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 12Valgrind 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 18hierarchy and (optionally) branch predictor. It simulates a machine with 19independent first-level instruction and data caches (I1 and D1), backed by a 20unified second-level cache (L2). This exactly matches the configuration of 21many modern machines.</para> 22 23<para>However, some modern machines have three or four levels of cache. For these 24machines (in the cases where Cachegrind can auto-detect the cache 25configuration) Cachegrind simulates the first-level and last-level caches. 26The reason for this choice is that the last-level cache has the most influence on 27runtime, as it masks accesses to main memory. Furthermore, the L1 caches 28often have low associativity, so simulating them can detect cases where the 29code interacts badly with this cache (eg. traversing a matrix column-wise 30with the row length being a power of 2).</para> 31 32<para>Therefore, Cachegrind always refers to the I1, D1 and LL (last-level) 33caches.</para> 34 35<para> 36Cachegrind gathers the following statistics (abbreviations used for each statistic 37is 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 75accesses 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 81function in the program. You can also annotate each line of source code in 82the program with the counts that were caused directly by it.</para> 83 84<para>On a modern machine, an L1 miss will typically cost 85around 10 cycles, an LL miss can cost as much as 200 86cycles, and a mispredicted branch costs in the region of 10 87to 30 cycles. Detailed cache and branch profiling can be very useful 88for understanding how your program interacts with the machine and thus how 89to make it faster.</para> 90 91<para>Also, since one instruction cache read is performed per 92instruction executed, you can find out how many instructions are 93executed 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 104compile with debugging info (the 105<option>-g</option> option). But by contrast with 106normal Valgrind use, you probably do want to turn 107optimisation on, since you should profile your program as it will 108be normally run.</para> 109 110<para>Then, you need to run Cachegrind itself to gather the profiling 111information, and then run cg_annotate to get a detailed presentation of that 112information. As an optional intermediate step, you can use cg_merge to sum 113together the outputs of multiple Cachegrind runs into a single file which 114you then use as the input for cg_annotate. Alternatively, you can use 115cg_diff to difference the outputs of two Cachegrind runs into a single file 116which 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[ 124valgrind --tool=cachegrind prog 125]]></screen> 126 127<para>The program will execute (slowly). Upon completion, 128summary 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 147first, giving the number of fetches made (this is the number of 148instructions executed, which can be useful to know in its own 149right), 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 153to that of the instruction fetches, except that the values are 154also 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 157total).</para> 158 159<para>Combined instruction and data figures for the LL cache 160follow that. Note that the LL miss rate is computed relative to the total 161number of memory accesses, not the number of L1 misses. I.e. it is 162<computeroutput>(ILmr + DLmr + DLmw) / (Ir + Dr + Dw)</computeroutput> 163not 164<computeroutput>(ILmr + DLmr + DLmw) / (I1mr + D1mr + D1mw)</computeroutput> 165</para> 166 167<para>Branch prediction statistics are not collected by default. 168To 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 177more 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 180can be changed with the <option>--cachegrind-out-file</option> option. This 181file is human-readable, but is intended to be interpreted by the 182accompanying program cg_annotate, described in the next section.</para> 183 184<para>The default <computeroutput>.<pid></computeroutput> suffix 185on the output file name serves two purposes. Firstly, it means you 186don't have to rename old log files that you don't want to overwrite. 187Secondly, and more importantly, it allows correct profiling with the 188<option>--trace-children=yes</option> option of 189programs that spawn child processes.</para> 190 191<para>The output file can be big, many megabytes for large applications 192built 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, 202it is worth widening your window to be at least 120-characters 203wide 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-------------------------------------------------------------------------------- 221I1 cache: 65536 B, 64 B, 2-way associative 222D1 cache: 65536 B, 64 B, 2-way associative 223LL cache: 262144 B, 64 B, 8-way associative 224Command: concord vg_to_ucode.c 225Events recorded: Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 226Events shown: Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 227Event sort order: Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 228Threshold: 99% 229Chosen for annotation: 230Auto-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 311program:</para> 312 313<programlisting><![CDATA[ 314-------------------------------------------------------------------------------- 315Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw 316-------------------------------------------------------------------------------- 