string constant that tells the GCC driver program how 24001 and when to place a reference to `libgcc.a' into the linker 24002 command line. This constant is placed both before and after the 24003 value of `LIB_SPEC'. 24004 24005 If this macro is not defined, the GCC driver provides a default 24006 that passes the string `-lgcc' to the linker. 24007 24008 -- Macro: REAL_LIBGCC_SPEC 24009 By default, if `ENABLE_SHARED_LIBGCC' is defined, the 24010 `LIBGCC_SPEC' is not directly used by the driver program but is 24011 instead modified to refer to different versions of `libgcc.a' 24012 depending on the values of the command line flags `-static', 24013 `-shared', `-static-libgcc', and `-shared-libgcc'. On targets 24014 where these modifications are inappropriate, define 24015 `REAL_LIBGCC_SPEC' instead. `REAL_LIBGCC_SPEC' tells the driver 24016 how to place a reference to `libgcc' on the link command line, 24017 but, unlike `LIBGCC_SPEC', it is used unmodified. 24018 24019 -- Macro: USE_LD_AS_NEEDED 24020 A macro that controls the modifications to `LIBGCC_SPEC' mentioned 24021 in `REAL_LIBGCC_SPEC'. If nonzero, a spec will be generated that 24022 uses -as-needed and the shared libgcc in place of the static 24023 exception handler library, when linking without any of `-static', 24024 `-static-libgcc', or `-shared-libgcc'. 24025 24026 -- Macro: LINK_EH_SPEC 24027 If defined, this C string constant is added to `LINK_SPEC'. When 24028 `USE_LD_AS_NEEDED' is zero or undefined, it also affects the 24029 modifications to `LIBGCC_SPEC' mentioned in `REAL_LIBGCC_SPEC'. 24030 24031 -- Macro: STARTFILE_SPEC 24032 Another C string constant used much like `LINK_SPEC'. The 24033 difference between the two is that `STARTFILE_SPEC' is used at the 24034 very beginning of the command given to the linker. 24035 24036 If this macro is not defined, a default is provided that loads the 24037 standard C startup file from the usual place. See `gcc.c'. 24038 24039 -- Macro: ENDFILE_SPEC 24040 Another C string constant used much like `LINK_SPEC'. The 24041 difference between the two is that `ENDFILE_SPEC' is used at the 24042 very end of the command given to the linker. 24043 24044 Do not define this macro if it does not need to do anything. 24045 24046 -- Macro: THREAD_MODEL_SPEC 24047 GCC `-v' will print the thread model GCC was configured to use. 24048 However, this doesn't work on platforms that are multilibbed on 24049 thread models, such as AIX 4.3. On such platforms, define 24050 `THREAD_MODEL_SPEC' such that it evaluates to a string without 24051 blanks that names one of the recognized thread models. `%*', the 24052 default value of this macro, will expand to the value of 24053 `thread_file' set in `config.gcc'. 24054 24055 -- Macro: SYSROOT_SUFFIX_SPEC 24056 Define this macro to add a suffix to the target sysroot when GCC is 24057 configured with a sysroot. This will cause GCC to search for 24058 usr/lib, et al, within sysroot+suffix. 24059 24060 -- Macro: SYSROOT_HEADERS_SUFFIX_SPEC 24061 Define this macro to add a headers_suffix to the target sysroot 24062 when GCC is configured with a sysroot. This will cause GCC to 24063 pass the updated sysroot+headers_suffix to CPP, causing it to 24064 search for usr/include, et al, within sysroot+headers_suffix. 24065 24066 -- Macro: EXTRA_SPECS 24067 Define this macro to provide additional specifications to put in 24068 the `specs' file that can be used in various specifications like 24069 `CC1_SPEC'. 24070 24071 The definition should be an initializer for an array of structures, 24072 containing a string constant, that defines the specification name, 24073 and a string constant that provides the specification. 24074 24075 Do not define this macro if it does not need to do anything. 24076 24077 `EXTRA_SPECS' is useful when an architecture contains several 24078 related targets, which have various `..._SPECS' which are similar 24079 to each other, and the maintainer would like one central place to 24080 keep these definitions. 24081 24082 For example, the PowerPC System V.4 targets use `EXTRA_SPECS' to 24083 define either `_CALL_SYSV' when the System V calling sequence is 24084 used or `_CALL_AIX' when the older AIX-based calling sequence is 24085 used. 24086 24087 The `config/rs6000/rs6000.h' target file defines: 24088 24089 #define EXTRA_SPECS \ 24090 { "cpp_sysv_default", CPP_SYSV_DEFAULT }, 24091 24092 #define CPP_SYS_DEFAULT "" 24093 24094 The `config/rs6000/sysv.h' target file defines: 24095 #undef CPP_SPEC 24096 #define CPP_SPEC \ 24097 "%{posix: -D_POSIX_SOURCE } \ 24098 %{mcall-sysv: -D_CALL_SYSV } \ 24099 %{!mcall-sysv: %(cpp_sysv_default) } \ 24100 %{msoft-float: -D_SOFT_FLOAT} %{mcpu=403: -D_SOFT_FLOAT}" 24101 24102 #undef CPP_SYSV_DEFAULT 24103 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV" 24104 24105 while the `config/rs6000/eabiaix.h' target file defines 24106 `CPP_SYSV_DEFAULT' as: 24107 24108 #undef CPP_SYSV_DEFAULT 24109 #define CPP_SYSV_DEFAULT "-D_CALL_AIX" 24110 24111 -- Macro: LINK_LIBGCC_SPECIAL_1 24112 Define this macro if the driver program should find the library 24113 `libgcc.a'. If you do not define this macro, the driver program 24114 will pass the argument `-lgcc' to tell the linker to do the search. 24115 24116 -- Macro: LINK_GCC_C_SEQUENCE_SPEC 24117 The sequence in which libgcc and libc are specified to the linker. 24118 By default this is `%G %L %G'. 24119 24120 -- Macro: LINK_COMMAND_SPEC 24121 A C string constant giving the complete command line need to 24122 execute the linker. When you do this, you will need to update 24123 your port each time a change is made to the link command line 24124 within `gcc.c'. Therefore, define this macro only if you need to 24125 completely redefine the command line for invoking the linker and 24126 there is no other way to accomplish the effect you need. 24127 Overriding this macro may be avoidable by overriding 24128 `LINK_GCC_C_SEQUENCE_SPEC' instead. 24129 24130 -- Macro: LINK_ELIMINATE_DUPLICATE_LDIRECTORIES 24131 A nonzero value causes `collect2' to remove duplicate 24132 `-LDIRECTORY' search directories from linking commands. Do not 24133 give it a nonzero value if removing duplicate search directories 24134 changes the linker's semantics. 24135 24136 -- Macro: MULTILIB_DEFAULTS 24137 Define this macro as a C expression for the initializer of an 24138 array of string to tell the driver program which options are 24139 defaults for this target and thus do not need to be handled 24140 specially when using `MULTILIB_OPTIONS'. 24141 24142 Do not define this macro if `MULTILIB_OPTIONS' is not defined in 24143 the target makefile fragment or if none of the options listed in 24144 `MULTILIB_OPTIONS' are set by default. *Note Target Fragment::. 24145 24146 -- Macro: RELATIVE_PREFIX_NOT_LINKDIR 24147 Define this macro to tell `gcc' that it should only translate a 24148 `-B' prefix into a `-L' linker option if the prefix indicates an 24149 absolute file name. 24150 24151 -- Macro: MD_EXEC_PREFIX 24152 If defined, this macro is an additional prefix to try after 24153 `STANDARD_EXEC_PREFIX'. `MD_EXEC_PREFIX' is not searched when the 24154 `-b' option is used, or the compiler is built as a cross compiler. 24155 If you define `MD_EXEC_PREFIX', then be sure to add it to the list 24156 of directories used to find the assembler in `configure.in'. 24157 24158 -- Macro: STANDARD_STARTFILE_PREFIX 24159 Define this macro as a C string constant if you wish to override 24160 the standard choice of `libdir' as the default prefix to try when 24161 searching for startup files such as `crt0.o'. 24162 `STANDARD_STARTFILE_PREFIX' is not searched when the compiler is 24163 built as a cross compiler. 24164 24165 -- Macro: STANDARD_STARTFILE_PREFIX_1 24166 Define this macro as a C string constant if you wish to override 24167 the standard choice of `/lib' as a prefix to try after the default 24168 prefix when searching for startup files such as `crt0.o'. 24169 `STANDARD_STARTFILE_PREFIX_1' is not searched when the compiler is 24170 built as a cross compiler. 24171 24172 -- Macro: STANDARD_STARTFILE_PREFIX_2 24173 Define this macro as a C string constant if you wish to override 24174 the standard choice of `/lib' as yet another prefix to try after 24175 the default prefix when searching for startup files such as 24176 `crt0.o'. `STANDARD_STARTFILE_PREFIX_2' is not searched when the 24177 compiler is built as a cross compiler. 24178 24179 -- Macro: MD_STARTFILE_PREFIX 24180 If defined, this macro supplies an additional prefix to try after 24181 the standard prefixes. `MD_EXEC_PREFIX' is not searched when the 24182 `-b' option is used, or when the compiler is built as a cross 24183 compiler. 24184 24185 -- Macro: MD_STARTFILE_PREFIX_1 24186 If defined, this macro supplies yet another prefix to try after the 24187 standard prefixes. It is not searched when the `-b' option is 24188 used, or when the compiler is built as a cross compiler. 24189 24190 -- Macro: INIT_ENVIRONMENT 24191 Define this macro as a C string constant if you wish to set 24192 environment variables for programs called by the driver, such as 24193 the assembler and loader. The driver passes the value of this 24194 macro to `putenv' to initialize the necessary environment 24195 variables. 24196 24197 -- Macro: LOCAL_INCLUDE_DIR 24198 Define this macro as a C string constant if you wish to override 24199 the standard choice of `/usr/local/include' as the default prefix 24200 to try when searching for local header files. `LOCAL_INCLUDE_DIR' 24201 comes before `SYSTEM_INCLUDE_DIR' in the search order. 24202 24203 Cross compilers do not search either `/usr/local/include' or its 24204 replacement. 24205 24206 -- Macro: MODIFY_TARGET_NAME 24207 Define this macro if you wish to define command-line switches that 24208 modify the default target name. 24209 24210 For each switch, you can include a string to be appended to the 24211 first part of the configuration name or a string to be deleted 24212 from the configuration name, if present. The definition should be 24213 an initializer for an array of structures. Each array element 24214 should have three elements: the switch name (a string constant, 24215 including the initial dash), one of the enumeration codes `ADD' or 24216 `DELETE' to indicate whether the string should be inserted or 24217 deleted, and the string to be inserted or deleted (a string 24218 constant). 24219 24220 For example, on a machine where `64' at the end of the 24221 configuration name denotes a 64-bit target and you want the `-32' 24222 and `-64' switches to select between 32- and 64-bit targets, you 24223 would code 24224 24225 #define MODIFY_TARGET_NAME \ 24226 { { "-32", DELETE, "64"}, \ 24227 {"-64", ADD, "64"}} 24228 24229 -- Macro: SYSTEM_INCLUDE_DIR 24230 Define this macro as a C string constant if you wish to specify a 24231 system-specific directory to search for header files before the 24232 standard directory. `SYSTEM_INCLUDE_DIR' comes before 24233 `STANDARD_INCLUDE_DIR' in the search order. 24234 24235 Cross compilers do not use this macro and do not search the 24236 directory specified. 24237 24238 -- Macro: STANDARD_INCLUDE_DIR 24239 Define this macro as a C string constant if you wish to override 24240 the standard choice of `/usr/include' as the default prefix to try 24241 when searching for header files. 24242 24243 Cross compilers ignore this macro and do not search either 24244 `/usr/include' or its replacement. 24245 24246 -- Macro: STANDARD_INCLUDE_COMPONENT 24247 The "component" corresponding to `STANDARD_INCLUDE_DIR'. See 24248 `INCLUDE_DEFAULTS', below, for the description of components. If 24249 you do not define this macro, no component is used. 24250 24251 -- Macro: INCLUDE_DEFAULTS 24252 Define this macro if you wish to override the entire default 24253 search path for include files. For a native compiler, the default 24254 search path usually consists of `GCC_INCLUDE_DIR', 24255 `LOCAL_INCLUDE_DIR', `SYSTEM_INCLUDE_DIR', 24256 `GPLUSPLUS_INCLUDE_DIR', and `STANDARD_INCLUDE_DIR'. In addition, 24257 `GPLUSPLUS_INCLUDE_DIR' and `GCC_INCLUDE_DIR' are defined 24258 automatically by `Makefile', and specify private search areas for 24259 GCC. The directory `GPLUSPLUS_INCLUDE_DIR' is used only for C++ 24260 programs. 24261 24262 The definition should be an initializer for an array of structures. 24263 Each array element should have four elements: the directory name (a 24264 string constant), the component name (also a string constant), a 24265 flag for C++-only directories, and a flag showing that the 24266 includes in the directory don't need to be wrapped in `extern `C'' 24267 when compiling C++. Mark the end of the array with a null element. 24268 24269 The component name denotes what GNU package the include file is 24270 part of, if any, in all uppercase letters. For example, it might 24271 be `GCC' or `BINUTILS'. If the package is part of a 24272 vendor-supplied operating system, code the component name as `0'. 24273 24274 For example, here is the definition used for VAX/VMS: 24275 24276 #define INCLUDE_DEFAULTS \ 24277 { \ 24278 { "GNU_GXX_INCLUDE:", "G++", 1, 1}, \ 24279 { "GNU_CC_INCLUDE:", "GCC", 0, 0}, \ 24280 { "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0}, \ 24281 { ".", 0, 0, 0}, \ 24282 { 0, 0, 0, 0} \ 24283 } 24284 24285 Here is the order of prefixes tried for exec files: 24286 24287 1. Any prefixes specified by the user with `-B'. 24288 24289 2. The environment variable `GCC_EXEC_PREFIX' or, if `GCC_EXEC_PREFIX' 24290 is not set and the compiler has not been installed in the 24291 configure-time PREFIX, the location in which the compiler has 24292 actually been installed. 24293 24294 3. The directories specified by the environment variable 24295 `COMPILER_PATH'. 24296 24297 4. The macro `STANDARD_EXEC_PREFIX', if the compiler has been 24298 installed in the configured-time PREFIX. 24299 24300 5. The location `/usr/libexec/gcc/', but only if this is a native 24301 compiler. 24302 24303 6. The location `/usr/lib/gcc/', but only if this is a native 24304 compiler. 24305 24306 7. The macro `MD_EXEC_PREFIX', if defined, but only if this is a 24307 native compiler. 24308 24309 Here is the order of prefixes tried for startfiles: 24310 24311 1. Any prefixes specified by the user with `-B'. 24312 24313 2. The environment variable `GCC_EXEC_PREFIX' or its automatically 24314 determined value based on the installed toolchain location. 24315 24316 3. The directories specified by the environment variable 24317 `LIBRARY_PATH' (or port-specific name; native only, cross 24318 compilers do not use this). 24319 24320 4. The macro `STANDARD_EXEC_PREFIX', but only if the toolchain is 24321 installed in the configured PREFIX or this is a native compiler. 24322 24323 5. The location `/usr/lib/gcc/', but only if this is a native 24324 compiler. 24325 24326 6. The macro `MD_EXEC_PREFIX', if defined, but only if this is a 24327 native compiler. 24328 24329 7. The macro `MD_STARTFILE_PREFIX', if defined, but only if this is a 24330 native compiler, or we have a target system root. 24331 24332 8. The macro `MD_STARTFILE_PREFIX_1', if defined, but only if this is 24333 a native compiler, or we have a target system root. 24334 24335 9. The macro `STANDARD_STARTFILE_PREFIX', with any sysroot 24336 modifications. If this path is relative it will be prefixed by 24337 `GCC_EXEC_PREFIX' and the machine suffix or `STANDARD_EXEC_PREFIX' 24338 and the machine suffix. 24339 24340 10. The macro `STANDARD_STARTFILE_PREFIX_1', but only if this is a 24341 native compiler, or we have a target system root. The default for 24342 this macro is `/lib/'. 24343 24344 11. The macro `STANDARD_STARTFILE_PREFIX_2', but only if this is a 24345 native compiler, or we have a target system root. The default for 24346 this macro is `/usr/lib/'. 24347 24348 24349 File: gccint.info, Node: Run-time Target, Next: Per-Function Data, Prev: Driver, Up: Target Macros 24350 24351 17.3 Run-time Target Specification 24352 ================================== 24353 24354 Here are run-time target specifications. 24355 24356 -- Macro: TARGET_CPU_CPP_BUILTINS () 24357 This function-like macro expands to a block of code that defines 24358 built-in preprocessor macros and assertions for the target CPU, 24359 using the functions `builtin_define', `builtin_define_std' and 24360 `builtin_assert'. When the front end calls this macro it provides 24361 a trailing semicolon, and since it has finished command line 24362 option processing your code can use those results freely. 24363 24364 `builtin_assert' takes a string in the form you pass to the 24365 command-line option `-A', such as `cpu=mips', and creates the 24366 assertion. `builtin_define' takes a string in the form accepted 24367 by option `-D' and unconditionally defines the macro. 24368 24369 `builtin_define_std' takes a string representing the name of an 24370 object-like macro. If it doesn't lie in the user's namespace, 24371 `builtin_define_std' defines it unconditionally. Otherwise, it 24372 defines a version with two leading underscores, and another version 24373 with two leading and trailing underscores, and defines the original 24374 only if an ISO standard was not requested on the command line. For 24375 example, passing `unix' defines `__unix', `__unix__' and possibly 24376 `unix'; passing `_mips' defines `__mips', `__mips__' and possibly 24377 `_mips', and passing `_ABI64' defines only `_ABI64'. 24378 24379 You can also test for the C dialect being compiled. The variable 24380 `c_language' is set to one of `clk_c', `clk_cplusplus' or 24381 `clk_objective_c'. Note that if we are preprocessing assembler, 24382 this variable will be `clk_c' but the function-like macro 24383 `preprocessing_asm_p()' will return true, so you might want to 24384 check for that first. If you need to check for strict ANSI, the 24385 variable `flag_iso' can be used. The function-like macro 24386 `preprocessing_trad_p()' can be used to check for traditional 24387 preprocessing. 24388 24389 -- Macro: TARGET_OS_CPP_BUILTINS () 24390 Similarly to `TARGET_CPU_CPP_BUILTINS' but this macro is optional 24391 and is used for the target operating system instead. 24392 24393 -- Macro: TARGET_OBJFMT_CPP_BUILTINS () 24394 Similarly to `TARGET_CPU_CPP_BUILTINS' but this macro is optional 24395 and is used for the target object format. `elfos.h' uses this 24396 macro to define `__ELF__', so you probably do not need to define 24397 it yourself. 24398 24399 -- Variable: extern int target_flags 24400 This variable is declared in `options.h', which is included before 24401 any target-specific headers. 24402 24403 -- Variable: Target Hook int TARGET_DEFAULT_TARGET_FLAGS 24404 This variable specifies the initial value of `target_flags'. Its 24405 default setting is 0. 24406 24407 -- Target Hook: bool TARGET_HANDLE_OPTION (size_t CODE, const char 24408 *ARG, int VALUE) 24409 This hook is called whenever the user specifies one of the 24410 target-specific options described by the `.opt' definition files 24411 (*note Options::). It has the opportunity to do some 24412 option-specific processing and should return true if the option is 24413 valid. The default definition does nothing but return true. 24414 24415 CODE specifies the `OPT_NAME' enumeration value associated with 24416 the selected option; NAME is just a rendering of the option name 24417 in which non-alphanumeric characters are replaced by underscores. 24418 ARG specifies the string argument and is null if no argument was 24419 given. If the option is flagged as a `UInteger' (*note Option 24420 properties::), VALUE is the numeric value of the argument. 24421 Otherwise VALUE is 1 if the positive form of the option was used 24422 and 0 if the "no-" form was. 24423 24424 -- Target Hook: bool TARGET_HANDLE_C_OPTION (size_t CODE, const char 24425 *ARG, int VALUE) 24426 This target hook is called whenever the user specifies one of the 24427 target-specific C language family options described by the `.opt' 24428 definition files(*note Options::). It has the opportunity to do 24429 some option-specific processing and should return true if the 24430 option is valid. The default definition does nothing but return 24431 false. 24432 24433 In general, you should use `TARGET_HANDLE_OPTION' to handle 24434 options. However, if processing an option requires routines that 24435 are only available in the C (and related language) front ends, 24436 then you should use `TARGET_HANDLE_C_OPTION' instead. 24437 24438 -- Macro: TARGET_VERSION 24439 This macro is a C statement to print on `stderr' a string 24440 describing the particular machine description choice. Every 24441 machine description should define `TARGET_VERSION'. For example: 24442 24443 #ifdef MOTOROLA 24444 #define TARGET_VERSION \ 24445 fprintf (stderr, " (68k, Motorola syntax)"); 24446 #else 24447 #define TARGET_VERSION \ 24448 fprintf (stderr, " (68k, MIT syntax)"); 24449 #endif 24450 24451 -- Macro: OVERRIDE_OPTIONS 24452 Sometimes certain combinations of command options do not make 24453 sense on a particular target machine. You can define a macro 24454 `OVERRIDE_OPTIONS' to take account of this. This macro, if 24455 defined, is executed once just after all the command options have 24456 been parsed. 24457 24458 Don't use this macro to turn on various extra optimizations for 24459 `-O'. That is what `OPTIMIZATION_OPTIONS' is for. 24460 24461 -- Macro: C_COMMON_OVERRIDE_OPTIONS 24462 This is similar to `OVERRIDE_OPTIONS' but is only used in the C 24463 language frontends (C, Objective-C, C++, Objective-C++) and so can 24464 be used to alter option flag variables which only exist in those 24465 frontends. 24466 24467 -- Macro: OPTIMIZATION_OPTIONS (LEVEL, SIZE) 24468 Some machines may desire to change what optimizations are 24469 performed for various optimization levels. This macro, if 24470 defined, is executed once just after the optimization level is 24471 determined and before the remainder of the command options have 24472 been parsed. Values set in this macro are used as the default 24473 values for the other command line options. 24474 24475 LEVEL is the optimization level specified; 2 if `-O2' is 24476 specified, 1 if `-O' is specified, and 0 if neither is specified. 24477 24478 SIZE is nonzero if `-Os' is specified and zero otherwise. 24479 24480 This macro is run once at program startup and when the optimization 24481 options are changed via `#pragma GCC optimize' or by using the 24482 `optimize' attribute. 24483 24484 *Do not examine `write_symbols' in this macro!* The debugging 24485 options are not supposed to alter the generated code. 24486 24487 -- Target Hook: bool TARGET_HELP (void) 24488 This hook is called in response to the user invoking 24489 `--target-help' on the command line. It gives the target a chance 24490 to display extra information on the target specific command line 24491 options found in its `.opt' file. 24492 24493 -- Macro: CAN_DEBUG_WITHOUT_FP 24494 Define this macro if debugging can be performed even without a 24495 frame pointer. If this macro is defined, GCC will turn on the 24496 `-fomit-frame-pointer' option whenever `-O' is specified. 24497 24498 24499 File: gccint.info, Node: Per-Function Data, Next: Storage Layout, Prev: Run-time Target, Up: Target Macros 24500 24501 17.4 Defining data structures for per-function information. 24502 =========================================================== 24503 24504 If the target needs to store information on a per-function basis, GCC 24505 provides a macro and a couple of variables to allow this. Note, just 24506 using statics to store the information is a bad idea, since GCC supports 24507 nested functions, so you can be halfway through encoding one function 24508 when another one comes along. 24509 24510 GCC defines a data structure called `struct function' which contains 24511 all of the data specific to an individual function. This structure 24512 contains a field called `machine' whose type is `struct 24513 machine_function *', which can be used by targets to point to their own 24514 specific data. 24515 24516 If a target needs per-function specific data it should define the type 24517 `struct machine_function' and also the macro `INIT_EXPANDERS'. This 24518 macro should be used to initialize the function pointer 24519 `init_machine_status'. This pointer is explained below. 24520 24521 One typical use of per-function, target specific data is to create an 24522 RTX to hold the register containing the function's return address. This 24523 RTX can then be used to implement the `__builtin_return_address' 24524 function, for level 0. 24525 24526 Note--earlier implementations of GCC used a single data area to hold 24527 all of the per-function information. Thus when processing of a nested 24528 function began the old per-function data had to be pushed onto a stack, 24529 and when the processing was finished, it had to be popped off the 24530 stack. GCC used to provide function pointers called 24531 `save_machine_status' and `restore_machine_status' to handle the saving 24532 and restoring of the target specific information. Since the single 24533 data area approach is no longer used, these pointers are no longer 24534 supported. 24535 24536 -- Macro: INIT_EXPANDERS 24537 Macro called to initialize any target specific information. This 24538 macro is called once per function, before generation of any RTL 24539 has begun. The intention of this macro is to allow the 24540 initialization of the function pointer `init_machine_status'. 24541 24542 -- Variable: void (*)(struct function *) init_machine_status 24543 If this function pointer is non-`NULL' it will be called once per 24544 function, before function compilation starts, in order to allow the 24545 target to perform any target specific initialization of the 24546 `struct function' structure. It is intended that this would be 24547 used to initialize the `machine' of that structure. 24548 24549 `struct machine_function' structures are expected to be freed by 24550 GC. Generally, any memory that they reference must be allocated 24551 by using `ggc_alloc', including the structure itself. 24552 24553 24554 File: gccint.info, Node: Storage Layout, Next: Type Layout, Prev: Per-Function Data, Up: Target Macros 24555 24556 17.5 Storage Layout 24557 =================== 24558 24559 Note that the definitions of the macros in this table which are sizes or 24560 alignments measured in bits do not need to be constant. They can be C 24561 expressions that refer to static variables, such as the `target_flags'. 24562 *Note Run-time Target::. 24563 24564 -- Macro: BITS_BIG_ENDIAN 24565 Define this macro to have the value 1 if the most significant bit 24566 in a byte has the lowest number; otherwise define it to have the 24567 value zero. This means that bit-field instructions count from the 24568 most significant bit. If the machine has no bit-field 24569 instructions, then this must still be defined, but it doesn't 24570 matter which value it is defined to. This macro need not be a 24571 constant. 24572 24573 This macro does not affect the way structure fields are packed into 24574 bytes or words; that is controlled by `BYTES_BIG_ENDIAN'. 24575 24576 -- Macro: BYTES_BIG_ENDIAN 24577 Define this macro to have the value 1 if the most significant byte 24578 in a word has the lowest number. This macro need not be a 24579 constant. 24580 24581 -- Macro: WORDS_BIG_ENDIAN 24582 Define this macro to have the value 1 if, in a multiword object, 24583 the most significant word has the lowest number. This applies to 24584 both memory locations and registers; GCC fundamentally assumes 24585 that the order of words in memory is the same as the order in 24586 registers. This macro need not be a constant. 24587 24588 -- Macro: LIBGCC2_WORDS_BIG_ENDIAN 24589 Define this macro if `WORDS_BIG_ENDIAN' is not constant. This 24590 must be a constant value with the same meaning as 24591 `WORDS_BIG_ENDIAN', which will be used only when compiling 24592 `libgcc2.c'. Typically the value will be set based on 24593 preprocessor defines. 24594 24595 -- Macro: FLOAT_WORDS_BIG_ENDIAN 24596 Define this macro to have the value 1 if `DFmode', `XFmode' or 24597 `TFmode' floating point numbers are stored in memory with the word 24598 containing the sign bit at the lowest address; otherwise define it 24599 to have the value 0. This macro need not be a constant. 24600 24601 You need not define this macro if the ordering is the same as for 24602 multi-word integers. 24603 24604 -- Macro: BITS_PER_UNIT 24605 Define this macro to be the number of bits in an addressable 24606 storage unit (byte). If you do not define this macro the default 24607 is 8. 24608 24609 -- Macro: BITS_PER_WORD 24610 Number of bits in a word. If you do not define this macro, the 24611 default is `BITS_PER_UNIT * UNITS_PER_WORD'. 24612 24613 -- Macro: MAX_BITS_PER_WORD 24614 Maximum number of bits in a word. If this is undefined, the 24615 default is `BITS_PER_WORD'. Otherwise, it is the constant value 24616 that is the largest value that `BITS_PER_WORD' can have at 24617 run-time. 24618 24619 -- Macro: UNITS_PER_WORD 24620 Number of storage units in a word; normally the size of a 24621 general-purpose register, a power of two from 1 or 8. 24622 24623 -- Macro: MIN_UNITS_PER_WORD 24624 Minimum number of units in a word. If this is undefined, the 24625 default is `UNITS_PER_WORD'. Otherwise, it is the constant value 24626 that is the smallest value that `UNITS_PER_WORD' can have at 24627 run-time. 24628 24629 -- Macro: UNITS_PER_SIMD_WORD (MODE) 24630 Number of units in the vectors that the vectorizer can produce for 24631 scalar mode MODE. The default is equal to `UNITS_PER_WORD', 24632 because the vectorizer can do some transformations even in absence 24633 of specialized SIMD hardware. 24634 24635 -- Macro: POINTER_SIZE 24636 Width of a pointer, in bits. You must specify a value no wider 24637 than the width of `Pmode'. If it is not equal to the width of 24638 `Pmode', you must define `POINTERS_EXTEND_UNSIGNED'. If you do 24639 not specify a value the default is `BITS_PER_WORD'. 24640 24641 -- Macro: POINTERS_EXTEND_UNSIGNED 24642 A C expression that determines how pointers should be extended from 24643 `ptr_mode' to either `Pmode' or `word_mode'. It is greater than 24644 zero if pointers should be zero-extended, zero if they should be 24645 sign-extended, and negative if some other sort of conversion is 24646 needed. In the last case, the extension is done by the target's 24647 `ptr_extend' instruction. 24648 24649 You need not define this macro if the `ptr_mode', `Pmode' and 24650 `word_mode' are all the same width. 24651 24652 -- Macro: PROMOTE_MODE (M, UNSIGNEDP, TYPE) 24653 A macro to update M and UNSIGNEDP when an object whose type is 24654 TYPE and which has the specified mode and signedness is to be 24655 stored in a register. This macro is only called when TYPE is a 24656 scalar type. 24657 24658 On most RISC machines, which only have operations that operate on 24659 a full register, define this macro to set M to `word_mode' if M is 24660 an integer mode narrower than `BITS_PER_WORD'. In most cases, 24661 only integer modes should be widened because wider-precision 24662 floating-point operations are usually more expensive than their 24663 narrower counterparts. 24664 24665 For most machines, the macro definition does not change UNSIGNEDP. 24666 However, some machines, have instructions that preferentially 24667 handle either signed or unsigned quantities of certain modes. For 24668 example, on the DEC Alpha, 32-bit loads from memory and 32-bit add 24669 instructions sign-extend the result to 64 bits. On such machines, 24670 set UNSIGNEDP according to which kind of extension is more 24671 efficient. 24672 24673 Do not define this macro if it would never modify M. 24674 24675 -- Macro: PROMOTE_FUNCTION_MODE 24676 Like `PROMOTE_MODE', but is applied to outgoing function arguments 24677 or function return values, as specified by 24678 `TARGET_PROMOTE_FUNCTION_ARGS' and 24679 `TARGET_PROMOTE_FUNCTION_RETURN', respectively. 24680 24681 The default is `PROMOTE_MODE'. 24682 24683 -- Target Hook: bool TARGET_PROMOTE_FUNCTION_ARGS (tree FNTYPE) 24684 This target hook should return `true' if the promotion described by 24685 `PROMOTE_FUNCTION_MODE' should be done for outgoing function 24686 arguments. 24687 24688 -- Target Hook: bool TARGET_PROMOTE_FUNCTION_RETURN (tree FNTYPE) 24689 This target hook should return `true' if the promotion described by 24690 `PROMOTE_FUNCTION_MODE' should be done for the return value of 24691 functions. 24692 24693 If this target hook returns `true', `TARGET_FUNCTION_VALUE' must 24694 perform the same promotions done by `PROMOTE_FUNCTION_MODE'. 24695 24696 -- Macro: PARM_BOUNDARY 24697 Normal alignment required for function parameters on the stack, in 24698 bits. All stack parameters receive at least this much alignment 24699 regardless of data type. On most machines, this is the same as the 24700 size of an integer. 24701 24702 -- Macro: STACK_BOUNDARY 24703 Define this macro to the minimum alignment enforced by hardware 24704 for the stack pointer on this machine. The definition is a C 24705 expression for the desired alignment (measured in bits). This 24706 value is used as a default if `PREFERRED_STACK_BOUNDARY' is not 24707 defined. On most machines, this should be the same as 24708 `PARM_BOUNDARY'. 24709 24710 -- Macro: PREFERRED_STACK_BOUNDARY 24711 Define this macro if you wish to preserve a certain alignment for 24712 the stack pointer, greater than what the hardware enforces. The 24713 definition is a C expression for the desired alignment (measured 24714 in bits). This macro must evaluate to a value equal to or larger 24715 than `STACK_BOUNDARY'. 24716 24717 -- Macro: INCOMING_STACK_BOUNDARY 24718 Define this macro if the incoming stack boundary may be different 24719 from `PREFERRED_STACK_BOUNDARY'. This macro must evaluate to a 24720 value equal to or larger than `STACK_BOUNDARY'. 24721 24722 -- Macro: FUNCTION_BOUNDARY 24723 Alignment required for a function entry point, in bits. 24724 24725 -- Macro: BIGGEST_ALIGNMENT 24726 Biggest alignment that any data type can require on this machine, 24727 in bits. Note that this is not the biggest alignment that is 24728 supported, just the biggest alignment that, when violated, may 24729 cause a fault. 24730 24731 -- Macro: MALLOC_ABI_ALIGNMENT 24732 Alignment, in bits, a C conformant malloc implementation has to 24733 provide. If not defined, the default value is `BITS_PER_WORD'. 24734 24735 -- Macro: ATTRIBUTE_ALIGNED_VALUE 24736 Alignment used by the `__attribute__ ((aligned))' construct. If 24737 not defined, the default value is `BIGGEST_ALIGNMENT'. 24738 24739 -- Macro: MINIMUM_ATOMIC_ALIGNMENT 24740 If defined, the smallest alignment, in bits, that can be given to 24741 an object that can be referenced in one operation, without 24742 disturbing any nearby object. Normally, this is `BITS_PER_UNIT', 24743 but may be larger on machines that don't have byte or half-word 24744 store operations. 24745 24746 -- Macro: BIGGEST_FIELD_ALIGNMENT 24747 Biggest alignment that any structure or union field can require on 24748 this machine, in bits. If defined, this overrides 24749 `BIGGEST_ALIGNMENT' for structure and union fields only, unless 24750 the field alignment has been set by the `__attribute__ ((aligned 24751 (N)))' construct. 24752 24753 -- Macro: ADJUST_FIELD_ALIGN (FIELD, COMPUTED) 24754 An expression for the alignment of a structure field FIELD if the 24755 alignment computed in the usual way (including applying of 24756 `BIGGEST_ALIGNMENT' and `BIGGEST_FIELD_ALIGNMENT' to the 24757 alignment) is COMPUTED. It overrides alignment only if the field 24758 alignment has not been set by the `__attribute__ ((aligned (N)))' 24759 construct. 24760 24761 -- Macro: MAX_STACK_ALIGNMENT 24762 Biggest stack alignment guaranteed by the backend. Use this macro 24763 to specify the maximum alignment of a variable on stack. 24764 24765 If not defined, the default value is `STACK_BOUNDARY'. 24766 24767 24768 -- Macro: MAX_OFILE_ALIGNMENT 24769 Biggest alignment supported by the object file format of this 24770 machine. Use this macro to limit the alignment which can be 24771 specified using the `__attribute__ ((aligned (N)))' construct. If 24772 not defined, the default value is `BIGGEST_ALIGNMENT'. 24773 24774 On systems that use ELF, the default (in `config/elfos.h') is the 24775 largest supported 32-bit ELF section alignment representable on a 24776 32-bit host e.g. `(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)'. On 24777 32-bit ELF the largest supported section alignment in bits is 24778 `(0x80000000 * 8)', but this is not representable on 32-bit hosts. 24779 24780 -- Macro: DATA_ALIGNMENT (TYPE, BASIC-ALIGN) 24781 If defined, a C expression to compute the alignment for a variable 24782 in the static store. TYPE is the data type, and BASIC-ALIGN is 24783 the alignment that the object would ordinarily have. The value of 24784 this macro is used instead of that alignment to align the object. 24785 24786 If this macro is not defined, then BASIC-ALIGN is used. 24787 24788 One use of this macro is to increase alignment of medium-size data 24789 to make it all fit in fewer cache lines. Another is to cause 24790 character arrays to be word-aligned so that `strcpy' calls that 24791 copy constants to character arrays can be done inline. 24792 24793 -- Macro: CONSTANT_ALIGNMENT (CONSTANT, BASIC-ALIGN) 24794 If defined, a C expression to compute the alignment given to a 24795 constant that is being placed in memory. CONSTANT is the constant 24796 and BASIC-ALIGN is the alignment that the object would ordinarily 24797 have. The value of this macro is used instead of that alignment to 24798 align the object. 24799 24800 If this macro is not defined, then BASIC-ALIGN is used. 24801 24802 The typical use of this macro is to increase alignment for string 24803 constants to be word aligned so that `strcpy' calls that copy 24804 constants can be done inline. 24805 24806 -- Macro: LOCAL_ALIGNMENT (TYPE, BASIC-ALIGN) 24807 If defined, a C expression to compute the alignment for a variable 24808 in the local store. TYPE is the data type, and BASIC-ALIGN is the 24809 alignment that the object would ordinarily have. The value of this 24810 macro is used instead of that alignment to align the object. 24811 24812 If this macro is not defined, then BASIC-ALIGN is used. 24813 24814 One use of this macro is to increase alignment of medium-size data 24815 to make it all fit in fewer cache lines. 24816 24817 -- Macro: STACK_SLOT_ALIGNMENT (TYPE, MODE, BASIC-ALIGN) 24818 If defined, a C expression to compute the alignment for stack slot. 24819 TYPE is the data type, MODE is the widest mode available, and 24820 BASIC-ALIGN is the alignment that the slot would ordinarily have. 24821 The value of this macro is used instead of that alignment to align 24822 the slot. 24823 24824 If this macro is not defined, then BASIC-ALIGN is used when TYPE 24825 is `NULL'. Otherwise, `LOCAL_ALIGNMENT' will be used. 24826 24827 This macro is to set alignment of stack slot to the maximum 24828 alignment of all possible modes which the slot may have. 24829 24830 -- Macro: LOCAL_DECL_ALIGNMENT (DECL) 24831 If defined, a C expression to compute the alignment for a local 24832 variable DECL. 24833 24834 If this macro is not defined, then `LOCAL_ALIGNMENT (TREE_TYPE 24835 (DECL), DECL_ALIGN (DECL))' is used. 24836 24837 One use of this macro is to increase alignment of medium-size data 24838 to make it all fit in fewer cache lines. 24839 24840 -- Macro: MINIMUM_ALIGNMENT (EXP, MODE, ALIGN) 24841 If defined, a C expression to compute the minimum required 24842 alignment for dynamic stack realignment purposes for EXP (a type 24843 or decl), MODE, assuming normal alignment ALIGN. 24844 24845 If this macro is not defined, then ALIGN will be used. 24846 24847 -- Macro: EMPTY_FIELD_BOUNDARY 24848 Alignment in bits to be given to a structure bit-field that 24849 follows an empty field such as `int : 0;'. 24850 24851 If `PCC_BITFIELD_TYPE_MATTERS' is true, it overrides this macro. 24852 24853 -- Macro: STRUCTURE_SIZE_BOUNDARY 24854 Number of bits which any structure or union's size must be a 24855 multiple of. Each structure or union's size is rounded up to a 24856 multiple of this. 24857 24858 If you do not define this macro, the default is the same as 24859 `BITS_PER_UNIT'. 24860 24861 -- Macro: STRICT_ALIGNMENT 24862 Define this macro to be the value 1 if instructions will fail to 24863 work if given data not on the nominal alignment. If instructions 24864 will merely go slower in that case, define this macro as 0. 24865 24866 -- Macro: PCC_BITFIELD_TYPE_MATTERS 24867 Define this if you wish to imitate the way many other C compilers 24868 handle alignment of bit-fields and the structures that contain 24869 them. 24870 24871 The behavior is that the type written for a named bit-field (`int', 24872 `short', or other integer type) imposes an alignment for the entire 24873 structure, as if the structure really did contain an ordinary 24874 field of that type. In addition, the bit-field is placed within 24875 the structure so that it would fit within such a field, not 24876 crossing a boundary for it. 24877 24878 Thus, on most machines, a named bit-field whose type is written as 24879 `int' would not cross a four-byte boundary, and would force 24880 four-byte alignment for the whole structure. (The alignment used 24881 may not be four bytes; it is controlled by the other alignment 24882 parameters.) 24883 24884 An unnamed bit-field will not affect the alignment of the 24885 containing structure. 24886 24887 If the macro is defined, its definition should be a C expression; 24888 a nonzero value for the expression enables this behavior. 24889 24890 Note that if this macro is not defined, or its value is zero, some 24891 bit-fields may cross more than one alignment boundary. The 24892 compiler can support such references if there are `insv', `extv', 24893 and `extzv' insns that can directly reference memory. 24894 24895 The other known way of making bit-fields work is to define 24896 `STRUCTURE_SIZE_BOUNDARY' as large as `BIGGEST_ALIGNMENT'. Then 24897 every structure can be accessed with fullwords. 24898 24899 Unless the machine has bit-field instructions or you define 24900 `STRUCTURE_SIZE_BOUNDARY' that way, you must define 24901 `PCC_BITFIELD_TYPE_MATTERS' to have a nonzero value. 24902 24903 If your aim is to make GCC use the same conventions for laying out 24904 bit-fields as are used by another compiler, here is how to 24905 investigate what the other compiler does. Compile and run this 24906 program: 24907 24908 struct foo1 24909 { 24910 char x; 24911 char :0; 24912 char y; 24913 }; 24914 24915 struct foo2 24916 { 24917 char x; 24918 int :0; 24919 char y; 24920 }; 24921 24922 main () 24923 { 24924 printf ("Size of foo1 is %d\n", 24925 sizeof (struct foo1)); 24926 printf ("Size of foo2 is %d\n", 24927 sizeof (struct foo2)); 24928 exit (0); 24929 } 24930 24931 If this prints 2 and 5, then the compiler's behavior is what you 24932 would get from `PCC_BITFIELD_TYPE_MATTERS'. 24933 24934 -- Macro: BITFIELD_NBYTES_LIMITED 24935 Like `PCC_BITFIELD_TYPE_MATTERS' except that its effect is limited 24936 to aligning a bit-field within the structure. 24937 24938 -- Target Hook: bool TARGET_ALIGN_ANON_BITFIELD (void) 24939 When `PCC_BITFIELD_TYPE_MATTERS' is true this hook will determine 24940 whether unnamed bitfields affect the alignment of the containing 24941 structure. The hook should return true if the structure should 24942 inherit the alignment requirements of an unnamed bitfield's type. 24943 24944 -- Target Hook: bool TARGET_NARROW_VOLATILE_BITFIELD (void) 24945 This target hook should return `true' if accesses to volatile 24946 bitfields should use the narrowest mode possible. It should 24947 return `false' if these accesses should use the bitfield container 24948 type. 24949 24950 The default is `!TARGET_STRICT_ALIGN'. 24951 24952 -- Macro: MEMBER_TYPE_FORCES_BLK (FIELD, MODE) 24953 Return 1 if a structure or array containing FIELD should be 24954 accessed using `BLKMODE'. 24955 24956 If FIELD is the only field in the structure, MODE is its mode, 24957 otherwise MODE is VOIDmode. MODE is provided in the case where 24958 structures of one field would require the structure's mode to 24959 retain the field's mode. 24960 24961 Normally, this is not needed. 24962 24963 -- Macro: ROUND_TYPE_ALIGN (TYPE, COMPUTED, SPECIFIED) 24964 Define this macro as an expression for the alignment of a type 24965 (given by TYPE as a tree node) if the alignment computed in the 24966 usual way is COMPUTED and the alignment explicitly specified was 24967 SPECIFIED. 24968 24969 The default is to use SPECIFIED if it is larger; otherwise, use 24970 the smaller of COMPUTED and `BIGGEST_ALIGNMENT' 24971 24972 -- Macro: MAX_FIXED_MODE_SIZE 24973 An integer expression for the size in bits of the largest integer 24974 machine mode that should actually be used. All integer machine 24975 modes of this size or smaller can be used for structures and 24976 unions with the appropriate sizes. If this macro is undefined, 24977 `GET_MODE_BITSIZE (DImode)' is assumed. 24978 24979 -- Macro: STACK_SAVEAREA_MODE (SAVE_LEVEL) 24980 If defined, an expression of type `enum machine_mode' that 24981 specifies the mode of the save area operand of a 24982 `save_stack_LEVEL' named pattern (*note Standard Names::). 24983 SAVE_LEVEL is one of `SAVE_BLOCK', `SAVE_FUNCTION', or 24984 `SAVE_NONLOCAL' and selects which of the three named patterns is 24985 having its mode specified. 24986 24987 You need not define this macro if it always returns `Pmode'. You 24988 would most commonly define this macro if the `save_stack_LEVEL' 24989 patterns need to support both a 32- and a 64-bit mode. 24990 24991 -- Macro: STACK_SIZE_MODE 24992 If defined, an expression of type `enum machine_mode' that 24993 specifies the mode of the size increment operand of an 24994 `allocate_stack' named pattern (*note Standard Names::). 24995 24996 You need not define this macro if it always returns `word_mode'. 24997 You would most commonly define this macro if the `allocate_stack' 24998 pattern needs to support both a 32- and a 64-bit mode. 24999 25000 -- Target Hook: enum machine_mode TARGET_LIBGCC_CMP_RETURN_MODE () 25001 This target hook should return the mode to be used for the return 25002 value of compare instructions expanded to libgcc calls. If not 25003 defined `word_mode' is returned which is the right choice for a 25004 majority of targets. 25005 25006 -- Target Hook: enum machine_mode TARGET_LIBGCC_SHIFT_COUNT_MODE () 25007 This target hook should return the mode to be used for the shift 25008 count operand of shift instructions expanded to libgcc calls. If 25009 not defined `word_mode' is returned which is the right choice for 25010 a majority of targets. 25011 25012 -- Macro: ROUND_TOWARDS_ZERO 25013 If defined, this macro should be true if the prevailing rounding 25014 mode is towards zero. 25015 25016 Defining this macro only affects the way `libgcc.a' emulates 25017 floating-point arithmetic. 25018 25019 Not defining this macro is equivalent to returning zero. 25020 25021 -- Macro: LARGEST_EXPONENT_IS_NORMAL (SIZE) 25022 This macro should return true if floats with SIZE bits do not have 25023 a NaN or infinity representation, but use the largest exponent for 25024 normal numbers instead. 25025 25026 Defining this macro only affects the way `libgcc.a' emulates 25027 floating-point arithmetic. 25028 25029 The default definition of this macro returns false for all sizes. 25030 25031 -- Target Hook: bool TARGET_VECTOR_OPAQUE_P (tree TYPE) 25032 This target hook should return `true' a vector is opaque. That 25033 is, if no cast is needed when copying a vector value of type TYPE 25034 into another vector lvalue of the same size. Vector opaque types 25035 cannot be initialized. The default is that there are no such 25036 types. 25037 25038 -- Target Hook: bool TARGET_MS_BITFIELD_LAYOUT_P (tree RECORD_TYPE) 25039 This target hook returns `true' if bit-fields in the given 25040 RECORD_TYPE are to be laid out following the rules of Microsoft 25041 Visual C/C++, namely: (i) a bit-field won't share the same storage 25042 unit with the previous bit-field if their underlying types have 25043 different sizes, and the bit-field will be aligned to the highest 25044 alignment of the underlying types of itself and of the previous 25045 bit-field; (ii) a zero-sized bit-field will affect the alignment of 25046 the whole enclosing structure, even if it is unnamed; except that 25047 (iii) a zero-sized bit-field will be disregarded unless it follows 25048 another bit-field of nonzero size. If this hook returns `true', 25049 other macros that control bit-field layout are ignored. 25050 25051 When a bit-field is inserted into a packed record, the whole size 25052 of the underlying type is used by one or more same-size adjacent 25053 bit-fields (that is, if its long:3, 32 bits is used in the record, 25054 and any additional adjacent long bit-fields are packed into the 25055 same chunk of 32 bits. However, if the size changes, a new field 25056 of that size is allocated). In an unpacked record, this is the 25057 same as using alignment, but not equivalent when packing. 25058 25059 If both MS bit-fields and `__attribute__((packed))' are used, the 25060 latter will take precedence. If `__attribute__((packed))' is used 25061 on a single field when MS bit-fields are in use, it will take 25062 precedence for that field, but the alignment of the rest of the 25063 structure may affect its placement. 25064 25065 -- Target Hook: bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void) 25066 Returns true if the target supports decimal floating point. 25067 25068 -- Target Hook: bool TARGET_FIXED_POINT_SUPPORTED_P (void) 25069 Returns true if the target supports fixed-point arithmetic. 25070 25071 -- Target Hook: void TARGET_EXPAND_TO_RTL_HOOK (void) 25072 This hook is called just before expansion into rtl, allowing the 25073 target to perform additional initializations or analysis before 25074 the expansion. For example, the rs6000 port uses it to allocate a 25075 scratch stack slot for use in copying SDmode values between memory 25076 and floating point registers whenever the function being expanded 25077 has any SDmode usage. 25078 25079 -- Target Hook: void TARGET_INSTANTIATE_DECLS (void) 25080 This hook allows the backend to perform additional instantiations 25081 on rtl that are not actually in any insns yet, but will be later. 25082 25083 -- Target Hook: const char * TARGET_MANGLE_TYPE (tree TYPE) 25084 If your target defines any fundamental types, or any types your 25085 target uses should be mangled differently from the default, define 25086 this hook to return the appropriate encoding for these types as 25087 part of a C++ mangled name. The TYPE argument is the tree 25088 structure representing the type to be mangled. The hook may be 25089 applied to trees which are not target-specific fundamental types; 25090 it should return `NULL' for all such types, as well as arguments 25091 it does not recognize. If the return value is not `NULL', it must 25092 point to a statically-allocated string constant. 25093 25094 Target-specific fundamental types might be new fundamental types or 25095 qualified versions of ordinary fundamental types. Encode new 25096 fundamental types as `u N NAME', where NAME is the name used for 25097 the type in source code, and N is the length of NAME in decimal. 25098 Encode qualified versions of ordinary types as `U N NAME CODE', 25099 where NAME is the name used for the type qualifier in source code, 25100 N is the length of NAME as above, and CODE is the code used to 25101 represent the unqualified version of this type. (See 25102 `write_builtin_type' in `cp/mangle.c' for the list of codes.) In 25103 both cases the spaces are for clarity; do not include any spaces 25104 in your string. 25105 25106 This hook is applied to types prior to typedef resolution. If the 25107 mangled name for a particular type depends only on that type's 25108 main variant, you can perform typedef resolution yourself using 25109 `TYPE_MAIN_VARIANT' before mangling. 25110 25111 The default version of this hook always returns `NULL', which is 25112 appropriate for a target that does not define any new fundamental 25113 types. 25114 25115 25116 File: gccint.info, Node: Type Layout, Next: Registers, Prev: Storage Layout, Up: Target Macros 25117 25118 17.6 Layout of Source Language Data Types 25119 ========================================= 25120 25121 These macros define the sizes and other characteristics of the standard 25122 basic data types used in programs being compiled. Unlike the macros in 25123 the previous section, these apply to specific features of C and related 25124 languages, rather than to fundamental aspects of storage layout. 25125 25126 -- Macro: INT_TYPE_SIZE 25127 A C expression for the size in bits of the type `int' on the 25128 target machine. If you don't define this, the default is one word. 25129 25130 -- Macro: SHORT_TYPE_SIZE 25131 A C expression for the size in bits of the type `short' on the 25132 target machine. If you don't define this, the default is half a 25133 word. (If this would be less than one storage unit, it is rounded 25134 up to one unit.) 25135 25136 -- Macro: LONG_TYPE_SIZE 25137 A C expression for the size in bits of the type `long' on the 25138 target machine. If you don't define this, the default is one word. 25139 25140 -- Macro: ADA_LONG_TYPE_SIZE 25141 On some machines, the size used for the Ada equivalent of the type 25142 `long' by a native Ada compiler differs from that used by C. In 25143 that situation, define this macro to be a C expression to be used 25144 for the size of that type. If you don't define this, the default 25145 is the value of `LONG_TYPE_SIZE'. 25146 25147 -- Macro: LONG_LONG_TYPE_SIZE 25148 A C expression for the size in bits of the type `long long' on the 25149 target machine. If you don't define this, the default is two 25150 words. If you want to support GNU Ada on your machine, the value 25151 of this macro must be at least 64. 25152 25153 -- Macro: CHAR_TYPE_SIZE 25154 A C expression for the size in bits of the type `char' on the 25155 target machine. If you don't define this, the default is 25156 `BITS_PER_UNIT'. 25157 25158 -- Macro: BOOL_TYPE_SIZE 25159 A C expression for the size in bits of the C++ type `bool' and C99 25160 type `_Bool' on the target machine. If you don't define this, and 25161 you probably shouldn't, the default is `CHAR_TYPE_SIZE'. 25162 25163 -- Macro: FLOAT_TYPE_SIZE 25164 A C expression for the size in bits of the type `float' on the 25165 target machine. If you don't define this, the default is one word. 25166 25167 -- Macro: DOUBLE_TYPE_SIZE 25168 A C expression for the size in bits of the type `double' on the 25169 target machine. If you don't define this, the default is two 25170 words. 25171 25172 -- Macro: LONG_DOUBLE_TYPE_SIZE 25173 A C expression for the size in bits of the type `long double' on 25174 the target machine. If you don't define this, the default is two 25175 words. 25176 25177 -- Macro: SHORT_FRACT_TYPE_SIZE 25178 A C expression for the size in bits of the type `short _Fract' on 25179 the target machine. If you don't define this, the default is 25180 `BITS_PER_UNIT'. 25181 25182 -- Macro: FRACT_TYPE_SIZE 25183 A C expression for the size in bits of the type `_Fract' on the 25184 target machine. If you don't define this, the default is 25185 `BITS_PER_UNIT * 2'. 25186 25187 -- Macro: LONG_FRACT_TYPE_SIZE 25188 A C expression for the size in bits of the type `long _Fract' on 25189 the target machine. If you don't define this, the default is 25190 `BITS_PER_UNIT * 4'. 25191 25192 -- Macro: LONG_LONG_FRACT_TYPE_SIZE 25193 A C expression for the size in bits of the type `long long _Fract' 25194 on the target machine. If you don't define this, the default is 25195 `BITS_PER_UNIT * 8'. 25196 25197 -- Macro: SHORT_ACCUM_TYPE_SIZE 25198 A C expression for the size in bits of the type `short _Accum' on 25199 the target machine. If you don't define this, the default is 25200 `BITS_PER_UNIT * 2'. 25201 25202 -- Macro: ACCUM_TYPE_SIZE 25203 A C expression for the size in bits of the type `_Accum' on the 25204 target machine. If you don't define this, the default is 25205 `BITS_PER_UNIT * 4'. 25206 25207 -- Macro: LONG_ACCUM_TYPE_SIZE 25208 A C expression for the size in bits of the type `long _Accum' on 25209 the target machine. If you don't define this, the default is 25210 `BITS_PER_UNIT * 8'. 25211 25212 -- Macro: LONG_LONG_ACCUM_TYPE_SIZE 25213 A C expression for the size in bits of the type `long long _Accum' 25214 on the target machine. If you don't define this, the default is 25215 `BITS_PER_UNIT * 16'. 25216 25217 -- Macro: LIBGCC2_LONG_DOUBLE_TYPE_SIZE 25218 Define this macro if `LONG_DOUBLE_TYPE_SIZE' is not constant or if 25219 you want routines in `libgcc2.a' for a size other than 25220 `LONG_DOUBLE_TYPE_SIZE'. If you don't define this, the default is 25221 `LONG_DOUBLE_TYPE_SIZE'. 25222 25223 -- Macro: LIBGCC2_HAS_DF_MODE 25224 Define this macro if neither `LIBGCC2_DOUBLE_TYPE_SIZE' nor 25225 `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is `DFmode' but you want `DFmode' 25226 routines in `libgcc2.a' anyway. If you don't define this and 25227 either `LIBGCC2_DOUBLE_TYPE_SIZE' or 25228 `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 64 then the default is 1, 25229 otherwise it is 0. 25230 25231 -- Macro: LIBGCC2_HAS_XF_MODE 25232 Define this macro if `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is not 25233 `XFmode' but you want `XFmode' routines in `libgcc2.a' anyway. If 25234 you don't define this and `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 80 25235 then the default is 1, otherwise it is 0. 25236 25237 -- Macro: LIBGCC2_HAS_TF_MODE 25238 Define this macro if `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is not 25239 `TFmode' but you want `TFmode' routines in `libgcc2.a' anyway. If 25240 you don't define this and `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 128 25241 then the default is 1, otherwise it is 0. 25242 25243 -- Macro: SF_SIZE 25244 -- Macro: DF_SIZE 25245 -- Macro: XF_SIZE 25246 -- Macro: TF_SIZE 25247 Define these macros to be the size in bits of the mantissa of 25248 `SFmode', `DFmode', `XFmode' and `TFmode' values, if the defaults 25249 in `libgcc2.h' are inappropriate. By default, `FLT_MANT_DIG' is 25250 used for `SF_SIZE', `LDBL_MANT_DIG' for `XF_SIZE' and `TF_SIZE', 25251 and `DBL_MANT_DIG' or `LDBL_MANT_DIG' for `DF_SIZE' according to 25252 whether `LIBGCC2_DOUBLE_TYPE_SIZE' or 25253 `LIBGCC2_LONG_DOUBLE_TYPE_SIZE' is 64. 25254 25255 -- Macro: TARGET_FLT_EVAL_METHOD 25256 A C expression for the value for `FLT_EVAL_METHOD' in `float.h', 25257 assuming, if applicable, that the floating-point control word is 25258 in its default state. If you do not define this macro the value of 25259 `FLT_EVAL_METHOD' will be zero. 25260 25261 -- Macro: WIDEST_HARDWARE_FP_SIZE 25262 A C expression for the size in bits of the widest floating-point 25263 format supported by the hardware. If you define this macro, you 25264 must specify a value less than or equal to the value of 25265 `LONG_DOUBLE_TYPE_SIZE'. If you do not define this macro, the 25266 value of `LONG_DOUBLE_TYPE_SIZE' is the default. 25267 25268 -- Macro: DEFAULT_SIGNED_CHAR 25269 An expression whose value is 1 or 0, according to whether the type 25270 `char' should be signed or unsigned by default. The user can 25271 always override this default with the options `-fsigned-char' and 25272 `-funsigned-char'. 25273 25274 -- Target Hook: bool TARGET_DEFAULT_SHORT_ENUMS (void) 25275 This target hook should return true if the compiler should give an 25276 `enum' type only as many bytes as it takes to represent the range 25277 of possible values of that type. It should return false if all 25278 `enum' types should be allocated like `int'. 25279 25280 The default is to return false. 25281 25282 -- Macro: SIZE_TYPE 25283 A C expression for a string describing the name of the data type 25284 to use for size values. The typedef name `size_t' is defined 25285 using the contents of the string. 25286 25287 The string can contain more than one keyword. If so, separate 25288 them with spaces, and write first any length keyword, then 25289 `unsigned' if appropriate, and finally `int'. The string must 25290 exactly match one of the data type names defined in the function 25291 `init_decl_processing' in the file `c-decl.c'. You may not omit 25292 `int' or change the order--that would cause the compiler to crash 25293 on startup. 25294 25295 If you don't define this macro, the default is `"long unsigned 25296 int"'. 25297 25298 -- Macro: PTRDIFF_TYPE 25299 A C expression for a string describing the name of the data type 25300 to use for the result of subtracting two pointers. The typedef 25301 name `ptrdiff_t' is defined using the contents of the string. See 25302 `SIZE_TYPE' above for more information. 25303 25304 If you don't define this macro, the default is `"long int"'. 25305 25306 -- Macro: WCHAR_TYPE 25307 A C expression for a string describing the name of the data type 25308 to use for wide characters. The typedef name `wchar_t' is defined 25309 using the contents of the string. See `SIZE_TYPE' above for more 25310 information. 25311 25312 If you don't define this macro, the default is `"int"'. 25313 25314 -- Macro: WCHAR_TYPE_SIZE 25315 A C expression for the size in bits of the data type for wide 25316 characters. This is used in `cpp', which cannot make use of 25317 `WCHAR_TYPE'. 25318 25319 -- Macro: WINT_TYPE 25320 A C expression for a string describing the name of the data type to 25321 use for wide characters passed to `printf' and returned from 25322 `getwc'. The typedef name `wint_t' is defined using the contents 25323 of the string. See `SIZE_TYPE' above for more information. 25324 25325 If you don't define this macro, the default is `"unsigned int"'. 25326 25327 -- Macro: INTMAX_TYPE 25328 A C expression for a string describing the name of the data type 25329 that can represent any value of any standard or extended signed 25330 integer type. The typedef name `intmax_t' is defined using the 25331 contents of the string. See `SIZE_TYPE' above for more 25332 information. 25333 25334 If you don't define this macro, the default is the first of 25335 `"int"', `"long int"', or `"long long int"' that has as much 25336 precision as `long long int'. 25337 25338 -- Macro: UINTMAX_TYPE 25339 A C expression for a string describing the name of the data type 25340 that can represent any value of any standard or extended unsigned 25341 integer type. The typedef name `uintmax_t' is defined using the 25342 contents of the string. See `SIZE_TYPE' above for more 25343 information. 25344 25345 If you don't define this macro, the default is the first of 25346 `"unsigned int"', `"long unsigned int"', or `"long long unsigned 25347 int"' that has as much precision as `long long unsigned int'. 25348 25349 -- Macro: TARGET_PTRMEMFUNC_VBIT_LOCATION 25350 The C++ compiler represents a pointer-to-member-function with a 25351 struct that looks like: 25352 25353 struct { 25354 union { 25355 void (*fn)(); 25356 ptrdiff_t vtable_index; 25357 }; 25358 ptrdiff_t delta; 25359 }; 25360 25361 The C++ compiler must use one bit to indicate whether the function 25362 that will be called through a pointer-to-member-function is 25363 virtual. Normally, we assume that the low-order bit of a function 25364 pointer must always be zero. Then, by ensuring that the 25365 vtable_index is odd, we can distinguish which variant of the union 25366 is in use. But, on some platforms function pointers can be odd, 25367 and so this doesn't work. In that case, we use the low-order bit 25368 of the `delta' field, and shift the remainder of the `delta' field 25369 to the left. 25370 25371 GCC will automatically make the right selection about where to 25372 store this bit using the `FUNCTION_BOUNDARY' setting for your 25373 platform. However, some platforms such as ARM/Thumb have 25374 `FUNCTION_BOUNDARY' set such that functions always start at even 25375 addresses, but the lowest bit of pointers to functions indicate 25376 whether the function at that address is in ARM or Thumb mode. If 25377 this is the case of your architecture, you should define this 25378 macro to `ptrmemfunc_vbit_in_delta'. 25379 25380 In general, you should not have to define this macro. On 25381 architectures in which function addresses are always even, 25382 according to `FUNCTION_BOUNDARY', GCC will automatically define 25383 this macro to `ptrmemfunc_vbit_in_pfn'. 25384 25385 -- Macro: TARGET_VTABLE_USES_DESCRIPTORS 25386 Normally, the C++ compiler uses function pointers in vtables. This 25387 macro allows the target to change to use "function descriptors" 25388 instead. Function descriptors are found on targets for whom a 25389 function pointer is actually a small data structure. Normally the 25390 data structure consists of the actual code address plus a data 25391 pointer to which the function's data is relative. 25392 25393 If vtables are used, the value of this macro should be the number 25394 of words that the function descriptor occupies. 25395 25396 -- Macro: TARGET_VTABLE_ENTRY_ALIGN 25397 By default, the vtable entries are void pointers, the so the 25398 alignment is the same as pointer alignment. The value of this 25399 macro specifies the alignment of the vtable entry in bits. It 25400 should be defined only when special alignment is necessary. */ 25401 25402 -- Macro: TARGET_VTABLE_DATA_ENTRY_DISTANCE 25403 There are a few non-descriptor entries in the vtable at offsets 25404 below zero. If these entries must be padded (say, to preserve the 25405 alignment specified by `TARGET_VTABLE_ENTRY_ALIGN'), set this to 25406 the number of words in each data entry. 25407 25408 25409 File: gccint.info, Node: Registers, Next: Register Classes, Prev: Type Layout, Up: Target Macros 25410 25411 17.7 Register Usage 25412 =================== 25413 25414 This section explains how to describe what registers the target machine 25415 has, and how (in general) they can be used. 25416 25417 The description of which registers a specific instruction can use is 25418 done with register classes; see *note Register Classes::. For 25419 information on using registers to access a stack frame, see *note Frame 25420 Registers::. For passing values in registers, see *note Register 25421 Arguments::. For returning values in registers, see *note Scalar 25422 Return::. 25423 25424 * Menu: 25425 25426 * Register Basics:: Number and kinds of registers. 25427 * Allocation Order:: Order in which registers are allocated. 25428 * Values in Registers:: What kinds of values each reg can hold. 25429 * Leaf Functions:: Renumbering registers for leaf functions. 25430 * Stack Registers:: Handling a register stack such as 80387. 25431 25432 25433 File: gccint.info, Node: Register Basics, Next: Allocation Order, Up: Registers 25434 25435 17.7.1 Basic Characteristics of Registers 25436 ----------------------------------------- 25437 25438 Registers have various characteristics. 25439 25440 -- Macro: FIRST_PSEUDO_REGISTER 25441 Number of hardware registers known to the compiler. They receive 25442 numbers 0 through `FIRST_PSEUDO_REGISTER-1'; thus, the first 25443 pseudo register's number really is assigned the number 25444 `FIRST_PSEUDO_REGISTER'. 25445 25446 -- Macro: FIXED_REGISTERS 25447 An initializer that says which registers are used for fixed 25448 purposes all throughout the compiled code and are therefore not 25449 available for general allocation. These would include the stack 25450 pointer, the frame pointer (except on machines where that can be 25451 used as a general register when no frame pointer is needed), the 25452 program counter on machines where that is considered one of the 25453 addressable registers, and any other numbered register with a 25454 standard use. 25455 25456 This information is expressed as a sequence of numbers, separated 25457 by commas and surrounded by braces. The Nth number is 1 if 25458 register N is fixed, 0 otherwise. 25459 25460 The table initialized from this macro, and the table initialized by 25461 the following one, may be overridden at run time either 25462 automatically, by the actions of the macro 25463 `CONDITIONAL_REGISTER_USAGE', or by the user with the command 25464 options `-ffixed-REG', `-fcall-used-REG' and `-fcall-saved-REG'. 25465 25466 -- Macro: CALL_USED_REGISTERS 25467 Like `FIXED_REGISTERS' but has 1 for each register that is 25468 clobbered (in general) by function calls as well as for fixed 25469 registers. This macro therefore identifies the registers that are 25470 not available for general allocation of values that must live 25471 across function calls. 25472 25473 If a register has 0 in `CALL_USED_REGISTERS', the compiler 25474 automatically saves it on function entry and restores it on 25475 function exit, if the register is used within the function. 25476 25477 -- Macro: CALL_REALLY_USED_REGISTERS 25478 Like `CALL_USED_REGISTERS' except this macro doesn't require that 25479 the entire set of `FIXED_REGISTERS' be included. 25480 (`CALL_USED_REGISTERS' must be a superset of `FIXED_REGISTERS'). 25481 This macro is optional. If not specified, it defaults to the value 25482 of `CALL_USED_REGISTERS'. 25483 25484 -- Macro: HARD_REGNO_CALL_PART_CLOBBERED (REGNO, MODE) 25485 A C expression that is nonzero if it is not permissible to store a 25486 value of mode MODE in hard register number REGNO across a call 25487 without some part of it being clobbered. For most machines this 25488 macro need not be defined. It is only required for machines that 25489 do not preserve the entire contents of a register across a call. 25490 25491 -- Macro: CONDITIONAL_REGISTER_USAGE 25492 Zero or more C statements that may conditionally modify five 25493 variables `fixed_regs', `call_used_regs', `global_regs', 25494 `reg_names', and `reg_class_contents', to take into account any 25495 dependence of these register sets on target flags. The first three 25496 of these are of type `char []' (interpreted as Boolean vectors). 25497 `global_regs' is a `const char *[]', and `reg_class_contents' is a 25498 `HARD_REG_SET'. Before the macro is called, `fixed_regs', 25499 `call_used_regs', `reg_class_contents', and `reg_names' have been 25500 initialized from `FIXED_REGISTERS', `CALL_USED_REGISTERS', 25501 `REG_CLASS_CONTENTS', and `REGISTER_NAMES', respectively. 25502 `global_regs' has been cleared, and any `-ffixed-REG', 25503 `-fcall-used-REG' and `-fcall-saved-REG' command options have been 25504 applied. 25505 25506 You need not define this macro if it has no work to do. 25507 25508 If the usage of an entire class of registers depends on the target 25509 flags, you may indicate this to GCC by using this macro to modify 25510 `fixed_regs' and `call_used_regs' to 1 for each of the registers 25511 in the classes which should not be used by GCC. Also define the 25512 macro `REG_CLASS_FROM_LETTER' / `REG_CLASS_FROM_CONSTRAINT' to 25513 return `NO_REGS' if it is called with a letter for a class that 25514 shouldn't be used. 25515 25516 (However, if this class is not included in `GENERAL_REGS' and all 25517 of the insn patterns whose constraints permit this class are 25518 controlled by target switches, then GCC will automatically avoid 25519 using these registers when the target switches are opposed to 25520 them.) 25521 25522 -- Macro: INCOMING_REGNO (OUT) 25523 Define this macro if the target machine has register windows. 25524 This C expression returns the register number as seen by the 25525 called function corresponding to the register number OUT as seen 25526 by the calling function. Return OUT if register number OUT is not 25527 an outbound register. 25528 25529 -- Macro: OUTGOING_REGNO (IN) 25530 Define this macro if the target machine has register windows. 25531 This C expression returns the register number as seen by the 25532 calling function corresponding to the register number IN as seen 25533 by the called function. Return IN if register number IN is not an 25534 inbound register. 25535 25536 -- Macro: LOCAL_REGNO (REGNO) 25537 Define this macro if the target machine has register windows. 25538 This C expression returns true if the register is call-saved but 25539 is in the register window. Unlike most call-saved registers, such 25540 registers need not be explicitly restored on function exit or 25541 during non-local gotos. 25542 25543 -- Macro: PC_REGNUM 25544 If the program counter has a register number, define this as that 25545 register number. Otherwise, do not define it. 25546 25547 25548 File: gccint.info, Node: Allocation Order, Next: Values in Registers, Prev: Register Basics, Up: Registers 25549 25550 17.7.2 Order of Allocation of Registers 25551 --------------------------------------- 25552 25553 Registers are allocated in order. 25554 25555 -- Macro: REG_ALLOC_ORDER 25556 If defined, an initializer for a vector of integers, containing the 25557 numbers of hard registers in the order in which GCC should prefer 25558 to use them (from most preferred to least). 25559 25560 If this macro is not defined, registers are used lowest numbered 25561 first (all else being equal). 25562 25563 One use of this macro is on machines where the highest numbered 25564 registers must always be saved and the save-multiple-registers 25565 instruction supports only sequences of consecutive registers. On 25566 such machines, define `REG_ALLOC_ORDER' to be an initializer that 25567 lists the highest numbered allocable register first. 25568 25569 -- Macro: ADJUST_REG_ALLOC_ORDER 25570 A C statement (sans semicolon) to choose the order in which to 25571 allocate hard registers for pseudo-registers local to a basic 25572 block. 25573 25574 Store the desired register order in the array `reg_alloc_order'. 25575 Element 0 should be the register to allocate first; element 1, the 25576 next register; and so on. 25577 25578 The macro body should not assume anything about the contents of 25579 `reg_alloc_order' before execution of the macro. 25580 25581 On most machines, it is not necessary to define this macro. 25582 25583 -- Macro: HONOR_REG_ALLOC_ORDER 25584 Normally, IRA tries to estimate the costs for saving a register in 25585 the prologue and restoring it in the epilogue. This discourages 25586 it from using call-saved registers. If a machine wants to ensure 25587 that IRA allocates registers in the order given by REG_ALLOC_ORDER 25588 even if some call-saved registers appear earlier than call-used 25589 ones, this macro should be defined. 25590 25591 -- Macro: IRA_HARD_REGNO_ADD_COST_MULTIPLIER (REGNO) 25592 In some case register allocation order is not enough for the 25593 Integrated Register Allocator (IRA) to generate a good code. If 25594 this macro is defined, it should return a floating point value 25595 based on REGNO. The cost of using REGNO for a pseudo will be 25596 increased by approximately the pseudo's usage frequency times the 25597 value returned by this macro. Not defining this macro is 25598 equivalent to having it always return `0.0'. 25599 25600 On most machines, it is not necessary to define this macro. 25601 25602 25603 File: gccint.info, Node: Values in Registers, Next: Leaf Functions, Prev: Allocation Order, Up: Registers 25604 25605 17.7.3 How Values Fit in Registers 25606 ---------------------------------- 25607 25608 This section discusses the macros that describe which kinds of values 25609 (specifically, which machine modes) each register can hold, and how many 25610 consecutive registers are needed for a given mode. 25611 25612 -- Macro: HARD_REGNO_NREGS (REGNO, MODE) 25613 A C expression for the number of consecutive hard registers, 25614 starting at register number REGNO, required to hold a value of mode 25615 MODE. This macro must never return zero, even if a register 25616 cannot hold the requested mode - indicate that with 25617 HARD_REGNO_MODE_OK and/or CANNOT_CHANGE_MODE_CLASS instead. 25618 25619 On a machine where all registers are exactly one word, a suitable 25620 definition of this macro is 25621 25622 #define HARD_REGNO_NREGS(REGNO, MODE) \ 25623 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ 25624 / UNITS_PER_WORD) 25625 25626 -- Macro: HARD_REGNO_NREGS_HAS_PADDING (REGNO, MODE) 25627 A C expression that is nonzero if a value of mode MODE, stored in 25628 memory, ends with padding that causes it to take up more space than 25629 in registers starting at register number REGNO (as determined by 25630 multiplying GCC's notion of the size of the register when 25631 containing this mode by the number of registers returned by 25632 `HARD_REGNO_NREGS'). By default this is zero. 25633 25634 For example, if a floating-point value is stored in three 32-bit 25635 registers but takes up 128 bits in memory, then this would be 25636 nonzero. 25637 25638 This macros only needs to be defined if there are cases where 25639 `subreg_get_info' would otherwise wrongly determine that a 25640 `subreg' can be represented by an offset to the register number, 25641 when in fact such a `subreg' would contain some of the padding not 25642 stored in registers and so not be representable. 25643 25644 -- Macro: HARD_REGNO_NREGS_WITH_PADDING (REGNO, MODE) 25645 For values of REGNO and MODE for which 25646 `HARD_REGNO_NREGS_HAS_PADDING' returns nonzero, a C expression 25647 returning the greater number of registers required to hold the 25648 value including any padding. In the example above, the value 25649 would be four. 25650 25651 -- Macro: REGMODE_NATURAL_SIZE (MODE) 25652 Define this macro if the natural size of registers that hold values 25653 of mode MODE is not the word size. It is a C expression that 25654 should give the natural size in bytes for the specified mode. It 25655 is used by the register allocator to try to optimize its results. 25656 This happens for example on SPARC 64-bit where the natural size of 25657 floating-point registers is still 32-bit. 25658 25659 -- Macro: HARD_REGNO_MODE_OK (REGNO, MODE) 25660 A C expression that is nonzero if it is permissible to store a 25661 value of mode MODE in hard register number REGNO (or in several 25662 registers starting with that one). For a machine where all 25663 registers are equivalent, a suitable definition is 25664 25665 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1 25666 25667 You need not include code to check for the numbers of fixed 25668 registers, because the allocation mechanism considers them to be 25669 always occupied. 25670 25671 On some machines, double-precision values must be kept in even/odd 25672 register pairs. You can implement that by defining this macro to 25673 reject odd register numbers for such modes. 25674 25675 The minimum requirement for a mode to be OK in a register is that 25676 the `movMODE' instruction pattern support moves between the 25677 register and other hard register in the same class and that moving 25678 a value into the register and back out not alter it. 25679 25680 Since the same instruction used to move `word_mode' will work for 25681 all narrower integer modes, it is not necessary on any machine for 25682 `HARD_REGNO_MODE_OK' to distinguish between these modes, provided 25683 you define patterns `movhi', etc., to take advantage of this. This 25684 is useful because of the interaction between `HARD_REGNO_MODE_OK' 25685 and `MODES_TIEABLE_P'; it is very desirable for all integer modes 25686 to be tieable. 25687 25688 Many machines have special registers for floating point arithmetic. 25689 Often people assume that floating point machine modes are allowed 25690 only in floating point registers. This is not true. Any 25691 registers that can hold integers can safely _hold_ a floating 25692 point machine mode, whether or not floating arithmetic can be done 25693 on it in those registers. Integer move instructions can be used 25694 to move the values. 25695 25696 On some machines, though, the converse is true: fixed-point machine 25697 modes may not go in floating registers. This is true if the 25698 floating registers normalize any value stored in them, because 25699 storing a non-floating value there would garble it. In this case, 25700 `HARD_REGNO_MODE_OK' should reject fixed-point machine modes in 25701 floating registers. But if the floating registers do not 25702 automatically normalize, if you can store any bit pattern in one 25703 and retrieve it unchanged without a trap, then any machine mode 25704 may go in a floating register, so you can define this macro to say 25705 so. 25706 25707 The primary significance of special floating registers is rather 25708 that they are the registers acceptable in floating point arithmetic 25709 instructions. However, this is of no concern to 25710 `HARD_REGNO_MODE_OK'. You handle it by writing the proper 25711 constraints for those instructions. 25712 25713 On some machines, the floating registers are especially slow to 25714 access, so that it is better to store a value in a stack frame 25715 than in such a register if floating point arithmetic is not being 25716 done. As long as the floating registers are not in class 25717 `GENERAL_REGS', they will not be used unless some pattern's 25718 constraint asks for one. 25719 25720 -- Macro: HARD_REGNO_RENAME_OK (FROM, TO) 25721 A C expression that is nonzero if it is OK to rename a hard 25722 register FROM to another hard register TO. 25723 25724 One common use of this macro is to prevent renaming of a register 25725 to another register that is not saved by a prologue in an interrupt 25726 handler. 25727 25728 The default is always nonzero. 25729 25730 -- Macro: MODES_TIEABLE_P (MODE1, MODE2) 25731 A C expression that is nonzero if a value of mode MODE1 is 25732 accessible in mode MODE2 without copying. 25733 25734 If `HARD_REGNO_MODE_OK (R, MODE1)' and `HARD_REGNO_MODE_OK (R, 25735 MODE2)' are always the same for any R, then `MODES_TIEABLE_P 25736 (MODE1, MODE2)' should be nonzero. If they differ for any R, you 25737 should define this macro to return zero unless some other 25738 mechanism ensures the accessibility of the value in a narrower 25739 mode. 25740 25741 You should define this macro to return nonzero in as many cases as 25742 possible since doing so will allow GCC to perform better register 25743 allocation. 25744 25745 -- Target Hook: bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int REGNO) 25746 This target hook should return `true' if it is OK to use a hard 25747 register REGNO as scratch reg in peephole2. 25748 25749 One common use of this macro is to prevent using of a register that 25750 is not saved by a prologue in an interrupt handler. 25751 25752 The default version of this hook always returns `true'. 25753 25754 -- Macro: AVOID_CCMODE_COPIES 25755 Define this macro if the compiler should avoid copies to/from 25756 `CCmode' registers. You should only define this macro if support 25757 for copying to/from `CCmode' is incomplete. 25758 25759 25760 File: gccint.info, Node: Leaf Functions, Next: Stack Registers, Prev: Values in Registers, Up: Registers 25761 25762 17.7.4 Handling Leaf Functions 25763 ------------------------------ 25764 25765 On some machines, a leaf function (i.e., one which makes no calls) can 25766 run more efficiently if it does not make its own register window. 25767 Often this means it is required to receive its arguments in the 25768 registers where they are passed by the caller, instead of the registers 25769 where they would normally arrive. 25770 25771 The special treatment for leaf functions generally applies only when 25772 other conditions are met; for example, often they may use only those 25773 registers for its own variables and temporaries. We use the term "leaf 25774 function" to mean a function that is suitable for this special 25775 handling, so that functions with no calls are not necessarily "leaf 25776 functions". 25777 25778 GCC assigns register numbers before it knows whether the function is 25779 suitable for leaf function treatment. So it needs to renumber the 25780 registers in order to output a leaf function. The following macros 25781 accomplish this. 25782 25783 -- Macro: LEAF_REGISTERS 25784 Name of a char vector, indexed by hard register number, which 25785 contains 1 for a register that is allowable in a candidate for leaf 25786 function treatment. 25787 25788 If leaf function treatment involves renumbering the registers, 25789 then the registers marked here should be the ones before 25790 renumbering--those that GCC would ordinarily allocate. The 25791 registers which will actually be used in the assembler code, after 25792 renumbering, should not be marked with 1 in this vector. 25793 25794 Define this macro only if the target machine offers a way to 25795 optimize the treatment of leaf functions. 25796 25797 -- Macro: LEAF_REG_REMAP (REGNO) 25798 A C expression whose value is the register number to which REGNO 25799 should be renumbered, when a function is treated as a leaf 25800 function. 25801 25802 If REGNO is a register number which should not appear in a leaf 25803 function before renumbering, then the expression should yield -1, 25804 which will cause the compiler to abort. 25805 25806 Define this macro only if the target machine offers a way to 25807 optimize the treatment of leaf functions, and registers need to be 25808 renumbered to do this. 25809 25810 `TARGET_ASM_FUNCTION_PROLOGUE' and `TARGET_ASM_FUNCTION_EPILOGUE' must 25811 usually treat leaf functions specially. They can test the C variable 25812 `current_function_is_leaf' which is nonzero for leaf functions. 25813 `current_function_is_leaf' is set prior to local register allocation 25814 and is valid for the remaining compiler passes. They can also test the 25815 C variable `current_function_uses_only_leaf_regs' which is nonzero for 25816 leaf functions which only use leaf registers. 25817 `current_function_uses_only_leaf_regs' is valid after all passes that 25818 modify the instructions have been run and is only useful if 25819 `LEAF_REGISTERS' is defined. 25820 25821 25822 File: gccint.info, Node: Stack Registers, Prev: Leaf Functions, Up: Registers 25823 25824 17.7.5 Registers That Form a Stack 25825 ---------------------------------- 25826 25827 There are special features to handle computers where some of the 25828 "registers" form a stack. Stack registers are normally written by 25829 pushing onto the stack, and are numbered relative to the top of the 25830 stack. 25831 25832 Currently, GCC can only handle one group of stack-like registers, and 25833 they must be consecutively numbered. Furthermore, the existing support 25834 for stack-like registers is specific to the 80387 floating point 25835 coprocessor. If you have a new architecture that uses stack-like 25836 registers, you will need to do substantial work on `reg-stack.c' and 25837 write your machine description to cooperate with it, as well as 25838 defining these macros. 25839 25840 -- Macro: STACK_REGS 25841 Define this if the machine has any stack-like registers. 25842 25843 -- Macro: FIRST_STACK_REG 25844 The number of the first stack-like register. This one is the top 25845 of the stack. 25846 25847 -- Macro: LAST_STACK_REG 25848 The number of the last stack-like register. This one is the 25849 bottom of the stack. 25850 25851 25852 File: gccint.info, Node: Register Classes, Next: Old Constraints, Prev: Registers, Up: Target Macros 25853 25854 17.8 Register Classes 25855 ===================== 25856 25857 On many machines, the numbered registers are not all equivalent. For 25858 example, certain registers may not be allowed for indexed addressing; 25859 certain registers may not be allowed in some instructions. These 25860 machine restrictions are described to the compiler using "register 25861 classes". 25862 25863 You define a number of register classes, giving each one a name and 25864 saying which of the registers belong to it. Then you can specify 25865 register classes that are allowed as operands to particular instruction 25866 patterns. 25867 25868 In general, each register will belong to several classes. In fact, one 25869 class must be named `ALL_REGS' and contain all the registers. Another 25870 class must be named `NO_REGS' and contain no registers. Often the 25871 union of two classes will be another class; however, this is not 25872 required. 25873 25874 One of the classes must be named `GENERAL_REGS'. There is nothing 25875 terribly special about the name, but the operand constraint letters `r' 25876 and `g' specify this class. If `GENERAL_REGS' is the same as 25877 `ALL_REGS', just define it as a macro which expands to `ALL_REGS'. 25878 25879 Order the classes so that if class X is contained in class Y then X 25880 has a lower class number than Y. 25881 25882 The way classes other than `GENERAL_REGS' are specified in operand 25883 constraints is through machine-dependent operand constraint letters. 25884 You can define such letters to correspond to various classes, then use 25885 them in operand constraints. 25886 25887 You should define a class for the union of two classes whenever some 25888 instruction allows both classes. For example, if an instruction allows 25889 either a floating point (coprocessor) register or a general register 25890 for a certain operand, you should define a class `FLOAT_OR_GENERAL_REGS' 25891 which includes both of them. Otherwise you will get suboptimal code. 25892 25893 You must also specify certain redundant information about the register 25894 classes: for each class, which classes contain it and which ones are 25895 contained in it; for each pair of classes, the largest class contained 25896 in their union. 25897 25898 When a value occupying several consecutive registers is expected in a 25899 certain class, all the registers used must belong to that class. 25900 Therefore, register classes cannot be used to enforce a requirement for 25901 a register pair to start with an even-numbered register. The way to 25902 specify this requirement is with `HARD_REGNO_MODE_OK'. 25903 25904 Register classes used for input-operands of bitwise-and or shift 25905 instructions have a special requirement: each such class must have, for 25906 each fixed-point machine mode, a subclass whose registers can transfer 25907 that mode to or from memory. For example, on some machines, the 25908 operations for single-byte values (`QImode') are limited to certain 25909 registers. When this is so, each register class that is used in a 25910 bitwise-and or shift instruction must have a subclass consisting of 25911 registers from which single-byte values can be loaded or stored. This 25912 is so that `PREFERRED_RELOAD_CLASS' can always have a possible value to 25913 return. 25914 25915 -- Data type: enum reg_class 25916 An enumerated type that must be defined with all the register 25917 class names as enumerated values. `NO_REGS' must be first. 25918 `ALL_REGS' must be the last register class, followed by one more 25919 enumerated value, `LIM_REG_CLASSES', which is not a register class 25920 but rather tells how many classes there are. 25921 25922 Each register class has a number, which is the value of casting 25923 the class name to type `int'. The number serves as an index in 25924 many of the tables described below. 25925 25926 -- Macro: N_REG_CLASSES 25927 The number of distinct register classes, defined as follows: 25928 25929 #define N_REG_CLASSES (int) LIM_REG_CLASSES 25930 25931 -- Macro: REG_CLASS_NAMES 25932 An initializer containing the names of the register classes as C 25933 string constants. These names are used in writing some of the 25934 debugging dumps. 25935 25936 -- Macro: REG_CLASS_CONTENTS 25937 An initializer containing the contents of the register classes, as 25938 integers which are bit masks. The Nth integer specifies the 25939 contents of class N. The way the integer MASK is interpreted is 25940 that register R is in the class if `MASK & (1 << R)' is 1. 25941 25942 When the machine has more than 32 registers, an integer does not 25943 suffice. Then the integers are replaced by sub-initializers, 25944 braced groupings containing several integers. Each 25945 sub-initializer must be suitable as an initializer for the type 25946 `HARD_REG_SET' which is defined in `hard-reg-set.h'. In this 25947 situation, the first integer in each sub-initializer corresponds to 25948 registers 0 through 31, the second integer to registers 32 through 25949 63, and so on. 25950 25951 -- Macro: REGNO_REG_CLASS (REGNO) 25952 A C expression whose value is a register class containing hard 25953 register REGNO. In general there is more than one such class; 25954 choose a class which is "minimal", meaning that no smaller class 25955 also contains the register. 25956 25957 -- Macro: BASE_REG_CLASS 25958 A macro whose definition is the name of the class to which a valid 25959 base register must belong. A base register is one used in an 25960 address which is the register value plus a displacement. 25961 25962 -- Macro: MODE_BASE_REG_CLASS (MODE) 25963 This is a variation of the `BASE_REG_CLASS' macro which allows the 25964 selection of a base register in a mode dependent manner. If MODE 25965 is VOIDmode then it should return the same value as 25966 `BASE_REG_CLASS'. 25967 25968 -- Macro: MODE_BASE_REG_REG_CLASS (MODE) 25969 A C expression whose value is the register class to which a valid 25970 base register must belong in order to be used in a base plus index 25971 register address. You should define this macro if base plus index 25972 addresses have different requirements than other base register 25973 uses. 25974 25975 -- Macro: MODE_CODE_BASE_REG_CLASS (MODE, OUTER_CODE, INDEX_CODE) 25976 A C expression whose value is the register class to which a valid 25977 base register must belong. OUTER_CODE and INDEX_CODE define the 25978 context in which the base register occurs. OUTER_CODE is the code 25979 of the immediately enclosing expression (`MEM' for the top level 25980 of an address, `ADDRESS' for something that occurs in an 25981 `address_operand'). INDEX_CODE is the code of the corresponding 25982 index expression if OUTER_CODE is `PLUS'; `SCRATCH' otherwise. 25983 25984 -- Macro: INDEX_REG_CLASS 25985 A macro whose definition is the name of the class to which a valid 25986 index register must belong. An index register is one used in an 25987 address where its value is either multiplied by a scale factor or 25988 added to another register (as well as added to a displacement). 25989 25990 -- Macro: REGNO_OK_FOR_BASE_P (NUM) 25991 A C expression which is nonzero if register number NUM is suitable 25992 for use as a base register in operand addresses. It may be either 25993 a suitable hard register or a pseudo register that has been 25994 allocated such a hard register. 25995 25996 -- Macro: REGNO_MODE_OK_FOR_BASE_P (NUM, MODE) 25997 A C expression that is just like `REGNO_OK_FOR_BASE_P', except that 25998 that expression may examine the mode of the memory reference in 25999 MODE. You should define this macro if the mode of the memory 26000 reference affects whether a register may be used as a base 26001 register. If you define this macro, the compiler will use it 26002 instead of `REGNO_OK_FOR_BASE_P'. The mode may be `VOIDmode' for 26003 addresses that appear outside a `MEM', i.e., as an 26004 `address_operand'. 26005 26006 26007 -- Macro: REGNO_MODE_OK_FOR_REG_BASE_P (NUM, MODE) 26008 A C expression which is nonzero if register number NUM is suitable 26009 for use as a base register in base plus index operand addresses, 26010 accessing memory in mode MODE. It may be either a suitable hard 26011 register or a pseudo register that has been allocated such a hard 26012 register. You should define this macro if base plus index 26013 addresses have different requirements than other base register 26014 uses. 26015 26016 Use of this macro is deprecated; please use the more general 26017 `REGNO_MODE_CODE_OK_FOR_BASE_P'. 26018 26019 -- Macro: REGNO_MODE_CODE_OK_FOR_BASE_P (NUM, MODE, OUTER_CODE, 26020 INDEX_CODE) 26021 A C expression that is just like `REGNO_MODE_OK_FOR_BASE_P', except 26022 that that expression may examine the context in which the register 26023 appears in the memory reference. OUTER_CODE is the code of the 26024 immediately enclosing expression (`MEM' if at the top level of the 26025 address, `ADDRESS' for something that occurs in an 26026 `address_operand'). INDEX_CODE is the code of the corresponding 26027 index expression if OUTER_CODE is `PLUS'; `SCRATCH' otherwise. 26028 The mode may be `VOIDmode' for addresses that appear outside a 26029 `MEM', i.e., as an `address_operand'. 26030 26031 -- Macro: REGNO_OK_FOR_INDEX_P (NUM) 26032 A C expression which is nonzero if register number NUM is suitable 26033 for use as an index register in operand addresses. It may be 26034 either a suitable hard register or a pseudo register that has been 26035 allocated such a hard register. 26036 26037 The difference between an index register and a base register is 26038 that the index register may be scaled. If an address involves the 26039 sum of two registers, neither one of them scaled, then either one 26040 may be labeled the "base" and the other the "index"; but whichever 26041 labeling is used must fit the machine's constraints of which 26042 registers may serve in each capacity. The compiler will try both 26043 labelings, looking for one that is valid, and will reload one or 26044 both registers only if neither labeling works. 26045 26046 -- Macro: PREFERRED_RELOAD_CLASS (X, CLASS) 26047 A C expression that places additional restrictions on the register 26048 class to use when it is necessary to copy value X into a register 26049 in class CLASS. The value is a register class; perhaps CLASS, or 26050 perhaps another, smaller class. On many machines, the following 26051 definition is safe: 26052 26053 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS 26054 26055 Sometimes returning a more restrictive class makes better code. 26056 For example, on the 68000, when X is an integer constant that is 26057 in range for a `moveq' instruction, the value of this macro is 26058 always `DATA_REGS' as long as CLASS includes the data registers. 26059 Requiring a data register guarantees that a `moveq' will be used. 26060 26061 One case where `PREFERRED_RELOAD_CLASS' must not return CLASS is 26062 if X is a legitimate constant which cannot be loaded into some 26063 register class. By returning `NO_REGS' you can force X into a 26064 memory location. For example, rs6000 can load immediate values 26065 into general-purpose registers, but does not have an instruction 26066 for loading an immediate value into a floating-point register, so 26067 `PREFERRED_RELOAD_CLASS' returns `NO_REGS' when X is a 26068 floating-point constant. If the constant can't be loaded into any 26069 kind of register, code generation will be better if 26070 `LEGITIMATE_CONSTANT_P' makes the constant illegitimate instead of 26071 using `PREFERRED_RELOAD_CLASS'. 26072 26073 If an insn has pseudos in it after register allocation, reload 26074 will go through the alternatives and call repeatedly 26075 `PREFERRED_RELOAD_CLASS' to find the best one. Returning 26076 `NO_REGS', in this case, makes reload add a `!' in front of the 26077 constraint: the x86 back-end uses this feature to discourage usage 26078 of 387 registers when math is done in the SSE registers (and vice 26079 versa). 26080 26081 -- Macro: PREFERRED_OUTPUT_RELOAD_CLASS (X, CLASS) 26082 Like `PREFERRED_RELOAD_CLASS', but for output reloads instead of 26083 input reloads. If you don't define this macro, the default is to 26084 use CLASS, unchanged. 26085 26086 You can also use `PREFERRED_OUTPUT_RELOAD_CLASS' to discourage 26087 reload from using some alternatives, like `PREFERRED_RELOAD_CLASS'. 26088 26089 -- Macro: LIMIT_RELOAD_CLASS (MODE, CLASS) 26090 A C expression that places additional restrictions on the register 26091 class to use when it is necessary to be able to hold a value of 26092 mode MODE in a reload register for which class CLASS would 26093 ordinarily be used. 26094 26095 Unlike `PREFERRED_RELOAD_CLASS', this macro should be used when 26096 there are certain modes that simply can't go in certain reload 26097 classes. 26098 26099 The value is a register class; perhaps CLASS, or perhaps another, 26100 smaller class. 26101 26102 Don't define this macro unless the target machine has limitations 26103 which require the macro to do something nontrivial. 26104 26105 -- Target Hook: enum reg_class TARGET_SECONDARY_RELOAD (bool IN_P, rtx 26106 X, enum reg_class RELOAD_CLASS, enum machine_mode 26107 RELOAD_MODE, secondary_reload_info *SRI) 26108 Many machines have some registers that cannot be copied directly 26109 to or from memory or even from other types of registers. An 26110 example is the `MQ' register, which on most machines, can only be 26111 copied to or from general registers, but not memory. Below, we 26112 shall be using the term 'intermediate register' when a move 26113 operation cannot be performed directly, but has to be done by 26114 copying the source into the intermediate register first, and then 26115 copying the intermediate register to the destination. An 26116 intermediate register always has the same mode as source and 26117 destination. Since it holds the actual value being copied, reload 26118 might apply optimizations to re-use an intermediate register and 26119 eliding the copy from the source when it can determine that the 26120 intermediate register still holds the required value. 26121 26122 Another kind of secondary reload is required on some machines which 26123 allow copying all registers to and from memory, but require a 26124 scratch register for stores to some memory locations (e.g., those 26125 with symbolic address on the RT, and those with certain symbolic 26126 address on the SPARC when compiling PIC). Scratch registers need 26127 not have the same mode as the value being copied, and usually hold 26128 a different value that that being copied. Special patterns in the 26129 md file are needed to describe how the copy is performed with the 26130 help of the scratch register; these patterns also describe the 26131 number, register class(es) and mode(s) of the scratch register(s). 26132 26133 In some cases, both an intermediate and a scratch register are 26134 required. 26135 26136 For input reloads, this target hook is called with nonzero IN_P, 26137 and X is an rtx that needs to be copied to a register of class 26138 RELOAD_CLASS in RELOAD_MODE. For output reloads, this target hook 26139 is called with zero IN_P, and a register of class RELOAD_CLASS 26140 needs to be copied to rtx X in RELOAD_MODE. 26141 26142 If copying a register of RELOAD_CLASS from/to X requires an 26143 intermediate register, the hook `secondary_reload' should return 26144 the register class required for this intermediate register. If no 26145 intermediate register is required, it should return NO_REGS. If 26146 more than one intermediate register is required, describe the one 26147 that is closest in the copy chain to the reload register. 26148 26149 If scratch registers are needed, you also have to describe how to 26150 perform the copy from/to the reload register to/from this closest 26151 intermediate register. Or if no intermediate register is 26152 required, but still a scratch register is needed, describe the 26153 copy from/to the reload register to/from the reload operand X. 26154 26155 You do this by setting `sri->icode' to the instruction code of a 26156 pattern in the md file which performs the move. Operands 0 and 1 26157 are the output and input of this copy, respectively. Operands 26158 from operand 2 onward are for scratch operands. These scratch 26159 operands must have a mode, and a single-register-class output 26160 constraint. 26161 26162 When an intermediate register is used, the `secondary_reload' hook 26163 will be called again to determine how to copy the intermediate 26164 register to/from the reload operand X, so your hook must also have 26165 code to handle the register class of the intermediate operand. 26166 26167 X might be a pseudo-register or a `subreg' of a pseudo-register, 26168 which could either be in a hard register or in memory. Use 26169 `true_regnum' to find out; it will return -1 if the pseudo is in 26170 memory and the hard register number if it is in a register. 26171 26172 Scratch operands in memory (constraint `"=m"' / `"=&m"') are 26173 currently not supported. For the time being, you will have to 26174 continue to use `SECONDARY_MEMORY_NEEDED' for that purpose. 26175 26176 `copy_cost' also uses this target hook to find out how values are 26177 copied. If you want it to include some extra cost for the need to 26178 allocate (a) scratch register(s), set `sri->extra_cost' to the 26179 additional cost. Or if two dependent moves are supposed to have a 26180 lower cost than the sum of the individual moves due to expected 26181 fortuitous scheduling and/or special forwarding logic, you can set 26182 `sri->extra_cost' to a negative amount. 26183 26184 -- Macro: SECONDARY_RELOAD_CLASS (CLASS, MODE, X) 26185 -- Macro: SECONDARY_INPUT_RELOAD_CLASS (CLASS, MODE, X) 26186 -- Macro: SECONDARY_OUTPUT_RELOAD_CLASS (CLASS, MODE, X) 26187 These macros are obsolete, new ports should use the target hook 26188 `TARGET_SECONDARY_RELOAD' instead. 26189 26190 These are obsolete macros, replaced by the 26191 `TARGET_SECONDARY_RELOAD' target hook. Older ports still define 26192 these macros to indicate to the reload phase that it may need to 26193 allocate at least one register for a reload in addition to the 26194 register to contain the data. Specifically, if copying X to a 26195 register CLASS in MODE requires an intermediate register, you were 26196 supposed to define `SECONDARY_INPUT_RELOAD_CLASS' to return the 26197 largest register class all of whose registers can be used as 26198 intermediate registers or scratch registers. 26199 26200 If copying a register CLASS in MODE to X requires an intermediate 26201 or scratch register, `SECONDARY_OUTPUT_RELOAD_CLASS' was supposed 26202 to be defined be defined to return the largest register class 26203 required. If the requirements for input and output reloads were 26204 the same, the macro `SECONDARY_RELOAD_CLASS' should have been used 26205 instead of defining both macros identically. 26206 26207 The values returned by these macros are often `GENERAL_REGS'. 26208 Return `NO_REGS' if no spare register is needed; i.e., if X can be 26209 directly copied to or from a register of CLASS in MODE without 26210 requiring a scratch register. Do not define this macro if it 26211 would always return `NO_REGS'. 26212 26213 If a scratch register is required (either with or without an 26214 intermediate register), you were supposed to define patterns for 26215 `reload_inM' or `reload_outM', as required (*note Standard 26216 Names::. These patterns, which were normally implemented with a 26217 `define_expand', should be similar to the `movM' patterns, except 26218 that operand 2 is the scratch register. 26219 26220 These patterns need constraints for the reload register and scratch 26221 register that contain a single register class. If the original 26222 reload register (whose class is CLASS) can meet the constraint 26223 given in the pattern, the value returned by these macros is used 26224 for the class of the scratch register. Otherwise, two additional 26225 reload registers are required. Their classes are obtained from 26226 the constraints in the insn pattern. 26227 26228 X might be a pseudo-register or a `subreg' of a pseudo-register, 26229 which could either be in a hard register or in memory. Use 26230 `true_regnum' to find out; it will return -1 if the pseudo is in 26231 memory and the hard register number if it is in a register. 26232 26233 These macros should not be used in the case where a particular 26234 class of registers can only be copied to memory and not to another 26235 class of registers. In that case, secondary reload registers are 26236 not needed and would not be helpful. Instead, a stack location 26237 must be used to perform the copy and the `movM' pattern should use 26238 memory as an intermediate storage. This case often occurs between 26239 floating-point and general registers. 26240 26241 -- Macro: SECONDARY_MEMORY_NEEDED (CLASS1, CLASS2, M) 26242 Certain machines have the property that some registers cannot be 26243 copied to some other registers without using memory. Define this 26244 macro on those machines to be a C expression that is nonzero if 26245 objects of mode M in registers of CLASS1 can only be copied to 26246 registers of class CLASS2 by storing a register of CLASS1 into 26247 memory and loading that memory location into a register of CLASS2. 26248 26249 Do not define this macro if its value would always be zero. 26250 26251 -- Macro: SECONDARY_MEMORY_NEEDED_RTX (MODE) 26252 Normally when `SECONDARY_MEMORY_NEEDED' is defined, the compiler 26253 allocates a stack slot for a memory location needed for register 26254 copies. If this macro is defined, the compiler instead uses the 26255 memory location defined by this macro. 26256 26257 Do not define this macro if you do not define 26258 `SECONDARY_MEMORY_NEEDED'. 26259 26260 -- Macro: SECONDARY_MEMORY_NEEDED_MODE (MODE) 26261 When the compiler needs a secondary memory location to copy 26262 between two registers of mode MODE, it normally allocates 26263 sufficient memory to hold a quantity of `BITS_PER_WORD' bits and 26264 performs the store and load operations in a mode that many bits 26265 wide and whose class is the same as that of MODE. 26266 26267 This is right thing to do on most machines because it ensures that 26268 all bits of the register are copied and prevents accesses to the 26269 registers in a narrower mode, which some machines prohibit for 26270 floating-point registers. 26271 26272 However, this default behavior is not correct on some machines, 26273 such as the DEC Alpha, that store short integers in floating-point 26274 registers differently than in integer registers. On those 26275 machines, the default widening will not work correctly and you 26276 must define this macro to suppress that widening in some cases. 26277 See the file `alpha.h' for details. 26278 26279 Do not define this macro if you do not define 26280 `SECONDARY_MEMORY_NEEDED' or if widening MODE to a mode that is 26281 `BITS_PER_WORD' bits wide is correct for your machine. 26282 26283 -- Macro: SMALL_REGISTER_CLASSES 26284 On some machines, it is risky to let hard registers live across 26285 arbitrary insns. Typically, these machines have instructions that 26286 require values to be in specific registers (like an accumulator), 26287 and reload will fail if the required hard register is used for 26288 another purpose across such an insn. 26289 26290 Define `SMALL_REGISTER_CLASSES' to be an expression with a nonzero 26291 value on these machines. When this macro has a nonzero value, the 26292 compiler will try to minimize the lifetime of hard registers. 26293 26294 It is always safe to define this macro with a nonzero value, but 26295 if you unnecessarily define it, you will reduce the amount of 26296 optimizations that can be performed in some cases. If you do not 26297 define this macro with a nonzero value when it is required, the 26298 compiler will run out of spill registers and print a fatal error 26299 message. For most machines, you should not define this macro at 26300 all. 26301 26302 -- Macro: CLASS_LIKELY_SPILLED_P (CLASS) 26303 A C expression whose value is nonzero if pseudos that have been 26304 assigned to registers of class CLASS would likely be spilled 26305 because registers of CLASS are needed for spill registers. 26306 26307 The default value of this macro returns 1 if CLASS has exactly one 26308 register and zero otherwise. On most machines, this default 26309 should be used. Only define this macro to some other expression 26310 if pseudos allocated by `local-alloc.c' end up in memory because 26311 their hard registers were needed for spill registers. If this 26312 macro returns nonzero for those classes, those pseudos will only 26313 be allocated by `global.c', which knows how to reallocate the 26314 pseudo to another register. If there would not be another 26315 register available for reallocation, you should not change the 26316 definition of this macro since the only effect of such a 26317 definition would be to slow down register allocation. 26318 26319 -- Macro: CLASS_MAX_NREGS (CLASS, MODE) 26320 A C expression for the maximum number of consecutive registers of 26321 class CLASS needed to hold a value of mode MODE. 26322 26323 This is closely related to the macro `HARD_REGNO_NREGS'. In fact, 26324 the value of the macro `CLASS_MAX_NREGS (CLASS, MODE)' should be 26325 the maximum value of `HARD_REGNO_NREGS (REGNO, MODE)' for all 26326 REGNO values in the class CLASS. 26327 26328 This macro helps control the handling of multiple-word values in 26329 the reload pass. 26330 26331 -- Macro: CANNOT_CHANGE_MODE_CLASS (FROM, TO, CLASS) 26332 If defined, a C expression that returns nonzero for a CLASS for 26333 which a change from mode FROM to mode TO is invalid. 26334 26335 For the example, loading 32-bit integer or floating-point objects 26336 into floating-point registers on the Alpha extends them to 64 bits. 26337 Therefore loading a 64-bit object and then storing it as a 32-bit 26338 object does not store the low-order 32 bits, as would be the case 26339 for a normal register. Therefore, `alpha.h' defines 26340 `CANNOT_CHANGE_MODE_CLASS' as below: 26341 26342 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \ 26343 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \ 26344 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0) 26345 26346 -- Target Hook: const enum reg_class * TARGET_IRA_COVER_CLASSES () 26347 Return an array of cover classes for the Integrated Register 26348 Allocator (IRA). Cover classes are a set of non-intersecting 26349 register classes covering all hard registers used for register 26350 allocation purposes. If a move between two registers in the same 26351 cover class is possible, it should be cheaper than a load or store 26352 of the registers. The array is terminated by a `LIM_REG_CLASSES' 26353 element. 26354 26355 This hook is called once at compiler startup, after the 26356 command-line options have been processed. It is then re-examined 26357 by every call to `target_reinit'. 26358 26359 The default implementation returns `IRA_COVER_CLASSES', if defined, 26360 otherwise there is no default implementation. You must define 26361 either this macro or `IRA_COVER_CLASSES' in order to use the 26362 integrated register allocator with Chaitin-Briggs coloring. If the 26363 macro is not defined, the only available coloring algorithm is 26364 Chow's priority coloring. 26365 26366 -- Macro: IRA_COVER_CLASSES 26367 See the documentation for `TARGET_IRA_COVER_CLASSES'. 26368 26369 26370 File: gccint.info, Node: Old Constraints, Next: Stack and Calling, Prev: Register Classes, Up: Target Macros 26371 26372 17.9 Obsolete Macros for Defining Constraints 26373 ============================================= 26374 26375 Machine-specific constraints can be defined with these macros instead 26376 of the machine description constructs described in *note Define 26377 Constraints::. This mechanism is obsolete. New ports should not use 26378 it; old ports should convert to the new mechanism. 26379 26380 -- Macro: CONSTRAINT_LEN (CHAR, STR) 26381 For the constraint at the start of STR, which starts with the 26382 letter C, return the length. This allows you to have register 26383 class / constant / extra constraints that are longer than a single 26384 letter; you don't need to define this macro if you can do with 26385 single-letter constraints only. The definition of this macro 26386 should use DEFAULT_CONSTRAINT_LEN for all the characters that you 26387 don't want to handle specially. There are some sanity checks in 26388 genoutput.c that check the constraint lengths for the md file, so 26389 you can also use this macro to help you while you are 26390 transitioning from a byzantine single-letter-constraint scheme: 26391 when you return a negative length for a constraint you want to 26392 re-use, genoutput will complain about every instance where it is 26393 used in the md file. 26394 26395 -- Macro: REG_CLASS_FROM_LETTER (CHAR) 26396 A C expression which defines the machine-dependent operand 26397 constraint letters for register classes. If CHAR is such a 26398 letter, the value should be the register class corresponding to 26399 it. Otherwise, the value should be `NO_REGS'. The register 26400 letter `r', corresponding to class `GENERAL_REGS', will not be 26401 passed to this macro; you do not need to handle it. 26402 26403 -- Macro: REG_CLASS_FROM_CONSTRAINT (CHAR, STR) 26404 Like `REG_CLASS_FROM_LETTER', but you also get the constraint 26405 string passed in STR, so that you can use suffixes to distinguish 26406 between different variants. 26407 26408 -- Macro: CONST_OK_FOR_LETTER_P (VALUE, C) 26409 A C expression that defines the machine-dependent operand 26410 constraint letters (`I', `J', `K', ... `P') that specify 26411 particular ranges of integer values. If C is one of those 26412 letters, the expression should check that VALUE, an integer, is in 26413 the appropriate range and return 1 if so, 0 otherwise. If C is 26414 not one of those letters, the value should be 0 regardless of 26415 VALUE. 26416 26417 -- Macro: CONST_OK_FOR_CONSTRAINT_P (VALUE, C, STR) 26418 Like `CONST_OK_FOR_LETTER_P', but you also get the constraint 26419 string passed in STR, so that you can use suffixes to distinguish 26420 between different variants. 26421 26422 -- Macro: CONST_DOUBLE_OK_FOR_LETTER_P (VALUE, C) 26423 A C expression that defines the machine-dependent operand 26424 constraint letters that specify particular ranges of 26425 `const_double' values (`G' or `H'). 26426 26427 If C is one of those letters, the expression should check that 26428 VALUE, an RTX of code `const_double', is in the appropriate range 26429 and return 1 if so, 0 otherwise. If C is not one of those 26430 letters, the value should be 0 regardless of VALUE. 26431 26432 `const_double' is used for all floating-point constants and for 26433 `DImode' fixed-point constants. A given letter can accept either 26434 or both kinds of values. It can use `GET_MODE' to distinguish 26435 between these kinds. 26436 26437 -- Macro: CONST_DOUBLE_OK_FOR_CONSTRAINT_P (VALUE, C, STR) 26438 Like `CONST_DOUBLE_OK_FOR_LETTER_P', but you also get the 26439 constraint string passed in STR, so that you can use suffixes to 26440 distinguish between different variants. 