1 <?xml version="1.0"?> <!-- -*- sgml -*- --> 2 <!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" 3 "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd" 4 [ <!ENTITY % vg-entities SYSTEM "../../docs/xml/vg-entities.xml"> %vg-entities; ]> 5 6 <!-- Referenced from both the manual and manpage --> 7 <chapter id="&vg-cg-manual-id;" xreflabel="&vg-cg-manual-label;"> 8 <title>Cachegrind: a cache and branch-prediction profiler</title> 9 10 <para>To use this tool, you must specify 11 <option>--tool=cachegrind</option> on the 12 Valgrind command line.</para> 13 14 <sect1 id="cg-manual.overview" xreflabel="Overview"> 15 <title>Overview</title> 16 17 <para>Cachegrind simulates how your program interacts with a machine's cache 18 hierarchy and (optionally) branch predictor. It simulates a machine with 19 independent first-level instruction and data caches (I1 and D1), backed by a 20 unified second-level cache (L2). This exactly matches the configuration of 21 many modern machines.</para> 22 23 <para>However, some modern machines have three or four levels of cache. For these 24 machines (in the cases where Cachegrind can auto-detect the cache 25 configuration) Cachegrind simulates the first-level and last-level caches. 26 The reason for this choice is that the last-level cache has the most influence on 27 runtime, as it masks accesses to main memory. Furthermore, the L1 caches 28 often have low associativity, so simulating them can detect cases where the 29 code interacts badly with this cache (eg. traversing a matrix column-wise 30 with the row length being a power of 2).</para> 31 32 <para>Therefore, Cachegrind always refers to the I1, D1 and LL (last-level) 33 caches.</para> 34 35 <para> 36 Cachegrind gathers the following statistics (abbreviations used for each statistic 37 is given in parentheses):</para> 38 <itemizedlist> 39 <listitem> 40 <para>I cache reads (<computeroutput>Ir</computeroutput>, 41 which equals the number of instructions executed), 42 I1 cache read misses (<computeroutput>I1mr</computeroutput>) and 43 LL cache instruction read misses (<computeroutput>ILmr</computeroutput>). 44 </para> 45 </listitem> 46 <listitem> 47 <para>D cache reads (<computeroutput>Dr</computeroutput>, which 48 equals the number of memory reads), 49 D1 cache read misses (<computeroutput>D1mr</computeroutput>), and 50 LL cache data read misses (<computeroutput>DLmr</computeroutput>). 51 </para> 52 </listitem> 53 <listitem> 54 <para>D cache writes (<computeroutput>Dw</computeroutput>, which equals 55 the number of memory writes), 56 D1 cache write misses (<computeroutput>D1mw</computeroutput>), and 57 LL cache data write misses (<computeroutput>DLmw</computeroutput>). 58 </para> 59 </listitem> 60 <listitem> 61 <para>Conditional branches executed (<computeroutput>Bc</computeroutput>) and 62 conditional branches mispredicted (<computeroutput>Bcm</computeroutput>). 63 </para> 64 </listitem> 65 <listitem> 66 <para>Indirect branches executed (<computeroutput>Bi</computeroutput>) and 67 indirect branches mispredicted (<computeroutput>Bim</computeroutput>). 68 </para> 69 </listitem> 70 </itemizedlist> 71 72 <para>Note that D1 total accesses is given by 73 <computeroutput>D1mr</computeroutput> + 74 <computeroutput>D1mw</computeroutput>, and that LL total 75 accesses is given by <computeroutput>ILmr</computeroutput> + 76 <computeroutput>DLmr</computeroutput> + 77 <computeroutput>DLmw</computeroutput>. 78 </para> 79 80 <para>These statistics are presented for the entire program and for each 81 function in the program. You can also annotate each line of source code in 82 the program with the counts that were caused directly by it.</para> 83 84 <para>On a modern machine, an L1 miss will typically cost 85 around 10 cycles, an LL miss can cost as much as 200 86 cycles, and a mispredicted branch costs in the region of 10 87 to 30 cycles. Detailed cache and branch profiling can be very useful 88 for understanding how your program interacts with the machine and thus how 89 to make it faster.</para> 90 91 <para>Also, since one instruction cache read is performed per 92 instruction executed, you can find out how many instructions are 93 executed per line, which can be useful for traditional profiling.</para> 94 95 </sect1> 96 97 98 99 <sect1 id="cg-manual.profile" 100 xreflabel="Using Cachegrind, cg_annotate and cg_merge"> 101 <title>Using Cachegrind, cg_annotate and cg_merge</title> 102 103 <para>First off, as for normal Valgrind use, you probably want to 104 compile with debugging info (the 105 <option>-g</option> option). But by contrast with 106 normal Valgrind use, you probably do want to turn 107 optimisation on, since you should profile your program as it will 108 be normally run.</para> 109 110 <para>Then, you need to run Cachegrind itself to gather the profiling 111 information, and then run cg_annotate to get a detailed presentation of that 112 information. As an optional intermediate step, you can use cg_merge to sum 113 together the outputs of multiple Cachegrind runs into a single file which 114 you then use as the input for cg_annotate. Alternatively, you can use 115 cg_diff to difference the outputs of two Cachegrind runs into a single file 116 which you then use as the input for cg_annotate.</para> 117 118 119 <sect2 id="cg-manual.running-cachegrind" xreflabel="Running Cachegrind"> 120 <title>Running Cachegrind</title> 121 122 <para>To run Cachegrind on a program <filename>prog</filename>, run:</para> 123 <screen><![CDATA[ 124 valgrind --tool=cachegrind prog 125 ]]></screen> 126 127 <para>The program will execute (slowly). Upon completion, 128 summary statistics that look like this will be printed:</para> 129 130 <programlisting><![CDATA[ 131 ==31751== I refs: 27,742,716 132 ==31751== I1 misses: 276 133 ==31751== LLi misses: 275 134 ==31751== I1 miss rate: 0.0% 135 ==31751== LLi miss rate: 0.0% 136 ==31751== 137 ==31751== D refs: 15,430,290 (10,955,517 rd + 4,474,773 wr) 138 ==31751== D1 misses: 41,185 ( 21,905 rd + 19,280 wr) 139 ==31751== LLd misses: 23,085 ( 3,987 rd + 19,098 wr) 140 ==31751== D1 miss rate: 0.2% ( 0.1% + 0.4%) 141 ==31751== LLd miss rate: 0.1% ( 0.0% + 0.4%) 142 ==31751== 143 ==31751== LL misses: 23,360 ( 4,262 rd + 19,098 wr) 144 ==31751== LL miss rate: 0.0% ( 0.0% + 0.4%)]]></programlisting> 145 146 <para>Cache accesses for instruction fetches are summarised 147 first, giving the number of fetches made (this is the number of 148 instructions executed, which can be useful to know in its own 149 right), the number of I1 misses, and the number of LL instruction 150 (<computeroutput>LLi</computeroutput>) misses.</para> 151 152 <para>Cache accesses for data follow. The information is similar 153 to that of the instruction fetches, except that the values are 154 also shown split between reads and writes (note each row's 155 <computeroutput>rd</computeroutput> and 156 <computeroutput>wr</computeroutput> values add up to the row's 157 total).</para> 158 159 <para>Combined instruction and data figures for the LL cache 160 follow that. Note that the LL miss rate is computed relative to the total 161 number of memory accesses, not the number of L1 misses. I.e. it is 162 <computeroutput>(ILmr + DLmr + DLmw) / (Ir + Dr + Dw)</computeroutput> 163 not 164 <computeroutput>(ILmr + DLmr + DLmw) / (I1mr + D1mr + D1mw)</computeroutput> 165 </para> 166 167 <para>Branch prediction statistics are not collected by default. 168 To do so, add the option <option>--branch-sim=yes</option>.