1.\" Copyright (c) 1983, 1990, 1993 2.\" The Regents of the University of California. All rights reserved. 3.\" 4.\" Redistribution and use in source and binary forms, with or without 5.\" modification, are permitted provided that the following conditions 6.\" are met: 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.\" 3. 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. 15.\" 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 26.\" SUCH DAMAGE. 27.\" 28.\" @(#)gprof.1 8.1 (Berkeley) 6/6/93 29.\" $FreeBSD$ 30.\" 31.Dd December 25, 2008 32.Dt GPROF 1 33.Os 34.Sh NAME 35.Nm gprof 36.Nd display call graph profile data 37.Sh SYNOPSIS 38.Nm 39.Op Fl abKlLsuz 40.Op Fl C Ar count 41.Op Fl e Ar name 42.Op Fl E Ar name 43.Op Fl f Ar name 44.Op Fl F Ar name 45.Op Fl k Ar fromname toname 46.Op Ar a.out Op Ar a.out.gmon ... 47.Sh DESCRIPTION 48The 49.Nm 50utility produces an execution profile of C, Pascal, or Fortran77 programs. 51The effect of called routines is incorporated in the profile of each caller. 52The profile data is taken from the call graph profile file 53which is created by programs that are compiled with the 54.Fl pg 55option of 56.Xr cc 1 , 57.Xr pc 1 , 58and 59.Xr f77 1 . 60The 61.Fl pg 62option also links in versions of the library routines 63that are compiled for profiling. 64By convention these libraries have their name suffixed with 65.Pa _p , 66i.e., the profiled version of 67.Pa libc.a 68is 69.Pa libc_p.a 70and if you specify libraries directly to the 71compiler or linker you can use 72.Fl l Ns Ar c_p 73instead of 74.Fl l Ns Ar c . 75Read the given object file (the default is 76.Pa a.out ) 77and establishes the relation between its symbol table 78and the call graph profile. 79The default graph profile file name is the name 80of the executable with the suffix 81.Pa .gmon 82appended. 83If more than one profile file is specified, 84the 85.Nm 86output shows the sum of the profile information in the given profile files. 87.Pp 88The 89.Nm 90utility calculates the amount of time spent in each routine. 91Next, these times are propagated along the edges of the call graph. 92Cycles are discovered, and calls into a cycle are made to share the time 93of the cycle. 94The first listing shows the functions 95sorted according to the time they represent 96including the time of their call graph descendants. 97Below each function entry is shown its (direct) call graph children, 98and how their times are propagated to this function. 99A similar display above the function shows how this function's time and the 100time of its descendants is propagated to its (direct) call graph parents. 101.Pp 102Cycles are also shown, with an entry for the cycle as a whole and 103a listing of the members of the cycle and their contributions to the 104time and call counts of the cycle. 105.Pp 106Second, a flat profile is given, 107similar to that provided by 108.Xr prof 1 . 109This listing gives the total execution times, the call counts, 110the time that the call spent in the routine itself, and 111the time that the call spent in the routine itself including 112its descendants. 113The units for the per-call times are normally milliseconds, 114but they are nanoseconds if the profiling clock frequency 115is 10 million or larger, 116and if a function appears to be never called then its total self time 117is printed as a percentage in the self time per call column. 118The very high profiling clock frequencies needed to get sufficient 119accuracy in the per-call times for short-lived programs are only 120implemented for 121.Dq high resolution 122(non-statistical) kernel profiling. 123.Pp 124Finally, an index of the function names is provided. 125.Pp 126The following options are available: 127.Bl -tag -width indent 128.It Fl a 129Suppress the printing of statically declared functions. 130If this option is given, all relevant information about the static function 131(e.g., time samples, calls to other functions, calls from other functions) 132belongs to the function loaded just before the static function in the 133.Pa a.out 134file. 135.It Fl b 136Suppress the printing of a description of each field in the profile. 137.It Fl C Ar count 138Find a minimal set of arcs that can be broken to eliminate all cycles with 139.Ar count 140or more members. 141Caution: the algorithm used to break cycles is exponential, 142so using this option may cause 143.Nm 144to run for a very long time. 145.It Fl e Ar name 146Suppress the printing of the graph profile entry for routine 147.Ar name 148and all its descendants 149(unless they have other ancestors that are not suppressed). 150More than one 151.Fl e 152option may be given. 