31727,742,716 276 275 10,955,517 21,905 3,987 4,474,773 19,280 19,098 PROGRAM TOTALS]]></programlisting> 318 319<para> 320These 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-------------------------------------------------------------------------------- 327Ir I1mr ILmr Dr D1mr DLmr Dw D1mw DLmw file:function 328-------------------------------------------------------------------------------- 3298,821,482 5 5 2,242,702 1,621 73 1,794,230 0 0 getc.c:_IO_getc 3305,222,023 4 4 2,276,334 16 12 875,959 1 1 concord.c:get_word 3312,649,248 2 2 1,344,810 7,326 1,385 . . . vg_main.c:strcmp 3322,521,927 2 2 591,215 0 0 179,398 0 0 concord.c:hash 3332,242,740 2 2 1,046,612 568 22 448,548 0 0 ctype.c:tolower 3341,496,937 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 349is identified by a 350<computeroutput>file_name:function_name</computeroutput> pair. If 351a column contains only a dot it means the function never performs 352that event (e.g. the third row shows that 353<computeroutput>strcmp()</computeroutput> contains no 354instructions that write to memory). The name 355<computeroutput>???</computeroutput> is used if the file name 356and/or function name could not be determined from debugging 357information. If most of the entries have the form 358<computeroutput>???:???</computeroutput> the program probably 359wasn't compiled with <option>-g</option>.</para> 360 361<para>It is worth noting that functions will come both from 362the profiled program (e.g. <filename>concord.c</filename>) 363and 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 372manually 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 375produces 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-------------------------------------------------------------------------------- 382Ir 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 410terminal is clearly useful.)</para> 411 412<para>Each source file is clearly marked 413(<computeroutput>User-annotated source</computeroutput>) as 414having been chosen manually for annotation. If the file was 415found in one of the directories specified with the 416<option>-I</option>/<option>--include</option> option, the directory 417and file are both given.</para> 418 419<para>Each line is annotated with its event counts. Events not 420applicable for a line are represented by a dot. This is useful 421for distinguishing between an event which cannot happen, and one 422which can but did not.</para> 423 424<para>Sometimes only a small section of a source file is 425executed. To minimise uninteresting output, Cachegrind only shows 426annotated lines and lines within a small distance of annotated 427lines. Gaps are marked with the line numbers so you know which 428part 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 437controlled by the <option>--context</option> 438option.</para> 439 440<para>To get automatic annotation, use the <option>--auto=yes</option> option. 441cg_annotate will automatically annotate every source file it can 442find that is mentioned in the function-by-function summary. 443Therefore, the files chosen for auto-annotation are affected by 444the <option>--sort</option> and 445<option>--threshold</option> options. Each 446source file is clearly marked (<computeroutput>Auto-annotated 447source</computeroutput>) as being chosen automatically. Any 448files that could not be found are mentioned at the end of the 449output, eg:</para> 450 451<programlisting><![CDATA[ 452------------------------------------------------------------------ 453The 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 460usually compiled with debugging information, but the source files 461are often not present on a system. If a file is chosen for 462annotation both manually and automatically, it 463is marked as <computeroutput>User-annotated 464source</computeroutput>. Use the 465<option>-I</option>/<option>--include</option> option to tell Valgrind where 466to look for source files if the filenames found from the debugging 467information 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, 471e.g. 30 seconds or more. Also beware that auto-annotation can 472produce 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 481the assembly code generated for your C program. Sometimes this is 482useful for understanding what is really happening when an 483interesting line of C code is translated into multiple 484instructions.</para> 485 486<para>To do this, you just need to assemble your 487<computeroutput>.s</computeroutput> files with assembly-level debug 488information. You can use compile with the <option>-S</option> to compile C/C++ 489programs 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 491assembly 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 498has 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>, 502then the outputs from the parent and child will be intermingled in a single 503output file, which will almost certainly make it unreadable by 504cg_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 512cg_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>Files with more than 65,535 lines cause difficulties 627 for the Stabs-format debug info reader. This is because the line 628 number in the <computeroutput>struct nlist</computeroutput> 629 defined in <filename>a.out.h</filename> under Linux is only a 630 16-bit value. Valgrind can handle some files with more than 631 65,535 lines correctly by making some guesses to identify 632 line number overflows. But some cases are beyond it, in 633 which case you'll get a warning message explaining that 634 annotations for the file might be incorrect.