</para> 169 170 </sect2> 171 172 173 <sect2 id="cg-manual.outputfile" xreflabel="Output File"> 174 <title>Output File</title> 175 176 <para>As well as printing summary information, Cachegrind also writes 177 more detailed profiling information to a file. By default this file is named 178 <filename>cachegrind.out.<pid></filename> (where 179 <filename><pid></filename> is the program's process ID), but its name 180 can be changed with the <option>--cachegrind-out-file</option> option. This 181 file is human-readable, but is intended to be interpreted by the 182 accompanying program cg_annotate, described in the next section.</para> 183 184 <para>The default <computeroutput>.<pid></computeroutput> suffix 185 on the output file name serves two purposes. Firstly, it means you 186 don't have to rename old log files that you don't want to overwrite. 187 Secondly, and more importantly, it allows correct profiling with the 188 <option>--trace-children=yes</option> option of 189 programs that spawn child processes.</para> 190 191 <para>The output file can be big, many megabytes for large applications 192 built with full debugging information.</para> 193 194 </sect2> 195 196 197 198 <sect2 id="cg-manual.running-cg_annotate" xreflabel="Running cg_annotate"> 199 <title>Running cg_annotate</title> 200 201 <para>Before using cg_annotate, 202 it is worth widening your window to be at least 120-characters 203 wide if possible, as the output lines can be quite long.</para> 204 205 <para>To get a function-by-function summary, run:</para> 206 207 <screen>cg_annotate <filename></screen> 208 209 <para>on a Cachegrind output file.</para> 210 211 </sect2> 212 213 214 <sect2 id="cg-manual.the-output-preamble" xreflabel="The Output Preamble"> 215 <title>The Output Preamble</title> 216 217 <para>The first part of the output looks like this:</para> 218 219 <programlisting><![CDATA[ 220 -------------------------------------------------------------------------------- 221 I1 cache: 65536 B, 64 B, 2-way associative 222 D1 cache: 65536 B, 64 B, 2-way associative 223 LL cache: 262144 B, 64 B, 8-way associative 224 Command: concord vg_to_ucode.c 225 Events recorded: Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 226 Events shown: Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 227 Event sort order: Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 228 Threshold: 99% 229 Chosen for annotation: 230 Auto-annotation: off 231 ]]></programlisting> 232 233 234 <para>This is a summary of the annotation options:</para> 235 236 <itemizedlist> 237 238 <listitem> 239 <para>I1 cache, D1 cache, LL cache: cache configuration. So 240 you know the configuration with which these results were 241 obtained.</para> 242 </listitem> 243 244 <listitem> 245 <para>Command: the command line invocation of the program 246 under examination.</para> 247 </listitem> 248 249 <listitem> 250 <para>Events recorded: which events were recorded.</para> 251 252 </listitem> 253 254 <listitem> 255 <para>Events shown: the events shown, which is a subset of the events 256 gathered. This can be adjusted with the 257 <option>--show</option> option.</para> 258 </listitem> 259 260 <listitem> 261 <para>Event sort order: the sort order in which functions are 262 shown. For example, in this case the functions are sorted 263 from highest <computeroutput>Ir</computeroutput> counts to 264 lowest. If two functions have identical 265 <computeroutput>Ir</computeroutput> counts, they will then be 266 sorted by <computeroutput>I1mr</computeroutput> counts, and 267 so on. This order can be adjusted with the 268 <option>--sort</option> option.</para> 269 270 <para>Note that this dictates the order the functions appear. 271 It is <emphasis>not</emphasis> the order in which the columns 272 appear; that is dictated by the "events shown" line (and can 273 be changed with the <option>--show</option> 274 option).</para> 275 </listitem> 276 277 <listitem> 278 <para>Threshold: cg_annotate 279 by default omits functions that cause very low counts 280 to avoid drowning you in information. In this case, 281 cg_annotate shows summaries the functions that account for 282 99% of the <computeroutput>Ir</computeroutput> counts; 283 <computeroutput>Ir</computeroutput> is chosen as the 284 threshold event since it is the primary sort event. The 285 threshold can be adjusted with the 286 <option>--threshold</option> 287 option.</para> 288 </listitem> 289 290 <listitem> 291 <para>Chosen for annotation: names of files specified 292 manually for annotation; in this case none.</para> 293 </listitem> 294 295 <listitem> 296 <para>Auto-annotation: whether auto-annotation was requested 297 via the <option>--auto=yes</option> 298 option. In this case no.</para> 299 </listitem> 300 301 </itemizedlist> 302 303 </sect2> 304 305 306 <sect2 id="cg-manual.the-global" 307 xreflabel="The Global and Function-level Counts"> 308 <title>The Global and Function-level Counts</title> 309 310 <para>Then follows summary statistics for the whole 311 program:</para> 312 313 <programlisting><![CDATA[ 314 -------------------------------------------------------------------------------- 315 Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 316 -------------------------------------------------------------------------------- 317 27,742,716 276 275 10,955,517 21,905 3,987 4,474,773 19,280 19,098 PROGRAM TOTALS]]></programlisting> 318 319 <para> 320 These are similar to the summary provided when Cachegrind finishes running. 321 </para> 322 323 <para>Then comes function-by-function statistics:</para> 324 325 <programlisting><![CDATA[ 326 -------------------------------------------------------------------------------- 327 Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw file:function 328 -------------------------------------------------------------------------------- 329 8,821,482 5 5 2,242,702 1,621 73 1,794,230 0 0 getc.c:_IO_getc 330 5,222,023 4 4 2,276,334 16 12 875,959 1 1 concord.c:get_word 331 2,649,248 2 2 1,344,810 7,326 1,385 . . . vg_main.c:strcmp 332 2,521,927 2 2 591,215 0 0 179,398 0 0 concord.c:hash 333 2,242,740 2 2 1,046,612 568 22 448,548 0 0 ctype.c:tolower 334 1,496,937 4 4 630,874 9,000 1,400 279,388 0 0 concord.c:insert 335 897,991 51 51 897,831 95 30 62 1 1 ???:??? 336 598,068 1 1 299,034 0 0 149,517 0 0 ../sysdeps/generic/lockfile.c:__flockfile 337 598,068 0 0 299,034 0 0 149,517 0 0 ../sysdeps/generic/lockfile.c:__funlockfile 338 598,024 4 4 213,580 35 16 149,506 0 0 vg_clientmalloc.c:malloc 339 446,587 1 1 215,973 2,167 430 129,948 14,057 13,957 concord.c:add_existing 340 341,760 2 2 128,160 0 0 128,160 0 0 vg_clientmalloc.c:vg_trap_here_WRAPPER 341 320,782 4 4 150,711 276 0 56,027 53 53 concord.c:init_hash_table 342 298,998 1 1 106,785 0 0 64,071 1 1 concord.c:create 343 149,518 0 0 149,516 0 0 1 0 0 ???:tolower@@GLIBC_2.0 344 149,518 0 0 149,516 0 0 1 0 0 ???:fgetc@@GLIBC_2.0 345 95,983 4 4 38,031 0 0 34,409 3,152 3,150 concord.c:new_word_node 346 85,440 0 0 42,720 0 0 21,360 0 0 vg_clientmalloc.c:vg_bogus_epilogue]]></programlisting> 347 348 <para>Each function 349 is identified by a 350 <computeroutput>file_name:function_name</computeroutput> pair. If 351 a column contains only a dot it means the function never performs 352 that event (e.g. the third row shows that 353 <computeroutput>strcmp()</computeroutput> contains no 354 instructions that write to memory). The name 355 <computeroutput>???</computeroutput> is used if the file name 356 and/or function name could not be determined from debugging 357 information. If most of the entries have the form 358 <computeroutput>???:???