1 .\" Copyright (c) 1983, 1990, 1993
2 .\" The Regents of the University of California. All rights reserved.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice, this list of conditions and the following disclaimer.
9 .\" 2. Redistributions in binary form must reproduce the above copyright
10 .\" notice, this list of conditions and the following disclaimer in the
11 .\" documentation and/or other materials provided with the distribution.
12 .\" 4. Neither the name of the University nor the names of its contributors
13 .\" may be used to endorse or promote products derived from this software
14 .\" without specific prior written permission.
16 .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
17 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
20 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 .\" @(#)gprof.1 8.1 (Berkeley) 6/6/93
36 .Nd display call graph profile data
45 .Op Fl k Ar fromname toname
46 .Op Ar a.out Op Ar a.out.gmon ...
50 utility produces an execution profile of C, Pascal, or Fortran77 programs.
51 The effect of called routines is incorporated in the profile of each caller.
52 The profile data is taken from the call graph profile file
53 which is created by programs that are compiled with the
62 option also links in versions of the library routines
63 that are compiled for profiling.
64 By convention these libraries have their name suffixed with
66 i.e., the profiled version of
70 and if you specify libraries directly to the
71 compiler or linker you can use
75 Read the given object file (the default is
77 and establishes the relation between its symbol table
78 and the call graph profile.
79 The default graph profile file name is the name
80 of the executable with the suffix
83 If more than one profile file is specified,
86 output shows the sum of the profile information in the given profile files.
90 utility calculates the amount of time spent in each routine.
91 Next, these times are propagated along the edges of the call graph.
92 Cycles are discovered, and calls into a cycle are made to share the time
94 The first listing shows the functions
95 sorted according to the time they represent
96 including the time of their call graph descendants.
97 Below each function entry is shown its (direct) call graph children,
98 and how their times are propagated to this function.
99 A similar display above the function shows how this function's time and the
100 time of its descendants is propagated to its (direct) call graph parents.
102 Cycles are also shown, with an entry for the cycle as a whole and
103 a listing of the members of the cycle and their contributions to the
104 time and call counts of the cycle.
106 Second, a flat profile is given,
107 similar to that provided by
109 This listing gives the total execution times, the call counts,
110 the time that the call spent in the routine itself, and
111 the time that the call spent in the routine itself including
113 The units for the per-call times are normally milliseconds,
114 but they are nanoseconds if the profiling clock frequency
115 is 10 million or larger,
116 and if a function appears to be never called then its total self time
117 is printed as a percentage in the self time per call column.
118 The very high profiling clock frequencies needed to get sufficient
119 accuracy in the per-call times for short-lived programs are only
122 (non-statistical) kernel profiling.
124 Finally, an index of the function names is provided.
126 The following options are available:
127 .Bl -tag -width indent
129 Suppress the printing of statically declared functions.
130 If this option is given, all relevant information about the static function
131 (e.g., time samples, calls to other functions, calls from other functions)
132 belongs to the function loaded just before the static function in the
136 Suppress the printing of a description of each field in the profile.
138 Find a minimal set of arcs that can be broken to eliminate all cycles with
141 Caution: the algorithm used to break cycles is exponential,
142 so using this option may cause
144 to run for a very long time.
146 Suppress the printing of the graph profile entry for routine
148 and all its descendants
149 (unless they have other ancestors that are not suppressed).
155 may be given with each
159 Suppress the printing of the graph profile entry for routine
161 (and its descendants) as
163 above, and also excludes the time spent in
165 (and its descendants) from the total and percentage time computations.
173 Print the graph profile entry of only the specified routine
181 may be given with each
185 Print the graph profile entry of only the routine
187 and its descendants (as
189 above) and also uses only the times of the printed routines
190 in total time and percentage computations.
196 may be given with each
206 .It Fl k Ar fromname Ar toname
207 Will delete any arcs from routine
211 This can be used to break undesired cycles.
215 Only one pair of routine names may be given with each
219 Gather information about symbols from the currently-running kernel using the
226 argument to be ignored, and allows for symbols in
230 Suppress the printing of the call-graph profile.
232 Suppress the printing of the flat profile.
236 is produced that represents
237 the sum of the profile information in all the specified profile files.
238 This summary profile file may be given to later
239 executions of gprof (probably also with a
241 to accumulate profile data across several runs of an
245 Suppress the printing of functions whose names are not visible to
247 For the ELF object format, this means names that
251 For the a.out object format, it means names that do not
255 All relevant information about such functions belongs to the
256 (non-suppressed) function with the next lowest address.
257 This is useful for eliminating "functions" that are just labels
258 inside other functions.
260 Display routines that have zero usage (as shown by call counts
261 and accumulated time).
264 .Bl -tag -width a.out.gmon -compact
266 The namelist and text space.
268 Dynamic call graph and profile.
270 Summarized dynamic call graph and profile.
279 .%T "An Execution Profiler for Modular Programs"
283 .%J "Software - Practice and Experience"
289 .%T "gprof: A Call Graph Execution Profiler"
293 .%J "Proceedings of the SIGPLAN '82 Symposium on Compiler Construction, SIGPLAN Notices"
306 The granularity of the sampling is shown, but remains
308 We assume that the time for each execution of a function
309 can be expressed by the total time for the function divided
310 by the number of times the function is called.
311 Thus the time propagated along the call graph arcs to the function's
312 parents is directly proportional to the number of times that
315 Parents that are not themselves profiled will have the time of
316 their profiled children propagated to them, but they will appear
317 to be spontaneously invoked in the call graph listing, and will
318 not have their time propagated further.
319 Similarly, signal catchers, even though profiled, will appear
320 to be spontaneous (although for more obscure reasons).
321 Any profiled children of signal catchers should have their times
322 propagated properly, unless the signal catcher was invoked during
323 the execution of the profiling routine, in which case all is lost.
325 The profiled program must call
327 or return normally for the profiling information to be saved
328 in the graph profile file.