Commit | Line | Data |
---|---|---|
252b5132 RH |
1 | \input texinfo @c -*-texinfo-*- |
2 | @setfilename gprof.info | |
250d07de | 3 | @c Copyright (C) 1988-2021 Free Software Foundation, Inc. |
252b5132 RH |
4 | @settitle GNU gprof |
5 | @setchapternewpage odd | |
6 | ||
e49e529d JM |
7 | @c man begin INCLUDE |
8 | @include bfdver.texi | |
9 | @c man end | |
10 | ||
9160ea82 | 11 | @ifnottex |
252b5132 RH |
12 | @c This is a dir.info fragment to support semi-automated addition of |
13 | @c manuals to an info tree. zoo@cygnus.com is developing this facility. | |
9160ea82 AM |
14 | @dircategory Software development |
15 | @direntry | |
252b5132 | 16 | * gprof: (gprof). Profiling your program's execution |
9160ea82 AM |
17 | @end direntry |
18 | @end ifnottex | |
252b5132 | 19 | |
0e9517a9 | 20 | @copying |
252b5132 RH |
21 | This file documents the gprof profiler of the GNU system. |
22 | ||
40f90528 | 23 | @c man begin COPYRIGHT |
250d07de | 24 | Copyright @copyright{} 1988-2021 Free Software Foundation, Inc. |
252b5132 | 25 | |
40f90528 | 26 | Permission is granted to copy, distribute and/or modify this document |
793c5807 | 27 | under the terms of the GNU Free Documentation License, Version 1.3 |
40f90528 AM |
28 | or any later version published by the Free Software Foundation; |
29 | with no Invariant Sections, with no Front-Cover Texts, and with no | |
30 | Back-Cover Texts. A copy of the license is included in the | |
afb17569 | 31 | section entitled ``GNU Free Documentation License''. |
40f90528 AM |
32 | |
33 | @c man end | |
0e9517a9 | 34 | @end copying |
252b5132 RH |
35 | |
36 | @finalout | |
37 | @smallbook | |
38 | ||
39 | @titlepage | |
40 | @title GNU gprof | |
f3445b37 | 41 | @subtitle The @sc{gnu} Profiler |
e49e529d JM |
42 | @ifset VERSION_PACKAGE |
43 | @subtitle @value{VERSION_PACKAGE} | |
44 | @end ifset | |
45 | @subtitle Version @value{VERSION} | |
252b5132 RH |
46 | @author Jay Fenlason and Richard Stallman |
47 | ||
48 | @page | |
49 | ||
50 | This manual describes the @sc{gnu} profiler, @code{gprof}, and how you | |
51 | can use it to determine which parts of a program are taking most of the | |
52 | execution time. We assume that you know how to write, compile, and | |
53 | execute programs. @sc{gnu} @code{gprof} was written by Jay Fenlason. | |
83aeabb6 | 54 | Eric S. Raymond made some minor corrections and additions in 2003. |
252b5132 | 55 | |
252b5132 | 56 | @vskip 0pt plus 1filll |
250d07de | 57 | Copyright @copyright{} 1988-2021 Free Software Foundation, Inc. |
252b5132 | 58 | |
cf055d54 | 59 | Permission is granted to copy, distribute and/or modify this document |
793c5807 | 60 | under the terms of the GNU Free Documentation License, Version 1.3 |
cf055d54 NC |
61 | or any later version published by the Free Software Foundation; |
62 | with no Invariant Sections, with no Front-Cover Texts, and with no | |
63 | Back-Cover Texts. A copy of the license is included in the | |
afb17569 | 64 | section entitled ``GNU Free Documentation License''. |
252b5132 RH |
65 | |
66 | @end titlepage | |
4ecceb71 | 67 | @contents |
252b5132 | 68 | |
913b4d4b | 69 | @ifnottex |
252b5132 RH |
70 | @node Top |
71 | @top Profiling a Program: Where Does It Spend Its Time? | |
72 | ||
73 | This manual describes the @sc{gnu} profiler, @code{gprof}, and how you | |
74 | can use it to determine which parts of a program are taking most of the | |
75 | execution time. We assume that you know how to write, compile, and | |
76 | execute programs. @sc{gnu} @code{gprof} was written by Jay Fenlason. | |
77 | ||
e49e529d JM |
78 | This manual is for @code{gprof} |
79 | @ifset VERSION_PACKAGE | |
80 | @value{VERSION_PACKAGE} | |
81 | @end ifset | |
82 | version @value{VERSION}. | |
83 | ||
cf055d54 | 84 | This document is distributed under the terms of the GNU Free |
793c5807 NC |
85 | Documentation License version 1.3. A copy of the license is included |
86 | in the section entitled ``GNU Free Documentation License''. | |
cf055d54 | 87 | |
252b5132 RH |
88 | @menu |
89 | * Introduction:: What profiling means, and why it is useful. | |
90 | ||
91 | * Compiling:: How to compile your program for profiling. | |
92 | * Executing:: Executing your program to generate profile data | |
93 | * Invoking:: How to run @code{gprof}, and its options | |
94 | ||
afb17569 | 95 | * Output:: Interpreting @code{gprof}'s output |
252b5132 RH |
96 | |
97 | * Inaccuracy:: Potential problems you should be aware of | |
98 | * How do I?:: Answers to common questions | |
99 | * Incompatibilities:: (between @sc{gnu} @code{gprof} and Unix @code{gprof}.) | |
100 | * Details:: Details of how profiling is done | |
cf055d54 | 101 | * GNU Free Documentation License:: GNU Free Documentation License |
252b5132 | 102 | @end menu |
913b4d4b | 103 | @end ifnottex |
252b5132 RH |
104 | |
105 | @node Introduction | |
106 | @chapter Introduction to Profiling | |
107 | ||
40f90528 AM |
108 | @ifset man |
109 | @c man title gprof display call graph profile data | |
110 | ||
111 | @smallexample | |
112 | @c man begin SYNOPSIS | |
efa97609 | 113 | gprof [ -[abcDhilLrsTvwxyz] ] [ -[ACeEfFJnNOpPqQRStZ][@var{name}] ] |
40f90528 | 114 | [ -I @var{dirs} ] [ -d[@var{num}] ] [ -k @var{from/to} ] |
a1c21132 | 115 | [ -m @var{min-count} ] [ -R @var{map_file} ] [ -t @var{table-length} ] |
f3445b37 | 116 | [ --[no-]annotated-source[=@var{name}] ] |
40f90528 AM |
117 | [ --[no-]exec-counts[=@var{name}] ] |
118 | [ --[no-]flat-profile[=@var{name}] ] [ --[no-]graph[=@var{name}] ] | |
f3445b37 L |
119 | [ --[no-]time=@var{name}] [ --all-lines ] [ --brief ] |
120 | [ --debug[=@var{level}] ] [ --function-ordering ] | |
afb17569 | 121 | [ --file-ordering @var{map_file} ] [ --directory-path=@var{dirs} ] |
40f90528 | 122 | [ --display-unused-functions ] [ --file-format=@var{name} ] |
630b0510 CH |
123 | [ --file-info ] [ --help ] [ --line ] [ --inline-file-names ] |
124 | [ --min-count=@var{n} ] [ --no-static ] [ --print-path ] | |
125 | [ --separate-files ] [ --static-call-graph ] [ --sum ] | |
126 | [ --table-length=@var{len} ] [ --traditional ] [ --version ] | |
127 | [ --width=@var{n} ] [ --ignore-non-functions ] | |
128 | [ --demangle[=@var{STYLE}] ] [ --no-demangle ] | |
129 | [--external-symbol-table=name] | |
0e27a8f6 | 130 | [ @var{image-file} ] [ @var{profile-file} @dots{} ] |
40f90528 AM |
131 | @c man end |
132 | @end smallexample | |
133 | ||
134 | @c man begin DESCRIPTION | |
f3445b37 L |
135 | @code{gprof} produces an execution profile of C, Pascal, or Fortran77 |
136 | programs. The effect of called routines is incorporated in the profile | |
40f90528 AM |
137 | of each caller. The profile data is taken from the call graph profile file |
138 | (@file{gmon.out} default) which is created by programs | |
139 | that are compiled with the @samp{-pg} option of | |
140 | @code{cc}, @code{pc}, and @code{f77}. | |
141 | The @samp{-pg} option also links in versions of the library routines | |
f3445b37 | 142 | that are compiled for profiling. @code{Gprof} reads the given object |
40f90528 AM |
143 | file (the default is @code{a.out}) and establishes the relation between |
144 | its symbol table and the call graph profile from @file{gmon.out}. | |
145 | If more than one profile file is specified, the @code{gprof} | |
146 | output shows the sum of the profile information in the given profile files. | |
147 | ||
148 | @code{Gprof} calculates the amount of time spent in each routine. | |
149 | Next, these times are propagated along the edges of the call graph. | |
150 | Cycles are discovered, and calls into a cycle are made to share the time | |
151 | of the cycle. | |
152 | ||
153 | @c man end | |
154 | ||
155 | @c man begin BUGS | |
156 | The granularity of the sampling is shown, but remains | |
157 | statistical at best. | |
158 | We assume that the time for each execution of a function | |
159 | can be expressed by the total time for the function divided | |
160 | by the number of times the function is called. | |
161 | Thus the time propagated along the call graph arcs to the function's | |
162 | parents is directly proportional to the number of times that | |
163 | arc is traversed. | |
164 | ||
165 | Parents that are not themselves profiled will have the time of | |
166 | their profiled children propagated to them, but they will appear | |
167 | to be spontaneously invoked in the call graph listing, and will | |
168 | not have their time propagated further. | |
169 | Similarly, signal catchers, even though profiled, will appear | |
170 | to be spontaneous (although for more obscure reasons). | |
171 | Any profiled children of signal catchers should have their times | |
172 | propagated properly, unless the signal catcher was invoked during | |
173 | the execution of the profiling routine, in which case all is lost. | |
174 | ||
175 | The profiled program must call @code{exit}(2) | |
176 | or return normally for the profiling information to be saved | |
177 | in the @file{gmon.out} file. | |
178 | @c man end | |
179 | ||
180 | @c man begin FILES | |
181 | @table @code | |
182 | @item @file{a.out} | |
183 | the namelist and text space. | |
184 | @item @file{gmon.out} | |
185 | dynamic call graph and profile. | |
186 | @item @file{gmon.sum} | |
f3445b37 | 187 | summarized dynamic call graph and profile. |
40f90528 AM |
188 | @end table |
189 | @c man end | |
190 | ||
191 | @c man begin SEEALSO | |
192 | monitor(3), profil(2), cc(1), prof(1), and the Info entry for @file{gprof}. | |
193 | ||
194 | ``An Execution Profiler for Modular Programs'', | |
195 | by S. Graham, P. Kessler, M. McKusick; | |
196 | Software - Practice and Experience, | |
197 | Vol. 13, pp. 671-685, 1983. | |
198 | ||
199 | ``gprof: A Call Graph Execution Profiler'', | |
200 | by S. Graham, P. Kessler, M. McKusick; | |
201 | Proceedings of the SIGPLAN '82 Symposium on Compiler Construction, | |
202 | SIGPLAN Notices, Vol. 17, No 6, pp. 120-126, June 1982. | |
203 | @c man end | |
204 | @end ifset | |
205 | ||
252b5132 RH |
206 | Profiling allows you to learn where your program spent its time and which |
207 | functions called which other functions while it was executing. This | |
208 | information can show you which pieces of your program are slower than you | |
209 | expected, and might be candidates for rewriting to make your program | |
210 | execute faster. It can also tell you which functions are being called more | |
211 | or less often than you expected. This may help you spot bugs that had | |
212 | otherwise been unnoticed. | |
213 | ||
214 | Since the profiler uses information collected during the actual execution | |
215 | of your program, it can be used on programs that are too large or too | |
216 | complex to analyze by reading the source. However, how your program is run | |
217 | will affect the information that shows up in the profile data. If you | |
218 | don't use some feature of your program while it is being profiled, no | |
219 | profile information will be generated for that feature. | |
220 | ||
221 | Profiling has several steps: | |
222 | ||
223 | @itemize @bullet | |
224 | @item | |
225 | You must compile and link your program with profiling enabled. | |
afb17569 | 226 | @xref{Compiling, ,Compiling a Program for Profiling}. |
252b5132 RH |
227 | |
228 | @item | |
229 | You must execute your program to generate a profile data file. | |
afb17569 | 230 | @xref{Executing, ,Executing the Program}. |
252b5132 RH |
231 | |
232 | @item | |
233 | You must run @code{gprof} to analyze the profile data. | |
afb17569 | 234 | @xref{Invoking, ,@code{gprof} Command Summary}. |
252b5132 RH |
235 | @end itemize |
236 | ||
237 | The next three chapters explain these steps in greater detail. | |
238 | ||
40f90528 AM |
239 | @c man begin DESCRIPTION |
240 | ||
252b5132 RH |
241 | Several forms of output are available from the analysis. |
242 | ||
243 | The @dfn{flat profile} shows how much time your program spent in each function, | |
244 | and how many times that function was called. If you simply want to know | |
245 | which functions burn most of the cycles, it is stated concisely here. | |
afb17569 | 246 | @xref{Flat Profile, ,The Flat Profile}. |
252b5132 RH |
247 | |
248 | The @dfn{call graph} shows, for each function, which functions called it, which | |
249 | other functions it called, and how many times. There is also an estimate | |
250 | of how much time was spent in the subroutines of each function. This can | |
251 | suggest places where you might try to eliminate function calls that use a | |
afb17569 | 252 | lot of time. @xref{Call Graph, ,The Call Graph}. |
252b5132 RH |
253 | |
254 | The @dfn{annotated source} listing is a copy of the program's | |
255 | source code, labeled with the number of times each line of the | |
afb17569 BW |
256 | program was executed. @xref{Annotated Source, ,The Annotated Source |
257 | Listing}. | |
40f90528 | 258 | @c man end |
252b5132 RH |
259 | |
260 | To better understand how profiling works, you may wish to read | |
261 | a description of its implementation. | |
afb17569 | 262 | @xref{Implementation, ,Implementation of Profiling}. |
252b5132 RH |
263 | |
264 | @node Compiling | |
265 | @chapter Compiling a Program for Profiling | |
266 | ||
267 | The first step in generating profile information for your program is | |
268 | to compile and link it with profiling enabled. | |
269 | ||
270 | To compile a source file for profiling, specify the @samp{-pg} option when | |
271 | you run the compiler. (This is in addition to the options you normally | |
272 | use.) | |
273 | ||
274 | To link the program for profiling, if you use a compiler such as @code{cc} | |
275 | to do the linking, simply specify @samp{-pg} in addition to your usual | |
276 | options. The same option, @samp{-pg}, alters either compilation or linking | |
277 | to do what is necessary for profiling. Here are examples: | |
278 | ||
279 | @example | |
280 | cc -g -c myprog.c utils.c -pg | |
281 | cc -o myprog myprog.o utils.o -pg | |
282 | @end example | |
283 | ||
284 | The @samp{-pg} option also works with a command that both compiles and links: | |
285 | ||
286 | @example | |
287 | cc -o myprog myprog.c utils.c -g -pg | |
288 | @end example | |
289 | ||
83aeabb6 | 290 | Note: The @samp{-pg} option must be part of your compilation options |
83b6e7e8 NC |
291 | as well as your link options. If it is not then no call-graph data |