</computeroutput> the program probably 359 wasn't compiled with <option>-g</option>.</para> 360 361 <para>It is worth noting that functions will come both from 362 the profiled program (e.g. <filename>concord.c</filename>) 363 and from libraries (e.g. <filename>getc.c</filename>)</para> 364 365 </sect2> 366 367 368 <sect2 id="cg-manual.line-by-line" xreflabel="Line-by-line Counts"> 369 <title>Line-by-line Counts</title> 370 371 <para>There are two ways to annotate source files -- by specifying them 372 manually as arguments to cg_annotate, or with the 373 <option>--auto=yes</option> option. For example, the output from running 374 <filename>cg_annotate <filename> concord.c</filename> for our example 375 produces the same output as above followed by an annotated version of 376 <filename>concord.c</filename>, a section of which looks like:</para> 377 378 <programlisting><![CDATA[ 379 -------------------------------------------------------------------------------- 380 -- User-annotated source: concord.c 381 -------------------------------------------------------------------------------- 382 Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 383 384 . . . . . . . . . void init_hash_table(char *file_name, Word_Node *table[]) 385 3 1 1 . . . 1 0 0 { 386 . . . . . . . . . FILE *file_ptr; 387 . . . . . . . . . Word_Info *data; 388 1 0 0 . . . 1 1 1 int line = 1, i; 389 . . . . . . . . . 390 5 0 0 . . . 3 0 0 data = (Word_Info *) create(sizeof(Word_Info)); 391 . . . . . . . . . 392 4,991 0 0 1,995 0 0 998 0 0 for (i = 0; i < TABLE_SIZE; i++) 393 3,988 1 1 1,994 0 0 997 53 52 table[i] = NULL; 394 . . . . . . . . . 395 . . . . . . . . . /* Open file, check it. */ 396 6 0 0 1 0 0 4 0 0 file_ptr = fopen(file_name, "r"); 397 2 0 0 1 0 0 . . . if (!(file_ptr)) { 398 . . . . . . . . . fprintf(stderr, "Couldn't open '%s'.\n", file_name); 399 1 1 1 . . . . . . exit(EXIT_FAILURE); 400 . . . . . . . . . } 401 . . . . . . . . . 402 165,062 1 1 73,360 0 0 91,700 0 0 while ((line = get_word(data, line, file_ptr)) != EOF) 403 146,712 0 0 73,356 0 0 73,356 0 0 insert(data->;word, data->line, table); 404 . . . . . . . . . 405 4 0 0 1 0 0 2 0 0 free(data); 406 4 0 0 1 0 0 2 0 0 fclose(file_ptr); 407 3 0 0 2 0 0 . . . }]]></programlisting> 408 409 <para>(Although column widths are automatically minimised, a wide 410 terminal is clearly useful.)</para> 411 412 <para>Each source file is clearly marked 413 (<computeroutput>User-annotated source</computeroutput>) as 414 having been chosen manually for annotation. If the file was 415 found in one of the directories specified with the 416 <option>-I</option>/<option>--include</option> option, the directory 417 and file are both given.</para> 418 419 <para>Each line is annotated with its event counts. Events not 420 applicable for a line are represented by a dot. This is useful 421 for distinguishing between an event which cannot happen, and one 422 which can but did not.</para> 423 424 <para>Sometimes only a small section of a source file is 425 executed. To minimise uninteresting output, Cachegrind only shows 426 annotated lines and lines within a small distance of annotated 427 lines. Gaps are marked with the line numbers so you know which 428 part of a file the shown code comes from, eg:</para> 429 430 <programlisting><![CDATA[ 431 (figures and code for line 704) 432 -- line 704 ---------------------------------------- 433 -- line 878 ---------------------------------------- 434 (figures and code for line 878)]]></programlisting> 435 436 <para>The amount of context to show around annotated lines is 437 controlled by the <option>--context</option> 438 option.</para> 439 440 <para>To get automatic annotation, use the <option>--auto=yes</option> option. 441 cg_annotate will automatically annotate every source file it can 442 find that is mentioned in the function-by-function summary. 443 Therefore, the files chosen for auto-annotation are affected by 444 the <option>--sort</option> and 445 <option>--threshold</option> options. Each 446 source file is clearly marked (<computeroutput>Auto-annotated 447 source</computeroutput>) as being chosen automatically. Any 448 files that could not be found are mentioned at the end of the 449 output, eg:</para> 450 451 <programlisting><![CDATA[ 452 ------------------------------------------------------------------ 453 The following files chosen for auto-annotation could not be found: 454 ------------------------------------------------------------------ 455 getc.c 456 ctype.c 457 ../sysdeps/generic/lockfile.c]]></programlisting> 458 459 <para>This is quite common for library files, since libraries are 460 usually compiled with debugging information, but the source files 461 are often not present on a system. If a file is chosen for 462 annotation both manually and automatically, it 463 is marked as <computeroutput>User-annotated 464 source</computeroutput>. Use the 465 <option>-I</option>/<option>--include</option> option to tell Valgrind where 466 to look for source files if the filenames found from the debugging 467 information aren't specific enough.</para> 468 469 <para>Beware that cg_annotate can take some time to digest large 470 <filename>cachegrind.out.<pid></filename> files, 471 e.g. 30 seconds or more. Also beware that auto-annotation can 472 produce a lot of output if your program is large!</para> 473 474 </sect2> 475 476 477 <sect2 id="cg-manual.assembler" xreflabel="Annotating Assembly Code Programs"> 478 <title>Annotating Assembly Code Programs</title> 479 480 <para>Valgrind can annotate assembly code programs too, or annotate 481 the assembly code generated for your C program. Sometimes this is 482 useful for understanding what is really happening when an 483 interesting line of C code is translated into multiple 484 instructions.</para> 485 486 <para>To do this, you just need to assemble your 487 <computeroutput>.s</computeroutput> files with assembly-level debug 488 information. You can use compile with the <option>-S</option> to compile C/C++ 489 programs to assembly code, and then assemble the assembly code files with 490 <option>-g</option> to achieve this. You can then profile and annotate the 491 assembly code source files in the same way as C/C++ source files.</para> 492 493 </sect2> 494 495 <sect2 id="ms-manual.forkingprograms" xreflabel="Forking Programs"> 496 <title>Forking Programs</title> 497 <para>If your program forks, the child will inherit all the profiling data that 498 has been gathered for the parent.</para> 499 500 <para>If the output file format string (controlled by 501 <option>--cachegrind-out-file</option>) does not contain <option>%p</option>, 502 then the outputs from the parent and child will be intermingled in a single 503 output file, which will almost certainly make it unreadable by 504 cg_annotate.</para> 505 </sect2> 506 507 508 <sect2 id="cg-manual.annopts.warnings" xreflabel="cg_annotate Warnings"> 509 <title>cg_annotate Warnings</title> 510 511 <para>There are a couple of situations in which 512 cg_annotate issues warnings.</para> 513 514 <itemizedlist> 515 <listitem> 516 <para>If a source file is more recent than the 517 <filename>cachegrind.out.<pid></filename> file. 518 This is because the information in 519 <filename>cachegrind.out.<pid></filename> is only 520 recorded with line numbers, so if the line numbers change at 521 all in the source (e.g. lines added, deleted, swapped), any 522 annotations will be incorrect.