head 1.12; access; symbols netbsd-11-0-RC4:1.10 netbsd-11-0-RC3:1.10 netbsd-11-0-RC2:1.10 binutils-2-46:1.1.1.12 netbsd-11-0-RC1:1.10 binutils-2-45:1.1.1.11 perseant-exfatfs-base-20250801:1.10 netbsd-11:1.10.0.2 netbsd-11-base:1.10 netbsd-10-1-RELEASE:1.8 perseant-exfatfs-base-20240630:1.10 binutils-2-42:1.1.1.10 perseant-exfatfs:1.9.0.2 perseant-exfatfs-base:1.9 netbsd-8-3-RELEASE:1.5 netbsd-9-4-RELEASE:1.7 netbsd-10-0-RELEASE:1.8 netbsd-10-0-RC6:1.8 netbsd-10-0-RC5:1.8 netbsd-10-0-RC4:1.8 netbsd-10-0-RC3:1.8 netbsd-10-0-RC2:1.8 netbsd-10-0-RC1:1.8 binutils-2-39:1.1.1.9 netbsd-10:1.8.0.6 netbsd-10-base:1.8 netbsd-9-3-RELEASE:1.7 cjep_sun2x-base1:1.8 cjep_sun2x:1.8.0.4 cjep_sun2x-base:1.8 cjep_staticlib_x-base1:1.8 netbsd-9-2-RELEASE:1.7 cjep_staticlib_x:1.8.0.2 cjep_staticlib_x-base:1.8 netbsd-9-1-RELEASE:1.7 phil-wifi-20200421:1.8 phil-wifi-20200411:1.8 is-mlppp:1.7.0.4 is-mlppp-base:1.7 phil-wifi-20200406:1.8 binutils-2-34:1.1.1.8 netbsd-8-2-RELEASE:1.5 netbsd-9-0-RELEASE:1.7 netbsd-9-0-RC2:1.7 netbsd-9-0-RC1:1.7 phil-wifi-20191119:1.7 netbsd-9:1.7.0.2 netbsd-9-base:1.7 phil-wifi-20190609:1.7 netbsd-8-1-RELEASE:1.5 netbsd-8-1-RC1:1.5 pgoyette-compat-merge-20190127:1.5.12.2 pgoyette-compat-20190127:1.7 pgoyette-compat-20190118:1.7 pgoyette-compat-1226:1.7 pgoyette-compat-1126:1.7 binutils-2-31-1:1.1.1.7 pgoyette-compat-1020:1.6 pgoyette-compat-0930:1.6 pgoyette-compat-0906:1.6 netbsd-7-2-RELEASE:1.3 pgoyette-compat-0728:1.6 netbsd-8-0-RELEASE:1.5 phil-wifi:1.6.0.2 phil-wifi-base:1.6 pgoyette-compat-0625:1.6 netbsd-8-0-RC2:1.5 pgoyette-compat-0521:1.6 pgoyette-compat-0502:1.6 pgoyette-compat-0422:1.6 netbsd-8-0-RC1:1.5 pgoyette-compat-0415:1.5 binutils-2-30:1.1.1.6 pgoyette-compat-0407:1.5 pgoyette-compat-0330:1.5 pgoyette-compat-0322:1.5 pgoyette-compat-0315:1.5 netbsd-7-1-2-RELEASE:1.3 pgoyette-compat:1.5.0.12 pgoyette-compat-base:1.5 netbsd-7-1-1-RELEASE:1.3 matt-nb8-mediatek:1.5.0.10 matt-nb8-mediatek-base:1.5 perseant-stdc-iso10646:1.5.0.8 perseant-stdc-iso10646-base:1.5 netbsd-8:1.5.0.6 netbsd-8-base:1.5 prg-localcount2-base3:1.5 prg-localcount2-base2:1.5 prg-localcount2-base1:1.5 prg-localcount2:1.5.0.4 prg-localcount2-base:1.5 pgoyette-localcount-20170426:1.5 bouyer-socketcan-base1:1.5 pgoyette-localcount-20170320:1.5 netbsd-7-1:1.3.0.10 netbsd-7-1-RELEASE:1.3 netbsd-7-1-RC2:1.3 netbsd-7-nhusb-base-20170116:1.3 bouyer-socketcan:1.5.0.2 bouyer-socketcan-base:1.5 pgoyette-localcount-20170107:1.5 netbsd-7-1-RC1:1.3 pgoyette-localcount-20161104:1.5 binutils-2-27:1.1.1.5 netbsd-7-0-2-RELEASE:1.3 localcount-20160914:1.4 netbsd-7-nhusb:1.3.0.8 netbsd-7-nhusb-base:1.3 pgoyette-localcount-20160806:1.4 binutils-2-26-1:1.1.1.4 pgoyette-localcount-20160726:1.4 pgoyette-localcount:1.4.0.2 pgoyette-localcount-base:1.4 netbsd-7-0-1-RELEASE:1.3 binutils-2-26:1.1.1.4 netbsd-7-0:1.3.0.6 netbsd-7-0-RELEASE:1.3 netbsd-7-0-RC3:1.3 netbsd-7-0-RC2:1.3 netbsd-7-0-RC1:1.3 netbsd-6-0-6-RELEASE:1.2 netbsd-6-1-5-RELEASE:1.2 netbsd-7:1.3.0.4 netbsd-7-base:1.3 yamt-pagecache-base9:1.3 yamt-pagecache-tag8:1.2 netbsd-6-1-4-RELEASE:1.2 netbsd-6-0-5-RELEASE:1.2 tls-earlyentropy:1.3.0.2 tls-earlyentropy-base:1.3 riastradh-xf86-video-intel-2-7-1-pre-2-21-15:1.3 riastradh-drm2-base3:1.3 netbsd-6-1-3-RELEASE:1.2 netbsd-6-0-4-RELEASE:1.2 binutils-2-23-2:1.1.1.3 netbsd-6-1-2-RELEASE:1.2 netbsd-6-0-3-RELEASE:1.2 netbsd-6-1-1-RELEASE:1.2 riastradh-drm2-base2:1.2 riastradh-drm2-base1:1.2 riastradh-drm2:1.2.0.12 riastradh-drm2-base:1.2 netbsd-6-1:1.2.0.16 netbsd-6-0-2-RELEASE:1.2 netbsd-6-1-RELEASE:1.2 netbsd-6-1-RC4:1.2 netbsd-6-1-RC3:1.2 agc-symver:1.2.0.14 agc-symver-base:1.2 netbsd-6-1-RC2:1.2 netbsd-6-1-RC1:1.2 yamt-pagecache-base8:1.2 netbsd-6-0-1-RELEASE:1.2 yamt-pagecache-base7:1.2 matt-nb6-plus-nbase:1.2 yamt-pagecache-base6:1.2 netbsd-6-0:1.2.0.10 netbsd-6-0-RELEASE:1.2 netbsd-6-0-RC2:1.2 tls-maxphys:1.2.0.8 tls-maxphys-base:1.3 matt-nb6-plus:1.2.0.6 matt-nb6-plus-base:1.2 netbsd-6-0-RC1:1.2 yamt-pagecache-base5:1.2 yamt-pagecache-base4:1.2 netbsd-6:1.2.0.4 netbsd-6-base:1.2 yamt-pagecache-base3:1.2 yamt-pagecache-base2:1.2 yamt-pagecache:1.2.0.2 yamt-pagecache-base:1.2 binutils-2-21-1a:1.1.1.2 cherry-xenmp:1.1.1.1.0.4 cherry-xenmp-base:1.1.1.1 bouyer-quota2-nbase:1.1.1.1 bouyer-quota2:1.1.1.1.0.2 bouyer-quota2-base:1.1.1.1 matt-mips64-premerge-20101231:1.1.1.1 binutils-2-19-1-20090402:1.1.1.1 matt-premerge-20091211:1.1.1.1 binutils-2-19-1:1.1.1.1 FSF:1.1.1; locks; strict; comment @# @; 1.12 date 2026.02.11.20.15.01; author christos; state Exp; branches; next 1.11; commitid 2ZnkVhvajDjbVZtG; 1.11 date 2025.08.25.15.47.39; author christos; state Exp; branches; next 1.10; commitid 83YQOH8LeFn9V78G; 1.10 date 2024.06.30.16.10.18; author christos; state Exp; branches; next 1.9; commitid YLzDDRlomJFmz1gF; 1.9 date 2022.12.24.20.17.06; author christos; state Exp; branches 1.9.2.1; next 1.8; commitid y5TadmP38ddHKQ6E; 1.8 date 2020.04.03.23.48.48; author christos; state Exp; branches; next 1.7; commitid 23XN8arVUk87UZ2C; 1.7 date 2018.11.07.01.13.54; author christos; state Exp; branches; next 1.6; commitid fBTCYRhE5Q1XTWYA; 1.6 date 2018.04.15.20.06.03; author christos; state Exp; branches 1.6.2.1; next 1.5; commitid MpiHxWBBfLwGMzyA; 1.5 date 2016.10.26.18.43.25; author christos; state Exp; branches 1.5.12.1; next 1.4; commitid icdZ2D2VouhuyGrz; 1.4 date 2016.01.29.14.42.46; author christos; state Exp; branches 1.4.2.1; next 1.3; commitid QJ5IMUNnZ6y2WPSy; 1.3 date 2013.09.29.14.03.31; author christos; state Exp; branches; next 1.2; commitid M0UvZ6fnBgRH4l7x; 1.2 date 2011.09.25.04.32.42; author christos; state Exp; branches 1.2.2.1 1.2.8.1; next 1.1; 1.1 date 2009.08.18.10.16.42; author skrll; state Exp; branches 1.1.1.1; next ; 1.9.2.1 date 2024.07.01.01.00.13; author perseant; state Exp; branches; next ; commitid NkoYLLCQWWw9v4gF; 1.6.2.1 date 2019.06.10.21.52.20; author christos; state Exp; branches; next 1.6.2.2; commitid jtc8rnCzWiEEHGqB; 1.6.2.2 date 2020.04.08.14.04.43; author martin; state Exp; branches; next ; commitid Qli2aW9E74UFuA3C; 1.5.12.1 date 2018.04.22.07.20.13; author pgoyette; state Exp; branches; next 1.5.12.2; commitid W6xykws0Zbl4kpzA; 1.5.12.2 date 2018.11.26.01.50.18; author pgoyette; state Exp; branches; next ; commitid Zj4q5SspGdKXto1B; 1.4.2.1 date 2016.11.04.14.43.18; author pgoyette; state Exp; branches; next ; commitid 2m1JRwYmpwPkOOsz; 1.2.2.1 date 2014.05.22.15.54.29; author yamt; state Exp; branches; next ; commitid LcfsdHQ4HK08ayBx; 1.2.8.1 date 2014.08.19.23.52.43; author tls; state Exp; branches; next ; commitid jTnpym9Qu0o4R1Nx; 1.1.1.1 date 2009.08.18.10.16.42; author skrll; state Exp; branches; next 1.1.1.2; 1.1.1.2 date 2011.09.24.22.35.33; author christos; state Exp; branches; next 1.1.1.3; 1.1.1.3 date 2013.09.29.13.45.04; author christos; state Exp; branches; next 1.1.1.4; commitid TyuuzSx830x3Yk7x; 1.1.1.4 date 2016.01.29.12.45.09; author christos; state Exp; branches; next 1.1.1.5; commitid B0qHLT45wK2yhPSy; 1.1.1.5 date 2016.10.26.17.30.15; author christos; state Exp; branches; next 1.1.1.6; commitid su5ccdRz6qdh9Grz; 1.1.1.6 date 2018.04.14.22.17.15; author christos; state Exp; branches; next 1.1.1.7; commitid ShNerhdmcUZyxsyA; 1.1.1.7 date 2018.11.06.21.19.16; author christos; state Exp; branches; next 1.1.1.8; commitid 6Ndf2Zuj4yZRAVYA; 1.1.1.8 date 2020.04.03.23.40.28; author christos; state Exp; branches; next 1.1.1.9; commitid qJOWzDWs5P5cQZ2C; 1.1.1.9 date 2022.12.23.19.01.40; author christos; state Exp; branches; next 1.1.1.10; commitid AaAP7ITTx5JnmI6E; 1.1.1.10 date 2024.06.30.16.00.58; author christos; state Exp; branches; next 1.1.1.11; commitid Eqxmo2uubpM9w1gF; 1.1.1.11 date 2025.08.25.14.43.13; author christos; state Exp; branches; next 1.1.1.12; commitid aUBKFf4jmEzPv78G; 1.1.1.12 date 2026.02.11.19.50.14; author christos; state Exp; branches; next ; commitid vybCSY05tblOMZtG; desc @@ 1.12 log @Merge conflicts between 2.45 and 2.46, regen x86 @ text @This is gprof.info, produced by makeinfo version 7.2 from gprof.texi. This file documents the gprof profiler of the GNU system. Copyright © 1988-2026 Free Software Foundation, Inc. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover Texts, and with no Back-Cover Texts. A copy of the license is included in the section entitled "GNU Free Documentation License". INFO-DIR-SECTION Software development START-INFO-DIR-ENTRY * gprof: (gprof). Profiling your program's execution END-INFO-DIR-ENTRY  File: gprof.info, Node: Top, Next: Introduction, Up: (dir) Profiling a Program: Where Does It Spend Its Time? ************************************************** This manual describes the GNU profiler, ‘gprof’, and how you can use it to determine which parts of a program are taking most of the execution time. We assume that you know how to write, compile, and execute programs. GNU ‘gprof’ was written by Jay Fenlason. This manual is for ‘gprof’ (GNU Binutils) version 2.46.0. This document is distributed under the terms of the GNU Free Documentation License version 1.3. A copy of the license is included in the section entitled "GNU Free Documentation License". * Menu: * Introduction:: What profiling means, and why it is useful. * Compiling:: How to compile your program for profiling. * Executing:: Executing your program to generate profile data * Invoking:: How to run ‘gprof’, and its options * Output:: Interpreting ‘gprof’'s output * Inaccuracy:: Potential problems you should be aware of * How do I?:: Answers to common questions * Incompatibilities:: (between GNU ‘gprof’ and Unix ‘gprof’.) * Details:: Details of how profiling is done * GNU Free Documentation License:: GNU Free Documentation License  File: gprof.info, Node: Introduction, Next: Compiling, Prev: Top, Up: Top 1 Introduction to Profiling *************************** Profiling allows you to learn where your program spent its time and which functions called which other functions while it was executing. This information can show you which pieces of your program are slower than you expected, and might be candidates for rewriting to make your program execute faster. It can also tell you which functions are being called more or less often than you expected. This may help you spot bugs that had otherwise been unnoticed. Since the profiler uses information collected during the actual execution of your program, it can be used on programs that are too large or too complex to analyze by reading the source. However, how your program is run will affect the information that shows up in the profile data. If you don't use some feature of your program while it is being profiled, no profile information will be generated for that feature. Profiling has several steps: • You must compile and link your program with profiling enabled. *Note Compiling a Program for Profiling: Compiling. • You must execute your program to generate a profile data file. *Note Executing the Program: Executing. • You must run ‘gprof’ to analyze the profile data. *Note ‘gprof’ Command Summary: Invoking. The next three chapters explain these steps in greater detail. Several forms of output are available from the analysis. The “flat profile” shows how much time your program spent in each function, and how many times that function was called. If you simply want to know which functions burn most of the cycles, it is stated concisely here. *Note The Flat Profile: Flat Profile. The “call graph” shows, for each function, which functions called it, which other functions it called, and how many times. There is also an estimate of how much time was spent in the subroutines of each function. This can suggest places where you might try to eliminate function calls that use a lot of time. *Note The Call Graph: Call Graph. The “annotated source” listing is a copy of the program's source code, labeled with the number of times each line of the program was executed. *Note The Annotated Source Listing: Annotated Source. To better understand how profiling works, you may wish to read a description of its implementation. *Note Implementation of Profiling: Implementation.  File: gprof.info, Node: Compiling, Next: Executing, Prev: Introduction, Up: Top 2 Compiling a Program for Profiling *********************************** The first step in generating profile information for your program is to compile and link it with profiling enabled. To compile a source file for profiling, specify the ‘-pg’ option when you run the compiler. (This is in addition to the options you normally use.) To link the program for profiling, if you use a compiler such as ‘cc’ to do the linking, simply specify ‘-pg’ in addition to your usual options. The same option, ‘-pg’, alters either compilation or linking to do what is necessary for profiling. Here are examples: cc -g -c myprog.c utils.c -pg cc -o myprog myprog.o utils.o -pg The ‘-pg’ option also works with a command that both compiles and links: cc -o myprog myprog.c utils.c -g -pg Note: The ‘-pg’ option must be part of your compilation options as well as your link options. If it is not then no call-graph data will be gathered and when you run ‘gprof’ you will get an error message like this: gprof: gmon.out file is missing call-graph data If you add the ‘-Q’ switch to suppress the printing of the call graph data you will still be able to see the time samples: Flat profile: Each sample counts as 0.01 seconds. % cumulative self self total time seconds seconds calls Ts/call Ts/call name 44.12 0.07 0.07 zazLoop 35.29 0.14 0.06 main 20.59 0.17 0.04 bazMillion If you run the linker ‘ld’ directly instead of through a compiler such as ‘cc’, you may have to specify a profiling startup file ‘gcrt0.o’ as the first input file instead of the usual startup file ‘crt0.o’. In addition, you would probably want to specify the profiling C library, ‘libc_p.a’, by writing ‘-lc_p’ instead of the usual ‘-lc’. This is not absolutely necessary, but doing this gives you number-of-calls information for standard library functions such as ‘read’ and ‘open’. For example: ld -o myprog /lib/gcrt0.o myprog.o utils.o -lc_p If you are running the program on a system which supports shared libraries you may run into problems with the profiling support code in a shared library being called before that library has been fully initialised. This is usually detected by the program encountering a segmentation fault as soon as it is run. The solution is to link against a static version of the library containing the profiling support code, which for ‘gcc’ users can be done via the ‘-static’ or ‘-static-libgcc’ command-line option. For example: gcc -g -pg -static-libgcc myprog.c utils.c -o myprog If you compile only some of the modules of the program with ‘-pg’, you can still profile the program, but you won't get complete information about the modules that were compiled without ‘-pg’. The only information you get for the functions in those modules is the total time spent in them; there is no record of how many times they were called, or from where. This will not affect the flat profile (except that the ‘calls’ field for the functions will be blank), but will greatly reduce the usefulness of the call graph. If you wish to perform line-by-line profiling you should use the ‘gcov’ tool instead of ‘gprof’. See that tool's manual or info pages for more details of how to do this. Note, older versions of ‘gcc’ produce line-by-line profiling information that works with ‘gprof’ rather than ‘gcov’ so there is still support for displaying this kind of information in ‘gprof’. *Note Line-by-line Profiling: Line-by-line. It also worth noting that ‘gcc’ implements a ‘-finstrument-functions’ command-line option which will insert calls to special user supplied instrumentation routines at the entry and exit of every function in their program. This can be used to implement an alternative profiling scheme.  File: gprof.info, Node: Executing, Next: Invoking, Prev: Compiling, Up: Top 3 Executing the Program *********************** Once the program is compiled for profiling, you must run it in order to generate the information that ‘gprof’ needs. Simply run the program as usual, using the normal arguments, file names, etc. The program should run normally, producing the same output as usual. It will, however, run somewhat slower than normal because of the time spent collecting and writing the profile data. The way you run the program--the arguments and input that you give it--may have a dramatic effect on what the profile information shows. The profile data will describe the parts of the program that were activated for the particular input you use. For example, if the first command you give to your program is to quit, the profile data will show the time used in initialization and in cleanup, but not much else. Your program will write the profile data into a file called ‘gmon.out’ just before exiting. If there is already a file called ‘gmon.out’, its contents are overwritten. You can rename the file afterwards if you are concerned that it may be overwritten. If your system libc allows you may be able to write the profile data under a different name. Set the GMON_OUT_PREFIX environment variable; this name will be appended with the PID of the running program. In order to write the ‘gmon.out’ file properly, your program must exit normally: by returning from ‘main’ or by calling ‘exit’. Calling the low-level function ‘_exit’ does not write the profile data, and neither does abnormal termination due to an unhandled signal. The ‘gmon.out’ file is written in the program's _current working directory_ at the time it exits. This means that if your program calls ‘chdir’, the ‘gmon.out’ file will be left in the last directory your program ‘chdir’'d to. If you don't have permission to write in this directory, the file is not written, and you will get an error message. Older versions of the GNU profiling library may also write a file called ‘bb.out’. This file, if present, contains an human-readable listing of the basic-block execution counts. Unfortunately, the appearance of a human-readable ‘bb.out’ means the basic-block counts didn't get written into ‘gmon.out’. The Perl script ‘bbconv.pl’, included with the ‘gprof’ source distribution, will convert a ‘bb.out’ file into a format readable by ‘gprof’. Invoke it like this: bbconv.pl < bb.out > BH-DATA This translates the information in ‘bb.out’ into a form that ‘gprof’ can understand. But you still need to tell ‘gprof’ about the existence of this translated information. To do that, include BB-DATA on the ‘gprof’ command line, _along with ‘gmon.out’_, like this: gprof OPTIONS EXECUTABLE-FILE gmon.out BB-DATA [YET-MORE-PROFILE-DATA-FILES...] [> OUTFILE]  File: gprof.info, Node: Invoking, Next: Output, Prev: Executing, Up: Top 4 ‘gprof’ Command Summary ************************* After you have a profile data file ‘gmon.out’, you can run ‘gprof’ to interpret the information in it. The ‘gprof’ program prints a flat profile and a call graph on standard output. Typically you would redirect the output of ‘gprof’ into a file with ‘>’. Note that ‘gprofng’, a more modern profiler which is also part of binutils, can also read gmon files and provides more functionality than ‘gprof’. You run ‘gprof’ like this: gprof OPTIONS [EXECUTABLE-FILE [PROFILE-DATA-FILES...]] [> OUTFILE] Here square-brackets indicate optional arguments. If you omit the executable file name, the file ‘a.out’ is used. If you give no profile data file name, the file ‘gmon.out’ is used. If any file is not in the proper format, or if the profile data file does not appear to belong to the executable file, an error message is printed. You can give more than one profile data file by entering all their names after the executable file name; then the statistics in all the data files are summed together. The order of these options does not matter. * Menu: * Output Options:: Controlling ‘gprof’'s output style * Analysis Options:: Controlling how ‘gprof’ analyzes its data * Miscellaneous Options:: * Deprecated Options:: Options you no longer need to use, but which have been retained for compatibility * Symspecs:: Specifying functions to include or exclude  File: gprof.info, Node: Output Options, Next: Analysis Options, Up: Invoking 4.1 Output Options ================== These options specify which of several output formats ‘gprof’ should produce. Many of these options take an optional “symspec” to specify functions to be included or excluded. These options can be specified multiple times, with different symspecs, to include or exclude sets of symbols. *Note Symspecs: Symspecs. Specifying any of these options overrides the default (‘-p -q’), which prints a flat profile and call graph analysis for all functions. ‘-A[SYMSPEC]’ ‘--annotated-source[=SYMSPEC]’ The ‘-A’ option causes ‘gprof’ to print annotated source code. If SYMSPEC is specified, print output only for matching symbols. *Note The Annotated Source Listing: Annotated Source. ‘-b’ ‘--brief’ If the ‘-b’ option is given, ‘gprof’ doesn't print the verbose blurbs that try to explain the meaning of all of the fields in the tables. This is useful if you intend to print out the output, or are tired of seeing the blurbs. ‘-B’ The ‘-B’ option causes ‘gprof’ to print the call graph analysis. ‘-C[SYMSPEC]’ ‘--exec-counts[=SYMSPEC]’ The ‘-C’ option causes ‘gprof’ to print a tally of functions and the number of times each was called. If SYMSPEC is specified, print tally only for matching symbols. If the profile data file contains basic-block count records, specifying the ‘-l’ option, along with ‘-C’, will cause basic-block execution counts to be tallied and displayed. ‘-i’ ‘--file-info’ The ‘-i’ option causes ‘gprof’ to display summary information about the profile data file(s) and then exit. The number of histogram, call graph, and basic-block count records is displayed. ‘-I DIRS’ ‘--directory-path=DIRS’ The ‘-I’ option specifies a list of search directories in which to find source files. Environment variable GPROF_PATH can also be used to convey this information. Used mostly for annotated source output. ‘-J[SYMSPEC]’ ‘--no-annotated-source[=SYMSPEC]’ The ‘-J’ option causes ‘gprof’ not to print annotated source code. If SYMSPEC is specified, ‘gprof’ prints annotated source, but excludes matching symbols. ‘-L’ ‘--print-path’ Normally, source filenames are printed with the path component suppressed. The ‘-L’ option causes ‘gprof’ to print the full pathname of source filenames, which is determined from symbolic debugging information in the image file and is relative to the directory in which the compiler was invoked. ‘-p[SYMSPEC]’ ‘--flat-profile[=SYMSPEC]’ The ‘-p’ option causes ‘gprof’ to print a flat profile. If SYMSPEC is specified, print flat profile only for matching symbols. *Note The Flat Profile: Flat Profile. ‘-P[SYMSPEC]’ ‘--no-flat-profile[=SYMSPEC]’ The ‘-P’ option causes ‘gprof’ to suppress printing a flat profile. If SYMSPEC is specified, ‘gprof’ prints a flat profile, but excludes matching symbols. ‘-q[SYMSPEC]’ ‘--graph[=SYMSPEC]’ The ‘-q’ option causes ‘gprof’ to print the call graph analysis. If SYMSPEC is specified, print call graph only for matching symbols and their children. *Note The Call Graph: Call Graph. ‘-Q[SYMSPEC]’ ‘--no-graph[=SYMSPEC]’ The ‘-Q’ option causes ‘gprof’ to suppress printing the call graph. If SYMSPEC is specified, ‘gprof’ prints a call graph, but excludes matching symbols. ‘-t’ ‘--table-length=NUM’ The ‘-t’ option causes the NUM most active source lines in each source file to be listed when source annotation is enabled. The default is 10. ‘-y’ ‘--separate-files’ This option affects annotated source output only. Normally, ‘gprof’ prints annotated source files to standard-output. If this option is specified, annotated source for a file named ‘path/FILENAME’ is generated in the file ‘FILENAME-ann’. If the underlying file system would truncate ‘FILENAME-ann’ so that it overwrites the original ‘FILENAME’, ‘gprof’ generates annotated source in the file ‘FILENAME.ann’ instead (if the original file name has an extension, that extension is _replaced_ with ‘.ann’). ‘-Z[SYMSPEC]’ ‘--no-exec-counts[=SYMSPEC]’ The ‘-Z’ option causes ‘gprof’ not to print a tally of functions and the number of times each was called. If SYMSPEC is specified, print tally, but exclude matching symbols. ‘-r’ ‘--function-ordering’ The ‘--function-ordering’ option causes ‘gprof’ to print a suggested function ordering for the program based on profiling data. This option suggests an ordering which may improve paging, tlb and cache behavior for the program on systems which support arbitrary ordering of functions in an executable. The exact details of how to force the linker to place functions in a particular order is system dependent and out of the scope of this manual. ‘-R MAP_FILE’ ‘--file-ordering MAP_FILE’ The ‘--file-ordering’ option causes ‘gprof’ to print a suggested .o link line ordering for the program based on profiling data. This option suggests an ordering which may improve paging, tlb and cache behavior for the program on systems which do not support arbitrary ordering of functions in an executable. Use of the ‘-a’ argument is highly recommended with this option. The MAP_FILE argument is a pathname to a file which provides function name to object file mappings. The format of the file is similar to the output of the program ‘nm’. c-parse.o:00000000 T yyparse c-parse.o:00000004 C yyerrflag c-lang.o:00000000 T maybe_objc_method_name c-lang.o:00000000 T print_lang_statistics c-lang.o:00000000 T recognize_objc_keyword c-decl.o:00000000 T print_lang_identifier c-decl.o:00000000 T print_lang_type ... To create a MAP_FILE with GNU ‘nm’, type a command like ‘nm --extern-only --defined-only -v --print-file-name program-name’. ‘-T’ ‘--traditional’ The ‘-T’ option causes ‘gprof’ to print its output in "traditional" BSD style. ‘-w WIDTH’ ‘--width=WIDTH’ Sets width of output lines to WIDTH. Currently only used when printing the function index at the bottom of the call graph. ‘-x’ ‘--all-lines’ This option affects annotated source output only. By default, only the lines at the beginning of a basic-block are annotated. If this option is specified, every line in a basic-block is annotated by repeating the annotation for the first line. This behavior is similar to ‘tcov’'s ‘-a’. ‘--demangle[=STYLE]’ ‘--no-demangle’ These options control whether C++ symbol names should be demangled when printing output. The default is to demangle symbols. The ‘--no-demangle’ option may be used to turn off demangling. Different compilers have different mangling styles. The optional demangling style argument can be used to choose an appropriate demangling style for your compiler.  File: gprof.info, Node: Analysis Options, Next: Miscellaneous Options, Prev: Output Options, Up: Invoking 4.2 Analysis Options ==================== ‘-a’ ‘--no-static’ The ‘-a’ option causes ‘gprof’ to suppress the printing of statically declared (private) functions. (These are functions whose names are not listed as global, and which are not visible outside the file/function/block where they were defined.) Time spent in these functions, calls to/from them, etc., will all be attributed to the function that was loaded directly before it in the executable file. This option affects both the flat profile and the call graph. ‘-c’ ‘--static-call-graph’ The ‘-c’ option causes the call graph of the program to be augmented by a heuristic which examines the text space of the object file and identifies function calls in the binary machine code. Since normal call graph records are only generated when functions are entered, this option identifies children that could have been called, but never were. Calls to functions that were not compiled with profiling enabled are also identified, but only if symbol table entries are present for them. Calls to dynamic library routines are typically _not_ found by this option. Parents or children identified via this heuristic are indicated in the call graph with call counts of ‘0’. ‘-D’ ‘--ignore-non-functions’ The ‘-D’ option causes ‘gprof’ to ignore symbols which are not known to be functions. This option will give more accurate profile data on systems where it is supported (Solaris and HPUX for example). ‘-k FROM/TO’ The ‘-k’ option allows you to delete from the call graph any arcs from symbols matching symspec FROM to those matching symspec TO. ‘-l’ ‘--line’ The ‘-l’ option enables line-by-line profiling, which causes histogram hits to be charged to individual source code lines, instead of functions. This feature only works with programs compiled by older versions of the ‘gcc’ compiler. Newer versions of ‘gcc’ are designed to work with the ‘gcov’ tool instead. If the program was compiled with basic-block counting enabled, this option will also identify how many times each line of code was executed. While line-by-line profiling can help isolate where in a large function a program is spending its time, it also significantly increases the running time of ‘gprof’, and magnifies statistical inaccuracies. *Note Statistical Sampling Error: Sampling Error. ‘--inline-file-names’ This option causes ‘gprof’ to print the source file after each symbol in both the flat profile and the call graph. The full path to the file is printed if used with the ‘-L’ option. ‘-m NUM’ ‘--min-count=NUM’ This option affects execution count output only. Symbols that are executed less than NUM times are suppressed. ‘-nSYMSPEC’ ‘--time=SYMSPEC’ The ‘-n’ option causes ‘gprof’, in its call graph analysis, to only propagate times for symbols matching SYMSPEC. ‘-NSYMSPEC’ ‘--no-time=SYMSPEC’ The ‘-n’ option causes ‘gprof’, in its call graph analysis, not to propagate times for symbols matching SYMSPEC. ‘-SFILENAME’ ‘--external-symbol-table=FILENAME’ The ‘-S’ option causes ‘gprof’ to read an external symbol table file, such as ‘/proc/kallsyms’, rather than read the symbol table from the given object file (the default is ‘a.out’). This is useful for profiling kernel modules. ‘-z’ ‘--display-unused-functions’ If you give the ‘-z’ option, ‘gprof’ will mention all functions in the flat profile, even those that were never called, and that had no time spent in them. This is useful in conjunction with the ‘-c’ option for discovering which routines were never called.  