</para> 635 636 <para>If you are using GCC 3.1 or later, this is most likely 637 irrelevant, since GCC switched to using the more modern DWARF2 638 format by default at version 3.1. DWARF2 does not have any such 639 limitations on line numbers.</para> 640 </listitem> 641 642 <listitem> 643 <para>If you compile some files with 644 <option>-g</option> and some without, some 645 events that take place in a file without debug info could be 646 attributed to the last line of a file with debug info 647 (whichever one gets placed before the non-debug-info file in 648 the executable).</para> 649 </listitem> 650 651</itemizedlist> 652 653<para>This list looks long, but these cases should be fairly 654rare.</para> 655 656</sect2> 657 658 659<sect2 id="cg-manual.cg_merge" xreflabel="cg_merge"> 660<title>Merging Profiles with cg_merge</title> 661 662<para> 663cg_merge is a simple program which 664reads multiple profile files, as created by Cachegrind, merges them 665together, and writes the results into another file in the same format. 666You can then examine the merged results using 667<computeroutput>cg_annotate <filename></computeroutput>, as 668described above. The merging functionality might be useful if you 669want to aggregate costs over multiple runs of the same program, or 670from a single parallel run with multiple instances of the same 671program.</para> 672 673<para> 674cg_merge is invoked as follows: 675</para> 676 677<programlisting><![CDATA[ 678cg_merge -o outputfile file1 file2 file3 ...]]></programlisting> 679 680<para> 681It reads and checks <computeroutput>file1</computeroutput>, then read 682and checks <computeroutput>file2</computeroutput> and merges it into 683the running totals, then the same with 684<computeroutput>file3</computeroutput>, etc. The final results are 685written to <computeroutput>outputfile</computeroutput>, or to standard 686out if no output file is specified.</para> 687 688<para> 689Costs are summed on a per-function, per-line and per-instruction 690basis. Because of this, the order in which the input files does not 691matter, although you should take care to only mention each file once, 692since any file mentioned twice will be added in twice.</para> 693 694<para> 695cg_merge does not attempt to check 696that the input files come from runs of the same executable. It will 697happily merge together profile files from completely unrelated 698programs. It does however check that the 699<computeroutput>Events:</computeroutput> lines of all the inputs are 700identical, so as to ensure that the addition of costs makes sense. 701For example, it would be nonsensical for it to add a number indicating 702D1 read references to a number from a different file indicating LL 703write misses.</para> 704 705<para> 706A number of other syntax and sanity checks are done whilst reading the 707inputs. cg_merge will stop and 708attempt to print a helpful error message if any of the input files 709fail these checks.</para> 710 711</sect2> 712 713 714<sect2 id="cg-manual.cg_diff" xreflabel="cg_diff"> 715<title>Differencing Profiles with cg_diff</title> 716 717<para> 718cg_diff is a simple program which 719reads two profile files, as created by Cachegrind, finds the difference 720between them, and writes the results into another file in the same format. 721You can then examine the merged results using 722<computeroutput>cg_annotate <filename></computeroutput>, as 723described above. This is very useful if you want to measure how a change to 724a program affected its performance. 725</para> 726 727<para> 728cg_diff is invoked as follows: 729</para> 730 731<programlisting><![CDATA[ 732cg_diff file1 file2]]></programlisting> 733 734<para> 735It reads and checks <computeroutput>file1</computeroutput>, then read 736and checks <computeroutput>file2</computeroutput>, then computes the 737difference (effectively <computeroutput>file1</computeroutput> - 738<computeroutput>file2</computeroutput>). The final results are written to 739standard output.</para> 740 741<para> 742Costs are summed on a per-function basis. Per-line costs are not summed, 743because doing so is too difficult. For example, consider differencing two 744profiles, one from a single-file program A, and one from the same program A 745where a single blank line was inserted at the top of the file. Every single 746per-line count has changed. In comparison, the per-function counts have not 747changed. The per-function count differences are still very useful for 748determining differences between programs. Note that because the result is 749the difference of two profiles, many of the counts will be negative; this 750indicates that the counts for the relevant function are fewer in the second 751version than those in the first version.</para> 752 753<para> 754cg_diff does not attempt to check 755that the input files come from runs of the same executable. It will 756happily merge together profile files from completely unrelated 757programs. It does however check that the 758<computeroutput>Events:</computeroutput> lines of all the inputs are 759identical, so as to ensure that the addition of costs makes sense. 760For example, it would be nonsensical for it to add a number indicating 761D1 read references to a number from a different file indicating LL 762write misses.</para> 763 764<para> 765A number of other syntax and sanity checks are done whilst reading the 766inputs. cg_diff will stop and 767attempt to print a helpful error message if any of the input files 768fail these checks.