26441 26442 -- Macro: EXTRA_CONSTRAINT (VALUE, C) 26443 A C expression that defines the optional machine-dependent 26444 constraint letters that can be used to segregate specific types of 26445 operands, usually memory references, for the target machine. Any 26446 letter that is not elsewhere defined and not matched by 26447 `REG_CLASS_FROM_LETTER' / `REG_CLASS_FROM_CONSTRAINT' may be used. 26448 Normally this macro will not be defined. 26449 26450 If it is required for a particular target machine, it should 26451 return 1 if VALUE corresponds to the operand type represented by 26452 the constraint letter C. If C is not defined as an extra 26453 constraint, the value returned should be 0 regardless of VALUE. 26454 26455 For example, on the ROMP, load instructions cannot have their 26456 output in r0 if the memory reference contains a symbolic address. 26457 Constraint letter `Q' is defined as representing a memory address 26458 that does _not_ contain a symbolic address. An alternative is 26459 specified with a `Q' constraint on the input and `r' on the 26460 output. The next alternative specifies `m' on the input and a 26461 register class that does not include r0 on the output. 26462 26463 -- Macro: EXTRA_CONSTRAINT_STR (VALUE, C, STR) 26464 Like `EXTRA_CONSTRAINT', but you also get the constraint string 26465 passed in STR, so that you can use suffixes to distinguish between 26466 different variants. 26467 26468 -- Macro: EXTRA_MEMORY_CONSTRAINT (C, STR) 26469 A C expression that defines the optional machine-dependent 26470 constraint letters, amongst those accepted by `EXTRA_CONSTRAINT', 26471 that should be treated like memory constraints by the reload pass. 26472 26473 It should return 1 if the operand type represented by the 26474 constraint at the start of STR, the first letter of which is the 26475 letter C, comprises a subset of all memory references including 26476 all those whose address is simply a base register. This allows 26477 the reload pass to reload an operand, if it does not directly 26478 correspond to the operand type of C, by copying its address into a 26479 base register. 26480 26481 For example, on the S/390, some instructions do not accept 26482 arbitrary memory references, but only those that do not make use 26483 of an index register. The constraint letter `Q' is defined via 26484 `EXTRA_CONSTRAINT' as representing a memory address of this type. 26485 If the letter `Q' is marked as `EXTRA_MEMORY_CONSTRAINT', a `Q' 26486 constraint can handle any memory operand, because the reload pass 26487 knows it can be reloaded by copying the memory address into a base 26488 register if required. This is analogous to the way a `o' 26489 constraint can handle any memory operand. 26490 26491 -- Macro: EXTRA_ADDRESS_CONSTRAINT (C, STR) 26492 A C expression that defines the optional machine-dependent 26493 constraint letters, amongst those accepted by `EXTRA_CONSTRAINT' / 26494 `EXTRA_CONSTRAINT_STR', that should be treated like address 26495 constraints by the reload pass. 26496 26497 It should return 1 if the operand type represented by the 26498 constraint at the start of STR, which starts with the letter C, 26499 comprises a subset of all memory addresses including all those 26500 that consist of just a base register. This allows the reload pass 26501 to reload an operand, if it does not directly correspond to the 26502 operand type of STR, by copying it into a base register. 26503 26504 Any constraint marked as `EXTRA_ADDRESS_CONSTRAINT' can only be 26505 used with the `address_operand' predicate. It is treated 26506 analogously to the `p' constraint. 26507 26508 26509 File: gccint.info, Node: Stack and Calling, Next: Varargs, Prev: Old Constraints, Up: Target Macros 26510 26511 17.10 Stack Layout and Calling Conventions 26512 ========================================== 26513 26514 This describes the stack layout and calling conventions. 26515 26516 * Menu: 26517 26518 * Frame Layout:: 26519 * Exception Handling:: 26520 * Stack Checking:: 26521 * Frame Registers:: 26522 * Elimination:: 26523 * Stack Arguments:: 26524 * Register Arguments:: 26525 * Scalar Return:: 26526 * Aggregate Return:: 26527 * Caller Saves:: 26528 * Function Entry:: 26529 * Profiling:: 26530 * Tail Calls:: 26531 * Stack Smashing Protection:: 26532 26533 26534 File: gccint.info, Node: Frame Layout, Next: Exception Handling, Up: Stack and Calling 26535 26536 17.10.1 Basic Stack Layout 26537 -------------------------- 26538 26539 Here is the basic stack layout. 26540 26541 -- Macro: STACK_GROWS_DOWNWARD 26542 Define this macro if pushing a word onto the stack moves the stack 26543 pointer to a smaller address. 26544 26545 When we say, "define this macro if ...", it means that the 26546 compiler checks this macro only with `#ifdef' so the precise 26547 definition used does not matter. 26548 26549 -- Macro: STACK_PUSH_CODE 26550 This macro defines the operation used when something is pushed on 26551 the stack. In RTL, a push operation will be `(set (mem 26552 (STACK_PUSH_CODE (reg sp))) ...)' 26553 26554 The choices are `PRE_DEC', `POST_DEC', `PRE_INC', and `POST_INC'. 26555 Which of these is correct depends on the stack direction and on 26556 whether the stack pointer points to the last item on the stack or 26557 whether it points to the space for the next item on the stack. 26558 26559 The default is `PRE_DEC' when `STACK_GROWS_DOWNWARD' is defined, 26560 which is almost always right, and `PRE_INC' otherwise, which is 26561 often wrong. 26562 26563 -- Macro: FRAME_GROWS_DOWNWARD 26564 Define this macro to nonzero value if the addresses of local 26565 variable slots are at negative offsets from the frame pointer. 26566 26567 -- Macro: ARGS_GROW_DOWNWARD 26568 Define this macro if successive arguments to a function occupy 26569 decreasing addresses on the stack. 26570 26571 -- Macro: STARTING_FRAME_OFFSET 26572 Offset from the frame pointer to the first local variable slot to 26573 be allocated. 26574 26575 If `FRAME_GROWS_DOWNWARD', find the next slot's offset by 26576 subtracting the first slot's length from `STARTING_FRAME_OFFSET'. 26577 Otherwise, it is found by adding the length of the first slot to 26578 the value `STARTING_FRAME_OFFSET'. 26579 26580 -- Macro: STACK_ALIGNMENT_NEEDED 26581 Define to zero to disable final alignment of the stack during 26582 reload. The nonzero default for this macro is suitable for most 26583 ports. 26584 26585 On ports where `STARTING_FRAME_OFFSET' is nonzero or where there 26586 is a register save block following the local block that doesn't 26587 require alignment to `STACK_BOUNDARY', it may be beneficial to 26588 disable stack alignment and do it in the backend. 26589 26590 -- Macro: STACK_POINTER_OFFSET 26591 Offset from the stack pointer register to the first location at 26592 which outgoing arguments are placed. If not specified, the 26593 default value of zero is used. This is the proper value for most 26594 machines. 26595 26596 If `ARGS_GROW_DOWNWARD', this is the offset to the location above 26597 the first location at which outgoing arguments are placed. 26598 26599 -- Macro: FIRST_PARM_OFFSET (FUNDECL) 26600 Offset from the argument pointer register to the first argument's 26601 address. On some machines it may depend on the data type of the 26602 function. 26603 26604 If `ARGS_GROW_DOWNWARD', this is the offset to the location above 26605 the first argument's address. 26606 26607 -- Macro: STACK_DYNAMIC_OFFSET (FUNDECL) 26608 Offset from the stack pointer register to an item dynamically 26609 allocated on the stack, e.g., by `alloca'. 26610 26611 The default value for this macro is `STACK_POINTER_OFFSET' plus the 26612 length of the outgoing arguments. The default is correct for most 26613 machines. See `function.c' for details. 26614 26615 -- Macro: INITIAL_FRAME_ADDRESS_RTX 26616 A C expression whose value is RTL representing the address of the 26617 initial stack frame. This address is passed to `RETURN_ADDR_RTX' 26618 and `DYNAMIC_CHAIN_ADDRESS'. If you don't define this macro, a 26619 reasonable default value will be used. Define this macro in order 26620 to make frame pointer elimination work in the presence of 26621 `__builtin_frame_address (count)' and `__builtin_return_address 26622 (count)' for `count' not equal to zero. 26623 26624 -- Macro: DYNAMIC_CHAIN_ADDRESS (FRAMEADDR) 26625 A C expression whose value is RTL representing the address in a 26626 stack frame where the pointer to the caller's frame is stored. 26627 Assume that FRAMEADDR is an RTL expression for the address of the 26628 stack frame itself. 26629 26630 If you don't define this macro, the default is to return the value 26631 of FRAMEADDR--that is, the stack frame address is also the address 26632 of the stack word that points to the previous frame. 26633 26634 -- Macro: SETUP_FRAME_ADDRESSES 26635 If defined, a C expression that produces the machine-specific code 26636 to setup the stack so that arbitrary frames can be accessed. For 26637 example, on the SPARC, we must flush all of the register windows 26638 to the stack before we can access arbitrary stack frames. You 26639 will seldom need to define this macro. 26640 26641 -- Target Hook: bool TARGET_BUILTIN_SETJMP_FRAME_VALUE () 26642 This target hook should return an rtx that is used to store the 26643 address of the current frame into the built in `setjmp' buffer. 26644 The default value, `virtual_stack_vars_rtx', is correct for most 26645 machines. One reason you may need to define this target hook is if 26646 `hard_frame_pointer_rtx' is the appropriate value on your machine. 26647 26648 -- Macro: FRAME_ADDR_RTX (FRAMEADDR) 26649 A C expression whose value is RTL representing the value of the 26650 frame address for the current frame. FRAMEADDR is the frame 26651 pointer of the current frame. This is used for 26652 __builtin_frame_address. You need only define this macro if the 26653 frame address is not the same as the frame pointer. Most machines 26654 do not need to define it. 26655 26656 -- Macro: RETURN_ADDR_RTX (COUNT, FRAMEADDR) 26657 A C expression whose value is RTL representing the value of the 26658 return address for the frame COUNT steps up from the current 26659 frame, after the prologue. FRAMEADDR is the frame pointer of the 26660 COUNT frame, or the frame pointer of the COUNT - 1 frame if 26661 `RETURN_ADDR_IN_PREVIOUS_FRAME' is defined. 26662 26663 The value of the expression must always be the correct address when 26664 COUNT is zero, but may be `NULL_RTX' if there is no way to 26665 determine the return address of other frames. 26666 26667 -- Macro: RETURN_ADDR_IN_PREVIOUS_FRAME 26668 Define this if the return address of a particular stack frame is 26669 accessed from the frame pointer of the previous stack frame. 26670 26671 -- Macro: INCOMING_RETURN_ADDR_RTX 26672 A C expression whose value is RTL representing the location of the 26673 incoming return address at the beginning of any function, before 26674 the prologue. This RTL is either a `REG', indicating that the 26675 return value is saved in `REG', or a `MEM' representing a location 26676 in the stack. 26677 26678 You only need to define this macro if you want to support call 26679 frame debugging information like that provided by DWARF 2. 26680 26681 If this RTL is a `REG', you should also define 26682 `DWARF_FRAME_RETURN_COLUMN' to `DWARF_FRAME_REGNUM (REGNO)'. 26683 26684 -- Macro: DWARF_ALT_FRAME_RETURN_COLUMN 26685 A C expression whose value is an integer giving a DWARF 2 column 26686 number that may be used as an alternative return column. The 26687 column must not correspond to any gcc hard register (that is, it 26688 must not be in the range of `DWARF_FRAME_REGNUM'). 26689 26690 This macro can be useful if `DWARF_FRAME_RETURN_COLUMN' is set to a 26691 general register, but an alternative column needs to be used for 26692 signal frames. Some targets have also used different frame return 26693 columns over time. 26694 26695 -- Macro: DWARF_ZERO_REG 26696 A C expression whose value is an integer giving a DWARF 2 register 26697 number that is considered to always have the value zero. This 26698 should only be defined if the target has an architected zero 26699 register, and someone decided it was a good idea to use that 26700 register number to terminate the stack backtrace. New ports 26701 should avoid this. 26702 26703 -- Target Hook: void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char 26704 *LABEL, rtx PATTERN, int INDEX) 26705 This target hook allows the backend to emit frame-related insns 26706 that contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame 26707 debugging info engine will invoke it on insns of the form 26708 (set (reg) (unspec [...] UNSPEC_INDEX)) 26709 and 26710 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)). 26711 to let the backend emit the call frame instructions. LABEL is the 26712 CFI label attached to the insn, PATTERN is the pattern of the insn 26713 and INDEX is `UNSPEC_INDEX' or `UNSPECV_INDEX'. 26714 26715 -- Macro: INCOMING_FRAME_SP_OFFSET 26716 A C expression whose value is an integer giving the offset, in 26717 bytes, from the value of the stack pointer register to the top of 26718 the stack frame at the beginning of any function, before the 26719 prologue. The top of the frame is defined to be the value of the 26720 stack pointer in the previous frame, just before the call 26721 instruction. 26722 26723 You only need to define this macro if you want to support call 26724 frame debugging information like that provided by DWARF 2. 26725 26726 -- Macro: ARG_POINTER_CFA_OFFSET (FUNDECL) 26727 A C expression whose value is an integer giving the offset, in 26728 bytes, from the argument pointer to the canonical frame address 26729 (cfa). The final value should coincide with that calculated by 26730 `INCOMING_FRAME_SP_OFFSET'. Which is unfortunately not usable 26731 during virtual register instantiation. 26732 26733 The default value for this macro is `FIRST_PARM_OFFSET (fundecl)', 26734 which is correct for most machines; in general, the arguments are 26735 found immediately before the stack frame. Note that this is not 26736 the case on some targets that save registers into the caller's 26737 frame, such as SPARC and rs6000, and so such targets need to 26738 define this macro. 26739 26740 You only need to define this macro if the default is incorrect, 26741 and you want to support call frame debugging information like that 26742 provided by DWARF 2. 26743 26744 -- Macro: FRAME_POINTER_CFA_OFFSET (FUNDECL) 26745 If defined, a C expression whose value is an integer giving the 26746 offset in bytes from the frame pointer to the canonical frame 26747 address (cfa). The final value should coincide with that 26748 calculated by `INCOMING_FRAME_SP_OFFSET'. 26749 26750 Normally the CFA is calculated as an offset from the argument 26751 pointer, via `ARG_POINTER_CFA_OFFSET', but if the argument pointer 26752 is variable due to the ABI, this may not be possible. If this 26753 macro is defined, it implies that the virtual register 26754 instantiation should be based on the frame pointer instead of the 26755 argument pointer. Only one of `FRAME_POINTER_CFA_OFFSET' and 26756 `ARG_POINTER_CFA_OFFSET' should be defined. 26757 26758 -- Macro: CFA_FRAME_BASE_OFFSET (FUNDECL) 26759 If defined, a C expression whose value is an integer giving the 26760 offset in bytes from the canonical frame address (cfa) to the 26761 frame base used in DWARF 2 debug information. The default is 26762 zero. A different value may reduce the size of debug information 26763 on some ports. 26764 26765 26766 File: gccint.info, Node: Exception Handling, Next: Stack Checking, Prev: Frame Layout, Up: Stack and Calling 26767 26768 17.10.2 Exception Handling Support 26769 ---------------------------------- 26770 26771 -- Macro: EH_RETURN_DATA_REGNO (N) 26772 A C expression whose value is the Nth register number used for 26773 data by exception handlers, or `INVALID_REGNUM' if fewer than N 26774 registers are usable. 26775 26776 The exception handling library routines communicate with the 26777 exception handlers via a set of agreed upon registers. Ideally 26778 these registers should be call-clobbered; it is possible to use 26779 call-saved registers, but may negatively impact code size. The 26780 target must support at least 2 data registers, but should define 4 26781 if there are enough free registers. 26782 26783 You must define this macro if you want to support call frame 26784 exception handling like that provided by DWARF 2. 26785 26786 -- Macro: EH_RETURN_STACKADJ_RTX 26787 A C expression whose value is RTL representing a location in which 26788 to store a stack adjustment to be applied before function return. 26789 This is used to unwind the stack to an exception handler's call 26790 frame. It will be assigned zero on code paths that return 26791 normally. 26792 26793 Typically this is a call-clobbered hard register that is otherwise 26794 untouched by the epilogue, but could also be a stack slot. 26795 26796 Do not define this macro if the stack pointer is saved and restored 26797 by the regular prolog and epilog code in the call frame itself; in 26798 this case, the exception handling library routines will update the 26799 stack location to be restored in place. Otherwise, you must define 26800 this macro if you want to support call frame exception handling 26801 like that provided by DWARF 2. 26802 26803 -- Macro: EH_RETURN_HANDLER_RTX 26804 A C expression whose value is RTL representing a location in which 26805 to store the address of an exception handler to which we should 26806 return. It will not be assigned on code paths that return 26807 normally. 26808 26809 Typically this is the location in the call frame at which the 26810 normal return address is stored. For targets that return by 26811 popping an address off the stack, this might be a memory address 26812 just below the _target_ call frame rather than inside the current 26813 call frame. If defined, `EH_RETURN_STACKADJ_RTX' will have already 26814 been assigned, so it may be used to calculate the location of the 26815 target call frame. 26816 26817 Some targets have more complex requirements than storing to an 26818 address calculable during initial code generation. In that case 26819 the `eh_return' instruction pattern should be used instead. 26820 26821 If you want to support call frame exception handling, you must 26822 define either this macro or the `eh_return' instruction pattern. 26823 26824 -- Macro: RETURN_ADDR_OFFSET 26825 If defined, an integer-valued C expression for which rtl will be 26826 generated to add it to the exception handler address before it is 26827 searched in the exception handling tables, and to subtract it 26828 again from the address before using it to return to the exception 26829 handler. 26830 26831 -- Macro: ASM_PREFERRED_EH_DATA_FORMAT (CODE, GLOBAL) 26832 This macro chooses the encoding of pointers embedded in the 26833 exception handling sections. If at all possible, this should be 26834 defined such that the exception handling section will not require 26835 dynamic relocations, and so may be read-only. 26836 26837 CODE is 0 for data, 1 for code labels, 2 for function pointers. 26838 GLOBAL is true if the symbol may be affected by dynamic 26839 relocations. The macro should return a combination of the 26840 `DW_EH_PE_*' defines as found in `dwarf2.h'. 26841 26842 If this macro is not defined, pointers will not be encoded but 26843 represented directly. 26844 26845 -- Macro: ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (FILE, ENCODING, SIZE, 26846 ADDR, DONE) 26847 This macro allows the target to emit whatever special magic is 26848 required to represent the encoding chosen by 26849 `ASM_PREFERRED_EH_DATA_FORMAT'. Generic code takes care of 26850 pc-relative and indirect encodings; this must be defined if the 26851 target uses text-relative or data-relative encodings. 26852 26853 This is a C statement that branches to DONE if the format was 26854 handled. ENCODING is the format chosen, SIZE is the number of 26855 bytes that the format occupies, ADDR is the `SYMBOL_REF' to be 26856 emitted. 26857 26858 -- Macro: MD_UNWIND_SUPPORT 26859 A string specifying a file to be #include'd in unwind-dw2.c. The 26860 file so included typically defines `MD_FALLBACK_FRAME_STATE_FOR'. 26861 26862 -- Macro: MD_FALLBACK_FRAME_STATE_FOR (CONTEXT, FS) 26863 This macro allows the target to add CPU and operating system 26864 specific code to the call-frame unwinder for use when there is no 26865 unwind data available. The most common reason to implement this 26866 macro is to unwind through signal frames. 26867 26868 This macro is called from `uw_frame_state_for' in `unwind-dw2.c', 26869 `unwind-dw2-xtensa.c' and `unwind-ia64.c'. CONTEXT is an 26870 `_Unwind_Context'; FS is an `_Unwind_FrameState'. Examine 26871 `context->ra' for the address of the code being executed and 26872 `context->cfa' for the stack pointer value. If the frame can be 26873 decoded, the register save addresses should be updated in FS and 26874 the macro should evaluate to `_URC_NO_REASON'. If the frame 26875 cannot be decoded, the macro should evaluate to 26876 `_URC_END_OF_STACK'. 26877 26878 For proper signal handling in Java this macro is accompanied by 26879 `MAKE_THROW_FRAME', defined in `libjava/include/*-signal.h' 26880 headers. 26881 26882 -- Macro: MD_HANDLE_UNWABI (CONTEXT, FS) 26883 This macro allows the target to add operating system specific code 26884 to the call-frame unwinder to handle the IA-64 `.unwabi' unwinding 26885 directive, usually used for signal or interrupt frames. 26886 26887 This macro is called from `uw_update_context' in `unwind-ia64.c'. 26888 CONTEXT is an `_Unwind_Context'; FS is an `_Unwind_FrameState'. 26889 Examine `fs->unwabi' for the abi and context in the `.unwabi' 26890 directive. If the `.unwabi' directive can be handled, the 26891 register save addresses should be updated in FS. 26892 26893 -- Macro: TARGET_USES_WEAK_UNWIND_INFO 26894 A C expression that evaluates to true if the target requires unwind 26895 info to be given comdat linkage. Define it to be `1' if comdat 26896 linkage is necessary. The default is `0'. 26897 26898 26899 File: gccint.info, Node: Stack Checking, Next: Frame Registers, Prev: Exception Handling, Up: Stack and Calling 26900 26901 17.10.3 Specifying How Stack Checking is Done 26902 --------------------------------------------- 26903 26904 GCC will check that stack references are within the boundaries of the 26905 stack, if the option `-fstack-check' is specified, in one of three ways: 26906 26907 1. If the value of the `STACK_CHECK_BUILTIN' macro is nonzero, GCC 26908 will assume that you have arranged for full stack checking to be 26909 done at appropriate places in the configuration files. GCC will 26910 not do other special processing. 26911 26912 2. If `STACK_CHECK_BUILTIN' is zero and the value of the 26913 `STACK_CHECK_STATIC_BUILTIN' macro is nonzero, GCC will assume 26914 that you have arranged for static stack checking (checking of the 26915 static stack frame of functions) to be done at appropriate places 26916 in the configuration files. GCC will only emit code to do dynamic 26917 stack checking (checking on dynamic stack allocations) using the 26918 third approach below. 26919 26920 3. If neither of the above are true, GCC will generate code to 26921 periodically "probe" the stack pointer using the values of the 26922 macros defined below. 26923 26924 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is 26925 defined, GCC will change its allocation strategy for large objects if 26926 the option `-fstack-check' is specified: they will always be allocated 26927 dynamically if their size exceeds `STACK_CHECK_MAX_VAR_SIZE' bytes. 26928 26929 -- Macro: STACK_CHECK_BUILTIN 26930 A nonzero value if stack checking is done by the configuration 26931 files in a machine-dependent manner. You should define this macro 26932 if stack checking is require by the ABI of your machine or if you 26933 would like to do stack checking in some more efficient way than 26934 the generic approach. The default value of this macro is zero. 26935 26936 -- Macro: STACK_CHECK_STATIC_BUILTIN 26937 A nonzero value if static stack checking is done by the 26938 configuration files in a machine-dependent manner. You should 26939 define this macro if you would like to do static stack checking in 26940 some more efficient way than the generic approach. The default 26941 value of this macro is zero. 26942 26943 -- Macro: STACK_CHECK_PROBE_INTERVAL 26944 An integer representing the interval at which GCC must generate 26945 stack probe instructions. You will normally define this macro to 26946 be no larger than the size of the "guard pages" at the end of a 26947 stack area. The default value of 4096 is suitable for most 26948 systems. 26949 26950 -- Macro: STACK_CHECK_PROBE_LOAD 26951 An integer which is nonzero if GCC should perform the stack probe 26952 as a load instruction and zero if GCC should use a store 26953 instruction. The default is zero, which is the most efficient 26954 choice on most systems. 26955 26956 -- Macro: STACK_CHECK_PROTECT 26957 The number of bytes of stack needed to recover from a stack 26958 overflow, for languages where such a recovery is supported. The 26959 default value of 75 words should be adequate for most machines. 26960 26961 The following macros are relevant only if neither STACK_CHECK_BUILTIN 26962 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether 26963 in the opposite case. 26964 26965 -- Macro: STACK_CHECK_MAX_FRAME_SIZE 26966 The maximum size of a stack frame, in bytes. GCC will generate 26967 probe instructions in non-leaf functions to ensure at least this 26968 many bytes of stack are available. If a stack frame is larger 26969 than this size, stack checking will not be reliable and GCC will 26970 issue a warning. The default is chosen so that GCC only generates 26971 one instruction on most systems. You should normally not change 26972 the default value of this macro. 26973 26974 -- Macro: STACK_CHECK_FIXED_FRAME_SIZE 26975 GCC uses this value to generate the above warning message. It 26976 represents the amount of fixed frame used by a function, not 26977 including space for any callee-saved registers, temporaries and 26978 user variables. You need only specify an upper bound for this 26979 amount and will normally use the default of four words. 26980 26981 -- Macro: STACK_CHECK_MAX_VAR_SIZE 26982 The maximum size, in bytes, of an object that GCC will place in the 26983 fixed area of the stack frame when the user specifies 26984 `-fstack-check'. GCC computed the default from the values of the 26985 above macros and you will normally not need to override that 26986 default. 26987 26988 26989 File: gccint.info, Node: Frame Registers, Next: Elimination, Prev: Stack Checking, Up: Stack and Calling 26990 26991 17.10.4 Registers That Address the Stack Frame 26992 ---------------------------------------------- 26993 26994 This discusses registers that address the stack frame. 26995 26996 -- Macro: STACK_POINTER_REGNUM 26997 The register number of the stack pointer register, which must also 26998 be a fixed register according to `FIXED_REGISTERS'. On most 26999 machines, the hardware determines which register this is. 27000 27001 -- Macro: FRAME_POINTER_REGNUM 27002 The register number of the frame pointer register, which is used to 27003 access automatic variables in the stack frame. On some machines, 27004 the hardware determines which register this is. On other 27005 machines, you can choose any register you wish for this purpose. 27006 27007 -- Macro: HARD_FRAME_POINTER_REGNUM 27008 On some machines the offset between the frame pointer and starting 27009 offset of the automatic variables is not known until after register 27010 allocation has been done (for example, because the saved registers 27011 are between these two locations). On those machines, define 27012 `FRAME_POINTER_REGNUM' the number of a special, fixed register to 27013 be used internally until the offset is known, and define 27014 `HARD_FRAME_POINTER_REGNUM' to be the actual hard register number 27015 used for the frame pointer. 27016 27017 You should define this macro only in the very rare circumstances 27018 when it is not possible to calculate the offset between the frame 27019 pointer and the automatic variables until after register 27020 allocation has been completed. When this macro is defined, you 27021 must also indicate in your definition of `ELIMINABLE_REGS' how to 27022 eliminate `FRAME_POINTER_REGNUM' into either 27023 `HARD_FRAME_POINTER_REGNUM' or `STACK_POINTER_REGNUM'. 27024 27025 Do not define this macro if it would be the same as 27026 `FRAME_POINTER_REGNUM'. 27027 27028 -- Macro: ARG_POINTER_REGNUM 27029 The register number of the arg pointer register, which is used to 27030 access the function's argument list. On some machines, this is 27031 the same as the frame pointer register. On some machines, the 27032 hardware determines which register this is. On other machines, 27033 you can choose any register you wish for this purpose. If this is 27034 not the same register as the frame pointer register, then you must 27035 mark it as a fixed register according to `FIXED_REGISTERS', or 27036 arrange to be able to eliminate it (*note Elimination::). 27037 27038 -- Macro: RETURN_ADDRESS_POINTER_REGNUM 27039 The register number of the return address pointer register, which 27040 is used to access the current function's return address from the 27041 stack. On some machines, the return address is not at a fixed 27042 offset from the frame pointer or stack pointer or argument 27043 pointer. This register can be defined to point to the return 27044 address on the stack, and then be converted by `ELIMINABLE_REGS' 27045 into either the frame pointer or stack pointer. 27046 27047 Do not define this macro unless there is no other way to get the 27048 return address from the stack. 27049 27050 -- Macro: STATIC_CHAIN_REGNUM 27051 -- Macro: STATIC_CHAIN_INCOMING_REGNUM 27052 Register numbers used for passing a function's static chain 27053 pointer. If register windows are used, the register number as 27054 seen by the called function is `STATIC_CHAIN_INCOMING_REGNUM', 27055 while the register number as seen by the calling function is 27056 `STATIC_CHAIN_REGNUM'. If these registers are the same, 27057 `STATIC_CHAIN_INCOMING_REGNUM' need not be defined. 27058 27059 The static chain register need not be a fixed register. 27060 27061 If the static chain is passed in memory, these macros should not be 27062 defined; instead, the next two macros should be defined. 27063 27064 -- Macro: STATIC_CHAIN 27065 -- Macro: STATIC_CHAIN_INCOMING 27066 If the static chain is passed in memory, these macros provide rtx 27067 giving `mem' expressions that denote where they are stored. 27068 `STATIC_CHAIN' and `STATIC_CHAIN_INCOMING' give the locations as 27069 seen by the calling and called functions, respectively. Often the 27070 former will be at an offset from the stack pointer and the latter 27071 at an offset from the frame pointer. 27072 27073 The variables `stack_pointer_rtx', `frame_pointer_rtx', and 27074 `arg_pointer_rtx' will have been initialized prior to the use of 27075 these macros and should be used to refer to those items. 27076 27077 If the static chain is passed in a register, the two previous 27078 macros should be defined instead. 27079 27080 -- Macro: DWARF_FRAME_REGISTERS 27081 This macro specifies the maximum number of hard registers that can 27082 be saved in a call frame. This is used to size data structures 27083 used in DWARF2 exception handling. 27084 27085 Prior to GCC 3.0, this macro was needed in order to establish a 27086 stable exception handling ABI in the face of adding new hard 27087 registers for ISA extensions. In GCC 3.0 and later, the EH ABI is 27088 insulated from changes in the number of hard registers. 27089 Nevertheless, this macro can still be used to reduce the runtime 27090 memory requirements of the exception handling routines, which can 27091 be substantial if the ISA contains a lot of registers that are not 27092 call-saved. 27093 27094 If this macro is not defined, it defaults to 27095 `FIRST_PSEUDO_REGISTER'. 27096 27097 -- Macro: PRE_GCC3_DWARF_FRAME_REGISTERS 27098 This macro is similar to `DWARF_FRAME_REGISTERS', but is provided 27099 for backward compatibility in pre GCC 3.0 compiled code. 27100 27101 If this macro is not defined, it defaults to 27102 `DWARF_FRAME_REGISTERS'. 27103 27104 -- Macro: DWARF_REG_TO_UNWIND_COLUMN (REGNO) 27105 Define this macro if the target's representation for dwarf 27106 registers is different than the internal representation for unwind 27107 column. Given a dwarf register, this macro should return the 27108 internal unwind column number to use instead. 27109 27110 See the PowerPC's SPE target for an example. 27111 27112 -- Macro: DWARF_FRAME_REGNUM (REGNO) 27113 Define this macro if the target's representation for dwarf 27114 registers used in .eh_frame or .debug_frame is different from that 27115 used in other debug info sections. Given a GCC hard register 27116 number, this macro should return the .eh_frame register number. 27117 The default is `DBX_REGISTER_NUMBER (REGNO)'. 27118 27119 27120 -- Macro: DWARF2_FRAME_REG_OUT (REGNO, FOR_EH) 27121 Define this macro to map register numbers held in the call frame 27122 info that GCC has collected using `DWARF_FRAME_REGNUM' to those 27123 that should be output in .debug_frame (`FOR_EH' is zero) and 27124 .eh_frame (`FOR_EH' is nonzero). The default is to return `REGNO'. 27125 27126 27127 27128 File: gccint.info, Node: Elimination, Next: Stack Arguments, Prev: Frame Registers, Up: Stack and Calling 27129 27130 17.10.5 Eliminating Frame Pointer and Arg Pointer 27131 ------------------------------------------------- 27132 27133 This is about eliminating the frame pointer and arg pointer. 27134 27135 -- Macro: FRAME_POINTER_REQUIRED 27136 A C expression which is nonzero if a function must have and use a 27137 frame pointer. This expression is evaluated in the reload pass. 27138 If its value is nonzero the function will have a frame pointer. 27139 27140 The expression can in principle examine the current function and 27141 decide according to the facts, but on most machines the constant 0 27142 or the constant 1 suffices. Use 0 when the machine allows code to 27143 be generated with no frame pointer, and doing so saves some time 27144 or space. Use 1 when there is no possible advantage to avoiding a 27145 frame pointer. 27146 27147 In certain cases, the compiler does not know how to produce valid 27148 code without a frame pointer. The compiler recognizes those cases 27149 and automatically gives the function a frame pointer regardless of 27150 what `FRAME_POINTER_REQUIRED' says. You don't need to worry about 27151 them. 27152 27153 In a function that does not require a frame pointer, the frame 27154 pointer register can be allocated for ordinary usage, unless you 27155 mark it as a fixed register. See `FIXED_REGISTERS' for more 27156 information. 27157 27158 -- Macro: INITIAL_FRAME_POINTER_OFFSET (DEPTH-VAR) 27159 A C statement to store in the variable DEPTH-VAR the difference 27160 between the frame pointer and the stack pointer values immediately 27161 after the function prologue. The value would be computed from 27162 information such as the result of `get_frame_size ()' and the 27163 tables of registers `regs_ever_live' and `call_used_regs'. 27164 27165 If `ELIMINABLE_REGS' is defined, this macro will be not be used and 27166 need not be defined. Otherwise, it must be defined even if 27167 `FRAME_POINTER_REQUIRED' is defined to always be true; in that 27168 case, you may set DEPTH-VAR to anything. 27169 27170 -- Macro: ELIMINABLE_REGS 27171 If defined, this macro specifies a table of register pairs used to 27172 eliminate unneeded registers that point into the stack frame. If 27173 it is not defined, the only elimination attempted by the compiler 27174 is to replace references to the frame pointer with references to 27175 the stack pointer. 27176 27177 The definition of this macro is a list of structure 27178 initializations, each of which specifies an original and 27179 replacement register. 27180 27181 On some machines, the position of the argument pointer is not 27182 known until the compilation is completed. In such a case, a 27183 separate hard register must be used for the argument pointer. 27184 This register can be eliminated by replacing it with either the 27185 frame pointer or the argument pointer, depending on whether or not 27186 the frame pointer has been eliminated. 27187 27188 In this case, you might specify: 27189 #define ELIMINABLE_REGS \ 27190 {{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM}, \ 27191 {ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM}, \ 27192 {FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM}} 27193 27194 Note that the elimination of the argument pointer with the stack 27195 pointer is specified first since that is the preferred elimination. 27196 27197 -- Macro: CAN_ELIMINATE (FROM-REG, TO-REG) 27198 A C expression that returns nonzero if the compiler is allowed to 27199 try to replace register number FROM-REG with register number 27200 TO-REG. This macro need only be defined if `ELIMINABLE_REGS' is 27201 defined, and will usually be the constant 1, since most of the 27202 cases preventing register elimination are things that the compiler 27203 already knows about. 27204 27205 -- Macro: INITIAL_ELIMINATION_OFFSET (FROM-REG, TO-REG, OFFSET-VAR) 27206 This macro is similar to `INITIAL_FRAME_POINTER_OFFSET'. It 27207 specifies the initial difference between the specified pair of 27208 registers. This macro must be defined if `ELIMINABLE_REGS' is 27209 defined. 27210 27211 27212 File: gccint.info, Node: Stack Arguments, Next: Register Arguments, Prev: Elimination, Up: Stack and Calling 27213 27214 17.10.6 Passing Function Arguments on the Stack 27215 ----------------------------------------------- 27216 27217 The macros in this section control how arguments are passed on the 27218 stack. See the following section for other macros that control passing 27219 certain arguments in registers. 27220 27221 -- Target Hook: bool TARGET_PROMOTE_PROTOTYPES (tree FNTYPE) 27222 This target hook returns `true' if an argument declared in a 27223 prototype as an integral type smaller than `int' should actually be 27224 passed as an `int'. In addition to avoiding errors in certain 27225 cases of mismatch, it also makes for better code on certain 27226 machines. The default is to not promote prototypes. 27227 27228 -- Macro: PUSH_ARGS 27229 A C expression. If nonzero, push insns will be used to pass 27230 outgoing arguments. If the target machine does not have a push 27231 instruction, set it to zero. That directs GCC to use an alternate 27232 strategy: to allocate the entire argument block and then store the 27233 arguments into it. When `PUSH_ARGS' is nonzero, `PUSH_ROUNDING' 27234 must be defined too. 27235 27236 -- Macro: PUSH_ARGS_REVERSED 27237 A C expression. If nonzero, function arguments will be evaluated 27238 from last to first, rather than from first to last. If this macro 27239 is not defined, it defaults to `PUSH_ARGS' on targets where the 27240 stack and args grow in opposite directions, and 0 otherwise. 27241 27242 -- Macro: PUSH_ROUNDING (NPUSHED) 27243 A C expression that is the number of bytes actually pushed onto the 27244 stack when an instruction attempts to push NPUSHED bytes. 27245 27246 On some machines, the definition 27247 27248 #define PUSH_ROUNDING(BYTES) (BYTES) 27249 27250 will suffice. But on other machines, instructions that appear to 27251 push one byte actually push two bytes in an attempt to maintain 27252 alignment. Then the definition should be 27253 27254 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) 27255 27256 -- Macro: ACCUMULATE_OUTGOING_ARGS 27257 A C expression. If nonzero, the maximum amount of space required 27258 for outgoing arguments will be computed and placed into the 27259 variable `current_function_outgoing_args_size'. No space will be 27260 pushed onto the stack for each call; instead, the function 27261 prologue should increase the stack frame size by this amount. 27262 27263 Setting both `PUSH_ARGS' and `ACCUMULATE_OUTGOING_ARGS' is not 27264 proper. 27265 27266 -- Macro: REG_PARM_STACK_SPACE (FNDECL) 27267 Define this macro if functions should assume that stack space has 27268 been allocated for arguments even when their values are passed in 27269 registers. 27270 27271 The value of this macro is the size, in bytes, of the area 27272 reserved for arguments passed in registers for the function 27273 represented by FNDECL, which can be zero if GCC is calling a 27274 library function. The argument FNDECL can be the FUNCTION_DECL, 27275 or the type itself of the function. 27276 27277 This space can be allocated by the caller, or be a part of the 27278 machine-dependent stack frame: `OUTGOING_REG_PARM_STACK_SPACE' says 27279 which. 27280 27281 -- Macro: OUTGOING_REG_PARM_STACK_SPACE (FNTYPE) 27282 Define this to a nonzero value if it is the responsibility of the 27283 caller to allocate the area reserved for arguments passed in 27284 registers when calling a function of FNTYPE. FNTYPE may be NULL 27285 if the function called is a library function. 27286 27287 If `ACCUMULATE_OUTGOING_ARGS' is defined, this macro controls 27288 whether the space for these arguments counts in the value of 27289 `current_function_outgoing_args_size'. 27290 27291 -- Macro: STACK_PARMS_IN_REG_PARM_AREA 27292 Define this macro if `REG_PARM_STACK_SPACE' is defined, but the 27293 stack parameters don't skip the area specified by it. 27294 27295 Normally, when a parameter is not passed in registers, it is 27296 placed on the stack beyond the `REG_PARM_STACK_SPACE' area. 27297 Defining this macro suppresses this behavior and causes the 27298 parameter to be passed on the stack in its natural location. 27299 27300 -- Macro: RETURN_POPS_ARGS (FUNDECL, FUNTYPE, STACK-SIZE) 27301 A C expression that should indicate the number of bytes of its own 27302 arguments that a function pops on returning, or 0 if the function 27303 pops no arguments and the caller must therefore pop them all after 27304 the function returns. 27305 27306 FUNDECL is a C variable whose value is a tree node that describes 27307 the function in question. Normally it is a node of type 27308 `FUNCTION_DECL' that describes the declaration of the function. 27309 From this you can obtain the `DECL_ATTRIBUTES' of the function. 27310 27311 FUNTYPE is a C variable whose value is a tree node that describes 27312 the function in question. Normally it is a node of type 27313 `FUNCTION_TYPE' that describes the data type of the function. 27314 From this it is possible to obtain the data types of the value and 27315 arguments (if known). 27316 27317 When a call to a library function is being considered, FUNDECL 27318 will contain an identifier node for the library function. Thus, if 27319 you need to distinguish among various library functions, you can 27320 do so by their names. Note that "library function" in this 27321 context means a function used to perform arithmetic, whose name is 27322 known specially in the compiler and was not mentioned in the C 27323 code being compiled. 27324 27325 STACK-SIZE is the number of bytes of arguments passed on the 27326 stack. If a variable number of bytes is passed, it is zero, and 27327 argument popping will always be the responsibility of the calling 27328 function. 27329 27330 On the VAX, all functions always pop their arguments, so the 27331 definition of this macro is STACK-SIZE. On the 68000, using the 27332 standard calling convention, no functions pop their arguments, so 27333 the value of the macro is always 0 in this case. But an 27334 alternative calling convention is available in which functions 27335 that take a fixed number of arguments pop them but other functions 27336 (such as `printf') pop nothing (the caller pops all). When this 27337 convention is in use, FUNTYPE is examined to determine whether a 27338 function takes a fixed number of arguments. 27339 27340 -- Macro: CALL_POPS_ARGS (CUM) 27341 A C expression that should indicate the number of bytes a call 27342 sequence pops off the stack. It is added to the value of 27343 `RETURN_POPS_ARGS' when compiling a function call. 27344 27345 CUM is the variable in which all arguments to the called function 27346 have been accumulated. 27347 27348 On certain architectures, such as the SH5, a call trampoline is 27349 used that pops certain registers off the stack, depending on the 27350 arguments that have been passed to the function. Since this is a 27351 property of the call site, not of the called function, 27352 `RETURN_POPS_ARGS' is not appropriate. 27353 27354 27355 File: gccint.info, Node: Register Arguments, Next: Scalar Return, Prev: Stack Arguments, Up: Stack and Calling 27356 27357 17.10.7 Passing Arguments in Registers 27358 -------------------------------------- 27359 27360 This section describes the macros which let you control how various 27361 types of arguments are passed in registers or how they are arranged in 27362 the stack. 27363 27364 -- Macro: FUNCTION_ARG (CUM, MODE, TYPE, NAMED) 27365 A C expression that controls whether a function argument is passed 27366 in a register, and which register. 27367 27368 The arguments are CUM, which summarizes all the previous 27369 arguments; MODE, the machine mode of the argument; TYPE, the data 27370 type of the argument as a tree node or 0 if that is not known 27371 (which happens for C support library functions); and NAMED, which 27372 is 1 for an ordinary argument and 0 for nameless arguments that 27373 correspond to `...' in the called function's prototype. TYPE can 27374 be an incomplete type if a syntax error has previously occurred. 27375 27376 The value of the expression is usually either a `reg' RTX for the 27377 hard register in which to pass the argument, or zero to pass the 27378 argument on the stack. 27379 27380 For machines like the VAX and 68000, where normally all arguments 27381 are pushed, zero suffices as a definition. 27382 27383 The value of the expression can also be a `parallel' RTX. This is 27384 used when an argument is passed in multiple locations. The mode 27385 of the `parallel' should be the mode of the entire argument. The 27386 `parallel' holds any number of `expr_list' pairs; each one 27387 describes where part of the argument is passed. In each 27388 `expr_list' the first operand must be a `reg' RTX for the hard 27389 register in which to pass this part of the argument, and the mode 27390 of the register RTX indicates how large this part of the argument 27391 is. The second operand of the `expr_list' is a `const_int' which 27392 gives the offset in bytes into the entire argument of where this 27393 part starts. As a special exception the first `expr_list' in the 27394 `parallel' RTX may have a first operand of zero. This indicates 27395 that the entire argument is also stored on the stack. 27396 27397 The last time this macro is called, it is called with `MODE == 27398 VOIDmode', and its result is passed to the `call' or `call_value' 27399 pattern as operands 2 and 3 respectively. 27400 27401 The usual way to make the ISO library `stdarg.h' work on a machine 27402 where some arguments are usually passed in registers, is to cause 27403 nameless arguments to be passed on the stack instead. This is done 27404 by making `FUNCTION_ARG' return 0 whenever NAMED is 0. 27405 27406 You may use the hook `targetm.calls.must_pass_in_stack' in the 27407 definition of this macro to determine if this argument is of a 27408 type that must be passed in the stack. If `REG_PARM_STACK_SPACE' 27409 is not defined and `FUNCTION_ARG' returns nonzero for such an 27410 argument, the compiler will abort. If `REG_PARM_STACK_SPACE' is 27411 defined, the argument will be computed in the stack and then 27412 loaded into a register. 27413 27414 -- Target Hook: bool TARGET_MUST_PASS_IN_STACK (enum machine_mode 27415 MODE, tree TYPE) 27416 This target hook should return `true' if we should not pass TYPE 27417 solely in registers. The file `expr.h' defines a definition that 27418 is usually appropriate, refer to `expr.h' for additional 27419 documentation. 27420 27421 -- Macro: FUNCTION_INCOMING_ARG (CUM, MODE, TYPE, NAMED) 27422 Define this macro if the target machine has "register windows", so 27423 that the register in which a function sees an arguments is not 27424 necessarily the same as the one in which the caller passed the 27425 argument. 27426 27427 For such machines, `FUNCTION_ARG' computes the register in which 27428 the caller passes the value, and `FUNCTION_INCOMING_ARG' should be 27429 defined in a similar fashion to tell the function being called 27430 where the arguments will arrive. 27431 27432 If `FUNCTION_INCOMING_ARG' is not defined, `FUNCTION_ARG' serves 27433 both purposes. 27434 27435 -- Target Hook: int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *CUM, 27436 enum machine_mode MODE, tree TYPE, bool NAMED) 27437 This target hook returns the number of bytes at the beginning of an 27438 argument that must be put in registers. The value must be zero for 27439 arguments that are passed entirely in registers or that are 27440 entirely pushed on the stack. 27441 27442 On some machines, certain arguments must be passed partially in 27443 registers and partially in memory. On these machines, typically 27444 the first few words of arguments are passed in registers, and the 27445 rest on the stack. If a multi-word argument (a `double' or a 27446 structure) crosses that boundary, its first few words must be 27447 passed in registers and the rest must be pushed. This macro tells 27448 the compiler when this occurs, and how many bytes should go in 27449 registers. 27450 27451 `FUNCTION_ARG' for these arguments should return the first 27452 register to be used by the caller for this argument; likewise 27453 `FUNCTION_INCOMING_ARG', for the called function. 27454 27455 -- Target Hook: bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *CUM, 27456 enum machine_mode MODE, tree TYPE, bool NAMED) 27457 This target hook should return `true' if an argument at the 27458 position indicated by CUM should be passed by reference. This 27459 predicate is queried after target independent reasons for being 27460 passed by reference, such as `TREE_ADDRESSABLE (type)'. 27461 27462 If the hook returns true, a copy of that argument is made in 27463 memory and a pointer to the argument is passed instead of the 27464 argument itself. The pointer is passed in whatever way is 27465 appropriate for passing a pointer to that type. 27466 27467 -- Target Hook: bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *CUM, enum 27468 machine_mode MODE, tree TYPE, bool NAMED) 27469 The function argument described by the parameters to this hook is 27470 known to be passed by reference. The hook should return true if 27471 the function argument should be copied by the callee instead of 27472 copied by the caller. 27473 27474 For any argument for which the hook returns true, if it can be 27475 determined that the argument is not modified, then a copy need not 27476 be generated. 27477 27478 The default version of this hook always returns false. 27479 27480 -- Macro: CUMULATIVE_ARGS 27481 A C type for declaring a variable that is used as the first 27482 argument of `FUNCTION_ARG' and other related values. For some 27483 target machines, the type `int' suffices and can hold the number 27484 of bytes of argument so far. 27485 27486 There is no need to record in `CUMULATIVE_ARGS' anything about the 27487 arguments that have been passed on the stack. The compiler has 27488 other variables to keep track of that. For target machines on 27489 which all arguments are passed on the stack, there is no need to 27490 store anything in `CUMULATIVE_ARGS'; however, the data structure 27491 must exist and should not be empty, so use `int'. 27492 27493 -- Macro: OVERRIDE_ABI_FORMAT (FNDECL) 27494 If defined, this macro is called before generating any code for a 27495 function, but after the CFUN descriptor for the function has been 27496 created. The back end may use this macro to update CFUN to 27497 reflect an ABI other than that which would normally be used by 27498 default. If the compiler is generating code for a 27499 compiler-generated function, FNDECL may be `NULL'. 27500 27501 -- Macro: INIT_CUMULATIVE_ARGS (CUM, FNTYPE, LIBNAME, FNDECL, 27502 N_NAMED_ARGS) 27503 A C statement (sans semicolon) for initializing the variable CUM 27504 for the state at the beginning of the argument list. The variable 27505 has type `CUMULATIVE_ARGS'. The value of FNTYPE is the tree node 27506 for the data type of the function which will receive the args, or 27507 0 if the args are to a compiler support library function. For 27508 direct calls that are not libcalls, FNDECL contain the declaration 27509 node of the function. FNDECL is also set when 27510 `INIT_CUMULATIVE_ARGS' is used to find arguments for the function 27511 being compiled. N_NAMED_ARGS is set to the number of named 27512 arguments, including a structure return address if it is passed as 27513 a parameter, when making a call. When processing incoming 27514 arguments, N_NAMED_ARGS is set to -1. 27515 27516 When processing a call to a compiler support library function, 27517 LIBNAME identifies which one. It is a `symbol_ref' rtx which 27518 contains the name of the function, as a string. LIBNAME is 0 when 27519 an ordinary C function call is being processed. Thus, each time 27520 this macro is called, either LIBNAME or FNTYPE is nonzero, but 27521 never both of them at once. 27522 27523 -- Macro: INIT_CUMULATIVE_LIBCALL_ARGS (CUM, MODE, LIBNAME) 27524 Like `INIT_CUMULATIVE_ARGS' but only used for outgoing libcalls, 27525 it gets a `MODE' argument instead of FNTYPE, that would be `NULL'. 27526 INDIRECT would always be zero, too. If this macro is not defined, 27527 `INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, 0)' is used instead. 27528 27529 -- Macro: INIT_CUMULATIVE_INCOMING_ARGS (CUM, FNTYPE, LIBNAME) 27530 Like `INIT_CUMULATIVE_ARGS' but overrides it for the purposes of 27531 finding the arguments for the function being compiled. If this 27532 macro is undefined, `INIT_CUMULATIVE_ARGS' is used instead. 27533 27534 The value passed for LIBNAME is always 0, since library routines 27535 with special calling conventions are never compiled with GCC. The 27536 argument LIBNAME exists for symmetry with `INIT_CUMULATIVE_ARGS'. 27537 27538 -- Macro: FUNCTION_ARG_ADVANCE (CUM, MODE, TYPE, NAMED) 27539 A C statement (sans semicolon) to update the summarizer variable 27540 CUM to advance past an argument in the argument list. The values 27541 MODE, TYPE and NAMED describe that argument. Once this is done, 27542 the variable CUM is suitable for analyzing the _following_ 27543 argument with `FUNCTION_ARG', etc. 27544 27545 This macro need not do anything if the argument in question was 27546 passed on the stack. The compiler knows how to track the amount 27547 of stack space used for arguments without any special help. 27548 27549 -- Macro: FUNCTION_ARG_OFFSET (MODE, TYPE) 27550 If defined, a C expression that is the number of bytes to add to 27551 the offset of the argument passed in memory. This is needed for 27552 the SPU, which passes `char' and `short' arguments in the preferred 27553 slot that is in the middle of the quad word instead of starting at 27554 the top. 27555 27556 -- Macro: FUNCTION_ARG_PADDING (MODE, TYPE) 27557 If defined, a C expression which determines whether, and in which 27558 direction, to pad out an argument with extra space. The value 27559 should be of type `enum direction': either `upward' to pad above 27560 the argument, `downward' to pad below, or `none' to inhibit 27561 padding. 27562 27563 The _amount_ of padding is always just enough to reach the next 27564 multiple of `FUNCTION_ARG_BOUNDARY'; this macro does not control 27565 it. 27566 27567 This macro has a default definition which is right for most 27568 systems. For little-endian machines, the default is to pad 27569 upward. For big-endian machines, the default is to pad downward 27570 for an argument of constant size shorter than an `int', and upward 27571 otherwise. 27572 27573 -- Macro: PAD_VARARGS_DOWN 27574 If defined, a C expression which determines whether the default 27575 implementation of va_arg will attempt to pad down before reading 27576 the next argument, if that argument is smaller than its aligned 27577 space as controlled by `PARM_BOUNDARY'. If this macro is not 27578 defined, all such arguments are padded down if `BYTES_BIG_ENDIAN' 27579 is true. 27580 27581 -- Macro: BLOCK_REG_PADDING (MODE, TYPE, FIRST) 27582 Specify padding for the last element of a block move between 27583 registers and memory. FIRST is nonzero if this is the only 27584 element. Defining this macro allows better control of register 27585 function parameters on big-endian machines, without using 27586 `PARALLEL' rtl. In particular, `MUST_PASS_IN_STACK' need not test 27587 padding and mode of types in registers, as there is no longer a 27588 "wrong" part of a register; For example, a three byte aggregate 27589 may be passed in the high part of a register if so required. 27590 27591 -- Macro: FUNCTION_ARG_BOUNDARY (MODE, TYPE) 27592 If defined, a C expression that gives the alignment boundary, in 27593 bits, of an argument with the specified mode and type. If it is 27594 not defined, `PARM_BOUNDARY' is used for all arguments. 27595 27596 -- Macro: FUNCTION_ARG_REGNO_P (REGNO) 27597 A C expression that is nonzero if REGNO is the number of a hard 27598 register in which function arguments are sometimes passed. This 27599 does _not_ include implicit arguments such as the static chain and 27600 the structure-value address. On many machines, no registers can be 27601 used for this purpose since all function arguments are pushed on 27602 the stack. 27603 27604 -- Target Hook: bool TARGET_SPLIT_COMPLEX_ARG (tree TYPE) 27605 This hook should return true if parameter of type TYPE are passed 27606 as two scalar parameters. By default, GCC will attempt to pack 27607 complex arguments into the target's word size. Some ABIs require 27608 complex arguments to be split and treated as their individual 27609 components. For example, on AIX64, complex floats should be 27610 passed in a pair of floating point registers, even though a 27611 complex float would fit in one 64-bit floating point register. 27612 27613 The default value of this hook is `NULL', which is treated as 27614 always false. 27615 27616 -- Target Hook: tree TARGET_BUILD_BUILTIN_VA_LIST (void) 27617 This hook returns a type node for `va_list' for the target. The 27618 default version of the hook returns `void*'. 27619 27620 -- Target Hook: tree TARGET_FN_ABI_VA_LIST (tree FNDECL) 27621 This hook returns the va_list type of the calling convention 27622 specified by FNDECL. The default version of this hook returns 27623 `va_list_type_node'. 27624 27625 -- Target Hook: tree TARGET_CANONICAL_VA_LIST_TYPE (tree TYPE) 27626 This hook returns the va_list type of the calling convention 27627 specified by the type of TYPE. If TYPE is not a valid va_list 27628 type, it returns `NULL_TREE'. 27629 27630 -- Target Hook: tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree VALIST, tree 27631 TYPE, tree *PRE_P, tree *POST_P) 27632 This hook performs target-specific gimplification of 27633 `VA_ARG_EXPR'. The first two parameters correspond to the 27634 arguments to `va_arg'; the latter two are as in 27635 `gimplify.c:gimplify_expr'. 27636 27637 -- Target Hook: bool TARGET_VALID_POINTER_MODE (enum machine_mode MODE) 27638 Define this to return nonzero if the port can handle pointers with 27639 machine mode MODE. The default version of this hook returns true 27640 for both `ptr_mode' and `Pmode'. 27641 27642 -- Target Hook: bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode 27643 MODE) 27644 Define this to return nonzero if the port is prepared to handle 27645 insns involving scalar mode MODE. For a scalar mode to be 27646 considered supported, all the basic arithmetic and comparisons 27647 must work. 27648 27649 The default version of this hook returns true for any mode 27650 required to handle the basic C types (as defined by the port). 27651 Included here are the double-word arithmetic supported by the code 27652 in `optabs.c'. 27653 27654 -- Target Hook: bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode 27655 MODE) 27656 Define this to return nonzero if the port is prepared to handle 27657 insns involving vector mode MODE. At the very least, it must have 27658 move patterns for this mode. 27659 27660 27661 File: gccint.info, Node: Scalar Return, Next: Aggregate Return, Prev: Register Arguments, Up: Stack and Calling 27662 27663 17.10.8 How Scalar Function Values Are Returned 27664 ----------------------------------------------- 27665 27666 This section discusses the macros that control returning scalars as 27667 values--values that can fit in registers. 27668 27669 -- Target Hook: rtx TARGET_FUNCTION_VALUE (tree RET_TYPE, tree 27670 FN_DECL_OR_TYPE, bool OUTGOING) 27671 Define this to return an RTX representing the place where a 27672 function returns or receives a value of data type RET_TYPE, a tree 27673 node node representing a data type. FN_DECL_OR_TYPE is a tree node 27674 representing `FUNCTION_DECL' or `FUNCTION_TYPE' of a function 27675 being called. If OUTGOING is false, the hook should compute the 27676 register in which the caller will see the return value. 27677 Otherwise, the hook should return an RTX representing the place 27678 where a function returns a value. 27679 27680 On many machines, only `TYPE_MODE (RET_TYPE)' is relevant. 27681 (Actually, on most machines, scalar values are returned in the same 27682 place regardless of mode.) The value of the expression is usually 27683 a `reg' RTX for the hard register where the return value is stored. 27684 The value can also be a `parallel' RTX, if the return value is in 27685 multiple places. See `FUNCTION_ARG' for an explanation of the 27686 `parallel' form. Note that the callee will populate every 27687 location specified in the `parallel', but if the first element of 27688 the `parallel' contains the whole return value, callers will use 27689 that element as the canonical location and ignore the others. The 27690 m68k port uses this type of `parallel' to return pointers in both 27691 `%a0' (the canonical location) and `%d0'. 27692 27693 If `TARGET_PROMOTE_FUNCTION_RETURN' returns true, you must apply 27694 the same promotion rules specified in `PROMOTE_MODE' if VALTYPE is 27695 a scalar type. 27696 27697 If the precise function being called is known, FUNC is a tree node 27698 (`FUNCTION_DECL') for it; otherwise, FUNC is a null pointer. This 27699 makes it possible to use a different value-returning convention 27700 for specific functions when all their calls are known. 27701 27702 Some target machines have "register windows" so that the register 27703 in which a function returns its value is not the same as the one 27704 in which the caller sees the value. For such machines, you should 27705 return different RTX depending on OUTGOING. 27706 27707 `TARGET_FUNCTION_VALUE' is not used for return values with 27708 aggregate data types, because these are returned in another way. 27709 See `TARGET_STRUCT_VALUE_RTX' and related macros, below. 27710 27711 -- Macro: FUNCTION_VALUE (VALTYPE, FUNC) 27712 This macro has been deprecated. Use `TARGET_FUNCTION_VALUE' for a 27713 new target instead. 27714 27715 -- Macro: FUNCTION_OUTGOING_VALUE (VALTYPE, FUNC) 27716 This macro has been deprecated. Use `TARGET_FUNCTION_VALUE' for a 27717 new target instead. 27718 27719 -- Macro: LIBCALL_VALUE (MODE) 27720 A C expression to create an RTX representing the place where a 27721 library function returns a value of mode MODE. 27722 27723 Note that "library function" in this context means a compiler 27724 support routine, used to perform arithmetic, whose name is known 27725 specially by the compiler and was not mentioned in the C code being 27726 compiled. 27727 27728 -- Target Hook: rtx TARGET_LIBCALL_VALUE (enum machine_mode 27729 MODE, rtx FUN) Define this hook if the back-end needs to know the 27730 name of the libcall function in order to determine where the 27731 result should be returned. 27732 27733 The mode of the result is given by MODE and the name of the called 27734 library function is given by FUN. The hook should return an RTX 27735 representing the place where the library function result will be 27736 returned. 27737 27738 If this hook is not defined, then LIBCALL_VALUE will be used. 27739 27740 -- Macro: FUNCTION_VALUE_REGNO_P (REGNO) 27741 A C expression that is nonzero if REGNO is the number of a hard 27742 register in which the values of called function may come back. 27743 27744 A register whose use for returning values is limited to serving as 27745 the second of a pair (for a value of type `double', say) need not 27746 be recognized by this macro. So for most machines, this definition 27747 suffices: 27748 27749 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) 27750 27751 If the machine has register windows, so that the caller and the 27752 called function use different registers for the return value, this 27753 macro should recognize only the caller's register numbers. 27754 27755 -- Macro: TARGET_ENUM_VA_LIST (IDX, PNAME, PTYPE) 27756 This target macro is used in function `c_common_nodes_and_builtins' 27757 to iterate through the target specific builtin types for va_list. 27758 The variable IDX is used as iterator. PNAME has to be a pointer to 27759 a `const char *' and PTYPE a pointer to a `tree' typed variable. 27760 The arguments PNAME and PTYPE are used to store the result of this 27761 macro and are set to the name of the va_list builtin type and its 27762 internal type. If the return value of this macro is zero, then 27763 there is no more element. Otherwise the IDX should be increased 27764 for the next call of this macro to iterate through all types. 27765 27766 -- Macro: APPLY_RESULT_SIZE 27767 Define this macro if `untyped_call' and `untyped_return' need more 27768 space than is implied by `FUNCTION_VALUE_REGNO_P' for saving and 27769 restoring an arbitrary return value. 27770 27771 -- Target Hook: bool TARGET_RETURN_IN_MSB (tree TYPE) 27772 This hook should return true if values of type TYPE are returned 27773 at the most significant end of a register (in other words, if they 27774 are padded at the least significant end). You can assume that TYPE 27775 is returned in a register; the caller is required to check this. 27776 27777 Note that the register provided by `TARGET_FUNCTION_VALUE' must be 27778 able to hold the complete return value. For example, if a 1-, 2- 27779 or 3-byte structure is returned at the most significant end of a 27780 4-byte register, `TARGET_FUNCTION_VALUE' should provide an 27781 `SImode' rtx. 27782 27783 27784 File: gccint.info, Node: Aggregate Return, Next: Caller Saves, Prev: Scalar Return, Up: Stack and Calling 27785 27786 17.10.9 How Large Values Are Returned 27787 ------------------------------------- 27788 27789 When a function value's mode is `BLKmode' (and in some other cases), 27790 the value is not returned according to `TARGET_FUNCTION_VALUE' (*note 27791 Scalar Return::). Instead, the caller passes the address of a block of 27792 memory in which the value should be stored. This address is called the 27793 "structure value address". 27794 27795 This section describes how to control returning structure values in 27796 memory. 27797 27798 -- Target Hook: bool TARGET_RETURN_IN_MEMORY (tree TYPE, tree FNTYPE) 27799 This target hook should return a nonzero value to say to return the 27800 function value in memory, just as large structures are always 27801 returned. Here TYPE will be the data type of the value, and FNTYPE 27802 will be the type of the function doing the returning, or `NULL' for 27803 libcalls. 27804 27805 Note that values of mode `BLKmode' must be explicitly handled by 27806 this function. Also, the option `-fpcc-struct-return' takes 27807 effect regardless of this macro. On most systems, it is possible 27808 to leave the hook undefined; this causes a default definition to 27809 be used, whose value is the constant 1 for `BLKmode' values, and 0 27810 otherwise. 27811 27812 Do not use this hook to indicate that structures and unions should 27813 always be returned in memory. You should instead use 27814 `DEFAULT_PCC_STRUCT_RETURN' to indicate this. 27815 27816 -- Macro: DEFAULT_PCC_STRUCT_RETURN 27817 Define this macro to be 1 if all structure and union return values 27818 must be in memory. Since this results in slower code, this should 27819 be defined only if needed for compatibility with other compilers 27820 or with an ABI. If you define this macro to be 0, then the 27821 conventions used for structure and union return values are decided 27822 by the `TARGET_RETURN_IN_MEMORY' target hook. 27823 27824 If not defined, this defaults to the value 1. 27825 27826 -- Target Hook: rtx TARGET_STRUCT_VALUE_RTX (tree FNDECL, int INCOMING) 27827 This target hook should return the location of the structure value 27828 address (normally a `mem' or `reg'), or 0 if the address is passed 27829 as an "invisible" first argument. Note that FNDECL may be `NULL', 27830 for libcalls. You do not need to define this target hook if the 27831 address is always passed as an "invisible" first argument. 27832 27833 On some architectures the place where the structure value address 27834 is found by the called function is not the same place that the 27835 caller put it. This can be due to register windows, or it could 27836 be because the function prologue moves it to a different place. 27837 INCOMING is `1' or `2' when the location is needed in the context 27838 of the called function, and `0' in the context of the caller. 27839 27840 If INCOMING is nonzero and the address is to be found on the 27841 stack, return a `mem' which refers to the frame pointer. If 27842 INCOMING is `2', the result is being used to fetch the structure 27843 value address at the beginning of a function. If you need to emit 27844 adjusting code, you should do it at this point. 27845 27846 -- Macro: PCC_STATIC_STRUCT_RETURN 27847 Define this macro if the usual system convention on the target 27848 machine for returning structures and unions is for the called 27849 function to return the address of a static variable containing the 27850 value. 27851 27852 Do not define this if the usual system convention is for the 27853 caller to pass an address to the subroutine. 27854 27855 This macro has effect in `-fpcc-struct-return' mode, but it does 27856 nothing when you use `-freg-struct-return' mode. 27857 27858 27859 File: gccint.info, Node: Caller Saves, Next: Function Entry, Prev: Aggregate Return, Up: Stack and Calling 27860 27861 17.10.10 Caller-Saves Register Allocation 27862 ----------------------------------------- 27863 27864 If you enable it, GCC can save registers around function calls. This 27865 makes it possible to use call-clobbered registers to hold variables that 27866 must live across calls. 27867 27868 -- Macro: CALLER_SAVE_PROFITABLE (REFS, CALLS) 27869 A C expression to determine whether it is worthwhile to consider 27870 placing a pseudo-register in a call-clobbered hard register and 27871 saving and restoring it around each function call. The expression 27872 should be 1 when this is worth doing, and 0 otherwise. 27873 27874 If you don't define this macro, a default is used which is good on 27875 most machines: `4 * CALLS < REFS'. 27876 27877 -- Macro: HARD_REGNO_CALLER_SAVE_MODE (REGNO, NREGS) 27878 A C expression specifying which mode is required for saving NREGS 27879 of a pseudo-register in call-clobbered hard register REGNO. If 27880 REGNO is unsuitable for caller save, `VOIDmode' should be 27881 returned. For most machines this macro need not be defined since 27882 GCC will select the smallest suitable mode. 27883 27884 27885 File: gccint.info, Node: Function Entry, Next: Profiling, Prev: Caller Saves, Up: Stack and Calling 27886 27887 17.10.11 Function Entry and Exit 27888 -------------------------------- 27889 27890 This section describes the macros that output function entry 27891 ("prologue") and exit ("epilogue") code. 27892 27893 -- Target Hook: void TARGET_ASM_FUNCTION_PROLOGUE (FILE *FILE, 27894 HOST_WIDE_INT SIZE) 27895 If defined, a function that outputs the assembler code for entry 27896 to a function. The prologue is responsible for setting up the 27897 stack frame, initializing the frame pointer register, saving 27898 registers that must be saved, and allocating SIZE additional bytes 27899 of storage for the local variables. SIZE is an integer. FILE is 27900 a stdio stream to which the assembler code should be output. 27901 27902 The label for the beginning of the function need not be output by 27903 this macro. That has already been done when the macro is run. 27904 27905 To determine which registers to save, the macro can refer to the 27906 array `regs_ever_live': element R is nonzero if hard register R is 27907 used anywhere within the function. This implies the function 27908 prologue should save register R, provided it is not one of the 27909 call-used registers. (`TARGET_ASM_FUNCTION_EPILOGUE' must 27910 likewise use `regs_ever_live'.) 27911 27912 On machines that have "register windows", the function entry code 27913 does not save on the stack the registers that are in the windows, 27914 even if they are supposed to be preserved by function calls; 27915 instead it takes appropriate steps to "push" the register stack, 27916 if any non-call-used registers are used in the function. 27917 27918 On machines where functions may or may not have frame-pointers, the 27919 function entry code must vary accordingly; it must set up the frame 27920 pointer if one is wanted, and not otherwise. To determine whether 27921 a frame pointer is in wanted, the macro can refer to the variable 27922 `frame_pointer_needed'. The variable's value will be 1 at run 27923 time in a function that needs a frame pointer. *Note 27924 Elimination::. 27925 27926 The function entry code is responsible for allocating any stack 27927 space required for the function. This stack space consists of the 27928 regions listed below. In most cases, these regions are allocated 27929 in the order listed, with the last listed region closest to the 27930 top of the stack (the lowest address if `STACK_GROWS_DOWNWARD' is 27931 defined, and the highest address if it is not defined). You can 27932 use a different order for a machine if doing so is more convenient 27933 or required for compatibility reasons. Except in cases where 27934 required by standard or by a debugger, there is no reason why the 27935 stack layout used by GCC need agree with that used by other 27936 compilers for a machine. 27937 27938 -- Target Hook: void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *FILE) 27939 If defined, a function that outputs assembler code at the end of a 27940 prologue. This should be used when the function prologue is being 27941 emitted as RTL, and you have some extra assembler that needs to be 27942 emitted. *Note prologue instruction pattern::. 27943 27944 -- Target Hook: void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *FILE) 27945 If defined, a function that outputs assembler code at the start of 27946 an epilogue. This should be used when the function epilogue is 27947 being emitted as RTL, and you have some extra assembler that needs 27948 to be emitted. *Note epilogue instruction pattern::. 27949 27950 -- Target Hook: void TARGET_ASM_FUNCTION_EPILOGUE (FILE *FILE, 27951 HOST_WIDE_INT SIZE) 27952 If defined, a function that outputs the assembler code for exit 27953 from a function. The epilogue is responsible for restoring the 27954 saved registers and stack pointer to their values when the 27955 function was called, and returning control to the caller. This 27956 macro takes the same arguments as the macro 27957 `TARGET_ASM_FUNCTION_PROLOGUE', and the registers to restore are 27958 determined from `regs_ever_live' and `CALL_USED_REGISTERS' in the 27959 same way. 27960 27961 On some machines, there is a single instruction that does all the 27962 work of returning from the function. On these machines, give that 27963 instruction the name `return' and do not define the macro 27964 `TARGET_ASM_FUNCTION_EPILOGUE' at all. 27965 27966 Do not define a pattern named `return' if you want the 27967 `TARGET_ASM_FUNCTION_EPILOGUE' to be used. If you want the target 27968 switches to control whether return instructions or epilogues are 27969 used, define a `return' pattern with a validity condition that 27970 tests the target switches appropriately. If the `return' 27971 pattern's validity condition is false, epilogues will be used. 27972 27973 On machines where functions may or may not have frame-pointers, the 27974 function exit code must vary accordingly. Sometimes the code for 27975 these two cases is completely different. To determine whether a 27976 frame pointer is wanted, the macro can refer to the variable 27977 `frame_pointer_needed'. The variable's value will be 1 when 27978 compiling a function that needs a frame pointer. 27979 27980 Normally, `TARGET_ASM_FUNCTION_PROLOGUE' and 27981 `TARGET_ASM_FUNCTION_EPILOGUE' must treat leaf functions specially. 27982 The C variable `current_function_is_leaf' is nonzero for such a 27983 function. *Note Leaf Functions::. 27984 27985 On some machines, some functions pop their arguments on exit while 27986 others leave that for the caller to do. For example, the 68020 27987 when given `-mrtd' pops arguments in functions that take a fixed 27988 number of arguments. 27989 27990 Your definition of the macro `RETURN_POPS_ARGS' decides which 27991 functions pop their own arguments. `TARGET_ASM_FUNCTION_EPILOGUE' 27992 needs to know what was decided. The variable that is called 27993 `current_function_pops_args' is the number of bytes of its 27994 arguments that a function should pop. *Note Scalar Return::. 27995 27996 * A region of `current_function_pretend_args_size' bytes of 27997 uninitialized space just underneath the first argument arriving on 27998 the stack. (This may not be at the very start of the allocated 27999 stack region if the calling sequence has pushed anything else 28000 since pushing the stack arguments. But usually, on such machines, 28001 nothing else has been pushed yet, because the function prologue 28002 itself does all the pushing.) This region is used on machines 28003 where an argument may be passed partly in registers and partly in 28004 memory, and, in some cases to support the features in `<stdarg.h>'. 28005 28006 * An area of memory used to save certain registers used by the 28007 function. The size of this area, which may also include space for 28008 such things as the return address and pointers to previous stack 28009 frames, is machine-specific and usually depends on which registers 28010 have been used in the function. Machines with register windows 28011 often do not require a save area. 28012 28013 * A region of at least SIZE bytes, possibly rounded up to an 28014 allocation boundary, to contain the local variables of the 28015 function. On some machines, this region and the save area may 28016 occur in the opposite order, with the save area closer to the top 28017 of the stack. 28018 28019 * Optionally, when `ACCUMULATE_OUTGOING_ARGS' is defined, a region of 28020 `current_function_outgoing_args_size' bytes to be used for outgoing 28021 argument lists of the function. *Note Stack Arguments::. 28022 28023 -- Macro: EXIT_IGNORE_STACK 28024 Define this macro as a C expression that is nonzero if the return 28025 instruction or the function epilogue ignores the value of the stack 28026 pointer; in other words, if it is safe to delete an instruction to 28027 adjust the stack pointer before a return from the function. The 28028 default is 0. 28029 28030 Note that this macro's value is relevant only for functions for 28031 which frame pointers are maintained. It is never safe to delete a 28032 final stack adjustment in a function that has no frame pointer, 28033 and the compiler knows this regardless of `EXIT_IGNORE_STACK'. 28034 28035 -- Macro: EPILOGUE_USES (REGNO) 28036 Define this macro as a C expression that is nonzero for registers 28037 that are used by the epilogue or the `return' pattern. The stack 28038 and frame pointer registers are already assumed to be used as 28039 needed. 28040 28041 -- Macro: EH_USES (REGNO) 28042 Define this macro as a C expression that is nonzero for registers 28043 that are used by the exception handling mechanism, and so should 28044 be considered live on entry to an exception edge. 28045 28046 -- Macro: DELAY_SLOTS_FOR_EPILOGUE 28047 Define this macro if the function epilogue contains delay slots to 28048 which instructions from the rest of the function can be "moved". 28049 The definition should be a C expression whose value is an integer 28050 representing the number of delay slots there. 28051 28052 -- Macro: ELIGIBLE_FOR_EPILOGUE_DELAY (INSN, N) 28053 A C expression that returns 1 if INSN can be placed in delay slot 28054 number N of the epilogue. 28055 28056 The argument N is an integer which identifies the delay slot now 28057 being considered (since different slots may have different rules of 28058 eligibility). It is never negative and is always less than the 28059 number of epilogue delay slots (what `DELAY_SLOTS_FOR_EPILOGUE' 28060 returns). If you reject a particular insn for a given delay slot, 28061 in principle, it may be reconsidered for a subsequent delay slot. 28062 Also, other insns may (at least in principle) be considered for 28063 the so far unfilled delay slot. 28064 28065 The insns accepted to fill the epilogue delay slots are put in an 28066 RTL list made with `insn_list' objects, stored in the variable 28067 `current_function_epilogue_delay_list'. The insn for the first 28068 delay slot comes first in the list. Your definition of the macro 28069 `TARGET_ASM_FUNCTION_EPILOGUE' should fill the delay slots by 28070 outputting the insns in this list, usually by calling 28071 `final_scan_insn'. 28072 28073 You need not define this macro if you did not define 28074 `DELAY_SLOTS_FOR_EPILOGUE'. 28075 28076 -- Target Hook: void TARGET_ASM_OUTPUT_MI_THUNK (FILE *FILE, tree 28077 THUNK_FNDECL, HOST_WIDE_INT DELTA, HOST_WIDE_INT 28078 VCALL_OFFSET, tree FUNCTION) 28079 A function that outputs the assembler code for a thunk function, 28080 used to implement C++ virtual function calls with multiple 28081 inheritance. The thunk acts as a wrapper around a virtual 28082 function, adjusting the implicit object parameter before handing 28083 control off to the real function. 28084 28085 First, emit code to add the integer DELTA to the location that 28086 contains the incoming first argument. Assume that this argument 28087 contains a pointer, and is the one used to pass the `this' pointer 28088 in C++. This is the incoming argument _before_ the function 28089 prologue, e.g. `%o0' on a sparc. The addition must preserve the 28090 values of all other incoming arguments. 28091 28092 Then, if VCALL_OFFSET is nonzero, an additional adjustment should 28093 be made after adding `delta'. In particular, if P is the adjusted 28094 pointer, the following adjustment should be made: 28095 28096 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)] 28097 28098 After the additions, emit code to jump to FUNCTION, which is a 28099 `FUNCTION_DECL'. This is a direct pure jump, not a call, and does 28100 not touch the return address. Hence returning from FUNCTION will 28101 return to whoever called the current `thunk'. 28102 28103 The effect must be as if FUNCTION had been called directly with 28104 the adjusted first argument. This macro is responsible for 28105 emitting all of the code for a thunk function; 28106 `TARGET_ASM_FUNCTION_PROLOGUE' and `TARGET_ASM_FUNCTION_EPILOGUE' 28107 are not invoked. 28108 28109 The THUNK_FNDECL is redundant. (DELTA and FUNCTION have already 28110 been extracted from it.) It might possibly be useful on some 28111 targets, but probably not. 28112 28113 If you do not define this macro, the target-independent code in 28114 the C++ front end will generate a less efficient heavyweight thunk 28115 that calls FUNCTION instead of jumping to it. The generic 28116 approach does not support varargs. 28117 28118 -- Target Hook: bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (tree 28119 THUNK_FNDECL, HOST_WIDE_INT DELTA, HOST_WIDE_INT 28120 VCALL_OFFSET, tree FUNCTION) 28121 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would 28122 be able to output the assembler code for the thunk function 28123 specified by the arguments it is passed, and false otherwise. In 28124 the latter case, the generic approach will be used by the C++ 28125 front end, with the limitations previously exposed. 28126 28127 28128 File: gccint.info, Node: Profiling, Next: Tail Calls, Prev: Function Entry, Up: Stack and Calling 28129 28130 17.10.12 Generating Code for Profiling 28131 -------------------------------------- 28132 28133 These macros will help you generate code for profiling. 28134 28135 -- Macro: FUNCTION_PROFILER (FILE, LABELNO) 28136 A C statement or compound statement to output to FILE some 28137 assembler code to call the profiling subroutine `mcount'. 28138 28139 The details of how `mcount' expects to be called are determined by 28140 your operating system environment, not by GCC. To figure them out, 28141 compile a small program for profiling using the system's installed 28142 C compiler and look at the assembler code that results. 28143 28144 Older implementations of `mcount' expect the address of a counter 28145 variable to be loaded into some register. The name of this 28146 variable is `LP' followed by the number LABELNO, so you would 28147 generate the name using `LP%d' in a `fprintf'. 28148 28149 -- Macro: PROFILE_HOOK 28150 A C statement or compound statement to output to FILE some assembly 28151 code to call the profiling subroutine `mcount' even the target does 28152 not support profiling. 28153 28154 -- Macro: NO_PROFILE_COUNTERS 28155 Define this macro to be an expression with a nonzero value if the 28156 `mcount' subroutine on your system does not need a counter variable 28157 allocated for each function. This is true for almost all modern 28158 implementations. If you define this macro, you must not use the 28159 LABELNO argument to `FUNCTION_PROFILER'. 28160 28161 -- Macro: PROFILE_BEFORE_PROLOGUE 28162 Define this macro if the code for function profiling should come 28163 before the function prologue. Normally, the profiling code comes 28164 after. 28165 28166 28167 File: gccint.info, Node: Tail Calls, Next: Stack Smashing Protection, Prev: Profiling, Up: Stack and Calling 28168 28169 17.10.13 Permitting tail calls 28170 ------------------------------ 28171 28172 -- Target Hook: bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree DECL, tree 28173 EXP) 28174 True if it is ok to do sibling call optimization for the specified 28175 call expression EXP. DECL will be the called function, or `NULL' 28176 if this is an indirect call. 28177 28178 It is not uncommon for limitations of calling conventions to 28179 prevent tail calls to functions outside the current unit of 28180 translation, or during PIC compilation. The hook is used to 28181 enforce these restrictions, as the `sibcall' md pattern can not 28182 fail, or fall over to a "normal" call. The criteria for 28183 successful sibling call optimization may vary greatly between 28184 different architectures. 28185 28186 -- Target Hook: void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *REGS) 28187 Add any hard registers to REGS that are live on entry to the 28188 function. This hook only needs to be defined to provide registers 28189 that cannot be found by examination of FUNCTION_ARG_REGNO_P, the 28190 callee saved registers, STATIC_CHAIN_INCOMING_REGNUM, 28191 STATIC_CHAIN_REGNUM, TARGET_STRUCT_VALUE_RTX, 28192 FRAME_POINTER_REGNUM, EH_USES, FRAME_POINTER_REGNUM, 28193 ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM. 28194 28195 28196 File: gccint.info, Node: Stack Smashing Protection, Prev: Tail Calls, Up: Stack and Calling 28197 28198 17.10.14 Stack smashing protection 28199 ---------------------------------- 28200 28201 -- Target Hook: tree TARGET_STACK_PROTECT_GUARD (void) 28202 This hook returns a `DECL' node for the external variable to use 28203 for the stack protection guard. This variable is initialized by 28204 the runtime to some random value and is used to initialize the 28205 guard value that is placed at the top of the local stack frame. 28206 The type of this variable must be `ptr_type_node'. 28207 28208 The default version of this hook creates a variable called 28209 `__stack_chk_guard', which is normally defined in `libgcc2.c'. 28210 28211 -- Target Hook: tree TARGET_STACK_PROTECT_FAIL (void) 28212 This hook returns a tree expression that alerts the runtime that 28213 the stack protect guard variable has been modified. This 28214 expression should involve a call to a `noreturn' function. 28215 28216 The default version of this hook invokes a function called 28217 `__stack_chk_fail', taking no arguments. This function is 28218 normally defined in `libgcc2.c'. 28219 28220 28221 File: gccint.info, Node: Varargs, Next: Trampolines, Prev: Stack and Calling, Up: Target Macros 28222 28223 17.11 Implementing the Varargs Macros 28224 ===================================== 28225 28226 GCC comes with an implementation of `<varargs.h>' and `<stdarg.h>' that 28227 work without change on machines that pass arguments on the stack. 28228 Other machines require their own implementations of varargs, and the 28229 two machine independent header files must have conditionals to include 28230 it. 28231 28232 ISO `<stdarg.h>' differs from traditional `<varargs.h>' mainly in the 28233 calling convention for `va_start'. The traditional implementation 28234 takes just one argument, which is the variable in which to store the 28235 argument pointer. The ISO implementation of `va_start' takes an 28236 additional second argument. The user is supposed to write the last 28237 named argument of the function here. 28238 28239 However, `va_start' should not use this argument. The way to find the 28240 end of the named arguments is with the built-in functions described 28241 below. 28242 28243 -- Macro: __builtin_saveregs () 28244 Use this built-in function to save the argument registers in 28245 memory so that the varargs mechanism can access them. Both ISO 28246 and traditional versions of `va_start' must use 28247 `__builtin_saveregs', unless you use 28248 `TARGET_SETUP_INCOMING_VARARGS' (see below) instead. 28249 28250 On some machines, `__builtin_saveregs' is open-coded under the 28251 control of the target hook `TARGET_EXPAND_BUILTIN_SAVEREGS'. On 28252 other machines, it calls a routine written in assembler language, 28253 found in `libgcc2.c'. 28254 28255 Code generated for the call to `__builtin_saveregs' appears at the 28256 beginning of the function, as opposed to where the call to 28257 `__builtin_saveregs' is written, regardless of what the code is. 28258 This is because the registers must be saved before the function 28259 starts to use them for its own purposes. 28260 28261 -- Macro: __builtin_args_info (CATEGORY) 28262 Use this built-in function to find the first anonymous arguments in 28263 registers. 28264 28265 In general, a machine may have several categories of registers 28266 used for arguments, each for a particular category of data types. 28267 (For example, on some machines, floating-point registers are used 28268 for floating-point arguments while other arguments are passed in 28269 the general registers.) To make non-varargs functions use the 28270 proper calling convention, you have defined the `CUMULATIVE_ARGS' 28271 data type to record how many registers in each category have been 28272 used so far 28273 28274 `__builtin_args_info' accesses the same data structure of type 28275 `CUMULATIVE_ARGS' after the ordinary argument layout is finished 28276 with it, with CATEGORY specifying which word to access. Thus, the 28277 value indicates the first unused register in a given category. 28278 28279 Normally, you would use `__builtin_args_info' in the implementation 28280 of `va_start', accessing each category just once and storing the 28281 value in the `va_list' object. This is because `va_list' will 28282 have to update the values, and there is no way to alter the values 28283 accessed by `__builtin_args_info'. 28284 28285 -- Macro: __builtin_next_arg (LASTARG) 28286 This is the equivalent of `__builtin_args_info', for stack 28287 arguments. It returns the address of the first anonymous stack 28288 argument, as type `void *'. If `ARGS_GROW_DOWNWARD', it returns 28289 the address of the location above the first anonymous stack 28290 argument. Use it in `va_start' to initialize the pointer for 28291 fetching arguments from the stack. Also use it in `va_start' to 28292 verify that the second parameter LASTARG is the last named argument 28293 of the current function. 28294 28295 -- Macro: __builtin_classify_type (OBJECT) 28296 Since each machine has its own conventions for which data types are 28297 passed in which kind of register, your implementation of `va_arg' 28298 has to embody these conventions. The easiest way to categorize the 28299 specified data type is to use `__builtin_classify_type' together 28300 with `sizeof' and `__alignof__'. 28301 28302 `__builtin_classify_type' ignores the value of OBJECT, considering 28303 only its data type. It returns an integer describing what kind of 28304 type that is--integer, floating, pointer, structure, and so on. 28305 28306 The file `typeclass.h' defines an enumeration that you can use to 28307 interpret the values of `__builtin_classify_type'. 28308 28309 These machine description macros help implement varargs: 28310 28311 -- Target Hook: rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void) 28312 If defined, this hook produces the machine-specific code for a 28313 call to `__builtin_saveregs'. This code will be moved to the very 28314 beginning of the function, before any parameter access are made. 28315 The return value of this function should be an RTX that contains 28316 the value to use as the return of `__builtin_saveregs'. 28317 28318 -- Target Hook: void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS 28319 *ARGS_SO_FAR, enum machine_mode MODE, tree TYPE, int 28320 *PRETEND_ARGS_SIZE, int SECOND_TIME) 28321 This target hook offers an alternative to using 28322 `__builtin_saveregs' and defining the hook 28323 `TARGET_EXPAND_BUILTIN_SAVEREGS'. Use it to store the anonymous 28324 register arguments into the stack so that all the arguments appear 28325 to have been passed consecutively on the stack. Once this is 28326 done, you can use the standard implementation of varargs that 28327 works for machines that pass all their arguments on the stack. 28328 28329 The argument ARGS_SO_FAR points to the `CUMULATIVE_ARGS' data 28330 structure, containing the values that are obtained after 28331 processing the named arguments. The arguments MODE and TYPE 28332 describe the last named argument--its machine mode and its data 28333 type as a tree node. 28334 28335 The target hook should do two things: first, push onto the stack 28336 all the argument registers _not_ used for the named arguments, and 28337 second, store the size of the data thus pushed into the 28338 `int'-valued variable pointed to by PRETEND_ARGS_SIZE. The value 28339 that you store here will serve as additional offset for setting up 28340 the stack frame. 28341 28342 Because you must generate code to push the anonymous arguments at 28343 compile time without knowing their data types, 28344 `TARGET_SETUP_INCOMING_VARARGS' is only useful on machines that 28345 have just a single category of argument register and use it 28346 uniformly for all data types. 28347 28348 If the argument SECOND_TIME is nonzero, it means that the 28349 arguments of the function are being analyzed for the second time. 28350 This happens for an inline function, which is not actually 28351 compiled until the end of the source file. The hook 28352 `TARGET_SETUP_INCOMING_VARARGS' should not generate any 28353 instructions in this case. 28354 28355 -- Target Hook: bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS 28356 *CA) 28357 Define this hook to return `true' if the location where a function 28358 argument is passed depends on whether or not it is a named 28359 argument. 28360 28361 This hook controls how the NAMED argument to `FUNCTION_ARG' is set 28362 for varargs and stdarg functions. If this hook returns `true', 28363 the NAMED argument is always true for named arguments, and false 28364 for unnamed arguments. If it returns `false', but 28365 `TARGET_PRETEND_OUTGOING_VARARGS_NAMED' returns `true', then all 28366 arguments are treated as named. Otherwise, all named arguments 28367 except the last are treated as named. 28368 28369 You need not define this hook if it always returns zero. 28370 28371 -- Target Hook: bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED 28372 If you need to conditionally change ABIs so that one works with 28373 `TARGET_SETUP_INCOMING_VARARGS', but the other works like neither 28374 `TARGET_SETUP_INCOMING_VARARGS' nor 28375 `TARGET_STRICT_ARGUMENT_NAMING' was defined, then define this hook 28376 to return `true' if `TARGET_SETUP_INCOMING_VARARGS' is used, 28377 `false' otherwise. Otherwise, you should not define this hook. 28378 28379 28380 File: gccint.info, Node: Trampolines, Next: Library Calls, Prev: Varargs, Up: Target Macros 28381 28382 17.12 Trampolines for Nested Functions 28383 ====================================== 28384 28385 A "trampoline" is a small piece of code that is created at run time 28386 when the address of a nested function is taken. It normally resides on 28387 the stack, in the stack frame of the containing function. These macros 28388 tell GCC how to generate code to allocate and initialize a trampoline. 28389 28390 The instructions in the trampoline must do two things: load a constant 28391 address into the static chain register, and jump to the real address of 28392 the nested function. On CISC machines such as the m68k, this requires 28393 two instructions, a move immediate and a jump. Then the two addresses 28394 exist in the trampoline as word-long immediate operands. On RISC 28395 machines, it is often necessary to load each address into a register in 28396 two parts. Then pieces of each address form separate immediate 28397 operands. 28398 28399 The code generated to initialize the trampoline must store the variable 28400 parts--the static chain value and the function address--into the 28401 immediate operands of the instructions. On a CISC machine, this is 28402 simply a matter of copying each address to a memory reference at the 28403 proper offset from the start of the trampoline. On a RISC machine, it 28404 may be necessary to take out pieces of the address and store them 28405 separately. 28406 28407 -- Macro: TRAMPOLINE_TEMPLATE (FILE) 28408 A C statement to output, on the stream FILE, assembler code for a 28409 block of data that contains the constant parts of a trampoline. 28410 This code should not include a label--the label is taken care of 28411 automatically. 28412 28413 If you do not define this macro, it means no template is needed 28414 for the target. Do not define this macro on systems where the 28415 block move code to copy the trampoline into place would be larger 28416 than the code to generate it on the spot. 28417 28418 -- Macro: TRAMPOLINE_SECTION 28419 Return the section into which the trampoline template is to be 28420 placed (*note Sections::). The default value is 28421 `readonly_data_section'. 28422 28423 -- Macro: TRAMPOLINE_SIZE 28424 A C expression for the size in bytes of the trampoline, as an 28425 integer. 28426 28427 -- Macro: TRAMPOLINE_ALIGNMENT 28428 Alignment required for trampolines, in bits. 28429 28430 If you don't define this macro, the value of `BIGGEST_ALIGNMENT' 28431 is used for aligning trampolines. 28432 28433 -- Macro: INITIALIZE_TRAMPOLINE (ADDR, FNADDR, STATIC_CHAIN) 28434 A C statement to initialize the variable parts of a trampoline. 28435 ADDR is an RTX for the address of the trampoline; FNADDR is an RTX 28436 for the address of the nested function; STATIC_CHAIN is an RTX for 28437 the static chain value that should be passed to the function when 28438 it is called. 28439 28440 -- Macro: TRAMPOLINE_ADJUST_ADDRESS (ADDR) 28441 A C statement that should perform any machine-specific adjustment 28442 in the address of the trampoline. Its argument contains the 28443 address that was passed to `INITIALIZE_TRAMPOLINE'. In case the 28444 address to be used for a function call should be different from 28445 the address in which the template was stored, the different 28446 address should be assigned to ADDR. If this macro is not defined, 28447 ADDR will be used for function calls. 28448 28449 If this macro is not defined, by default the trampoline is 28450 allocated as a stack slot. This default is right for most 28451 machines. The exceptions are machines where it is impossible to 28452 execute instructions in the stack area. On such machines, you may 28453 have to implement a separate stack, using this macro in 28454 conjunction with `TARGET_ASM_FUNCTION_PROLOGUE' and 28455 `TARGET_ASM_FUNCTION_EPILOGUE'. 28456 28457 FP points to a data structure, a `struct function', which 28458 describes the compilation status of the immediate containing 28459 function of the function which the trampoline is for. The stack 28460 slot for the trampoline is in the stack frame of this containing 28461 function. Other allocation strategies probably must do something 28462 analogous with this information. 28463 28464 Implementing trampolines is difficult on many machines because they 28465 have separate instruction and data caches. Writing into a stack 28466 location fails to clear the memory in the instruction cache, so when 28467 the program jumps to that location, it executes the old contents. 28468 28469 Here are two possible solutions. One is to clear the relevant parts of 28470 the instruction cache whenever a trampoline is set up. The other is to 28471 make all trampolines identical, by having them jump to a standard 28472 subroutine. The former technique makes trampoline execution faster; the 28473 latter makes initialization faster. 28474 28475 To clear the instruction cache when a trampoline is initialized, define 28476 the following macro. 28477 28478 -- Macro: CLEAR_INSN_CACHE (BEG, END) 28479 If defined, expands to a C expression clearing the _instruction 28480 cache_ in the specified interval. The definition of this macro 28481 would typically be a series of `asm' statements. Both BEG and END 28482 are both pointer expressions. 28483 28484 The operating system may also require the stack to be made executable 28485 before calling the trampoline. To implement this requirement, define 28486 the following macro. 28487 28488 -- Macro: ENABLE_EXECUTE_STACK 28489 Define this macro if certain operations must be performed before 28490 executing code located on the stack. The macro should expand to a 28491 series of C file-scope constructs (e.g. functions) and provide a 28492 unique entry point named `__enable_execute_stack'. The target is 28493 responsible for emitting calls to the entry point in the code, for 28494 example from the `INITIALIZE_TRAMPOLINE' macro. 28495 28496 To use a standard subroutine, define the following macro. In addition, 28497 you must make sure that the instructions in a trampoline fill an entire 28498 cache line with identical instructions, or else ensure that the 28499 beginning of the trampoline code is always aligned at the same point in 28500 its cache line. Look in `m68k.h' as a guide. 28501 28502 -- Macro: TRANSFER_FROM_TRAMPOLINE 28503 Define this macro if trampolines need a special subroutine to do 28504 their work. The macro should expand to a series of `asm' 28505 statements which will be compiled with GCC. They go in a library 28506 function named `__transfer_from_trampoline'. 28507 28508 If you need to avoid executing the ordinary prologue code of a 28509 compiled C function when you jump to the subroutine, you can do so 28510 by placing a special label of your own in the assembler code. Use 28511 one `asm' statement to generate an assembler label, and another to 28512 make the label global. Then trampolines can use that label to 28513 jump directly to your special assembler code. 28514 28515 28516 File: gccint.info, Node: Library Calls, Next: Addressing Modes, Prev: Trampolines, Up: Target Macros 28517 28518 17.13 Implicit Calls to Library Routines 28519 ======================================== 28520 28521 Here is an explanation of implicit calls to library routines. 28522 28523 -- Macro: DECLARE_LIBRARY_RENAMES 28524 This macro, if defined, should expand to a piece of C code that 28525 will get expanded when compiling functions for libgcc.a. It can 28526 be used to provide alternate names for GCC's internal library 28527 functions if there are ABI-mandated names that the compiler should 28528 provide. 28529 28530 -- Target Hook: void TARGET_INIT_LIBFUNCS (void) 28531 This hook should declare additional library routines or rename 28532 existing ones, using the functions `set_optab_libfunc' and 28533 `init_one_libfunc' defined in `optabs.c'. `init_optabs' calls 28534 this macro after initializing all the normal library routines. 28535 28536 The default is to do nothing. Most ports don't need to define 28537 this hook. 28538 28539 -- Macro: FLOAT_LIB_COMPARE_RETURNS_BOOL (MODE, COMPARISON) 28540 This macro should return `true' if the library routine that 28541 implements the floating point comparison operator COMPARISON in 28542 mode MODE will return a boolean, and FALSE if it will return a 28543 tristate. 28544 28545 GCC's own floating point libraries return tristates from the 28546 comparison operators, so the default returns false always. Most 28547 ports don't need to define this macro. 28548 28549 -- Macro: TARGET_LIB_INT_CMP_BIASED 28550 This macro should evaluate to `true' if the integer comparison 28551 functions (like `__cmpdi2') return 0 to indicate that the first 28552 operand is smaller than the second, 1 to indicate that they are 28553 equal, and 2 to indicate that the first operand is greater than 28554 the second. If this macro evaluates to `false' the comparison 28555 functions return -1, 0, and 1 instead of 0, 1, and 2. If the 28556 target uses the routines in `libgcc.a', you do not need to define 28557 this macro. 28558 28559 -- Macro: US_SOFTWARE_GOFAST 28560 Define this macro if your system C library uses the US Software 28561 GOFAST library to provide floating point emulation. 28562 28563 In addition to defining this macro, your architecture must set 28564 `TARGET_INIT_LIBFUNCS' to `gofast_maybe_init_libfuncs', or else 28565 call that function from its version of that hook. It is defined 28566 in `config/gofast.h', which must be included by your 28567 architecture's `CPU.c' file. See `sparc/sparc.c' for an example. 28568 28569 If this macro is defined, the 28570 `TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL' target hook must return 28571 false for `SFmode' and `DFmode' comparisons. 28572 28573 -- Macro: TARGET_EDOM 28574 The value of `EDOM' on the target machine, as a C integer constant 28575 expression. If you don't define this macro, GCC does not attempt 28576 to deposit the value of `EDOM' into `errno' directly. Look in 28577 `/usr/include/errno.h' to find the value of `EDOM' on your system. 28578 28579 If you do not define `TARGET_EDOM', then compiled code reports 28580 domain errors by calling the library function and letting it 28581 report the error. If mathematical functions on your system use 28582 `matherr' when there is an error, then you should leave 28583 `TARGET_EDOM' undefined so that `matherr' is used normally. 28584 28585 -- Macro: GEN_ERRNO_RTX 28586 Define this macro as a C expression to create an rtl expression 28587 that refers to the global "variable" `errno'. (On certain systems, 28588 `errno' may not actually be a variable.) If you don't define this 28589 macro, a reasonable default is used. 28590 28591 -- Macro: TARGET_C99_FUNCTIONS 28592 When this macro is nonzero, GCC will implicitly optimize `sin' 28593 calls into `sinf' and similarly for other functions defined by C99 28594 standard. The default is zero because a number of existing 28595 systems lack support for these functions in their runtime so this 28596 macro needs to be redefined to one on systems that do support the 28597 C99 runtime. 28598 28599 -- Macro: TARGET_HAS_SINCOS 28600 When this macro is nonzero, GCC will implicitly optimize calls to 28601 `sin' and `cos' with the same argument to a call to `sincos'. The 28602 default is zero. The target has to provide the following 28603 functions: 28604 void sincos(double x, double *sin, double *cos); 28605 void sincosf(float x, float *sin, float *cos); 28606 void sincosl(long double x, long double *sin, long double *cos); 28607 28608 -- Macro: NEXT_OBJC_RUNTIME 28609 Define this macro to generate code for Objective-C message sending 28610 using the calling convention of the NeXT system. This calling 28611 convention involves passing the object, the selector and the 28612 method arguments all at once to the method-lookup library function. 28613 28614 The default calling convention passes just the object and the 28615 selector to the lookup function, which returns a pointer to the 28616 method. 28617 28618 28619 File: gccint.info, Node: Addressing Modes, Next: Anchored Addresses, Prev: Library Calls, Up: Target Macros 28620 28621 17.14 Addressing Modes 28622 ====================== 28623 28624 This is about addressing modes. 28625 28626 -- Macro: HAVE_PRE_INCREMENT 28627 -- Macro: HAVE_PRE_DECREMENT 28628 -- Macro: HAVE_POST_INCREMENT 28629 -- Macro: HAVE_POST_DECREMENT 28630 A C expression that is nonzero if the machine supports 28631 pre-increment, pre-decrement, post-increment, or post-decrement 28632 addressing respectively. 28633 28634 -- Macro: HAVE_PRE_MODIFY_DISP 28635 -- Macro: HAVE_POST_MODIFY_DISP 28636 A C expression that is nonzero if the machine supports pre- or 28637 post-address side-effect generation involving constants other than 28638 the size of the memory operand. 28639 28640 -- Macro: HAVE_PRE_MODIFY_REG 28641 -- Macro: HAVE_POST_MODIFY_REG 28642 A C expression that is nonzero if the machine supports pre- or 28643 post-address side-effect generation involving a register 28644 displacement. 28645 28646 -- Macro: CONSTANT_ADDRESS_P (X) 28647 A C expression that is 1 if the RTX X is a constant which is a 28648 valid address. On most machines, this can be defined as 28649 `CONSTANT_P (X)', but a few machines are more restrictive in which 28650 constant addresses are supported. 28651 28652 -- Macro: CONSTANT_P (X) 28653 `CONSTANT_P', which is defined by target-independent code, accepts 28654 integer-values expressions whose values are not explicitly known, 28655 such as `symbol_ref', `label_ref', and `high' expressions and 28656 `const' arithmetic expressions, in addition to `const_int' and 28657 `const_double' expressions. 28658 28659 -- Macro: MAX_REGS_PER_ADDRESS 28660 A number, the maximum number of registers that can appear in a 28661 valid memory address. Note that it is up to you to specify a 28662 value equal to the maximum number that `GO_IF_LEGITIMATE_ADDRESS' 28663 would ever accept. 28664 28665 -- Macro: GO_IF_LEGITIMATE_ADDRESS (MODE, X, LABEL) 28666 A C compound statement with a conditional `goto LABEL;' executed 28667 if X (an RTX) is a legitimate memory address on the target machine 28668 for a memory operand of mode MODE. 28669 28670 It usually pays to define several simpler macros to serve as 28671 subroutines for this one. Otherwise it may be too complicated to 28672 understand. 28673 28674 This macro must exist in two variants: a strict variant and a 28675 non-strict one. The strict variant is used in the reload pass. It 28676 must be defined so that any pseudo-register that has not been 28677 allocated a hard register is considered a memory reference. In 28678 contexts where some kind of register is required, a pseudo-register 28679 with no hard register must be rejected. 28680 28681 The non-strict variant is used in other passes. It must be 28682 defined to accept all pseudo-registers in every context where some 28683 kind of register is required. 28684 28685 Compiler source files that want to use the strict variant of this 28686 macro define the macro `REG_OK_STRICT'. You should use an `#ifdef 28687 REG_OK_STRICT' conditional to define the strict variant in that 28688 case and the non-strict variant otherwise. 28689 28690 Subroutines to check for acceptable registers for various purposes 28691 (one for base registers, one for index registers, and so on) are 28692 typically among the subroutines used to define 28693 `GO_IF_LEGITIMATE_ADDRESS'. Then only these subroutine macros 28694 need have two variants; the higher levels of macros may be the 28695 same whether strict or not. 28696 28697 Normally, constant addresses which are the sum of a `symbol_ref' 28698 and an integer are stored inside a `const' RTX to mark them as 28699 constant. Therefore, there is no need to recognize such sums 28700 specifically as legitimate addresses. Normally you would simply 28701 recognize any `const' as legitimate. 28702 28703 Usually `PRINT_OPERAND_ADDRESS' is not prepared to handle constant 28704 sums that are not marked with `const'. It assumes that a naked 28705 `plus' indicates indexing. If so, then you _must_ reject such 28706 naked constant sums as illegitimate addresses, so that none of 28707 them will be given to `PRINT_OPERAND_ADDRESS'. 28708 28709 On some machines, whether a symbolic address is legitimate depends 28710 on the section that the address refers to. On these machines, 28711 define the target hook `TARGET_ENCODE_SECTION_INFO' to store the 28712 information into the `symbol_ref', and then check for it here. 28713 When you see a `const', you will have to look inside it to find the 28714 `symbol_ref' in order to determine the section. *Note Assembler 28715 Format::. 28716 28717 -- Macro: TARGET_MEM_CONSTRAINT 28718 A single character to be used instead of the default `'m'' 28719 character for general memory addresses. This defines the 28720 constraint letter which matches the memory addresses accepted by 28721 `GO_IF_LEGITIMATE_ADDRESS_P'. Define this macro if you want to 28722 support new address formats in your back end without changing the 28723 semantics of the `'m'' constraint. This is necessary in order to 28724 preserve functionality of inline assembly constructs using the 28725 `'m'' constraint. 28726 28727 -- Macro: FIND_BASE_TERM (X) 28728 A C expression to determine the base term of address X, or to 28729 provide a simplified version of X from which `alias.c' can easily 28730 find the base term. This macro is used in only two places: 28731 `find_base_value' and `find_base_term' in `alias.c'. 28732 28733 It is always safe for this macro to not be defined. It exists so 28734 that alias analysis can understand machine-dependent addresses. 28735 28736 The typical use of this macro is to handle addresses containing a 28737 label_ref or symbol_ref within an UNSPEC. 28738 28739 -- Macro: LEGITIMIZE_ADDRESS (X, OLDX, MODE, WIN) 28740 A C compound statement that attempts to replace X with a valid 28741 memory address for an operand of mode MODE. WIN will be a C 28742 statement label elsewhere in the code; the macro definition may use 28743 28744 GO_IF_LEGITIMATE_ADDRESS (MODE, X, WIN); 28745 28746 to avoid further processing if the address has become legitimate. 28747 28748 X will always be the result of a call to `break_out_memory_refs', 28749 and OLDX will be the operand that was given to that function to 28750 produce X. 28751 28752 The code generated by this macro should not alter the substructure 28753 of X. If it transforms X into a more legitimate form, it should 28754 assign X (which will always be a C variable) a new value. 28755 28756 It is not necessary for this macro to come up with a legitimate 28757 address. The compiler has standard ways of doing so in all cases. 28758 In fact, it is safe to omit this macro. But often a 28759 machine-dependent strategy can generate better code. 28760 28761 -- Macro: LEGITIMIZE_RELOAD_ADDRESS (X, MODE, OPNUM, TYPE, IND_LEVELS, 28762 WIN) 28763 A C compound statement that attempts to replace X, which is an 28764 address that needs reloading, with a valid memory address for an 28765 operand of mode MODE. WIN will be a C statement label elsewhere 28766 in the code. It is not necessary to define this macro, but it 28767 might be useful for performance reasons. 28768 28769 For example, on the i386, it is sometimes possible to use a single 28770 reload register instead of two by reloading a sum of two pseudo 28771 registers into a register. On the other hand, for number of RISC 28772 processors offsets are limited so that often an intermediate 28773 address needs to be generated in order to address a stack slot. 28774 By defining `LEGITIMIZE_RELOAD_ADDRESS' appropriately, the 28775 intermediate addresses generated for adjacent some stack slots can 28776 be made identical, and thus be shared. 28777 28778 _Note_: This macro should be used with caution. It is necessary 28779 to know something of how reload works in order to effectively use 28780 this, and it is quite easy to produce macros that build in too 28781 much knowledge of reload internals. 28782 28783 _Note_: This macro must be able to reload an address created by a 28784 previous invocation of this macro. If it fails to handle such 28785 addresses then the compiler may generate incorrect code or abort. 28786 28787 The macro definition should use `push_reload' to indicate parts 28788 that need reloading; OPNUM, TYPE and IND_LEVELS are usually 28789 suitable to be passed unaltered to `push_reload'. 28790 28791 The code generated by this macro must not alter the substructure of 28792 X. If it transforms X into a more legitimate form, it should 28793 assign X (which will always be a C variable) a new value. This 28794 also applies to parts that you change indirectly by calling 28795 `push_reload'. 28796 28797 The macro definition may use `strict_memory_address_p' to test if 28798 the address has become legitimate. 