153Only one 154.Ar name 155may be given with each 156.Fl e 157option. 158.It Fl E Ar name 159Suppress the printing of the graph profile entry for routine 160.Ar name 161(and its descendants) as 162.Fl e , 163above, and also excludes the time spent in 164.Ar name 165(and its descendants) from the total and percentage time computations. 166(For example, 167.Fl E 168.Ar mcount 169.Fl E 170.Ar mcleanup 171is the default.) 172.It Fl f Ar name 173Print the graph profile entry of only the specified routine 174.Ar name 175and its descendants. 176More than one 177.Fl f 178option may be given. 179Only one 180.Ar name 181may be given with each 182.Fl f 183option. 184.It Fl F Ar name 185Print the graph profile entry of only the routine 186.Ar name 187and its descendants (as 188.Fl f , 189above) and also uses only the times of the printed routines 190in total time and percentage computations. 191More than one 192.Fl F 193option may be given. 194Only one 195.Ar name 196may be given with each 197.Fl F 198option. 199The 200.Fl F 201option 202overrides 203the 204.Fl E 205option. 206.It Fl k Ar fromname Ar toname 207Will delete any arcs from routine 208.Ar fromname 209to routine 210.Ar toname . 211This can be used to break undesired cycles. 212More than one 213.Fl k 214option may be given. 215Only one pair of routine names may be given with each 216.Fl k 217option. 218.It Fl K 219Gather information about symbols from the currently-running kernel using the 220.Xr sysctl 3 221and 222.Xr kldsym 2 223interfaces. 224This forces the 225.Pa a.out 226argument to be ignored, and allows for symbols in 227.Xr kld 4 228modules to be used. 229.It Fl l 230Suppress the printing of the call-graph profile. 231.It Fl L 232Suppress the printing of the flat profile. 233.It Fl s 234A profile file 235.Pa gmon.sum 236is produced that represents 237the sum of the profile information in all the specified profile files. 238This summary profile file may be given to later 239executions of gprof (probably also with a 240.Fl s ) 241to accumulate profile data across several runs of an 242.Pa a.out 243file. 244.It Fl u 245Suppress the printing of functions whose names are not visible to 246C programs. 247For the ELF object format, this means names that 248contain the 249.Ql .\& 250character. 251For the a.out object format, it means names that do not 252begin with a 253.Ql _ 254character. 255All relevant information about such functions belongs to the 256(non-suppressed) function with the next lowest address. 257This is useful for eliminating "functions" that are just labels 258inside other functions. 259.It Fl z 260Display routines that have zero usage (as shown by call counts 261and accumulated time). 262.El 263.Sh FILES 264.Bl -tag -width a.out.gmon -compact 265.It Pa a.out 266The namelist and text space. 267.It Pa a.out.gmon 268Dynamic call graph and profile. 269.It Pa gmon.sum 270Summarized dynamic call graph and profile. 271.El 272.Sh SEE ALSO 273.Xr cc 1 , 274.Xr profil 2 , 275.Xr clocks 7 276.\" .Xr monitor 3 , 277.\" .Xr prof 1 278.Rs 279.%T "An Execution Profiler for Modular Programs" 280.%A S. Graham 281.%A P. Kessler 282.%A M. McKusick 283.%J "Software - Practice and Experience" 284.%V 13 285.%P pp. 671-685 286.%D 1983 287.Re 288.Rs 289.%T "gprof: A Call Graph Execution Profiler" 290.%A S. Graham 291.%A P. Kessler 292.%A M. McKusick 293.%J "Proceedings of the SIGPLAN '82 Symposium on Compiler Construction, SIGPLAN Notices" 294.%V 17 295.%N 6 296.%P pp. 120-126 297.%D June 1982 298.Re 299.Sh HISTORY 300The 301.Nm 302profiler 303appeared in 304.Bx 4.2 . 305.Sh BUGS 306The granularity of the sampling is shown, but remains 307statistical at best. 308We assume that the time for each execution of a function 309can be expressed by the total time for the function divided 310by the number of times the function is called. 311Thus the time propagated along the call graph arcs to the function's 312parents is directly proportional to the number of times that 313arc is traversed. 314.Pp 315Parents that are not themselves profiled will have the time of 316their profiled children propagated to them, but they will appear 317to be spontaneously invoked in the call graph listing, and will 318not have their time propagated further. 319Similarly, signal catchers, even though profiled, will appear 320to be spontaneous (although for more obscure reasons). 321Any profiled children of signal catchers should have their times 322propagated properly, unless the signal catcher was invoked during 323the execution of the profiling routine, in which case all is lost. 324.Pp 325The profiled program must call 326.Xr exit 3 327or return normally for the profiling information to be saved 328in the graph profile file. 329