File: gprof.info, Node: Miscellaneous Options, Next: Deprecated Options, Prev: Analysis Options, Up: Invoking 4.3 Miscellaneous Options ========================= ‘-d[NUM]’ ‘--debug[=NUM]’ The ‘-d NUM’ option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging ‘gprof’: Debugging. ‘-h’ ‘--help’ The ‘-h’ option prints command line usage. ‘-ONAME’ ‘--file-format=NAME’ Selects the format of the profile data files. Recognized formats are ‘auto’ (the default), ‘bsd’, ‘4.4bsd’, ‘magic’, and ‘prof’ (not yet supported). ‘-s’ ‘--sum’ The ‘-s’ option causes ‘gprof’ to summarize the information in the profile data files it read in, and write out a profile data file called ‘gmon.sum’, which contains all the information from the profile data files that ‘gprof’ read in. The file ‘gmon.sum’ may be one of the specified input files; the effect of this is to merge the data in the other input files into ‘gmon.sum’. Eventually you can run ‘gprof’ again without ‘-s’ to analyze the cumulative data in the file ‘gmon.sum’. ‘-v’ ‘--version’ The ‘-v’ flag causes ‘gprof’ to print the current version number, and then exit.  File: gprof.info, Node: Deprecated Options, Next: Symspecs, Prev: Miscellaneous Options, Up: Invoking 4.4 Deprecated Options ====================== These options have been replaced with newer versions that use symspecs. ‘-e FUNCTION_NAME’ The ‘-e FUNCTION’ option tells ‘gprof’ to not print information about the function FUNCTION_NAME (and its children...) in the call graph. The function will still be listed as a child of any functions that call it, but its index number will be shown as ‘[not printed]’. More than one ‘-e’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-e’ option. ‘-E FUNCTION_NAME’ The ‘-E FUNCTION’ option works like the ‘-e’ option, but time spent in the function (and children who were not called from anywhere else), will not be used to compute the percentages-of-time for the call graph. More than one ‘-E’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-E’ option. ‘-f FUNCTION_NAME’ The ‘-f FUNCTION’ option causes ‘gprof’ to limit the call graph to the function FUNCTION_NAME and its children (and their children...). More than one ‘-f’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-f’ option. ‘-F FUNCTION_NAME’ The ‘-F FUNCTION’ option works like the ‘-f’ option, but only time spent in the function and its children (and their children...) will be used to determine total-time and percentages-of-time for the call graph. More than one ‘-F’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-F’ option. The ‘-F’ option overrides the ‘-E’ option. Note that only one function can be specified with each ‘-e’, ‘-E’, ‘-f’ or ‘-F’ option. To specify more than one function, use multiple options. For example, this command: gprof -e boring -f foo -f bar myprogram > gprof.output lists in the call graph all functions that were reached from either ‘foo’ or ‘bar’ and were not reachable from ‘boring’.  File: gprof.info, Node: Symspecs, Prev: Deprecated Options, Up: Invoking 4.5 Symspecs ============ Many of the output options allow functions to be included or excluded using “symspecs” (symbol specifications), which observe the following syntax: filename_containing_a_dot | funcname_not_containing_a_dot | linenumber | ( [ any_filename ] `:' ( any_funcname | linenumber ) ) Here are some sample symspecs: ‘main.c’ Selects everything in file ‘main.c’--the dot in the string tells ‘gprof’ to interpret the string as a filename, rather than as a function name. To select a file whose name does not contain a dot, a trailing colon should be specified. For example, ‘odd:’ is interpreted as the file named ‘odd’. ‘main’ Selects all functions named ‘main’. Note that there may be multiple instances of the same function name because some of the definitions may be local (i.e., static). Unless a function name is unique in a program, you must use the colon notation explained below to specify a function from a specific source file. Sometimes, function names contain dots. In such cases, it is necessary to add a leading colon to the name. For example, ‘:.mul’ selects function ‘.mul’. In some object file formats, symbols have a leading underscore. ‘gprof’ will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as ‘gprof’ prints it in its output. For example, if the compiler produces a symbol ‘_main’ from your ‘main’ function, ‘gprof’ still prints it as ‘main’ in its output, so you should use ‘main’ in symspecs. ‘main.c:main’ Selects function ‘main’ in file ‘main.c’. ‘main.c:134’ Selects line 134 in file ‘main.c’.  File: gprof.info, Node: Output, Next: Inaccuracy, Prev: Invoking, Up: Top 5 Interpreting ‘gprof’'s Output ******************************* ‘gprof’ can produce several different output styles, the most important of which are described below. The simplest output styles (file information, execution count, and function and file ordering) are not described here, but are documented with the respective options that trigger them. *Note Output Options: Output Options. * Menu: * Flat Profile:: The flat profile shows how much time was spent executing directly in each function. * Call Graph:: The call graph shows which functions called which others, and how much time each function used when its subroutine calls are included. * Line-by-line:: ‘gprof’ can analyze individual source code lines * Annotated Source:: The annotated source listing displays source code labeled with execution counts  File: gprof.info, Node: Flat Profile, Next: Call Graph, Up: Output 5.1 The Flat Profile ==================== The “flat profile” shows the total amount of time your program spent executing each function. Unless the ‘-z’ option is given, functions with no apparent time spent in them, and no apparent calls to them, are not mentioned. Note that if a function was not compiled for profiling, and didn't run long enough to show up on the program counter histogram, it will be indistinguishable from a function that was never called. This is part of a flat profile for a small program: Flat profile: Each sample counts as 0.01 seconds. % cumulative self self total time seconds seconds calls ms/call ms/call name 33.34 0.02 0.02 7208 0.00 0.00 open 16.67 0.03 0.01 244 0.04 0.12 offtime 16.67 0.04 0.01 8 1.25 1.25 memccpy 16.67 0.05 0.01 7 1.43 1.43 write 16.67 0.06 0.01 mcount 0.00 0.06 0.00 236 0.00 0.00 tzset 0.00 0.06 0.00 192 0.00 0.00 tolower 0.00 0.06 0.00 47 0.00 0.00 strlen 0.00 0.06 0.00 45 0.00 0.00 strchr 0.00 0.06 0.00 1 0.00 50.00 main 0.00 0.06 0.00 1 0.00 0.00 memcpy 0.00 0.06 0.00 1 0.00 10.11 print 0.00 0.06 0.00 1 0.00 0.00 profil 0.00 0.06 0.00 1 0.00 50.00 report ... The functions are sorted first by decreasing run-time spent in them, then by decreasing number of calls, then alphabetically by name. The functions ‘mcount’ and ‘profil’ are part of the profiling apparatus and appear in every flat profile; their time gives a measure of the amount of overhead due to profiling. Just before the column headers, a statement appears indicating how much time each sample counted as. This “sampling period” estimates the margin of error in each of the time figures. A time figure that is not much larger than this is not reliable. In this example, each sample counted as 0.01 seconds, suggesting a 100 Hz sampling rate. The program's total execution time was 0.06 seconds, as indicated by the ‘cumulative seconds’ field. Since each sample counted for 0.01 seconds, this means only six samples were taken during the run. Two of the samples occurred while the program was in the ‘open’ function, as indicated by the ‘self seconds’ field. Each of the other four samples occurred one each in ‘offtime’, ‘memccpy’, ‘write’, and ‘mcount’. Since only six samples were taken, none of these values can be regarded as particularly reliable. In another run, the ‘self seconds’ field for ‘mcount’ might well be ‘0.00’ or ‘0.02’. *Note Statistical Sampling Error: Sampling Error, for a complete discussion. The remaining functions in the listing (those whose ‘self seconds’ field is ‘0.00’) didn't appear in the histogram samples at all. However, the call graph indicated that they were called, so therefore they are listed, sorted in decreasing order by the ‘calls’ field. Clearly some time was spent executing these functions, but the paucity of histogram samples prevents any determination of how much time each took. Here is what the fields in each line mean: ‘% time’ This is the percentage of the total execution time your program spent in this function. These should all add up to 100%. ‘cumulative seconds’ This is the cumulative total number of seconds the computer spent executing this functions, plus the time spent in all the functions above this one in this table. ‘self seconds’ This is the number of seconds accounted for by this function alone. The flat profile listing is sorted first by this number. ‘calls’ This is the total number of times the function was called. If the function was never called, or the number of times it was called cannot be determined (probably because the function was not compiled with profiling enabled), the “calls” field is blank. ‘self ms/call’ This represents the average number of milliseconds spent in this function per call, if this function is profiled. Otherwise, this field is blank for this function. ‘total ms/call’ This represents the average number of milliseconds spent in this function and its descendants per call, if this function is profiled. Otherwise, this field is blank for this function. This is the only field in the flat profile that uses call graph analysis. ‘name’ This is the name of the function. The flat profile is sorted by this field alphabetically after the “self seconds” and “calls” fields are sorted.  File: gprof.info, Node: Call Graph, Next: Line-by-line, Prev: Flat Profile, Up: Output 5.2 The Call Graph ================== The “call graph” shows how much time was spent in each function and its children. From this information, you can find functions that, while they themselves may not have used much time, called other functions that did use unusual amounts of time. Here is a sample call from a small program. This call came from the same ‘gprof’ run as the flat profile example in the previous section. granularity: each sample hit covers 2 byte(s) for 20.00% of 0.05 seconds index % time self children called name [1] 100.0 0.00 0.05 start [1] 0.00 0.05 1/1 main [2] 0.00 0.00 1/2 on_exit [28] 0.00 0.00 1/1 exit [59] ----------------------------------------------- 0.00 0.05 1/1 start [1] [2] 100.0 0.00 0.05 1 main [2] 0.00 0.05 1/1 report [3] ----------------------------------------------- 0.00 0.05 1/1 main [2] [3] 100.0 0.00 0.05 1 report [3] 0.00 0.03 8/8 timelocal [6] 0.00 0.01 1/1 print [9] 0.00 0.01 9/9 fgets [12] 0.00 0.00 12/34 strncmp [40] 0.00 0.00 8/8 lookup [20] 0.00 0.00 1/1 fopen [21] 0.00 0.00 8/8 chewtime [24] 0.00 0.00 8/16 skipspace [44] ----------------------------------------------- [4] 59.8 0.01 0.02 8+472 [4] 0.01 0.02 244+260 offtime [7] 0.00 0.00 236+1 tzset [26] ----------------------------------------------- The lines full of dashes divide this table into “entries”, one for each function. Each entry has one or more lines. In each entry, the primary line is the one that starts with an index number in square brackets. The end of this line says which function the entry is for. The preceding lines in the entry describe the callers of this function and the following lines describe its subroutines (also called “children” when we speak of the call graph). The entries are sorted by time spent in the function and its subroutines. The internal profiling function ‘mcount’ (*note The Flat Profile: Flat Profile.) is never mentioned in the call graph. * Menu: * Primary:: Details of the primary line's contents. * Callers:: Details of caller-lines' contents. * Subroutines:: Details of subroutine-lines' contents. * Cycles:: When there are cycles of recursion, such as ‘a’ calls ‘b’ calls ‘a’...  File: gprof.info, Node: Primary, Next: Callers, Up: Call Graph 5.2.1 The Primary Line ---------------------- The “primary line” in a call graph entry is the line that describes the function which the entry is about and gives the overall statistics for this function. For reference, we repeat the primary line from the entry for function ‘report’ in our main example, together with the heading line that shows the names of the fields: index % time self children called name ... [3] 100.0 0.00 0.05 1 report [3] Here is what the fields in the primary line mean: ‘index’ Entries are numbered with consecutive integers. Each function therefore has an index number, which appears at the beginning of its primary line. Each cross-reference to a function, as a caller or subroutine of another, gives its index number as well as its name. The index number guides you if you wish to look for the entry for that function. ‘% time’ This is the percentage of the total time that was spent in this function, including time spent in subroutines called from this function. The time spent in this function is counted again for the callers of this function. Therefore, adding up these percentages is meaningless. ‘self’ This is the total amount of time spent in this function. This should be identical to the number printed in the ‘seconds’ field for this function in the flat profile. ‘children’ This is the total amount of time spent in the subroutine calls made by this function. This should be equal to the sum of all the ‘self’ and ‘children’ entries of the children listed directly below this function. ‘called’ This is the number of times the function was called. If the function called itself recursively, there are two numbers, separated by a ‘+’. The first number counts non-recursive calls, and the second counts recursive calls. In the example above, the function ‘report’ was called once from ‘main’. ‘name’ This is the name of the current function. The index number is repeated after it. If the function is part of a cycle of recursion, the cycle number is printed between the function's name and the index number (*note How Mutually Recursive Functions Are Described: Cycles.). For example, if function ‘gnurr’ is part of cycle number one, and has index number twelve, its primary line would be end like this: gnurr [12]  File: gprof.info, Node: Callers, Next: Subroutines, Prev: Primary, Up: Call Graph 5.2.2 Lines for a Function's Callers ------------------------------------ A function's entry has a line for each function it was called by. These lines' fields correspond to the fields of the primary line, but their meanings are different because of the difference in context. For reference, we repeat two lines from the entry for the function ‘report’, the primary line and one caller-line preceding it, together with the heading line that shows the names of the fields: index % time self children called name ... 0.00 0.05 1/1 main [2] [3] 100.0 0.00 0.05 1 report [3] Here are the meanings of the fields in the caller-line for ‘report’ called from ‘main’: ‘self’ An estimate of the amount of time spent in ‘report’ itself when it was called from ‘main’. ‘children’ An estimate of the amount of time spent in subroutines of ‘report’ when ‘report’ was called from ‘main’. The sum of the ‘self’ and ‘children’ fields is an estimate of the amount of time spent within calls to ‘report’ from ‘main’. ‘called’ Two numbers: the number of times ‘report’ was called from ‘main’, followed by the total number of non-recursive calls to ‘report’ from all its callers. ‘name and index number’ The name of the caller of ‘report’ to which this line applies, followed by the caller's index number. Not all functions have entries in the call graph; some options to ‘gprof’ request the omission of certain functions. When a caller has no entry of its own, it still has caller-lines in the entries of the functions it calls. If the caller is part of a recursion cycle, the cycle number is printed between the name and the index number. If the identity of the callers of a function cannot be determined, a dummy caller-line is printed which has ‘’ as the "caller's name" and all other fields blank. This can happen for signal handlers.  File: gprof.info, Node: Subroutines, Next: Cycles, Prev: Callers, Up: Call Graph 5.2.3 Lines for a Function's Subroutines ---------------------------------------- A function's entry has a line for each of its subroutines--in other words, a line for each other function that it called. These lines' fields correspond to the fields of the primary line, but their meanings are different because of the difference in context. For reference, we repeat two lines from the entry for the function ‘main’, the primary line and a line for a subroutine, together with the heading line that shows the names of the fields: index % time self children called name ... [2] 100.0 0.00 0.05 1 main [2] 0.00 0.05 1/1 report [3] Here are the meanings of the fields in the subroutine-line for ‘main’ calling ‘report’: ‘self’ An estimate of the amount of time spent directly within ‘report’ when ‘report’ was called from ‘main’. ‘children’ An estimate of the amount of time spent in subroutines of ‘report’ when ‘report’ was called from ‘main’. The sum of the ‘self’ and ‘children’ fields is an estimate of the total time spent in calls to ‘report’ from ‘main’. ‘called’ Two numbers, the number of calls to ‘report’ from ‘main’ followed by the total number of non-recursive calls to ‘report’. This ratio is used to determine how much of ‘report’'s ‘self’ and ‘children’ time gets credited to ‘main’. *Note Estimating ‘children’ Times: Assumptions. ‘name’ The name of the subroutine of ‘main’ to which this line applies, followed by the subroutine's index number. If the caller is part of a recursion cycle, the cycle number is printed between the name and the index number.  File: gprof.info, Node: Cycles, Prev: Subroutines, Up: Call Graph 5.2.4 How Mutually Recursive Functions Are Described ---------------------------------------------------- The graph may be complicated by the presence of “cycles of recursion” in the call graph. A cycle exists if a function calls another function that (directly or indirectly) calls (or appears to call) the original function. For example: if ‘a’ calls ‘b’, and ‘b’ calls ‘a’, then ‘a’ and ‘b’ form a cycle. Whenever there are call paths both ways between a pair of functions, they belong to the same cycle. If ‘a’ and ‘b’ call each other and ‘b’ and ‘c’ call each other, all three make one cycle. Note that even if ‘b’ only calls ‘a’ if it was not called from ‘a’, ‘gprof’ cannot determine this, so ‘a’ and ‘b’ are still considered a cycle. The cycles are numbered with consecutive integers. When a function belongs to a cycle, each time the function name appears in the call graph it is followed by ‘’. The reason cycles matter is that they make the time values in the call graph paradoxical. The "time spent in children" of ‘a’ should include the time spent in its subroutine ‘b’ and in ‘b’'s subroutines--but one of ‘b’'s subroutines is ‘a’! How much of ‘a’'s time should be included in the children of ‘a’, when ‘a’ is indirectly recursive? The way ‘gprof’ resolves this paradox is by creating a single entry for the cycle as a whole. The primary line of this entry describes the total time spent directly in the functions of the cycle. The "subroutines" of the cycle are the individual functions of the cycle, and all other functions that were called directly by them. The "callers" of the cycle are the functions, outside the cycle, that called functions in the cycle. Here is an example portion of a call graph which shows a cycle containing functions ‘a’ and ‘b’. The cycle was entered by a call to ‘a’ from ‘main’; both ‘a’ and ‘b’ called ‘c’. index % time self children called name ---------------------------------------- 1.77 0 1/1 main [2] [3] 91.71 1.77 0 1+5 [3] 1.02 0 3 b [4] 0.75 0 2 a [5] ---------------------------------------- 3 a [5] [4] 52.85 1.02 0 0 b [4] 2 a [5] 0 0 3/6 c [6] ---------------------------------------- 1.77 0 1/1 main [2] 2 b [4] [5] 38.86 0.75 0 1 a [5] 3 b [4] 0 0 3/6 c [6] ---------------------------------------- (The entire call graph for this program contains in addition an entry for ‘main’, which calls ‘a’, and an entry for ‘c’, with callers ‘a’ and ‘b’.) index % time self children called name [1] 100.00 0 1.93 0 start [1] 0.16 1.77 1/1 main [2] ---------------------------------------- 0.16 1.77 1/1 start [1] [2] 100.00 0.16 1.77 1 main [2] 1.77 0 1/1 a [5] ---------------------------------------- 1.77 0 1/1 main [2] [3] 91.71 1.77 0 1+5 [3] 1.02 0 3 b [4] 0.75 0 2 a [5] 0 0 6/6 c [6] ---------------------------------------- 3 a [5] [4] 52.85 1.02 0 0 b [4] 2 a [5] 0 0 3/6 c [6] ---------------------------------------- 1.77 0 1/1 main [2] 2 b [4] [5] 38.86 0.75 0 1 a [5] 3 b [4] 0 0 3/6 c [6] ---------------------------------------- 0 0 3/6 b [4] 0 0 3/6 a [5] [6] 0.00 0 0 6 c [6] ---------------------------------------- The ‘self’ field of the cycle's primary line is the total time spent in all the functions of the cycle. It equals the sum of the ‘self’ fields for the individual functions in the cycle, found in the entry in the subroutine lines for these functions. The ‘children’ fields of the cycle's primary line and subroutine lines count only subroutines outside the cycle. Even though ‘a’ calls ‘b’, the time spent in those calls to ‘b’ is not counted in ‘a’'s ‘children’ time. Thus, we do not encounter the problem of what to do when the time in those calls to ‘b’ includes indirect recursive calls back to ‘a’. The ‘children’ field of a caller-line in the cycle's entry estimates the amount of time spent _in the whole cycle_, and its other subroutines, on the times when that caller called a function in the cycle. The ‘called’ field in the primary line for the cycle has two numbers: first, the number of times functions in the cycle were called by functions outside the cycle; second, the number of times they were called by functions in the cycle (including times when a function in the cycle calls itself). This is a generalization of the usual split into non-recursive and recursive calls. The ‘called’ field of a subroutine-line for a cycle member in the cycle's entry says how many time that function was called from functions in the cycle. The total of all these is the second number in the primary line's ‘called’ field. In the individual entry for a function in a cycle, the other functions in the same cycle can appear as subroutines and as callers. These lines show how many times each function in the cycle called or was called from each other function in the cycle. The ‘self’ and ‘children’ fields in these lines are blank because of the difficulty of defining meanings for them when recursion is going on.  File: gprof.info, Node: Line-by-line, Next: Annotated Source, Prev: Call Graph, Up: Output 5.3 Line-by-line Profiling ========================== ‘gprof’'s ‘-l’ option causes the program to perform “line-by-line” profiling. In this mode, histogram samples are assigned not to functions, but to individual lines of source code. This only works with programs compiled with older versions of the ‘gcc’ compiler. Newer versions of ‘gcc’ use a different program - ‘gcov’ - to display line-by-line profiling information. With the older versions of ‘gcc’ the program usually has to be compiled with a ‘-g’ option, in addition to ‘-pg’, in order to generate debugging symbols for tracking source code lines. Note, in much older versions of ‘gcc’ the program had to be compiled with the ‘-a’ command-line option as well. The flat profile is the most useful output table in line-by-line mode. The call graph isn't as useful as normal, since the current version of ‘gprof’ does not propagate call graph arcs from source code lines to the enclosing function. The call graph does, however, show each line of code that called each function, along with a count. Here is a section of ‘gprof’'s output, without line-by-line profiling. Note that ‘ct_init’ accounted for four histogram hits, and 13327 calls to ‘init_block’. Flat profile: Each sample counts as 0.01 seconds. % cumulative self self total time seconds seconds calls us/call us/call name 30.77 0.13 0.04 6335 6.31 6.31 ct_init Call graph (explanation follows) granularity: each sample hit covers 4 byte(s) for 7.69% of 0.13 seconds index % time self children called name 0.00 0.00 1/13496 name_too_long 0.00 0.00 40/13496 deflate 0.00 0.00 128/13496 deflate_fast 0.00 0.00 13327/13496 ct_init [7] 0.0 0.00 0.00 13496 init_block Now let's look at some of ‘gprof’'s output from the same program run, this time with line-by-line profiling enabled. Note that ‘ct_init’'s four histogram hits are broken down into four lines of source code--one hit occurred on each of lines 349, 351, 382 and 385. In the call graph, note how ‘ct_init’'s 13327 calls to ‘init_block’ are broken down into one call from line 396, 3071 calls from line 384, 3730 calls from line 385, and 6525 calls from 387. Flat profile: Each sample counts as 0.01 seconds. % cumulative self time seconds seconds calls name 7.69 0.10 0.01 ct_init (trees.c:349) 7.69 0.11 0.01 ct_init (trees.c:351) 7.69 0.12 0.01 ct_init (trees.c:382) 7.69 0.13 0.01 ct_init (trees.c:385) Call graph (explanation follows) granularity: each sample hit covers 4 byte(s) for 7.69% of 0.13 seconds % time self children called name 0.00 0.00 1/13496 name_too_long (gzip.c:1440) 0.00 0.00 1/13496 deflate (deflate.c:763) 0.00 0.00 1/13496 ct_init (trees.c:396) 0.00 0.00 2/13496 deflate (deflate.c:727) 0.00 0.00 4/13496 deflate (deflate.c:686) 0.00 0.00 5/13496 deflate (deflate.c:675) 0.00 0.00 12/13496 deflate (deflate.c:679) 0.00 0.00 16/13496 deflate (deflate.c:730) 0.00 0.00 128/13496 deflate_fast (deflate.c:654) 0.00 0.00 3071/13496 ct_init (trees.c:384) 0.00 0.00 3730/13496 ct_init (trees.c:385) 0.00 0.00 6525/13496 ct_init (trees.c:387) [6] 0.0 0.00 0.00 13496 init_block (trees.c:408)  File: gprof.info, Node: Annotated Source, Prev: Line-by-line, Up: Output 5.4 The Annotated Source Listing ================================ ‘gprof’'s ‘-A’ option triggers an annotated source listing, which lists the program's source code, each function labeled with the number of times it was called. You may also need to specify the ‘-I’ option, if ‘gprof’ can't find the source code files. With older versions of ‘gcc’ compiling with ‘gcc ... -g -pg -a’ augments your program with basic-block counting code, in addition to function counting code. This enables ‘gprof’ to determine how many times each line of code was executed. With newer versions of ‘gcc’ support for displaying basic-block counts is provided by the ‘gcov’ program. For example, consider the following function, taken from gzip, with line numbers added: 1 ulg updcrc(s, n) 2 uch *s; 3 unsigned n; 4 { 5 register ulg c; 6 7 static ulg crc = (ulg)0xffffffffL; 8 9 if (s == NULL) { 10 c = 0xffffffffL; 11 } else { 12 c = crc; 13 if (n) do { 14 c = crc_32_tab[...]; 15 } while (--n); 16 } 17 crc = c; 18 return c ^ 0xffffffffL; 19 } ‘updcrc’ has at least five basic-blocks. One is the function itself. The ‘if’ statement on line 9 generates two more basic-blocks, one for each branch of the ‘if’. A fourth basic-block results from the ‘if’ on line 13, and the contents of the ‘do’ loop form the fifth basic-block. The compiler may also generate additional basic-blocks to handle various special cases. A program augmented for basic-block counting can be analyzed with ‘gprof -l -A’. The ‘-x’ option is also helpful, to ensure that each line of code is labeled at least once. Here is ‘updcrc’'s annotated source listing for a sample ‘gzip’ run: ulg updcrc(s, n) uch *s; unsigned n; 2 ->{ register ulg c; static ulg crc = (ulg)0xffffffffL; 2 -> if (s == NULL) { 1 -> c = 0xffffffffL; 1 -> } else { 1 -> c = crc; 1 -> if (n) do { 26312 -> c = crc_32_tab[...]; 26312,1,26311 -> } while (--n); } 2 -> crc = c; 2 -> return c ^ 0xffffffffL; 2 ->} In this example, the function was called twice, passing once through each branch of the ‘if’ statement. The body of the ‘do’ loop was executed a total of 26312 times. Note how the ‘while’ statement is annotated. It began execution 26312 times, once for each iteration through the loop. One of those times (the last time) it exited, while it branched back to the beginning of the loop 26311 times.  