</para> 523 </listitem> 524 <listitem> 525 <para>If information is recorded about line numbers past the 526 end of a file. This can be caused by the above problem, 527 i.e. shortening the source file while using an old 528 <filename>cachegrind.out.<pid></filename> file. If 529 this happens, the figures for the bogus lines are printed 530 anyway (clearly marked as bogus) in case they are 531 important.</para> 532 </listitem> 533 </itemizedlist> 534 535 </sect2> 536 537 538 539 <sect2 id="cg-manual.annopts.things-to-watch-out-for" 540 xreflabel="Unusual Annotation Cases"> 541 <title>Unusual Annotation Cases</title> 542 543 <para>Some odd things that can occur during annotation:</para> 544 545 <itemizedlist> 546 <listitem> 547 <para>If annotating at the assembler level, you might see 548 something like this:</para> 549 <programlisting><![CDATA[ 550 1 0 0 . . . . . . leal -12(%ebp),%eax 551 1 0 0 . . . 1 0 0 movl %eax,84(%ebx) 552 2 0 0 0 0 0 1 0 0 movl $1,-20(%ebp) 553 . . . . . . . . . .align 4,0x90 554 1 0 0 . . . . . . movl $.LnrB,%eax 555 1 0 0 . . . 1 0 0 movl %eax,-16(%ebp)]]></programlisting> 556 557 <para>How can the third instruction be executed twice when 558 the others are executed only once? As it turns out, it 559 isn't. Here's a dump of the executable, using 560 <computeroutput>objdump -d</computeroutput>:</para> 561 <programlisting><![CDATA[ 562 8048f25: 8d 45 f4 lea 0xfffffff4(%ebp),%eax 563 8048f28: 89 43 54 mov %eax,0x54(%ebx) 564 8048f2b: c7 45 ec 01 00 00 00 movl $0x1,0xffffffec(%ebp) 565 8048f32: 89 f6 mov %esi,%esi 566 8048f34: b8 08 8b 07 08 mov $0x8078b08,%eax 567 8048f39: 89 45 f0 mov %eax,0xfffffff0(%ebp)]]></programlisting> 568 569 <para>Notice the extra <computeroutput>mov 570 %esi,%esi</computeroutput> instruction. Where did this come 571 from? The GNU assembler inserted it to serve as the two 572 bytes of padding needed to align the <computeroutput>movl 573 $.LnrB,%eax</computeroutput> instruction on a four-byte 574 boundary, but pretended it didn't exist when adding debug 575 information. Thus when Valgrind reads the debug info it 576 thinks that the <computeroutput>movl 577 $0x1,0xffffffec(%ebp)</computeroutput> instruction covers the 578 address range 0x8048f2b--0x804833 by itself, and attributes 579 the counts for the <computeroutput>mov 580 %esi,%esi</computeroutput> to it.</para> 581 </listitem> 582 583 <!-- 584 I think this isn't true any more, not since cost centres were moved from 585 being associated with instruction addresses to being associated with 586 source line numbers. 587 <listitem> 588 <para>Inlined functions can cause strange results in the 589 function-by-function summary. If a function 590 <computeroutput>inline_me()</computeroutput> is defined in 591 <filename>foo.h</filename> and inlined in the functions 592 <computeroutput>f1()</computeroutput>, 593 <computeroutput>f2()</computeroutput> and 594 <computeroutput>f3()</computeroutput> in 595 <filename>bar.c</filename>, there will not be a 596 <computeroutput>foo.h:inline_me()</computeroutput> function 597 entry. Instead, there will be separate function entries for 598 each inlining site, i.e. 599 <computeroutput>foo.h:f1()</computeroutput>, 600 <computeroutput>foo.h:f2()</computeroutput> and 601 <computeroutput>foo.h:f3()</computeroutput>. To find the 602 total counts for 603 <computeroutput>foo.h:inline_me()</computeroutput>, add up 604 the counts from each entry.</para> 605 606 <para>The reason for this is that although the debug info 607 output by GCC indicates the switch from 608 <filename>bar.c</filename> to <filename>foo.h</filename>, it 609 doesn't indicate the name of the function in 610 <filename>foo.h</filename>, so Valgrind keeps using the old 611 one.</para> 612 </listitem> 613 --> 614 615 <listitem> 616 <para>Sometimes, the same filename might be represented with 617 a relative name and with an absolute name in different parts 618 of the debug info, eg: 619 <filename>/home/user/proj/proj.h</filename> and 620 <filename>../proj.h</filename>. In this case, if you use 621 auto-annotation, the file will be annotated twice with the 622 counts split between the two.</para> 623 </listitem> 624 625 <listitem> 626 <para>If you compile some files with 627 <option>-g</option> and some without, some 628 events that take place in a file without debug info could be 629 attributed to the last line of a file with debug info 630 (whichever one gets placed before the non-debug-info file in 631 the executable).</para> 632 </listitem> 633 634 </itemizedlist> 635 636 <para>This list looks long, but these cases should be fairly 637 rare.</para> 638 639 </sect2> 640 641 642 <sect2 id="cg-manual.cg_merge" xreflabel="cg_merge"> 643 <title>Merging Profiles with cg_merge</title> 644 645 <para> 646 cg_merge is a simple program which 647 reads multiple profile files, as created by Cachegrind, merges them 648 together, and writes the results into another file in the same format. 649 You can then examine the merged results using 650 <computeroutput>cg_annotate <filename></computeroutput>, as 651 described above. The merging functionality might be useful if you 652 want to aggregate costs over multiple runs of the same program, or 653 from a single parallel run with multiple instances of the same 654 program.</para> 655 656 <para> 657 cg_merge is invoked as follows: 658 </para> 659 660 <programlisting><![CDATA[ 661 cg_merge -o outputfile file1 file2 file3 ...]]></programlisting> 662 663 <para> 664 It reads and checks <computeroutput>file1</computeroutput>, then read 665 and checks <computeroutput>file2</computeroutput> and merges it into 666 the running totals, then the same with 667 <computeroutput>file3</computeroutput>, etc. The final results are 668 written to <computeroutput>outputfile</computeroutput>, or to standard 669 out if no output file is specified.</para> 670 671 <para> 672 Costs are summed on a per-function, per-line and per-instruction 673 basis. Because of this, the order in which the input files does not 674 matter, although you should take care to only mention each file once, 675 since any file mentioned twice will be added in twice.</para> 676 677 <para> 678 cg_merge does not attempt to check 679 that the input files come from runs of the same executable. It will 680 happily merge together profile files from completely unrelated 681 programs. It does however check that the 682 <computeroutput>Events:</computeroutput> lines of all the inputs are 683 identical, so as to ensure that the addition of costs makes sense. 684 For example, it would be nonsensical for it to add a number indicating 685 D1 read references to a number from a different file indicating LL 686 write misses.</para> 687 688 <para> 689 A number of other syntax and sanity checks are done whilst reading the 690 inputs. cg_merge will stop and 691 attempt to print a helpful error message if any of the input files 692 fail these checks.</para> 693 694 </sect2> 695 696 697 <sect2 id="cg-manual.cg_diff" xreflabel="cg_diff"> 698 <title>Differencing Profiles with cg_diff</title> 699 700 <para> 701 cg_diff is a simple program which 702 reads two profile files, as created by Cachegrind, finds the difference 703 between them, and writes the results into another file in the same format. 704 You can then examine the merged results using 705 <computeroutput>cg_annotate <filename></computeroutput>, as 706 described above. This is very useful if you want to measure how a change to 707 a program affected its performance. 