</para> 769 770<para> 771Sometimes you will want to compare Cachegrind profiles of two versions of a 772program that you have sitting side-by-side. For example, you might have 773<computeroutput>version1/prog.c</computeroutput> and 774<computeroutput>version2/prog.c</computeroutput>, where the second is 775slightly different to the first. A straight comparison of the two will not 776be useful -- because functions are qualified with filenames, a function 777<function>f</function> will be listed as 778<computeroutput>version1/prog.c:f</computeroutput> for the first version but 779<computeroutput>version2/prog.c:f</computeroutput> for the second 780version.</para> 781 782<para> 783When this happens, you can use the <option>--mod-filename</option> option. 784Its argument is a Perl search-and-replace expression that will be applied 785to all the filenames in both Cachegrind output files. It can be used to 786remove minor differences in filenames. For example, the option 787<option>--mod-filename='s/version[0-9]/versionN/'</option> will suffice for 788this case.</para> 789 790<para> 791Similarly, sometimes compilers auto-generate certain functions and give them 792randomized names. For example, GCC sometimes auto-generates functions with 793names like <function>T.1234</function>, and the suffixes vary from build to 794build. You can use the <option>--mod-funcname</option> option to remove 795small differences like these; it works in the same way as 796<option>--mod-filename</option>.</para> 797 798</sect2> 799 800 801</sect1> 802 803 804 805<sect1 id="cg-manual.cgopts" xreflabel="Cachegrind Command-line Options"> 806<title>Cachegrind Command-line Options</title> 807 808<!-- start of xi:include in the manpage --> 809<para>Cachegrind-specific options are:</para> 810 811<variablelist id="cg.opts.list"> 812 813 <varlistentry id="opt.I1" xreflabel="--I1"> 814 <term> 815 <option><![CDATA[--I1=<size>,<associativity>,<line size> ]]></option> 816 </term> 817 <listitem> 818 <para>Specify the size, associativity and line size of the level 1 819 instruction cache. </para> 820 </listitem> 821 </varlistentry> 822 823 <varlistentry id="opt.D1" xreflabel="--D1"> 824 <term> 825 <option><![CDATA[--D1=<size>,<associativity>,<line size> ]]></option> 826 </term> 827 <listitem> 828 <para>Specify the size, associativity and line size of the level 1 829 data cache.</para> 830 </listitem> 831 </varlistentry> 832 833 <varlistentry id="opt.LL" xreflabel="--LL"> 834 <term> 835 <option><![CDATA[--LL=<size>,<associativity>,<line size> ]]></option> 836 </term> 837 <listitem> 838 <para>Specify the size, associativity and line size of the last-level 839 cache.</para> 840 </listitem> 841 </varlistentry> 842 843 <varlistentry id="opt.cache-sim" xreflabel="--cache-sim"> 844 <term> 845 <option><![CDATA[--cache-sim=no|yes [yes] ]]></option> 846 </term> 847 <listitem> 848 <para>Enables or disables collection of cache access and miss 849 counts.</para> 850 </listitem> 851 </varlistentry> 852 853 <varlistentry id="opt.branch-sim" xreflabel="--branch-sim"> 854 <term> 855 <option><![CDATA[--branch-sim=no|yes [no] ]]></option> 856 </term> 857 <listitem> 858 <para>Enables or disables collection of branch instruction and 859 misprediction counts. By default this is disabled as it 860 slows Cachegrind down by approximately 25%. Note that you 861 cannot specify <option>--cache-sim=no</option> 862 and <option>--branch-sim=no</option> 863 together, as that would leave Cachegrind with no 864 information to collect.</para> 865 </listitem> 866 </varlistentry> 867 868 <varlistentry id="opt.cachegrind-out-file" xreflabel="--cachegrind-out-file"> 869 <term> 870 <option><![CDATA[--cachegrind-out-file=<file> ]]></option> 871 </term> 872 <listitem> 873 <para>Write the profile data to 874 <computeroutput>file</computeroutput> rather than to the default 875 output file, 876 <filename>cachegrind.out.<pid></filename>. The 877 <option>%p</option> and <option>%q</option> format specifiers 878 can be used to embed the process ID and/or the contents of an 879 environment variable in the name, as is the case for the core 880 option <option><xref linkend="opt.log-file"/></option>. 881 </para> 882 </listitem> 883 </varlistentry> 884 885</variablelist> 886<!-- end of xi:include in the manpage --> 887 888</sect1> 889 890 891 892<sect1 id="cg-manual.annopts" xreflabel="cg_annotate Command-line Options"> 893<title>cg_annotate Command-line Options</title> 894 895<!-- start of xi:include in the manpage --> 896<variablelist id="cg_annotate.opts.list"> 897 898 <varlistentry> 899 <term> 900 <option><![CDATA[-h --help ]]></option> 901 </term> 902 <listitem> 903 <para>Show the help message.</para> 904 </listitem> 905 </varlistentry> 906 907 <varlistentry> 908 <term> 909 <option><![CDATA[--version ]]></option> 910 </term> 911 <listitem> 912 <para>Show the version number.</para> 913 </listitem> 914 </varlistentry> 915 916 <varlistentry> 917 <term> 918 <option><![CDATA[--show=A,B,C [default: all, using order in 919 cachegrind.out.<pid>] ]]></option> 920 </term> 921 <listitem> 922 <para>Specifies which events to show (and the column 923 order). Default is to use all present in the 924 <filename>cachegrind.out.<pid></filename> file (and 925 use the order in the file). Useful if you want to concentrate on, for 926 example, I cache misses (<option>--show=I1mr,ILmr</option>), or data 927 read misses (<option>--show=D1mr,DLmr</option>), or LL data misses 928 (<option>--show=DLmr,DLmw</option>). Best used in conjunction with 929 <option>--sort</option>.