File: gprof.info, Node: Inaccuracy, Next: How do I?, Prev: Output, Up: Top 6 Inaccuracy of ‘gprof’ Output ****************************** * Menu: * Sampling Error:: Statistical margins of error * Assumptions:: Estimating children times  File: gprof.info, Node: Sampling Error, Next: Assumptions, Up: Inaccuracy 6.1 Statistical Sampling Error ============================== The run-time figures that ‘gprof’ gives you are based on a sampling process, so they are subject to statistical inaccuracy. If a function runs only a small amount of time, so that on the average the sampling process ought to catch that function in the act only once, there is a pretty good chance it will actually find that function zero times, or twice. By contrast, the number-of-calls and basic-block figures are derived by counting, not sampling. They are completely accurate and will not vary from run to run if your program is deterministic and single threaded. In multi-threaded applications, or single threaded applications that link with multi-threaded libraries, the counts are only deterministic if the counting function is thread-safe. (Note: beware that the mcount counting function in glibc is _not_ thread-safe). *Note Implementation of Profiling: Implementation. The “sampling period” that is printed at the beginning of the flat profile says how often samples are taken. The rule of thumb is that a run-time figure is accurate if it is considerably bigger than the sampling period. The actual amount of error can be predicted. For N samples, the _expected_ error is the square-root of N. For example, if the sampling period is 0.01 seconds and ‘foo’'s run-time is 1 second, N is 100 samples (1 second/0.01 seconds), sqrt(N) is 10 samples, so the expected error in ‘foo’'s run-time is 0.1 seconds (10*0.01 seconds), or ten percent of the observed value. Again, if the sampling period is 0.01 seconds and ‘bar’'s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in ‘bar’'s run-time is 1 second, or one percent of the observed value. It is likely to vary this much _on the average_ from one profiling run to the next. (_Sometimes_ it will vary more.) This does not mean that a small run-time figure is devoid of information. If the program's _total_ run-time is large, a small run-time for one function does tell you that that function used an insignificant fraction of the whole program's time. Usually this means it is not worth optimizing. One way to get more accuracy is to give your program more (but similar) input data so it will take longer. Another way is to combine the data from several runs, using the ‘-s’ option of ‘gprof’. Here is how: 1. Run your program once. 2. Issue the command ‘mv gmon.out gmon.sum’. 3. Run your program again, the same as before. 4. Merge the new data in ‘gmon.out’ into ‘gmon.sum’ with this command: gprof -s EXECUTABLE-FILE gmon.out gmon.sum 5. Repeat the last two steps as often as you wish. 6. Analyze the cumulative data using this command: gprof EXECUTABLE-FILE gmon.sum > OUTPUT-FILE  File: gprof.info, Node: Assumptions, Prev: Sampling Error, Up: Inaccuracy 6.2 Estimating ‘children’ Times =============================== Some of the figures in the call graph are estimates--for example, the ‘children’ time values and all the time figures in caller and subroutine lines. There is no direct information about these measurements in the profile data itself. Instead, ‘gprof’ estimates them by making an assumption about your program that might or might not be true. The assumption made is that the average time spent in each call to any function ‘foo’ is not correlated with who called ‘foo’. If ‘foo’ used 5 seconds in all, and 2/5 of the calls to ‘foo’ came from ‘a’, then ‘foo’ contributes 2 seconds to ‘a’'s ‘children’ time, by assumption. This assumption is usually true enough, but for some programs it is far from true. Suppose that ‘foo’ returns very quickly when its argument is zero; suppose that ‘a’ always passes zero as an argument, while other callers of ‘foo’ pass other arguments. In this program, all the time spent in ‘foo’ is in the calls from callers other than ‘a’. But ‘gprof’ has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in ‘foo’ to the children of ‘a’. We hope some day to put more complete data into ‘gmon.out’, so that this assumption is no longer needed, if we can figure out how. For the novice, the estimated figures are usually more useful than misleading.  File: gprof.info, Node: How do I?, Next: Incompatibilities, Prev: Inaccuracy, Up: Top 7 Answers to Common Questions ***************************** How can I get more exact information about hot spots in my program? Looking at the per-line call counts only tells part of the story. Because ‘gprof’ can only report call times and counts by function, the best way to get finer-grained information on where the program is spending its time is to re-factor large functions into sequences of calls to smaller ones. Beware however that this can introduce artificial hot spots since compiling with ‘-pg’ adds a significant overhead to function calls. An alternative solution is to use a non-intrusive profiler, e.g. oprofile. How do I find which lines in my program were executed the most times? Use the ‘gcov’ program. How do I find which lines in my program called a particular function? Use ‘gprof -l’ and lookup the function in the call graph. The callers will be broken down by function and line number. How do I analyze a program that runs for less than a second? Try using a shell script like this one: for i in `seq 1 100`; do fastprog mv gmon.out gmon.out.$i done gprof -s fastprog gmon.out.* gprof fastprog gmon.sum If your program is completely deterministic, all the call counts will be simple multiples of 100 (i.e., a function called once in each run will appear with a call count of 100).  File: gprof.info, Node: Incompatibilities, Next: Details, Prev: How do I?, Up: Top 8 Incompatibilities with Unix ‘gprof’ ************************************* GNU ‘gprof’ and Berkeley Unix ‘gprof’ use the same data file ‘gmon.out’, and provide essentially the same information. But there are a few differences. • GNU ‘gprof’ uses a new, generalized file format with support for basic-block execution counts and non-realtime histograms. A magic cookie and version number allows ‘gprof’ to easily identify new style files. Old BSD-style files can still be read. *Note Profiling Data File Format: File Format. • For a recursive function, Unix ‘gprof’ lists the function as a parent and as a child, with a ‘calls’ field that lists the number of recursive calls. GNU ‘gprof’ omits these lines and puts the number of recursive calls in the primary line. • When a function is suppressed from the call graph with ‘-e’, GNU ‘gprof’ still lists it as a subroutine of functions that call it. • GNU ‘gprof’ accepts the ‘-k’ with its argument in the form ‘from/to’, instead of ‘from to’. • In the annotated source listing, if there are multiple basic blocks on the same line, GNU ‘gprof’ prints all of their counts, separated by commas. • The blurbs, field widths, and output formats are different. GNU ‘gprof’ prints blurbs after the tables, so that you can see the tables without skipping the blurbs.  File: gprof.info, Node: Details, Next: GNU Free Documentation License, Prev: Incompatibilities, Up: Top 9 Details of Profiling ********************** * Menu: * Implementation:: How a program collects profiling information * File Format:: Format of ‘gmon.out’ files * Internals:: ‘gprof’'s internal operation * Debugging:: Using ‘gprof’'s ‘-d’ option  File: gprof.info, Node: Implementation, Next: File Format, Up: Details 9.1 Implementation of Profiling =============================== Profiling works by changing how every function in your program is compiled so that when it is called, it will stash away some information about where it was called from. From this, the profiler can figure out what function called it, and can count how many times it was called. This change is made by the compiler when your program is compiled with the ‘-pg’ option, which causes every function to call ‘mcount’ (or ‘_mcount’, or ‘__mcount’, depending on the OS and compiler) as one of its first operations. The ‘mcount’ routine, included in the profiling library, is responsible for recording in an in-memory call graph table both its parent routine (the child) and its parent's parent. This is typically done by examining the stack frame to find both the address of the child, and the return address in the original parent. Since this is a very machine-dependent operation, ‘mcount’ itself is typically a short assembly-language stub routine that extracts the required information, and then calls ‘__mcount_internal’ (a normal C function) with two arguments--‘frompc’ and ‘selfpc’. ‘__mcount_internal’ is responsible for maintaining the in-memory call graph, which records ‘frompc’, ‘selfpc’, and the number of times each of these call arcs was traversed. GCC Version 2 provides a magical function (‘__builtin_return_address’), which allows a generic ‘mcount’ function to extract the required information from the stack frame. However, on some architectures, most notably the SPARC, using this builtin can be very computationally expensive, and an assembly language version of ‘mcount’ is used for performance reasons. Number-of-calls information for library routines is collected by using a special version of the C library. The programs in it are the same as in the usual C library, but they were compiled with ‘-pg’. If you link your program with ‘gcc ... -pg’, it automatically uses the profiling version of the library. Profiling also involves watching your program as it runs, and keeping a histogram of where the program counter happens to be every now and then. Typically the program counter is looked at around 100 times per second of run time, but the exact frequency may vary from system to system. This is done is one of two ways. Most UNIX-like operating systems provide a ‘profil()’ system call, which registers a memory array with the kernel, along with a scale factor that determines how the program's address space maps into the array. Typical scaling values cause every 2 to 8 bytes of address space to map into a single array slot. On every tick of the system clock (assuming the profiled program is running), the value of the program counter is examined and the corresponding slot in the memory array is incremented. Since this is done in the kernel, which had to interrupt the process anyway to handle the clock interrupt, very little additional system overhead is required. However, some operating systems, most notably Linux 2.0 (and earlier), do not provide a ‘profil()’ system call. On such a system, arrangements are made for the kernel to periodically deliver a signal to the process (typically via ‘setitimer()’), which then performs the same operation of examining the program counter and incrementing a slot in the memory array. Since this method requires a signal to be delivered to user space every time a sample is taken, it uses considerably more overhead than kernel-based profiling. Also, due to the added delay required to deliver the signal, this method is less accurate as well. A special startup routine allocates memory for the histogram and either calls ‘profil()’ or sets up a clock signal handler. This routine (‘monstartup’) can be invoked in several ways. On Linux systems, a special profiling startup file ‘gcrt0.o’, which invokes ‘monstartup’ before ‘main’, is used instead of the default ‘crt0.o’. Use of this special startup file is one of the effects of using ‘gcc ... -pg’ to link. On SPARC systems, no special startup files are used. Rather, the ‘mcount’ routine, when it is invoked for the first time (typically when ‘main’ is called), calls ‘monstartup’. If the compiler's ‘-a’ option was used, basic-block counting is also enabled. Each object file is then compiled with a static array of counts, initially zero. In the executable code, every time a new basic-block begins (i.e., when an ‘if’ statement appears), an extra instruction is inserted to increment the corresponding count in the array. At compile time, a paired array was constructed that recorded the starting address of each basic-block. Taken together, the two arrays record the starting address of every basic-block, along with the number of times it was executed. The profiling library also includes a function (‘mcleanup’) which is typically registered using ‘atexit()’ to be called as the program exits, and is responsible for writing the file ‘gmon.out’. Profiling is turned off, various headers are output, and the histogram is written, followed by the call-graph arcs and the basic-block counts. The output from ‘gprof’ gives no indication of parts of your program that are limited by I/O or swapping bandwidth. This is because samples of the program counter are taken at fixed intervals of the program's run time. Therefore, the time measurements in ‘gprof’ output say nothing about time that your program was not running. For example, a part of the program that creates so much data that it cannot all fit in physical memory at once may run very slowly due to thrashing, but ‘gprof’ will say it uses little time. On the other hand, sampling by run time has the advantage that the amount of load due to other users won't directly affect the output you get.  File: gprof.info, Node: File Format, Next: Internals, Prev: Implementation, Up: Details 9.2 Profiling Data File Format ============================== The old BSD-derived file format used for profile data does not contain a magic cookie that allows one to check whether a data file really is a ‘gprof’ file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU ‘gprof’ uses a new file format that provides these features. For backward compatibility, GNU ‘gprof’ continues to support the old BSD-derived format, but not all features are supported with it. For example, basic-block execution counts cannot be accommodated by the old file format. The new file format is defined in header file ‘gmon_out.h’. It consists of a header containing the magic cookie and a version number, as well as some spare bytes available for future extensions. All data in a profile data file is in the native format of the target for which the profile was collected. GNU ‘gprof’ adapts automatically to the byte-order in use. In the new file format, the header is followed by a sequence of records. Currently, there are three different record types: histogram records, call-graph arc records, and basic-block execution count records. Each file can contain any number of each record type. When reading a file, GNU ‘gprof’ will ensure records of the same type are compatible with each other and compute the union of all records. For example, for basic-block execution counts, the union is simply the sum of all execution counts for each basic-block. 9.2.1 Histogram Records ----------------------- Histogram records consist of a header that is followed by an array of bins. The header contains the text-segment range that the histogram spans, the size of the histogram in bytes (unlike in the old BSD format, this does not include the size of the header), the rate of the profiling clock, and the physical dimension that the bin counts represent after being scaled by the profiling clock rate. The physical dimension is specified in two parts: a long name of up to 15 characters and a single character abbreviation. For example, a histogram representing real-time would specify the long name as "seconds" and the abbreviation as "s". This feature is useful for architectures that support performance monitor hardware (which, fortunately, is becoming increasingly common). For example, under DEC OSF/1, the "uprofile" command can be used to produce a histogram of, say, instruction cache misses. In this case, the dimension in the histogram header could be set to "i-cache misses" and the abbreviation could be set to "1" (because it is simply a count, not a physical dimension). Also, the profiling rate would have to be set to 1 in this case. Histogram bins are 16-bit numbers and each bin represent an equal amount of text-space. For example, if the text-segment is one thousand bytes long and if there are ten bins in the histogram, each bin represents one hundred bytes. 9.2.2 Call-Graph Records ------------------------ Call-graph records have a format that is identical to the one used in the BSD-derived file format. It consists of an arc in the call graph and a count indicating the number of times the arc was traversed during program execution. Arcs are specified by a pair of addresses: the first must be within caller's function and the second must be within the callee's function. When performing profiling at the function level, these addresses can point anywhere within the respective function. However, when profiling at the line-level, it is better if the addresses are as close to the call-site/entry-point as possible. This will ensure that the line-level call-graph is able to identify exactly which line of source code performed calls to a function. 9.2.3 Basic-Block Execution Count Records ----------------------------------------- Basic-block execution count records consist of a header followed by a sequence of address/count pairs. The header simply specifies the length of the sequence. In an address/count pair, the address identifies a basic-block and the count specifies the number of times that basic-block was executed. Any address within the basic-address can be used.  File: gprof.info, Node: Internals, Next: Debugging, Prev: File Format, Up: Details 9.3 ‘gprof’'s Internal Operation ================================ Like most programs, ‘gprof’ begins by processing its options. During this stage, it may building its symspec list (‘sym_ids.c:sym_id_add’), if options are specified which use symspecs. ‘gprof’ maintains a single linked list of symspecs, which will eventually get turned into 12 symbol tables, organized into six include/exclude pairs--one pair each for the flat profile (INCL_FLAT/EXCL_FLAT), the call graph arcs (INCL_ARCS/EXCL_ARCS), printing in the call graph (INCL_GRAPH/EXCL_GRAPH), timing propagation in the call graph (INCL_TIME/EXCL_TIME), the annotated source listing (INCL_ANNO/EXCL_ANNO), and the execution count listing (INCL_EXEC/EXCL_EXEC). After option processing, ‘gprof’ finishes building the symspec list by adding all the symspecs in ‘default_excluded_list’ to the exclude lists EXCL_TIME and EXCL_GRAPH, and if line-by-line profiling is specified, EXCL_FLAT as well. These default excludes are not added to EXCL_ANNO, EXCL_ARCS, and EXCL_EXEC. Next, the BFD library is called to open the object file, verify that it is an object file, and read its symbol table (‘core.c:core_init’), using ‘bfd_canonicalize_symtab’ after mallocing an appropriately sized array of symbols. At this point, function mappings are read (if the ‘--file-ordering’ option has been specified), and the core text space is read into memory (if the ‘-c’ option was given). ‘gprof’'s own symbol table, an array of Sym structures, is now built. This is done in one of two ways, by one of two routines, depending on whether line-by-line profiling (‘-l’ option) has been enabled. For normal profiling, the BFD canonical symbol table is scanned. For line-by-line profiling, every text space address is examined, and a new symbol table entry gets created every time the line number changes. In either case, two passes are made through the symbol table--one to count the size of the symbol table required, and the other to actually read the symbols. In between the two passes, a single array of type ‘Sym’ is created of the appropriate length. Finally, ‘symtab.c:symtab_finalize’ is called to sort the symbol table and remove duplicate entries (entries with the same memory address). The symbol table must be a contiguous array for two reasons. First, the ‘qsort’ library function (which sorts an array) will be used to sort the symbol table. Also, the symbol lookup routine (‘symtab.c:sym_lookup’), which finds symbols based on memory address, uses a binary search algorithm which requires the symbol table to be a sorted array. Function symbols are indicated with an ‘is_func’ flag. Line number symbols have no special flags set. Additionally, a symbol can have an ‘is_static’ flag to indicate that it is a local symbol. With the symbol table read, the symspecs can now be translated into Syms (‘sym_ids.c:sym_id_parse’). Remember that a single symspec can match multiple symbols. An array of symbol tables (‘syms’) is created, each entry of which is a symbol table of Syms to be included or excluded from a particular listing. The master symbol table and the symspecs are examined by nested loops, and every symbol that matches a symspec is inserted into the appropriate syms table. This is done twice, once to count the size of each required symbol table, and again to build the tables, which have been malloced between passes. From now on, to determine whether a symbol is on an include or exclude symspec list, ‘gprof’ simply uses its standard symbol lookup routine on the appropriate table in the ‘syms’ array. Now the profile data file(s) themselves are read (‘gmon_io.c:gmon_out_read’), first by checking for a new-style ‘gmon.out’ header, then assuming this is an old-style BSD ‘gmon.out’ if the magic number test failed. New-style histogram records are read by ‘hist.c:hist_read_rec’. For the first histogram record, allocate a memory array to hold all the bins, and read them in. When multiple profile data files (or files with multiple histogram records) are read, the memory ranges of each pair of histogram records must be either equal, or non-overlapping. For each pair of histogram records, the resolution (memory region size divided by the number of bins) must be the same. The time unit must be the same for all histogram records. If the above containts are met, all histograms for the same memory range are merged. As each call graph record is read (‘call_graph.c:cg_read_rec’), the parent and child addresses are matched to symbol table entries, and a call graph arc is created by ‘cg_arcs.c:arc_add’, unless the arc fails a symspec check against INCL_ARCS/EXCL_ARCS. As each arc is added, a linked list is maintained of the parent's child arcs, and of the child's parent arcs. Both the child's call count and the arc's call count are incremented by the record's call count. Basic-block records are read (‘basic_blocks.c:bb_read_rec’), but only if line-by-line profiling has been selected. Each basic-block address is matched to a corresponding line symbol in the symbol table, and an entry made in the symbol's bb_addr and bb_calls arrays. Again, if multiple basic-block records are present for the same address, the call counts are cumulative. A gmon.sum file is dumped, if requested (‘gmon_io.c:gmon_out_write’). If histograms were present in the data files, assign them to symbols (‘hist.c:hist_assign_samples’) by iterating over all the sample bins and assigning them to symbols. Since the symbol table is sorted in order of ascending memory addresses, we can simple follow along in the symbol table as we make our pass over the sample bins. This step includes a symspec check against INCL_FLAT/EXCL_FLAT. Depending on the histogram scale factor, a sample bin may span multiple symbols, in which case a fraction of the sample count is allocated to each symbol, proportional to the degree of overlap. This effect is rare for normal profiling, but overlaps are more common during line-by-line profiling, and can cause each of two adjacent lines to be credited with half a hit, for example. If call graph data is present, ‘cg_arcs.c:cg_assemble’ is called. First, if ‘-c’ was specified, a machine-dependent routine (‘find_call’) scans through each symbol's machine code, looking for subroutine call instructions, and adding them to the call graph with a zero call count. A topological sort is performed by depth-first numbering all the symbols (‘cg_dfn.c:cg_dfn’), so that children are always numbered less than their parents, then making a array of pointers into the symbol table and sorting it into numerical order, which is reverse topological order (children appear before parents). Cycles are also detected at this point, all members of which are assigned the same topological number. Two passes are now made through this sorted array of symbol pointers. The first pass, from end to beginning (parents to children), computes the fraction of child time to propagate to each parent and a print flag. The print flag reflects symspec handling of INCL_GRAPH/EXCL_GRAPH, with a parent's include or exclude (print or no print) property being propagated to its children, unless they themselves explicitly appear in INCL_GRAPH or EXCL_GRAPH. A second pass, from beginning to end (children to parents) actually propagates the timings along the call graph, subject to a check against INCL_TIME/EXCL_TIME. With the print flag, fractions, and timings now stored in the symbol structures, the topological sort array is now discarded, and a new array of pointers is assembled, this time sorted by propagated time. Finally, print the various outputs the user requested, which is now fairly straightforward. The call graph (‘cg_print.c:cg_print’) and flat profile (‘hist.c:hist_print’) are regurgitations of values already computed. The annotated source listing (‘basic_blocks.c:print_annotated_source’) uses basic-block information, if present, to label each line of code with call counts, otherwise only the function call counts are presented. The function ordering code is marginally well documented in the source code itself (‘cg_print.c’). Basically, the functions with the most use and the most parents are placed first, followed by other functions with the most use, followed by lower use functions, followed by unused functions at the end.  File: gprof.info, Node: Debugging, Prev: Internals, Up: Details 9.4 Debugging ‘gprof’ ===================== If ‘gprof’ was compiled with debugging enabled, the ‘-d’ option triggers debugging output (to stdout) which can be helpful in understanding its operation. The debugging number specified is interpreted as a sum of the following options: 2 - Topological sort Monitor depth-first numbering of symbols during call graph analysis 4 - Cycles Shows symbols as they are identified as cycle heads 16 - Tallying As the call graph arcs are read, show each arc and how the total calls to each function are tallied 32 - Call graph arc sorting Details sorting individual parents/children within each call graph entry 64 - Reading histogram and call graph records Shows address ranges of histograms as they are read, and each call graph arc 128 - Symbol table Reading, classifying, and sorting the symbol table from the object file. For line-by-line profiling (‘-l’ option), also shows line numbers being assigned to memory addresses. 