292 | will be gathered and when you run @code{gprof} you will get an error | |
293 | message like this: | |
83aeabb6 NC |
294 | |
295 | @example | |
296 | gprof: gmon.out file is missing call-graph data | |
297 | @end example | |
298 | ||
83b6e7e8 NC |
299 | If you add the @samp{-Q} switch to suppress the printing of the call |
300 | graph data you will still be able to see the time samples: | |
301 | ||
302 | @example | |
303 | Flat profile: | |
304 | ||
305 | Each sample counts as 0.01 seconds. | |
f3445b37 L |
306 | % cumulative self self total |
307 | time seconds seconds calls Ts/call Ts/call name | |
83b6e7e8 NC |
308 | 44.12 0.07 0.07 zazLoop |
309 | 35.29 0.14 0.06 main | |
310 | 20.59 0.17 0.04 bazMillion | |
83b6e7e8 NC |
311 | @end example |
312 | ||
252b5132 RH |
313 | If you run the linker @code{ld} directly instead of through a compiler |
314 | such as @code{cc}, you may have to specify a profiling startup file | |
315 | @file{gcrt0.o} as the first input file instead of the usual startup | |
316 | file @file{crt0.o}. In addition, you would probably want to | |
317 | specify the profiling C library, @file{libc_p.a}, by writing | |
318 | @samp{-lc_p} instead of the usual @samp{-lc}. This is not absolutely | |
319 | necessary, but doing this gives you number-of-calls information for | |
320 | standard library functions such as @code{read} and @code{open}. For | |
321 | example: | |
322 | ||
323 | @example | |
324 | ld -o myprog /lib/gcrt0.o myprog.o utils.o -lc_p | |
325 | @end example | |
326 | ||
c8d1c0af NC |
327 | If you are running the program on a system which supports shared |
328 | libraries you may run into problems with the profiling support code in | |
329 | a shared library being called before that library has been fully | |
330 | initialised. This is usually detected by the program encountering a | |
331 | segmentation fault as soon as it is run. The solution is to link | |
332 | against a static version of the library containing the profiling | |
333 | support code, which for @code{gcc} users can be done via the | |
a05a5b64 | 334 | @samp{-static} or @samp{-static-libgcc} command-line option. For |
c8d1c0af NC |
335 | example: |
336 | ||
337 | @example | |
338 | gcc -g -pg -static-libgcc myprog.c utils.c -o myprog | |
339 | @end example | |
340 | ||
252b5132 RH |
341 | If you compile only some of the modules of the program with @samp{-pg}, you |
342 | can still profile the program, but you won't get complete information about | |
343 | the modules that were compiled without @samp{-pg}. The only information | |
344 | you get for the functions in those modules is the total time spent in them; | |
345 | there is no record of how many times they were called, or from where. This | |
346 | will not affect the flat profile (except that the @code{calls} field for | |
347 | the functions will be blank), but will greatly reduce the usefulness of the | |
348 | call graph. | |
349 | ||
25c909f1 NC |
350 | If you wish to perform line-by-line profiling you should use the |
351 | @code{gcov} tool instead of @code{gprof}. See that tool's manual or | |
352 | info pages for more details of how to do this. | |
353 | ||
354 | Note, older versions of @code{gcc} produce line-by-line profiling | |
355 | information that works with @code{gprof} rather than @code{gcov} so | |
356 | there is still support for displaying this kind of information in | |
357 | @code{gprof}. @xref{Line-by-line, ,Line-by-line Profiling}. | |
358 | ||
359 | It also worth noting that @code{gcc} implements a | |
a05a5b64 | 360 | @samp{-finstrument-functions} command-line option which will insert |
83b6e7e8 NC |
361 | calls to special user supplied instrumentation routines at the entry |
362 | and exit of every function in their program. This can be used to | |
363 | implement an alternative profiling scheme. | |
364 | ||
252b5132 RH |
365 | @node Executing |
366 | @chapter Executing the Program | |
367 | ||
368 | Once the program is compiled for profiling, you must run it in order to | |
369 | generate the information that @code{gprof} needs. Simply run the program | |
370 | as usual, using the normal arguments, file names, etc. The program should | |
371 | run normally, producing the same output as usual. It will, however, run | |
afb17569 | 372 | somewhat slower than normal because of the time spent collecting and |
252b5132 RH |
373 | writing the profile data. |
374 | ||
375 | The way you run the program---the arguments and input that you give | |
376 | it---may have a dramatic effect on what the profile information shows. The | |
377 | profile data will describe the parts of the program that were activated for | |
378 | the particular input you use. For example, if the first command you give | |
379 | to your program is to quit, the profile data will show the time used in | |
380 | initialization and in cleanup, but not much else. | |
381 | ||
382 | Your program will write the profile data into a file called @file{gmon.out} | |
383 | just before exiting. If there is already a file called @file{gmon.out}, | |
bbd47c12 BA |
384 | its contents are overwritten. You can rename the file afterwards if you |
385 | are concerned that it may be overwritten. If your system libc allows you | |
386 | may be able to write the profile data under a different name. Set the | |
387 | GMON_OUT_PREFIX environment variable; this name will be appended with | |
388 | the PID of the running program. | |
252b5132 RH |
389 | |
390 | In order to write the @file{gmon.out} file properly, your program must exit | |
391 | normally: by returning from @code{main} or by calling @code{exit}. Calling | |
392 | the low-level function @code{_exit} does not write the profile data, and | |
393 | neither does abnormal termination due to an unhandled signal. | |
394 | ||
395 | The @file{gmon.out} file is written in the program's @emph{current working | |
396 | directory} at the time it exits. This means that if your program calls | |
397 | @code{chdir}, the @file{gmon.out} file will be left in the last directory | |
398 | your program @code{chdir}'d to. If you don't have permission to write in | |
399 | this directory, the file is not written, and you will get an error message. | |
400 | ||
401 | Older versions of the @sc{gnu} profiling library may also write a file | |
402 | called @file{bb.out}. This file, if present, contains an human-readable | |
403 | listing of the basic-block execution counts. Unfortunately, the | |
404 | appearance of a human-readable @file{bb.out} means the basic-block | |
405 | counts didn't get written into @file{gmon.out}. | |
406 | The Perl script @code{bbconv.pl}, included with the @code{gprof} | |
407 | source distribution, will convert a @file{bb.out} file into | |
0c24a63f UD |
408 | a format readable by @code{gprof}. Invoke it like this: |
409 | ||
410 | @smallexample | |
411 | bbconv.pl < bb.out > @var{bh-data} | |
412 | @end smallexample | |
413 | ||
414 | This translates the information in @file{bb.out} into a form that | |
415 | @code{gprof} can understand. But you still need to tell @code{gprof} | |
416 | about the existence of this translated information. To do that, include | |
417 | @var{bb-data} on the @code{gprof} command line, @emph{along with | |
418 | @file{gmon.out}}, like this: | |
419 | ||
420 | @smallexample | |
421 | gprof @var{options} @var{executable-file} gmon.out @var{bb-data} [@var{yet-more-profile-data-files}@dots{}] [> @var{outfile}] | |
422 | @end smallexample | |
252b5132 RH |
423 | |
424 | @node Invoking | |
425 | @chapter @code{gprof} Command Summary | |
426 | ||
427 | After you have a profile data file @file{gmon.out}, you can run @code{gprof} | |
428 | to interpret the information in it. The @code{gprof} program prints a | |
429 | flat profile and a call graph on standard output. Typically you would | |
430 | redirect the output of @code{gprof} into a file with @samp{>}. | |
431 | ||
432 | You run @code{gprof} like this: | |
433 | ||
434 | @smallexample | |
435 | gprof @var{options} [@var{executable-file} [@var{profile-data-files}@dots{}]] [> @var{outfile}] | |
436 | @end smallexample | |
437 | ||
438 | @noindent | |
439 | Here square-brackets indicate optional arguments. | |
440 | ||
441 | If you omit the executable file name, the file @file{a.out} is used. If | |
442 | you give no profile data file name, the file @file{gmon.out} is used. If | |
443 | any file is not in the proper format, or if the profile data file does not | |
444 | appear to belong to the executable file, an error message is printed. | |
445 | ||
446 | You can give more than one profile data file by entering all their names | |
447 | after the executable file name; then the statistics in all the data files | |
448 | are summed together. | |
449 | ||
450 | The order of these options does not matter. | |
451 | ||
452 | @menu | |
453 | * Output Options:: Controlling @code{gprof}'s output style | |
b45619c0 | 454 | * Analysis Options:: Controlling how @code{gprof} analyzes its data |
252b5132 | 455 | * Miscellaneous Options:: |
5af11cab | 456 | * Deprecated Options:: Options you no longer need to use, but which |
252b5132 RH |
457 | have been retained for compatibility |
458 | * Symspecs:: Specifying functions to include or exclude | |
459 | @end menu | |
460 | ||
afb17569 | 461 | @node Output Options |
252b5132 RH |
462 | @section Output Options |
463 | ||
40f90528 | 464 | @c man begin OPTIONS |
252b5132 RH |
465 | These options specify which of several output formats |
466 | @code{gprof} should produce. | |
467 | ||
468 | Many of these options take an optional @dfn{symspec} to specify | |
469 | functions to be included or excluded. These options can be | |
470 | specified multiple times, with different symspecs, to include | |
afb17569 | 471 | or exclude sets of symbols. @xref{Symspecs, ,Symspecs}. |
252b5132 RH |
472 | |
473 | Specifying any of these options overrides the default (@samp{-p -q}), | |
474 | which prints a flat profile and call graph analysis | |
475 | for all functions. | |
476 | ||
477 | @table @code | |
478 | ||
479 | @item -A[@var{symspec}] | |
480 | @itemx --annotated-source[=@var{symspec}] | |
481 | The @samp{-A} option causes @code{gprof} to print annotated source code. | |
482 | If @var{symspec} is specified, print output only for matching symbols. | |
afb17569 | 483 | @xref{Annotated Source, ,The Annotated Source Listing}. |
252b5132 RH |
484 | |
485 | @item -b | |
486 | @itemx --brief | |
487 | If the @samp{-b} option is given, @code{gprof} doesn't print the | |
488 | verbose blurbs that try to explain the meaning of all of the fields in | |
489 | the tables. This is useful if you intend to print out the output, or | |
490 | are tired of seeing the blurbs. | |
491 | ||
492 | @item -C[@var{symspec}] | |
493 | @itemx --exec-counts[=@var{symspec}] | |
494 | The @samp{-C} option causes @code{gprof} to | |
495 | print a tally of functions and the number of times each was called. | |
496 | If @var{symspec} is specified, print tally only for matching symbols. | |
497 | ||
5af11cab | 498 | If the profile data file contains basic-block count records, specifying |
252b5132 RH |
499 | the @samp{-l} option, along with @samp{-C}, will cause basic-block |
500 | execution counts to be tallied and displayed. | |
501 | ||
502 | @item -i | |
503 | @itemx --file-info | |
504 | The @samp{-i} option causes @code{gprof} to display summary information | |
505 | about the profile data file(s) and then exit. The number of histogram, | |
506 | call graph, and basic-block count records is displayed. | |
507 | ||
508 | @item -I @var{dirs} | |
509 | @itemx --directory-path=@var{dirs} | |
510 | The @samp{-I} option specifies a list of search directories in | |
511 | which to find source files. Environment variable @var{GPROF_PATH} | |
5af11cab | 512 | can also be used to convey this information. |
252b5132 RH |
513 | Used mostly for annotated source output. |
514 | ||
515 | @item -J[@var{symspec}] | |
516 | @itemx --no-annotated-source[=@var{symspec}] | |
517 | The @samp{-J} option causes @code{gprof} not to | |
518 | print annotated source code. | |
519 | If @var{symspec} is specified, @code{gprof} prints annotated source, | |
520 | but excludes matching symbols. | |
521 | ||
522 | @item -L | |
523 | @itemx --print-path | |
524 | Normally, source filenames are printed with the path | |
525 | component suppressed. The @samp{-L} option causes @code{gprof} | |
526 | to print the full pathname of | |
527 | source filenames, which is determined | |
528 | from symbolic debugging information in the image file | |
529 | and is relative to the directory in which the compiler | |
530 | was invoked. | |
531 | ||
532 | @item -p[@var{symspec}] | |
533 | @itemx --flat-profile[=@var{symspec}] | |
534 | The @samp{-p} option causes @code{gprof} to print a flat profile. | |
535 | If @var{symspec} is specified, print flat profile only for matching symbols. | |
afb17569 | 536 | @xref{Flat Profile, ,The Flat Profile}. |
252b5132 RH |
537 | |
538 | @item -P[@var{symspec}] | |
539 | @itemx --no-flat-profile[=@var{symspec}] | |
540 | The @samp{-P} option causes @code{gprof} to suppress printing a flat profile. | |
541 | If @var{symspec} is specified, @code{gprof} prints a flat profile, | |
542 | but excludes matching symbols. | |
543 | ||
544 | @item -q[@var{symspec}] | |
545 | @itemx --graph[=@var{symspec}] | |
546 | The @samp{-q} option causes @code{gprof} to print the call graph analysis. | |
547 | If @var{symspec} is specified, print call graph only for matching symbols | |
548 | and their children. | |
afb17569 | 549 | @xref{Call Graph, ,The Call Graph}. |
252b5132 RH |
550 | |
551 | @item -Q[@var{symspec}] | |
552 | @itemx --no-graph[=@var{symspec}] | |
553 | The @samp{-Q} option causes @code{gprof} to suppress printing the | |
554 | call graph. | |
555 | If @var{symspec} is specified, @code{gprof} prints a call graph, | |
556 | but excludes matching symbols. | |
557 | ||
a1c21132 BE |
558 | @item -t |
559 | @itemx --table-length=@var{num} | |
560 | The @samp{-t} option causes the @var{num} most active source lines in | |
561 | each source file to be listed when source annotation is enabled. The | |
562 | default is 10. | |
563 | ||
252b5132 RH |
564 | @item -y |
565 | @itemx --separate-files | |
566 | This option affects annotated source output only. | |