28799 28800 If you want to change only a part of X, one standard way of doing 28801 this is to use `copy_rtx'. Note, however, that it unshares only a 28802 single level of rtl. Thus, if the part to be changed is not at the 28803 top level, you'll need to replace first the top level. It is not 28804 necessary for this macro to come up with a legitimate address; 28805 but often a machine-dependent strategy can generate better code. 28806 28807 -- Macro: GO_IF_MODE_DEPENDENT_ADDRESS (ADDR, LABEL) 28808 A C statement or compound statement with a conditional `goto 28809 LABEL;' executed if memory address X (an RTX) can have different 28810 meanings depending on the machine mode of the memory reference it 28811 is used for or if the address is valid for some modes but not 28812 others. 28813 28814 Autoincrement and autodecrement addresses typically have 28815 mode-dependent effects because the amount of the increment or 28816 decrement is the size of the operand being addressed. Some 28817 machines have other mode-dependent addresses. Many RISC machines 28818 have no mode-dependent addresses. 28819 28820 You may assume that ADDR is a valid address for the machine. 28821 28822 -- Macro: LEGITIMATE_CONSTANT_P (X) 28823 A C expression that is nonzero if X is a legitimate constant for 28824 an immediate operand on the target machine. You can assume that X 28825 satisfies `CONSTANT_P', so you need not check this. In fact, `1' 28826 is a suitable definition for this macro on machines where anything 28827 `CONSTANT_P' is valid. 28828 28829 -- Target Hook: rtx TARGET_DELEGITIMIZE_ADDRESS (rtx X) 28830 This hook is used to undo the possibly obfuscating effects of the 28831 `LEGITIMIZE_ADDRESS' and `LEGITIMIZE_RELOAD_ADDRESS' target 28832 macros. Some backend implementations of these macros wrap symbol 28833 references inside an `UNSPEC' rtx to represent PIC or similar 28834 addressing modes. This target hook allows GCC's optimizers to 28835 understand the semantics of these opaque `UNSPEC's by converting 28836 them back into their original form. 28837 28838 -- Target Hook: bool TARGET_CANNOT_FORCE_CONST_MEM (rtx X) 28839 This hook should return true if X is of a form that cannot (or 28840 should not) be spilled to the constant pool. The default version 28841 of this hook returns false. 28842 28843 The primary reason to define this hook is to prevent reload from 28844 deciding that a non-legitimate constant would be better reloaded 28845 from the constant pool instead of spilling and reloading a register 28846 holding the constant. This restriction is often true of addresses 28847 of TLS symbols for various targets. 28848 28849 -- Target Hook: bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum 28850 machine_mode MODE, rtx X) 28851 This hook should return true if pool entries for constant X can be 28852 placed in an `object_block' structure. MODE is the mode of X. 28853 28854 The default version returns false for all constants. 28855 28856 -- Target Hook: tree TARGET_BUILTIN_RECIPROCAL (enum tree_code FN, 28857 bool TM_FN, bool SQRT) 28858 This hook should return the DECL of a function that implements 28859 reciprocal of the builtin function with builtin function code FN, 28860 or `NULL_TREE' if such a function is not available. TM_FN is true 28861 when FN is a code of a machine-dependent builtin function. When 28862 SQRT is true, additional optimizations that apply only to the 28863 reciprocal of a square root function are performed, and only 28864 reciprocals of `sqrt' function are valid. 28865 28866 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void) 28867 This hook should return the DECL of a function F that given an 28868 address ADDR as an argument returns a mask M that can be used to 28869 extract from two vectors the relevant data that resides in ADDR in 28870 case ADDR is not properly aligned. 28871 28872 The autovectorizer, when vectorizing a load operation from an 28873 address ADDR that may be unaligned, will generate two vector loads 28874 from the two aligned addresses around ADDR. It then generates a 28875 `REALIGN_LOAD' operation to extract the relevant data from the two 28876 loaded vectors. The first two arguments to `REALIGN_LOAD', V1 and 28877 V2, are the two vectors, each of size VS, and the third argument, 28878 OFF, defines how the data will be extracted from these two 28879 vectors: if OFF is 0, then the returned vector is V2; otherwise, 28880 the returned vector is composed from the last VS-OFF elements of 28881 V1 concatenated to the first OFF elements of V2. 28882 28883 If this hook is defined, the autovectorizer will generate a call 28884 to F (using the DECL tree that this hook returns) and will use the 28885 return value of F as the argument OFF to `REALIGN_LOAD'. 28886 Therefore, the mask M returned by F should comply with the 28887 semantics expected by `REALIGN_LOAD' described above. If this 28888 hook is not defined, then ADDR will be used as the argument OFF to 28889 `REALIGN_LOAD', in which case the low log2(VS)-1 bits of ADDR will 28890 be considered. 28891 28892 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree X) 28893 This hook should return the DECL of a function F that implements 28894 widening multiplication of the even elements of two input vectors 28895 of type X. 28896 28897 If this hook is defined, the autovectorizer will use it along with 28898 the `TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD' target hook when 28899 vectorizing widening multiplication in cases that the order of the 28900 results does not have to be preserved (e.g. used only by a 28901 reduction computation). Otherwise, the `widen_mult_hi/lo' idioms 28902 will be used. 28903 28904 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree X) 28905 This hook should return the DECL of a function F that implements 28906 widening multiplication of the odd elements of two input vectors 28907 of type X. 28908 28909 If this hook is defined, the autovectorizer will use it along with 28910 the `TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN' target hook when 28911 vectorizing widening multiplication in cases that the order of the 28912 results does not have to be preserved (e.g. used only by a 28913 reduction computation). Otherwise, the `widen_mult_hi/lo' idioms 28914 will be used. 28915 28916 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum 28917 tree_code CODE, tree TYPE) 28918 This hook should return the DECL of a function that implements 28919 conversion of the input vector of type TYPE. If TYPE is an 28920 integral type, the result of the conversion is a vector of 28921 floating-point type of the same size. If TYPE is a floating-point 28922 type, the result of the conversion is a vector of integral type of 28923 the same size. CODE specifies how the conversion is to be applied 28924 (truncation, rounding, etc.). 28925 28926 If this hook is defined, the autovectorizer will use the 28927 `TARGET_VECTORIZE_BUILTIN_CONVERSION' target hook when vectorizing 28928 conversion. Otherwise, it will return `NULL_TREE'. 28929 28930 -- Target Hook: tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION 28931 (enum built_in_function CODE, tree VEC_TYPE_OUT, tree 28932 VEC_TYPE_IN) 28933 This hook should return the decl of a function that implements the 28934 vectorized variant of the builtin function with builtin function 28935 code CODE or `NULL_TREE' if such a function is not available. The 28936 return type of the vectorized function shall be of vector type 28937 VEC_TYPE_OUT and the argument types should be VEC_TYPE_IN. 28938 28939 28940 File: gccint.info, Node: Anchored Addresses, Next: Condition Code, Prev: Addressing Modes, Up: Target Macros 28941 28942 17.15 Anchored Addresses 28943 ======================== 28944 28945 GCC usually addresses every static object as a separate entity. For 28946 example, if we have: 28947 28948 static int a, b, c; 28949 int foo (void) { return a + b + c; } 28950 28951 the code for `foo' will usually calculate three separate symbolic 28952 addresses: those of `a', `b' and `c'. On some targets, it would be 28953 better to calculate just one symbolic address and access the three 28954 variables relative to it. The equivalent pseudocode would be something 28955 like: 28956 28957 int foo (void) 28958 { 28959 register int *xr = &x; 28960 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 28961 } 28962 28963 (which isn't valid C). We refer to shared addresses like `x' as 28964 "section anchors". Their use is controlled by `-fsection-anchors'. 28965 28966 The hooks below describe the target properties that GCC needs to know 28967 in order to make effective use of section anchors. It won't use 28968 section anchors at all unless either `TARGET_MIN_ANCHOR_OFFSET' or 28969 `TARGET_MAX_ANCHOR_OFFSET' is set to a nonzero value. 28970 28971 -- Variable: Target Hook HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET 28972 The minimum offset that should be applied to a section anchor. On 28973 most targets, it should be the smallest offset that can be applied 28974 to a base register while still giving a legitimate address for 28975 every mode. The default value is 0. 28976 28977 -- Variable: Target Hook HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET 28978 Like `TARGET_MIN_ANCHOR_OFFSET', but the maximum (inclusive) 28979 offset that should be applied to section anchors. The default 28980 value is 0. 28981 28982 -- Target Hook: void TARGET_ASM_OUTPUT_ANCHOR (rtx X) 28983 Write the assembly code to define section anchor X, which is a 28984 `SYMBOL_REF' for which `SYMBOL_REF_ANCHOR_P (X)' is true. The 28985 hook is called with the assembly output position set to the 28986 beginning of `SYMBOL_REF_BLOCK (X)'. 28987 28988 If `ASM_OUTPUT_DEF' is available, the hook's default definition 28989 uses it to define the symbol as `. + SYMBOL_REF_BLOCK_OFFSET (X)'. 28990 If `ASM_OUTPUT_DEF' is not available, the hook's default definition 28991 is `NULL', which disables the use of section anchors altogether. 28992 28993 -- Target Hook: bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx X) 28994 Return true if GCC should attempt to use anchors to access 28995 `SYMBOL_REF' X. You can assume `SYMBOL_REF_HAS_BLOCK_INFO_P (X)' 28996 and `!SYMBOL_REF_ANCHOR_P (X)'. 28997 28998 The default version is correct for most targets, but you might 28999 need to intercept this hook to handle things like target-specific 29000 attributes or target-specific sections. 29001 29002 29003 File: gccint.info, Node: Condition Code, Next: Costs, Prev: Anchored Addresses, Up: Target Macros 29004 29005 17.16 Condition Code Status 29006 =========================== 29007 29008 This describes the condition code status. 29009 29010 The file `conditions.h' defines a variable `cc_status' to describe how 29011 the condition code was computed (in case the interpretation of the 29012 condition code depends on the instruction that it was set by). This 29013 variable contains the RTL expressions on which the condition code is 29014 currently based, and several standard flags. 29015 29016 Sometimes additional machine-specific flags must be defined in the 29017 machine description header file. It can also add additional 29018 machine-specific information by defining `CC_STATUS_MDEP'. 29019 29020 -- Macro: CC_STATUS_MDEP 29021 C code for a data type which is used for declaring the `mdep' 29022 component of `cc_status'. It defaults to `int'. 29023 29024 This macro is not used on machines that do not use `cc0'. 29025 29026 -- Macro: CC_STATUS_MDEP_INIT 29027 A C expression to initialize the `mdep' field to "empty". The 29028 default definition does nothing, since most machines don't use the 29029 field anyway. If you want to use the field, you should probably 29030 define this macro to initialize it. 29031 29032 This macro is not used on machines that do not use `cc0'. 29033 29034 -- Macro: NOTICE_UPDATE_CC (EXP, INSN) 29035 A C compound statement to set the components of `cc_status' 29036 appropriately for an insn INSN whose body is EXP. It is this 29037 macro's responsibility to recognize insns that set the condition 29038 code as a byproduct of other activity as well as those that 29039 explicitly set `(cc0)'. 29040 29041 This macro is not used on machines that do not use `cc0'. 29042 29043 If there are insns that do not set the condition code but do alter 29044 other machine registers, this macro must check to see whether they 29045 invalidate the expressions that the condition code is recorded as 29046 reflecting. For example, on the 68000, insns that store in address 29047 registers do not set the condition code, which means that usually 29048 `NOTICE_UPDATE_CC' can leave `cc_status' unaltered for such insns. 29049 But suppose that the previous insn set the condition code based on 29050 location `a4@(102)' and the current insn stores a new value in 29051 `a4'. Although the condition code is not changed by this, it will 29052 no longer be true that it reflects the contents of `a4@(102)'. 29053 Therefore, `NOTICE_UPDATE_CC' must alter `cc_status' in this case 29054 to say that nothing is known about the condition code value. 29055 29056 The definition of `NOTICE_UPDATE_CC' must be prepared to deal with 29057 the results of peephole optimization: insns whose patterns are 29058 `parallel' RTXs containing various `reg', `mem' or constants which 29059 are just the operands. The RTL structure of these insns is not 29060 sufficient to indicate what the insns actually do. What 29061 `NOTICE_UPDATE_CC' should do when it sees one is just to run 29062 `CC_STATUS_INIT'. 29063 29064 A possible definition of `NOTICE_UPDATE_CC' is to call a function 29065 that looks at an attribute (*note Insn Attributes::) named, for 29066 example, `cc'. This avoids having detailed information about 29067 patterns in two places, the `md' file and in `NOTICE_UPDATE_CC'. 29068 29069 -- Macro: SELECT_CC_MODE (OP, X, Y) 29070 Returns a mode from class `MODE_CC' to be used when comparison 29071 operation code OP is applied to rtx X and Y. For example, on the 29072 SPARC, `SELECT_CC_MODE' is defined as (see *note Jump Patterns:: 29073 for a description of the reason for this definition) 29074 29075 #define SELECT_CC_MODE(OP,X,Y) \ 29076 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ 29077 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \ 29078 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ 29079 || GET_CODE (X) == NEG) \ 29080 ? CC_NOOVmode : CCmode)) 29081 29082 You should define this macro if and only if you define extra CC 29083 modes in `MACHINE-modes.def'. 29084 29085 -- Macro: CANONICALIZE_COMPARISON (CODE, OP0, OP1) 29086 On some machines not all possible comparisons are defined, but you 29087 can convert an invalid comparison into a valid one. For example, 29088 the Alpha does not have a `GT' comparison, but you can use an `LT' 29089 comparison instead and swap the order of the operands. 29090 29091 On such machines, define this macro to be a C statement to do any 29092 required conversions. CODE is the initial comparison code and OP0 29093 and OP1 are the left and right operands of the comparison, 29094 respectively. You should modify CODE, OP0, and OP1 as required. 29095 29096 GCC will not assume that the comparison resulting from this macro 29097 is valid but will see if the resulting insn matches a pattern in 29098 the `md' file. 29099 29100 You need not define this macro if it would never change the 29101 comparison code or operands. 29102 29103 -- Macro: REVERSIBLE_CC_MODE (MODE) 29104 A C expression whose value is one if it is always safe to reverse a 29105 comparison whose mode is MODE. If `SELECT_CC_MODE' can ever 29106 return MODE for a floating-point inequality comparison, then 29107 `REVERSIBLE_CC_MODE (MODE)' must be zero. 29108 29109 You need not define this macro if it would always returns zero or 29110 if the floating-point format is anything other than 29111 `IEEE_FLOAT_FORMAT'. For example, here is the definition used on 29112 the SPARC, where floating-point inequality comparisons are always 29113 given `CCFPEmode': 29114 29115 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode) 29116 29117 -- Macro: REVERSE_CONDITION (CODE, MODE) 29118 A C expression whose value is reversed condition code of the CODE 29119 for comparison done in CC_MODE MODE. The macro is used only in 29120 case `REVERSIBLE_CC_MODE (MODE)' is nonzero. Define this macro in 29121 case machine has some non-standard way how to reverse certain 29122 conditionals. For instance in case all floating point conditions 29123 are non-trapping, compiler may freely convert unordered compares 29124 to ordered one. Then definition may look like: 29125 29126 #define REVERSE_CONDITION(CODE, MODE) \ 29127 ((MODE) != CCFPmode ? reverse_condition (CODE) \ 29128 : reverse_condition_maybe_unordered (CODE)) 29129 29130 -- Macro: REVERSE_CONDEXEC_PREDICATES_P (OP1, OP2) 29131 A C expression that returns true if the conditional execution 29132 predicate OP1, a comparison operation, is the inverse of OP2 and 29133 vice versa. Define this to return 0 if the target has conditional 29134 execution predicates that cannot be reversed safely. There is no 29135 need to validate that the arguments of op1 and op2 are the same, 29136 this is done separately. If no expansion is specified, this macro 29137 is defined as follows: 29138 29139 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \ 29140 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL)) 29141 29142 -- Target Hook: bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, 29143 unsigned int *) 29144 On targets which do not use `(cc0)', and which use a hard register 29145 rather than a pseudo-register to hold condition codes, the regular 29146 CSE passes are often not able to identify cases in which the hard 29147 register is set to a common value. Use this hook to enable a 29148 small pass which optimizes such cases. This hook should return 29149 true to enable this pass, and it should set the integers to which 29150 its arguments point to the hard register numbers used for 29151 condition codes. When there is only one such register, as is true 29152 on most systems, the integer pointed to by the second argument 29153 should be set to `INVALID_REGNUM'. 29154 29155 The default version of this hook returns false. 29156 29157 -- Target Hook: enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum 29158 machine_mode, enum machine_mode) 29159 On targets which use multiple condition code modes in class 29160 `MODE_CC', it is sometimes the case that a comparison can be 29161 validly done in more than one mode. On such a system, define this 29162 target hook to take two mode arguments and to return a mode in 29163 which both comparisons may be validly done. If there is no such 29164 mode, return `VOIDmode'. 29165 29166 The default version of this hook checks whether the modes are the 29167 same. If they are, it returns that mode. If they are different, 29168 it returns `VOIDmode'. 29169 29170 29171 File: gccint.info, Node: Costs, Next: Scheduling, Prev: Condition Code, Up: Target Macros 29172 29173 17.17 Describing Relative Costs of Operations 29174 ============================================= 29175 29176 These macros let you describe the relative speed of various operations 29177 on the target machine. 29178 29179 -- Macro: REGISTER_MOVE_COST (MODE, FROM, TO) 29180 A C expression for the cost of moving data of mode MODE from a 29181 register in class FROM to one in class TO. The classes are 29182 expressed using the enumeration values such as `GENERAL_REGS'. A 29183 value of 2 is the default; other values are interpreted relative to 29184 that. 29185 29186 It is not required that the cost always equal 2 when FROM is the 29187 same as TO; on some machines it is expensive to move between 29188 registers if they are not general registers. 29189 29190 If reload sees an insn consisting of a single `set' between two 29191 hard registers, and if `REGISTER_MOVE_COST' applied to their 29192 classes returns a value of 2, reload does not check to ensure that 29193 the constraints of the insn are met. Setting a cost of other than 29194 2 will allow reload to verify that the constraints are met. You 29195 should do this if the `movM' pattern's constraints do not allow 29196 such copying. 29197 29198 -- Macro: MEMORY_MOVE_COST (MODE, CLASS, IN) 29199 A C expression for the cost of moving data of mode MODE between a 29200 register of class CLASS and memory; IN is zero if the value is to 29201 be written to memory, nonzero if it is to be read in. This cost 29202 is relative to those in `REGISTER_MOVE_COST'. If moving between 29203 registers and memory is more expensive than between two registers, 29204 you should define this macro to express the relative cost. 29205 29206 If you do not define this macro, GCC uses a default cost of 4 plus 29207 the cost of copying via a secondary reload register, if one is 29208 needed. If your machine requires a secondary reload register to 29209 copy between memory and a register of CLASS but the reload 29210 mechanism is more complex than copying via an intermediate, define 29211 this macro to reflect the actual cost of the move. 29212 29213 GCC defines the function `memory_move_secondary_cost' if secondary 29214 reloads are needed. It computes the costs due to copying via a 29215 secondary register. If your machine copies from memory using a 29216 secondary register in the conventional way but the default base 29217 value of 4 is not correct for your machine, define this macro to 29218 add some other value to the result of that function. The 29219 arguments to that function are the same as to this macro. 29220 29221 -- Macro: BRANCH_COST (SPEED_P, PREDICTABLE_P) 29222 A C expression for the cost of a branch instruction. A value of 1 29223 is the default; other values are interpreted relative to that. 29224 Parameter SPEED_P is true when the branch in question should be 29225 optimized for speed. When it is false, `BRANCH_COST' should be 29226 returning value optimal for code size rather then performance 29227 considerations. PREDICTABLE_P is true for well predictable 29228 branches. On many architectures the `BRANCH_COST' can be reduced 29229 then. 29230 29231 Here are additional macros which do not specify precise relative costs, 29232 but only that certain actions are more expensive than GCC would 29233 ordinarily expect. 29234 29235 -- Macro: SLOW_BYTE_ACCESS 29236 Define this macro as a C expression which is nonzero if accessing 29237 less than a word of memory (i.e. a `char' or a `short') is no 29238 faster than accessing a word of memory, i.e., if such access 29239 require more than one instruction or if there is no difference in 29240 cost between byte and (aligned) word loads. 29241 29242 When this macro is not defined, the compiler will access a field by 29243 finding the smallest containing object; when it is defined, a 29244 fullword load will be used if alignment permits. Unless bytes 29245 accesses are faster than word accesses, using word accesses is 29246 preferable since it may eliminate subsequent memory access if 29247 subsequent accesses occur to other fields in the same word of the 29248 structure, but to different bytes. 29249 29250 -- Macro: SLOW_UNALIGNED_ACCESS (MODE, ALIGNMENT) 29251 Define this macro to be the value 1 if memory accesses described 29252 by the MODE and ALIGNMENT parameters have a cost many times greater 29253 than aligned accesses, for example if they are emulated in a trap 29254 handler. 29255 29256 When this macro is nonzero, the compiler will act as if 29257 `STRICT_ALIGNMENT' were nonzero when generating code for block 29258 moves. This can cause significantly more instructions to be 29259 produced. Therefore, do not set this macro nonzero if unaligned 29260 accesses only add a cycle or two to the time for a memory access. 29261 29262 If the value of this macro is always zero, it need not be defined. 29263 If this macro is defined, it should produce a nonzero value when 29264 `STRICT_ALIGNMENT' is nonzero. 29265 29266 -- Macro: MOVE_RATIO 29267 The threshold of number of scalar memory-to-memory move insns, 29268 _below_ which a sequence of insns should be generated instead of a 29269 string move insn or a library call. Increasing the value will 29270 always make code faster, but eventually incurs high cost in 29271 increased code size. 29272 29273 Note that on machines where the corresponding move insn is a 29274 `define_expand' that emits a sequence of insns, this macro counts 29275 the number of such sequences. 29276 29277 If you don't define this, a reasonable default is used. 29278 29279 -- Macro: MOVE_BY_PIECES_P (SIZE, ALIGNMENT) 29280 A C expression used to determine whether `move_by_pieces' will be 29281 used to copy a chunk of memory, or whether some other block move 29282 mechanism will be used. Defaults to 1 if `move_by_pieces_ninsns' 29283 returns less than `MOVE_RATIO'. 29284 29285 -- Macro: MOVE_MAX_PIECES 29286 A C expression used by `move_by_pieces' to determine the largest 29287 unit a load or store used to copy memory is. Defaults to 29288 `MOVE_MAX'. 29289 29290 -- Macro: CLEAR_RATIO 29291 The threshold of number of scalar move insns, _below_ which a 29292 sequence of insns should be generated to clear memory instead of a 29293 string clear insn or a library call. Increasing the value will 29294 always make code faster, but eventually incurs high cost in 29295 increased code size. 29296 29297 If you don't define this, a reasonable default is used. 29298 29299 -- Macro: CLEAR_BY_PIECES_P (SIZE, ALIGNMENT) 29300 A C expression used to determine whether `clear_by_pieces' will be 29301 used to clear a chunk of memory, or whether some other block clear 29302 mechanism will be used. Defaults to 1 if `move_by_pieces_ninsns' 29303 returns less than `CLEAR_RATIO'. 29304 29305 -- Macro: SET_RATIO 29306 The threshold of number of scalar move insns, _below_ which a 29307 sequence of insns should be generated to set memory to a constant 29308 value, instead of a block set insn or a library call. Increasing 29309 the value will always make code faster, but eventually incurs high 29310 cost in increased code size. 29311 29312 If you don't define this, it defaults to the value of `MOVE_RATIO'. 29313 29314 -- Macro: SET_BY_PIECES_P (SIZE, ALIGNMENT) 29315 A C expression used to determine whether `store_by_pieces' will be 29316 used to set a chunk of memory to a constant value, or whether some 29317 other mechanism will be used. Used by `__builtin_memset' when 29318 storing values other than constant zero. Defaults to 1 if 29319 `move_by_pieces_ninsns' returns less than `SET_RATIO'. 29320 29321 -- Macro: STORE_BY_PIECES_P (SIZE, ALIGNMENT) 29322 A C expression used to determine whether `store_by_pieces' will be 29323 used to set a chunk of memory to a constant string value, or 29324 whether some other mechanism will be used. Used by 29325 `__builtin_strcpy' when called with a constant source string. 29326 Defaults to 1 if `move_by_pieces_ninsns' returns less than 29327 `MOVE_RATIO'. 29328 29329 -- Macro: USE_LOAD_POST_INCREMENT (MODE) 29330 A C expression used to determine whether a load postincrement is a 29331 good thing to use for a given mode. Defaults to the value of 29332 `HAVE_POST_INCREMENT'. 29333 29334 -- Macro: USE_LOAD_POST_DECREMENT (MODE) 29335 A C expression used to determine whether a load postdecrement is a 29336 good thing to use for a given mode. Defaults to the value of 29337 `HAVE_POST_DECREMENT'. 29338 29339 -- Macro: USE_LOAD_PRE_INCREMENT (MODE) 29340 A C expression used to determine whether a load preincrement is a 29341 good thing to use for a given mode. Defaults to the value of 29342 `HAVE_PRE_INCREMENT'. 29343 29344 -- Macro: USE_LOAD_PRE_DECREMENT (MODE) 29345 A C expression used to determine whether a load predecrement is a 29346 good thing to use for a given mode. Defaults to the value of 29347 `HAVE_PRE_DECREMENT'. 29348 29349 -- Macro: USE_STORE_POST_INCREMENT (MODE) 29350 A C expression used to determine whether a store postincrement is 29351 a good thing to use for a given mode. Defaults to the value of 29352 `HAVE_POST_INCREMENT'. 29353 29354 -- Macro: USE_STORE_POST_DECREMENT (MODE) 29355 A C expression used to determine whether a store postdecrement is 29356 a good thing to use for a given mode. Defaults to the value of 29357 `HAVE_POST_DECREMENT'. 29358 29359 -- Macro: USE_STORE_PRE_INCREMENT (MODE) 29360 This macro is used to determine whether a store preincrement is a 29361 good thing to use for a given mode. Defaults to the value of 29362 `HAVE_PRE_INCREMENT'. 29363 29364 -- Macro: USE_STORE_PRE_DECREMENT (MODE) 29365 This macro is used to determine whether a store predecrement is a 29366 good thing to use for a given mode. Defaults to the value of 29367 `HAVE_PRE_DECREMENT'. 29368 29369 -- Macro: NO_FUNCTION_CSE 29370 Define this macro if it is as good or better to call a constant 29371 function address than to call an address kept in a register. 29372 29373 -- Macro: RANGE_TEST_NON_SHORT_CIRCUIT 29374 Define this macro if a non-short-circuit operation produced by 29375 `fold_range_test ()' is optimal. This macro defaults to true if 29376 `BRANCH_COST' is greater than or equal to the value 2. 29377 29378 -- Target Hook: bool TARGET_RTX_COSTS (rtx X, int CODE, int 29379 OUTER_CODE, int *TOTAL) 29380 This target hook describes the relative costs of RTL expressions. 29381 29382 The cost may depend on the precise form of the expression, which is 29383 available for examination in X, and the rtx code of the expression 29384 in which it is contained, found in OUTER_CODE. CODE is the 29385 expression code--redundant, since it can be obtained with 29386 `GET_CODE (X)'. 29387 29388 In implementing this hook, you can use the construct 29389 `COSTS_N_INSNS (N)' to specify a cost equal to N fast instructions. 29390 29391 On entry to the hook, `*TOTAL' contains a default estimate for the 29392 cost of the expression. The hook should modify this value as 29393 necessary. Traditionally, the default costs are `COSTS_N_INSNS 29394 (5)' for multiplications, `COSTS_N_INSNS (7)' for division and 29395 modulus operations, and `COSTS_N_INSNS (1)' for all other 29396 operations. 29397 29398 When optimizing for code size, i.e. when `optimize_size' is 29399 nonzero, this target hook should be used to estimate the relative 29400 size cost of an expression, again relative to `COSTS_N_INSNS'. 29401 29402 The hook returns true when all subexpressions of X have been 29403 processed, and false when `rtx_cost' should recurse. 29404 29405 -- Target Hook: int TARGET_ADDRESS_COST (rtx ADDRESS) 29406 This hook computes the cost of an addressing mode that contains 29407 ADDRESS. If not defined, the cost is computed from the ADDRESS 29408 expression and the `TARGET_RTX_COST' hook. 29409 29410 For most CISC machines, the default cost is a good approximation 29411 of the true cost of the addressing mode. However, on RISC 29412 machines, all instructions normally have the same length and 29413 execution time. Hence all addresses will have equal costs. 29414 29415 In cases where more than one form of an address is known, the form 29416 with the lowest cost will be used. If multiple forms have the 29417 same, lowest, cost, the one that is the most complex will be used. 29418 29419 For example, suppose an address that is equal to the sum of a 29420 register and a constant is used twice in the same basic block. 29421 When this macro is not defined, the address will be computed in a 29422 register and memory references will be indirect through that 29423 register. On machines where the cost of the addressing mode 29424 containing the sum is no higher than that of a simple indirect 29425 reference, this will produce an additional instruction and 29426 possibly require an additional register. Proper specification of 29427 this macro eliminates this overhead for such machines. 29428 29429 This hook is never called with an invalid address. 29430 29431 On machines where an address involving more than one register is as 29432 cheap as an address computation involving only one register, 29433 defining `TARGET_ADDRESS_COST' to reflect this can cause two 29434 registers to be live over a region of code where only one would 29435 have been if `TARGET_ADDRESS_COST' were not defined in that 29436 manner. This effect should be considered in the definition of 29437 this macro. Equivalent costs should probably only be given to 29438 addresses with different numbers of registers on machines with 29439 lots of registers. 29440 29441 29442 File: gccint.info, Node: Scheduling, Next: Sections, Prev: Costs, Up: Target Macros 29443 29444 17.18 Adjusting the Instruction Scheduler 29445 ========================================= 29446 29447 The instruction scheduler may need a fair amount of machine-specific 29448 adjustment in order to produce good code. GCC provides several target 29449 hooks for this purpose. It is usually enough to define just a few of 29450 them: try the first ones in this list first. 29451 29452 -- Target Hook: int TARGET_SCHED_ISSUE_RATE (void) 29453 This hook returns the maximum number of instructions that can ever 29454 issue at the same time on the target machine. The default is one. 29455 Although the insn scheduler can define itself the possibility of 29456 issue an insn on the same cycle, the value can serve as an 29457 additional constraint to issue insns on the same simulated 29458 processor cycle (see hooks `TARGET_SCHED_REORDER' and 29459 `TARGET_SCHED_REORDER2'). This value must be constant over the 29460 entire compilation. If you need it to vary depending on what the 29461 instructions are, you must use `TARGET_SCHED_VARIABLE_ISSUE'. 29462 29463 -- Target Hook: int TARGET_SCHED_VARIABLE_ISSUE (FILE *FILE, int 29464 VERBOSE, rtx INSN, int MORE) 29465 This hook is executed by the scheduler after it has scheduled an 29466 insn from the ready list. It should return the number of insns 29467 which can still be issued in the current cycle. The default is 29468 `MORE - 1' for insns other than `CLOBBER' and `USE', which 29469 normally are not counted against the issue rate. You should 29470 define this hook if some insns take more machine resources than 29471 others, so that fewer insns can follow them in the same cycle. 29472 FILE is either a null pointer, or a stdio stream to write any 29473 debug output to. VERBOSE is the verbose level provided by 29474 `-fsched-verbose-N'. INSN is the instruction that was scheduled. 29475 29476 -- Target Hook: int TARGET_SCHED_ADJUST_COST (rtx INSN, rtx LINK, rtx 29477 DEP_INSN, int COST) 29478 This function corrects the value of COST based on the relationship 29479 between INSN and DEP_INSN through the dependence LINK. It should 29480 return the new value. The default is to make no adjustment to 29481 COST. This can be used for example to specify to the scheduler 29482 using the traditional pipeline description that an output- or 29483 anti-dependence does not incur the same cost as a data-dependence. 29484 If the scheduler using the automaton based pipeline description, 29485 the cost of anti-dependence is zero and the cost of 29486 output-dependence is maximum of one and the difference of latency 29487 times of the first and the second insns. If these values are not 29488 acceptable, you could use the hook to modify them too. See also 29489 *note Processor pipeline description::. 29490 29491 -- Target Hook: int TARGET_SCHED_ADJUST_PRIORITY (rtx INSN, int 29492 PRIORITY) 29493 This hook adjusts the integer scheduling priority PRIORITY of 29494 INSN. It should return the new priority. Increase the priority to 29495 execute INSN earlier, reduce the priority to execute INSN later. 29496 Do not define this hook if you do not need to adjust the 29497 scheduling priorities of insns. 29498 29499 -- Target Hook: int TARGET_SCHED_REORDER (FILE *FILE, int VERBOSE, rtx 29500 *READY, int *N_READYP, int CLOCK) 29501 This hook is executed by the scheduler after it has scheduled the 29502 ready list, to allow the machine description to reorder it (for 29503 example to combine two small instructions together on `VLIW' 29504 machines). FILE is either a null pointer, or a stdio stream to 29505 write any debug output to. VERBOSE is the verbose level provided 29506 by `-fsched-verbose-N'. READY is a pointer to the ready list of 29507 instructions that are ready to be scheduled. N_READYP is a 29508 pointer to the number of elements in the ready list. The scheduler 29509 reads the ready list in reverse order, starting with 29510 READY[*N_READYP-1] and going to READY[0]. CLOCK is the timer tick 29511 of the scheduler. You may modify the ready list and the number of 29512 ready insns. The return value is the number of insns that can 29513 issue this cycle; normally this is just `issue_rate'. See also 29514 `TARGET_SCHED_REORDER2'. 29515 29516 -- Target Hook: int TARGET_SCHED_REORDER2 (FILE *FILE, int VERBOSE, 29517 rtx *READY, int *N_READY, CLOCK) 29518 Like `TARGET_SCHED_REORDER', but called at a different time. That 29519 function is called whenever the scheduler starts a new cycle. 29520 This one is called once per iteration over a cycle, immediately 29521 after `TARGET_SCHED_VARIABLE_ISSUE'; it can reorder the ready list 29522 and return the number of insns to be scheduled in the same cycle. 29523 Defining this hook can be useful if there are frequent situations 29524 where scheduling one insn causes other insns to become ready in 29525 the same cycle. These other insns can then be taken into account 29526 properly. 29527 29528 -- Target Hook: void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx 29529 HEAD, rtx TAIL) 29530 This hook is called after evaluation forward dependencies of insns 29531 in chain given by two parameter values (HEAD and TAIL 29532 correspondingly) but before insns scheduling of the insn chain. 29533 For example, it can be used for better insn classification if it 29534 requires analysis of dependencies. This hook can use backward and 29535 forward dependencies of the insn scheduler because they are already 29536 calculated. 29537 29538 -- Target Hook: void TARGET_SCHED_INIT (FILE *FILE, int VERBOSE, int 29539 MAX_READY) 29540 This hook is executed by the scheduler at the beginning of each 29541 block of instructions that are to be scheduled. FILE is either a 29542 null pointer, or a stdio stream to write any debug output to. 29543 VERBOSE is the verbose level provided by `-fsched-verbose-N'. 29544 MAX_READY is the maximum number of insns in the current scheduling 29545 region that can be live at the same time. This can be used to 29546 allocate scratch space if it is needed, e.g. by 29547 `TARGET_SCHED_REORDER'. 29548 29549 -- Target Hook: void TARGET_SCHED_FINISH (FILE *FILE, int VERBOSE) 29550 This hook is executed by the scheduler at the end of each block of 29551 instructions that are to be scheduled. It can be used to perform 29552 cleanup of any actions done by the other scheduling hooks. FILE 29553 is either a null pointer, or a stdio stream to write any debug 29554 output to. VERBOSE is the verbose level provided by 29555 `-fsched-verbose-N'. 29556 29557 -- Target Hook: void TARGET_SCHED_INIT_GLOBAL (FILE *FILE, int 29558 VERBOSE, int OLD_MAX_UID) 29559 This hook is executed by the scheduler after function level 29560 initializations. FILE is either a null pointer, or a stdio stream 29561 to write any debug output to. VERBOSE is the verbose level 29562 provided by `-fsched-verbose-N'. OLD_MAX_UID is the maximum insn 29563 uid when scheduling begins. 29564 29565 -- Target Hook: void TARGET_SCHED_FINISH_GLOBAL (FILE *FILE, int 29566 VERBOSE) 29567 This is the cleanup hook corresponding to 29568 `TARGET_SCHED_INIT_GLOBAL'. FILE is either a null pointer, or a 29569 stdio stream to write any debug output to. VERBOSE is the verbose 29570 level provided by `-fsched-verbose-N'. 29571 29572 -- Target Hook: int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void) 29573 The hook returns an RTL insn. The automaton state used in the 29574 pipeline hazard recognizer is changed as if the insn were scheduled 29575 when the new simulated processor cycle starts. Usage of the hook 29576 may simplify the automaton pipeline description for some VLIW 29577 processors. If the hook is defined, it is used only for the 29578 automaton based pipeline description. The default is not to 29579 change the state when the new simulated processor cycle starts. 29580 29581 -- Target Hook: void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void) 29582 The hook can be used to initialize data used by the previous hook. 29583 29584 -- Target Hook: int TARGET_SCHED_DFA_POST_CYCLE_INSN (void) 29585 The hook is analogous to `TARGET_SCHED_DFA_PRE_CYCLE_INSN' but used 29586 to changed the state as if the insn were scheduled when the new 29587 simulated processor cycle finishes. 29588 29589 -- Target Hook: void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void) 29590 The hook is analogous to `TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN' but 29591 used to initialize data used by the previous hook. 29592 29593 -- Target Hook: void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void) 29594 The hook to notify target that the current simulated cycle is 29595 about to finish. The hook is analogous to 29596 `TARGET_SCHED_DFA_PRE_CYCLE_INSN' but used to change the state in 29597 more complicated situations - e.g., when advancing state on a 29598 single insn is not enough. 29599 29600 -- Target Hook: void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void) 29601 The hook to notify target that new simulated cycle has just 29602 started. The hook is analogous to 29603 `TARGET_SCHED_DFA_POST_CYCLE_INSN' but used to change the state in 29604 more complicated situations - e.g., when advancing state on a 29605 single insn is not enough. 29606 29607 -- Target Hook: int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD 29608 (void) 29609 This hook controls better choosing an insn from the ready insn 29610 queue for the DFA-based insn scheduler. Usually the scheduler 29611 chooses the first insn from the queue. If the hook returns a 29612 positive value, an additional scheduler code tries all 29613 permutations of `TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD 29614 ()' subsequent ready insns to choose an insn whose issue will 29615 result in maximal number of issued insns on the same cycle. For 29616 the VLIW processor, the code could actually solve the problem of 29617 packing simple insns into the VLIW insn. Of course, if the rules 29618 of VLIW packing are described in the automaton. 29619 29620 This code also could be used for superscalar RISC processors. Let 29621 us consider a superscalar RISC processor with 3 pipelines. Some 29622 insns can be executed in pipelines A or B, some insns can be 29623 executed only in pipelines B or C, and one insn can be executed in 29624 pipeline B. The processor may issue the 1st insn into A and the 29625 2nd one into B. In this case, the 3rd insn will wait for freeing B 29626 until the next cycle. If the scheduler issues the 3rd insn the 29627 first, the processor could issue all 3 insns per cycle. 29628 29629 Actually this code demonstrates advantages of the automaton based 29630 pipeline hazard recognizer. We try quickly and easy many insn 29631 schedules to choose the best one. 29632 29633 The default is no multipass scheduling. 29634 29635 -- Target Hook: int 29636 TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx) 29637 This hook controls what insns from the ready insn queue will be 29638 considered for the multipass insn scheduling. If the hook returns 29639 zero for insn passed as the parameter, the insn will be not chosen 29640 to be issued. 29641 29642 The default is that any ready insns can be chosen to be issued. 29643 29644 -- Target Hook: int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, 29645 int, int *) 29646 This hook is called by the insn scheduler before issuing insn 29647 passed as the third parameter on given cycle. If the hook returns 29648 nonzero, the insn is not issued on given processors cycle. 29649 Instead of that, the processor cycle is advanced. If the value 29650 passed through the last parameter is zero, the insn ready queue is 29651 not sorted on the new cycle start as usually. The first parameter 29652 passes file for debugging output. The second one passes the 29653 scheduler verbose level of the debugging output. The forth and 29654 the fifth parameter values are correspondingly processor cycle on 29655 which the previous insn has been issued and the current processor 29656 cycle. 29657 29658 -- Target Hook: bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def 29659 *_DEP, int COST, int DISTANCE) 29660 This hook is used to define which dependences are considered 29661 costly by the target, so costly that it is not advisable to 29662 schedule the insns that are involved in the dependence too close 29663 to one another. The parameters to this hook are as follows: The 29664 first parameter _DEP is the dependence being evaluated. The 29665 second parameter COST is the cost of the dependence, and the third 29666 parameter DISTANCE is the distance in cycles between the two insns. 29667 The hook returns `true' if considering the distance between the two 29668 insns the dependence between them is considered costly by the 29669 target, and `false' otherwise. 29670 29671 Defining this hook can be useful in multiple-issue out-of-order 29672 machines, where (a) it's practically hopeless to predict the 29673 actual data/resource delays, however: (b) there's a better chance 29674 to predict the actual grouping that will be formed, and (c) 29675 correctly emulating the grouping can be very important. In such 29676 targets one may want to allow issuing dependent insns closer to 29677 one another--i.e., closer than the dependence distance; however, 29678 not in cases of "costly dependences", which this hooks allows to 29679 define. 29680 29681 -- Target Hook: void TARGET_SCHED_H_I_D_EXTENDED (void) 29682 This hook is called by the insn scheduler after emitting a new 29683 instruction to the instruction stream. The hook notifies a target 29684 backend to extend its per instruction data structures. 29685 29686 -- Target Hook: void * TARGET_SCHED_ALLOC_SCHED_CONTEXT (void) 29687 Return a pointer to a store large enough to hold target scheduling 29688 context. 29689 29690 -- Target Hook: void TARGET_SCHED_INIT_SCHED_CONTEXT (void *TC, bool 29691 CLEAN_P) 29692 Initialize store pointed to by TC to hold target scheduling 29693 context. It CLEAN_P is true then initialize TC as if scheduler is 29694 at the beginning of the block. Otherwise, make a copy of the 29695 current context in TC. 29696 29697 -- Target Hook: void TARGET_SCHED_SET_SCHED_CONTEXT (void *TC) 29698 Copy target scheduling context pointer to by TC to the current 29699 context. 29700 29701 -- Target Hook: void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *TC) 29702 Deallocate internal data in target scheduling context pointed to 29703 by TC. 29704 29705 -- Target Hook: void TARGET_SCHED_FREE_SCHED_CONTEXT (void *TC) 29706 Deallocate a store for target scheduling context pointed to by TC. 29707 29708 -- Target Hook: void * TARGET_SCHED_ALLOC_SCHED_CONTEXT (void) 29709 Return a pointer to a store large enough to hold target scheduling 29710 context. 29711 29712 -- Target Hook: void TARGET_SCHED_INIT_SCHED_CONTEXT (void *TC, bool 29713 CLEAN_P) 29714 Initialize store pointed to by TC to hold target scheduling 29715 context. It CLEAN_P is true then initialize TC as if scheduler is 29716 at the beginning of the block. Otherwise, make a copy of the 29717 current context in TC. 29718 29719 -- Target Hook: void TARGET_SCHED_SET_SCHED_CONTEXT (void *TC) 29720 Copy target scheduling context pointer to by TC to the current 29721 context. 29722 29723 -- Target Hook: void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *TC) 29724 Deallocate internal data in target scheduling context pointed to 29725 by TC. 29726 29727 -- Target Hook: void TARGET_SCHED_FREE_SCHED_CONTEXT (void *TC) 29728 Deallocate a store for target scheduling context pointed to by TC. 29729 29730 -- Target Hook: int TARGET_SCHED_SPECULATE_INSN (rtx INSN, int 29731 REQUEST, rtx *NEW_PAT) 29732 This hook is called by the insn scheduler when INSN has only 29733 speculative dependencies and therefore can be scheduled 29734 speculatively. The hook is used to check if the pattern of INSN 29735 has a speculative version and, in case of successful check, to 29736 generate that speculative pattern. The hook should return 1, if 29737 the instruction has a speculative form, or -1, if it doesn't. 29738 REQUEST describes the type of requested speculation. If the 29739 return value equals 1 then NEW_PAT is assigned the generated 29740 speculative pattern. 29741 29742 -- Target Hook: int TARGET_SCHED_NEEDS_BLOCK_P (rtx INSN) 29743 This hook is called by the insn scheduler during generation of 29744 recovery code for INSN. It should return nonzero, if the 29745 corresponding check instruction should branch to recovery code, or 29746 zero otherwise. 29747 29748 -- Target Hook: rtx TARGET_SCHED_GEN_CHECK (rtx INSN, rtx LABEL, int 29749 MUTATE_P) 29750 This hook is called by the insn scheduler to generate a pattern 29751 for recovery check instruction. If MUTATE_P is zero, then INSN is 29752 a speculative instruction for which the check should be generated. 29753 LABEL is either a label of a basic block, where recovery code 29754 should be emitted, or a null pointer, when requested check doesn't 29755 branch to recovery code (a simple check). If MUTATE_P is nonzero, 29756 then a pattern for a branchy check corresponding to a simple check 29757 denoted by INSN should be generated. In this case LABEL can't be 29758 null. 29759 29760 -- Target Hook: int 29761 TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx INSN) 29762 This hook is used as a workaround for 29763 `TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD' not being 29764 called on the first instruction of the ready list. The hook is 29765 used to discard speculative instruction that stand first in the 29766 ready list from being scheduled on the current cycle. For 29767 non-speculative instructions, the hook should always return 29768 nonzero. For example, in the ia64 backend the hook is used to 29769 cancel data speculative insns when the ALAT table is nearly full. 29770 29771 -- Target Hook: void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int 29772 *FLAGS, spec_info_t SPEC_INFO) 29773 This hook is used by the insn scheduler to find out what features 29774 should be enabled/used. FLAGS initially may have either the 29775 SCHED_RGN or SCHED_EBB bit set. This denotes the scheduler pass 29776 for which the data should be provided. The target backend should 29777 modify FLAGS by modifying the bits corresponding to the following 29778 features: USE_DEPS_LIST, USE_GLAT, DETACH_LIFE_INFO, and 29779 DO_SPECULATION. For the DO_SPECULATION feature an additional 29780 structure SPEC_INFO should be filled by the target. The structure 29781 describes speculation types that can be used in the scheduler. 29782 29783 -- Target Hook: int TARGET_SCHED_SMS_RES_MII (struct ddg *G) 29784 This hook is called by the swing modulo scheduler to calculate a 29785 resource-based lower bound which is based on the resources 29786 available in the machine and the resources required by each 29787 instruction. The target backend can use G to calculate such 29788 bound. A very simple lower bound will be used in case this hook 29789 is not implemented: the total number of instructions divided by 29790 the issue rate. 29791 29792 29793 File: gccint.info, Node: Sections, Next: PIC, Prev: Scheduling, Up: Target Macros 29794 29795 17.19 Dividing the Output into Sections (Texts, Data, ...) 29796 ========================================================== 29797 29798 An object file is divided into sections containing different types of 29799 data. In the most common case, there are three sections: the "text 29800 section", which holds instructions and read-only data; the "data 29801 section", which holds initialized writable data; and the "bss section", 29802 which holds uninitialized data. Some systems have other kinds of 29803 sections. 29804 29805 `varasm.c' provides several well-known sections, such as 29806 `text_section', `data_section' and `bss_section'. The normal way of 29807 controlling a `FOO_section' variable is to define the associated 29808 `FOO_SECTION_ASM_OP' macro, as described below. The macros are only 29809 read once, when `varasm.c' initializes itself, so their values must be 29810 run-time constants. They may however depend on command-line flags. 29811 29812 _Note:_ Some run-time files, such `crtstuff.c', also make use of the 29813 `FOO_SECTION_ASM_OP' macros, and expect them to be string literals. 29814 29815 Some assemblers require a different string to be written every time a 29816 section is selected. If your assembler falls into this category, you 29817 should define the `TARGET_ASM_INIT_SECTIONS' hook and use 29818 `get_unnamed_section' to set up the sections. 29819 29820 You must always create a `text_section', either by defining 29821 `TEXT_SECTION_ASM_OP' or by initializing `text_section' in 29822 `TARGET_ASM_INIT_SECTIONS'. The same is true of `data_section' and 29823 `DATA_SECTION_ASM_OP'. If you do not create a distinct 29824 `readonly_data_section', the default is to reuse `text_section'. 29825 29826 All the other `varasm.c' sections are optional, and are null if the 29827 target does not provide them. 29828 29829 -- Macro: TEXT_SECTION_ASM_OP 29830 A C expression whose value is a string, including spacing, 29831 containing the assembler operation that should precede 29832 instructions and read-only data. Normally `"\t.text"' is right. 29833 29834 -- Macro: HOT_TEXT_SECTION_NAME 29835 If defined, a C string constant for the name of the section 29836 containing most frequently executed functions of the program. If 29837 not defined, GCC will provide a default definition if the target 29838 supports named sections. 29839 29840 -- Macro: UNLIKELY_EXECUTED_TEXT_SECTION_NAME 29841 If defined, a C string constant for the name of the section 29842 containing unlikely executed functions in the program. 29843 29844 -- Macro: DATA_SECTION_ASM_OP 29845 A C expression whose value is a string, including spacing, 29846 containing the assembler operation to identify the following data 29847 as writable initialized data. Normally `"\t.data"' is right. 29848 29849 -- Macro: SDATA_SECTION_ASM_OP 29850 If defined, a C expression whose value is a string, including 29851 spacing, containing the assembler operation to identify the 29852 following data as initialized, writable small data. 29853 29854 -- Macro: READONLY_DATA_SECTION_ASM_OP 29855 A C expression whose value is a string, including spacing, 29856 containing the assembler operation to identify the following data 29857 as read-only initialized data. 29858 29859 -- Macro: BSS_SECTION_ASM_OP 29860 If defined, a C expression whose value is a string, including 29861 spacing, containing the assembler operation to identify the 29862 following data as uninitialized global data. If not defined, and 29863 neither `ASM_OUTPUT_BSS' nor `ASM_OUTPUT_ALIGNED_BSS' are defined, 29864 uninitialized global data will be output in the data section if 29865 `-fno-common' is passed, otherwise `ASM_OUTPUT_COMMON' will be 29866 used. 29867 29868 -- Macro: SBSS_SECTION_ASM_OP 29869 If defined, a C expression whose value is a string, including 29870 spacing, containing the assembler operation to identify the 29871 following data as uninitialized, writable small data. 29872 29873 -- Macro: INIT_SECTION_ASM_OP 29874 If defined, a C expression whose value is a string, including 29875 spacing, containing the assembler operation to identify the 29876 following data as initialization code. If not defined, GCC will 29877 assume such a section does not exist. This section has no 29878 corresponding `init_section' variable; it is used entirely in 29879 runtime code. 29880 29881 -- Macro: FINI_SECTION_ASM_OP 29882 If defined, a C expression whose value is a string, including 29883 spacing, containing the assembler operation to identify the 29884 following data as finalization code. If not defined, GCC will 29885 assume such a section does not exist. This section has no 29886 corresponding `fini_section' variable; it is used entirely in 29887 runtime code. 29888 29889 -- Macro: INIT_ARRAY_SECTION_ASM_OP 29890 If defined, a C expression whose value is a string, including 29891 spacing, containing the assembler operation to identify the 29892 following data as part of the `.init_array' (or equivalent) 29893 section. If not defined, GCC will assume such a section does not 29894 exist. Do not define both this macro and `INIT_SECTION_ASM_OP'. 29895 29896 -- Macro: FINI_ARRAY_SECTION_ASM_OP 29897 If defined, a C expression whose value is a string, including 29898 spacing, containing the assembler operation to identify the 29899 following data as part of the `.fini_array' (or equivalent) 29900 section. If not defined, GCC will assume such a section does not 29901 exist. Do not define both this macro and `FINI_SECTION_ASM_OP'. 29902 29903 -- Macro: CRT_CALL_STATIC_FUNCTION (SECTION_OP, FUNCTION) 29904 If defined, an ASM statement that switches to a different section 29905 via SECTION_OP, calls FUNCTION, and switches back to the text 29906 section. This is used in `crtstuff.c' if `INIT_SECTION_ASM_OP' or 29907 `FINI_SECTION_ASM_OP' to calls to initialization and finalization 29908 functions from the init and fini sections. By default, this macro 29909 uses a simple function call. Some ports need hand-crafted 29910 assembly code to avoid dependencies on registers initialized in 29911 the function prologue or to ensure that constant pools don't end 29912 up too far way in the text section. 29913 29914 -- Macro: TARGET_LIBGCC_SDATA_SECTION 29915 If defined, a string which names the section into which small 29916 variables defined in crtstuff and libgcc should go. This is useful 29917 when the target has options for optimizing access to small data, 29918 and you want the crtstuff and libgcc routines to be conservative 29919 in what they expect of your application yet liberal in what your 29920 application expects. For example, for targets with a `.sdata' 29921 section (like MIPS), you could compile crtstuff with `-G 0' so 29922 that it doesn't require small data support from your application, 29923 but use this macro to put small data into `.sdata' so that your 29924 application can access these variables whether it uses small data 29925 or not. 29926 29927 -- Macro: FORCE_CODE_SECTION_ALIGN 29928 If defined, an ASM statement that aligns a code section to some 29929 arbitrary boundary. This is used to force all fragments of the 29930 `.init' and `.fini' sections to have to same alignment and thus 29931 prevent the linker from having to add any padding. 29932 29933 -- Macro: JUMP_TABLES_IN_TEXT_SECTION 29934 Define this macro to be an expression with a nonzero value if jump 29935 tables (for `tablejump' insns) should be output in the text 29936 section, along with the assembler instructions. Otherwise, the 29937 readonly data section is used. 29938 29939 This macro is irrelevant if there is no separate readonly data 29940 section. 29941 29942 -- Target Hook: void TARGET_ASM_INIT_SECTIONS (void) 29943 Define this hook if you need to do something special to set up the 29944 `varasm.c' sections, or if your target has some special sections 29945 of its own that you need to create. 29946 29947 GCC calls this hook after processing the command line, but before 29948 writing any assembly code, and before calling any of the 29949 section-returning hooks described below. 29950 29951 -- Target Hook: TARGET_ASM_RELOC_RW_MASK (void) 29952 Return a mask describing how relocations should be treated when 29953 selecting sections. Bit 1 should be set if global relocations 29954 should be placed in a read-write section; bit 0 should be set if 29955 local relocations should be placed in a read-write section. 29956 29957 The default version of this function returns 3 when `-fpic' is in 29958 effect, and 0 otherwise. The hook is typically redefined when the 29959 target cannot support (some kinds of) dynamic relocations in 29960 read-only sections even in executables. 29961 29962 -- Target Hook: section * TARGET_ASM_SELECT_SECTION (tree EXP, int 29963 RELOC, unsigned HOST_WIDE_INT ALIGN) 29964 Return the section into which EXP should be placed. You can 29965 assume that EXP is either a `VAR_DECL' node or a constant of some 29966 sort. RELOC indicates whether the initial value of EXP requires 29967 link-time relocations. Bit 0 is set when variable contains local 29968 relocations only, while bit 1 is set for global relocations. 29969 ALIGN is the constant alignment in bits. 29970 29971 The default version of this function takes care of putting 29972 read-only variables in `readonly_data_section'. 29973 29974 See also USE_SELECT_SECTION_FOR_FUNCTIONS. 29975 29976 -- Macro: USE_SELECT_SECTION_FOR_FUNCTIONS 29977 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be 29978 called for `FUNCTION_DECL's as well as for variables and constants. 29979 29980 In the case of a `FUNCTION_DECL', RELOC will be zero if the 29981 function has been determined to be likely to be called, and 29982 nonzero if it is unlikely to be called. 29983 29984 -- Target Hook: void TARGET_ASM_UNIQUE_SECTION (tree DECL, int RELOC) 29985 Build up a unique section name, expressed as a `STRING_CST' node, 29986 and assign it to `DECL_SECTION_NAME (DECL)'. As with 29987 `TARGET_ASM_SELECT_SECTION', RELOC indicates whether the initial 29988 value of EXP requires link-time relocations. 29989 29990 The default version of this function appends the symbol name to the 29991 ELF section name that would normally be used for the symbol. For 29992 example, the function `foo' would be placed in `.text.foo'. 29993 Whatever the actual target object format, this is often good 29994 enough. 29995 29996 -- Target Hook: section * TARGET_ASM_FUNCTION_RODATA_SECTION (tree 29997 DECL) 29998 Return the readonly data section associated with 29999 `DECL_SECTION_NAME (DECL)'. The default version of this function 30000 selects `.gnu.linkonce.r.name' if the function's section is 30001 `.gnu.linkonce.t.name', `.rodata.name' if function is in 30002 `.text.name', and the normal readonly-data section otherwise. 30003 30004 -- Target Hook: section * TARGET_ASM_SELECT_RTX_SECTION (enum 30005 machine_mode MODE, rtx X, unsigned HOST_WIDE_INT ALIGN) 30006 Return the section into which a constant X, of mode MODE, should 30007 be placed. You can assume that X is some kind of constant in RTL. 30008 The argument MODE is redundant except in the case of a `const_int' 30009 rtx. ALIGN is the constant alignment in bits. 30010 30011 The default version of this function takes care of putting symbolic 30012 constants in `flag_pic' mode in `data_section' and everything else 30013 in `readonly_data_section'. 30014 30015 -- Target Hook: void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree DECL, 30016 tree ID) 30017 Define this hook if you need to postprocess the assembler name 30018 generated by target-independent code. The ID provided to this 30019 hook will be the computed name (e.g., the macro `DECL_NAME' of the 30020 DECL in C, or the mangled name of the DECL in C++). The return 30021 value of the hook is an `IDENTIFIER_NODE' for the appropriate 30022 mangled name on your target system. The default implementation of 30023 this hook just returns the ID provided. 30024 30025 -- Target Hook: void TARGET_ENCODE_SECTION_INFO (tree DECL, rtx RTL, 30026 int NEW_DECL_P) 30027 Define this hook if references to a symbol or a constant must be 30028 treated differently depending on something about the variable or 30029 function named by the symbol (such as what section it is in). 30030 30031 The hook is executed immediately after rtl has been created for 30032 DECL, which may be a variable or function declaration or an entry 30033 in the constant pool. In either case, RTL is the rtl in question. 30034 Do _not_ use `DECL_RTL (DECL)' in this hook; that field may not 30035 have been initialized yet. 30036 30037 In the case of a constant, it is safe to assume that the rtl is a 30038 `mem' whose address is a `symbol_ref'. Most decls will also have 30039 this form, but that is not guaranteed. Global register variables, 30040 for instance, will have a `reg' for their rtl. (Normally the 30041 right thing to do with such unusual rtl is leave it alone.) 30042 30043 The NEW_DECL_P argument will be true if this is the first time 30044 that `TARGET_ENCODE_SECTION_INFO' has been invoked on this decl. 30045 It will be false for subsequent invocations, which will happen for 30046 duplicate declarations. Whether or not anything must be done for 30047 the duplicate declaration depends on whether the hook examines 30048 `DECL_ATTRIBUTES'. NEW_DECL_P is always true when the hook is 30049 called for a constant. 30050 30051 The usual thing for this hook to do is to record flags in the 30052 `symbol_ref', using `SYMBOL_REF_FLAG' or `SYMBOL_REF_FLAGS'. 30053 Historically, the name string was modified if it was necessary to 30054 encode more than one bit of information, but this practice is now 30055 discouraged; use `SYMBOL_REF_FLAGS'. 30056 30057 The default definition of this hook, `default_encode_section_info' 30058 in `varasm.c', sets a number of commonly-useful bits in 30059 `SYMBOL_REF_FLAGS'. Check whether the default does what you need 30060 before overriding it. 30061 30062 -- Target Hook: const char *TARGET_STRIP_NAME_ENCODING (const char 30063 *name) 30064 Decode NAME and return the real name part, sans the characters 30065 that `TARGET_ENCODE_SECTION_INFO' may have added. 30066 30067 -- Target Hook: bool TARGET_IN_SMALL_DATA_P (tree EXP) 30068 Returns true if EXP should be placed into a "small data" section. 30069 The default version of this hook always returns false. 30070 30071 -- Variable: Target Hook bool TARGET_HAVE_SRODATA_SECTION 30072 Contains the value true if the target places read-only "small 30073 data" into a separate section. The default value is false. 30074 30075 -- Target Hook: bool TARGET_BINDS_LOCAL_P (tree EXP) 30076 Returns true if EXP names an object for which name resolution 30077 rules must resolve to the current "module" (dynamic shared library 30078 or executable image). 30079 30080 The default version of this hook implements the name resolution 30081 rules for ELF, which has a looser model of global name binding 30082 than other currently supported object file formats. 30083 30084 -- Variable: Target Hook bool TARGET_HAVE_TLS 30085 Contains the value true if the target supports thread-local 30086 storage. The default value is false. 30087 30088 30089 File: gccint.info, Node: PIC, Next: Assembler Format, Prev: Sections, Up: Target Macros 30090 30091 17.20 Position Independent Code 30092 =============================== 30093 30094 This section describes macros that help implement generation of position 30095 independent code. Simply defining these macros is not enough to 30096 generate valid PIC; you must also add support to the macros 30097 `GO_IF_LEGITIMATE_ADDRESS' and `PRINT_OPERAND_ADDRESS', as well as 30098 `LEGITIMIZE_ADDRESS'. You must modify the definition of `movsi' to do 30099 something appropriate when the source operand contains a symbolic 30100 address. You may also need to alter the handling of switch statements 30101 so that they use relative addresses. 30102 30103 -- Macro: PIC_OFFSET_TABLE_REGNUM 30104 The register number of the register used to address a table of 30105 static data addresses in memory. In some cases this register is 30106 defined by a processor's "application binary interface" (ABI). 30107 When this macro is defined, RTL is generated for this register 30108 once, as with the stack pointer and frame pointer registers. If 30109 this macro is not defined, it is up to the machine-dependent files 30110 to allocate such a register (if necessary). Note that this 30111 register must be fixed when in use (e.g. when `flag_pic' is true). 30112 30113 -- Macro: PIC_OFFSET_TABLE_REG_CALL_CLOBBERED 30114 Define this macro if the register defined by 30115 `PIC_OFFSET_TABLE_REGNUM' is clobbered by calls. Do not define 30116 this macro if `PIC_OFFSET_TABLE_REGNUM' is not defined. 30117 30118 -- Macro: LEGITIMATE_PIC_OPERAND_P (X) 30119 A C expression that is nonzero if X is a legitimate immediate 30120 operand on the target machine when generating position independent 30121 code. You can assume that X satisfies `CONSTANT_P', so you need 30122 not check this. You can also assume FLAG_PIC is true, so you need 30123 not check it either. You need not define this macro if all 30124 constants (including `SYMBOL_REF') can be immediate operands when 30125 generating position independent code. 30126 30127 30128 File: gccint.info, Node: Assembler Format, Next: Debugging Info, Prev: PIC, Up: Target Macros 30129 30130 17.21 Defining the Output Assembler Language 30131 ============================================ 30132 30133 This section describes macros whose principal purpose is to describe how 30134 to write instructions in assembler language--rather than what the 30135 instructions do. 30136 30137 * Menu: 30138 30139 * File Framework:: Structural information for the assembler file. 30140 * Data Output:: Output of constants (numbers, strings, addresses). 30141 * Uninitialized Data:: Output of uninitialized variables. 30142 * Label Output:: Output and generation of labels. 30143 * Initialization:: General principles of initialization 30144 and termination routines. 30145 * Macros for Initialization:: 30146 Specific macros that control the handling of 30147 initialization and termination routines. 30148 * Instruction Output:: Output of actual instructions. 30149 * Dispatch Tables:: Output of jump tables. 30150 * Exception Region Output:: Output of exception region code. 30151 * Alignment Output:: Pseudo ops for alignment and skipping data. 30152 30153 30154 File: gccint.info, Node: File Framework, Next: Data Output, Up: Assembler Format 30155 30156 17.21.1 The Overall Framework of an Assembler File 30157 -------------------------------------------------- 30158 30159 This describes the overall framework of an assembly file. 30160 30161 -- Target Hook: void TARGET_ASM_FILE_START () 30162 Output to `asm_out_file' any text which the assembler expects to 30163 find at the beginning of a file. The default behavior is 30164 controlled by two flags, documented below. Unless your target's 30165 assembler is quite unusual, if you override the default, you 30166 should call `default_file_start' at some point in your target 30167 hook. This lets other target files rely on these variables. 30168 30169 -- Target Hook: bool TARGET_ASM_FILE_START_APP_OFF 30170 If this flag is true, the text of the macro `ASM_APP_OFF' will be 30171 printed as the very first line in the assembly file, unless 30172 `-fverbose-asm' is in effect. (If that macro has been defined to 30173 the empty string, this variable has no effect.) With the normal 30174 definition of `ASM_APP_OFF', the effect is to notify the GNU 30175 assembler that it need not bother stripping comments or extra 30176 whitespace from its input. This allows it to work a bit faster. 30177 30178 The default is false. You should not set it to true unless you 30179 have verified that your port does not generate any extra 30180 whitespace or comments that will cause GAS to issue errors in 30181 NO_APP mode. 30182 30183 -- Target Hook: bool TARGET_ASM_FILE_START_FILE_DIRECTIVE 30184 If this flag is true, `output_file_directive' will be called for 30185 the primary source file, immediately after printing `ASM_APP_OFF' 30186 (if that is enabled). Most ELF assemblers expect this to be done. 30187 The default is false. 30188 30189 -- Target Hook: void TARGET_ASM_FILE_END () 30190 Output to `asm_out_file' any text which the assembler expects to 30191 find at the end of a file. The default is to output nothing. 30192 30193 -- Function: void file_end_indicate_exec_stack () 30194 Some systems use a common convention, the `.note.GNU-stack' 30195 special section, to indicate whether or not an object file relies 30196 on the stack being executable. If your system uses this 30197 convention, you should define `TARGET_ASM_FILE_END' to this 30198 function. If you need to do other things in that hook, have your 30199 hook function call this function. 30200 30201 -- Macro: ASM_COMMENT_START 30202 A C string constant describing how to begin a comment in the target 30203 assembler language. The compiler assumes that the comment will 30204 end at the end of the line. 30205 30206 -- Macro: ASM_APP_ON 30207 A C string constant for text to be output before each `asm' 30208 statement or group of consecutive ones. Normally this is 30209 `"#APP"', which is a comment that has no effect on most assemblers 30210 but tells the GNU assembler that it must check the lines that 30211 follow for all valid assembler constructs. 30212 30213 -- Macro: ASM_APP_OFF 30214 A C string constant for text to be output after each `asm' 30215 statement or group of consecutive ones. Normally this is 30216 `"#NO_APP"', which tells the GNU assembler to resume making the 30217 time-saving assumptions that are valid for ordinary compiler 30218 output. 30219 30220 -- Macro: ASM_OUTPUT_SOURCE_FILENAME (STREAM, NAME) 30221 A C statement to output COFF information or DWARF debugging 30222 information which indicates that filename NAME is the current 30223 source file to the stdio stream STREAM. 30224 30225 This macro need not be defined if the standard form of output for 30226 the file format in use is appropriate. 30227 30228 -- Macro: OUTPUT_QUOTED_STRING (STREAM, STRING) 30229 A C statement to output the string STRING to the stdio stream 30230 STREAM. If you do not call the function `output_quoted_string' in 30231 your config files, GCC will only call it to output filenames to 30232 the assembler source. So you can use it to canonicalize the format 30233 of the filename using this macro. 30234 30235 -- Macro: ASM_OUTPUT_IDENT (STREAM, STRING) 30236 A C statement to output something to the assembler file to handle a 30237 `#ident' directive containing the text STRING. If this macro is 30238 not defined, nothing is output for a `#ident' directive. 30239 30240 -- Target Hook: void TARGET_ASM_NAMED_SECTION (const char *NAME, 30241 unsigned int FLAGS, unsigned int ALIGN) 30242 Output assembly directives to switch to section NAME. The section 30243 should have attributes as specified by FLAGS, which is a bit mask 30244 of the `SECTION_*' flags defined in `output.h'. If ALIGN is 30245 nonzero, it contains an alignment in bytes to be used for the 30246 section, otherwise some target default should be used. Only 30247 targets that must specify an alignment within the section 30248 directive need pay attention to ALIGN - we will still use 30249 `ASM_OUTPUT_ALIGN'. 30250 30251 -- Target Hook: bool TARGET_HAVE_NAMED_SECTIONS 30252 This flag is true if the target supports 30253 `TARGET_ASM_NAMED_SECTION'. 30254 30255 -- Target Hook: bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS 30256 This flag is true if we can create zeroed data by switching to a 30257 BSS section and then using `ASM_OUTPUT_SKIP' to allocate the space. 30258 This is true on most ELF targets. 30259 30260 -- Target Hook: unsigned int TARGET_SECTION_TYPE_FLAGS (tree DECL, 30261 const char *NAME, int RELOC) 30262 Choose a set of section attributes for use by 30263 `TARGET_ASM_NAMED_SECTION' based on a variable or function decl, a 30264 section name, and whether or not the declaration's initializer may 30265 contain runtime relocations. DECL may be null, in which case 30266 read-write data should be assumed. 30267 30268 The default version of this function handles choosing code vs data, 30269 read-only vs read-write data, and `flag_pic'. You should only 30270 need to override this if your target has special flags that might 30271 be set via `__attribute__'. 30272 30273 -- Target Hook: int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type 30274 TYPE, const char * TEXT) 30275 Provides the target with the ability to record the gcc command line 30276 switches that have been passed to the compiler, and options that 30277 are enabled. The TYPE argument specifies what is being recorded. 30278 It can take the following values: 30279 30280 `SWITCH_TYPE_PASSED' 30281 TEXT is a command line switch that has been set by the user. 30282 30283 `SWITCH_TYPE_ENABLED' 30284 TEXT is an option which has been enabled. This might be as a 30285 direct result of a command line switch, or because it is 30286 enabled by default or because it has been enabled as a side 30287 effect of a different command line switch. For example, the 30288 `-O2' switch enables various different individual 30289 optimization passes. 30290 30291 `SWITCH_TYPE_DESCRIPTIVE' 30292 TEXT is either NULL or some descriptive text which should be 30293 ignored. If TEXT is NULL then it is being used to warn the 30294 target hook that either recording is starting or ending. The 30295 first time TYPE is SWITCH_TYPE_DESCRIPTIVE and TEXT is NULL, 30296 the warning is for start up and the second time the warning 30297 is for wind down. This feature is to allow the target hook 30298 to make any necessary preparations before it starts to record 30299 switches and to perform any necessary tidying up after it has 30300 finished recording switches. 30301 30302 `SWITCH_TYPE_LINE_START' 30303 This option can be ignored by this target hook. 30304 30305 `SWITCH_TYPE_LINE_END' 30306 This option can be ignored by this target hook. 30307 30308 The hook's return value must be zero. Other return values may be 30309 supported in the future. 30310 30311 By default this hook is set to NULL, but an example implementation 30312 is provided for ELF based targets. Called ELF_RECORD_GCC_SWITCHES, 30313 it records the switches as ASCII text inside a new, string 30314 mergeable section in the assembler output file. The name of the 30315 new section is provided by the 30316 `TARGET_ASM_RECORD_GCC_SWITCHES_SECTION' target hook. 30317 30318 -- Target Hook: const char * TARGET_ASM_RECORD_GCC_SWITCHES_SECTION 30319 This is the name of the section that will be created by the example 30320 ELF implementation of the `TARGET_ASM_RECORD_GCC_SWITCHES' target 30321 hook. 30322 30323 30324 File: gccint.info, Node: Data Output, Next: Uninitialized Data, Prev: File Framework, Up: Assembler Format 30325 30326 17.21.2 Output of Data 30327 ---------------------- 30328 30329 -- Target Hook: const char * TARGET_ASM_BYTE_OP 30330 -- Target Hook: const char * TARGET_ASM_ALIGNED_HI_OP 30331 -- Target Hook: const char * TARGET_ASM_ALIGNED_SI_OP 30332 -- Target Hook: const char * TARGET_ASM_ALIGNED_DI_OP 30333 -- Target Hook: const char * TARGET_ASM_ALIGNED_TI_OP 30334 -- Target Hook: const char * TARGET_ASM_UNALIGNED_HI_OP 30335 -- Target Hook: const char * TARGET_ASM_UNALIGNED_SI_OP 30336 -- Target Hook: const char * TARGET_ASM_UNALIGNED_DI_OP 30337 -- Target Hook: const char * TARGET_ASM_UNALIGNED_TI_OP 30338 These hooks specify assembly directives for creating certain kinds 30339 of integer object. The `TARGET_ASM_BYTE_OP' directive creates a 30340 byte-sized object, the `TARGET_ASM_ALIGNED_HI_OP' one creates an 30341 aligned two-byte object, and so on. Any of the hooks may be 30342 `NULL', indicating that no suitable directive is available. 30343 30344 The compiler will print these strings at the start of a new line, 30345 followed immediately by the object's initial value. In most cases, 30346 the string should contain a tab, a pseudo-op, and then another tab. 30347 30348 -- Target Hook: bool TARGET_ASM_INTEGER (rtx X, unsigned int SIZE, int 30349 ALIGNED_P) 30350 The `assemble_integer' function uses this hook to output an 30351 integer object. X is the object's value, SIZE is its size in 30352 bytes and ALIGNED_P indicates whether it is aligned. The function 30353 should return `true' if it was able to output the object. If it 30354 returns false, `assemble_integer' will try to split the object 30355 into smaller parts. 30356 30357 The default implementation of this hook will use the 30358 `TARGET_ASM_BYTE_OP' family of strings, returning `false' when the 30359 relevant string is `NULL'. 30360 30361 -- Macro: OUTPUT_ADDR_CONST_EXTRA (STREAM, X, FAIL) 30362 A C statement to recognize RTX patterns that `output_addr_const' 30363 can't deal with, and output assembly code to STREAM corresponding 30364 to the pattern X. This may be used to allow machine-dependent 30365 `UNSPEC's to appear within constants. 30366 30367 If `OUTPUT_ADDR_CONST_EXTRA' fails to recognize a pattern, it must 30368 `goto fail', so that a standard error message is printed. If it 30369 prints an error message itself, by calling, for example, 30370 `output_operand_lossage', it may just complete normally. 30371 30372 -- Macro: ASM_OUTPUT_ASCII (STREAM, PTR, LEN) 30373 A C statement to output to the stdio stream STREAM an assembler 30374 instruction to assemble a string constant containing the LEN bytes 30375 at PTR. PTR will be a C expression of type `char *' and LEN a C 30376 expression of type `int'. 30377 30378 If the assembler has a `.ascii' pseudo-op as found in the Berkeley 30379 Unix assembler, do not define the macro `ASM_OUTPUT_ASCII'. 30380 30381 -- Macro: ASM_OUTPUT_FDESC (STREAM, DECL, N) 30382 A C statement to output word N of a function descriptor for DECL. 30383 This must be defined if `TARGET_VTABLE_USES_DESCRIPTORS' is 30384 defined, and is otherwise unused. 30385 30386 -- Macro: CONSTANT_POOL_BEFORE_FUNCTION 30387 You may define this macro as a C expression. You should define the 30388 expression to have a nonzero value if GCC should output the 30389 constant pool for a function before the code for the function, or 30390 a zero value if GCC should output the constant pool after the 30391 function. If you do not define this macro, the usual case, GCC 30392 will output the constant pool before the function. 30393 30394 -- Macro: ASM_OUTPUT_POOL_PROLOGUE (FILE, FUNNAME, FUNDECL, SIZE) 30395 A C statement to output assembler commands to define the start of 30396 the constant pool for a function. FUNNAME is a string giving the 30397 name of the function. Should the return type of the function be 30398 required, it can be obtained via FUNDECL. SIZE is the size, in 30399 bytes, of the constant pool that will be written immediately after 30400 this call. 30401 30402 If no constant-pool prefix is required, the usual case, this macro 30403 need not be defined. 30404 30405 -- Macro: ASM_OUTPUT_SPECIAL_POOL_ENTRY (FILE, X, MODE, ALIGN, 30406 LABELNO, JUMPTO) 30407 A C statement (with or without semicolon) to output a constant in 30408 the constant pool, if it needs special treatment. (This macro 30409 need not do anything for RTL expressions that can be output 30410 normally.) 30411 30412 The argument FILE is the standard I/O stream to output the 30413 assembler code on. X is the RTL expression for the constant to 30414 output, and MODE is the machine mode (in case X is a `const_int'). 30415 ALIGN is the required alignment for the value X; you should output 30416 an assembler directive to force this much alignment. 30417 30418 The argument LABELNO is a number to use in an internal label for 30419 the address of this pool entry. The definition of this macro is 30420 responsible for outputting the label definition at the proper 30421 place. Here is how to do this: 30422 30423 `(*targetm.asm_out.internal_label)' (FILE, "LC", LABELNO); 30424 30425 When you output a pool entry specially, you should end with a 30426 `goto' to the label JUMPTO. This will prevent the same pool entry 30427 from being output a second time in the usual manner. 30428 30429 You need not define this macro if it would do nothing. 30430 30431 -- Macro: ASM_OUTPUT_POOL_EPILOGUE (FILE FUNNAME FUNDECL SIZE) 30432 A C statement to output assembler commands to at the end of the 30433 constant pool for a function. FUNNAME is a string giving the name 30434 of the function. Should the return type of the function be 30435 required, you can obtain it via FUNDECL. SIZE is the size, in 30436 bytes, of the constant pool that GCC wrote immediately before this 30437 call. 30438 30439 If no constant-pool epilogue is required, the usual case, you need 30440 not define this macro. 30441 30442 -- Macro: IS_ASM_LOGICAL_LINE_SEPARATOR (C, STR) 30443 Define this macro as a C expression which is nonzero if C is used 30444 as a logical line separator by the assembler. STR points to the 30445 position in the string where C was found; this can be used if a 30446 line separator uses multiple characters. 30447 30448 If you do not define this macro, the default is that only the 30449 character `;' is treated as a logical line separator. 30450 30451 -- Target Hook: const char * TARGET_ASM_OPEN_PAREN 30452 -- Target Hook: const char * TARGET_ASM_CLOSE_PAREN 30453 These target hooks are C string constants, describing the syntax 30454 in the assembler for grouping arithmetic expressions. If not 30455 overridden, they default to normal parentheses, which is correct 30456 for most assemblers. 30457 30458 These macros are provided by `real.h' for writing the definitions of 30459 `ASM_OUTPUT_DOUBLE' and the like: 30460 30461 -- Macro: REAL_VALUE_TO_TARGET_SINGLE (X, L) 30462 -- Macro: REAL_VALUE_TO_TARGET_DOUBLE (X, L) 30463 -- Macro: REAL_VALUE_TO_TARGET_LONG_DOUBLE (X, L) 30464 -- Macro: REAL_VALUE_TO_TARGET_DECIMAL32 (X, L) 30465 -- Macro: REAL_VALUE_TO_TARGET_DECIMAL64 (X, L) 30466 -- Macro: REAL_VALUE_TO_TARGET_DECIMAL128 (X, L) 30467 These translate X, of type `REAL_VALUE_TYPE', to the target's 30468 floating point representation, and store its bit pattern in the 30469 variable L. For `REAL_VALUE_TO_TARGET_SINGLE' and 30470 `REAL_VALUE_TO_TARGET_DECIMAL32', this variable should be a simple 30471 `long int'. For the others, it should be an array of `long int'. 30472 The number of elements in this array is determined by the size of 30473 the desired target floating point data type: 32 bits of it go in 30474 each `long int' array element. Each array element holds 32 bits 30475 of the result, even if `long int' is wider than 32 bits on the 30476 host machine. 30477 30478 The array element values are designed so that you can print them 30479 out using `fprintf' in the order they should appear in the target 30480 machine's memory. 30481 30482 30483 File: gccint.info, Node: Uninitialized Data, Next: Label Output, Prev: Data Output, Up: Assembler Format 30484 30485 17.21.3 Output of Uninitialized Variables 30486 ----------------------------------------- 30487 30488 Each of the macros in this section is used to do the whole job of 30489 outputting a single uninitialized variable. 30490 30491 -- Macro: ASM_OUTPUT_COMMON (STREAM, NAME, SIZE, ROUNDED) 30492 A C statement (sans semicolon) to output to the stdio stream 30493 STREAM the assembler definition of a common-label named NAME whose 30494 size is SIZE bytes. The variable ROUNDED is the size rounded up 30495 to whatever alignment the caller wants. 30496 30497 Use the expression `assemble_name (STREAM, NAME)' to output the 30498 name itself; before and after that, output the additional 30499 assembler syntax for defining the name, and a newline. 30500 30501 This macro controls how the assembler definitions of uninitialized 30502 common global variables are output. 30503 30504 -- Macro: ASM_OUTPUT_ALIGNED_COMMON (STREAM, NAME, SIZE, ALIGNMENT) 30505 Like `ASM_OUTPUT_COMMON' except takes the required alignment as a 30506 separate, explicit argument. If you define this macro, it is used 30507 in place of `ASM_OUTPUT_COMMON', and gives you more flexibility in 30508 handling the required alignment of the variable. The alignment is 30509 specified as the number of bits. 30510 30511 -- Macro: ASM_OUTPUT_ALIGNED_DECL_COMMON (STREAM, DECL, NAME, SIZE, 30512 ALIGNMENT) 30513 Like `ASM_OUTPUT_ALIGNED_COMMON' except that DECL of the variable 30514 to be output, if there is one, or `NULL_TREE' if there is no 30515 corresponding variable. If you define this macro, GCC will use it 30516 in place of both `ASM_OUTPUT_COMMON' and 30517 `ASM_OUTPUT_ALIGNED_COMMON'. Define this macro when you need to 30518 see the variable's decl in order to chose what to output. 30519 30520 -- Macro: ASM_OUTPUT_BSS (STREAM, DECL, NAME, SIZE, ROUNDED) 30521 A C statement (sans semicolon) to output to the stdio stream 30522 STREAM the assembler definition of uninitialized global DECL named 30523 NAME whose size is SIZE bytes. The variable ROUNDED is the size 30524 rounded up to whatever alignment the caller wants. 30525 30526 Try to use function `asm_output_bss' defined in `varasm.c' when 30527 defining this macro. If unable, use the expression `assemble_name 30528 (STREAM, NAME)' to output the name itself; before and after that, 30529 output the additional assembler syntax for defining the name, and 30530 a newline. 30531 30532 There are two ways of handling global BSS. One is to define either 30533 this macro or its aligned counterpart, `ASM_OUTPUT_ALIGNED_BSS'. 30534 The other is to have `TARGET_ASM_SELECT_SECTION' return a 30535 switchable BSS section (*note 30536 TARGET_HAVE_SWITCHABLE_BSS_SECTIONS::). You do not need to do 30537 both. 30538 30539 Some languages do not have `common' data, and require a non-common 30540 form of global BSS in order to handle uninitialized globals 30541 efficiently. C++ is one example of this. However, if the target 30542 does not support global BSS, the front end may choose to make 30543 globals common in order to save space in the object file. 30544 30545 -- Macro: ASM_OUTPUT_ALIGNED_BSS (STREAM, DECL, NAME, SIZE, ALIGNMENT) 30546 Like `ASM_OUTPUT_BSS' except takes the required alignment as a 30547 separate, explicit argument. If you define this macro, it is used 30548 in place of `ASM_OUTPUT_BSS', and gives you more flexibility in 30549 handling the required alignment of the variable. The alignment is 30550 specified as the number of bits. 30551 30552 Try to use function `asm_output_aligned_bss' defined in file 30553 `varasm.c' when defining this macro. 30554 30555 -- Macro: ASM_OUTPUT_LOCAL (STREAM, NAME, SIZE, ROUNDED) 30556 A C statement (sans semicolon) to output to the stdio stream 30557 STREAM the assembler definition of a local-common-label named NAME 30558 whose size is SIZE bytes. The variable ROUNDED is the size 30559 rounded up to whatever alignment the caller wants. 30560 30561 Use the expression `assemble_name (STREAM, NAME)' to output the 30562 name itself; before and after that, output the additional 30563 assembler syntax for defining the name, and a newline. 30564 30565 This macro controls how the assembler definitions of uninitialized 30566 static variables are output. 30567 30568 -- Macro: ASM_OUTPUT_ALIGNED_LOCAL (STREAM, NAME, SIZE, ALIGNMENT) 30569 Like `ASM_OUTPUT_LOCAL' except takes the required alignment as a 30570 separate, explicit argument. If you define this macro, it is used 30571 in place of `ASM_OUTPUT_LOCAL', and gives you more flexibility in 30572 handling the required alignment of the variable. The alignment is 30573 specified as the number of bits. 30574 30575 -- Macro: ASM_OUTPUT_ALIGNED_DECL_LOCAL (STREAM, DECL, NAME, SIZE, 30576 ALIGNMENT) 30577 Like `ASM_OUTPUT_ALIGNED_DECL' except that DECL of the variable to 30578 be output, if there is one, or `NULL_TREE' if there is no 30579 corresponding variable. If you define this macro, GCC will use it 30580 in place of both `ASM_OUTPUT_DECL' and `ASM_OUTPUT_ALIGNED_DECL'. 30581 Define this macro when you need to see the variable's decl in 30582 order to chose what to output. 30583 30584 30585 File: gccint.info, Node: Label Output, Next: Initialization, Prev: Uninitialized Data, Up: Assembler Format 30586 30587 17.21.4 Output and Generation of Labels 30588 --------------------------------------- 30589 30590 This is about outputting labels. 30591 30592 -- Macro: ASM_OUTPUT_LABEL (STREAM, NAME) 30593 A C statement (sans semicolon) to output to the stdio stream 30594 STREAM the assembler definition of a label named NAME. Use the 30595 expression `assemble_name (STREAM, NAME)' to output the name 30596 itself; before and after that, output the additional assembler 30597 syntax for defining the name, and a newline. A default definition 30598 of this macro is provided which is correct for most systems. 30599 30600 -- Macro: ASM_OUTPUT_INTERNAL_LABEL (STREAM, NAME) 30601 Identical to `ASM_OUTPUT_LABEL', except that NAME is known to 30602 refer to a compiler-generated label. The default definition uses 30603 `assemble_name_raw', which is like `assemble_name' except that it 30604 is more efficient. 30605 30606 -- Macro: SIZE_ASM_OP 30607 A C string containing the appropriate assembler directive to 30608 specify the size of a symbol, without any arguments. On systems 30609 that use ELF, the default (in `config/elfos.h') is `"\t.size\t"'; 30610 on other systems, the default is not to define this macro. 30611 30612 Define this macro only if it is correct to use the default 30613 definitions of `ASM_OUTPUT_SIZE_DIRECTIVE' and 30614 `ASM_OUTPUT_MEASURED_SIZE' for your system. If you need your own 30615 custom definitions of those macros, or if you do not need explicit 30616 symbol sizes at all, do not define this macro. 30617 30618 -- Macro: ASM_OUTPUT_SIZE_DIRECTIVE (STREAM, NAME, SIZE) 30619 A C statement (sans semicolon) to output to the stdio stream 30620 STREAM a directive telling the assembler that the size of the 30621 symbol NAME is SIZE. SIZE is a `HOST_WIDE_INT'. If you define 30622 `SIZE_ASM_OP', a default definition of this macro is provided. 30623 30624 -- Macro: ASM_OUTPUT_MEASURED_SIZE (STREAM, NAME) 30625 A C statement (sans semicolon) to output to the stdio stream 30626 STREAM a directive telling the assembler to calculate the size of 30627 the symbol NAME by subtracting its address from the current 30628 address. 30629 30630 If you define `SIZE_ASM_OP', a default definition of this macro is 30631 provided. The default assumes that the assembler recognizes a 30632 special `.' symbol as referring to the current address, and can 30633 calculate the difference between this and another symbol. If your 30634 assembler does not recognize `.' or cannot do calculations with 30635 it, you will need to redefine `ASM_OUTPUT_MEASURED_SIZE' to use 30636 some other technique. 30637 30638 -- Macro: TYPE_ASM_OP 30639 A C string containing the appropriate assembler directive to 30640 specify the type of a symbol, without any arguments. On systems 30641 that use ELF, the default (in `config/elfos.h') is `"\t.type\t"'; 30642 on other systems, the default is not to define this macro. 30643 30644 Define this macro only if it is correct to use the default 30645 definition of `ASM_OUTPUT_TYPE_DIRECTIVE' for your system. If you 30646 need your own custom definition of this macro, or if you do not 30647 need explicit symbol types at all, do not define this macro. 30648 30649 -- Macro: TYPE_OPERAND_FMT 30650 A C string which specifies (using `printf' syntax) the format of 30651 the second operand to `TYPE_ASM_OP'. On systems that use ELF, the 30652 default (in `config/elfos.h') is `"@%s"'; on other systems, the 30653 default is not to define this macro. 30654 30655 Define this macro only if it is correct to use the default 30656 definition of `ASM_OUTPUT_TYPE_DIRECTIVE' for your system. If you 30657 need your own custom definition of this macro, or if you do not 30658 need explicit symbol types at all, do not define this macro. 30659 30660 -- Macro: ASM_OUTPUT_TYPE_DIRECTIVE (STREAM, TYPE) 30661 A C statement (sans semicolon) to output to the stdio stream 30662 STREAM a directive telling the assembler that the type of the 30663 symbol NAME is TYPE. TYPE is a C string; currently, that string 30664 is always either `"function"' or `"object"', but you should not 30665 count on this. 30666 30667 If you define `TYPE_ASM_OP' and `TYPE_OPERAND_FMT', a default 30668 definition of this macro is provided. 30669 30670 -- Macro: ASM_DECLARE_FUNCTION_NAME (STREAM, NAME, DECL) 30671 A C statement (sans semicolon) to output to the stdio stream 30672 STREAM any text necessary for declaring the name NAME of a 30673 function which is being defined. This macro is responsible for 30674 outputting the label definition (perhaps using 30675 `ASM_OUTPUT_LABEL'). The argument DECL is the `FUNCTION_DECL' 30676 tree node representing the function. 30677 30678 If this macro is not defined, then the function name is defined in 30679 the usual manner as a label (by means of `ASM_OUTPUT_LABEL'). 30680 30681 You may wish to use `ASM_OUTPUT_TYPE_DIRECTIVE' in the definition 30682 of this macro. 30683 30684 -- Macro: ASM_DECLARE_FUNCTION_SIZE (STREAM, NAME, DECL) 30685 A C statement (sans semicolon) to output to the stdio stream 30686 STREAM any text necessary for declaring the size of a function 30687 which is being defined. The argument NAME is the name of the 30688 function. The argument DECL is the `FUNCTION_DECL' tree node 30689 representing the function. 30690 30691 If this macro is not defined, then the function size is not 30692 defined. 30693 30694 You may wish to use `ASM_OUTPUT_MEASURED_SIZE' in the definition 30695 of this macro. 30696 30697 -- Macro: ASM_DECLARE_OBJECT_NAME (STREAM, NAME, DECL) 30698 A C statement (sans semicolon) to output to the stdio stream 30699 STREAM any text necessary for declaring the name NAME of an 30700 initialized variable which is being defined. This macro must 30701 output the label definition (perhaps using `ASM_OUTPUT_LABEL'). 30702 The argument DECL is the `VAR_DECL' tree node representing the 30703 variable. 30704 30705 If this macro is not defined, then the variable name is defined in 30706 the usual manner as a label (by means of `ASM_OUTPUT_LABEL'). 30707 30708 You may wish to use `ASM_OUTPUT_TYPE_DIRECTIVE' and/or 30709 `ASM_OUTPUT_SIZE_DIRECTIVE' in the definition of this macro. 30710 30711 -- Macro: ASM_DECLARE_CONSTANT_NAME (STREAM, NAME, EXP, SIZE) 30712 A C statement (sans semicolon) to output to the stdio stream 30713 STREAM any text necessary for declaring the name NAME of a 30714 constant which is being defined. This macro is responsible for 30715 outputting the label definition (perhaps using 30716 `ASM_OUTPUT_LABEL'). The argument EXP is the value of the 30717 constant, and SIZE is the size of the constant in bytes. NAME 30718 will be an internal label. 30719 30720 If this macro is not defined, then the NAME is defined in the 30721 usual manner as a label (by means of `ASM_OUTPUT_LABEL'). 30722 30723 You may wish to use `ASM_OUTPUT_TYPE_DIRECTIVE' in the definition 30724 of this macro. 30725 30726 -- Macro: ASM_DECLARE_REGISTER_GLOBAL (STREAM, DECL, REGNO, NAME) 30727 A C statement (sans semicolon) to output to the stdio stream 30728 STREAM any text necessary for claiming a register REGNO for a 30729 global variable DECL with name NAME. 30730 30731 If you don't define this macro, that is equivalent to defining it 30732 to do nothing. 30733 30734 -- Macro: ASM_FINISH_DECLARE_OBJECT (STREAM, DECL, TOPLEVEL, ATEND) 30735 A C statement (sans semicolon) to finish up declaring a variable 30736 name once the compiler has processed its initializer fully and 30737 thus has had a chance to determine the size of an array when 30738 controlled by an initializer. This is used on systems where it's 30739 necessary to declare something about the size of the object. 30740 30741 If you don't define this macro, that is equivalent to defining it 30742 to do nothing. 30743 30744 You may wish to use `ASM_OUTPUT_SIZE_DIRECTIVE' and/or 30745 `ASM_OUTPUT_MEASURED_SIZE' in the definition of this macro. 30746 30747 -- Target Hook: void TARGET_ASM_GLOBALIZE_LABEL (FILE *STREAM, const 30748 char *NAME) 30749 This target hook is a function to output to the stdio stream 30750 STREAM some commands that will make the label NAME global; that 30751 is, available for reference from other files. 30752 30753 The default implementation relies on a proper definition of 30754 `GLOBAL_ASM_OP'. 30755 30756 -- Target Hook: void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *STREAM, 30757 tree DECL) 30758 This target hook is a function to output to the stdio stream 30759 STREAM some commands that will make the name associated with DECL 30760 global; that is, available for reference from other files. 30761 30762 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL 30763 target hook. 30764 30765 -- Macro: ASM_WEAKEN_LABEL (STREAM, NAME) 30766 A C statement (sans semicolon) to output to the stdio stream 30767 STREAM some commands that will make the label NAME weak; that is, 30768 available for reference from other files but only used if no other 30769 definition is available. Use the expression `assemble_name 30770 (STREAM, NAME)' to output the name itself; before and after that, 30771 output the additional assembler syntax for making that name weak, 30772 and a newline. 30773 30774 If you don't define this macro or `ASM_WEAKEN_DECL', GCC will not 30775 support weak symbols and you should not define the `SUPPORTS_WEAK' 30776 macro. 30777 30778 -- Macro: ASM_WEAKEN_DECL (STREAM, DECL, NAME, VALUE) 30779 Combines (and replaces) the function of `ASM_WEAKEN_LABEL' and 30780 `ASM_OUTPUT_WEAK_ALIAS', allowing access to the associated function 30781 or variable decl. If VALUE is not `NULL', this C statement should 30782 output to the stdio stream STREAM assembler code which defines 30783 (equates) the weak symbol NAME to have the value VALUE. If VALUE 30784 is `NULL', it should output commands to make NAME weak. 30785 30786 -- Macro: ASM_OUTPUT_WEAKREF (STREAM, DECL, NAME, VALUE) 30787 Outputs a directive that enables NAME to be used to refer to 30788 symbol VALUE with weak-symbol semantics. `decl' is the 30789 declaration of `name'. 30790 30791 -- Macro: SUPPORTS_WEAK 30792 A C expression which evaluates to true if the target supports weak 30793 symbols. 30794 30795 If you don't define this macro, `defaults.h' provides a default 30796 definition. If either `ASM_WEAKEN_LABEL' or `ASM_WEAKEN_DECL' is 30797 defined, the default definition is `1'; otherwise, it is `0'. 30798 Define this macro if you want to control weak symbol support with 30799 a compiler flag such as `-melf'. 30800 30801 -- Macro: MAKE_DECL_ONE_ONLY (DECL) 30802 A C statement (sans semicolon) to mark DECL to be emitted as a 30803 public symbol such that extra copies in multiple translation units 30804 will be discarded by the linker. Define this macro if your object 30805 file format provides support for this concept, such as the `COMDAT' 30806 section flags in the Microsoft Windows PE/COFF format, and this 30807 support requires changes to DECL, such as putting it in a separate 30808 section. 30809 30810 -- Macro: SUPPORTS_ONE_ONLY 30811 A C expression which evaluates to true if the target supports 30812 one-only semantics. 30813 30814 If you don't define this macro, `varasm.c' provides a default 30815 definition. If `MAKE_DECL_ONE_ONLY' is defined, the default 30816 definition is `1'; otherwise, it is `0'. Define this macro if you 30817 want to control one-only symbol support with a compiler flag, or if 30818 setting the `DECL_ONE_ONLY' flag is enough to mark a declaration to 30819 be emitted as one-only. 30820 30821 -- Target Hook: void TARGET_ASM_ASSEMBLE_VISIBILITY (tree DECL, const 30822 char *VISIBILITY) 30823 This target hook is a function to output to ASM_OUT_FILE some 30824 commands that will make the symbol(s) associated with DECL have 30825 hidden, protected or internal visibility as specified by 30826 VISIBILITY. 30827 30828 -- Macro: TARGET_WEAK_NOT_IN_ARCHIVE_TOC 30829 A C expression that evaluates to true if the target's linker 30830 expects that weak symbols do not appear in a static archive's 30831 table of contents. The default is `0'. 30832 30833 Leaving weak symbols out of an archive's table of contents means 30834 that, if a symbol will only have a definition in one translation 30835 unit and will have undefined references from other translation 30836 units, that symbol should not be weak. Defining this macro to be 30837 nonzero will thus have the effect that certain symbols that would 30838 normally be weak (explicit template instantiations, and vtables 30839 for polymorphic classes with noninline key methods) will instead 30840 be nonweak. 30841 30842 The C++ ABI requires this macro to be zero. Define this macro for 30843 targets where full C++ ABI compliance is impossible and where 30844 linker restrictions require weak symbols to be left out of a 30845 static archive's table of contents. 30846 30847 -- Macro: ASM_OUTPUT_EXTERNAL (STREAM, DECL, NAME) 30848 A C statement (sans semicolon) to output to the stdio stream 30849 STREAM any text necessary for declaring the name of an external 30850 symbol named NAME which is referenced in this compilation but not 30851 defined. The value of DECL is the tree node for the declaration. 30852 30853 This macro need not be defined if it does not need to output 30854 anything. The GNU assembler and most Unix assemblers don't 30855 require anything. 30856 30857 -- Target Hook: void TARGET_ASM_EXTERNAL_LIBCALL (rtx SYMREF) 30858 This target hook is a function to output to ASM_OUT_FILE an 30859 assembler pseudo-op to declare a library function name external. 30860 The name of the library function is given by SYMREF, which is a 30861 `symbol_ref'. 30862 30863 -- Target Hook: void TARGET_ASM_MARK_DECL_PRESERVED (tree DECL) 30864 This target hook is a function to output to ASM_OUT_FILE an 30865 assembler directive to annotate used symbol. Darwin target use 30866 .no_dead_code_strip directive. 30867 30868 -- Macro: ASM_OUTPUT_LABELREF (STREAM, NAME) 30869 A C statement (sans semicolon) to output to the stdio stream 30870 STREAM a reference in assembler syntax to a label named NAME. 30871 This should add `_' to the front of the name, if that is customary 30872 on your operating system, as it is in most Berkeley Unix systems. 30873 This macro is used in `assemble_name'. 30874 30875 -- Macro: ASM_OUTPUT_SYMBOL_REF (STREAM, SYM) 30876 A C statement (sans semicolon) to output a reference to 30877 `SYMBOL_REF' SYM. If not defined, `assemble_name' will be used to 30878 output the name of the symbol. This macro may be used to modify 30879 the way a symbol is referenced depending on information encoded by 30880 `TARGET_ENCODE_SECTION_INFO'. 30881 30882 -- Macro: ASM_OUTPUT_LABEL_REF (STREAM, BUF) 30883 A C statement (sans semicolon) to output a reference to BUF, the 30884 result of `ASM_GENERATE_INTERNAL_LABEL'. If not defined, 30885 `assemble_name' will be used to output the name of the symbol. 30886 This macro is not used by `output_asm_label', or the `%l' 30887 specifier that calls it; the intention is that this macro should 30888 be set when it is necessary to output a label differently when its 30889 address is being taken. 30890 30891 -- Target Hook: void TARGET_ASM_INTERNAL_LABEL (FILE *STREAM, const 30892 char *PREFIX, unsigned long LABELNO) 30893 A function to output to the stdio stream STREAM a label whose name 30894 is made from the string PREFIX and the number LABELNO. 30895 30896 It is absolutely essential that these labels be distinct from the 30897 labels used for user-level functions and variables. Otherwise, 30898 certain programs will have name conflicts with internal labels. 30899 30900 It is desirable to exclude internal labels from the symbol table 30901 of the object file. Most assemblers have a naming convention for 30902 labels that should be excluded; on many systems, the letter `L' at 30903 the beginning of a label has this effect. You should find out what 30904 convention your system uses, and follow it. 30905 30906 The default version of this function utilizes 30907 `ASM_GENERATE_INTERNAL_LABEL'. 30908 30909 -- Macro: ASM_OUTPUT_DEBUG_LABEL (STREAM, PREFIX, NUM) 30910 A C statement to output to the stdio stream STREAM a debug info 30911 label whose name is made from the string PREFIX and the number 30912 NUM. This is useful for VLIW targets, where debug info labels may 30913 need to be treated differently than branch target labels. On some 30914 systems, branch target labels must be at the beginning of 30915 instruction bundles, but debug info labels can occur in the middle 30916 of instruction bundles. 30917 30918 If this macro is not defined, then 30919 `(*targetm.asm_out.internal_label)' will be used. 30920 30921 -- Macro: ASM_GENERATE_INTERNAL_LABEL (STRING, PREFIX, NUM) 30922 A C statement to store into the string STRING a label whose name 30923 is made from the string PREFIX and the number NUM. 30924 30925 This string, when output subsequently by `assemble_name', should 30926 produce the output that `(*targetm.asm_out.internal_label)' would 30927 produce with the same PREFIX and NUM. 30928 30929 If the string begins with `*', then `assemble_name' will output 30930 the rest of the string unchanged. It is often convenient for 30931 `ASM_GENERATE_INTERNAL_LABEL' to use `*' in this way. If the 30932 string doesn't start with `*', then `ASM_OUTPUT_LABELREF' gets to 30933 output the string, and may change it. (Of course, 30934 `ASM_OUTPUT_LABELREF' is also part of your machine description, so 30935 you should know what it does on your machine.) 30936 30937 -- Macro: ASM_FORMAT_PRIVATE_NAME (OUTVAR, NAME, NUMBER) 30938 A C expression to assign to OUTVAR (which is a variable of type 30939 `char *') a newly allocated string made from the string NAME and 30940 the number NUMBER, with some suitable punctuation added. Use 30941 `alloca' to get space for the string. 30942 30943 The string will be used as an argument to `ASM_OUTPUT_LABELREF' to 30944 produce an assembler label for an internal static variable whose 30945 name is NAME. Therefore, the string must be such as to result in 30946 valid assembler code. The argument NUMBER is different each time 30947 this macro is executed; it prevents conflicts between 30948 similarly-named internal static variables in different scopes. 30949 30950 Ideally this string should not be a valid C identifier, to prevent 30951 any conflict with the user's own symbols. Most assemblers allow 30952 periods or percent signs in assembler symbols; putting at least 30953 one of these between the name and the number will suffice. 30954 30955 If this macro is not defined, a default definition will be provided 30956 which is correct for most systems. 30957 30958 -- Macro: ASM_OUTPUT_DEF (STREAM, NAME, VALUE) 30959 A C statement to output to the stdio stream STREAM assembler code 30960 which defines (equates) the symbol NAME to have the value VALUE. 30961 30962 If `SET_ASM_OP' is defined, a default definition is provided which 30963 is correct for most systems. 30964 30965 -- Macro: ASM_OUTPUT_DEF_FROM_DECLS (STREAM, DECL_OF_NAME, 30966 DECL_OF_VALUE) 30967 A C statement to output to the stdio stream STREAM assembler code 30968 which defines (equates) the symbol whose tree node is DECL_OF_NAME 30969 to have the value of the tree node DECL_OF_VALUE. This macro will 30970 be used in preference to `ASM_OUTPUT_DEF' if it is defined and if 30971 the tree nodes are available. 30972 30973 If `SET_ASM_OP' is defined, a default definition is provided which 30974 is correct for most systems. 30975 30976 -- Macro: TARGET_DEFERRED_OUTPUT_DEFS (DECL_OF_NAME, DECL_OF_VALUE) 30977 A C statement that evaluates to true if the assembler code which 30978 defines (equates) the symbol whose tree node is DECL_OF_NAME to 30979 have the value of the tree node DECL_OF_VALUE should be emitted 30980 near the end of the current compilation unit. The default is to 30981 not defer output of defines. This macro affects defines output by 30982 `ASM_OUTPUT_DEF' and `ASM_OUTPUT_DEF_FROM_DECLS'. 30983 30984 -- Macro: ASM_OUTPUT_WEAK_ALIAS (STREAM, NAME, VALUE) 30985 A C statement to output to the stdio stream STREAM assembler code 30986 which defines (equates) the weak symbol NAME to have the value 30987 VALUE. If VALUE is `NULL', it defines NAME as an undefined weak 30988 symbol. 30989 30990 Define this macro if the target only supports weak aliases; define 30991 `ASM_OUTPUT_DEF' instead if possible. 30992 30993 -- Macro: OBJC_GEN_METHOD_LABEL (BUF, IS_INST, CLASS_NAME, CAT_NAME, 30994 SEL_NAME) 30995 Define this macro to override the default assembler names used for 30996 Objective-C methods. 30997 30998 The default name is a unique method number followed by the name of 30999 the class (e.g. `_1_Foo'). For methods in categories, the name of 31000 the category is also included in the assembler name (e.g. 31001 `_1_Foo_Bar'). 31002 31003 These names are safe on most systems, but make debugging difficult 31004 since the method's selector is not present in the name. 31005 Therefore, particular systems define other ways of computing names. 31006 31007 BUF is an expression of type `char *' which gives you a buffer in 31008 which to store the name; its length is as long as CLASS_NAME, 31009 CAT_NAME and SEL_NAME put together, plus 50 characters extra. 31010 31011 The argument IS_INST specifies whether the method is an instance 31012 method or a class method; CLASS_NAME is the name of the class; 31013 CAT_NAME is the name of the category (or `NULL' if the method is 31014 not in a category); and SEL_NAME is the name of the selector. 31015 31016 On systems where the assembler can handle quoted names, you can 31017 use this macro to provide more human-readable names. 31018 31019 -- Macro: ASM_DECLARE_CLASS_REFERENCE (STREAM, NAME) 31020 A C statement (sans semicolon) to output to the stdio stream 31021 STREAM commands to declare that the label NAME is an Objective-C 31022 class reference. This is only needed for targets whose linkers 31023 have special support for NeXT-style runtimes. 31024 31025 -- Macro: ASM_DECLARE_UNRESOLVED_REFERENCE (STREAM, NAME) 31026 A C statement (sans semicolon) to output to the stdio stream 31027 STREAM commands to declare that the label NAME is an unresolved 31028 Objective-C class reference. This is only needed for targets 31029 whose linkers have special support for NeXT-style runtimes. 31030 31031 31032 File: gccint.info, Node: Initialization, Next: Macros for Initialization, Prev: Label Output, Up: Assembler Format 31033 31034 17.21.5 How Initialization Functions Are Handled 31035 ------------------------------------------------ 31036 31037 The compiled code for certain languages includes "constructors" (also 31038 called "initialization routines")--functions to initialize data in the 31039 program when the program is started. These functions need to be called 31040 before the program is "started"--that is to say, before `main' is 31041 called. 31042 31043 Compiling some languages generates "destructors" (also called 31044 "termination routines") that should be called when the program 31045 terminates. 31046 31047 To make the initialization and termination functions work, the compiler 31048 must output something in the assembler code to cause those functions to 31049 be called at the appropriate time. When you port the compiler to a new 31050 system, you need to specify how to do this. 31051 31052 There are two major ways that GCC currently supports the execution of 31053 initialization and termination functions. Each way has two variants. 31054 Much of the structure is common to all four variations. 31055 31056 The linker must build two lists of these functions--a list of 31057 initialization functions, called `__CTOR_LIST__', and a list of 31058 termination functions, called `__DTOR_LIST__'. 31059 31060 Each list always begins with an ignored function pointer (which may 31061 hold 0, -1, or a count of the function pointers after it, depending on 31062 the environment). This is followed by a series of zero or more function 31063 pointers to constructors (or destructors), followed by a function 31064 pointer containing zero. 31065 31066 Depending on the operating system and its executable file format, 31067 either `crtstuff.c' or `libgcc2.c' traverses these lists at startup 31068 time and exit time. Constructors are called in reverse order of the 31069 list; destructors in forward order. 31070 31071 The best way to handle static constructors works only for object file 31072 formats which provide arbitrarily-named sections. A section is set 31073 aside for a list of constructors, and another for a list of destructors. 31074 Traditionally these are called `.ctors' and `.dtors'. Each object file 31075 that defines an initialization function also puts a word in the 31076 constructor section to point to that function. The linker accumulates 31077 all these words into one contiguous `.ctors' section. Termination 31078 functions are handled similarly. 31079 31080 This method will be chosen as the default by `target-def.h' if 31081 `TARGET_ASM_NAMED_SECTION' is defined. A target that does not support 31082 arbitrary sections, but does support special designated constructor and 31083 destructor sections may define `CTORS_SECTION_ASM_OP' and 31084 `DTORS_SECTION_ASM_OP' to achieve the same effect. 31085 31086 When arbitrary sections are available, there are two variants, 31087 depending upon how the code in `crtstuff.c' is called. On systems that 31088 support a ".init" section which is executed at program startup, parts 31089 of `crtstuff.c' are compiled into that section. The program is linked 31090 by the `gcc' driver like this: 31091 31092 ld -o OUTPUT_FILE crti.o crtbegin.o ... -lgcc crtend.o crtn.o 31093 31094 The prologue of a function (`__init') appears in the `.init' section 31095 of `crti.o'; the epilogue appears in `crtn.o'. Likewise for the 31096 function `__fini' in the ".fini" section. Normally these files are 31097 provided by the operating system or by the GNU C library, but are 31098 provided by GCC for a few targets. 31099 31100 The objects `crtbegin.o' and `crtend.o' are (for most targets) 31101 compiled from `crtstuff.c'. They contain, among other things, code 31102 fragments within the `.init' and `.fini' sections that branch to 31103 routines in the `.text' section. The linker will pull all parts of a 31104 section together, which results in a complete `__init' function that 31105 invokes the routines we need at startup. 31106 31107 To use this variant, you must define the `INIT_SECTION_ASM_OP' macro 31108 properly. 31109 31110 If no init section is available, when GCC compiles any function called 31111 `main' (or more accurately, any function designated as a program entry 31112 point by the language front end calling `expand_main_function'), it 31113 inserts a procedure call to `__main' as the first executable code after 31114 the function prologue. The `__main' function is defined in `libgcc2.c' 31115 and runs the global constructors. 