256 - Static call graph Trace operation of ‘-c’ option 512 - Symbol table and arc table lookups Detail operation of lookup routines 1024 - Call graph propagation Shows how function times are propagated along the call graph 2048 - Basic-blocks Shows basic-block records as they are read from profile data (only meaningful with ‘-l’ option) 4096 - Symspecs Shows symspec-to-symbol pattern matching operation 8192 - Annotate source Tracks operation of ‘-A’ option  File: gprof.info, Node: GNU Free Documentation License, Prev: Details, Up: Top Appendix A GNU Free Documentation License ***************************************** Version 1.3, 3 November 2008 Copyright © 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. 0. PREAMBLE The purpose of this License is to make a manual, textbook, or other functional and useful document “free” in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others. This License is a kind of "copyleft", which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software. We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference. 1. APPLICABILITY AND DEFINITIONS This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that work under the conditions stated herein. The "Document", below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "you". You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. A "Modified Version" of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modifications and/or translated into another language. A "Secondary Section" is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document's overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them. The "Invariant Sections" are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. The "Cover Texts" are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most 25 words. A "Transparent" copy of the Document means a machine-readable copy, represented in a format whose specification is available to the general public, that is suitable for revising the document straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not "Transparent" is called "Opaque". Examples of suitable formats for Transparent copies include plain ASCII without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. The "Title Page" means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as such, "Title Page" means the text near the most prominent appearance of the work's title, preceding the beginning of the body of the text. The "publisher" means any person or entity that distributes copies of the Document to the public. A section "Entitled XYZ" means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specific section name mentioned below, such as "Acknowledgements", "Dedications", "Endorsements", or "History".) To "Preserve the Title" of such a section when you modify the Document means that it remains a section "Entitled XYZ" according to this definition. The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License. 2. VERBATIM COPYING You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3. You may also lend copies, under the same conditions stated above, and you may publicly display copies. 3. COPYING IN QUANTITY If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Document's license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects. If the required texts for either cover are too voluminous to fit legibly, you should put the first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto adjacent pages. If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public. It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document. 4. MODIFICATIONS You may copy and distribute a Modified Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version: A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. B. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors, if it has fewer than five), unless they release you from this requirement. C. State on the Title page the name of the publisher of the Modified Version, as the publisher. D. Preserve all the copyright notices of the Document. E. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices. F. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below. G. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document's license notice. H. Include an unaltered copy of this License. I. Preserve the section Entitled "History", Preserve its Title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. J. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. K. For any section Entitled "Acknowledgements" or "Dedications", Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein. L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. M. Delete any section Entitled "Endorsements". Such a section may not be included in the Modified Version. N. Do not retitle any existing section to be Entitled "Endorsements" or to conflict in title with any Invariant Section. O. Preserve any Warranty Disclaimers. If the Modified Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles. You may add a section Entitled "Endorsements", provided it contains nothing but endorsements of your Modified Version by various parties--for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. You may add a passage of up to five words as a Front-Cover Text, and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one. The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modified Version. 5. COMBINING DOCUMENTS You may combine the Document with other documents released under this License, under the terms defined in section 4 above for modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, unmodified, and list them all as Invariant Sections of your combined work in its license notice, and that you preserve all their Warranty Disclaimers. The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but different contents, make the title of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in the license notice of the combined work. In the combination, you must combine any sections Entitled "History" in the various original documents, forming one section Entitled "History"; likewise combine any sections Entitled "Acknowledgements", and any sections Entitled "Dedications". You must delete all sections Entitled "Endorsements." 6. COLLECTIONS OF DOCUMENTS You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection, provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects. You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document. 7. AGGREGATION WITH INDEPENDENT WORKS A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an "aggregate" if the copyright resulting from the compilation is not used to limit the legal rights of the compilation's users beyond what the individual works permit. When the Document is included in an aggregate, this License does not apply to the other works in the aggregate which are not themselves derivative works of the Document. If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one half of the entire aggregate, the Document's Cover Texts may be placed on covers that bracket the Document within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate. 8. TRANSLATION Translation is considered a kind of modification, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License, and all the license notices in the Document, and any Warranty Disclaimers, provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer, the original version will prevail. If a section in the Document is Entitled "Acknowledgements", "Dedications", or "History", the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title. 9. TERMINATION You may not copy, modify, sublicense, or distribute the Document except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense, or distribute it is void, and will automatically terminate your rights under this License. However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation. Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice. Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it. 10. FUTURE REVISIONS OF THIS LICENSE The Free Software Foundation may publish new, revised versions of the GNU Free Documentation License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. See . Each version of the License is given a distinguishing version number. If the Document specifies that a particular numbered version of this License "or any later version" applies to it, you have the option of following the terms and conditions either of that specified version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Document. 11. RELICENSING "Massive Multiauthor Collaboration Site" (or "MMC Site") means any World Wide Web server that publishes copyrightable works and also provides prominent facilities for anybody to edit those works. A public wiki that anybody can edit is an example of such a server. A "Massive Multiauthor Collaboration" (or "MMC") contained in the site means any set of copyrightable works thus published on the MMC site. "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 license published by Creative Commons Corporation, a not-for-profit corporation with a principal place of business in San Francisco, California, as well as future copyleft versions of that license published by that same organization. "Incorporate" means to publish or republish a Document, in whole or in part, as part of another Document. An MMC is "eligible for relicensing" if it is licensed under this License, and if all works that were first published under this License somewhere other than this MMC, and subsequently incorporated in whole or in part into the MMC, (1) had no cover texts or invariant sections, and (2) were thus incorporated prior to November 1, 2008. The operator of an MMC Site may republish an MMC contained in the site under CC-BY-SA on the same site at any time before August 1, 2009, provided the MMC is eligible for relicensing. ADDENDUM: How to use this License for your documents ==================================================== To use this License in a document you have written, include a copy of the License in the document and put the following copyright and license notices just after the title page: Copyright (C) YEAR YOUR NAME. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled ``GNU Free Documentation License''. If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace the "with...Texts." line with this: with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover Texts being LIST, and with the Back-Cover Texts being LIST. If you have Invariant Sections without Cover Texts, or some other combination of the three, merge those two alternatives to suit the situation. If your document contains nontrivial examples of program code, we recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software.  Tag Table: Node: Top718 Node: Introduction2071 Node: Compiling4588 Node: Executing8768 Node: Invoking11753 Node: Output Options13371 Node: Analysis Options20922 Node: Miscellaneous Options25022 Node: Deprecated Options26392 Node: Symspecs28575 Node: Output30489 Node: Flat Profile31541 Node: Call Graph36610 Node: Primary39874 Node: Callers42522 Node: Subroutines44728 Node: Cycles46673 Node: Line-by-line53670 Node: Annotated Source57820 Node: Inaccuracy60901 Node: Sampling Error61163 Node: Assumptions64111 Node: How do I?65661 Node: Incompatibilities67234 Node: Details68808 Node: Implementation69217 Node: File Format75268 Node: Internals79584 Node: Debugging88230 Node: GNU Free Documentation License89848  End Tag Table  Local Variables: coding: utf-8 End: @ 1.11 log @merge conflicts @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 7.1.1 from gprof.texi. d5 1 a5 1 Copyright © 1988-2025 Free Software Foundation, Inc. d30 1 a30 1 This manual is for ‘gprof’ (GNU Binutils) version 2.45. d265 4 a446 1 a1251 1 a1291 1 a1332 1 a2427 1 d2430 1 a2430 1 Node: Top720 d2435 25 a2459 25 Node: Output Options13212 Node: Analysis Options20764 Node: Miscellaneous Options24864 Node: Deprecated Options26234 Node: Symspecs28417 Node: Output30331 Node: Flat Profile31383 Node: Call Graph36452 Node: Primary39716 Node: Callers42364 Node: Subroutines44570 Node: Cycles46515 Node: Line-by-line53512 Node: Annotated Source57664 Node: Inaccuracy60746 Node: Sampling Error61008 Node: Assumptions63956 Node: How do I?65506 Node: Incompatibilities67079 Node: Details68653 Node: Implementation69062 Node: File Format75113 Node: Internals79429 Node: Debugging88075 Node: GNU Free Documentation License89693 @ 1.10 log @Merge conflicts between 2.39 and 2.42 @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 7.0.2 from gprof.texi. d5 1 a5 1 Copyright © 1988-2024 Free Software Foundation, Inc. d11 2 a12 2 Texts. A copy of the license is included in the section entitled “GNU Free Documentation License”. d16 1 a16 1 * gprof: (gprof). Profiling your program’s execution d30 1 a30 1 This manual is for ‘gprof’ (GNU Binutils) version 2.42. d34 1 a34 1 the section entitled “GNU Free Documentation License”. d44 1 a44 1 * Output:: Interpreting ‘gprof’’s output d70 1 a70 1 data. If you don’t use some feature of your program while it is being d99 1 a99 1 The “annotated source” listing is a copy of the program’s source d175 1 a175 1 you can still profile the program, but you won’t get complete d184 1 a184 1 ‘gcov’ tool instead of ‘gprof’. See that tool’s manual or info pages d211 2 a212 2 The way you run the program—the arguments and input that you give it—may have a dramatic effect on what the profile information shows. d231 1 a231 1 The ‘gmon.out’ file is written in the program’s _current working d234 1 a234 1 program ‘chdir’’d to. If you don’t have permission to write in this d241 1 a241 1 didn’t get written into ‘gmon.out’. The Perl script ‘bbconv.pl’, d284 1 a284 1 * Output Options:: Controlling ‘gprof’’s output style d316 1 a316 1 If the ‘-b’ option is given, ‘gprof’ doesn’t print the verbose d449 1 a449 1 The ‘-T’ option causes ‘gprof’ to print its output in “traditional” d463 1 a463 1 similar to ‘tcov’’s ‘-a’. d669 1 a669 1 Selects everything in file ‘main.c’—the dot in the string tells d704 1 a704 1 5 Interpreting ‘gprof’’s Output d734 1 a734 1 and didn’t run long enough to show up on the program counter histogram, d771 1 a771 1 program’s total execution time was 0.06 seconds, as indicated by the d783 1 a783 1 field is ‘0.00’) didn’t appear in the histogram samples at all. d888 3 a890 3 * Primary:: Details of the primary line’s contents. * Callers:: Details of caller-lines’ contents. * Subroutines:: Details of subroutine-lines’ contents. d959 1 a959 1 is printed between the function’s name and the index number (*note d969 1 a969 1 5.2.2 Lines for a Function’s Callers d972 2 a973 2 A function’s entry has a line for each function it was called by. These lines’ fields correspond to the fields of the primary line, but their d1006 1 a1006 1 followed by the caller’s index number. d1017 2 a1018 2 dummy caller-line is printed which has ‘’ as the “caller’s name” and all other fields blank. This can happen for signal handlers. d1023 1 a1023 1 5.2.3 Lines for a Function’s Subroutines d1026 2 a1027 2 A function’s entry has a line for each of its subroutines—in other words, a line for each other function that it called. These lines’ d1057 1 a1057 1 is used to determine how much of ‘report’’s ‘self’ and ‘children’ d1063 1 a1063 1 followed by the subroutine’s index number. d1091 4 a1094 4 call graph paradoxical. The “time spent in children” of ‘a’ should include the time spent in its subroutine ‘b’ and in ‘b’’s subroutines—but one of ‘b’’s subroutines is ‘a’! How much of ‘a’’s time should be included in the children of ‘a’, when ‘a’ is indirectly d1100 1 a1100 1 “subroutines” of the cycle are the individual functions of the cycle, d1102 1 a1102 1 “callers” of the cycle are the functions, outside the cycle, that called d1163 1 a1163 1 The ‘self’ field of the cycle’s primary line is the total time spent d1168 1 a1168 1 The ‘children’ fields of the cycle’s primary line and subroutine d1170 1 a1170 1 ‘b’, the time spent in those calls to ‘b’ is not counted in ‘a’’s d1175 1 a1175 1 The ‘children’ field of a caller-line in the cycle’s entry estimates d1188 1 a1188 1 cycle’s entry says how many time that function was called from functions d1190 1 a1190 1 primary line’s ‘called’ field. d1205 1 a1205 1 ‘gprof’’s ‘-l’ option causes the program to perform “line-by-line” d1219 1 a1219 1 mode. The call graph isn’t as useful as normal, since the current d1224 1 a1224 1 Here is a section of ‘gprof’’s output, without line-by-line d1250 3 a1252 3 Now let’s look at some of ‘gprof’’s output from the same program run, this time with line-by-line profiling enabled. Note that ‘ct_init’’s four histogram hits are broken down into four lines of source code—one d1254 1 a1254 1 note how ‘ct_init’’s 13327 calls to ‘init_block’ are broken down into d1297 2 a1298 2 ‘gprof’’s ‘-A’ option triggers an annotated source listing, which lists the program’s source code, each function labeled with the number of d1300 1 a1300 1 ‘gprof’ can’t find the source code files. d1342 1 a1342 1 line of code is labeled at least once. Here is ‘updcrc’’s annotated d1412 1 a1412 1 period is 0.01 seconds and ‘foo’’s run-time is 1 second, N is 100 d1414 1 a1414 1 error in ‘foo’’s run-time is 0.1 seconds (10*0.01 seconds), or ten d1416 2 a1417 2 seconds and ‘bar’’s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in ‘bar’’s run-time is 1 second, d1423 1 a1423 1 information. If the program’s _total_ run-time is large, a small d1425 1 a1425 1 insignificant fraction of the whole program’s time. Usually this means d1455 1 a1455 1 Some of the figures in the call graph are estimates—for example, the d1466 1 a1466 1 ‘foo’ contributes 2 seconds to ‘a’’s ‘children’ time, by assumption. d1568 2 a1569 2 * Internals:: ‘gprof’’s internal operation * Debugging:: Using ‘gprof’’s ‘-d’ option d1588 1 a1588 1 parent routine (the child) and its parent’s parent. This is typically d1594 3 a1596 3 arguments—‘frompc’ and ‘selfpc’. ‘__mcount_internal’ is responsible for maintaining the in-memory call graph, which records ‘frompc’, ‘selfpc’, and the number of times each of these call arcs was traversed. d1619 1 a1619 1 the kernel, along with a scale factor that determines how the program’s d1648 1 a1648 1 If the compiler’s ‘-a’ option was used, basic-block counting is also d1666 1 a1666 1 of the program counter are taken at fixed intervals of the program’s run d1672 1 a1672 1 the advantage that the amount of load due to other users won’t directly d1718 1 a1718 1 would specify the long name as “seconds” and the abbreviation as “s”. d1721 1 a1721 1 For example, under DEC OSF/1, the “uprofile” command can be used to d1723 2 a1724 2 the dimension in the histogram header could be set to “i-cache misses” and the abbreviation could be set to “1” (because it is simply a count, d1740 2 a1741 2 must be within caller’s function and the second must be within the callee’s function. When performing profiling at the function level, d1760 1 a1760 1 9.3 ‘gprof’’s Internal Operation d1767 1 a1767 1 tables, organized into six include/exclude pairs—one pair each for the d1788 1 a1788 1 ‘gprof’’s own symbol table, an array of Sym structures, is now built. d1794 1 a1794 1 either case, two passes are made through the symbol table—one to count d1842 3 a1844 3 linked list is maintained of the parent’s child arcs, and of the child’s parent arcs. Both the child’s call count and the arc’s call count are incremented by the record’s call count. d1849 1 a1849 1 entry made in the symbol’s bb_addr and bb_calls arrays. Again, if d1869 1 a1869 1 scans through each symbol’s machine code, looking for subroutine call d1881 1 a1881 1 a parent’s include or exclude (print or no print) property being d1970 1 a1970 1 This License is a kind of “copyleft”, which means that derivative d1991 2 a1992 2 “Document”, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as “you”. You accept d1996 1 a1996 1 A “Modified Version” of the Document means any work containing the d2000 1 a2000 1 A “Secondary Section” is a named appendix or a front-matter section d2002 1 a2002 1 publishers or authors of the Document to the Document’s overall d2011 1 a2011 1 The “Invariant Sections” are certain Secondary Sections whose d2019 1 a2019 1 The “Cover Texts” are certain short passages of text that are d2025 1 a2025 1 A “Transparent” copy of the Document means a machine-readable copy, d2037 1 a2037 1 “Transparent” is called “Opaque”. d2050 1 a2050 1 The “Title Page” means, for a printed book, the title page itself, d2053 3 a2055 3 works in formats which do not have any title page as such, “Title Page” means the text near the most prominent appearance of the work’s title, preceding the beginning of the body of the text. d2057 1 a2057 1 The “publisher” means any person or entity that distributes copies d2060 1 a2060 1 A section “Entitled XYZ” means a named subunit of the Document d2064 3 a2066 3 “Acknowledgements”, “Dedications”, “Endorsements”, or “History”.) To “Preserve the Title” of such a section when you modify the Document means that it remains a section “Entitled XYZ” according d2096 1 a2096 1 the Document’s license notice requires Cover Texts, you must d2168 1 a2168 1 Sections and required Cover Texts given in the Document’s d2173 1 a2173 1 I. Preserve the section Entitled “History”, Preserve its Title, d2176 1 a2176 1 Title Page. If there is no section Entitled “History” in the d2186 1 a2186 1 “History” section. You may omit a network location for a work d2191 1 a2191 1 K. For any section Entitled “Acknowledgements” or “Dedications”, d2200 1 a2200 1 M. Delete any section Entitled “Endorsements”. Such a section d2204 1 a2204 1 “Endorsements” or to conflict in title with any Invariant d2213 1 a2213 1 titles to the list of Invariant Sections in the Modified Version’s d2217 1 a2217 1 You may add a section Entitled “Endorsements”, provided it contains d2219 3 a2221 3 parties—for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2259 4 a2262 4 “History” in the various original documents, forming one section Entitled “History”; likewise combine any sections Entitled “Acknowledgements”, and any sections Entitled “Dedications”. You must delete all sections Entitled “Endorsements.” d2283 1 a2283 1 storage or distribution medium, is called an “aggregate” if the d2285 1 a2285 1 legal rights of the compilation’s users beyond what the individual d2292 1 a2292 1 of the entire aggregate, the Document’s Cover Texts may be placed d2314 2 a2315 2 If a section in the Document is Entitled “Acknowledgements”, “Dedications”, or “History”, the requirement (section 4) to d2356 1 a2356 1 version of this License “or any later version” applies to it, you d2364 1 a2364 1 proxy’s public statement of acceptance of a version permanently d2369 1 a2369 1 “Massive Multiauthor Collaboration Site” (or “MMC Site”) means any d2373 1 a2373 1 A “Massive Multiauthor Collaboration” (or “MMC”) contained in the d2377 1 a2377 1 “CC-BY-SA” means the Creative Commons Attribution-Share Alike 3.0 d2383 1 a2383 1 “Incorporate” means to publish or republish a Document, in whole or d2386 1 a2386 1 An MMC is “eligible for relicensing” if it is licensed under this d2413 1 a2413 1 Texts, replace the “with...Texts.” line with this: d2431 30 a2460 30 Node: Top726 Node: Introduction2083 Node: Compiling4604 Node: Executing8788 Node: Invoking11783 Node: Output Options13244 Node: Analysis Options20804 Node: Miscellaneous Options24904 Node: Deprecated Options26274 Node: Symspecs28457 Node: Output30372 Node: Flat Profile31426 Node: Call Graph36501 Node: Primary39771 Node: Callers42421 Node: Subroutines44641 Node: Cycles46597 Node: Line-by-line53625 Node: Annotated Source57792 Node: Inaccuracy60882 Node: Sampling Error61144 Node: Assumptions64104 Node: How do I?65657 Node: Incompatibilities67230 Node: Details68804 Node: Implementation69217 Node: File Format75279 Node: Internals79619 Node: Debugging88287 Node: GNU Free Documentation License89905 @ 1.9 log @Merge conflicts between 2.34 and 2.39, and regen for x86_64 @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.8 from gprof.texi. d5 1 a5 1 Copyright (C) 1988-2022 Free Software Foundation, Inc. d11 2 a12 2 Texts. A copy of the license is included in the section entitled "GNU Free Documentation License". d16 1 a16 1 * gprof: (gprof). Profiling your program's execution d25 1 a25 1 This manual describes the GNU profiler, 'gprof', and how you can use it d28 1 a28 1 programs. GNU 'gprof' was written by Jay Fenlason. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.39. d34 1 a34 1 the section entitled "GNU Free Documentation License". d42 1 a42 1 * Invoking:: How to run 'gprof', and its options d44 1 a44 1 * Output:: Interpreting 'gprof''s output d48 1 a48 1 * Incompatibilities:: (between GNU 'gprof' and Unix 'gprof'.) d70 1 a70 1 data. If you don't use some feature of your program while it is being d75 1 a75 1 * You must compile and link your program with profiling enabled. d78 1 a78 1 * You must execute your program to generate a profile data file. d81 1 a81 1 * You must run 'gprof' to analyze the profile data. *Note 'gprof' d88 1 a88 1 The "flat profile" shows how much time your program spent in each d93 1 a93 1 The "call graph" shows, for each function, which functions called it, d99 1 a99 1 The "annotated source" listing is a copy of the program's source d116 1 a116 1 To compile a source file for profiling, specify the '-pg' option when d120 3 a122 3 To link the program for profiling, if you use a compiler such as 'cc' to do the linking, simply specify '-pg' in addition to your usual options. The same option, '-pg', alters either compilation or linking d128 1 a128 1 The '-pg' option also works with a command that both compiles and d133 1 a133 1 Note: The '-pg' option must be part of your compilation options as d135 1 a135 1 gathered and when you run 'gprof' you will get an error message like d140 1 a140 1 If you add the '-Q' switch to suppress the printing of the call graph d152 3 a154 3 If you run the linker 'ld' directly instead of through a compiler such as 'cc', you may have to specify a profiling startup file 'gcrt0.o' as the first input file instead of the usual startup file 'crt0.o'. In d156 1 a156 1 'libc_p.a', by writing '-lc_p' instead of the usual '-lc'. This is not d158 1 a158 1 information for standard library functions such as 'read' and 'open'. d169 2 a170 2 code, which for 'gcc' users can be done via the '-static' or '-static-libgcc' command-line option. For example: d174 3 a176 3 If you compile only some of the modules of the program with '-pg', you can still profile the program, but you won't get complete information about the modules that were compiled without '-pg'. The d180 1 a180 1 that the 'calls' field for the functions will be blank), but will d184 1 a184 1 'gcov' tool instead of 'gprof'. See that tool's manual or info pages d187 3 a189 3 Note, older versions of 'gcc' produce line-by-line profiling information that works with 'gprof' rather than 'gcov' so there is still support for displaying this kind of information in 'gprof'. *Note d192 1 a192 1 It also worth noting that 'gcc' implements a '-finstrument-functions' d205 1 a205 1 generate the information that 'gprof' needs. Simply run the program as d211 2 a212 2 The way you run the program--the arguments and input that you give it--may have a dramatic effect on what the profile information shows. d219 2 a220 2 'gmon.out' just before exiting. If there is already a file called 'gmon.out', its contents are overwritten. You can rename the file d226 3 a228 3 In order to write the 'gmon.out' file properly, your program must exit normally: by returning from 'main' or by calling 'exit'. Calling the low-level function '_exit' does not write the profile data, and d231 1 a231 1 The 'gmon.out' file is written in the program's _current working d233 2 a234 2 'chdir', the 'gmon.out' file will be left in the last directory your program 'chdir''d to. If you don't have permission to write in this d238 1 a238 1 called 'bb.out'. This file, if present, contains an human-readable d240 4 a243 4 appearance of a human-readable 'bb.out' means the basic-block counts didn't get written into 'gmon.out'. The Perl script 'bbconv.pl', included with the 'gprof' source distribution, will convert a 'bb.out' file into a format readable by 'gprof'. Invoke it like this: d247 2 a248 2 This translates the information in 'bb.out' into a form that 'gprof' can understand. But you still need to tell 'gprof' about the existence d250 1 a250 1 'gprof' command line, _along with 'gmon.out'_, like this: d257 1 a257 1 4 'gprof' Command Summary d260 2 a261 2 After you have a profile data file 'gmon.out', you can run 'gprof' to interpret the information in it. The 'gprof' program prints a flat d263 1 a263 1 redirect the output of 'gprof' into a file with '>'. d265 