</para> 930 </listitem> 931 </varlistentry> 932 933 <varlistentry> 934 <term> 935 <option><![CDATA[--sort=A,B,C [default: order in 936 cachegrind.out.<pid>] ]]></option> 937 </term> 938 <listitem> 939 <para>Specifies the events upon which the sorting of the 940 function-by-function entries will be based.</para> 941 </listitem> 942 </varlistentry> 943 944 <varlistentry> 945 <term> 946 <option><![CDATA[--threshold=X [default: 0.1%] ]]></option> 947 </term> 948 <listitem> 949 <para>Sets the threshold for the function-by-function 950 summary. A function is shown if it accounts for more than X% 951 of the counts for the primary sort event. If auto-annotating, also 952 affects which files are annotated.</para> 953 954 <para>Note: thresholds can be set for more than one of the 955 events by appending any events for the 956 <option>--sort</option> option with a colon 957 and a number (no spaces, though). E.g. if you want to see 958 each function that covers more than 1% of LL read misses or 1% of LL 959 write misses, use this option:</para> 960 <para><option>--sort=DLmr:1,DLmw:1</option></para> 961 </listitem> 962 </varlistentry> 963 964 <varlistentry> 965 <term> 966 <option><![CDATA[--auto=<no|yes> [default: no] ]]></option> 967 </term> 968 <listitem> 969 <para>When enabled, automatically annotates every file that 970 is mentioned in the function-by-function summary that can be 971 found. Also gives a list of those that couldn't be found.</para> 972 </listitem> 973 </varlistentry> 974 975 <varlistentry> 976 <term> 977 <option><![CDATA[--context=N [default: 8] ]]></option> 978 </term> 979 <listitem> 980 <para>Print N lines of context before and after each 981 annotated line. Avoids printing large sections of source 982 files that were not executed. Use a large number 983 (e.g. 100000) to show all source lines.</para> 984 </listitem> 985 </varlistentry> 986 987 <varlistentry> 988 <term> 989 <option><![CDATA[-I<dir> --include=<dir> [default: none] ]]></option> 990 </term> 991 <listitem> 992 <para>Adds a directory to the list in which to search for 993 files. Multiple <option>-I</option>/<option>--include</option> 994 options can be given to add multiple directories.</para> 995 </listitem> 996 </varlistentry> 997 998</variablelist> 999<!-- end of xi:include in the manpage --> 1000 1001</sect1> 1002 1003 1004<sect1 id="cg-manual.mergeopts" xreflabel="cg_merge Command-line Options"> 1005<title>cg_merge Command-line Options</title> 1006 1007<!-- start of xi:include in the manpage --> 1008<variablelist id="cg_merge.opts.list"> 1009 1010 <varlistentry> 1011 <term> 1012 <option><![CDATA[-o outfile]]></option> 1013 </term> 1014 <listitem> 1015 <para>Write the profile data to <computeroutput>outfile</computeroutput> 1016 rather than to standard output. 1017 </para> 1018 </listitem> 1019 </varlistentry> 1020 1021</variablelist> 1022<!-- end of xi:include in the manpage --> 1023 1024</sect1> 1025 1026 1027<sect1 id="cg-manual.diffopts" xreflabel="cg_diff Command-line Options"> 1028<title>cg_diff Command-line Options</title> 1029 1030<!-- start of xi:include in the manpage --> 1031<variablelist id="cg_diff.opts.list"> 1032 1033 <varlistentry> 1034 <term> 1035 <option><![CDATA[-h --help ]]></option> 1036 </term> 1037 <listitem> 1038 <para>Show the help message.</para> 1039 </listitem> 1040 </varlistentry> 1041 1042 <varlistentry> 1043 <term> 1044 <option><![CDATA[--version ]]></option> 1045 </term> 1046 <listitem> 1047 <para>Show the version number.</para> 1048 </listitem> 1049 </varlistentry> 1050 1051 <varlistentry> 1052 <term> 1053 <option><![CDATA[--mod-filename=<expr> [default: none]]]></option> 1054 </term> 1055 <listitem> 1056 <para>Specifies a Perl search-and-replace expression that is applied 1057 to all filenames. Useful for removing minor differences in paths 1058 between two different versions of a program that are sitting in 1059 different directories.</para> 1060 </listitem> 1061 </varlistentry> 1062 1063 <varlistentry> 1064 <term> 1065 <option><![CDATA[--mod-funcname=<expr> [default: none]]]></option> 1066 </term> 1067 <listitem> 1068 <para>Like <option>--mod-filename</option>, but for filenames. 1069 Useful for removing minor differences in randomized names of 1070 auto-generated functions generated by some compilers.</para> 1071 </listitem> 1072 </varlistentry> 1073 1074</variablelist> 1075<!-- end of xi:include in the manpage --> 1076 1077</sect1> 1078 1079 1080 1081 1082<sect1 id="cg-manual.acting-on" 1083 xreflabel="Acting on Cachegrind's Information"> 1084<title>Acting on Cachegrind's Information</title> 1085<para> 1086Cachegrind gives you lots of information, but acting on that information 1087isn't always easy. Here are some rules of thumb that we have found to be 1088useful.</para> 1089 1090<para> 1091First of all, the global hit/miss counts and miss rates are not that useful. 1092If you have multiple programs or multiple runs of a program, comparing the 1093numbers might identify if any are outliers and worthy of closer 1094investigation. Otherwise, they're not enough to act on.</para> 1095 1096<para> 1097The function-by-function counts are more useful to look at, as they pinpoint 1098which functions are causing large numbers of counts. However, beware that 1099inlining can make these counts misleading. If a function 1100<function>f</function> is always inlined, counts will be attributed to the 1101functions it is inlined into, rather than itself. However, if you look at 1102the line-by-line annotations for <function>f</function> you'll see the 1103counts that belong to <function>f</function>. (This is hard to avoid, it's 1104how the debug info is structured.) So it's worth looking for large numbers 1105in the line-by-line annotations.