31116 31117 In file formats that don't support arbitrary sections, there are again 31118 two variants. In the simplest variant, the GNU linker (GNU `ld') and 31119 an `a.out' format must be used. In this case, `TARGET_ASM_CONSTRUCTOR' 31120 is defined to produce a `.stabs' entry of type `N_SETT', referencing 31121 the name `__CTOR_LIST__', and with the address of the void function 31122 containing the initialization code as its value. The GNU linker 31123 recognizes this as a request to add the value to a "set"; the values 31124 are accumulated, and are eventually placed in the executable as a 31125 vector in the format described above, with a leading (ignored) count 31126 and a trailing zero element. `TARGET_ASM_DESTRUCTOR' is handled 31127 similarly. Since no init section is available, the absence of 31128 `INIT_SECTION_ASM_OP' causes the compilation of `main' to call `__main' 31129 as above, starting the initialization process. 31130 31131 The last variant uses neither arbitrary sections nor the GNU linker. 31132 This is preferable when you want to do dynamic linking and when using 31133 file formats which the GNU linker does not support, such as `ECOFF'. In 31134 this case, `TARGET_HAVE_CTORS_DTORS' is false, initialization and 31135 termination functions are recognized simply by their names. This 31136 requires an extra program in the linkage step, called `collect2'. This 31137 program pretends to be the linker, for use with GCC; it does its job by 31138 running the ordinary linker, but also arranges to include the vectors of 31139 initialization and termination functions. These functions are called 31140 via `__main' as described above. In order to use this method, 31141 `use_collect2' must be defined in the target in `config.gcc'. 31142 31143 The following section describes the specific macros that control and 31144 customize the handling of initialization and termination functions. 31145 31146 31147 File: gccint.info, Node: Macros for Initialization, Next: Instruction Output, Prev: Initialization, Up: Assembler Format 31148 31149 17.21.6 Macros Controlling Initialization Routines 31150 -------------------------------------------------- 31151 31152 Here are the macros that control how the compiler handles initialization 31153 and termination functions: 31154 31155 -- Macro: INIT_SECTION_ASM_OP 31156 If defined, a C string constant, including spacing, for the 31157 assembler operation to identify the following data as 31158 initialization code. If not defined, GCC will assume such a 31159 section does not exist. When you are using special sections for 31160 initialization and termination functions, this macro also controls 31161 how `crtstuff.c' and `libgcc2.c' arrange to run the initialization 31162 functions. 31163 31164 -- Macro: HAS_INIT_SECTION 31165 If defined, `main' will not call `__main' as described above. 31166 This macro should be defined for systems that control start-up code 31167 on a symbol-by-symbol basis, such as OSF/1, and should not be 31168 defined explicitly for systems that support `INIT_SECTION_ASM_OP'. 31169 31170 -- Macro: LD_INIT_SWITCH 31171 If defined, a C string constant for a switch that tells the linker 31172 that the following symbol is an initialization routine. 31173 31174 -- Macro: LD_FINI_SWITCH 31175 If defined, a C string constant for a switch that tells the linker 31176 that the following symbol is a finalization routine. 31177 31178 -- Macro: COLLECT_SHARED_INIT_FUNC (STREAM, FUNC) 31179 If defined, a C statement that will write a function that can be 31180 automatically called when a shared library is loaded. The function 31181 should call FUNC, which takes no arguments. If not defined, and 31182 the object format requires an explicit initialization function, 31183 then a function called `_GLOBAL__DI' will be generated. 31184 31185 This function and the following one are used by collect2 when 31186 linking a shared library that needs constructors or destructors, 31187 or has DWARF2 exception tables embedded in the code. 31188 31189 -- Macro: COLLECT_SHARED_FINI_FUNC (STREAM, FUNC) 31190 If defined, a C statement that will write a function that can be 31191 automatically called when a shared library is unloaded. The 31192 function should call FUNC, which takes no arguments. If not 31193 defined, and the object format requires an explicit finalization 31194 function, then a function called `_GLOBAL__DD' will be generated. 31195 31196 -- Macro: INVOKE__main 31197 If defined, `main' will call `__main' despite the presence of 31198 `INIT_SECTION_ASM_OP'. This macro should be defined for systems 31199 where the init section is not actually run automatically, but is 31200 still useful for collecting the lists of constructors and 31201 destructors. 31202 31203 -- Macro: SUPPORTS_INIT_PRIORITY 31204 If nonzero, the C++ `init_priority' attribute is supported and the 31205 compiler should emit instructions to control the order of 31206 initialization of objects. If zero, the compiler will issue an 31207 error message upon encountering an `init_priority' attribute. 31208 31209 -- Target Hook: bool TARGET_HAVE_CTORS_DTORS 31210 This value is true if the target supports some "native" method of 31211 collecting constructors and destructors to be run at startup and 31212 exit. It is false if we must use `collect2'. 31213 31214 -- Target Hook: void TARGET_ASM_CONSTRUCTOR (rtx SYMBOL, int PRIORITY) 31215 If defined, a function that outputs assembler code to arrange to 31216 call the function referenced by SYMBOL at initialization time. 31217 31218 Assume that SYMBOL is a `SYMBOL_REF' for a function taking no 31219 arguments and with no return value. If the target supports 31220 initialization priorities, PRIORITY is a value between 0 and 31221 `MAX_INIT_PRIORITY'; otherwise it must be `DEFAULT_INIT_PRIORITY'. 31222 31223 If this macro is not defined by the target, a suitable default will 31224 be chosen if (1) the target supports arbitrary section names, (2) 31225 the target defines `CTORS_SECTION_ASM_OP', or (3) `USE_COLLECT2' 31226 is not defined. 31227 31228 -- Target Hook: void TARGET_ASM_DESTRUCTOR (rtx SYMBOL, int PRIORITY) 31229 This is like `TARGET_ASM_CONSTRUCTOR' but used for termination 31230 functions rather than initialization functions. 31231 31232 If `TARGET_HAVE_CTORS_DTORS' is true, the initialization routine 31233 generated for the generated object file will have static linkage. 31234 31235 If your system uses `collect2' as the means of processing 31236 constructors, then that program normally uses `nm' to scan an object 31237 file for constructor functions to be called. 31238 31239 On certain kinds of systems, you can define this macro to make 31240 `collect2' work faster (and, in some cases, make it work at all): 31241 31242 -- Macro: OBJECT_FORMAT_COFF 31243 Define this macro if the system uses COFF (Common Object File 31244 Format) object files, so that `collect2' can assume this format 31245 and scan object files directly for dynamic constructor/destructor 31246 functions. 31247 31248 This macro is effective only in a native compiler; `collect2' as 31249 part of a cross compiler always uses `nm' for the target machine. 31250 31251 -- Macro: REAL_NM_FILE_NAME 31252 Define this macro as a C string constant containing the file name 31253 to use to execute `nm'. The default is to search the path 31254 normally for `nm'. 31255 31256 If your system supports shared libraries and has a program to list 31257 the dynamic dependencies of a given library or executable, you can 31258 define these macros to enable support for running initialization 31259 and termination functions in shared libraries: 31260 31261 -- Macro: LDD_SUFFIX 31262 Define this macro to a C string constant containing the name of 31263 the program which lists dynamic dependencies, like `"ldd"' under 31264 SunOS 4. 31265 31266 -- Macro: PARSE_LDD_OUTPUT (PTR) 31267 Define this macro to be C code that extracts filenames from the 31268 output of the program denoted by `LDD_SUFFIX'. PTR is a variable 31269 of type `char *' that points to the beginning of a line of output 31270 from `LDD_SUFFIX'. If the line lists a dynamic dependency, the 31271 code must advance PTR to the beginning of the filename on that 31272 line. Otherwise, it must set PTR to `NULL'. 31273 31274 -- Macro: SHLIB_SUFFIX 31275 Define this macro to a C string constant containing the default 31276 shared library extension of the target (e.g., `".so"'). `collect2' 31277 strips version information after this suffix when generating global 31278 constructor and destructor names. This define is only needed on 31279 targets that use `collect2' to process constructors and 31280 destructors. 31281 31282 31283 File: gccint.info, Node: Instruction Output, Next: Dispatch Tables, Prev: Macros for Initialization, Up: Assembler Format 31284 31285 17.21.7 Output of Assembler Instructions 31286 ---------------------------------------- 31287 31288 This describes assembler instruction output. 31289 31290 -- Macro: REGISTER_NAMES 31291 A C initializer containing the assembler's names for the machine 31292 registers, each one as a C string constant. This is what 31293 translates register numbers in the compiler into assembler 31294 language. 31295 31296 -- Macro: ADDITIONAL_REGISTER_NAMES 31297 If defined, a C initializer for an array of structures containing 31298 a name and a register number. This macro defines additional names 31299 for hard registers, thus allowing the `asm' option in declarations 31300 to refer to registers using alternate names. 31301 31302 -- Macro: ASM_OUTPUT_OPCODE (STREAM, PTR) 31303 Define this macro if you are using an unusual assembler that 31304 requires different names for the machine instructions. 31305 31306 The definition is a C statement or statements which output an 31307 assembler instruction opcode to the stdio stream STREAM. The 31308 macro-operand PTR is a variable of type `char *' which points to 31309 the opcode name in its "internal" form--the form that is written 31310 in the machine description. The definition should output the 31311 opcode name to STREAM, performing any translation you desire, and 31312 increment the variable PTR to point at the end of the opcode so 31313 that it will not be output twice. 31314 31315 In fact, your macro definition may process less than the entire 31316 opcode name, or more than the opcode name; but if you want to 31317 process text that includes `%'-sequences to substitute operands, 31318 you must take care of the substitution yourself. Just be sure to 31319 increment PTR over whatever text should not be output normally. 31320 31321 If you need to look at the operand values, they can be found as the 31322 elements of `recog_data.operand'. 31323 31324 If the macro definition does nothing, the instruction is output in 31325 the usual way. 31326 31327 -- Macro: FINAL_PRESCAN_INSN (INSN, OPVEC, NOPERANDS) 31328 If defined, a C statement to be executed just prior to the output 31329 of assembler code for INSN, to modify the extracted operands so 31330 they will be output differently. 31331 31332 Here the argument OPVEC is the vector containing the operands 31333 extracted from INSN, and NOPERANDS is the number of elements of 31334 the vector which contain meaningful data for this insn. The 31335 contents of this vector are what will be used to convert the insn 31336 template into assembler code, so you can change the assembler 31337 output by changing the contents of the vector. 31338 31339 This macro is useful when various assembler syntaxes share a single 31340 file of instruction patterns; by defining this macro differently, 31341 you can cause a large class of instructions to be output 31342 differently (such as with rearranged operands). Naturally, 31343 variations in assembler syntax affecting individual insn patterns 31344 ought to be handled by writing conditional output routines in 31345 those patterns. 31346 31347 If this macro is not defined, it is equivalent to a null statement. 31348 31349 -- Macro: PRINT_OPERAND (STREAM, X, CODE) 31350 A C compound statement to output to stdio stream STREAM the 31351 assembler syntax for an instruction operand X. X is an RTL 31352 expression. 31353 31354 CODE is a value that can be used to specify one of several ways of 31355 printing the operand. It is used when identical operands must be 31356 printed differently depending on the context. CODE comes from the 31357 `%' specification that was used to request printing of the 31358 operand. If the specification was just `%DIGIT' then CODE is 0; 31359 if the specification was `%LTR DIGIT' then CODE is the ASCII code 31360 for LTR. 31361 31362 If X is a register, this macro should print the register's name. 31363 The names can be found in an array `reg_names' whose type is `char 31364 *[]'. `reg_names' is initialized from `REGISTER_NAMES'. 31365 31366 When the machine description has a specification `%PUNCT' (a `%' 31367 followed by a punctuation character), this macro is called with a 31368 null pointer for X and the punctuation character for CODE. 31369 31370 -- Macro: PRINT_OPERAND_PUNCT_VALID_P (CODE) 31371 A C expression which evaluates to true if CODE is a valid 31372 punctuation character for use in the `PRINT_OPERAND' macro. If 31373 `PRINT_OPERAND_PUNCT_VALID_P' is not defined, it means that no 31374 punctuation characters (except for the standard one, `%') are used 31375 in this way. 31376 31377 -- Macro: PRINT_OPERAND_ADDRESS (STREAM, X) 31378 A C compound statement to output to stdio stream STREAM the 31379 assembler syntax for an instruction operand that is a memory 31380 reference whose address is X. X is an RTL expression. 31381 31382 On some machines, the syntax for a symbolic address depends on the 31383 section that the address refers to. On these machines, define the 31384 hook `TARGET_ENCODE_SECTION_INFO' to store the information into the 31385 `symbol_ref', and then check for it here. *Note Assembler 31386 Format::. 31387 31388 -- Macro: DBR_OUTPUT_SEQEND (FILE) 31389 A C statement, to be executed after all slot-filler instructions 31390 have been output. If necessary, call `dbr_sequence_length' to 31391 determine the number of slots filled in a sequence (zero if not 31392 currently outputting a sequence), to decide how many no-ops to 31393 output, or whatever. 31394 31395 Don't define this macro if it has nothing to do, but it is helpful 31396 in reading assembly output if the extent of the delay sequence is 31397 made explicit (e.g. with white space). 31398 31399 Note that output routines for instructions with delay slots must be 31400 prepared to deal with not being output as part of a sequence (i.e. when 31401 the scheduling pass is not run, or when no slot fillers could be 31402 found.) The variable `final_sequence' is null when not processing a 31403 sequence, otherwise it contains the `sequence' rtx being output. 31404 31405 -- Macro: REGISTER_PREFIX 31406 -- Macro: LOCAL_LABEL_PREFIX 31407 -- Macro: USER_LABEL_PREFIX 31408 -- Macro: IMMEDIATE_PREFIX 31409 If defined, C string expressions to be used for the `%R', `%L', 31410 `%U', and `%I' options of `asm_fprintf' (see `final.c'). These 31411 are useful when a single `md' file must support multiple assembler 31412 formats. In that case, the various `tm.h' files can define these 31413 macros differently. 31414 31415 -- Macro: ASM_FPRINTF_EXTENSIONS (FILE, ARGPTR, FORMAT) 31416 If defined this macro should expand to a series of `case' 31417 statements which will be parsed inside the `switch' statement of 31418 the `asm_fprintf' function. This allows targets to define extra 31419 printf formats which may useful when generating their assembler 31420 statements. Note that uppercase letters are reserved for future 31421 generic extensions to asm_fprintf, and so are not available to 31422 target specific code. The output file is given by the parameter 31423 FILE. The varargs input pointer is ARGPTR and the rest of the 31424 format string, starting the character after the one that is being 31425 switched upon, is pointed to by FORMAT. 31426 31427 -- Macro: ASSEMBLER_DIALECT 31428 If your target supports multiple dialects of assembler language 31429 (such as different opcodes), define this macro as a C expression 31430 that gives the numeric index of the assembler language dialect to 31431 use, with zero as the first variant. 31432 31433 If this macro is defined, you may use constructs of the form 31434 `{option0|option1|option2...}' 31435 in the output templates of patterns (*note Output Template::) or 31436 in the first argument of `asm_fprintf'. This construct outputs 31437 `option0', `option1', `option2', etc., if the value of 31438 `ASSEMBLER_DIALECT' is zero, one, two, etc. Any special characters 31439 within these strings retain their usual meaning. If there are 31440 fewer alternatives within the braces than the value of 31441 `ASSEMBLER_DIALECT', the construct outputs nothing. 31442 31443 If you do not define this macro, the characters `{', `|' and `}' 31444 do not have any special meaning when used in templates or operands 31445 to `asm_fprintf'. 31446 31447 Define the macros `REGISTER_PREFIX', `LOCAL_LABEL_PREFIX', 31448 `USER_LABEL_PREFIX' and `IMMEDIATE_PREFIX' if you can express the 31449 variations in assembler language syntax with that mechanism. 31450 Define `ASSEMBLER_DIALECT' and use the `{option0|option1}' syntax 31451 if the syntax variant are larger and involve such things as 31452 different opcodes or operand order. 31453 31454 -- Macro: ASM_OUTPUT_REG_PUSH (STREAM, REGNO) 31455 A C expression to output to STREAM some assembler code which will 31456 push hard register number REGNO onto the stack. The code need not 31457 be optimal, since this macro is used only when profiling. 31458 31459 -- Macro: ASM_OUTPUT_REG_POP (STREAM, REGNO) 31460 A C expression to output to STREAM some assembler code which will 31461 pop hard register number REGNO off of the stack. The code need 31462 not be optimal, since this macro is used only when profiling. 31463 31464 31465 File: gccint.info, Node: Dispatch Tables, Next: Exception Region Output, Prev: Instruction Output, Up: Assembler Format 31466 31467 17.21.8 Output of Dispatch Tables 31468 --------------------------------- 31469 31470 This concerns dispatch tables. 31471 31472 -- Macro: ASM_OUTPUT_ADDR_DIFF_ELT (STREAM, BODY, VALUE, REL) 31473 A C statement to output to the stdio stream STREAM an assembler 31474 pseudo-instruction to generate a difference between two labels. 31475 VALUE and REL are the numbers of two internal labels. The 31476 definitions of these labels are output using 31477 `(*targetm.asm_out.internal_label)', and they must be printed in 31478 the same way here. For example, 31479 31480 fprintf (STREAM, "\t.word L%d-L%d\n", 31481 VALUE, REL) 31482 31483 You must provide this macro on machines where the addresses in a 31484 dispatch table are relative to the table's own address. If 31485 defined, GCC will also use this macro on all machines when 31486 producing PIC. BODY is the body of the `ADDR_DIFF_VEC'; it is 31487 provided so that the mode and flags can be read. 31488 31489 -- Macro: ASM_OUTPUT_ADDR_VEC_ELT (STREAM, VALUE) 31490 This macro should be provided on machines where the addresses in a 31491 dispatch table are absolute. 31492 31493 The definition should be a C statement to output to the stdio 31494 stream STREAM an assembler pseudo-instruction to generate a 31495 reference to a label. VALUE is the number of an internal label 31496 whose definition is output using 31497 `(*targetm.asm_out.internal_label)'. For example, 31498 31499 fprintf (STREAM, "\t.word L%d\n", VALUE) 31500 31501 -- Macro: ASM_OUTPUT_CASE_LABEL (STREAM, PREFIX, NUM, TABLE) 31502 Define this if the label before a jump-table needs to be output 31503 specially. The first three arguments are the same as for 31504 `(*targetm.asm_out.internal_label)'; the fourth argument is the 31505 jump-table which follows (a `jump_insn' containing an `addr_vec' 31506 or `addr_diff_vec'). 31507 31508 This feature is used on system V to output a `swbeg' statement for 31509 the table. 31510 31511 If this macro is not defined, these labels are output with 31512 `(*targetm.asm_out.internal_label)'. 31513 31514 -- Macro: ASM_OUTPUT_CASE_END (STREAM, NUM, TABLE) 31515 Define this if something special must be output at the end of a 31516 jump-table. The definition should be a C statement to be executed 31517 after the assembler code for the table is written. It should write 31518 the appropriate code to stdio stream STREAM. The argument TABLE 31519 is the jump-table insn, and NUM is the label-number of the 31520 preceding label. 31521 31522 If this macro is not defined, nothing special is output at the end 31523 of the jump-table. 31524 31525 -- Target Hook: void TARGET_ASM_EMIT_UNWIND_LABEL (STREAM, DECL, 31526 FOR_EH, EMPTY) 31527 This target hook emits a label at the beginning of each FDE. It 31528 should be defined on targets where FDEs need special labels, and it 31529 should write the appropriate label, for the FDE associated with the 31530 function declaration DECL, to the stdio stream STREAM. The third 31531 argument, FOR_EH, is a boolean: true if this is for an exception 31532 table. The fourth argument, EMPTY, is a boolean: true if this is 31533 a placeholder label for an omitted FDE. 31534 31535 The default is that FDEs are not given nonlocal labels. 31536 31537 -- Target Hook: void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (STREAM) 31538 This target hook emits a label at the beginning of the exception 31539 table. It should be defined on targets where it is desirable for 31540 the table to be broken up according to function. 31541 31542 The default is that no label is emitted. 31543 31544 -- Target Hook: void TARGET_UNWIND_EMIT (FILE * STREAM, rtx INSN) 31545 This target hook emits and assembly directives required to unwind 31546 the given instruction. This is only used when TARGET_UNWIND_INFO 31547 is set. 31548 31549 31550 File: gccint.info, Node: Exception Region Output, Next: Alignment Output, Prev: Dispatch Tables, Up: Assembler Format 31551 31552 17.21.9 Assembler Commands for Exception Regions 31553 ------------------------------------------------ 31554 31555 This describes commands marking the start and the end of an exception 31556 region. 31557 31558 -- Macro: EH_FRAME_SECTION_NAME 31559 If defined, a C string constant for the name of the section 31560 containing exception handling frame unwind information. If not 31561 defined, GCC will provide a default definition if the target 31562 supports named sections. `crtstuff.c' uses this macro to switch 31563 to the appropriate section. 31564 31565 You should define this symbol if your target supports DWARF 2 frame 31566 unwind information and the default definition does not work. 31567 31568 -- Macro: EH_FRAME_IN_DATA_SECTION 31569 If defined, DWARF 2 frame unwind information will be placed in the 31570 data section even though the target supports named sections. This 31571 might be necessary, for instance, if the system linker does garbage 31572 collection and sections cannot be marked as not to be collected. 31573 31574 Do not define this macro unless `TARGET_ASM_NAMED_SECTION' is also 31575 defined. 31576 31577 -- Macro: EH_TABLES_CAN_BE_READ_ONLY 31578 Define this macro to 1 if your target is such that no frame unwind 31579 information encoding used with non-PIC code will ever require a 31580 runtime relocation, but the linker may not support merging 31581 read-only and read-write sections into a single read-write section. 31582 31583 -- Macro: MASK_RETURN_ADDR 31584 An rtx used to mask the return address found via 31585 `RETURN_ADDR_RTX', so that it does not contain any extraneous set 31586 bits in it. 31587 31588 -- Macro: DWARF2_UNWIND_INFO 31589 Define this macro to 0 if your target supports DWARF 2 frame unwind 31590 information, but it does not yet work with exception handling. 31591 Otherwise, if your target supports this information (if it defines 31592 `INCOMING_RETURN_ADDR_RTX' and either `UNALIGNED_INT_ASM_OP' or 31593 `OBJECT_FORMAT_ELF'), GCC will provide a default definition of 1. 31594 31595 If `TARGET_UNWIND_INFO' is defined, the target specific unwinder 31596 will be used in all cases. Defining this macro will enable the 31597 generation of DWARF 2 frame debugging information. 31598 31599 If `TARGET_UNWIND_INFO' is not defined, and this macro is defined 31600 to 1, the DWARF 2 unwinder will be the default exception handling 31601 mechanism; otherwise, the `setjmp'/`longjmp'-based scheme will be 31602 used by default. 31603 31604 -- Macro: TARGET_UNWIND_INFO 31605 Define this macro if your target has ABI specified unwind tables. 31606 Usually these will be output by `TARGET_UNWIND_EMIT'. 31607 31608 -- Variable: Target Hook bool TARGET_UNWIND_TABLES_DEFAULT 31609 This variable should be set to `true' if the target ABI requires 31610 unwinding tables even when exceptions are not used. 31611 31612 -- Macro: MUST_USE_SJLJ_EXCEPTIONS 31613 This macro need only be defined if `DWARF2_UNWIND_INFO' is 31614 runtime-variable. In that case, `except.h' cannot correctly 31615 determine the corresponding definition of 31616 `MUST_USE_SJLJ_EXCEPTIONS', so the target must provide it directly. 31617 31618 -- Macro: DONT_USE_BUILTIN_SETJMP 31619 Define this macro to 1 if the `setjmp'/`longjmp'-based scheme 31620 should use the `setjmp'/`longjmp' functions from the C library 31621 instead of the `__builtin_setjmp'/`__builtin_longjmp' machinery. 31622 31623 -- Macro: DWARF_CIE_DATA_ALIGNMENT 31624 This macro need only be defined if the target might save registers 31625 in the function prologue at an offset to the stack pointer that is 31626 not aligned to `UNITS_PER_WORD'. The definition should be the 31627 negative minimum alignment if `STACK_GROWS_DOWNWARD' is defined, 31628 and the positive minimum alignment otherwise. *Note SDB and 31629 DWARF::. Only applicable if the target supports DWARF 2 frame 31630 unwind information. 31631 31632 -- Variable: Target Hook bool TARGET_TERMINATE_DW2_EH_FRAME_INFO 31633 Contains the value true if the target should add a zero word onto 31634 the end of a Dwarf-2 frame info section when used for exception 31635 handling. Default value is false if `EH_FRAME_SECTION_NAME' is 31636 defined, and true otherwise. 31637 31638 -- Target Hook: rtx TARGET_DWARF_REGISTER_SPAN (rtx REG) 31639 Given a register, this hook should return a parallel of registers 31640 to represent where to find the register pieces. Define this hook 31641 if the register and its mode are represented in Dwarf in 31642 non-contiguous locations, or if the register should be represented 31643 in more than one register in Dwarf. Otherwise, this hook should 31644 return `NULL_RTX'. If not defined, the default is to return 31645 `NULL_RTX'. 31646 31647 -- Target Hook: void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree ADDRESS) 31648 If some registers are represented in Dwarf-2 unwind information in 31649 multiple pieces, define this hook to fill in information about the 31650 sizes of those pieces in the table used by the unwinder at runtime. 31651 It will be called by `expand_builtin_init_dwarf_reg_sizes' after 31652 filling in a single size corresponding to each hard register; 31653 ADDRESS is the address of the table. 31654 31655 -- Target Hook: bool TARGET_ASM_TTYPE (rtx SYM) 31656 This hook is used to output a reference from a frame unwinding 31657 table to the type_info object identified by SYM. It should return 31658 `true' if the reference was output. Returning `false' will cause 31659 the reference to be output using the normal Dwarf2 routines. 31660 31661 -- Target Hook: bool TARGET_ARM_EABI_UNWINDER 31662 This hook should be set to `true' on targets that use an ARM EABI 31663 based unwinding library, and `false' on other targets. This 31664 effects the format of unwinding tables, and how the unwinder in 31665 entered after running a cleanup. The default is `false'. 31666 31667 31668 File: gccint.info, Node: Alignment Output, Prev: Exception Region Output, Up: Assembler Format 31669 31670 17.21.10 Assembler Commands for Alignment 31671 ----------------------------------------- 31672 31673 This describes commands for alignment. 31674 31675 -- Macro: JUMP_ALIGN (LABEL) 31676 The alignment (log base 2) to put in front of LABEL, which is a 31677 common destination of jumps and has no fallthru incoming edge. 31678 31679 This macro need not be defined if you don't want any special 31680 alignment to be done at such a time. Most machine descriptions do 31681 not currently define the macro. 31682 31683 Unless it's necessary to inspect the LABEL parameter, it is better 31684 to set the variable ALIGN_JUMPS in the target's 31685 `OVERRIDE_OPTIONS'. Otherwise, you should try to honor the user's 31686 selection in ALIGN_JUMPS in a `JUMP_ALIGN' implementation. 31687 31688 -- Macro: LABEL_ALIGN_AFTER_BARRIER (LABEL) 31689 The alignment (log base 2) to put in front of LABEL, which follows 31690 a `BARRIER'. 31691 31692 This macro need not be defined if you don't want any special 31693 alignment to be done at such a time. Most machine descriptions do 31694 not currently define the macro. 31695 31696 -- Macro: LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP 31697 The maximum number of bytes to skip when applying 31698 `LABEL_ALIGN_AFTER_BARRIER'. This works only if 31699 `ASM_OUTPUT_MAX_SKIP_ALIGN' is defined. 31700 31701 -- Macro: LOOP_ALIGN (LABEL) 31702 The alignment (log base 2) to put in front of LABEL, which follows 31703 a `NOTE_INSN_LOOP_BEG' note. 31704 31705 This macro need not be defined if you don't want any special 31706 alignment to be done at such a time. Most machine descriptions do 31707 not currently define the macro. 31708 31709 Unless it's necessary to inspect the LABEL parameter, it is better 31710 to set the variable `align_loops' in the target's 31711 `OVERRIDE_OPTIONS'. Otherwise, you should try to honor the user's 31712 selection in `align_loops' in a `LOOP_ALIGN' implementation. 31713 31714 -- Macro: LOOP_ALIGN_MAX_SKIP 31715 The maximum number of bytes to skip when applying `LOOP_ALIGN'. 31716 This works only if `ASM_OUTPUT_MAX_SKIP_ALIGN' is defined. 31717 31718 -- Macro: LABEL_ALIGN (LABEL) 31719 The alignment (log base 2) to put in front of LABEL. If 31720 `LABEL_ALIGN_AFTER_BARRIER' / `LOOP_ALIGN' specify a different 31721 alignment, the maximum of the specified values is used. 31722 31723 Unless it's necessary to inspect the LABEL parameter, it is better 31724 to set the variable `align_labels' in the target's 31725 `OVERRIDE_OPTIONS'. Otherwise, you should try to honor the user's 31726 selection in `align_labels' in a `LABEL_ALIGN' implementation. 31727 31728 -- Macro: LABEL_ALIGN_MAX_SKIP 31729 The maximum number of bytes to skip when applying `LABEL_ALIGN'. 31730 This works only if `ASM_OUTPUT_MAX_SKIP_ALIGN' is defined. 31731 31732 -- Macro: ASM_OUTPUT_SKIP (STREAM, NBYTES) 31733 A C statement to output to the stdio stream STREAM an assembler 31734 instruction to advance the location counter by NBYTES bytes. 31735 Those bytes should be zero when loaded. NBYTES will be a C 31736 expression of type `unsigned HOST_WIDE_INT'. 31737 31738 -- Macro: ASM_NO_SKIP_IN_TEXT 31739 Define this macro if `ASM_OUTPUT_SKIP' should not be used in the 31740 text section because it fails to put zeros in the bytes that are 31741 skipped. This is true on many Unix systems, where the pseudo-op 31742 to skip bytes produces no-op instructions rather than zeros when 31743 used in the text section. 31744 31745 -- Macro: ASM_OUTPUT_ALIGN (STREAM, POWER) 31746 A C statement to output to the stdio stream STREAM an assembler 31747 command to advance the location counter to a multiple of 2 to the 31748 POWER bytes. POWER will be a C expression of type `int'. 31749 31750 -- Macro: ASM_OUTPUT_ALIGN_WITH_NOP (STREAM, POWER) 31751 Like `ASM_OUTPUT_ALIGN', except that the "nop" instruction is used 31752 for padding, if necessary. 31753 31754 -- Macro: ASM_OUTPUT_MAX_SKIP_ALIGN (STREAM, POWER, MAX_SKIP) 31755 A C statement to output to the stdio stream STREAM an assembler 31756 command to advance the location counter to a multiple of 2 to the 31757 POWER bytes, but only if MAX_SKIP or fewer bytes are needed to 31758 satisfy the alignment request. POWER and MAX_SKIP will be a C 31759 expression of type `int'. 31760 31761 31762 File: gccint.info, Node: Debugging Info, Next: Floating Point, Prev: Assembler Format, Up: Target Macros 31763 31764 17.22 Controlling Debugging Information Format 31765 ============================================== 31766 31767 This describes how to specify debugging information. 31768 31769 * Menu: 31770 31771 * All Debuggers:: Macros that affect all debugging formats uniformly. 31772 * DBX Options:: Macros enabling specific options in DBX format. 31773 * DBX Hooks:: Hook macros for varying DBX format. 31774 * File Names and DBX:: Macros controlling output of file names in DBX format. 31775 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. 31776 * VMS Debug:: Macros for VMS debug format. 31777 31778 31779 File: gccint.info, Node: All Debuggers, Next: DBX Options, Up: Debugging Info 31780 31781 17.22.1 Macros Affecting All Debugging Formats 31782 ---------------------------------------------- 31783 31784 These macros affect all debugging formats. 31785 31786 -- Macro: DBX_REGISTER_NUMBER (REGNO) 31787 A C expression that returns the DBX register number for the 31788 compiler register number REGNO. In the default macro provided, 31789 the value of this expression will be REGNO itself. But sometimes 31790 there are some registers that the compiler knows about and DBX 31791 does not, or vice versa. In such cases, some register may need to 31792 have one number in the compiler and another for DBX. 31793 31794 If two registers have consecutive numbers inside GCC, and they can 31795 be used as a pair to hold a multiword value, then they _must_ have 31796 consecutive numbers after renumbering with `DBX_REGISTER_NUMBER'. 31797 Otherwise, debuggers will be unable to access such a pair, because 31798 they expect register pairs to be consecutive in their own 31799 numbering scheme. 31800 31801 If you find yourself defining `DBX_REGISTER_NUMBER' in way that 31802 does not preserve register pairs, then what you must do instead is 31803 redefine the actual register numbering scheme. 31804 31805 -- Macro: DEBUGGER_AUTO_OFFSET (X) 31806 A C expression that returns the integer offset value for an 31807 automatic variable having address X (an RTL expression). The 31808 default computation assumes that X is based on the frame-pointer 31809 and gives the offset from the frame-pointer. This is required for 31810 targets that produce debugging output for DBX or COFF-style 31811 debugging output for SDB and allow the frame-pointer to be 31812 eliminated when the `-g' options is used. 31813 31814 -- Macro: DEBUGGER_ARG_OFFSET (OFFSET, X) 31815 A C expression that returns the integer offset value for an 31816 argument having address X (an RTL expression). The nominal offset 31817 is OFFSET. 31818 31819 -- Macro: PREFERRED_DEBUGGING_TYPE 31820 A C expression that returns the type of debugging output GCC should 31821 produce when the user specifies just `-g'. Define this if you 31822 have arranged for GCC to support more than one format of debugging 31823 output. Currently, the allowable values are `DBX_DEBUG', 31824 `SDB_DEBUG', `DWARF_DEBUG', `DWARF2_DEBUG', `XCOFF_DEBUG', 31825 `VMS_DEBUG', and `VMS_AND_DWARF2_DEBUG'. 31826 31827 When the user specifies `-ggdb', GCC normally also uses the value 31828 of this macro to select the debugging output format, but with two 31829 exceptions. If `DWARF2_DEBUGGING_INFO' is defined, GCC uses the 31830 value `DWARF2_DEBUG'. Otherwise, if `DBX_DEBUGGING_INFO' is 31831 defined, GCC uses `DBX_DEBUG'. 31832 31833 The value of this macro only affects the default debugging output; 31834 the user can always get a specific type of output by using 31835 `-gstabs', `-gcoff', `-gdwarf-2', `-gxcoff', or `-gvms'. 31836 31837 31838 File: gccint.info, Node: DBX Options, Next: DBX Hooks, Prev: All Debuggers, Up: Debugging Info 31839 31840 17.22.2 Specific Options for DBX Output 31841 --------------------------------------- 31842 31843 These are specific options for DBX output. 31844 31845 -- Macro: DBX_DEBUGGING_INFO 31846 Define this macro if GCC should produce debugging output for DBX 31847 in response to the `-g' option. 31848 31849 -- Macro: XCOFF_DEBUGGING_INFO 31850 Define this macro if GCC should produce XCOFF format debugging 31851 output in response to the `-g' option. This is a variant of DBX 31852 format. 31853 31854 -- Macro: DEFAULT_GDB_EXTENSIONS 31855 Define this macro to control whether GCC should by default generate 31856 GDB's extended version of DBX debugging information (assuming 31857 DBX-format debugging information is enabled at all). If you don't 31858 define the macro, the default is 1: always generate the extended 31859 information if there is any occasion to. 31860 31861 -- Macro: DEBUG_SYMS_TEXT 31862 Define this macro if all `.stabs' commands should be output while 31863 in the text section. 31864 31865 -- Macro: ASM_STABS_OP 31866 A C string constant, including spacing, naming the assembler 31867 pseudo op to use instead of `"\t.stabs\t"' to define an ordinary 31868 debugging symbol. If you don't define this macro, `"\t.stabs\t"' 31869 is used. This macro applies only to DBX debugging information 31870 format. 31871 31872 -- Macro: ASM_STABD_OP 31873 A C string constant, including spacing, naming the assembler 31874 pseudo op to use instead of `"\t.stabd\t"' to define a debugging 31875 symbol whose value is the current location. If you don't define 31876 this macro, `"\t.stabd\t"' is used. This macro applies only to 31877 DBX debugging information format. 31878 31879 -- Macro: ASM_STABN_OP 31880 A C string constant, including spacing, naming the assembler 31881 pseudo op to use instead of `"\t.stabn\t"' to define a debugging 31882 symbol with no name. If you don't define this macro, 31883 `"\t.stabn\t"' is used. This macro applies only to DBX debugging 31884 information format. 31885 31886 -- Macro: DBX_NO_XREFS 31887 Define this macro if DBX on your system does not support the 31888 construct `xsTAGNAME'. On some systems, this construct is used to 31889 describe a forward reference to a structure named TAGNAME. On 31890 other systems, this construct is not supported at all. 31891 31892 -- Macro: DBX_CONTIN_LENGTH 31893 A symbol name in DBX-format debugging information is normally 31894 continued (split into two separate `.stabs' directives) when it 31895 exceeds a certain length (by default, 80 characters). On some 31896 operating systems, DBX requires this splitting; on others, 31897 splitting must not be done. You can inhibit splitting by defining 31898 this macro with the value zero. You can override the default 31899 splitting-length by defining this macro as an expression for the 31900 length you desire. 31901 31902 -- Macro: DBX_CONTIN_CHAR 31903 Normally continuation is indicated by adding a `\' character to 31904 the end of a `.stabs' string when a continuation follows. To use 31905 a different character instead, define this macro as a character 31906 constant for the character you want to use. Do not define this 31907 macro if backslash is correct for your system. 31908 31909 -- Macro: DBX_STATIC_STAB_DATA_SECTION 31910 Define this macro if it is necessary to go to the data section 31911 before outputting the `.stabs' pseudo-op for a non-global static 31912 variable. 31913 31914 -- Macro: DBX_TYPE_DECL_STABS_CODE 31915 The value to use in the "code" field of the `.stabs' directive for 31916 a typedef. The default is `N_LSYM'. 31917 31918 -- Macro: DBX_STATIC_CONST_VAR_CODE 31919 The value to use in the "code" field of the `.stabs' directive for 31920 a static variable located in the text section. DBX format does not 31921 provide any "right" way to do this. The default is `N_FUN'. 31922 31923 -- Macro: DBX_REGPARM_STABS_CODE 31924 The value to use in the "code" field of the `.stabs' directive for 31925 a parameter passed in registers. DBX format does not provide any 31926 "right" way to do this. The default is `N_RSYM'. 31927 31928 -- Macro: DBX_REGPARM_STABS_LETTER 31929 The letter to use in DBX symbol data to identify a symbol as a 31930 parameter passed in registers. DBX format does not customarily 31931 provide any way to do this. The default is `'P''. 31932 31933 -- Macro: DBX_FUNCTION_FIRST 31934 Define this macro if the DBX information for a function and its 31935 arguments should precede the assembler code for the function. 31936 Normally, in DBX format, the debugging information entirely 31937 follows the assembler code. 31938 31939 -- Macro: DBX_BLOCKS_FUNCTION_RELATIVE 31940 Define this macro, with value 1, if the value of a symbol 31941 describing the scope of a block (`N_LBRAC' or `N_RBRAC') should be 31942 relative to the start of the enclosing function. Normally, GCC 31943 uses an absolute address. 31944 31945 -- Macro: DBX_LINES_FUNCTION_RELATIVE 31946 Define this macro, with value 1, if the value of a symbol 31947 indicating the current line number (`N_SLINE') should be relative 31948 to the start of the enclosing function. Normally, GCC uses an 31949 absolute address. 31950 31951 -- Macro: DBX_USE_BINCL 31952 Define this macro if GCC should generate `N_BINCL' and `N_EINCL' 31953 stabs for included header files, as on Sun systems. This macro 31954 also directs GCC to output a type number as a pair of a file 31955 number and a type number within the file. Normally, GCC does not 31956 generate `N_BINCL' or `N_EINCL' stabs, and it outputs a single 31957 number for a type number. 31958 31959 31960 File: gccint.info, Node: DBX Hooks, Next: File Names and DBX, Prev: DBX Options, Up: Debugging Info 31961 31962 17.22.3 Open-Ended Hooks for DBX Format 31963 --------------------------------------- 31964 31965 These are hooks for DBX format. 31966 31967 -- Macro: DBX_OUTPUT_LBRAC (STREAM, NAME) 31968 Define this macro to say how to output to STREAM the debugging 31969 information for the start of a scope level for variable names. The 31970 argument NAME is the name of an assembler symbol (for use with 31971 `assemble_name') whose value is the address where the scope begins. 31972 31973 -- Macro: DBX_OUTPUT_RBRAC (STREAM, NAME) 31974 Like `DBX_OUTPUT_LBRAC', but for the end of a scope level. 31975 31976 -- Macro: DBX_OUTPUT_NFUN (STREAM, LSCOPE_LABEL, DECL) 31977 Define this macro if the target machine requires special handling 31978 to output an `N_FUN' entry for the function DECL. 31979 31980 -- Macro: DBX_OUTPUT_SOURCE_LINE (STREAM, LINE, COUNTER) 31981 A C statement to output DBX debugging information before code for 31982 line number LINE of the current source file to the stdio stream 31983 STREAM. COUNTER is the number of time the macro was invoked, 31984 including the current invocation; it is intended to generate 31985 unique labels in the assembly output. 31986 31987 This macro should not be defined if the default output is correct, 31988 or if it can be made correct by defining 31989 `DBX_LINES_FUNCTION_RELATIVE'. 31990 31991 -- Macro: NO_DBX_FUNCTION_END 31992 Some stabs encapsulation formats (in particular ECOFF), cannot 31993 handle the `.stabs "",N_FUN,,0,0,Lscope-function-1' gdb dbx 31994 extension construct. On those machines, define this macro to turn 31995 this feature off without disturbing the rest of the gdb extensions. 31996 31997 -- Macro: NO_DBX_BNSYM_ENSYM 31998 Some assemblers cannot handle the `.stabd BNSYM/ENSYM,0,0' gdb dbx 31999 extension construct. On those machines, define this macro to turn 32000 this feature off without disturbing the rest of the gdb extensions. 32001 32002 32003 File: gccint.info, Node: File Names and DBX, Next: SDB and DWARF, Prev: DBX Hooks, Up: Debugging Info 32004 32005 17.22.4 File Names in DBX Format 32006 -------------------------------- 32007 32008 This describes file names in DBX format. 32009 32010 -- Macro: DBX_OUTPUT_MAIN_SOURCE_FILENAME (STREAM, NAME) 32011 A C statement to output DBX debugging information to the stdio 32012 stream STREAM, which indicates that file NAME is the main source 32013 file--the file specified as the input file for compilation. This 32014 macro is called only once, at the beginning of compilation. 32015 32016 This macro need not be defined if the standard form of output for 32017 DBX debugging information is appropriate. 32018 32019 It may be necessary to refer to a label equal to the beginning of 32020 the text section. You can use `assemble_name (stream, 32021 ltext_label_name)' to do so. If you do this, you must also set 32022 the variable USED_LTEXT_LABEL_NAME to `true'. 32023 32024 -- Macro: NO_DBX_MAIN_SOURCE_DIRECTORY 32025 Define this macro, with value 1, if GCC should not emit an 32026 indication of the current directory for compilation and current 32027 source language at the beginning of the file. 32028 32029 -- Macro: NO_DBX_GCC_MARKER 32030 Define this macro, with value 1, if GCC should not emit an 32031 indication that this object file was compiled by GCC. The default 32032 is to emit an `N_OPT' stab at the beginning of every source file, 32033 with `gcc2_compiled.' for the string and value 0. 32034 32035 -- Macro: DBX_OUTPUT_MAIN_SOURCE_FILE_END (STREAM, NAME) 32036 A C statement to output DBX debugging information at the end of 32037 compilation of the main source file NAME. Output should be 32038 written to the stdio stream STREAM. 32039 32040 If you don't define this macro, nothing special is output at the 32041 end of compilation, which is correct for most machines. 32042 32043 -- Macro: DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END 32044 Define this macro _instead of_ defining 32045 `DBX_OUTPUT_MAIN_SOURCE_FILE_END', if what needs to be output at 32046 the end of compilation is a `N_SO' stab with an empty string, 32047 whose value is the highest absolute text address in the file. 32048 32049 32050 File: gccint.info, Node: SDB and DWARF, Next: VMS Debug, Prev: File Names and DBX, Up: Debugging Info 32051 32052 17.22.5 Macros for SDB and DWARF Output 32053 --------------------------------------- 32054 32055 Here are macros for SDB and DWARF output. 32056 32057 -- Macro: SDB_DEBUGGING_INFO 32058 Define this macro if GCC should produce COFF-style debugging output 32059 for SDB in response to the `-g' option. 32060 32061 -- Macro: DWARF2_DEBUGGING_INFO 32062 Define this macro if GCC should produce dwarf version 2 format 32063 debugging output in response to the `-g' option. 32064 32065 -- Target Hook: int TARGET_DWARF_CALLING_CONVENTION (tree 32066 FUNCTION) 32067 Define this to enable the dwarf attribute 32068 `DW_AT_calling_convention' to be emitted for each function. 32069 Instead of an integer return the enum value for the `DW_CC_' 32070 tag. 32071 32072 To support optional call frame debugging information, you must also 32073 define `INCOMING_RETURN_ADDR_RTX' and either set 32074 `RTX_FRAME_RELATED_P' on the prologue insns if you use RTL for the 32075 prologue, or call `dwarf2out_def_cfa' and `dwarf2out_reg_save' as 32076 appropriate from `TARGET_ASM_FUNCTION_PROLOGUE' if you don't. 32077 32078 -- Macro: DWARF2_FRAME_INFO 32079 Define this macro to a nonzero value if GCC should always output 32080 Dwarf 2 frame information. If `DWARF2_UNWIND_INFO' (*note 32081 Exception Region Output:: is nonzero, GCC will output this 32082 information not matter how you define `DWARF2_FRAME_INFO'. 32083 32084 -- Macro: DWARF2_ASM_LINE_DEBUG_INFO 32085 Define this macro to be a nonzero value if the assembler can 32086 generate Dwarf 2 line debug info sections. This will result in 32087 much more compact line number tables, and hence is desirable if it 32088 works. 32089 32090 -- Macro: ASM_OUTPUT_DWARF_DELTA (STREAM, SIZE, LABEL1, LABEL2) 32091 A C statement to issue assembly directives that create a difference 32092 LAB1 minus LAB2, using an integer of the given SIZE. 32093 32094 -- Macro: ASM_OUTPUT_DWARF_OFFSET (STREAM, SIZE, LABEL, SECTION) 32095 A C statement to issue assembly directives that create a 32096 section-relative reference to the given LABEL, using an integer of 32097 the given SIZE. The label is known to be defined in the given 32098 SECTION. 32099 32100 -- Macro: ASM_OUTPUT_DWARF_PCREL (STREAM, SIZE, LABEL) 32101 A C statement to issue assembly directives that create a 32102 self-relative reference to the given LABEL, using an integer of 32103 the given SIZE. 32104 32105 -- Target Hook: void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *FILE, int 32106 SIZE, rtx X) 32107 If defined, this target hook is a function which outputs a 32108 DTP-relative reference to the given TLS symbol of the specified 32109 size. 32110 32111 -- Macro: PUT_SDB_... 32112 Define these macros to override the assembler syntax for the 32113 special SDB assembler directives. See `sdbout.c' for a list of 32114 these macros and their arguments. If the standard syntax is used, 32115 you need not define them yourself. 32116 32117 -- Macro: SDB_DELIM 32118 Some assemblers do not support a semicolon as a delimiter, even 32119 between SDB assembler directives. In that case, define this macro 32120 to be the delimiter to use (usually `\n'). It is not necessary to 32121 define a new set of `PUT_SDB_OP' macros if this is the only change 32122 required. 32123 32124 -- Macro: SDB_ALLOW_UNKNOWN_REFERENCES 32125 Define this macro to allow references to unknown structure, union, 32126 or enumeration tags to be emitted. Standard COFF does not allow 32127 handling of unknown references, MIPS ECOFF has support for it. 32128 32129 -- Macro: SDB_ALLOW_FORWARD_REFERENCES 32130 Define this macro to allow references to structure, union, or 32131 enumeration tags that have not yet been seen to be handled. Some 32132 assemblers choke if forward tags are used, while some require it. 32133 32134 -- Macro: SDB_OUTPUT_SOURCE_LINE (STREAM, LINE) 32135 A C statement to output SDB debugging information before code for 32136 line number LINE of the current source file to the stdio stream 32137 STREAM. The default is to emit an `.ln' directive. 32138 32139 32140 File: gccint.info, Node: VMS Debug, Prev: SDB and DWARF, Up: Debugging Info 32141 32142 17.22.6 Macros for VMS Debug Format 32143 ----------------------------------- 32144 32145 Here are macros for VMS debug format. 32146 32147 -- Macro: VMS_DEBUGGING_INFO 32148 Define this macro if GCC should produce debugging output for VMS 32149 in response to the `-g' option. The default behavior for VMS is 32150 to generate minimal debug info for a traceback in the absence of 32151 `-g' unless explicitly overridden with `-g0'. This behavior is 32152 controlled by `OPTIMIZATION_OPTIONS' and `OVERRIDE_OPTIONS'. 32153 32154 32155 File: gccint.info, Node: Floating Point, Next: Mode Switching, Prev: Debugging Info, Up: Target Macros 32156 32157 17.23 Cross Compilation and Floating Point 32158 ========================================== 32159 32160 While all modern machines use twos-complement representation for 32161 integers, there are a variety of representations for floating point 32162 numbers. This means that in a cross-compiler the representation of 32163 floating point numbers in the compiled program may be different from 32164 that used in the machine doing the compilation. 32165 32166 Because different representation systems may offer different amounts of 32167 range and precision, all floating point constants must be represented in 32168 the target machine's format. Therefore, the cross compiler cannot 32169 safely use the host machine's floating point arithmetic; it must emulate 32170 the target's arithmetic. To ensure consistency, GCC always uses 32171 emulation to work with floating point values, even when the host and 32172 target floating point formats are identical. 32173 32174 The following macros are provided by `real.h' for the compiler to use. 32175 All parts of the compiler which generate or optimize floating-point 32176 calculations must use these macros. They may evaluate their operands 32177 more than once, so operands must not have side effects. 32178 32179 -- Macro: REAL_VALUE_TYPE 32180 The C data type to be used to hold a floating point value in the 32181 target machine's format. Typically this is a `struct' containing 32182 an array of `HOST_WIDE_INT', but all code should treat it as an 32183 opaque quantity. 32184 32185 -- Macro: int REAL_VALUES_EQUAL (REAL_VALUE_TYPE X, REAL_VALUE_TYPE Y) 32186 Compares for equality the two values, X and Y. If the target 32187 floating point format supports negative zeroes and/or NaNs, 32188 `REAL_VALUES_EQUAL (-0.0, 0.0)' is true, and `REAL_VALUES_EQUAL 32189 (NaN, NaN)' is false. 32190 32191 -- Macro: int REAL_VALUES_LESS (REAL_VALUE_TYPE X, REAL_VALUE_TYPE Y) 32192 Tests whether X is less than Y. 32193 32194 -- Macro: HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE X) 32195 Truncates X to a signed integer, rounding toward zero. 32196 32197 -- Macro: unsigned HOST_WIDE_INT REAL_VALUE_UNSIGNED_FIX 32198 (REAL_VALUE_TYPE X) 32199 Truncates X to an unsigned integer, rounding toward zero. If X is 32200 negative, returns zero. 32201 32202 -- Macro: REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *STRING, enum 32203 machine_mode MODE) 32204 Converts STRING into a floating point number in the target 32205 machine's representation for mode MODE. This routine can handle 32206 both decimal and hexadecimal floating point constants, using the 32207 syntax defined by the C language for both. 32208 32209 -- Macro: int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE X) 32210 Returns 1 if X is negative (including negative zero), 0 otherwise. 32211 32212 -- Macro: int REAL_VALUE_ISINF (REAL_VALUE_TYPE X) 32213 Determines whether X represents infinity (positive or negative). 32214 32215 -- Macro: int REAL_VALUE_ISNAN (REAL_VALUE_TYPE X) 32216 Determines whether X represents a "NaN" (not-a-number). 32217 32218 -- Macro: void REAL_ARITHMETIC (REAL_VALUE_TYPE OUTPUT, enum tree_code 32219 CODE, REAL_VALUE_TYPE X, REAL_VALUE_TYPE Y) 32220 Calculates an arithmetic operation on the two floating point values 32221 X and Y, storing the result in OUTPUT (which must be a variable). 32222 32223 The operation to be performed is specified by CODE. Only the 32224 following codes are supported: `PLUS_EXPR', `MINUS_EXPR', 32225 `MULT_EXPR', `RDIV_EXPR', `MAX_EXPR', `MIN_EXPR'. 32226 32227 If `REAL_ARITHMETIC' is asked to evaluate division by zero and the 32228 target's floating point format cannot represent infinity, it will 32229 call `abort'. Callers should check for this situation first, using 32230 `MODE_HAS_INFINITIES'. *Note Storage Layout::. 32231 32232 -- Macro: REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE X) 32233 Returns the negative of the floating point value X. 32234 32235 -- Macro: REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE X) 32236 Returns the absolute value of X. 32237 32238 -- Macro: REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE MODE, 32239 enum machine_mode X) 32240 Truncates the floating point value X to fit in MODE. The return 32241 value is still a full-size `REAL_VALUE_TYPE', but it has an 32242 appropriate bit pattern to be output as a floating constant whose 32243 precision accords with mode MODE. 32244 32245 -- Macro: void REAL_VALUE_TO_INT (HOST_WIDE_INT LOW, HOST_WIDE_INT 32246 HIGH, REAL_VALUE_TYPE X) 32247 Converts a floating point value X into a double-precision integer 32248 which is then stored into LOW and HIGH. If the value is not 32249 integral, it is truncated. 32250 32251 -- Macro: void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE X, HOST_WIDE_INT 32252 LOW, HOST_WIDE_INT HIGH, enum machine_mode MODE) 32253 Converts a double-precision integer found in LOW and HIGH, into a 32254 floating point value which is then stored into X. The value is 32255 truncated to fit in mode MODE. 32256 32257 32258 File: gccint.info, Node: Mode Switching, Next: Target Attributes, Prev: Floating Point, Up: Target Macros 32259 32260 17.24 Mode Switching Instructions 32261 ================================= 32262 32263 The following macros control mode switching optimizations: 32264 32265 -- Macro: OPTIMIZE_MODE_SWITCHING (ENTITY) 32266 Define this macro if the port needs extra instructions inserted 32267 for mode switching in an optimizing compilation. 32268 32269 For an example, the SH4 can perform both single and double 32270 precision floating point operations, but to perform a single 32271 precision operation, the FPSCR PR bit has to be cleared, while for 32272 a double precision operation, this bit has to be set. Changing 32273 the PR bit requires a general purpose register as a scratch 32274 register, hence these FPSCR sets have to be inserted before 32275 reload, i.e. you can't put this into instruction emitting or 32276 `TARGET_MACHINE_DEPENDENT_REORG'. 32277 32278 You can have multiple entities that are mode-switched, and select 32279 at run time which entities actually need it. 32280 `OPTIMIZE_MODE_SWITCHING' should return nonzero for any ENTITY 32281 that needs mode-switching. If you define this macro, you also 32282 have to define `NUM_MODES_FOR_MODE_SWITCHING', `MODE_NEEDED', 32283 `MODE_PRIORITY_TO_MODE' and `EMIT_MODE_SET'. `MODE_AFTER', 32284 `MODE_ENTRY', and `MODE_EXIT' are optional. 32285 32286 -- Macro: NUM_MODES_FOR_MODE_SWITCHING 32287 If you define `OPTIMIZE_MODE_SWITCHING', you have to define this as 32288 initializer for an array of integers. Each initializer element N 32289 refers to an entity that needs mode switching, and specifies the 32290 number of different modes that might need to be set for this 32291 entity. The position of the initializer in the 32292 initializer--starting counting at zero--determines the integer 32293 that is used to refer to the mode-switched entity in question. In 32294 macros that take mode arguments / yield a mode result, modes are 32295 represented as numbers 0 ... N - 1. N is used to specify that no 32296 mode switch is needed / supplied. 32297 32298 -- Macro: MODE_NEEDED (ENTITY, INSN) 32299 ENTITY is an integer specifying a mode-switched entity. If 32300 `OPTIMIZE_MODE_SWITCHING' is defined, you must define this macro to 32301 return an integer value not larger than the corresponding element 32302 in `NUM_MODES_FOR_MODE_SWITCHING', to denote the mode that ENTITY 32303 must be switched into prior to the execution of INSN. 32304 32305 -- Macro: MODE_AFTER (MODE, INSN) 32306 If this macro is defined, it is evaluated for every INSN during 32307 mode switching. It determines the mode that an insn results in (if 32308 different from the incoming mode). 32309 32310 -- Macro: MODE_ENTRY (ENTITY) 32311 If this macro is defined, it is evaluated for every ENTITY that 32312 needs mode switching. It should evaluate to an integer, which is 32313 a mode that ENTITY is assumed to be switched to at function entry. 32314 If `MODE_ENTRY' is defined then `MODE_EXIT' must be defined. 32315 32316 -- Macro: MODE_EXIT (ENTITY) 32317 If this macro is defined, it is evaluated for every ENTITY that 32318 needs mode switching. It should evaluate to an integer, which is 32319 a mode that ENTITY is assumed to be switched to at function exit. 32320 If `MODE_EXIT' is defined then `MODE_ENTRY' must be defined. 32321 32322 -- Macro: MODE_PRIORITY_TO_MODE (ENTITY, N) 32323 This macro specifies the order in which modes for ENTITY are 32324 processed. 0 is the highest priority, 32325 `NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1' the lowest. The value 32326 of the macro should be an integer designating a mode for ENTITY. 32327 For any fixed ENTITY, `mode_priority_to_mode' (ENTITY, N) shall be 32328 a bijection in 0 ... `num_modes_for_mode_switching[ENTITY] - 1'. 32329 32330 -- Macro: EMIT_MODE_SET (ENTITY, MODE, HARD_REGS_LIVE) 32331 Generate one or more insns to set ENTITY to MODE. HARD_REG_LIVE 32332 is the set of hard registers live at the point where the insn(s) 32333 are to be inserted. 32334 32335 32336 File: gccint.info, Node: Target Attributes, Next: Emulated TLS, Prev: Mode Switching, Up: Target Macros 32337 32338 17.25 Defining target-specific uses of `__attribute__' 32339 ====================================================== 32340 32341 Target-specific attributes may be defined for functions, data and types. 32342 These are described using the following target hooks; they also need to 32343 be documented in `extend.texi'. 32344 32345 -- Target Hook: const struct attribute_spec * TARGET_ATTRIBUTE_TABLE 32346 If defined, this target hook points to an array of `struct 32347 attribute_spec' (defined in `tree.h') specifying the machine 32348 specific attributes for this target and some of the restrictions 32349 on the entities to which these attributes are applied and the 32350 arguments they take. 32351 32352 -- Target Hook: int TARGET_COMP_TYPE_ATTRIBUTES (tree TYPE1, tree 32353 TYPE2) 32354 If defined, this target hook is a function which returns zero if 32355 the attributes on TYPE1 and TYPE2 are incompatible, one if they 32356 are compatible, and two if they are nearly compatible (which 32357 causes a warning to be generated). If this is not defined, 32358 machine-specific attributes are supposed always to be compatible. 32359 32360 -- Target Hook: void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree TYPE) 32361 If defined, this target hook is a function which assigns default 32362 attributes to newly defined TYPE. 32363 32364 -- Target Hook: tree TARGET_MERGE_TYPE_ATTRIBUTES (tree TYPE1, tree 32365 TYPE2) 32366 Define this target hook if the merging of type attributes needs 32367 special handling. If defined, the result is a list of the combined 32368 `TYPE_ATTRIBUTES' of TYPE1 and TYPE2. It is assumed that 32369 `comptypes' has already been called and returned 1. This function 32370 may call `merge_attributes' to handle machine-independent merging. 32371 32372 -- Target Hook: tree TARGET_MERGE_DECL_ATTRIBUTES (tree OLDDECL, tree 32373 NEWDECL) 32374 Define this target hook if the merging of decl attributes needs 32375 special handling. If defined, the result is a list of the combined 32376 `DECL_ATTRIBUTES' of OLDDECL and NEWDECL. NEWDECL is a duplicate 32377 declaration of OLDDECL. Examples of when this is needed are when 32378 one attribute overrides another, or when an attribute is nullified 32379 by a subsequent definition. This function may call 32380 `merge_attributes' to handle machine-independent merging. 32381 32382 If the only target-specific handling you require is `dllimport' 32383 for Microsoft Windows targets, you should define the macro 32384 `TARGET_DLLIMPORT_DECL_ATTRIBUTES' to `1'. The compiler will then 32385 define a function called `merge_dllimport_decl_attributes' which 32386 can then be defined as the expansion of 32387 `TARGET_MERGE_DECL_ATTRIBUTES'. You can also add 32388 `handle_dll_attribute' in the attribute table for your port to 32389 perform initial processing of the `dllimport' and `dllexport' 32390 attributes. This is done in `i386/cygwin.h' and `i386/i386.c', 32391 for example. 32392 32393 -- Target Hook: bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree DECL) 32394 DECL is a variable or function with `__attribute__((dllimport))' 32395 specified. Use this hook if the target needs to add extra 32396 validation checks to `handle_dll_attribute'. 32397 32398 -- Macro: TARGET_DECLSPEC 32399 Define this macro to a nonzero value if you want to treat 32400 `__declspec(X)' as equivalent to `__attribute((X))'. By default, 32401 this behavior is enabled only for targets that define 32402 `TARGET_DLLIMPORT_DECL_ATTRIBUTES'. The current implementation of 32403 `__declspec' is via a built-in macro, but you should not rely on 32404 this implementation detail. 32405 32406 -- Target Hook: void TARGET_INSERT_ATTRIBUTES (tree NODE, tree 32407 *ATTR_PTR) 32408 Define this target hook if you want to be able to add attributes 32409 to a decl when it is being created. This is normally useful for 32410 back ends which wish to implement a pragma by using the attributes 32411 which correspond to the pragma's effect. The NODE argument is the 32412 decl which is being created. The ATTR_PTR argument is a pointer 32413 to the attribute list for this decl. The list itself should not 32414 be modified, since it may be shared with other decls, but 32415 attributes may be chained on the head of the list and `*ATTR_PTR' 32416 modified to point to the new attributes, or a copy of the list may 32417 be made if further changes are needed. 32418 32419 -- Target Hook: bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree 32420 FNDECL) 32421 This target hook returns `true' if it is ok to inline FNDECL into 32422 the current function, despite its having target-specific 32423 attributes, `false' otherwise. By default, if a function has a 32424 target specific attribute attached to it, it will not be inlined. 32425 32426 -- Target Hook: bool TARGET_VALID_OPTION_ATTRIBUTE_P (tree FNDECL, 32427 tree NAME, tree ARGS, int FLAGS) 32428 This hook is called to parse the `attribute(option("..."))', and 32429 it allows the function to set different target machine compile time 32430 options for the current function that might be different than the 32431 options specified on the command line. The hook should return 32432 `true' if the options are valid. 32433 32434 The hook should set the DECL_FUNCTION_SPECIFIC_TARGET field in the 32435 function declaration to hold a pointer to a target specific STRUCT 32436 CL_TARGET_OPTION structure. 32437 32438 -- Target Hook: void TARGET_OPTION_SAVE (struct cl_target_option *PTR) 32439 This hook is called to save any additional target specific 32440 information in the STRUCT CL_TARGET_OPTION structure for function 32441 specific options. *Note Option file format::. 32442 32443 -- Target Hook: void TARGET_OPTION_RESTORE (struct cl_target_option 32444 *PTR) 32445 This hook is called to restore any additional target specific 32446 information in the STRUCT CL_TARGET_OPTION structure for function 32447 specific options. 32448 32449 -- Target Hook: void TARGET_OPTION_PRINT (struct cl_target_option *PTR) 32450 This hook is called to print any additional target specific 32451 information in the STRUCT CL_TARGET_OPTION structure for function 32452 specific options. 32453 32454 -- Target Hook: bool TARGET_OPTION_PRAGMA_PARSE (target ARGS) 32455 This target hook parses the options for `#pragma GCC option' to 32456 set the machine specific options for functions that occur later in 32457 the input stream. The options should be the same as handled by the 32458 `TARGET_VALID_OPTION_ATTRIBUTE_P' hook. 32459 32460 -- Target Hook: bool TARGET_CAN_INLINE_P (tree CALLER, tree CALLEE) 32461 This target hook returns `false' if the CALLER function cannot 32462 inline CALLEE, based on target specific information. By default, 32463 inlining is not allowed if the callee function has function 32464 specific target options and the caller does not use the same 32465 options. 32466 32467 32468 File: gccint.info, Node: Emulated TLS, Next: MIPS Coprocessors, Prev: Target Attributes, Up: Target Macros 32469 32470 17.26 Emulating TLS 32471 =================== 32472 32473 For targets whose psABI does not provide Thread Local Storage via 32474 specific relocations and instruction sequences, an emulation layer is 32475 used. A set of target hooks allows this emulation layer to be 32476 configured for the requirements of a particular target. For instance 32477 the psABI may in fact specify TLS support in terms of an emulation 32478 layer. 32479 32480 The emulation layer works by creating a control object for every TLS 32481 object. To access the TLS object, a lookup function is provided which, 32482 when given the address of the control object, will return the address 32483 of the current thread's instance of the TLS object. 32484 32485 -- Target Hook: const char * TARGET_EMUTLS_GET_ADDRESS 32486 Contains the name of the helper function that uses a TLS control 32487 object to locate a TLS instance. The default causes libgcc's 32488 emulated TLS helper function to be used. 32489 32490 -- Target Hook: const char * TARGET_EMUTLS_REGISTER_COMMON 32491 Contains the name of the helper function that should be used at 32492 program startup to register TLS objects that are implicitly 32493 initialized to zero. If this is `NULL', all TLS objects will have 32494 explicit initializers. The default causes libgcc's emulated TLS 32495 registration function to be used. 32496 32497 -- Target Hook: const char * TARGET_EMUTLS_VAR_SECTION 32498 Contains the name of the section in which TLS control variables 32499 should be placed. The default of `NULL' allows these to be placed 32500 in any section. 32501 32502 -- Target Hook: const char * TARGET_EMUTLS_TMPL_SECTION 32503 Contains the name of the section in which TLS initializers should 32504 be placed. The default of `NULL' allows these to be placed in any 32505 section. 32506 32507 -- Target Hook: const char * TARGET_EMUTLS_VAR_PREFIX 32508 Contains the prefix to be prepended to TLS control variable names. 32509 The default of `NULL' uses a target-specific prefix. 32510 32511 -- Target Hook: const char * TARGET_EMUTLS_TMPL_PREFIX 32512 Contains the prefix to be prepended to TLS initializer objects. 32513 The default of `NULL' uses a target-specific prefix. 32514 32515 -- Target Hook: tree TARGET_EMUTLS_VAR_FIELDS (tree TYPE, tree *NAME) 32516 Specifies a function that generates the FIELD_DECLs for a TLS 32517 control object type. TYPE is the RECORD_TYPE the fields are for 32518 and NAME should be filled with the structure tag, if the default of 32519 `__emutls_object' is unsuitable. The default creates a type 32520 suitable for libgcc's emulated TLS function. 32521 32522 -- Target Hook: tree TARGET_EMUTLS_VAR_INIT (tree VAR, tree DECL, tree 32523 TMPL_ADDR) 32524 Specifies a function that generates the CONSTRUCTOR to initialize a 32525 TLS control object. VAR is the TLS control object, DECL is the 32526 TLS object and TMPL_ADDR is the address of the initializer. The 32527 default initializes libgcc's emulated TLS control object. 32528 32529 -- Target Hook: bool TARGET_EMUTLS_VAR_ALIGN_FIXED 32530 Specifies whether the alignment of TLS control variable objects is 32531 fixed and should not be increased as some backends may do to 32532 optimize single objects. The default is false. 32533 32534 -- Target Hook: bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS 32535 Specifies whether a DWARF `DW_OP_form_tls_address' location 32536 descriptor may be used to describe emulated TLS control objects. 32537 32538 32539 File: gccint.info, Node: MIPS Coprocessors, Next: PCH Target, Prev: Emulated TLS, Up: Target Macros 32540 32541 17.27 Defining coprocessor specifics for MIPS targets. 32542 ====================================================== 32543 32544 The MIPS specification allows MIPS implementations to have as many as 4 32545 coprocessors, each with as many as 32 private registers. GCC supports 32546 accessing these registers and transferring values between the registers 32547 and memory using asm-ized variables. For example: 32548 32549 register unsigned int cp0count asm ("c0r1"); 32550 unsigned int d; 32551 32552 d = cp0count + 3; 32553 32554 ("c0r1" is the default name of register 1 in coprocessor 0; alternate 32555 names may be added as described below, or the default names may be 32556 overridden entirely in `SUBTARGET_CONDITIONAL_REGISTER_USAGE'.) 32557 32558 Coprocessor registers are assumed to be epilogue-used; sets to them 32559 will be preserved even if it does not appear that the register is used 32560 again later in the function. 32561 32562 Another note: according to the MIPS spec, coprocessor 1 (if present) is 32563 the FPU. One accesses COP1 registers through standard mips 32564 floating-point support; they are not included in this mechanism. 32565 32566 There is one macro used in defining the MIPS coprocessor interface 32567 which you may want to override in subtargets; it is described below. 32568 32569 -- Macro: ALL_COP_ADDITIONAL_REGISTER_NAMES 32570 A comma-separated list (with leading comma) of pairs describing the 32571 alternate names of coprocessor registers. The format of each 32572 entry should be 32573 { ALTERNATENAME, REGISTER_NUMBER} 32574 Default: empty. 32575 32576 32577 File: gccint.info, Node: PCH Target, Next: C++ ABI, Prev: MIPS Coprocessors, Up: Target Macros 32578 32579 17.28 Parameters for Precompiled Header Validity Checking 32580 ========================================================= 32581 32582 -- Target Hook: void *TARGET_GET_PCH_VALIDITY (size_t *SZ) 32583 This hook returns the data needed by `TARGET_PCH_VALID_P' and sets 32584 `*SZ' to the size of the data in bytes. 32585 32586 -- Target Hook: const char *TARGET_PCH_VALID_P (const void *DATA, 32587 size_t SZ) 32588 This hook checks whether the options used to create a PCH file are 32589 compatible with the current settings. It returns `NULL' if so and 32590 a suitable error message if not. Error messages will be presented 32591 to the user and must be localized using `_(MSG)'. 32592 32593 DATA is the data that was returned by `TARGET_GET_PCH_VALIDITY' 32594 when the PCH file was created and SZ is the size of that data in 32595 bytes. It's safe to assume that the data was created by the same 32596 version of the compiler, so no format checking is needed. 32597 32598 The default definition of `default_pch_valid_p' should be suitable 32599 for most targets. 32600 32601 -- Target Hook: const char *TARGET_CHECK_PCH_TARGET_FLAGS (int 32602 PCH_FLAGS) 32603 If this hook is nonnull, the default implementation of 32604 `TARGET_PCH_VALID_P' will use it to check for compatible values of 32605 `target_flags'. PCH_FLAGS specifies the value that `target_flags' 32606 had when the PCH file was created. The return value is the same 32607 as for `TARGET_PCH_VALID_P'. 32608 32609 32610 File: gccint.info, Node: C++ ABI, Next: Misc, Prev: PCH Target, Up: Target Macros 32611 32612 17.29 C++ ABI parameters 32613 ======================== 32614 32615 -- Target Hook: tree TARGET_CXX_GUARD_TYPE (void) 32616 Define this hook to override the integer type used for guard 32617 variables. These are used to implement one-time construction of 32618 static objects. The default is long_long_integer_type_node. 32619 32620 -- Target Hook: bool TARGET_CXX_GUARD_MASK_BIT (void) 32621 This hook determines how guard variables are used. It should 32622 return `false' (the default) if first byte should be used. A 32623 return value of `true' indicates the least significant bit should 32624 be used. 32625 32626 -- Target Hook: tree TARGET_CXX_GET_COOKIE_SIZE (tree TYPE) 32627 This hook returns the size of the cookie to use when allocating an 32628 array whose elements have the indicated TYPE. Assumes that it is 32629 already known that a cookie is needed. The default is `max(sizeof 32630 (size_t), alignof(type))', as defined in section 2.7 of the 32631 IA64/Generic C++ ABI. 32632 32633 -- Target Hook: bool TARGET_CXX_COOKIE_HAS_SIZE (void) 32634 This hook should return `true' if the element size should be 32635 stored in array cookies. The default is to return `false'. 32636 32637 -- Target Hook: int TARGET_CXX_IMPORT_EXPORT_CLASS (tree TYPE, int 32638 IMPORT_EXPORT) 32639 If defined by a backend this hook allows the decision made to 32640 export class TYPE to be overruled. Upon entry IMPORT_EXPORT will 32641 contain 1 if the class is going to be exported, -1 if it is going 32642 to be imported and 0 otherwise. This function should return the 32643 modified value and perform any other actions necessary to support 32644 the backend's targeted operating system. 32645 32646 -- Target Hook: bool TARGET_CXX_CDTOR_RETURNS_THIS (void) 32647 This hook should return `true' if constructors and destructors 32648 return the address of the object created/destroyed. The default 32649 is to return `false'. 32650 32651 -- Target Hook: bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void) 32652 This hook returns true if the key method for a class (i.e., the 32653 method which, if defined in the current translation unit, causes 32654 the virtual table to be emitted) may be an inline function. Under 32655 the standard Itanium C++ ABI the key method may be an inline 32656 function so long as the function is not declared inline in the 32657 class definition. Under some variants of the ABI, an inline 32658 function can never be the key method. The default is to return 32659 `true'. 32660 32661 -- Target Hook: void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree 32662 DECL) 32663 DECL is a virtual table, virtual table table, typeinfo object, or 32664 other similar implicit class data object that will be emitted with 32665 external linkage in this translation unit. No ELF visibility has 32666 been explicitly specified. If the target needs to specify a 32667 visibility other than that of the containing class, use this hook 32668 to set `DECL_VISIBILITY' and `DECL_VISIBILITY_SPECIFIED'. 32669 32670 -- Target Hook: bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void) 32671 This hook returns true (the default) if virtual tables and other 32672 similar implicit class data objects are always COMDAT if they have 32673 external linkage. If this hook returns false, then class data for 32674 classes whose virtual table will be emitted in only one translation 32675 unit will not be COMDAT. 32676 32677 -- Target Hook: bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void) 32678 This hook returns true (the default) if the RTTI information for 32679 the basic types which is defined in the C++ runtime should always 32680 be COMDAT, false if it should not be COMDAT. 32681 32682 -- Target Hook: bool TARGET_CXX_USE_AEABI_ATEXIT (void) 32683 This hook returns true if `__aeabi_atexit' (as defined by the ARM 32684 EABI) should be used to register static destructors when 32685 `-fuse-cxa-atexit' is in effect. The default is to return false 32686 to use `__cxa_atexit'. 32687 32688 -- Target Hook: bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void) 32689 This hook returns true if the target `atexit' function can be used 32690 in the same manner as `__cxa_atexit' to register C++ static 32691 destructors. This requires that `atexit'-registered functions in 32692 shared libraries are run in the correct order when the libraries 32693 are unloaded. The default is to return false. 32694 32695 -- Target Hook: void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree TYPE) 32696 TYPE is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has 32697 just been defined. Use this hook to make adjustments to the class 32698 (eg, tweak visibility or perform any other required target 32699 modifications). 32700 32701 32702 File: gccint.info, Node: Misc, Prev: C++ ABI, Up: Target Macros 32703 32704 17.30 Miscellaneous Parameters 32705 ============================== 32706 32707 Here are several miscellaneous parameters. 32708 32709 -- Macro: HAS_LONG_COND_BRANCH 32710 Define this boolean macro to indicate whether or not your 32711 architecture has conditional branches that can span all of memory. 32712 It is used in conjunction with an optimization that partitions hot 32713 and cold basic blocks into separate sections of the executable. 32714 If this macro is set to false, gcc will convert any conditional 32715 branches that attempt to cross between sections into unconditional 32716 branches or indirect jumps. 32717 32718 -- Macro: HAS_LONG_UNCOND_BRANCH 32719 Define this boolean macro to indicate whether or not your 32720 architecture has unconditional branches that can span all of 32721 memory. It is used in conjunction with an optimization that 32722 partitions hot and cold basic blocks into separate sections of the 32723 executable. If this macro is set to false, gcc will convert any 32724 unconditional branches that attempt to cross between sections into 32725 indirect jumps. 32726 32727 -- Macro: CASE_VECTOR_MODE 32728 An alias for a machine mode name. This is the machine mode that 32729 elements of a jump-table should have. 32730 32731 -- Macro: CASE_VECTOR_SHORTEN_MODE (MIN_OFFSET, MAX_OFFSET, BODY) 32732 Optional: return the preferred mode for an `addr_diff_vec' when 32733 the minimum and maximum offset are known. If you define this, it 32734 enables extra code in branch shortening to deal with 32735 `addr_diff_vec'. To make this work, you also have to define 32736 `INSN_ALIGN' and make the alignment for `addr_diff_vec' explicit. 32737 The BODY argument is provided so that the offset_unsigned and scale 32738 flags can be updated. 32739 32740 -- Macro: CASE_VECTOR_PC_RELATIVE 32741 Define this macro to be a C expression to indicate when jump-tables 32742 should contain relative addresses. You need not define this macro 32743 if jump-tables never contain relative addresses, or jump-tables 32744 should contain relative addresses only when `-fPIC' or `-fPIC' is 32745 in effect. 32746 32747 -- Macro: CASE_VALUES_THRESHOLD 32748 Define this to be the smallest number of different values for 32749 which it is best to use a jump-table instead of a tree of 32750 conditional branches. The default is four for machines with a 32751 `casesi' instruction and five otherwise. This is best for most 32752 machines. 32753 32754 -- Macro: CASE_USE_BIT_TESTS 32755 Define this macro to be a C expression to indicate whether C switch 32756 statements may be implemented by a sequence of bit tests. This is 32757 advantageous on processors that can efficiently implement left 32758 shift of 1 by the number of bits held in a register, but 32759 inappropriate on targets that would require a loop. By default, 32760 this macro returns `true' if the target defines an `ashlsi3' 32761 pattern, and `false' otherwise. 32762 32763 -- Macro: WORD_REGISTER_OPERATIONS 32764 Define this macro if operations between registers with integral 32765 mode smaller than a word are always performed on the entire 32766 register. Most RISC machines have this property and most CISC 32767 machines do not. 32768 32769 -- Macro: LOAD_EXTEND_OP (MEM_MODE) 32770 Define this macro to be a C expression indicating when insns that 32771 read memory in MEM_MODE, an integral mode narrower than a word, 32772 set the bits outside of MEM_MODE to be either the sign-extension 32773 or the zero-extension of the data read. Return `SIGN_EXTEND' for 32774 values of MEM_MODE for which the insn sign-extends, `ZERO_EXTEND' 32775 for which it zero-extends, and `UNKNOWN' for other modes. 32776 32777 This macro is not called with MEM_MODE non-integral or with a width 32778 greater than or equal to `BITS_PER_WORD', so you may return any 32779 value in this case. Do not define this macro if it would always 32780 return `UNKNOWN'. On machines where this macro is defined, you 32781 will normally define it as the constant `SIGN_EXTEND' or 32782 `ZERO_EXTEND'. 32783 32784 You may return a non-`UNKNOWN' value even if for some hard 32785 registers the sign extension is not performed, if for the 32786 `REGNO_REG_CLASS' of these hard registers 32787 `CANNOT_CHANGE_MODE_CLASS' returns nonzero when the FROM mode is 32788 MEM_MODE and the TO mode is any integral mode larger than this but 32789 not larger than `word_mode'. 32790 32791 You must return `UNKNOWN' if for some hard registers that allow 32792 this mode, `CANNOT_CHANGE_MODE_CLASS' says that they cannot change 32793 to `word_mode', but that they can change to another integral mode 32794 that is larger then MEM_MODE but still smaller than `word_mode'. 32795 32796 -- Macro: SHORT_IMMEDIATES_SIGN_EXTEND 32797 Define this macro if loading short immediate values into registers 32798 sign extends. 32799 32800 -- Macro: FIXUNS_TRUNC_LIKE_FIX_TRUNC 32801 Define this macro if the same instructions that convert a floating 32802 point number to a signed fixed point number also convert validly 32803 to an unsigned one. 32804 32805 -- Target Hook: int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum 32806 machine_mode MODE) 32807 When `-ffast-math' is in effect, GCC tries to optimize divisions 32808 by the same divisor, by turning them into multiplications by the 32809 reciprocal. This target hook specifies the minimum number of 32810 divisions that should be there for GCC to perform the optimization 32811 for a variable of mode MODE. The default implementation returns 3 32812 if the machine has an instruction for the division, and 2 if it 32813 does not. 32814 32815 -- Macro: MOVE_MAX 32816 The maximum number of bytes that a single instruction can move 32817 quickly between memory and registers or between two memory 32818 locations. 32819 32820 -- Macro: MAX_MOVE_MAX 32821 The maximum number of bytes that a single instruction can move 32822 quickly between memory and registers or between two memory 32823 locations. If this is undefined, the default is `MOVE_MAX'. 32824 Otherwise, it is the constant value that is the largest value that 32825 `MOVE_MAX' can have at run-time. 32826 32827 -- Macro: SHIFT_COUNT_TRUNCATED 32828 A C expression that is nonzero if on this machine the number of 32829 bits actually used for the count of a shift operation is equal to 32830 the number of bits needed to represent the size of the object 32831 being shifted. When this macro is nonzero, the compiler will 32832 assume that it is safe to omit a sign-extend, zero-extend, and 32833 certain bitwise `and' instructions that truncates the count of a 32834 shift operation. On machines that have instructions that act on 32835 bit-fields at variable positions, which may include `bit test' 32836 instructions, a nonzero `SHIFT_COUNT_TRUNCATED' also enables 32837 deletion of truncations of the values that serve as arguments to 32838 bit-field instructions. 32839 32840 If both types of instructions truncate the count (for shifts) and 32841 position (for bit-field operations), or if no variable-position 32842 bit-field instructions exist, you should define this macro. 32843 32844 However, on some machines, such as the 80386 and the 680x0, 32845 truncation only applies to shift operations and not the (real or 32846 pretended) bit-field operations. Define `SHIFT_COUNT_TRUNCATED' 32847 to be zero on such machines. Instead, add patterns to the `md' 32848 file that include the implied truncation of the shift instructions. 32849 32850 You need not define this macro if it would always have the value 32851 of zero. 32852 32853 -- Target Hook: int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode 32854 MODE) 32855 This function describes how the standard shift patterns for MODE 32856 deal with shifts by negative amounts or by more than the width of 32857 the mode. *Note shift patterns::. 32858 32859 On many machines, the shift patterns will apply a mask M to the 32860 shift count, meaning that a fixed-width shift of X by Y is 32861 equivalent to an arbitrary-width shift of X by Y & M. If this is 32862 true for mode MODE, the function should return M, otherwise it 32863 should return 0. A return value of 0 indicates that no particular 32864 behavior is guaranteed. 32865 32866 Note that, unlike `SHIFT_COUNT_TRUNCATED', this function does 32867 _not_ apply to general shift rtxes; it applies only to instructions 32868 that are generated by the named shift patterns. 32869 32870 The default implementation of this function returns 32871 `GET_MODE_BITSIZE (MODE) - 1' if `SHIFT_COUNT_TRUNCATED' and 0 32872 otherwise. This definition is always safe, but if 32873 `SHIFT_COUNT_TRUNCATED' is false, and some shift patterns 32874 nevertheless truncate the shift count, you may get better code by 32875 overriding it. 32876 32877 -- Macro: TRULY_NOOP_TRUNCATION (OUTPREC, INPREC) 32878 A C expression which is nonzero if on this machine it is safe to 32879 "convert" an integer of INPREC bits to one of OUTPREC bits (where 32880 OUTPREC is smaller than INPREC) by merely operating on it as if it 32881 had only OUTPREC bits. 32882 32883 On many machines, this expression can be 1. 32884 32885 When `TRULY_NOOP_TRUNCATION' returns 1 for a pair of sizes for 32886 modes for which `MODES_TIEABLE_P' is 0, suboptimal code can result. 32887 If this is the case, making `TRULY_NOOP_TRUNCATION' return 0 in 32888 such cases may improve things. 32889 32890 -- Target Hook: int TARGET_MODE_REP_EXTENDED (enum machine_mode MODE, 32891 enum machine_mode REP_MODE) 32892 The representation of an integral mode can be such that the values 32893 are always extended to a wider integral mode. Return 32894 `SIGN_EXTEND' if values of MODE are represented in sign-extended 32895 form to REP_MODE. Return `UNKNOWN' otherwise. (Currently, none 32896 of the targets use zero-extended representation this way so unlike 32897 `LOAD_EXTEND_OP', `TARGET_MODE_REP_EXTENDED' is expected to return 32898 either `SIGN_EXTEND' or `UNKNOWN'. Also no target extends MODE to 32899 MODE_REP so that MODE_REP is not the next widest integral mode and 32900 currently we take advantage of this fact.) 32901 32902 Similarly to `LOAD_EXTEND_OP' you may return a non-`UNKNOWN' value 32903 even if the extension is not performed on certain hard registers 32904 as long as for the `REGNO_REG_CLASS' of these hard registers 32905 `CANNOT_CHANGE_MODE_CLASS' returns nonzero. 32906 32907 Note that `TARGET_MODE_REP_EXTENDED' and `LOAD_EXTEND_OP' describe 32908 two related properties. If you define `TARGET_MODE_REP_EXTENDED 32909 (mode, word_mode)' you probably also want to define 32910 `LOAD_EXTEND_OP (mode)' to return the same type of extension. 32911 32912 In order to enforce the representation of `mode', 32913 `TRULY_NOOP_TRUNCATION' should return false when truncating to 32914 `mode'. 32915 32916 -- Macro: STORE_FLAG_VALUE 32917 A C expression describing the value returned by a comparison 32918 operator with an integral mode and stored by a store-flag 32919 instruction (`sCOND') when the condition is true. This 32920 description must apply to _all_ the `sCOND' patterns and all the 32921 comparison operators whose results have a `MODE_INT' mode. 32922 32923 A value of 1 or -1 means that the instruction implementing the 32924 comparison operator returns exactly 1 or -1 when the comparison is 32925 true and 0 when the comparison is false. Otherwise, the value 32926 indicates which bits of the result are guaranteed to be 1 when the 32927 comparison is true. This value is interpreted in the mode of the 32928 comparison operation, which is given by the mode of the first 32929 operand in the `sCOND' pattern. Either the low bit or the sign 32930 bit of `STORE_FLAG_VALUE' be on. Presently, only those bits are 32931 used by the compiler. 32932 32933 If `STORE_FLAG_VALUE' is neither 1 or -1, the compiler will 32934 generate code that depends only on the specified bits. It can also 32935 replace comparison operators with equivalent operations if they 32936 cause the required bits to be set, even if the remaining bits are 32937 undefined. For example, on a machine whose comparison operators 32938 return an `SImode' value and where `STORE_FLAG_VALUE' is defined as 32939 `0x80000000', saying that just the sign bit is relevant, the 32940 expression 32941 32942 (ne:SI (and:SI X (const_int POWER-OF-2)) (const_int 0)) 32943 32944 can be converted to 32945 32946 (ashift:SI X (const_int N)) 32947 32948 where N is the appropriate shift count to move the bit being 32949 tested into the sign bit. 32950 32951 There is no way to describe a machine that always sets the 32952 low-order bit for a true value, but does not guarantee the value 32953 of any other bits, but we do not know of any machine that has such 32954 an instruction. If you are trying to port GCC to such a machine, 32955 include an instruction to perform a logical-and of the result with 32956 1 in the pattern for the comparison operators and let us know at 32957 <gcc (a] gcc.gnu.org>. 32958 32959 Often, a machine will have multiple instructions that obtain a 32960 value from a comparison (or the condition codes). Here are rules 32961 to guide the choice of value for `STORE_FLAG_VALUE', and hence the 32962 instructions to be used: 32963 32964 * Use the shortest sequence that yields a valid definition for 32965 `STORE_FLAG_VALUE'. It is more efficient for the compiler to 32966 "normalize" the value (convert it to, e.g., 1 or 0) than for 32967 the comparison operators to do so because there may be 32968 opportunities to combine the normalization with other 32969 operations. 32970 32971 * For equal-length sequences, use a value of 1 or -1, with -1 32972 being slightly preferred on machines with expensive jumps and 32973 1 preferred on other machines. 32974 32975 * As a second choice, choose a value of `0x80000001' if 32976 instructions exist that set both the sign and low-order bits 32977 but do not define the others. 32978 32979 * Otherwise, use a value of `0x80000000'. 32980 32981 Many machines can produce both the value chosen for 32982 `STORE_FLAG_VALUE' and its negation in the same number of 32983 instructions. On those machines, you should also define a pattern 32984 for those cases, e.g., one matching 32985 32986 (set A (neg:M (ne:M B C))) 32987 32988 Some machines can also perform `and' or `plus' operations on 32989 condition code values with less instructions than the corresponding 32990 `sCOND' insn followed by `and' or `plus'. On those machines, 32991 define the appropriate patterns. Use the names `incscc' and 32992 `decscc', respectively, for the patterns which perform `plus' or 32993 `minus' operations on condition code values. See `rs6000.md' for 32994 some examples. The GNU Superoptizer can be used to find such 32995 instruction sequences on other machines. 32996 32997 If this macro is not defined, the default value, 1, is used. You 32998 need not define `STORE_FLAG_VALUE' if the machine has no store-flag 32999 instructions, or if the value generated by these instructions is 1. 33000 33001 -- Macro: FLOAT_STORE_FLAG_VALUE (MODE) 33002 A C expression that gives a nonzero `REAL_VALUE_TYPE' value that is 33003 returned when comparison operators with floating-point results are 33004 true. Define this macro on machines that have comparison 33005 operations that return floating-point values. If there are no 33006 such operations, do not define this macro. 33007 33008 -- Macro: VECTOR_STORE_FLAG_VALUE (MODE) 33009 A C expression that gives a rtx representing the nonzero true 33010 element for vector comparisons. The returned rtx should be valid 33011 for the inner mode of MODE which is guaranteed to be a vector 33012 mode. Define this macro on machines that have vector comparison 33013 operations that return a vector result. If there are no such 33014 operations, do not define this macro. Typically, this macro is 33015 defined as `const1_rtx' or `constm1_rtx'. This macro may return 33016 `NULL_RTX' to prevent the compiler optimizing such vector 33017 comparison operations for the given mode. 33018 33019 -- Macro: CLZ_DEFINED_VALUE_AT_ZERO (MODE, VALUE) 33020 -- Macro: CTZ_DEFINED_VALUE_AT_ZERO (MODE, VALUE) 33021 A C expression that indicates whether the architecture defines a 33022 value for `clz' or `ctz' with a zero operand. A result of `0' 33023 indicates the value is undefined. If the value is defined for 33024 only the RTL expression, the macro should evaluate to `1'; if the 33025 value applies also to the corresponding optab entry (which is 33026 normally the case if it expands directly into the corresponding 33027 RTL), then the macro should evaluate to `2'. In the cases where 33028 the value is defined, VALUE should be set to this value. 33029 33030 If this macro is not defined, the value of `clz' or `ctz' at zero 33031 is assumed to be undefined. 33032 33033 This macro must be defined if the target's expansion for `ffs' 33034 relies on a particular value to get correct results. Otherwise it 33035 is not necessary, though it may be used to optimize some corner 33036 cases, and to provide a default expansion for the `ffs' optab. 33037 33038 Note that regardless of this macro the "definedness" of `clz' and 33039 `ctz' at zero do _not_ extend to the builtin functions visible to 33040 the user. Thus one may be free to adjust the value at will to 33041 match the target expansion of these operations without fear of 33042 breaking the API. 33043 33044 -- Macro: Pmode 33045 An alias for the machine mode for pointers. On most machines, 33046 define this to be the integer mode corresponding to the width of a 33047 hardware pointer; `SImode' on 32-bit machine or `DImode' on 64-bit 33048 machines. On some machines you must define this to be one of the 33049 partial integer modes, such as `PSImode'. 33050 33051 The width of `Pmode' must be at least as large as the value of 33052 `POINTER_SIZE'. If it is not equal, you must define the macro 33053 `POINTERS_EXTEND_UNSIGNED' to specify how pointers are extended to 33054 `Pmode'. 33055 33056 -- Macro: FUNCTION_MODE 33057 An alias for the machine mode used for memory references to 33058 functions being called, in `call' RTL expressions. On most CISC 33059 machines, where an instruction can begin at any byte address, this 33060 should be `QImode'. On most RISC machines, where all instructions 33061 have fixed size and alignment, this should be a mode with the same 33062 size and alignment as the machine instruction words - typically 33063 `SImode' or `HImode'. 33064 33065 -- Macro: STDC_0_IN_SYSTEM_HEADERS 33066 In normal operation, the preprocessor expands `__STDC__' to the 33067 constant 1, to signify that GCC conforms to ISO Standard C. On 33068 some hosts, like Solaris, the system compiler uses a different 33069 convention, where `__STDC__' is normally 0, but is 1 if the user 33070 specifies strict conformance to the C Standard. 33071 33072 Defining `STDC_0_IN_SYSTEM_HEADERS' makes GNU CPP follows the host 33073 convention when processing system header files, but when 33074 processing user files `__STDC__' will always expand to 1. 33075 33076 -- Macro: NO_IMPLICIT_EXTERN_C 33077 Define this macro if the system header files support C++ as well 33078 as C. This macro inhibits the usual method of using system header 33079 files in C++, which is to pretend that the file's contents are 33080 enclosed in `extern "C" {...}'. 33081 33082 -- Macro: REGISTER_TARGET_PRAGMAS () 33083 Define this macro if you want to implement any target-specific 33084 pragmas. If defined, it is a C expression which makes a series of 33085 calls to `c_register_pragma' or `c_register_pragma_with_expansion' 33086 for each pragma. The macro may also do any setup required for the 33087 pragmas. 33088 33089 The primary reason to define this macro is to provide 33090 compatibility with other compilers for the same target. In 33091 general, we discourage definition of target-specific pragmas for 33092 GCC. 33093 33094 If the pragma can be implemented by attributes then you should 33095 consider defining the target hook `TARGET_INSERT_ATTRIBUTES' as 33096 well. 33097 33098 Preprocessor macros that appear on pragma lines are not expanded. 33099 All `#pragma' directives that do not match any registered pragma 33100 are silently ignored, unless the user specifies 33101 `-Wunknown-pragmas'. 33102 33103 -- Function: void c_register_pragma (const char *SPACE, const char 33104 *NAME, void (*CALLBACK) (struct cpp_reader *)) 33105 -- Function: void c_register_pragma_with_expansion (const char *SPACE, 33106 const char *NAME, void (*CALLBACK) (struct cpp_reader *)) 33107 Each call to `c_register_pragma' or 33108 `c_register_pragma_with_expansion' establishes one pragma. The 33109 CALLBACK routine will be called when the preprocessor encounters a 33110 pragma of the form 33111 33112 #pragma [SPACE] NAME ... 33113 33114 SPACE is the case-sensitive namespace of the pragma, or `NULL' to 33115 put the pragma in the global namespace. The callback routine 33116 receives PFILE as its first argument, which can be passed on to 33117 cpplib's functions if necessary. You can lex tokens after the 33118 NAME by calling `pragma_lex'. Tokens that are not read by the 33119 callback will be silently ignored. The end of the line is 33120 indicated by a token of type `CPP_EOF'. Macro expansion occurs on 33121 the arguments of pragmas registered with 33122 `c_register_pragma_with_expansion' but not on the arguments of 33123 pragmas registered with `c_register_pragma'. 33124 33125 Note that the use of `pragma_lex' is specific to the C and C++ 33126 compilers. It will not work in the Java or Fortran compilers, or 33127 any other language compilers for that matter. Thus if 33128 `pragma_lex' is going to be called from target-specific code, it 33129 must only be done so when building the C and C++ compilers. This 33130 can be done by defining the variables `c_target_objs' and 33131 `cxx_target_objs' in the target entry in the `config.gcc' file. 33132 These variables should name the target-specific, language-specific 33133 object file which contains the code that uses `pragma_lex'. Note 33134 it will also be necessary to add a rule to the makefile fragment 33135 pointed to by `tmake_file' that shows how to build this object 33136 file. 33137 33138 -- Macro: HANDLE_SYSV_PRAGMA 33139 Define this macro (to a value of 1) if you want the System V style 33140 pragmas `#pragma pack(<n>)' and `#pragma weak <name> [=<value>]' 33141 to be supported by gcc. 33142 33143 The pack pragma specifies the maximum alignment (in bytes) of 33144 fields within a structure, in much the same way as the 33145 `__aligned__' and `__packed__' `__attribute__'s do. A pack value 33146 of zero resets the behavior to the default. 33147 33148 A subtlety for Microsoft Visual C/C++ style bit-field packing 33149 (e.g. -mms-bitfields) for targets that support it: When a 33150 bit-field is inserted into a packed record, the whole size of the 33151 underlying type is used by one or more same-size adjacent 33152 bit-fields (that is, if its long:3, 32 bits is used in the record, 33153 and any additional adjacent long bit-fields are packed into the 33154 same chunk of 32 bits. However, if the size changes, a new field 33155 of that size is allocated). 33156 33157 If both MS bit-fields and `__attribute__((packed))' are used, the 33158 latter will take precedence. If `__attribute__((packed))' is used 33159 on a single field when MS bit-fields are in use, it will take 33160 precedence for that field, but the alignment of the rest of the 33161 structure may affect its placement. 33162 33163 The weak pragma only works if `SUPPORTS_WEAK' and 33164 `ASM_WEAKEN_LABEL' are defined. If enabled it allows the creation 33165 of specifically named weak labels, optionally with a value. 33166 33167 -- Macro: HANDLE_PRAGMA_PACK_PUSH_POP 33168 Define this macro (to a value of 1) if you want to support the 33169 Win32 style pragmas `#pragma pack(push[,N])' and `#pragma 33170 pack(pop)'. The `pack(push,[N])' pragma specifies the maximum 33171 alignment (in bytes) of fields within a structure, in much the 33172 same way as the `__aligned__' and `__packed__' `__attribute__'s 33173 do. A pack value of zero resets the behavior to the default. 33174 Successive invocations of this pragma cause the previous values to 33175 be stacked, so that invocations of `#pragma pack(pop)' will return 33176 to the previous value. 33177 33178 -- Macro: HANDLE_PRAGMA_PACK_WITH_EXPANSION 33179 Define this macro, as well as `HANDLE_SYSV_PRAGMA', if macros 33180 should be expanded in the arguments of `#pragma pack'. 33181 33182 -- Macro: TARGET_DEFAULT_PACK_STRUCT 33183 If your target requires a structure packing default other than 0 33184 (meaning the machine default), define this macro to the necessary 33185 value (in bytes). This must be a value that would also be valid 33186 to use with `#pragma pack()' (that is, a small power of two). 33187 33188 -- Macro: DOLLARS_IN_IDENTIFIERS 33189 Define this macro to control use of the character `$' in 33190 identifier names for the C family of languages. 0 means `$' is 33191 not allowed by default; 1 means it is allowed. 1 is the default; 33192 there is no need to define this macro in that case. 33193 33194 -- Macro: NO_DOLLAR_IN_LABEL 33195 Define this macro if the assembler does not accept the character 33196 `$' in label names. By default constructors and destructors in 33197 G++ have `$' in the identifiers. If this macro is defined, `.' is 33198 used instead. 33199 33200 -- Macro: NO_DOT_IN_LABEL 33201 Define this macro if the assembler does not accept the character 33202 `.' in label names. By default constructors and destructors in G++ 33203 have names that use `.'. If this macro is defined, these names 33204 are rewritten to avoid `.'. 33205 33206 -- Macro: INSN_SETS_ARE_DELAYED (INSN) 33207 Define this macro as a C expression that is nonzero if it is safe 33208 for the delay slot scheduler to place instructions in the delay 33209 slot of INSN, even if they appear to use a resource set or 33210 clobbered in INSN. INSN is always a `jump_insn' or an `insn'; GCC 33211 knows that every `call_insn' has this behavior. On machines where 33212 some `insn' or `jump_insn' is really a function call and hence has 33213 this behavior, you should define this macro. 33214 33215 You need not define this macro if it would always return zero. 33216 33217 -- Macro: INSN_REFERENCES_ARE_DELAYED (INSN) 33218 Define this macro as a C expression that is nonzero if it is safe 33219 for the delay slot scheduler to place instructions in the delay 33220 slot of INSN, even if they appear to set or clobber a resource 33221 referenced in INSN. INSN is always a `jump_insn' or an `insn'. 33222 On machines where some `insn' or `jump_insn' is really a function 33223 call and its operands are registers whose use is actually in the 33224 subroutine it calls, you should define this macro. Doing so 33225 allows the delay slot scheduler to move instructions which copy 33226 arguments into the argument registers into the delay slot of INSN. 33227 33228 You need not define this macro if it would always return zero. 33229 33230 -- Macro: MULTIPLE_SYMBOL_SPACES 33231 Define this macro as a C expression that is nonzero if, in some 33232 cases, global symbols from one translation unit may not be bound 33233 to undefined symbols in another translation unit without user 33234 intervention. For instance, under Microsoft Windows symbols must 33235 be explicitly imported from shared libraries (DLLs). 33236 33237 You need not define this macro if it would always evaluate to zero. 33238 33239 -- Target Hook: tree TARGET_MD_ASM_CLOBBERS (tree OUTPUTS, tree 33240 INPUTS, tree CLOBBERS) 33241 This target hook should add to CLOBBERS `STRING_CST' trees for any 33242 hard regs the port wishes to automatically clobber for an asm. It 33243 should return the result of the last `tree_cons' used to add a 33244 clobber. The OUTPUTS, INPUTS and CLOBBER lists are the 33245 corresponding parameters to the asm and may be inspected to avoid 33246 clobbering a register that is an input or output of the asm. You 33247 can use `tree_overlaps_hard_reg_set', declared in `tree.h', to test 33248 for overlap with regards to asm-declared registers. 33249 33250 -- Macro: MATH_LIBRARY 33251 Define this macro as a C string constant for the linker argument 33252 to link in the system math library, or `""' if the target does not 33253 have a separate math library. 33254 33255 You need only define this macro if the default of `"-lm"' is wrong. 33256 33257 -- Macro: LIBRARY_PATH_ENV 33258 Define this macro as a C string constant for the environment 33259 variable that specifies where the linker should look for libraries. 33260 33261 You need only define this macro if the default of `"LIBRARY_PATH"' 33262 is wrong. 33263 33264 -- Macro: TARGET_POSIX_IO 33265 Define this macro if the target supports the following POSIX file 33266 functions, access, mkdir and file locking with fcntl / F_SETLKW. 33267 Defining `TARGET_POSIX_IO' will enable the test coverage code to 33268 use file locking when exiting a program, which avoids race 33269 conditions if the program has forked. It will also create 33270 directories at run-time for cross-profiling. 33271 33272 -- Macro: MAX_CONDITIONAL_EXECUTE 33273 A C expression for the maximum number of instructions to execute 33274 via conditional execution instructions instead of a branch. A 33275 value of `BRANCH_COST'+1 is the default if the machine does not 33276 use cc0, and 1 if it does use cc0. 33277 33278 -- Macro: IFCVT_MODIFY_TESTS (CE_INFO, TRUE_EXPR, FALSE_EXPR) 33279 Used if the target needs to perform machine-dependent 33280 modifications on the conditionals used for turning basic blocks 33281 into conditionally executed code. CE_INFO points to a data 33282 structure, `struct ce_if_block', which contains information about 33283 the currently processed blocks. TRUE_EXPR and FALSE_EXPR are the 33284 tests that are used for converting the then-block and the 33285 else-block, respectively. Set either TRUE_EXPR or FALSE_EXPR to a 33286 null pointer if the tests cannot be converted. 33287 33288 -- Macro: IFCVT_MODIFY_MULTIPLE_TESTS (CE_INFO, BB, TRUE_EXPR, 33289 FALSE_EXPR) 33290 Like `IFCVT_MODIFY_TESTS', but used when converting more 33291 complicated if-statements into conditions combined by `and' and 33292 `or' operations. BB contains the basic block that contains the 33293 test that is currently being processed and about to be turned into 33294 a condition. 33295 33296 -- Macro: IFCVT_MODIFY_INSN (CE_INFO, PATTERN, INSN) 33297 A C expression to modify the PATTERN of an INSN that is to be 33298 converted to conditional execution format. CE_INFO points to a 33299 data structure, `struct ce_if_block', which contains information 33300 about the currently processed blocks. 33301 33302 -- Macro: IFCVT_MODIFY_FINAL (CE_INFO) 33303 A C expression to perform any final machine dependent 33304 modifications in converting code to conditional execution. The 33305 involved basic blocks can be found in the `struct ce_if_block' 33306 structure that is pointed to by CE_INFO. 33307 33308 -- Macro: IFCVT_MODIFY_CANCEL (CE_INFO) 33309 A C expression to cancel any machine dependent modifications in 33310 converting code to conditional execution. The involved basic 33311 blocks can be found in the `struct ce_if_block' structure that is 33312 pointed to by CE_INFO. 33313 33314 -- Macro: IFCVT_INIT_EXTRA_FIELDS (CE_INFO) 33315 A C expression to initialize any extra fields in a `struct 33316 ce_if_block' structure, which are defined by the 33317 `IFCVT_EXTRA_FIELDS' macro. 33318 33319 -- Macro: IFCVT_EXTRA_FIELDS 33320 If defined, it should expand to a set of field declarations that 33321 will be added to the `struct ce_if_block' structure. These should 33322 be initialized by the `IFCVT_INIT_EXTRA_FIELDS' macro. 33323 33324 -- Target Hook: void TARGET_MACHINE_DEPENDENT_REORG () 33325 If non-null, this hook performs a target-specific pass over the 33326 instruction stream. The compiler will run it at all optimization 33327 levels, just before the point at which it normally does 33328 delayed-branch scheduling. 33329 33330 The exact purpose of the hook varies from target to target. Some 33331 use it to do transformations that are necessary for correctness, 33332 such as laying out in-function constant pools or avoiding hardware 33333 hazards. Others use it as an opportunity to do some 33334 machine-dependent optimizations. 33335 33336 You need not implement the hook if it has nothing to do. The 33337 default definition is null. 33338 33339 -- Target Hook: void TARGET_INIT_BUILTINS () 33340 Define this hook if you have any machine-specific built-in 33341 functions that need to be defined. It should be a function that 33342 performs the necessary setup. 33343 33344 Machine specific built-in functions can be useful to expand 33345 special machine instructions that would otherwise not normally be 33346 generated because they have no equivalent in the source language 33347 (for example, SIMD vector instructions or prefetch instructions). 33348 33349 To create a built-in function, call the function 33350 `lang_hooks.builtin_function' which is defined by the language 33351 front end. You can use any type nodes set up by 33352 `build_common_tree_nodes' and `build_common_tree_nodes_2'; only 33353 language front ends that use those two functions will call 33354 `TARGET_INIT_BUILTINS'. 33355 33356 -- Target Hook: rtx TARGET_EXPAND_BUILTIN (tree EXP, rtx TARGET, rtx 33357 SUBTARGET, enum machine_mode MODE, int IGNORE) 33358 Expand a call to a machine specific built-in function that was set 33359 up by `TARGET_INIT_BUILTINS'. EXP is the expression for the 33360 function call; the result should go to TARGET if that is 33361 convenient, and have mode MODE if that is convenient. SUBTARGET 33362 may be used as the target for computing one of EXP's operands. 33363 IGNORE is nonzero if the value is to be ignored. This function 33364 should return the result of the call to the built-in function. 33365 33366 -- Target Hook: tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree FNDECL, 33367 tree ARGLIST) 33368 Select a replacement for a machine specific built-in function that 33369 was set up by `TARGET_INIT_BUILTINS'. This is done _before_ 33370 regular type checking, and so allows the target to implement a 33371 crude form of function overloading. FNDECL is the declaration of 33372 the built-in function. ARGLIST is the list of arguments passed to 33373 the built-in function. The result is a complete expression that 33374 implements the operation, usually another `CALL_EXPR'. 33375 33376 -- Target Hook: tree TARGET_FOLD_BUILTIN (tree FNDECL, tree ARGLIST, 33377 bool IGNORE) 33378 Fold a call to a machine specific built-in function that was set 33379 up by `TARGET_INIT_BUILTINS'. FNDECL is the declaration of the 33380 built-in function. ARGLIST is the list of arguments passed to the 33381 built-in function. The result is another tree containing a 33382 simplified expression for the call's result. If IGNORE is true 33383 the value will be ignored. 33384 33385 -- Target Hook: const char * TARGET_INVALID_WITHIN_DOLOOP (rtx INSN) 33386 Take an instruction in INSN and return NULL if it is valid within a 33387 low-overhead loop, otherwise return a string why doloop could not 33388 be applied. 33389 33390 Many targets use special registers for low-overhead looping. For 33391 any instruction that clobbers these this function should return a 33392 string indicating the reason why the doloop could not be applied. 33393 By default, the RTL loop optimizer does not use a present doloop 33394 pattern for loops containing function calls or branch on table 33395 instructions. 33396 33397 -- Macro: MD_CAN_REDIRECT_BRANCH (BRANCH1, BRANCH2) 33398 Take a branch insn in BRANCH1 and another in BRANCH2. Return true 33399 if redirecting BRANCH1 to the destination of BRANCH2 is possible. 33400 33401 On some targets, branches may have a limited range. Optimizing the 33402 filling of delay slots can result in branches being redirected, 33403 and this may in turn cause a branch offset to overflow. 33404 33405 -- Target Hook: bool TARGET_COMMUTATIVE_P (rtx X, OUTER_CODE) 33406 This target hook returns `true' if X is considered to be 33407 commutative. Usually, this is just COMMUTATIVE_P (X), but the HP 33408 PA doesn't consider PLUS to be commutative inside a MEM. 33409 OUTER_CODE is the rtx code of the enclosing rtl, if known, 33410 otherwise it is UNKNOWN. 33411 33412 -- Target Hook: rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx HARD_REG) 33413 When the initial value of a hard register has been copied in a 33414 pseudo register, it is often not necessary to actually allocate 33415 another register to this pseudo register, because the original 33416 hard register or a stack slot it has been saved into can be used. 33417 `TARGET_ALLOCATE_INITIAL_VALUE' is called at the start of register 33418 allocation once for each hard register that had its initial value 33419 copied by using `get_func_hard_reg_initial_val' or 33420 `get_hard_reg_initial_val'. Possible values are `NULL_RTX', if 33421 you don't want to do any special allocation, a `REG' rtx--that 33422 would typically be the hard register itself, if it is known not to 33423 be clobbered--or a `MEM'. If you are returning a `MEM', this is 33424 only a hint for the allocator; it might decide to use another 33425 register anyways. You may use `current_function_leaf_function' in 33426 the hook, functions that use `REG_N_SETS', to determine if the hard 33427 register in question will not be clobbered. The default value of 33428 this hook is `NULL', which disables any special allocation. 33429 33430 -- Target Hook: int TARGET_UNSPEC_MAY_TRAP_P (const_rtx X, unsigned 33431 FLAGS) 33432 This target hook returns nonzero if X, an `unspec' or 33433 `unspec_volatile' operation, might cause a trap. Targets can use 33434 this hook to enhance precision of analysis for `unspec' and 33435 `unspec_volatile' operations. You may call `may_trap_p_1' to 33436 analyze inner elements of X in which case FLAGS should be passed 33437 along. 33438 33439 -- Target Hook: void TARGET_SET_CURRENT_FUNCTION (tree DECL) 33440 The compiler invokes this hook whenever it changes its current 33441 function context (`cfun'). You can define this function if the 33442 back end needs to perform any initialization or reset actions on a 33443 per-function basis. For example, it may be used to implement 33444 function attributes that affect register usage or code generation 33445 patterns. The argument DECL is the declaration for the new 33446 function context, and may be null to indicate that the compiler 33447 has left a function context and is returning to processing at the 33448 top level. The default hook function does nothing. 33449 33450 GCC sets `cfun' to a dummy function context during initialization 33451 of some parts of the back end. The hook function is not invoked 33452 in this situation; you need not worry about the hook being invoked 33453 recursively, or when the back end is in a partially-initialized 33454 state. 33455 33456 -- Macro: TARGET_OBJECT_SUFFIX 33457 Define this macro to be a C string representing the suffix for 33458 object files on your target machine. If you do not define this 33459 macro, GCC will use `.o' as the suffix for object files. 33460 33461 -- Macro: TARGET_EXECUTABLE_SUFFIX 33462 Define this macro to be a C string representing the suffix to be 33463 automatically added to executable files on your target machine. 33464 If you do not define this macro, GCC will use the null string as 33465 the suffix for executable files. 33466 33467 -- Macro: COLLECT_EXPORT_LIST 33468 If defined, `collect2' will scan the individual object files 33469 specified on its command line and create an export list for the 33470 linker. Define this macro for systems like AIX, where the linker 33471 discards object files that are not referenced from `main' and uses 33472 export lists. 