1 a265 1 You run 'gprof' like this: d271 2 a272 2 If you omit the executable file name, the file 'a.out' is used. If you give no profile data file name, the file 'gmon.out' is used. If any d284 2 a285 2 * Output Options:: Controlling 'gprof''s output style * Analysis Options:: Controlling how 'gprof' analyzes its data d297 1 a297 1 These options specify which of several output formats 'gprof' should d300 1 a300 1 Many of these options take an optional "symspec" to specify functions d305 1 a305 1 Specifying any of these options overrides the default ('-p -q'), d308 3 a310 3 '-A[SYMSPEC]' '--annotated-source[=SYMSPEC]' The '-A' option causes 'gprof' to print annotated source code. If d314 3 a316 3 '-b' '--brief' If the '-b' option is given, 'gprof' doesn't print the verbose d321 6 a326 3 '-C[SYMSPEC]' '--exec-counts[=SYMSPEC]' The '-C' option causes 'gprof' to print a tally of functions and d331 1 a331 1 specifying the '-l' option, along with '-C', will cause basic-block d334 3 a336 3 '-i' '--file-info' The '-i' option causes 'gprof' to display summary information about d340 3 a342 3 '-I DIRS' '--directory-path=DIRS' The '-I' option specifies a list of search directories in which to d347 4 a350 4 '-J[SYMSPEC]' '--no-annotated-source[=SYMSPEC]' The '-J' option causes 'gprof' not to print annotated source code. If SYMSPEC is specified, 'gprof' prints annotated source, but d353 2 a354 2 '-L' '--print-path' d356 1 a356 1 suppressed. The '-L' option causes 'gprof' to print the full d361 3 a363 3 '-p[SYMSPEC]' '--flat-profile[=SYMSPEC]' The '-p' option causes 'gprof' to print a flat profile. If SYMSPEC d367 4 a370 4 '-P[SYMSPEC]' '--no-flat-profile[=SYMSPEC]' The '-P' option causes 'gprof' to suppress printing a flat profile. If SYMSPEC is specified, 'gprof' prints a flat profile, but d373 3 a375 3 '-q[SYMSPEC]' '--graph[=SYMSPEC]' The '-q' option causes 'gprof' to print the call graph analysis. d379 4 a382 4 '-Q[SYMSPEC]' '--no-graph[=SYMSPEC]' The '-Q' option causes 'gprof' to suppress printing the call graph. If SYMSPEC is specified, 'gprof' prints a call graph, but excludes d385 3 a387 3 '-t' '--table-length=NUM' The '-t' option causes the NUM most active source lines in each d391 2 a392 2 '-y' '--separate-files' d394 1 a394 1 'gprof' prints annotated source files to standard-output. If this d396 9 a404 9 'path/FILENAME' is generated in the file 'FILENAME-ann'. If the underlying file system would truncate 'FILENAME-ann' so that it overwrites the original 'FILENAME', 'gprof' generates annotated source in the file 'FILENAME.ann' instead (if the original file name has an extension, that extension is _replaced_ with '.ann'). '-Z[SYMSPEC]' '--no-exec-counts[=SYMSPEC]' The '-Z' option causes 'gprof' not to print a tally of functions d408 3 a410 3 '-r' '--function-ordering' The '--function-ordering' option causes 'gprof' to print a d420 3 a422 3 '-R MAP_FILE' '--file-ordering MAP_FILE' The '--file-ordering' option causes 'gprof' to print a suggested .o d428 1 a428 1 Use of the '-a' argument is highly recommended with this option. d432 1 a432 1 similar to the output of the program 'nm'. d444 2 a445 2 To create a MAP_FILE with GNU 'nm', type a command like 'nm --extern-only --defined-only -v --print-file-name program-name'. d447 3 a449 3 '-T' '--traditional' The '-T' option causes 'gprof' to print its output in "traditional" d452 2 a453 2 '-w WIDTH' '--width=WIDTH' d457 2 a458 2 '-x' '--all-lines' d463 1 a463 1 similar to 'tcov''s '-a'. d465 2 a466 2 '--demangle[=STYLE]' '--no-demangle' d469 1 a469 1 '--no-demangle' option may be used to turn off demangling. d480 3 a482 3 '-a' '--no-static' The '-a' option causes 'gprof' to suppress the printing of d491 3 a493 3 '-c' '--static-call-graph' The '-c' option causes the call graph of the program to be d503 1 a503 1 graph with call counts of '0'. d505 3 a507 3 '-D' '--ignore-non-functions' The '-D' option causes 'gprof' to ignore symbols which are not d512 2 a513 2 '-k FROM/TO' The '-k' option allows you to delete from the call graph any arcs d516 3 a518 3 '-l' '--line' The '-l' option enables line-by-line profiling, which causes d521 2 a522 2 compiled by older versions of the 'gcc' compiler. Newer versions of 'gcc' are designed to work with the 'gcov' tool instead. d528 1 a528 1 significantly increases the running time of 'gprof', and magnifies d532 2 a533 2 '--inline-file-names' This option causes 'gprof' to print the source file after each d535 1 a535 1 to the file is printed if used with the '-L' option. d537 2 a538 2 '-m NUM' '--min-count=NUM' d542 3 a544 3 '-nSYMSPEC' '--time=SYMSPEC' The '-n' option causes 'gprof', in its call graph analysis, to only d547 3 a549 3 '-NSYMSPEC' '--no-time=SYMSPEC' The '-n' option causes 'gprof', in its call graph analysis, not to d552 5 a556 5 '-SFILENAME' '--external-symbol-table=FILENAME' The '-S' option causes 'gprof' to read an external symbol table file, such as '/proc/kallsyms', rather than read the symbol table from the given object file (the default is 'a.out'). This is d559 3 a561 3 '-z' '--display-unused-functions' If you give the '-z' option, 'gprof' will mention all functions in d563 1 a563 1 no time spent in them. This is useful in conjunction with the '-c' d572 4 a575 4 '-d[NUM]' '--debug[=NUM]' The '-d NUM' option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging 'gprof': d578 3 a580 3 '-h' '--help' The '-h' option prints command line usage. d582 2 a583 2 '-ONAME' '--file-format=NAME' d585 1 a585 1 are 'auto' (the default), 'bsd', '4.4bsd', 'magic', and 'prof' (not d588 3 a590 3 '-s' '--sum' The '-s' option causes 'gprof' to summarize the information in the d592 2 a593 2 called 'gmon.sum', which contains all the information from the profile data files that 'gprof' read in. The file 'gmon.sum' may d595 1 a595 1 the data in the other input files into 'gmon.sum'. d597 2 a598 2 Eventually you can run 'gprof' again without '-s' to analyze the cumulative data in the file 'gmon.sum'. d600 3 a602 3 '-v' '--version' The '-v' flag causes 'gprof' to print the current version number, d613 2 a614 2 '-e FUNCTION_NAME' The '-e FUNCTION' option tells 'gprof' to not print information d617 3 a619 3 functions that call it, but its index number will be shown as '[not printed]'. More than one '-e' option may be given; only one FUNCTION_NAME may be indicated with each '-e' option. d621 2 a622 2 '-E FUNCTION_NAME' The '-E FUNCTION' option works like the '-e' option, but time spent d625 2 a626 2 call graph. More than one '-E' option may be given; only one FUNCTION_NAME may be indicated with each '-E' option. d628 2 a629 2 '-f FUNCTION_NAME' The '-f FUNCTION' option causes 'gprof' to limit the call graph to d631 2 a632 2 children...). More than one '-f' option may be given; only one FUNCTION_NAME may be indicated with each '-f' option. d634 2 a635 2 '-F FUNCTION_NAME' The '-F FUNCTION' option works like the '-f' option, but only time d638 3 a640 3 call graph. More than one '-F' option may be given; only one FUNCTION_NAME may be indicated with each '-F' option. The '-F' option overrides the '-E' option. d642 2 a643 2 Note that only one function can be specified with each '-e', '-E', '-f' or '-F' option. To specify more than one function, use multiple d649 1 a649 1 'foo' or 'bar' and were not reachable from 'boring'. d658 1 a658 1 using "symspecs" (symbol specifications), which observe the following d668 3 a670 3 'main.c' Selects everything in file 'main.c'--the dot in the string tells 'gprof' to interpret the string as a filename, rather than as a d672 2 a673 2 a trailing colon should be specified. For example, 'odd:' is interpreted as the file named 'odd'. d675 2 a676 2 'main' Selects all functions named 'main'. d685 2 a686 2 necessary to add a leading colon to the name. For example, ':.mul' selects function '.mul'. d689 2 a690 2 'gprof' will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as 'gprof' prints d692 2 a693 2 '_main' from your 'main' function, 'gprof' still prints it as 'main' in its output, so you should use 'main' in symspecs. d695 2 a696 2 'main.c:main' Selects function 'main' in file 'main.c'. d698 2 a699 2 'main.c:134' Selects line 134 in file 'main.c'. d704 1 a704 1 5 Interpreting 'gprof''s Output d707 1 a707 1 'gprof' can produce several different output styles, the most important d720 1 a720 1 * Line-by-line:: 'gprof' can analyze individual source code lines d730 2 a731 2 The "flat profile" shows the total amount of time your program spent executing each function. Unless the '-z' option is given, functions d734 1 a734 1 and didn't run long enough to show up on the program counter histogram, d762 1 a762 1 functions 'mcount' and 'profil' are part of the profiling apparatus and d767 1 a767 1 much time each sample counted as. This "sampling period" estimates the d771 2 a772 2 program's total execution time was 0.06 seconds, as indicated by the 'cumulative seconds' field. Since each sample counted for 0.01 seconds, d774 3 a776 3 samples occurred while the program was in the 'open' function, as indicated by the 'self seconds' field. Each of the other four samples occurred one each in 'offtime', 'memccpy', 'write', and 'mcount'. Since d778 2 a779 2 particularly reliable. In another run, the 'self seconds' field for 'mcount' might well be '0.00' or '0.02'. *Note Statistical Sampling d782 2 a783 2 The remaining functions in the listing (those whose 'self seconds' field is '0.00') didn't appear in the histogram samples at all. d785 1 a785 1 they are listed, sorted in decreasing order by the 'calls' field. d792 1 a792 1 '% time' d796 1 a796 1 'cumulative seconds' d801 1 a801 1 'self seconds' d805 1 a805 1 'calls' d809 1 a809 1 compiled with profiling enabled), the "calls" field is blank. d811 1 a811 1 'self ms/call' d816 1 a816 1 'total ms/call' d823 1 a823 1 'name' d825 1 a825 1 this field alphabetically after the "self seconds" and "calls" d834 1 a834 1 The "call graph" shows how much time was spent in each function and its d840 1 a840 1 same 'gprof' run as the flat profile example in the previous section. d871 1 a871 1 The lines full of dashes divide this table into "entries", one for d878 1 a878 1 called "children" when we speak of the call graph). d883 1 a883 1 The internal profiling function 'mcount' (*note The Flat Profile: d888 3 a890 3 * Primary:: Details of the primary line's contents. * Callers:: Details of caller-lines' contents. * Subroutines:: Details of subroutine-lines' contents. d892 1 a892 1 such as 'a' calls 'b' calls 'a'... d900 1 a900 1 The "primary line" in a call graph entry is the line that describes the d905 1 a905 1 'report' in our main example, together with the heading line that shows d914 1 a914 1 'index' d924 1 a924 1 '% time' d933 1 a933 1 'self' d935 1 a935 1 should be identical to the number printed in the 'seconds' field d938 1 a938 1 'children' d941 1 a941 1 'self' and 'children' entries of the children listed directly below d944 1 a944 1 'called' d948 1 a948 1 separated by a '+'. The first number counts non-recursive calls, d951 2 a952 2 In the example above, the function 'report' was called once from 'main'. d954 1 a954 1 'name' d959 1 a959 1 is printed between the function's name and the index number (*note d961 1 a961 1 example, if function 'gnurr' is part of cycle number one, and has d969 1 a969 1 5.2.2 Lines for a Function's Callers d972 2 a973 2 A function's entry has a line for each function it was called by. These lines' fields correspond to the fields of the primary line, but their d977 1 a977 1 'report', the primary line and one caller-line preceding it, together d985 2 a986 2 Here are the meanings of the fields in the caller-line for 'report' called from 'main': d988 14 a1001 14 'self' An estimate of the amount of time spent in 'report' itself when it was called from 'main'. 'children' An estimate of the amount of time spent in subroutines of 'report' when 'report' was called from 'main'. The sum of the 'self' and 'children' fields is an estimate of the amount of time spent within calls to 'report' from 'main'. 'called' Two numbers: the number of times 'report' was called from 'main', followed by the total number of non-recursive calls to 'report' d1004 3 a1006 3 'name and index number' The name of the caller of 'report' to which this line applies, followed by the caller's index number. d1009 1 a1009 1 'gprof' request the omission of certain functions. When a caller d1017 2 a1018 2 dummy caller-line is printed which has '' as the "caller's name" and all other fields blank. This can happen for signal handlers. d1023 1 a1023 1 5.2.3 Lines for a Function's Subroutines d1026 2 a1027 2 A function's entry has a line for each of its subroutines--in other words, a line for each other function that it called. These lines' d1032 1 a1032 1 'main', the primary line and a line for a subroutine, together with the d1040 2 a1041 2 Here are the meanings of the fields in the subroutine-line for 'main' calling 'report': d1043 16 a1058 16 'self' An estimate of the amount of time spent directly within 'report' when 'report' was called from 'main'. 'children' An estimate of the amount of time spent in subroutines of 'report' when 'report' was called from 'main'. The sum of the 'self' and 'children' fields is an estimate of the total time spent in calls to 'report' from 'main'. 'called' Two numbers, the number of calls to 'report' from 'main' followed by the total number of non-recursive calls to 'report'. This ratio is used to determine how much of 'report''s 'self' and 'children' time gets credited to 'main'. *Note Estimating 'children' Times: d1061 3 a1063 3 'name' The name of the subroutine of 'main' to which this line applies, followed by the subroutine's index number. d1074 1 a1074 1 The graph may be complicated by the presence of "cycles of recursion" in d1077 2 a1078 2 function. For example: if 'a' calls 'b', and 'b' calls 'a', then 'a' and 'b' form a cycle. d1081 4 a1084 4 they belong to the same cycle. If 'a' and 'b' call each other and 'b' and 'c' call each other, all three make one cycle. Note that even if 'b' only calls 'a' if it was not called from 'a', 'gprof' cannot determine this, so 'a' and 'b' are still considered a cycle. d1088 1 a1088 1 graph it is followed by ''. d1091 4 a1094 4 call graph paradoxical. The "time spent in children" of 'a' should include the time spent in its subroutine 'b' and in 'b''s subroutines--but one of 'b''s subroutines is 'a'! How much of 'a''s time should be included in the children of 'a', when 'a' is indirectly d1097 1 a1097 1 The way 'gprof' resolves this paradox is by creating a single entry d1100 1 a1100 1 "subroutines" of the cycle are the individual functions of the cycle, d1102 1 a1102 1 "callers" of the cycle are the functions, outside the cycle, that called d1106 2 a1107 2 containing functions 'a' and 'b'. The cycle was entered by a call to 'a' from 'main'; both 'a' and 'b' called 'c'. d1129 2 a1130 2 for 'main', which calls 'a', and an entry for 'c', with callers 'a' and 'b'.) d1163 2 a1164 2 The 'self' field of the cycle's primary line is the total time spent in all the functions of the cycle. It equals the sum of the 'self' d1168 6 a1173 6 The 'children' fields of the cycle's primary line and subroutine lines count only subroutines outside the cycle. Even though 'a' calls 'b', the time spent in those calls to 'b' is not counted in 'a''s 'children' time. Thus, we do not encounter the problem of what to do when the time in those calls to 'b' includes indirect recursive calls back to 'a'. d1175 1 a1175 1 The 'children' field of a caller-line in the cycle's entry estimates d1180 1 a1180 1 The 'called' field in the primary line for the cycle has two numbers: d1187 2 a1188 2 The 'called' field of a subroutine-line for a cycle member in the cycle's entry says how many time that function was called from functions d1190 1 a1190 1 primary line's 'called' field. d1195 1 a1195 1 called from each other function in the cycle. The 'self' and 'children' d1205 1 a1205 1 'gprof''s '-l' option causes the program to perform "line-by-line" d1208 2 a1209 2 programs compiled with older versions of the 'gcc' compiler. Newer versions of 'gcc' use a different program - 'gcov' - to display d1212 2 a1213 2 With the older versions of 'gcc' the program usually has to be compiled with a '-g' option, in addition to '-pg', in order to generate d1215 1 a1215 1 versions of 'gcc' the program had to be compiled with the '-a' d1219 2 a1220 2 mode. The call graph isn't as useful as normal, since the current version of 'gprof' does not propagate call graph arcs from source code d1224 3 a1226 3 Here is a section of 'gprof''s output, without line-by-line profiling. Note that 'ct_init' accounted for four histogram hits, and 13327 calls to 'init_block'. d1250 3 a1252 3 Now let's look at some of 'gprof''s output from the same program run, this time with line-by-line profiling enabled. Note that 'ct_init''s four histogram hits are broken down into four lines of source code--one d1254 1 a1254 1 note how 'ct_init''s 13327 calls to 'init_block' are broken down into d1297 4 a1300 4 'gprof''s '-A' option triggers an annotated source listing, which lists the program's source code, each function labeled with the number of times it was called. You may also need to specify the '-I' option, if 'gprof' can't find the source code files. d1302 1 a1302 1 With older versions of 'gcc' compiling with 'gcc ... -g -pg -a' d1304 3 a1306 3 function counting code. This enables 'gprof' to determine how many times each line of code was executed. With newer versions of 'gcc' support for displaying basic-block counts is provided by the 'gcov' d1333 4 a1336 4 'updcrc' has at least five basic-blocks. One is the function itself. The 'if' statement on line 9 generates two more basic-blocks, one for each branch of the 'if'. A fourth basic-block results from the 'if' on line 13, and the contents of the 'do' loop form the fifth basic-block. d1341 3 a1343 3 'gprof -l -A'. The '-x' option is also helpful, to ensure that each line of code is labeled at least once. Here is 'updcrc''s annotated source listing for a sample 'gzip' run: d1366 2 a1367 2 each branch of the 'if' statement. The body of the 'do' loop was executed a total of 26312 times. Note how the 'while' statement is d1375 1 a1375 1 6 Inaccuracy of 'gprof' Output d1389 1 a1389 1 The run-time figures that 'gprof' gives you are based on a sampling d1405 1 a1405 1 The "sampling period" that is printed at the beginning of the flat d1412 1 a1412 1 period is 0.01 seconds and 'foo''s run-time is 1 second, N is 100 d1414 1 a1414 1 error in 'foo''s run-time is 0.1 seconds (10*0.01 seconds), or ten d1416 2 a1417 2 seconds and 'bar''s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in 'bar''s run-time is 1 second, d1423 1 a1423 1 information. If the program's _total_ run-time is large, a small d1425 1 a1425 1 insignificant fraction of the whole program's time. Usually this means d1430 1 a1430 1 the data from several runs, using the '-s' option of 'gprof'. Here is d1435 1 a1435 1 2. Issue the command 'mv gmon.out gmon.sum'. d1439 1 a1439 1 4. Merge the new data in 'gmon.out' into 'gmon.sum' with this command: d1452 1 a1452 1 6.2 Estimating 'children' Times d1455 2 a1456 2 Some of the figures in the call graph are estimates--for example, the 'children' time values and all the time figures in caller and subroutine d1460 1 a1460 1 profile data itself. Instead, 'gprof' estimates them by making an d1464 3 a1466 3 any function 'foo' is not correlated with who called 'foo'. If 'foo' used 5 seconds in all, and 2/5 of the calls to 'foo' came from 'a', then 'foo' contributes 2 seconds to 'a''s 'children' time, by assumption. d1469 6 a1474 6 far from true. Suppose that 'foo' returns very quickly when its argument is zero; suppose that 'a' always passes zero as an argument, while other callers of 'foo' pass other arguments. In this program, all the time spent in 'foo' is in the calls from callers other than 'a'. But 'gprof' has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in 'foo' to the children of 'a'. d1476 1 a1476 1 We hope some day to put more complete data into 'gmon.out', so that d1489 1 a1489 1 Because 'gprof' can only report call times and counts by function, d1493 1 a1493 1 artificial hot spots since compiling with '-pg' adds a significant d1499 1 a1499 1 Use the 'gcov' program. d1503 1 a1503 1 Use 'gprof -l' and lookup the function in the call graph. The d1526 1 a1526 1 8 Incompatibilities with Unix 'gprof' d1529 1 a1529 1 GNU 'gprof' and Berkeley Unix 'gprof' use the same data file 'gmon.out', d1533 1 a1533 1 * GNU 'gprof' uses a new, generalized file format with support for d1535 1 a1535 1 cookie and version number allows 'gprof' to easily identify new d1539 3 a1541 3 * For a recursive function, Unix 'gprof' lists the function as a parent and as a child, with a 'calls' field that lists the number of recursive calls. GNU 'gprof' omits these lines and puts the d1544 2 a1545 2 * When a function is suppressed from the call graph with '-e', GNU 'gprof' still lists it as a subroutine of functions that call it. d1547 2 a1548 2 * GNU 'gprof' accepts the '-k' with its argument in the form 'from/to', instead of 'from to'. d1550 2 a1551 2 * In the annotated source listing, if there are multiple basic blocks on the same line, GNU 'gprof' prints all of their counts, separated d1554 2 a1555 2 * The blurbs, field widths, and output formats are different. GNU 'gprof' prints blurbs after the tables, so that you can see the d1567 3 a1569 3 * File Format:: Format of 'gmon.out' files * Internals:: 'gprof''s internal operation * Debugging:: Using 'gprof''s '-d' option d1582 2 a1583 2 the '-pg' option, which causes every function to call 'mcount' (or '_mcount', or '__mcount', depending on the OS and compiler) as one of d1586 1 a1586 1 The 'mcount' routine, included in the profiling library, is d1588 1 a1588 1 parent routine (the child) and its parent's parent. This is typically d1591 1 a1591 1 machine-dependent operation, 'mcount' itself is typically a short d1593 4 a1596 4 and then calls '__mcount_internal' (a normal C function) with two arguments--'frompc' and 'selfpc'. '__mcount_internal' is responsible for maintaining the in-memory call graph, which records 'frompc', 'selfpc', and the number of times each of these call arcs was traversed. d1599 1 a1599 1 ('__builtin_return_address'), which allows a generic 'mcount' function d1603 1 a1603 1 'mcount' is used for performance reasons. d1607 2 a1608 2 same as in the usual C library, but they were compiled with '-pg'. If you link your program with 'gcc ... -pg', it automatically uses the d1618 2 a1619 2 provide a 'profil()' system call, which registers a memory array with the kernel, along with a scale factor that determines how the program's d1629 1 a1629 1 earlier), do not provide a 'profil()' system call. On such a system, d1631 1 a1631 1 the process (typically via 'setitimer()'), which then performs the same d1639 5 a1643 5 either calls 'profil()' or sets up a clock signal handler. This routine ('monstartup') can be invoked in several ways. On Linux systems, a special profiling startup file 'gcrt0.o', which invokes 'monstartup' before 'main', is used instead of the default 'crt0.o'. Use of this special startup file is one of the effects of using 'gcc ... -pg' to d1645 2 a1646 2 'mcount' routine, when it is invoked for the first time (typically when 'main' is called), calls 'monstartup'. d1648 1 a1648 1 If the compiler's '-a' option was used, basic-block counting is also d1651 1 a1651 1 basic-block begins (i.e., when an 'if' statement appears), an extra d1658 3 a1660 3 The profiling library also includes a function ('mcleanup') which is typically registered using 'atexit()' to be called as the program exits, and is responsible for writing the file 'gmon.out'. Profiling is turned d1664 1 a1664 1 The output from 'gprof' gives no indication of parts of your program d1666 2 a1667 2 of the program counter are taken at fixed intervals of the program's run time. Therefore, the time measurements in 'gprof' output say nothing d1670 1 a1670 1 memory at once may run very slowly due to thrashing, but 'gprof' will d1672 1 a1672 1 the advantage that the amount of load due to other users won't directly d1683 2 a1684 2 'gprof' file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU 'gprof' d1686 1 a1686 1 compatibility, GNU 'gprof' continues to support the old BSD-derived d1691 1 a1691 1 The new file format is defined in header file 'gmon_out.h'. It d1695 1 a1695 1 the profile was collected. GNU 'gprof' adapts automatically to the d1702 1 a1702 1 reading a file, GNU 'gprof' will ensure records of the same type are d1718 1 a1718 1 would specify the long name as "seconds" and the abbreviation as "s". d1721 1 a1721 1 For example, under DEC OSF/1, the "uprofile" command can be used to d1723 2 a1724 2 the dimension in the histogram header could be set to "i-cache misses" and the abbreviation could be set to "1" (because it is simply a count, d1740 2 a1741 2 must be within caller's function and the second must be within the callee's function. When performing profiling at the function level, d1760 1 a1760 1 9.3 'gprof''s Internal Operation d1763 3 a1765 3 Like most programs, 'gprof' begins by processing its options. During this stage, it may building its symspec list ('sym_ids.c:sym_id_add'), if options are specified which use symspecs. 