5af11cab | 567 | Normally, @code{gprof} prints annotated source files |
252b5132 | 568 | to standard-output. If this option is specified, |
5af11cab AM |
569 | annotated source for a file named @file{path/@var{filename}} |
570 | is generated in the file @file{@var{filename}-ann}. If the underlying | |
b45619c0 | 571 | file system would truncate @file{@var{filename}-ann} so that it |
5af11cab AM |
572 | overwrites the original @file{@var{filename}}, @code{gprof} generates |
573 | annotated source in the file @file{@var{filename}.ann} instead (if the | |
574 | original file name has an extension, that extension is @emph{replaced} | |
575 | with @file{.ann}). | |
252b5132 RH |
576 | |
577 | @item -Z[@var{symspec}] | |
578 | @itemx --no-exec-counts[=@var{symspec}] | |
579 | The @samp{-Z} option causes @code{gprof} not to | |
580 | print a tally of functions and the number of times each was called. | |
581 | If @var{symspec} is specified, print tally, but exclude matching symbols. | |
582 | ||
a1c21132 | 583 | @item -r |
242b2571 | 584 | @itemx --function-ordering |
252b5132 RH |
585 | The @samp{--function-ordering} option causes @code{gprof} to print a |
586 | suggested function ordering for the program based on profiling data. | |
587 | This option suggests an ordering which may improve paging, tlb and | |
588 | cache behavior for the program on systems which support arbitrary | |
589 | ordering of functions in an executable. | |
590 | ||
591 | The exact details of how to force the linker to place functions | |
592 | in a particular order is system dependent and out of the scope of this | |
593 | manual. | |
594 | ||
a1c21132 | 595 | @item -R @var{map_file} |
242b2571 | 596 | @itemx --file-ordering @var{map_file} |
252b5132 RH |
597 | The @samp{--file-ordering} option causes @code{gprof} to print a |
598 | suggested .o link line ordering for the program based on profiling data. | |
599 | This option suggests an ordering which may improve paging, tlb and | |
600 | cache behavior for the program on systems which do not support arbitrary | |
601 | ordering of functions in an executable. | |
602 | ||
603 | Use of the @samp{-a} argument is highly recommended with this option. | |
604 | ||
605 | The @var{map_file} argument is a pathname to a file which provides | |
606 | function name to object file mappings. The format of the file is similar to | |
607 | the output of the program @code{nm}. | |
608 | ||
609 | @smallexample | |
610 | @group | |
611 | c-parse.o:00000000 T yyparse | |
612 | c-parse.o:00000004 C yyerrflag | |
613 | c-lang.o:00000000 T maybe_objc_method_name | |
614 | c-lang.o:00000000 T print_lang_statistics | |
615 | c-lang.o:00000000 T recognize_objc_keyword | |
616 | c-decl.o:00000000 T print_lang_identifier | |
617 | c-decl.o:00000000 T print_lang_type | |
618 | @dots{} | |
619 | ||
620 | @end group | |
621 | @end smallexample | |
622 | ||
5af11cab AM |
623 | To create a @var{map_file} with @sc{gnu} @code{nm}, type a command like |
624 | @kbd{nm --extern-only --defined-only -v --print-file-name program-name}. | |
252b5132 RH |
625 | |
626 | @item -T | |
627 | @itemx --traditional | |
628 | The @samp{-T} option causes @code{gprof} to print its output in | |
629 | ``traditional'' BSD style. | |
630 | ||
631 | @item -w @var{width} | |
632 | @itemx --width=@var{width} | |
633 | Sets width of output lines to @var{width}. | |
634 | Currently only used when printing the function index at the bottom | |
635 | of the call graph. | |
636 | ||
637 | @item -x | |
638 | @itemx --all-lines | |
639 | This option affects annotated source output only. | |
640 | By default, only the lines at the beginning of a basic-block | |
641 | are annotated. If this option is specified, every line in | |
642 | a basic-block is annotated by repeating the annotation for the | |
643 | first line. This behavior is similar to @code{tcov}'s @samp{-a}. | |
644 | ||
28c309a2 | 645 | @item --demangle[=@var{style}] |
252b5132 RH |
646 | @itemx --no-demangle |
647 | These options control whether C++ symbol names should be demangled when | |
648 | printing output. The default is to demangle symbols. The | |
f3445b37 L |
649 | @code{--no-demangle} option may be used to turn off demangling. Different |
650 | compilers have different mangling styles. The optional demangling style | |
651 | argument can be used to choose an appropriate demangling style for your | |
28c309a2 | 652 | compiler. |
252b5132 RH |
653 | @end table |
654 | ||
afb17569 | 655 | @node Analysis Options |
252b5132 RH |
656 | @section Analysis Options |
657 | ||
658 | @table @code | |
659 | ||
660 | @item -a | |
661 | @itemx --no-static | |
662 | The @samp{-a} option causes @code{gprof} to suppress the printing of | |
663 | statically declared (private) functions. (These are functions whose | |
664 | names are not listed as global, and which are not visible outside the | |
665 | file/function/block where they were defined.) Time spent in these | |
b45619c0 | 666 | functions, calls to/from them, etc., will all be attributed to the |
252b5132 | 667 | function that was loaded directly before it in the executable file. |
f3445b37 | 668 | @c This is compatible with Unix @code{gprof}, but a bad idea. |
252b5132 RH |
669 | This option affects both the flat profile and the call graph. |
670 | ||
671 | @item -c | |
672 | @itemx --static-call-graph | |
673 | The @samp{-c} option causes the call graph of the program to be | |
674 | augmented by a heuristic which examines the text space of the object | |
675 | file and identifies function calls in the binary machine code. | |
676 | Since normal call graph records are only generated when functions are | |
677 | entered, this option identifies children that could have been called, | |
678 | but never were. Calls to functions that were not compiled with | |
679 | profiling enabled are also identified, but only if symbol table | |
680 | entries are present for them. | |
681 | Calls to dynamic library routines are typically @emph{not} found | |
682 | by this option. | |
683 | Parents or children identified via this heuristic | |
684 | are indicated in the call graph with call counts of @samp{0}. | |
685 | ||
686 | @item -D | |
687 | @itemx --ignore-non-functions | |
688 | The @samp{-D} option causes @code{gprof} to ignore symbols which | |
689 | are not known to be functions. This option will give more accurate | |
690 | profile data on systems where it is supported (Solaris and HPUX for | |
691 | example). | |
692 | ||
693 | @item -k @var{from}/@var{to} | |
694 | The @samp{-k} option allows you to delete from the call graph any arcs from | |
695 | symbols matching symspec @var{from} to those matching symspec @var{to}. | |
696 | ||
697 | @item -l | |
698 | @itemx --line | |
699 | The @samp{-l} option enables line-by-line profiling, which causes | |
700 | histogram hits to be charged to individual source code lines, | |
25c909f1 NC |
701 | instead of functions. This feature only works with programs compiled |
702 | by older versions of the @code{gcc} compiler. Newer versions of | |
703 | @code{gcc} are designed to work with the @code{gcov} tool instead. | |
704 | ||
252b5132 RH |
705 | If the program was compiled with basic-block counting enabled, |
706 | this option will also identify how many times each line of | |
707 | code was executed. | |
708 | While line-by-line profiling can help isolate where in a large function | |
709 | a program is spending its time, it also significantly increases | |
710 | the running time of @code{gprof}, and magnifies statistical | |
711 | inaccuracies. | |
afb17569 | 712 | @xref{Sampling Error, ,Statistical Sampling Error}. |
252b5132 | 713 | |
630b0510 CH |
714 | @item --inline-file-names |
715 | This option causes @code{gprof} to print the source file after each | |
716 | symbol in both the flat profile and the call graph. The full path to the | |
717 | file is printed if used with the @samp{-L} option. | |
718 | ||
252b5132 RH |
719 | @item -m @var{num} |
720 | @itemx --min-count=@var{num} | |
721 | This option affects execution count output only. | |
722 | Symbols that are executed less than @var{num} times are suppressed. | |
723 | ||
6bacc34d BW |
724 | @item -n@var{symspec} |
725 | @itemx --time=@var{symspec} | |
252b5132 RH |
726 | The @samp{-n} option causes @code{gprof}, in its call graph analysis, |
727 | to only propagate times for symbols matching @var{symspec}. | |
728 | ||
6bacc34d BW |
729 | @item -N@var{symspec} |
730 | @itemx --no-time=@var{symspec} | |
252b5132 RH |
731 | The @samp{-n} option causes @code{gprof}, in its call graph analysis, |
732 | not to propagate times for symbols matching @var{symspec}. | |
733 | ||
0e27a8f6 NC |
734 | @item -S@var{filename} |
735 | @itemx --external-symbol-table=@var{filename} | |
736 | The @samp{-S} option causes @code{gprof} to read an external symbol table | |
f3445b37 L |
737 | file, such as @file{/proc/kallsyms}, rather than read the symbol table |
738 | from the given object file (the default is @code{a.out}). This is useful | |
0e27a8f6 NC |
739 | for profiling kernel modules. |
740 | ||
252b5132 RH |
741 | @item -z |
742 | @itemx --display-unused-functions | |
743 | If you give the @samp{-z} option, @code{gprof} will mention all | |
744 | functions in the flat profile, even those that were never called, and | |
745 | that had no time spent in them. This is useful in conjunction with the | |
746 | @samp{-c} option for discovering which routines were never called. | |
747 | ||
748 | @end table | |
749 | ||
afb17569 | 750 | @node Miscellaneous Options |
252b5132 RH |
751 | @section Miscellaneous Options |
752 | ||
753 | @table @code | |
754 | ||
755 | @item -d[@var{num}] | |
756 | @itemx --debug[=@var{num}] | |
757 | The @samp{-d @var{num}} option specifies debugging options. | |
758 | If @var{num} is not specified, enable all debugging. | |
afb17569 | 759 | @xref{Debugging, ,Debugging @code{gprof}}. |
252b5132 | 760 | |
a1c21132 BE |
761 | @item -h |
762 | @itemx --help | |
763 | The @samp{-h} option prints command line usage. | |
764 | ||
252b5132 RH |
765 | @item -O@var{name} |
766 | @itemx --file-format=@var{name} | |
767 | Selects the format of the profile data files. Recognized formats are | |
768 | @samp{auto} (the default), @samp{bsd}, @samp{4.4bsd}, @samp{magic}, and | |
769 | @samp{prof} (not yet supported). | |
770 | ||
771 | @item -s | |
772 | @itemx --sum | |
773 | The @samp{-s} option causes @code{gprof} to summarize the information | |
774 | in the profile data files it read in, and write out a profile data | |
775 | file called @file{gmon.sum}, which contains all the information from | |
776 | the profile data files that @code{gprof} read in. The file @file{gmon.sum} | |
777 | may be one of the specified input files; the effect of this is to | |
778 | merge the data in the other input files into @file{gmon.sum}. | |
779 | ||
780 | Eventually you can run @code{gprof} again without @samp{-s} to analyze the | |
781 | cumulative data in the file @file{gmon.sum}. | |
782 | ||
783 | @item -v | |
784 | @itemx --version | |
785 | The @samp{-v} flag causes @code{gprof} to print the current version | |
786 | number, and then exit. | |
787 | ||
788 | @end table | |
789 | ||
afb17569 | 790 | @node Deprecated Options |
5af11cab | 791 | @section Deprecated Options |
252b5132 | 792 | |
252b5132 RH |
793 | These options have been replaced with newer versions that use symspecs. |
794 | ||
2e7d562e AM |
795 | @table @code |
796 | ||
252b5132 RH |
797 | @item -e @var{function_name} |
798 | The @samp{-e @var{function}} option tells @code{gprof} to not print | |
799 | information about the function @var{function_name} (and its | |
800 | children@dots{}) in the call graph. The function will still be listed | |
801 | as a child of any functions that call it, but its index number will be | |
802 | shown as @samp{[not printed]}. More than one @samp{-e} option may be | |
803 | given; only one @var{function_name} may be indicated with each @samp{-e} | |
f3445b37 | 804 | option. |
252b5132 RH |
805 | |
806 | @item -E @var{function_name} | |
807 | The @code{-E @var{function}} option works like the @code{-e} option, but | |
808 | time spent in the function (and children who were not called from | |
809 | anywhere else), will not be used to compute the percentages-of-time for | |
810 | the call graph. More than one @samp{-E} option may be given; only one | |
811 | @var{function_name} may be indicated with each @samp{-E} option. | |
812 | ||
813 | @item -f @var{function_name} | |
814 | The @samp{-f @var{function}} option causes @code{gprof} to limit the | |
815 | call graph to the function @var{function_name} and its children (and | |
816 | their children@dots{}). More than one @samp{-f} option may be given; | |
817 | only one @var{function_name} may be indicated with each @samp{-f} | |
f3445b37 | 818 | option. |
252b5132 RH |
819 | |
820 | @item -F @var{function_name} | |
821 | The @samp{-F @var{function}} option works like the @code{-f} option, but | |
822 | only time spent in the function and its children (and their | |
823 | children@dots{}) will be used to determine total-time and | |
824 | percentages-of-time for the call graph. More than one @samp{-F} option | |
825 | may be given; only one @var{function_name} may be indicated with each | |
826 | @samp{-F} option. The @samp{-F} option overrides the @samp{-E} option. | |
827 | ||
828 | @end table | |
829 | ||
40f90528 AM |
830 | @c man end |
831 | ||
252b5132 RH |
832 | Note that only one function can be specified with each @code{-e}, |
833 | @code{-E}, @code{-f} or @code{-F} option. To specify more than one | |
834 | function, use multiple options. For example, this command: | |
835 | ||
836 | @example | |
837 | gprof -e boring -f foo -f bar myprogram > gprof.output | |
838 | @end example | |
839 | ||
840 | @noindent | |
841 | lists in the call graph all functions that were reached from either | |
842 | @code{foo} or @code{bar} and were not reachable from @code{boring}. | |
843 | ||
afb17569 | 844 | @node Symspecs |
252b5132 RH |
845 | @section Symspecs |
846 | ||
847 | Many of the output options allow functions to be included or excluded | |
848 | using @dfn{symspecs} (symbol specifications), which observe the | |
849 | following syntax: | |
850 | ||
851 | @example | |
852 | filename_containing_a_dot | |
853 | | funcname_not_containing_a_dot | |
854 | | linenumber | |