33473 33474 -- Macro: MODIFY_JNI_METHOD_CALL (MDECL) 33475 Define this macro to a C expression representing a variant of the 33476 method call MDECL, if Java Native Interface (JNI) methods must be 33477 invoked differently from other methods on your target. For 33478 example, on 32-bit Microsoft Windows, JNI methods must be invoked 33479 using the `stdcall' calling convention and this macro is then 33480 defined as this expression: 33481 33482 build_type_attribute_variant (MDECL, 33483 build_tree_list 33484 (get_identifier ("stdcall"), 33485 NULL)) 33486 33487 -- Target Hook: bool TARGET_CANNOT_MODIFY_JUMPS_P (void) 33488 This target hook returns `true' past the point in which new jump 33489 instructions could be created. On machines that require a 33490 register for every jump such as the SHmedia ISA of SH5, this point 33491 would typically be reload, so this target hook should be defined 33492 to a function such as: 33493 33494 static bool 33495 cannot_modify_jumps_past_reload_p () 33496 { 33497 return (reload_completed || reload_in_progress); 33498 } 33499 33500 -- Target Hook: int TARGET_BRANCH_TARGET_REGISTER_CLASS (void) 33501 This target hook returns a register class for which branch target 33502 register optimizations should be applied. All registers in this 33503 class should be usable interchangeably. After reload, registers 33504 in this class will be re-allocated and loads will be hoisted out 33505 of loops and be subjected to inter-block scheduling. 33506 33507 -- Target Hook: bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool 33508 AFTER_PROLOGUE_EPILOGUE_GEN) 33509 Branch target register optimization will by default exclude 33510 callee-saved registers that are not already live during the 33511 current function; if this target hook returns true, they will be 33512 included. The target code must than make sure that all target 33513 registers in the class returned by 33514 `TARGET_BRANCH_TARGET_REGISTER_CLASS' that might need saving are 33515 saved. AFTER_PROLOGUE_EPILOGUE_GEN indicates if prologues and 33516 epilogues have already been generated. Note, even if you only 33517 return true when AFTER_PROLOGUE_EPILOGUE_GEN is false, you still 33518 are likely to have to make special provisions in 33519 `INITIAL_ELIMINATION_OFFSET' to reserve space for caller-saved 33520 target registers. 33521 33522 -- Target Hook: bool TARGET_HAVE_CONDITIONAL_EXECUTION (void) 33523 This target hook returns true if the target supports conditional 33524 execution. This target hook is required only when the target has 33525 several different modes and they have different conditional 33526 execution capability, such as ARM. 33527 33528 -- Macro: POWI_MAX_MULTS 33529 If defined, this macro is interpreted as a signed integer C 33530 expression that specifies the maximum number of floating point 33531 multiplications that should be emitted when expanding 33532 exponentiation by an integer constant inline. When this value is 33533 defined, exponentiation requiring more than this number of 33534 multiplications is implemented by calling the system library's 33535 `pow', `powf' or `powl' routines. The default value places no 33536 upper bound on the multiplication count. 33537 33538 -- Macro: void TARGET_EXTRA_INCLUDES (const char *SYSROOT, const char 33539 *IPREFIX, int STDINC) 33540 This target hook should register any extra include files for the 33541 target. The parameter STDINC indicates if normal include files 33542 are present. The parameter SYSROOT is the system root directory. 33543 The parameter IPREFIX is the prefix for the gcc directory. 33544 33545 -- Macro: void TARGET_EXTRA_PRE_INCLUDES (const char *SYSROOT, const 33546 char *IPREFIX, int STDINC) 33547 This target hook should register any extra include files for the 33548 target before any standard headers. The parameter STDINC 33549 indicates if normal include files are present. The parameter 33550 SYSROOT is the system root directory. The parameter IPREFIX is 33551 the prefix for the gcc directory. 33552 33553 -- Macro: void TARGET_OPTF (char *PATH) 33554 This target hook should register special include paths for the 33555 target. The parameter PATH is the include to register. On Darwin 33556 systems, this is used for Framework includes, which have semantics 33557 that are different from `-I'. 33558 33559 -- Target Hook: bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree FNDECL) 33560 This target hook returns `true' if it is safe to use a local alias 33561 for a virtual function FNDECL when constructing thunks, `false' 33562 otherwise. By default, the hook returns `true' for all functions, 33563 if a target supports aliases (i.e. defines `ASM_OUTPUT_DEF'), 33564 `false' otherwise, 33565 33566 -- Macro: TARGET_FORMAT_TYPES 33567 If defined, this macro is the name of a global variable containing 33568 target-specific format checking information for the `-Wformat' 33569 option. The default is to have no target-specific format checks. 33570 33571 -- Macro: TARGET_N_FORMAT_TYPES 33572 If defined, this macro is the number of entries in 33573 `TARGET_FORMAT_TYPES'. 33574 33575 -- Macro: TARGET_OVERRIDES_FORMAT_ATTRIBUTES 33576 If defined, this macro is the name of a global variable containing 33577 target-specific format overrides for the `-Wformat' option. The 33578 default is to have no target-specific format overrides. If defined, 33579 `TARGET_FORMAT_TYPES' must be defined, too. 33580 33581 -- Macro: TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT 33582 If defined, this macro specifies the number of entries in 33583 `TARGET_OVERRIDES_FORMAT_ATTRIBUTES'. 33584 33585 -- Macro: TARGET_OVERRIDES_FORMAT_INIT 33586 If defined, this macro specifies the optional initialization 33587 routine for target specific customizations of the system printf 33588 and scanf formatter settings. 33589 33590 -- Target Hook: bool TARGET_RELAXED_ORDERING 33591 If set to `true', means that the target's memory model does not 33592 guarantee that loads which do not depend on one another will access 33593 main memory in the order of the instruction stream; if ordering is 33594 important, an explicit memory barrier must be used. This is true 33595 of many recent processors which implement a policy of "relaxed," 33596 "weak," or "release" memory consistency, such as Alpha, PowerPC, 33597 and ia64. The default is `false'. 33598 33599 -- Target Hook: const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN 33600 (tree TYPELIST, tree FUNCDECL, tree VAL) 33601 If defined, this macro returns the diagnostic message when it is 33602 illegal to pass argument VAL to function FUNCDECL with prototype 33603 TYPELIST. 33604 33605 -- Target Hook: const char * TARGET_INVALID_CONVERSION (tree FROMTYPE, 33606 tree TOTYPE) 33607 If defined, this macro returns the diagnostic message when it is 33608 invalid to convert from FROMTYPE to TOTYPE, or `NULL' if validity 33609 should be determined by the front end. 33610 33611 -- Target Hook: const char * TARGET_INVALID_UNARY_OP (int OP, tree 33612 TYPE) 33613 If defined, this macro returns the diagnostic message when it is 33614 invalid to apply operation OP (where unary plus is denoted by 33615 `CONVERT_EXPR') to an operand of type TYPE, or `NULL' if validity 33616 should be determined by the front end. 33617 33618 -- Target Hook: const char * TARGET_INVALID_BINARY_OP (int OP, tree 33619 TYPE1, tree TYPE2) 33620 If defined, this macro returns the diagnostic message when it is 33621 invalid to apply operation OP to operands of types TYPE1 and 33622 TYPE2, or `NULL' if validity should be determined by the front end. 33623 33624 -- Macro: TARGET_USE_JCR_SECTION 33625 This macro determines whether to use the JCR section to register 33626 Java classes. By default, TARGET_USE_JCR_SECTION is defined to 1 33627 if both SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, 33628 else 0. 33629 33630 -- Macro: OBJC_JBLEN 33631 This macro determines the size of the objective C jump buffer for 33632 the NeXT runtime. By default, OBJC_JBLEN is defined to an 33633 innocuous value. 33634 33635 -- Macro: LIBGCC2_UNWIND_ATTRIBUTE 33636 Define this macro if any target-specific attributes need to be 33637 attached to the functions in `libgcc' that provide low-level 33638 support for call stack unwinding. It is used in declarations in 33639 `unwind-generic.h' and the associated definitions of those 33640 functions. 33641 33642 -- Target Hook: void TARGET_UPDATE_STACK_BOUNDARY (void) 33643 Define this macro to update the current function stack boundary if 33644 necessary. 33645 33646 -- Target Hook: rtx TARGET_GET_DRAP_RTX (void) 33647 Define this macro to an rtx for Dynamic Realign Argument Pointer 33648 if a different argument pointer register is needed to access the 33649 function's argument list when stack is aligned. 33650 33651 -- Target Hook: bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void) 33652 When optimization is disabled, this hook indicates whether or not 33653 arguments should be allocated to stack slots. Normally, GCC 33654 allocates stacks slots for arguments when not optimizing in order 33655 to make debugging easier. However, when a function is declared 33656 with `__attribute__((naked))', there is no stack frame, and the 33657 compiler cannot safely move arguments from the registers in which 33658 they are passed to the stack. Therefore, this hook should return 33659 true in general, but false for naked functions. The default 33660 implementation always returns true. 33661 33662 -- Target Hook: rtx TARGET_GET_PIC_REG (void) 33663 Return the pic_reg pseudo register which holds the base address of 33664 GOT. It is only required by the simplify-got optimization. 33665 33666 -- Target Hook: void TARGET_CLEAR_PIC_REG (void) 33667 After successful simplify-got optimization, the pic_reg is 33668 useless. So a target can use this hook to clear pic_reg. 33669 33670 -- Target Hook: rtx TARGET_LOADED_GLOBAL_VAR (rtx INSN, rtx * 33671 OFFSET_REG, rtx * OFFSET_INSN) 33672 This hook is used to detect if the given INSN loads a global 33673 variable's address from GOT with the form of 33674 33675 (set ADDRESS_REG (mem (plus pic_reg OFFSET_REG))) 33676 33677 If so return the global variable whose address will be loaded and 33678 fill in OFFSET_INSN and OFFSET_REG. OFFSET_REG is set at 33679 OFFSET_INSN to hold the offset from GOT base to the GOT entry of 33680 the global variable. Otherwise return `NULL_RTX'. 33681 33682 -- Target Hook: bool TARGET_CAN_SIMPLIFY_GOT_ACCESS (int N_SYMBOL, int 33683 N_ACCESS) 33684 This hook determines if it satisfy the target dependent conditions 33685 to do simplify-got when given the number of global variable 33686 accessing and the number of accessed symbols. If the returned 33687 value is false the GOT access insns will not be rewritten. 33688 Otherwise we will rewrite these insns. 33689 33690 -- Target Hook: void TARGET_LOAD_GLOBAL_ADDRESS (rtx SYMBOL, rtx 33691 OFFSET_REG, rtx ADDRESS_REG, rtx LOAD_INSN, rtx OFFSET_INSN) 33692 This hook does the actual rewriting of GOT access insn LOAD_INSN. 33693 The global variable is SYMBOL. The global address should be loaded 33694 into ADDRESS_REG. The register OFFSET_REG was previously set in 33695 insn OFFSET_INSN to hold the offset from GOT base to the GOT entry 33696 of the global variable. Now it can be used as a scratch register. 33697 33698 33699 File: gccint.info, Node: Host Config, Next: Fragments, Prev: Target Macros, Up: Top 33700 33701 18 Host Configuration 33702 ********************* 33703 33704 Most details about the machine and system on which the compiler is 33705 actually running are detected by the `configure' script. Some things 33706 are impossible for `configure' to detect; these are described in two 33707 ways, either by macros defined in a file named `xm-MACHINE.h' or by 33708 hook functions in the file specified by the OUT_HOST_HOOK_OBJ variable 33709 in `config.gcc'. (The intention is that very few hosts will need a 33710 header file but nearly every fully supported host will need to override 33711 some hooks.) 33712 33713 If you need to define only a few macros, and they have simple 33714 definitions, consider using the `xm_defines' variable in your 33715 `config.gcc' entry instead of creating a host configuration header. 33716 *Note System Config::. 33717 33718 * Menu: 33719 33720 * Host Common:: Things every host probably needs implemented. 33721 * Filesystem:: Your host can't have the letter `a' in filenames? 33722 * Host Misc:: Rare configuration options for hosts. 33723 33724 33725 File: gccint.info, Node: Host Common, Next: Filesystem, Up: Host Config 33726 33727 18.1 Host Common 33728 ================ 33729 33730 Some things are just not portable, even between similar operating 33731 systems, and are too difficult for autoconf to detect. They get 33732 implemented using hook functions in the file specified by the 33733 HOST_HOOK_OBJ variable in `config.gcc'. 33734 33735 -- Host Hook: void HOST_HOOKS_EXTRA_SIGNALS (void) 33736 This host hook is used to set up handling for extra signals. The 33737 most common thing to do in this hook is to detect stack overflow. 33738 33739 -- Host Hook: void * HOST_HOOKS_GT_PCH_GET_ADDRESS (size_t SIZE, int 33740 FD) 33741 This host hook returns the address of some space that is likely to 33742 be free in some subsequent invocation of the compiler. We intend 33743 to load the PCH data at this address such that the data need not 33744 be relocated. The area should be able to hold SIZE bytes. If the 33745 host uses `mmap', FD is an open file descriptor that can be used 33746 for probing. 33747 33748 -- Host Hook: int HOST_HOOKS_GT_PCH_USE_ADDRESS (void * ADDRESS, 33749 size_t SIZE, int FD, size_t OFFSET) 33750 This host hook is called when a PCH file is about to be loaded. 33751 We want to load SIZE bytes from FD at OFFSET into memory at 33752 ADDRESS. The given address will be the result of a previous 33753 invocation of `HOST_HOOKS_GT_PCH_GET_ADDRESS'. Return -1 if we 33754 couldn't allocate SIZE bytes at ADDRESS. Return 0 if the memory 33755 is allocated but the data is not loaded. Return 1 if the hook has 33756 performed everything. 33757 33758 If the implementation uses reserved address space, free any 33759 reserved space beyond SIZE, regardless of the return value. If no 33760 PCH will be loaded, this hook may be called with SIZE zero, in 33761 which case all reserved address space should be freed. 33762 33763 Do not try to handle values of ADDRESS that could not have been 33764 returned by this executable; just return -1. Such values usually 33765 indicate an out-of-date PCH file (built by some other GCC 33766 executable), and such a PCH file won't work. 33767 33768 -- Host Hook: size_t HOST_HOOKS_GT_PCH_ALLOC_GRANULARITY (void); 33769 This host hook returns the alignment required for allocating 33770 virtual memory. Usually this is the same as getpagesize, but on 33771 some hosts the alignment for reserving memory differs from the 33772 pagesize for committing memory. 33773 33774 33775 File: gccint.info, Node: Filesystem, Next: Host Misc, Prev: Host Common, Up: Host Config 33776 33777 18.2 Host Filesystem 33778 ==================== 33779 33780 GCC needs to know a number of things about the semantics of the host 33781 machine's filesystem. Filesystems with Unix and MS-DOS semantics are 33782 automatically detected. For other systems, you can define the 33783 following macros in `xm-MACHINE.h'. 33784 33785 `HAVE_DOS_BASED_FILE_SYSTEM' 33786 This macro is automatically defined by `system.h' if the host file 33787 system obeys the semantics defined by MS-DOS instead of Unix. DOS 33788 file systems are case insensitive, file specifications may begin 33789 with a drive letter, and both forward slash and backslash (`/' and 33790 `\') are directory separators. 33791 33792 `DIR_SEPARATOR' 33793 `DIR_SEPARATOR_2' 33794 If defined, these macros expand to character constants specifying 33795 separators for directory names within a file specification. 33796 `system.h' will automatically give them appropriate values on Unix 33797 and MS-DOS file systems. If your file system is neither of these, 33798 define one or both appropriately in `xm-MACHINE.h'. 33799 33800 However, operating systems like VMS, where constructing a pathname 33801 is more complicated than just stringing together directory names 33802 separated by a special character, should not define either of these 33803 macros. 33804 33805 `PATH_SEPARATOR' 33806 If defined, this macro should expand to a character constant 33807 specifying the separator for elements of search paths. The default 33808 value is a colon (`:'). DOS-based systems usually, but not 33809 always, use semicolon (`;'). 33810 33811 `VMS' 33812 Define this macro if the host system is VMS. 33813 33814 `HOST_OBJECT_SUFFIX' 33815 Define this macro to be a C string representing the suffix for 33816 object files on your host machine. If you do not define this 33817 macro, GCC will use `.o' as the suffix for object files. 33818 33819 `HOST_EXECUTABLE_SUFFIX' 33820 Define this macro to be a C string representing the suffix for 33821 executable files on your host machine. If you do not define this 33822 macro, GCC will use the null string as the suffix for executable 33823 files. 33824 33825 `HOST_BIT_BUCKET' 33826 A pathname defined by the host operating system, which can be 33827 opened as a file and written to, but all the information written 33828 is discarded. This is commonly known as a "bit bucket" or "null 33829 device". If you do not define this macro, GCC will use 33830 `/dev/null' as the bit bucket. If the host does not support a bit 33831 bucket, define this macro to an invalid filename. 33832 33833 `UPDATE_PATH_HOST_CANONICALIZE (PATH)' 33834 If defined, a C statement (sans semicolon) that performs 33835 host-dependent canonicalization when a path used in a compilation 33836 driver or preprocessor is canonicalized. PATH is a malloc-ed path 33837 to be canonicalized. If the C statement does canonicalize PATH 33838 into a different buffer, the old path should be freed and the new 33839 buffer should have been allocated with malloc. 33840 33841 `DUMPFILE_FORMAT' 33842 Define this macro to be a C string representing the format to use 33843 for constructing the index part of debugging dump file names. The 33844 resultant string must fit in fifteen bytes. The full filename 33845 will be the concatenation of: the prefix of the assembler file 33846 name, the string resulting from applying this format to an index 33847 number, and a string unique to each dump file kind, e.g. `rtl'. 33848 33849 If you do not define this macro, GCC will use `.%02d.'. You should 33850 define this macro if using the default will create an invalid file 33851 name. 33852 33853 `DELETE_IF_ORDINARY' 33854 Define this macro to be a C statement (sans semicolon) that 33855 performs host-dependent removal of ordinary temp files in the 33856 compilation driver. 33857 33858 If you do not define this macro, GCC will use the default version. 33859 You should define this macro if the default version does not 33860 reliably remove the temp file as, for example, on VMS which allows 33861 multiple versions of a file. 33862 33863 `HOST_LACKS_INODE_NUMBERS' 33864 Define this macro if the host filesystem does not report 33865 meaningful inode numbers in struct stat. 33866 33867 33868 File: gccint.info, Node: Host Misc, Prev: Filesystem, Up: Host Config 33869 33870 18.3 Host Misc 33871 ============== 33872 33873 `FATAL_EXIT_CODE' 33874 A C expression for the status code to be returned when the compiler 33875 exits after serious errors. The default is the system-provided 33876 macro `EXIT_FAILURE', or `1' if the system doesn't define that 33877 macro. Define this macro only if these defaults are incorrect. 33878 33879 `SUCCESS_EXIT_CODE' 33880 A C expression for the status code to be returned when the compiler 33881 exits without serious errors. (Warnings are not serious errors.) 33882 The default is the system-provided macro `EXIT_SUCCESS', or `0' if 33883 the system doesn't define that macro. Define this macro only if 33884 these defaults are incorrect. 33885 33886 `USE_C_ALLOCA' 33887 Define this macro if GCC should use the C implementation of 33888 `alloca' provided by `libiberty.a'. This only affects how some 33889 parts of the compiler itself allocate memory. It does not change 33890 code generation. 33891 33892 When GCC is built with a compiler other than itself, the C `alloca' 33893 is always used. This is because most other implementations have 33894 serious bugs. You should define this macro only on a system where 33895 no stack-based `alloca' can possibly work. For instance, if a 33896 system has a small limit on the size of the stack, GCC's builtin 33897 `alloca' will not work reliably. 33898 33899 `COLLECT2_HOST_INITIALIZATION' 33900 If defined, a C statement (sans semicolon) that performs 33901 host-dependent initialization when `collect2' is being initialized. 33902 33903 `GCC_DRIVER_HOST_INITIALIZATION' 33904 If defined, a C statement (sans semicolon) that performs 33905 host-dependent initialization when a compilation driver is being 33906 initialized. 33907 33908 `HOST_LONG_LONG_FORMAT' 33909 If defined, the string used to indicate an argument of type `long 33910 long' to functions like `printf'. The default value is `"ll"'. 33911 33912 In addition, if `configure' generates an incorrect definition of any 33913 of the macros in `auto-host.h', you can override that definition in a 33914 host configuration header. If you need to do this, first see if it is 33915 possible to fix `configure'. 33916 33917 33918 File: gccint.info, Node: Fragments, Next: Collect2, Prev: Host Config, Up: Top 33919 33920 19 Makefile Fragments 33921 ********************* 33922 33923 When you configure GCC using the `configure' script, it will construct 33924 the file `Makefile' from the template file `Makefile.in'. When it does 33925 this, it can incorporate makefile fragments from the `config' 33926 directory. These are used to set Makefile parameters that are not 33927 amenable to being calculated by autoconf. The list of fragments to 33928 incorporate is set by `config.gcc' (and occasionally `config.build' and 33929 `config.host'); *Note System Config::. 33930 33931 Fragments are named either `t-TARGET' or `x-HOST', depending on 33932 whether they are relevant to configuring GCC to produce code for a 33933 particular target, or to configuring GCC to run on a particular host. 33934 Here TARGET and HOST are mnemonics which usually have some relationship 33935 to the canonical system name, but no formal connection. 33936 33937 If these files do not exist, it means nothing needs to be added for a 33938 given target or host. Most targets need a few `t-TARGET' fragments, 33939 but needing `x-HOST' fragments is rare. 33940 33941 * Menu: 33942 33943 * Target Fragment:: Writing `t-TARGET' files. 33944 * Host Fragment:: Writing `x-HOST' files. 33945 33946 33947 File: gccint.info, Node: Target Fragment, Next: Host Fragment, Up: Fragments 33948 33949 19.1 Target Makefile Fragments 33950 ============================== 33951 33952 Target makefile fragments can set these Makefile variables. 33953 33954 `LIBGCC2_CFLAGS' 33955 Compiler flags to use when compiling `libgcc2.c'. 33956 33957 `LIB2FUNCS_EXTRA' 33958 A list of source file names to be compiled or assembled and 33959 inserted into `libgcc.a'. 33960 33961 `Floating Point Emulation' 33962 To have GCC include software floating point libraries in `libgcc.a' 33963 define `FPBIT' and `DPBIT' along with a few rules as follows: 33964 # We want fine grained libraries, so use the new code 33965 # to build the floating point emulation libraries. 33966 FPBIT = fp-bit.c 33967 DPBIT = dp-bit.c 33968 33969 33970 fp-bit.c: $(srcdir)/config/fp-bit.c 33971 echo '#define FLOAT' > fp-bit.c 33972 cat $(srcdir)/config/fp-bit.c >> fp-bit.c 33973 33974 dp-bit.c: $(srcdir)/config/fp-bit.c 33975 cat $(srcdir)/config/fp-bit.c > dp-bit.c 33976 33977 You may need to provide additional #defines at the beginning of 33978 `fp-bit.c' and `dp-bit.c' to control target endianness and other 33979 options. 33980 33981 `CRTSTUFF_T_CFLAGS' 33982 Special flags used when compiling `crtstuff.c'. *Note 33983 Initialization::. 33984 33985 `CRTSTUFF_T_CFLAGS_S' 33986 Special flags used when compiling `crtstuff.c' for shared linking. 33987 Used if you use `crtbeginS.o' and `crtendS.o' in `EXTRA-PARTS'. 33988 *Note Initialization::. 33989 33990 `MULTILIB_OPTIONS' 33991 For some targets, invoking GCC in different ways produces objects 33992 that can not be linked together. For example, for some targets GCC 33993 produces both big and little endian code. For these targets, you 33994 must arrange for multiple versions of `libgcc.a' to be compiled, 33995 one for each set of incompatible options. When GCC invokes the 33996 linker, it arranges to link in the right version of `libgcc.a', 33997 based on the command line options used. 33998 33999 The `MULTILIB_OPTIONS' macro lists the set of options for which 34000 special versions of `libgcc.a' must be built. Write options that 34001 are mutually incompatible side by side, separated by a slash. 34002 Write options that may be used together separated by a space. The 34003 build procedure will build all combinations of compatible options. 34004 34005 For example, if you set `MULTILIB_OPTIONS' to `m68000/m68020 34006 msoft-float', `Makefile' will build special versions of `libgcc.a' 34007 using the following sets of options: `-m68000', `-m68020', 34008 `-msoft-float', `-m68000 -msoft-float', and `-m68020 -msoft-float'. 34009 34010 `MULTILIB_DIRNAMES' 34011 If `MULTILIB_OPTIONS' is used, this variable specifies the 34012 directory names that should be used to hold the various libraries. 34013 Write one element in `MULTILIB_DIRNAMES' for each element in 34014 `MULTILIB_OPTIONS'. If `MULTILIB_DIRNAMES' is not used, the 34015 default value will be `MULTILIB_OPTIONS', with all slashes treated 34016 as spaces. 34017 34018 For example, if `MULTILIB_OPTIONS' is set to `m68000/m68020 34019 msoft-float', then the default value of `MULTILIB_DIRNAMES' is 34020 `m68000 m68020 msoft-float'. You may specify a different value if 34021 you desire a different set of directory names. 34022 34023 `MULTILIB_MATCHES' 34024 Sometimes the same option may be written in two different ways. 34025 If an option is listed in `MULTILIB_OPTIONS', GCC needs to know 34026 about any synonyms. In that case, set `MULTILIB_MATCHES' to a 34027 list of items of the form `option=option' to describe all relevant 34028 synonyms. For example, `m68000=mc68000 m68020=mc68020'. 34029 34030 `MULTILIB_EXCEPTIONS' 34031 Sometimes when there are multiple sets of `MULTILIB_OPTIONS' being 34032 specified, there are combinations that should not be built. In 34033 that case, set `MULTILIB_EXCEPTIONS' to be all of the switch 34034 exceptions in shell case syntax that should not be built. 34035 34036 For example the ARM processor cannot execute both hardware floating 34037 point instructions and the reduced size THUMB instructions at the 34038 same time, so there is no need to build libraries with both of 34039 these options enabled. Therefore `MULTILIB_EXCEPTIONS' is set to: 34040 *mthumb/*mhard-float* 34041 34042 `MULTILIB_EXTRA_OPTS' 34043 Sometimes it is desirable that when building multiple versions of 34044 `libgcc.a' certain options should always be passed on to the 34045 compiler. In that case, set `MULTILIB_EXTRA_OPTS' to be the list 34046 of options to be used for all builds. If you set this, you should 34047 probably set `CRTSTUFF_T_CFLAGS' to a dash followed by it. 34048 34049 `NATIVE_SYSTEM_HEADER_DIR' 34050 If the default location for system headers is not `/usr/include', 34051 you must set this to the directory containing the headers. This 34052 value should match the value of the `SYSTEM_INCLUDE_DIR' macro. 34053 34054 `SPECS' 34055 Unfortunately, setting `MULTILIB_EXTRA_OPTS' is not enough, since 34056 it does not affect the build of target libraries, at least not the 34057 build of the default multilib. One possible work-around is to use 34058 `DRIVER_SELF_SPECS' to bring options from the `specs' file as if 34059 they had been passed in the compiler driver command line. 34060 However, you don't want to be adding these options after the 34061 toolchain is installed, so you can instead tweak the `specs' file 34062 that will be used during the toolchain build, while you still 34063 install the original, built-in `specs'. The trick is to set 34064 `SPECS' to some other filename (say `specs.install'), that will 34065 then be created out of the built-in specs, and introduce a 34066 `Makefile' rule to generate the `specs' file that's going to be 34067 used at build time out of your `specs.install'. 34068 34069 `T_CFLAGS' 34070 These are extra flags to pass to the C compiler. They are used 34071 both when building GCC, and when compiling things with the 34072 just-built GCC. This variable is deprecated and should not be 34073 used. 34074 34075 34076 File: gccint.info, Node: Host Fragment, Prev: Target Fragment, Up: Fragments 34077 34078 19.2 Host Makefile Fragments 34079 ============================ 34080 34081 The use of `x-HOST' fragments is discouraged. You should only use it 34082 for makefile dependencies. 34083 34084 34085 File: gccint.info, Node: Collect2, Next: Header Dirs, Prev: Fragments, Up: Top 34086 34087 20 `collect2' 34088 ************* 34089 34090 GCC uses a utility called `collect2' on nearly all systems to arrange 34091 to call various initialization functions at start time. 34092 34093 The program `collect2' works by linking the program once and looking 34094 through the linker output file for symbols with particular names 34095 indicating they are constructor functions. If it finds any, it creates 34096 a new temporary `.c' file containing a table of them, compiles it, and 34097 links the program a second time including that file. 34098 34099 The actual calls to the constructors are carried out by a subroutine 34100 called `__main', which is called (automatically) at the beginning of 34101 the body of `main' (provided `main' was compiled with GNU CC). Calling 34102 `__main' is necessary, even when compiling C code, to allow linking C 34103 and C++ object code together. (If you use `-nostdlib', you get an 34104 unresolved reference to `__main', since it's defined in the standard 34105 GCC library. Include `-lgcc' at the end of your compiler command line 34106 to resolve this reference.) 34107 34108 The program `collect2' is installed as `ld' in the directory where the 34109 passes of the compiler are installed. When `collect2' needs to find 34110 the _real_ `ld', it tries the following file names: 34111 34112 * `real-ld' in the directories listed in the compiler's search 34113 directories. 34114 34115 * `real-ld' in the directories listed in the environment variable 34116 `PATH'. 34117 34118 * The file specified in the `REAL_LD_FILE_NAME' configuration macro, 34119 if specified. 34120 34121 * `ld' in the compiler's search directories, except that `collect2' 34122 will not execute itself recursively. 34123 34124 * `ld' in `PATH'. 34125 34126 "The compiler's search directories" means all the directories where 34127 `gcc' searches for passes of the compiler. This includes directories 34128 that you specify with `-B'. 34129 34130 Cross-compilers search a little differently: 34131 34132 * `real-ld' in the compiler's search directories. 34133 34134 * `TARGET-real-ld' in `PATH'. 34135 34136 * The file specified in the `REAL_LD_FILE_NAME' configuration macro, 34137 if specified. 34138 34139 * `ld' in the compiler's search directories. 34140 34141 * `TARGET-ld' in `PATH'. 34142 34143 `collect2' explicitly avoids running `ld' using the file name under 34144 which `collect2' itself was invoked. In fact, it remembers up a list 34145 of such names--in case one copy of `collect2' finds another copy (or 34146 version) of `collect2' installed as `ld' in a second place in the 34147 search path. 34148 34149 `collect2' searches for the utilities `nm' and `strip' using the same 34150 algorithm as above for `ld'. 34151 34152 34153 File: gccint.info, Node: Header Dirs, Next: Type Information, Prev: Collect2, Up: Top 34154 34155 21 Standard Header File Directories 34156 *********************************** 34157 34158 `GCC_INCLUDE_DIR' means the same thing for native and cross. It is 34159 where GCC stores its private include files, and also where GCC stores 34160 the fixed include files. A cross compiled GCC runs `fixincludes' on 34161 the header files in `$(tooldir)/include'. (If the cross compilation 34162 header files need to be fixed, they must be installed before GCC is 34163 built. If the cross compilation header files are already suitable for 34164 GCC, nothing special need be done). 34165 34166 `GPLUSPLUS_INCLUDE_DIR' means the same thing for native and cross. It 34167 is where `g++' looks first for header files. The C++ library installs 34168 only target independent header files in that directory. 34169 34170 `LOCAL_INCLUDE_DIR' is used only by native compilers. GCC doesn't 34171 install anything there. It is normally `/usr/local/include'. This is 34172 where local additions to a packaged system should place header files. 34173 34174 `CROSS_INCLUDE_DIR' is used only by cross compilers. GCC doesn't 34175 install anything there. 34176 34177 `TOOL_INCLUDE_DIR' is used for both native and cross compilers. It is 34178 the place for other packages to install header files that GCC will use. 34179 For a cross-compiler, this is the equivalent of `/usr/include'. When 34180 you build a cross-compiler, `fixincludes' processes any header files in 34181 this directory. 34182 34183 34184 File: gccint.info, Node: Type Information, Next: Plugins, Prev: Header Dirs, Up: Top 34185 34186 22 Memory Management and Type Information 34187 ***************************************** 34188 34189 GCC uses some fairly sophisticated memory management techniques, which 34190 involve determining information about GCC's data structures from GCC's 34191 source code and using this information to perform garbage collection and 34192 implement precompiled headers. 34193 34194 A full C parser would be too complicated for this task, so a limited 34195 subset of C is interpreted and special markers are used to determine 34196 what parts of the source to look at. All `struct' and `union' 34197 declarations that define data structures that are allocated under 34198 control of the garbage collector must be marked. All global variables 34199 that hold pointers to garbage-collected memory must also be marked. 34200 Finally, all global variables that need to be saved and restored by a 34201 precompiled header must be marked. (The precompiled header mechanism 34202 can only save static variables if they're scalar. Complex data 34203 structures must be allocated in garbage-collected memory to be saved in 34204 a precompiled header.) 34205 34206 The full format of a marker is 34207 GTY (([OPTION] [(PARAM)], [OPTION] [(PARAM)] ...)) 34208 but in most cases no options are needed. The outer double parentheses 34209 are still necessary, though: `GTY(())'. Markers can appear: 34210 34211 * In a structure definition, before the open brace; 34212 34213 * In a global variable declaration, after the keyword `static' or 34214 `extern'; and 34215 34216 * In a structure field definition, before the name of the field. 34217 34218 Here are some examples of marking simple data structures and globals. 34219 34220 struct TAG GTY(()) 34221 { 34222 FIELDS... 34223 }; 34224 34225 typedef struct TAG GTY(()) 34226 { 34227 FIELDS... 34228 } *TYPENAME; 34229 34230 static GTY(()) struct TAG *LIST; /* points to GC memory */ 34231 static GTY(()) int COUNTER; /* save counter in a PCH */ 34232 34233 The parser understands simple typedefs such as `typedef struct TAG 34234 *NAME;' and `typedef int NAME;'. These don't need to be marked. 34235 34236 * Menu: 34237 34238 * GTY Options:: What goes inside a `GTY(())'. 34239 * GGC Roots:: Making global variables GGC roots. 34240 * Files:: How the generated files work. 34241 * Invoking the garbage collector:: How to invoke the garbage collector. 34242 34243 34244 File: gccint.info, Node: GTY Options, Next: GGC Roots, Up: Type Information 34245 34246 22.1 The Inside of a `GTY(())' 34247 ============================== 34248 34249 Sometimes the C code is not enough to fully describe the type 34250 structure. Extra information can be provided with `GTY' options and 34251 additional markers. Some options take a parameter, which may be either 34252 a string or a type name, depending on the parameter. If an option 34253 takes no parameter, it is acceptable either to omit the parameter 34254 entirely, or to provide an empty string as a parameter. For example, 34255 `GTY ((skip))' and `GTY ((skip ("")))' are equivalent. 34256 34257 When the parameter is a string, often it is a fragment of C code. Four 34258 special escapes may be used in these strings, to refer to pieces of the 34259 data structure being marked: 34260 34261 `%h' 34262 The current structure. 34263 34264 `%1' 34265 The structure that immediately contains the current structure. 34266 34267 `%0' 34268 The outermost structure that contains the current structure. 34269 34270 `%a' 34271 A partial expression of the form `[i1][i2]...' that indexes the 34272 array item currently being marked. 34273 34274 For instance, suppose that you have a structure of the form 34275 struct A { 34276 ... 34277 }; 34278 struct B { 34279 struct A foo[12]; 34280 }; 34281 and `b' is a variable of type `struct B'. When marking `b.foo[11]', 34282 `%h' would expand to `b.foo[11]', `%0' and `%1' would both expand to 34283 `b', and `%a' would expand to `[11]'. 34284 34285 As in ordinary C, adjacent strings will be concatenated; this is 34286 helpful when you have a complicated expression. 34287 GTY ((chain_next ("TREE_CODE (&%h.generic) == INTEGER_TYPE" 34288 " ? TYPE_NEXT_VARIANT (&%h.generic)" 34289 " : TREE_CHAIN (&%h.generic)"))) 34290 34291 The available options are: 34292 34293 `length ("EXPRESSION")' 34294 There are two places the type machinery will need to be explicitly 34295 told the length of an array. The first case is when a structure 34296 ends in a variable-length array, like this: 34297 struct rtvec_def GTY(()) { 34298 int num_elem; /* number of elements */ 34299 rtx GTY ((length ("%h.num_elem"))) elem[1]; 34300 }; 34301 34302 In this case, the `length' option is used to override the specified 34303 array length (which should usually be `1'). The parameter of the 34304 option is a fragment of C code that calculates the length. 34305 34306 The second case is when a structure or a global variable contains a 34307 pointer to an array, like this: 34308 tree * 34309 GTY ((length ("%h.regno_pointer_align_length"))) regno_decl; 34310 In this case, `regno_decl' has been allocated by writing something 34311 like 34312 x->regno_decl = 34313 ggc_alloc (x->regno_pointer_align_length * sizeof (tree)); 34314 and the `length' provides the length of the field. 34315 34316 This second use of `length' also works on global variables, like: static GTY((length ("reg_base_value_size"))) 34317 rtx *reg_base_value; 34318 34319 `skip' 34320 If `skip' is applied to a field, the type machinery will ignore it. 34321 This is somewhat dangerous; the only safe use is in a union when 34322 one field really isn't ever used. 34323 34324 `desc ("EXPRESSION")' 34325 `tag ("CONSTANT")' 34326 `default' 34327 The type machinery needs to be told which field of a `union' is 34328 currently active. This is done by giving each field a constant 34329 `tag' value, and then specifying a discriminator using `desc'. 34330 The value of the expression given by `desc' is compared against 34331 each `tag' value, each of which should be different. If no `tag' 34332 is matched, the field marked with `default' is used if there is 34333 one, otherwise no field in the union will be marked. 34334 34335 In the `desc' option, the "current structure" is the union that it 34336 discriminates. Use `%1' to mean the structure containing it. 34337 There are no escapes available to the `tag' option, since it is a 34338 constant. 34339 34340 For example, 34341 struct tree_binding GTY(()) 34342 { 34343 struct tree_common common; 34344 union tree_binding_u { 34345 tree GTY ((tag ("0"))) scope; 34346 struct cp_binding_level * GTY ((tag ("1"))) level; 34347 } GTY ((desc ("BINDING_HAS_LEVEL_P ((tree)&%0)"))) xscope; 34348 tree value; 34349 }; 34350 34351 In this example, the value of BINDING_HAS_LEVEL_P when applied to a 34352 `struct tree_binding *' is presumed to be 0 or 1. If 1, the type 34353 mechanism will treat the field `level' as being present and if 0, 34354 will treat the field `scope' as being present. 34355 34356 `param_is (TYPE)' 34357 `use_param' 34358 Sometimes it's convenient to define some data structure to work on 34359 generic pointers (that is, `PTR') and then use it with a specific 34360 type. `param_is' specifies the real type pointed to, and 34361 `use_param' says where in the generic data structure that type 34362 should be put. 34363 34364 For instance, to have a `htab_t' that points to trees, one would 34365 write the definition of `htab_t' like this: 34366 typedef struct GTY(()) { 34367 ... 34368 void ** GTY ((use_param, ...)) entries; 34369 ... 34370 } htab_t; 34371 and then declare variables like this: 34372 static htab_t GTY ((param_is (union tree_node))) ict; 34373 34374 `paramN_is (TYPE)' 34375 `use_paramN' 34376 In more complicated cases, the data structure might need to work on 34377 several different types, which might not necessarily all be 34378 pointers. For this, `param1_is' through `param9_is' may be used to 34379 specify the real type of a field identified by `use_param1' through 34380 `use_param9'. 34381 34382 `use_params' 34383 When a structure contains another structure that is parameterized, 34384 there's no need to do anything special, the inner structure 34385 inherits the parameters of the outer one. When a structure 34386 contains a pointer to a parameterized structure, the type 34387 machinery won't automatically detect this (it could, it just 34388 doesn't yet), so it's necessary to tell it that the pointed-to 34389 structure should use the same parameters as the outer structure. 34390 This is done by marking the pointer with the `use_params' option. 34391 34392 `deletable' 34393 `deletable', when applied to a global variable, indicates that when 34394 garbage collection runs, there's no need to mark anything pointed 34395 to by this variable, it can just be set to `NULL' instead. This 34396 is used to keep a list of free structures around for re-use. 34397 34398 `if_marked ("EXPRESSION")' 34399 Suppose you want some kinds of object to be unique, and so you put 34400 them in a hash table. If garbage collection marks the hash table, 34401 these objects will never be freed, even if the last other 34402 reference to them goes away. GGC has special handling to deal 34403 with this: if you use the `if_marked' option on a global hash 34404 table, GGC will call the routine whose name is the parameter to 34405 the option on each hash table entry. If the routine returns 34406 nonzero, the hash table entry will be marked as usual. If the 34407 routine returns zero, the hash table entry will be deleted. 34408 34409 The routine `ggc_marked_p' can be used to determine if an element 34410 has been marked already; in fact, the usual case is to use 34411 `if_marked ("ggc_marked_p")'. 34412 34413 `mark_hook ("HOOK-ROUTINE-NAME")' 34414 If provided for a structure or union type, the given 34415 HOOK-ROUTINE-NAME (between double-quotes) is the name of a routine 34416 called when the garbage collector has just marked the data as 34417 reachable. This routine should not change the data, or call any ggc 34418 routine. Its only argument is a pointer to the just marked (const) 34419 structure or union. 34420 34421 `maybe_undef' 34422 When applied to a field, `maybe_undef' indicates that it's OK if 34423 the structure that this fields points to is never defined, so long 34424 as this field is always `NULL'. This is used to avoid requiring 34425 backends to define certain optional structures. It doesn't work 34426 with language frontends. 34427 34428 `nested_ptr (TYPE, "TO EXPRESSION", "FROM EXPRESSION")' 34429 The type machinery expects all pointers to point to the start of an 34430 object. Sometimes for abstraction purposes it's convenient to have 34431 a pointer which points inside an object. So long as it's possible 34432 to convert the original object to and from the pointer, such 34433 pointers can still be used. TYPE is the type of the original 34434 object, the TO EXPRESSION returns the pointer given the original 34435 object, and the FROM EXPRESSION returns the original object given 34436 the pointer. The pointer will be available using the `%h' escape. 34437 34438 `chain_next ("EXPRESSION")' 34439 `chain_prev ("EXPRESSION")' 34440 `chain_circular ("EXPRESSION")' 34441 It's helpful for the type machinery to know if objects are often 34442 chained together in long lists; this lets it generate code that 34443 uses less stack space by iterating along the list instead of 34444 recursing down it. `chain_next' is an expression for the next 34445 item in the list, `chain_prev' is an expression for the previous 34446 item. For singly linked lists, use only `chain_next'; for doubly 34447 linked lists, use both. The machinery requires that taking the 34448 next item of the previous item gives the original item. 34449 `chain_circular' is similar to `chain_next', but can be used for 34450 circular single linked lists. 34451 34452 `reorder ("FUNCTION NAME")' 34453 Some data structures depend on the relative ordering of pointers. 34454 If the precompiled header machinery needs to change that ordering, 34455 it will call the function referenced by the `reorder' option, 34456 before changing the pointers in the object that's pointed to by 34457 the field the option applies to. The function must take four 34458 arguments, with the signature 34459 `void *, void *, gt_pointer_operator, void *'. The first 34460 parameter is a pointer to the structure that contains the object 34461 being updated, or the object itself if there is no containing 34462 structure. The second parameter is a cookie that should be 34463 ignored. The third parameter is a routine that, given a pointer, 34464 will update it to its correct new value. The fourth parameter is 34465 a cookie that must be passed to the second parameter. 34466 34467 PCH cannot handle data structures that depend on the absolute 34468 values of pointers. `reorder' functions can be expensive. When 34469 possible, it is better to depend on properties of the data, like 34470 an ID number or the hash of a string instead. 34471 34472 `special ("NAME")' 34473 The `special' option is used to mark types that have to be dealt 34474 with by special case machinery. The parameter is the name of the 34475 special case. See `gengtype.c' for further details. Avoid adding 34476 new special cases unless there is no other alternative. 34477 34478 34479 File: gccint.info, Node: GGC Roots, Next: Files, Prev: GTY Options, Up: Type Information 34480 34481 22.2 Marking Roots for the Garbage Collector 34482 ============================================ 34483 34484 In addition to keeping track of types, the type machinery also locates 34485 the global variables ("roots") that the garbage collector starts at. 34486 Roots must be declared using one of the following syntaxes: 34487 34488 * `extern GTY(([OPTIONS])) TYPE NAME;' 34489 34490 * `static GTY(([OPTIONS])) TYPE NAME;' 34491 The syntax 34492 * `GTY(([OPTIONS])) TYPE NAME;' 34493 is _not_ accepted. There should be an `extern' declaration of such a 34494 variable in a header somewhere--mark that, not the definition. Or, if 34495 the variable is only used in one file, make it `static'. 34496 34497 34498 File: gccint.info, Node: Files, Next: Invoking the garbage collector, Prev: GGC Roots, Up: Type Information 34499 34500 22.3 Source Files Containing Type Information 34501 ============================================= 34502 34503 Whenever you add `GTY' markers to a source file that previously had 34504 none, or create a new source file containing `GTY' markers, there are 34505 three things you need to do: 34506 34507 1. You need to add the file to the list of source files the type 34508 machinery scans. There are four cases: 34509 34510 a. For a back-end file, this is usually done automatically; if 34511 not, you should add it to `target_gtfiles' in the appropriate 34512 port's entries in `config.gcc'. 34513 34514 b. For files shared by all front ends, add the filename to the 34515 `GTFILES' variable in `Makefile.in'. 34516 34517 c. For files that are part of one front end, add the filename to 34518 the `gtfiles' variable defined in the appropriate 34519 `config-lang.in'. For C, the file is `c-config-lang.in'. 34520 Headers should appear before non-headers in this list. 34521 34522 d. For files that are part of some but not all front ends, add 34523 the filename to the `gtfiles' variable of _all_ the front ends 34524 that use it. 34525 34526 2. If the file was a header file, you'll need to check that it's 34527 included in the right place to be visible to the generated files. 34528 For a back-end header file, this should be done automatically. 34529 For a front-end header file, it needs to be included by the same 34530 file that includes `gtype-LANG.h'. For other header files, it 34531 needs to be included in `gtype-desc.c', which is a generated file, 34532 so add it to `ifiles' in `open_base_file' in `gengtype.c'. 34533 34534 For source files that aren't header files, the machinery will 34535 generate a header file that should be included in the source file 34536 you just changed. The file will be called `gt-PATH.h' where PATH 34537 is the pathname relative to the `gcc' directory with slashes 34538 replaced by -, so for example the header file to be included in 34539 `cp/parser.c' is called `gt-cp-parser.c'. The generated header 34540 file should be included after everything else in the source file. 34541 Don't forget to mention this file as a dependency in the 34542 `Makefile'! 34543 34544 34545 For language frontends, there is another file that needs to be included 34546 somewhere. It will be called `gtype-LANG.h', where LANG is the name of 34547 the subdirectory the language is contained in. 34548 34549 Plugins can add additional root tables. Run the `gengtype' utility in 34550 plugin mode as `gengtype -p SOURCE-DIR FILE-LIST PLUGIN*.C' with your 34551 plugin files PLUGIN*.C using `GTY' to generate the corresponding 34552 GT-PLUGIN*.H files. The GCC build tree is needed to be present in that 34553 mode. 34554 34555 34556 File: gccint.info, Node: Invoking the garbage collector, Prev: Files, Up: Type Information 34557 34558 22.4 How to invoke the garbage collector 34559 ======================================== 34560 34561 The GCC garbage collector GGC is only invoked explicitly. In contrast 34562 with many other garbage collectors, it is not implicitly invoked by 34563 allocation routines when a lot of memory has been consumed. So the only 34564 way to have GGC reclaim storage it to call the `ggc_collect' function 34565 explicitly. This call is an expensive operation, as it may have to scan 34566 the entire heap. Beware that local variables (on the GCC call stack) 34567 are not followed by such an invocation (as many other garbage 34568 collectors do): you should reference all your data from static or 34569 external `GTY'-ed variables, and it is advised to call `ggc_collect' 34570 with a shallow call stack. The GGC is an exact mark and sweep garbage 34571 collector (so it does not scan the call stack for pointers). In 34572 practice GCC passes don't often call `ggc_collect' themselves, because 34573 it is called by the pass manager between passes. 34574 34575 34576 File: gccint.info, Node: Plugins, Next: Funding, Prev: Type Information, Up: Top 34577 34578 23 Plugins 34579 ********** 34580 34581 23.1 Loading Plugins 34582 ==================== 34583 34584 Plugins are supported on platforms that support `-ldl -rdynamic'. They 34585 are loaded by the compiler using `dlopen' and invoked at pre-determined 34586 locations in the compilation process. 34587 34588 Plugins are loaded with 34589 34590 `-fplugin=/path/to/NAME.so' `-fplugin-arg-NAME-<key1>[=<value1>]' 34591 34592 The plugin arguments are parsed by GCC and passed to respective 34593 plugins as key-value pairs. Multiple plugins can be invoked by 34594 specifying multiple `-fplugin' arguments. 34595 34596 23.2 Plugin API 34597 =============== 34598 34599 Plugins are activated by the compiler at specific events as defined in 34600 `gcc-plugin.h'. For each event of interest, the plugin should call 34601 `register_callback' specifying the name of the event and address of the 34602 callback function that will handle that event. 34603 34604 The header `gcc-plugin.h' must be the first gcc header to be included. 34605 34606 23.2.1 Plugin initialization 34607 ---------------------------- 34608 34609 Every plugin should export a function called `plugin_init' that is 34610 called right after the plugin is loaded. This function is responsible 34611 for registering all the callbacks required by the plugin and do any 34612 other required initialization. 34613 34614 This function is called from `compile_file' right before invoking the 34615 parser. The arguments to `plugin_init' are: 34616 34617 * `plugin_info': Plugin invocation information. 34618 34619 * `version': GCC version. 34620 34621 The `plugin_info' struct is defined as follows: 34622 34623 struct plugin_name_args 34624 { 34625 char *base_name; /* Short name of the plugin 34626 (filename without .so suffix). */ 34627 const char *full_name; /* Path to the plugin as specified with 34628 -fplugin=. */ 34629 int argc; /* Number of arguments specified with 34630 -fplugin-arg-.... */ 34631 struct plugin_argument *argv; /* Array of ARGC key-value pairs. */ 34632 const char *version; /* Version string provided by plugin. */ 34633 const char *help; /* Help string provided by plugin. */ 34634 } 34635 34636 If initialization fails, `plugin_init' must return a non-zero value. 34637 Otherwise, it should return 0. 34638 34639 The version of the GCC compiler loading the plugin is described by the 34640 following structure: 34641 34642 struct plugin_gcc_version 34643 { 34644 const char *basever; 34645 const char *datestamp; 34646 const char *devphase; 34647 const char *revision; 34648 const char *configuration_arguments; 34649 }; 34650 34651 The function `plugin_default_version_check' takes two pointers to such 34652 structure and compare them field by field. It can be used by the 34653 plugin's `plugin_init' function. 34654 34655 23.2.2 Plugin callbacks 34656 ----------------------- 34657 34658 Callback functions have the following prototype: 34659 34660 /* The prototype for a plugin callback function. 34661 gcc_data - event-specific data provided by GCC 34662 user_data - plugin-specific data provided by the plug-in. */ 34663 typedef void (*plugin_callback_func)(void *gcc_data, void *user_data); 34664 34665 Callbacks can be invoked at the following pre-determined events: 34666 34667 enum plugin_event 34668 { 34669 PLUGIN_PASS_MANAGER_SETUP, /* To hook into pass manager. */ 34670 PLUGIN_FINISH_TYPE, /* After finishing parsing a type. */ 34671 PLUGIN_FINISH_UNIT, /* Useful for summary processing. */ 34672 PLUGIN_CXX_CP_PRE_GENERICIZE, /* Allows to see low level AST in C++ FE. */ 34673 PLUGIN_FINISH, /* Called before GCC exits. */ 34674 PLUGIN_INFO, /* Information about the plugin. */ 34675 PLUGIN_GGC_START, /* Called at start of GCC Garbage Collection. */ 34676 PLUGIN_GGC_MARKING, /* Extend the GGC marking. */ 34677 PLUGIN_GGC_END, /* Called at end of GGC. */ 34678 PLUGIN_REGISTER_GGC_ROOTS, /* Register an extra GGC root table. */ 34679 PLUGIN_ATTRIBUTES, /* Called during attribute registration */ 34680 PLUGIN_START_UNIT, /* Called before processing a translation unit. */ 34681 PLUGIN_EVENT_LAST /* Dummy event used for indexing callback 34682 array. */ 34683 }; 34684 34685 To register a callback, the plugin calls `register_callback' with the 34686 arguments: 34687 34688 * `char *name': Plugin name. 34689 34690 * `enum plugin_event event': The event code. 34691 34692 * `plugin_callback_func callback': The function that handles `event'. 34693 34694 * `void *user_data': Pointer to plugin-specific data. 34695 34696 For the PLUGIN_PASS_MANAGER_SETUP, PLUGIN_INFO, and 34697 PLUGIN_REGISTER_GGC_ROOTS pseudo-events the `callback' should be null, 34698 and the `user_data' is specific. 34699 34700 23.3 Interacting with the pass manager 34701 ====================================== 34702 34703 There needs to be a way to add/reorder/remove passes dynamically. This 34704 is useful for both analysis plugins (plugging in after a certain pass 34705 such as CFG or an IPA pass) and optimization plugins. 34706 34707 Basic support for inserting new passes or replacing existing passes is 34708 provided. A plugin registers a new pass with GCC by calling 34709 `register_callback' with the `PLUGIN_PASS_MANAGER_SETUP' event and a 34710 pointer to a `struct plugin_pass' object defined as follows 34711 34712 enum pass_positioning_ops 34713 { 34714 PASS_POS_INSERT_AFTER, // Insert after the reference pass. 34715 PASS_POS_INSERT_BEFORE, // Insert before the reference pass. 34716 PASS_POS_REPLACE // Replace the reference pass. 34717 }; 34718 34719 struct plugin_pass 34720 { 34721 struct opt_pass *pass; /* New pass provided by the plugin. */ 34722 const char *reference_pass_name; /* Name of the reference pass for hooking 34723 up the new pass. */ 34724 int ref_pass_instance_number; /* Insert the pass at the specified 34725 instance number of the reference pass. */ 34726 /* Do it for every instance if it is 0. */ 34727 enum pass_positioning_ops pos_op; /* how to insert the new pass. */ 34728 }; 34729 34730 34731 /* Sample plugin code that registers a new pass. */ 34732 int 34733 plugin_init (struct plugin_name_args *plugin_info, 34734 struct plugin_gcc_version *version) 34735 { 34736 struct plugin_pass pass_info; 34737 34738 ... 34739 34740 /* Code to fill in the pass_info object with new pass information. */ 34741 34742 ... 34743 34744 /* Register the new pass. */ 34745 register_callback (plugin_info->base_name, PLUGIN_PASS_MANAGER_SETUP, NULL, &pass_info); 34746 34747 ... 34748 } 34749 34750 23.4 Interacting with the GCC Garbage Collector 34751 =============================================== 34752 34753 Some plugins may want to be informed when GGC (the GCC Garbage 34754 Collector) is running. They can register callbacks for the 34755 `PLUGIN_GGC_START' and `PLUGIN_GGC_END' events (for which the callback 34756 is called with a null `gcc_data') to be notified of the start or end of 34757 the GCC garbage collection. 34758 34759 Some plugins may need to have GGC mark additional data. This can be 34760 done by registering a callback (called with a null `gcc_data') for the 34761 `PLUGIN_GGC_MARKING' event. Such callbacks can call the `ggc_set_mark' 34762 routine, preferably thru the `ggc_mark' macro (and conversely, these 34763 routines should usually not be used in plugins outside of the 34764 `PLUGIN_GGC_MARKING' event). 34765 34766 Some plugins may need to add extra GGC root tables, e.g. to handle 34767 their own `GTY'-ed data. This can be done with the 34768 `PLUGIN_REGISTER_GGC_ROOTS' pseudo-event with a null callback and the 34769 extra root table as `user_data'. Running the `gengtype -p SOURCE-DIR 34770 FILE-LIST PLUGIN*.C ...' utility generates this extra root table. 34771 34772 You should understand the details of memory management inside GCC 34773 before using `PLUGIN_GGC_MARKING' or `PLUGIN_REGISTER_GGC_ROOTS'. 34774 34775 23.5 Giving information about a plugin 34776 ====================================== 34777 34778 A plugin should give some information to the user about itself. This 34779 uses the following structure: 34780 34781 struct plugin_info 34782 { 34783 const char *version; 34784 const char *help; 34785 }; 34786 34787 Such a structure is passed as the `user_data' by the plugin's init 34788 routine using `register_callback' with the `PLUGIN_INFO' pseudo-event 34789 and a null callback. 34790 34791 23.6 Registering custom attributes 34792 ================================== 34793 34794 For analysis purposes it is useful to be able to add custom attributes. 34795 34796 The `PLUGIN_ATTRIBUTES' callback is called during attribute 34797 registration. Use the `register_attribute' function to register custom 34798 attributes. 34799 34800 /* Attribute handler callback */ 34801 static tree 34802 handle_user_attribute (tree *node, tree name, tree args, 34803 int flags, bool *no_add_attrs) 34804 { 34805 return NULL_TREE; 34806 } 34807 34808 /* Attribute definition */ 34809 static struct attribute_spec user_attr = 34810 { "user", 1, 1, false, false, false, handle_user_attribute }; 34811 34812 /* Plugin callback called during attribute registration. 34813 Registered with register_callback (plugin_name, PLUGIN_ATTRIBUTES, register_attributes, NULL) 34814 */ 34815 static void 34816 register_attributes (void *event_data, void *data) 34817 { 34818 warning (0, G_("Callback to register attributes")); 34819 register_attribute (&user_attr); 34820 } 34821 34822 23.7 Building GCC plugins 34823 ========================= 34824 34825 If plugins are enabled, GCC installs the headers needed to build a 34826 plugin (somehwere in the installation tree, e.g. under `/usr/local'). 34827 In particular a `plugin/include' directory is installed, containing all 34828 the header files needed to build plugins. 34829 34830 On most systems, you can query this `plugin' directory by invoking 34831 `gcc -print-file-name=plugin' (replace if needed `gcc' with the 34832 appropriate program path). 34833 34834 The following GNU Makefile excerpt shows how to build a simple plugin: 34835 34836 GCC=gcc 34837 PLUGIN_SOURCE_FILES= plugin1.c plugin2.c 34838 PLUGIN_OBJECT_FILES= $(patsubst %.c,%.o,$(PLUGIN_SOURCE_FILES)) 34839 GCCPLUGINS_DIR:= $(shell $(GCC) -print-file-name=plugin) 34840 CFLAGS+= -I$(GCCPLUGINS_DIR)/include -fPIC -O2 34841 34842 plugin.so: $(PLUGIN_OBJECT_FILES) 34843 $(GCC) -shared $^ -o $ 34844 34845 A single source file plugin may be built with `gcc -I`gcc 34846 -print-file-name=plugin`/include -fPIC -shared -O2 plugin.c -o 34847 plugin.so', using backquote shell syntax to query the `plugin' 34848 directory. 34849 34850 Plugins needing to use `gengtype' require a GCC build directory for 34851 the same version of GCC that they will be linked against. 34852 34853 34854 File: gccint.info, Node: Funding, Next: GNU Project, Prev: Plugins, Up: Top 34855 34856 Funding Free Software 34857 ********************* 34858 34859 If you want to have more free software a few years from now, it makes 34860 sense for you to help encourage people to contribute funds for its 34861 development. The most effective approach known is to encourage 34862 commercial redistributors to donate. 34863 34864 Users of free software systems can boost the pace of development by 34865 encouraging for-a-fee distributors to donate part of their selling price 34866 to free software developers--the Free Software Foundation, and others. 34867 34868 The way to convince distributors to do this is to demand it and expect 34869 it from them. So when you compare distributors, judge them partly by 34870 how much they give to free software development. Show distributors 34871 they must compete to be the one who gives the most. 34872 34873 To make this approach work, you must insist on numbers that you can 34874 compare, such as, "We will donate ten dollars to the Frobnitz project 34875 for each disk sold." Don't be satisfied with a vague promise, such as 34876 "A portion of the profits are donated," since it doesn't give a basis 34877 for comparison. 34878 34879 Even a precise fraction "of the profits from this disk" is not very 34880 meaningful, since creative accounting and unrelated business decisions 34881 can greatly alter what fraction of the sales price counts as profit. 34882 If the price you pay is $50, ten percent of the profit is probably less 34883 than a dollar; it might be a few cents, or nothing at all. 34884 34885 Some redistributors do development work themselves. This is useful 34886 too; but to keep everyone honest, you need to inquire how much they do, 34887 and what kind. Some kinds of development make much more long-term 34888 difference than others. For example, maintaining a separate version of 34889 a program contributes very little; maintaining the standard version of a 34890 program for the whole community contributes much. Easy new ports 34891 contribute little, since someone else would surely do them; difficult 34892 ports such as adding a new CPU to the GNU Compiler Collection 34893 contribute more; major new features or packages contribute the most. 34894 34895 By establishing the idea that supporting further development is "the 34896 proper thing to do" when distributing free software for a fee, we can 34897 assure a steady flow of resources into making more free software. 34898 34899 Copyright (C) 1994 Free Software Foundation, Inc. 34900 Verbatim copying and redistribution of this section is permitted 34901 without royalty; alteration is not permitted. 34902 34903 34904 File: gccint.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top 34905 34906 The GNU Project and GNU/Linux 34907 ***************************** 34908 34909 The GNU Project was launched in 1984 to develop a complete Unix-like 34910 operating system which is free software: the GNU system. (GNU is a 34911 recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".) 34912 Variants of the GNU operating system, which use the kernel Linux, are 34913 now widely used; though these systems are often referred to as "Linux", 34914 they are more accurately called GNU/Linux systems. 34915 34916 For more information, see: 34917 `http://www.gnu.org/' 34918 `http://www.gnu.org/gnu/linux-and-gnu.html' 34919 34920 34921 File: gccint.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top 34922 34923 GNU General Public License 34924 ************************** 34925 34926 Version 3, 29 June 2007 34927 34928 Copyright (C) 2007 Free Software Foundation, Inc. `http://fsf.org/' 34929 34930 Everyone is permitted to copy and distribute verbatim copies of this 34931 license document, but changing it is not allowed. 34932 34933 Preamble 34934 ======== 34935 34936 The GNU General Public License is a free, copyleft license for software 34937 and other kinds of works. 34938 34939 The licenses for most software and other practical works are designed 34940 to take away your freedom to share and change the works. By contrast, 34941 the GNU General Public License is intended to guarantee your freedom to 34942 share and change all versions of a program-to make sure it remains free 34943 software for all its users. We, the Free Software Foundation, use the 34944 GNU General Public License for most of our software; it applies also to 34945 any other work released this way by its authors. You can apply it to 34946 your programs, too. 34947 34948 When we speak of free software, we are referring to freedom, not 34949 price. Our General Public Licenses are designed to make sure that you 34950 have the freedom to distribute copies of free software (and charge for 34951 them if you wish), that you receive source code or can get it if you 34952 want it, that you can change the software or use pieces of it in new 34953 free programs, and that you know you can do these things. 34954 34955 To protect your rights, we need to prevent others from denying you 34956 these rights or asking you to surrender the rights. Therefore, you 34957 have certain responsibilities if you distribute copies of the software, 34958 or if you modify it: responsibilities to respect the freedom of others. 34959 34960 For example, if you distribute copies of such a program, whether 34961 gratis or for a fee, you must pass on to the recipients the same 34962 freedoms that you received. You must make sure that they, too, receive 34963 or can get the source code. And you must show them these terms so they 34964 know their rights. 34965 34966 Developers that use the GNU GPL protect your rights with two steps: 34967 (1) assert copyright on the software, and (2) offer you this License 34968 giving you legal permission to copy, distribute and/or modify it. 34969 34970 For the developers' and authors' protection, the GPL clearly explains 34971 that there is no warranty for this free software. For both users' and 34972 authors' sake, the GPL requires that modified versions be marked as 34973 changed, so that their problems will not be attributed erroneously to 34974 authors of previous versions. 34975 34976 Some devices are designed to deny users access to install or run 34977 modified versions of the software inside them, although the 34978 manufacturer can do so. This is fundamentally incompatible with the 34979 aim of protecting users' freedom to change the software. The 34980 systematic pattern of such abuse occurs in the area of products for 34981 individuals to use, which is precisely where it is most unacceptable. 34982 Therefore, we have designed this version of the GPL to prohibit the 34983 practice for those products. If such problems arise substantially in 34984 other domains, we stand ready to extend this provision to those domains 34985 in future versions of the GPL, as needed to protect the freedom of 34986 users. 34987 34988 Finally, every program is threatened constantly by software patents. 34989 States should not allow patents to restrict development and use of 34990 software on general-purpose computers, but in those that do, we wish to 34991 avoid the special danger that patents applied to a free program could 34992 make it effectively proprietary. To prevent this, the GPL assures that 34993 patents cannot be used to render the program non-free. 34994 34995 The precise terms and conditions for copying, distribution and 34996 modification follow. 34997 34998 TERMS AND CONDITIONS 34999 ==================== 35000 35001 0. Definitions. 35002 35003 "This License" refers to version 3 of the GNU General Public 35004 License. 35005 35006 "Copyright" also means copyright-like laws that apply to other 35007 kinds of works, such as semiconductor masks. 35008 35009 "The Program" refers to any copyrightable work licensed under this 35010 License. Each licensee is addressed as "you". "Licensees" and 35011 "recipients" may be individuals or organizations. 35012 35013 To "modify" a work means to copy from or adapt all or part of the 35014 work in a fashion requiring copyright permission, other than the 35015 making of an exact copy. The resulting work is called a "modified 35016 version" of the earlier work or a work "based on" the earlier work. 35017 35018 A "covered work" means either the unmodified Program or a work 35019 based on the Program. 35020 35021 To "propagate" a work means to do anything with it that, without 35022 permission, would make you directly or secondarily liable for 35023 infringement under applicable copyright law, except executing it 35024 on a computer or modifying a private copy. Propagation includes 35025 copying, distribution (with or without modification), making 35026 available to the public, and in some countries other activities as 35027 well. 35028 35029 To "convey" a work means any kind of propagation that enables other 35030 parties to make or receive copies. Mere interaction with a user 35031 through a computer network, with no transfer of a copy, is not 35032 conveying. 35033 35034 An interactive user interface displays "Appropriate Legal Notices" 35035 to the extent that it includes a convenient and prominently visible 35036 feature that (1) displays an appropriate copyright notice, and (2) 35037 tells the user that there is no warranty for the work (except to 35038 the extent that warranties are provided), that licensees may 35039 convey the work under this License, and how to view a copy of this 35040 License. If the interface presents a list of user commands or 35041 options, such as a menu, a prominent item in the list meets this 35042 criterion. 35043 35044 1. Source Code. 35045 35046 The "source code" for a work means the preferred form of the work 35047 for making modifications to it. "Object code" means any 35048 non-source form of a work. 35049 35050 A "Standard Interface" means an interface that either is an 35051 official standard defined by a recognized standards body, or, in 35052 the case of interfaces specified for a particular programming 35053 language, one that is widely used among developers working in that 35054 language. 35055 35056 The "System Libraries" of an executable work include anything, 35057 other than the work as a whole, that (a) is included in the normal 35058 form of packaging a Major Component, but which is not part of that 35059 Major Component, and (b) serves only to enable use of the work 35060 with that Major Component, or to implement a Standard Interface 35061 for which an implementation is available to the public in source 35062 code form. A "Major Component", in this context, means a major 35063 essential component (kernel, window system, and so on) of the 35064 specific operating system (if any) on which the executable work 35065 runs, or a compiler used to produce the work, or an object code 35066 interpreter used to run it. 35067 35068 The "Corresponding Source" for a work in object code form means all 35069 the source code needed to generate, install, and (for an executable 35070 work) run the object code and to modify the work, including 35071 scripts to control those activities. However, it does not include 35072 the work's System Libraries, or general-purpose tools or generally 35073 available free programs which are used unmodified in performing 35074 those activities but which are not part of the work. For example, 35075 Corresponding Source includes interface definition files 35076 associated with source files for the work, and the source code for 35077 shared libraries and dynamically linked subprograms that the work 35078 is specifically designed to require, such as by intimate data 35079 communication or control flow between those subprograms and other 35080 parts of the work. 35081 35082 The Corresponding Source need not include anything that users can 35083 regenerate automatically from other parts of the Corresponding 35084 Source. 35085 35086 The Corresponding Source for a work in source code form is that 35087 same work. 35088 35089 2. Basic Permissions. 35090 35091 All rights granted under this License are granted for the term of 35092 copyright on the Program, and are irrevocable provided the stated 35093 conditions are met. This License explicitly affirms your unlimited 35094 permission to run the unmodified Program. The output from running 35095 a covered work is covered by this License only if the output, 35096 given its content, constitutes a covered work. This License 35097 acknowledges your rights of fair use or other equivalent, as 35098 provided by copyright law. 35099 35100 You may make, run and propagate covered works that you do not 35101 convey, without conditions so long as your license otherwise 35102 remains in force. You may convey covered works to others for the 35103 sole purpose of having them make modifications exclusively for 35104 you, or provide you with facilities for running those works, 35105 provided that you comply with the terms of this License in 35106 conveying all material for which you do not control copyright. 35107 Those thus making or running the covered works for you must do so 35108 exclusively on your behalf, under your direction and control, on 35109 terms that prohibit them from making any copies of your 35110 copyrighted material outside their relationship with you. 35111 35112 Conveying under any other circumstances is permitted solely under 35113 the conditions stated below. Sublicensing is not allowed; section 35114 10 makes it unnecessary. 35115 35116 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 35117 35118 No covered work shall be deemed part of an effective technological 35119 measure under any applicable law fulfilling obligations under 35120 article 11 of the WIPO copyright treaty adopted on 20 December 35121 1996, or similar laws prohibiting or restricting circumvention of 35122 such measures. 35123 35124 When you convey a covered work, you waive any legal power to forbid 35125 circumvention of technological measures to the extent such 35126 circumvention is effected by exercising rights under this License 35127 with respect to the covered work, and you disclaim any intention 35128 to limit operation or modification of the work as a means of 35129 enforcing, against the work's users, your or third parties' legal 35130 rights to forbid circumvention of technological measures. 35131 35132 4. Conveying Verbatim Copies. 35133 35134 You may convey verbatim copies of the Program's source code as you 35135 receive it, in any medium, provided that you conspicuously and 35136 appropriately publish on each copy an appropriate copyright notice; 35137 keep intact all notices stating that this License and any 35138 non-permissive terms added in accord with section 7 apply to the 35139 code; keep intact all notices of the absence of any warranty; and 35140 give all recipients a copy of this License along with the Program. 35141 35142 You may charge any price or no price for each copy that you convey, 35143 and you may offer support or warranty protection for a fee. 35144 35145 5. Conveying Modified Source Versions. 35146 35147 You may convey a work based on the Program, or the modifications to 35148 produce it from the Program, in the form of source code under the 35149 terms of section 4, provided that you also meet all of these 35150 conditions: 35151 35152 a. The work must carry prominent notices stating that you 35153 modified it, and giving a relevant date. 35154 35155 b. The work must carry prominent notices stating that it is 35156 released under this License and any conditions added under 35157 section 7. This requirement modifies the requirement in 35158 section 4 to "keep intact all notices". 35159 35160 c. You must license the entire work, as a whole, under this 35161 License to anyone who comes into possession of a copy. This 35162 License will therefore apply, along with any applicable 35163 section 7 additional terms, to the whole of the work, and all 35164 its parts, regardless of how they are packaged. This License 35165 gives no permission to license the work in any other way, but 35166 it does not invalidate such permission if you have separately 35167 received it. 35168 35169 d. If the work has interactive user interfaces, each must display 35170 Appropriate Legal Notices; however, if the Program has 35171 interactive interfaces that do not display Appropriate Legal 35172 Notices, your work need not make them do so. 35173 35174 A compilation of a covered work with other separate and independent 35175 works, which are not by their nature extensions of the covered 35176 work, and which are not combined with it such as to form a larger 35177 program, in or on a volume of a storage or distribution medium, is 35178 called an "aggregate" if the compilation and its resulting 35179 copyright are not used to limit the access or legal rights of the 35180 compilation's users beyond what the individual works permit. 35181 Inclusion of a covered work in an aggregate does not cause this 35182 License to apply to the other parts of the aggregate. 35183 35184 6. Conveying Non-Source Forms. 35185 35186 You may convey a covered work in object code form under the terms 35187 of sections 4 and 5, provided that you also convey the 35188 machine-readable Corresponding Source under the terms of this 35189 License, in one of these ways: 35190 35191 a. Convey the object code in, or embodied in, a physical product 35192 (including a physical distribution medium), accompanied by the 35193 Corresponding Source fixed on a durable physical medium 35194 customarily used for software interchange. 35195 35196 b. Convey the object code in, or embodied in, a physical product 35197 (including a physical distribution medium), accompanied by a 35198 written offer, valid for at least three years and valid for 35199 as long as you offer spare parts or customer support for that 35200 product model, to give anyone who possesses the object code 35201 either (1) a copy of the Corresponding Source for all the 35202 software in the product that is covered by this License, on a 35203 durable physical medium customarily used for software 35204 interchange, for a price no more than your reasonable cost of 35205 physically performing this conveying of source, or (2) access 35206 to copy the Corresponding Source from a network server at no 35207 charge. 35208 35209 c. Convey individual copies of the object code with a copy of 35210 the written offer to provide the Corresponding Source. This 35211 alternative is allowed only occasionally and noncommercially, 35212 and only if you received the object code with such an offer, 35213 in accord with subsection 6b. 35214 35215 d. Convey the object code by offering access from a designated 35216 place (gratis or for a charge), and offer equivalent access 35217 to the Corresponding Source in the same way through the same 35218 place at no further charge. You need not require recipients 35219 to copy the Corresponding Source along with the object code. 35220 If the place to copy the object code is a network server, the 35221 Corresponding Source may be on a different server (operated 35222 by you or a third party) that supports equivalent copying 35223 facilities, provided you maintain clear directions next to 35224 the object code saying where to find the Corresponding Source. 35225 Regardless of what server hosts the Corresponding Source, you 35226 remain obligated to ensure that it is available for as long 35227 as needed to satisfy these requirements. 35228 35229 e. Convey the object code using peer-to-peer transmission, 35230 provided you inform other peers where the object code and 35231 Corresponding Source of the work are being offered to the 35232 general public at no charge under subsection 6d. 35233 35234 35235 A separable portion of the object code, whose source code is 35236 excluded from the Corresponding Source as a System Library, need 35237 not be included in conveying the object code work. 35238 35239 A "User Product" is either (1) a "consumer product", which means 35240 any tangible personal property which is normally used for personal, 35241 family, or household purposes, or (2) anything designed or sold for 35242 incorporation into a dwelling. In determining whether a product 35243 is a consumer product, doubtful cases shall be resolved in favor of 35244 coverage. For a particular product received by a particular user, 35245 "normally used" refers to a typical or common use of that class of 35246 product, regardless of the status of the particular user or of the 35247 way in which the particular user actually uses, or expects or is 35248 expected to use, the product. A product is a consumer product 35249 regardless of whether the product has substantial commercial, 35250 industrial or non-consumer uses, unless such uses represent the 35251 only significant mode of use of the product. 35252 35253 "Installation Information" for a User Product means any methods, 35254 procedures, authorization keys, or other information required to 35255 install and execute modified versions of a covered work in that 35256 User Product from a modified version of its Corresponding Source. 35257 The information must suffice to ensure that the continued 35258 functioning of the modified object code is in no case prevented or 35259 interfered with solely because modification has been made. 35260 35261 If you convey an object code work under this section in, or with, 35262 or specifically for use in, a User Product, and the conveying 35263 occurs as part of a transaction in which the right of possession 35264 and use of the User Product is transferred to the recipient in 35265 perpetuity or for a fixed term (regardless of how the transaction 35266 is characterized), the Corresponding Source conveyed under this 35267 section must be accompanied by the Installation Information. But 35268 this requirement does not apply if neither you nor any third party 35269 retains the ability to install modified object code on the User 35270 Product (for example, the work has been installed in ROM). 35271 35272 The requirement to provide Installation Information does not 35273 include a requirement to continue to provide support service, 35274 warranty, or updates for a work that has been modified or 35275 installed by the recipient, or for the User Product in which it 35276 has been modified or installed. Access to a network may be denied 35277 when the modification itself materially and adversely affects the 35278 operation of the network or violates the rules and protocols for 35279 communication across the network. 35280 35281 Corresponding Source conveyed, and Installation Information 35282 provided, in accord with this section must be in a format that is 35283 publicly documented (and with an implementation available to the 35284 public in source code form), and must require no special password 35285 or key for unpacking, reading or copying. 35286 35287 7. Additional Terms. 35288 35289 "Additional permissions" are terms that supplement the terms of 35290 this License by making exceptions from one or more of its 35291 conditions. Additional permissions that are applicable to the 35292 entire Program shall be treated as though they were included in 35293 this License, to the extent that they are valid under applicable 35294 law. If additional permissions apply only to part of the Program, 35295 that part may be used separately under those permissions, but the 35296 entire Program remains governed by this License without regard to 35297 the additional permissions. 35298 35299 When you convey a copy of a covered work, you may at your option 35300 remove any additional permissions from that copy, or from any part 35301 of it. (Additional permissions may be written to require their own 35302 removal in certain cases when you modify the work.) You may place 35303 additional permissions on material, added by you to a covered work, 35304 for which you have or can give appropriate copyright permission. 35305 35306 Notwithstanding any other provision of this License, for material 35307 you add to a covered work, you may (if authorized by the copyright 35308 holders of that material) supplement the terms of this License 35309 with terms: 35310 35311 a. Disclaiming warranty or limiting liability differently from 35312 the terms of sections 15 and 16 of this License; or 35313 35314 b. Requiring preservation of specified reasonable legal notices 35315 or author attributions in that material or in the Appropriate 35316 Legal Notices displayed by works containing it; or 35317 35318 c. Prohibiting misrepresentation of the origin of that material, 35319 or requiring that modified versions of such material be 35320 marked in reasonable ways as different from the original 35321 version; or 35322 35323 d. Limiting the use for publicity purposes of names of licensors 35324 or authors of the material; or 35325 35326 e. Declining to grant rights under trademark law for use of some 35327 trade names, trademarks, or service marks; or 35328 35329 f. Requiring indemnification of licensors and authors of that 35330 material by anyone who conveys the material (or modified 35331 versions of it) with contractual assumptions of liability to 35332 the recipient, for any liability that these contractual 35333 assumptions directly impose on those licensors and authors. 35334 35335 All other non-permissive additional terms are considered "further 35336 restrictions" within the meaning of section 10. If the Program as 35337 you received it, or any part of it, contains a notice stating that 35338 it is governed by this License along with a term that is a further 35339 restriction, you may remove that term. If a license document 35340 contains a further restriction but permits relicensing or 35341 conveying under this License, you may add to a covered work 35342 material governed by the terms of that license document, provided 35343 that the further restriction does not survive such relicensing or 35344 conveying. 35345 35346 If you add terms to a covered work in accord with this section, you 35347 must place, in the relevant source files, a statement of the 35348 additional terms that apply to those files, or a notice indicating 35349 where to find the applicable terms. 35350 35351 Additional terms, permissive or non-permissive, may be stated in 35352 the form of a separately written license, or stated as exceptions; 35353 the above requirements apply either way. 35354 35355 8. Termination. 35356 35357 You may not propagate or modify a covered work except as expressly 35358 provided under this License. Any attempt otherwise to propagate or 35359 modify it is void, and will automatically terminate your rights 35360 under this License (including any patent licenses granted under 35361 the third paragraph of section 11). 35362 35363 However, if you cease all violation of this License, then your 35364 license from a particular copyright holder is reinstated (a) 35365 provisionally, unless and until the copyright holder explicitly 35366 and finally terminates your license, and (b) permanently, if the 35367 copyright holder fails to notify you of the violation by some 35368 reasonable means prior to 60 days after the cessation. 35369 35370 Moreover, your license from a particular copyright holder is 35371 reinstated permanently if the copyright holder notifies you of the 35372 violation by some reasonable means, this is the first time you have 35373 received notice of violation of this License (for any work) from 35374 that copyright holder, and you cure the violation prior to 30 days 35375 after your receipt of the notice. 35376 35377 Termination of your rights under this section does not terminate 35378 the licenses of parties who have received copies or rights from 35379 you under this License. If your rights have been terminated and 35380 not permanently reinstated, you do not qualify to receive new 35381 licenses for the same material under section 10. 35382 35383 9. Acceptance Not Required for Having Copies. 35384 35385 You are not required to accept this License in order to receive or 35386 run a copy of the Program. Ancillary propagation of a covered work 35387 occurring solely as a consequence of using peer-to-peer 35388 transmission to receive a copy likewise does not require 35389 acceptance. However, nothing other than this License grants you 35390 permission to propagate or modify any covered work. These actions 35391 infringe copyright if you do not accept this License. Therefore, 35392 by modifying or propagating a covered work, you indicate your 35393 acceptance of this License to do so. 35394 35395 10. Automatic Licensing of Downstream Recipients. 35396 35397 Each time you convey a covered work, the recipient automatically 35398 receives a license from the original licensors, to run, modify and 35399 propagate that work, subject to this License. You are not 35400 responsible for enforcing compliance by third parties with this 35401 License. 35402 35403 An "entity transaction" is a transaction transferring control of an 35404 organization, or substantially all assets of one, or subdividing an 35405 organization, or merging organizations. If propagation of a 35406 covered work results from an entity transaction, each party to that 35407 transaction who receives a copy of the work also receives whatever 35408 licenses to the work the party's predecessor in interest had or 35409 could give under the previous paragraph, plus a right to 35410 possession of the Corresponding Source of the work from the 35411 predecessor in interest, if the predecessor has it or can get it 35412 with reasonable efforts. 35413 35414 You may not impose any further restrictions on the exercise of the 35415 rights granted or affirmed under this License. For example, you 35416 may not impose a license fee, royalty, or other charge for 35417 exercise of rights granted under this License, and you may not 35418 initiate litigation (including a cross-claim or counterclaim in a 35419 lawsuit) alleging that any patent claim is infringed by making, 35420 using, selling, offering for sale, or importing the Program or any 35421 portion of it. 35422 35423 11. Patents. 35424 35425 A "contributor" is a copyright holder who authorizes use under this 35426 License of the Program or a work on which the Program is based. 35427 The work thus licensed is called the contributor's "contributor 35428 version". 35429 35430 A contributor's "essential patent claims" are all patent claims 35431 owned or controlled by the contributor, whether already acquired or 35432 hereafter acquired, that would be infringed by some manner, 35433 permitted by this License, of making, using, or selling its 35434 contributor version, but do not include claims that would be 35435 infringed only as a consequence of further modification of the 35436 contributor version. For purposes of this definition, "control" 35437 includes the right to grant patent sublicenses in a manner 35438 consistent with the requirements of this License. 35439 35440 Each contributor grants you a non-exclusive, worldwide, 35441 royalty-free patent license under the contributor's essential 35442 patent claims, to make, use, sell, offer for sale, import and 35443 otherwise run, modify and propagate the contents of its 35444 contributor version. 35445 35446 In the following three paragraphs, a "patent license" is any 35447 express agreement or commitment, however denominated, not to 35448 enforce a patent (such as an express permission to practice a 35449 patent or covenant not to sue for patent infringement). To 35450 "grant" such a patent license to a party means to make such an 35451 agreement or commitment not to enforce a patent against the party. 35452 35453 If you convey a covered work, knowingly relying on a patent 35454 license, and the Corresponding Source of the work is not available 35455 for anyone to copy, free of charge and under the terms of this 35456 License, through a publicly available network server or other 35457 readily accessible means, then you must either (1) cause the 35458 Corresponding Source to be so available, or (2) arrange to deprive 35459 yourself of the benefit of the patent license for this particular 35460 work, or (3) arrange, in a manner consistent with the requirements 35461 of this License, to extend the patent license to downstream 35462 recipients. "Knowingly relying" means you have actual knowledge 35463 that, but for the patent license, your conveying the covered work 35464 in a country, or your recipient's use of the covered work in a 35465 country, would infringe one or more identifiable patents in that 35466 country that you have reason to believe are valid. 35467 35468 If, pursuant to or in connection with a single transaction or 35469 arrangement, you convey, or propagate by procuring conveyance of, a 35470 covered work, and grant a patent license to some of the parties 35471 receiving the covered work authorizing them to use, propagate, 35472 modify or convey a specific copy of the covered work, then the 35473 patent license you grant is automatically extended to all 35474 recipients of the covered work and works based on it. 35475 35476 A patent license is "discriminatory" if it does not include within 35477 the scope of its coverage, prohibits the exercise of, or is 35478 conditioned on the non-exercise of one or more of the rights that 35479 are specifically granted under this License. You may not convey a 35480 covered work if you are a party to an arrangement with a third 35481 party that is in the business of distributing software, under 35482 which you make payment to the third party based on the extent of 35483 your activity of conveying the work, and under which the third 35484 party grants, to any of the parties who would receive the covered 35485 work from you, a discriminatory patent license (a) in connection 35486 with copies of the covered work conveyed by you (or copies made 35487 from those copies), or (b) primarily for and in connection with 35488 specific products or compilations that contain the covered work, 35489 unless you entered into that arrangement, or that patent license 35490 was granted, prior to 28 March 2007. 35491 35492 Nothing in this License shall be construed as excluding or limiting 35493 any implied license or other defenses to infringement that may 35494 otherwise be available to you under applicable patent law. 35495 35496 12. No Surrender of Others' Freedom. 35497 35498 If conditions are imposed on you (whether by court order, 35499 agreement or otherwise) that contradict the conditions of this 35500 License, they do not excuse you from the conditions of this 35501 License. If you cannot convey a covered work so as to satisfy 35502 simultaneously your obligations under this License and any other 35503 pertinent obligations, then as a consequence you may not convey it 35504 at all. For example, if you agree to terms that obligate you to 35505 collect a royalty for further conveying from those to whom you 35506 convey the Program, the only way you could satisfy both those 35507 terms and this License would be to refrain entirely from conveying 35508 the Program. 35509 35510 13. Use with the GNU Affero General Public License. 35511 35512 Notwithstanding any other provision of this License, you have 35513 permission to link or combine any covered work with a work licensed 35514 under version 3 of the GNU Affero General Public License into a 35515 single combined work, and to convey the resulting work. The terms 35516 of this License will continue to apply to the part which is the 35517 covered work, but the special requirements of the GNU Affero 35518 General Public License, section 13, concerning interaction through 35519 a network will apply to the combination as such. 35520 35521 14. Revised Versions of this License. 35522 35523 The Free Software Foundation may publish revised and/or new 35524 versions of the GNU General Public License from time to time. 35525 Such new versions will be similar in spirit to the present 35526 version, but may differ in detail to address new problems or 35527 concerns. 35528 35529 Each version is given a distinguishing version number. If the 35530 Program specifies that a certain numbered version of the GNU 35531 General Public License "or any later version" applies to it, you 35532 have the option of following the terms and conditions either of 35533 that numbered version or of any later version published by the 35534 Free Software Foundation. If the Program does not specify a 35535 version number of the GNU General Public License, you may choose 35536 any version ever published by the Free Software Foundation. 35537 35538 If the Program specifies that a proxy can decide which future 35539 versions of the GNU General Public License can be used, that 35540 proxy's public statement of acceptance of a version permanently 35541 authorizes you to choose that version for the Program. 35542 35543 Later license versions may give you additional or different 35544 permissions. However, no additional obligations are imposed on any 35545 author or copyright holder as a result of your choosing to follow a 35546 later version. 35547 35548 15. Disclaimer of Warranty. 35549 35550 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 35551 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE 35552 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" 35553 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, 35554 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 35555 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE 35556 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. 35557 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL 35558 NECESSARY SERVICING, REPAIR OR CORRECTION. 35559 35560 16. Limitation of Liability. 35561 35562 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN 35563 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES 35564 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU 35565 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR 35566 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE 35567 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA 35568 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 35569 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 35570 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF 35571 THE POSSIBILITY OF SUCH DAMAGES. 35572 35573 17. Interpretation of Sections 15 and 16. 35574 35575 If the disclaimer of warranty and limitation of liability provided 35576 above cannot be given local legal effect according to their terms, 35577 reviewing courts shall apply local law that most closely 35578 approximates an absolute waiver of all civil liability in 35579 connection with the Program, unless a warranty or assumption of 35580 liability accompanies a copy of the Program in return for a fee. 35581 35582 35583 END OF TERMS AND CONDITIONS 35584 =========================== 35585 35586 How to Apply These Terms to Your New Programs 35587 ============================================= 35588 35589 If you develop a new program, and you want it to be of the greatest 35590 possible use to the public, the best way to achieve this is to make it 35591 free software which everyone can redistribute and change under these 35592 terms. 35593 35594 To do so, attach the following notices to the program. It is safest 35595 to attach them to the start of each source file to most effectively 35596 state the exclusion of warranty; and each file should have at least the 35597 "copyright" line and a pointer to where the full notice is found. 35598 35599 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 35600 Copyright (C) YEAR NAME OF AUTHOR 35601 35602 This program is free software: you can redistribute it and/or modify 35603 it under the terms of the GNU General Public License as published by 35604 the Free Software Foundation, either version 3 of the License, or (at 35605 your option) any later version. 35606 35607 This program is distributed in the hope that it will be useful, but 35608 WITHOUT ANY WARRANTY; without even the implied warranty of 35609 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 35610 General Public License for more details. 35611 35612 You should have received a copy of the GNU General Public License 35613 along with this program. If not, see `http://www.gnu.org/licenses/'. 35614 35615 Also add information on how to contact you by electronic and paper 35616 mail. 35617 35618 If the program does terminal interaction, make it output a short 35619 notice like this when it starts in an interactive mode: 35620 35621 PROGRAM Copyright (C) YEAR NAME OF AUTHOR 35622 This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 35623 This is free software, and you are welcome to redistribute it 35624 under certain conditions; type `show c' for details. 35625 35626 The hypothetical commands `show w' and `show c' should show the 35627 appropriate parts of the General Public License. Of course, your 35628 program's commands might be different; for a GUI interface, you would 35629 use an "about box". 35630 35631 You should also get your employer (if you work as a programmer) or 35632 school, if any, to sign a "copyright disclaimer" for the program, if 35633 necessary. For more information on this, and how to apply and follow 35634 the GNU GPL, see `http://www.gnu.org/licenses/'. 35635 35636 The GNU General Public License does not permit incorporating your 35637 program into proprietary programs. If your program is a subroutine 35638 library, you may consider it more useful to permit linking proprietary 35639 applications with the library. If this is what you want to do, use the 35640 GNU Lesser General Public License instead of this License. But first, 35641 please read `http://www.gnu.org/philosophy/why-not-lgpl.html'. 35642 35643 35644 File: gccint.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top 35645 35646 GNU Free Documentation License 35647 ****************************** 35648 35649 Version 1.2, November 2002 35650 35651 Copyright (C) 2000,2001,2002 Free Software Foundation, Inc. 35652 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA 35653 35654 Everyone is permitted to copy and distribute verbatim copies 35655 of this license document, but changing it is not allowed. 35656 35657 0. PREAMBLE 35658 35659 The purpose of this License is to make a manual, textbook, or other 35660 functional and useful document "free" in the sense of freedom: to 35661 assure everyone the effective freedom to copy and redistribute it, 35662 with or without modifying it, either commercially or 35663 noncommercially. Secondarily, this License preserves for the 35664 author and publisher a way to get credit for their work, while not 35665 being considered responsible for modifications made by others. 35666 35667 This License is a kind of "copyleft", which means that derivative 35668 works of the document must themselves be free in the same sense. 35669 It complements the GNU General Public License, which is a copyleft 35670 license designed for free software. 35671 35672 We have designed this License in order to use it for manuals for 35673 free software, because free software needs free documentation: a 35674 free program should come with manuals providing the same freedoms 35675 that the software does. But this License is not limited to 35676 software manuals; it can be used for any textual work, regardless 35677 of subject matter or whether it is published as a printed book. 35678 We recommend this License principally for works whose purpose is 35679 instruction or reference. 35680 35681 1. APPLICABILITY AND DEFINITIONS 35682 35683 This License applies to any manual or other work, in any medium, 35684 that contains a notice placed by the copyright holder saying it 35685 can be distributed under the terms of this License. Such a notice 35686 grants a world-wide, royalty-free license, unlimited in duration, 35687 to use that work under the conditions stated herein. The 35688 "Document", below, refers to any such manual or work. Any member 35689 of the public is a licensee, and is addressed as "you". You 35690 accept the license if you copy, modify or distribute the work in a 35691 way requiring permission under copyright law. 35692 35693 A "Modified Version" of the Document means any work containing the 35694 Document or a portion of it, either copied verbatim, or with 35695 modifications and/or translated into another language. 35696 35697 A "Secondary Section" is a named appendix or a front-matter section 35698 of the Document that deals exclusively with the relationship of the 35699 publishers or authors of the Document to the Document's overall 35700 subject (or to related matters) and contains nothing that could 35701 fall directly within that overall subject. (Thus, if the Document 35702 is in part a textbook of mathematics, a Secondary Section may not 35703 explain any mathematics.) The relationship could be a matter of 35704 historical connection with the subject or with related matters, or 35705 of legal, commercial, philosophical, ethical or political position 35706 regarding them. 35707 35708 The "Invariant Sections" are certain Secondary Sections whose 35709 titles are designated, as being those of Invariant Sections, in 35710 the notice that says that the Document is released under this 35711 License. If a section does not fit the above definition of 35712 Secondary then it is not allowed to be designated as Invariant. 35713 The Document may contain zero Invariant Sections. If the Document 35714 does not identify any Invariant Sections then there are none. 35715 35716 The "Cover Texts" are certain short passages of text that are 35717 listed, as Front-Cover Texts or Back-Cover Texts, in the notice 35718 that says that the Document is released under this License. A 35719 Front-Cover Text may be at most 5 words, and a Back-Cover Text may 35720 be at most 25 words. 35721 35722 A "Transparent" copy of the Document means a machine-readable copy, 35723 represented in a format whose specification is available to the 35724 general public, that is suitable for revising the document 35725 straightforwardly with generic text editors or (for images 35726 composed of pixels) generic paint programs or (for drawings) some 35727 widely available drawing editor, and that is suitable for input to 35728 text formatters or for automatic translation to a variety of 35729 formats suitable for input to text formatters. A copy made in an 35730 otherwise Transparent file format whose markup, or absence of 35731 markup, has been arranged to thwart or discourage subsequent 35732 modification by readers is not Transparent. An image format is 35733 not Transparent if used for any substantial amount of text. A 35734 copy that is not "Transparent" is called "Opaque". 35735 35736 Examples of suitable formats for Transparent copies include plain 35737 ASCII without markup, Texinfo input format, LaTeX input format, 35738 SGML or XML using a publicly available DTD, and 35739 standard-conforming simple HTML, PostScript or PDF designed for 35740 human modification. Examples of transparent image formats include 35741 PNG, XCF and JPG. Opaque formats include proprietary formats that 35742 can be read and edited only by proprietary word processors, SGML or 35743 XML for which the DTD and/or processing tools are not generally 35744 available, and the machine-generated HTML, PostScript or PDF 35745 produced by some word processors for output purposes only. 35746 35747 The "Title Page" means, for a printed book, the title page itself, 35748 plus such following pages as are needed to hold, legibly, the 35749 material this License requires to appear in the title page. For 35750 works in formats which do not have any title page as such, "Title 35751 Page" means the text near the most prominent appearance of the 35752 work's title, preceding the beginning of the body of the text. 35753 35754 A section "Entitled XYZ" means a named subunit of the Document 35755 whose title either is precisely XYZ or contains XYZ in parentheses 35756 following text that translates XYZ in another language. (Here XYZ 35757 stands for a specific section name mentioned below, such as 35758 "Acknowledgements", "Dedications", "Endorsements", or "History".) 35759 To "Preserve the Title" of such a section when you modify the 35760 Document means that it remains a section "Entitled XYZ" according 35761 to this definition. 35762 35763 The Document may include Warranty Disclaimers next to the notice 35764 which states that this License applies to the Document. These 35765 Warranty Disclaimers are considered to be included by reference in 35766 this License, but only as regards disclaiming warranties: any other 35767 implication that these Warranty Disclaimers may have is void and 35768 has no effect on the meaning of this License. 35769 35770 2. VERBATIM COPYING 35771 35772 You may copy and distribute the Document in any medium, either 35773 commercially or noncommercially, provided that this License, the 35774 copyright notices, and the license notice saying this License 35775 applies to the Document are reproduced in all copies, and that you 35776 add no other conditions whatsoever to those of this License. You 35777 may not use technical measures to obstruct or control the reading 35778 or further copying of the copies you make or distribute. However, 35779 you may accept compensation in exchange for copies. If you 35780 distribute a large enough number of copies you must also follow 35781 the conditions in section 3. 35782 35783 You may also lend copies, under the same conditions stated above, 35784 and you may publicly display copies. 35785 35786 3. COPYING IN QUANTITY 35787 35788 If you publish printed copies (or copies in media that commonly 35789 have printed covers) of the Document, numbering more than 100, and 35790 the Document's license notice requires Cover Texts, you must 35791 enclose the copies in covers that carry, clearly and legibly, all 35792 these Cover Texts: Front-Cover Texts on the front cover, and 35793 Back-Cover Texts on the back cover. Both covers must also clearly 35794 and legibly identify you as the publisher of these copies. The 35795 front cover must present the full title with all words of the 35796 title equally prominent and visible. You may add other material 35797 on the covers in addition. Copying with changes limited to the 35798 covers, as long as they preserve the title of the Document and 35799 satisfy these conditions, can be treated as verbatim copying in 35800 other respects. 35801 35802 If the required texts for either cover are too voluminous to fit 35803 legibly, you should put the first ones listed (as many as fit 35804 reasonably) on the actual cover, and continue the rest onto 35805 adjacent pages. 35806 35807 If you publish or distribute Opaque copies of the Document 35808 numbering more than 100, you must either include a 35809 machine-readable Transparent copy along with each Opaque copy, or 35810 state in or with each Opaque copy a computer-network location from 35811 which the general network-using public has access to download 35812 using public-standard network protocols a complete Transparent 35813 copy of the Document, free of added material. If you use the 35814 latter option, you must take reasonably prudent steps, when you 35815 begin distribution of Opaque copies in quantity, to ensure that 35816 this Transparent copy will remain thus accessible at the stated 35817 location until at least one year after the last time you 35818 distribute an Opaque copy (directly or through your agents or 35819 retailers) of that edition to the public. 35820 35821 It is requested, but not required, that you contact the authors of 35822 the Document well before redistributing any large number of 35823 copies, to give them a chance to provide you with an updated 35824 version of the Document. 35825 35826 4. MODIFICATIONS 35827 35828 You may copy and distribute a Modified Version of the Document 35829 under the conditions of sections 2 and 3 above, provided that you 35830 release the Modified Version under precisely this License, with 35831 the Modified Version filling the role of the Document, thus 35832 licensing distribution and modification of the Modified Version to 35833 whoever possesses a copy of it. In addition, you must do these 35834 things in the Modified Version: 35835 35836 A. Use in the Title Page (and on the covers, if any) a title 35837 distinct from that of the Document, and from those of 35838 previous versions (which should, if there were any, be listed 35839 in the History section of the Document). You may use the 35840 same title as a previous version if the original publisher of 35841 that version gives permission. 35842 35843 B. List on the Title Page, as authors, one or more persons or 35844 entities responsible for authorship of the modifications in 35845 the Modified Version, together with at least five of the 35846 principal authors of the Document (all of its principal 35847 authors, if it has fewer than five), unless they release you 35848 from this requirement. 35849 35850 C. State on the Title page the name of the publisher of the 35851 Modified Version, as the publisher. 35852 35853 D. Preserve all the copyright notices of the Document. 35854 35855 E. Add an appropriate copyright notice for your modifications 35856 adjacent to the other copyright notices. 35857 35858 F. Include, immediately after the copyright notices, a license 35859 notice giving the public permission to use the Modified 35860 Version under the terms of this License, in the form shown in 35861 the Addendum below. 35862 35863 G. Preserve in that license notice the full lists of Invariant 35864 Sections and required Cover Texts given in the Document's 35865 license notice. 35866 35867 H. Include an unaltered copy of this License. 35868 35869 I. Preserve the section Entitled "History", Preserve its Title, 35870 and add to it an item stating at least the title, year, new 35871 authors, and publisher of the Modified Version as given on 35872 the Title Page. If there is no section Entitled "History" in 35873 the Document, create one stating the title, year, authors, 35874 and publisher of the Document as given on its Title Page, 35875 then add an item describing the Modified Version as stated in 35876 the previous sentence. 35877 35878 J. Preserve the network location, if any, given in the Document 35879 for public access to a Transparent copy of the Document, and 35880 likewise the network locations given in the Document for 35881 previous versions it was based on. These may be placed in 35882 the "History" section. You may omit a network location for a 35883 work that was published at least four years before the 35884 Document itself, or if the original publisher of the version 35885 it refers to gives permission. 35886 35887 K. For any section Entitled "Acknowledgements" or "Dedications", 35888 Preserve the Title of the section, and preserve in the 35889 section all the substance and tone of each of the contributor 35890 acknowledgements and/or dedications given therein. 35891 35892 L. Preserve all the Invariant Sections of the Document, 35893 unaltered in their text and in their titles. Section numbers 35894 or the equivalent are not considered part of the section 35895 titles. 35896 35897 M. Delete any section Entitled "Endorsements". Such a section 35898 may not be included in the Modified Version. 35899 35900 N. Do not retitle any existing section to be Entitled 35901 "Endorsements" or to conflict in title with any Invariant 35902 Section. 35903 35904 O. Preserve any Warranty Disclaimers. 35905 35906 If the Modified Version includes new front-matter sections or 35907 appendices that qualify as Secondary Sections and contain no 35908 material copied from the Document, you may at your option 35909 designate some or all of these sections as invariant. To do this, 35910 add their titles to the list of Invariant Sections in the Modified 35911 Version's license notice. These titles must be distinct from any 35912 other section titles. 