'gprof' maintains a single d1767 1 a1767 1 tables, organized into six include/exclude pairs--one pair each for the d1775 2 a1776 2 After option processing, 'gprof' finishes building the symspec list by adding all the symspecs in 'default_excluded_list' to the exclude d1782 2 a1783 2 it is an object file, and read its symbol table ('core.c:core_init'), using 'bfd_canonicalize_symtab' after mallocing an appropriately sized d1785 2 a1786 2 '--file-ordering' option has been specified), and the core text space is read into memory (if the '-c' option was given). d1788 1 a1788 1 'gprof''s own symbol table, an array of Sym structures, is now built. d1790 1 a1790 1 whether line-by-line profiling ('-l' option) has been enabled. For d1794 1 a1794 1 either case, two passes are made through the symbol table--one to count d1796 2 a1797 2 the symbols. In between the two passes, a single array of type 'Sym' is created of the appropriate length. Finally, 'symtab.c:symtab_finalize' d1802 1 a1802 1 the 'qsort' library function (which sorts an array) will be used to sort d1804 1 a1804 1 ('symtab.c:sym_lookup'), which finds symbols based on memory address, d1806 1 a1806 1 sorted array. Function symbols are indicated with an 'is_func' flag. d1808 1 a1808 1 can have an 'is_static' flag to indicate that it is a local symbol. d1811 2 a1812 2 Syms ('sym_ids.c:sym_id_parse'). Remember that a single symspec can match multiple symbols. An array of symbol tables ('syms') is created, d1820 2 a1821 2 'gprof' simply uses its standard symbol lookup routine on the appropriate table in the 'syms' array. d1824 2 a1825 2 ('gmon_io.c:gmon_out_read'), first by checking for a new-style 'gmon.out' header, then assuming this is an old-style BSD 'gmon.out' if d1828 1 a1828 1 New-style histogram records are read by 'hist.c:hist_read_rec'. For d1838 1 a1838 1 As each call graph record is read ('call_graph.c:cg_read_rec'), the d1840 1 a1840 1 call graph arc is created by 'cg_arcs.c:arc_add', unless the arc fails a d1842 3 a1844 3 linked list is maintained of the parent's child arcs, and of the child's parent arcs. Both the child's call count and the arc's call count are incremented by the record's call count. d1846 1 a1846 1 Basic-block records are read ('basic_blocks.c:bb_read_rec'), but only d1849 1 a1849 1 entry made in the symbol's bb_addr and bb_calls arrays. Again, if d1853 1 a1853 1 A gmon.sum file is dumped, if requested ('gmon_io.c:gmon_out_write'). d1856 1 a1856 1 ('hist.c:hist_assign_samples') by iterating over all the sample bins and d1867 3 a1869 3 If call graph data is present, 'cg_arcs.c:cg_assemble' is called. First, if '-c' was specified, a machine-dependent routine ('find_call') scans through each symbol's machine code, looking for subroutine call d1872 1 a1872 1 ('cg_dfn.c:cg_dfn'), so that children are always numbered less than d1881 1 a1881 1 a parent's include or exclude (print or no print) property being d1891 2 a1892 2 fairly straightforward. The call graph ('cg_print.c:cg_print') and flat profile ('hist.c:hist_print') are regurgitations of values already d1894 1 a1894 1 ('basic_blocks.c:print_annotated_source') uses basic-block information, d1899 1 a1899 1 source code itself ('cg_print.c'). Basically, the functions with the d1907 1 a1907 1 9.4 Debugging 'gprof' d1910 1 a1910 1 If 'gprof' was compiled with debugging enabled, the '-d' option triggers d1930 1 a1930 1 file. For line-by-line profiling ('-l' option), also shows line d1933 1 a1933 1 Trace operation of '-c' option d1940 1 a1940 1 meaningful with '-l' option) d1944 1 a1944 1 Tracks operation of '-A' option d1954 1 a1954 1 Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. d1963 1 a1963 1 functional and useful document "free" in the sense of freedom: to d1970 1 a1970 1 This License is a kind of "copyleft", which means that derivative d1991 2 a1992 2 "Document", below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "you". You accept d1996 1 a1996 1 A "Modified Version" of the Document means any work containing the d2000 1 a2000 1 A "Secondary Section" is a named appendix or a front-matter section d2002 1 a2002 1 publishers or authors of the Document to the Document's overall d2011 1 a2011 1 The "Invariant Sections" are certain Secondary Sections whose d2019 1 a2019 1 The "Cover Texts" are certain short passages of text that are d2025 1 a2025 1 A "Transparent" copy of the Document means a machine-readable copy, d2037 1 a2037 1 "Transparent" is called "Opaque". d2050 1 a2050 1 The "Title Page" means, for a printed book, the title page itself, d2053 3 a2055 3 works in formats which do not have any title page as such, "Title Page" means the text near the most prominent appearance of the work's title, preceding the beginning of the body of the text. d2057 1 a2057 1 The "publisher" means any person or entity that distributes copies d2060 1 a2060 1 A section "Entitled XYZ" means a named subunit of the Document d2064 3 a2066 3 "Acknowledgements", "Dedications", "Endorsements", or "History".) To "Preserve the Title" of such a section when you modify the Document means that it remains a section "Entitled XYZ" according d2096 1 a2096 1 the Document's license notice requires Cover Texts, you must d2168 1 a2168 1 Sections and required Cover Texts given in the Document's d2173 1 a2173 1 I. Preserve the section Entitled "History", Preserve its Title, d2176 1 a2176 1 Title Page. If there is no section Entitled "History" in the d2186 1 a2186 1 "History" section. You may omit a network location for a work d2191 1 a2191 1 K. For any section Entitled "Acknowledgements" or "Dedications", d2200 1 a2200 1 M. Delete any section Entitled "Endorsements". Such a section d2204 1 a2204 1 "Endorsements" or to conflict in title with any Invariant d2213 1 a2213 1 titles to the list of Invariant Sections in the Modified Version's d2217 1 a2217 1 You may add a section Entitled "Endorsements", provided it contains d2219 3 a2221 3 parties--for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2259 4 a2262 4 "History" in the various original documents, forming one section Entitled "History"; likewise combine any sections Entitled "Acknowledgements", and any sections Entitled "Dedications". You must delete all sections Entitled "Endorsements." d2283 1 a2283 1 storage or distribution medium, is called an "aggregate" if the d2285 1 a2285 1 legal rights of the compilation's users beyond what the individual d2292 1 a2292 1 of the entire aggregate, the Document's Cover Texts may be placed d2314 2 a2315 2 If a section in the Document is Entitled "Acknowledgements", "Dedications", or "History", the requirement (section 4) to d2356 1 a2356 1 version of this License "or any later version" applies to it, you d2364 1 a2364 1 proxy's public statement of acceptance of a version permanently d2369 1 a2369 1 "Massive Multiauthor Collaboration Site" (or "MMC Site") means any d2373 1 a2373 1 A "Massive Multiauthor Collaboration" (or "MMC") contained in the d2377 1 a2377 1 "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 d2383 1 a2383 1 "Incorporate" means to publish or republish a Document, in whole or d2386 1 a2386 1 An MMC is "eligible for relicensing" if it is licensed under this d2413 1 a2413 1 Texts, replace the "with...Texts." line with this: d2431 30 a2460 30 Node: Top719 Node: Introduction2042 Node: Compiling4533 Node: Executing8589 Node: Invoking11482 Node: Output Options12897 Node: Analysis Options19989 Node: Miscellaneous Options23909 Node: Deprecated Options25163 Node: Symspecs27226 Node: Output29052 Node: Flat Profile30092 Node: Call Graph35045 Node: Primary38277 Node: Callers40865 Node: Subroutines42983 Node: Cycles44824 Node: Line-by-line51601 Node: Annotated Source55677 Node: Inaccuracy58675 Node: Sampling Error58933 Node: Assumptions61837 Node: How do I?63307 Node: Incompatibilities64864 Node: Details66358 Node: Implementation66751 Node: File Format72650 Node: Internals76942 Node: Debugging85432 Node: GNU Free Documentation License87022 @ 1.9.2.1 log @Sync with HEAD. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 7.0.2 from gprof.texi. d5 1 a5 1 Copyright © 1988-2024 Free Software Foundation, Inc. d11 2 a12 2 Texts. A copy of the license is included in the section entitled “GNU Free Documentation License”. d16 1 a16 1 * gprof: (gprof). Profiling your program’s execution d25 1 a25 1 This manual describes the GNU profiler, ‘gprof’, and how you can use it d28 1 a28 1 programs. GNU ‘gprof’ was written by Jay Fenlason. d30 1 a30 1 This manual is for ‘gprof’ (GNU Binutils) version 2.42. d34 1 a34 1 the section entitled “GNU Free Documentation License”. d42 1 a42 1 * Invoking:: How to run ‘gprof’, and its options d44 1 a44 1 * Output:: Interpreting ‘gprof’’s output d48 1 a48 1 * Incompatibilities:: (between GNU ‘gprof’ and Unix ‘gprof’.) d70 1 a70 1 data. If you don’t use some feature of your program while it is being d75 1 a75 1 • You must compile and link your program with profiling enabled. d78 1 a78 1 • You must execute your program to generate a profile data file. d81 1 a81 1 • You must run ‘gprof’ to analyze the profile data. *Note ‘gprof’ d88 1 a88 1 The “flat profile” shows how much time your program spent in each d93 1 a93 1 The “call graph” shows, for each function, which functions called it, d99 1 a99 1 The “annotated source” listing is a copy of the program’s source d116 1 a116 1 To compile a source file for profiling, specify the ‘-pg’ option when d120 3 a122 3 To link the program for profiling, if you use a compiler such as ‘cc’ to do the linking, simply specify ‘-pg’ in addition to your usual options. The same option, ‘-pg’, alters either compilation or linking d128 1 a128 1 The ‘-pg’ option also works with a command that both compiles and d133 1 a133 1 Note: The ‘-pg’ option must be part of your compilation options as d135 1 a135 1 gathered and when you run ‘gprof’ you will get an error message like d140 1 a140 1 If you add the ‘-Q’ switch to suppress the printing of the call graph d152 3 a154 3 If you run the linker ‘ld’ directly instead of through a compiler such as ‘cc’, you may have to specify a profiling startup file ‘gcrt0.o’ as the first input file instead of the usual startup file ‘crt0.o’. In d156 1 a156 1 ‘libc_p.a’, by writing ‘-lc_p’ instead of the usual ‘-lc’. This is not d158 1 a158 1 information for standard library functions such as ‘read’ and ‘open’. d169 2 a170 2 code, which for ‘gcc’ users can be done via the ‘-static’ or ‘-static-libgcc’ command-line option. For example: d174 3 a176 3 If you compile only some of the modules of the program with ‘-pg’, you can still profile the program, but you won’t get complete information about the modules that were compiled without ‘-pg’. The d180 1 a180 1 that the ‘calls’ field for the functions will be blank), but will d184 1 a184 1 ‘gcov’ tool instead of ‘gprof’. See that tool’s manual or info pages d187 3 a189 3 Note, older versions of ‘gcc’ produce line-by-line profiling information that works with ‘gprof’ rather than ‘gcov’ so there is still support for displaying this kind of information in ‘gprof’. *Note d192 1 a192 1 It also worth noting that ‘gcc’ implements a ‘-finstrument-functions’ d205 1 a205 1 generate the information that ‘gprof’ needs. Simply run the program as d211 2 a212 2 The way you run the program—the arguments and input that you give it—may have a dramatic effect on what the profile information shows. d219 2 a220 2 ‘gmon.out’ just before exiting. If there is already a file called ‘gmon.out’, its contents are overwritten. You can rename the file d226 3 a228 3 In order to write the ‘gmon.out’ file properly, your program must exit normally: by returning from ‘main’ or by calling ‘exit’. Calling the low-level function ‘_exit’ does not write the profile data, and d231 1 a231 1 The ‘gmon.out’ file is written in the program’s _current working d233 2 a234 2 ‘chdir’, the ‘gmon.out’ file will be left in the last directory your program ‘chdir’’d to. If you don’t have permission to write in this d238 1 a238 1 called ‘bb.out’. This file, if present, contains an human-readable d240 4 a243 4 appearance of a human-readable ‘bb.out’ means the basic-block counts didn’t get written into ‘gmon.out’. The Perl script ‘bbconv.pl’, included with the ‘gprof’ source distribution, will convert a ‘bb.out’ file into a format readable by ‘gprof’. Invoke it like this: d247 2 a248 2 This translates the information in ‘bb.out’ into a form that ‘gprof’ can understand. But you still need to tell ‘gprof’ about the existence d250 1 a250 1 ‘gprof’ command line, _along with ‘gmon.out’_, like this: d257 1 a257 1 4 ‘gprof’ Command Summary d260 2 a261 2 After you have a profile data file ‘gmon.out’, you can run ‘gprof’ to interpret the information in it. The ‘gprof’ program prints a flat d263 1 a263 1 redirect the output of ‘gprof’ into a file with ‘>’. d265 1 a265 1 You run ‘gprof’ like this: d271 2 a272 2 If you omit the executable file name, the file ‘a.out’ is used. If you give no profile data file name, the file ‘gmon.out’ is used. If any d284 2 a285 2 * Output Options:: Controlling ‘gprof’’s output style * Analysis Options:: Controlling how ‘gprof’ analyzes its data d297 1 a297 1 These options specify which of several output formats ‘gprof’ should d300 1 a300 1 Many of these options take an optional “symspec” to specify functions d305 1 a305 1 Specifying any of these options overrides the default (‘-p -q’), d308 3 a310 3 ‘-A[SYMSPEC]’ ‘--annotated-source[=SYMSPEC]’ The ‘-A’ option causes ‘gprof’ to print annotated source code. If d314 3 a316 3 ‘-b’ ‘--brief’ If the ‘-b’ option is given, ‘gprof’ doesn’t print the verbose d321 3 a323 6 ‘-B’ The ‘-B’ option causes ‘gprof’ to print the call graph analysis. ‘-C[SYMSPEC]’ ‘--exec-counts[=SYMSPEC]’ The ‘-C’ option causes ‘gprof’ to print a tally of functions and d328 1 a328 1 specifying the ‘-l’ option, along with ‘-C’, will cause basic-block d331 3 a333 3 ‘-i’ ‘--file-info’ The ‘-i’ option causes ‘gprof’ to display summary information about d337 3 a339 3 ‘-I DIRS’ ‘--directory-path=DIRS’ The ‘-I’ option specifies a list of search directories in which to d344 4 a347 4 ‘-J[SYMSPEC]’ ‘--no-annotated-source[=SYMSPEC]’ The ‘-J’ option causes ‘gprof’ not to print annotated source code. If SYMSPEC is specified, ‘gprof’ prints annotated source, but d350 2 a351 2 ‘-L’ ‘--print-path’ d353 1 a353 1 suppressed. The ‘-L’ option causes ‘gprof’ to print the full d358 3 a360 3 ‘-p[SYMSPEC]’ ‘--flat-profile[=SYMSPEC]’ The ‘-p’ option causes ‘gprof’ to print a flat profile. If SYMSPEC d364 4 a367 4 ‘-P[SYMSPEC]’ ‘--no-flat-profile[=SYMSPEC]’ The ‘-P’ option causes ‘gprof’ to suppress printing a flat profile. If SYMSPEC is specified, ‘gprof’ prints a flat profile, but d370 3 a372 3 ‘-q[SYMSPEC]’ ‘--graph[=SYMSPEC]’ The ‘-q’ option causes ‘gprof’ to print the call graph analysis. d376 4 a379 4 ‘-Q[SYMSPEC]’ ‘--no-graph[=SYMSPEC]’ The ‘-Q’ option causes ‘gprof’ to suppress printing the call graph. If SYMSPEC is specified, ‘gprof’ prints a call graph, but excludes d382 3 a384 3 ‘-t’ ‘--table-length=NUM’ The ‘-t’ option causes the NUM most active source lines in each d388 2 a389 2 ‘-y’ ‘--separate-files’ d391 1 a391 1 ‘gprof’ prints annotated source files to standard-output. If this d393 9 a401 9 ‘path/FILENAME’ is generated in the file ‘FILENAME-ann’. If the underlying file system would truncate ‘FILENAME-ann’ so that it overwrites the original ‘FILENAME’, ‘gprof’ generates annotated source in the file ‘FILENAME.ann’ instead (if the original file name has an extension, that extension is _replaced_ with ‘.ann’). ‘-Z[SYMSPEC]’ ‘--no-exec-counts[=SYMSPEC]’ The ‘-Z’ option causes ‘gprof’ not to print a tally of functions d405 3 a407 3 ‘-r’ ‘--function-ordering’ The ‘--function-ordering’ option causes ‘gprof’ to print a d417 3 a419 3 ‘-R MAP_FILE’ ‘--file-ordering MAP_FILE’ The ‘--file-ordering’ option causes ‘gprof’ to print a suggested .o d425 1 a425 1 Use of the ‘-a’ argument is highly recommended with this option. d429 1 a429 1 similar to the output of the program ‘nm’. d441 2 a442 2 To create a MAP_FILE with GNU ‘nm’, type a command like ‘nm --extern-only --defined-only -v --print-file-name program-name’. d444 3 a446 3 ‘-T’ ‘--traditional’ The ‘-T’ option causes ‘gprof’ to print its output in “traditional” d449 2 a450 2 ‘-w WIDTH’ ‘--width=WIDTH’ d454 2 a455 2 ‘-x’ ‘--all-lines’ d460 1 a460 1 similar to ‘tcov’’s ‘-a’. d462 2 a463 2 ‘--demangle[=STYLE]’ ‘--no-demangle’ d466 1 a466 1 ‘--no-demangle’ option may be used to turn off demangling. d477 3 a479 3 ‘-a’ ‘--no-static’ The ‘-a’ option causes ‘gprof’ to suppress the printing of d488 3 a490 3 ‘-c’ ‘--static-call-graph’ The ‘-c’ option causes the call graph of the program to be d500 1 a500 1 graph with call counts of ‘0’. d502 3 a504 3 ‘-D’ ‘--ignore-non-functions’ The ‘-D’ option causes ‘gprof’ to ignore symbols which are not d509 2 a510 2 ‘-k FROM/TO’ The ‘-k’ option allows you to delete from the call graph any arcs d513 3 a515 3 ‘-l’ ‘--line’ The ‘-l’ option enables line-by-line profiling, which causes d518 2 a519 2 compiled by older versions of the ‘gcc’ compiler. Newer versions of ‘gcc’ are designed to work with the ‘gcov’ tool instead. d525 1 a525 1 significantly increases the running time of ‘gprof’, and magnifies d529 2 a530 2 ‘--inline-file-names’ This option causes ‘gprof’ to print the source file after each d532 1 a532 1 to the file is printed if used with the ‘-L’ option. d534 2 a535 2 ‘-m NUM’ ‘--min-count=NUM’ d539 3 a541 3 ‘-nSYMSPEC’ ‘--time=SYMSPEC’ The ‘-n’ option causes ‘gprof’, in its call graph analysis, to only d544 3 a546 3 ‘-NSYMSPEC’ ‘--no-time=SYMSPEC’ The ‘-n’ option causes ‘gprof’, in its call graph analysis, not to d549 5 a553 5 ‘-SFILENAME’ ‘--external-symbol-table=FILENAME’ The ‘-S’ option causes ‘gprof’ to read an external symbol table file, such as ‘/proc/kallsyms’, rather than read the symbol table from the given object file (the default is ‘a.out’). This is d556 3 a558 3 ‘-z’ ‘--display-unused-functions’ If you give the ‘-z’ option, ‘gprof’ will mention all functions in d560 1 a560 1 no time spent in them. This is useful in conjunction with the ‘-c’ d569 4 a572 4 ‘-d[NUM]’ ‘--debug[=NUM]’ The ‘-d NUM’ option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging ‘gprof’: d575 3 a577 3 ‘-h’ ‘--help’ The ‘-h’ option prints command line usage. d579 2 a580 2 ‘-ONAME’ ‘--file-format=NAME’ d582 1 a582 1 are ‘auto’ (the default), ‘bsd’, ‘4.4bsd’, ‘magic’, and ‘prof’ (not d585 3 a587 3 ‘-s’ ‘--sum’ The ‘-s’ option causes ‘gprof’ to summarize the information in the d589 2 a590 2 called ‘gmon.sum’, which contains all the information from the profile data files that ‘gprof’ read in. The file ‘gmon.sum’ may d592 1 a592 1 the data in the other input files into ‘gmon.sum’. d594 2 a595 2 Eventually you can run ‘gprof’ again without ‘-s’ to analyze the cumulative data in the file ‘gmon.sum’. d597 3 a599 3 ‘-v’ ‘--version’ The ‘-v’ flag causes ‘gprof’ to print the current version number, d610 2 a611 2 ‘-e FUNCTION_NAME’ The ‘-e FUNCTION’ option tells ‘gprof’ to not print information d614 3 a616 3 functions that call it, but its index number will be shown as ‘[not printed]’. More than one ‘-e’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-e’ option. d618 2 a619 2 ‘-E FUNCTION_NAME’ The ‘-E FUNCTION’ option works like the ‘-e’ option, but time spent d622 2 a623 2 call graph. More than one ‘-E’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-E’ option. d625 2 a626 2 ‘-f FUNCTION_NAME’ The ‘-f FUNCTION’ option causes ‘gprof’ to limit the call graph to d628 2 a629 2 children...). More than one ‘-f’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-f’ option. d631 2 a632 2 ‘-F FUNCTION_NAME’ The ‘-F FUNCTION’ option works like the ‘-f’ option, but only time d635 3 a637 3 call graph. More than one ‘-F’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-F’ option. The ‘-F’ option overrides the ‘-E’ option. d639 2 a640 2 Note that only one function can be specified with each ‘-e’, ‘-E’, ‘-f’ or ‘-F’ option. To specify more than one function, use multiple d646 1 a646 1 ‘foo’ or ‘bar’ and were not reachable from ‘boring’. d655 1 a655 1 using “symspecs” (symbol specifications), which observe the following d665 3 a667 3 ‘main.c’ Selects everything in file ‘main.c’—the dot in the string tells ‘gprof’ to interpret the string as a filename, rather than as a d669 2 a670 2 a trailing colon should be specified. For example, ‘odd:’ is interpreted as the file named ‘odd’. d672 2 a673 2 ‘main’ Selects all functions named ‘main’. d682 2 a683 2 necessary to add a leading colon to the name. For example, ‘:.mul’ selects function ‘.mul’. d686 2 a687 2 ‘gprof’ will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as ‘gprof’ prints d689 2 a690 2 ‘_main’ from your ‘main’ function, ‘gprof’ still prints it as ‘main’ in its output, so you should use ‘main’ in symspecs. d692 2 a693 2 ‘main.c:main’ Selects function ‘main’ in file ‘main.c’. d695 2 a696 2 ‘main.c:134’ Selects line 134 in file ‘main.c’. d701 1 a701 1 5 Interpreting ‘gprof’’s Output d704 1 a704 1 ‘gprof’ can produce several different output styles, the most important d717 1 a717 1 * Line-by-line:: ‘gprof’ can analyze individual source code lines d727 2 a728 2 The “flat profile” shows the total amount of time your program spent executing each function. Unless the ‘-z’ option is given, functions d731 1 a731 1 and didn’t run long enough to show up on the program counter histogram, d759 1 a759 1 functions ‘mcount’ and ‘profil’ are part of the profiling apparatus and d764 1 a764 1 much time each sample counted as. This “sampling period” estimates the d768 2 a769 2 program’s total execution time was 0.06 seconds, as indicated by the ‘cumulative seconds’ field. Since each sample counted for 0.01 seconds, d771 3 a773 3 samples occurred while the program was in the ‘open’ function, as indicated by the ‘self seconds’ field. Each of the other four samples occurred one each in ‘offtime’, ‘memccpy’, ‘write’, and ‘mcount’. Since d775 2 a776 2 particularly reliable. In another run, the ‘self seconds’ field for ‘mcount’ might well be ‘0.00’ or ‘0.02’. *Note Statistical Sampling d779 2 a780 2 The remaining functions in the listing (those whose ‘self seconds’ field is ‘0.00’) didn’t appear in the histogram samples at all. d782 1 a782 1 they are listed, sorted in decreasing order by the ‘calls’ field. d789 1 a789 1 ‘% time’ d793 1 a793 1 ‘cumulative seconds’ d798 1 a798 1 ‘self seconds’ d802 1 a802 1 ‘calls’ d806 1 a806 1 compiled with profiling enabled), the “calls” field is blank. d808 1 a808 1 ‘self ms/call’ d813 1 a813 1 ‘total ms/call’ d820 1 a820 1 ‘name’ d822 1 a822 1 this field alphabetically after the “self seconds” and “calls” d831 1 a831 1 The “call graph” shows how much time was spent in each function and its d837 1 a837 1 same ‘gprof’ run as the flat profile example in the previous section. d868 1 a868 1 The lines full of dashes divide this table into “entries”, one for d875 1 a875 1 called “children” when we speak of the call graph). d880 1 a880 1 The internal profiling function ‘mcount’ (*note The Flat Profile: d885 3 a887 3 * Primary:: Details of the primary line’s contents. * Callers:: Details of caller-lines’ contents. * Subroutines:: Details of subroutine-lines’ contents. d889 1 a889 1 such as ‘a’ calls ‘b’ calls ‘a’... d897 1 a897 1 The “primary line” in a call graph entry is the line that describes the d902 1 a902 1 ‘report’ in our main example, together with the heading line that shows d911 1 a911 1 ‘index’ d921 1 a921 1 ‘% time’ d930 1 a930 1 ‘self’ d932 1 a932 1 should be identical to the number printed in the ‘seconds’ field d935 1 a935 1 ‘children’ d938 1 a938 1 ‘self’ and ‘children’ entries of the children listed directly below d941 1 a941 1 ‘called’ d945 1 a945 1 separated by a ‘+’. The first number counts non-recursive calls, d948 2 a949 2 In the example above, the function ‘report’ was called once from ‘main’. d951 1 a951 1 ‘name’ d956 1 a956 1 is printed between the function’s name and the index number (*note d958 1 a958 1 example, if function ‘gnurr’ is part of cycle number one, and has d966 1 a966 1 5.2.2 Lines for a Function’s Callers d969 2 a970 2 A function’s entry has a line for each function it was called by. These lines’ fields correspond to the fields of the primary line, but their d974 1 a974 1 ‘report’, the primary line and one caller-line preceding it, together d982 2 a983 2 Here are the meanings of the fields in the caller-line for ‘report’ called from ‘main’: d985 14 a998 14 ‘self’ An estimate of the amount of time spent in ‘report’ itself when it was called from ‘main’. ‘children’ An estimate of the amount of time spent in subroutines of ‘report’ when ‘report’ was called from ‘main’. The sum of the ‘self’ and ‘children’ fields is an estimate of the amount of time spent within calls to ‘report’ from ‘main’. ‘called’ Two numbers: the number of times ‘report’ was called from ‘main’, followed by the total number of non-recursive calls to ‘report’ d1001 3 a1003 3 ‘name and index number’ The name of the caller of ‘report’ to which this line applies, followed by the caller’s index number. d1006 1 a1006 1 ‘gprof’ request the omission of certain functions. When a caller d1014 2 a1015 2 dummy caller-line is printed which has ‘’ as the “caller’s name” and all other fields blank. This can happen for signal handlers. d1020 1 a1020 1 5.2.3 Lines for a Function’s Subroutines d1023 2 a1024 2 A function’s entry has a line for each of its subroutines—in other words, a line for each other function that it called. These lines’ d1029 1 a1029 1 ‘main’, the primary line and a line for a subroutine, together with the d1037 2 a1038 2 Here are the meanings of the fields in the subroutine-line for ‘main’ calling ‘report’: d1040 16 a1055 16 ‘self’ An estimate of the amount of time spent directly within ‘report’ when ‘report’ was called from ‘main’. ‘children’ An estimate of the amount of time spent in subroutines of ‘report’ when ‘report’ was called from ‘main’. The sum of the ‘self’ and ‘children’ fields is an estimate of the total time spent in calls to ‘report’ from ‘main’. ‘called’ Two numbers, the number of calls to ‘report’ from ‘main’ followed by the total number of non-recursive calls to ‘report’. This ratio is used to determine how much of ‘report’’s ‘self’ and ‘children’ time gets credited to ‘main’. *Note Estimating ‘children’ Times: d1058 3 a1060 3 ‘name’ The name of the subroutine of ‘main’ to which this line applies, followed by the subroutine’s index number. d1071 1 a1071 1 The graph may be complicated by the presence of “cycles of recursion” in d1074 2 a1075 2 function. For example: if ‘a’ calls ‘b’, and ‘b’ calls ‘a’, then ‘a’ and ‘b’ form a cycle. d1078 4 a1081 4 they belong to the same cycle. If ‘a’ and ‘b’ call each other and ‘b’ and ‘c’ call each other, all three make one cycle. Note that even if ‘b’ only calls ‘a’ if it was not called from ‘a’, ‘gprof’ cannot determine this, so ‘a’ and ‘b’ are still considered a cycle. d1085 1 a1085 1 graph it is followed by ‘’. d1088 4 a1091 4 call graph paradoxical. The “time spent in children” of ‘a’ should include the time spent in its subroutine ‘b’ and in ‘b’’s subroutines—but one of ‘b’’s subroutines is ‘a’! How much of ‘a’’s time should be included in the children of ‘a’, when ‘a’ is indirectly d1094 1 a1094 1 The way ‘gprof’ resolves this paradox is by creating a single entry d1097 1 a1097 1 “subroutines” of the cycle are the individual functions of the cycle, d1099 1 a1099 1 “callers” of the cycle are the functions, outside the cycle, that called d1103 2 a1104 2 containing functions ‘a’ and ‘b’. The cycle was entered by a call to ‘a’ from ‘main’; both ‘a’ and ‘b’ called ‘c’. d1126 2 a1127 2 for ‘main’, which calls ‘a’, and an entry for ‘c’, with callers ‘a’ and ‘b’.) d1160 2 a1161 2 The ‘self’ field of the cycle’s primary line is the total time spent in all the functions of the cycle. It equals the sum of the ‘self’ d1165 6 a1170 6 The ‘children’ fields of the cycle’s primary line and subroutine lines count only subroutines outside the cycle. Even though ‘a’ calls ‘b’, the time spent in those calls to ‘b’ is not counted in ‘a’’s ‘children’ time. Thus, we do not encounter the problem of what to do when the time in those calls to ‘b’ includes indirect recursive calls back to ‘a’. d1172 1 a1172 1 The ‘children’ field of a caller-line in the cycle’s entry estimates d1177 1 a1177 1 The ‘called’ field in the primary line for the cycle has two numbers: d1184 2 a1185 2 The ‘called’ field of a subroutine-line for a cycle member in the cycle’s entry says how many time that function was called from functions d1187 1 a1187 1 primary line’s ‘called’ field. d1192 1 a1192 1 called from each other function in the cycle. The ‘self’ and ‘children’ d1202 1 a1202 1 ‘gprof’’s ‘-l’ option causes the program to perform “line-by-line” d1205 2 a1206 2 programs compiled with older versions of the ‘gcc’ compiler. Newer versions of ‘gcc’ use a different program - ‘gcov’ - to display d1209 2 a1210 2 With the older versions of ‘gcc’ the program usually has to be compiled with a ‘-g’ option, in addition to ‘-pg’, in order to generate d1212 1 a1212 1 versions of ‘gcc’ the program had to be compiled with the ‘-a’ d1216 2 a1217 2 mode. The call graph isn’t as useful as normal, since the current version of ‘gprof’ does not propagate call graph arcs from source code d1221 3 a1223 3 Here is a section of ‘gprof’’s output, without line-by-line profiling. Note that ‘ct_init’ accounted for four histogram hits, and 13327 calls to ‘init_block’. d1247 3 a1249 3 Now let’s look at some of ‘gprof’’s output from the same program run, this time with line-by-line profiling enabled. Note that ‘ct_init’’s four histogram hits are broken down into four lines of source code—one d1251 1 a1251 1 note how ‘ct_init’’s 13327 calls to ‘init_block’ are broken down into d1294 4 a1297 4 ‘gprof’’s ‘-A’ option triggers an annotated source listing, which lists the program’s source code, each function labeled with the number of times it was called. You may also need to specify the ‘-I’ option, if ‘gprof’ can’t find the source code files. d1299 1 a1299 1 With older versions of ‘gcc’ compiling with ‘gcc ... -g -pg -a’ d1301 3 a1303 3 function counting code. This enables ‘gprof’ to determine how many times each line of code was executed. With newer versions of ‘gcc’ support for displaying basic-block counts is provided by the ‘gcov’ d1330 4 a1333 4 ‘updcrc’ has at least five basic-blocks. One is the function itself. The ‘if’ statement on line 9 generates two more basic-blocks, one for each branch of the ‘if’. A fourth basic-block results from the ‘if’ on line 13, and the contents of the ‘do’ loop form the fifth basic-block. d1338 3 a1340 3 ‘gprof -l -A’. The ‘-x’ option is also helpful, to ensure that each line of code is labeled at least once. Here is ‘updcrc’’s annotated source listing for a sample ‘gzip’ run: d1363 2 a1364 2 each branch of the ‘if’ statement. The body of the ‘do’ loop was executed a total of 26312 times. Note how the ‘while’ statement is d1372 1 a1372 1 6 Inaccuracy of ‘gprof’ Output d1386 1 a1386 1 The run-time figures that ‘gprof’ gives you are based on a sampling d1402 1 a1402 1 The “sampling period” that is printed at the beginning of the flat d1409 1 a1409 1 period is 0.01 seconds and ‘foo’’s run-time is 1 second, N is 100 d1411 1 a1411 1 error in ‘foo’’s run-time is 0.1 seconds (10*0.01 seconds), or ten d1413 2 a1414 2 seconds and ‘bar’’s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in ‘bar’’s run-time is 1 second, d1420 1 a1420 1 information. If the program’s _total_ run-time is large, a small d1422 1 a1422 1 insignificant fraction of the whole program’s time. Usually this means d1427 1 