</para> 1106 1107<para> 1108The line-by-line source code annotations are much more useful. In our 1109experience, the best place to start is by looking at the 1110<computeroutput>Ir</computeroutput> numbers. They simply measure how many 1111instructions were executed for each line, and don't include any cache 1112information, but they can still be very useful for identifying 1113bottlenecks.</para> 1114 1115<para> 1116After that, we have found that LL misses are typically a much bigger source 1117of slow-downs than L1 misses. So it's worth looking for any snippets of 1118code with high <computeroutput>DLmr</computeroutput> or 1119<computeroutput>DLmw</computeroutput> counts. (You can use 1120<option>--show=DLmr 1121--sort=DLmr</option> with cg_annotate to focus just on 1122<literal>DLmr</literal> counts, for example.) If you find any, it's still 1123not always easy to work out how to improve things. You need to have a 1124reasonable understanding of how caches work, the principles of locality, and 1125your program's data access patterns. Improving things may require 1126redesigning a data structure, for example.</para> 1127 1128<para> 1129Looking at the <computeroutput>Bcm</computeroutput> and 1130<computeroutput>Bim</computeroutput> misses can also be helpful. 1131In particular, <computeroutput>Bim</computeroutput> misses are often caused 1132by <literal>switch</literal> statements, and in some cases these 1133<literal>switch</literal> statements can be replaced with table-driven code. 1134For example, you might replace code like this:</para> 1135 1136<programlisting><![CDATA[ 1137enum E { A, B, C }; 1138enum E e; 1139int i; 1140... 1141switch (e) 1142{ 1143 case A: i += 1; break; 1144 case B: i += 2; break; 1145 case C: i += 3; break; 1146} 1147]]></programlisting> 1148 1149<para>with code like this:</para> 1150 1151<programlisting><![CDATA[ 1152enum E { A, B, C }; 1153enum E e; 1154enum E table[] = { 1, 2, 3 }; 1155int i; 1156... 1157i += table[e]; 1158]]></programlisting> 1159 1160<para> 1161This is obviously a contrived example, but the basic principle applies in a 1162wide variety of situations.</para> 1163 1164<para> 1165In short, Cachegrind can tell you where some of the bottlenecks in your code 1166are, but it can't tell you how to fix them. You have to work that out for 1167yourself. But at least you have the information! 1168</para> 1169 1170</sect1> 1171 1172 1173<sect1 id="cg-manual.sim-details" 1174 xreflabel="Simulation Details"> 1175<title>Simulation Details</title> 1176<para> 1177This section talks about details you don't need to know about in order to 1178use Cachegrind, but may be of interest to some people. 1179</para> 1180 1181<sect2 id="cache-sim" xreflabel="Cache Simulation Specifics"> 1182<title>Cache Simulation Specifics</title> 1183 1184<para>Specific characteristics of the cache simulation are as 1185follows:</para> 1186 1187<itemizedlist> 1188 1189 <listitem> 1190 <para>Write-allocate: when a write miss occurs, the block 1191 written to is brought into the D1 cache. Most modern caches 1192 have this property.</para> 1193 </listitem> 1194 1195 <listitem> 1196 <para>Bit-selection hash function: the set of line(s) in the cache 1197 to which a memory block maps is chosen by the middle bits 1198 M--(M+N-1) of the byte address, where:</para> 1199 <itemizedlist> 1200 <listitem> 1201 <para>line size = 2^M bytes</para> 1202 </listitem> 1203 <listitem> 1204 <para>(cache size / line size / associativity) = 2^N bytes</para> 1205 </listitem> 1206 </itemizedlist> 1207 </listitem> 1208 1209 <listitem> 1210 <para>Inclusive LL cache: the LL cache typically replicates all 1211 the entries of the L1 caches, because fetching into L1 involves 1212 fetching into LL first (this does not guarantee strict inclusiveness, 1213 as lines evicted from LL still could reside in L1). This is 1214 standard on Pentium chips, but AMD Opterons, Athlons and Durons 1215 use an exclusive LL cache that only holds 1216 blocks evicted from L1. Ditto most modern VIA CPUs.</para> 1217 </listitem> 1218 1219</itemizedlist> 1220 1221<para>The cache configuration simulated (cache size, 1222associativity and line size) is determined automatically using 1223the x86 CPUID instruction. If you have a machine that (a) 1224doesn't support the CPUID instruction, or (b) supports it in an 1225early incarnation that doesn't give any cache information, then 1226Cachegrind will fall back to using a default configuration (that 1227of a model 3/4 Athlon). Cachegrind will tell you if this 1228happens. You can manually specify one, two or all three levels 1229(I1/D1/LL) of the cache from the command line using the 1230<option>--I1</option>, 1231<option>--D1</option> and 1232<option>--LL</option> options. 1233For cache parameters to be valid for simulation, the number 1234of sets (with associativity being the number of cache lines in 1235each set) has to be a power of two.</para> 1236 1237<para>On PowerPC platforms 1238Cachegrind cannot automatically 1239determine the cache configuration, so you will 1240need to specify it with the 1241<option>--I1</option>, 1242<option>--D1</option> and 1243<option>--LL</option> options.</para> 1244 1245 1246<para>Other noteworthy behaviour:</para> 1247 1248<itemizedlist> 1249 <listitem> 1250 <para>References that straddle two cache lines are treated as 1251 follows:</para> 1252 <itemizedlist> 1253 <listitem> 1254 <para>If both blocks hit --> counted as one hit</para> 1255 </listitem> 1256 <listitem> 1257 <para>If one block hits, the other misses --> counted 1258 as one miss.