855 | | ( [ any_filename ] `:' ( any_funcname | linenumber ) ) | |
856 | @end example | |
857 | ||
858 | Here are some sample symspecs: | |
859 | ||
860 | @table @samp | |
861 | @item main.c | |
862 | Selects everything in file @file{main.c}---the | |
5af11cab | 863 | dot in the string tells @code{gprof} to interpret |
252b5132 RH |
864 | the string as a filename, rather than as |
865 | a function name. To select a file whose | |
866 | name does not contain a dot, a trailing colon | |
867 | should be specified. For example, @samp{odd:} is | |
868 | interpreted as the file named @file{odd}. | |
869 | ||
870 | @item main | |
871 | Selects all functions named @samp{main}. | |
872 | ||
873 | Note that there may be multiple instances of the same function name | |
874 | because some of the definitions may be local (i.e., static). Unless a | |
875 | function name is unique in a program, you must use the colon notation | |
876 | explained below to specify a function from a specific source file. | |
877 | ||
a53f781e | 878 | Sometimes, function names contain dots. In such cases, it is necessary |
252b5132 RH |
879 | to add a leading colon to the name. For example, @samp{:.mul} selects |
880 | function @samp{.mul}. | |
881 | ||
5af11cab AM |
882 | In some object file formats, symbols have a leading underscore. |
883 | @code{gprof} will normally not print these underscores. When you name a | |
884 | symbol in a symspec, you should type it exactly as @code{gprof} prints | |
885 | it in its output. For example, if the compiler produces a symbol | |
886 | @samp{_main} from your @code{main} function, @code{gprof} still prints | |
887 | it as @samp{main} in its output, so you should use @samp{main} in | |
888 | symspecs. | |
252b5132 RH |
889 | |
890 | @item main.c:main | |
891 | Selects function @samp{main} in file @file{main.c}. | |
892 | ||
893 | @item main.c:134 | |
894 | Selects line 134 in file @file{main.c}. | |
895 | @end table | |
896 | ||
897 | @node Output | |
898 | @chapter Interpreting @code{gprof}'s Output | |
899 | ||
900 | @code{gprof} can produce several different output styles, the | |
901 | most important of which are described below. The simplest output | |
902 | styles (file information, execution count, and function and file ordering) | |
903 | are not described here, but are documented with the respective options | |
904 | that trigger them. | |
afb17569 | 905 | @xref{Output Options, ,Output Options}. |
252b5132 RH |
906 | |
907 | @menu | |
908 | * Flat Profile:: The flat profile shows how much time was spent | |
909 | executing directly in each function. | |
910 | * Call Graph:: The call graph shows which functions called which | |
911 | others, and how much time each function used | |
912 | when its subroutine calls are included. | |
913 | * Line-by-line:: @code{gprof} can analyze individual source code lines | |
914 | * Annotated Source:: The annotated source listing displays source code | |
915 | labeled with execution counts | |
916 | @end menu | |
917 | ||
918 | ||
afb17569 | 919 | @node Flat Profile |
252b5132 RH |
920 | @section The Flat Profile |
921 | @cindex flat profile | |
922 | ||
923 | The @dfn{flat profile} shows the total amount of time your program | |
924 | spent executing each function. Unless the @samp{-z} option is given, | |
925 | functions with no apparent time spent in them, and no apparent calls | |
926 | to them, are not mentioned. Note that if a function was not compiled | |
927 | for profiling, and didn't run long enough to show up on the program | |
928 | counter histogram, it will be indistinguishable from a function that | |
929 | was never called. | |
930 | ||
931 | This is part of a flat profile for a small program: | |
932 | ||
933 | @smallexample | |
934 | @group | |
935 | Flat profile: | |
936 | ||
937 | Each sample counts as 0.01 seconds. | |
f3445b37 L |
938 | % cumulative self self total |
939 | time seconds seconds calls ms/call ms/call name | |
252b5132 RH |
940 | 33.34 0.02 0.02 7208 0.00 0.00 open |
941 | 16.67 0.03 0.01 244 0.04 0.12 offtime | |
942 | 16.67 0.04 0.01 8 1.25 1.25 memccpy | |
943 | 16.67 0.05 0.01 7 1.43 1.43 write | |
944 | 16.67 0.06 0.01 mcount | |
945 | 0.00 0.06 0.00 236 0.00 0.00 tzset | |
946 | 0.00 0.06 0.00 192 0.00 0.00 tolower | |
947 | 0.00 0.06 0.00 47 0.00 0.00 strlen | |
948 | 0.00 0.06 0.00 45 0.00 0.00 strchr | |
949 | 0.00 0.06 0.00 1 0.00 50.00 main | |
950 | 0.00 0.06 0.00 1 0.00 0.00 memcpy | |
951 | 0.00 0.06 0.00 1 0.00 10.11 print | |
952 | 0.00 0.06 0.00 1 0.00 0.00 profil | |
953 | 0.00 0.06 0.00 1 0.00 50.00 report | |
954 | @dots{} | |
955 | @end group | |
956 | @end smallexample | |
957 | ||
958 | @noindent | |
afb17569 | 959 | The functions are sorted first by decreasing run-time spent in them, |
252b5132 RH |
960 | then by decreasing number of calls, then alphabetically by name. The |
961 | functions @samp{mcount} and @samp{profil} are part of the profiling | |
5af11cab | 962 | apparatus and appear in every flat profile; their time gives a measure of |
252b5132 RH |
963 | the amount of overhead due to profiling. |
964 | ||
965 | Just before the column headers, a statement appears indicating | |
966 | how much time each sample counted as. | |
967 | This @dfn{sampling period} estimates the margin of error in each of the time | |
968 | figures. A time figure that is not much larger than this is not | |
969 | reliable. In this example, each sample counted as 0.01 seconds, | |
970 | suggesting a 100 Hz sampling rate. | |
971 | The program's total execution time was 0.06 | |
972 | seconds, as indicated by the @samp{cumulative seconds} field. Since | |
973 | each sample counted for 0.01 seconds, this means only six samples | |
5af11cab | 974 | were taken during the run. Two of the samples occurred while the |
252b5132 RH |
975 | program was in the @samp{open} function, as indicated by the |
976 | @samp{self seconds} field. Each of the other four samples | |
5af11cab | 977 | occurred one each in @samp{offtime}, @samp{memccpy}, @samp{write}, |
252b5132 RH |
978 | and @samp{mcount}. |
979 | Since only six samples were taken, none of these values can | |
980 | be regarded as particularly reliable. | |
981 | In another run, | |
982 | the @samp{self seconds} field for | |
983 | @samp{mcount} might well be @samp{0.00} or @samp{0.02}. | |
afb17569 BW |
984 | @xref{Sampling Error, ,Statistical Sampling Error}, |
985 | for a complete discussion. | |
252b5132 RH |
986 | |
987 | The remaining functions in the listing (those whose | |
988 | @samp{self seconds} field is @samp{0.00}) didn't appear | |
989 | in the histogram samples at all. However, the call graph | |
990 | indicated that they were called, so therefore they are listed, | |
991 | sorted in decreasing order by the @samp{calls} field. | |
992 | Clearly some time was spent executing these functions, | |
993 | but the paucity of histogram samples prevents any | |
994 | determination of how much time each took. | |
995 | ||
996 | Here is what the fields in each line mean: | |
997 | ||
998 | @table @code | |
999 | @item % time | |
1000 | This is the percentage of the total execution time your program spent | |
1001 | in this function. These should all add up to 100%. | |
1002 | ||
1003 | @item cumulative seconds | |
1004 | This is the cumulative total number of seconds the computer spent | |
1005 | executing this functions, plus the time spent in all the functions | |
1006 | above this one in this table. | |
1007 | ||
1008 | @item self seconds | |
1009 | This is the number of seconds accounted for by this function alone. | |
1010 | The flat profile listing is sorted first by this number. | |
1011 | ||
1012 | @item calls | |
1013 | This is the total number of times the function was called. If the | |
1014 | function was never called, or the number of times it was called cannot | |
1015 | be determined (probably because the function was not compiled with | |
1016 | profiling enabled), the @dfn{calls} field is blank. | |
1017 | ||
1018 | @item self ms/call | |
1019 | This represents the average number of milliseconds spent in this | |
1020 | function per call, if this function is profiled. Otherwise, this field | |
1021 | is blank for this function. | |
1022 | ||
1023 | @item total ms/call | |
1024 | This represents the average number of milliseconds spent in this | |
1025 | function and its descendants per call, if this function is profiled. | |
1026 | Otherwise, this field is blank for this function. | |
1027 | This is the only field in the flat profile that uses call graph analysis. | |
1028 | ||
1029 | @item name | |
1030 | This is the name of the function. The flat profile is sorted by this | |
1031 | field alphabetically after the @dfn{self seconds} and @dfn{calls} | |
1032 | fields are sorted. | |
1033 | @end table | |
1034 | ||
afb17569 | 1035 | @node Call Graph |
252b5132 RH |
1036 | @section The Call Graph |
1037 | @cindex call graph | |
1038 | ||
1039 | The @dfn{call graph} shows how much time was spent in each function | |
1040 | and its children. From this information, you can find functions that, | |
1041 | while they themselves may not have used much time, called other | |
1042 | functions that did use unusual amounts of time. | |
1043 | ||
1044 | Here is a sample call from a small program. This call came from the | |
1045 | same @code{gprof} run as the flat profile example in the previous | |
afb17569 | 1046 | section. |
252b5132 RH |
1047 | |
1048 | @smallexample | |
1049 | @group | |
1050 | granularity: each sample hit covers 2 byte(s) for 20.00% of 0.05 seconds | |
1051 | ||
1052 | index % time self children called name | |
1053 | <spontaneous> | |
1054 | [1] 100.0 0.00 0.05 start [1] | |
1055 | 0.00 0.05 1/1 main [2] | |
1056 | 0.00 0.00 1/2 on_exit [28] | |
1057 | 0.00 0.00 1/1 exit [59] | |
1058 | ----------------------------------------------- | |
1059 | 0.00 0.05 1/1 start [1] | |
1060 | [2] 100.0 0.00 0.05 1 main [2] | |
1061 | 0.00 0.05 1/1 report [3] | |
1062 | ----------------------------------------------- | |
1063 | 0.00 0.05 1/1 main [2] | |
1064 | [3] 100.0 0.00 0.05 1 report [3] | |
1065 | 0.00 0.03 8/8 timelocal [6] | |
1066 | 0.00 0.01 1/1 print [9] | |
1067 | 0.00 0.01 9/9 fgets [12] | |
1068 | 0.00 0.00 12/34 strncmp <cycle 1> [40] | |
1069 | 0.00 0.00 8/8 lookup [20] | |
1070 | 0.00 0.00 1/1 fopen [21] | |
1071 | 0.00 0.00 8/8 chewtime [24] | |
1072 | 0.00 0.00 8/16 skipspace [44] | |
1073 | ----------------------------------------------- | |
afb17569 | 1074 | [4] 59.8 0.01 0.02 8+472 <cycle 2 as a whole> [4] |
252b5132 RH |
1075 | 0.01 0.02 244+260 offtime <cycle 2> [7] |
1076 | 0.00 0.00 236+1 tzset <cycle 2> [26] | |
1077 | ----------------------------------------------- | |
1078 | @end group | |
1079 | @end smallexample | |
1080 | ||
1081 | The lines full of dashes divide this table into @dfn{entries}, one for each | |
1082 | function. Each entry has one or more lines. | |
1083 | ||
1084 | In each entry, the primary line is the one that starts with an index number | |
1085 | in square brackets. The end of this line says which function the entry is | |
1086 | for. The preceding lines in the entry describe the callers of this | |
1087 | function and the following lines describe its subroutines (also called | |
1088 | @dfn{children} when we speak of the call graph). | |
1089 | ||
1090 | The entries are sorted by time spent in the function and its subroutines. | |
1091 | ||
f3445b37 | 1092 | The internal profiling function @code{mcount} (@pxref{Flat Profile, ,The |
afb17569 | 1093 | Flat Profile}) is never mentioned in the call graph. |
252b5132 RH |
1094 | |
1095 | @menu | |
1096 | * Primary:: Details of the primary line's contents. | |
1097 | * Callers:: Details of caller-lines' contents. | |
1098 | * Subroutines:: Details of subroutine-lines' contents. | |
1099 | * Cycles:: When there are cycles of recursion, | |
1100 | such as @code{a} calls @code{b} calls @code{a}@dots{} | |
1101 | @end menu | |
1102 | ||
1103 | @node Primary | |
1104 | @subsection The Primary Line | |
1105 | ||
1106 | The @dfn{primary line} in a call graph entry is the line that | |
1107 | describes the function which the entry is about and gives the overall | |
1108 | statistics for this function. | |
1109 | ||
1110 | For reference, we repeat the primary line from the entry for function | |
1111 | @code{report} in our main example, together with the heading line that | |
1112 | shows the names of the fields: | |
1113 | ||
1114 | @smallexample | |
1115 | @group | |
1116 | index % time self children called name | |
1117 | @dots{} | |
1118 | [3] 100.0 0.00 0.05 1 report [3] | |
1119 | @end group | |
1120 | @end smallexample | |
1121 | ||
1122 | Here is what the fields in the primary line mean: | |
1123 | ||
1124 | @table @code | |
1125 | @item index | |
1126 | Entries are numbered with consecutive integers. Each function | |
1127 | therefore has an index number, which appears at the beginning of its | |
1128 | primary line. | |
1129 | ||
1130 | Each cross-reference to a function, as a caller or subroutine of | |
1131 | another, gives its index number as well as its name. The index number | |
1132 | guides you if you wish to look for the entry for that function. | |
1133 | ||
1134 | @item % time | |
1135 | This is the percentage of the total time that was spent in this | |
1136 | function, including time spent in subroutines called from this | |
1137 | function. | |
1138 | ||
1139 | The time spent in this function is counted again for the callers of | |
1140 | this function. Therefore, adding up these percentages is meaningless. | |
1141 | ||
1142 | @item self | |
1143 | This is the total amount of time spent in this function. This | |
1144 | should be identical to the number printed in the @code{seconds} field | |
1145 | for this function in the flat profile. | |
1146 | ||
1147 | @item children | |
1148 | This is the total amount of time spent in the subroutine calls made by | |
1149 | this function. This should be equal to the sum of all the @code{self} | |
1150 | and @code{children} entries of the children listed directly below this | |
1151 | function. | |
1152 | ||
1153 | @item called | |
1154 | This is the number of times the function was called. | |
1155 | ||
1156 | If the function called itself recursively, there are two numbers, | |
1157 | separated by a @samp{+}. The first number counts non-recursive calls, | |