35913 35914 You may add a section Entitled "Endorsements", provided it contains 35915 nothing but endorsements of your Modified Version by various 35916 parties--for example, statements of peer review or that the text 35917 has been approved by an organization as the authoritative 35918 definition of a standard. 35919 35920 You may add a passage of up to five words as a Front-Cover Text, 35921 and a passage of up to 25 words as a Back-Cover Text, to the end 35922 of the list of Cover Texts in the Modified Version. Only one 35923 passage of Front-Cover Text and one of Back-Cover Text may be 35924 added by (or through arrangements made by) any one entity. If the 35925 Document already includes a cover text for the same cover, 35926 previously added by you or by arrangement made by the same entity 35927 you are acting on behalf of, you may not add another; but you may 35928 replace the old one, on explicit permission from the previous 35929 publisher that added the old one. 35930 35931 The author(s) and publisher(s) of the Document do not by this 35932 License give permission to use their names for publicity for or to 35933 assert or imply endorsement of any Modified Version. 35934 35935 5. COMBINING DOCUMENTS 35936 35937 You may combine the Document with other documents released under 35938 this License, under the terms defined in section 4 above for 35939 modified versions, provided that you include in the combination 35940 all of the Invariant Sections of all of the original documents, 35941 unmodified, and list them all as Invariant Sections of your 35942 combined work in its license notice, and that you preserve all 35943 their Warranty Disclaimers. 35944 35945 The combined work need only contain one copy of this License, and 35946 multiple identical Invariant Sections may be replaced with a single 35947 copy. If there are multiple Invariant Sections with the same name 35948 but different contents, make the title of each such section unique 35949 by adding at the end of it, in parentheses, the name of the 35950 original author or publisher of that section if known, or else a 35951 unique number. Make the same adjustment to the section titles in 35952 the list of Invariant Sections in the license notice of the 35953 combined work. 35954 35955 In the combination, you must combine any sections Entitled 35956 "History" in the various original documents, forming one section 35957 Entitled "History"; likewise combine any sections Entitled 35958 "Acknowledgements", and any sections Entitled "Dedications". You 35959 must delete all sections Entitled "Endorsements." 35960 35961 6. COLLECTIONS OF DOCUMENTS 35962 35963 You may make a collection consisting of the Document and other 35964 documents released under this License, and replace the individual 35965 copies of this License in the various documents with a single copy 35966 that is included in the collection, provided that you follow the 35967 rules of this License for verbatim copying of each of the 35968 documents in all other respects. 35969 35970 You may extract a single document from such a collection, and 35971 distribute it individually under this License, provided you insert 35972 a copy of this License into the extracted document, and follow 35973 this License in all other respects regarding verbatim copying of 35974 that document. 35975 35976 7. AGGREGATION WITH INDEPENDENT WORKS 35977 35978 A compilation of the Document or its derivatives with other 35979 separate and independent documents or works, in or on a volume of 35980 a storage or distribution medium, is called an "aggregate" if the 35981 copyright resulting from the compilation is not used to limit the 35982 legal rights of the compilation's users beyond what the individual 35983 works permit. When the Document is included in an aggregate, this 35984 License does not apply to the other works in the aggregate which 35985 are not themselves derivative works of the Document. 35986 35987 If the Cover Text requirement of section 3 is applicable to these 35988 copies of the Document, then if the Document is less than one half 35989 of the entire aggregate, the Document's Cover Texts may be placed 35990 on covers that bracket the Document within the aggregate, or the 35991 electronic equivalent of covers if the Document is in electronic 35992 form. Otherwise they must appear on printed covers that bracket 35993 the whole aggregate. 35994 35995 8. TRANSLATION 35996 35997 Translation is considered a kind of modification, so you may 35998 distribute translations of the Document under the terms of section 35999 4. Replacing Invariant Sections with translations requires special 36000 permission from their copyright holders, but you may include 36001 translations of some or all Invariant Sections in addition to the 36002 original versions of these Invariant Sections. You may include a 36003 translation of this License, and all the license notices in the 36004 Document, and any Warranty Disclaimers, provided that you also 36005 include the original English version of this License and the 36006 original versions of those notices and disclaimers. In case of a 36007 disagreement between the translation and the original version of 36008 this License or a notice or disclaimer, the original version will 36009 prevail. 36010 36011 If a section in the Document is Entitled "Acknowledgements", 36012 "Dedications", or "History", the requirement (section 4) to 36013 Preserve its Title (section 1) will typically require changing the 36014 actual title. 36015 36016 9. TERMINATION 36017 36018 You may not copy, modify, sublicense, or distribute the Document 36019 except as expressly provided for under this License. Any other 36020 attempt to copy, modify, sublicense or distribute the Document is 36021 void, and will automatically terminate your rights under this 36022 License. However, parties who have received copies, or rights, 36023 from you under this License will not have their licenses 36024 terminated so long as such parties remain in full compliance. 36025 36026 10. FUTURE REVISIONS OF THIS LICENSE 36027 36028 The Free Software Foundation may publish new, revised versions of 36029 the GNU Free Documentation License from time to time. Such new 36030 versions will be similar in spirit to the present version, but may 36031 differ in detail to address new problems or concerns. See 36032 `http://www.gnu.org/copyleft/'. 36033 36034 Each version of the License is given a distinguishing version 36035 number. If the Document specifies that a particular numbered 36036 version of this License "or any later version" applies to it, you 36037 have the option of following the terms and conditions either of 36038 that specified version or of any later version that has been 36039 published (not as a draft) by the Free Software Foundation. If 36040 the Document does not specify a version number of this License, 36041 you may choose any version ever published (not as a draft) by the 36042 Free Software Foundation. 36043 36044 ADDENDUM: How to use this License for your documents 36045 ==================================================== 36046 36047 To use this License in a document you have written, include a copy of 36048 the License in the document and put the following copyright and license 36049 notices just after the title page: 36050 36051 Copyright (C) YEAR YOUR NAME. 36052 Permission is granted to copy, distribute and/or modify this document 36053 under the terms of the GNU Free Documentation License, Version 1.2 36054 or any later version published by the Free Software Foundation; 36055 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 36056 Texts. A copy of the license is included in the section entitled ``GNU 36057 Free Documentation License''. 36058 36059 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, 36060 replace the "with...Texts." line with this: 36061 36062 with the Invariant Sections being LIST THEIR TITLES, with 36063 the Front-Cover Texts being LIST, and with the Back-Cover Texts 36064 being LIST. 36065 36066 If you have Invariant Sections without Cover Texts, or some other 36067 combination of the three, merge those two alternatives to suit the 36068 situation. 36069 36070 If your document contains nontrivial examples of program code, we 36071 recommend releasing these examples in parallel under your choice of 36072 free software license, such as the GNU General Public License, to 36073 permit their use in free software. 36074 36075 36076 File: gccint.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top 36077 36078 Contributors to GCC 36079 ******************* 36080 36081 The GCC project would like to thank its many contributors. Without 36082 them the project would not have been nearly as successful as it has 36083 been. Any omissions in this list are accidental. Feel free to contact 36084 <law (a] redhat.com> or <gerald (a] pfeifer.com> if you have been left out or 36085 some of your contributions are not listed. Please keep this list in 36086 alphabetical order. 36087 36088 * Analog Devices helped implement the support for complex data types 36089 and iterators. 36090 36091 * John David Anglin for threading-related fixes and improvements to 36092 libstdc++-v3, and the HP-UX port. 36093 36094 * James van Artsdalen wrote the code that makes efficient use of the 36095 Intel 80387 register stack. 36096 36097 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta 36098 Series port. 36099 36100 * Alasdair Baird for various bug fixes. 36101 36102 * Giovanni Bajo for analyzing lots of complicated C++ problem 36103 reports. 36104 36105 * Peter Barada for his work to improve code generation for new 36106 ColdFire cores. 36107 36108 * Gerald Baumgartner added the signature extension to the C++ front 36109 end. 36110 36111 * Godmar Back for his Java improvements and encouragement. 36112 36113 * Scott Bambrough for help porting the Java compiler. 36114 36115 * Wolfgang Bangerth for processing tons of bug reports. 36116 36117 * Jon Beniston for his Microsoft Windows port of Java. 36118 36119 * Daniel Berlin for better DWARF2 support, faster/better 36120 optimizations, improved alias analysis, plus migrating GCC to 36121 Bugzilla. 36122 36123 * Geoff Berry for his Java object serialization work and various 36124 patches. 36125 36126 * Uros Bizjak for the implementation of x87 math built-in functions 36127 and for various middle end and i386 back end improvements and bug 36128 fixes. 36129 36130 * Eric Blake for helping to make GCJ and libgcj conform to the 36131 specifications. 36132 36133 * Janne Blomqvist for contributions to GNU Fortran. 36134 36135 * Segher Boessenkool for various fixes. 36136 36137 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and 36138 other Java work. 36139 36140 * Neil Booth for work on cpplib, lang hooks, debug hooks and other 36141 miscellaneous clean-ups. 36142 36143 * Steven Bosscher for integrating the GNU Fortran front end into GCC 36144 and for contributing to the tree-ssa branch. 36145 36146 * Eric Botcazou for fixing middle- and backend bugs left and right. 36147 36148 * Per Bothner for his direction via the steering committee and 36149 various improvements to the infrastructure for supporting new 36150 languages. Chill front end implementation. Initial 36151 implementations of cpplib, fix-header, config.guess, libio, and 36152 past C++ library (libg++) maintainer. Dreaming up, designing and 36153 implementing much of GCJ. 36154 36155 * Devon Bowen helped port GCC to the Tahoe. 36156 36157 * Don Bowman for mips-vxworks contributions. 36158 36159 * Dave Brolley for work on cpplib and Chill. 36160 36161 * Paul Brook for work on the ARM architecture and maintaining GNU 36162 Fortran. 36163 36164 * Robert Brown implemented the support for Encore 32000 systems. 36165 36166 * Christian Bruel for improvements to local store elimination. 36167 36168 * Herman A.J. ten Brugge for various fixes. 36169 36170 * Joerg Brunsmann for Java compiler hacking and help with the GCJ 36171 FAQ. 36172 36173 * Joe Buck for his direction via the steering committee. 36174 36175 * Craig Burley for leadership of the G77 Fortran effort. 36176 36177 * Stephan Buys for contributing Doxygen notes for libstdc++. 36178 36179 * Paolo Carlini for libstdc++ work: lots of efficiency improvements 36180 to the C++ strings, streambufs and formatted I/O, hard detective 36181 work on the frustrating localization issues, and keeping up with 36182 the problem reports. 36183 36184 * John Carr for his alias work, SPARC hacking, infrastructure 36185 improvements, previous contributions to the steering committee, 36186 loop optimizations, etc. 36187 36188 * Stephane Carrez for 68HC11 and 68HC12 ports. 36189 36190 * Steve Chamberlain for support for the Renesas SH and H8 processors 36191 and the PicoJava processor, and for GCJ config fixes. 36192 36193 * Glenn Chambers for help with the GCJ FAQ. 36194 36195 * John-Marc Chandonia for various libgcj patches. 36196 36197 * Scott Christley for his Objective-C contributions. 36198 36199 * Eric Christopher for his Java porting help and clean-ups. 36200 36201 * Branko Cibej for more warning contributions. 36202 36203 * The GNU Classpath project for all of their merged runtime code. 36204 36205 * Nick Clifton for arm, mcore, fr30, v850, m32r work, `--help', and 36206 other random hacking. 36207 36208 * Michael Cook for libstdc++ cleanup patches to reduce warnings. 36209 36210 * R. Kelley Cook for making GCC buildable from a read-only directory 36211 as well as other miscellaneous build process and documentation 36212 clean-ups. 36213 36214 * Ralf Corsepius for SH testing and minor bug fixing. 36215 36216 * Stan Cox for care and feeding of the x86 port and lots of behind 36217 the scenes hacking. 36218 36219 * Alex Crain provided changes for the 3b1. 36220 36221 * Ian Dall for major improvements to the NS32k port. 36222 36223 * Paul Dale for his work to add uClinux platform support to the m68k 36224 backend. 36225 36226 * Dario Dariol contributed the four varieties of sample programs 36227 that print a copy of their source. 36228 36229 * Russell Davidson for fstream and stringstream fixes in libstdc++. 36230 36231 * Bud Davis for work on the G77 and GNU Fortran compilers. 36232 36233 * Mo DeJong for GCJ and libgcj bug fixes. 36234 36235 * DJ Delorie for the DJGPP port, build and libiberty maintenance, 36236 various bug fixes, and the M32C port. 36237 36238 * Arnaud Desitter for helping to debug GNU Fortran. 36239 36240 * Gabriel Dos Reis for contributions to G++, contributions and 36241 maintenance of GCC diagnostics infrastructure, libstdc++-v3, 36242 including `valarray<>', `complex<>', maintaining the numerics 36243 library (including that pesky `<limits>' :-) and keeping 36244 up-to-date anything to do with numbers. 36245 36246 * Ulrich Drepper for his work on glibc, testing of GCC using glibc, 36247 ISO C99 support, CFG dumping support, etc., plus support of the 36248 C++ runtime libraries including for all kinds of C interface 36249 issues, contributing and maintaining `complex<>', sanity checking 36250 and disbursement, configuration architecture, libio maintenance, 36251 and early math work. 36252 36253 * Zdenek Dvorak for a new loop unroller and various fixes. 36254 36255 * Richard Earnshaw for his ongoing work with the ARM. 36256 36257 * David Edelsohn for his direction via the steering committee, 36258 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa 36259 loop changes, doing the entire AIX port of libstdc++ with his bare 36260 hands, and for ensuring GCC properly keeps working on AIX. 36261 36262 * Kevin Ediger for the floating point formatting of num_put::do_put 36263 in libstdc++. 36264 36265 * Phil Edwards for libstdc++ work including configuration hackery, 36266 documentation maintainer, chief breaker of the web pages, the 36267 occasional iostream bug fix, and work on shared library symbol 36268 versioning. 36269 36270 * Paul Eggert for random hacking all over GCC. 36271 36272 * Mark Elbrecht for various DJGPP improvements, and for libstdc++ 36273 configuration support for locales and fstream-related fixes. 36274 36275 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and 36276 iostreams. 36277 36278 * Christian Ehrhardt for dealing with bug reports. 36279 36280 * Ben Elliston for his work to move the Objective-C runtime into its 36281 own subdirectory and for his work on autoconf. 36282 36283 * Revital Eres for work on the PowerPC 750CL port. 36284 36285 * Marc Espie for OpenBSD support. 36286 36287 * Doug Evans for much of the global optimization framework, arc, 36288 m32r, and SPARC work. 36289 36290 * Christopher Faylor for his work on the Cygwin port and for caring 36291 and feeding the gcc.gnu.org box and saving its users tons of spam. 36292 36293 * Fred Fish for BeOS support and Ada fixes. 36294 36295 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ. 36296 36297 * Peter Gerwinski for various bug fixes and the Pascal front end. 36298 36299 * Kaveh R. Ghazi for his direction via the steering committee, 36300 amazing work to make `-W -Wall -W* -Werror' useful, and 36301 continuously testing GCC on a plethora of platforms. Kaveh 36302 extends his gratitude to the CAIP Center at Rutgers University for 36303 providing him with computing resources to work on Free Software 36304 since the late 1980s. 36305 36306 * John Gilmore for a donation to the FSF earmarked improving GNU 36307 Java. 36308 36309 * Judy Goldberg for c++ contributions. 36310 36311 * Torbjorn Granlund for various fixes and the c-torture testsuite, 36312 multiply- and divide-by-constant optimization, improved long long 36313 support, improved leaf function register allocation, and his 36314 direction via the steering committee. 36315 36316 * Anthony Green for his `-Os' contributions and Java front end work. 36317 36318 * Stu Grossman for gdb hacking, allowing GCJ developers to debug 36319 Java code. 36320 36321 * Michael K. Gschwind contributed the port to the PDP-11. 36322 36323 * Ron Guilmette implemented the `protoize' and `unprotoize' tools, 36324 the support for Dwarf symbolic debugging information, and much of 36325 the support for System V Release 4. He has also worked heavily on 36326 the Intel 386 and 860 support. 36327 36328 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload 36329 GCSE. 36330 36331 * Bruno Haible for improvements in the runtime overhead for EH, new 36332 warnings and assorted bug fixes. 36333 36334 * Andrew Haley for his amazing Java compiler and library efforts. 36335 36336 * Chris Hanson assisted in making GCC work on HP-UX for the 9000 36337 series 300. 36338 36339 * Michael Hayes for various thankless work he's done trying to get 36340 the c30/c40 ports functional. Lots of loop and unroll 36341 improvements and fixes. 36342 36343 * Dara Hazeghi for wading through myriads of target-specific bug 36344 reports. 36345 36346 * Kate Hedstrom for staking the G77 folks with an initial testsuite. 36347 36348 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64 36349 work, loop opts, and generally fixing lots of old problems we've 36350 ignored for years, flow rewrite and lots of further stuff, 36351 including reviewing tons of patches. 36352 36353 * Aldy Hernandez for working on the PowerPC port, SIMD support, and 36354 various fixes. 36355 36356 * Nobuyuki Hikichi of Software Research Associates, Tokyo, 36357 contributed the support for the Sony NEWS machine. 36358 36359 * Kazu Hirata for caring and feeding the Renesas H8/300 port and 36360 various fixes. 36361 36362 * Katherine Holcomb for work on GNU Fortran. 36363 36364 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots 36365 of testing and bug fixing, particularly of GCC configury code. 36366 36367 * Steve Holmgren for MachTen patches. 36368 36369 * Jan Hubicka for his x86 port improvements. 36370 36371 * Falk Hueffner for working on C and optimization bug reports. 36372 36373 * Bernardo Innocenti for his m68k work, including merging of 36374 ColdFire improvements and uClinux support. 36375 36376 * Christian Iseli for various bug fixes. 36377 36378 * Kamil Iskra for general m68k hacking. 36379 36380 * Lee Iverson for random fixes and MIPS testing. 36381 36382 * Andreas Jaeger for testing and benchmarking of GCC and various bug 36383 fixes. 36384 36385 * Jakub Jelinek for his SPARC work and sibling call optimizations as 36386 well as lots of bug fixes and test cases, and for improving the 36387 Java build system. 36388 36389 * Janis Johnson for ia64 testing and fixes, her quality improvement 36390 sidetracks, and web page maintenance. 36391 36392 * Kean Johnston for SCO OpenServer support and various fixes. 36393 36394 * Tim Josling for the sample language treelang based originally on 36395 Richard Kenner's "toy" language. 36396 36397 * Nicolai Josuttis for additional libstdc++ documentation. 36398 36399 * Klaus Kaempf for his ongoing work to make alpha-vms a viable 36400 target. 36401 36402 * Steven G. Kargl for work on GNU Fortran. 36403 36404 * David Kashtan of SRI adapted GCC to VMS. 36405 36406 * Ryszard Kabatek for many, many libstdc++ bug fixes and 36407 optimizations of strings, especially member functions, and for 36408 auto_ptr fixes. 36409 36410 * Geoffrey Keating for his ongoing work to make the PPC work for 36411 GNU/Linux and his automatic regression tester. 36412 36413 * Brendan Kehoe for his ongoing work with G++ and for a lot of early 36414 work in just about every part of libstdc++. 36415 36416 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the 36417 MIL-STD-1750A. 36418 36419 * Richard Kenner of the New York University Ultracomputer Research 36420 Laboratory wrote the machine descriptions for the AMD 29000, the 36421 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the 36422 support for instruction attributes. He also made changes to 36423 better support RISC processors including changes to common 36424 subexpression elimination, strength reduction, function calling 36425 sequence handling, and condition code support, in addition to 36426 generalizing the code for frame pointer elimination and delay slot 36427 scheduling. Richard Kenner was also the head maintainer of GCC 36428 for several years. 36429 36430 * Mumit Khan for various contributions to the Cygwin and Mingw32 36431 ports and maintaining binary releases for Microsoft Windows hosts, 36432 and for massive libstdc++ porting work to Cygwin/Mingw32. 36433 36434 * Robin Kirkham for cpu32 support. 36435 36436 * Mark Klein for PA improvements. 36437 36438 * Thomas Koenig for various bug fixes. 36439 36440 * Bruce Korb for the new and improved fixincludes code. 36441 36442 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3 36443 effort. 36444 36445 * Charles LaBrec contributed the support for the Integrated Solutions 36446 68020 system. 36447 36448 * Asher Langton and Mike Kumbera for contributing Cray pointer 36449 support to GNU Fortran, and for other GNU Fortran improvements. 36450 36451 * Jeff Law for his direction via the steering committee, 36452 coordinating the entire egcs project and GCC 2.95, rolling out 36453 snapshots and releases, handling merges from GCC2, reviewing tons 36454 of patches that might have fallen through the cracks else, and 36455 random but extensive hacking. 36456 36457 * Marc Lehmann for his direction via the steering committee and 36458 helping with analysis and improvements of x86 performance. 36459 36460 * Victor Leikehman for work on GNU Fortran. 36461 36462 * Ted Lemon wrote parts of the RTL reader and printer. 36463 36464 * Kriang Lerdsuwanakij for C++ improvements including template as 36465 template parameter support, and many C++ fixes. 36466 36467 * Warren Levy for tremendous work on libgcj (Java Runtime Library) 36468 and random work on the Java front end. 36469 36470 * Alain Lichnewsky ported GCC to the MIPS CPU. 36471 36472 * Oskar Liljeblad for hacking on AWT and his many Java bug reports 36473 and patches. 36474 36475 * Robert Lipe for OpenServer support, new testsuites, testing, etc. 36476 36477 * Chen Liqin for various S+core related fixes/improvement, and for 36478 maintaining the S+core port. 36479 36480 * Weiwen Liu for testing and various bug fixes. 36481 36482 * Manuel Lo'pez-Iba'n~ez for improving `-Wconversion' and many other 36483 diagnostics fixes and improvements. 36484 36485 * Dave Love for his ongoing work with the Fortran front end and 36486 runtime libraries. 36487 36488 * Martin von Lo"wis for internal consistency checking infrastructure, 36489 various C++ improvements including namespace support, and tons of 36490 assistance with libstdc++/compiler merges. 36491 36492 * H.J. Lu for his previous contributions to the steering committee, 36493 many x86 bug reports, prototype patches, and keeping the GNU/Linux 36494 ports working. 36495 36496 * Greg McGary for random fixes and (someday) bounded pointers. 36497 36498 * Andrew MacLeod for his ongoing work in building a real EH system, 36499 various code generation improvements, work on the global 36500 optimizer, etc. 36501 36502 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC 36503 hacking improvements to compile-time performance, overall 36504 knowledge and direction in the area of instruction scheduling, and 36505 design and implementation of the automaton based instruction 36506 scheduler. 36507 36508 * Bob Manson for his behind the scenes work on dejagnu. 36509 36510 * Philip Martin for lots of libstdc++ string and vector iterator 36511 fixes and improvements, and string clean up and testsuites. 36512 36513 * All of the Mauve project contributors, for Java test code. 36514 36515 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements. 36516 36517 * Adam Megacz for his work on the Microsoft Windows port of GCJ. 36518 36519 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS, 36520 powerpc, haifa, ECOFF debug support, and other assorted hacking. 36521 36522 * Jason Merrill for his direction via the steering committee and 36523 leading the G++ effort. 36524 36525 * Martin Michlmayr for testing GCC on several architectures using the 36526 entire Debian archive. 36527 36528 * David Miller for his direction via the steering committee, lots of 36529 SPARC work, improvements in jump.c and interfacing with the Linux 36530 kernel developers. 36531 36532 * Gary Miller ported GCC to Charles River Data Systems machines. 36533 36534 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning 36535 the entire libstdc++ testsuite namespace-compatible. 36536 36537 * Mark Mitchell for his direction via the steering committee, 36538 mountains of C++ work, load/store hoisting out of loops, alias 36539 analysis improvements, ISO C `restrict' support, and serving as 36540 release manager for GCC 3.x. 36541 36542 * Alan Modra for various GNU/Linux bits and testing. 36543 36544 * Toon Moene for his direction via the steering committee, Fortran 36545 maintenance, and his ongoing work to make us make Fortran run fast. 36546 36547 * Jason Molenda for major help in the care and feeding of all the 36548 services on the gcc.gnu.org (formerly egcs.cygnus.com) 36549 machine--mail, web services, ftp services, etc etc. Doing all 36550 this work on scrap paper and the backs of envelopes would have 36551 been... difficult. 36552 36553 * Catherine Moore for fixing various ugly problems we have sent her 36554 way, including the haifa bug which was killing the Alpha & PowerPC 36555 Linux kernels. 36556 36557 * Mike Moreton for his various Java patches. 36558 36559 * David Mosberger-Tang for various Alpha improvements, and for the 36560 initial IA-64 port. 36561 36562 * Stephen Moshier contributed the floating point emulator that 36563 assists in cross-compilation and permits support for floating 36564 point numbers wider than 64 bits and for ISO C99 support. 36565 36566 * Bill Moyer for his behind the scenes work on various issues. 36567 36568 * Philippe De Muyter for his work on the m68k port. 36569 36570 * Joseph S. Myers for his work on the PDP-11 port, format checking 36571 and ISO C99 support, and continuous emphasis on (and contributions 36572 to) documentation. 36573 36574 * Nathan Myers for his work on libstdc++-v3: architecture and 36575 authorship through the first three snapshots, including 36576 implementation of locale infrastructure, string, shadow C headers, 36577 and the initial project documentation (DESIGN, CHECKLIST, and so 36578 forth). Later, more work on MT-safe string and shadow headers. 36579 36580 * Felix Natter for documentation on porting libstdc++. 36581 36582 * Nathanael Nerode for cleaning up the configuration/build process. 36583 36584 * NeXT, Inc. donated the front end that supports the Objective-C 36585 language. 36586 36587 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to 36588 the search engine setup, various documentation fixes and other 36589 small fixes. 36590 36591 * Geoff Noer for his work on getting cygwin native builds working. 36592 36593 * Diego Novillo for his work on Tree SSA, OpenMP, SPEC performance 36594 tracking web pages, GIMPLE tuples, and assorted fixes. 36595 36596 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64, 36597 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and 36598 related infrastructure improvements. 36599 36600 * Alexandre Oliva for various build infrastructure improvements, 36601 scripts and amazing testing work, including keeping libtool issues 36602 sane and happy. 36603 36604 * Stefan Olsson for work on mt_alloc. 36605 36606 * Melissa O'Neill for various NeXT fixes. 36607 36608 * Rainer Orth for random MIPS work, including improvements to GCC's 36609 o32 ABI support, improvements to dejagnu's MIPS support, Java 36610 configuration clean-ups and porting work, etc. 36611 36612 * Hartmut Penner for work on the s390 port. 36613 36614 * Paul Petersen wrote the machine description for the Alliant FX/8. 36615 36616 * Alexandre Petit-Bianco for implementing much of the Java compiler 36617 and continued Java maintainership. 36618 36619 * Matthias Pfaller for major improvements to the NS32k port. 36620 36621 * Gerald Pfeifer for his direction via the steering committee, 36622 pointing out lots of problems we need to solve, maintenance of the 36623 web pages, and taking care of documentation maintenance in general. 36624 36625 * Andrew Pinski for processing bug reports by the dozen. 36626 36627 * Ovidiu Predescu for his work on the Objective-C front end and 36628 runtime libraries. 36629 36630 * Jerry Quinn for major performance improvements in C++ formatted 36631 I/O. 36632 36633 * Ken Raeburn for various improvements to checker, MIPS ports and 36634 various cleanups in the compiler. 36635 36636 * Rolf W. Rasmussen for hacking on AWT. 36637 36638 * David Reese of Sun Microsystems contributed to the Solaris on 36639 PowerPC port. 36640 36641 * Volker Reichelt for keeping up with the problem reports. 36642 36643 * Joern Rennecke for maintaining the sh port, loop, regmove & reload 36644 hacking. 36645 36646 * Loren J. Rittle for improvements to libstdc++-v3 including the 36647 FreeBSD port, threading fixes, thread-related configury changes, 36648 critical threading documentation, and solutions to really tricky 36649 I/O problems, as well as keeping GCC properly working on FreeBSD 36650 and continuous testing. 36651 36652 * Craig Rodrigues for processing tons of bug reports. 36653 36654 * Ola Ro"nnerup for work on mt_alloc. 36655 36656 * Gavin Romig-Koch for lots of behind the scenes MIPS work. 36657 36658 * David Ronis inspired and encouraged Craig to rewrite the G77 36659 documentation in texinfo format by contributing a first pass at a 36660 translation of the old `g77-0.5.16/f/DOC' file. 36661 36662 * Ken Rose for fixes to GCC's delay slot filling code. 36663 36664 * Paul Rubin wrote most of the preprocessor. 36665 36666 * Pe'tur Runo'lfsson for major performance improvements in C++ 36667 formatted I/O and large file support in C++ filebuf. 36668 36669 * Chip Salzenberg for libstdc++ patches and improvements to locales, 36670 traits, Makefiles, libio, libtool hackery, and "long long" support. 36671 36672 * Juha Sarlin for improvements to the H8 code generator. 36673 36674 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series 36675 300. 36676 36677 * Roger Sayle for improvements to constant folding and GCC's RTL 36678 optimizers as well as for fixing numerous bugs. 36679 36680 * Bradley Schatz for his work on the GCJ FAQ. 36681 36682 * Peter Schauer wrote the code to allow debugging to work on the 36683 Alpha. 36684 36685 * William Schelter did most of the work on the Intel 80386 support. 36686 36687 * Tobias Schlu"ter for work on GNU Fortran. 36688 36689 * Bernd Schmidt for various code generation improvements and major 36690 work in the reload pass as well a serving as release manager for 36691 GCC 2.95.3. 36692 36693 * Peter Schmid for constant testing of libstdc++--especially 36694 application testing, going above and beyond what was requested for 36695 the release criteria--and libstdc++ header file tweaks. 36696 36697 * Jason Schroeder for jcf-dump patches. 36698 36699 * Andreas Schwab for his work on the m68k port. 36700 36701 * Lars Segerlund for work on GNU Fortran. 36702 36703 * Joel Sherrill for his direction via the steering committee, RTEMS 36704 contributions and RTEMS testing. 36705 36706 * Nathan Sidwell for many C++ fixes/improvements. 36707 36708 * Jeffrey Siegal for helping RMS with the original design of GCC, 36709 some code which handles the parse tree and RTL data structures, 36710 constant folding and help with the original VAX & m68k ports. 36711 36712 * Kenny Simpson for prompting libstdc++ fixes due to defect reports 36713 from the LWG (thereby keeping GCC in line with updates from the 36714 ISO). 36715 36716 * Franz Sirl for his ongoing work with making the PPC port stable 36717 for GNU/Linux. 36718 36719 * Andrey Slepuhin for assorted AIX hacking. 36720 36721 * Trevor Smigiel for contributing the SPU port. 36722 36723 * Christopher Smith did the port for Convex machines. 36724 36725 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports. 36726 36727 * Randy Smith finished the Sun FPA support. 36728 36729 * Scott Snyder for queue, iterator, istream, and string fixes and 36730 libstdc++ testsuite entries. Also for providing the patch to G77 36731 to add rudimentary support for `INTEGER*1', `INTEGER*2', and 36732 `LOGICAL*1'. 36733 36734 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique. 36735 36736 * Richard Stallman, for writing the original GCC and launching the 36737 GNU project. 36738 36739 * Jan Stein of the Chalmers Computer Society provided support for 36740 Genix, as well as part of the 32000 machine description. 36741 36742 * Nigel Stephens for various mips16 related fixes/improvements. 36743 36744 * Jonathan Stone wrote the machine description for the Pyramid 36745 computer. 36746 36747 * Graham Stott for various infrastructure improvements. 36748 36749 * John Stracke for his Java HTTP protocol fixes. 36750 36751 * Mike Stump for his Elxsi port, G++ contributions over the years 36752 and more recently his vxworks contributions 36753 36754 * Jeff Sturm for Java porting help, bug fixes, and encouragement. 36755 36756 * Shigeya Suzuki for this fixes for the bsdi platforms. 36757 36758 * Ian Lance Taylor for his mips16 work, general configury hacking, 36759 fixincludes, etc. 36760 36761 * Holger Teutsch provided the support for the Clipper CPU. 36762 36763 * Gary Thomas for his ongoing work to make the PPC work for 36764 GNU/Linux. 36765 36766 * Philipp Thomas for random bug fixes throughout the compiler 36767 36768 * Jason Thorpe for thread support in libstdc++ on NetBSD. 36769 36770 * Kresten Krab Thorup wrote the run time support for the Objective-C 36771 language and the fantastic Java bytecode interpreter. 36772 36773 * Michael Tiemann for random bug fixes, the first instruction 36774 scheduler, initial C++ support, function integration, NS32k, SPARC 36775 and M88k machine description work, delay slot scheduling. 36776 36777 * Andreas Tobler for his work porting libgcj to Darwin. 36778 36779 * Teemu Torma for thread safe exception handling support. 36780 36781 * Leonard Tower wrote parts of the parser, RTL generator, and RTL 36782 definitions, and of the VAX machine description. 36783 36784 * Daniel Towner and Hariharan Sandanagobalane contributed and 36785 maintain the picoChip port. 36786 36787 * Tom Tromey for internationalization support and for his many Java 36788 contributions and libgcj maintainership. 36789 36790 * Lassi Tuura for improvements to config.guess to determine HP 36791 processor types. 36792 36793 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes. 36794 36795 * Andy Vaught for the design and initial implementation of the GNU 36796 Fortran front end. 36797 36798 * Brent Verner for work with the libstdc++ cshadow files and their 36799 associated configure steps. 36800 36801 * Todd Vierling for contributions for NetBSD ports. 36802 36803 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML 36804 guidance. 36805 36806 * Dean Wakerley for converting the install documentation from HTML 36807 to texinfo in time for GCC 3.0. 36808 36809 * Krister Walfridsson for random bug fixes. 36810 36811 * Feng Wang for contributions to GNU Fortran. 36812 36813 * Stephen M. Webb for time and effort on making libstdc++ shadow 36814 files work with the tricky Solaris 8+ headers, and for pushing the 36815 build-time header tree. 36816 36817 * John Wehle for various improvements for the x86 code generator, 36818 related infrastructure improvements to help x86 code generation, 36819 value range propagation and other work, WE32k port. 36820 36821 * Ulrich Weigand for work on the s390 port. 36822 36823 * Zack Weinberg for major work on cpplib and various other bug fixes. 36824 36825 * Matt Welsh for help with Linux Threads support in GCJ. 36826 36827 * Urban Widmark for help fixing java.io. 36828 36829 * Mark Wielaard for new Java library code and his work integrating 36830 with Classpath. 36831 36832 * Dale Wiles helped port GCC to the Tahoe. 36833 36834 * Bob Wilson from Tensilica, Inc. for the Xtensa port. 36835 36836 * Jim Wilson for his direction via the steering committee, tackling 36837 hard problems in various places that nobody else wanted to work 36838 on, strength reduction and other loop optimizations. 36839 36840 * Paul Woegerer and Tal Agmon for the CRX port. 36841 36842 * Carlo Wood for various fixes. 36843 36844 * Tom Wood for work on the m88k port. 36845 36846 * Canqun Yang for work on GNU Fortran. 36847 36848 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine 36849 description for the Tron architecture (specifically, the Gmicro). 36850 36851 * Kevin Zachmann helped port GCC to the Tahoe. 36852 36853 * Ayal Zaks for Swing Modulo Scheduling (SMS). 36854 36855 * Xiaoqiang Zhang for work on GNU Fortran. 36856 36857 * Gilles Zunino for help porting Java to Irix. 36858 36859 36860 The following people are recognized for their contributions to GNAT, 36861 the Ada front end of GCC: 36862 * Bernard Banner 36863 36864 * Romain Berrendonner 36865 36866 * Geert Bosch 36867 36868 * Emmanuel Briot 36869 36870 * Joel Brobecker 36871 36872 * Ben Brosgol 36873 36874 * Vincent Celier 36875 36876 * Arnaud Charlet 36877 36878 * Chien Chieng 36879 36880 * Cyrille Comar 36881 36882 * Cyrille Crozes 36883 36884 * Robert Dewar 36885 36886 * Gary Dismukes 36887 36888 * Robert Duff 36889 36890 * Ed Falis 36891 36892 * Ramon Fernandez 36893 36894 * Sam Figueroa 36895 36896 * Vasiliy Fofanov 36897 36898 * Michael Friess 36899 36900 * Franco Gasperoni 36901 36902 * Ted Giering 36903 36904 * Matthew Gingell 36905 36906 * Laurent Guerby 36907 36908 * Jerome Guitton 36909 36910 * Olivier Hainque 36911 36912 * Jerome Hugues 36913 36914 * Hristian Kirtchev 36915 36916 * Jerome Lambourg 36917 36918 * Bruno Leclerc 36919 36920 * Albert Lee 36921 36922 * Sean McNeil 36923 36924 * Javier Miranda 36925 36926 * Laurent Nana 36927 36928 * Pascal Obry 36929 36930 * Dong-Ik Oh 36931 36932 * Laurent Pautet 36933 36934 * Brett Porter 36935 36936 * Thomas Quinot 36937 36938 * Nicolas Roche 36939 36940 * Pat Rogers 36941 36942 * Jose Ruiz 36943 36944 * Douglas Rupp 36945 36946 * Sergey Rybin 36947 36948 * Gail Schenker 36949 36950 * Ed Schonberg 36951 36952 * Nicolas Setton 36953 36954 * Samuel Tardieu 36955 36956 36957 The following people are recognized for their contributions of new 36958 features, bug reports, testing and integration of classpath/libgcj for 36959 GCC version 4.1: 36960 * Lillian Angel for `JTree' implementation and lots Free Swing 36961 additions and bug fixes. 36962 36963 * Wolfgang Baer for `GapContent' bug fixes. 36964 36965 * Anthony Balkissoon for `JList', Free Swing 1.5 updates and mouse 36966 event fixes, lots of Free Swing work including `JTable' editing. 36967 36968 * Stuart Ballard for RMI constant fixes. 36969 36970 * Goffredo Baroncelli for `HTTPURLConnection' fixes. 36971 36972 * Gary Benson for `MessageFormat' fixes. 36973 36974 * Daniel Bonniot for `Serialization' fixes. 36975 36976 * Chris Burdess for lots of gnu.xml and http protocol fixes, `StAX' 36977 and `DOM xml:id' support. 36978 36979 * Ka-Hing Cheung for `TreePath' and `TreeSelection' fixes. 36980 36981 * Archie Cobbs for build fixes, VM interface updates, 36982 `URLClassLoader' updates. 36983 36984 * Kelley Cook for build fixes. 36985 36986 * Martin Cordova for Suggestions for better `SocketTimeoutException'. 36987 36988 * David Daney for `BitSet' bug fixes, `HttpURLConnection' rewrite 36989 and improvements. 36990 36991 * Thomas Fitzsimmons for lots of upgrades to the gtk+ AWT and Cairo 36992 2D support. Lots of imageio framework additions, lots of AWT and 36993 Free Swing bug fixes. 36994 36995 * Jeroen Frijters for `ClassLoader' and nio cleanups, serialization 36996 fixes, better `Proxy' support, bug fixes and IKVM integration. 36997 36998 * Santiago Gala for `AccessControlContext' fixes. 36999 37000 * Nicolas Geoffray for `VMClassLoader' and `AccessController' 37001 improvements. 37002 37003 * David Gilbert for `basic' and `metal' icon and plaf support and 37004 lots of documenting, Lots of Free Swing and metal theme additions. 37005 `MetalIconFactory' implementation. 37006 37007 * Anthony Green for `MIDI' framework, `ALSA' and `DSSI' providers. 37008 37009 * Andrew Haley for `Serialization' and `URLClassLoader' fixes, gcj 37010 build speedups. 37011 37012 * Kim Ho for `JFileChooser' implementation. 37013 37014 * Andrew John Hughes for `Locale' and net fixes, URI RFC2986 37015 updates, `Serialization' fixes, `Properties' XML support and 37016 generic branch work, VMIntegration guide update. 37017 37018 * Bastiaan Huisman for `TimeZone' bug fixing. 37019 37020 * Andreas Jaeger for mprec updates. 37021 37022 * Paul Jenner for better `-Werror' support. 37023 37024 * Ito Kazumitsu for `NetworkInterface' implementation and updates. 37025 37026 * Roman Kennke for `BoxLayout', `GrayFilter' and `SplitPane', plus 37027 bug fixes all over. Lots of Free Swing work including styled text. 37028 37029 * Simon Kitching for `String' cleanups and optimization suggestions. 37030 37031 * Michael Koch for configuration fixes, `Locale' updates, bug and 37032 build fixes. 37033 37034 * Guilhem Lavaux for configuration, thread and channel fixes and 37035 Kaffe integration. JCL native `Pointer' updates. Logger bug fixes. 37036 37037 * David Lichteblau for JCL support library global/local reference 37038 cleanups. 37039 37040 * Aaron Luchko for JDWP updates and documentation fixes. 37041 37042 * Ziga Mahkovec for `Graphics2D' upgraded to Cairo 0.5 and new regex 37043 features. 37044 37045 * Sven de Marothy for BMP imageio support, CSS and `TextLayout' 37046 fixes. `GtkImage' rewrite, 2D, awt, free swing and date/time fixes 37047 and implementing the Qt4 peers. 37048 37049 * Casey Marshall for crypto algorithm fixes, `FileChannel' lock, 37050 `SystemLogger' and `FileHandler' rotate implementations, NIO 37051 `FileChannel.map' support, security and policy updates. 37052 37053 * Bryce McKinlay for RMI work. 37054 37055 * Audrius Meskauskas for lots of Free Corba, RMI and HTML work plus 37056 testing and documenting. 37057 37058 * Kalle Olavi Niemitalo for build fixes. 37059 37060 * Rainer Orth for build fixes. 37061 37062 * Andrew Overholt for `File' locking fixes. 37063 37064 * Ingo Proetel for `Image', `Logger' and `URLClassLoader' updates. 37065 37066 * Olga Rodimina for `MenuSelectionManager' implementation. 37067 37068 * Jan Roehrich for `BasicTreeUI' and `JTree' fixes. 37069 37070 * Julian Scheid for documentation updates and gjdoc support. 37071 37072 * Christian Schlichtherle for zip fixes and cleanups. 37073 37074 * Robert Schuster for documentation updates and beans fixes, 37075 `TreeNode' enumerations and `ActionCommand' and various fixes, XML 37076 and URL, AWT and Free Swing bug fixes. 37077 37078 * Keith Seitz for lots of JDWP work. 37079 37080 * Christian Thalinger for 64-bit cleanups, Configuration and VM 37081 interface fixes and `CACAO' integration, `fdlibm' updates. 37082 37083 * Gael Thomas for `VMClassLoader' boot packages support suggestions. 37084 37085 * Andreas Tobler for Darwin and Solaris testing and fixing, `Qt4' 37086 support for Darwin/OS X, `Graphics2D' support, `gtk+' updates. 37087 37088 * Dalibor Topic for better `DEBUG' support, build cleanups and Kaffe 37089 integration. `Qt4' build infrastructure, `SHA1PRNG' and 37090 `GdkPixbugDecoder' updates. 37091 37092 * Tom Tromey for Eclipse integration, generics work, lots of bug 37093 fixes and gcj integration including coordinating The Big Merge. 37094 37095 * Mark Wielaard for bug fixes, packaging and release management, 37096 `Clipboard' implementation, system call interrupts and network 37097 timeouts and `GdkPixpufDecoder' fixes. 37098 37099 37100 In addition to the above, all of which also contributed time and 37101 energy in testing GCC, we would like to thank the following for their 37102 contributions to testing: 37103 37104 * Michael Abd-El-Malek 37105 37106 * Thomas Arend 37107 37108 * Bonzo Armstrong 37109 37110 * Steven Ashe 37111 37112 * Chris Baldwin 37113 37114 * David Billinghurst 37115 37116 * Jim Blandy 37117 37118 * Stephane Bortzmeyer 37119 37120 * Horst von Brand 37121 37122 * Frank Braun 37123 37124 * Rodney Brown 37125 37126 * Sidney Cadot 37127 37128 * Bradford Castalia 37129 37130 * Robert Clark 37131 37132 * Jonathan Corbet 37133 37134 * Ralph Doncaster 37135 37136 * Richard Emberson 37137 37138 * Levente Farkas 37139 37140 * Graham Fawcett 37141 37142 * Mark Fernyhough 37143 37144 * Robert A. French 37145 37146 * Jo"rgen Freyh 37147 37148 * Mark K. Gardner 37149 37150 * Charles-Antoine Gauthier 37151 37152 * Yung Shing Gene 37153 37154 * David Gilbert 37155 37156 * Simon Gornall 37157 37158 * Fred Gray 37159 37160 * John Griffin 37161 37162 * Patrik Hagglund 37163 37164 * Phil Hargett 37165 37166 * Amancio Hasty 37167 37168 * Takafumi Hayashi 37169 37170 * Bryan W. Headley 37171 37172 * Kevin B. Hendricks 37173 37174 * Joep Jansen 37175 37176 * Christian Joensson 37177 37178 * Michel Kern 37179 37180 * David Kidd 37181 37182 * Tobias Kuipers 37183 37184 * Anand Krishnaswamy 37185 37186 * A. O. V. Le Blanc 37187 37188 * llewelly 37189 37190 * Damon Love 37191 37192 * Brad Lucier 37193 37194 * Matthias Klose 37195 37196 * Martin Knoblauch 37197 37198 * Rick Lutowski 37199 37200 * Jesse Macnish 37201 37202 * Stefan Morrell 37203 37204 * Anon A. Mous 37205 37206 * Matthias Mueller 37207 37208 * Pekka Nikander 37209 37210 * Rick Niles 37211 37212 * Jon Olson 37213 37214 * Magnus Persson 37215 37216 * Chris Pollard 37217 37218 * Richard Polton 37219 37220 * Derk Reefman 37221 37222 * David Rees 37223 37224 * Paul Reilly 37225 37226 * Tom Reilly 37227 37228 * Torsten Rueger 37229 37230 * Danny Sadinoff 37231 37232 * Marc Schifer 37233 37234 * Erik Schnetter 37235 37236 * Wayne K. Schroll 37237 37238 * David Schuler 37239 37240 * Vin Shelton 37241 37242 * Tim Souder 37243 37244 * Adam Sulmicki 37245 37246 * Bill Thorson 37247 37248 * George Talbot 37249 37250 * Pedro A. M. Vazquez 37251 37252 * Gregory Warnes 37253 37254 * Ian Watson 37255 37256 * David E. Young 37257 37258 * And many others 37259 37260 And finally we'd like to thank everyone who uses the compiler, provides 37261 feedback and generally reminds us why we're doing this work in the first 37262 place. 37263 37264 37265 File: gccint.info, Node: Option Index, Next: Concept Index, Prev: Contributors, Up: Top 37266 37267 Option Index 37268 ************ 37269 37270 GCC's command line options are indexed here without any initial `-' or 37271 `--'. Where an option has both positive and negative forms (such as 37272 `-fOPTION' and `-fno-OPTION'), relevant entries in the manual are 37273 indexed under the most appropriate form; it may sometimes be useful to 37274 look up both forms. 37275 37276 [index] 37277 * Menu: 37278 37279 * msoft-float: Soft float library routines. 37280 (line 6) 37281 37282 37283 File: gccint.info, Node: Concept Index, Prev: Option Index, Up: Top 37284 37285 Concept Index 37286 ************* 37287 37288 [index] 37289 * Menu: 37290 37291 * ! in constraint: Multi-Alternative. (line 47) 37292 * # in constraint: Modifiers. (line 67) 37293 * # in template: Output Template. (line 66) 37294 * #pragma: Misc. (line 381) 37295 * % in constraint: Modifiers. (line 45) 37296 * % in GTY option: GTY Options. (line 18) 37297 * % in template: Output Template. (line 6) 37298 * & in constraint: Modifiers. (line 25) 37299 * ( <1>: Sections. (line 160) 37300 * ( <2>: GIMPLE_CALL. (line 63) 37301 * ( <3>: GIMPLE_ASM. (line 21) 37302 * (: Logical Operators. (line 107) 37303 * (nil): RTL Objects. (line 73) 37304 * * <1>: Host Common. (line 17) 37305 * *: Scheduling. (line 246) 37306 * * in constraint: Modifiers. (line 72) 37307 * * in template: Output Statement. (line 29) 37308 * *gimple_assign_lhs_ptr: GIMPLE_ASSIGN. (line 54) 37309 * *gimple_assign_rhs1_ptr: GIMPLE_ASSIGN. (line 60) 37310 * *gimple_assign_rhs2_ptr: GIMPLE_ASSIGN. (line 67) 37311 * *gimple_call_arg_ptr: GIMPLE_CALL. (line 71) 37312 * *gimple_call_lhs_ptr: GIMPLE_CALL. (line 32) 37313 * *gimple_catch_types_ptr: GIMPLE_CATCH. (line 16) 37314 * *gimple_cdt_location_ptr: GIMPLE_CHANGE_DYNAMIC_TYPE. 37315 (line 28) 37316 * *gimple_cdt_new_type_ptr: GIMPLE_CHANGE_DYNAMIC_TYPE. 37317 (line 15) 37318 * *gimple_eh_filter_types_ptr: GIMPLE_EH_FILTER. (line 15) 37319 * *gimple_omp_critical_name_ptr: GIMPLE_OMP_CRITICAL. 37320 (line 16) 37321 * *gimple_omp_for_clauses_ptr: GIMPLE_OMP_FOR. (line 23) 37322 * *gimple_omp_for_final_ptr: GIMPLE_OMP_FOR. (line 54) 37323 * *gimple_omp_for_incr_ptr: GIMPLE_OMP_FOR. (line 64) 37324 * *gimple_omp_for_index_ptr: GIMPLE_OMP_FOR. (line 34) 37325 * *gimple_omp_for_initial_ptr: GIMPLE_OMP_FOR. (line 44) 37326 * *gimple_omp_parallel_child_fn_ptr: GIMPLE_OMP_PARALLEL. 37327 (line 46) 37328 * *gimple_omp_parallel_clauses_ptr: GIMPLE_OMP_PARALLEL. 37329 (line 34) 37330 * *gimple_omp_parallel_data_arg_ptr: GIMPLE_OMP_PARALLEL. 37331 (line 58) 37332 * *gimple_omp_sections_clauses_ptr: GIMPLE_OMP_SECTIONS. 37333 (line 33) 37334 * *gimple_omp_sections_control_ptr: GIMPLE_OMP_SECTIONS. 37335 (line 21) 37336 * *gimple_omp_single_clauses_ptr: GIMPLE_OMP_SINGLE. (line 17) 37337 * *gimple_op_ptr: Manipulating GIMPLE statements. 37338 (line 84) 37339 * *gimple_ops <1>: Manipulating GIMPLE statements. 37340 (line 78) 37341 * *gimple_ops: Logical Operators. (line 82) 37342 * *gimple_phi_result_ptr: GIMPLE_PHI. (line 22) 37343 * *gsi_stmt_ptr: Sequence iterators. (line 80) 37344 * *TARGET_GET_PCH_VALIDITY: PCH Target. (line 7) 37345 * + in constraint: Modifiers. (line 12) 37346 * -fsection-anchors <1>: Anchored Addresses. (line 6) 37347 * -fsection-anchors: Special Accessors. (line 106) 37348 * /c in RTL dump: Flags. (line 234) 37349 * /f in RTL dump: Flags. (line 242) 37350 * /i in RTL dump: Flags. (line 294) 37351 * /j in RTL dump: Flags. (line 309) 37352 * /s in RTL dump: Flags. (line 258) 37353 * /u in RTL dump: Flags. (line 319) 37354 * /v in RTL dump: Flags. (line 351) 37355 * 0 in constraint: Simple Constraints. (line 120) 37356 * < in constraint: Simple Constraints. (line 48) 37357 * = in constraint: Modifiers. (line 8) 37358 * > in constraint: Simple Constraints. (line 52) 37359 * ? in constraint: Multi-Alternative. (line 41) 37360 * \: Output Template. (line 46) 37361 * __absvdi2: Integer library routines. 37362 (line 107) 37363 * __absvsi2: Integer library routines. 37364 (line 106) 37365 * __addda3: Fixed-point fractional library routines. 37366 (line 45) 37367 * __adddf3: Soft float library routines. 37368 (line 23) 37369 * __adddq3: Fixed-point fractional library routines. 37370 (line 33) 37371 * __addha3: Fixed-point fractional library routines. 37372 (line 43) 37373 * __addhq3: Fixed-point fractional library routines. 37374 (line 30) 37375 * __addqq3: Fixed-point fractional library routines. 37376 (line 29) 37377 * __addsa3: Fixed-point fractional library routines. 37378 (line 44) 37379 * __addsf3: Soft float library routines. 37380 (line 22) 37381 * __addsq3: Fixed-point fractional library routines. 37382 (line 31) 37383 * __addta3: Fixed-point fractional library routines. 37384 (line 47) 37385 * __addtf3: Soft float library routines. 37386 (line 25) 37387 * __adduda3: Fixed-point fractional library routines. 37388 (line 53) 37389 * __addudq3: Fixed-point fractional library routines. 37390 (line 41) 37391 * __adduha3: Fixed-point fractional library routines. 37392 (line 49) 37393 * __adduhq3: Fixed-point fractional library routines. 37394 (line 37) 37395 * __adduqq3: Fixed-point fractional library routines. 37396 (line 35) 37397 * __addusa3: Fixed-point fractional library routines. 37398 (line 51) 37399 * __addusq3: Fixed-point fractional library routines. 37400 (line 39) 37401 * __adduta3: Fixed-point fractional library routines. 37402 (line 55) 37403 * __addvdi3: Integer library routines. 37404 (line 111) 37405 * __addvsi3: Integer library routines. 37406 (line 110) 37407 * __addxf3: Soft float library routines. 37408 (line 27) 37409 * __ashlda3: Fixed-point fractional library routines. 37410 (line 351) 37411 * __ashldi3: Integer library routines. 37412 (line 14) 37413 * __ashldq3: Fixed-point fractional library routines. 37414 (line 340) 37415 * __ashlha3: Fixed-point fractional library routines. 37416 (line 349) 37417 * __ashlhq3: Fixed-point fractional library routines. 37418 (line 337) 37419 * __ashlqq3: Fixed-point fractional library routines. 37420 (line 336) 37421 * __ashlsa3: Fixed-point fractional library routines. 37422 (line 350) 37423 * __ashlsi3: Integer library routines. 37424 (line 13) 37425 * __ashlsq3: Fixed-point fractional library routines. 37426 (line 338) 37427 * __ashlta3: Fixed-point fractional library routines. 37428 (line 353) 37429 * __ashlti3: Integer library routines. 37430 (line 15) 37431 * __ashluda3: Fixed-point fractional library routines. 37432 (line 359) 37433 * __ashludq3: Fixed-point fractional library routines. 37434 (line 348) 37435 * __ashluha3: Fixed-point fractional library routines. 37436 (line 355) 37437 * __ashluhq3: Fixed-point fractional library routines. 37438 (line 344) 37439 * __ashluqq3: Fixed-point fractional library routines. 37440 (line 342) 37441 * __ashlusa3: Fixed-point fractional library routines. 37442 (line 357) 37443 * __ashlusq3: Fixed-point fractional library routines. 37444 (line 346) 37445 * __ashluta3: Fixed-point fractional library routines. 37446 (line 361) 37447 * __ashrda3: Fixed-point fractional library routines. 37448 (line 371) 37449 * __ashrdi3: Integer library routines. 37450 (line 19) 37451 * __ashrdq3: Fixed-point fractional library routines. 37452 (line 368) 37453 * __ashrha3: Fixed-point fractional library routines. 37454 (line 369) 37455 * __ashrhq3: Fixed-point fractional library routines. 37456 (line 365) 37457 * __ashrqq3: Fixed-point fractional library routines. 37458 (line 364) 37459 * __ashrsa3: Fixed-point fractional library routines. 37460 (line 370) 37461 * __ashrsi3: Integer library routines. 37462 (line 18) 37463 * __ashrsq3: Fixed-point fractional library routines. 37464 (line 366) 37465 * __ashrta3: Fixed-point fractional library routines. 37466 (line 373) 37467 * __ashrti3: Integer library routines. 37468 (line 20) 37469 * __bid_adddd3: Decimal float library routines. 37470 (line 25) 37471 * __bid_addsd3: Decimal float library routines. 37472 (line 21) 37473 * __bid_addtd3: Decimal float library routines. 37474 (line 29) 37475 * __bid_divdd3: Decimal float library routines. 37476 (line 68) 37477 * __bid_divsd3: Decimal float library routines. 37478 (line 64) 37479 * __bid_divtd3: Decimal float library routines. 37480 (line 72) 37481 * __bid_eqdd2: Decimal float library routines. 37482 (line 259) 37483 * __bid_eqsd2: Decimal float library routines. 37484 (line 257) 37485 * __bid_eqtd2: Decimal float library routines. 37486 (line 261) 37487 * __bid_extendddtd2: Decimal float library routines. 37488 (line 92) 37489 * __bid_extendddtf: Decimal float library routines. 37490 (line 140) 37491 * __bid_extendddxf: Decimal float library routines. 37492 (line 134) 37493 * __bid_extenddfdd: Decimal float library routines. 37494 (line 147) 37495 * __bid_extenddftd: Decimal float library routines. 37496 (line 107) 37497 * __bid_extendsddd2: Decimal float library routines. 37498 (line 88) 37499 * __bid_extendsddf: Decimal float library routines. 37500 (line 128) 37501 * __bid_extendsdtd2: Decimal float library routines. 37502 (line 90) 37503 * __bid_extendsdtf: Decimal float library routines. 37504 (line 138) 37505 * __bid_extendsdxf: Decimal float library routines. 37506 (line 132) 37507 * __bid_extendsfdd: Decimal float library routines. 37508 (line 103) 37509 * __bid_extendsfsd: Decimal float library routines. 37510 (line 145) 37511 * __bid_extendsftd: Decimal float library routines. 37512 (line 105) 37513 * __bid_extendtftd: Decimal float library routines. 37514 (line 149) 37515 * __bid_extendxftd: Decimal float library routines. 37516 (line 109) 37517 * __bid_fixdddi: Decimal float library routines. 37518 (line 170) 37519 * __bid_fixddsi: Decimal float library routines. 37520 (line 162) 37521 * __bid_fixsddi: Decimal float library routines. 37522 (line 168) 37523 * __bid_fixsdsi: Decimal float library routines. 37524 (line 160) 37525 * __bid_fixtddi: Decimal float library routines. 37526 (line 172) 37527 * __bid_fixtdsi: Decimal float library routines. 37528 (line 164) 37529 * __bid_fixunsdddi: Decimal float library routines. 37530 (line 187) 37531 * __bid_fixunsddsi: Decimal float library routines. 37532 (line 178) 37533 * __bid_fixunssddi: Decimal float library routines. 37534 (line 185) 37535 * __bid_fixunssdsi: Decimal float library routines. 37536 (line 176) 37537 * __bid_fixunstddi: Decimal float library routines. 37538 (line 189) 37539 * __bid_fixunstdsi: Decimal float library routines. 37540 (line 180) 37541 * __bid_floatdidd: Decimal float library routines. 37542 (line 205) 37543 * __bid_floatdisd: Decimal float library routines. 37544 (line 203) 37545 * __bid_floatditd: Decimal float library routines. 37546 (line 207) 37547 * __bid_floatsidd: Decimal float library routines. 37548 (line 196) 37549 * __bid_floatsisd: Decimal float library routines. 37550 (line 194) 37551 * __bid_floatsitd: Decimal float library routines. 37552 (line 198) 37553 * __bid_floatunsdidd: Decimal float library routines. 37554 (line 223) 37555 * __bid_floatunsdisd: Decimal float library routines. 37556 (line 221) 37557 * __bid_floatunsditd: Decimal float library routines. 37558 (line 225) 37559 * __bid_floatunssidd: Decimal float library routines. 37560 (line 214) 37561 * __bid_floatunssisd: Decimal float library routines. 37562 (line 212) 37563 * __bid_floatunssitd: Decimal float library routines. 37564 (line 216) 37565 * __bid_gedd2: Decimal float library routines. 37566 (line 277) 37567 * __bid_gesd2: Decimal float library routines. 37568 (line 275) 37569 * __bid_getd2: Decimal float library routines. 37570 (line 279) 37571 * __bid_gtdd2: Decimal float library routines. 37572 (line 304) 37573 * __bid_gtsd2: Decimal float library routines. 37574 (line 302) 37575 * __bid_gttd2: Decimal float library routines. 37576 (line 306) 37577 * __bid_ledd2: Decimal float library routines. 37578 (line 295) 37579 * __bid_lesd2: Decimal float library routines. 37580 (line 293) 37581 * __bid_letd2: Decimal float library routines. 37582 (line 297) 37583 * __bid_ltdd2: Decimal float library routines. 37584 (line 286) 37585 * __bid_ltsd2: Decimal float library routines. 37586 (line 284) 37587 * __bid_lttd2: Decimal float library routines. 37588 (line 288) 37589 * __bid_muldd3: Decimal float library routines. 37590 (line 54) 37591 * __bid_mulsd3: Decimal float library routines. 37592 (line 50) 37593 * __bid_multd3: Decimal float library routines. 37594 (line 58) 37595 * __bid_nedd2: Decimal float library routines. 37596 (line 268) 37597 * __bid_negdd2: Decimal float library routines. 37598 (line 78) 37599 * __bid_negsd2: Decimal float library routines. 37600 (line 76) 37601 * __bid_negtd2: Decimal float library routines. 37602 (line 80) 37603 * __bid_nesd2: Decimal float library routines. 37604 (line 266) 37605 * __bid_netd2: Decimal float library routines. 37606 (line 270) 37607 * __bid_subdd3: Decimal float library routines. 37608 (line 39) 37609 * __bid_subsd3: Decimal float library routines. 37610 (line 35) 37611 * __bid_subtd3: Decimal float library routines. 37612 (line 43) 37613 * __bid_truncdddf: Decimal float library routines. 37614 (line 153) 37615 * __bid_truncddsd2: Decimal float library routines. 37616 (line 94) 37617 * __bid_truncddsf: Decimal float library routines. 37618 (line 124) 37619 * __bid_truncdfsd: Decimal float library routines. 37620 (line 111) 37621 * __bid_truncsdsf: Decimal float library routines. 37622 (line 151) 37623 * __bid_trunctddd2: Decimal float library routines. 37624 (line 98) 37625 * __bid_trunctddf: Decimal float library routines. 37626 (line 130) 37627 * __bid_trunctdsd2: Decimal float library routines. 37628 (line 96) 37629 * __bid_trunctdsf: Decimal float library routines. 37630 (line 126) 37631 * __bid_trunctdtf: Decimal float library routines. 37632 (line 155) 37633 * __bid_trunctdxf: Decimal float library routines. 37634 (line 136) 37635 * __bid_trunctfdd: Decimal float library routines. 37636 (line 119) 37637 * __bid_trunctfsd: Decimal float library routines. 37638 (line 115) 37639 * __bid_truncxfdd: Decimal float library routines. 37640 (line 117) 37641 * __bid_truncxfsd: Decimal float library routines. 37642 (line 113) 37643 * __bid_unorddd2: Decimal float library routines. 37644 (line 235) 37645 * __bid_unordsd2: Decimal float library routines. 37646 (line 233) 37647 * __bid_unordtd2: Decimal float library routines. 37648 (line 237) 37649 * __bswapdi2: Integer library routines. 37650 (line 162) 37651 * __bswapsi2: Integer library routines. 37652 (line 161) 37653 * __builtin_args_info: Varargs. (line 42) 37654 * __builtin_classify_type: Varargs. (line 76) 37655 * __builtin_next_arg: Varargs. (line 66) 37656 * __builtin_saveregs: Varargs. (line 24) 37657 * __clear_cache: Miscellaneous routines. 37658 (line 10) 37659 * __clzdi2: Integer library routines. 37660 (line 131) 37661 * __clzsi2: Integer library routines. 37662 (line 130) 37663 * __clzti2: Integer library routines. 37664 (line 132) 37665 * __cmpda2: Fixed-point fractional library routines. 37666 (line 451) 37667 * __cmpdf2: Soft float library routines. 37668 (line 164) 37669 * __cmpdi2: Integer library routines. 37670 (line 87) 37671 * __cmpdq2: Fixed-point fractional library routines. 37672 (line 441) 37673 * __cmpha2: Fixed-point fractional library routines. 37674 (line 449) 37675 * __cmphq2: Fixed-point fractional library routines. 37676 (line 438) 37677 * __cmpqq2: Fixed-point fractional library routines. 37678 (line 437) 37679 * __cmpsa2: Fixed-point fractional library routines. 37680 (line 450) 37681 * __cmpsf2: Soft float library routines. 37682 (line 163) 37683 * __cmpsq2: Fixed-point fractional library routines. 37684 (line 439) 37685 * __cmpta2: Fixed-point fractional library routines. 37686 (line 453) 37687 * __cmptf2: Soft float library routines. 37688 (line 165) 37689 * __cmpti2: Integer library routines. 37690 (line 88) 37691 * __cmpuda2: Fixed-point fractional library routines. 37692 (line 458) 37693 * __cmpudq2: Fixed-point fractional library routines. 37694 (line 448) 37695 * __cmpuha2: Fixed-point fractional library routines. 37696 (line 455) 37697 * __cmpuhq2: Fixed-point fractional library routines. 37698 (line 444) 37699 * __cmpuqq2: Fixed-point fractional library routines. 37700 (line 443) 37701 * __cmpusa2: Fixed-point fractional library routines. 37702 (line 456) 37703 * __cmpusq2: Fixed-point fractional library routines. 37704 (line 446) 37705 * __cmputa2: Fixed-point fractional library routines. 37706 (line 460) 37707 * __CTOR_LIST__: Initialization. (line 25) 37708 * __ctzdi2: Integer library routines. 37709 (line 138) 37710 * __ctzsi2: Integer library routines. 37711 (line 137) 37712 * __ctzti2: Integer library routines. 37713 (line 139) 37714 * __divda3: Fixed-point fractional library routines. 37715 (line 227) 37716 * __divdc3: Soft float library routines. 37717 (line 252) 37718 * __divdf3: Soft float library routines. 37719 (line 48) 37720 * __divdi3: Integer library routines. 37721 (line 25) 37722 * __divdq3: Fixed-point fractional library routines. 37723 (line 223) 37724 * __divha3: Fixed-point fractional library routines. 37725 (line 225) 37726 * __divhq3: Fixed-point fractional library routines. 37727 (line 220) 37728 * __divqq3: Fixed-point fractional library routines. 37729 (line 219) 37730 * __divsa3: Fixed-point fractional library routines. 37731 (line 226) 37732 * __divsc3: Soft float library routines. 37733 (line 250) 37734 * __divsf3: Soft float library routines. 37735 (line 47) 37736 * __divsi3: Integer library routines. 37737 (line 24) 37738 * __divsq3: Fixed-point fractional library routines. 37739 (line 221) 37740 * __divta3: Fixed-point fractional library routines. 37741 (line 229) 37742 * __divtc3: Soft float library routines. 37743 (line 254) 37744 * __divtf3: Soft float library routines. 37745 (line 50) 37746 * __divti3: Integer library routines. 37747 (line 26) 37748 * __divxc3: Soft float library routines. 37749 (line 256) 37750 * __divxf3: Soft float library routines. 37751 (line 52) 37752 * __dpd_adddd3: Decimal float library routines. 37753 (line 23) 37754 * __dpd_addsd3: Decimal float library routines. 37755 (line 19) 37756 * __dpd_addtd3: Decimal float library routines. 37757 (line 27) 37758 * __dpd_divdd3: Decimal float library routines. 37759 (line 66) 37760 * __dpd_divsd3: Decimal float library routines. 37761 (line 62) 37762 * __dpd_divtd3: Decimal float library routines. 37763 (line 70) 37764 * __dpd_eqdd2: Decimal float library routines. 37765 (line 258) 37766 * __dpd_eqsd2: Decimal float library routines. 37767 (line 256) 37768 * __dpd_eqtd2: Decimal float library routines. 37769 (line 260) 37770 * __dpd_extendddtd2: Decimal float library routines. 37771 (line 91) 37772 * __dpd_extendddtf: Decimal float library routines. 37773 (line 139) 37774 * __dpd_extendddxf: Decimal float library routines. 37775 (line 133) 37776 * __dpd_extenddfdd: Decimal float library routines. 37777 (line 146) 37778 * __dpd_extenddftd: Decimal float library routines. 37779 (line 106) 37780 * __dpd_extendsddd2: Decimal float library routines. 37781 (line 87) 37782 * __dpd_extendsddf: Decimal float library routines. 37783 (line 127) 37784 * __dpd_extendsdtd2: Decimal float library routines. 37785 (line 89) 37786 * __dpd_extendsdtf: Decimal float library routines. 37787 (line 137) 37788 * __dpd_extendsdxf: Decimal float library routines. 37789 (line 131) 37790 * __dpd_extendsfdd: Decimal float library routines. 37791 (line 102) 37792 * __dpd_extendsfsd: Decimal float library routines. 37793 (line 144) 37794 * __dpd_extendsftd: Decimal float library routines. 37795 (line 104) 37796 * __dpd_extendtftd: Decimal float library routines. 37797 (line 148) 37798 * __dpd_extendxftd: Decimal float library routines. 37799 (line 108) 37800 * __dpd_fixdddi: Decimal float library routines. 37801 (line 169) 37802 * __dpd_fixddsi: Decimal float library routines. 37803 (line 161) 37804 * __dpd_fixsddi: Decimal float library routines. 37805 (line 167) 37806 * __dpd_fixsdsi: Decimal float library routines. 37807 (line 159) 37808 * __dpd_fixtddi: Decimal float library routines. 37809 (line 171) 37810 * __dpd_fixtdsi: Decimal float library routines. 37811 (line 163) 37812 * __dpd_fixunsdddi: Decimal float library routines. 37813 (line 186) 37814 * __dpd_fixunsddsi: Decimal float library routines. 37815 (line 177) 37816 * __dpd_fixunssddi: Decimal float library routines. 37817 (line 184) 37818 * __dpd_fixunssdsi: Decimal float library routines. 37819 (line 175) 37820 * __dpd_fixunstddi: Decimal float library routines. 37821 (line 188) 37822 * __dpd_fixunstdsi: Decimal float library routines. 37823 (line 179) 37824 * __dpd_floatdidd: Decimal float library routines. 37825 (line 204) 37826 * __dpd_floatdisd: Decimal float library routines. 37827 (line 202) 37828 * __dpd_floatditd: Decimal float library routines. 37829 (line 206) 37830 * __dpd_floatsidd: Decimal float library routines. 37831 (line 195) 37832 * __dpd_floatsisd: Decimal float library routines. 37833 (line 193) 37834 * __dpd_floatsitd: Decimal float library routines. 37835 (line 197) 37836 * __dpd_floatunsdidd: Decimal float library routines. 37837 (line 222) 37838 * __dpd_floatunsdisd: Decimal float library routines. 37839 (line 220) 37840 * __dpd_floatunsditd: Decimal float library routines. 37841 (line 224) 37842 * __dpd_floatunssidd: Decimal float library routines. 37843 (line 213) 37844 * __dpd_floatunssisd: Decimal float library routines. 37845 (line 211) 37846 * __dpd_floatunssitd: Decimal float library routines. 37847 (line 215) 37848 * __dpd_gedd2: Decimal float library routines. 37849 (line 276) 37850 * __dpd_gesd2: Decimal float library routines. 37851 (line 274) 37852 * __dpd_getd2: Decimal float library routines. 37853 (line 278) 37854 * __dpd_gtdd2: Decimal float library routines. 37855 (line 303) 37856 * __dpd_gtsd2: Decimal float library routines. 37857 (line 301) 37858 * __dpd_gttd2: Decimal float library routines. 37859 (line 305) 37860 * __dpd_ledd2: Decimal float library routines. 37861 (line 294) 37862 * __dpd_lesd2: Decimal float library routines. 37863 (line 292) 37864 * __dpd_letd2: Decimal float library routines. 37865 (line 296) 37866 * __dpd_ltdd2: Decimal float library routines. 37867 (line 285) 37868 * __dpd_ltsd2: Decimal float library routines. 37869 (line 283) 37870 * __dpd_lttd2: Decimal float library routines. 37871 (line 287) 37872 * __dpd_muldd3: Decimal float library routines. 37873 (line 52) 37874 * __dpd_mulsd3: Decimal float library routines. 37875 (line 48) 37876 * __dpd_multd3: Decimal float library routines. 37877 (line 56) 37878 * __dpd_nedd2: Decimal float library routines. 37879 (line 267) 37880 * __dpd_negdd2: Decimal float library routines. 37881 (line 77) 37882 * __dpd_negsd2: Decimal float library routines. 37883 (line 75) 37884 * __dpd_negtd2: Decimal float library routines. 37885 (line 79) 37886 * __dpd_nesd2: Decimal float library routines. 37887 (line 265) 37888 * __dpd_netd2: Decimal float library routines. 37889 (line 269) 37890 * __dpd_subdd3: Decimal float library routines. 37891 (line 37) 37892 * __dpd_subsd3: Decimal float library routines. 37893 (line 33) 37894 * __dpd_subtd3: Decimal float library routines. 37895 (line 41) 37896 * __dpd_truncdddf: Decimal float library routines. 37897 (line 152) 37898 * __dpd_truncddsd2: Decimal float library routines. 37899 (line 93) 37900 * __dpd_truncddsf: Decimal float library routines. 37901 (line 123) 37902 * __dpd_truncdfsd: Decimal float library routines. 37903 (line 110) 37904 * __dpd_truncsdsf: Decimal float library routines. 37905 (line 150) 37906 * __dpd_trunctddd2: Decimal float library routines. 37907 (line 97) 37908 * __dpd_trunctddf: Decimal float library routines. 37909 (line 129) 37910 * __dpd_trunctdsd2: Decimal float library routines. 37911 (line 95) 37912 * __dpd_trunctdsf: Decimal float library routines. 37913 (line 125) 37914 * __dpd_trunctdtf: Decimal float library routines. 37915 (line 154) 37916 * __dpd_trunctdxf: Decimal float library routines. 37917 (line 135) 37918 * __dpd_trunctfdd: Decimal float library routines. 37919 (line 118) 37920 * __dpd_trunctfsd: Decimal float library routines. 37921 (line 114) 37922 * __dpd_truncxfdd: Decimal float library routines. 37923 (line 116) 37924 * __dpd_truncxfsd: Decimal float library routines. 37925 (line 112) 37926 * __dpd_unorddd2: Decimal float library routines. 37927 (line 234) 37928 * __dpd_unordsd2: Decimal float library routines. 37929 (line 232) 37930 * __dpd_unordtd2: Decimal float library routines. 37931 (line 236) 37932 * __DTOR_LIST__: Initialization. (line 25) 37933 * __eqdf2: Soft float library routines. 37934 (line 194) 37935 * __eqsf2: Soft float library routines. 37936 (line 193) 37937 * __eqtf2: Soft float library routines. 37938 (line 195) 37939 * __extenddftf2: Soft float library routines. 37940 (line 68) 37941 * __extenddfxf2: Soft float library routines. 37942 (line 69) 37943 * __extendsfdf2: Soft float library routines. 37944 (line 65) 37945 * __extendsftf2: Soft float library routines. 37946 (line 66) 37947 * __extendsfxf2: Soft float library routines. 37948 (line 67) 37949 * __ffsdi2: Integer library routines. 37950 (line 144) 37951 * __ffsti2: Integer library routines. 37952 (line 145) 37953 * __fixdfdi: Soft float library routines. 37954 (line 88) 37955 * __fixdfsi: Soft float library routines. 37956 (line 81) 37957 * __fixdfti: Soft float library routines. 37958 (line 94) 37959 * __fixsfdi: Soft float library routines. 37960 (line 87) 37961 * __fixsfsi: Soft float library routines. 37962 (line 80) 37963 * __fixsfti: Soft float library routines. 37964 (line 93) 37965 * __fixtfdi: Soft float library routines. 37966 (line 89) 37967 * __fixtfsi: Soft float library routines. 37968 (line 82) 37969 * __fixtfti: Soft float library routines. 37970 (line 95) 37971 * __fixunsdfdi: Soft float library routines. 37972 (line 108) 37973 * __fixunsdfsi: Soft float library routines. 37974 (line 101) 37975 * __fixunsdfti: Soft float library routines. 37976 (line 115) 37977 * __fixunssfdi: Soft float library routines. 37978 (line 107) 37979 * __fixunssfsi: Soft float library routines. 37980 (line 100) 37981 * __fixunssfti: Soft float library routines. 37982 (line 114) 37983 * __fixunstfdi: Soft float library routines. 37984 (line 109) 37985 * __fixunstfsi: Soft float library routines. 37986 (line 102) 37987 * __fixunstfti: Soft float library routines. 37988 (line 116) 37989 * __fixunsxfdi: Soft float library routines. 37990 (line 110) 37991 * __fixunsxfsi: Soft float library routines. 37992 (line 103) 37993 * __fixunsxfti: Soft float library routines. 37994 (line 117) 37995 * __fixxfdi: Soft float library routines. 37996 (line 90) 37997 * __fixxfsi: Soft float library routines. 37998 (line 83) 37999 * __fixxfti: Soft float library routines. 38000 (line 96) 38001 * __floatdidf: Soft float library routines. 38002 (line 128) 38003 * __floatdisf: Soft float library routines. 38004 (line 127) 38005 * __floatditf: Soft float library routines. 38006 (line 129) 38007 * __floatdixf: Soft float library routines. 38008 (line 130) 38009 * __floatsidf: Soft float library routines. 38010 (line 122) 38011 * __floatsisf: Soft float library routines. 38012 (line 121) 38013 * __floatsitf: Soft float library routines. 38014 (line 123) 38015 * __floatsixf: Soft float library routines. 38016 (line 124) 38017 * __floattidf: Soft float library routines. 38018 (line 134) 38019 * __floattisf: Soft float library routines. 38020 (line 133) 38021 * __floattitf: Soft float library routines. 38022 (line 135) 38023 * __floattixf: Soft float library routines. 38024 (line 136) 38025 * __floatundidf: Soft float library routines. 38026 (line 146) 38027 * __floatundisf: Soft float library routines. 38028 (line 145) 38029 * __floatunditf: Soft float library routines. 38030 (line 147) 38031 * __floatundixf: Soft float library routines. 38032 (line 148) 38033 * __floatunsidf: Soft float library routines. 38034 (line 140) 38035 * __floatunsisf: Soft float library routines. 38036 (line 139) 38037 * __floatunsitf: Soft float library routines. 38038 (line 141) 38039 * __floatunsixf: Soft float library routines. 38040 (line 142) 38041 * __floatuntidf: Soft float library routines. 38042 (line 152) 38043 * __floatuntisf: Soft float library routines. 38044 (line 151) 38045 * __floatuntitf: Soft float library routines. 38046 (line 153) 38047 * __floatuntixf: Soft float library routines. 38048 (line 154) 38049 * __fractdadf: Fixed-point fractional library routines. 38050 (line 636) 38051 * __fractdadi: Fixed-point fractional library routines. 38052 (line 633) 38053 * __fractdadq: Fixed-point fractional library routines. 38054 (line 616) 38055 * __fractdaha2: Fixed-point fractional library routines. 38056 (line 617) 38057 * __fractdahi: Fixed-point fractional library routines. 38058 (line 631) 38059 * __fractdahq: Fixed-point fractional library routines. 38060 (line 614) 38061 * __fractdaqi: Fixed-point fractional library routines. 38062 (line 630) 38063 * __fractdaqq: Fixed-point fractional library routines. 38064 (line 613) 38065 * __fractdasa2: Fixed-point fractional library routines. 38066 (line 618) 38067 * __fractdasf: Fixed-point fractional library routines. 38068 (line 635) 38069 * __fractdasi: Fixed-point fractional library routines. 38070 (line 632) 38071 * __fractdasq: Fixed-point fractional library routines. 38072 (line 615) 38073 * __fractdata2: Fixed-point fractional library routines. 38074 (line 619) 38075 * __fractdati: Fixed-point fractional library routines. 38076 (line 634) 38077 * __fractdauda: Fixed-point fractional library routines. 38078 (line 627) 38079 * __fractdaudq: Fixed-point fractional library routines. 38080 (line 624) 38081 * __fractdauha: Fixed-point fractional library routines. 38082 (line 625) 38083 * __fractdauhq: Fixed-point fractional library routines. 38084 (line 621) 38085 * __fractdauqq: Fixed-point fractional library routines. 38086 (line 620) 38087 * __fractdausa: Fixed-point fractional library routines. 38088 (line 626) 38089 * __fractdausq: Fixed-point fractional library routines. 38090 (line 622) 38091 * __fractdauta: Fixed-point fractional library routines. 38092 (line 629) 38093 * __fractdfda: Fixed-point fractional library routines. 38094 (line 1025) 38095 * __fractdfdq: Fixed-point fractional library routines. 38096 (line 1022) 38097 * __fractdfha: Fixed-point fractional library routines. 38098 (line 1023) 38099 * __fractdfhq: Fixed-point fractional library routines. 38100 (line 1020) 38101 * __fractdfqq: Fixed-point fractional library routines. 38102 (line 1019) 38103 * __fractdfsa: Fixed-point fractional library routines. 38104 (line 1024) 38105 * __fractdfsq: Fixed-point fractional library routines. 38106 (line 1021) 38107 * __fractdfta: Fixed-point fractional library routines. 38108 (line 1026) 38109 * __fractdfuda: Fixed-point fractional library routines. 38110 (line 1033) 38111 * __fractdfudq: Fixed-point fractional library routines. 38112 (line 1030) 38113 * __fractdfuha: Fixed-point fractional library routines. 38114 (line 1031) 38115 * __fractdfuhq: Fixed-point fractional library routines. 38116 (line 1028) 38117 * __fractdfuqq: Fixed-point fractional library routines. 38118 (line 1027) 38119 * __fractdfusa: Fixed-point fractional library routines. 38120 (line 1032) 38121 * __fractdfusq: Fixed-point fractional library routines. 38122 (line 1029) 38123 * __fractdfuta: Fixed-point fractional library routines. 38124 (line 1034) 38125 * __fractdida: Fixed-point fractional library routines. 38126 (line 975) 38127 * __fractdidq: Fixed-point fractional library routines. 38128 (line 972) 38129 * __fractdiha: Fixed-point fractional library routines. 38130 (line 973) 38131 * __fractdihq: Fixed-point fractional library routines. 38132 (line 970) 38133 * __fractdiqq: Fixed-point fractional library routines. 38134 (line 969) 38135 * __fractdisa: Fixed-point fractional library routines. 38136 (line 974) 38137 * __fractdisq: Fixed-point fractional library routines. 38138 (line 971) 38139 * __fractdita: Fixed-point fractional library routines. 38140 (line 976) 38141 * __fractdiuda: Fixed-point fractional library routines. 38142 (line 983) 38143 * __fractdiudq: Fixed-point fractional library routines. 38144 (line 980) 38145 * __fractdiuha: Fixed-point fractional library routines. 38146 (line 981) 38147 * __fractdiuhq: Fixed-point fractional library routines. 38148 (line 978) 38149 * __fractdiuqq: Fixed-point fractional library routines. 38150 (line 977) 38151 * __fractdiusa: Fixed-point fractional library routines. 38152 (line 982) 38153 * __fractdiusq: Fixed-point fractional library routines. 38154 (line 979) 38155 * __fractdiuta: Fixed-point fractional library routines. 38156 (line 984) 38157 * __fractdqda: Fixed-point fractional library routines. 38158 (line 544) 38159 * __fractdqdf: Fixed-point fractional library routines. 38160 (line 566) 38161 * __fractdqdi: Fixed-point fractional library routines. 38162 (line 563) 38163 * __fractdqha: Fixed-point fractional library routines. 38164 (line 542) 38165 * __fractdqhi: Fixed-point fractional library routines. 38166 (line 561) 38167 * __fractdqhq2: Fixed-point fractional library routines. 38168 (line 540) 38169 * __fractdqqi: Fixed-point fractional library routines. 38170 (line 560) 38171 * __fractdqqq2: Fixed-point fractional library routines. 38172 (line 539) 38173 * __fractdqsa: Fixed-point fractional library routines. 38174 (line 543) 38175 * __fractdqsf: Fixed-point fractional library routines. 38176 (line 565) 38177 * __fractdqsi: Fixed-point fractional library routines. 38178 (line 562) 38179 * __fractdqsq2: Fixed-point fractional library routines. 38180 (line 541) 38181 * __fractdqta: Fixed-point fractional library routines. 38182 (line 545) 38183 * __fractdqti: Fixed-point fractional library routines. 38184 (line 564) 38185 * __fractdquda: Fixed-point fractional library routines. 38186 (line 557) 38187 * __fractdqudq: Fixed-point fractional library routines. 38188 (line 552) 38189 * __fractdquha: Fixed-point fractional library routines. 38190 (line 554) 38191 * __fractdquhq: Fixed-point fractional library routines. 38192 (line 548) 38193 * __fractdquqq: Fixed-point fractional library routines. 38194 (line 547) 38195 * __fractdqusa: Fixed-point fractional library routines. 38196 (line 555) 38197 * __fractdqusq: Fixed-point fractional library routines. 38198 (line 550) 38199 * __fractdquta: Fixed-point fractional library routines. 38200 (line 559) 38201 * __fracthada2: Fixed-point fractional library routines. 38202 (line 572) 38203 * __fracthadf: Fixed-point fractional library routines. 38204 (line 590) 38205 * __fracthadi: Fixed-point fractional library routines. 38206 (line 587) 38207 * __fracthadq: Fixed-point fractional library routines. 38208 (line 570) 38209 * __fracthahi: Fixed-point fractional library routines. 38210 (line 585) 38211 * __fracthahq: Fixed-point fractional library routines. 38212 (line 568) 38213 * __fracthaqi: Fixed-point fractional library routines. 38214 (line 584) 38215 * __fracthaqq: Fixed-point fractional library routines. 38216 (line 567) 38217 * __fracthasa2: Fixed-point fractional library routines. 38218 (line 571) 38219 * __fracthasf: Fixed-point fractional library routines. 38220 (line 589) 38221 * __fracthasi: Fixed-point fractional library routines. 38222 (line 586) 38223 * __fracthasq: Fixed-point fractional library routines. 38224 (line 569) 38225 * __fracthata2: Fixed-point fractional library routines. 38226 (line 573) 38227 * __fracthati: Fixed-point fractional library routines. 38228 (line 588) 38229 * __fracthauda: Fixed-point fractional library routines. 38230 (line 581) 38231 * __fracthaudq: Fixed-point fractional library routines. 38232 (line 578) 38233 * __fracthauha: Fixed-point fractional library routines. 38234 (line 579) 38235 * __fracthauhq: Fixed-point fractional library routines. 38236 (line 575) 38237 * __fracthauqq: Fixed-point fractional library routines. 38238 (line 574) 38239 * __fracthausa: Fixed-point fractional library routines. 38240 (line 580) 38241 * __fracthausq: Fixed-point fractional library routines. 38242 (line 576) 38243 * __fracthauta: Fixed-point fractional library routines. 38244 (line 583) 38245 * __fracthida: Fixed-point fractional library routines. 38246 (line 943) 38247 * __fracthidq: Fixed-point fractional library routines. 38248 (line 940) 38249 * __fracthiha: Fixed-point fractional library routines. 38250 (line 941) 38251 * __fracthihq: Fixed-point fractional library routines. 38252 (line 938) 38253 * __fracthiqq: Fixed-point fractional library routines. 38254 (line 937) 38255 * __fracthisa: Fixed-point fractional library routines. 38256 (line 942) 38257 * __fracthisq: Fixed-point fractional library routines. 38258 (line 939) 38259 * __fracthita: Fixed-point fractional library routines. 38260 (line 944) 38261 * __fracthiuda: Fixed-point fractional library routines. 38262 (line 951) 38263 * __fracthiudq: Fixed-point fractional library routines. 38264 (line 948) 38265 * __fracthiuha: Fixed-point fractional library routines. 38266 (line 949) 38267 * __fracthiuhq: Fixed-point fractional library routines. 38268 (line 946) 38269 * __fracthiuqq: Fixed-point fractional library routines. 38270 (line 945) 38271 * __fracthiusa: Fixed-point fractional library routines. 38272 (line 950) 38273 * __fracthiusq: Fixed-point fractional library routines. 38274 (line 947) 38275 * __fracthiuta: Fixed-point fractional library routines. 38276 (line 952) 38277 * __fracthqda: Fixed-point fractional library routines. 38278 (line 498) 38279 * __fracthqdf: Fixed-point fractional library routines. 38280 (line 514) 38281 * __fracthqdi: Fixed-point fractional library routines. 38282 (line 511) 38283 * __fracthqdq2: Fixed-point fractional library routines. 38284 (line 495) 38285 * __fracthqha: Fixed-point fractional library routines. 38286 (line 496) 38287 * __fracthqhi: Fixed-point fractional library routines. 38288 (line 509) 38289 * __fracthqqi: Fixed-point fractional library routines. 38290 (line 508) 38291 * __fracthqqq2: Fixed-point fractional library routines. 38292 (line 493) 38293 * __fracthqsa: Fixed-point fractional library routines. 38294 (line 497) 38295 * __fracthqsf: Fixed-point fractional library routines. 38296 (line 513) 38297 * __fracthqsi: Fixed-point fractional library routines. 38298 (line 510) 38299 * __fracthqsq2: Fixed-point fractional library routines. 38300 (line 494) 38301 * __fracthqta: Fixed-point fractional library routines. 38302 (line 499) 38303 * __fracthqti: Fixed-point fractional library routines. 38304 (line 512) 38305 * __fracthquda: Fixed-point fractional library routines. 38306 (line 506) 38307 * __fracthqudq: Fixed-point fractional library routines. 38308 (line 503) 38309 * __fracthquha: Fixed-point fractional library routines. 38310 (line 504) 38311 * __fracthquhq: Fixed-point fractional library routines. 38312 (line 501) 38313 * __fracthquqq: Fixed-point fractional library routines. 38314 (line 500) 38315 * __fracthqusa: Fixed-point fractional library routines. 38316 (line 505) 38317 * __fracthqusq: Fixed-point fractional library routines. 38318 (line 502) 38319 * __fracthquta: Fixed-point fractional library routines. 38320 (line 507) 38321 * __fractqida: Fixed-point fractional library routines. 38322 (line 925) 38323 * __fractqidq: Fixed-point fractional library routines. 38324 (line 922) 38325 * __fractqiha: Fixed-point fractional library routines. 38326 (line 923) 38327 * __fractqihq: Fixed-point fractional library routines. 38328 (line 920) 38329 * __fractqiqq: Fixed-point fractional library routines. 38330 (line 919) 38331 * __fractqisa: Fixed-point fractional library routines. 38332 (line 924) 38333 * __fractqisq: Fixed-point fractional library routines. 38334 (line 921) 38335 * __fractqita: Fixed-point fractional library routines. 38336 (line 926) 38337 * __fractqiuda: Fixed-point fractional library routines. 38338 (line 934) 38339 * __fractqiudq: Fixed-point fractional library routines. 38340 (line 931) 38341 * __fractqiuha: Fixed-point fractional library routines. 38342 (line 932) 38343 * __fractqiuhq: Fixed-point fractional library routines. 38344 (line 928) 38345 * __fractqiuqq: Fixed-point fractional library routines. 38346 (line 927) 38347 * __fractqiusa: Fixed-point fractional library routines. 38348 (line 933) 38349 * __fractqiusq: Fixed-point fractional library routines. 38350 (line 929) 38351 * __fractqiuta: Fixed-point fractional library routines. 38352 (line 936) 38353 * __fractqqda: Fixed-point fractional library routines. 38354 (line 474) 38355 * __fractqqdf: Fixed-point fractional library routines. 38356 (line 492) 38357 * __fractqqdi: Fixed-point fractional library routines. 38358 (line 489) 38359 * __fractqqdq2: Fixed-point fractional library routines. 38360 (line 471) 38361 * __fractqqha: Fixed-point fractional library routines. 38362 (line 472) 38363 * __fractqqhi: Fixed-point fractional library routines. 38364 (line 487) 38365 * __fractqqhq2: Fixed-point fractional library routines. 38366 (line 469) 38367 * __fractqqqi: Fixed-point fractional library routines. 38368 (line 486) 38369 * __fractqqsa: Fixed-point fractional library routines. 38370 (line 473) 38371 * __fractqqsf: Fixed-point fractional library routines. 38372 (line 491) 38373 * __fractqqsi: Fixed-point fractional library routines. 38374 (line 488) 38375 * __fractqqsq2: Fixed-point fractional library routines. 38376 (line 470) 38377 * __fractqqta: Fixed-point fractional library routines. 38378 (line 475) 38379 * __fractqqti: Fixed-point fractional library routines. 38380 (line 490) 38381 * __fractqquda: Fixed-point fractional library routines. 38382 (line 483) 38383 * __fractqqudq: Fixed-point fractional library routines. 38384 (line 480) 38385 * __fractqquha: Fixed-point fractional library routines. 38386 (line 481) 38387 * __fractqquhq: Fixed-point fractional library routines. 38388 (line 477) 38389 * __fractqquqq: Fixed-point fractional library routines. 38390 (line 476) 38391 * __fractqqusa: Fixed-point fractional library routines. 38392 (line 482) 38393 * __fractqqusq: Fixed-point fractional library routines. 38394 (line 478) 38395 * __fractqquta: Fixed-point fractional library routines. 38396 (line 485) 38397 * __fractsada2: Fixed-point fractional library routines. 38398 (line 596) 38399 * __fractsadf: Fixed-point fractional library routines. 38400 (line 612) 38401 * __fractsadi: Fixed-point fractional library routines. 38402 (line 609) 38403 * __fractsadq: Fixed-point fractional library routines. 38404 (line 594) 38405 * __fractsaha2: Fixed-point fractional library routines. 38406 (line 595) 38407 * __fractsahi: Fixed-point fractional library routines. 38408 (line 607) 38409 * __fractsahq: Fixed-point fractional library routines. 38410 (line 592) 38411 * __fractsaqi: Fixed-point fractional library routines. 38412 (line 606) 38413 * __fractsaqq: Fixed-point fractional library routines. 38414 (line 591) 38415 * __fractsasf: Fixed-point fractional library routines. 38416 (line 611) 38417 * __fractsasi: Fixed-point fractional library routines. 38418 (line 608) 38419 * __fractsasq: Fixed-point fractional library routines. 38420 (line 593) 38421 * __fractsata2: Fixed-point fractional library routines. 38422 (line 597) 38423 * __fractsati: Fixed-point fractional library routines. 38424 (line 610) 38425 * __fractsauda: Fixed-point fractional library routines. 38426 (line 604) 38427 * __fractsaudq: Fixed-point fractional library routines. 38428 (line 601) 38429 * __fractsauha: Fixed-point fractional library routines. 38430 (line 602) 38431 * __fractsauhq: Fixed-point fractional library routines. 38432 (line 599) 38433 * __fractsauqq: Fixed-point fractional library routines. 38434 (line 598) 38435 * __fractsausa: Fixed-point fractional library routines. 38436 (line 603) 38437 * __fractsausq: Fixed-point fractional library routines. 38438 (line 600) 38439 * __fractsauta: Fixed-point fractional library routines. 38440 (line 605) 38441 * __fractsfda: Fixed-point fractional library routines. 38442 (line 1009) 38443 * __fractsfdq: Fixed-point fractional library routines. 38444 (line 1006) 38445 * __fractsfha: Fixed-point fractional library routines. 38446 (line 1007) 38447 * __fractsfhq: Fixed-point fractional library routines. 38448 (line 1004) 38449 * __fractsfqq: Fixed-point fractional library routines. 38450 (line 1003) 38451 * __fractsfsa: Fixed-point fractional library routines. 38452 (line 1008) 38453 * __fractsfsq: Fixed-point fractional library routines. 38454 (line 1005) 38455 * __fractsfta: Fixed-point fractional library routines. 38456 (line 1010) 38457 * __fractsfuda: Fixed-point fractional library routines. 38458 (line 1017) 38459 * __fractsfudq: Fixed-point fractional library routines. 38460 (line 1014) 38461 * __fractsfuha: Fixed-point fractional library routines. 38462 (line 1015) 38463 * __fractsfuhq: Fixed-point fractional library routines. 38464 (line 1012) 38465 * __fractsfuqq: Fixed-point fractional library routines. 38466 (line 1011) 38467 * __fractsfusa: Fixed-point fractional library routines. 38468 (line 1016) 38469 * __fractsfusq: Fixed-point fractional library routines. 38470 (line 1013) 38471 * __fractsfuta: Fixed-point fractional library routines. 38472 (line 1018) 38473 * __fractsida: Fixed-point fractional library routines. 38474 (line 959) 38475 * __fractsidq: Fixed-point fractional library routines. 38476 (line 956) 38477 * __fractsiha: Fixed-point fractional library routines. 38478 (line 957) 38479 * __fractsihq: Fixed-point fractional library routines. 38480 (line 954) 38481 * __fractsiqq: Fixed-point fractional library routines. 38482 (line 953) 38483 * __fractsisa: Fixed-point fractional library routines. 38484 (line 958) 38485 * __fractsisq: Fixed-point fractional library routines. 38486 (line 955) 38487 * __fractsita: Fixed-point fractional library routines. 38488 (line 960) 38489 * __fractsiuda: Fixed-point fractional library routines. 38490 (line 967) 38491 * __fractsiudq: Fixed-point fractional library routines. 38492 (line 964) 38493 * __fractsiuha: Fixed-point fractional library routines. 38494 (line 965) 38495 * __fractsiuhq: Fixed-point fractional library routines. 38496 (line 962) 38497 * __fractsiuqq: Fixed-point fractional library routines. 38498 (line 961) 38499 * __fractsiusa: Fixed-point fractional library routines. 38500 (line 966) 38501 * __fractsiusq: Fixed-point fractional library routines. 38502 (line 963) 38503 * __fractsiuta: Fixed-point fractional library routines. 38504 (line 968) 38505 * __fractsqda: Fixed-point fractional library routines. 38506 (line 520) 38507 * __fractsqdf: Fixed-point fractional library routines. 38508 (line 538) 38509 * __fractsqdi: Fixed-point fractional library routines. 38510 (line 535) 38511 * __fractsqdq2: Fixed-point fractional library routines. 38512 (line 517) 38513 * __fractsqha: Fixed-point fractional library routines. 38514 (line 518) 38515 * __fractsqhi: Fixed-point fractional library routines. 38516 (line 533) 38517 * __fractsqhq2: Fixed-point fractional library routines. 38518 (line 516) 38519 * __fractsqqi: Fixed-point fractional library routines. 38520 (line 532) 38521 * __fractsqqq2: Fixed-point fractional library routines. 38522 (line 515) 38523 * __fractsqsa: Fixed-point fractional library routines. 38524 (line 519) 38525 * __fractsqsf: Fixed-point fractional library routines. 38526 (line 537) 38527 * __fractsqsi: Fixed-point fractional library routines. 38528 (line 534) 38529 * __fractsqta: Fixed-point fractional library routines. 38530 (line 521) 38531 * __fractsqti: Fixed-point fractional library routines. 38532 (line 536) 38533 * __fractsquda: Fixed-point fractional library routines. 38534 (line 529) 38535 * __fractsqudq: Fixed-point fractional library routines. 38536 (line 526) 38537 * __fractsquha: Fixed-point fractional library routines. 38538 (line 527) 38539 * __fractsquhq: Fixed-point fractional library routines. 38540 (line 523) 38541 * __fractsquqq: Fixed-point fractional library routines. 38542 (line 522) 38543 * __fractsqusa: Fixed-point fractional library routines. 38544 (line 528) 38545 * __fractsqusq: Fixed-point fractional library routines. 38546 (line 524) 38547 * __fractsquta: Fixed-point fractional library routines. 38548 (line 531) 38549 * __fracttada2: Fixed-point fractional library routines. 38550 (line 643) 38551 * __fracttadf: Fixed-point fractional library routines. 38552 (line 664) 38553 * __fracttadi: Fixed-point fractional library routines. 38554 (line 661) 38555 * __fracttadq: Fixed-point fractional library routines. 38556 (line 640) 38557 * __fracttaha2: Fixed-point fractional library routines. 38558 (line 641) 38559 * __fracttahi: Fixed-point fractional library routines. 38560 (line 659) 38561 * __fracttahq: Fixed-point fractional library routines. 38562 (line 638) 38563 * __fracttaqi: Fixed-point fractional library routines. 38564 (line 658) 38565 * __fracttaqq: Fixed-point fractional library routines. 38566 (line 637) 38567 * __fracttasa2: Fixed-point fractional library routines. 38568 (line 642) 38569 * __fracttasf: Fixed-point fractional library routines. 38570 (line 663) 38571 * __fracttasi: Fixed-point fractional library routines. 38572 (line 660) 38573 * __fracttasq: Fixed-point fractional library routines. 38574 (line 639) 38575 * __fracttati: Fixed-point fractional library routines. 38576 (line 662) 38577 * __fracttauda: Fixed-point fractional library routines. 38578 (line 655) 38579 * __fracttaudq: Fixed-point fractional library routines. 38580 (line 650) 38581 * __fracttauha: Fixed-point fractional library routines. 38582 (line 652) 38583 * __fracttauhq: Fixed-point fractional library routines. 38584 (line 646) 38585 * __fracttauqq: Fixed-point fractional library routines. 38586 (line 645) 38587 * __fracttausa: Fixed-point fractional library routines. 38588 (line 653) 38589 * __fracttausq: Fixed-point fractional library routines. 38590 (line 648) 38591 * __fracttauta: Fixed-point fractional library routines. 38592 (line 657) 38593 * __fracttida: Fixed-point fractional library routines. 38594 (line 991) 38595 * __fracttidq: Fixed-point fractional library routines. 38596 (line 988) 38597 * __fracttiha: Fixed-point fractional library routines. 38598 (line 989) 38599 * __fracttihq: Fixed-point fractional library routines. 38600 (line 986) 38601 * __fracttiqq: Fixed-point fractional library routines. 38602 (line 985) 38603 * __fracttisa: Fixed-point fractional library routines. 38604 (line 990) 38605 * __fracttisq: Fixed-point fractional library routines. 38606 (line 987) 38607 * __fracttita: Fixed-point fractional library routines. 38608 (line 992) 38609 * __fracttiuda: Fixed-point fractional library routines. 38610 (line 1000) 38611 * __fracttiudq: Fixed-point fractional library routines. 38612 (line 997) 38613 * __fracttiuha: Fixed-point fractional library routines. 38614 (line 998) 38615 * __fracttiuhq: Fixed-point fractional library routines. 38616 (line 994) 38617 * __fracttiuqq: Fixed-point fractional library routines. 38618 (line 993) 38619 * __fracttiusa: Fixed-point fractional library routines. 38620 (line 999) 38621 * __fracttiusq: Fixed-point fractional library routines. 38622 (line 995) 38623 * __fracttiuta: Fixed-point fractional library routines. 38624 (line 1002) 38625 * __fractudada: Fixed-point fractional library routines. 38626 (line 858) 38627 * __fractudadf: Fixed-point fractional library routines. 38628 (line 881) 38629 * __fractudadi: Fixed-point fractional library routines. 38630 (line 878) 38631 * __fractudadq: Fixed-point fractional library routines. 38632 (line 855) 38633 * __fractudaha: Fixed-point fractional library routines. 38634 (line 856) 38635 * __fractudahi: Fixed-point fractional library routines. 38636 (line 876) 38637 * __fractudahq: Fixed-point fractional library routines. 38638 (line 852) 38639 * __fractudaqi: Fixed-point fractional library routines. 38640 (line 875) 38641 * __fractudaqq: Fixed-point fractional library routines. 38642 (line 851) 38643 * __fractudasa: Fixed-point fractional library routines. 38644 (line 857) 38645 * __fractudasf: Fixed-point fractional library routines. 38646 (line 880) 38647 * __fractudasi: Fixed-point fractional library routines. 38648 (line 877) 38649 * __fractudasq: Fixed-point fractional library routines. 38650 (line 853) 38651 * __fractudata: Fixed-point fractional library routines. 38652 (line 860) 38653 * __fractudati: Fixed-point fractional library routines. 38654 (line 879) 38655 * __fractudaudq: Fixed-point fractional library routines. 38656 (line 868) 38657 * __fractudauha2: Fixed-point fractional library routines. 38658 (line 870) 38659 * __fractudauhq: Fixed-point fractional library routines. 38660 (line 864) 38661 * __fractudauqq: Fixed-point fractional library routines. 38662 (line 862) 38663 * __fractudausa2: Fixed-point fractional library routines. 38664 (line 872) 38665 * __fractudausq: Fixed-point fractional library routines. 38666 (line 866) 38667 * __fractudauta2: Fixed-point fractional library routines. 38668 (line 874) 38669 * __fractudqda: Fixed-point fractional library routines. 38670 (line 766) 38671 * __fractudqdf: Fixed-point fractional library routines. 38672 (line 791) 38673 * __fractudqdi: Fixed-point fractional library routines. 38674 (line 787) 38675 * __fractudqdq: Fixed-point fractional library routines. 38676 (line 761) 38677 * __fractudqha: Fixed-point fractional library routines. 38678 (line 763) 38679 * __fractudqhi: Fixed-point fractional library routines. 38680 (line 785) 38681 * __fractudqhq: Fixed-point fractional library routines. 38682 (line 757) 38683 * __fractudqqi: Fixed-point fractional library routines. 38684 (line 784) 38685 * __fractudqqq: Fixed-point fractional library routines. 38686 (line 756) 38687 * __fractudqsa: Fixed-point fractional library routines. 38688 (line 764) 38689 * __fractudqsf: Fixed-point fractional library routines. 38690 (line 790) 38691 * __fractudqsi: Fixed-point fractional library routines. 38692 (line 786) 38693 * __fractudqsq: Fixed-point fractional library routines. 38694 (line 759) 38695 * __fractudqta: Fixed-point fractional library routines. 38696 (line 768) 38697 * __fractudqti: Fixed-point fractional library routines. 38698 (line 789) 38699 * __fractudquda: Fixed-point fractional library routines. 38700 (line 780) 38701 * __fractudquha: Fixed-point fractional library routines. 38702 (line 776) 38703 * __fractudquhq2: Fixed-point fractional library routines. 38704 (line 772) 38705 * __fractudquqq2: Fixed-point fractional library routines. 38706 (line 770) 38707 * __fractudqusa: Fixed-point fractional library routines. 38708 (line 778) 38709 * __fractudqusq2: Fixed-point fractional library routines. 38710 (line 774) 38711 * __fractudquta: Fixed-point fractional library routines. 38712 (line 782) 38713 * __fractuhada: Fixed-point fractional library routines. 38714 (line 799) 38715 * __fractuhadf: Fixed-point fractional library routines. 38716 (line 822) 38717 * __fractuhadi: Fixed-point fractional library routines. 38718 (line 819) 38719 * __fractuhadq: Fixed-point fractional library routines. 38720 (line 796) 38721 * __fractuhaha: Fixed-point fractional library routines. 38722 (line 797) 38723 * __fractuhahi: Fixed-point fractional library routines. 38724 (line 817) 38725 * __fractuhahq: Fixed-point fractional library routines. 38726 (line 793) 38727 * __fractuhaqi: Fixed-point fractional library routines. 38728 (line 816) 38729 * __fractuhaqq: Fixed-point fractional library routines. 38730 (line 792) 38731 * __fractuhasa: Fixed-point fractional library routines. 38732 (line 798) 38733 * __fractuhasf: Fixed-point fractional library routines. 38734 (line 821) 38735 * __fractuhasi: Fixed-point fractional library routines. 38736 (line 818) 38737 * __fractuhasq: Fixed-point fractional library routines. 38738 (line 794) 38739 * __fractuhata: Fixed-point fractional library routines. 38740 (line 801) 38741 * __fractuhati: Fixed-point fractional library routines. 38742 (line 820) 38743 * __fractuhauda2: Fixed-point fractional library routines. 38744 (line 813) 38745 * __fractuhaudq: Fixed-point fractional library routines. 38746 (line 809) 38747 * __fractuhauhq: Fixed-point fractional library routines. 38748 (line 805) 38749 * __fractuhauqq: Fixed-point fractional library routines. 38750 (line 803) 38751 * __fractuhausa2: Fixed-point fractional library routines. 38752 (line 811) 38753 * __fractuhausq: Fixed-point fractional library routines. 38754 (line 807) 38755 * __fractuhauta2: Fixed-point fractional library routines. 38756 (line 815) 38757 * __fractuhqda: Fixed-point fractional library routines. 38758 (line 702) 38759 * __fractuhqdf: Fixed-point fractional library routines. 38760 (line 723) 38761 * __fractuhqdi: Fixed-point fractional library routines. 38762 (line 720) 38763 * __fractuhqdq: Fixed-point fractional library routines. 38764 (line 699) 38765 * __fractuhqha: Fixed-point fractional library routines. 38766 (line 700) 38767 * __fractuhqhi: Fixed-point fractional library routines. 38768 (line 718) 38769 * __fractuhqhq: Fixed-point fractional library routines. 38770 (line 697) 38771 * __fractuhqqi: Fixed-point fractional library routines. 38772 (line 717) 38773 * __fractuhqqq: Fixed-point fractional library routines. 38774 (line 696) 38775 * __fractuhqsa: Fixed-point fractional library routines. 38776 (line 701) 38777 * __fractuhqsf: Fixed-point fractional library routines. 38778 (line 722) 38779 * __fractuhqsi: Fixed-point fractional library routines. 38780 (line 719) 38781 * __fractuhqsq: Fixed-point fractional library routines. 38782 (line 698) 38783 * __fractuhqta: Fixed-point fractional library routines. 38784 (line 703) 38785 * __fractuhqti: Fixed-point fractional library routines. 38786 (line 721) 38787 * __fractuhquda: Fixed-point fractional library routines. 38788 (line 714) 38789 * __fractuhqudq2: Fixed-point fractional library routines. 38790 (line 709) 38791 * __fractuhquha: Fixed-point fractional library routines. 38792 (line 711) 38793 * __fractuhquqq2: Fixed-point fractional library routines. 38794 (line 705) 38795 * __fractuhqusa: Fixed-point fractional library routines. 38796 (line 712) 38797 * __fractuhqusq2: Fixed-point fractional library routines. 38798 (line 707) 38799 * __fractuhquta: Fixed-point fractional library routines. 38800 (line 716) 38801 * __fractunsdadi: Fixed-point fractional library routines. 38802 (line 1555) 38803 * __fractunsdahi: Fixed-point fractional library routines. 38804 (line 1553) 38805 * __fractunsdaqi: Fixed-point fractional library routines. 38806 (line 1552) 38807 * __fractunsdasi: Fixed-point fractional library routines. 38808 (line 1554) 38809 * __fractunsdati: Fixed-point fractional library routines. 38810 (line 1556) 38811 * __fractunsdida: Fixed-point fractional library routines. 38812 (line 1707) 38813 * __fractunsdidq: Fixed-point fractional library routines. 38814 (line 1704) 38815 * __fractunsdiha: Fixed-point fractional library routines. 38816 (line 1705) 38817 * __fractunsdihq: Fixed-point fractional library routines. 38818 (line 1702) 38819 * __fractunsdiqq: Fixed-point fractional library routines. 38820 (line 1701) 38821 * __fractunsdisa: Fixed-point fractional library routines. 38822 (line 1706) 38823 * __fractunsdisq: Fixed-point fractional library routines. 38824 (line 1703) 38825 * __fractunsdita: Fixed-point fractional library routines. 38826 (line 1708) 38827 * __fractunsdiuda: Fixed-point fractional library routines. 38828 (line 1720) 38829 * __fractunsdiudq: Fixed-point fractional library routines. 38830 (line 1715) 38831 * __fractunsdiuha: Fixed-point fractional library routines. 38832 (line 1717) 38833 * __fractunsdiuhq: Fixed-point fractional library routines. 38834 (line 1711) 38835 * __fractunsdiuqq: Fixed-point fractional library routines. 38836 (line 1710) 38837 * __fractunsdiusa: Fixed-point fractional library routines. 38838 (line 1718) 38839 * __fractunsdiusq: Fixed-point fractional library routines. 38840 (line 1713) 38841 * __fractunsdiuta: Fixed-point fractional library routines. 38842 (line 1722) 38843 * __fractunsdqdi: Fixed-point fractional library routines. 38844 (line 1539) 38845 * __fractunsdqhi: Fixed-point fractional library routines. 38846 (line 1537) 38847 * __fractunsdqqi: Fixed-point fractional library routines. 38848 (line 1536) 38849 * __fractunsdqsi: Fixed-point fractional library routines. 38850 (line 1538) 38851 * __fractunsdqti: Fixed-point fractional library routines. 38852 (line 1541) 38853 * __fractunshadi: Fixed-point fractional library routines. 38854 (line 1545) 38855 * __fractunshahi: Fixed-point fractional library routines. 38856 (line 1543) 38857 * __fractunshaqi: Fixed-point fractional library routines. 38858 (line 1542) 38859 * __fractunshasi: Fixed-point fractional library routines. 38860 (line 1544) 38861 * __fractunshati: Fixed-point fractional library routines. 38862 (line 1546) 38863 * __fractunshida: Fixed-point fractional library routines. 38864 (line 1663) 38865 * __fractunshidq: Fixed-point fractional library routines. 38866 (line 1660) 38867 * __fractunshiha: Fixed-point fractional library routines. 38868 (line 1661) 38869 * __fractunshihq: Fixed-point fractional library routines. 38870 (line 1658) 38871 * __fractunshiqq: Fixed-point fractional library routines. 38872 (line 1657) 38873 * __fractunshisa: Fixed-point fractional library routines. 38874 (line 1662) 38875 * __fractunshisq: Fixed-point fractional library routines. 38876 (line 1659) 38877 * __fractunshita: Fixed-point fractional library routines. 38878 (line 1664) 38879 * __fractunshiuda: Fixed-point fractional library routines. 38880 (line 1676) 38881 * __fractunshiudq: Fixed-point fractional library routines. 38882 (line 1671) 38883 * __fractunshiuha: Fixed-point fractional library routines. 38884 (line 1673) 38885 * __fractunshiuhq: Fixed-point fractional library routines. 38886 (line 1667) 38887 * __fractunshiuqq: Fixed-point fractional library routines. 38888 (line 1666) 38889 * __fractunshiusa: Fixed-point fractional library routines. 38890 (line 1674) 38891 * __fractunshiusq: Fixed-point fractional library routines. 38892 (line 1669) 38893 * __fractunshiuta: Fixed-point fractional library routines. 38894 (line 1678) 38895 * __fractunshqdi: Fixed-point fractional library routines. 38896 (line 1529) 38897 * __fractunshqhi: Fixed-point fractional library routines. 38898 (line 1527) 38899 * __fractunshqqi: Fixed-point fractional library routines. 38900 (line 1526) 38901 * __fractunshqsi: Fixed-point fractional library routines. 38902 (line 1528) 38903 * __fractunshqti: Fixed-point fractional library routines. 38904 (line 1530) 38905 * __fractunsqida: Fixed-point fractional library routines. 38906 (line 1641) 38907 * __fractunsqidq: Fixed-point fractional library routines. 38908 (line 1638) 38909 * __fractunsqiha: Fixed-point fractional library routines. 38910 (line 1639) 38911 * __fractunsqihq: Fixed-point fractional library routines. 38912 (line 1636) 38913 * __fractunsqiqq: Fixed-point fractional library routines. 38914 (line 1635) 38915 * __fractunsqisa: Fixed-point fractional library routines. 38916 (line 1640) 38917 * __fractunsqisq: Fixed-point fractional library routines. 38918 (line 1637) 38919 * __fractunsqita: Fixed-point fractional library routines. 38920 (line 1642) 38921 * __fractunsqiuda: Fixed-point fractional library routines. 38922 (line 1654) 38923 * __fractunsqiudq: Fixed-point fractional library routines. 38924 (line 1649) 38925 * __fractunsqiuha: Fixed-point fractional library routines. 38926 (line 1651) 38927 * __fractunsqiuhq: Fixed-point fractional library routines. 38928 (line 1645) 38929 * __fractunsqiuqq: Fixed-point fractional library routines. 38930 (line 1644) 38931 * __fractunsqiusa: Fixed-point fractional library routines. 38932 (line 1652) 38933 * __fractunsqiusq: Fixed-point fractional library routines. 38934 (line 1647) 38935 * __fractunsqiuta: Fixed-point fractional library routines. 38936 (line 1656) 38937 * __fractunsqqdi: Fixed-point fractional library routines. 38938 (line 1524) 38939 * __fractunsqqhi: Fixed-point fractional library routines. 38940 (line 1522) 38941 * __fractunsqqqi: Fixed-point fractional library routines. 38942 (line 1521) 38943 * __fractunsqqsi: Fixed-point fractional library routines. 38944 (line 1523) 38945 * __fractunsqqti: Fixed-point fractional library routines. 38946 (line 1525) 38947 * __fractunssadi: Fixed-point fractional library routines. 38948 (line 1550) 38949 * __fractunssahi: Fixed-point fractional library routines. 38950 (line 1548) 38951 * __fractunssaqi: Fixed-point fractional library routines. 38952 (line 1547) 38953 * __fractunssasi: Fixed-point fractional library routines. 38954 (line 1549) 38955 * __fractunssati: Fixed-point fractional library routines. 38956 (line 1551) 38957 * __fractunssida: Fixed-point fractional library routines. 38958 (line 1685) 38959 * __fractunssidq: Fixed-point fractional library routines. 38960 (line 1682) 38961 * __fractunssiha: Fixed-point fractional library routines. 38962 (line 1683) 38963 * __fractunssihq: Fixed-point fractional library routines. 38964 (line 1680) 38965 * __fractunssiqq: Fixed-point fractional library routines. 38966 (line 1679) 38967 * __fractunssisa: Fixed-point fractional library routines. 38968 (line 1684) 38969 * __fractunssisq: Fixed-point fractional library routines. 38970 (line 1681) 38971 * __fractunssita: Fixed-point fractional library routines. 38972 (line 1686) 38973 * __fractunssiuda: Fixed-point fractional library routines. 38974 (line 1698) 38975 * __fractunssiudq: Fixed-point fractional library routines. 38976 (line 1693) 38977 * __fractunssiuha: Fixed-point fractional library routines. 38978 (line 1695) 38979 * __fractunssiuhq: Fixed-point fractional library routines. 38980 (line 1689) 38981 * __fractunssiuqq: Fixed-point fractional library routines. 38982 (line 1688) 38983 * __fractunssiusa: Fixed-point fractional library routines. 38984 (line 1696) 38985 * __fractunssiusq: Fixed-point fractional library routines. 38986 (line 1691) 38987 * __fractunssiuta: Fixed-point fractional library routines. 38988 (line 1700) 38989 * __fractunssqdi: Fixed-point fractional library routines. 38990 (line 1534) 38991 * __fractunssqhi: Fixed-point fractional library routines. 38992 (line 1532) 38993 * __fractunssqqi: Fixed-point fractional library routines. 38994 (line 1531) 38995 * __fractunssqsi: Fixed-point fractional library routines. 38996 (line 1533) 38997 * __fractunssqti: Fixed-point fractional library routines. 38998 (line 1535) 38999 * __fractunstadi: Fixed-point fractional library routines. 39000 (line 1560) 39001 * __fractunstahi: Fixed-point fractional library routines. 39002 (line 1558) 39003 * __fractunstaqi: Fixed-point fractional library routines. 39004 (line 1557) 39005 * __fractunstasi: Fixed-point fractional library routines. 39006 (line 1559) 39007 * __fractunstati: Fixed-point fractional library routines. 39008 (line 1562) 39009 * __fractunstida: Fixed-point fractional library routines. 39010 (line 1730) 39011 * __fractunstidq: Fixed-point fractional library routines. 39012 (line 1727) 39013 * __fractunstiha: Fixed-point fractional library routines. 39014 (line 1728) 39015 * __fractunstihq: Fixed-point fractional library routines. 39016 (line 1724) 39017 * __fractunstiqq: Fixed-point fractional library routines. 39018 (line 1723) 39019 * __fractunstisa: Fixed-point fractional library routines. 39020 (line 1729) 39021 * __fractunstisq: Fixed-point fractional library routines. 39022 (line 1725) 39023 * __fractunstita: Fixed-point fractional library routines. 39024 (line 1732) 39025 * __fractunstiuda: Fixed-point fractional library routines. 39026 (line 1746) 39027 * __fractunstiudq: Fixed-point fractional library routines. 39028 (line 1740) 39029 * __fractunstiuha: Fixed-point fractional library routines. 39030 (line 1742) 39031 * __fractunstiuhq: Fixed-point fractional library routines. 39032 (line 1736) 39033 * __fractunstiuqq: Fixed-point fractional library routines. 39034 (line 1734) 39035 * __fractunstiusa: Fixed-point fractional library routines. 39036 (line 1744) 39037 * __fractunstiusq: Fixed-point fractional library routines. 39038 (line 1738) 39039 * __fractunstiuta: Fixed-point fractional library routines. 39040 (line 1748) 39041 * __fractunsudadi: Fixed-point fractional library routines. 39042 (line 1622) 39043 * __fractunsudahi: Fixed-point fractional library routines. 39044 (line 1618) 39045 * __fractunsudaqi: Fixed-point fractional library routines. 39046 (line 1616) 39047 * __fractunsudasi: Fixed-point fractional library routines. 39048 (line 1620) 39049 * __fractunsudati: Fixed-point fractional library routines. 39050 (line 1624) 39051 * __fractunsudqdi: Fixed-point fractional library routines. 39052 (line 1596) 39053 * __fractunsudqhi: Fixed-point fractional library routines. 39054 (line 1592) 39055 * __fractunsudqqi: Fixed-point fractional library routines. 39056 (line 1590) 39057 * __fractunsudqsi: Fixed-point fractional library routines. 39058 (line 1594) 39059 * __fractunsudqti: Fixed-point fractional library routines. 39060 (line 1598) 39061 * __fractunsuhadi: Fixed-point fractional library routines. 39062 (line 1606) 39063 * __fractunsuhahi: Fixed-point fractional library routines. 39064 (line 1602) 39065 * __fractunsuhaqi: Fixed-point fractional library routines. 39066 (line 1600) 39067 * __fractunsuhasi: Fixed-point fractional library routines. 39068 (line 1604) 39069 * __fractunsuhati: Fixed-point fractional library routines. 39070 (line 1608) 39071 * __fractunsuhqdi: Fixed-point fractional library routines. 39072 (line 1576) 39073 * __fractunsuhqhi: Fixed-point fractional library routines. 39074 (line 1574) 39075 * __fractunsuhqqi: Fixed-point fractional library routines. 39076 (line 1573) 39077 * __fractunsuhqsi: Fixed-point fractional library routines. 39078 (line 1575) 39079 * __fractunsuhqti: Fixed-point fractional library routines. 39080 (line 1578) 39081 * __fractunsuqqdi: Fixed-point fractional library routines. 39082 (line 1570) 39083 * __fractunsuqqhi: Fixed-point fractional library routines. 39084 (line 1566) 39085 * __fractunsuqqqi: Fixed-point fractional library routines. 39086 (line 1564) 39087 * __fractunsuqqsi: Fixed-point fractional library routines. 39088 (line 1568) 39089 * __fractunsuqqti: Fixed-point fractional library routines. 39090 (line 1572) 39091 * __fractunsusadi: Fixed-point fractional library routines. 39092 (line 1612) 39093 * __fractunsusahi: Fixed-point fractional library routines. 39094 (line 1610) 39095 * __fractunsusaqi: Fixed-point fractional library routines. 39096 (line 1609) 39097 * __fractunsusasi: Fixed-point fractional library routines. 39098 (line 1611) 39099 * __fractunsusati: Fixed-point fractional library routines. 