</para> 1259 </listitem> 1260 <listitem> 1261 <para>If both blocks miss --> counted as one miss (not 1262 two)</para> 1263 </listitem> 1264 </itemizedlist> 1265 </listitem> 1266 1267 <listitem> 1268 <para>Instructions that modify a memory location 1269 (e.g. <computeroutput>inc</computeroutput> and 1270 <computeroutput>dec</computeroutput>) are counted as doing 1271 just a read, i.e. a single data reference. This may seem 1272 strange, but since the write can never cause a miss (the read 1273 guarantees the block is in the cache) it's not very 1274 interesting.</para> 1275 1276 <para>Thus it measures not the number of times the data cache 1277 is accessed, but the number of times a data cache miss could 1278 occur.</para> 1279 </listitem> 1280 1281</itemizedlist> 1282 1283<para>If you are interested in simulating a cache with different 1284properties, it is not particularly hard to write your own cache 1285simulator, or to modify the existing ones in 1286<computeroutput>cg_sim.c</computeroutput>. We'd be 1287interested to hear from anyone who does.</para> 1288 1289</sect2> 1290 1291 1292<sect2 id="branch-sim" xreflabel="Branch Simulation Specifics"> 1293<title>Branch Simulation Specifics</title> 1294 1295<para>Cachegrind simulates branch predictors intended to be 1296typical of mainstream desktop/server processors of around 2004.</para> 1297 1298<para>Conditional branches are predicted using an array of 16384 2-bit 1299saturating counters. The array index used for a branch instruction is 1300computed partly from the low-order bits of the branch instruction's 1301address and partly using the taken/not-taken behaviour of the last few 1302conditional branches. As a result the predictions for any specific 1303branch depend both on its own history and the behaviour of previous 1304branches. This is a standard technique for improving prediction 1305accuracy.</para> 1306 1307<para>For indirect branches (that is, jumps to unknown destinations) 1308Cachegrind uses a simple branch target address predictor. Targets are 1309predicted using an array of 512 entries indexed by the low order 9 1310bits of the branch instruction's address. Each branch is predicted to 1311jump to the same address it did last time. Any other behaviour causes 1312a mispredict.</para> 1313 1314<para>More recent processors have better branch predictors, in 1315particular better indirect branch predictors. Cachegrind's predictor 1316design is deliberately conservative so as to be representative of the 1317large installed base of processors which pre-date widespread 1318deployment of more sophisticated indirect branch predictors. In 1319particular, late model Pentium 4s (Prescott), Pentium M, Core and Core 13202 have more sophisticated indirect branch predictors than modelled by 1321Cachegrind. </para> 1322 1323<para>Cachegrind does not simulate a return stack predictor. It 1324assumes that processors perfectly predict function return addresses, 1325an assumption which is probably close to being true.</para> 1326 1327<para>See Hennessy and Patterson's classic text "Computer 1328Architecture: A Quantitative Approach", 4th edition (2007), Section 13292.3 (pages 80-89) for background on modern branch predictors.</para> 1330 1331</sect2> 1332 1333<sect2 id="cg-manual.annopts.accuracy" xreflabel="Accuracy"> 1334<title>Accuracy</title> 1335 1336<para>Valgrind's cache profiling has a number of 1337shortcomings:</para> 1338 1339<itemizedlist> 1340 <listitem> 1341 <para>It doesn't account for kernel activity -- the effect of system 1342 calls on the cache and branch predictor contents is ignored.</para> 1343 </listitem> 1344 1345 <listitem> 1346 <para>It doesn't account for other process activity. 1347 This is probably desirable when considering a single 1348 program.</para> 1349 </listitem> 1350 1351 <listitem> 1352 <para>It doesn't account for virtual-to-physical address 1353 mappings. Hence the simulation is not a true 1354 representation of what's happening in the 1355 cache. Most caches and branch predictors are physically indexed, but 1356 Cachegrind simulates caches using virtual addresses.</para> 1357 </listitem> 1358 1359 <listitem> 1360 <para>It doesn't account for cache misses not visible at the 1361 instruction level, e.g. those arising from TLB misses, or 1362 speculative execution.</para> 1363 </listitem> 1364 1365 <listitem> 1366 <para>Valgrind will schedule 1367 threads differently from how they would be when running natively. 1368 This could warp the results for threaded programs.</para> 1369 </listitem> 1370 1371 <listitem> 1372 <para>The x86/amd64 instructions <computeroutput>bts</computeroutput>, 1373 <computeroutput>btr</computeroutput> and 1374 <computeroutput>btc</computeroutput> will incorrectly be 1375 counted as doing a data read if both the arguments are 1376 registers, eg:</para> 1377<programlisting><![CDATA[ 1378 btsl %eax, %edx]]></programlisting> 1379 1380 <para>This should only happen rarely.</para> 1381 </listitem> 1382 1383 <listitem> 1384 <para>x86/amd64 FPU instructions with data sizes of 28 and 108 bytes 1385 (e.g. <computeroutput>fsave</computeroutput>) are treated as 1386 though they only access 16 bytes. These instructions seem to 1387 be rare so hopefully this won't affect accuracy much.</para> 1388 </listitem> 1389 1390</itemizedlist> 1391 1392<para>Another thing worth noting is that results are very sensitive. 