a1427 1 the data from several runs, using the ‘-s’ option of ‘gprof’. Here is d1432 1 a1432 1 2. Issue the command ‘mv gmon.out gmon.sum’. d1436 1 a1436 1 4. Merge the new data in ‘gmon.out’ into ‘gmon.sum’ with this command: d1449 1 a1449 1 6.2 Estimating ‘children’ Times d1452 2 a1453 2 Some of the figures in the call graph are estimates—for example, the ‘children’ time values and all the time figures in caller and subroutine d1457 1 a1457 1 profile data itself. Instead, ‘gprof’ estimates them by making an d1461 3 a1463 3 any function ‘foo’ is not correlated with who called ‘foo’. If ‘foo’ used 5 seconds in all, and 2/5 of the calls to ‘foo’ came from ‘a’, then ‘foo’ contributes 2 seconds to ‘a’’s ‘children’ time, by assumption. d1466 6 a1471 6 far from true. Suppose that ‘foo’ returns very quickly when its argument is zero; suppose that ‘a’ always passes zero as an argument, while other callers of ‘foo’ pass other arguments. In this program, all the time spent in ‘foo’ is in the calls from callers other than ‘a’. But ‘gprof’ has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in ‘foo’ to the children of ‘a’. d1473 1 a1473 1 We hope some day to put more complete data into ‘gmon.out’, so that d1486 1 a1486 1 Because ‘gprof’ can only report call times and counts by function, d1490 1 a1490 1 artificial hot spots since compiling with ‘-pg’ adds a significant d1496 1 a1496 1 Use the ‘gcov’ program. d1500 1 a1500 1 Use ‘gprof -l’ and lookup the function in the call graph. The d1523 1 a1523 1 8 Incompatibilities with Unix ‘gprof’ d1526 1 a1526 1 GNU ‘gprof’ and Berkeley Unix ‘gprof’ use the same data file ‘gmon.out’, d1530 1 a1530 1 • GNU ‘gprof’ uses a new, generalized file format with support for d1532 1 a1532 1 cookie and version number allows ‘gprof’ to easily identify new d1536 3 a1538 3 • For a recursive function, Unix ‘gprof’ lists the function as a parent and as a child, with a ‘calls’ field that lists the number of recursive calls. GNU ‘gprof’ omits these lines and puts the d1541 2 a1542 2 • When a function is suppressed from the call graph with ‘-e’, GNU ‘gprof’ still lists it as a subroutine of functions that call it. d1544 2 a1545 2 • GNU ‘gprof’ accepts the ‘-k’ with its argument in the form ‘from/to’, instead of ‘from to’. d1547 2 a1548 2 • In the annotated source listing, if there are multiple basic blocks on the same line, GNU ‘gprof’ prints all of their counts, separated d1551 2 a1552 2 • The blurbs, field widths, and output formats are different. GNU ‘gprof’ prints blurbs after the tables, so that you can see the d1564 3 a1566 3 * File Format:: Format of ‘gmon.out’ files * Internals:: ‘gprof’’s internal operation * Debugging:: Using ‘gprof’’s ‘-d’ option d1579 2 a1580 2 the ‘-pg’ option, which causes every function to call ‘mcount’ (or ‘_mcount’, or ‘__mcount’, depending on the OS and compiler) as one of d1583 1 a1583 1 The ‘mcount’ routine, included in the profiling library, is d1585 1 a1585 1 parent routine (the child) and its parent’s parent. This is typically d1588 1 a1588 1 machine-dependent operation, ‘mcount’ itself is typically a short d1590 4 a1593 4 and then calls ‘__mcount_internal’ (a normal C function) with two arguments—‘frompc’ and ‘selfpc’. ‘__mcount_internal’ is responsible for maintaining the in-memory call graph, which records ‘frompc’, ‘selfpc’, and the number of times each of these call arcs was traversed. d1596 1 a1596 1 (‘__builtin_return_address’), which allows a generic ‘mcount’ function d1600 1 a1600 1 ‘mcount’ is used for performance reasons. d1604 2 a1605 2 same as in the usual C library, but they were compiled with ‘-pg’. If you link your program with ‘gcc ... -pg’, it automatically uses the d1615 2 a1616 2 provide a ‘profil()’ system call, which registers a memory array with the kernel, along with a scale factor that determines how the program’s d1626 1 a1626 1 earlier), do not provide a ‘profil()’ system call. On such a system, d1628 1 a1628 1 the process (typically via ‘setitimer()’), which then performs the same d1636 5 a1640 5 either calls ‘profil()’ or sets up a clock signal handler. This routine (‘monstartup’) can be invoked in several ways. On Linux systems, a special profiling startup file ‘gcrt0.o’, which invokes ‘monstartup’ before ‘main’, is used instead of the default ‘crt0.o’. Use of this special startup file is one of the effects of using ‘gcc ... -pg’ to d1642 2 a1643 2 ‘mcount’ routine, when it is invoked for the first time (typically when ‘main’ is called), calls ‘monstartup’. d1645 1 a1645 1 If the compiler’s ‘-a’ option was used, basic-block counting is also d1648 1 a1648 1 basic-block begins (i.e., when an ‘if’ statement appears), an extra d1655 3 a1657 3 The profiling library also includes a function (‘mcleanup’) which is typically registered using ‘atexit()’ to be called as the program exits, and is responsible for writing the file ‘gmon.out’. Profiling is turned d1661 1 a1661 1 The output from ‘gprof’ gives no indication of parts of your program d1663 2 a1664 2 of the program counter are taken at fixed intervals of the program’s run time. Therefore, the time measurements in ‘gprof’ output say nothing d1667 1 a1667 1 memory at once may run very slowly due to thrashing, but ‘gprof’ will d1669 1 a1669 1 the advantage that the amount of load due to other users won’t directly d1680 2 a1681 2 ‘gprof’ file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU ‘gprof’ d1683 1 a1683 1 compatibility, GNU ‘gprof’ continues to support the old BSD-derived d1688 1 a1688 1 The new file format is defined in header file ‘gmon_out.h’. It d1692 1 a1692 1 the profile was collected. GNU ‘gprof’ adapts automatically to the d1699 1 a1699 1 reading a file, GNU ‘gprof’ will ensure records of the same type are d1715 1 a1715 1 would specify the long name as “seconds” and the abbreviation as “s”. d1718 1 a1718 1 For example, under DEC OSF/1, the “uprofile” command can be used to d1720 2 a1721 2 the dimension in the histogram header could be set to “i-cache misses” and the abbreviation could be set to “1” (because it is simply a count, d1737 2 a1738 2 must be within caller’s function and the second must be within the callee’s function. When performing profiling at the function level, d1757 1 a1757 1 9.3 ‘gprof’’s Internal Operation d1760 3 a1762 3 Like most programs, ‘gprof’ begins by processing its options. During this stage, it may building its symspec list (‘sym_ids.c:sym_id_add’), if options are specified which use symspecs. ‘gprof’ maintains a single d1764 1 a1764 1 tables, organized into six include/exclude pairs—one pair each for the d1772 2 a1773 2 After option processing, ‘gprof’ finishes building the symspec list by adding all the symspecs in ‘default_excluded_list’ to the exclude d1779 2 a1780 2 it is an object file, and read its symbol table (‘core.c:core_init’), using ‘bfd_canonicalize_symtab’ after mallocing an appropriately sized d1782 2 a1783 2 ‘--file-ordering’ option has been specified), and the core text space is read into memory (if the ‘-c’ option was given). d1785 1 a1785 1 ‘gprof’’s own symbol table, an array of Sym structures, is now built. d1787 1 a1787 1 whether line-by-line profiling (‘-l’ option) has been enabled. For d1791 1 a1791 1 either case, two passes are made through the symbol table—one to count d1793 2 a1794 2 the symbols. In between the two passes, a single array of type ‘Sym’ is created of the appropriate length. Finally, ‘symtab.c:symtab_finalize’ d1799 1 a1799 1 the ‘qsort’ library function (which sorts an array) will be used to sort d1801 1 a1801 1 (‘symtab.c:sym_lookup’), which finds symbols based on memory address, d1803 1 a1803 1 sorted array. Function symbols are indicated with an ‘is_func’ flag. d1805 1 a1805 1 can have an ‘is_static’ flag to indicate that it is a local symbol. d1808 2 a1809 2 Syms (‘sym_ids.c:sym_id_parse’). Remember that a single symspec can match multiple symbols. An array of symbol tables (‘syms’) is created, d1817 2 a1818 2 ‘gprof’ simply uses its standard symbol lookup routine on the appropriate table in the ‘syms’ array. d1821 2 a1822 2 (‘gmon_io.c:gmon_out_read’), first by checking for a new-style ‘gmon.out’ header, then assuming this is an old-style BSD ‘gmon.out’ if d1825 1 a1825 1 New-style histogram records are read by ‘hist.c:hist_read_rec’. For d1835 1 a1835 1 As each call graph record is read (‘call_graph.c:cg_read_rec’), the d1837 1 a1837 1 call graph arc is created by ‘cg_arcs.c:arc_add’, unless the arc fails a d1839 3 a1841 3 linked list is maintained of the parent’s child arcs, and of the child’s parent arcs. Both the child’s call count and the arc’s call count are incremented by the record’s call count. d1843 1 a1843 1 Basic-block records are read (‘basic_blocks.c:bb_read_rec’), but only d1846 1 a1846 1 entry made in the symbol’s bb_addr and bb_calls arrays. Again, if d1850 1 a1850 1 A gmon.sum file is dumped, if requested (‘gmon_io.c:gmon_out_write’). d1853 1 a1853 1 (‘hist.c:hist_assign_samples’) by iterating over all the sample bins and d1864 3 a1866 3 If call graph data is present, ‘cg_arcs.c:cg_assemble’ is called. First, if ‘-c’ was specified, a machine-dependent routine (‘find_call’) scans through each symbol’s machine code, looking for subroutine call d1869 1 a1869 1 (‘cg_dfn.c:cg_dfn’), so that children are always numbered less than d1878 1 a1878 1 a parent’s include or exclude (print or no print) property being d1888 2 a1889 2 fairly straightforward. The call graph (‘cg_print.c:cg_print’) and flat profile (‘hist.c:hist_print’) are regurgitations of values already d1891 1 a1891 1 (‘basic_blocks.c:print_annotated_source’) uses basic-block information, d1896 1 a1896 1 source code itself (‘cg_print.c’). Basically, the functions with the d1904 1 a1904 1 9.4 Debugging ‘gprof’ d1907 1 a1907 1 If ‘gprof’ was compiled with debugging enabled, the ‘-d’ option triggers d1927 1 a1927 1 file. For line-by-line profiling (‘-l’ option), also shows line d1930 1 a1930 1 Trace operation of ‘-c’ option d1937 1 a1937 1 meaningful with ‘-l’ option) d1941 1 a1941 1 Tracks operation of ‘-A’ option d1951 1 a1951 1 Copyright © 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. d1960 1 a1960 1 functional and useful document “free” in the sense of freedom: to d1967 1 a1967 1 This License is a kind of “copyleft”, which means that derivative d1988 2 a1989 2 “Document”, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as “you”. You accept d1993 1 a1993 1 A “Modified Version” of the Document means any work containing the d1997 1 a1997 1 A “Secondary Section” is a named appendix or a front-matter section d1999 1 a1999 1 publishers or authors of the Document to the Document’s overall d2008 1 a2008 1 The “Invariant Sections” are certain Secondary Sections whose d2016 1 a2016 1 The “Cover Texts” are certain short passages of text that are d2022 1 a2022 1 A “Transparent” copy of the Document means a machine-readable copy, d2034 1 a2034 1 “Transparent” is called “Opaque”. d2047 1 a2047 1 The “Title Page” means, for a printed book, the title page itself, d2050 3 a2052 3 works in formats which do not have any title page as such, “Title Page” means the text near the most prominent appearance of the work’s title, preceding the beginning of the body of the text. d2054 1 a2054 1 The “publisher” means any person or entity that distributes copies d2057 1 a2057 1 A section “Entitled XYZ” means a named subunit of the Document d2061 3 a2063 3 “Acknowledgements”, “Dedications”, “Endorsements”, or “History”.) To “Preserve the Title” of such a section when you modify the Document means that it remains a section “Entitled XYZ” according d2093 1 a2093 1 the Document’s license notice requires Cover Texts, you must d2165 1 a2165 1 Sections and required Cover Texts given in the Document’s d2170 1 a2170 1 I. Preserve the section Entitled “History”, Preserve its Title, d2173 1 a2173 1 Title Page. If there is no section Entitled “History” in the d2183 1 a2183 1 “History” section. You may omit a network location for a work d2188 1 a2188 1 K. For any section Entitled “Acknowledgements” or “Dedications”, d2197 1 a2197 1 M. Delete any section Entitled “Endorsements”. Such a section d2201 1 a2201 1 “Endorsements” or to conflict in title with any Invariant d2210 1 a2210 1 titles to the list of Invariant Sections in the Modified Version’s d2214 1 a2214 1 You may add a section Entitled “Endorsements”, provided it contains d2216 3 a2218 3 parties—for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2256 4 a2259 4 “History” in the various original documents, forming one section Entitled “History”; likewise combine any sections Entitled “Acknowledgements”, and any sections Entitled “Dedications”. You must delete all sections Entitled “Endorsements.” d2280 1 a2280 1 storage or distribution medium, is called an “aggregate” if the d2282 1 a2282 1 legal rights of the compilation’s users beyond what the individual d2289 1 a2289 1 of the entire aggregate, the Document’s Cover Texts may be placed d2311 2 a2312 2 If a section in the Document is Entitled “Acknowledgements”, “Dedications”, or “History”, the requirement (section 4) to d2353 1 a2353 1 version of this License “or any later version” applies to it, you d2361 1 a2361 1 proxy’s public statement of acceptance of a version permanently d2366 1 a2366 1 “Massive Multiauthor Collaboration Site” (or “MMC Site”) means any d2370 1 a2370 1 A “Massive Multiauthor Collaboration” (or “MMC”) contained in the d2374 1 a2374 1 “CC-BY-SA” means the Creative Commons Attribution-Share Alike 3.0 d2380 1 a2380 1 “Incorporate” means to publish or republish a Document, in whole or d2383 1 a2383 1 An MMC is “eligible for relicensing” if it is licensed under this d2410 1 a2410 1 Texts, replace the “with...Texts.” line with this: d2428 30 a2457 30 Node: Top726 Node: Introduction2083 Node: Compiling4604 Node: Executing8788 Node: Invoking11783 Node: Output Options13244 Node: Analysis Options20804 Node: Miscellaneous Options24904 Node: Deprecated Options26274 Node: Symspecs28457 Node: Output30372 Node: Flat Profile31426 Node: Call Graph36501 Node: Primary39771 Node: Callers42421 Node: Subroutines44641 Node: Cycles46597 Node: Line-by-line53625 Node: Annotated Source57792 Node: Inaccuracy60882 Node: Sampling Error61144 Node: Assumptions64104 Node: How do I?65657 Node: Incompatibilities67230 Node: Details68804 Node: Implementation69217 Node: File Format75279 Node: Internals79619 Node: Debugging88287 Node: GNU Free Documentation License89905 @ 1.8 log @Merge conflicts @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.5 from gprof.texi. d5 1 a5 1 Copyright (C) 1988-2020 Free Software Foundation, Inc. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.34. d220 5 a224 4 'gmon.out', its contents are overwritten. There is currently no way to tell the program to write the profile data under a different name, but you can rename the file afterwards if you are concerned that it may be overwritten. d1679 1 a1679 1 magic cookie that allows to check whether a data file really is a d2432 26 a2457 26 Node: Invoking11377 Node: Output Options12792 Node: Analysis Options19884 Node: Miscellaneous Options23804 Node: Deprecated Options25058 Node: Symspecs27121 Node: Output28947 Node: Flat Profile29987 Node: Call Graph34940 Node: Primary38172 Node: Callers40760 Node: Subroutines42878 Node: Cycles44719 Node: Line-by-line51496 Node: Annotated Source55572 Node: Inaccuracy58570 Node: Sampling Error58828 Node: Assumptions61732 Node: How do I?63202 Node: Incompatibilities64759 Node: Details66253 Node: Implementation66646 Node: File Format72545 Node: Internals76833 Node: Debugging85323 Node: GNU Free Documentation License86913 d2460 5 @ 1.7 log @merge conflicts @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.4 from gprof.texi. d5 1 a5 1 Copyright (C) 1988-2018 Free Software Foundation, Inc. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.90. d170 1 a170 1 '-static-libgcc' command line option. For example: d193 1 a193 1 command line option which will insert calls to special user supplied d1211 2 a1212 2 versions of 'gcc' the program had to be compiled with the '-a' command line option as well. d2428 29 a2456 29 Node: Introduction2045 Node: Compiling4536 Node: Executing8592 Node: Invoking11380 Node: Output Options12795 Node: Analysis Options19887 Node: Miscellaneous Options23807 Node: Deprecated Options25061 Node: Symspecs27124 Node: Output28950 Node: Flat Profile29990 Node: Call Graph34943 Node: Primary38175 Node: Callers40763 Node: Subroutines42881 Node: Cycles44722 Node: Line-by-line51499 Node: Annotated Source55575 Node: Inaccuracy58573 Node: Sampling Error58831 Node: Assumptions61735 Node: How do I?63205 Node: Incompatibilities64762 Node: Details66256 Node: Implementation66649 Node: File Format72548 Node: Internals76836 Node: Debugging85326 Node: GNU Free Documentation License86916 @ 1.6 log @- merge conflicts for 2.30 - bump libraries - regen for x86 @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.3 from gprof.texi. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.0. d2428 29 a2456 29 Node: Introduction2044 Node: Compiling4535 Node: Executing8591 Node: Invoking11379 Node: Output Options12794 Node: Analysis Options19886 Node: Miscellaneous Options23806 Node: Deprecated Options25060 Node: Symspecs27123 Node: Output28949 Node: Flat Profile29989 Node: Call Graph34942 Node: Primary38174 Node: Callers40762 Node: Subroutines42880 Node: Cycles44721 Node: Line-by-line51498 Node: Annotated Source55574 Node: Inaccuracy58572 Node: Sampling Error58830 Node: Assumptions61734 Node: How do I?63204 Node: Incompatibilities64761 Node: Details66255 Node: Implementation66648 Node: File Format72547 Node: Internals76835 Node: Debugging85325 Node: GNU Free Documentation License86915 @ 1.6.2.1 log @Sync with HEAD @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.4 from gprof.texi. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.90. d2428 29 a2456 29 Node: Introduction2045 Node: Compiling4536 Node: Executing8592 Node: Invoking11380 Node: Output Options12795 Node: Analysis Options19887 Node: Miscellaneous Options23807 Node: Deprecated Options25061 Node: Symspecs27124 Node: Output28950 Node: Flat Profile29990 Node: Call Graph34943 Node: Primary38175 Node: Callers40763 Node: Subroutines42881 Node: Cycles44722 Node: Line-by-line51499 Node: Annotated Source55575 Node: Inaccuracy58573 Node: Sampling Error58831 Node: Assumptions61735 Node: How do I?63205 Node: Incompatibilities64762 Node: Details66256 Node: Implementation66649 Node: File Format72548 Node: Internals76836 Node: Debugging85326 Node: GNU Free Documentation License86916 @ 1.6.2.2 log @Merge changes from current as of 20200406 @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.5 from gprof.texi. d5 1 a5 1 Copyright (C) 1988-2020 Free Software Foundation, Inc. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.34. d170 1 a170 1 '-static-libgcc' command-line option. For example: d193 1 a193 1 command-line option which will insert calls to special user supplied d1211 2 a1212 2 versions of 'gcc' the program had to be compiled with the '-a' command-line option as well. d2428 29 a2456 29 Node: Introduction2042 Node: Compiling4533 Node: Executing8589 Node: Invoking11377 Node: Output Options12792 Node: Analysis Options19884 Node: Miscellaneous Options23804 Node: Deprecated Options25058 Node: Symspecs27121 Node: Output28947 Node: Flat Profile29987 Node: Call Graph34940 Node: Primary38172 Node: Callers40760 Node: Subroutines42878 Node: Cycles44719 Node: Line-by-line51496 Node: Annotated Source55572 Node: Inaccuracy58570 Node: Sampling Error58828 Node: Assumptions61732 Node: How do I?63202 Node: Incompatibilities64759 Node: Details66253 Node: Implementation66646 Node: File Format72545 Node: Internals76833 Node: Debugging85323 Node: GNU Free Documentation License86913 @ 1.5 log @merge conflicts, disable testsuite @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 4.8 from gprof.texi. d3 1 a3 6 INFO-DIR-SECTION Software development START-INFO-DIR-ENTRY * gprof: (gprof). Profiling your program's execution END-INFO-DIR-ENTRY This file documents the gprof profiler of the GNU system. d5 1 a5 1 Copyright (C) 1988-2016 Free Software Foundation, Inc. d14 5 d25 1 a25 1 This manual describes the GNU profiler, `gprof', and how you can use it d28 1 a28 1 programs. GNU `gprof' was written by Jay Fenlason. d30 1 a30 1 This manual is for `gprof' (GNU Binutils) version 2.27. d33 2 a34 2 Documentation License version 1.3. A copy of the license is included in the section entitled "GNU Free Documentation License". d42 1 a42 1 * Invoking:: How to run `gprof', and its options d44 1 a44 1 * Output:: Interpreting `gprof''s output d48 1 a48 1 * Incompatibilities:: (between GNU `gprof' and Unix `gprof'.) d67 5 a71 6 execution of your program, it can be used on programs that are too large or too complex to analyze by reading the source. However, how your program is run will affect the information that shows up in the profile data. If you don't use some feature of your program while it is being profiled, no profile information will be generated for that feature. d81 1 a81 1 * You must run `gprof' to analyze the profile data. *Note `gprof' d93 5 a97 6 The "call graph" shows, for each function, which functions called it, which other functions it called, and how many times. There is also an estimate of how much time was spent in the subroutines of each function. This can suggest places where you might try to eliminate function calls that use a lot of time. *Note The Call Graph: Call Graph. d116 1 a116 1 To compile a source file for profiling, specify the `-pg' option when d120 3 a122 3 To link the program for profiling, if you use a compiler such as `cc' to do the linking, simply specify `-pg' in addition to your usual options. The same option, `-pg', alters either compilation or linking d128 1 a128 1 The `-pg' option also works with a command that both compiles and d133 3 a135 3 Note: The `-pg' option must be part of your compilation options as well as your link options. If it is not then no call-graph data will be gathered and when you run `gprof' you will get an error message like d140 2 a141 2 If you add the `-Q' switch to suppress the printing of the call graph data you will still be able to see the time samples: d152 8 a159 8 If you run the linker `ld' directly instead of through a compiler such as `cc', you may have to specify a profiling startup file `gcrt0.o' as the first input file instead of the usual startup file `crt0.o'. In addition, you would probably want to specify the profiling C library, `libc_p.a', by writing `-lc_p' instead of the usual `-lc'. This is not absolutely necessary, but doing this gives you number-of-calls information for standard library functions such as `read' and `open'. For example: d164 2 a165 2 libraries you may run into problems with the profiling support code in a shared library being called before that library has been fully d168 3 a170 3 against a static version of the library containing the profiling support code, which for `gcc' users can be done via the `-static' or `-static-libgcc' command line option. For example: d174 1 a174 1 If you compile only some of the modules of the program with `-pg', d176 6 a181 6 information about the modules that were compiled without `-pg'. The only information you get for the functions in those modules is the total time spent in them; there is no record of how many times they were called, or from where. This will not affect the flat profile (except that the `calls' field for the functions will be blank), but will greatly reduce the usefulness of the call graph. d184 1 a184 1 `gcov' tool instead of `gprof'. See that tool's manual or info pages d187 3 a189 3 Note, older versions of `gcc' produce line-by-line profiling information that works with `gprof' rather than `gcov' so there is still support for displaying this kind of information in `gprof'. *Note d192 5 a196 5 It also worth noting that `gcc' implements a `-finstrument-functions' command line option which will insert calls to special user supplied instrumentation routines at the entry and exit of every function in their program. This can be used to implement an alternative profiling scheme. d205 1 a205 1 generate the information that `gprof' needs. Simply run the program as d219 2 a220 2 `gmon.out' just before exiting. If there is already a file called `gmon.out', its contents are overwritten. There is currently no way to d225 3 a227 3 In order to write the `gmon.out' file properly, your program must exit normally: by returning from `main' or by calling `exit'. Calling the low-level function `_exit' does not write the profile data, and d230 1 a230 1 The `gmon.out' file is written in the program's _current working d232 2 a233 2 `chdir', the `gmon.out' file will be left in the last directory your program `chdir''d to. If you don't have permission to write in this d237 1 a237 1 called `bb.out'. This file, if present, contains an human-readable d239 4 a242 4 appearance of a human-readable `bb.out' means the basic-block counts didn't get written into `gmon.out'. The Perl script `bbconv.pl', included with the `gprof' source distribution, will convert a `bb.out' file into a format readable by `gprof'. Invoke it like this: d246 2 a247 2 This translates the information in `bb.out' into a form that `gprof' can understand. But you still need to tell `gprof' about the existence d249 1 a249 1 `gprof' command line, _along with `gmon.out'_, like this: d256 1 a256 1 4 `gprof' Command Summary d259 2 a260 2 After you have a profile data file `gmon.out', you can run `gprof' to interpret the information in it. The `gprof' program prints a flat d262 1 a262 1 redirect the output of `gprof' into a file with `>'. d264 1 a264 1 You run `gprof' like this: d270 4 a273 5 If you omit the executable file name, the file `a.out' is used. If you give no profile data file name, the file `gmon.out' is used. If any file is not in the proper format, or if the profile data file does not appear to belong to the executable file, an error message is printed. d283 2 a284 2 * Output Options:: Controlling `gprof''s output style * Analysis Options:: Controlling how `gprof' analyzes its data d296 1 a296 1 These options specify which of several output formats `gprof' should d299 4 a302 4 Many of these options take an optional "symspec" to specify functions to be included or excluded. These options can be specified multiple times, with different symspecs, to include or exclude sets of symbols. *Note Symspecs: Symspecs. d304 1 a304 1 Specifying any of these options overrides the default (`-p -q'), d307 3 a309 3 `-A[SYMSPEC]' `--annotated-source[=SYMSPEC]' The `-A' option causes `gprof' to print annotated source code. If d313 3 a315 3 `-b' `--brief' If the `-b' option is given, `gprof' doesn't print the verbose d320 3 a322 3 `-C[SYMSPEC]' `--exec-counts[=SYMSPEC]' The `-C' option causes `gprof' to print a tally of functions and d327 1 a327 1 specifying the `-l' option, along with `-C', will cause basic-block d330 9 a338 9 `-i' `--file-info' The `-i' option causes `gprof' to display summary information about the profile data file(s) and then exit. The number of histogram, call graph, and basic-block count records is displayed. `-I DIRS' `--directory-path=DIRS' The `-I' option specifies a list of search directories in which to d343 4 a346 4 `-J[SYMSPEC]' `--no-annotated-source[=SYMSPEC]' The `-J' option causes `gprof' not to print annotated source code. If SYMSPEC is specified, `gprof' prints annotated source, but d349 2 a350 2 `-L' `--print-path' d352 1 a352 1 suppressed. The `-L' option causes `gprof' to print the full d357 10 a366 10 `-p[SYMSPEC]' `--flat-profile[=SYMSPEC]' The `-p' option causes `gprof' to print a flat profile. If SYMSPEC is specified, print flat profile only for matching symbols. *Note The Flat Profile: Flat Profile. `-P[SYMSPEC]' `--no-flat-profile[=SYMSPEC]' The `-P' option causes `gprof' to suppress printing a flat profile. If SYMSPEC is specified, `gprof' prints a flat profile, but d369 3 a371 3 `-q[SYMSPEC]' `--graph[=SYMSPEC]' The `-q' option causes `gprof' to print the call graph analysis. d375 4 a378 4 `-Q[SYMSPEC]' `--no-graph[=SYMSPEC]' The `-Q' option causes `gprof' to suppress printing the call graph. If SYMSPEC is specified, `gprof' prints a call graph, but excludes d381 3 a383 3 `-t' `--table-length=NUM' The `-t' option causes the NUM most active source lines in each d387 2 a388 2 `-y' `--separate-files' d390 1 a390 1 `gprof' prints annotated source files to standard-output. If this d392 