708 </para> 709 710 <para> 711 cg_diff is invoked as follows: 712 </para> 713 714 <programlisting><![CDATA[ 715 cg_diff file1 file2]]></programlisting> 716 717 <para> 718 It reads and checks <computeroutput>file1</computeroutput>, then read 719 and checks <computeroutput>file2</computeroutput>, then computes the 720 difference (effectively <computeroutput>file1</computeroutput> - 721 <computeroutput>file2</computeroutput>). The final results are written to 722 standard output.</para> 723 724 <para> 725 Costs are summed on a per-function basis. Per-line costs are not summed, 726 because doing so is too difficult. For example, consider differencing two 727 profiles, one from a single-file program A, and one from the same program A 728 where a single blank line was inserted at the top of the file. Every single 729 per-line count has changed. In comparison, the per-function counts have not 730 changed. The per-function count differences are still very useful for 731 determining differences between programs. Note that because the result is 732 the difference of two profiles, many of the counts will be negative; this 733 indicates that the counts for the relevant function are fewer in the second 734 version than those in the first version.</para> 735 736 <para> 737 cg_diff does not attempt to check 738 that the input files come from runs of the same executable. It will 739 happily merge together profile files from completely unrelated 740 programs. It does however check that the 741 <computeroutput>Events:</computeroutput> lines of all the inputs are 742 identical, so as to ensure that the addition of costs makes sense. 743 For example, it would be nonsensical for it to add a number indicating 744 D1 read references to a number from a different file indicating LL 745 write misses.</para> 746 747 <para> 748 A number of other syntax and sanity checks are done whilst reading the 749 inputs. cg_diff will stop and 750 attempt to print a helpful error message if any of the input files 751 fail these checks.</para> 752 753 <para> 754 Sometimes you will want to compare Cachegrind profiles of two versions of a 755 program that you have sitting side-by-side. For example, you might have 756 <computeroutput>version1/prog.c</computeroutput> and 757 <computeroutput>version2/prog.c</computeroutput>, where the second is 758 slightly different to the first. A straight comparison of the two will not 759 be useful -- because functions are qualified with filenames, a function 760 <function>f</function> will be listed as 761 <computeroutput>version1/prog.c:f</computeroutput> for the first version but 762 <computeroutput>version2/prog.c:f</computeroutput> for the second 763 version.</para> 764 765 <para> 766 When this happens, you can use the <option>--mod-filename</option> option. 767 Its argument is a Perl search-and-replace expression that will be applied 768 to all the filenames in both Cachegrind output files. It can be used to 769 remove minor differences in filenames. For example, the option 770 <option>--mod-filename='s/version[0-9]/versionN/'</option> will suffice for 771 this case.</para> 772 773 <para> 774 Similarly, sometimes compilers auto-generate certain functions and give them 775 randomized names. For example, GCC sometimes auto-generates functions with 776 names like <function>T.1234</function>, and the suffixes vary from build to 777 build. You can use the <option>--mod-funcname</option> option to remove 778 small differences like these; it works in the same way as 779 <option>--mod-filename</option>.</para> 780 781 </sect2> 782 783 784 </sect1> 785 786 787 788 <sect1 id="cg-manual.cgopts" xreflabel="Cachegrind Command-line Options"> 789 <title>Cachegrind Command-line Options</title> 790 791 <!-- start of xi:include in the manpage --> 792 <para>Cachegrind-specific options are:</para> 793 794 <variablelist id="cg.opts.list"> 795 796 <varlistentry id="opt.I1" xreflabel="--I1"> 797 <term> 798 <option><![CDATA[--I1=<size>,<associativity>,<line size> ]]></option> 799 </term> 800 <listitem> 801 <para>Specify the size, associativity and line size of the level 1 802 instruction cache. </para> 803 </listitem> 804 </varlistentry> 805 806 <varlistentry id="opt.D1" xreflabel="--D1"> 807 <term> 808 <option><![CDATA[--D1=<size>,<associativity>,<line size> ]]></option> 809 </term> 810 <listitem> 811 <para>Specify the size, associativity and line size of the level 1 812 data cache.</para> 813 </listitem> 814 </varlistentry> 815 816 <varlistentry id="opt.LL" xreflabel="--LL"> 817 <term> 818 <option><![CDATA[--LL=<size>,<associativity>,<line size> ]]></option> 819 </term> 820 <listitem> 821 <para>Specify the size, associativity and line size of the last-level 822 cache.</para> 823 </listitem> 824 </varlistentry> 825 826 <varlistentry id="opt.cache-sim" xreflabel="--cache-sim"> 827 <term> 828 <option><![CDATA[--cache-sim=no|yes [yes] ]]></option> 829 </term> 830 <listitem> 831 <para>Enables or disables collection of cache access and miss 832 counts.</para> 833 </listitem> 834 </varlistentry> 835 836 <varlistentry id="opt.branch-sim" xreflabel="--branch-sim"> 837 <term> 838 <option><![CDATA[--branch-sim=no|yes [no] ]]></option> 839 </term> 840 <listitem> 841 <para>Enables or disables collection of branch instruction and 842 misprediction counts. By default this is disabled as it 843 slows Cachegrind down by approximately 25%. Note that you 844 cannot specify <option>--cache-sim=no</option> 845 and <option>--branch-sim=no</option> 846 together, as that would leave Cachegrind with no 847 information to collect.</para> 848 </listitem> 849 </varlistentry> 850 851 <varlistentry id="opt.cachegrind-out-file" xreflabel="--cachegrind-out-file"> 852 <term> 853 <option><![CDATA[--cachegrind-out-file=<file> ]]></option> 854 </term> 855 <listitem> 856 <para>Write the profile data to 857 <computeroutput>file</computeroutput> rather than to the default 858 output file, 859 <filename>cachegrind.out.<pid></filename>. The 860 <option>%p</option> and <option>%q</option> format specifiers 861 can be used to embed the process ID and/or the contents of an 862 environment variable in the name, as is the case for the core 863 option <option><xref linkend="opt.log-file"/></option>. 864 </para> 865 </listitem> 866 </varlistentry> 867 868 </variablelist> 869 <!-- end of xi:include in the manpage --> 870 871 </sect1> 872 873 874 875 <sect1 id="cg-manual.annopts" xreflabel="cg_annotate Command-line Options"> 876 <title>cg_annotate Command-line Options</title> 877 878 <!-- start of xi:include in the manpage --> 879 <variablelist id="cg_annotate.opts.list"> 880 881 <varlistentry> 882 <term> 883 <option><![CDATA[-h --help ]]></option> 884 </term> 885 <listitem> 886 <para>Show the help message.</para> 887 </listitem> 888 </varlistentry> 889 890 <varlistentry> 891 <term> 892 <option><![CDATA[--version ]]></option> 893 </term> 894 <listitem> 895 <para>Show the version number.</para> 896 </listitem> 897 </varlistentry> 898 899 <varlistentry> 900 <term> 901 <option><![CDATA[--show=A,B,C [default: all, using order in 902 cachegrind.out.<pid>] ]]></option> 903 </term> 904 <listitem> 905 <para>Specifies which events to show (and the column 906 order). Default is to use all present in the 907 <filename>cachegrind.out.<pid></filename> file (and 908 use the order in the file). Useful if you want to concentrate on, for 909 example, I cache misses (<option>--show=I1mr,ILmr</option>), or data 910 read misses (<option>--show=D1mr,DLmr</option>), or LL data misses 911 (<option>--show=DLmr,DLmw</option>). Best used in conjunction with 912 <option>--sort</option>.