39100 (line 1614) 39101 * __fractunsusqdi: Fixed-point fractional library routines. 39102 (line 1586) 39103 * __fractunsusqhi: Fixed-point fractional library routines. 39104 (line 1582) 39105 * __fractunsusqqi: Fixed-point fractional library routines. 39106 (line 1580) 39107 * __fractunsusqsi: Fixed-point fractional library routines. 39108 (line 1584) 39109 * __fractunsusqti: Fixed-point fractional library routines. 39110 (line 1588) 39111 * __fractunsutadi: Fixed-point fractional library routines. 39112 (line 1632) 39113 * __fractunsutahi: Fixed-point fractional library routines. 39114 (line 1628) 39115 * __fractunsutaqi: Fixed-point fractional library routines. 39116 (line 1626) 39117 * __fractunsutasi: Fixed-point fractional library routines. 39118 (line 1630) 39119 * __fractunsutati: Fixed-point fractional library routines. 39120 (line 1634) 39121 * __fractuqqda: Fixed-point fractional library routines. 39122 (line 672) 39123 * __fractuqqdf: Fixed-point fractional library routines. 39124 (line 695) 39125 * __fractuqqdi: Fixed-point fractional library routines. 39126 (line 692) 39127 * __fractuqqdq: Fixed-point fractional library routines. 39128 (line 669) 39129 * __fractuqqha: Fixed-point fractional library routines. 39130 (line 670) 39131 * __fractuqqhi: Fixed-point fractional library routines. 39132 (line 690) 39133 * __fractuqqhq: Fixed-point fractional library routines. 39134 (line 666) 39135 * __fractuqqqi: Fixed-point fractional library routines. 39136 (line 689) 39137 * __fractuqqqq: Fixed-point fractional library routines. 39138 (line 665) 39139 * __fractuqqsa: Fixed-point fractional library routines. 39140 (line 671) 39141 * __fractuqqsf: Fixed-point fractional library routines. 39142 (line 694) 39143 * __fractuqqsi: Fixed-point fractional library routines. 39144 (line 691) 39145 * __fractuqqsq: Fixed-point fractional library routines. 39146 (line 667) 39147 * __fractuqqta: Fixed-point fractional library routines. 39148 (line 674) 39149 * __fractuqqti: Fixed-point fractional library routines. 39150 (line 693) 39151 * __fractuqquda: Fixed-point fractional library routines. 39152 (line 686) 39153 * __fractuqqudq2: Fixed-point fractional library routines. 39154 (line 680) 39155 * __fractuqquha: Fixed-point fractional library routines. 39156 (line 682) 39157 * __fractuqquhq2: Fixed-point fractional library routines. 39158 (line 676) 39159 * __fractuqqusa: Fixed-point fractional library routines. 39160 (line 684) 39161 * __fractuqqusq2: Fixed-point fractional library routines. 39162 (line 678) 39163 * __fractuqquta: Fixed-point fractional library routines. 39164 (line 688) 39165 * __fractusada: Fixed-point fractional library routines. 39166 (line 829) 39167 * __fractusadf: Fixed-point fractional library routines. 39168 (line 850) 39169 * __fractusadi: Fixed-point fractional library routines. 39170 (line 847) 39171 * __fractusadq: Fixed-point fractional library routines. 39172 (line 826) 39173 * __fractusaha: Fixed-point fractional library routines. 39174 (line 827) 39175 * __fractusahi: Fixed-point fractional library routines. 39176 (line 845) 39177 * __fractusahq: Fixed-point fractional library routines. 39178 (line 824) 39179 * __fractusaqi: Fixed-point fractional library routines. 39180 (line 844) 39181 * __fractusaqq: Fixed-point fractional library routines. 39182 (line 823) 39183 * __fractusasa: Fixed-point fractional library routines. 39184 (line 828) 39185 * __fractusasf: Fixed-point fractional library routines. 39186 (line 849) 39187 * __fractusasi: Fixed-point fractional library routines. 39188 (line 846) 39189 * __fractusasq: Fixed-point fractional library routines. 39190 (line 825) 39191 * __fractusata: Fixed-point fractional library routines. 39192 (line 830) 39193 * __fractusati: Fixed-point fractional library routines. 39194 (line 848) 39195 * __fractusauda2: Fixed-point fractional library routines. 39196 (line 841) 39197 * __fractusaudq: Fixed-point fractional library routines. 39198 (line 837) 39199 * __fractusauha2: Fixed-point fractional library routines. 39200 (line 839) 39201 * __fractusauhq: Fixed-point fractional library routines. 39202 (line 833) 39203 * __fractusauqq: Fixed-point fractional library routines. 39204 (line 832) 39205 * __fractusausq: Fixed-point fractional library routines. 39206 (line 835) 39207 * __fractusauta2: Fixed-point fractional library routines. 39208 (line 843) 39209 * __fractusqda: Fixed-point fractional library routines. 39210 (line 731) 39211 * __fractusqdf: Fixed-point fractional library routines. 39212 (line 754) 39213 * __fractusqdi: Fixed-point fractional library routines. 39214 (line 751) 39215 * __fractusqdq: Fixed-point fractional library routines. 39216 (line 728) 39217 * __fractusqha: Fixed-point fractional library routines. 39218 (line 729) 39219 * __fractusqhi: Fixed-point fractional library routines. 39220 (line 749) 39221 * __fractusqhq: Fixed-point fractional library routines. 39222 (line 725) 39223 * __fractusqqi: Fixed-point fractional library routines. 39224 (line 748) 39225 * __fractusqqq: Fixed-point fractional library routines. 39226 (line 724) 39227 * __fractusqsa: Fixed-point fractional library routines. 39228 (line 730) 39229 * __fractusqsf: Fixed-point fractional library routines. 39230 (line 753) 39231 * __fractusqsi: Fixed-point fractional library routines. 39232 (line 750) 39233 * __fractusqsq: Fixed-point fractional library routines. 39234 (line 726) 39235 * __fractusqta: Fixed-point fractional library routines. 39236 (line 733) 39237 * __fractusqti: Fixed-point fractional library routines. 39238 (line 752) 39239 * __fractusquda: Fixed-point fractional library routines. 39240 (line 745) 39241 * __fractusqudq2: Fixed-point fractional library routines. 39242 (line 739) 39243 * __fractusquha: Fixed-point fractional library routines. 39244 (line 741) 39245 * __fractusquhq2: Fixed-point fractional library routines. 39246 (line 737) 39247 * __fractusquqq2: Fixed-point fractional library routines. 39248 (line 735) 39249 * __fractusqusa: Fixed-point fractional library routines. 39250 (line 743) 39251 * __fractusquta: Fixed-point fractional library routines. 39252 (line 747) 39253 * __fractutada: Fixed-point fractional library routines. 39254 (line 893) 39255 * __fractutadf: Fixed-point fractional library routines. 39256 (line 918) 39257 * __fractutadi: Fixed-point fractional library routines. 39258 (line 914) 39259 * __fractutadq: Fixed-point fractional library routines. 39260 (line 888) 39261 * __fractutaha: Fixed-point fractional library routines. 39262 (line 890) 39263 * __fractutahi: Fixed-point fractional library routines. 39264 (line 912) 39265 * __fractutahq: Fixed-point fractional library routines. 39266 (line 884) 39267 * __fractutaqi: Fixed-point fractional library routines. 39268 (line 911) 39269 * __fractutaqq: Fixed-point fractional library routines. 39270 (line 883) 39271 * __fractutasa: Fixed-point fractional library routines. 39272 (line 891) 39273 * __fractutasf: Fixed-point fractional library routines. 39274 (line 917) 39275 * __fractutasi: Fixed-point fractional library routines. 39276 (line 913) 39277 * __fractutasq: Fixed-point fractional library routines. 39278 (line 886) 39279 * __fractutata: Fixed-point fractional library routines. 39280 (line 895) 39281 * __fractutati: Fixed-point fractional library routines. 39282 (line 916) 39283 * __fractutauda2: Fixed-point fractional library routines. 39284 (line 909) 39285 * __fractutaudq: Fixed-point fractional library routines. 39286 (line 903) 39287 * __fractutauha2: Fixed-point fractional library routines. 39288 (line 905) 39289 * __fractutauhq: Fixed-point fractional library routines. 39290 (line 899) 39291 * __fractutauqq: Fixed-point fractional library routines. 39292 (line 897) 39293 * __fractutausa2: Fixed-point fractional library routines. 39294 (line 907) 39295 * __fractutausq: Fixed-point fractional library routines. 39296 (line 901) 39297 * __gedf2: Soft float library routines. 39298 (line 206) 39299 * __gesf2: Soft float library routines. 39300 (line 205) 39301 * __getf2: Soft float library routines. 39302 (line 207) 39303 * __gtdf2: Soft float library routines. 39304 (line 224) 39305 * __gtsf2: Soft float library routines. 39306 (line 223) 39307 * __gttf2: Soft float library routines. 39308 (line 225) 39309 * __ledf2: Soft float library routines. 39310 (line 218) 39311 * __lesf2: Soft float library routines. 39312 (line 217) 39313 * __letf2: Soft float library routines. 39314 (line 219) 39315 * __lshrdi3: Integer library routines. 39316 (line 31) 39317 * __lshrsi3: Integer library routines. 39318 (line 30) 39319 * __lshrti3: Integer library routines. 39320 (line 32) 39321 * __lshruda3: Fixed-point fractional library routines. 39322 (line 390) 39323 * __lshrudq3: Fixed-point fractional library routines. 39324 (line 384) 39325 * __lshruha3: Fixed-point fractional library routines. 39326 (line 386) 39327 * __lshruhq3: Fixed-point fractional library routines. 39328 (line 380) 39329 * __lshruqq3: Fixed-point fractional library routines. 39330 (line 378) 39331 * __lshrusa3: Fixed-point fractional library routines. 39332 (line 388) 39333 * __lshrusq3: Fixed-point fractional library routines. 39334 (line 382) 39335 * __lshruta3: Fixed-point fractional library routines. 39336 (line 392) 39337 * __ltdf2: Soft float library routines. 39338 (line 212) 39339 * __ltsf2: Soft float library routines. 39340 (line 211) 39341 * __lttf2: Soft float library routines. 39342 (line 213) 39343 * __main: Collect2. (line 15) 39344 * __moddi3: Integer library routines. 39345 (line 37) 39346 * __modsi3: Integer library routines. 39347 (line 36) 39348 * __modti3: Integer library routines. 39349 (line 38) 39350 * __mulda3: Fixed-point fractional library routines. 39351 (line 171) 39352 * __muldc3: Soft float library routines. 39353 (line 241) 39354 * __muldf3: Soft float library routines. 39355 (line 40) 39356 * __muldi3: Integer library routines. 39357 (line 43) 39358 * __muldq3: Fixed-point fractional library routines. 39359 (line 159) 39360 * __mulha3: Fixed-point fractional library routines. 39361 (line 169) 39362 * __mulhq3: Fixed-point fractional library routines. 39363 (line 156) 39364 * __mulqq3: Fixed-point fractional library routines. 39365 (line 155) 39366 * __mulsa3: Fixed-point fractional library routines. 39367 (line 170) 39368 * __mulsc3: Soft float library routines. 39369 (line 239) 39370 * __mulsf3: Soft float library routines. 39371 (line 39) 39372 * __mulsi3: Integer library routines. 39373 (line 42) 39374 * __mulsq3: Fixed-point fractional library routines. 39375 (line 157) 39376 * __multa3: Fixed-point fractional library routines. 39377 (line 173) 39378 * __multc3: Soft float library routines. 39379 (line 243) 39380 * __multf3: Soft float library routines. 39381 (line 42) 39382 * __multi3: Integer library routines. 39383 (line 44) 39384 * __muluda3: Fixed-point fractional library routines. 39385 (line 179) 39386 * __muludq3: Fixed-point fractional library routines. 39387 (line 167) 39388 * __muluha3: Fixed-point fractional library routines. 39389 (line 175) 39390 * __muluhq3: Fixed-point fractional library routines. 39391 (line 163) 39392 * __muluqq3: Fixed-point fractional library routines. 39393 (line 161) 39394 * __mulusa3: Fixed-point fractional library routines. 39395 (line 177) 39396 * __mulusq3: Fixed-point fractional library routines. 39397 (line 165) 39398 * __muluta3: Fixed-point fractional library routines. 39399 (line 181) 39400 * __mulvdi3: Integer library routines. 39401 (line 115) 39402 * __mulvsi3: Integer library routines. 39403 (line 114) 39404 * __mulxc3: Soft float library routines. 39405 (line 245) 39406 * __mulxf3: Soft float library routines. 39407 (line 44) 39408 * __nedf2: Soft float library routines. 39409 (line 200) 39410 * __negda2: Fixed-point fractional library routines. 39411 (line 299) 39412 * __negdf2: Soft float library routines. 39413 (line 56) 39414 * __negdi2: Integer library routines. 39415 (line 47) 39416 * __negdq2: Fixed-point fractional library routines. 39417 (line 289) 39418 * __negha2: Fixed-point fractional library routines. 39419 (line 297) 39420 * __neghq2: Fixed-point fractional library routines. 39421 (line 287) 39422 * __negqq2: Fixed-point fractional library routines. 39423 (line 286) 39424 * __negsa2: Fixed-point fractional library routines. 39425 (line 298) 39426 * __negsf2: Soft float library routines. 39427 (line 55) 39428 * __negsq2: Fixed-point fractional library routines. 39429 (line 288) 39430 * __negta2: Fixed-point fractional library routines. 39431 (line 300) 39432 * __negtf2: Soft float library routines. 39433 (line 57) 39434 * __negti2: Integer library routines. 39435 (line 48) 39436 * __neguda2: Fixed-point fractional library routines. 39437 (line 305) 39438 * __negudq2: Fixed-point fractional library routines. 39439 (line 296) 39440 * __neguha2: Fixed-point fractional library routines. 39441 (line 302) 39442 * __neguhq2: Fixed-point fractional library routines. 39443 (line 292) 39444 * __neguqq2: Fixed-point fractional library routines. 39445 (line 291) 39446 * __negusa2: Fixed-point fractional library routines. 39447 (line 303) 39448 * __negusq2: Fixed-point fractional library routines. 39449 (line 294) 39450 * __neguta2: Fixed-point fractional library routines. 39451 (line 307) 39452 * __negvdi2: Integer library routines. 39453 (line 119) 39454 * __negvsi2: Integer library routines. 39455 (line 118) 39456 * __negxf2: Soft float library routines. 39457 (line 58) 39458 * __nesf2: Soft float library routines. 39459 (line 199) 39460 * __netf2: Soft float library routines. 39461 (line 201) 39462 * __paritydi2: Integer library routines. 39463 (line 151) 39464 * __paritysi2: Integer library routines. 39465 (line 150) 39466 * __parityti2: Integer library routines. 39467 (line 152) 39468 * __popcountdi2: Integer library routines. 39469 (line 157) 39470 * __popcountsi2: Integer library routines. 39471 (line 156) 39472 * __popcountti2: Integer library routines. 39473 (line 158) 39474 * __powidf2: Soft float library routines. 39475 (line 233) 39476 * __powisf2: Soft float library routines. 39477 (line 232) 39478 * __powitf2: Soft float library routines. 39479 (line 234) 39480 * __powixf2: Soft float library routines. 39481 (line 235) 39482 * __satfractdadq: Fixed-point fractional library routines. 39483 (line 1153) 39484 * __satfractdaha2: Fixed-point fractional library routines. 39485 (line 1154) 39486 * __satfractdahq: Fixed-point fractional library routines. 39487 (line 1151) 39488 * __satfractdaqq: Fixed-point fractional library routines. 39489 (line 1150) 39490 * __satfractdasa2: Fixed-point fractional library routines. 39491 (line 1155) 39492 * __satfractdasq: Fixed-point fractional library routines. 39493 (line 1152) 39494 * __satfractdata2: Fixed-point fractional library routines. 39495 (line 1156) 39496 * __satfractdauda: Fixed-point fractional library routines. 39497 (line 1166) 39498 * __satfractdaudq: Fixed-point fractional library routines. 39499 (line 1162) 39500 * __satfractdauha: Fixed-point fractional library routines. 39501 (line 1164) 39502 * __satfractdauhq: Fixed-point fractional library routines. 39503 (line 1159) 39504 * __satfractdauqq: Fixed-point fractional library routines. 39505 (line 1158) 39506 * __satfractdausa: Fixed-point fractional library routines. 39507 (line 1165) 39508 * __satfractdausq: Fixed-point fractional library routines. 39509 (line 1160) 39510 * __satfractdauta: Fixed-point fractional library routines. 39511 (line 1168) 39512 * __satfractdfda: Fixed-point fractional library routines. 39513 (line 1506) 39514 * __satfractdfdq: Fixed-point fractional library routines. 39515 (line 1503) 39516 * __satfractdfha: Fixed-point fractional library routines. 39517 (line 1504) 39518 * __satfractdfhq: Fixed-point fractional library routines. 39519 (line 1501) 39520 * __satfractdfqq: Fixed-point fractional library routines. 39521 (line 1500) 39522 * __satfractdfsa: Fixed-point fractional library routines. 39523 (line 1505) 39524 * __satfractdfsq: Fixed-point fractional library routines. 39525 (line 1502) 39526 * __satfractdfta: Fixed-point fractional library routines. 39527 (line 1507) 39528 * __satfractdfuda: Fixed-point fractional library routines. 39529 (line 1515) 39530 * __satfractdfudq: Fixed-point fractional library routines. 39531 (line 1512) 39532 * __satfractdfuha: Fixed-point fractional library routines. 39533 (line 1513) 39534 * __satfractdfuhq: Fixed-point fractional library routines. 39535 (line 1509) 39536 * __satfractdfuqq: Fixed-point fractional library routines. 39537 (line 1508) 39538 * __satfractdfusa: Fixed-point fractional library routines. 39539 (line 1514) 39540 * __satfractdfusq: Fixed-point fractional library routines. 39541 (line 1510) 39542 * __satfractdfuta: Fixed-point fractional library routines. 39543 (line 1517) 39544 * __satfractdida: Fixed-point fractional library routines. 39545 (line 1456) 39546 * __satfractdidq: Fixed-point fractional library routines. 39547 (line 1453) 39548 * __satfractdiha: Fixed-point fractional library routines. 39549 (line 1454) 39550 * __satfractdihq: Fixed-point fractional library routines. 39551 (line 1451) 39552 * __satfractdiqq: Fixed-point fractional library routines. 39553 (line 1450) 39554 * __satfractdisa: Fixed-point fractional library routines. 39555 (line 1455) 39556 * __satfractdisq: Fixed-point fractional library routines. 39557 (line 1452) 39558 * __satfractdita: Fixed-point fractional library routines. 39559 (line 1457) 39560 * __satfractdiuda: Fixed-point fractional library routines. 39561 (line 1464) 39562 * __satfractdiudq: Fixed-point fractional library routines. 39563 (line 1461) 39564 * __satfractdiuha: Fixed-point fractional library routines. 39565 (line 1462) 39566 * __satfractdiuhq: Fixed-point fractional library routines. 39567 (line 1459) 39568 * __satfractdiuqq: Fixed-point fractional library routines. 39569 (line 1458) 39570 * __satfractdiusa: Fixed-point fractional library routines. 39571 (line 1463) 39572 * __satfractdiusq: Fixed-point fractional library routines. 39573 (line 1460) 39574 * __satfractdiuta: Fixed-point fractional library routines. 39575 (line 1465) 39576 * __satfractdqda: Fixed-point fractional library routines. 39577 (line 1098) 39578 * __satfractdqha: Fixed-point fractional library routines. 39579 (line 1096) 39580 * __satfractdqhq2: Fixed-point fractional library routines. 39581 (line 1094) 39582 * __satfractdqqq2: Fixed-point fractional library routines. 39583 (line 1093) 39584 * __satfractdqsa: Fixed-point fractional library routines. 39585 (line 1097) 39586 * __satfractdqsq2: Fixed-point fractional library routines. 39587 (line 1095) 39588 * __satfractdqta: Fixed-point fractional library routines. 39589 (line 1099) 39590 * __satfractdquda: Fixed-point fractional library routines. 39591 (line 1111) 39592 * __satfractdqudq: Fixed-point fractional library routines. 39593 (line 1106) 39594 * __satfractdquha: Fixed-point fractional library routines. 39595 (line 1108) 39596 * __satfractdquhq: Fixed-point fractional library routines. 39597 (line 1102) 39598 * __satfractdquqq: Fixed-point fractional library routines. 39599 (line 1101) 39600 * __satfractdqusa: Fixed-point fractional library routines. 39601 (line 1109) 39602 * __satfractdqusq: Fixed-point fractional library routines. 39603 (line 1104) 39604 * __satfractdquta: Fixed-point fractional library routines. 39605 (line 1113) 39606 * __satfracthada2: Fixed-point fractional library routines. 39607 (line 1119) 39608 * __satfracthadq: Fixed-point fractional library routines. 39609 (line 1117) 39610 * __satfracthahq: Fixed-point fractional library routines. 39611 (line 1115) 39612 * __satfracthaqq: Fixed-point fractional library routines. 39613 (line 1114) 39614 * __satfracthasa2: Fixed-point fractional library routines. 39615 (line 1118) 39616 * __satfracthasq: Fixed-point fractional library routines. 39617 (line 1116) 39618 * __satfracthata2: Fixed-point fractional library routines. 39619 (line 1120) 39620 * __satfracthauda: Fixed-point fractional library routines. 39621 (line 1132) 39622 * __satfracthaudq: Fixed-point fractional library routines. 39623 (line 1127) 39624 * __satfracthauha: Fixed-point fractional library routines. 39625 (line 1129) 39626 * __satfracthauhq: Fixed-point fractional library routines. 39627 (line 1123) 39628 * __satfracthauqq: Fixed-point fractional library routines. 39629 (line 1122) 39630 * __satfracthausa: Fixed-point fractional library routines. 39631 (line 1130) 39632 * __satfracthausq: Fixed-point fractional library routines. 39633 (line 1125) 39634 * __satfracthauta: Fixed-point fractional library routines. 39635 (line 1134) 39636 * __satfracthida: Fixed-point fractional library routines. 39637 (line 1424) 39638 * __satfracthidq: Fixed-point fractional library routines. 39639 (line 1421) 39640 * __satfracthiha: Fixed-point fractional library routines. 39641 (line 1422) 39642 * __satfracthihq: Fixed-point fractional library routines. 39643 (line 1419) 39644 * __satfracthiqq: Fixed-point fractional library routines. 39645 (line 1418) 39646 * __satfracthisa: Fixed-point fractional library routines. 39647 (line 1423) 39648 * __satfracthisq: Fixed-point fractional library routines. 39649 (line 1420) 39650 * __satfracthita: Fixed-point fractional library routines. 39651 (line 1425) 39652 * __satfracthiuda: Fixed-point fractional library routines. 39653 (line 1432) 39654 * __satfracthiudq: Fixed-point fractional library routines. 39655 (line 1429) 39656 * __satfracthiuha: Fixed-point fractional library routines. 39657 (line 1430) 39658 * __satfracthiuhq: Fixed-point fractional library routines. 39659 (line 1427) 39660 * __satfracthiuqq: Fixed-point fractional library routines. 39661 (line 1426) 39662 * __satfracthiusa: Fixed-point fractional library routines. 39663 (line 1431) 39664 * __satfracthiusq: Fixed-point fractional library routines. 39665 (line 1428) 39666 * __satfracthiuta: Fixed-point fractional library routines. 39667 (line 1433) 39668 * __satfracthqda: Fixed-point fractional library routines. 39669 (line 1064) 39670 * __satfracthqdq2: Fixed-point fractional library routines. 39671 (line 1061) 39672 * __satfracthqha: Fixed-point fractional library routines. 39673 (line 1062) 39674 * __satfracthqqq2: Fixed-point fractional library routines. 39675 (line 1059) 39676 * __satfracthqsa: Fixed-point fractional library routines. 39677 (line 1063) 39678 * __satfracthqsq2: Fixed-point fractional library routines. 39679 (line 1060) 39680 * __satfracthqta: Fixed-point fractional library routines. 39681 (line 1065) 39682 * __satfracthquda: Fixed-point fractional library routines. 39683 (line 1072) 39684 * __satfracthqudq: Fixed-point fractional library routines. 39685 (line 1069) 39686 * __satfracthquha: Fixed-point fractional library routines. 39687 (line 1070) 39688 * __satfracthquhq: Fixed-point fractional library routines. 39689 (line 1067) 39690 * __satfracthquqq: Fixed-point fractional library routines. 39691 (line 1066) 39692 * __satfracthqusa: Fixed-point fractional library routines. 39693 (line 1071) 39694 * __satfracthqusq: Fixed-point fractional library routines. 39695 (line 1068) 39696 * __satfracthquta: Fixed-point fractional library routines. 39697 (line 1073) 39698 * __satfractqida: Fixed-point fractional library routines. 39699 (line 1402) 39700 * __satfractqidq: Fixed-point fractional library routines. 39701 (line 1399) 39702 * __satfractqiha: Fixed-point fractional library routines. 39703 (line 1400) 39704 * __satfractqihq: Fixed-point fractional library routines. 39705 (line 1397) 39706 * __satfractqiqq: Fixed-point fractional library routines. 39707 (line 1396) 39708 * __satfractqisa: Fixed-point fractional library routines. 39709 (line 1401) 39710 * __satfractqisq: Fixed-point fractional library routines. 39711 (line 1398) 39712 * __satfractqita: Fixed-point fractional library routines. 39713 (line 1403) 39714 * __satfractqiuda: Fixed-point fractional library routines. 39715 (line 1415) 39716 * __satfractqiudq: Fixed-point fractional library routines. 39717 (line 1410) 39718 * __satfractqiuha: Fixed-point fractional library routines. 39719 (line 1412) 39720 * __satfractqiuhq: Fixed-point fractional library routines. 39721 (line 1406) 39722 * __satfractqiuqq: Fixed-point fractional library routines. 39723 (line 1405) 39724 * __satfractqiusa: Fixed-point fractional library routines. 39725 (line 1413) 39726 * __satfractqiusq: Fixed-point fractional library routines. 39727 (line 1408) 39728 * __satfractqiuta: Fixed-point fractional library routines. 39729 (line 1417) 39730 * __satfractqqda: Fixed-point fractional library routines. 39731 (line 1043) 39732 * __satfractqqdq2: Fixed-point fractional library routines. 39733 (line 1040) 39734 * __satfractqqha: Fixed-point fractional library routines. 39735 (line 1041) 39736 * __satfractqqhq2: Fixed-point fractional library routines. 39737 (line 1038) 39738 * __satfractqqsa: Fixed-point fractional library routines. 39739 (line 1042) 39740 * __satfractqqsq2: Fixed-point fractional library routines. 39741 (line 1039) 39742 * __satfractqqta: Fixed-point fractional library routines. 39743 (line 1044) 39744 * __satfractqquda: Fixed-point fractional library routines. 39745 (line 1056) 39746 * __satfractqqudq: Fixed-point fractional library routines. 39747 (line 1051) 39748 * __satfractqquha: Fixed-point fractional library routines. 39749 (line 1053) 39750 * __satfractqquhq: Fixed-point fractional library routines. 39751 (line 1047) 39752 * __satfractqquqq: Fixed-point fractional library routines. 39753 (line 1046) 39754 * __satfractqqusa: Fixed-point fractional library routines. 39755 (line 1054) 39756 * __satfractqqusq: Fixed-point fractional library routines. 39757 (line 1049) 39758 * __satfractqquta: Fixed-point fractional library routines. 39759 (line 1058) 39760 * __satfractsada2: Fixed-point fractional library routines. 39761 (line 1140) 39762 * __satfractsadq: Fixed-point fractional library routines. 39763 (line 1138) 39764 * __satfractsaha2: Fixed-point fractional library routines. 39765 (line 1139) 39766 * __satfractsahq: Fixed-point fractional library routines. 39767 (line 1136) 39768 * __satfractsaqq: Fixed-point fractional library routines. 39769 (line 1135) 39770 * __satfractsasq: Fixed-point fractional library routines. 39771 (line 1137) 39772 * __satfractsata2: Fixed-point fractional library routines. 39773 (line 1141) 39774 * __satfractsauda: Fixed-point fractional library routines. 39775 (line 1148) 39776 * __satfractsaudq: Fixed-point fractional library routines. 39777 (line 1145) 39778 * __satfractsauha: Fixed-point fractional library routines. 39779 (line 1146) 39780 * __satfractsauhq: Fixed-point fractional library routines. 39781 (line 1143) 39782 * __satfractsauqq: Fixed-point fractional library routines. 39783 (line 1142) 39784 * __satfractsausa: Fixed-point fractional library routines. 39785 (line 1147) 39786 * __satfractsausq: Fixed-point fractional library routines. 39787 (line 1144) 39788 * __satfractsauta: Fixed-point fractional library routines. 39789 (line 1149) 39790 * __satfractsfda: Fixed-point fractional library routines. 39791 (line 1490) 39792 * __satfractsfdq: Fixed-point fractional library routines. 39793 (line 1487) 39794 * __satfractsfha: Fixed-point fractional library routines. 39795 (line 1488) 39796 * __satfractsfhq: Fixed-point fractional library routines. 39797 (line 1485) 39798 * __satfractsfqq: Fixed-point fractional library routines. 39799 (line 1484) 39800 * __satfractsfsa: Fixed-point fractional library routines. 39801 (line 1489) 39802 * __satfractsfsq: Fixed-point fractional library routines. 39803 (line 1486) 39804 * __satfractsfta: Fixed-point fractional library routines. 39805 (line 1491) 39806 * __satfractsfuda: Fixed-point fractional library routines. 39807 (line 1498) 39808 * __satfractsfudq: Fixed-point fractional library routines. 39809 (line 1495) 39810 * __satfractsfuha: Fixed-point fractional library routines. 39811 (line 1496) 39812 * __satfractsfuhq: Fixed-point fractional library routines. 39813 (line 1493) 39814 * __satfractsfuqq: Fixed-point fractional library routines. 39815 (line 1492) 39816 * __satfractsfusa: Fixed-point fractional library routines. 39817 (line 1497) 39818 * __satfractsfusq: Fixed-point fractional library routines. 39819 (line 1494) 39820 * __satfractsfuta: Fixed-point fractional library routines. 39821 (line 1499) 39822 * __satfractsida: Fixed-point fractional library routines. 39823 (line 1440) 39824 * __satfractsidq: Fixed-point fractional library routines. 39825 (line 1437) 39826 * __satfractsiha: Fixed-point fractional library routines. 39827 (line 1438) 39828 * __satfractsihq: Fixed-point fractional library routines. 39829 (line 1435) 39830 * __satfractsiqq: Fixed-point fractional library routines. 39831 (line 1434) 39832 * __satfractsisa: Fixed-point fractional library routines. 39833 (line 1439) 39834 * __satfractsisq: Fixed-point fractional library routines. 39835 (line 1436) 39836 * __satfractsita: Fixed-point fractional library routines. 39837 (line 1441) 39838 * __satfractsiuda: Fixed-point fractional library routines. 39839 (line 1448) 39840 * __satfractsiudq: Fixed-point fractional library routines. 39841 (line 1445) 39842 * __satfractsiuha: Fixed-point fractional library routines. 39843 (line 1446) 39844 * __satfractsiuhq: Fixed-point fractional library routines. 39845 (line 1443) 39846 * __satfractsiuqq: Fixed-point fractional library routines. 39847 (line 1442) 39848 * __satfractsiusa: Fixed-point fractional library routines. 39849 (line 1447) 39850 * __satfractsiusq: Fixed-point fractional library routines. 39851 (line 1444) 39852 * __satfractsiuta: Fixed-point fractional library routines. 39853 (line 1449) 39854 * __satfractsqda: Fixed-point fractional library routines. 39855 (line 1079) 39856 * __satfractsqdq2: Fixed-point fractional library routines. 39857 (line 1076) 39858 * __satfractsqha: Fixed-point fractional library routines. 39859 (line 1077) 39860 * __satfractsqhq2: Fixed-point fractional library routines. 39861 (line 1075) 39862 * __satfractsqqq2: Fixed-point fractional library routines. 39863 (line 1074) 39864 * __satfractsqsa: Fixed-point fractional library routines. 39865 (line 1078) 39866 * __satfractsqta: Fixed-point fractional library routines. 39867 (line 1080) 39868 * __satfractsquda: Fixed-point fractional library routines. 39869 (line 1090) 39870 * __satfractsqudq: Fixed-point fractional library routines. 39871 (line 1086) 39872 * __satfractsquha: Fixed-point fractional library routines. 39873 (line 1088) 39874 * __satfractsquhq: Fixed-point fractional library routines. 39875 (line 1083) 39876 * __satfractsquqq: Fixed-point fractional library routines. 39877 (line 1082) 39878 * __satfractsqusa: Fixed-point fractional library routines. 39879 (line 1089) 39880 * __satfractsqusq: Fixed-point fractional library routines. 39881 (line 1084) 39882 * __satfractsquta: Fixed-point fractional library routines. 39883 (line 1092) 39884 * __satfracttada2: Fixed-point fractional library routines. 39885 (line 1175) 39886 * __satfracttadq: Fixed-point fractional library routines. 39887 (line 1172) 39888 * __satfracttaha2: Fixed-point fractional library routines. 39889 (line 1173) 39890 * __satfracttahq: Fixed-point fractional library routines. 39891 (line 1170) 39892 * __satfracttaqq: Fixed-point fractional library routines. 39893 (line 1169) 39894 * __satfracttasa2: Fixed-point fractional library routines. 39895 (line 1174) 39896 * __satfracttasq: Fixed-point fractional library routines. 39897 (line 1171) 39898 * __satfracttauda: Fixed-point fractional library routines. 39899 (line 1187) 39900 * __satfracttaudq: Fixed-point fractional library routines. 39901 (line 1182) 39902 * __satfracttauha: Fixed-point fractional library routines. 39903 (line 1184) 39904 * __satfracttauhq: Fixed-point fractional library routines. 39905 (line 1178) 39906 * __satfracttauqq: Fixed-point fractional library routines. 39907 (line 1177) 39908 * __satfracttausa: Fixed-point fractional library routines. 39909 (line 1185) 39910 * __satfracttausq: Fixed-point fractional library routines. 39911 (line 1180) 39912 * __satfracttauta: Fixed-point fractional library routines. 39913 (line 1189) 39914 * __satfracttida: Fixed-point fractional library routines. 39915 (line 1472) 39916 * __satfracttidq: Fixed-point fractional library routines. 39917 (line 1469) 39918 * __satfracttiha: Fixed-point fractional library routines. 39919 (line 1470) 39920 * __satfracttihq: Fixed-point fractional library routines. 39921 (line 1467) 39922 * __satfracttiqq: Fixed-point fractional library routines. 39923 (line 1466) 39924 * __satfracttisa: Fixed-point fractional library routines. 39925 (line 1471) 39926 * __satfracttisq: Fixed-point fractional library routines. 39927 (line 1468) 39928 * __satfracttita: Fixed-point fractional library routines. 39929 (line 1473) 39930 * __satfracttiuda: Fixed-point fractional library routines. 39931 (line 1481) 39932 * __satfracttiudq: Fixed-point fractional library routines. 39933 (line 1478) 39934 * __satfracttiuha: Fixed-point fractional library routines. 39935 (line 1479) 39936 * __satfracttiuhq: Fixed-point fractional library routines. 39937 (line 1475) 39938 * __satfracttiuqq: Fixed-point fractional library routines. 39939 (line 1474) 39940 * __satfracttiusa: Fixed-point fractional library routines. 39941 (line 1480) 39942 * __satfracttiusq: Fixed-point fractional library routines. 39943 (line 1476) 39944 * __satfracttiuta: Fixed-point fractional library routines. 39945 (line 1483) 39946 * __satfractudada: Fixed-point fractional library routines. 39947 (line 1351) 39948 * __satfractudadq: Fixed-point fractional library routines. 39949 (line 1347) 39950 * __satfractudaha: Fixed-point fractional library routines. 39951 (line 1349) 39952 * __satfractudahq: Fixed-point fractional library routines. 39953 (line 1344) 39954 * __satfractudaqq: Fixed-point fractional library routines. 39955 (line 1343) 39956 * __satfractudasa: Fixed-point fractional library routines. 39957 (line 1350) 39958 * __satfractudasq: Fixed-point fractional library routines. 39959 (line 1345) 39960 * __satfractudata: Fixed-point fractional library routines. 39961 (line 1353) 39962 * __satfractudaudq: Fixed-point fractional library routines. 39963 (line 1361) 39964 * __satfractudauha2: Fixed-point fractional library routines. 39965 (line 1363) 39966 * __satfractudauhq: Fixed-point fractional library routines. 39967 (line 1357) 39968 * __satfractudauqq: Fixed-point fractional library routines. 39969 (line 1355) 39970 * __satfractudausa2: Fixed-point fractional library routines. 39971 (line 1365) 39972 * __satfractudausq: Fixed-point fractional library routines. 39973 (line 1359) 39974 * __satfractudauta2: Fixed-point fractional library routines. 39975 (line 1367) 39976 * __satfractudqda: Fixed-point fractional library routines. 39977 (line 1276) 39978 * __satfractudqdq: Fixed-point fractional library routines. 39979 (line 1271) 39980 * __satfractudqha: Fixed-point fractional library routines. 39981 (line 1273) 39982 * __satfractudqhq: Fixed-point fractional library routines. 39983 (line 1267) 39984 * __satfractudqqq: Fixed-point fractional library routines. 39985 (line 1266) 39986 * __satfractudqsa: Fixed-point fractional library routines. 39987 (line 1274) 39988 * __satfractudqsq: Fixed-point fractional library routines. 39989 (line 1269) 39990 * __satfractudqta: Fixed-point fractional library routines. 39991 (line 1278) 39992 * __satfractudquda: Fixed-point fractional library routines. 39993 (line 1290) 39994 * __satfractudquha: Fixed-point fractional library routines. 39995 (line 1286) 39996 * __satfractudquhq2: Fixed-point fractional library routines. 39997 (line 1282) 39998 * __satfractudquqq2: Fixed-point fractional library routines. 39999 (line 1280) 40000 * __satfractudqusa: Fixed-point fractional library routines. 40001 (line 1288) 40002 * __satfractudqusq2: Fixed-point fractional library routines. 40003 (line 1284) 40004 * __satfractudquta: Fixed-point fractional library routines. 40005 (line 1292) 40006 * __satfractuhada: Fixed-point fractional library routines. 40007 (line 1304) 40008 * __satfractuhadq: Fixed-point fractional library routines. 40009 (line 1299) 40010 * __satfractuhaha: Fixed-point fractional library routines. 40011 (line 1301) 40012 * __satfractuhahq: Fixed-point fractional library routines. 40013 (line 1295) 40014 * __satfractuhaqq: Fixed-point fractional library routines. 40015 (line 1294) 40016 * __satfractuhasa: Fixed-point fractional library routines. 40017 (line 1302) 40018 * __satfractuhasq: Fixed-point fractional library routines. 40019 (line 1297) 40020 * __satfractuhata: Fixed-point fractional library routines. 40021 (line 1306) 40022 * __satfractuhauda2: Fixed-point fractional library routines. 40023 (line 1318) 40024 * __satfractuhaudq: Fixed-point fractional library routines. 40025 (line 1314) 40026 * __satfractuhauhq: Fixed-point fractional library routines. 40027 (line 1310) 40028 * __satfractuhauqq: Fixed-point fractional library routines. 40029 (line 1308) 40030 * __satfractuhausa2: Fixed-point fractional library routines. 40031 (line 1316) 40032 * __satfractuhausq: Fixed-point fractional library routines. 40033 (line 1312) 40034 * __satfractuhauta2: Fixed-point fractional library routines. 40035 (line 1320) 40036 * __satfractuhqda: Fixed-point fractional library routines. 40037 (line 1224) 40038 * __satfractuhqdq: Fixed-point fractional library routines. 40039 (line 1221) 40040 * __satfractuhqha: Fixed-point fractional library routines. 40041 (line 1222) 40042 * __satfractuhqhq: Fixed-point fractional library routines. 40043 (line 1219) 40044 * __satfractuhqqq: Fixed-point fractional library routines. 40045 (line 1218) 40046 * __satfractuhqsa: Fixed-point fractional library routines. 40047 (line 1223) 40048 * __satfractuhqsq: Fixed-point fractional library routines. 40049 (line 1220) 40050 * __satfractuhqta: Fixed-point fractional library routines. 40051 (line 1225) 40052 * __satfractuhquda: Fixed-point fractional library routines. 40053 (line 1236) 40054 * __satfractuhqudq2: Fixed-point fractional library routines. 40055 (line 1231) 40056 * __satfractuhquha: Fixed-point fractional library routines. 40057 (line 1233) 40058 * __satfractuhquqq2: Fixed-point fractional library routines. 40059 (line 1227) 40060 * __satfractuhqusa: Fixed-point fractional library routines. 40061 (line 1234) 40062 * __satfractuhqusq2: Fixed-point fractional library routines. 40063 (line 1229) 40064 * __satfractuhquta: Fixed-point fractional library routines. 40065 (line 1238) 40066 * __satfractunsdida: Fixed-point fractional library routines. 40067 (line 1834) 40068 * __satfractunsdidq: Fixed-point fractional library routines. 40069 (line 1831) 40070 * __satfractunsdiha: Fixed-point fractional library routines. 40071 (line 1832) 40072 * __satfractunsdihq: Fixed-point fractional library routines. 40073 (line 1828) 40074 * __satfractunsdiqq: Fixed-point fractional library routines. 40075 (line 1827) 40076 * __satfractunsdisa: Fixed-point fractional library routines. 40077 (line 1833) 40078 * __satfractunsdisq: Fixed-point fractional library routines. 40079 (line 1829) 40080 * __satfractunsdita: Fixed-point fractional library routines. 40081 (line 1836) 40082 * __satfractunsdiuda: Fixed-point fractional library routines. 40083 (line 1850) 40084 * __satfractunsdiudq: Fixed-point fractional library routines. 40085 (line 1844) 40086 * __satfractunsdiuha: Fixed-point fractional library routines. 40087 (line 1846) 40088 * __satfractunsdiuhq: Fixed-point fractional library routines. 40089 (line 1840) 40090 * __satfractunsdiuqq: Fixed-point fractional library routines. 40091 (line 1838) 40092 * __satfractunsdiusa: Fixed-point fractional library routines. 40093 (line 1848) 40094 * __satfractunsdiusq: Fixed-point fractional library routines. 40095 (line 1842) 40096 * __satfractunsdiuta: Fixed-point fractional library routines. 40097 (line 1852) 40098 * __satfractunshida: Fixed-point fractional library routines. 40099 (line 1786) 40100 * __satfractunshidq: Fixed-point fractional library routines. 40101 (line 1783) 40102 * __satfractunshiha: Fixed-point fractional library routines. 40103 (line 1784) 40104 * __satfractunshihq: Fixed-point fractional library routines. 40105 (line 1780) 40106 * __satfractunshiqq: Fixed-point fractional library routines. 40107 (line 1779) 40108 * __satfractunshisa: Fixed-point fractional library routines. 40109 (line 1785) 40110 * __satfractunshisq: Fixed-point fractional library routines. 40111 (line 1781) 40112 * __satfractunshita: Fixed-point fractional library routines. 40113 (line 1788) 40114 * __satfractunshiuda: Fixed-point fractional library routines. 40115 (line 1802) 40116 * __satfractunshiudq: Fixed-point fractional library routines. 40117 (line 1796) 40118 * __satfractunshiuha: Fixed-point fractional library routines. 40119 (line 1798) 40120 * __satfractunshiuhq: Fixed-point fractional library routines. 40121 (line 1792) 40122 * __satfractunshiuqq: Fixed-point fractional library routines. 40123 (line 1790) 40124 * __satfractunshiusa: Fixed-point fractional library routines. 40125 (line 1800) 40126 * __satfractunshiusq: Fixed-point fractional library routines. 40127 (line 1794) 40128 * __satfractunshiuta: Fixed-point fractional library routines. 40129 (line 1804) 40130 * __satfractunsqida: Fixed-point fractional library routines. 40131 (line 1760) 40132 * __satfractunsqidq: Fixed-point fractional library routines. 40133 (line 1757) 40134 * __satfractunsqiha: Fixed-point fractional library routines. 40135 (line 1758) 40136 * __satfractunsqihq: Fixed-point fractional library routines. 40137 (line 1754) 40138 * __satfractunsqiqq: Fixed-point fractional library routines. 40139 (line 1753) 40140 * __satfractunsqisa: Fixed-point fractional library routines. 40141 (line 1759) 40142 * __satfractunsqisq: Fixed-point fractional library routines. 40143 (line 1755) 40144 * __satfractunsqita: Fixed-point fractional library routines. 40145 (line 1762) 40146 * __satfractunsqiuda: Fixed-point fractional library routines. 40147 (line 1776) 40148 * __satfractunsqiudq: Fixed-point fractional library routines. 40149 (line 1770) 40150 * __satfractunsqiuha: Fixed-point fractional library routines. 40151 (line 1772) 40152 * __satfractunsqiuhq: Fixed-point fractional library routines. 40153 (line 1766) 40154 * __satfractunsqiuqq: Fixed-point fractional library routines. 40155 (line 1764) 40156 * __satfractunsqiusa: Fixed-point fractional library routines. 40157 (line 1774) 40158 * __satfractunsqiusq: Fixed-point fractional library routines. 40159 (line 1768) 40160 * __satfractunsqiuta: Fixed-point fractional library routines. 40161 (line 1778) 40162 * __satfractunssida: Fixed-point fractional library routines. 40163 (line 1811) 40164 * __satfractunssidq: Fixed-point fractional library routines. 40165 (line 1808) 40166 * __satfractunssiha: Fixed-point fractional library routines. 40167 (line 1809) 40168 * __satfractunssihq: Fixed-point fractional library routines. 40169 (line 1806) 40170 * __satfractunssiqq: Fixed-point fractional library routines. 40171 (line 1805) 40172 * __satfractunssisa: Fixed-point fractional library routines. 40173 (line 1810) 40174 * __satfractunssisq: Fixed-point fractional library routines. 40175 (line 1807) 40176 * __satfractunssita: Fixed-point fractional library routines. 40177 (line 1812) 40178 * __satfractunssiuda: Fixed-point fractional library routines. 40179 (line 1824) 40180 * __satfractunssiudq: Fixed-point fractional library routines. 40181 (line 1819) 40182 * __satfractunssiuha: Fixed-point fractional library routines. 40183 (line 1821) 40184 * __satfractunssiuhq: Fixed-point fractional library routines. 40185 (line 1815) 40186 * __satfractunssiuqq: Fixed-point fractional library routines. 40187 (line 1814) 40188 * __satfractunssiusa: Fixed-point fractional library routines. 40189 (line 1822) 40190 * __satfractunssiusq: Fixed-point fractional library routines. 40191 (line 1817) 40192 * __satfractunssiuta: Fixed-point fractional library routines. 40193 (line 1826) 40194 * __satfractunstida: Fixed-point fractional library routines. 40195 (line 1864) 40196 * __satfractunstidq: Fixed-point fractional library routines. 40197 (line 1859) 40198 * __satfractunstiha: Fixed-point fractional library routines. 40199 (line 1861) 40200 * __satfractunstihq: Fixed-point fractional library routines. 40201 (line 1855) 40202 * __satfractunstiqq: Fixed-point fractional library routines. 40203 (line 1854) 40204 * __satfractunstisa: Fixed-point fractional library routines. 40205 (line 1862) 40206 * __satfractunstisq: Fixed-point fractional library routines. 40207 (line 1857) 40208 * __satfractunstita: Fixed-point fractional library routines. 40209 (line 1866) 40210 * __satfractunstiuda: Fixed-point fractional library routines. 40211 (line 1880) 40212 * __satfractunstiudq: Fixed-point fractional library routines. 40213 (line 1874) 40214 * __satfractunstiuha: Fixed-point fractional library routines. 40215 (line 1876) 40216 * __satfractunstiuhq: Fixed-point fractional library routines. 40217 (line 1870) 40218 * __satfractunstiuqq: Fixed-point fractional library routines. 40219 (line 1868) 40220 * __satfractunstiusa: Fixed-point fractional library routines. 40221 (line 1878) 40222 * __satfractunstiusq: Fixed-point fractional library routines. 40223 (line 1872) 40224 * __satfractunstiuta: Fixed-point fractional library routines. 40225 (line 1882) 40226 * __satfractuqqda: Fixed-point fractional library routines. 40227 (line 1201) 40228 * __satfractuqqdq: Fixed-point fractional library routines. 40229 (line 1196) 40230 * __satfractuqqha: Fixed-point fractional library routines. 40231 (line 1198) 40232 * __satfractuqqhq: Fixed-point fractional library routines. 40233 (line 1192) 40234 * __satfractuqqqq: Fixed-point fractional library routines. 40235 (line 1191) 40236 * __satfractuqqsa: Fixed-point fractional library routines. 40237 (line 1199) 40238 * __satfractuqqsq: Fixed-point fractional library routines. 40239 (line 1194) 40240 * __satfractuqqta: Fixed-point fractional library routines. 40241 (line 1203) 40242 * __satfractuqquda: Fixed-point fractional library routines. 40243 (line 1215) 40244 * __satfractuqqudq2: Fixed-point fractional library routines. 40245 (line 1209) 40246 * __satfractuqquha: Fixed-point fractional library routines. 40247 (line 1211) 40248 * __satfractuqquhq2: Fixed-point fractional library routines. 40249 (line 1205) 40250 * __satfractuqqusa: Fixed-point fractional library routines. 40251 (line 1213) 40252 * __satfractuqqusq2: Fixed-point fractional library routines. 40253 (line 1207) 40254 * __satfractuqquta: Fixed-point fractional library routines. 40255 (line 1217) 40256 * __satfractusada: Fixed-point fractional library routines. 40257 (line 1327) 40258 * __satfractusadq: Fixed-point fractional library routines. 40259 (line 1324) 40260 * __satfractusaha: Fixed-point fractional library routines. 40261 (line 1325) 40262 * __satfractusahq: Fixed-point fractional library routines. 40263 (line 1322) 40264 * __satfractusaqq: Fixed-point fractional library routines. 40265 (line 1321) 40266 * __satfractusasa: Fixed-point fractional library routines. 40267 (line 1326) 40268 * __satfractusasq: Fixed-point fractional library routines. 40269 (line 1323) 40270 * __satfractusata: Fixed-point fractional library routines. 40271 (line 1328) 40272 * __satfractusauda2: Fixed-point fractional library routines. 40273 (line 1339) 40274 * __satfractusaudq: Fixed-point fractional library routines. 40275 (line 1335) 40276 * __satfractusauha2: Fixed-point fractional library routines. 40277 (line 1337) 40278 * __satfractusauhq: Fixed-point fractional library routines. 40279 (line 1331) 40280 * __satfractusauqq: Fixed-point fractional library routines. 40281 (line 1330) 40282 * __satfractusausq: Fixed-point fractional library routines. 40283 (line 1333) 40284 * __satfractusauta2: Fixed-point fractional library routines. 40285 (line 1341) 40286 * __satfractusqda: Fixed-point fractional library routines. 40287 (line 1248) 40288 * __satfractusqdq: Fixed-point fractional library routines. 40289 (line 1244) 40290 * __satfractusqha: Fixed-point fractional library routines. 40291 (line 1246) 40292 * __satfractusqhq: Fixed-point fractional library routines. 40293 (line 1241) 40294 * __satfractusqqq: Fixed-point fractional library routines. 40295 (line 1240) 40296 * __satfractusqsa: Fixed-point fractional library routines. 40297 (line 1247) 40298 * __satfractusqsq: Fixed-point fractional library routines. 40299 (line 1242) 40300 * __satfractusqta: Fixed-point fractional library routines. 40301 (line 1250) 40302 * __satfractusquda: Fixed-point fractional library routines. 40303 (line 1262) 40304 * __satfractusqudq2: Fixed-point fractional library routines. 40305 (line 1256) 40306 * __satfractusquha: Fixed-point fractional library routines. 40307 (line 1258) 40308 * __satfractusquhq2: Fixed-point fractional library routines. 40309 (line 1254) 40310 * __satfractusquqq2: Fixed-point fractional library routines. 40311 (line 1252) 40312 * __satfractusqusa: Fixed-point fractional library routines. 40313 (line 1260) 40314 * __satfractusquta: Fixed-point fractional library routines. 40315 (line 1264) 40316 * __satfractutada: Fixed-point fractional library routines. 40317 (line 1379) 40318 * __satfractutadq: Fixed-point fractional library routines. 40319 (line 1374) 40320 * __satfractutaha: Fixed-point fractional library routines. 40321 (line 1376) 40322 * __satfractutahq: Fixed-point fractional library routines. 40323 (line 1370) 40324 * __satfractutaqq: Fixed-point fractional library routines. 40325 (line 1369) 40326 * __satfractutasa: Fixed-point fractional library routines. 40327 (line 1377) 40328 * __satfractutasq: Fixed-point fractional library routines. 40329 (line 1372) 40330 * __satfractutata: Fixed-point fractional library routines. 40331 (line 1381) 40332 * __satfractutauda2: Fixed-point fractional library routines. 40333 (line 1395) 40334 * __satfractutaudq: Fixed-point fractional library routines. 40335 (line 1389) 40336 * __satfractutauha2: Fixed-point fractional library routines. 40337 (line 1391) 40338 * __satfractutauhq: Fixed-point fractional library routines. 40339 (line 1385) 40340 * __satfractutauqq: Fixed-point fractional library routines. 40341 (line 1383) 40342 * __satfractutausa2: Fixed-point fractional library routines. 40343 (line 1393) 40344 * __satfractutausq: Fixed-point fractional library routines. 40345 (line 1387) 40346 * __ssaddda3: Fixed-point fractional library routines. 40347 (line 67) 40348 * __ssadddq3: Fixed-point fractional library routines. 40349 (line 63) 40350 * __ssaddha3: Fixed-point fractional library routines. 40351 (line 65) 40352 * __ssaddhq3: Fixed-point fractional library routines. 40353 (line 60) 40354 * __ssaddqq3: Fixed-point fractional library routines. 40355 (line 59) 40356 * __ssaddsa3: Fixed-point fractional library routines. 40357 (line 66) 40358 * __ssaddsq3: Fixed-point fractional library routines. 40359 (line 61) 40360 * __ssaddta3: Fixed-point fractional library routines. 40361 (line 69) 40362 * __ssashlda3: Fixed-point fractional library routines. 40363 (line 402) 40364 * __ssashldq3: Fixed-point fractional library routines. 40365 (line 399) 40366 * __ssashlha3: Fixed-point fractional library routines. 40367 (line 400) 40368 * __ssashlhq3: Fixed-point fractional library routines. 40369 (line 396) 40370 * __ssashlsa3: Fixed-point fractional library routines. 40371 (line 401) 40372 * __ssashlsq3: Fixed-point fractional library routines. 40373 (line 397) 40374 * __ssashlta3: Fixed-point fractional library routines. 40375 (line 404) 40376 * __ssdivda3: Fixed-point fractional library routines. 40377 (line 261) 40378 * __ssdivdq3: Fixed-point fractional library routines. 40379 (line 257) 40380 * __ssdivha3: Fixed-point fractional library routines. 40381 (line 259) 40382 * __ssdivhq3: Fixed-point fractional library routines. 40383 (line 254) 40384 * __ssdivqq3: Fixed-point fractional library routines. 40385 (line 253) 40386 * __ssdivsa3: Fixed-point fractional library routines. 40387 (line 260) 40388 * __ssdivsq3: Fixed-point fractional library routines. 40389 (line 255) 40390 * __ssdivta3: Fixed-point fractional library routines. 40391 (line 263) 40392 * __ssmulda3: Fixed-point fractional library routines. 40393 (line 193) 40394 * __ssmuldq3: Fixed-point fractional library routines. 40395 (line 189) 40396 * __ssmulha3: Fixed-point fractional library routines. 40397 (line 191) 40398 * __ssmulhq3: Fixed-point fractional library routines. 40399 (line 186) 40400 * __ssmulqq3: Fixed-point fractional library routines. 40401 (line 185) 40402 * __ssmulsa3: Fixed-point fractional library routines. 40403 (line 192) 40404 * __ssmulsq3: Fixed-point fractional library routines. 40405 (line 187) 40406 * __ssmulta3: Fixed-point fractional library routines. 40407 (line 195) 40408 * __ssnegda2: Fixed-point fractional library routines. 40409 (line 316) 40410 * __ssnegdq2: Fixed-point fractional library routines. 40411 (line 313) 40412 * __ssnegha2: Fixed-point fractional library routines. 40413 (line 314) 40414 * __ssneghq2: Fixed-point fractional library routines. 40415 (line 311) 40416 * __ssnegqq2: Fixed-point fractional library routines. 40417 (line 310) 40418 * __ssnegsa2: Fixed-point fractional library routines. 40419 (line 315) 40420 * __ssnegsq2: Fixed-point fractional library routines. 40421 (line 312) 40422 * __ssnegta2: Fixed-point fractional library routines. 40423 (line 317) 40424 * __sssubda3: Fixed-point fractional library routines. 40425 (line 129) 40426 * __sssubdq3: Fixed-point fractional library routines. 40427 (line 125) 40428 * __sssubha3: Fixed-point fractional library routines. 40429 (line 127) 40430 * __sssubhq3: Fixed-point fractional library routines. 40431 (line 122) 40432 * __sssubqq3: Fixed-point fractional library routines. 40433 (line 121) 40434 * __sssubsa3: Fixed-point fractional library routines. 40435 (line 128) 40436 * __sssubsq3: Fixed-point fractional library routines. 40437 (line 123) 40438 * __sssubta3: Fixed-point fractional library routines. 40439 (line 131) 40440 * __subda3: Fixed-point fractional library routines. 40441 (line 107) 40442 * __subdf3: Soft float library routines. 40443 (line 31) 40444 * __subdq3: Fixed-point fractional library routines. 40445 (line 95) 40446 * __subha3: Fixed-point fractional library routines. 40447 (line 105) 40448 * __subhq3: Fixed-point fractional library routines. 40449 (line 92) 40450 * __subqq3: Fixed-point fractional library routines. 40451 (line 91) 40452 * __subsa3: Fixed-point fractional library routines. 40453 (line 106) 40454 * __subsf3: Soft float library routines. 40455 (line 30) 40456 * __subsq3: Fixed-point fractional library routines. 40457 (line 93) 40458 * __subta3: Fixed-point fractional library routines. 40459 (line 109) 40460 * __subtf3: Soft float library routines. 40461 (line 33) 40462 * __subuda3: Fixed-point fractional library routines. 40463 (line 115) 40464 * __subudq3: Fixed-point fractional library routines. 40465 (line 103) 40466 * __subuha3: Fixed-point fractional library routines. 40467 (line 111) 40468 * __subuhq3: Fixed-point fractional library routines. 40469 (line 99) 40470 * __subuqq3: Fixed-point fractional library routines. 40471 (line 97) 40472 * __subusa3: Fixed-point fractional library routines. 40473 (line 113) 40474 * __subusq3: Fixed-point fractional library routines. 40475 (line 101) 40476 * __subuta3: Fixed-point fractional library routines. 40477 (line 117) 40478 * __subvdi3: Integer library routines. 40479 (line 123) 40480 * __subvsi3: Integer library routines. 40481 (line 122) 40482 * __subxf3: Soft float library routines. 40483 (line 35) 40484 * __truncdfsf2: Soft float library routines. 40485 (line 76) 40486 * __trunctfdf2: Soft float library routines. 40487 (line 73) 40488 * __trunctfsf2: Soft float library routines. 40489 (line 75) 40490 * __truncxfdf2: Soft float library routines. 40491 (line 72) 40492 * __truncxfsf2: Soft float library routines. 40493 (line 74) 40494 * __ucmpdi2: Integer library routines. 40495 (line 93) 40496 * __ucmpti2: Integer library routines. 40497 (line 95) 40498 * __udivdi3: Integer library routines. 40499 (line 54) 40500 * __udivmoddi3: Integer library routines. 40501 (line 61) 40502 * __udivsi3: Integer library routines. 40503 (line 52) 40504 * __udivti3: Integer library routines. 40505 (line 56) 40506 * __udivuda3: Fixed-point fractional library routines. 40507 (line 246) 40508 * __udivudq3: Fixed-point fractional library routines. 40509 (line 240) 40510 * __udivuha3: Fixed-point fractional library routines. 40511 (line 242) 40512 * __udivuhq3: Fixed-point fractional library routines. 40513 (line 236) 40514 * __udivuqq3: Fixed-point fractional library routines. 40515 (line 234) 40516 * __udivusa3: Fixed-point fractional library routines. 40517 (line 244) 40518 * __udivusq3: Fixed-point fractional library routines. 40519 (line 238) 40520 * __udivuta3: Fixed-point fractional library routines. 40521 (line 248) 40522 * __umoddi3: Integer library routines. 40523 (line 71) 40524 * __umodsi3: Integer library routines. 40525 (line 69) 40526 * __umodti3: Integer library routines. 40527 (line 73) 40528 * __unorddf2: Soft float library routines. 40529 (line 173) 40530 * __unordsf2: Soft float library routines. 40531 (line 172) 40532 * __unordtf2: Soft float library routines. 40533 (line 174) 40534 * __usadduda3: Fixed-point fractional library routines. 40535 (line 85) 40536 * __usaddudq3: Fixed-point fractional library routines. 40537 (line 79) 40538 * __usadduha3: Fixed-point fractional library routines. 40539 (line 81) 40540 * __usadduhq3: Fixed-point fractional library routines. 40541 (line 75) 40542 * __usadduqq3: Fixed-point fractional library routines. 40543 (line 73) 40544 * __usaddusa3: Fixed-point fractional library routines. 40545 (line 83) 40546 * __usaddusq3: Fixed-point fractional library routines. 40547 (line 77) 40548 * __usadduta3: Fixed-point fractional library routines. 40549 (line 87) 40550 * __usashluda3: Fixed-point fractional library routines. 40551 (line 421) 40552 * __usashludq3: Fixed-point fractional library routines. 40553 (line 415) 40554 * __usashluha3: Fixed-point fractional library routines. 40555 (line 417) 40556 * __usashluhq3: Fixed-point fractional library routines. 40557 (line 411) 40558 * __usashluqq3: Fixed-point fractional library routines. 40559 (line 409) 40560 * __usashlusa3: Fixed-point fractional library routines. 40561 (line 419) 40562 * __usashlusq3: Fixed-point fractional library routines. 40563 (line 413) 40564 * __usashluta3: Fixed-point fractional library routines. 40565 (line 423) 40566 * __usdivuda3: Fixed-point fractional library routines. 40567 (line 280) 40568 * __usdivudq3: Fixed-point fractional library routines. 40569 (line 274) 40570 * __usdivuha3: Fixed-point fractional library routines. 40571 (line 276) 40572 * __usdivuhq3: Fixed-point fractional library routines. 40573 (line 270) 40574 * __usdivuqq3: Fixed-point fractional library routines. 40575 (line 268) 40576 * __usdivusa3: Fixed-point fractional library routines. 40577 (line 278) 40578 * __usdivusq3: Fixed-point fractional library routines. 40579 (line 272) 40580 * __usdivuta3: Fixed-point fractional library routines. 40581 (line 282) 40582 * __usmuluda3: Fixed-point fractional library routines. 40583 (line 212) 40584 * __usmuludq3: Fixed-point fractional library routines. 40585 (line 206) 40586 * __usmuluha3: Fixed-point fractional library routines. 40587 (line 208) 40588 * __usmuluhq3: Fixed-point fractional library routines. 40589 (line 202) 40590 * __usmuluqq3: Fixed-point fractional library routines. 40591 (line 200) 40592 * __usmulusa3: Fixed-point fractional library routines. 40593 (line 210) 40594 * __usmulusq3: Fixed-point fractional library routines. 40595 (line 204) 40596 * __usmuluta3: Fixed-point fractional library routines. 40597 (line 214) 40598 * __usneguda2: Fixed-point fractional library routines. 40599 (line 331) 40600 * __usnegudq2: Fixed-point fractional library routines. 40601 (line 326) 40602 * __usneguha2: Fixed-point fractional library routines. 40603 (line 328) 40604 * __usneguhq2: Fixed-point fractional library routines. 40605 (line 322) 40606 * __usneguqq2: Fixed-point fractional library routines. 40607 (line 321) 40608 * __usnegusa2: Fixed-point fractional library routines. 40609 (line 329) 40610 * __usnegusq2: Fixed-point fractional library routines. 40611 (line 324) 40612 * __usneguta2: Fixed-point fractional library routines. 40613 (line 333) 40614 * __ussubuda3: Fixed-point fractional library routines. 40615 (line 148) 40616 * __ussubudq3: Fixed-point fractional library routines. 40617 (line 142) 40618 * __ussubuha3: Fixed-point fractional library routines. 40619 (line 144) 40620 * __ussubuhq3: Fixed-point fractional library routines. 40621 (line 138) 40622 * __ussubuqq3: Fixed-point fractional library routines. 40623 (line 136) 40624 * __ussubusa3: Fixed-point fractional library routines. 40625 (line 146) 40626 * __ussubusq3: Fixed-point fractional library routines. 40627 (line 140) 40628 * __ussubuta3: Fixed-point fractional library routines. 40629 (line 150) 40630 * abort: Portability. (line 21) 40631 * abs: Arithmetic. (line 195) 40632 * abs and attributes: Expressions. (line 64) 40633 * ABS_EXPR: Expression trees. (line 6) 40634 * absence_set: Processor pipeline description. 40635 (line 220) 40636 * absM2 instruction pattern: Standard Names. (line 452) 40637 * absolute value: Arithmetic. (line 195) 40638 * access to operands: Accessors. (line 6) 40639 * access to special operands: Special Accessors. (line 6) 40640 * accessors: Accessors. (line 6) 40641 * ACCUM_TYPE_SIZE: Type Layout. (line 88) 40642 * ACCUMULATE_OUTGOING_ARGS: Stack Arguments. (line 46) 40643 * ACCUMULATE_OUTGOING_ARGS and stack frames: Function Entry. (line 135) 40644 * ADA_LONG_TYPE_SIZE: Type Layout. (line 26) 40645 * Adding a new GIMPLE statement code: Adding a new GIMPLE statement code. 40646 (line 6) 40647 * ADDITIONAL_REGISTER_NAMES: Instruction Output. (line 15) 40648 * addM3 instruction pattern: Standard Names. (line 216) 40649 * addMODEcc instruction pattern: Standard Names. (line 904) 40650 * addr_diff_vec: Side Effects. (line 302) 40651 * addr_diff_vec, length of: Insn Lengths. (line 26) 40652 * ADDR_EXPR: Expression trees. (line 6) 40653 * addr_vec: Side Effects. (line 297) 40654 * addr_vec, length of: Insn Lengths. (line 26) 40655 * address constraints: Simple Constraints. (line 154) 40656 * address_operand <1>: Simple Constraints. (line 158) 40657 * address_operand: Machine-Independent Predicates. 40658 (line 63) 40659 * addressing modes: Addressing Modes. (line 6) 40660 * ADJUST_FIELD_ALIGN: Storage Layout. (line 201) 40661 * ADJUST_INSN_LENGTH: Insn Lengths. (line 35) 40662 * ADJUST_REG_ALLOC_ORDER: Allocation Order. (line 23) 40663 * AGGR_INIT_EXPR: Expression trees. (line 6) 40664 * aggregates as return values: Aggregate Return. (line 6) 40665 * alias: Alias analysis. (line 6) 40666 * ALL_COP_ADDITIONAL_REGISTER_NAMES: MIPS Coprocessors. (line 32) 40667 * ALL_REGS: Register Classes. (line 17) 40668 * allocate_stack instruction pattern: Standard Names. (line 1227) 40669 * alternate entry points: Insns. (line 140) 40670 * anchored addresses: Anchored Addresses. (line 6) 40671 * and: Arithmetic. (line 153) 40672 * and and attributes: Expressions. (line 50) 40673 * and, canonicalization of: Insn Canonicalizations. 40674 (line 57) 40675 * andM3 instruction pattern: Standard Names. (line 222) 40676 * annotations: Annotations. (line 6) 40677 * APPLY_RESULT_SIZE: Scalar Return. (line 107) 40678 * ARG_POINTER_CFA_OFFSET: Frame Layout. (line 194) 40679 * ARG_POINTER_REGNUM: Frame Registers. (line 41) 40680 * ARG_POINTER_REGNUM and virtual registers: Regs and Memory. (line 65) 40681 * arg_pointer_rtx: Frame Registers. (line 85) 40682 * ARGS_GROW_DOWNWARD: Frame Layout. (line 35) 40683 * argument passing: Interface. (line 36) 40684 * arguments in registers: Register Arguments. (line 6) 40685 * arguments on stack: Stack Arguments. (line 6) 40686 * arithmetic library: Soft float library routines. 40687 (line 6) 40688 * arithmetic shift: Arithmetic. (line 168) 40689 * arithmetic shift with signed saturation: Arithmetic. (line 168) 40690 * arithmetic shift with unsigned saturation: Arithmetic. (line 168) 40691 * arithmetic, in RTL: Arithmetic. (line 6) 40692 * ARITHMETIC_TYPE_P: Types. (line 76) 40693 * array: Types. (line 6) 40694 * ARRAY_RANGE_REF: Expression trees. (line 6) 40695 * ARRAY_REF: Expression trees. (line 6) 40696 * ARRAY_TYPE: Types. (line 6) 40697 * AS_NEEDS_DASH_FOR_PIPED_INPUT: Driver. (line 151) 40698 * ashift: Arithmetic. (line 168) 40699 * ashift and attributes: Expressions. (line 64) 40700 * ashiftrt: Arithmetic. (line 185) 40701 * ashiftrt and attributes: Expressions. (line 64) 40702 * ashlM3 instruction pattern: Standard Names. (line 431) 40703 * ashrM3 instruction pattern: Standard Names. (line 441) 40704 * ASM_APP_OFF: File Framework. (line 61) 40705 * ASM_APP_ON: File Framework. (line 54) 40706 * ASM_COMMENT_START: File Framework. (line 49) 40707 * ASM_DECLARE_CLASS_REFERENCE: Label Output. (line 436) 40708 * ASM_DECLARE_CONSTANT_NAME: Label Output. (line 128) 40709 * ASM_DECLARE_FUNCTION_NAME: Label Output. (line 87) 40710 * ASM_DECLARE_FUNCTION_SIZE: Label Output. (line 101) 40711 * ASM_DECLARE_OBJECT_NAME: Label Output. (line 114) 40712 * ASM_DECLARE_REGISTER_GLOBAL: Label Output. (line 143) 40713 * ASM_DECLARE_UNRESOLVED_REFERENCE: Label Output. (line 442) 40714 * ASM_FINAL_SPEC: Driver. (line 144) 40715 * ASM_FINISH_DECLARE_OBJECT: Label Output. (line 151) 40716 * ASM_FORMAT_PRIVATE_NAME: Label Output. (line 354) 40717 * asm_fprintf: Instruction Output. (line 123) 40718 * ASM_FPRINTF_EXTENSIONS: Instruction Output. (line 134) 40719 * ASM_GENERATE_INTERNAL_LABEL: Label Output. (line 338) 40720 * asm_input: Side Effects. (line 284) 40721 * asm_input and /v: Flags. (line 94) 40722 * ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX: Exception Handling. (line 82) 40723 * ASM_NO_SKIP_IN_TEXT: Alignment Output. (line 72) 40724 * asm_noperands: Insns. (line 266) 40725 * asm_operands and /v: Flags. (line 94) 40726 * asm_operands, RTL sharing: Sharing. (line 45) 40727 * asm_operands, usage: Assembler. (line 6) 40728 * ASM_OUTPUT_ADDR_DIFF_ELT: Dispatch Tables. (line 9) 40729 * ASM_OUTPUT_ADDR_VEC_ELT: Dispatch Tables. (line 26) 40730 * ASM_OUTPUT_ALIGN: Alignment Output. (line 79) 40731 * ASM_OUTPUT_ALIGN_WITH_NOP: Alignment Output. (line 84) 40732 * ASM_OUTPUT_ALIGNED_BSS: Uninitialized Data. (line 64) 40733 * ASM_OUTPUT_ALIGNED_COMMON: Uninitialized Data. (line 23) 40734 * ASM_OUTPUT_ALIGNED_DECL_COMMON: Uninitialized Data. (line 31) 40735 * ASM_OUTPUT_ALIGNED_DECL_LOCAL: Uninitialized Data. (line 95) 40736 * ASM_OUTPUT_ALIGNED_LOCAL: Uninitialized Data. (line 87) 40737 * ASM_OUTPUT_ASCII: Data Output. (line 50) 40738 * ASM_OUTPUT_BSS: Uninitialized Data. (line 39) 40739 * ASM_OUTPUT_CASE_END: Dispatch Tables. (line 51) 40740 * ASM_OUTPUT_CASE_LABEL: Dispatch Tables. (line 38) 40741 * ASM_OUTPUT_COMMON: Uninitialized Data. (line 10) 40742 * ASM_OUTPUT_DEBUG_LABEL: Label Output. (line 326) 40743 * ASM_OUTPUT_DEF: Label Output. (line 375) 40744 * ASM_OUTPUT_DEF_FROM_DECLS: Label Output. (line 383) 40745 * ASM_OUTPUT_DWARF_DELTA: SDB and DWARF. (line 42) 40746 * ASM_OUTPUT_DWARF_OFFSET: SDB and DWARF. (line 46) 40747 * ASM_OUTPUT_DWARF_PCREL: SDB and DWARF. (line 52) 40748 * ASM_OUTPUT_EXTERNAL: Label Output. (line 264) 40749 * ASM_OUTPUT_FDESC: Data Output. (line 59) 40750 * ASM_OUTPUT_IDENT: File Framework. (line 83) 40751 * ASM_OUTPUT_INTERNAL_LABEL: Label Output. (line 17) 40752 * ASM_OUTPUT_LABEL: Label Output. (line 9) 40753 * ASM_OUTPUT_LABEL_REF: Label Output. (line 299) 40754 * ASM_OUTPUT_LABELREF: Label Output. (line 285) 40755 * ASM_OUTPUT_LOCAL: Uninitialized Data. (line 74) 40756 * ASM_OUTPUT_MAX_SKIP_ALIGN: Alignment Output. (line 88) 40757 * ASM_OUTPUT_MEASURED_SIZE: Label Output. (line 41) 40758 * ASM_OUTPUT_OPCODE: Instruction Output. (line 21) 40759 * ASM_OUTPUT_POOL_EPILOGUE: Data Output. (line 109) 40760 * ASM_OUTPUT_POOL_PROLOGUE: Data Output. (line 72) 40761 * ASM_OUTPUT_REG_POP: Instruction Output. (line 178) 40762 * ASM_OUTPUT_REG_PUSH: Instruction Output. (line 173) 40763 * ASM_OUTPUT_SIZE_DIRECTIVE: Label Output. (line 35) 40764 * ASM_OUTPUT_SKIP: Alignment Output. (line 66) 40765 * ASM_OUTPUT_SOURCE_FILENAME: File Framework. (line 68) 40766 * ASM_OUTPUT_SPECIAL_POOL_ENTRY: Data Output. (line 84) 40767 * ASM_OUTPUT_SYMBOL_REF: Label Output. (line 292) 40768 * ASM_OUTPUT_TYPE_DIRECTIVE: Label Output. (line 77) 40769 * ASM_OUTPUT_WEAK_ALIAS: Label Output. (line 401) 40770 * ASM_OUTPUT_WEAKREF: Label Output. (line 203) 40771 * ASM_PREFERRED_EH_DATA_FORMAT: Exception Handling. (line 67) 40772 * ASM_SPEC: Driver. (line 136) 40773 * ASM_STABD_OP: DBX Options. (line 36) 40774 * ASM_STABN_OP: DBX Options. (line 43) 40775 * ASM_STABS_OP: DBX Options. (line 29) 40776 * ASM_WEAKEN_DECL: Label Output. (line 195) 40777 * ASM_WEAKEN_LABEL: Label Output. (line 182) 40778 * assemble_name: Label Output. (line 8) 40779 * assemble_name_raw: Label Output. (line 16) 40780 * assembler format: File Framework. (line 6) 40781 * assembler instructions in RTL: Assembler. (line 6) 40782 * ASSEMBLER_DIALECT: Instruction Output. (line 146) 40783 * assigning attribute values to insns: Tagging Insns. (line 6) 40784 * assignment operator: Function Basics. (line 6) 40785 * asterisk in template: Output Statement. (line 29) 40786 * atan2M3 instruction pattern: Standard Names. (line 522) 40787 * attr <1>: Tagging Insns. (line 54) 40788 * attr: Expressions. (line 154) 40789 * attr_flag: Expressions. (line 119) 40790 * attribute expressions: Expressions. (line 6) 40791 * attribute specifications: Attr Example. (line 6) 40792 * attribute specifications example: Attr Example. (line 6) 40793 * ATTRIBUTE_ALIGNED_VALUE: Storage Layout. (line 183) 40794 * attributes: Attributes. (line 6) 40795 * attributes, defining: Defining Attributes. 40796 (line 6) 40797 * attributes, target-specific: Target Attributes. (line 6) 40798 * autoincrement addressing, availability: Portability. (line 21) 40799 * autoincrement/decrement addressing: Simple Constraints. (line 30) 40800 * automata_option: Processor pipeline description. 40801 (line 301) 40802 * automaton based pipeline description: Processor pipeline description. 40803 (line 6) 40804 * automaton based scheduler: Processor pipeline description. 40805 (line 6) 40806 * AVOID_CCMODE_COPIES: Values in Registers. 40807 (line 153) 40808 * backslash: Output Template. (line 46) 40809 * barrier: Insns. (line 160) 40810 * barrier and /f: Flags. (line 125) 40811 * barrier and /v: Flags. (line 44) 40812 * BASE_REG_CLASS: Register Classes. (line 107) 40813 * basic block: Basic Blocks. (line 6) 40814 * basic-block.h: Control Flow. (line 6) 40815 * BASIC_BLOCK: Basic Blocks. (line 19) 40816 * basic_block: Basic Blocks. (line 6) 40817 * BB_HEAD, BB_END: Maintaining the CFG. 40818 (line 88) 40819 * bb_seq: GIMPLE sequences. (line 73) 40820 * bCOND instruction pattern: Standard Names. (line 941) 40821 * BIGGEST_ALIGNMENT: Storage Layout. (line 173) 40822 * BIGGEST_FIELD_ALIGNMENT: Storage Layout. (line 194) 40823 * BImode: Machine Modes. (line 22) 40824 * BIND_EXPR: Expression trees. (line 6) 40825 * BINFO_TYPE: Classes. (line 6) 40826 * bit-fields: Bit-Fields. (line 6) 40827 * BIT_AND_EXPR: Expression trees. (line 6) 40828 * BIT_IOR_EXPR: Expression trees. (line 6) 40829 * BIT_NOT_EXPR: Expression trees. (line 6) 40830 * BIT_XOR_EXPR: Expression trees. (line 6) 40831 * BITFIELD_NBYTES_LIMITED: Storage Layout. (line 382) 40832 * BITS_BIG_ENDIAN: Storage Layout. (line 12) 40833 * BITS_BIG_ENDIAN, effect on sign_extract: Bit-Fields. (line 8) 40834 * BITS_PER_UNIT: Storage Layout. (line 52) 40835 * BITS_PER_WORD: Storage Layout. (line 57) 40836 * bitwise complement: Arithmetic. (line 149) 40837 * bitwise exclusive-or: Arithmetic. (line 163) 40838 * bitwise inclusive-or: Arithmetic. (line 158) 40839 * bitwise logical-and: Arithmetic. (line 153) 40840 * BLKmode: Machine Modes. (line 183) 40841 * BLKmode, and function return values: Calls. (line 23) 40842 * block statement iterators <1>: Maintaining the CFG. 40843 (line 45) 40844 * block statement iterators: Basic Blocks. (line 68) 40845 * BLOCK_FOR_INSN, bb_for_stmt: Maintaining the CFG. 40846 (line 40) 40847 * BLOCK_REG_PADDING: Register Arguments. (line 228) 40848 * blockage instruction pattern: Standard Names. (line 1408) 40849 * Blocks: Blocks. (line 6) 40850 * bool <1>: Exception Region Output. 40851 (line 60) 40852 * bool: Sections. (line 280) 40853 * BOOL_TYPE_SIZE: Type Layout. (line 44) 40854 * BOOLEAN_TYPE: Types. (line 6) 40855 * branch prediction: Profile information. 40856 (line 24) 40857 * BRANCH_COST: Costs. (line 52) 40858 * break_out_memory_refs: Addressing Modes. (line 130) 40859 * BREAK_STMT: Function Bodies. (line 6) 40860 * bsi_commit_edge_inserts: Maintaining the CFG. 40861 (line 118) 40862 * bsi_end_p: Maintaining the CFG. 40863 (line 60) 40864 * bsi_insert_after: Maintaining the CFG. 40865 (line 72) 40866 * bsi_insert_before: Maintaining the CFG. 40867 (line 78) 40868 * bsi_insert_on_edge: Maintaining the CFG. 40869 (line 118) 40870 * bsi_last: Maintaining the CFG. 40871 (line 56) 40872 * bsi_next: Maintaining the CFG. 40873 (line 64) 40874 * bsi_prev: Maintaining the CFG. 40875 (line 68) 40876 * bsi_remove: Maintaining the CFG. 40877 (line 84) 40878 * bsi_start: Maintaining the CFG. 40879 (line 52) 40880 * BSS_SECTION_ASM_OP: Sections. (line 68) 40881 * bswap: Arithmetic. (line 232) 40882 * btruncM2 instruction pattern: Standard Names. (line 540) 40883 * builtin_longjmp instruction pattern: Standard Names. (line 1313) 40884 * builtin_setjmp_receiver instruction pattern: Standard Names. 40885 (line 1303) 40886 * builtin_setjmp_setup instruction pattern: Standard Names. (line 1292) 40887 * byte_mode: Machine Modes. (line 336) 40888 * BYTES_BIG_ENDIAN: Storage Layout. (line 24) 40889 * BYTES_BIG_ENDIAN, effect on subreg: Regs and Memory. (line 221) 40890 * C statements for assembler output: Output Statement. (line 6) 40891 * C/C++ Internal Representation: Trees. (line 6) 40892 * C99 math functions, implicit usage: Library Calls. (line 76) 40893 * C_COMMON_OVERRIDE_OPTIONS: Run-time Target. (line 114) 40894 * c_register_pragma: Misc. (line 404) 40895 * c_register_pragma_with_expansion: Misc. (line 406) 40896 * call <1>: Side Effects. (line 86) 40897 * call: Flags. (line 234) 40898 * call instruction pattern: Standard Names. (line 974) 40899 * call usage: Calls. (line 10) 40900 * call, in call_insn: Flags. (line 33) 40901 * call, in mem: Flags. (line 99) 40902 * call-clobbered register: Register Basics. (line 35) 40903 * call-saved register: Register Basics. (line 35) 40904 * call-used register: Register Basics. (line 35) 40905 * CALL_EXPR: Expression trees. (line 6) 40906 * call_insn: Insns. (line 95) 40907 * call_insn and /c: Flags. (line 33) 40908 * call_insn and /f: Flags. (line 125) 40909 * call_insn and /i: Flags. (line 24) 40910 * call_insn and /j: Flags. (line 179) 40911 * call_insn and /s: Flags. (line 49) 40912 * call_insn and /u: Flags. (line 19) 40913 * call_insn and /u or /i: Flags. (line 29) 40914 * call_insn and /v: Flags. (line 44) 40915 * CALL_INSN_FUNCTION_USAGE: Insns. (line 101) 40916 * call_pop instruction pattern: Standard Names. (line 1002) 40917 * CALL_POPS_ARGS: Stack Arguments. (line 130) 40918 * CALL_REALLY_USED_REGISTERS: Register Basics. (line 46) 40919 * CALL_USED_REGISTERS: Register Basics. (line 35) 40920 * call_used_regs: Register Basics. (line 59) 40921 * call_value instruction pattern: Standard Names. (line 994) 40922 * call_value_pop instruction pattern: Standard Names. (line 1002) 40923 * CALLER_SAVE_PROFITABLE: Caller Saves. (line 11) 40924 * calling conventions: Stack and Calling. (line 6) 40925 * calling functions in RTL: Calls. (line 6) 40926 * can_create_pseudo_p: Standard Names. (line 75) 40927 * CAN_DEBUG_WITHOUT_FP: Run-time Target. (line 146) 40928 * CAN_ELIMINATE: Elimination. (line 71) 40929 * can_fallthru: Basic Blocks. (line 57) 40930 * canadian: Configure Terms. (line 6) 40931 * CANNOT_CHANGE_MODE_CLASS: Register Classes. (line 481) 40932 * CANNOT_CHANGE_MODE_CLASS and subreg semantics: Regs and Memory. 40933 (line 280) 40934 * canonicalization of instructions: Insn Canonicalizations. 40935 (line 6) 40936 * CANONICALIZE_COMPARISON: Condition Code. (line 84) 40937 * canonicalize_funcptr_for_compare instruction pattern: Standard Names. 40938 (line 1158) 40939 * CASE_USE_BIT_TESTS: Misc. (line 54) 40940 * CASE_VALUES_THRESHOLD: Misc. (line 47) 40941 * CASE_VECTOR_MODE: Misc. (line 27) 40942 * CASE_VECTOR_PC_RELATIVE: Misc. (line 40) 40943 * CASE_VECTOR_SHORTEN_MODE: Misc. (line 31) 40944 * casesi instruction pattern: Standard Names. (line 1082) 40945 * cbranchMODE4 instruction pattern: Standard Names. (line 963) 40946 * cc0: Regs and Memory. (line 307) 40947 * cc0, RTL sharing: Sharing. (line 27) 40948 * cc0_rtx: Regs and Memory. (line 333) 40949 * CC1_SPEC: Driver. (line 118) 40950 * CC1PLUS_SPEC: Driver. (line 126) 40951 * cc_status: Condition Code. (line 8) 40952 * CC_STATUS_MDEP: Condition Code. (line 19) 40953 * CC_STATUS_MDEP_INIT: Condition Code. (line 25) 40954 * CCmode: Machine Modes. (line 176) 40955 * CDImode: Machine Modes. (line 202) 40956 * CEIL_DIV_EXPR: Expression trees. (line 6) 40957 * CEIL_MOD_EXPR: Expression trees. (line 6) 40958 * ceilM2 instruction pattern: Standard Names. (line 556) 40959 * CFA_FRAME_BASE_OFFSET: Frame Layout. (line 226) 40960 * CFG, Control Flow Graph: Control Flow. (line 6) 40961 * cfghooks.h: Maintaining the CFG. 40962 (line 6) 40963 * cgraph_finalize_function: Parsing pass. (line 52) 40964 * chain_circular: GTY Options. (line 195) 40965 * chain_next: GTY Options. (line 195) 40966 * chain_prev: GTY Options. (line 195) 40967 * change_address: Standard Names. (line 47) 40968 * CHANGE_DYNAMIC_TYPE_EXPR: Expression trees. (line 6) 40969 * char <1>: Misc. (line 685) 40970 * char <2>: PCH Target. (line 12) 40971 * char <3>: Sections. (line 272) 40972 * char: GIMPLE_ASM. (line 53) 40973 * CHAR_TYPE_SIZE: Type Layout. (line 39) 40974 * check_stack instruction pattern: Standard Names. (line 1245) 40975 * CHImode: Machine Modes. (line 202) 40976 * class: Classes. (line 6) 40977 * class definitions, register: Register Classes. (line 6) 40978 * class preference constraints: Class Preferences. (line 6) 40979 * CLASS_LIKELY_SPILLED_P: Register Classes. (line 452) 40980 * CLASS_MAX_NREGS: Register Classes. (line 469) 40981 * CLASS_TYPE_P: Types. (line 80) 40982 * classes of RTX codes: RTL Classes. (line 6) 40983 * CLASSTYPE_DECLARED_CLASS: Classes. (line 6) 40984 * CLASSTYPE_HAS_MUTABLE: Classes. (line 80) 40985 * CLASSTYPE_NON_POD_P: Classes. (line 85) 40986 * CLEANUP_DECL: Function Bodies. (line 6) 40987 * CLEANUP_EXPR: Function Bodies. (line 6) 40988 * CLEANUP_POINT_EXPR: Expression trees. (line 6) 40989 * CLEANUP_STMT: Function Bodies. (line 6) 40990 * Cleanups: Cleanups. (line 6) 40991 * CLEAR_BY_PIECES_P: Costs. (line 130) 40992 * clear_cache instruction pattern: Standard Names. (line 1553) 40993 * CLEAR_INSN_CACHE: Trampolines. (line 100) 40994 * CLEAR_RATIO: Costs. (line 121) 40995 * clobber: Side Effects. (line 100) 40996 * clz: Arithmetic. (line 208) 40997 * CLZ_DEFINED_VALUE_AT_ZERO: Misc. (line 319) 40998 * clzM2 instruction pattern: Standard Names. (line 621) 40999 * cmpM instruction pattern: Standard Names. (line 654) 41000 * cmpmemM instruction pattern: Standard Names. (line 769) 41001 * cmpstrM instruction pattern: Standard Names. (line 750) 41002 * cmpstrnM instruction pattern: Standard Names. (line 738) 41003 * code generation RTL sequences: Expander Definitions. 41004 (line 6) 41005 * code iterators in .md files: Code Iterators. (line 6) 41006 * code_label: Insns. (line 119) 41007 * code_label and /i: Flags. (line 59) 41008 * code_label and /v: Flags. (line 44) 41009 * CODE_LABEL_NUMBER: Insns. (line 119) 41010 * codes, RTL expression: RTL Objects. (line 47) 41011 * COImode: Machine Modes. (line 202) 41012 * COLLECT2_HOST_INITIALIZATION: Host Misc. (line 32) 41013 * COLLECT_EXPORT_LIST: Misc. (line 767) 41014 * COLLECT_SHARED_FINI_FUNC: Macros for Initialization. 41015 (line 44) 41016 * COLLECT_SHARED_INIT_FUNC: Macros for Initialization. 41017 (line 33) 41018 * commit_edge_insertions: Maintaining the CFG. 41019 (line 118) 41020 * compare: Arithmetic. (line 43) 41021 * compare, canonicalization of: Insn Canonicalizations. 41022 (line 37) 41023 * comparison_operator: Machine-Independent Predicates. 41024 (line 111) 41025 * compiler passes and files: Passes. (line 6) 41026 * complement, bitwise: Arithmetic. (line 149) 41027 * COMPLEX_CST: Expression trees. (line 6) 41028 * COMPLEX_EXPR: Expression trees. (line 6) 41029 * COMPLEX_TYPE: Types. (line 6) 41030 * COMPONENT_REF: Expression trees. (line 6) 41031 * Compound Expressions: Compound Expressions. 41032 (line 6) 41033 * Compound Lvalues: Compound Lvalues. (line 6) 41034 * COMPOUND_EXPR: Expression trees. (line 6) 41035 * COMPOUND_LITERAL_EXPR: Expression trees. (line 6) 41036 * COMPOUND_LITERAL_EXPR_DECL: Expression trees. (line 608) 41037 * COMPOUND_LITERAL_EXPR_DECL_STMT: Expression trees. (line 608) 41038 * computed jump: Edges. (line 128) 41039 * computing the length of an insn: Insn Lengths. (line 6) 41040 * concat: Regs and Memory. (line 385) 41041 * concatn: Regs and Memory. (line 391) 41042 * cond: Comparisons. (line 90) 41043 * cond and attributes: Expressions. (line 37) 41044 * cond_exec: Side Effects. (line 248) 41045 * COND_EXPR: Expression trees. (line 6) 41046 * condition code register: Regs and Memory. (line 307) 41047 * condition code status: Condition Code. (line 6) 41048 * condition codes: Comparisons. (line 20) 41049 * conditional execution: Conditional Execution. 41050 (line 6) 41051 * Conditional Expressions: Conditional Expressions. 41052 (line 6) 41053 * CONDITIONAL_REGISTER_USAGE: Register Basics. (line 60) 41054 * conditional_trap instruction pattern: Standard Names. (line 1379) 41055 * conditions, in patterns: Patterns. (line 43) 41056 * configuration file <1>: Host Misc. (line 6) 41057 * configuration file: Filesystem. (line 6) 41058 * configure terms: Configure Terms. (line 6) 41059 * CONJ_EXPR: Expression trees. (line 6) 41060 * const: Constants. (line 99) 41061 * CONST0_RTX: Constants. (line 119) 41062 * const0_rtx: Constants. (line 16) 41063 * CONST1_RTX: Constants. (line 119) 41064 * const1_rtx: Constants. (line 16) 41065 * CONST2_RTX: Constants. (line 119) 41066 * const2_rtx: Constants. (line 16) 41067 * CONST_DECL: Declarations. (line 6) 41068 * const_double: Constants. (line 32) 41069 * const_double, RTL sharing: Sharing. (line 29) 41070 * CONST_DOUBLE_LOW: Constants. (line 39) 41071 * CONST_DOUBLE_OK_FOR_CONSTRAINT_P: Old Constraints. (line 69) 41072 * CONST_DOUBLE_OK_FOR_LETTER_P: Old Constraints. (line 54) 41073 * const_double_operand: Machine-Independent Predicates. 41074 (line 21) 41075 * const_fixed: Constants. (line 52) 41076 * const_int: Constants. (line 8) 41077 * const_int and attribute tests: Expressions. (line 47) 41078 * const_int and attributes: Expressions. (line 10) 41079 * const_int, RTL sharing: Sharing. (line 23) 41080 * const_int_operand: Machine-Independent Predicates. 41081 (line 16) 41082 * CONST_OK_FOR_CONSTRAINT_P: Old Constraints. (line 49) 41083 * CONST_OK_FOR_LETTER_P: Old Constraints. (line 40) 41084 * const_string: Constants. (line 71) 41085 * const_string and attributes: Expressions. (line 20) 41086 * const_true_rtx: Constants. (line 26) 41087 * const_vector: Constants. (line 59) 41088 * const_vector, RTL sharing: Sharing. (line 32) 41089 * constant attributes: Constant Attributes. 41090 (line 6) 41091 * constant definitions: Constant Definitions. 41092 (line 6) 41093 * CONSTANT_ADDRESS_P: Addressing Modes. (line 29) 41094 * CONSTANT_ALIGNMENT: Storage Layout. (line 241) 41095 * CONSTANT_P: Addressing Modes. (line 35) 41096 * CONSTANT_POOL_ADDRESS_P: Flags. (line 10) 41097 * CONSTANT_POOL_BEFORE_FUNCTION: Data Output. (line 64) 41098 * constants in constraints: Simple Constraints. (line 60) 41099 * constm1_rtx: Constants. (line 16) 41100 * constraint modifier characters: Modifiers. (line 6) 41101 * constraint, matching: Simple Constraints. (line 132) 41102 * CONSTRAINT_LEN: Old Constraints. (line 12) 41103 * constraint_num: C Constraint Interface. 41104 (line 38) 41105 * constraint_satisfied_p: C Constraint Interface. 41106 (line 54) 41107 * constraints: Constraints. (line 6) 41108 * constraints, defining: Define Constraints. (line 6) 41109 * constraints, defining, obsolete method: Old Constraints. (line 6) 41110 * constraints, machine specific: Machine Constraints. 41111 (line 6) 41112 * constraints, testing: C Constraint Interface. 41113 (line 6) 41114 * CONSTRUCTOR: Expression trees. (line 6) 41115 * constructor: Function Basics. (line 6) 41116 * constructors, automatic calls: Collect2. (line 15) 41117 * constructors, output of: Initialization. (line 6) 41118 * container: Containers. (line 6) 41119 * CONTINUE_STMT: Function Bodies. (line 6) 41120 * contributors: Contributors. (line 6) 41121 * controlling register usage: Register Basics. (line 76) 41122 * controlling the compilation driver: Driver. (line 6) 41123 * conventions, run-time: Interface. (line 6) 41124 * conversions: Conversions. (line 6) 41125 * CONVERT_EXPR: Expression trees. (line 6) 41126 * copy constructor: Function Basics. (line 6) 41127 * copy_rtx: Addressing Modes. (line 182) 41128 * copy_rtx_if_shared: Sharing. (line 64) 41129 * copysignM3 instruction pattern: Standard Names. (line 602) 41130 * cosM2 instruction pattern: Standard Names. (line 481) 41131 * costs of instructions: Costs. (line 6) 41132 * CP_INTEGRAL_TYPE: Types. (line 72) 41133 * cp_namespace_decls: Namespaces. (line 44) 41134 * CP_TYPE_CONST_NON_VOLATILE_P: Types. (line 45) 41135 * CP_TYPE_CONST_P: Types. (line 36) 41136 * CP_TYPE_QUALS: Types. (line 6) 41137 * CP_TYPE_RESTRICT_P: Types. (line 42) 41138 * CP_TYPE_VOLATILE_P: Types. (line 39) 41139 * CPLUSPLUS_CPP_SPEC: Driver. (line 113) 41140 * CPP_SPEC: Driver. (line 106) 41141 * CQImode: Machine Modes. (line 202) 41142 * cross compilation and floating point: Floating Point. (line 6) 41143 * CRT_CALL_STATIC_FUNCTION: Sections. (line 112) 41144 * CRTSTUFF_T_CFLAGS: Target Fragment. (line 35) 41145 * CRTSTUFF_T_CFLAGS_S: Target Fragment. (line 39) 41146 * CSImode: Machine Modes. (line 202) 41147 * CTImode: Machine Modes. (line 202) 41148 * ctz: Arithmetic. (line 216) 41149 * CTZ_DEFINED_VALUE_AT_ZERO: Misc. (line 320) 41150 * ctzM2 instruction pattern: Standard Names. (line 630) 41151 * CUMULATIVE_ARGS: Register Arguments. (line 127) 41152 * current_function_epilogue_delay_list: Function Entry. (line 181) 41153 * current_function_is_leaf: Leaf Functions. (line 51) 41154 * current_function_outgoing_args_size: Stack Arguments. (line 45) 41155 * current_function_pops_args: Function Entry. (line 106) 41156 * current_function_pretend_args_size: Function Entry. (line 112) 41157 * current_function_uses_only_leaf_regs: Leaf Functions. (line 51) 41158 * current_insn_predicate: Conditional Execution. 41159 (line 26) 41160 * DAmode: Machine Modes. (line 152) 41161 * data bypass: Processor pipeline description. 41162 (line 106) 41163 * data dependence delays: Processor pipeline description. 41164 (line 6) 41165 * Data Dependency Analysis: Dependency analysis. 41166 (line 6) 41167 * data structures: Per-Function Data. (line 6) 41168 * DATA_ALIGNMENT: Storage Layout. (line 228) 41169 * DATA_SECTION_ASM_OP: Sections. (line 53) 41170 * DBR_OUTPUT_SEQEND: Instruction Output. (line 107) 41171 * dbr_sequence_length: Instruction Output. (line 106) 41172 * DBX_BLOCKS_FUNCTION_RELATIVE: DBX Options. (line 103) 41173 * DBX_CONTIN_CHAR: DBX Options. (line 66) 41174 * DBX_CONTIN_LENGTH: DBX Options. (line 56) 41175 * DBX_DEBUGGING_INFO: DBX Options. (line 9) 41176 * DBX_FUNCTION_FIRST: DBX Options. (line 97) 41177 * DBX_LINES_FUNCTION_RELATIVE: DBX Options. (line 109) 41178 * DBX_NO_XREFS: DBX Options. (line 50) 41179 * DBX_OUTPUT_LBRAC: DBX Hooks. (line 9) 41180 * DBX_OUTPUT_MAIN_SOURCE_FILE_END: File Names and DBX. (line 34) 41181 * DBX_OUTPUT_MAIN_SOURCE_FILENAME: File Names and DBX. (line 9) 41182 * DBX_OUTPUT_NFUN: DBX Hooks. (line 18) 41183 * DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END: File Names and DBX. 41184 (line 42) 41185 * DBX_OUTPUT_RBRAC: DBX Hooks. (line 15) 41186 * DBX_OUTPUT_SOURCE_LINE: DBX Hooks. (line 22) 41187 * DBX_REGISTER_NUMBER: All Debuggers. (line 9) 41188 * DBX_REGPARM_STABS_CODE: DBX Options. (line 87) 41189 * DBX_REGPARM_STABS_LETTER: DBX Options. (line 92) 41190 * DBX_STATIC_CONST_VAR_CODE: DBX Options. (line 82) 41191 * DBX_STATIC_STAB_DATA_SECTION: DBX Options. (line 73) 41192 * DBX_TYPE_DECL_STABS_CODE: DBX Options. (line 78) 41193 * DBX_USE_BINCL: DBX Options. (line 115) 41194 * DCmode: Machine Modes. (line 197) 41195 * DDmode: Machine Modes. (line 90) 41196 * De Morgan's law: Insn Canonicalizations. 41197 (line 57) 41198 * dead_or_set_p: define_peephole. (line 65) 41199 * DEBUG_SYMS_TEXT: DBX Options. (line 25) 41200 * DEBUGGER_ARG_OFFSET: All Debuggers. (line 37) 41201 * DEBUGGER_AUTO_OFFSET: All Debuggers. (line 28) 41202 * decimal float library: Decimal float library routines. 41203 (line 6) 41204 * DECL_ALIGN: Declarations. (line 6) 41205 * DECL_ANTICIPATED: Function Basics. (line 48) 41206 * DECL_ARGUMENTS: Function Basics. (line 163) 41207 * DECL_ARRAY_DELETE_OPERATOR_P: Function Basics. (line 184) 41208 * DECL_ARTIFICIAL <1>: Function Basics. (line 6) 41209 * DECL_ARTIFICIAL: Working with declarations. 41210 (line 24) 41211 * DECL_ASSEMBLER_NAME: Function Basics. (line 6) 41212 * DECL_ATTRIBUTES: Attributes. (line 22) 41213 * DECL_BASE_CONSTRUCTOR_P: Function Basics. (line 94) 41214 * DECL_CLASS_SCOPE_P: Working with declarations. 41215 (line 41) 41216 * DECL_COMPLETE_CONSTRUCTOR_P: Function Basics. (line 90) 41217 * DECL_COMPLETE_DESTRUCTOR_P: Function Basics. (line 104) 41218 * DECL_CONST_MEMFUNC_P: Function Basics. (line 77) 41219 * DECL_CONSTRUCTOR_P: Function Basics. (line 6) 41220 * DECL_CONTEXT: Namespaces. (line 26) 41221 * DECL_CONV_FN_P: Function Basics. (line 6) 41222 * DECL_COPY_CONSTRUCTOR_P: Function Basics. (line 98) 41223 * DECL_DESTRUCTOR_P: Function Basics. (line 6) 41224 * DECL_EXTERN_C_FUNCTION_P: Function Basics. (line 52) 41225 * DECL_EXTERNAL <1>: Function Basics. (line 38) 41226 * DECL_EXTERNAL: Declarations. (line 6) 41227 * DECL_FUNCTION_MEMBER_P: Function Basics. (line 6) 41228 * DECL_FUNCTION_SCOPE_P: Working with declarations. 41229 (line 44) 41230 * DECL_FUNCTION_SPECIFIC_OPTIMIZATION: Function Basics. (line 6) 41231 * DECL_FUNCTION_SPECIFIC_TARGET: Function Basics. (line 6) 41232 * DECL_GLOBAL_CTOR_P: Function Basics. (line 6) 41233 * DECL_GLOBAL_DTOR_P: Function Basics. (line 6) 41234 * DECL_INITIAL: Declarations. (line 6) 41235 * DECL_LINKONCE_P: Function Basics. (line 6) 41236 * DECL_LOCAL_FUNCTION_P: Function Basics. (line 44) 41237 * DECL_MAIN_P: Function Basics. (line 7) 41238 * DECL_NAME <1>: Function Basics. (line 6) 41239 * DECL_NAME <2>: Working with declarations. 41240 (line 7) 41241 * DECL_NAME: Namespaces. (line 15) 41242 * DECL_NAMESPACE_ALIAS: Namespaces. (line 30) 41243 * DECL_NAMESPACE_SCOPE_P: Working with declarations. 41244 (line 37) 41245 * DECL_NAMESPACE_STD_P: Namespaces. (line 40) 41246 * DECL_NON_THUNK_FUNCTION_P: Function Basics. (line 144) 41247 * DECL_NONCONVERTING_P: Function Basics. (line 86) 41248 * DECL_NONSTATIC_MEMBER_FUNCTION_P: Function Basics. (line 74) 41249 * DECL_OVERLOADED_OPERATOR_P: Function Basics. (line 6) 41250 * DECL_RESULT: Function Basics. (line 168) 41251 * DECL_SIZE: Declarations. (line 6) 41252 * DECL_STATIC_FUNCTION_P: Function Basics. (line 71) 41253 * DECL_STMT: Function Bodies. (line 6) 41254 * DECL_STMT_DECL: Function Bodies. (line 6) 41255 * DECL_THUNK_P: Function Basics. (line 122) 41256 * DECL_VOLATILE_MEMFUNC_P: Function Basics. (line 80) 41257 * declaration: Declarations. (line 6) 41258 * declarations, RTL: RTL Declarations. (line 6) 41259 * DECLARE_LIBRARY_RENAMES: Library Calls. (line 9) 41260 * decrement_and_branch_until_zero instruction pattern: Standard Names. 41261 (line 1120) 41262 * def_optype_d: Manipulating GIMPLE statements. 41263 (line 94) 41264 * default: GTY Options. (line 81) 41265 * default_file_start: File Framework. (line 9) 41266 * DEFAULT_GDB_EXTENSIONS: DBX Options. (line 18) 41267 * DEFAULT_PCC_STRUCT_RETURN: Aggregate Return. (line 34) 41268 * DEFAULT_SIGNED_CHAR: Type Layout. (line 154) 41269 * define_address_constraint: Define Constraints. (line 107) 41270 * define_asm_attributes: Tagging Insns. (line 73) 41271 * define_attr: Defining Attributes. 41272 (line 6) 41273 * define_automaton: Processor pipeline description. 41274 (line 53) 41275 * define_bypass: Processor pipeline description. 41276 (line 197) 41277 * define_code_attr: Code Iterators. (line 6) 41278 * define_code_iterator: Code Iterators. (line 6) 41279 * define_cond_exec: Conditional Execution. 41280 (line 13) 41281 * define_constants: Constant Definitions. 41282 (line 6) 41283 * define_constraint: Define Constraints. (line 48) 41284 * define_cpu_unit: Processor pipeline description. 41285 (line 68) 41286 * define_delay: Delay Slots. (line 25) 41287 * define_expand: Expander Definitions. 41288 (line 11) 41289 * define_insn: Patterns. (line 6) 41290 * define_insn example: Example. (line 6) 41291 * define_insn_and_split: Insn Splitting. (line 170) 41292 * define_insn_reservation: Processor pipeline description. 41293 (line 106) 41294 * define_memory_constraint: Define Constraints. (line 88) 41295 * define_mode_attr: Substitutions. (line 6) 41296 * define_mode_iterator: Defining Mode Iterators. 41297 (line 6) 41298 * define_peephole: define_peephole. (line 6) 41299 * define_peephole2: define_peephole2. (line 6) 41300 * define_predicate: Defining Predicates. 41301 (line 6) 41302 * define_query_cpu_unit: Processor pipeline description. 41303 (line 90) 41304 * define_register_constraint: Define Constraints. (line 28) 41305 * define_reservation: Processor pipeline description. 41306 (line 186) 41307 * define_special_predicate: Defining Predicates. 41308 (line 6) 41309 * define_split: Insn Splitting. (line 32) 41310 * defining attributes and their values: Defining Attributes. 41311 (line 6) 41312 * defining constraints: Define Constraints. (line 6) 41313 * defining constraints, obsolete method: Old Constraints. (line 6) 41314 * defining jump instruction patterns: Jump Patterns. (line 6) 41315 * defining looping instruction patterns: Looping Patterns. (line 6) 41316 * defining peephole optimizers: Peephole Definitions. 41317 (line 6) 41318 * defining predicates: Defining Predicates. 41319 (line 6) 41320 * defining RTL sequences for code generation: Expander Definitions. 41321 (line 6) 41322 * delay slots, defining: Delay Slots. (line 6) 41323 * DELAY_SLOTS_FOR_EPILOGUE: Function Entry. (line 163) 41324 * deletable: GTY Options. (line 149) 41325 * DELETE_IF_ORDINARY: Filesystem. (line 79) 41326 * Dependent Patterns: Dependent Patterns. (line 6) 41327 * desc: GTY Options. (line 81) 41328 * destructor: Function Basics. (line 6) 41329 * destructors, output of: Initialization. (line 6) 41330 * deterministic finite state automaton: Processor pipeline description. 41331 (line 6) 41332 * DF_SIZE: Type Layout. (line 130) 41333 * DFmode: Machine Modes. (line 73) 41334 * digits in constraint: Simple Constraints. (line 120) 41335 * DImode: Machine Modes. (line 45) 41336 * DIR_SEPARATOR: Filesystem. (line 18) 41337 * DIR_SEPARATOR_2: Filesystem. (line 19) 41338 * directory options .md: Including Patterns. (line 44) 41339 * disabling certain registers: Register Basics. (line 76) 41340 * dispatch table: Dispatch Tables. (line 8) 41341 * div: Arithmetic. (line 111) 41342 * div and attributes: Expressions. (line 64) 41343 * division: Arithmetic. (line 111) 41344 * divM3 instruction pattern: Standard Names. (line 222) 41345 * divmodM4 instruction pattern: Standard Names. (line 411) 41346 * DO_BODY: Function Bodies. (line 6) 41347 * DO_COND: Function Bodies. (line 6) 41348 * DO_STMT: Function Bodies. (line 6) 41349 * DOLLARS_IN_IDENTIFIERS: Misc. (line 488) 41350 * doloop_begin instruction pattern: Standard Names. (line 1151) 41351 * doloop_end instruction pattern: Standard Names. (line 1130) 41352 * DONE: Expander Definitions. 41353 (line 74) 41354 * DONT_USE_BUILTIN_SETJMP: Exception Region Output. 41355 (line 70) 41356 * DOUBLE_TYPE_SIZE: Type Layout. (line 53) 41357 * DQmode: Machine Modes. (line 115) 41358 * driver: Driver. (line 6) 41359 * DRIVER_SELF_SPECS: Driver. (line 71) 41360 * DUMPFILE_FORMAT: Filesystem. (line 67) 41361 * DWARF2_ASM_LINE_DEBUG_INFO: SDB and DWARF. (line 36) 41362 * DWARF2_DEBUGGING_INFO: SDB and DWARF. (line 13) 41363 * DWARF2_FRAME_INFO: SDB and DWARF. (line 30) 41364 * DWARF2_FRAME_REG_OUT: Frame Registers. (line 133) 41365 * DWARF2_UNWIND_INFO: Exception Region Output. 41366 (line 40) 41367 * DWARF_ALT_FRAME_RETURN_COLUMN: Frame Layout. (line 152) 41368 * DWARF_CIE_DATA_ALIGNMENT: Exception Region Output. 41369 (line 75) 41370 * DWARF_FRAME_REGISTERS: Frame Registers. (line 93) 41371 * DWARF_FRAME_REGNUM: Frame Registers. (line 125) 41372 * DWARF_REG_TO_UNWIND_COLUMN: Frame Registers. (line 117) 41373 * DWARF_ZERO_REG: Frame Layout. (line 163) 41374 * DYNAMIC_CHAIN_ADDRESS: Frame Layout. (line 92) 41375 * E in constraint: Simple Constraints. (line 79) 41376 * earlyclobber operand: Modifiers. (line 25) 41377 * edge: Edges. (line 6) 41378 * edge in the flow graph: Edges. (line 6) 41379 * edge iterators: Edges. (line 15) 41380 * edge splitting: Maintaining the CFG. 41381 (line 118) 41382 * EDGE_ABNORMAL: Edges. (line 128) 41383 * EDGE_ABNORMAL, EDGE_ABNORMAL_CALL: Edges. (line 171) 41384 * EDGE_ABNORMAL, EDGE_EH: Edges. (line 96) 41385 * EDGE_ABNORMAL, EDGE_SIBCALL: Edges. (line 122) 41386 * EDGE_FALLTHRU, force_nonfallthru: Edges. (line 86) 41387 * EDOM, implicit usage: Library Calls. (line 58) 41388 * EH_FRAME_IN_DATA_SECTION: Exception Region Output. 41389 (line 20) 41390 * EH_FRAME_SECTION_NAME: Exception Region Output. 41391 (line 10) 41392 * eh_return instruction pattern: Standard Names. (line 1319) 41393 * EH_RETURN_DATA_REGNO: Exception Handling. (line 7) 41394 * EH_RETURN_HANDLER_RTX: Exception Handling. (line 39) 41395 * EH_RETURN_STACKADJ_RTX: Exception Handling. (line 22) 41396 * EH_TABLES_CAN_BE_READ_ONLY: Exception Region Output. 41397 (line 29) 41398 * EH_USES: Function Entry. (line 158) 41399 * ei_edge: Edges. (line 43) 41400 * ei_end_p: Edges. (line 27) 41401 * ei_last: Edges. (line 23) 41402 * ei_next: Edges. (line 35) 41403 * ei_one_before_end_p: Edges. (line 31) 41404 * ei_prev: Edges. (line 39) 41405 * ei_safe_safe: Edges. (line 47) 41406 * ei_start: Edges. (line 19) 41407 * ELIGIBLE_FOR_EPILOGUE_DELAY: Function Entry. (line 169) 41408 * ELIMINABLE_REGS: Elimination. (line 44) 41409 * ELSE_CLAUSE: Function Bodies. (line 6) 41410 * Embedded C: Fixed-point fractional library routines. 41411 (line 6) 41412 * EMIT_MODE_SET: Mode Switching. (line 74) 41413 * Empty Statements: Empty Statements. (line 6) 41414 * EMPTY_CLASS_EXPR: Function Bodies. (line 6) 41415 * EMPTY_FIELD_BOUNDARY: Storage Layout. (line 295) 41416 * Emulated TLS: Emulated TLS. (line 6) 41417 * ENABLE_EXECUTE_STACK: Trampolines. (line 110) 41418 * enabled: Disable Insn Alternatives. 41419 (line 6) 41420 * ENDFILE_SPEC: Driver. (line 218) 41421 * endianness: Portability. (line 21) 41422 * ENTRY_BLOCK_PTR, EXIT_BLOCK_PTR: Basic Blocks. (line 28) 41423 * enum machine_mode: Machine Modes. (line 6) 41424 * enum reg_class: Register Classes. (line 65) 41425 * ENUMERAL_TYPE: Types. (line 6) 41426 * epilogue: Function Entry. (line 6) 41427 * epilogue instruction pattern: Standard Names. (line 1351) 41428 * EPILOGUE_USES: Function Entry. (line 152) 41429 * eq: Comparisons. (line 52) 41430 * eq and attributes: Expressions. (line 64) 41431 * eq_attr: Expressions. (line 85) 41432 * EQ_EXPR: Expression trees. (line 6) 41433 * equal: Comparisons. (line 52) 41434 * errno, implicit usage: Library Calls. (line 70) 41435 * EXACT_DIV_EXPR: Expression trees. (line 6) 41436 * examining SSA_NAMEs: SSA. (line 218) 41437 * exception handling <1>: Exception Handling. (line 6) 41438 * exception handling: Edges. (line 96) 41439 * exception_receiver instruction pattern: Standard Names. (line 1283) 41440 * exclamation point: Multi-Alternative. (line 47) 41441 * exclusion_set: Processor pipeline description. 41442 (line 220) 41443 * exclusive-or, bitwise: Arithmetic. (line 163) 41444 * EXIT_EXPR: Expression trees. (line 6) 41445 * EXIT_IGNORE_STACK: Function Entry. (line 140) 41446 * expander definitions: Expander Definitions. 41447 (line 6) 41448 * expM2 instruction pattern: Standard Names. (line 497) 41449 * expr_list: Insns. (line 505) 41450 * EXPR_STMT: Function Bodies. (line 6) 41451 * EXPR_STMT_EXPR: Function Bodies. (line 6) 41452 * expression: Expression trees. (line 6) 41453 * expression codes: RTL Objects. (line 47) 41454 * extendMN2 instruction pattern: Standard Names. (line 826) 41455 * extensible constraints: Simple Constraints. (line 163) 41456 * EXTRA_ADDRESS_CONSTRAINT: Old Constraints. (line 123) 41457 * EXTRA_CONSTRAINT: Old Constraints. (line 74) 41458 * EXTRA_CONSTRAINT_STR: Old Constraints. (line 95) 41459 * EXTRA_MEMORY_CONSTRAINT: Old Constraints. (line 100) 41460 * EXTRA_SPECS: Driver. (line 245) 41461 * extv instruction pattern: Standard Names. (line 862) 41462 * extzv instruction pattern: Standard Names. (line 877) 41463 * F in constraint: Simple Constraints. (line 84) 41464 * FAIL: Expander Definitions. 41465 (line 80) 41466 * fall-thru: Edges. (line 69) 41467 * FATAL_EXIT_CODE: Host Misc. (line 6) 41468 * FDL, GNU Free Documentation License: GNU Free Documentation License. 41469 (line 6) 41470 * features, optional, in system conventions: Run-time Target. 41471 (line 59) 41472 * ffs: Arithmetic. (line 202) 41473 * ffsM2 instruction pattern: Standard Names. (line 611) 41474 * FIELD_DECL: Declarations. (line 6) 41475 * file_end_indicate_exec_stack: File Framework. (line 41) 41476 * files and passes of the compiler: Passes. (line 6) 41477 * files, generated: Files. (line 6) 41478 * final_absence_set: Processor pipeline description. 41479 (line 220) 41480 * FINAL_PRESCAN_INSN: Instruction Output. (line 46) 41481 * final_presence_set: Processor pipeline description. 41482 (line 220) 41483 * final_scan_insn: Function Entry. (line 181) 41484 * final_sequence: Instruction Output. (line 117) 41485 * FIND_BASE_TERM: Addressing Modes. (line 110) 41486 * FINI_ARRAY_SECTION_ASM_OP: Sections. (line 105) 41487 * FINI_SECTION_ASM_OP: Sections. (line 90) 41488 * finite state automaton minimization: Processor pipeline description. 41489 (line 301) 41490 * FIRST_PARM_OFFSET: Frame Layout. (line 67) 41491 * FIRST_PARM_OFFSET and virtual registers: Regs and Memory. (line 65) 41492 * FIRST_PSEUDO_REGISTER: Register Basics. (line 9) 41493 * FIRST_STACK_REG: Stack Registers. (line 23) 41494 * FIRST_VIRTUAL_REGISTER: Regs and Memory. (line 51) 41495 * fix: Conversions. (line 66) 41496 * FIX_TRUNC_EXPR: Expression trees. (line 6) 41497 * fix_truncMN2 instruction pattern: Standard Names. (line 813) 41498 * fixed register: Register Basics. (line 15) 41499 * fixed-point fractional library: Fixed-point fractional library routines. 41500 (line 6) 41501 * FIXED_CONVERT_EXPR: Expression trees. (line 6) 41502 * FIXED_CST: Expression trees. (line 6) 41503 * FIXED_POINT_TYPE: Types. (line 6) 41504 * FIXED_REGISTERS: Register Basics. (line 15) 41505 * fixed_regs: Register Basics. (line 59) 41506 * fixMN2 instruction pattern: Standard Names. (line 793) 41507 * FIXUNS_TRUNC_LIKE_FIX_TRUNC: Misc. (line 100) 41508 * fixuns_truncMN2 instruction pattern: Standard Names. (line 817) 41509 * fixunsMN2 instruction pattern: Standard Names. (line 802) 41510 * flags in RTL expression: Flags. (line 6) 41511 * float: Conversions. (line 58) 41512 * FLOAT_EXPR: Expression trees. (line 6) 41513 * float_extend: Conversions. (line 33) 41514 * FLOAT_LIB_COMPARE_RETURNS_BOOL: Library Calls. (line 25) 41515 * FLOAT_STORE_FLAG_VALUE: Misc. (line 301) 41516 * float_truncate: Conversions. (line 53) 41517 * FLOAT_TYPE_SIZE: Type Layout. (line 49) 41518 * FLOAT_WORDS_BIG_ENDIAN: Storage Layout. (line 43) 41519 * FLOAT_WORDS_BIG_ENDIAN, (lack of) effect on subreg: Regs and Memory. 41520 (line 226) 41521 * floating point and cross compilation: Floating Point. (line 6) 41522 * Floating Point Emulation: Target Fragment. (line 15) 41523 * floating point emulation library, US Software GOFAST: Library Calls. 41524 (line 44) 41525 * floatMN2 instruction pattern: Standard Names. (line 785) 41526 * floatunsMN2 instruction pattern: Standard Names. (line 789) 41527 * FLOOR_DIV_EXPR: Expression trees. (line 6) 41528 * FLOOR_MOD_EXPR: Expression trees. (line 6) 41529 * floorM2 instruction pattern: Standard Names. (line 532) 41530 * flow-insensitive alias analysis: Alias analysis. (line 6) 41531 * flow-sensitive alias analysis: Alias analysis. (line 6) 41532 * fmodM3 instruction pattern: Standard Names. (line 463) 41533 * FOR_BODY: Function Bodies. (line 6) 41534 * FOR_COND: Function Bodies. (line 6) 41535 * FOR_EXPR: Function Bodies. (line 6) 41536 * FOR_INIT_STMT: Function Bodies. (line 6) 41537 * FOR_STMT: Function Bodies. (line 6) 41538 * FORCE_CODE_SECTION_ALIGN: Sections. (line 136) 41539 * force_reg: Standard Names. (line 36) 41540 * fract_convert: Conversions. (line 82) 41541 * FRACT_TYPE_SIZE: Type Layout. (line 68) 41542 * fractional types: Fixed-point fractional library routines. 41543 (line 6) 41544 * fractMN2 instruction pattern: Standard Names. (line 835) 41545 * fractunsMN2 instruction pattern: Standard Names. (line 850) 41546 * frame layout: Frame Layout. (line 6) 41547 * FRAME_ADDR_RTX: Frame Layout. (line 116) 41548 * FRAME_GROWS_DOWNWARD: Frame Layout. (line 31) 41549 * FRAME_GROWS_DOWNWARD and virtual registers: Regs and Memory. 41550 (line 69) 41551 * FRAME_POINTER_CFA_OFFSET: Frame Layout. (line 212) 41552 * frame_pointer_needed: Function Entry. (line 34) 41553 * FRAME_POINTER_REGNUM: Frame Registers. (line 14) 41554 * FRAME_POINTER_REGNUM and virtual registers: Regs and Memory. 41555 (line 74) 41556 * FRAME_POINTER_REQUIRED: Elimination. (line 9) 41557 * frame_pointer_rtx: Frame Registers. (line 85) 41558 * frame_related: Flags. (line 242) 41559 * frame_related, in insn, call_insn, jump_insn, barrier, and set: Flags. 41560 (line 125) 41561 * frame_related, in mem: Flags. (line 103) 41562 * frame_related, in reg: Flags. (line 112) 41563 * frame_related, in symbol_ref: Flags. (line 183) 41564 * frequency, count, BB_FREQ_BASE: Profile information. 41565 (line 30) 41566 * ftruncM2 instruction pattern: Standard Names. (line 808) 41567 * function: Functions. (line 6) 41568 * function body: Function Bodies. (line 6) 41569 * function call conventions: Interface. (line 6) 41570 * function entry and exit: Function Entry. (line 6) 41571 * function entry point, alternate function entry point: Edges. 41572 (line 180) 41573 * function-call insns: Calls. (line 6) 41574 * FUNCTION_ARG: Register Arguments. (line 11) 41575 * FUNCTION_ARG_ADVANCE: Register Arguments. (line 185) 41576 * FUNCTION_ARG_BOUNDARY: Register Arguments. (line 238) 41577 * FUNCTION_ARG_OFFSET: Register Arguments. (line 196) 41578 * FUNCTION_ARG_PADDING: Register Arguments. (line 203) 41579 * FUNCTION_ARG_REGNO_P: Register Arguments. (line 243) 41580 * FUNCTION_BOUNDARY: Storage Layout. (line 170) 41581 * FUNCTION_DECL: Functions. (line 6) 41582 * FUNCTION_INCOMING_ARG: Register Arguments. (line 68) 41583 * FUNCTION_MODE: Misc. (line 356) 41584 * FUNCTION_OUTGOING_VALUE: Scalar Return. (line 56) 41585 * FUNCTION_PROFILER: Profiling. (line 9) 41586 * FUNCTION_TYPE: Types. (line 6) 41587 * FUNCTION_VALUE: Scalar Return. (line 52) 41588 * FUNCTION_VALUE_REGNO_P: Scalar Return. (line 81) 41589 * functions, leaf: Leaf Functions. (line 6) 41590 * fundamental type: Types. (line 6) 41591 * g in constraint: Simple Constraints. (line 110) 41592 * G in constraint: Simple Constraints. (line 88) 41593 * garbage collector, invocation: Invoking the garbage collector. 41594 (line 6) 41595 * GCC and portability: Portability. (line 6) 41596 * GCC_DRIVER_HOST_INITIALIZATION: Host Misc. (line 36) 41597 * gcov_type: Profile information. 41598 (line 41) 41599 * ge: Comparisons. (line 72) 41600 * ge and attributes: Expressions. (line 64) 41601 * GE_EXPR: Expression trees. (line 6) 41602 * GEN_ERRNO_RTX: Library Calls. (line 71) 41603 * gencodes: RTL passes. (line 18) 41604 * general_operand: Machine-Independent Predicates. 41605 (line 105) 41606 * GENERAL_REGS: Register Classes. (line 23) 41607 * generated files: Files. (line 6) 41608 * generating assembler output: Output Statement. (line 6) 41609 * generating insns: RTL Template. (line 6) 41610 * GENERIC <1>: GENERIC. (line 6) 41611 * GENERIC <2>: Gimplification pass. 41612 (line 12) 41613 * GENERIC: Parsing pass. (line 6) 41614 * generic predicates: Machine-Independent Predicates. 41615 (line 6) 41616 * genflags: RTL passes. (line 18) 41617 * get_attr: Expressions. (line 80) 41618 * get_attr_length: Insn Lengths. (line 46) 41619 * GET_CLASS_NARROWEST_MODE: Machine Modes. (line 333) 41620 * GET_CODE: RTL Objects. (line 47) 41621 * get_frame_size: Elimination. (line 31) 41622 * get_insns: Insns. (line 34) 41623 * get_last_insn: Insns. (line 34) 41624 * GET_MODE: Machine Modes. (line 280) 41625 * GET_MODE_ALIGNMENT: Machine Modes. (line 320) 41626 * GET_MODE_BITSIZE: Machine Modes. (line 304) 41627 * GET_MODE_CLASS: Machine Modes. (line 294) 41628 * GET_MODE_FBIT: Machine Modes. (line 311) 41629 * GET_MODE_IBIT: Machine Modes. (line 307) 41630 * GET_MODE_MASK: Machine Modes. (line 315) 41631 * GET_MODE_NAME: Machine Modes. (line 291) 41632 * GET_MODE_NUNITS: Machine Modes. (line 329) 41633 * GET_MODE_SIZE: Machine Modes. (line 301) 41634 * GET_MODE_UNIT_SIZE: Machine Modes. (line 323) 41635 * GET_MODE_WIDER_MODE: Machine Modes. (line 297) 41636 * GET_RTX_CLASS: RTL Classes. (line 6) 41637 * GET_RTX_FORMAT: RTL Classes. (line 130) 41638 * GET_RTX_LENGTH: RTL Classes. (line 127) 41639 * geu: Comparisons. (line 72) 41640 * geu and attributes: Expressions. (line 64) 41641 * GGC: Type Information. (line 6) 41642 * ggc_collect: Invoking the garbage collector. 41643 (line 6) 41644 * GIMPLE <1>: GIMPLE. (line 6) 41645 * GIMPLE <2>: Gimplification pass. 41646 (line 6) 41647 * GIMPLE: Parsing pass. (line 14) 41648 * GIMPLE Exception Handling: GIMPLE Exception Handling. 41649 (line 6) 41650 * GIMPLE instruction set: GIMPLE instruction set. 41651 (line 6) 41652 * GIMPLE sequences: GIMPLE sequences. (line 6) 41653 * gimple_addresses_taken: Manipulating GIMPLE statements. 41654 (line 90) 41655 * GIMPLE_ASM: GIMPLE_ASM. (line 6) 41656 * gimple_asm_clear_volatile: GIMPLE_ASM. (line 63) 41657 * gimple_asm_clobber_op: GIMPLE_ASM. (line 46) 41658 * gimple_asm_input_op: GIMPLE_ASM. (line 30) 41659 * gimple_asm_output_op: GIMPLE_ASM. (line 38) 41660 * gimple_asm_set_clobber_op: GIMPLE_ASM. (line 50) 41661 * gimple_asm_set_input_op: GIMPLE_ASM. (line 34) 41662 * gimple_asm_set_output_op: GIMPLE_ASM. (line 42) 41663 * gimple_asm_set_volatile: GIMPLE_ASM. (line 60) 41664 * gimple_asm_volatile_p: GIMPLE_ASM. (line 57) 41665 * GIMPLE_ASSIGN: GIMPLE_ASSIGN. (line 6) 41666 * gimple_assign_cast_p: GIMPLE_ASSIGN. (line 89) 41667 * gimple_assign_lhs: GIMPLE_ASSIGN. (line 51) 41668 * gimple_assign_rhs1: GIMPLE_ASSIGN. (line 57) 41669 * gimple_assign_rhs2: GIMPLE_ASSIGN. (line 64) 41670 * gimple_assign_set_lhs: GIMPLE_ASSIGN. (line 71) 41671 * gimple_assign_set_rhs1: GIMPLE_ASSIGN. (line 74) 41672 * gimple_assign_set_rhs2: GIMPLE_ASSIGN. (line 85) 41673 * gimple_bb: Manipulating GIMPLE statements. 41674 (line 18) 41675 * GIMPLE_BIND: GIMPLE_BIND. (line 6) 41676 * gimple_bind_add_seq: GIMPLE_BIND. (line 36) 41677 * gimple_bind_add_stmt: GIMPLE_BIND. (line 32) 41678 * gimple_bind_append_vars: GIMPLE_BIND. (line 19) 41679 * gimple_bind_block: GIMPLE_BIND. (line 40) 41680 * gimple_bind_body: GIMPLE_BIND. (line 23) 41681 * gimple_bind_set_block: GIMPLE_BIND. (line 45) 41682 * gimple_bind_set_body: GIMPLE_BIND. (line 28) 41683 * gimple_bind_set_vars: GIMPLE_BIND. (line 15) 41684 * gimple_bind_vars: GIMPLE_BIND. (line 12) 41685 * gimple_block: Manipulating GIMPLE statements. 41686 (line 21) 41687 * gimple_build_asm: GIMPLE_ASM. (line 8) 41688 * gimple_build_asm_vec: GIMPLE_ASM. (line 17) 41689 * gimple_build_assign: GIMPLE_ASSIGN. (line 7) 41690 * gimple_build_assign_with_ops: GIMPLE_ASSIGN. (line 30) 41691 * gimple_build_bind: GIMPLE_BIND. (line 8) 41692 * gimple_build_call: GIMPLE_CALL. (line 8) 41693 * gimple_build_call_from_tree: GIMPLE_CALL. (line 16) 41694 * gimple_build_call_vec: GIMPLE_CALL. (line 25) 41695 * gimple_build_catch: GIMPLE_CATCH. (line 8) 41696 * gimple_build_cdt: GIMPLE_CHANGE_DYNAMIC_TYPE. 41697 (line 7) 41698 * gimple_build_cond: GIMPLE_COND. (line 8) 41699 * gimple_build_cond_from_tree: GIMPLE_COND. (line 16) 41700 * gimple_build_eh_filter: GIMPLE_EH_FILTER. (line 8) 41701 * gimple_build_goto: GIMPLE_LABEL. (line 18) 41702 * gimple_build_label: GIMPLE_LABEL. (line 7) 41703 * gimple_build_nop: GIMPLE_NOP. (line 7) 41704 * gimple_build_omp_atomic_load: GIMPLE_OMP_ATOMIC_LOAD. 41705 (line 8) 41706 * gimple_build_omp_atomic_store: GIMPLE_OMP_ATOMIC_STORE. 41707 (line 7) 41708 * gimple_build_omp_continue: GIMPLE_OMP_CONTINUE. 41709 (line 8) 41710 * gimple_build_omp_critical: GIMPLE_OMP_CRITICAL. 41711 (line 8) 41712 * gimple_build_omp_for: GIMPLE_OMP_FOR. (line 9) 41713 * gimple_build_omp_master: GIMPLE_OMP_MASTER. (line 7) 41714 * gimple_build_omp_ordered: GIMPLE_OMP_ORDERED. (line 7) 41715 * gimple_build_omp_parallel: GIMPLE_OMP_PARALLEL. 41716 (line 8) 41717 * gimple_build_omp_return: GIMPLE_OMP_RETURN. (line 7) 41718 * gimple_build_omp_section: GIMPLE_OMP_SECTION. (line 7) 41719 * gimple_build_omp_sections: GIMPLE_OMP_SECTIONS. 41720 (line 8) 41721 * gimple_build_omp_sections_switch: GIMPLE_OMP_SECTIONS. 41722 (line 14) 41723 * gimple_build_omp_single: GIMPLE_OMP_SINGLE. (line 8) 41724 * gimple_build_resx: GIMPLE_RESX. (line 7) 41725 * gimple_build_return: GIMPLE_RETURN. (line 7) 41726 * gimple_build_switch: GIMPLE_SWITCH. (line 8) 41727 * gimple_build_switch_vec: GIMPLE_SWITCH. (line 16) 41728 * gimple_build_try: GIMPLE_TRY. (line 8) 41729 * gimple_build_wce: GIMPLE_WITH_CLEANUP_EXPR. 41730 (line 7) 41731 * GIMPLE_CALL: GIMPLE_CALL. (line 6) 41732 * gimple_call_arg: GIMPLE_CALL. (line 66) 41733 * gimple_call_cannot_inline_p: GIMPLE_CALL. (line 91) 41734 * gimple_call_chain: GIMPLE_CALL. (line 57) 41735 * gimple_call_copy_skip_args: GIMPLE_CALL. (line 98) 41736 * gimple_call_fn: GIMPLE_CALL. (line 38) 41737 * gimple_call_fndecl: GIMPLE_CALL. (line 46) 41738 * gimple_call_lhs: GIMPLE_CALL. (line 29) 41739 * gimple_call_mark_uninlinable: GIMPLE_CALL. (line 88) 41740 * gimple_call_noreturn_p: GIMPLE_CALL. (line 94) 41741 * gimple_call_return_type: GIMPLE_CALL. (line 54) 41742 * gimple_call_set_arg: GIMPLE_CALL. (line 76) 41743 * gimple_call_set_chain: GIMPLE_CALL. (line 60) 41744 * gimple_call_set_fn: GIMPLE_CALL. (line 42) 41745 * gimple_call_set_fndecl: GIMPLE_CALL. (line 51) 41746 * gimple_call_set_lhs: GIMPLE_CALL. (line 35) 41747 * gimple_call_set_tail: GIMPLE_CALL. (line 80) 41748 * gimple_call_tail_p: GIMPLE_CALL. (line 85) 41749 * GIMPLE_CATCH: GIMPLE_CATCH. (line 6) 41750 * gimple_catch_handler: GIMPLE_CATCH. (line 20) 41751 * gimple_catch_set_handler: GIMPLE_CATCH. (line 28) 41752 * gimple_catch_set_types: GIMPLE_CATCH. (line 24) 41753 * gimple_catch_types: GIMPLE_CATCH. (line 13) 41754 * gimple_cdt_location: GIMPLE_CHANGE_DYNAMIC_TYPE. 41755 (line 24) 41756 * gimple_cdt_new_type: GIMPLE_CHANGE_DYNAMIC_TYPE. 41757 (line 11) 41758 * gimple_cdt_set_location: GIMPLE_CHANGE_DYNAMIC_TYPE. 41759 (line 32) 41760 * gimple_cdt_set_new_type: GIMPLE_CHANGE_DYNAMIC_TYPE. 41761 (line 20) 41762 * GIMPLE_CHANGE_DYNAMIC_TYPE: GIMPLE_CHANGE_DYNAMIC_TYPE. 41763 (line 6) 41764 * gimple_code: Manipulating GIMPLE statements. 41765 (line 15) 41766 * GIMPLE_COND: GIMPLE_COND. (line 6) 41767 * gimple_cond_false_label: GIMPLE_COND. (line 60) 41768 * gimple_cond_lhs: GIMPLE_COND. (line 30) 41769 * gimple_cond_make_false: GIMPLE_COND. (line 64) 41770 * gimple_cond_make_true: GIMPLE_COND. (line 67) 41771 * gimple_cond_rhs: GIMPLE_COND. (line 38) 41772 * gimple_cond_set_code: GIMPLE_COND. (line 26) 41773 * gimple_cond_set_false_label: GIMPLE_COND. (line 56) 41774 * gimple_cond_set_lhs: GIMPLE_COND. (line 34) 41775 * gimple_cond_set_rhs: GIMPLE_COND. (line 42) 41776 * gimple_cond_set_true_label: GIMPLE_COND. (line 51) 41777 * gimple_cond_true_label: GIMPLE_COND. (line 46) 41778 * gimple_copy: Manipulating GIMPLE statements. 41779 (line 147) 41780 * GIMPLE_EH_FILTER: GIMPLE_EH_FILTER. (line 6) 41781 * gimple_eh_filter_failure: GIMPLE_EH_FILTER. (line 19) 41782 * gimple_eh_filter_must_not_throw: GIMPLE_EH_FILTER. (line 33) 41783 * gimple_eh_filter_set_failure: GIMPLE_EH_FILTER. (line 29) 41784 * gimple_eh_filter_set_must_not_throw: GIMPLE_EH_FILTER. (line 37) 41785 * gimple_eh_filter_set_types: GIMPLE_EH_FILTER. (line 24) 41786 * gimple_eh_filter_types: GIMPLE_EH_FILTER. (line 12) 41787 * gimple_expr_type: Manipulating GIMPLE statements. 41788 (line 24) 41789 * gimple_goto_dest: GIMPLE_LABEL. (line 21) 41790 * gimple_goto_set_dest: GIMPLE_LABEL. (line 24) 41791 * gimple_has_mem_ops: Manipulating GIMPLE statements. 41792 (line 72) 41793 * gimple_has_ops: Manipulating GIMPLE statements. 41794 (line 69) 41795 * gimple_has_volatile_ops: Manipulating GIMPLE statements. 41796 (line 134) 41797 * GIMPLE_LABEL: GIMPLE_LABEL. (line 6) 41798 * gimple_label_label: GIMPLE_LABEL. (line 11) 41799 * gimple_label_set_label: GIMPLE_LABEL. (line 14) 41800 * gimple_loaded_syms: Manipulating GIMPLE statements. 41801 (line 122) 41802 * gimple_locus: Manipulating GIMPLE statements. 41803 (line 42) 41804 * gimple_locus_empty_p: Manipulating GIMPLE statements. 41805 (line 48) 41806 * gimple_modified_p: Manipulating GIMPLE statements. 41807 (line 130) 41808 * gimple_no_warning_p: Manipulating GIMPLE statements. 41809 (line 51) 41810 * GIMPLE_NOP: GIMPLE_NOP. (line 6) 41811 * gimple_nop_p: GIMPLE_NOP. (line 10) 41812 * gimple_num_ops <1>: Manipulating GIMPLE statements. 41813 (line 75) 41814 * gimple_num_ops: Logical Operators. (line 76) 41815 * GIMPLE_OMP_ATOMIC_LOAD: GIMPLE_OMP_ATOMIC_LOAD. 41816 (line 6) 41817 * gimple_omp_atomic_load_lhs: GIMPLE_OMP_ATOMIC_LOAD. 41818 (line 17) 41819 * gimple_omp_atomic_load_rhs: GIMPLE_OMP_ATOMIC_LOAD. 41820 (line 24) 41821 * gimple_omp_atomic_load_set_lhs: GIMPLE_OMP_ATOMIC_LOAD. 41822 (line 14) 41823 * gimple_omp_atomic_load_set_rhs: GIMPLE_OMP_ATOMIC_LOAD. 41824 (line 21) 41825 * GIMPLE_OMP_ATOMIC_STORE: GIMPLE_OMP_ATOMIC_STORE. 