1158 | and the second counts recursive calls. | |
1159 | ||
1160 | In the example above, the function @code{report} was called once from | |
1161 | @code{main}. | |
1162 | ||
1163 | @item name | |
1164 | This is the name of the current function. The index number is | |
1165 | repeated after it. | |
1166 | ||
1167 | If the function is part of a cycle of recursion, the cycle number is | |
1168 | printed between the function's name and the index number | |
afb17569 BW |
1169 | (@pxref{Cycles, ,How Mutually Recursive Functions Are Described}). |
1170 | For example, if function @code{gnurr} is part of | |
252b5132 RH |
1171 | cycle number one, and has index number twelve, its primary line would |
1172 | be end like this: | |
1173 | ||
1174 | @example | |
1175 | gnurr <cycle 1> [12] | |
1176 | @end example | |
1177 | @end table | |
1178 | ||
afb17569 | 1179 | @node Callers |
252b5132 RH |
1180 | @subsection Lines for a Function's Callers |
1181 | ||
1182 | A function's entry has a line for each function it was called by. | |
1183 | These lines' fields correspond to the fields of the primary line, but | |
1184 | their meanings are different because of the difference in context. | |
1185 | ||
1186 | For reference, we repeat two lines from the entry for the function | |
1187 | @code{report}, the primary line and one caller-line preceding it, together | |
1188 | with the heading line that shows the names of the fields: | |
1189 | ||
1190 | @smallexample | |
1191 | index % time self children called name | |
1192 | @dots{} | |
1193 | 0.00 0.05 1/1 main [2] | |
1194 | [3] 100.0 0.00 0.05 1 report [3] | |
1195 | @end smallexample | |
1196 | ||
1197 | Here are the meanings of the fields in the caller-line for @code{report} | |
1198 | called from @code{main}: | |
1199 | ||
1200 | @table @code | |
1201 | @item self | |
1202 | An estimate of the amount of time spent in @code{report} itself when it was | |
1203 | called from @code{main}. | |
1204 | ||
1205 | @item children | |
1206 | An estimate of the amount of time spent in subroutines of @code{report} | |
1207 | when @code{report} was called from @code{main}. | |
1208 | ||
1209 | The sum of the @code{self} and @code{children} fields is an estimate | |
1210 | of the amount of time spent within calls to @code{report} from @code{main}. | |
1211 | ||
1212 | @item called | |
1213 | Two numbers: the number of times @code{report} was called from @code{main}, | |
5af11cab | 1214 | followed by the total number of non-recursive calls to @code{report} from |
252b5132 RH |
1215 | all its callers. |
1216 | ||
1217 | @item name and index number | |
1218 | The name of the caller of @code{report} to which this line applies, | |
1219 | followed by the caller's index number. | |
1220 | ||
1221 | Not all functions have entries in the call graph; some | |
1222 | options to @code{gprof} request the omission of certain functions. | |
1223 | When a caller has no entry of its own, it still has caller-lines | |
1224 | in the entries of the functions it calls. | |
1225 | ||
1226 | If the caller is part of a recursion cycle, the cycle number is | |
1227 | printed between the name and the index number. | |
1228 | @end table | |
1229 | ||
1230 | If the identity of the callers of a function cannot be determined, a | |
1231 | dummy caller-line is printed which has @samp{<spontaneous>} as the | |
1232 | ``caller's name'' and all other fields blank. This can happen for | |
1233 | signal handlers. | |
1234 | @c What if some calls have determinable callers' names but not all? | |
1235 | @c FIXME - still relevant? | |
1236 | ||
afb17569 | 1237 | @node Subroutines |
252b5132 RH |
1238 | @subsection Lines for a Function's Subroutines |
1239 | ||
1240 | A function's entry has a line for each of its subroutines---in other | |
1241 | words, a line for each other function that it called. These lines' | |
1242 | fields correspond to the fields of the primary line, but their meanings | |
1243 | are different because of the difference in context. | |
1244 | ||
1245 | For reference, we repeat two lines from the entry for the function | |
1246 | @code{main}, the primary line and a line for a subroutine, together | |
1247 | with the heading line that shows the names of the fields: | |
1248 | ||
1249 | @smallexample | |
1250 | index % time self children called name | |
1251 | @dots{} | |
1252 | [2] 100.0 0.00 0.05 1 main [2] | |
1253 | 0.00 0.05 1/1 report [3] | |
1254 | @end smallexample | |
1255 | ||
1256 | Here are the meanings of the fields in the subroutine-line for @code{main} | |
1257 | calling @code{report}: | |
1258 | ||
1259 | @table @code | |
1260 | @item self | |
1261 | An estimate of the amount of time spent directly within @code{report} | |
1262 | when @code{report} was called from @code{main}. | |
1263 | ||
1264 | @item children | |
1265 | An estimate of the amount of time spent in subroutines of @code{report} | |
1266 | when @code{report} was called from @code{main}. | |
1267 | ||
1268 | The sum of the @code{self} and @code{children} fields is an estimate | |
1269 | of the total time spent in calls to @code{report} from @code{main}. | |
1270 | ||
1271 | @item called | |
1272 | Two numbers, the number of calls to @code{report} from @code{main} | |
5af11cab | 1273 | followed by the total number of non-recursive calls to @code{report}. |
252b5132 RH |
1274 | This ratio is used to determine how much of @code{report}'s @code{self} |
1275 | and @code{children} time gets credited to @code{main}. | |
afb17569 | 1276 | @xref{Assumptions, ,Estimating @code{children} Times}. |
252b5132 RH |
1277 | |
1278 | @item name | |
1279 | The name of the subroutine of @code{main} to which this line applies, | |
1280 | followed by the subroutine's index number. | |
1281 | ||
1282 | If the caller is part of a recursion cycle, the cycle number is | |
1283 | printed between the name and the index number. | |
1284 | @end table | |
1285 | ||
afb17569 | 1286 | @node Cycles |
252b5132 RH |
1287 | @subsection How Mutually Recursive Functions Are Described |
1288 | @cindex cycle | |
1289 | @cindex recursion cycle | |
1290 | ||
1291 | The graph may be complicated by the presence of @dfn{cycles of | |
1292 | recursion} in the call graph. A cycle exists if a function calls | |
1293 | another function that (directly or indirectly) calls (or appears to | |
1294 | call) the original function. For example: if @code{a} calls @code{b}, | |
1295 | and @code{b} calls @code{a}, then @code{a} and @code{b} form a cycle. | |
1296 | ||
1297 | Whenever there are call paths both ways between a pair of functions, they | |
1298 | belong to the same cycle. If @code{a} and @code{b} call each other and | |
1299 | @code{b} and @code{c} call each other, all three make one cycle. Note that | |
1300 | even if @code{b} only calls @code{a} if it was not called from @code{a}, | |
1301 | @code{gprof} cannot determine this, so @code{a} and @code{b} are still | |
1302 | considered a cycle. | |
1303 | ||
1304 | The cycles are numbered with consecutive integers. When a function | |
1305 | belongs to a cycle, each time the function name appears in the call graph | |
1306 | it is followed by @samp{<cycle @var{number}>}. | |
1307 | ||
1308 | The reason cycles matter is that they make the time values in the call | |
1309 | graph paradoxical. The ``time spent in children'' of @code{a} should | |
1310 | include the time spent in its subroutine @code{b} and in @code{b}'s | |
1311 | subroutines---but one of @code{b}'s subroutines is @code{a}! How much of | |
1312 | @code{a}'s time should be included in the children of @code{a}, when | |
1313 | @code{a} is indirectly recursive? | |
1314 | ||
1315 | The way @code{gprof} resolves this paradox is by creating a single entry | |
1316 | for the cycle as a whole. The primary line of this entry describes the | |
1317 | total time spent directly in the functions of the cycle. The | |
1318 | ``subroutines'' of the cycle are the individual functions of the cycle, and | |
1319 | all other functions that were called directly by them. The ``callers'' of | |
1320 | the cycle are the functions, outside the cycle, that called functions in | |
1321 | the cycle. | |
1322 | ||
1323 | Here is an example portion of a call graph which shows a cycle containing | |
1324 | functions @code{a} and @code{b}. The cycle was entered by a call to | |
1325 | @code{a} from @code{main}; both @code{a} and @code{b} called @code{c}. | |
1326 | ||
1327 | @smallexample | |
1328 | index % time self children called name | |
1329 | ---------------------------------------- | |
1330 | 1.77 0 1/1 main [2] | |
1331 | [3] 91.71 1.77 0 1+5 <cycle 1 as a whole> [3] | |
1332 | 1.02 0 3 b <cycle 1> [4] | |
1333 | 0.75 0 2 a <cycle 1> [5] | |
1334 | ---------------------------------------- | |
1335 | 3 a <cycle 1> [5] | |
1336 | [4] 52.85 1.02 0 0 b <cycle 1> [4] | |
1337 | 2 a <cycle 1> [5] | |
1338 | 0 0 3/6 c [6] | |
1339 | ---------------------------------------- | |
1340 | 1.77 0 1/1 main [2] | |
1341 | 2 b <cycle 1> [4] | |
1342 | [5] 38.86 0.75 0 1 a <cycle 1> [5] | |
1343 | 3 b <cycle 1> [4] | |
1344 | 0 0 3/6 c [6] | |
1345 | ---------------------------------------- | |
1346 | @end smallexample | |
1347 | ||
1348 | @noindent | |
1349 | (The entire call graph for this program contains in addition an entry for | |
1350 | @code{main}, which calls @code{a}, and an entry for @code{c}, with callers | |
1351 | @code{a} and @code{b}.) | |
1352 | ||
1353 | @smallexample | |
1354 | index % time self children called name | |
1355 | <spontaneous> | |
1356 | [1] 100.00 0 1.93 0 start [1] | |
1357 | 0.16 1.77 1/1 main [2] | |
1358 | ---------------------------------------- | |
1359 | 0.16 1.77 1/1 start [1] | |
1360 | [2] 100.00 0.16 1.77 1 main [2] | |
1361 | 1.77 0 1/1 a <cycle 1> [5] | |
1362 | ---------------------------------------- | |
1363 | 1.77 0 1/1 main [2] | |
1364 | [3] 91.71 1.77 0 1+5 <cycle 1 as a whole> [3] | |
1365 | 1.02 0 3 b <cycle 1> [4] | |
1366 | 0.75 0 2 a <cycle 1> [5] | |
1367 | 0 0 6/6 c [6] | |
1368 | ---------------------------------------- | |
1369 | 3 a <cycle 1> [5] | |
1370 | [4] 52.85 1.02 0 0 b <cycle 1> [4] | |
1371 | 2 a <cycle 1> [5] | |
1372 | 0 0 3/6 c [6] | |
1373 | ---------------------------------------- | |
1374 | 1.77 0 1/1 main [2] | |
1375 | 2 b <cycle 1> [4] | |
1376 | [5] 38.86 0.75 0 1 a <cycle 1> [5] | |
1377 | 3 b <cycle 1> [4] | |
1378 | 0 0 3/6 c [6] | |
1379 | ---------------------------------------- | |
1380 | 0 0 3/6 b <cycle 1> [4] | |
1381 | 0 0 3/6 a <cycle 1> [5] | |
1382 | [6] 0.00 0 0 6 c [6] | |
1383 | ---------------------------------------- | |
1384 | @end smallexample | |
1385 | ||
1386 | The @code{self} field of the cycle's primary line is the total time | |
1387 | spent in all the functions of the cycle. It equals the sum of the | |
1388 | @code{self} fields for the individual functions in the cycle, found | |
1389 | in the entry in the subroutine lines for these functions. | |
1390 | ||
1391 | The @code{children} fields of the cycle's primary line and subroutine lines | |
1392 | count only subroutines outside the cycle. Even though @code{a} calls | |
1393 | @code{b}, the time spent in those calls to @code{b} is not counted in | |
1394 | @code{a}'s @code{children} time. Thus, we do not encounter the problem of | |
1395 | what to do when the time in those calls to @code{b} includes indirect | |
1396 | recursive calls back to @code{a}. | |
1397 | ||
1398 | The @code{children} field of a caller-line in the cycle's entry estimates | |
1399 | the amount of time spent @emph{in the whole cycle}, and its other | |
1400 | subroutines, on the times when that caller called a function in the cycle. | |
1401 | ||
afb17569 | 1402 | The @code{called} field in the primary line for the cycle has two numbers: |
252b5132 RH |
1403 | first, the number of times functions in the cycle were called by functions |
1404 | outside the cycle; second, the number of times they were called by | |
1405 | functions in the cycle (including times when a function in the cycle calls | |
5af11cab | 1406 | itself). This is a generalization of the usual split into non-recursive and |
252b5132 RH |
1407 | recursive calls. |
1408 | ||
afb17569 | 1409 | The @code{called} field of a subroutine-line for a cycle member in the |
252b5132 RH |
1410 | cycle's entry says how many time that function was called from functions in |
1411 | the cycle. The total of all these is the second number in the primary line's | |
afb17569 | 1412 | @code{called} field. |
252b5132 RH |
1413 | |
1414 | In the individual entry for a function in a cycle, the other functions in | |
1415 | the same cycle can appear as subroutines and as callers. These lines show | |
1416 | how many times each function in the cycle called or was called from each other | |
1417 | function in the cycle. The @code{self} and @code{children} fields in these | |
1418 | lines are blank because of the difficulty of defining meanings for them | |
1419 | when recursion is going on. | |
1420 | ||
afb17569 | 1421 | @node Line-by-line |
252b5132 RH |
1422 | @section Line-by-line Profiling |
1423 | ||
1424 | @code{gprof}'s @samp{-l} option causes the program to perform | |
1425 | @dfn{line-by-line} profiling. In this mode, histogram | |
1426 | samples are assigned not to functions, but to individual | |
25c909f1 NC |
1427 | lines of source code. This only works with programs compiled with |
1428 | older versions of the @code{gcc} compiler. Newer versions of @code{gcc} | |
1429 | use a different program - @code{gcov} - to display line-by-line | |
1430 | profiling information. | |
1431 | ||
1432 | With the older versions of @code{gcc} the program usually has to be | |
1433 | compiled with a @samp{-g} option, in addition to @samp{-pg}, in order | |
252b5132 | 1434 | to generate debugging symbols for tracking source code lines. |
25c909f1 | 1435 | Note, in much older versions of @code{gcc} the program had to be |
a05a5b64 | 1436 | compiled with the @samp{-a} command-line option as well. |
252b5132 RH |
1437 | |
1438 | The flat profile is the most useful output table | |
1439 | in line-by-line mode. | |
1440 | The call graph isn't as useful as normal, since | |
1441 | the current version of @code{gprof} does not propagate | |
1442 | call graph arcs from source code lines to the enclosing function. | |
1443 | The call graph does, however, show each line of code | |