</para> 913 </listitem> 914 </varlistentry> 915 916 <varlistentry> 917 <term> 918 <option><![CDATA[--sort=A,B,C [default: order in 919 cachegrind.out.<pid>] ]]></option> 920 </term> 921 <listitem> 922 <para>Specifies the events upon which the sorting of the 923 function-by-function entries will be based.</para> 924 </listitem> 925 </varlistentry> 926 927 <varlistentry> 928 <term> 929 <option><![CDATA[--threshold=X [default: 0.1%] ]]></option> 930 </term> 931 <listitem> 932 <para>Sets the threshold for the function-by-function 933 summary. A function is shown if it accounts for more than X% 934 of the counts for the primary sort event. If auto-annotating, also 935 affects which files are annotated.</para> 936 937 <para>Note: thresholds can be set for more than one of the 938 events by appending any events for the 939 <option>--sort</option> option with a colon 940 and a number (no spaces, though). E.g. if you want to see 941 each function that covers more than 1% of LL read misses or 1% of LL 942 write misses, use this option:</para> 943 <para><option>--sort=DLmr:1,DLmw:1</option></para> 944 </listitem> 945 </varlistentry> 946 947 <varlistentry> 948 <term> 949 <option><![CDATA[--auto=<no|yes> [default: no] ]]></option> 950 </term> 951 <listitem> 952 <para>When enabled, automatically annotates every file that 953 is mentioned in the function-by-function summary that can be 954 found. Also gives a list of those that couldn't be found.</para> 955 </listitem> 956 </varlistentry> 957 958 <varlistentry> 959 <term> 960 <option><![CDATA[--context=N [default: 8] ]]></option> 961 </term> 962 <listitem> 963 <para>Print N lines of context before and after each 964 annotated line. Avoids printing large sections of source 965 files that were not executed. Use a large number 966 (e.g. 100000) to show all source lines.</para> 967 </listitem> 968 </varlistentry> 969 970 <varlistentry> 971 <term> 972 <option><![CDATA[-I<dir> --include=<dir> [default: none] ]]></option> 973 </term> 974 <listitem> 975 <para>Adds a directory to the list in which to search for 976 files. Multiple <option>-I</option>/<option>--include</option> 977 options can be given to add multiple directories.</para> 978 </listitem> 979 </varlistentry> 980 981 </variablelist> 982 <!-- end of xi:include in the manpage --> 983 984 </sect1> 985 986 987 <sect1 id="cg-manual.mergeopts" xreflabel="cg_merge Command-line Options"> 988 <title>cg_merge Command-line Options</title> 989 990 <!-- start of xi:include in the manpage --> 991 <variablelist id="cg_merge.opts.list"> 992 993 <varlistentry> 994 <term> 995 <option><![CDATA[-o outfile]]></option> 996 </term> 997 <listitem> 998 <para>Write the profile data to <computeroutput>outfile</computeroutput> 999 rather than to standard output. 1000 </para> 1001 </listitem> 1002 </varlistentry> 1003 1004 </variablelist> 1005 <!-- end of xi:include in the manpage --> 1006 1007 </sect1> 1008 1009 1010 <sect1 id="cg-manual.diffopts" xreflabel="cg_diff Command-line Options"> 1011 <title>cg_diff Command-line Options</title> 1012 1013 <!-- start of xi:include in the manpage --> 1014 <variablelist id="cg_diff.opts.list"> 1015 1016 <varlistentry> 1017 <term> 1018 <option><![CDATA[-h --help ]]></option> 1019 </term> 1020 <listitem> 1021 <para>Show the help message.</para> 1022 </listitem> 1023 </varlistentry> 1024 1025 <varlistentry> 1026 <term> 1027 <option><![CDATA[--version ]]></option> 1028 </term> 1029 <listitem> 1030 <para>Show the version number.</para> 1031 </listitem> 1032 </varlistentry> 1033 1034 <varlistentry> 1035 <term> 1036 <option><![CDATA[--mod-filename=<expr> [default: none]]]></option> 1037 </term> 1038 <listitem> 1039 <para>Specifies a Perl search-and-replace expression that is applied 1040 to all filenames. Useful for removing minor differences in paths 1041 between two different versions of a program that are sitting in 1042 different directories.</para> 1043 </listitem> 1044 </varlistentry> 1045 1046 <varlistentry> 1047 <term> 1048 <option><![CDATA[--mod-funcname=<expr> [default: none]]]></option> 1049 </term> 1050 <listitem> 1051 <para>Like <option>--mod-filename</option>, but for filenames. 1052 Useful for removing minor differences in randomized names of 1053 auto-generated functions generated by some compilers.</para> 1054 </listitem> 1055 </varlistentry> 1056 1057 </variablelist> 1058 <!-- end of xi:include in the manpage --> 1059 1060 </sect1> 1061 1062 1063 1064 1065 <sect1 id="cg-manual.acting-on" 1066 xreflabel="Acting on Cachegrind's Information"> 1067 <title>Acting on Cachegrind's Information</title> 1068 <para> 1069 Cachegrind gives you lots of information, but acting on that information 1070 isn't always easy. Here are some rules of thumb that we have found to be 1071 useful.</para> 1072 1073 <para> 1074 First of all, the global hit/miss counts and miss rates are not that useful. 1075 If you have multiple programs or multiple runs of a program, comparing the 1076 numbers might identify if any are outliers and worthy of closer 1077 investigation. Otherwise, they're not enough to act on.</para> 1078 1079 <para> 1080 The function-by-function counts are more useful to look at, as they pinpoint 1081 which functions are causing large numbers of counts. However, beware that 1082 inlining can make these counts misleading. If a function 1083 <function>f</function> is always inlined, counts will be attributed to the 1084 functions it is inlined into, rather than itself. However, if you look at 1085 the line-by-line annotations for <function>f</function> you'll see the 1086 counts that belong to <function>f</function>. (This is hard to avoid, it's 1087 how the debug info is structured.) So it's worth looking for large numbers 1088 in the line-by-line annotations.</para> 1089 1090 <para> 1091 The line-by-line source code annotations are much more useful. In our 1092 experience, the best place to start is by looking at the 1093 <computeroutput>Ir</computeroutput> numbers. They simply measure how many 1094 instructions were executed for each line, and don't include any cache 1095 information, but they can still be very useful for identifying 1096 bottlenecks.</para> 1097 1098 <para> 1099 After that, we have found that LL misses are typically a much bigger source 1100 of slow-downs than L1 misses. So it's worth looking for any snippets of 1101 code with high <computeroutput>DLmr</computeroutput> or 1102 <computeroutput>DLmw</computeroutput> counts. (You can use 1103 <option>--show=DLmr 1104 --sort=DLmr</option> with cg_annotate to focus just on 1105 <literal>DLmr</literal> counts, for example.) If you find any, it's still 1106 not always easy to work out how to improve things. You need to have a 1107 reasonable understanding of how caches work, the principles of locality, and 1108 your program's data access patterns. Improving things may require 1109 redesigning a data structure, for example.</para> 1110 1111 <para> 1112 Looking at the <computeroutput>Bcm</computeroutput> and 1113 <computeroutput>Bim</computeroutput> misses can also be helpful. 1114 In particular, <computeroutput>Bim</computeroutput> misses are often caused 1115 by <literal>switch</literal> statements, and in some cases these 1116 <literal>switch</literal> statements can be replaced with table-driven code. 1117 For example, you might replace code like this:</para> 1118 1119 <programlisting><![CDATA[ 1120 enum E { A, B, C }; 1121 enum E e; 1122 int i; 1123 ... 