41826 (line 6) 41827 * gimple_omp_atomic_store_set_val: GIMPLE_OMP_ATOMIC_STORE. 41828 (line 12) 41829 * gimple_omp_atomic_store_val: GIMPLE_OMP_ATOMIC_STORE. 41830 (line 15) 41831 * gimple_omp_body: GIMPLE_OMP_PARALLEL. 41832 (line 24) 41833 * GIMPLE_OMP_CONTINUE: GIMPLE_OMP_CONTINUE. 41834 (line 6) 41835 * gimple_omp_continue_control_def: GIMPLE_OMP_CONTINUE. 41836 (line 13) 41837 * gimple_omp_continue_control_def_ptr: GIMPLE_OMP_CONTINUE. 41838 (line 17) 41839 * gimple_omp_continue_control_use: GIMPLE_OMP_CONTINUE. 41840 (line 24) 41841 * gimple_omp_continue_control_use_ptr: GIMPLE_OMP_CONTINUE. 41842 (line 28) 41843 * gimple_omp_continue_set_control_def: GIMPLE_OMP_CONTINUE. 41844 (line 20) 41845 * gimple_omp_continue_set_control_use: GIMPLE_OMP_CONTINUE. 41846 (line 31) 41847 * GIMPLE_OMP_CRITICAL: GIMPLE_OMP_CRITICAL. 41848 (line 6) 41849 * gimple_omp_critical_name: GIMPLE_OMP_CRITICAL. 41850 (line 13) 41851 * gimple_omp_critical_set_name: GIMPLE_OMP_CRITICAL. 41852 (line 21) 41853 * GIMPLE_OMP_FOR: GIMPLE_OMP_FOR. (line 6) 41854 * gimple_omp_for_clauses: GIMPLE_OMP_FOR. (line 20) 41855 * gimple_omp_for_final: GIMPLE_OMP_FOR. (line 51) 41856 * gimple_omp_for_incr: GIMPLE_OMP_FOR. (line 61) 41857 * gimple_omp_for_index: GIMPLE_OMP_FOR. (line 31) 41858 * gimple_omp_for_initial: GIMPLE_OMP_FOR. (line 41) 41859 * gimple_omp_for_pre_body: GIMPLE_OMP_FOR. (line 70) 41860 * gimple_omp_for_set_clauses: GIMPLE_OMP_FOR. (line 27) 41861 * gimple_omp_for_set_cond: GIMPLE_OMP_FOR. (line 80) 41862 * gimple_omp_for_set_final: GIMPLE_OMP_FOR. (line 58) 41863 * gimple_omp_for_set_incr: GIMPLE_OMP_FOR. (line 67) 41864 * gimple_omp_for_set_index: GIMPLE_OMP_FOR. (line 38) 41865 * gimple_omp_for_set_initial: GIMPLE_OMP_FOR. (line 48) 41866 * gimple_omp_for_set_pre_body: GIMPLE_OMP_FOR. (line 75) 41867 * GIMPLE_OMP_MASTER: GIMPLE_OMP_MASTER. (line 6) 41868 * GIMPLE_OMP_ORDERED: GIMPLE_OMP_ORDERED. (line 6) 41869 * GIMPLE_OMP_PARALLEL: GIMPLE_OMP_PARALLEL. 41870 (line 6) 41871 * gimple_omp_parallel_child_fn: GIMPLE_OMP_PARALLEL. 41872 (line 42) 41873 * gimple_omp_parallel_clauses: GIMPLE_OMP_PARALLEL. 41874 (line 31) 41875 * gimple_omp_parallel_combined_p: GIMPLE_OMP_PARALLEL. 41876 (line 16) 41877 * gimple_omp_parallel_data_arg: GIMPLE_OMP_PARALLEL. 41878 (line 54) 41879 * gimple_omp_parallel_set_child_fn: GIMPLE_OMP_PARALLEL. 41880 (line 51) 41881 * gimple_omp_parallel_set_clauses: GIMPLE_OMP_PARALLEL. 41882 (line 38) 41883 * gimple_omp_parallel_set_combined_p: GIMPLE_OMP_PARALLEL. 41884 (line 20) 41885 * gimple_omp_parallel_set_data_arg: GIMPLE_OMP_PARALLEL. 41886 (line 62) 41887 * GIMPLE_OMP_RETURN: GIMPLE_OMP_RETURN. (line 6) 41888 * gimple_omp_return_nowait_p: GIMPLE_OMP_RETURN. (line 14) 41889 * gimple_omp_return_set_nowait: GIMPLE_OMP_RETURN. (line 11) 41890 * GIMPLE_OMP_SECTION: GIMPLE_OMP_SECTION. (line 6) 41891 * gimple_omp_section_last_p: GIMPLE_OMP_SECTION. (line 12) 41892 * gimple_omp_section_set_last: GIMPLE_OMP_SECTION. (line 16) 41893 * GIMPLE_OMP_SECTIONS: GIMPLE_OMP_SECTIONS. 41894 (line 6) 41895 * gimple_omp_sections_clauses: GIMPLE_OMP_SECTIONS. 41896 (line 30) 41897 * gimple_omp_sections_control: GIMPLE_OMP_SECTIONS. 41898 (line 17) 41899 * gimple_omp_sections_set_clauses: GIMPLE_OMP_SECTIONS. 41900 (line 37) 41901 * gimple_omp_sections_set_control: GIMPLE_OMP_SECTIONS. 41902 (line 26) 41903 * gimple_omp_set_body: GIMPLE_OMP_PARALLEL. 41904 (line 28) 41905 * GIMPLE_OMP_SINGLE: GIMPLE_OMP_SINGLE. (line 6) 41906 * gimple_omp_single_clauses: GIMPLE_OMP_SINGLE. (line 14) 41907 * gimple_omp_single_set_clauses: GIMPLE_OMP_SINGLE. (line 21) 41908 * gimple_op <1>: Manipulating GIMPLE statements. 41909 (line 81) 41910 * gimple_op: Logical Operators. (line 79) 41911 * GIMPLE_PHI: GIMPLE_PHI. (line 6) 41912 * gimple_phi_capacity: GIMPLE_PHI. (line 10) 41913 * gimple_phi_num_args: GIMPLE_PHI. (line 14) 41914 * gimple_phi_result: GIMPLE_PHI. (line 19) 41915 * gimple_phi_set_arg: GIMPLE_PHI. (line 33) 41916 * gimple_phi_set_result: GIMPLE_PHI. (line 25) 41917 * GIMPLE_RESX: GIMPLE_RESX. (line 6) 41918 * gimple_resx_region: GIMPLE_RESX. (line 13) 41919 * gimple_resx_set_region: GIMPLE_RESX. (line 16) 41920 * GIMPLE_RETURN: GIMPLE_RETURN. (line 6) 41921 * gimple_return_retval: GIMPLE_RETURN. (line 10) 41922 * gimple_return_set_retval: GIMPLE_RETURN. (line 14) 41923 * gimple_rhs_class: GIMPLE_ASSIGN. (line 46) 41924 * gimple_seq_add_seq: GIMPLE sequences. (line 32) 41925 * gimple_seq_add_stmt: GIMPLE sequences. (line 26) 41926 * gimple_seq_alloc: GIMPLE sequences. (line 62) 41927 * gimple_seq_copy: GIMPLE sequences. (line 67) 41928 * gimple_seq_deep_copy: GIMPLE sequences. (line 37) 41929 * gimple_seq_empty_p: GIMPLE sequences. (line 70) 41930 * gimple_seq_first: GIMPLE sequences. (line 44) 41931 * gimple_seq_init: GIMPLE sequences. (line 59) 41932 * gimple_seq_last: GIMPLE sequences. (line 47) 41933 * gimple_seq_reverse: GIMPLE sequences. (line 40) 41934 * gimple_seq_set_first: GIMPLE sequences. (line 55) 41935 * gimple_seq_set_last: GIMPLE sequences. (line 51) 41936 * gimple_seq_singleton_p: GIMPLE sequences. (line 79) 41937 * gimple_set_block: Manipulating GIMPLE statements. 41938 (line 39) 41939 * gimple_set_def_ops: Manipulating GIMPLE statements. 41940 (line 98) 41941 * gimple_set_has_volatile_ops: Manipulating GIMPLE statements. 41942 (line 138) 41943 * gimple_set_locus: Manipulating GIMPLE statements. 41944 (line 45) 41945 * gimple_set_op: Manipulating GIMPLE statements. 41946 (line 87) 41947 * gimple_set_plf: Manipulating GIMPLE statements. 41948 (line 62) 41949 * gimple_set_use_ops: Manipulating GIMPLE statements. 41950 (line 105) 41951 * gimple_set_vdef_ops: Manipulating GIMPLE statements. 41952 (line 119) 41953 * gimple_set_visited: Manipulating GIMPLE statements. 41954 (line 55) 41955 * gimple_set_vuse_ops: Manipulating GIMPLE statements. 41956 (line 112) 41957 * gimple_statement_base: Tuple representation. 41958 (line 14) 41959 * gimple_statement_with_ops: Tuple representation. 41960 (line 96) 41961 * gimple_stored_syms: Manipulating GIMPLE statements. 41962 (line 126) 41963 * GIMPLE_SWITCH: GIMPLE_SWITCH. (line 6) 41964 * gimple_switch_default_label: GIMPLE_SWITCH. (line 46) 41965 * gimple_switch_index: GIMPLE_SWITCH. (line 31) 41966 * gimple_switch_label: GIMPLE_SWITCH. (line 37) 41967 * gimple_switch_num_labels: GIMPLE_SWITCH. (line 22) 41968 * gimple_switch_set_default_label: GIMPLE_SWITCH. (line 50) 41969 * gimple_switch_set_index: GIMPLE_SWITCH. (line 34) 41970 * gimple_switch_set_label: GIMPLE_SWITCH. (line 42) 41971 * gimple_switch_set_num_labels: GIMPLE_SWITCH. (line 27) 41972 * GIMPLE_TRY: GIMPLE_TRY. (line 6) 41973 * gimple_try_catch_is_cleanup: GIMPLE_TRY. (line 20) 41974 * gimple_try_cleanup: GIMPLE_TRY. (line 27) 41975 * gimple_try_eval: GIMPLE_TRY. (line 23) 41976 * gimple_try_flags: GIMPLE_TRY. (line 16) 41977 * gimple_try_set_catch_is_cleanup: GIMPLE_TRY. (line 32) 41978 * gimple_try_set_cleanup: GIMPLE_TRY. (line 41) 41979 * gimple_try_set_eval: GIMPLE_TRY. (line 36) 41980 * gimple_visited_p: Manipulating GIMPLE statements. 41981 (line 58) 41982 * gimple_wce_cleanup: GIMPLE_WITH_CLEANUP_EXPR. 41983 (line 11) 41984 * gimple_wce_cleanup_eh_only: GIMPLE_WITH_CLEANUP_EXPR. 41985 (line 18) 41986 * gimple_wce_set_cleanup: GIMPLE_WITH_CLEANUP_EXPR. 41987 (line 15) 41988 * gimple_wce_set_cleanup_eh_only: GIMPLE_WITH_CLEANUP_EXPR. 41989 (line 22) 41990 * GIMPLE_WITH_CLEANUP_EXPR: GIMPLE_WITH_CLEANUP_EXPR. 41991 (line 6) 41992 * gimplification <1>: Gimplification pass. 41993 (line 6) 41994 * gimplification: Parsing pass. (line 14) 41995 * gimplifier: Parsing pass. (line 14) 41996 * gimplify_assign: GIMPLE_ASSIGN. (line 19) 41997 * gimplify_expr: Gimplification pass. 41998 (line 18) 41999 * gimplify_function_tree: Gimplification pass. 42000 (line 18) 42001 * GLOBAL_INIT_PRIORITY: Function Basics. (line 6) 42002 * global_regs: Register Basics. (line 59) 42003 * GO_IF_LEGITIMATE_ADDRESS: Addressing Modes. (line 48) 42004 * GO_IF_MODE_DEPENDENT_ADDRESS: Addressing Modes. (line 190) 42005 * GOFAST, floating point emulation library: Library Calls. (line 44) 42006 * gofast_maybe_init_libfuncs: Library Calls. (line 44) 42007 * greater than: Comparisons. (line 60) 42008 * gsi_after_labels: Sequence iterators. (line 76) 42009 * gsi_bb: Sequence iterators. (line 83) 42010 * gsi_commit_edge_inserts: Sequence iterators. (line 194) 42011 * gsi_commit_one_edge_insert: Sequence iterators. (line 190) 42012 * gsi_end_p: Sequence iterators. (line 60) 42013 * gsi_for_stmt: Sequence iterators. (line 157) 42014 * gsi_insert_after: Sequence iterators. (line 147) 42015 * gsi_insert_before: Sequence iterators. (line 136) 42016 * gsi_insert_on_edge: Sequence iterators. (line 174) 42017 * gsi_insert_on_edge_immediate: Sequence iterators. (line 185) 42018 * gsi_insert_seq_after: Sequence iterators. (line 154) 42019 * gsi_insert_seq_before: Sequence iterators. (line 143) 42020 * gsi_insert_seq_on_edge: Sequence iterators. (line 179) 42021 * gsi_last: Sequence iterators. (line 50) 42022 * gsi_last_bb: Sequence iterators. (line 56) 42023 * gsi_link_after: Sequence iterators. (line 115) 42024 * gsi_link_before: Sequence iterators. (line 105) 42025 * gsi_link_seq_after: Sequence iterators. (line 110) 42026 * gsi_link_seq_before: Sequence iterators. (line 99) 42027 * gsi_move_after: Sequence iterators. (line 161) 42028 * gsi_move_before: Sequence iterators. (line 166) 42029 * gsi_move_to_bb_end: Sequence iterators. (line 171) 42030 * gsi_next: Sequence iterators. (line 66) 42031 * gsi_one_before_end_p: Sequence iterators. (line 63) 42032 * gsi_prev: Sequence iterators. (line 69) 42033 * gsi_remove: Sequence iterators. (line 90) 42034 * gsi_replace: Sequence iterators. (line 130) 42035 * gsi_seq: Sequence iterators. (line 86) 42036 * gsi_split_seq_after: Sequence iterators. (line 120) 42037 * gsi_split_seq_before: Sequence iterators. (line 125) 42038 * gsi_start: Sequence iterators. (line 40) 42039 * gsi_start_bb: Sequence iterators. (line 46) 42040 * gsi_stmt: Sequence iterators. (line 72) 42041 * gt: Comparisons. (line 60) 42042 * gt and attributes: Expressions. (line 64) 42043 * GT_EXPR: Expression trees. (line 6) 42044 * gtu: Comparisons. (line 64) 42045 * gtu and attributes: Expressions. (line 64) 42046 * GTY: Type Information. (line 6) 42047 * H in constraint: Simple Constraints. (line 88) 42048 * HAmode: Machine Modes. (line 144) 42049 * HANDLE_PRAGMA_PACK_PUSH_POP: Misc. (line 467) 42050 * HANDLE_PRAGMA_PACK_WITH_EXPANSION: Misc. (line 478) 42051 * HANDLE_SYSV_PRAGMA: Misc. (line 438) 42052 * HANDLER: Function Bodies. (line 6) 42053 * HANDLER_BODY: Function Bodies. (line 6) 42054 * HANDLER_PARMS: Function Bodies. (line 6) 42055 * hard registers: Regs and Memory. (line 9) 42056 * HARD_FRAME_POINTER_REGNUM: Frame Registers. (line 20) 42057 * HARD_REGNO_CALL_PART_CLOBBERED: Register Basics. (line 53) 42058 * HARD_REGNO_CALLER_SAVE_MODE: Caller Saves. (line 20) 42059 * HARD_REGNO_MODE_OK: Values in Registers. 42060 (line 58) 42061 * HARD_REGNO_NREGS: Values in Registers. 42062 (line 11) 42063 * HARD_REGNO_NREGS_HAS_PADDING: Values in Registers. 42064 (line 25) 42065 * HARD_REGNO_NREGS_WITH_PADDING: Values in Registers. 42066 (line 43) 42067 * HARD_REGNO_RENAME_OK: Values in Registers. 42068 (line 119) 42069 * HAS_INIT_SECTION: Macros for Initialization. 42070 (line 19) 42071 * HAS_LONG_COND_BRANCH: Misc. (line 9) 42072 * HAS_LONG_UNCOND_BRANCH: Misc. (line 18) 42073 * HAVE_DOS_BASED_FILE_SYSTEM: Filesystem. (line 11) 42074 * HAVE_POST_DECREMENT: Addressing Modes. (line 12) 42075 * HAVE_POST_INCREMENT: Addressing Modes. (line 11) 42076 * HAVE_POST_MODIFY_DISP: Addressing Modes. (line 18) 42077 * HAVE_POST_MODIFY_REG: Addressing Modes. (line 24) 42078 * HAVE_PRE_DECREMENT: Addressing Modes. (line 10) 42079 * HAVE_PRE_INCREMENT: Addressing Modes. (line 9) 42080 * HAVE_PRE_MODIFY_DISP: Addressing Modes. (line 17) 42081 * HAVE_PRE_MODIFY_REG: Addressing Modes. (line 23) 42082 * HCmode: Machine Modes. (line 197) 42083 * HFmode: Machine Modes. (line 58) 42084 * high: Constants. (line 109) 42085 * HImode: Machine Modes. (line 29) 42086 * HImode, in insn: Insns. (line 231) 42087 * HONOR_REG_ALLOC_ORDER: Allocation Order. (line 37) 42088 * host configuration: Host Config. (line 6) 42089 * host functions: Host Common. (line 6) 42090 * host hooks: Host Common. (line 6) 42091 * host makefile fragment: Host Fragment. (line 6) 42092 * HOST_BIT_BUCKET: Filesystem. (line 51) 42093 * HOST_EXECUTABLE_SUFFIX: Filesystem. (line 45) 42094 * HOST_HOOKS_EXTRA_SIGNALS: Host Common. (line 12) 42095 * HOST_HOOKS_GT_PCH_ALLOC_GRANULARITY: Host Common. (line 45) 42096 * HOST_HOOKS_GT_PCH_USE_ADDRESS: Host Common. (line 26) 42097 * HOST_LACKS_INODE_NUMBERS: Filesystem. (line 89) 42098 * HOST_LONG_LONG_FORMAT: Host Misc. (line 41) 42099 * HOST_OBJECT_SUFFIX: Filesystem. (line 40) 42100 * HOST_WIDE_INT: Anchored Addresses. (line 33) 42101 * HOT_TEXT_SECTION_NAME: Sections. (line 43) 42102 * HQmode: Machine Modes. (line 107) 42103 * I in constraint: Simple Constraints. (line 71) 42104 * i in constraint: Simple Constraints. (line 60) 42105 * identifier: Identifiers. (line 6) 42106 * IDENTIFIER_LENGTH: Identifiers. (line 20) 42107 * IDENTIFIER_NODE: Identifiers. (line 6) 42108 * IDENTIFIER_OPNAME_P: Identifiers. (line 25) 42109 * IDENTIFIER_POINTER: Identifiers. (line 15) 42110 * IDENTIFIER_TYPENAME_P: Identifiers. (line 31) 42111 * IEEE 754-2008: Decimal float library routines. 42112 (line 6) 42113 * IF_COND: Function Bodies. (line 6) 42114 * if_marked: GTY Options. (line 155) 42115 * IF_STMT: Function Bodies. (line 6) 42116 * if_then_else: Comparisons. (line 80) 42117 * if_then_else and attributes: Expressions. (line 32) 42118 * if_then_else usage: Side Effects. (line 56) 42119 * IFCVT_EXTRA_FIELDS: Misc. (line 619) 42120 * IFCVT_INIT_EXTRA_FIELDS: Misc. (line 614) 42121 * IFCVT_MODIFY_CANCEL: Misc. (line 608) 42122 * IFCVT_MODIFY_FINAL: Misc. (line 602) 42123 * IFCVT_MODIFY_INSN: Misc. (line 596) 42124 * IFCVT_MODIFY_MULTIPLE_TESTS: Misc. (line 589) 42125 * IFCVT_MODIFY_TESTS: Misc. (line 578) 42126 * IMAGPART_EXPR: Expression trees. (line 6) 42127 * Immediate Uses: SSA Operands. (line 274) 42128 * immediate_operand: Machine-Independent Predicates. 42129 (line 11) 42130 * IMMEDIATE_PREFIX: Instruction Output. (line 127) 42131 * in_struct: Flags. (line 258) 42132 * in_struct, in code_label and note: Flags. (line 59) 42133 * in_struct, in insn and jump_insn and call_insn: Flags. (line 49) 42134 * in_struct, in insn, jump_insn and call_insn: Flags. (line 166) 42135 * in_struct, in mem: Flags. (line 70) 42136 * in_struct, in subreg: Flags. (line 205) 42137 * include: Including Patterns. (line 6) 42138 * INCLUDE_DEFAULTS: Driver. (line 430) 42139 * inclusive-or, bitwise: Arithmetic. (line 158) 42140 * INCOMING_FRAME_SP_OFFSET: Frame Layout. (line 183) 42141 * INCOMING_REGNO: Register Basics. (line 91) 42142 * INCOMING_RETURN_ADDR_RTX: Frame Layout. (line 139) 42143 * INCOMING_STACK_BOUNDARY: Storage Layout. (line 165) 42144 * INDEX_REG_CLASS: Register Classes. (line 134) 42145 * indirect_jump instruction pattern: Standard Names. (line 1078) 42146 * indirect_operand: Machine-Independent Predicates. 42147 (line 71) 42148 * INDIRECT_REF: Expression trees. (line 6) 42149 * INIT_ARRAY_SECTION_ASM_OP: Sections. (line 98) 42150 * INIT_CUMULATIVE_ARGS: Register Arguments. (line 149) 42151 * INIT_CUMULATIVE_INCOMING_ARGS: Register Arguments. (line 176) 42152 * INIT_CUMULATIVE_LIBCALL_ARGS: Register Arguments. (line 170) 42153 * INIT_ENVIRONMENT: Driver. (line 369) 42154 * INIT_EXPANDERS: Per-Function Data. (line 39) 42155 * INIT_EXPR: Expression trees. (line 6) 42156 * init_machine_status: Per-Function Data. (line 45) 42157 * init_one_libfunc: Library Calls. (line 15) 42158 * INIT_SECTION_ASM_OP <1>: Macros for Initialization. 42159 (line 10) 42160 * INIT_SECTION_ASM_OP: Sections. (line 82) 42161 * INITIAL_ELIMINATION_OFFSET: Elimination. (line 79) 42162 * INITIAL_FRAME_ADDRESS_RTX: Frame Layout. (line 83) 42163 * INITIAL_FRAME_POINTER_OFFSET: Elimination. (line 32) 42164 * initialization routines: Initialization. (line 6) 42165 * INITIALIZE_TRAMPOLINE: Trampolines. (line 55) 42166 * inlining: Target Attributes. (line 86) 42167 * insert_insn_on_edge: Maintaining the CFG. 42168 (line 118) 42169 * insn: Insns. (line 63) 42170 * insn and /f: Flags. (line 125) 42171 * insn and /j: Flags. (line 175) 42172 * insn and /s: Flags. (line 49) 42173 * insn and /u: Flags. (line 39) 42174 * insn and /v: Flags. (line 44) 42175 * insn attributes: Insn Attributes. (line 6) 42176 * insn canonicalization: Insn Canonicalizations. 42177 (line 6) 42178 * insn includes: Including Patterns. (line 6) 42179 * insn lengths, computing: Insn Lengths. (line 6) 42180 * insn splitting: Insn Splitting. (line 6) 42181 * insn-attr.h: Defining Attributes. 42182 (line 24) 42183 * INSN_ANNULLED_BRANCH_P: Flags. (line 39) 42184 * INSN_CODE: Insns. (line 257) 42185 * INSN_DELETED_P: Flags. (line 44) 42186 * INSN_FROM_TARGET_P: Flags. (line 49) 42187 * insn_list: Insns. (line 505) 42188 * INSN_REFERENCES_ARE_DELAYED: Misc. (line 517) 42189 * INSN_SETS_ARE_DELAYED: Misc. (line 506) 42190 * INSN_UID: Insns. (line 23) 42191 * insns: Insns. (line 6) 42192 * insns, generating: RTL Template. (line 6) 42193 * insns, recognizing: RTL Template. (line 6) 42194 * instruction attributes: Insn Attributes. (line 6) 42195 * instruction latency time: Processor pipeline description. 42196 (line 6) 42197 * instruction patterns: Patterns. (line 6) 42198 * instruction splitting: Insn Splitting. (line 6) 42199 * insv instruction pattern: Standard Names. (line 880) 42200 * int <1>: Run-time Target. (line 56) 42201 * int: Manipulating GIMPLE statements. 42202 (line 66) 42203 * INT_TYPE_SIZE: Type Layout. (line 12) 42204 * INTEGER_CST: Expression trees. (line 6) 42205 * INTEGER_TYPE: Types. (line 6) 42206 * Interdependence of Patterns: Dependent Patterns. (line 6) 42207 * interfacing to GCC output: Interface. (line 6) 42208 * interlock delays: Processor pipeline description. 42209 (line 6) 42210 * intermediate representation lowering: Parsing pass. (line 14) 42211 * INTMAX_TYPE: Type Layout. (line 213) 42212 * introduction: Top. (line 6) 42213 * INVOKE__main: Macros for Initialization. 42214 (line 51) 42215 * ior: Arithmetic. (line 158) 42216 * ior and attributes: Expressions. (line 50) 42217 * ior, canonicalization of: Insn Canonicalizations. 42218 (line 57) 42219 * iorM3 instruction pattern: Standard Names. (line 222) 42220 * IRA_COVER_CLASSES: Register Classes. (line 516) 42221 * IRA_HARD_REGNO_ADD_COST_MULTIPLIER: Allocation Order. (line 45) 42222 * IS_ASM_LOGICAL_LINE_SEPARATOR: Data Output. (line 120) 42223 * is_gimple_omp: GIMPLE_OMP_PARALLEL. 42224 (line 65) 42225 * iterators in .md files: Iterators. (line 6) 42226 * IV analysis on GIMPLE: Scalar evolutions. (line 6) 42227 * IV analysis on RTL: loop-iv. (line 6) 42228 * jump: Flags. (line 309) 42229 * jump instruction pattern: Standard Names. (line 969) 42230 * jump instruction patterns: Jump Patterns. (line 6) 42231 * jump instructions and set: Side Effects. (line 56) 42232 * jump, in call_insn: Flags. (line 179) 42233 * jump, in insn: Flags. (line 175) 42234 * jump, in mem: Flags. (line 79) 42235 * JUMP_ALIGN: Alignment Output. (line 9) 42236 * jump_insn: Insns. (line 73) 42237 * jump_insn and /f: Flags. (line 125) 42238 * jump_insn and /s: Flags. (line 49) 42239 * jump_insn and /u: Flags. (line 39) 42240 * jump_insn and /v: Flags. (line 44) 42241 * JUMP_LABEL: Insns. (line 80) 42242 * JUMP_TABLES_IN_TEXT_SECTION: Sections. (line 142) 42243 * Jumps: Jumps. (line 6) 42244 * LABEL_ALIGN: Alignment Output. (line 52) 42245 * LABEL_ALIGN_AFTER_BARRIER: Alignment Output. (line 22) 42246 * LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP: Alignment Output. (line 30) 42247 * LABEL_ALIGN_MAX_SKIP: Alignment Output. (line 62) 42248 * LABEL_ALT_ENTRY_P: Insns. (line 140) 42249 * LABEL_ALTERNATE_NAME: Edges. (line 180) 42250 * LABEL_DECL: Declarations. (line 6) 42251 * LABEL_KIND: Insns. (line 140) 42252 * LABEL_NUSES: Insns. (line 136) 42253 * LABEL_PRESERVE_P: Flags. (line 59) 42254 * label_ref: Constants. (line 86) 42255 * label_ref and /v: Flags. (line 65) 42256 * label_ref, RTL sharing: Sharing. (line 35) 42257 * LABEL_REF_NONLOCAL_P: Flags. (line 65) 42258 * lang_hooks.gimplify_expr: Gimplification pass. 42259 (line 18) 42260 * lang_hooks.parse_file: Parsing pass. (line 6) 42261 * language-independent intermediate representation: Parsing pass. 42262 (line 14) 42263 * large return values: Aggregate Return. (line 6) 42264 * LARGEST_EXPONENT_IS_NORMAL: Storage Layout. (line 469) 42265 * LAST_STACK_REG: Stack Registers. (line 27) 42266 * LAST_VIRTUAL_REGISTER: Regs and Memory. (line 51) 42267 * lceilMN2: Standard Names. (line 597) 42268 * LCSSA: LCSSA. (line 6) 42269 * LD_FINI_SWITCH: Macros for Initialization. 42270 (line 29) 42271 * LD_INIT_SWITCH: Macros for Initialization. 42272 (line 25) 42273 * LDD_SUFFIX: Macros for Initialization. 42274 (line 116) 42275 * le: Comparisons. (line 76) 42276 * le and attributes: Expressions. (line 64) 42277 * LE_EXPR: Expression trees. (line 6) 42278 * leaf functions: Leaf Functions. (line 6) 42279 * leaf_function_p: Standard Names. (line 1040) 42280 * LEAF_REG_REMAP: Leaf Functions. (line 39) 42281 * LEAF_REGISTERS: Leaf Functions. (line 25) 42282 * left rotate: Arithmetic. (line 190) 42283 * left shift: Arithmetic. (line 168) 42284 * LEGITIMATE_CONSTANT_P: Addressing Modes. (line 205) 42285 * LEGITIMATE_PIC_OPERAND_P: PIC. (line 31) 42286 * LEGITIMIZE_ADDRESS: Addressing Modes. (line 122) 42287 * LEGITIMIZE_RELOAD_ADDRESS: Addressing Modes. (line 145) 42288 * length: GTY Options. (line 50) 42289 * less than: Comparisons. (line 68) 42290 * less than or equal: Comparisons. (line 76) 42291 * leu: Comparisons. (line 76) 42292 * leu and attributes: Expressions. (line 64) 42293 * lfloorMN2: Standard Names. (line 592) 42294 * LIB2FUNCS_EXTRA: Target Fragment. (line 11) 42295 * LIB_SPEC: Driver. (line 170) 42296 * LIBCALL_VALUE: Scalar Return. (line 60) 42297 * libgcc.a: Library Calls. (line 6) 42298 * LIBGCC2_CFLAGS: Target Fragment. (line 8) 42299 * LIBGCC2_HAS_DF_MODE: Type Layout. (line 109) 42300 * LIBGCC2_HAS_TF_MODE: Type Layout. (line 123) 42301 * LIBGCC2_HAS_XF_MODE: Type Layout. (line 117) 42302 * LIBGCC2_LONG_DOUBLE_TYPE_SIZE: Type Layout. (line 103) 42303 * LIBGCC2_UNWIND_ATTRIBUTE: Misc. (line 935) 42304 * LIBGCC2_WORDS_BIG_ENDIAN: Storage Layout. (line 36) 42305 * LIBGCC_SPEC: Driver. (line 178) 42306 * library subroutine names: Library Calls. (line 6) 42307 * LIBRARY_PATH_ENV: Misc. (line 557) 42308 * LIMIT_RELOAD_CLASS: Register Classes. (line 239) 42309 * Linear loop transformations framework: Lambda. (line 6) 42310 * LINK_COMMAND_SPEC: Driver. (line 299) 42311 * LINK_EH_SPEC: Driver. (line 205) 42312 * LINK_ELIMINATE_DUPLICATE_LDIRECTORIES: Driver. (line 309) 42313 * LINK_GCC_C_SEQUENCE_SPEC: Driver. (line 295) 42314 * LINK_LIBGCC_SPECIAL_1: Driver. (line 290) 42315 * LINK_SPEC: Driver. (line 163) 42316 * linkage: Function Basics. (line 6) 42317 * list: Containers. (line 6) 42318 * Liveness representation: Liveness information. 42319 (line 6) 42320 * lo_sum: Arithmetic. (line 24) 42321 * load address instruction: Simple Constraints. (line 154) 42322 * LOAD_EXTEND_OP: Misc. (line 69) 42323 * load_multiple instruction pattern: Standard Names. (line 137) 42324 * LOCAL_ALIGNMENT: Storage Layout. (line 254) 42325 * LOCAL_CLASS_P: Classes. (line 68) 42326 * LOCAL_DECL_ALIGNMENT: Storage Layout. (line 278) 42327 * LOCAL_INCLUDE_DIR: Driver. (line 376) 42328 * LOCAL_LABEL_PREFIX: Instruction Output. (line 125) 42329 * LOCAL_REGNO: Register Basics. (line 105) 42330 * LOG_LINKS: Insns. (line 276) 42331 * Logical Operators: Logical Operators. (line 6) 42332 * logical-and, bitwise: Arithmetic. (line 153) 42333 * logM2 instruction pattern: Standard Names. (line 505) 42334 * LONG_ACCUM_TYPE_SIZE: Type Layout. (line 93) 42335 * LONG_DOUBLE_TYPE_SIZE: Type Layout. (line 58) 42336 * LONG_FRACT_TYPE_SIZE: Type Layout. (line 73) 42337 * LONG_LONG_ACCUM_TYPE_SIZE: Type Layout. (line 98) 42338 * LONG_LONG_FRACT_TYPE_SIZE: Type Layout. (line 78) 42339 * LONG_LONG_TYPE_SIZE: Type Layout. (line 33) 42340 * LONG_TYPE_SIZE: Type Layout. (line 22) 42341 * longjmp and automatic variables: Interface. (line 52) 42342 * Loop analysis: Loop representation. 42343 (line 6) 42344 * Loop manipulation: Loop manipulation. (line 6) 42345 * Loop querying: Loop querying. (line 6) 42346 * Loop representation: Loop representation. 42347 (line 6) 42348 * Loop-closed SSA form: LCSSA. (line 6) 42349 * LOOP_ALIGN: Alignment Output. (line 35) 42350 * LOOP_ALIGN_MAX_SKIP: Alignment Output. (line 48) 42351 * LOOP_EXPR: Expression trees. (line 6) 42352 * looping instruction patterns: Looping Patterns. (line 6) 42353 * lowering, language-dependent intermediate representation: Parsing pass. 42354 (line 14) 42355 * lrintMN2: Standard Names. (line 582) 42356 * lroundMN2: Standard Names. (line 587) 42357 * LSHIFT_EXPR: Expression trees. (line 6) 42358 * lshiftrt: Arithmetic. (line 185) 42359 * lshiftrt and attributes: Expressions. (line 64) 42360 * lshrM3 instruction pattern: Standard Names. (line 441) 42361 * lt: Comparisons. (line 68) 42362 * lt and attributes: Expressions. (line 64) 42363 * LT_EXPR: Expression trees. (line 6) 42364 * LTGT_EXPR: Expression trees. (line 6) 42365 * ltu: Comparisons. (line 68) 42366 * m in constraint: Simple Constraints. (line 17) 42367 * machine attributes: Target Attributes. (line 6) 42368 * machine description macros: Target Macros. (line 6) 42369 * machine descriptions: Machine Desc. (line 6) 42370 * machine mode conversions: Conversions. (line 6) 42371 * machine modes: Machine Modes. (line 6) 42372 * machine specific constraints: Machine Constraints. 42373 (line 6) 42374 * machine-independent predicates: Machine-Independent Predicates. 42375 (line 6) 42376 * machine_mode: Condition Code. (line 157) 42377 * macros, target description: Target Macros. (line 6) 42378 * maddMN4 instruction pattern: Standard Names. (line 364) 42379 * MAKE_DECL_ONE_ONLY: Label Output. (line 218) 42380 * make_phi_node: GIMPLE_PHI. (line 7) 42381 * make_safe_from: Expander Definitions. 42382 (line 148) 42383 * makefile fragment: Fragments. (line 6) 42384 * makefile targets: Makefile. (line 6) 42385 * MALLOC_ABI_ALIGNMENT: Storage Layout. (line 179) 42386 * Manipulating GIMPLE statements: Manipulating GIMPLE statements. 42387 (line 6) 42388 * mark_hook: GTY Options. (line 170) 42389 * marking roots: GGC Roots. (line 6) 42390 * MASK_RETURN_ADDR: Exception Region Output. 42391 (line 35) 42392 * match_dup <1>: define_peephole2. (line 28) 42393 * match_dup: RTL Template. (line 73) 42394 * match_dup and attributes: Insn Lengths. (line 16) 42395 * match_op_dup: RTL Template. (line 163) 42396 * match_operand: RTL Template. (line 16) 42397 * match_operand and attributes: Expressions. (line 55) 42398 * match_operator: RTL Template. (line 95) 42399 * match_par_dup: RTL Template. (line 219) 42400 * match_parallel: RTL Template. (line 172) 42401 * match_scratch <1>: define_peephole2. (line 28) 42402 * match_scratch: RTL Template. (line 58) 42403 * matching constraint: Simple Constraints. (line 132) 42404 * matching operands: Output Template. (line 49) 42405 * math library: Soft float library routines. 42406 (line 6) 42407 * math, in RTL: Arithmetic. (line 6) 42408 * MATH_LIBRARY: Misc. (line 550) 42409 * matherr: Library Calls. (line 58) 42410 * MAX_BITS_PER_WORD: Storage Layout. (line 61) 42411 * MAX_CONDITIONAL_EXECUTE: Misc. (line 572) 42412 * MAX_FIXED_MODE_SIZE: Storage Layout. (line 420) 42413 * MAX_MOVE_MAX: Misc. (line 120) 42414 * MAX_OFILE_ALIGNMENT: Storage Layout. (line 216) 42415 * MAX_REGS_PER_ADDRESS: Addressing Modes. (line 42) 42416 * MAX_STACK_ALIGNMENT: Storage Layout. (line 209) 42417 * maxM3 instruction pattern: Standard Names. (line 234) 42418 * may_trap_p, tree_could_trap_p: Edges. (line 115) 42419 * maybe_undef: GTY Options. (line 178) 42420 * mcount: Profiling. (line 12) 42421 * MD_CAN_REDIRECT_BRANCH: Misc. (line 697) 42422 * MD_EXEC_PREFIX: Driver. (line 330) 42423 * MD_FALLBACK_FRAME_STATE_FOR: Exception Handling. (line 98) 42424 * MD_HANDLE_UNWABI: Exception Handling. (line 118) 42425 * MD_STARTFILE_PREFIX: Driver. (line 358) 42426 * MD_STARTFILE_PREFIX_1: Driver. (line 364) 42427 * MD_UNWIND_SUPPORT: Exception Handling. (line 94) 42428 * mem: Regs and Memory. (line 374) 42429 * mem and /c: Flags. (line 99) 42430 * mem and /f: Flags. (line 103) 42431 * mem and /i: Flags. (line 85) 42432 * mem and /j: Flags. (line 79) 42433 * mem and /s: Flags. (line 70) 42434 * mem and /u: Flags. (line 152) 42435 * mem and /v: Flags. (line 94) 42436 * mem, RTL sharing: Sharing. (line 40) 42437 * MEM_ALIAS_SET: Special Accessors. (line 9) 42438 * MEM_ALIGN: Special Accessors. (line 36) 42439 * MEM_EXPR: Special Accessors. (line 20) 42440 * MEM_IN_STRUCT_P: Flags. (line 70) 42441 * MEM_KEEP_ALIAS_SET_P: Flags. (line 79) 42442 * MEM_NOTRAP_P: Flags. (line 99) 42443 * MEM_OFFSET: Special Accessors. (line 28) 42444 * MEM_POINTER: Flags. (line 103) 42445 * MEM_READONLY_P: Flags. (line 152) 42446 * MEM_SCALAR_P: Flags. (line 85) 42447 * MEM_SIZE: Special Accessors. (line 31) 42448 * MEM_VOLATILE_P: Flags. (line 94) 42449 * MEMBER_TYPE_FORCES_BLK: Storage Layout. (line 400) 42450 * memory reference, nonoffsettable: Simple Constraints. (line 246) 42451 * memory references in constraints: Simple Constraints. (line 17) 42452 * memory_barrier instruction pattern: Standard Names. (line 1413) 42453 * MEMORY_MOVE_COST: Costs. (line 29) 42454 * memory_operand: Machine-Independent Predicates. 42455 (line 58) 42456 * METHOD_TYPE: Types. (line 6) 42457 * MIN_UNITS_PER_WORD: Storage Layout. (line 71) 42458 * MINIMUM_ALIGNMENT: Storage Layout. (line 288) 42459 * MINIMUM_ATOMIC_ALIGNMENT: Storage Layout. (line 187) 42460 * minM3 instruction pattern: Standard Names. (line 234) 42461 * minus: Arithmetic. (line 36) 42462 * minus and attributes: Expressions. (line 64) 42463 * minus, canonicalization of: Insn Canonicalizations. 42464 (line 27) 42465 * MINUS_EXPR: Expression trees. (line 6) 42466 * MIPS coprocessor-definition macros: MIPS Coprocessors. (line 6) 42467 * mod: Arithmetic. (line 131) 42468 * mod and attributes: Expressions. (line 64) 42469 * mode classes: Machine Modes. (line 219) 42470 * mode iterators in .md files: Mode Iterators. (line 6) 42471 * mode switching: Mode Switching. (line 6) 42472 * MODE_ACCUM: Machine Modes. (line 249) 42473 * MODE_AFTER: Mode Switching. (line 49) 42474 * MODE_BASE_REG_CLASS: Register Classes. (line 112) 42475 * MODE_BASE_REG_REG_CLASS: Register Classes. (line 118) 42476 * MODE_CC: Machine Modes. (line 268) 42477 * MODE_CODE_BASE_REG_CLASS: Register Classes. (line 125) 42478 * MODE_COMPLEX_FLOAT: Machine Modes. (line 260) 42479 * MODE_COMPLEX_INT: Machine Modes. (line 257) 42480 * MODE_DECIMAL_FLOAT: Machine Modes. (line 237) 42481 * MODE_ENTRY: Mode Switching. (line 54) 42482 * MODE_EXIT: Mode Switching. (line 60) 42483 * MODE_FLOAT: Machine Modes. (line 233) 42484 * MODE_FRACT: Machine Modes. (line 241) 42485 * MODE_FUNCTION: Machine Modes. (line 264) 42486 * MODE_INT: Machine Modes. (line 225) 42487 * MODE_NEEDED: Mode Switching. (line 42) 42488 * MODE_PARTIAL_INT: Machine Modes. (line 229) 42489 * MODE_PRIORITY_TO_MODE: Mode Switching. (line 66) 42490 * MODE_RANDOM: Machine Modes. (line 273) 42491 * MODE_UACCUM: Machine Modes. (line 253) 42492 * MODE_UFRACT: Machine Modes. (line 245) 42493 * MODES_TIEABLE_P: Values in Registers. 42494 (line 129) 42495 * modifiers in constraints: Modifiers. (line 6) 42496 * MODIFY_EXPR: Expression trees. (line 6) 42497 * MODIFY_JNI_METHOD_CALL: Misc. (line 774) 42498 * MODIFY_TARGET_NAME: Driver. (line 385) 42499 * modM3 instruction pattern: Standard Names. (line 222) 42500 * modulo scheduling: RTL passes. (line 131) 42501 * MOVE_BY_PIECES_P: Costs. (line 110) 42502 * MOVE_MAX: Misc. (line 115) 42503 * MOVE_MAX_PIECES: Costs. (line 116) 42504 * MOVE_RATIO: Costs. (line 97) 42505 * movM instruction pattern: Standard Names. (line 11) 42506 * movmemM instruction pattern: Standard Names. (line 672) 42507 * movmisalignM instruction pattern: Standard Names. (line 126) 42508 * movMODEcc instruction pattern: Standard Names. (line 891) 42509 * movstr instruction pattern: Standard Names. (line 707) 42510 * movstrictM instruction pattern: Standard Names. (line 120) 42511 * msubMN4 instruction pattern: Standard Names. (line 387) 42512 * mulhisi3 instruction pattern: Standard Names. (line 340) 42513 * mulM3 instruction pattern: Standard Names. (line 222) 42514 * mulqihi3 instruction pattern: Standard Names. (line 344) 42515 * mulsidi3 instruction pattern: Standard Names. (line 344) 42516 * mult: Arithmetic. (line 92) 42517 * mult and attributes: Expressions. (line 64) 42518 * mult, canonicalization of: Insn Canonicalizations. 42519 (line 27) 42520 * MULT_EXPR: Expression trees. (line 6) 42521 * MULTILIB_DEFAULTS: Driver. (line 315) 42522 * MULTILIB_DIRNAMES: Target Fragment. (line 64) 42523 * MULTILIB_EXCEPTIONS: Target Fragment. (line 84) 42524 * MULTILIB_EXTRA_OPTS: Target Fragment. (line 96) 42525 * MULTILIB_MATCHES: Target Fragment. (line 77) 42526 * MULTILIB_OPTIONS: Target Fragment. (line 44) 42527 * multiple alternative constraints: Multi-Alternative. (line 6) 42528 * MULTIPLE_SYMBOL_SPACES: Misc. (line 530) 42529 * multiplication: Arithmetic. (line 92) 42530 * multiplication with signed saturation: Arithmetic. (line 92) 42531 * multiplication with unsigned saturation: Arithmetic. (line 92) 42532 * MUST_USE_SJLJ_EXCEPTIONS: Exception Region Output. 42533 (line 64) 42534 * n in constraint: Simple Constraints. (line 65) 42535 * N_REG_CLASSES: Register Classes. (line 76) 42536 * name: Identifiers. (line 6) 42537 * named patterns and conditions: Patterns. (line 47) 42538 * names, pattern: Standard Names. (line 6) 42539 * namespace: Namespaces. (line 6) 42540 * namespace, class, scope: Scopes. (line 6) 42541 * NAMESPACE_DECL <1>: Declarations. (line 6) 42542 * NAMESPACE_DECL: Namespaces. (line 6) 42543 * NATIVE_SYSTEM_HEADER_DIR: Target Fragment. (line 103) 42544 * ne: Comparisons. (line 56) 42545 * ne and attributes: Expressions. (line 64) 42546 * NE_EXPR: Expression trees. (line 6) 42547 * nearbyintM2 instruction pattern: Standard Names. (line 564) 42548 * neg: Arithmetic. (line 81) 42549 * neg and attributes: Expressions. (line 64) 42550 * neg, canonicalization of: Insn Canonicalizations. 42551 (line 27) 42552 * NEGATE_EXPR: Expression trees. (line 6) 42553 * negation: Arithmetic. (line 81) 42554 * negation with signed saturation: Arithmetic. (line 81) 42555 * negation with unsigned saturation: Arithmetic. (line 81) 42556 * negM2 instruction pattern: Standard Names. (line 449) 42557 * nested functions, trampolines for: Trampolines. (line 6) 42558 * nested_ptr: GTY Options. (line 185) 42559 * next_bb, prev_bb, FOR_EACH_BB: Basic Blocks. (line 10) 42560 * next_cc0_user: Jump Patterns. (line 64) 42561 * NEXT_INSN: Insns. (line 30) 42562 * NEXT_OBJC_RUNTIME: Library Calls. (line 94) 42563 * nil: RTL Objects. (line 73) 42564 * NO_DBX_BNSYM_ENSYM: DBX Hooks. (line 39) 42565 * NO_DBX_FUNCTION_END: DBX Hooks. (line 33) 42566 * NO_DBX_GCC_MARKER: File Names and DBX. (line 28) 42567 * NO_DBX_MAIN_SOURCE_DIRECTORY: File Names and DBX. (line 23) 42568 * NO_DOLLAR_IN_LABEL: Misc. (line 494) 42569 * NO_DOT_IN_LABEL: Misc. (line 500) 42570 * NO_FUNCTION_CSE: Costs. (line 200) 42571 * NO_IMPLICIT_EXTERN_C: Misc. (line 376) 42572 * NO_PROFILE_COUNTERS: Profiling. (line 28) 42573 * NO_REGS: Register Classes. (line 17) 42574 * NON_LVALUE_EXPR: Expression trees. (line 6) 42575 * nondeterministic finite state automaton: Processor pipeline description. 42576 (line 301) 42577 * nonimmediate_operand: Machine-Independent Predicates. 42578 (line 101) 42579 * nonlocal goto handler: Edges. (line 171) 42580 * nonlocal_goto instruction pattern: Standard Names. (line 1255) 42581 * nonlocal_goto_receiver instruction pattern: Standard Names. 42582 (line 1272) 42583 * nonmemory_operand: Machine-Independent Predicates. 42584 (line 97) 42585 * nonoffsettable memory reference: Simple Constraints. (line 246) 42586 * nop instruction pattern: Standard Names. (line 1073) 42587 * NOP_EXPR: Expression trees. (line 6) 42588 * normal predicates: Predicates. (line 31) 42589 * not: Arithmetic. (line 149) 42590 * not and attributes: Expressions. (line 50) 42591 * not equal: Comparisons. (line 56) 42592 * not, canonicalization of: Insn Canonicalizations. 42593 (line 27) 42594 * note: Insns. (line 168) 42595 * note and /i: Flags. (line 59) 42596 * note and /v: Flags. (line 44) 42597 * NOTE_INSN_BASIC_BLOCK, CODE_LABEL, notes: Basic Blocks. (line 41) 42598 * NOTE_INSN_BLOCK_BEG: Insns. (line 193) 42599 * NOTE_INSN_BLOCK_END: Insns. (line 193) 42600 * NOTE_INSN_DELETED: Insns. (line 183) 42601 * NOTE_INSN_DELETED_LABEL: Insns. (line 188) 42602 * NOTE_INSN_EH_REGION_BEG: Insns. (line 199) 42603 * NOTE_INSN_EH_REGION_END: Insns. (line 199) 42604 * NOTE_INSN_FUNCTION_BEG: Insns. (line 223) 42605 * NOTE_INSN_LOOP_BEG: Insns. (line 207) 42606 * NOTE_INSN_LOOP_CONT: Insns. (line 213) 42607 * NOTE_INSN_LOOP_END: Insns. (line 207) 42608 * NOTE_INSN_LOOP_VTOP: Insns. (line 217) 42609 * NOTE_LINE_NUMBER: Insns. (line 168) 42610 * NOTE_SOURCE_FILE: Insns. (line 168) 42611 * NOTICE_UPDATE_CC: Condition Code. (line 33) 42612 * NUM_MACHINE_MODES: Machine Modes. (line 286) 42613 * NUM_MODES_FOR_MODE_SWITCHING: Mode Switching. (line 30) 42614 * Number of iterations analysis: Number of iterations. 42615 (line 6) 42616 * o in constraint: Simple Constraints. (line 23) 42617 * OBJC_GEN_METHOD_LABEL: Label Output. (line 411) 42618 * OBJC_JBLEN: Misc. (line 930) 42619 * OBJECT_FORMAT_COFF: Macros for Initialization. 42620 (line 97) 42621 * OFFSET_TYPE: Types. (line 6) 42622 * offsettable address: Simple Constraints. (line 23) 42623 * OImode: Machine Modes. (line 51) 42624 * Omega a solver for linear programming problems: Omega. (line 6) 42625 * OMP_ATOMIC: Expression trees. (line 6) 42626 * OMP_CLAUSE: Expression trees. (line 6) 42627 * OMP_CONTINUE: Expression trees. (line 6) 42628 * OMP_CRITICAL: Expression trees. (line 6) 42629 * OMP_FOR: Expression trees. (line 6) 42630 * OMP_MASTER: Expression trees. (line 6) 42631 * OMP_ORDERED: Expression trees. (line 6) 42632 * OMP_PARALLEL: Expression trees. (line 6) 42633 * OMP_RETURN: Expression trees. (line 6) 42634 * OMP_SECTION: Expression trees. (line 6) 42635 * OMP_SECTIONS: Expression trees. (line 6) 42636 * OMP_SINGLE: Expression trees. (line 6) 42637 * one_cmplM2 instruction pattern: Standard Names. (line 651) 42638 * operand access: Accessors. (line 6) 42639 * Operand Access Routines: SSA Operands. (line 119) 42640 * operand constraints: Constraints. (line 6) 42641 * Operand Iterators: SSA Operands. (line 119) 42642 * operand predicates: Predicates. (line 6) 42643 * operand substitution: Output Template. (line 6) 42644 * operands <1>: Patterns. (line 53) 42645 * operands: SSA Operands. (line 6) 42646 * Operands: Operands. (line 6) 42647 * operator predicates: Predicates. (line 6) 42648 * optc-gen.awk: Options. (line 6) 42649 * Optimization infrastructure for GIMPLE: Tree SSA. (line 6) 42650 * OPTIMIZATION_OPTIONS: Run-time Target. (line 120) 42651 * OPTIMIZE_MODE_SWITCHING: Mode Switching. (line 9) 42652 * option specification files: Options. (line 6) 42653 * OPTION_DEFAULT_SPECS: Driver. (line 88) 42654 * optional hardware or system features: Run-time Target. (line 59) 42655 * options, directory search: Including Patterns. (line 44) 42656 * order of register allocation: Allocation Order. (line 6) 42657 * ORDERED_EXPR: Expression trees. (line 6) 42658 * Ordering of Patterns: Pattern Ordering. (line 6) 42659 * ORIGINAL_REGNO: Special Accessors. (line 40) 42660 * other register constraints: Simple Constraints. (line 163) 42661 * OUTGOING_REG_PARM_STACK_SPACE: Stack Arguments. (line 71) 42662 * OUTGOING_REGNO: Register Basics. (line 98) 42663 * output of assembler code: File Framework. (line 6) 42664 * output statements: Output Statement. (line 6) 42665 * output templates: Output Template. (line 6) 42666 * OUTPUT_ADDR_CONST_EXTRA: Data Output. (line 39) 42667 * output_asm_insn: Output Statement. (line 53) 42668 * OUTPUT_QUOTED_STRING: File Framework. (line 76) 42669 * OVERLOAD: Functions. (line 6) 42670 * OVERRIDE_ABI_FORMAT: Register Arguments. (line 140) 42671 * OVERRIDE_OPTIONS: Run-time Target. (line 104) 42672 * OVL_CURRENT: Functions. (line 6) 42673 * OVL_NEXT: Functions. (line 6) 42674 * p in constraint: Simple Constraints. (line 154) 42675 * PAD_VARARGS_DOWN: Register Arguments. (line 220) 42676 * parallel: Side Effects. (line 204) 42677 * param_is: GTY Options. (line 113) 42678 * parameters, c++ abi: C++ ABI. (line 6) 42679 * parameters, miscellaneous: Misc. (line 6) 42680 * parameters, precompiled headers: PCH Target. (line 6) 42681 * paramN_is: GTY Options. (line 131) 42682 * parity: Arithmetic. (line 228) 42683 * parityM2 instruction pattern: Standard Names. (line 645) 42684 * PARM_BOUNDARY: Storage Layout. (line 144) 42685 * PARM_DECL: Declarations. (line 6) 42686 * PARSE_LDD_OUTPUT: Macros for Initialization. 42687 (line 121) 42688 * passes and files of the compiler: Passes. (line 6) 42689 * passing arguments: Interface. (line 36) 42690 * PATH_SEPARATOR: Filesystem. (line 31) 42691 * PATTERN: Insns. (line 247) 42692 * pattern conditions: Patterns. (line 43) 42693 * pattern names: Standard Names. (line 6) 42694 * Pattern Ordering: Pattern Ordering. (line 6) 42695 * patterns: Patterns. (line 6) 42696 * pc: Regs and Memory. (line 361) 42697 * pc and attributes: Insn Lengths. (line 20) 42698 * pc, RTL sharing: Sharing. (line 25) 42699 * PC_REGNUM: Register Basics. (line 112) 42700 * pc_rtx: Regs and Memory. (line 366) 42701 * PCC_BITFIELD_TYPE_MATTERS: Storage Layout. (line 314) 42702 * PCC_STATIC_STRUCT_RETURN: Aggregate Return. (line 64) 42703 * PDImode: Machine Modes. (line 40) 42704 * peephole optimization, RTL representation: Side Effects. (line 238) 42705 * peephole optimizer definitions: Peephole Definitions. 42706 (line 6) 42707 * per-function data: Per-Function Data. (line 6) 42708 * percent sign: Output Template. (line 6) 42709 * PHI nodes: SSA. (line 31) 42710 * phi_arg_d: GIMPLE_PHI. (line 28) 42711 * PHI_ARG_DEF: SSA. (line 71) 42712 * PHI_ARG_EDGE: SSA. (line 68) 42713 * PHI_ARG_ELT: SSA. (line 63) 42714 * PHI_NUM_ARGS: SSA. (line 59) 42715 * PHI_RESULT: SSA. (line 56) 42716 * PIC: PIC. (line 6) 42717 * PIC_OFFSET_TABLE_REG_CALL_CLOBBERED: PIC. (line 26) 42718 * PIC_OFFSET_TABLE_REGNUM: PIC. (line 16) 42719 * pipeline hazard recognizer: Processor pipeline description. 42720 (line 6) 42721 * Plugins: Plugins. (line 6) 42722 * plus: Arithmetic. (line 14) 42723 * plus and attributes: Expressions. (line 64) 42724 * plus, canonicalization of: Insn Canonicalizations. 42725 (line 27) 42726 * PLUS_EXPR: Expression trees. (line 6) 42727 * Pmode: Misc. (line 344) 42728 * pmode_register_operand: Machine-Independent Predicates. 42729 (line 35) 42730 * pointer: Types. (line 6) 42731 * POINTER_PLUS_EXPR: Expression trees. (line 6) 42732 * POINTER_SIZE: Storage Layout. (line 83) 42733 * POINTER_TYPE: Types. (line 6) 42734 * POINTERS_EXTEND_UNSIGNED: Storage Layout. (line 89) 42735 * pop_operand: Machine-Independent Predicates. 42736 (line 88) 42737 * popcount: Arithmetic. (line 224) 42738 * popcountM2 instruction pattern: Standard Names. (line 639) 42739 * portability: Portability. (line 6) 42740 * position independent code: PIC. (line 6) 42741 * post_dec: Incdec. (line 25) 42742 * post_inc: Incdec. (line 30) 42743 * post_modify: Incdec. (line 33) 42744 * POSTDECREMENT_EXPR: Expression trees. (line 6) 42745 * POSTINCREMENT_EXPR: Expression trees. (line 6) 42746 * POWI_MAX_MULTS: Misc. (line 828) 42747 * powM3 instruction pattern: Standard Names. (line 513) 42748 * pragma: Misc. (line 381) 42749 * pre_dec: Incdec. (line 8) 42750 * PRE_GCC3_DWARF_FRAME_REGISTERS: Frame Registers. (line 110) 42751 * pre_inc: Incdec. (line 22) 42752 * pre_modify: Incdec. (line 51) 42753 * PREDECREMENT_EXPR: Expression trees. (line 6) 42754 * predefined macros: Run-time Target. (line 6) 42755 * predicates: Predicates. (line 6) 42756 * predicates and machine modes: Predicates. (line 31) 42757 * predication: Conditional Execution. 42758 (line 6) 42759 * predict.def: Profile information. 42760 (line 24) 42761 * PREFERRED_DEBUGGING_TYPE: All Debuggers. (line 42) 42762 * PREFERRED_OUTPUT_RELOAD_CLASS: Register Classes. (line 231) 42763 * PREFERRED_RELOAD_CLASS: Register Classes. (line 196) 42764 * PREFERRED_STACK_BOUNDARY: Storage Layout. (line 158) 42765 * prefetch: Side Effects. (line 312) 42766 * prefetch instruction pattern: Standard Names. (line 1392) 42767 * PREINCREMENT_EXPR: Expression trees. (line 6) 42768 * presence_set: Processor pipeline description. 42769 (line 220) 42770 * preserving SSA form: SSA. (line 76) 42771 * preserving virtual SSA form: SSA. (line 186) 42772 * prev_active_insn: define_peephole. (line 60) 42773 * prev_cc0_setter: Jump Patterns. (line 64) 42774 * PREV_INSN: Insns. (line 26) 42775 * PRINT_OPERAND: Instruction Output. (line 68) 42776 * PRINT_OPERAND_ADDRESS: Instruction Output. (line 96) 42777 * PRINT_OPERAND_PUNCT_VALID_P: Instruction Output. (line 89) 42778 * processor functional units: Processor pipeline description. 42779 (line 6) 42780 * processor pipeline description: Processor pipeline description. 42781 (line 6) 42782 * product: Arithmetic. (line 92) 42783 * profile feedback: Profile information. 42784 (line 14) 42785 * profile representation: Profile information. 42786 (line 6) 42787 * PROFILE_BEFORE_PROLOGUE: Profiling. (line 35) 42788 * PROFILE_HOOK: Profiling. (line 23) 42789 * profiling, code generation: Profiling. (line 6) 42790 * program counter: Regs and Memory. (line 362) 42791 * prologue: Function Entry. (line 6) 42792 * prologue instruction pattern: Standard Names. (line 1338) 42793 * PROMOTE_FUNCTION_MODE: Storage Layout. (line 123) 42794 * PROMOTE_MODE: Storage Layout. (line 100) 42795 * pseudo registers: Regs and Memory. (line 9) 42796 * PSImode: Machine Modes. (line 32) 42797 * PTRDIFF_TYPE: Type Layout. (line 184) 42798 * PTRMEM_CST: Expression trees. (line 6) 42799 * PTRMEM_CST_CLASS: Expression trees. (line 6) 42800 * PTRMEM_CST_MEMBER: Expression trees. (line 6) 42801 * purge_dead_edges <1>: Maintaining the CFG. 42802 (line 93) 42803 * purge_dead_edges: Edges. (line 104) 42804 * push address instruction: Simple Constraints. (line 154) 42805 * PUSH_ARGS: Stack Arguments. (line 18) 42806 * PUSH_ARGS_REVERSED: Stack Arguments. (line 26) 42807 * push_operand: Machine-Independent Predicates. 42808 (line 81) 42809 * push_reload: Addressing Modes. (line 169) 42810 * PUSH_ROUNDING: Stack Arguments. (line 32) 42811 * pushM1 instruction pattern: Standard Names. (line 209) 42812 * PUT_CODE: RTL Objects. (line 47) 42813 * PUT_MODE: Machine Modes. (line 283) 42814 * PUT_REG_NOTE_KIND: Insns. (line 309) 42815 * PUT_SDB_: SDB and DWARF. (line 63) 42816 * QCmode: Machine Modes. (line 197) 42817 * QFmode: Machine Modes. (line 54) 42818 * QImode: Machine Modes. (line 25) 42819 * QImode, in insn: Insns. (line 231) 42820 * QQmode: Machine Modes. (line 103) 42821 * qualified type: Types. (line 6) 42822 * querying function unit reservations: Processor pipeline description. 42823 (line 90) 42824 * question mark: Multi-Alternative. (line 41) 42825 * quotient: Arithmetic. (line 111) 42826 * r in constraint: Simple Constraints. (line 56) 42827 * RANGE_TEST_NON_SHORT_CIRCUIT: Costs. (line 204) 42828 * RDIV_EXPR: Expression trees. (line 6) 42829 * READONLY_DATA_SECTION_ASM_OP: Sections. (line 63) 42830 * real operands: SSA Operands. (line 6) 42831 * REAL_ARITHMETIC: Floating Point. (line 66) 42832 * REAL_CST: Expression trees. (line 6) 42833 * REAL_LIBGCC_SPEC: Driver. (line 187) 42834 * REAL_NM_FILE_NAME: Macros for Initialization. 42835 (line 106) 42836 * REAL_TYPE: Types. (line 6) 42837 * REAL_VALUE_ABS: Floating Point. (line 82) 42838 * REAL_VALUE_ATOF: Floating Point. (line 50) 42839 * REAL_VALUE_FIX: Floating Point. (line 41) 42840 * REAL_VALUE_FROM_INT: Floating Point. (line 99) 42841 * REAL_VALUE_ISINF: Floating Point. (line 59) 42842 * REAL_VALUE_ISNAN: Floating Point. (line 62) 42843 * REAL_VALUE_NEGATE: Floating Point. (line 79) 42844 * REAL_VALUE_NEGATIVE: Floating Point. (line 56) 42845 * REAL_VALUE_TO_INT: Floating Point. (line 93) 42846 * REAL_VALUE_TO_TARGET_DECIMAL128: Data Output. (line 144) 42847 * REAL_VALUE_TO_TARGET_DECIMAL32: Data Output. (line 142) 42848 * REAL_VALUE_TO_TARGET_DECIMAL64: Data Output. (line 143) 42849 * REAL_VALUE_TO_TARGET_DOUBLE: Data Output. (line 140) 42850 * REAL_VALUE_TO_TARGET_LONG_DOUBLE: Data Output. (line 141) 42851 * REAL_VALUE_TO_TARGET_SINGLE: Data Output. (line 139) 42852 * REAL_VALUE_TRUNCATE: Floating Point. (line 86) 42853 * REAL_VALUE_TYPE: Floating Point. (line 26) 42854 * REAL_VALUE_UNSIGNED_FIX: Floating Point. (line 45) 42855 * REAL_VALUES_EQUAL: Floating Point. (line 32) 42856 * REAL_VALUES_LESS: Floating Point. (line 38) 42857 * REALPART_EXPR: Expression trees. (line 6) 42858 * recog_data.operand: Instruction Output. (line 39) 42859 * recognizing insns: RTL Template. (line 6) 42860 * RECORD_TYPE <1>: Classes. (line 6) 42861 * RECORD_TYPE: Types. (line 6) 42862 * redirect_edge_and_branch: Profile information. 42863 (line 71) 42864 * redirect_edge_and_branch, redirect_jump: Maintaining the CFG. 42865 (line 103) 42866 * reduc_smax_M instruction pattern: Standard Names. (line 240) 42867 * reduc_smin_M instruction pattern: Standard Names. (line 240) 42868 * reduc_splus_M instruction pattern: Standard Names. (line 252) 42869 * reduc_umax_M instruction pattern: Standard Names. (line 246) 42870 * reduc_umin_M instruction pattern: Standard Names. (line 246) 42871 * reduc_uplus_M instruction pattern: Standard Names. (line 258) 42872 * reference: Types. (line 6) 42873 * REFERENCE_TYPE: Types. (line 6) 42874 * reg: Regs and Memory. (line 9) 42875 * reg and /f: Flags. (line 112) 42876 * reg and /i: Flags. (line 107) 42877 * reg and /v: Flags. (line 116) 42878 * reg, RTL sharing: Sharing. (line 17) 42879 * REG_ALLOC_ORDER: Allocation Order. (line 9) 42880 * REG_BR_PRED: Insns. (line 491) 42881 * REG_BR_PROB: Insns. (line 485) 42882 * REG_BR_PROB_BASE, BB_FREQ_BASE, count: Profile information. 42883 (line 82) 42884 * REG_BR_PROB_BASE, EDGE_FREQUENCY: Profile information. 42885 (line 52) 42886 * REG_CC_SETTER: Insns. (line 456) 42887 * REG_CC_USER: Insns. (line 456) 42888 * REG_CLASS_CONTENTS: Register Classes. (line 86) 42889 * reg_class_contents: Register Basics. (line 59) 42890 * REG_CLASS_FROM_CONSTRAINT: Old Constraints. (line 35) 42891 * REG_CLASS_FROM_LETTER: Old Constraints. (line 27) 42892 * REG_CLASS_NAMES: Register Classes. (line 81) 42893 * REG_CROSSING_JUMP: Insns. (line 368) 42894 * REG_DEAD: Insns. (line 320) 42895 * REG_DEAD, REG_UNUSED: Liveness information. 42896 (line 32) 42897 * REG_DEP_ANTI: Insns. (line 478) 42898 * REG_DEP_OUTPUT: Insns. (line 474) 42899 * REG_DEP_TRUE: Insns. (line 471) 42900 * REG_EH_REGION, EDGE_ABNORMAL_CALL: Edges. (line 110) 42901 * REG_EQUAL: Insns. (line 384) 42902 * REG_EQUIV: Insns. (line 384) 42903 * REG_EXPR: Special Accessors. (line 46) 42904 * REG_FRAME_RELATED_EXPR: Insns. (line 497) 42905 * REG_FUNCTION_VALUE_P: Flags. (line 107) 42906 * REG_INC: Insns. (line 336) 42907 * reg_label and /v: Flags. (line 65) 42908 * REG_LABEL_OPERAND: Insns. (line 350) 42909 * REG_LABEL_TARGET: Insns. (line 359) 42910 * reg_names <1>: Instruction Output. (line 80) 42911 * reg_names: Register Basics. (line 59) 42912 * REG_NONNEG: Insns. (line 342) 42913 * REG_NOTE_KIND: Insns. (line 309) 42914 * REG_NOTES: Insns. (line 283) 42915 * REG_OFFSET: Special Accessors. (line 50) 42916 * REG_OK_STRICT: Addressing Modes. (line 67) 42917 * REG_PARM_STACK_SPACE: Stack Arguments. (line 56) 42918 * REG_PARM_STACK_SPACE, and FUNCTION_ARG: Register Arguments. 42919 (line 52) 42920 * REG_POINTER: Flags. (line 112) 42921 * REG_SETJMP: Insns. (line 378) 42922 * REG_UNUSED: Insns. (line 329) 42923 * REG_USERVAR_P: Flags. (line 116) 42924 * regclass_for_constraint: C Constraint Interface. 42925 (line 60) 42926 * register allocation order: Allocation Order. (line 6) 42927 * register class definitions: Register Classes. (line 6) 42928 * register class preference constraints: Class Preferences. (line 6) 42929 * register pairs: Values in Registers. 42930 (line 69) 42931 * Register Transfer Language (RTL): RTL. (line 6) 42932 * register usage: Registers. (line 6) 42933 * REGISTER_MOVE_COST: Costs. (line 10) 42934 * REGISTER_NAMES: Instruction Output. (line 9) 42935 * register_operand: Machine-Independent Predicates. 42936 (line 30) 42937 * REGISTER_PREFIX: Instruction Output. (line 124) 42938 * REGISTER_TARGET_PRAGMAS: Misc. (line 382) 42939 * registers arguments: Register Arguments. (line 6) 42940 * registers in constraints: Simple Constraints. (line 56) 42941 * REGMODE_NATURAL_SIZE: Values in Registers. 42942 (line 50) 42943 * REGNO_MODE_CODE_OK_FOR_BASE_P: Register Classes. (line 170) 42944 * REGNO_MODE_OK_FOR_BASE_P: Register Classes. (line 146) 42945 * REGNO_MODE_OK_FOR_REG_BASE_P: Register Classes. (line 157) 42946 * REGNO_OK_FOR_BASE_P: Register Classes. (line 140) 42947 * REGNO_OK_FOR_INDEX_P: Register Classes. (line 181) 42948 * REGNO_REG_CLASS: Register Classes. (line 101) 42949 * regs_ever_live: Function Entry. (line 21) 42950 * regular expressions: Processor pipeline description. 42951 (line 6) 42952 * relative costs: Costs. (line 6) 42953 * RELATIVE_PREFIX_NOT_LINKDIR: Driver. (line 325) 42954 * reload_completed: Standard Names. (line 1040) 42955 * reload_in instruction pattern: Standard Names. (line 99) 42956 * reload_in_progress: Standard Names. (line 57) 42957 * reload_out instruction pattern: Standard Names. (line 99) 42958 * reloading: RTL passes. (line 182) 42959 * remainder: Arithmetic. (line 131) 42960 * remainderM3 instruction pattern: Standard Names. (line 472) 42961 * reorder: GTY Options. (line 209) 42962 * representation of RTL: RTL. (line 6) 42963 * reservation delays: Processor pipeline description. 42964 (line 6) 42965 * rest_of_decl_compilation: Parsing pass. (line 52) 42966 * rest_of_type_compilation: Parsing pass. (line 52) 42967 * restore_stack_block instruction pattern: Standard Names. (line 1174) 42968 * restore_stack_function instruction pattern: Standard Names. 42969 (line 1174) 42970 * restore_stack_nonlocal instruction pattern: Standard Names. 42971 (line 1174) 42972 * RESULT_DECL: Declarations. (line 6) 42973 * return: Side Effects. (line 72) 42974 * return instruction pattern: Standard Names. (line 1027) 42975 * return values in registers: Scalar Return. (line 6) 42976 * RETURN_ADDR_IN_PREVIOUS_FRAME: Frame Layout. (line 135) 42977 * RETURN_ADDR_OFFSET: Exception Handling. (line 60) 42978 * RETURN_ADDR_RTX: Frame Layout. (line 124) 42979 * RETURN_ADDRESS_POINTER_REGNUM: Frame Registers. (line 51) 42980 * RETURN_EXPR: Function Bodies. (line 6) 42981 * RETURN_POPS_ARGS: Stack Arguments. (line 90) 42982 * RETURN_STMT: Function Bodies. (line 6) 42983 * return_val: Flags. (line 294) 42984 * return_val, in call_insn: Flags. (line 24) 42985 * return_val, in mem: Flags. (line 85) 42986 * return_val, in reg: Flags. (line 107) 42987 * return_val, in symbol_ref: Flags. (line 220) 42988 * returning aggregate values: Aggregate Return. (line 6) 42989 * returning structures and unions: Interface. (line 10) 42990 * reverse probability: Profile information. 42991 (line 66) 42992 * REVERSE_CONDEXEC_PREDICATES_P: Condition Code. (line 129) 42993 * REVERSE_CONDITION: Condition Code. (line 116) 42994 * REVERSIBLE_CC_MODE: Condition Code. (line 102) 42995 * right rotate: Arithmetic. (line 190) 42996 * right shift: Arithmetic. (line 185) 42997 * rintM2 instruction pattern: Standard Names. (line 572) 42998 * RISC: Processor pipeline description. 42999 (line 6) 43000 * roots, marking: GGC Roots. (line 6) 43001 * rotate: Arithmetic. (line 190) 43002 * rotatert: Arithmetic. (line 190) 43003 * rotlM3 instruction pattern: Standard Names. (line 441) 43004 * rotrM3 instruction pattern: Standard Names. (line 441) 43005 * ROUND_DIV_EXPR: Expression trees. (line 6) 43006 * ROUND_MOD_EXPR: Expression trees. (line 6) 43007 * ROUND_TOWARDS_ZERO: Storage Layout. (line 460) 43008 * ROUND_TYPE_ALIGN: Storage Layout. (line 411) 43009 * roundM2 instruction pattern: Standard Names. (line 548) 43010 * RSHIFT_EXPR: Expression trees. (line 6) 43011 * RTL addition: Arithmetic. (line 14) 43012 * RTL addition with signed saturation: Arithmetic. (line 14) 43013 * RTL addition with unsigned saturation: Arithmetic. (line 14) 43014 * RTL classes: RTL Classes. (line 6) 43015 * RTL comparison: Arithmetic. (line 43) 43016 * RTL comparison operations: Comparisons. (line 6) 43017 * RTL constant expression types: Constants. (line 6) 43018 * RTL constants: Constants. (line 6) 43019 * RTL declarations: RTL Declarations. (line 6) 43020 * RTL difference: Arithmetic. (line 36) 43021 * RTL expression: RTL Objects. (line 6) 43022 * RTL expressions for arithmetic: Arithmetic. (line 6) 43023 * RTL format: RTL Classes. (line 71) 43024 * RTL format characters: RTL Classes. (line 76) 43025 * RTL function-call insns: Calls. (line 6) 43026 * RTL insn template: RTL Template. (line 6) 43027 * RTL integers: RTL Objects. (line 6) 43028 * RTL memory expressions: Regs and Memory. (line 6) 43029 * RTL object types: RTL Objects. (line 6) 43030 * RTL postdecrement: Incdec. (line 6) 43031 * RTL postincrement: Incdec. (line 6) 43032 * RTL predecrement: Incdec. (line 6) 43033 * RTL preincrement: Incdec. (line 6) 43034 * RTL register expressions: Regs and Memory. (line 6) 43035 * RTL representation: RTL. (line 6) 43036 * RTL side effect expressions: Side Effects. (line 6) 43037 * RTL strings: RTL Objects. (line 6) 43038 * RTL structure sharing assumptions: Sharing. (line 6) 43039 * RTL subtraction: Arithmetic. (line 36) 43040 * RTL subtraction with signed saturation: Arithmetic. (line 36) 43041 * RTL subtraction with unsigned saturation: Arithmetic. (line 36) 43042 * RTL sum: Arithmetic. (line 14) 43043 * RTL vectors: RTL Objects. (line 6) 43044 * RTL_CONST_CALL_P: Flags. (line 19) 43045 * RTL_CONST_OR_PURE_CALL_P: Flags. (line 29) 43046 * RTL_LOOPING_CONST_OR_PURE_CALL_P: Flags. (line 33) 43047 * RTL_PURE_CALL_P: Flags. (line 24) 43048 * RTX (See RTL): RTL Objects. (line 6) 43049 * RTX codes, classes of: RTL Classes. (line 6) 43050 * RTX_FRAME_RELATED_P: Flags. (line 125) 43051 * run-time conventions: Interface. (line 6) 43052 * run-time target specification: Run-time Target. (line 6) 43053 * s in constraint: Simple Constraints. (line 92) 43054 * same_type_p: Types. (line 148) 43055 * SAmode: Machine Modes. (line 148) 43056 * sat_fract: Conversions. (line 90) 43057 * satfractMN2 instruction pattern: Standard Names. (line 843) 43058 * satfractunsMN2 instruction pattern: Standard Names. (line 856) 43059 * satisfies_constraint_: C Constraint Interface. 43060 (line 47) 43061 * SAVE_EXPR: Expression trees. (line 6) 43062 * save_stack_block instruction pattern: Standard Names. (line 1174) 43063 * save_stack_function instruction pattern: Standard Names. (line 1174) 43064 * save_stack_nonlocal instruction pattern: Standard Names. (line 1174) 43065 * SBSS_SECTION_ASM_OP: Sections. (line 77) 43066 * Scalar evolutions: Scalar evolutions. (line 6) 43067 * scalars, returned as values: Scalar Return. (line 6) 43068 * SCHED_GROUP_P: Flags. (line 166) 43069 * SCmode: Machine Modes. (line 197) 43070 * sCOND instruction pattern: Standard Names. (line 911) 43071 * scratch: Regs and Memory. (line 298) 43072 * scratch operands: Regs and Memory. (line 298) 43073 * scratch, RTL sharing: Sharing. (line 35) 43074 * scratch_operand: Machine-Independent Predicates. 43075 (line 50) 43076 * SDATA_SECTION_ASM_OP: Sections. (line 58) 43077 * SDB_ALLOW_FORWARD_REFERENCES: SDB and DWARF. (line 81) 43078 * SDB_ALLOW_UNKNOWN_REFERENCES: SDB and DWARF. (line 76) 43079 * SDB_DEBUGGING_INFO: SDB and DWARF. (line 9) 43080 * SDB_DELIM: SDB and DWARF. (line 69) 43081 * SDB_OUTPUT_SOURCE_LINE: SDB and DWARF. (line 86) 43082 * SDmode: Machine Modes. (line 85) 43083 * sdot_prodM instruction pattern: Standard Names. (line 264) 43084 * search options: Including Patterns. (line 44) 43085 * SECONDARY_INPUT_RELOAD_CLASS: Register Classes. (line 335) 43086 * SECONDARY_MEMORY_NEEDED: Register Classes. (line 391) 43087 * SECONDARY_MEMORY_NEEDED_MODE: Register Classes. (line 410) 43088 * SECONDARY_MEMORY_NEEDED_RTX: Register Classes. (line 401) 43089 * SECONDARY_OUTPUT_RELOAD_CLASS: Register Classes. (line 336) 43090 * SECONDARY_RELOAD_CLASS: Register Classes. (line 334) 43091 * SELECT_CC_MODE: Condition Code. (line 68) 43092 * sequence: Side Effects. (line 254) 43093 * Sequence iterators: Sequence iterators. (line 6) 43094 * set: Side Effects. (line 15) 43095 * set and /f: Flags. (line 125) 43096 * SET_ASM_OP: Label Output. (line 378) 43097 * set_attr: Tagging Insns. (line 31) 43098 * set_attr_alternative: Tagging Insns. (line 49) 43099 * set_bb_seq: GIMPLE sequences. (line 76) 43100 * SET_BY_PIECES_P: Costs. (line 145) 43101 * SET_DEST: Side Effects. (line 69) 43102 * SET_IS_RETURN_P: Flags. (line 175) 43103 * SET_LABEL_KIND: Insns. (line 140) 43104 * set_optab_libfunc: Library Calls. (line 15) 43105 * SET_RATIO: Costs. (line 136) 43106 * SET_SRC: Side Effects. (line 69) 43107 * SET_TYPE_STRUCTURAL_EQUALITY: Types. (line 6) 43108 * setmemM instruction pattern: Standard Names. (line 715) 43109 * SETUP_FRAME_ADDRESSES: Frame Layout. (line 102) 43110 * SF_SIZE: Type Layout. (line 129) 43111 * SFmode: Machine Modes. (line 66) 43112 * sharing of RTL components: Sharing. (line 6) 43113 * shift: Arithmetic. (line 168) 43114 * SHIFT_COUNT_TRUNCATED: Misc. (line 127) 43115 * SHLIB_SUFFIX: Macros for Initialization. 43116 (line 129) 43117 * SHORT_ACCUM_TYPE_SIZE: Type Layout. (line 83) 43118 * SHORT_FRACT_TYPE_SIZE: Type Layout. (line 63) 43119 * SHORT_IMMEDIATES_SIGN_EXTEND: Misc. (line 96) 43120 * SHORT_TYPE_SIZE: Type Layout. (line 16) 43121 * sibcall_epilogue instruction pattern: Standard Names. (line 1364) 43122 * sibling call: Edges. (line 122) 43123 * SIBLING_CALL_P: Flags. (line 179) 43124 * sign_extend: Conversions. (line 23) 43125 * sign_extract: Bit-Fields. (line 8) 43126 * sign_extract, canonicalization of: Insn Canonicalizations. 43127 (line 96) 43128 * signed division: Arithmetic. (line 111) 43129 * signed division with signed saturation: Arithmetic. (line 111) 43130 * signed maximum: Arithmetic. (line 136) 43131 * signed minimum: Arithmetic. (line 136) 43132 * SImode: Machine Modes. (line 37) 43133 * simple constraints: Simple Constraints. (line 6) 43134 * sincos math function, implicit usage: Library Calls. (line 84) 43135 * sinM2 instruction pattern: Standard Names. (line 489) 43136 * SIZE_ASM_OP: Label Output. (line 23) 43137 * SIZE_TYPE: Type Layout. (line 168) 43138 * skip: GTY Options. (line 76) 43139 * SLOW_BYTE_ACCESS: Costs. (line 66) 43140 * SLOW_UNALIGNED_ACCESS: Costs. (line 81) 43141 * SMALL_REGISTER_CLASSES: Register Classes. (line 433) 43142 * smax: Arithmetic. (line 136) 43143 * smin: Arithmetic. (line 136) 43144 * sms, swing, software pipelining: RTL passes. (line 131) 43145 * smulM3_highpart instruction pattern: Standard Names. (line 356) 43146 * soft float library: Soft float library routines. 43147 (line 6) 43148 * special: GTY Options. (line 229) 43149 * special predicates: Predicates. (line 31) 43150 * SPECS: Target Fragment. (line 108) 43151 * speed of instructions: Costs. (line 6) 43152 * split_block: Maintaining the CFG. 43153 (line 110) 43154 * splitting instructions: Insn Splitting. (line 6) 43155 * SQmode: Machine Modes. (line 111) 43156 * sqrt: Arithmetic. (line 198) 43157 * sqrtM2 instruction pattern: Standard Names. (line 455) 43158 * square root: Arithmetic. (line 198) 43159 * ss_ashift: Arithmetic. (line 168) 43160 * ss_div: Arithmetic. (line 111) 43161 * ss_minus: Arithmetic. (line 36) 43162 * ss_mult: Arithmetic. (line 92) 43163 * ss_neg: Arithmetic. (line 81) 43164 * ss_plus: Arithmetic. (line 14) 43165 * ss_truncate: Conversions. (line 43) 43166 * SSA: SSA. (line 6) 43167 * SSA_NAME_DEF_STMT: SSA. (line 221) 43168 * SSA_NAME_VERSION: SSA. (line 226) 43169 * ssaddM3 instruction pattern: Standard Names. (line 222) 43170 * ssashlM3 instruction pattern: Standard Names. (line 431) 43171 * ssdivM3 instruction pattern: Standard Names. (line 222) 43172 * ssmaddMN4 instruction pattern: Standard Names. (line 379) 43173 * ssmsubMN4 instruction pattern: Standard Names. (line 403) 43174 * ssmulM3 instruction pattern: Standard Names. (line 222) 43175 * ssnegM2 instruction pattern: Standard Names. (line 449) 43176 * sssubM3 instruction pattern: Standard Names. (line 222) 43177 * ssum_widenM3 instruction pattern: Standard Names. (line 274) 43178 * stack arguments: Stack Arguments. (line 6) 43179 * stack frame layout: Frame Layout. (line 6) 43180 * stack smashing protection: Stack Smashing Protection. 43181 (line 6) 43182 * STACK_ALIGNMENT_NEEDED: Frame Layout. (line 48) 43183 * STACK_BOUNDARY: Storage Layout. (line 150) 43184 * STACK_CHECK_BUILTIN: Stack Checking. (line 32) 43185 * STACK_CHECK_FIXED_FRAME_SIZE: Stack Checking. (line 77) 43186 * STACK_CHECK_MAX_FRAME_SIZE: Stack Checking. (line 68) 43187 * STACK_CHECK_MAX_VAR_SIZE: Stack Checking. (line 84) 43188 * STACK_CHECK_PROBE_INTERVAL: Stack Checking. (line 46) 43189 * STACK_CHECK_PROBE_LOAD: Stack Checking. (line 53) 43190 * STACK_CHECK_PROTECT: Stack Checking. (line 59) 43191 * STACK_CHECK_STATIC_BUILTIN: Stack Checking. (line 39) 43192 * STACK_DYNAMIC_OFFSET: Frame Layout. (line 75) 43193 * STACK_DYNAMIC_OFFSET and virtual registers: Regs and Memory. 43194 (line 83) 43195 * STACK_GROWS_DOWNWARD: Frame Layout. (line 9) 43196 * STACK_PARMS_IN_REG_PARM_AREA: Stack Arguments. (line 81) 43197 * STACK_POINTER_OFFSET: Frame Layout. (line 58) 43198 * STACK_POINTER_OFFSET and virtual registers: Regs and Memory. 43199 (line 93) 43200 * STACK_POINTER_REGNUM: Frame Registers. (line 9) 43201 * STACK_POINTER_REGNUM and virtual registers: Regs and Memory. 43202 (line 83) 43203 * stack_pointer_rtx: Frame Registers. (line 85) 43204 * stack_protect_set instruction pattern: Standard Names. (line 1534) 43205 * stack_protect_test instruction pattern: Standard Names. (line 1544) 43206 * STACK_PUSH_CODE: Frame Layout. (line 17) 43207 * STACK_REGS: Stack Registers. (line 20) 43208 * STACK_SAVEAREA_MODE: Storage Layout. (line 427) 43209 * STACK_SIZE_MODE: Storage Layout. (line 439) 43210 * STACK_SLOT_ALIGNMENT: Storage Layout. (line 265) 43211 * standard pattern names: Standard Names. (line 6) 43212 * STANDARD_INCLUDE_COMPONENT: Driver. (line 425) 43213 * STANDARD_INCLUDE_DIR: Driver. (line 417) 43214 * STANDARD_STARTFILE_PREFIX: Driver. (line 337) 43215 * STANDARD_STARTFILE_PREFIX_1: Driver. (line 344) 43216 * STANDARD_STARTFILE_PREFIX_2: Driver. (line 351) 43217 * STARTFILE_SPEC: Driver. (line 210) 43218 * STARTING_FRAME_OFFSET: Frame Layout. (line 39) 43219 * STARTING_FRAME_OFFSET and virtual registers: Regs and Memory. 43220 (line 74) 43221 * Statement and operand traversals: Statement and operand traversals. 43222 (line 6) 43223 * Statement Sequences: Statement Sequences. 43224 (line 6) 43225 * Statements: Statements. (line 6) 43226 * statements: Function Bodies. (line 6) 43227 * Static profile estimation: Profile information. 43228 (line 24) 43229 * static single assignment: SSA. (line 6) 43230 * STATIC_CHAIN: Frame Registers. (line 77) 43231 * STATIC_CHAIN_INCOMING: Frame Registers. (line 78) 43232 * STATIC_CHAIN_INCOMING_REGNUM: Frame Registers. (line 64) 43233 * STATIC_CHAIN_REGNUM: Frame Registers. (line 63) 43234 * stdarg.h and register arguments: Register Arguments. (line 47) 43235 * STDC_0_IN_SYSTEM_HEADERS: Misc. (line 365) 43236 * STMT_EXPR: Expression trees. (line 6) 43237 * STMT_IS_FULL_EXPR_P: Function Bodies. (line 22) 43238 * storage layout: Storage Layout. (line 6) 43239 * STORE_BY_PIECES_P: Costs. (line 152) 43240 * STORE_FLAG_VALUE: Misc. (line 216) 43241 * store_multiple instruction pattern: Standard Names. (line 160) 43242 * strcpy: Storage Layout. (line 235) 43243 * STRICT_ALIGNMENT: Storage Layout. (line 309) 43244 * strict_low_part: RTL Declarations. (line 9) 43245 * strict_memory_address_p: Addressing Modes. (line 179) 43246 * STRING_CST: Expression trees. (line 6) 43247 * STRING_POOL_ADDRESS_P: Flags. (line 183) 43248 * strlenM instruction pattern: Standard Names. (line 778) 43249 * structure value address: Aggregate Return. (line 6) 43250 * STRUCTURE_SIZE_BOUNDARY: Storage Layout. (line 301) 43251 * structures, returning: Interface. (line 10) 43252 * subM3 instruction pattern: Standard Names. (line 222) 43253 * SUBOBJECT: Function Bodies. (line 6) 43254 * SUBOBJECT_CLEANUP: Function Bodies. (line 6) 43255 * subreg: Regs and Memory. (line 97) 43256 * subreg and /s: Flags. (line 205) 43257 * subreg and /u: Flags. (line 198) 43258 * subreg and /u and /v: Flags. (line 188) 43259 * subreg, in strict_low_part: RTL Declarations. (line 9) 43260 * SUBREG_BYTE: Regs and Memory. (line 289) 43261 * SUBREG_PROMOTED_UNSIGNED_P: Flags. (line 188) 43262 * SUBREG_PROMOTED_UNSIGNED_SET: Flags. (line 198) 43263 * SUBREG_PROMOTED_VAR_P: Flags. (line 205) 43264 * SUBREG_REG: Regs and Memory. (line 289) 43265 * SUCCESS_EXIT_CODE: Host Misc. (line 12) 43266 * SUPPORTS_INIT_PRIORITY: Macros for Initialization. 43267 (line 58) 43268 * SUPPORTS_ONE_ONLY: Label Output. (line 227) 43269 * SUPPORTS_WEAK: Label Output. (line 208) 43270 * SWITCH_BODY: Function Bodies. (line 6) 43271 * SWITCH_COND: Function Bodies. (line 6) 43272 * SWITCH_CURTAILS_COMPILATION: Driver. (line 33) 43273 * SWITCH_STMT: Function Bodies. (line 6) 43274 * SWITCH_TAKES_ARG: Driver. (line 9) 43275 * SWITCHES_NEED_SPACES: Driver. (line 47) 43276 * SYMBOL_FLAG_ANCHOR: Special Accessors. (line 106) 43277 * SYMBOL_FLAG_EXTERNAL: Special Accessors. (line 88) 43278 * SYMBOL_FLAG_FUNCTION: Special Accessors. (line 81) 43279 * SYMBOL_FLAG_HAS_BLOCK_INFO: Special Accessors. (line 102) 43280 * SYMBOL_FLAG_LOCAL: Special Accessors. (line 84) 43281 * SYMBOL_FLAG_SMALL: Special Accessors. (line 93) 43282 * SYMBOL_FLAG_TLS_SHIFT: Special Accessors. (line 97) 43283 * symbol_ref: Constants. (line 76) 43284 * symbol_ref and /f: Flags. (line 183) 43285 * symbol_ref and /i: Flags. (line 220) 43286 * symbol_ref and /u: Flags. (line 10) 43287 * symbol_ref and /v: Flags. (line 224) 43288 * symbol_ref, RTL sharing: Sharing. (line 20) 43289 * SYMBOL_REF_ANCHOR_P: Special Accessors. (line 106) 43290 * SYMBOL_REF_BLOCK: Special Accessors. (line 119) 43291 * SYMBOL_REF_BLOCK_OFFSET: Special Accessors. (line 124) 43292 * SYMBOL_REF_CONSTANT: Special Accessors. (line 67) 43293 * SYMBOL_REF_DATA: Special Accessors. (line 71) 43294 * SYMBOL_REF_DECL: Special Accessors. (line 55) 43295 * SYMBOL_REF_EXTERNAL_P: Special Accessors. (line 88) 43296 * SYMBOL_REF_FLAG: Flags. (line 224) 43297 * SYMBOL_REF_FLAG, in TARGET_ENCODE_SECTION_INFO: Sections. (line 259) 43298 * SYMBOL_REF_FLAGS: Special Accessors. (line 75) 43299 * SYMBOL_REF_FUNCTION_P: Special Accessors. (line 81) 43300 * SYMBOL_REF_HAS_BLOCK_INFO_P: Special Accessors. (line 102) 43301 * SYMBOL_REF_LOCAL_P: Special Accessors. (line 84) 43302 * SYMBOL_REF_SMALL_P: Special Accessors. (line 93) 43303 * SYMBOL_REF_TLS_MODEL: Special Accessors. (line 97) 43304 * SYMBOL_REF_USED: Flags. (line 215) 43305 * SYMBOL_REF_WEAK: Flags. (line 220) 43306 * symbolic label: Sharing. (line 20) 43307 * sync_addMODE instruction pattern: Standard Names. (line 1450) 43308 * sync_andMODE instruction pattern: Standard Names. (line 1450) 43309 * sync_compare_and_swap_ccMODE instruction pattern: Standard Names. 43310 (line 1437) 43311 * sync_compare_and_swapMODE instruction pattern: Standard Names. 43312 (line 1419) 43313 * sync_iorMODE instruction pattern: Standard Names. (line 1450) 43314 * sync_lock_releaseMODE instruction pattern: Standard Names. (line 1515) 43315 * sync_lock_test_and_setMODE instruction pattern: Standard Names. 43316 (line 1489) 43317 * sync_nandMODE instruction pattern: Standard Names. (line 1450) 43318 * sync_new_addMODE instruction pattern: Standard Names. (line 1482) 43319 * sync_new_andMODE instruction pattern: Standard Names. (line 1482) 43320 * sync_new_iorMODE instruction pattern: Standard Names. (line 1482) 43321 * sync_new_nandMODE instruction pattern: Standard Names. (line 1482) 43322 * sync_new_subMODE instruction pattern: Standard Names. (line 1482) 43323 * sync_new_xorMODE instruction pattern: Standard Names. (line 1482) 43324 * sync_old_addMODE instruction pattern: Standard Names. (line 1465) 43325 * sync_old_andMODE instruction pattern: Standard Names. (line 1465) 43326 * sync_old_iorMODE instruction pattern: Standard Names. (line 1465) 43327 * sync_old_nandMODE instruction pattern: Standard Names. (line 1465) 43328 * sync_old_subMODE instruction pattern: Standard Names. (line 1465) 43329 * sync_old_xorMODE instruction pattern: Standard Names. (line 1465) 43330 * sync_subMODE instruction pattern: Standard Names. (line 1450) 43331 * sync_xorMODE instruction pattern: Standard Names. (line 1450) 43332 * SYSROOT_HEADERS_SUFFIX_SPEC: Driver. (line 239) 43333 * SYSROOT_SUFFIX_SPEC: Driver. (line 234) 43334 * SYSTEM_INCLUDE_DIR: Driver. (line 408) 43335 * t-TARGET: Target Fragment. (line 6) 43336 * table jump: Basic Blocks. (line 57) 43337 * tablejump instruction pattern: Standard Names. (line 1102) 43338 * tag: GTY Options. (line 81) 43339 * tagging insns: Tagging Insns. (line 6) 43340 * tail calls: Tail Calls. (line 6) 43341 * TAmode: Machine Modes. (line 156) 43342 * target attributes: Target Attributes. (line 6) 43343 * target description macros: Target Macros. (line 6) 43344 * target functions: Target Structure. (line 6) 43345 * target hooks: Target Structure. (line 6) 43346 * target makefile fragment: Target Fragment. (line 6) 43347 * target specifications: Run-time Target. (line 6) 43348 * TARGET_ADDRESS_COST: Costs. (line 236) 43349 * TARGET_ALIGN_ANON_BITFIELD: Storage Layout. (line 386) 43350 * TARGET_ALLOCATE_INITIAL_VALUE: Misc. (line 712) 43351 * TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS: Misc. (line 951) 43352 * TARGET_ARG_PARTIAL_BYTES: Register Arguments. (line 83) 43353 * TARGET_ARM_EABI_UNWINDER: Exception Region Output. 43354 (line 113) 43355 * TARGET_ASM_ALIGNED_DI_OP: Data Output. (line 10) 43356 * TARGET_ASM_ALIGNED_HI_OP: Data Output. (line 8) 43357 * TARGET_ASM_ALIGNED_SI_OP: Data Output. (line 9) 43358 * TARGET_ASM_ALIGNED_TI_OP: Data Output. (line 11) 43359 * TARGET_ASM_ASSEMBLE_VISIBILITY: Label Output. (line 239) 43360 * TARGET_ASM_BYTE_OP: Data Output. (line 7) 43361 * TARGET_ASM_CAN_OUTPUT_MI_THUNK: Function Entry. (line 237) 43362 * TARGET_ASM_CLOSE_PAREN: Data Output. (line 130) 43363 * TARGET_ASM_CONSTRUCTOR: Macros for Initialization. 43364 (line 69) 43365 * TARGET_ASM_DESTRUCTOR: Macros for Initialization. 43366 (line 83) 43367 * TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL: Dispatch Tables. (line 74) 43368 * TARGET_ASM_EMIT_UNWIND_LABEL: Dispatch Tables. (line 63) 43369 * TARGET_ASM_EXTERNAL_LIBCALL: Label Output. (line 274) 43370 * TARGET_ASM_FILE_END: File Framework. (line 37) 43371 * TARGET_ASM_FILE_START: File Framework. (line 9) 43372 * TARGET_ASM_FILE_START_APP_OFF: File Framework. (line 17) 43373 * TARGET_ASM_FILE_START_FILE_DIRECTIVE: File Framework. (line 31) 43374 * TARGET_ASM_FUNCTION_BEGIN_EPILOGUE: Function Entry. (line 61) 43375 * TARGET_ASM_FUNCTION_END_PROLOGUE: Function Entry. (line 55) 43376 * TARGET_ASM_FUNCTION_EPILOGUE: Function Entry. (line 68) 43377 * TARGET_ASM_FUNCTION_EPILOGUE and trampolines: Trampolines. (line 70) 43378 * TARGET_ASM_FUNCTION_PROLOGUE: Function Entry. (line 11) 43379 * TARGET_ASM_FUNCTION_PROLOGUE and trampolines: Trampolines. (line 70) 43380 * TARGET_ASM_FUNCTION_RODATA_SECTION: Sections. (line 206) 43381 * TARGET_ASM_GLOBALIZE_DECL_NAME: Label Output. (line 174) 43382 * TARGET_ASM_GLOBALIZE_LABEL: Label Output. (line 165) 43383 * TARGET_ASM_INIT_SECTIONS: Sections. (line 151) 43384 * TARGET_ASM_INTEGER: Data Output. (line 27) 43385 * TARGET_ASM_INTERNAL_LABEL: Label Output. (line 309) 43386 * TARGET_ASM_MARK_DECL_PRESERVED: Label Output. (line 280) 43387 * TARGET_ASM_NAMED_SECTION: File Framework. (line 89) 43388 * TARGET_ASM_OPEN_PAREN: Data Output. (line 129) 43389 * TARGET_ASM_OUTPUT_ANCHOR: Anchored Addresses. (line 44) 43390 * TARGET_ASM_OUTPUT_DWARF_DTPREL: SDB and DWARF. (line 58) 43391 * TARGET_ASM_OUTPUT_MI_THUNK: Function Entry. (line 195) 43392 * TARGET_ASM_RECORD_GCC_SWITCHES: File Framework. (line 122) 43393 * TARGET_ASM_RECORD_GCC_SWITCHES_SECTION: File Framework. (line 166) 43394 * TARGET_ASM_SELECT_RTX_SECTION: Sections. (line 214) 43395 * TARGET_ASM_SELECT_SECTION: Sections. (line 172) 43396 * TARGET_ASM_TTYPE: Exception Region Output. 43397 (line 107) 43398 * TARGET_ASM_UNALIGNED_DI_OP: Data Output. (line 14) 43399 * TARGET_ASM_UNALIGNED_HI_OP: Data Output. (line 12) 43400 * TARGET_ASM_UNALIGNED_SI_OP: Data Output. (line 13) 43401 * TARGET_ASM_UNALIGNED_TI_OP: Data Output. (line 15) 43402 * TARGET_ASM_UNIQUE_SECTION: Sections. (line 193) 43403 * TARGET_ATTRIBUTE_TABLE: Target Attributes. (line 11) 43404 * TARGET_BINDS_LOCAL_P: Sections. (line 284) 43405 * TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED: Misc. (line 808) 43406 * TARGET_BRANCH_TARGET_REGISTER_CLASS: Misc. (line 800) 43407 * TARGET_BUILD_BUILTIN_VA_LIST: Register Arguments. (line 263) 43408 * TARGET_BUILTIN_RECIPROCAL: Addressing Modes. (line 240) 43409 * TARGET_BUILTIN_SETJMP_FRAME_VALUE: Frame Layout. (line 109) 43410 * TARGET_C99_FUNCTIONS: Library Calls. (line 77) 43411 * TARGET_CALLEE_COPIES: Register Arguments. (line 115) 43412 * TARGET_CAN_INLINE_P: Target Attributes. (line 126) 43413 * TARGET_CAN_SIMPLIFY_GOT_ACCESS: Misc. (line 983) 43414 * TARGET_CANNOT_FORCE_CONST_MEM: Addressing Modes. (line 221) 43415 * TARGET_CANNOT_MODIFY_JUMPS_P: Misc. (line 787) 43416 * TARGET_CANONICAL_VA_LIST_TYPE: Register Arguments. (line 272) 43417 * TARGET_CLEAR_PIC_REG: Misc. (line 966) 43418 * TARGET_COMMUTATIVE_P: Misc. (line 705) 43419 * TARGET_COMP_TYPE_ATTRIBUTES: Target Attributes. (line 19) 43420 * TARGET_CPU_CPP_BUILTINS: Run-time Target. (line 9) 43421 * TARGET_CXX_ADJUST_CLASS_AT_DEFINITION: C++ ABI. (line 87) 43422 * TARGET_CXX_CDTOR_RETURNS_THIS: C++ ABI. (line 38) 43423 * TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT: C++ ABI. (line 62) 43424 * TARGET_CXX_COOKIE_HAS_SIZE: C++ ABI. (line 25) 43425 * TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY: C++ ABI. (line 54) 43426 * TARGET_CXX_GET_COOKIE_SIZE: C++ ABI. (line 18) 43427 * TARGET_CXX_GUARD_MASK_BIT: C++ ABI. (line 12) 43428 * TARGET_CXX_GUARD_TYPE: C++ ABI. (line 7) 43429 * TARGET_CXX_IMPORT_EXPORT_CLASS: C++ ABI. (line 30) 43430 * TARGET_CXX_KEY_METHOD_MAY_BE_INLINE: C++ ABI. (line 43) 43431 * TARGET_CXX_LIBRARY_RTTI_COMDAT: C++ ABI. (line 69) 43432 * TARGET_CXX_USE_AEABI_ATEXIT: C++ ABI. (line 74) 43433 * TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT: C++ ABI. (line 80) 43434 * TARGET_DECIMAL_FLOAT_SUPPORTED_P: Storage Layout. (line 513) 43435 * TARGET_DECLSPEC: Target Attributes. (line 64) 43436 * TARGET_DEFAULT_PACK_STRUCT: Misc. (line 482) 43437 * TARGET_DEFAULT_SHORT_ENUMS: Type Layout. (line 160) 43438 * TARGET_DEFERRED_OUTPUT_DEFS: Label Output. (line 393) 43439 * TARGET_DELEGITIMIZE_ADDRESS: Addressing Modes. (line 212) 43440 * TARGET_DLLIMPORT_DECL_ATTRIBUTES: Target Attributes. (line 47) 43441 * TARGET_DWARF_CALLING_CONVENTION: SDB and DWARF. (line 18) 43442 * TARGET_DWARF_HANDLE_FRAME_UNSPEC: Frame Layout. (line 172) 43443 * TARGET_DWARF_REGISTER_SPAN: Exception Region Output. 43444 (line 90) 43445 * TARGET_EDOM: Library Calls. (line 59) 43446 * TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS: Emulated TLS. (line 68) 43447 * TARGET_EMUTLS_GET_ADDRESS: Emulated TLS. (line 19) 43448 * TARGET_EMUTLS_REGISTER_COMMON: Emulated TLS. (line 24) 43449 * TARGET_EMUTLS_TMPL_PREFIX: Emulated TLS. (line 45) 43450 * TARGET_EMUTLS_TMPL_SECTION: Emulated TLS. (line 36) 43451 * TARGET_EMUTLS_VAR_ALIGN_FIXED: Emulated TLS. (line 63) 43452 * TARGET_EMUTLS_VAR_FIELDS: Emulated TLS. (line 49) 43453 * TARGET_EMUTLS_VAR_INIT: Emulated TLS. (line 57) 43454 * TARGET_EMUTLS_VAR_PREFIX: Emulated TLS. (line 41) 43455 * TARGET_EMUTLS_VAR_SECTION: Emulated TLS. (line 31) 43456 * TARGET_ENCODE_SECTION_INFO: Sections. (line 235) 43457 * TARGET_ENCODE_SECTION_INFO and address validation: Addressing Modes. 43458 (line 91) 43459 * TARGET_ENCODE_SECTION_INFO usage: Instruction Output. (line 100) 43460 * TARGET_ENUM_VA_LIST: Scalar Return. (line 96) 43461 * TARGET_EXECUTABLE_SUFFIX: Misc. (line 761) 43462 * TARGET_EXPAND_BUILTIN: Misc. (line 657) 43463 * TARGET_EXPAND_BUILTIN_SAVEREGS: Varargs. (line 92) 43464 * TARGET_EXPAND_TO_RTL_HOOK: Storage Layout. (line 519) 43465 * TARGET_EXPR: Expression trees. (line 6) 43466 * TARGET_EXTRA_INCLUDES: Misc. (line 839) 43467 * TARGET_EXTRA_LIVE_ON_ENTRY: Tail Calls. (line 21) 43468 * TARGET_EXTRA_PRE_INCLUDES: Misc. (line 846) 43469 * TARGET_FIXED_CONDITION_CODE_REGS: Condition Code. (line 142) 43470 * TARGET_FIXED_POINT_SUPPORTED_P: Storage Layout. (line 516) 43471 * target_flags: Run-time Target. (line 52) 43472 * TARGET_FLT_EVAL_METHOD: Type Layout. (line 141) 43473 * TARGET_FN_ABI_VA_LIST: Register Arguments. (line 267) 43474 * TARGET_FOLD_BUILTIN: Misc. (line 677) 43475 * TARGET_FORMAT_TYPES: Misc. (line 866) 43476 * TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P: Target Attributes. (line 86) 43477 * TARGET_FUNCTION_OK_FOR_SIBCALL: Tail Calls. (line 8) 43478 * TARGET_FUNCTION_VALUE: Scalar Return. (line 11) 43479 * TARGET_GET_DRAP_RTX: Misc. (line 946) 43480 * TARGET_GET_PIC_REG: Misc. (line 962) 43481 * TARGET_GIMPLIFY_VA_ARG_EXPR: Register Arguments. (line 278) 43482 * TARGET_HANDLE_C_OPTION: Run-time Target. (line 78) 43483 * TARGET_HANDLE_OPTION: Run-time Target. (line 61) 43484 * TARGET_HARD_REGNO_SCRATCH_OK: Values in Registers. 43485 (line 144) 43486 * TARGET_HAS_SINCOS: Library Calls. (line 85) 43487 * TARGET_HAVE_CONDITIONAL_EXECUTION: Misc. (line 822) 43488 * TARGET_HAVE_CTORS_DTORS: Macros for Initialization. 43489 (line 64) 43490 * TARGET_HAVE_NAMED_SECTIONS: File Framework. (line 99) 43491 * TARGET_HAVE_SWITCHABLE_BSS_SECTIONS: File Framework. (line 103) 43492 * TARGET_HELP: Run-time Target. (line 140) 43493 * TARGET_IN_SMALL_DATA_P: Sections. (line 276) 43494 * TARGET_INIT_BUILTINS: Misc. (line 639) 43495 * TARGET_INIT_DWARF_REG_SIZES_EXTRA: Exception Region Output. 43496 (line 99) 43497 * TARGET_INIT_LIBFUNCS: Library Calls. (line 16) 43498 * TARGET_INSERT_ATTRIBUTES: Target Attributes. (line 73) 43499 * TARGET_INSTANTIATE_DECLS: Storage Layout. (line 527) 43500 * TARGET_INVALID_BINARY_OP: Misc. (line 919) 43501 * TARGET_INVALID_CONVERSION: Misc. (line 906) 43502 * TARGET_INVALID_UNARY_OP: Misc. (line 912) 43503 * TARGET_IRA_COVER_CLASSES: Register Classes. (line 496) 43504 * TARGET_LIB_INT_CMP_BIASED: Library Calls. (line 35) 43505 * TARGET_LIBCALL_VALUE: Scalar Return. (line 69) 43506 * TARGET_LIBGCC_CMP_RETURN_MODE: Storage Layout. (line 448) 43507 * TARGET_LIBGCC_SDATA_SECTION: Sections. (line 123) 43508 * TARGET_LIBGCC_SHIFT_COUNT_MODE: Storage Layout. (line 454) 43509 * TARGET_LOAD_GLOBAL_ADDRESS: Misc. (line 991) 43510 * TARGET_LOADED_GLOBAL_VAR: Misc. (line 971) 43511 * TARGET_MACHINE_DEPENDENT_REORG: Misc. (line 624) 43512 * TARGET_MANGLE_DECL_ASSEMBLER_NAME: Sections. (line 225) 43513 * TARGET_MANGLE_TYPE: Storage Layout. (line 531) 43514 * TARGET_MD_ASM_CLOBBERS: Misc. (line 540) 43515 * TARGET_MEM_CONSTRAINT: Addressing Modes. (line 100) 43516 * TARGET_MEM_REF: Expression trees. (line 6) 43517 * TARGET_MERGE_DECL_ATTRIBUTES: Target Attributes. (line 39) 43518 * TARGET_MERGE_TYPE_ATTRIBUTES: Target Attributes. (line 31) 43519 * TARGET_MIN_DIVISIONS_FOR_RECIP_MUL: Misc. (line 106) 43520 * TARGET_MODE_REP_EXTENDED: Misc. (line 191) 43521 * TARGET_MS_BITFIELD_LAYOUT_P: Storage Layout. (line 486) 43522 * TARGET_MUST_PASS_IN_STACK: Register Arguments. (line 62) 43523 * TARGET_MUST_PASS_IN_STACK, and FUNCTION_ARG: Register Arguments. 43524 (line 52) 43525 * TARGET_N_FORMAT_TYPES: Misc. (line 871) 43526 * TARGET_NARROW_VOLATILE_BITFIELD: Storage Layout. (line 392) 43527 * TARGET_OBJECT_SUFFIX: Misc. (line 756) 43528 * TARGET_OBJFMT_CPP_BUILTINS: Run-time Target. (line 46) 43529 * TARGET_OPTF: Misc. (line 853) 43530 * TARGET_OPTION_PRAGMA_PARSE: Target Attributes. (line 120) 43531 * TARGET_OPTION_PRINT: Target Attributes. (line 115) 43532 * TARGET_OPTION_RESTORE: Target Attributes. (line 110) 43533 * TARGET_OPTION_SAVE: Target Attributes. (line 104) 43534 * TARGET_OPTION_TRANSLATE_TABLE: Driver. (line 53) 43535 * TARGET_OS_CPP_BUILTINS: Run-time Target. (line 42) 43536 * TARGET_OVERRIDES_FORMAT_ATTRIBUTES: Misc. (line 875) 43537 * TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT: Misc. (line 881) 43538 * TARGET_OVERRIDES_FORMAT_INIT: Misc. (line 885) 43539 * TARGET_PASS_BY_REFERENCE: Register Arguments. (line 103) 43540 * TARGET_POSIX_IO: Misc. (line 564) 43541 * TARGET_PRETEND_OUTGOING_VARARGS_NAMED: Varargs. (line 152) 43542 * TARGET_PROMOTE_FUNCTION_ARGS: Storage Layout. (line 131) 43543 * TARGET_PROMOTE_FUNCTION_RETURN: Storage Layout. (line 136) 43544 * TARGET_PROMOTE_PROTOTYPES: Stack Arguments. (line 11) 43545 * TARGET_PTRMEMFUNC_VBIT_LOCATION: Type Layout. (line 235) 43546 * TARGET_RELAXED_ORDERING: Misc. (line 890) 43547 * TARGET_RESOLVE_OVERLOADED_BUILTIN: Misc. (line 667) 43548 * TARGET_RETURN_IN_MEMORY: Aggregate Return. (line 16) 43549 * TARGET_RETURN_IN_MSB: Scalar Return. (line 112) 43550 * TARGET_RTX_COSTS: Costs. (line 210) 43551 * TARGET_SCALAR_MODE_SUPPORTED_P: Register Arguments. (line 290) 43552 * TARGET_SCHED_ADJUST_COST: Scheduling. (line 37) 43553 * TARGET_SCHED_ADJUST_PRIORITY: Scheduling. (line 52) 43554 * TARGET_SCHED_CLEAR_SCHED_CONTEXT: Scheduling. (line 261) 43555 * TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK: Scheduling. (line 89) 43556 * TARGET_SCHED_DFA_NEW_CYCLE: Scheduling. (line 205) 43557 * TARGET_SCHED_DFA_POST_CYCLE_ADVANCE: Scheduling. (line 160) 43558 * TARGET_SCHED_DFA_POST_CYCLE_INSN: Scheduling. (line 144) 43559 * TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE: Scheduling. (line 153) 43560 * TARGET_SCHED_DFA_PRE_CYCLE_INSN: Scheduling. (line 132) 43561 * TARGET_SCHED_FINISH: Scheduling. (line 109) 43562 * TARGET_SCHED_FINISH_GLOBAL: Scheduling. (line 126) 43563 * TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD: Scheduling. 43564 (line 168) 43565 * TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD: Scheduling. 43566 (line 196) 43567 * TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC: Scheduling. 43568 (line 321) 43569 * TARGET_SCHED_FREE_SCHED_CONTEXT: Scheduling. (line 265) 43570 * TARGET_SCHED_GEN_CHECK: Scheduling. (line 309) 43571 * TARGET_SCHED_H_I_D_EXTENDED: Scheduling. (line 241) 43572 * TARGET_SCHED_INIT: Scheduling. (line 99) 43573 * TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN: Scheduling. (line 149) 43574 * TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN: Scheduling. (line 141) 43575 * TARGET_SCHED_INIT_GLOBAL: Scheduling. (line 118) 43576 * TARGET_SCHED_INIT_SCHED_CONTEXT: Scheduling. (line 251) 43577 * TARGET_SCHED_IS_COSTLY_DEPENDENCE: Scheduling. (line 219) 43578 * TARGET_SCHED_ISSUE_RATE: Scheduling. (line 12) 43579 * TARGET_SCHED_NEEDS_BLOCK_P: Scheduling. (line 302) 43580 * TARGET_SCHED_REORDER: Scheduling. (line 60) 43581 * TARGET_SCHED_REORDER2: Scheduling. (line 77) 43582 * TARGET_SCHED_SET_SCHED_CONTEXT: Scheduling. (line 257) 43583 * TARGET_SCHED_SET_SCHED_FLAGS: Scheduling. (line 332) 43584 * TARGET_SCHED_SMS_RES_MII: Scheduling. (line 343) 43585 * TARGET_SCHED_SPECULATE_INSN: Scheduling. (line 291) 43586 * TARGET_SCHED_VARIABLE_ISSUE: Scheduling. (line 24) 43587 * TARGET_SECONDARY_RELOAD: Register Classes. (line 257) 43588 * TARGET_SECTION_TYPE_FLAGS: File Framework. (line 109) 43589 * TARGET_SET_CURRENT_FUNCTION: Misc. (line 739) 43590 * TARGET_SET_DEFAULT_TYPE_ATTRIBUTES: Target Attributes. (line 26) 43591 * TARGET_SETUP_INCOMING_VARARGS: Varargs. (line 101) 43592 * TARGET_SHIFT_TRUNCATION_MASK: Misc. (line 154) 43593 * TARGET_SPLIT_COMPLEX_ARG: Register Arguments. (line 251) 43594 * TARGET_STACK_PROTECT_FAIL: Stack Smashing Protection. 43595 (line 17) 43596 * TARGET_STACK_PROTECT_GUARD: Stack Smashing Protection. 43597 (line 7) 43598 * TARGET_STRICT_ARGUMENT_NAMING: Varargs. (line 137) 43599 * TARGET_STRUCT_VALUE_RTX: Aggregate Return. (line 44) 43600 * TARGET_UNSPEC_MAY_TRAP_P: Misc. (line 731) 43601 * TARGET_UNWIND_EMIT: Dispatch Tables. (line 81) 43602 * TARGET_UNWIND_INFO: Exception Region Output. 43603 (line 56) 43604 * TARGET_UPDATE_STACK_BOUNDARY: Misc. (line 942) 43605 * TARGET_USE_ANCHORS_FOR_SYMBOL_P: Anchored Addresses. (line 55) 43606 * TARGET_USE_BLOCKS_FOR_CONSTANT_P: Addressing Modes. (line 233) 43607 * TARGET_USE_JCR_SECTION: Misc. (line 924) 43608 * TARGET_USE_LOCAL_THUNK_ALIAS_P: Misc. (line 859) 43609 * TARGET_USES_WEAK_UNWIND_INFO: Exception Handling. (line 129) 43610 * TARGET_VALID_DLLIMPORT_ATTRIBUTE_P: Target Attributes. (line 59) 43611 * TARGET_VALID_OPTION_ATTRIBUTE_P: Target Attributes. (line 93) 43612 * TARGET_VALID_POINTER_MODE: Register Arguments. (line 284) 43613 * TARGET_VECTOR_MODE_SUPPORTED_P: Register Arguments. (line 302) 43614 * TARGET_VECTOR_OPAQUE_P: Storage Layout. (line 479) 43615 * TARGET_VECTORIZE_BUILTIN_CONVERSION: Addressing Modes. (line 300) 43616 * TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD: Addressing Modes. (line 249) 43617 * TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN: Addressing Modes. (line 275) 43618 * TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD: Addressing Modes. (line 287) 43619 * TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION: Addressing Modes. 43620 (line 315) 43621 * TARGET_VERSION: Run-time Target. (line 91) 43622 * TARGET_VTABLE_DATA_ENTRY_DISTANCE: Type Layout. (line 288) 43623 * TARGET_VTABLE_ENTRY_ALIGN: Type Layout. (line 282) 43624 * TARGET_VTABLE_USES_DESCRIPTORS: Type Layout. (line 271) 43625 * TARGET_WEAK_NOT_IN_ARCHIVE_TOC: Label Output. (line 245) 43626 * targetm: Target Structure. (line 7) 43627 * targets, makefile: Makefile. (line 6) 43628 * TCmode: Machine Modes. (line 197) 43629 * TDmode: Machine Modes. (line 94) 43630 * TEMPLATE_DECL: Declarations. (line 6) 43631 * Temporaries: Temporaries. (line 6) 43632 * termination routines: Initialization. (line 6) 43633 * testing constraints: C Constraint Interface. 43634 (line 6) 43635 * TEXT_SECTION_ASM_OP: Sections. (line 38) 43636 * TF_SIZE: Type Layout. (line 132) 43637 * TFmode: Machine Modes. (line 98) 43638 * THEN_CLAUSE: Function Bodies. (line 6) 43639 * THREAD_MODEL_SPEC: Driver. (line 225) 43640 * THROW_EXPR: Expression trees. (line 6) 43641 * THUNK_DECL: Declarations. (line 6) 43642 * THUNK_DELTA: Declarations. (line 6) 43643 * TImode: Machine Modes. (line 48) 43644 * TImode, in insn: Insns. (line 231) 43645 * tm.h macros: Target Macros. (line 6) 43646 * TQFmode: Machine Modes. (line 62) 43647 * TQmode: Machine Modes. (line 119) 43648 * TRAMPOLINE_ADJUST_ADDRESS: Trampolines. (line 62) 43649 * TRAMPOLINE_ALIGNMENT: Trampolines. (line 49) 43650 * TRAMPOLINE_SECTION: Trampolines. (line 40) 43651 * TRAMPOLINE_SIZE: Trampolines. (line 45) 43652 * TRAMPOLINE_TEMPLATE: Trampolines. (line 29) 43653 * trampolines for nested functions: Trampolines. (line 6) 43654 * TRANSFER_FROM_TRAMPOLINE: Trampolines. (line 124) 43655 * trap instruction pattern: Standard Names. (line 1374) 43656 * tree <1>: Macros and Functions. 43657 (line 6) 43658 * tree: Tree overview. (line 6) 43659 * Tree SSA: Tree SSA. (line 6) 43660 * tree_code <1>: GIMPLE_OMP_FOR. (line 83) 43661 * tree_code <2>: GIMPLE_COND. (line 21) 43662 * tree_code <3>: GIMPLE_ASSIGN. (line 41) 43663 * tree_code: Manipulating GIMPLE statements. 43664 (line 31) 43665 * TREE_CODE: Tree overview. (line 6) 43666 * TREE_FILENAME: Working with declarations. 43667 (line 14) 43668 * tree_int_cst_equal: Expression trees. (line 6) 43669 * TREE_INT_CST_HIGH: Expression trees. (line 6) 43670 * TREE_INT_CST_LOW: Expression trees. (line 6) 43671 * tree_int_cst_lt: Expression trees. (line 6) 43672 * TREE_LINENO: Working with declarations. 43673 (line 20) 43674 * TREE_LIST: Containers. (line 6) 43675 * TREE_OPERAND: Expression trees. (line 6) 43676 * TREE_PUBLIC: Function Basics. (line 6) 43677 * TREE_PURPOSE: Containers. (line 6) 43678 * TREE_STRING_LENGTH: Expression trees. (line 6) 43679 * TREE_STRING_POINTER: Expression trees. (line 6) 43680 * TREE_TYPE <1>: Expression trees. (line 6) 43681 * TREE_TYPE <2>: Function Basics. (line 171) 43682 * TREE_TYPE <3>: Working with declarations. 43683 (line 11) 43684 * TREE_TYPE: Types. (line 6) 43685 * TREE_VALUE: Containers. (line 6) 43686 * TREE_VEC: Containers. (line 6) 43687 * TREE_VEC_ELT: Containers. (line 6) 43688 * TREE_VEC_LENGTH: Containers. (line 6) 43689 * Trees: Trees. (line 6) 43690 * TRULY_NOOP_TRUNCATION: Misc. (line 177) 43691 * TRUNC_DIV_EXPR: Expression trees. (line 6) 43692 * TRUNC_MOD_EXPR: Expression trees. (line 6) 43693 * truncate: Conversions. (line 38) 43694 * truncMN2 instruction pattern: Standard Names. (line 821) 43695 * TRUTH_AND_EXPR: Expression trees. (line 6) 43696 * TRUTH_ANDIF_EXPR: Expression trees. (line 6) 43697 * TRUTH_NOT_EXPR: Expression trees. (line 6) 43698 * TRUTH_OR_EXPR: Expression trees. (line 6) 43699 * TRUTH_ORIF_EXPR: Expression trees. (line 6) 43700 * TRUTH_XOR_EXPR: Expression trees. (line 6) 43701 * TRY_BLOCK: Function Bodies. (line 6) 43702 * TRY_HANDLERS: Function Bodies. (line 6) 43703 * TRY_STMTS: Function Bodies. (line 6) 43704 * tstM instruction pattern: Standard Names. (line 661) 43705 * Tuple specific accessors: Tuple specific accessors. 43706 (line 6) 43707 * tuples: Tuple representation. 43708 (line 6) 43709 * type: Types. (line 6) 43710 * type declaration: Declarations. (line 6) 43711 * TYPE_ALIGN: Types. (line 6) 43712 * TYPE_ARG_TYPES: Types. (line 6) 43713 * TYPE_ASM_OP: Label Output. (line 55) 43714 * TYPE_ATTRIBUTES: Attributes. (line 25) 43715 * TYPE_BINFO: Classes. (line 6) 43716 * TYPE_BUILT_IN: Types. (line 83) 43717 * TYPE_CANONICAL: Types. (line 6) 43718 * TYPE_CONTEXT: Types. (line 6) 43719 * TYPE_DECL: Declarations. (line 6) 43720 * TYPE_FIELDS <1>: Classes. (line 6) 43721 * TYPE_FIELDS: Types. (line 6) 43722 * TYPE_HAS_ARRAY_NEW_OPERATOR: Classes. (line 91) 43723 * TYPE_HAS_DEFAULT_CONSTRUCTOR: Classes. (line 76) 43724 * TYPE_HAS_MUTABLE_P: Classes. (line 81) 43725 * TYPE_HAS_NEW_OPERATOR: Classes. (line 88) 43726 * TYPE_MAIN_VARIANT: Types. (line 6) 43727 * TYPE_MAX_VALUE: Types. (line 6) 43728 * TYPE_METHOD_BASETYPE: Types. (line 6) 43729 * TYPE_METHODS: Classes. (line 6) 43730 * TYPE_MIN_VALUE: Types. (line 6) 43731 * TYPE_NAME: Types. (line 6) 43732 * TYPE_NOTHROW_P: Function Basics. (line 180) 43733 * TYPE_OFFSET_BASETYPE: Types. (line 6) 43734 * TYPE_OPERAND_FMT: Label Output. (line 66) 43735 * TYPE_OVERLOADS_ARRAY_REF: Classes. (line 99) 43736 * TYPE_OVERLOADS_ARROW: Classes. (line 102) 43737 * TYPE_OVERLOADS_CALL_EXPR: Classes. (line 95) 43738 * TYPE_POLYMORPHIC_P: Classes. (line 72) 43739 * TYPE_PRECISION: Types. (line 6) 43740 * TYPE_PTR_P: Types. (line 89) 43741 * TYPE_PTRFN_P: Types. (line 93) 43742 * TYPE_PTRMEM_P: Types. (line 6) 43743 * TYPE_PTROB_P: Types. (line 96) 43744 * TYPE_PTROBV_P: Types. (line 6) 43745 * TYPE_QUAL_CONST: Types. (line 6) 43746 * TYPE_QUAL_RESTRICT: Types. (line 6) 43747 * TYPE_QUAL_VOLATILE: Types. (line 6) 43748 * TYPE_RAISES_EXCEPTIONS: Function Basics. (line 175) 43749 * TYPE_SIZE: Types. (line 6) 43750 * TYPE_STRUCTURAL_EQUALITY_P: Types. (line 6) 43751 * TYPE_UNQUALIFIED: Types. (line 6) 43752 * TYPE_VFIELD: Classes. (line 6) 43753 * TYPENAME_TYPE: Types. (line 6) 43754 * TYPENAME_TYPE_FULLNAME: Types. (line 6) 43755 * TYPEOF_TYPE: Types. (line 6) 43756 * UDAmode: Machine Modes. (line 168) 43757 * udiv: Arithmetic. (line 125) 43758 * udivM3 instruction pattern: Standard Names. (line 222) 43759 * udivmodM4 instruction pattern: Standard Names. (line 428) 43760 * udot_prodM instruction pattern: Standard Names. (line 265) 43761 * UDQmode: Machine Modes. (line 136) 43762 * UHAmode: Machine Modes. (line 160) 43763 * UHQmode: Machine Modes. (line 128) 43764 * UINTMAX_TYPE: Type Layout. (line 224) 43765 * umaddMN4 instruction pattern: Standard Names. (line 375) 43766 * umax: Arithmetic. (line 144) 43767 * umaxM3 instruction pattern: Standard Names. (line 222) 43768 * umin: Arithmetic. (line 144) 43769 * uminM3 instruction pattern: Standard Names. (line 222) 43770 * umod: Arithmetic. (line 131) 43771 * umodM3 instruction pattern: Standard Names. (line 222) 43772 * umsubMN4 instruction pattern: Standard Names. (line 399) 43773 * umulhisi3 instruction pattern: Standard Names. (line 347) 43774 * umulM3_highpart instruction pattern: Standard Names. (line 361) 43775 * umulqihi3 instruction pattern: Standard Names. (line 347) 43776 * umulsidi3 instruction pattern: Standard Names. (line 347) 43777 * unchanging: Flags. (line 319) 43778 * unchanging, in call_insn: Flags. (line 19) 43779 * unchanging, in jump_insn, call_insn and insn: Flags. (line 39) 43780 * unchanging, in mem: Flags. (line 152) 43781 * unchanging, in subreg: Flags. (line 188) 43782 * unchanging, in symbol_ref: Flags. (line 10) 43783 * UNEQ_EXPR: Expression trees. (line 6) 43784 * UNGE_EXPR: Expression trees. (line 6) 43785 * UNGT_EXPR: Expression trees. (line 6) 43786 * UNION_TYPE <1>: Classes. (line 6) 43787 * UNION_TYPE: Types. (line 6) 43788 * unions, returning: Interface. (line 10) 43789 * UNITS_PER_SIMD_WORD: Storage Layout. (line 77) 43790 * UNITS_PER_WORD: Storage Layout. (line 67) 43791 * UNKNOWN_TYPE: Types. (line 6) 43792 * UNLE_EXPR: Expression trees. (line 6) 43793 * UNLIKELY_EXECUTED_TEXT_SECTION_NAME: Sections. (line 49) 43794 * UNLT_EXPR: Expression trees. (line 6) 43795 * UNORDERED_EXPR: Expression trees. (line 6) 43796 * unshare_all_rtl: Sharing. (line 58) 43797 * unsigned division: Arithmetic. (line 125) 43798 * unsigned division with unsigned saturation: Arithmetic. (line 125) 43799 * unsigned greater than: Comparisons. (line 64) 43800 * unsigned less than: Comparisons. (line 68) 43801 * unsigned minimum and maximum: Arithmetic. (line 144) 43802 * unsigned_fix: Conversions. (line 77) 43803 * unsigned_float: Conversions. (line 62) 43804 * unsigned_fract_convert: Conversions. (line 97) 43805 * unsigned_sat_fract: Conversions. (line 103) 43806 * unspec: Side Effects. (line 287) 43807 * unspec_volatile: Side Effects. (line 287) 43808 * untyped_call instruction pattern: Standard Names. (line 1012) 43809 * untyped_return instruction pattern: Standard Names. (line 1062) 43810 * UPDATE_PATH_HOST_CANONICALIZE (PATH): Filesystem. (line 59) 43811 * update_ssa: SSA. (line 76) 43812 * update_stmt <1>: SSA Operands. (line 6) 43813 * update_stmt: Manipulating GIMPLE statements. 43814 (line 141) 43815 * update_stmt_if_modified: Manipulating GIMPLE statements. 43816 (line 144) 43817 * UQQmode: Machine Modes. (line 123) 43818 * US Software GOFAST, floating point emulation library: Library Calls. 43819 (line 44) 43820 * us_ashift: Arithmetic. (line 168) 43821 * us_minus: Arithmetic. (line 36) 43822 * us_mult: Arithmetic. (line 92) 43823 * us_neg: Arithmetic. (line 81) 43824 * us_plus: Arithmetic. (line 14) 43825 * US_SOFTWARE_GOFAST: Library Calls. (line 45) 43826 * us_truncate: Conversions. (line 48) 43827 * usaddM3 instruction pattern: Standard Names. (line 222) 43828 * USAmode: Machine Modes. (line 164) 43829 * usashlM3 instruction pattern: Standard Names. (line 431) 43830 * usdivM3 instruction pattern: Standard Names. (line 222) 43831 * use: Side Effects. (line 162) 43832 * USE_C_ALLOCA: Host Misc. (line 19) 43833 * USE_LD_AS_NEEDED: Driver. (line 198) 43834 * USE_LOAD_POST_DECREMENT: Costs. (line 165) 43835 * USE_LOAD_POST_INCREMENT: Costs. (line 160) 43836 * USE_LOAD_PRE_DECREMENT: Costs. (line 175) 43837 * USE_LOAD_PRE_INCREMENT: Costs. (line 170) 43838 * use_optype_d: Manipulating GIMPLE statements. 43839 (line 101) 43840 * use_param: GTY Options. (line 113) 43841 * use_paramN: GTY Options. (line 131) 43842 * use_params: GTY Options. (line 139) 43843 * USE_SELECT_SECTION_FOR_FUNCTIONS: Sections. (line 185) 43844 * USE_STORE_POST_DECREMENT: Costs. (line 185) 43845 * USE_STORE_POST_INCREMENT: Costs. (line 180) 43846 * USE_STORE_PRE_DECREMENT: Costs. (line 195) 43847 * USE_STORE_PRE_INCREMENT: Costs. (line 190) 43848 * used: Flags. (line 337) 43849 * used, in symbol_ref: Flags. (line 215) 43850 * USER_LABEL_PREFIX: Instruction Output. (line 126) 43851 * USING_DECL: Declarations. (line 6) 43852 * USING_STMT: Function Bodies. (line 6) 43853 * usmaddMN4 instruction pattern: Standard Names. (line 383) 43854 * usmsubMN4 instruction pattern: Standard Names. (line 407) 43855 * usmulhisi3 instruction pattern: Standard Names. (line 351) 43856 * usmulM3 instruction pattern: Standard Names. (line 222) 43857 * usmulqihi3 instruction pattern: Standard Names. (line 351) 43858 * usmulsidi3 instruction pattern: Standard Names. (line 351) 43859 * usnegM2 instruction pattern: Standard Names. (line 449) 43860 * USQmode: Machine Modes. (line 132) 43861 * ussubM3 instruction pattern: Standard Names. (line 222) 43862 * usum_widenM3 instruction pattern: Standard Names. (line 275) 43863 * UTAmode: Machine Modes. (line 172) 43864 * UTQmode: Machine Modes. (line 140) 43865 * V in constraint: Simple Constraints. (line 43) 43866 * VA_ARG_EXPR: Expression trees. (line 6) 43867 * values, returned by functions: Scalar Return. (line 6) 43868 * VAR_DECL <1>: Expression trees. (line 6) 43869 * VAR_DECL: Declarations. (line 6) 43870 * varargs implementation: Varargs. (line 6) 43871 * variable: Declarations. (line 6) 43872 * vashlM3 instruction pattern: Standard Names. (line 445) 43873 * vashrM3 instruction pattern: Standard Names. (line 445) 43874 * vec_concat: Vector Operations. (line 25) 43875 * vec_duplicate: Vector Operations. (line 30) 43876 * VEC_EXTRACT_EVEN_EXPR: Expression trees. (line 6) 43877 * vec_extract_evenM instruction pattern: Standard Names. (line 176) 43878 * VEC_EXTRACT_ODD_EXPR: Expression trees. (line 6) 43879 * vec_extract_oddM instruction pattern: Standard Names. (line 183) 43880 * vec_extractM instruction pattern: Standard Names. (line 171) 43881 * vec_initM instruction pattern: Standard Names. (line 204) 43882 * VEC_INTERLEAVE_HIGH_EXPR: Expression trees. (line 6) 43883 * vec_interleave_highM instruction pattern: Standard Names. (line 190) 43884 * VEC_INTERLEAVE_LOW_EXPR: Expression trees. (line 6) 43885 * vec_interleave_lowM instruction pattern: Standard Names. (line 197) 43886 * VEC_LSHIFT_EXPR: Expression trees. (line 6) 43887 * vec_merge: Vector Operations. (line 11) 43888 * VEC_PACK_FIX_TRUNC_EXPR: Expression trees. (line 6) 43889 * VEC_PACK_SAT_EXPR: Expression trees. (line 6) 43890 * vec_pack_sfix_trunc_M instruction pattern: Standard Names. (line 302) 43891 * vec_pack_ssat_M instruction pattern: Standard Names. (line 295) 43892 * VEC_PACK_TRUNC_EXPR: Expression trees. (line 6) 43893 * vec_pack_trunc_M instruction pattern: Standard Names. (line 288) 43894 * vec_pack_ufix_trunc_M instruction pattern: Standard Names. (line 302) 43895 * vec_pack_usat_M instruction pattern: Standard Names. (line 295) 43896 * VEC_RSHIFT_EXPR: Expression trees. (line 6) 43897 * vec_select: Vector Operations. (line 19) 43898 * vec_setM instruction pattern: Standard Names. (line 166) 43899 * vec_shl_M instruction pattern: Standard Names. (line 282) 43900 * vec_shr_M instruction pattern: Standard Names. (line 282) 43901 * VEC_UNPACK_FLOAT_HI_EXPR: Expression trees. (line 6) 43902 * VEC_UNPACK_FLOAT_LO_EXPR: Expression trees. (line 6) 43903 * VEC_UNPACK_HI_EXPR: Expression trees. (line 6) 43904 * VEC_UNPACK_LO_EXPR: Expression trees. (line 6) 43905 * vec_unpacks_float_hi_M instruction pattern: Standard Names. 43906 (line 324) 43907 * vec_unpacks_float_lo_M instruction pattern: Standard Names. 43908 (line 324) 43909 * vec_unpacks_hi_M instruction pattern: Standard Names. (line 309) 43910 * vec_unpacks_lo_M instruction pattern: Standard Names. (line 309) 43911 * vec_unpacku_float_hi_M instruction pattern: Standard Names. 43912 (line 324) 43913 * vec_unpacku_float_lo_M instruction pattern: Standard Names. 43914 (line 324) 43915 * vec_unpacku_hi_M instruction pattern: Standard Names. (line 317) 43916 * vec_unpacku_lo_M instruction pattern: Standard Names. (line 317) 43917 * VEC_WIDEN_MULT_HI_EXPR: Expression trees. (line 6) 43918 * VEC_WIDEN_MULT_LO_EXPR: Expression trees. (line 6) 43919 * vec_widen_smult_hi_M instruction pattern: Standard Names. (line 333) 43920 * vec_widen_smult_lo_M instruction pattern: Standard Names. (line 333) 43921 * vec_widen_umult_hi_M instruction pattern: Standard Names. (line 333) 43922 * vec_widen_umult_lo__M instruction pattern: Standard Names. (line 333) 43923 * vector: Containers. (line 6) 43924 * vector operations: Vector Operations. (line 6) 43925 * VECTOR_CST: Expression trees. (line 6) 43926 * VECTOR_STORE_FLAG_VALUE: Misc. (line 308) 43927 * virtual operands: SSA Operands. (line 6) 43928 * VIRTUAL_INCOMING_ARGS_REGNUM: Regs and Memory. (line 59) 43929 * VIRTUAL_OUTGOING_ARGS_REGNUM: Regs and Memory. (line 87) 43930 * VIRTUAL_STACK_DYNAMIC_REGNUM: Regs and Memory. (line 78) 43931 * VIRTUAL_STACK_VARS_REGNUM: Regs and Memory. (line 69) 43932 * VLIW: Processor pipeline description. 43933 (line 6) 43934 * vlshrM3 instruction pattern: Standard Names. (line 445) 43935 * VMS: Filesystem. (line 37) 43936 * VMS_DEBUGGING_INFO: VMS Debug. (line 9) 43937 * VOID_TYPE: Types. (line 6) 43938 * VOIDmode: Machine Modes. (line 190) 43939 * volatil: Flags. (line 351) 43940 * volatil, in insn, call_insn, jump_insn, code_label, barrier, and note: Flags. 43941 (line 44) 43942 * volatil, in label_ref and reg_label: Flags. (line 65) 43943 * volatil, in mem, asm_operands, and asm_input: Flags. (line 94) 43944 * volatil, in reg: Flags. (line 116) 43945 * volatil, in subreg: Flags. (line 188) 43946 * volatil, in symbol_ref: Flags. (line 224) 43947 * volatile memory references: Flags. (line 352) 43948 * voptype_d: Manipulating GIMPLE statements. 43949 (line 108) 43950 * voting between constraint alternatives: Class Preferences. (line 6) 43951 * vrotlM3 instruction pattern: Standard Names. (line 445) 43952 * vrotrM3 instruction pattern: Standard Names. (line 445) 43953 * walk_dominator_tree: SSA. (line 256) 43954 * walk_gimple_op: Statement and operand traversals. 43955 (line 32) 43956 * walk_gimple_seq: Statement and operand traversals. 43957 (line 50) 43958 * walk_gimple_stmt: Statement and operand traversals. 43959 (line 13) 43960 * walk_use_def_chains: SSA. (line 232) 43961 * WCHAR_TYPE: Type Layout. (line 192) 43962 * WCHAR_TYPE_SIZE: Type Layout. (line 200) 43963 * which_alternative: Output Statement. (line 59) 43964 * WHILE_BODY: Function Bodies. (line 6) 43965 * WHILE_COND: Function Bodies. (line 6) 43966 * WHILE_STMT: Function Bodies. (line 6) 43967 * WIDEST_HARDWARE_FP_SIZE: Type Layout. (line 147) 43968 * WINT_TYPE: Type Layout. (line 205) 43969 * word_mode: Machine Modes. (line 336) 43970 * WORD_REGISTER_OPERATIONS: Misc. (line 63) 43971 * WORD_SWITCH_TAKES_ARG: Driver. (line 20) 43972 * WORDS_BIG_ENDIAN: Storage Layout. (line 29) 43973 * WORDS_BIG_ENDIAN, effect on subreg: Regs and Memory. (line 217) 43974 * X in constraint: Simple Constraints. (line 114) 43975 * x-HOST: Host Fragment. (line 6) 43976 * XCmode: Machine Modes. (line 197) 43977 * XCOFF_DEBUGGING_INFO: DBX Options. (line 13) 43978 * XEXP: Accessors. (line 6) 43979 * XF_SIZE: Type Layout. (line 131) 43980 * XFmode: Machine Modes. (line 79) 43981 * XINT: Accessors. (line 6) 43982 * xm-MACHINE.h <1>: Host Misc. (line 6) 43983 * xm-MACHINE.h: Filesystem. (line 6) 43984 * xor: Arithmetic. (line 163) 43985 * xor, canonicalization of: Insn Canonicalizations. 43986 (line 84) 43987 * xorM3 instruction pattern: Standard Names. (line 222) 43988 * XSTR: Accessors. (line 6) 43989 * XVEC: Accessors. (line 41) 43990 * XVECEXP: Accessors. (line 48) 43991 * XVECLEN: Accessors. (line 44) 43992 * XWINT: Accessors. (line 6) 43993 * zero_extend: Conversions. (line 28) 43994 * zero_extendMN2 instruction pattern: Standard Names. (line 831) 43995 * zero_extract: Bit-Fields. (line 30) 43996 * zero_extract, canonicalization of: Insn Canonicalizations. 43997 (line 96) 43998 43999 44000 44001 Tag Table: 44002 Node: Top2090 44003 Node: Contributing5172 44004 Node: Portability5913 44005 Node: Interface7701 44006 Node: Libgcc10741 44007 Node: Integer library routines12582 44008 Node: Soft float library routines19421 44009 Node: Decimal float library routines31358 44010 Node: Fixed-point fractional library routines47115 44011 Node: Exception handling routines147513 44012 Node: Miscellaneous routines148620 44013 Node: Languages149003 44014 Node: Source Tree150550 44015 Node: Configure Terms151169 44016 Node: Top Level154127 44017 Node: gcc Directory156897 44018 Node: Subdirectories157866 44019 Node: Configuration159716 44020 Node: Config Fragments160436 44021 Node: System Config161665 44022 Node: Configuration Files162601 44023 Node: Build165176 44024 Node: Makefile165588 44025 Ref: Makefile-Footnote-1172306 44026 Ref: Makefile-Footnote-2172451 44027 Node: Library Files172523 44028 Node: Headers173085 44029 Node: Documentation175168 44030 Node: Texinfo Manuals176027 44031 Node: Man Page Generation178365 44032 Node: Miscellaneous Docs180280 44033 Node: Front End181579 44034 Node: Front End Directory185280 44035 Node: Front End Config190470 44036 Node: Back End193384 44037 Node: Testsuites197061 44038 Node: Test Idioms197925 44039 Node: Test Directives201326 44040 Node: Ada Tests213390 44041 Node: C Tests214682 44042 Node: libgcj Tests219037 44043 Node: gcov Testing220169 44044 Node: profopt Testing223153 44045 Node: compat Testing224596 44046 Node: Torture Tests228840 44047 Node: Options230472 44048 Node: Option file format230913 44049 Node: Option properties233662 44050 Node: Passes239718 44051 Node: Parsing pass240460 44052 Node: Gimplification pass243988 44053 Node: Pass manager245815 44054 Node: Tree SSA passes247297 44055 Node: RTL passes269123 44056 Node: Trees281466 44057 Node: Deficiencies284196 44058 Node: Tree overview284433 44059 Node: Macros and Functions288556 44060 Node: Identifiers288702 44061 Node: Containers290227 44062 Node: Types291382 44063 Node: Scopes307085 44064 Node: Namespaces307847 44065 Node: Classes310659 44066 Node: Declarations315416 44067 Node: Working with declarations315911 44068 Node: Internal structure322368 44069 Node: Current structure hierarchy322750 44070 Node: Adding new DECL node types324842 44071 Node: Functions328913 44072 Node: Function Basics331316 44073 Node: Function Bodies339046 44074 Node: Attributes350288 44075 Node: Expression trees351529 44076 Node: RTL394138 44077 Node: RTL Objects396256 44078 Node: RTL Classes400130 44079 Node: Accessors405082 44080 Node: Special Accessors407476 44081 Node: Flags412694 44082 Node: Machine Modes428562 44083 Node: Constants440878 44084 Node: Regs and Memory446907 44085 Node: Arithmetic464808 44086 Node: Comparisons474328 44087 Node: Bit-Fields478620 44088 Node: Vector Operations480172 44089 Node: Conversions481798 44090 Node: RTL Declarations486296 44091 Node: Side Effects487117 44092 Node: Incdec503440 44093 Node: Assembler506775 44094 Node: Insns508307 44095 Node: Calls532196 44096 Node: Sharing534789 44097 Node: Reading RTL537899 44098 Node: GENERIC538889 44099 Node: Statements540528 44100 Node: Blocks540973 44101 Node: Statement Sequences542226 44102 Node: Empty Statements542559 44103 Node: Jumps543133 44104 Node: Cleanups543786 44105 Node: GIMPLE545539 44106 Node: Tuple representation549160 44107 Node: GIMPLE instruction set557815 44108 Node: GIMPLE Exception Handling559483 44109 Node: Temporaries561398 44110 Ref: Temporaries-Footnote-1562717 44111 Node: Operands562780 44112 Node: Compound Expressions563554 44113 Node: Compound Lvalues563788 44114 Node: Conditional Expressions564554 44115 Node: Logical Operators565224 44116 Node: Manipulating GIMPLE statements571315 44117 Node: Tuple specific accessors577243 44118 Node: `GIMPLE_ASM'578076 44119 Node: `GIMPLE_ASSIGN'580681 44120 Node: `GIMPLE_BIND'584627 44121 Node: `GIMPLE_CALL'586434 44122 Node: `GIMPLE_CATCH'590693 44123 Node: `GIMPLE_CHANGE_DYNAMIC_TYPE'591851 44124 Node: `GIMPLE_COND'593184 44125 Node: `GIMPLE_EH_FILTER'595990 44126 Node: `GIMPLE_LABEL'597476 44127 Node: `GIMPLE_NOP'598451 44128 Node: `GIMPLE_OMP_ATOMIC_LOAD'598820 44129 Node: `GIMPLE_OMP_ATOMIC_STORE'599730 44130 Node: `GIMPLE_OMP_CONTINUE'600369 44131 Node: `GIMPLE_OMP_CRITICAL'601719 44132 Node: `GIMPLE_OMP_FOR'602655 44133 Node: `GIMPLE_OMP_MASTER'606165 44134 Node: `GIMPLE_OMP_ORDERED'606548 44135 Node: `GIMPLE_OMP_PARALLEL'606948 44136 Node: `GIMPLE_OMP_RETURN'609717 44137 Node: `GIMPLE_OMP_SECTION'610367 44138 Node: `GIMPLE_OMP_SECTIONS'611033 44139 Node: `GIMPLE_OMP_SINGLE'612637 44140 Node: `GIMPLE_PHI'613573 44141 Node: `GIMPLE_RESX'614986 44142 Node: `GIMPLE_RETURN'615705 44143 Node: `GIMPLE_SWITCH'616273 44144 Node: `GIMPLE_TRY'618403 44145 Node: `GIMPLE_WITH_CLEANUP_EXPR'620193 44146 Node: GIMPLE sequences621076 44147 Node: Sequence iterators624282 44148 Node: Adding a new GIMPLE statement code632737 44149 Node: Statement and operand traversals634017 44150 Node: Tree SSA636627 44151 Node: Annotations638347 44152 Node: SSA Operands638873 44153 Node: SSA653404 44154 Node: Alias analysis665695 44155 Node: Loop Analysis and Representation673151 44156 Node: Loop representation674332 44157 Node: Loop querying681252 44158 Node: Loop manipulation684085 44159 Node: LCSSA686453 44160 Node: Scalar evolutions688525 44161 Node: loop-iv691769 44162 Node: Number of iterations693695 44163 Node: Dependency analysis696504 44164 Node: Lambda702872 44165 Node: Omega704542 44166 Node: Control Flow706107 44167 Node: Basic Blocks707102 44168 Node: Edges711670 44169 Node: Profile information720232 44170 Node: Maintaining the CFG724918 44171 Node: Liveness information731800 44172 Node: Machine Desc733927 44173 Node: Overview736395 44174 Node: Patterns738436 44175 Node: Example741874 44176 Node: RTL Template743309 44177 Node: Output Template753964 44178 Node: Output Statement757930 44179 Node: Predicates761892 44180 Node: Machine-Independent Predicates764810 44181 Node: Defining Predicates769442 44182 Node: Constraints775407 44183 Node: Simple Constraints776655 44184 Node: Multi-Alternative788861 44185 Node: Class Preferences791702 44186 Node: Modifiers792594 44187 Node: Machine Constraints796726 44188 Node: Disable Insn Alternatives829449 44189 Node: Define Constraints832342 44190 Node: C Constraint Interface839122 44191 Node: Standard Names842763 44192 Ref: shift patterns861691 44193 Ref: prologue instruction pattern902709 44194 Ref: epilogue instruction pattern903202 44195 Node: Pattern Ordering912701 44196 Node: Dependent Patterns913937 44197 Node: Jump Patterns916751 44198 Node: Looping Patterns922447 44199 Node: Insn Canonicalizations927175 44200 Node: Expander Definitions931559 44201 Node: Insn Splitting939677 44202 Node: Including Patterns949280 44203 Node: Peephole Definitions951060 44204 Node: define_peephole952313 44205 Node: define_peephole2958644 44206 Node: Insn Attributes961711 44207 Node: Defining Attributes962817 44208 Node: Expressions965337 44209 Node: Tagging Insns971939 44210 Node: Attr Example976292 44211 Node: Insn Lengths978666 44212 Node: Constant Attributes981725 44213 Node: Delay Slots982894 44214 Node: Processor pipeline description986118 44215 Ref: Processor pipeline description-Footnote-11003736 44216 Node: Conditional Execution1004058 44217 Node: Constant Definitions1006911 44218 Node: Iterators1008506 44219 Node: Mode Iterators1008953 44220 Node: Defining Mode Iterators1009931 44221 Node: Substitutions1011425 44222 Node: Examples1013666 44223 Node: Code Iterators1015114 44224 Node: Target Macros1017371 44225 Node: Target Structure1020394 44226 Node: Driver1021663 44227 Node: Run-time Target1045344 44228 Node: Per-Function Data1052468 44229 Node: Storage Layout1055231 44230 Node: Type Layout1080645 44231 Node: Registers1093602 44232 Node: Register Basics1094576 44233 Node: Allocation Order1100143 44234 Node: Values in Registers1102589 44235 Node: Leaf Functions1110078 44236 Node: Stack Registers1112936 44237 Node: Register Classes1114052 44238 Node: Old Constraints1140764 44239 Node: Stack and Calling1147915 44240 Node: Frame Layout1148449 44241 Node: Exception Handling1159295 44242 Node: Stack Checking1165673 44243 Node: Frame Registers1170060 44244 Node: Elimination1176666 44245 Node: Stack Arguments1180697 44246 Node: Register Arguments1187500 44247 Node: Scalar Return1202948 44248 Node: Aggregate Return1209021 44249 Node: Caller Saves1212680 44250 Node: Function Entry1213858 44251 Node: Profiling1226473 44252 Node: Tail Calls1228172 44253 Node: Stack Smashing Protection1229539 44254 Node: Varargs1230651 44255 Node: Trampolines1238611 44256 Node: Library Calls1245277 44257 Node: Addressing Modes1250127 44258 Node: Anchored Addresses1266045 44259 Node: Condition Code1268706 44260 Node: Costs1276995 44261 Node: Scheduling1290094 44262 Node: Sections1308655 44263 Node: PIC1323305 44264 Node: Assembler Format1325295 44265 Node: File Framework1326433 44266 Ref: TARGET_HAVE_SWITCHABLE_BSS_SECTIONS1331339 44267 Node: Data Output1334605 44268 Node: Uninitialized Data1342364 44269 Node: Label Output1347435 44270 Node: Initialization1369102 44271 Node: Macros for Initialization1375064 44272 Node: Instruction Output1381516 44273 Node: Dispatch Tables1390510 44274 Node: Exception Region Output1394305 44275 Node: Alignment Output1400065 44276 Node: Debugging Info1404228 44277 Node: All Debuggers1404898 44278 Node: DBX Options1407753 44279 Node: DBX Hooks1413202 44280 Node: File Names and DBX1415128 44281 Node: SDB and DWARF1417239 44282 Node: VMS Debug1421231 44283 Node: Floating Point1421801 44284 Node: Mode Switching1426624 44285 Node: Target Attributes1430550 44286 Node: Emulated TLS1437314 44287 Node: MIPS Coprocessors1440704 44288 Node: PCH Target1442273 44289 Node: C++ ABI1443794 44290 Node: Misc1448413 44291 Ref: TARGET_SHIFT_TRUNCATION_MASK1455783 44292 Node: Host Config1498447 44293 Node: Host Common1499515 44294 Node: Filesystem1501894 44295 Node: Host Misc1506009 44296 Node: Fragments1508148 44297 Node: Target Fragment1509343 44298 Node: Host Fragment1515233 44299 Node: Collect21515473 44300 Node: Header Dirs1518016 44301 Node: Type Information1519439 44302 Node: GTY Options1521730 44303 Node: GGC Roots1532405 44304 Node: Files1533125 44305 Node: Invoking the garbage collector1535875 44306 Node: Plugins1536928 44307 Node: Funding1547293 44308 Node: GNU Project1549780 44309 Node: Copying1550429 44310 Node: GNU Free Documentation License1587960 44311 Node: Contributors1610369 44312 Node: Option Index1646699 44313 Node: Concept Index1647284 44314 44315 End Tag Table 44316