1444 | that called each function, along with a count. | |
1445 | ||
1446 | Here is a section of @code{gprof}'s output, without line-by-line profiling. | |
1447 | Note that @code{ct_init} accounted for four histogram hits, and | |
1448 | 13327 calls to @code{init_block}. | |
1449 | ||
1450 | @smallexample | |
1451 | Flat profile: | |
1452 | ||
1453 | Each sample counts as 0.01 seconds. | |
f3445b37 L |
1454 | % cumulative self self total |
1455 | time seconds seconds calls us/call us/call name | |
252b5132 RH |
1456 | 30.77 0.13 0.04 6335 6.31 6.31 ct_init |
1457 | ||
1458 | ||
1459 | Call graph (explanation follows) | |
1460 | ||
1461 | ||
1462 | granularity: each sample hit covers 4 byte(s) for 7.69% of 0.13 seconds | |
1463 | ||
1464 | index % time self children called name | |
1465 | ||
1466 | 0.00 0.00 1/13496 name_too_long | |
1467 | 0.00 0.00 40/13496 deflate | |
1468 | 0.00 0.00 128/13496 deflate_fast | |
1469 | 0.00 0.00 13327/13496 ct_init | |
1470 | [7] 0.0 0.00 0.00 13496 init_block | |
1471 | ||
1472 | @end smallexample | |
1473 | ||
1474 | Now let's look at some of @code{gprof}'s output from the same program run, | |
1475 | this time with line-by-line profiling enabled. Note that @code{ct_init}'s | |
afb17569 | 1476 | four histogram hits are broken down into four lines of source code---one hit |
5af11cab | 1477 | occurred on each of lines 349, 351, 382 and 385. In the call graph, |
252b5132 RH |
1478 | note how |
1479 | @code{ct_init}'s 13327 calls to @code{init_block} are broken down | |
1480 | into one call from line 396, 3071 calls from line 384, 3730 calls | |
1481 | from line 385, and 6525 calls from 387. | |
1482 | ||
1483 | @smallexample | |
1484 | Flat profile: | |
1485 | ||
1486 | Each sample counts as 0.01 seconds. | |
f3445b37 L |
1487 | % cumulative self |
1488 | time seconds seconds calls name | |
252b5132 RH |
1489 | 7.69 0.10 0.01 ct_init (trees.c:349) |
1490 | 7.69 0.11 0.01 ct_init (trees.c:351) | |
1491 | 7.69 0.12 0.01 ct_init (trees.c:382) | |
1492 | 7.69 0.13 0.01 ct_init (trees.c:385) | |
1493 | ||
1494 | ||
1495 | Call graph (explanation follows) | |
1496 | ||
1497 | ||
1498 | granularity: each sample hit covers 4 byte(s) for 7.69% of 0.13 seconds | |
1499 | ||
1500 | % time self children called name | |
1501 | ||
1502 | 0.00 0.00 1/13496 name_too_long (gzip.c:1440) | |
1503 | 0.00 0.00 1/13496 deflate (deflate.c:763) | |
1504 | 0.00 0.00 1/13496 ct_init (trees.c:396) | |
1505 | 0.00 0.00 2/13496 deflate (deflate.c:727) | |
1506 | 0.00 0.00 4/13496 deflate (deflate.c:686) | |
1507 | 0.00 0.00 5/13496 deflate (deflate.c:675) | |
1508 | 0.00 0.00 12/13496 deflate (deflate.c:679) | |
1509 | 0.00 0.00 16/13496 deflate (deflate.c:730) | |
1510 | 0.00 0.00 128/13496 deflate_fast (deflate.c:654) | |
1511 | 0.00 0.00 3071/13496 ct_init (trees.c:384) | |
1512 | 0.00 0.00 3730/13496 ct_init (trees.c:385) | |
1513 | 0.00 0.00 6525/13496 ct_init (trees.c:387) | |
1514 | [6] 0.0 0.00 0.00 13496 init_block (trees.c:408) | |
1515 | ||
1516 | @end smallexample | |
1517 | ||
1518 | ||
afb17569 | 1519 | @node Annotated Source |
252b5132 RH |
1520 | @section The Annotated Source Listing |
1521 | ||
1522 | @code{gprof}'s @samp{-A} option triggers an annotated source listing, | |
1523 | which lists the program's source code, each function labeled with the | |
1524 | number of times it was called. You may also need to specify the | |
1525 | @samp{-I} option, if @code{gprof} can't find the source code files. | |
1526 | ||
25c909f1 NC |
1527 | With older versions of @code{gcc} compiling with @samp{gcc @dots{} -g |
1528 | -pg -a} augments your program with basic-block counting code, in | |
1529 | addition to function counting code. This enables @code{gprof} to | |
1530 | determine how many times each line of code was executed. With newer | |
1531 | versions of @code{gcc} support for displaying basic-block counts is | |
1532 | provided by the @code{gcov} program. | |
1533 | ||
252b5132 RH |
1534 | For example, consider the following function, taken from gzip, |
1535 | with line numbers added: | |
1536 | ||
1537 | @smallexample | |
1538 | 1 ulg updcrc(s, n) | |
1539 | 2 uch *s; | |
1540 | 3 unsigned n; | |
1541 | 4 @{ | |
1542 | 5 register ulg c; | |
1543 | 6 | |
1544 | 7 static ulg crc = (ulg)0xffffffffL; | |
1545 | 8 | |
1546 | 9 if (s == NULL) @{ | |
1547 | 10 c = 0xffffffffL; | |
1548 | 11 @} else @{ | |
1549 | 12 c = crc; | |
1550 | 13 if (n) do @{ | |
1551 | 14 c = crc_32_tab[...]; | |
1552 | 15 @} while (--n); | |
1553 | 16 @} | |
1554 | 17 crc = c; | |
1555 | 18 return c ^ 0xffffffffL; | |
1556 | 19 @} | |
1557 | ||
1558 | @end smallexample | |
1559 | ||
1560 | @code{updcrc} has at least five basic-blocks. | |
1561 | One is the function itself. The | |
1562 | @code{if} statement on line 9 generates two more basic-blocks, one | |
1563 | for each branch of the @code{if}. A fourth basic-block results from | |
1564 | the @code{if} on line 13, and the contents of the @code{do} loop form | |
1565 | the fifth basic-block. The compiler may also generate additional | |
1566 | basic-blocks to handle various special cases. | |
1567 | ||
1568 | A program augmented for basic-block counting can be analyzed with | |
afb17569 BW |
1569 | @samp{gprof -l -A}. |
1570 | The @samp{-x} option is also helpful, | |
1571 | to ensure that each line of code is labeled at least once. | |
252b5132 RH |
1572 | Here is @code{updcrc}'s |
1573 | annotated source listing for a sample @code{gzip} run: | |
1574 | ||
1575 | @smallexample | |
1576 | ulg updcrc(s, n) | |
1577 | uch *s; | |
1578 | unsigned n; | |
1579 | 2 ->@{ | |
1580 | register ulg c; | |
f3445b37 | 1581 | |
252b5132 | 1582 | static ulg crc = (ulg)0xffffffffL; |
f3445b37 | 1583 | |
252b5132 | 1584 | 2 -> if (s == NULL) @{ |
afb17569 | 1585 | 1 -> c = 0xffffffffL; |
252b5132 | 1586 | 1 -> @} else @{ |
afb17569 | 1587 | 1 -> c = crc; |
252b5132 RH |
1588 | 1 -> if (n) do @{ |
1589 | 26312 -> c = crc_32_tab[...]; | |
1590 | 26312,1,26311 -> @} while (--n); | |
1591 | @} | |
1592 | 2 -> crc = c; | |
1593 | 2 -> return c ^ 0xffffffffL; | |
1594 | 2 ->@} | |
1595 | @end smallexample | |
1596 | ||
1597 | In this example, the function was called twice, passing once through | |
1598 | each branch of the @code{if} statement. The body of the @code{do} | |
1599 | loop was executed a total of 26312 times. Note how the @code{while} | |
1600 | statement is annotated. It began execution 26312 times, once for | |
1601 | each iteration through the loop. One of those times (the last time) | |
1602 | it exited, while it branched back to the beginning of the loop 26311 times. | |
1603 | ||
1604 | @node Inaccuracy | |
1605 | @chapter Inaccuracy of @code{gprof} Output | |
1606 | ||
1607 | @menu | |
1608 | * Sampling Error:: Statistical margins of error | |
1609 | * Assumptions:: Estimating children times | |
1610 | @end menu | |
1611 | ||
afb17569 | 1612 | @node Sampling Error |
252b5132 RH |
1613 | @section Statistical Sampling Error |
1614 | ||
1615 | The run-time figures that @code{gprof} gives you are based on a sampling | |
1616 | process, so they are subject to statistical inaccuracy. If a function runs | |
1617 | only a small amount of time, so that on the average the sampling process | |
1618 | ought to catch that function in the act only once, there is a pretty good | |
1619 | chance it will actually find that function zero times, or twice. | |
1620 | ||
ede501f4 NC |
1621 | By contrast, the number-of-calls and basic-block figures are derived |
1622 | by counting, not sampling. They are completely accurate and will not | |
1623 | vary from run to run if your program is deterministic and single | |
1624 | threaded. In multi-threaded applications, or single threaded | |
1625 | applications that link with multi-threaded libraries, the counts are | |
1626 | only deterministic if the counting function is thread-safe. (Note: | |
1627 | beware that the mcount counting function in glibc is @emph{not} | |
1628 | thread-safe). @xref{Implementation, ,Implementation of Profiling}. | |
252b5132 RH |
1629 | |
1630 | The @dfn{sampling period} that is printed at the beginning of the flat | |
1631 | profile says how often samples are taken. The rule of thumb is that a | |
1632 | run-time figure is accurate if it is considerably bigger than the sampling | |
1633 | period. | |
1634 | ||
1635 | The actual amount of error can be predicted. | |
1636 | For @var{n} samples, the @emph{expected} error | |
1637 | is the square-root of @var{n}. For example, | |
1638 | if the sampling period is 0.01 seconds and @code{foo}'s run-time is 1 second, | |
1639 | @var{n} is 100 samples (1 second/0.01 seconds), sqrt(@var{n}) is 10 samples, so | |
1640 | the expected error in @code{foo}'s run-time is 0.1 seconds (10*0.01 seconds), | |
1641 | or ten percent of the observed value. | |
1642 | Again, if the sampling period is 0.01 seconds and @code{bar}'s run-time is | |
1643 | 100 seconds, @var{n} is 10000 samples, sqrt(@var{n}) is 100 samples, so | |
1644 | the expected error in @code{bar}'s run-time is 1 second, | |
1645 | or one percent of the observed value. | |
1646 | It is likely to | |
1647 | vary this much @emph{on the average} from one profiling run to the next. | |
1648 | (@emph{Sometimes} it will vary more.) | |
1649 | ||
1650 | This does not mean that a small run-time figure is devoid of information. | |
1651 | If the program's @emph{total} run-time is large, a small run-time for one | |
1652 | function does tell you that that function used an insignificant fraction of | |
1653 | the whole program's time. Usually this means it is not worth optimizing. | |
1654 | ||
1655 | One way to get more accuracy is to give your program more (but similar) | |
1656 | input data so it will take longer. Another way is to combine the data from | |
1657 | several runs, using the @samp{-s} option of @code{gprof}. Here is how: | |
1658 | ||
1659 | @enumerate | |
1660 | @item | |
1661 | Run your program once. | |
1662 | ||
1663 | @item | |
1664 | Issue the command @samp{mv gmon.out gmon.sum}. | |
1665 | ||
1666 | @item | |
1667 | Run your program again, the same as before. | |
1668 | ||
1669 | @item | |
1670 | Merge the new data in @file{gmon.out} into @file{gmon.sum} with this command: | |
1671 | ||
1672 | @example | |
1673 | gprof -s @var{executable-file} gmon.out gmon.sum | |
1674 | @end example | |
1675 | ||
1676 | @item | |
1677 | Repeat the last two steps as often as you wish. | |
1678 | ||
1679 | @item | |
1680 | Analyze the cumulative data using this command: | |
1681 | ||
1682 | @example | |
1683 | gprof @var{executable-file} gmon.sum > @var{output-file} | |
1684 | @end example | |
1685 | @end enumerate | |
1686 | ||
afb17569 | 1687 | @node Assumptions |
252b5132 RH |
1688 | @section Estimating @code{children} Times |
1689 | ||
1690 | Some of the figures in the call graph are estimates---for example, the | |
1be59579 | 1691 | @code{children} time values and all the time figures in caller and |
252b5132 RH |
1692 | subroutine lines. |
1693 | ||
1694 | There is no direct information about these measurements in the profile | |
1695 | data itself. Instead, @code{gprof} estimates them by making an assumption | |
1696 | about your program that might or might not be true. | |
1697 | ||
1698 | The assumption made is that the average time spent in each call to any | |
1699 | function @code{foo} is not correlated with who called @code{foo}. If | |
1700 | @code{foo} used 5 seconds in all, and 2/5 of the calls to @code{foo} came | |
1701 | from @code{a}, then @code{foo} contributes 2 seconds to @code{a}'s | |
1702 | @code{children} time, by assumption. | |
1703 | ||
1704 | This assumption is usually true enough, but for some programs it is far | |
1705 | from true. Suppose that @code{foo} returns very quickly when its argument | |
1706 | is zero; suppose that @code{a} always passes zero as an argument, while | |
1707 | other callers of @code{foo} pass other arguments. In this program, all the | |
1708 | time spent in @code{foo} is in the calls from callers other than @code{a}. | |
1709 | But @code{gprof} has no way of knowing this; it will blindly and | |
1710 | incorrectly charge 2 seconds of time in @code{foo} to the children of | |
1711 | @code{a}. | |
1712 | ||
1713 | @c FIXME - has this been fixed? | |
1714 | We hope some day to put more complete data into @file{gmon.out}, so that | |
1715 | this assumption is no longer needed, if we can figure out how. For the | |
afb17569 | 1716 | novice, the estimated figures are usually more useful than misleading. |
252b5132 RH |
1717 | |
1718 | @node How do I? | |
1719 | @chapter Answers to Common Questions | |
1720 | ||
1721 | @table @asis | |
83aeabb6 NC |
1722 | @item How can I get more exact information about hot spots in my program? |
1723 | ||
1724 | Looking at the per-line call counts only tells part of the story. | |
1725 | Because @code{gprof} can only report call times and counts by function, | |
1726 | the best way to get finer-grained information on where the program | |
1727 | is spending its time is to re-factor large functions into sequences | |
83b6e7e8 | 1728 | of calls to smaller ones. Beware however that this can introduce |
b45619c0 | 1729 | artificial hot spots since compiling with @samp{-pg} adds a significant |
83b6e7e8 NC |
1730 | overhead to function calls. An alternative solution is to use a |
1731 | non-intrusive profiler, e.g.@: oprofile. | |
83aeabb6 | 1732 | |
252b5132 RH |
1733 | @item How do I find which lines in my program were executed the most times? |
1734 | ||
25c909f1 | 1735 | Use the @code{gcov} program. |
252b5132 RH |
1736 | |
1737 | @item How do I find which lines in my program called a particular function? | |
1738 | ||
5af11cab | 1739 | Use @samp{gprof -l} and lookup the function in the call graph. |
252b5132 RH |
1740 | The callers will be broken down by function and line number. |
1741 | ||
1742 | @item How do I analyze a program that runs for less than a second? | |
1743 | ||
1744 | Try using a shell script like this one: | |
1745 | ||
1746 | @example | |
1747 | for i in `seq 1 100`; do | |
1748 | fastprog | |
1749 | mv gmon.out gmon.out.$i | |
1750 | done | |
1751 | ||
1752 | gprof -s fastprog gmon.out.* | |
1753 | ||
1754 | gprof fastprog gmon.sum | |
1755 | @end example | |
1756 | ||
1757 | If your program is completely deterministic, all the call counts | |