1124 switch (e) 1125 { 1126 case A: i += 1; break; 1127 case B: i += 2; break; 1128 case C: i += 3; break; 1129 } 1130 ]]></programlisting> 1131 1132 <para>with code like this:</para> 1133 1134 <programlisting><![CDATA[ 1135 enum E { A, B, C }; 1136 enum E e; 1137 enum E table[] = { 1, 2, 3 }; 1138 int i; 1139 ... 1140 i += table[e]; 1141 ]]></programlisting> 1142 1143 <para> 1144 This is obviously a contrived example, but the basic principle applies in a 1145 wide variety of situations.</para> 1146 1147 <para> 1148 In short, Cachegrind can tell you where some of the bottlenecks in your code 1149 are, but it can't tell you how to fix them. You have to work that out for 1150 yourself. But at least you have the information! 1151 </para> 1152 1153 </sect1> 1154 1155 1156 <sect1 id="cg-manual.sim-details" 1157 xreflabel="Simulation Details"> 1158 <title>Simulation Details</title> 1159 <para> 1160 This section talks about details you don't need to know about in order to 1161 use Cachegrind, but may be of interest to some people. 1162 </para> 1163 1164 <sect2 id="cache-sim" xreflabel="Cache Simulation Specifics"> 1165 <title>Cache Simulation Specifics</title> 1166 1167 <para>Specific characteristics of the cache simulation are as 1168 follows:</para> 1169 1170 <itemizedlist> 1171 1172 <listitem> 1173 <para>Write-allocate: when a write miss occurs, the block 1174 written to is brought into the D1 cache. Most modern caches 1175 have this property.</para> 1176 </listitem> 1177 1178 <listitem> 1179 <para>Bit-selection hash function: the set of line(s) in the cache 1180 to which a memory block maps is chosen by the middle bits 1181 M--(M+N-1) of the byte address, where:</para> 1182 <itemizedlist> 1183 <listitem> 1184 <para>line size = 2^M bytes</para> 1185 </listitem> 1186 <listitem> 1187 <para>(cache size / line size / associativity) = 2^N bytes</para> 1188 </listitem> 1189 </itemizedlist> 1190 </listitem> 1191 1192 <listitem> 1193 <para>Inclusive LL cache: the LL cache typically replicates all 1194 the entries of the L1 caches, because fetching into L1 involves 1195 fetching into LL first (this does not guarantee strict inclusiveness, 1196 as lines evicted from LL still could reside in L1). This is 1197 standard on Pentium chips, but AMD Opterons, Athlons and Durons 1198 use an exclusive LL cache that only holds 1199 blocks evicted from L1. Ditto most modern VIA CPUs.</para> 1200 </listitem> 1201 1202 </itemizedlist> 1203 1204 <para>The cache configuration simulated (cache size, 1205 associativity and line size) is determined automatically using 1206 the x86 CPUID instruction. If you have a machine that (a) 1207 doesn't support the CPUID instruction, or (b) supports it in an 1208 early incarnation that doesn't give any cache information, then 1209 Cachegrind will fall back to using a default configuration (that 1210 of a model 3/4 Athlon). Cachegrind will tell you if this 1211 happens. You can manually specify one, two or all three levels 1212 (I1/D1/LL) of the cache from the command line using the 1213 <option>--I1</option>, 1214 <option>--D1</option> and 1215 <option>--LL</option> options. 1216 For cache parameters to be valid for simulation, the number 1217 of sets (with associativity being the number of cache lines in 1218 each set) has to be a power of two.</para> 1219 1220 <para>On PowerPC platforms 1221 Cachegrind cannot automatically 1222 determine the cache configuration, so you will 1223 need to specify it with the 1224 <option>--I1</option>, 1225 <option>--D1</option> and 1226 <option>--LL</option> options.</para> 1227 1228 1229 <para>Other noteworthy behaviour:</para> 1230 1231 <itemizedlist> 1232 <listitem> 1233 <para>References that straddle two cache lines are treated as 1234 follows:</para> 1235 <itemizedlist> 1236 <listitem> 1237 <para>If both blocks hit --> counted as one hit</para> 1238 </listitem> 1239 <listitem> 1240 <para>If one block hits, the other misses --> counted 1241 as one miss.</para> 1242 </listitem> 1243 <listitem> 1244 <para>If both blocks miss --> counted as one miss (not 1245 two)</para> 1246 </listitem> 1247 </itemizedlist> 1248 </listitem> 1249 1250 <listitem> 1251 <para>Instructions that modify a memory location 1252 (e.g. <computeroutput>inc</computeroutput> and 1253 <computeroutput>dec</computeroutput>) are counted as doing 1254 just a read, i.e. a single data reference. This may seem 1255 strange, but since the write can never cause a miss (the read 1256 guarantees the block is in the cache) it's not very 1257 interesting.</para> 1258 1259 <para>Thus it measures not the number of times the data cache 1260 is accessed, but the number of times a data cache miss could 1261 occur.</para> 1262 </listitem> 1263 1264 </itemizedlist> 1265 1266 <para>If you are interested in simulating a cache with different 1267 properties, it is not particularly hard to write your own cache 1268 simulator, or to modify the existing ones in 1269 <computeroutput>cg_sim.c</computeroutput>. We'd be 1270 interested to hear from anyone who does.</para> 1271 1272 </sect2> 1273 1274 1275 <sect2 id="branch-sim" xreflabel="Branch Simulation Specifics"> 1276 <title>Branch Simulation Specifics</title> 1277 1278 <para>Cachegrind simulates branch predictors intended to be 1279 typical of mainstream desktop/server processors of around 2004.</para> 1280 1281 <para>Conditional branches are predicted using an array of 16384 2-bit 1282 saturating counters. The array index used for a branch instruction is 1283 computed partly from the low-order bits of the branch instruction's 1284 address and partly using the taken/not-taken behaviour of the last few 1285 conditional branches. As a result the predictions for any specific 1286 branch depend both on its own history and the behaviour of previous 1287 branches. This is a standard technique for improving prediction 1288 accuracy.</para> 1289 1290 <para>For indirect branches (that is, jumps to unknown destinations) 1291 Cachegrind uses a simple branch target address predictor. Targets are 1292 predicted using an array of 512 entries indexed by the low order 9 1293 bits of the branch instruction's address. Each branch is predicted to 1294 jump to the same address it did last time. Any other behaviour causes 1295 a mispredict.</para> 1296 1297 <para>More recent processors have better branch predictors, in 1298 particular better indirect branch predictors. Cachegrind's predictor 1299 design is deliberately conservative so as to be representative of the 1300 large installed base of processors which pre-date widespread 1301 deployment of more sophisticated indirect branch predictors. In 1302 particular, late model Pentium 4s (Prescott), Pentium M, Core and Core 1303 2 have more sophisticated indirect branch predictors than modelled by 1304 Cachegrind. </para> 1305 1306 <para>Cachegrind does not simulate a return stack predictor. It 1307 assumes that processors perfectly predict function return addresses, 1308 an assumption which is probably close to being true.</para> 1309 1310 <para>See Hennessy and Patterson's classic text "Computer 1311 Architecture: A Quantitative Approach", 4th edition (2007), Section 1312 2.3 (pages 80-89) for background on modern branch predictors.</para> 1313 1314 </sect2> 1315 1316 <sect2 id="cg-manual.annopts.accuracy" xreflabel="Accuracy"> 1317 <title>Accuracy</title> 1318 1319 <para>Valgrind's cache profiling has a number of 1320 shortcomings:</para> 1321 1322 <itemizedlist> 1323 <listitem> 1324 <para>It doesn't account for kernel activity -- the effect of system 1325 calls on the cache and branch predictor contents is ignored.