1393Changing the size of the executable being profiled, or the sizes 1394of any of the shared libraries it uses, or even the length of their 1395file names, can perturb the results. Variations will be small, but 1396don't expect perfectly repeatable results if your program changes at 1397all.</para> 1398 1399<para>More recent GNU/Linux distributions do address space 1400randomisation, in which identical runs of the same program have their 1401shared libraries loaded at different locations, as a security measure. 1402This also perturbs the results.</para> 1403 1404<para>While these factors mean you shouldn't trust the results to 1405be super-accurate, they should be close enough to be useful.</para> 1406 1407</sect2> 1408 1409</sect1> 1410 1411 1412 1413<sect1 id="cg-manual.impl-details" 1414 xreflabel="Implementation Details"> 1415<title>Implementation Details</title> 1416<para> 1417This section talks about details you don't need to know about in order to 1418use Cachegrind, but may be of interest to some people. 1419</para> 1420 1421<sect2 id="cg-manual.impl-details.how-cg-works" 1422 xreflabel="How Cachegrind Works"> 1423<title>How Cachegrind Works</title> 1424<para>The best reference for understanding how Cachegrind works is chapter 3 of 1425"Dynamic Binary Analysis and Instrumentation", by Nicholas Nethercote. It 1426is available on the <ulink url="&vg-pubs-url;">Valgrind publications 1427page</ulink>.</para> 1428</sect2> 1429 1430<sect2 id="cg-manual.impl-details.file-format" 1431 xreflabel="Cachegrind Output File Format"> 1432<title>Cachegrind Output File Format</title> 1433<para>The file format is fairly straightforward, basically giving the 1434cost centre for every line, grouped by files and 1435functions. It's also totally generic and self-describing, in the sense that 1436it can be used for any events that can be counted on a line-by-line basis, 1437not just cache and branch predictor events. For example, earlier versions 1438of Cachegrind didn't have a branch predictor simulation. When this was 1439added, the file format didn't need to change at all. So the format (and 1440consequently, cg_annotate) could be used by other tools.</para> 1441 1442<para>The file format:</para> 1443<programlisting><![CDATA[ 1444file ::= desc_line* cmd_line events_line data_line+ summary_line 1445desc_line ::= "desc:" ws? non_nl_string 1446cmd_line ::= "cmd:" ws? cmd 1447events_line ::= "events:" ws? (event ws)+ 1448data_line ::= file_line | fn_line | count_line 1449file_line ::= "fl=" filename 1450fn_line ::= "fn=" fn_name 1451count_line ::= line_num ws? (count ws)+ 1452summary_line ::= "summary:" ws? (count ws)+ 1453count ::= num | "."]]></programlisting> 1454 1455<para>Where:</para> 1456<itemizedlist> 1457 <listitem> 1458 <para><computeroutput>non_nl_string</computeroutput> is any 1459 string not containing a newline.</para> 1460 </listitem> 1461 <listitem> 1462 <para><computeroutput>cmd</computeroutput> is a string holding the 1463 command line of the profiled program.</para> 1464 </listitem> 1465 <listitem> 1466 <para><computeroutput>event</computeroutput> is a string containing 1467 no whitespace.</para> 1468 </listitem> 1469 <listitem> 1470 <para><computeroutput>filename</computeroutput> and 1471 <computeroutput>fn_name</computeroutput> are strings.</para> 1472 </listitem> 1473 <listitem> 1474 <para><computeroutput>num</computeroutput> and 1475 <computeroutput>line_num</computeroutput> are decimal 1476 numbers.</para> 1477 </listitem> 1478 <listitem> 1479 <para><computeroutput>ws</computeroutput> is whitespace.</para> 1480 </listitem> 1481</itemizedlist> 1482 1483<para>The contents of the "desc:" lines are printed out at the top 1484of the summary. This is a generic way of providing simulation 1485specific information, e.g. for giving the cache configuration for 1486cache simulation.</para> 1487 1488<para>More than one line of info can be presented for each file/fn/line number. 1489In such cases, the counts for the named events will be accumulated.</para> 1490 1491<para>Counts can be "." to represent zero. This makes the files easier for 1492humans to read.</para> 1493 1494<para>The number of counts in each 1495<computeroutput>line</computeroutput> and the 1496<computeroutput>summary_line</computeroutput> should not exceed 1497the number of events in the 1498<computeroutput>event_line</computeroutput>. If the number in 1499each <computeroutput>line</computeroutput> is less, cg_annotate 1500treats those missing as though they were a "." entry. This saves space. 1501</para> 1502 1503<para>A <computeroutput>file_line</computeroutput> changes the 1504current file name. A <computeroutput>fn_line</computeroutput> 1505changes the current function name. A 1506<computeroutput>count_line</computeroutput> contains counts that 1507pertain to the current filename/fn_name. A "fn=" 1508<computeroutput>file_line</computeroutput> and a 1509<computeroutput>fn_line</computeroutput> must appear before any 1510<computeroutput>count_line</computeroutput>s to give the context 1511of the first <computeroutput>count_line</computeroutput>s.</para> 1512 1513<para>Each <computeroutput>file_line</computeroutput> will normally be 1514immediately followed by a <computeroutput>fn_line</computeroutput>. But it 1515doesn't have to be.</para> 1516 1517<para>The summary line is redundant, because it just holds the total counts 1518for each event. But this serves as a useful sanity check of the data; if 1519the totals for each event don't match the summary line, something has gone 1520wrong.</para> 1521 1522</sect2> 1523 1524</sect1> 1525</chapter> 1526