b45619c0 | 1758 | will be simple multiples of 100 (i.e., a function called once in |
252b5132 RH |
1759 | each run will appear with a call count of 100). |
1760 | ||
1761 | @end table | |
1762 | ||
1763 | @node Incompatibilities | |
1764 | @chapter Incompatibilities with Unix @code{gprof} | |
1765 | ||
1766 | @sc{gnu} @code{gprof} and Berkeley Unix @code{gprof} use the same data | |
1767 | file @file{gmon.out}, and provide essentially the same information. But | |
1768 | there are a few differences. | |
1769 | ||
1770 | @itemize @bullet | |
1771 | @item | |
1772 | @sc{gnu} @code{gprof} uses a new, generalized file format with support | |
1773 | for basic-block execution counts and non-realtime histograms. A magic | |
1774 | cookie and version number allows @code{gprof} to easily identify | |
1775 | new style files. Old BSD-style files can still be read. | |
afb17569 | 1776 | @xref{File Format, ,Profiling Data File Format}. |
252b5132 RH |
1777 | |
1778 | @item | |
1779 | For a recursive function, Unix @code{gprof} lists the function as a | |
1780 | parent and as a child, with a @code{calls} field that lists the number | |
1781 | of recursive calls. @sc{gnu} @code{gprof} omits these lines and puts | |
1782 | the number of recursive calls in the primary line. | |
1783 | ||
1784 | @item | |
1785 | When a function is suppressed from the call graph with @samp{-e}, @sc{gnu} | |
1786 | @code{gprof} still lists it as a subroutine of functions that call it. | |
1787 | ||
1788 | @item | |
1789 | @sc{gnu} @code{gprof} accepts the @samp{-k} with its argument | |
1790 | in the form @samp{from/to}, instead of @samp{from to}. | |
1791 | ||
1792 | @item | |
1793 | In the annotated source listing, | |
1794 | if there are multiple basic blocks on the same line, | |
5af11cab | 1795 | @sc{gnu} @code{gprof} prints all of their counts, separated by commas. |
252b5132 RH |
1796 | |
1797 | @ignore - it does this now | |
1798 | @item | |
1799 | The function names printed in @sc{gnu} @code{gprof} output do not include | |
1800 | the leading underscores that are added internally to the front of all | |
1801 | C identifiers on many operating systems. | |
1802 | @end ignore | |
1803 | ||
1804 | @item | |
1805 | The blurbs, field widths, and output formats are different. @sc{gnu} | |
1806 | @code{gprof} prints blurbs after the tables, so that you can see the | |
1807 | tables without skipping the blurbs. | |
1808 | @end itemize | |
1809 | ||
1810 | @node Details | |
1811 | @chapter Details of Profiling | |
1812 | ||
1813 | @menu | |
5af11cab | 1814 | * Implementation:: How a program collects profiling information |
252b5132 RH |
1815 | * File Format:: Format of @samp{gmon.out} files |
1816 | * Internals:: @code{gprof}'s internal operation | |
1817 | * Debugging:: Using @code{gprof}'s @samp{-d} option | |
1818 | @end menu | |
1819 | ||
afb17569 | 1820 | @node Implementation |
252b5132 RH |
1821 | @section Implementation of Profiling |
1822 | ||
1823 | Profiling works by changing how every function in your program is compiled | |
1824 | so that when it is called, it will stash away some information about where | |
1825 | it was called from. From this, the profiler can figure out what function | |
1826 | called it, and can count how many times it was called. This change is made | |
1827 | by the compiler when your program is compiled with the @samp{-pg} option, | |
1828 | which causes every function to call @code{mcount} | |
1829 | (or @code{_mcount}, or @code{__mcount}, depending on the OS and compiler) | |
1830 | as one of its first operations. | |
1831 | ||
1832 | The @code{mcount} routine, included in the profiling library, | |
1833 | is responsible for recording in an in-memory call graph table | |
1834 | both its parent routine (the child) and its parent's parent. This is | |
1835 | typically done by examining the stack frame to find both | |
1836 | the address of the child, and the return address in the original parent. | |
5af11cab | 1837 | Since this is a very machine-dependent operation, @code{mcount} |
252b5132 RH |
1838 | itself is typically a short assembly-language stub routine |
1839 | that extracts the required | |
1840 | information, and then calls @code{__mcount_internal} | |
afb17569 | 1841 | (a normal C function) with two arguments---@code{frompc} and @code{selfpc}. |
252b5132 RH |
1842 | @code{__mcount_internal} is responsible for maintaining |
1843 | the in-memory call graph, which records @code{frompc}, @code{selfpc}, | |
5af11cab | 1844 | and the number of times each of these call arcs was traversed. |
252b5132 RH |
1845 | |
1846 | GCC Version 2 provides a magical function (@code{__builtin_return_address}), | |
1847 | which allows a generic @code{mcount} function to extract the | |
1848 | required information from the stack frame. However, on some | |
1849 | architectures, most notably the SPARC, using this builtin can be | |
1850 | very computationally expensive, and an assembly language version | |
1851 | of @code{mcount} is used for performance reasons. | |
1852 | ||
1853 | Number-of-calls information for library routines is collected by using a | |
1854 | special version of the C library. The programs in it are the same as in | |
1855 | the usual C library, but they were compiled with @samp{-pg}. If you | |
1856 | link your program with @samp{gcc @dots{} -pg}, it automatically uses the | |
1857 | profiling version of the library. | |
1858 | ||
1859 | Profiling also involves watching your program as it runs, and keeping a | |
1860 | histogram of where the program counter happens to be every now and then. | |
1861 | Typically the program counter is looked at around 100 times per second of | |
1862 | run time, but the exact frequency may vary from system to system. | |
1863 | ||
1864 | This is done is one of two ways. Most UNIX-like operating systems | |
1865 | provide a @code{profil()} system call, which registers a memory | |
1866 | array with the kernel, along with a scale | |
1867 | factor that determines how the program's address space maps | |
1868 | into the array. | |
1869 | Typical scaling values cause every 2 to 8 bytes of address space | |
1870 | to map into a single array slot. | |
1871 | On every tick of the system clock | |
1872 | (assuming the profiled program is running), the value of the | |
1873 | program counter is examined and the corresponding slot in | |
1874 | the memory array is incremented. Since this is done in the kernel, | |
1875 | which had to interrupt the process anyway to handle the clock | |
1876 | interrupt, very little additional system overhead is required. | |
1877 | ||
1878 | However, some operating systems, most notably Linux 2.0 (and earlier), | |
1879 | do not provide a @code{profil()} system call. On such a system, | |
1880 | arrangements are made for the kernel to periodically deliver | |
1881 | a signal to the process (typically via @code{setitimer()}), | |
1882 | which then performs the same operation of examining the | |
1883 | program counter and incrementing a slot in the memory array. | |
1884 | Since this method requires a signal to be delivered to | |
1885 | user space every time a sample is taken, it uses considerably | |
1886 | more overhead than kernel-based profiling. Also, due to the | |
1887 | added delay required to deliver the signal, this method is | |
1888 | less accurate as well. | |
1889 | ||
f3445b37 | 1890 | A special startup routine allocates memory for the histogram and |
252b5132 RH |
1891 | either calls @code{profil()} or sets up |
1892 | a clock signal handler. | |
1893 | This routine (@code{monstartup}) can be invoked in several ways. | |
1894 | On Linux systems, a special profiling startup file @code{gcrt0.o}, | |
1895 | which invokes @code{monstartup} before @code{main}, | |
1896 | is used instead of the default @code{crt0.o}. | |
1897 | Use of this special startup file is one of the effects | |
1898 | of using @samp{gcc @dots{} -pg} to link. | |
1899 | On SPARC systems, no special startup files are used. | |
1900 | Rather, the @code{mcount} routine, when it is invoked for | |
1901 | the first time (typically when @code{main} is called), | |
1902 | calls @code{monstartup}. | |
1903 | ||
1904 | If the compiler's @samp{-a} option was used, basic-block counting | |
1905 | is also enabled. Each object file is then compiled with a static array | |
1906 | of counts, initially zero. | |
1907 | In the executable code, every time a new basic-block begins | |
afb17569 | 1908 | (i.e., when an @code{if} statement appears), an extra instruction |
252b5132 RH |
1909 | is inserted to increment the corresponding count in the array. |
1910 | At compile time, a paired array was constructed that recorded | |
1911 | the starting address of each basic-block. Taken together, | |
1912 | the two arrays record the starting address of every basic-block, | |
1913 | along with the number of times it was executed. | |
1914 | ||
1915 | The profiling library also includes a function (@code{mcleanup}) which is | |
1916 | typically registered using @code{atexit()} to be called as the | |
1917 | program exits, and is responsible for writing the file @file{gmon.out}. | |
1918 | Profiling is turned off, various headers are output, and the histogram | |
1919 | is written, followed by the call-graph arcs and the basic-block counts. | |
1920 | ||
1921 | The output from @code{gprof} gives no indication of parts of your program that | |
1922 | are limited by I/O or swapping bandwidth. This is because samples of the | |
1923 | program counter are taken at fixed intervals of the program's run time. | |
1924 | Therefore, the | |
1925 | time measurements in @code{gprof} output say nothing about time that your | |
1926 | program was not running. For example, a part of the program that creates | |
1927 | so much data that it cannot all fit in physical memory at once may run very | |
1928 | slowly due to thrashing, but @code{gprof} will say it uses little time. On | |
1929 | the other hand, sampling by run time has the advantage that the amount of | |
1930 | load due to other users won't directly affect the output you get. | |
1931 | ||
afb17569 | 1932 | @node File Format |
252b5132 RH |
1933 | @section Profiling Data File Format |
1934 | ||
1935 | The old BSD-derived file format used for profile data does not contain a | |
f9a6a8f0 | 1936 | magic cookie that allows one to check whether a data file really is a |
5af11cab | 1937 | @code{gprof} file. Furthermore, it does not provide a version number, thus |
252b5132 RH |
1938 | rendering changes to the file format almost impossible. @sc{gnu} @code{gprof} |
1939 | uses a new file format that provides these features. For backward | |
1940 | compatibility, @sc{gnu} @code{gprof} continues to support the old BSD-derived | |
1941 | format, but not all features are supported with it. For example, | |
1942 | basic-block execution counts cannot be accommodated by the old file | |
1943 | format. | |
1944 | ||
1945 | The new file format is defined in header file @file{gmon_out.h}. It | |
1946 | consists of a header containing the magic cookie and a version number, | |
1947 | as well as some spare bytes available for future extensions. All data | |
dbdec02b NC |
1948 | in a profile data file is in the native format of the target for which |
1949 | the profile was collected. @sc{gnu} @code{gprof} adapts automatically | |
1950 | to the byte-order in use. | |
252b5132 RH |
1951 | |
1952 | In the new file format, the header is followed by a sequence of | |
1953 | records. Currently, there are three different record types: histogram | |
1954 | records, call-graph arc records, and basic-block execution count | |
1955 | records. Each file can contain any number of each record type. When | |
1956 | reading a file, @sc{gnu} @code{gprof} will ensure records of the same type are | |
1957 | compatible with each other and compute the union of all records. For | |
1958 | example, for basic-block execution counts, the union is simply the sum | |
1959 | of all execution counts for each basic-block. | |
1960 | ||
1961 | @subsection Histogram Records | |
1962 | ||
1963 | Histogram records consist of a header that is followed by an array of | |
1964 | bins. The header contains the text-segment range that the histogram | |
1965 | spans, the size of the histogram in bytes (unlike in the old BSD | |
1966 | format, this does not include the size of the header), the rate of the | |
1967 | profiling clock, and the physical dimension that the bin counts | |
1968 | represent after being scaled by the profiling clock rate. The | |
1969 | physical dimension is specified in two parts: a long name of up to 15 | |
1970 | characters and a single character abbreviation. For example, a | |
1971 | histogram representing real-time would specify the long name as | |
afb17569 | 1972 | ``seconds'' and the abbreviation as ``s''. This feature is useful for |
252b5132 RH |
1973 | architectures that support performance monitor hardware (which, |
1974 | fortunately, is becoming increasingly common). For example, under DEC | |
afb17569 | 1975 | OSF/1, the ``uprofile'' command can be used to produce a histogram of, |
252b5132 | 1976 | say, instruction cache misses. In this case, the dimension in the |
afb17569 BW |
1977 | histogram header could be set to ``i-cache misses'' and the abbreviation |
1978 | could be set to ``1'' (because it is simply a count, not a physical | |
252b5132 RH |
1979 | dimension). Also, the profiling rate would have to be set to 1 in |
1980 | this case. | |
1981 | ||
1982 | Histogram bins are 16-bit numbers and each bin represent an equal | |
1983 | amount of text-space. For example, if the text-segment is one | |
1984 | thousand bytes long and if there are ten bins in the histogram, each | |
1985 | bin represents one hundred bytes. | |
1986 | ||
1987 | ||
1988 | @subsection Call-Graph Records | |
1989 | ||
1990 | Call-graph records have a format that is identical to the one used in | |
1991 | the BSD-derived file format. It consists of an arc in the call graph | |
1992 | and a count indicating the number of times the arc was traversed | |
1993 | during program execution. Arcs are specified by a pair of addresses: | |
1994 | the first must be within caller's function and the second must be | |
1995 | within the callee's function. When performing profiling at the | |
1996 | function level, these addresses can point anywhere within the | |