</para> 1326 </listitem> 1327 1328 <listitem> 1329 <para>It doesn't account for other process activity. 1330 This is probably desirable when considering a single 1331 program.</para> 1332 </listitem> 1333 1334 <listitem> 1335 <para>It doesn't account for virtual-to-physical address 1336 mappings. Hence the simulation is not a true 1337 representation of what's happening in the 1338 cache. Most caches and branch predictors are physically indexed, but 1339 Cachegrind simulates caches using virtual addresses.</para> 1340 </listitem> 1341 1342 <listitem> 1343 <para>It doesn't account for cache misses not visible at the 1344 instruction level, e.g. those arising from TLB misses, or 1345 speculative execution.</para> 1346 </listitem> 1347 1348 <listitem> 1349 <para>Valgrind will schedule 1350 threads differently from how they would be when running natively. 1351 This could warp the results for threaded programs.</para> 1352 </listitem> 1353 1354 <listitem> 1355 <para>The x86/amd64 instructions <computeroutput>bts</computeroutput>, 1356 <computeroutput>btr</computeroutput> and 1357 <computeroutput>btc</computeroutput> will incorrectly be 1358 counted as doing a data read if both the arguments are 1359 registers, eg:</para> 1360 <programlisting><![CDATA[ 1361 btsl %eax, %edx]]></programlisting> 1362 1363 <para>This should only happen rarely.</para> 1364 </listitem> 1365 1366 <listitem> 1367 <para>x86/amd64 FPU instructions with data sizes of 28 and 108 bytes 1368 (e.g. <computeroutput>fsave</computeroutput>) are treated as 1369 though they only access 16 bytes. These instructions seem to 1370 be rare so hopefully this won't affect accuracy much.</para> 1371 </listitem> 1372 1373 </itemizedlist> 1374 1375 <para>Another thing worth noting is that results are very sensitive. 1376 Changing the size of the executable being profiled, or the sizes 1377 of any of the shared libraries it uses, or even the length of their 1378 file names, can perturb the results. Variations will be small, but 1379 don't expect perfectly repeatable results if your program changes at 1380 all.</para> 1381 1382 <para>More recent GNU/Linux distributions do address space 1383 randomisation, in which identical runs of the same program have their 1384 shared libraries loaded at different locations, as a security measure. 1385 This also perturbs the results.</para> 1386 1387 <para>While these factors mean you shouldn't trust the results to 1388 be super-accurate, they should be close enough to be useful.</para> 1389 1390 </sect2> 1391 1392 </sect1> 1393 1394 1395 1396 <sect1 id="cg-manual.impl-details" 1397 xreflabel="Implementation Details"> 1398 <title>Implementation Details</title> 1399 <para> 1400 This section talks about details you don't need to know about in order to 1401 use Cachegrind, but may be of interest to some people. 1402 </para> 1403 1404 <sect2 id="cg-manual.impl-details.how-cg-works" 1405 xreflabel="How Cachegrind Works"> 1406 <title>How Cachegrind Works</title> 1407 <para>The best reference for understanding how Cachegrind works is chapter 3 of 1408 "Dynamic Binary Analysis and Instrumentation", by Nicholas Nethercote. It 1409 is available on the <ulink url="&vg-pubs-url;">Valgrind publications 1410 page</ulink>.</para> 1411 </sect2> 1412 1413 <sect2 id="cg-manual.impl-details.file-format" 1414 xreflabel="Cachegrind Output File Format"> 1415 <title>Cachegrind Output File Format</title> 1416 <para>The file format is fairly straightforward, basically giving the 1417 cost centre for every line, grouped by files and 1418 functions. It's also totally generic and self-describing, in the sense that 1419 it can be used for any events that can be counted on a line-by-line basis, 1420 not just cache and branch predictor events. For example, earlier versions 1421 of Cachegrind didn't have a branch predictor simulation. When this was 1422 added, the file format didn't need to change at all. So the format (and 1423 consequently, cg_annotate) could be used by other tools.</para> 1424 1425 <para>The file format:</para> 1426 <programlisting><![CDATA[ 1427 file ::= desc_line* cmd_line events_line data_line+ summary_line 1428 desc_line ::= "desc:" ws? non_nl_string 1429 cmd_line ::= "cmd:" ws? cmd 1430 events_line ::= "events:" ws? (event ws)+ 1431 data_line ::= file_line | fn_line | count_line 1432 file_line ::= "fl=" filename 1433 fn_line ::= "fn=" fn_name 1434 count_line ::= line_num ws? (count ws)+ 1435 summary_line ::= "summary:" ws? (count ws)+ 1436 count ::= num | "."]]></programlisting> 1437 1438 <para>Where:</para> 1439 <itemizedlist> 1440 <listitem> 1441 <para><computeroutput>non_nl_string</computeroutput> is any 1442 string not containing a newline.</para> 1443 </listitem> 1444 <listitem> 1445 <para><computeroutput>cmd</computeroutput> is a string holding the 1446 command line of the profiled program.</para> 1447 </listitem> 1448 <listitem> 1449 <para><computeroutput>event</computeroutput> is a string containing 1450 no whitespace.</para> 1451 </listitem> 1452 <listitem> 1453 <para><computeroutput>filename</computeroutput> and 1454 <computeroutput>fn_name</computeroutput> are strings.</para> 1455 </listitem> 1456 <listitem> 1457 <para><computeroutput>num</computeroutput> and 1458 <computeroutput>line_num</computeroutput> are decimal 1459 numbers.</para> 1460 </listitem> 1461 <listitem> 1462 <para><computeroutput>ws</computeroutput> is whitespace.</para> 1463 </listitem> 1464 </itemizedlist> 1465 1466 <para>The contents of the "desc:" lines are printed out at the top 1467 of the summary. This is a generic way of providing simulation 1468 specific information, e.g. for giving the cache configuration for 1469 cache simulation.</para> 1470 1471 <para>More than one line of info can be presented for each file/fn/line number. 1472 In such cases, the counts for the named events will be accumulated.</para> 1473 1474 <para>Counts can be "." to represent zero. This makes the files easier for 1475 humans to read.</para> 1476 1477 <para>The number of counts in each 1478 <computeroutput>line</computeroutput> and the 1479 <computeroutput>summary_line</computeroutput> should not exceed 1480 the number of events in the 1481 <computeroutput>event_line</computeroutput>. If the number in 1482 each <computeroutput>line</computeroutput> is less, cg_annotate 1483 treats those missing as though they were a "." entry. This saves space. 1484 </para> 1485 1486 <para>A <computeroutput>file_line</computeroutput> changes the 1487 current file name. A <computeroutput>fn_line</computeroutput> 1488 changes the current function name. A 1489 <computeroutput>count_line</computeroutput> contains counts that 1490 pertain to the current filename/fn_name. A "fn=" 1491 <computeroutput>file_line</computeroutput> and a 1492 <computeroutput>fn_line</computeroutput> must appear before any 1493 <computeroutput>count_line</computeroutput>s to give the context 1494 of the first <computeroutput>count_line</computeroutput>s.</para> 1495 1496 <para>Each <computeroutput>file_line</computeroutput> will normally be 1497 immediately followed by a <computeroutput>fn_line</computeroutput>. But it 1498 doesn't have to be.</para> 1499 1500 <para>The summary line is redundant, because it just holds the total counts 1501 for each event. But this serves as a useful sanity check of the data; if 1502 the totals for each event don't match the summary line, something has gone 1503 wrong.</para> 1504 1505 </sect2> 1506 1507 </sect1> 1508 </chapter> 1509