1997 | respective function. However, when profiling at the line-level, it is | |
1998 | better if the addresses are as close to the call-site/entry-point as | |
1999 | possible. This will ensure that the line-level call-graph is able to | |
2000 | identify exactly which line of source code performed calls to a | |
2001 | function. | |
2002 | ||
2003 | @subsection Basic-Block Execution Count Records | |
2004 | ||
2005 | Basic-block execution count records consist of a header followed by a | |
2006 | sequence of address/count pairs. The header simply specifies the | |
2007 | length of the sequence. In an address/count pair, the address | |
2008 | identifies a basic-block and the count specifies the number of times | |
2009 | that basic-block was executed. Any address within the basic-address can | |
2010 | be used. | |
2011 | ||
afb17569 | 2012 | @node Internals |
252b5132 RH |
2013 | @section @code{gprof}'s Internal Operation |
2014 | ||
2015 | Like most programs, @code{gprof} begins by processing its options. | |
2016 | During this stage, it may building its symspec list | |
afb17569 | 2017 | (@code{sym_ids.c:@-sym_id_add}), if |
252b5132 RH |
2018 | options are specified which use symspecs. |
2019 | @code{gprof} maintains a single linked list of symspecs, | |
2020 | which will eventually get turned into 12 symbol tables, | |
afb17569 | 2021 | organized into six include/exclude pairs---one |
252b5132 RH |
2022 | pair each for the flat profile (INCL_FLAT/EXCL_FLAT), |
2023 | the call graph arcs (INCL_ARCS/EXCL_ARCS), | |
2024 | printing in the call graph (INCL_GRAPH/EXCL_GRAPH), | |
2025 | timing propagation in the call graph (INCL_TIME/EXCL_TIME), | |
2026 | the annotated source listing (INCL_ANNO/EXCL_ANNO), | |
2027 | and the execution count listing (INCL_EXEC/EXCL_EXEC). | |
2028 | ||
2029 | After option processing, @code{gprof} finishes | |
2030 | building the symspec list by adding all the symspecs in | |
2031 | @code{default_excluded_list} to the exclude lists | |
2032 | EXCL_TIME and EXCL_GRAPH, and if line-by-line profiling is specified, | |
2033 | EXCL_FLAT as well. | |
2034 | These default excludes are not added to EXCL_ANNO, EXCL_ARCS, and EXCL_EXEC. | |
2035 | ||
2036 | Next, the BFD library is called to open the object file, | |
2037 | verify that it is an object file, | |
afb17569 | 2038 | and read its symbol table (@code{core.c:@-core_init}), |
252b5132 | 2039 | using @code{bfd_canonicalize_symtab} after mallocing |
5af11cab | 2040 | an appropriately sized array of symbols. At this point, |
252b5132 RH |
2041 | function mappings are read (if the @samp{--file-ordering} option |
2042 | has been specified), and the core text space is read into | |
2043 | memory (if the @samp{-c} option was given). | |
2044 | ||
2045 | @code{gprof}'s own symbol table, an array of Sym structures, | |
2046 | is now built. | |
2047 | This is done in one of two ways, by one of two routines, depending | |
2048 | on whether line-by-line profiling (@samp{-l} option) has been | |
2049 | enabled. | |
2050 | For normal profiling, the BFD canonical symbol table is scanned. | |
2051 | For line-by-line profiling, every | |
2052 | text space address is examined, and a new symbol table entry | |
2053 | gets created every time the line number changes. | |
2054 | In either case, two passes are made through the symbol | |
afb17569 | 2055 | table---one to count the size of the symbol table required, |
252b5132 RH |
2056 | and the other to actually read the symbols. In between the |
2057 | two passes, a single array of type @code{Sym} is created of | |
5af11cab | 2058 | the appropriate length. |
afb17569 | 2059 | Finally, @code{symtab.c:@-symtab_finalize} |
252b5132 RH |
2060 | is called to sort the symbol table and remove duplicate entries |
2061 | (entries with the same memory address). | |
2062 | ||
2063 | The symbol table must be a contiguous array for two reasons. | |
2064 | First, the @code{qsort} library function (which sorts an array) | |
2065 | will be used to sort the symbol table. | |
afb17569 | 2066 | Also, the symbol lookup routine (@code{symtab.c:@-sym_lookup}), |
252b5132 RH |
2067 | which finds symbols |
2068 | based on memory address, uses a binary search algorithm | |
2069 | which requires the symbol table to be a sorted array. | |
2070 | Function symbols are indicated with an @code{is_func} flag. | |
2071 | Line number symbols have no special flags set. | |
2072 | Additionally, a symbol can have an @code{is_static} flag | |
2073 | to indicate that it is a local symbol. | |
2074 | ||
2075 | With the symbol table read, the symspecs can now be translated | |
afb17569 | 2076 | into Syms (@code{sym_ids.c:@-sym_id_parse}). Remember that a single |
252b5132 RH |
2077 | symspec can match multiple symbols. |
2078 | An array of symbol tables | |
2079 | (@code{syms}) is created, each entry of which is a symbol table | |
2080 | of Syms to be included or excluded from a particular listing. | |
2081 | The master symbol table and the symspecs are examined by nested | |
2082 | loops, and every symbol that matches a symspec is inserted | |
2083 | into the appropriate syms table. This is done twice, once to | |
2084 | count the size of each required symbol table, and again to build | |
2085 | the tables, which have been malloced between passes. | |
2086 | From now on, to determine whether a symbol is on an include | |
2087 | or exclude symspec list, @code{gprof} simply uses its | |
2088 | standard symbol lookup routine on the appropriate table | |
2089 | in the @code{syms} array. | |
2090 | ||
2091 | Now the profile data file(s) themselves are read | |
afb17569 | 2092 | (@code{gmon_io.c:@-gmon_out_read}), |
252b5132 RH |
2093 | first by checking for a new-style @samp{gmon.out} header, |
2094 | then assuming this is an old-style BSD @samp{gmon.out} | |
2095 | if the magic number test failed. | |
2096 | ||
afb17569 | 2097 | New-style histogram records are read by @code{hist.c:@-hist_read_rec}. |
252b5132 RH |
2098 | For the first histogram record, allocate a memory array to hold |
2099 | all the bins, and read them in. | |
2100 | When multiple profile data files (or files with multiple histogram | |
b3296dc5 VP |
2101 | records) are read, the memory ranges of each pair of histogram records |
2102 | must be either equal, or non-overlapping. For each pair of histogram | |
2103 | records, the resolution (memory region size divided by the number of | |
f3445b37 | 2104 | bins) must be the same. The time unit must be the same for all |
b3296dc5 VP |
2105 | histogram records. If the above containts are met, all histograms |
2106 | for the same memory range are merged. | |
252b5132 | 2107 | |
afb17569 | 2108 | As each call graph record is read (@code{call_graph.c:@-cg_read_rec}), |
252b5132 RH |
2109 | the parent and child addresses |
2110 | are matched to symbol table entries, and a call graph arc is | |
afb17569 | 2111 | created by @code{cg_arcs.c:@-arc_add}, unless the arc fails a symspec |
252b5132 RH |
2112 | check against INCL_ARCS/EXCL_ARCS. As each arc is added, |
2113 | a linked list is maintained of the parent's child arcs, and of the child's | |
2114 | parent arcs. | |
2115 | Both the child's call count and the arc's call count are | |
2116 | incremented by the record's call count. | |
2117 | ||
afb17569 | 2118 | Basic-block records are read (@code{basic_blocks.c:@-bb_read_rec}), |
252b5132 RH |
2119 | but only if line-by-line profiling has been selected. |
2120 | Each basic-block address is matched to a corresponding line | |
2121 | symbol in the symbol table, and an entry made in the symbol's | |
2122 | bb_addr and bb_calls arrays. Again, if multiple basic-block | |
2123 | records are present for the same address, the call counts | |
2124 | are cumulative. | |
2125 | ||
afb17569 | 2126 | A gmon.sum file is dumped, if requested (@code{gmon_io.c:@-gmon_out_write}). |
252b5132 RH |
2127 | |
2128 | If histograms were present in the data files, assign them to symbols | |
afb17569 | 2129 | (@code{hist.c:@-hist_assign_samples}) by iterating over all the sample |
252b5132 RH |
2130 | bins and assigning them to symbols. Since the symbol table |
2131 | is sorted in order of ascending memory addresses, we can | |
2132 | simple follow along in the symbol table as we make our pass | |
2133 | over the sample bins. | |
2134 | This step includes a symspec check against INCL_FLAT/EXCL_FLAT. | |
2135 | Depending on the histogram | |
2136 | scale factor, a sample bin may span multiple symbols, | |
2137 | in which case a fraction of the sample count is allocated | |
2138 | to each symbol, proportional to the degree of overlap. | |
2139 | This effect is rare for normal profiling, but overlaps | |
2140 | are more common during line-by-line profiling, and can | |
2141 | cause each of two adjacent lines to be credited with half | |
2142 | a hit, for example. | |
2143 | ||
afb17569 | 2144 | If call graph data is present, @code{cg_arcs.c:@-cg_assemble} is called. |
5af11cab | 2145 | First, if @samp{-c} was specified, a machine-dependent |
252b5132 RH |
2146 | routine (@code{find_call}) scans through each symbol's machine code, |
2147 | looking for subroutine call instructions, and adding them | |
2148 | to the call graph with a zero call count. | |
2149 | A topological sort is performed by depth-first numbering | |
afb17569 | 2150 | all the symbols (@code{cg_dfn.c:@-cg_dfn}), so that |
252b5132 RH |
2151 | children are always numbered less than their parents, |
2152 | then making a array of pointers into the symbol table and sorting it into | |
2153 | numerical order, which is reverse topological | |
2154 | order (children appear before parents). | |
2155 | Cycles are also detected at this point, all members | |
2156 | of which are assigned the same topological number. | |
2157 | Two passes are now made through this sorted array of symbol pointers. | |
2158 | The first pass, from end to beginning (parents to children), | |
5af11cab | 2159 | computes the fraction of child time to propagate to each parent |
252b5132 RH |
2160 | and a print flag. |
2161 | The print flag reflects symspec handling of INCL_GRAPH/EXCL_GRAPH, | |
2162 | with a parent's include or exclude (print or no print) property | |
2163 | being propagated to its children, unless they themselves explicitly appear | |
2164 | in INCL_GRAPH or EXCL_GRAPH. | |
2165 | A second pass, from beginning to end (children to parents) actually | |
5af11cab | 2166 | propagates the timings along the call graph, subject |
252b5132 RH |
2167 | to a check against INCL_TIME/EXCL_TIME. |
2168 | With the print flag, fractions, and timings now stored in the symbol | |
2169 | structures, the topological sort array is now discarded, and a | |
2170 | new array of pointers is assembled, this time sorted by propagated time. | |
2171 | ||
2172 | Finally, print the various outputs the user requested, which is now fairly | |
afb17569 BW |
2173 | straightforward. The call graph (@code{cg_print.c:@-cg_print}) and |
2174 | flat profile (@code{hist.c:@-hist_print}) are regurgitations of values | |
252b5132 | 2175 | already computed. The annotated source listing |
afb17569 | 2176 | (@code{basic_blocks.c:@-print_annotated_source}) uses basic-block |
252b5132 RH |
2177 | information, if present, to label each line of code with call counts, |
2178 | otherwise only the function call counts are presented. | |
2179 | ||
2180 | The function ordering code is marginally well documented | |
2181 | in the source code itself (@code{cg_print.c}). Basically, | |
2182 | the functions with the most use and the most parents are | |
2183 | placed first, followed by other functions with the most use, | |
2184 | followed by lower use functions, followed by unused functions | |
2185 | at the end. | |
2186 | ||
afb17569 | 2187 | @node Debugging |
19c6af1e | 2188 | @section Debugging @code{gprof} |
252b5132 RH |
2189 | |
2190 | If @code{gprof} was compiled with debugging enabled, | |
2191 | the @samp{-d} option triggers debugging output | |
2192 | (to stdout) which can be helpful in understanding its operation. | |
2193 | The debugging number specified is interpreted as a sum of the following | |
2194 | options: | |
2195 | ||
2196 | @table @asis | |
2197 | @item 2 - Topological sort | |
2198 | Monitor depth-first numbering of symbols during call graph analysis | |
2199 | @item 4 - Cycles | |
2200 | Shows symbols as they are identified as cycle heads | |
2201 | @item 16 - Tallying | |
2202 | As the call graph arcs are read, show each arc and how | |
2203 | the total calls to each function are tallied | |
2204 | @item 32 - Call graph arc sorting | |
2205 | Details sorting individual parents/children within each call graph entry | |
2206 | @item 64 - Reading histogram and call graph records | |
2207 | Shows address ranges of histograms as they are read, and each | |
2208 | call graph arc | |
2209 | @item 128 - Symbol table | |
2210 | Reading, classifying, and sorting the symbol table from the object file. | |
2211 | For line-by-line profiling (@samp{-l} option), also shows line numbers | |
2212 | being assigned to memory addresses. | |
2213 | @item 256 - Static call graph | |
2214 | Trace operation of @samp{-c} option | |
2215 | @item 512 - Symbol table and arc table lookups | |
2216 | Detail operation of lookup routines | |
2217 | @item 1024 - Call graph propagation | |
2218 | Shows how function times are propagated along the call graph | |
2219 | @item 2048 - Basic-blocks | |
2220 | Shows basic-block records as they are read from profile data | |
2221 | (only meaningful with @samp{-l} option) | |
2222 | @item 4096 - Symspecs | |
2223 | Shows symspec-to-symbol pattern matching operation | |
2224 | @item 8192 - Annotate source | |
2225 | Tracks operation of @samp{-A} option | |
2226 | @end table | |
2227 | ||
cf055d54 | 2228 | @node GNU Free Documentation License |
afb17569 | 2229 | @appendix GNU Free Documentation License |
793c5807 | 2230 | @include fdl.texi |
cf055d54 | 2231 | |
252b5132 RH |
2232 | @bye |
2233 | ||
2234 | NEEDS AN INDEX | |
2235 | ||
2236 | -T - "traditional BSD style": How is it different? Should the | |
2237 | differences be documented? | |
2238 | ||
2239 | example flat file adds up to 100.01%... | |
2240 | ||
2241 | note: time estimates now only go out to one decimal place (0.0), where | |
2242 | they used to extend two (78.67). |