9 a400 9 `path/FILENAME' is generated in the file `FILENAME-ann'. If the underlying file system would truncate `FILENAME-ann' so that it overwrites the original `FILENAME', `gprof' generates annotated source in the file `FILENAME.ann' instead (if the original file name has an extension, that extension is _replaced_ with `.ann'). `-Z[SYMSPEC]' `--no-exec-counts[=SYMSPEC]' The `-Z' option causes `gprof' not to print a tally of functions d404 3 a406 3 `-r' `--function-ordering' The `--function-ordering' option causes `gprof' to print a d416 7 a422 7 `-R MAP_FILE' `--file-ordering MAP_FILE' The `--file-ordering' option causes `gprof' to print a suggested .o link line ordering for the program based on profiling data. This option suggests an ordering which may improve paging, tlb and cache behavior for the program on systems which do not support arbitrary ordering of functions in an executable. d424 1 a424 1 Use of the `-a' argument is highly recommended with this option. d428 1 a428 1 similar to the output of the program `nm'. d439 2 a440 1 To create a MAP_FILE with GNU `nm', type a command like `nm d443 4 a446 4 `-T' `--traditional' The `-T' option causes `gprof' to print its output in "traditional" BSD style. d448 2 a449 2 `-w WIDTH' `--width=WIDTH' d453 7 a459 7 `-x' `--all-lines' This option affects annotated source output only. By default, only the lines at the beginning of a basic-block are annotated. If this option is specified, every line in a basic-block is annotated by repeating the annotation for the first line. This behavior is similar to `tcov''s `-a'. d461 2 a462 2 `--demangle[=STYLE]' `--no-demangle' d465 1 a465 1 `--no-demangle' option may be used to turn off demangling. d476 3 a478 3 `-a' `--no-static' The `-a' option causes `gprof' to suppress the printing of d484 2 a485 2 the executable file. This option affects both the flat profile and the call graph. d487 3 a489 3 `-c' `--static-call-graph' The `-c' option causes the call graph of the program to be d494 12 a505 12 have been called, but never were. Calls to functions that were not compiled with profiling enabled are also identified, but only if symbol table entries are present for them. Calls to dynamic library routines are typically _not_ found by this option. Parents or children identified via this heuristic are indicated in the call graph with call counts of `0'. `-D' `--ignore-non-functions' The `-D' option causes `gprof' to ignore symbols which are not known to be functions. This option will give more accurate profile data on systems where it is supported (Solaris and HPUX for d508 2 a509 2 `-k FROM/TO' The `-k' option allows you to delete from the call graph any arcs d512 3 a514 3 `-l' `--line' The `-l' option enables line-by-line profiling, which causes d517 2 a518 2 compiled by older versions of the `gcc' compiler. Newer versions of `gcc' are designed to work with the `gcov' tool instead. d520 5 a524 5 If the program was compiled with basic-block counting enabled, this option will also identify how many times each line of code was executed. While line-by-line profiling can help isolate where in a large function a program is spending its time, it also significantly increases the running time of `gprof', and magnifies d528 4 a531 4 `--inline-file-names' This option causes `gprof' to print the source file after each symbol in both the flat profile and the call graph. The full path to the file is printed if used with the `-L' option. d533 2 a534 2 `-m NUM' `--min-count=NUM' d538 3 a540 8 `-nSYMSPEC' `--time=SYMSPEC' The `-n' option causes `gprof', in its call graph analysis, to only propagate times for symbols matching SYMSPEC. `-NSYMSPEC' `--no-time=SYMSPEC' The `-n' option causes `gprof', in its call graph analysis, not to d543 15 a557 10 `-SFILENAME' `--external-symbol-table=FILENAME' The `-S' option causes `gprof' to read an external symbol table file, such as `/proc/kallsyms', rather than read the symbol table from the given object file (the default is `a.out'). This is useful for profiling kernel modules. `-z' `--display-unused-functions' If you give the `-z' option, `gprof' will mention all functions in d559 2 a560 3 no time spent in them. This is useful in conjunction with the `-c' option for discovering which routines were never called. d568 4 a571 4 `-d[NUM]' `--debug[=NUM]' The `-d NUM' option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging `gprof': d574 3 a576 3 `-h' `--help' The `-h' option prints command line usage. d578 2 a579 2 `-ONAME' `--file-format=NAME' d581 2 a582 2 are `auto' (the default), `bsd', `4.4bsd', `magic', and `prof' (not yet supported). d584 3 a586 3 `-s' `--sum' The `-s' option causes `gprof' to summarize the information in the d588 11 a598 11 called `gmon.sum', which contains all the information from the profile data files that `gprof' read in. The file `gmon.sum' may be one of the specified input files; the effect of this is to merge the data in the other input files into `gmon.sum'. Eventually you can run `gprof' again without `-s' to analyze the cumulative data in the file `gmon.sum'. `-v' `--version' The `-v' flag causes `gprof' to print the current version number, a600 1 d609 2 a610 2 `-e FUNCTION_NAME' The `-e FUNCTION' option tells `gprof' to not print information d613 10 a622 11 functions that call it, but its index number will be shown as `[not printed]'. More than one `-e' option may be given; only one FUNCTION_NAME may be indicated with each `-e' option. `-E FUNCTION_NAME' The `-E FUNCTION' option works like the `-e' option, but time spent in the function (and children who were not called from anywhere else), will not be used to compute the percentages-of-time for the call graph. More than one `-E' option may be given; only one FUNCTION_NAME may be indicated with each `-E' option. d624 2 a625 2 `-f FUNCTION_NAME' The `-f FUNCTION' option causes `gprof' to limit the call graph to d627 2 a628 10 children...). More than one `-f' option may be given; only one FUNCTION_NAME may be indicated with each `-f' option. `-F FUNCTION_NAME' The `-F FUNCTION' option works like the `-f' option, but only time spent in the function and its children (and their children...) will be used to determine total-time and percentages-of-time for the call graph. More than one `-F' option may be given; only one FUNCTION_NAME may be indicated with each `-F' option. The `-F' option overrides the `-E' option. d630 7 d638 2 a639 2 Note that only one function can be specified with each `-e', `-E', `-f' or `-F' option. To specify more than one function, use multiple d645 1 a645 1 `foo' or `bar' and were not reachable from `boring'. d664 6 a669 6 `main.c' Selects everything in file `main.c'--the dot in the string tells `gprof' to interpret the string as a filename, rather than as a function name. To select a file whose name does not contain a dot, a trailing colon should be specified. For example, `odd:' is interpreted as the file named `odd'. d671 2 a672 2 `main' Selects all functions named `main'. d681 2 a682 2 necessary to add a leading colon to the name. For example, `:.mul' selects function `.mul'. d685 2 a686 2 `gprof' will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as `gprof' prints d688 2 a689 2 `_main' from your `main' function, `gprof' still prints it as `main' in its output, so you should use `main' in symspecs. d691 2 a692 2 `main.c:main' Selects function `main' in file `main.c'. d694 2 a695 2 `main.c:134' Selects line 134 in file `main.c'. d700 1 a700 1 5 Interpreting `gprof''s Output d703 1 a703 1 `gprof' can produce several different output styles, the most important d716 1 a716 1 * Line-by-line:: `gprof' can analyze individual source code lines d727 1 a727 1 executing each function. Unless the `-z' option is given, functions d758 1 a758 1 functions `mcount' and `profil' are part of the profiling apparatus and d768 8 a775 8 `cumulative seconds' field. Since each sample counted for 0.01 seconds, this means only six samples were taken during the run. Two of the samples occurred while the program was in the `open' function, as indicated by the `self seconds' field. Each of the other four samples occurred one each in `offtime', `memccpy', `write', and `mcount'. Since only six samples were taken, none of these values can be regarded as particularly reliable. In another run, the `self seconds' field for `mcount' might well be `0.00' or `0.02'. *Note Statistical Sampling d778 2 a779 2 The remaining functions in the listing (those whose `self seconds' field is `0.00') didn't appear in the histogram samples at all. d781 1 a781 1 they are listed, sorted in decreasing order by the `calls' field. d788 1 a788 1 `% time' d792 1 a792 1 `cumulative seconds' d797 1 a797 1 `self seconds' d801 1 a801 1 `calls' d807 1 a807 1 `self ms/call' d812 1 a812 1 `total ms/call' d819 2 a820 2 `name' This is the name of the function. The flat profile is sorted by d832 2 a833 2 they themselves may not have used much time, called other functions that did use unusual amounts of time. d836 1 a836 1 same `gprof' run as the flat profile example in the previous section. d871 4 a874 4 number in square brackets. The end of this line says which function the entry is for. The preceding lines in the entry describe the callers of this function and the following lines describe its subroutines (also called "children" when we speak of the call graph). d879 1 a879 1 The internal profiling function `mcount' (*note The Flat Profile: d888 1 a888 1 such as `a' calls `b' calls `a'... d901 1 a901 1 `report' in our main example, together with the heading line that shows d910 1 a910 1 `index' d920 1 a920 1 `% time' d929 1 a929 1 `self' d931 1 a931 1 should be identical to the number printed in the `seconds' field d934 5 a938 5 `children' This is the total amount of time spent in the subroutine calls made by this function. This should be equal to the sum of all the `self' and `children' entries of the children listed directly below this function. d940 1 a940 1 `called' d944 1 a944 1 separated by a `+'. The first number counts non-recursive calls, d947 2 a948 2 In the example above, the function `report' was called once from `main'. d950 1 a950 1 `name' d957 1 a957 1 example, if function `gnurr' is part of cycle number one, and has d968 3 a970 3 A function's entry has a line for each function it was called by. These lines' fields correspond to the fields of the primary line, but their meanings are different because of the difference in context. d973 1 a973 1 `report', the primary line and one caller-line preceding it, together d981 2 a982 2 Here are the meanings of the fields in the caller-line for `report' called from `main': d984 14 a997 14 `self' An estimate of the amount of time spent in `report' itself when it was called from `main'. `children' An estimate of the amount of time spent in subroutines of `report' when `report' was called from `main'. The sum of the `self' and `children' fields is an estimate of the amount of time spent within calls to `report' from `main'. `called' Two numbers: the number of times `report' was called from `main', followed by the total number of non-recursive calls to `report' d1000 2 a1001 2 `name and index number' The name of the caller of `report' to which this line applies, d1005 1 a1005 1 `gprof' request the omission of certain functions. When a caller d1013 1 a1013 1 dummy caller-line is printed which has `' as the "caller's d1028 1 a1028 1 `main', the primary line and a line for a subroutine, together with the d1036 2 a1037 2 Here are the meanings of the fields in the subroutine-line for `main' calling `report': d1039 17 a1055 17 `self' An estimate of the amount of time spent directly within `report' when `report' was called from `main'. `children' An estimate of the amount of time spent in subroutines of `report' when `report' was called from `main'. The sum of the `self' and `children' fields is an estimate of the total time spent in calls to `report' from `main'. `called' Two numbers, the number of calls to `report' from `main' followed by the total number of non-recursive calls to `report'. This ratio is used to determine how much of `report''s `self' and `children' time gets credited to `main'. *Note Estimating `children' Times: Assumptions. d1057 2 a1058 2 `name' The name of the subroutine of `main' to which this line applies, d1070 2 a1071 2 The graph may be complicated by the presence of "cycles of recursion" in the call graph. A cycle exists if a function calls another function d1073 2 a1074 2 function. For example: if `a' calls `b', and `b' calls `a', then `a' and `b' form a cycle. d1077 4 a1080 4 they belong to the same cycle. If `a' and `b' call each other and `b' and `c' call each other, all three make one cycle. Note that even if `b' only calls `a' if it was not called from `a', `gprof' cannot determine this, so `a' and `b' are still considered a cycle. d1084 1 a1084 1 graph it is followed by `'. d1087 4 a1090 4 call graph paradoxical. The "time spent in children" of `a' should include the time spent in its subroutine `b' and in `b''s subroutines--but one of `b''s subroutines is `a'! How much of `a''s time should be included in the children of `a', when `a' is indirectly d1093 1 a1093 1 The way `gprof' resolves this paradox is by creating a single entry d1098 2 a1099 2 "callers" of the cycle are the functions, outside the cycle, that called functions in the cycle. d1102 2 a1103 2 containing functions `a' and `b'. The cycle was entered by a call to `a' from `main'; both `a' and `b' called `c'. d1125 2 a1126 2 for `main', which calls `a', and an entry for `c', with callers `a' and `b'.) d1159 2 a1160 2 The `self' field of the cycle's primary line is the total time spent in all the functions of the cycle. It equals the sum of the `self' d1164 6 a1169 6 The `children' fields of the cycle's primary line and subroutine lines count only subroutines outside the cycle. Even though `a' calls `b', the time spent in those calls to `b' is not counted in `a''s `children' time. Thus, we do not encounter the problem of what to do when the time in those calls to `b' includes indirect recursive calls back to `a'. d1171 1 a1171 1 The `children' field of a caller-line in the cycle's entry estimates d1176 1 a1176 1 The `called' field in the primary line for the cycle has two numbers: d1179 8 a1186 8 called by functions in the cycle (including times when a function in the cycle calls itself). This is a generalization of the usual split into non-recursive and recursive calls. The `called' field of a subroutine-line for a cycle member in the cycle's entry says how many time that function was called from functions in the cycle. The total of all these is the second number in the primary line's `called' field. d1190 4 a1193 4 These lines show how many times each function in the cycle called or was called from each other function in the cycle. The `self' and `children' fields in these lines are blank because of the difficulty of defining meanings for them when recursion is going on. d1201 1 a1201 1 `gprof''s `-l' option causes the program to perform "line-by-line" d1203 3 a1205 3 functions, but to individual lines of source code. This only works with programs compiled with older versions of the `gcc' compiler. Newer versions of `gcc' use a different program - `gcov' - to display d1208 2 a1209 2 With the older versions of `gcc' the program usually has to be compiled with a `-g' option, in addition to `-pg', in order to generate d1211 1 a1211 1 versions of `gcc' the program had to be compiled with the `-a' command d1216 1 a1216 1 version of `gprof' does not propagate call graph arcs from source code d1220 3 a1222 3 Here is a section of `gprof''s output, without line-by-line profiling. Note that `ct_init' accounted for four histogram hits, and 13327 calls to `init_block'. d1245 3 a1247 2 Now let's look at some of `gprof''s output from the same program run, this time with line-by-line profiling enabled. Note that `ct_init''s d1250 1 a1250 1 note how `ct_init''s 13327 calls to `init_block' are broken down into d1286 1 d1293 1 a1293 1 `gprof''s `-A' option triggers an annotated source listing, which lists d1295 2 a1296 2 times it was called. You may also need to specify the `-I' option, if `gprof' can't find the source code files. d1298 1 a1298 1 With older versions of `gcc' compiling with `gcc ... -g -pg -a' d1300 3 a1302 3 function counting code. This enables `gprof' to determine how many times each line of code was executed. With newer versions of `gcc' support for displaying basic-block counts is provided by the `gcov' d1328 7 a1334 6 `updcrc' has at least five basic-blocks. One is the function itself. The `if' statement on line 9 generates two more basic-blocks, one for each branch of the `if'. A fourth basic-block results from the `if' on line 13, and the contents of the `do' loop form the fifth basic-block. The compiler may also generate additional basic-blocks to handle various special cases. d1337 3 a1339 3 `gprof -l -A'. The `-x' option is also helpful, to ensure that each line of code is labeled at least once. Here is `updcrc''s annotated source listing for a sample `gzip' run: d1362 2 a1363 2 each branch of the `if' statement. The body of the `do' loop was executed a total of 26312 times. Note how the `while' statement is d1371 1 a1371 1 6 Inaccuracy of `gprof' Output d1385 1 a1385 1 The run-time figures that `gprof' gives you are based on a sampling d1398 2 a1399 2 beware that the mcount counting function in glibc is _not_ thread-safe). *Note Implementation of Profiling: Implementation. d1408 1 a1408 1 period is 0.01 seconds and `foo''s run-time is 1 second, N is 100 d1410 1 a1410 1 error in `foo''s run-time is 0.1 seconds (10*0.01 seconds), or ten d1412 5 a1416 5 seconds and `bar''s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in `bar''s run-time is 1 second, or one percent of the observed value. It is likely to vary this much _on the average_ from one profiling run to the next. (_Sometimes_ it will vary more.) d1426 1 a1426 1 the data from several runs, using the `-s' option of `gprof'. Here is d1431 1 a1431 1 2. Issue the command `mv gmon.out gmon.sum'. d1435 1 a1435 1 4. Merge the new data in `gmon.out' into `gmon.sum' with this command: d1448 1 a1448 1 6.2 Estimating `children' Times d1452 2 a1453 2 `children' time values and all the time figures in caller and subroutine lines. d1456 1 a1456 1 profile data itself. Instead, `gprof' estimates them by making an d1460 3 a1462 4 any function `foo' is not correlated with who called `foo'. If `foo' used 5 seconds in all, and 2/5 of the calls to `foo' came from `a', then `foo' contributes 2 seconds to `a''s `children' time, by assumption. d1465 6 a1470 6 far from true. Suppose that `foo' returns very quickly when its argument is zero; suppose that `a' always passes zero as an argument, while other callers of `foo' pass other arguments. In this program, all the time spent in `foo' is in the calls from callers other than `a'. But `gprof' has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in `foo' to the children of `a'. d1472 1 a1472 1 We hope some day to put more complete data into `gmon.out', so that d1483 1 d1485 1 a1485 1 Because `gprof' can only report call times and counts by function, d1489 1 a1489 1 artificial hot spots since compiling with `-pg' adds a significant d1494 2 a1495 1 Use the `gcov' program. d1498 2 a1499 1 Use `gprof -l' and lookup the function in the call graph. The d1503 1 a1518 1 d1522 1 a1522 1 8 Incompatibilities with Unix `gprof' d1525 3 a1527 3 GNU `gprof' and Berkeley Unix `gprof' use the same data file `gmon.out', and provide essentially the same information. But there are a few differences. d1529 1 a1529 1 * GNU `gprof' uses a new, generalized file format with support for d1531 1 a1531 1 cookie and version number allows `gprof' to easily identify new d1535 3 a1537 3 * For a recursive function, Unix `gprof' lists the function as a parent and as a child, with a `calls' field that lists the number of recursive calls. GNU `gprof' omits these lines and puts the d1540 2 a1541 2 * When a function is suppressed from the call graph with `-e', GNU `gprof' still lists it as a subroutine of functions that call it. d1543 2 a1544 2 * GNU `gprof' accepts the `-k' with its argument in the form `from/to', instead of `from to'. d1546 3 a1548 3 * In the annotated source listing, if there are multiple basic blocks on the same line, GNU `gprof' prints all of their counts, separated by commas. d1551 1 a1551 1 `gprof' prints blurbs after the tables, so that you can see the d1563 3 a1565 3 * File Format:: Format of `gmon.out' files * Internals:: `gprof''s internal operation * Debugging:: Using `gprof''s `-d' option d1578 2 a1579 2 the `-pg' option, which causes every function to call `mcount' (or `_mcount', or `__mcount', depending on the OS and compiler) as one of d1582 1 a1582 1 The `mcount' routine, included in the profiling library, is d1585 3 a1587 3 done by examining the stack frame to find both the address of the child, and the return address in the original parent. Since this is a very machine-dependent operation, `mcount' itself is typically a short d1589 4 a1592 4 and then calls `__mcount_internal' (a normal C function) with two arguments--`frompc' and `selfpc'. `__mcount_internal' is responsible for maintaining the in-memory call graph, which records `frompc', `selfpc', and the number of times each of these call arcs was traversed. d1595 1 a1595 1 (`__builtin_return_address'), which allows a generic `mcount' function d1599 1 a1599 1 `mcount' is used for performance reasons. d1603 2 a1604 2 same as in the usual C library, but they were compiled with `-pg'. If you link your program with `gcc ... -pg', it automatically uses the d1607 5 a1611 5 Profiling also involves watching your program as it runs, and keeping a histogram of where the program counter happens to be every now and then. Typically the program counter is looked at around 100 times per second of run time, but the exact frequency may vary from system to system. d1614 1 a1614 1 provide a `profil()' system call, which registers a memory array with d1616 7 a1622 7 address space maps into the array. Typical scaling values cause every 2 to 8 bytes of address space to map into a single array slot. On every tick of the system clock (assuming the profiled program is running), the value of the program counter is examined and the corresponding slot in the memory array is incremented. Since this is done in the kernel, which had to interrupt the process anyway to handle the clock interrupt, very little additional system overhead is required. d1625 8 a1632 9 earlier), do not provide a `profil()' system call. On such a system, arrangements are made for the kernel to periodically deliver a signal to the process (typically via `setitimer()'), which then performs the same operation of examining the program counter and incrementing a slot in the memory array. Since this method requires a signal to be delivered to user space every time a sample is taken, it uses considerably more overhead than kernel-based profiling. Also, due to the added delay required to deliver the signal, this method is less accurate as well. d1635 8 a1642 8 either calls `profil()' or sets up a clock signal handler. This routine (`monstartup') can be invoked in several ways. On Linux systems, a special profiling startup file `gcrt0.o', which invokes `monstartup' before `main', is used instead of the default `crt0.o'. Use of this special startup file is one of the effects of using `gcc ... -pg' to link. On SPARC systems, no special startup files are used. Rather, the `mcount' routine, when it is invoked for the first time (typically when `main' is called), calls `monstartup'. d1644 1 a1644 1 If the compiler's `-a' option was used, basic-block counting is also d1647 1 a1647 1 basic-block begins (i.e., when an `if' statement appears), an extra d1654 5 a1658 5 The profiling library also includes a function (`mcleanup') which is typically registered using `atexit()' to be called as the program exits, and is responsible for writing the file `gmon.out'. Profiling is turned off, various headers are output, and the histogram is written, followed by the call-graph arcs and the basic-block counts. d1660 1 a1660 1 The output from `gprof' gives no indication of parts of your program d1662 8 a1669 8 of the program counter are taken at fixed intervals of the program's run time. Therefore, the time measurements in `gprof' output say nothing about time that your program was not running. For example, a part of the program that creates so much data that it cannot all fit in physical memory at once may run very slowly due to thrashing, but `gprof' will say it uses little time. On the other hand, sampling by run time has the advantage that the amount of load due to other users won't directly affect the output you get. d1679 2 a1680 2 `gprof' file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU `gprof' d1682 1 a1682 1 compatibility, GNU `gprof' continues to support the old BSD-derived d1687 1 a1687 1 The new file format is defined in header file `gmon_out.h'. It d1691 1 a1691 1 the profile was collected. GNU `gprof' adapts automatically to the d1698 1 a1698 1 reading a file, GNU `gprof' will ensure records of the same type are d1708 15 a1722 15 spans, the size of the histogram in bytes (unlike in the old BSD format, this does not include the size of the header), the rate of the profiling clock, and the physical dimension that the bin counts represent after being scaled by the profiling clock rate. The physical dimension is specified in two parts: a long name of up to 15 characters and a single character abbreviation. For example, a histogram representing real-time would specify the long name as "seconds" and the abbreviation as "s". This feature is useful for architectures that support performance monitor hardware (which, fortunately, is becoming increasingly common). For example, under DEC OSF/1, the "uprofile" command can be used to produce a histogram of, say, instruction cache misses. In this case, the dimension in the histogram header could be set to "i-cache misses" and the abbreviation could be set to "1" (because it is simply a count, not a physical dimension). Also, the profiling rate would have to be set to 1 in this case. d1735 8 a1742 8 program execution. Arcs are specified by a pair of addresses: the first must be within caller's function and the second must be within the callee's function. When performing profiling at the function level, these addresses can point anywhere within the respective function. However, when profiling at the line-level, it is better if the addresses are as close to the call-site/entry-point as possible. This will ensure that the line-level call-graph is able to identify exactly which line of source code performed calls to a function. d1748 4 a1751 5 sequence of address/count pairs. The header simply specifies the length of the sequence. In an address/count pair, the address identifies a basic-block and the count specifies the number of times that basic-block was executed. Any address within the basic-address can be used. d1756 1 a1756 1 9.3 `gprof''s Internal Operation d1759 6 a1764 6 Like most programs, `gprof' begins by processing its options. During this stage, it may building its symspec list (`sym_ids.c:sym_id_add'), if options are specified which use symspecs. `gprof' maintains a single linked list of symspecs, which will eventually get turned into 12 symbol tables, organized into six include/exclude pairs--one pair each for the flat profile (INCL_FLAT/EXCL_FLAT), the call graph arcs d1771 2 a1772 2 After option processing, `gprof' finishes building the symspec list by adding all the symspecs in `default_excluded_list' to the exclude d1778 2 a1779 2 it is an object file, and read its symbol table (`core.c:core_init'), using `bfd_canonicalize_symtab' after mallocing an appropriately sized d1781 2 a1782 2 `--file-ordering' option has been specified), and the core text space is read into memory (if the `-c' option was given). d1784 1 a1784 1 `gprof''s own symbol table, an array of Sym structures, is now built. d1786 1 a1786 1 whether line-by-line profiling (`-l' option) has been enabled. For d1792 4 a1795 4 the symbols. In between the two passes, a single array of type `Sym' is created of the appropriate length. Finally, `symtab.c:symtab_finalize' is called to sort the symbol table and remove duplicate entries (entries with the same memory address). d1798 3 a1800 3 the `qsort' library function (which sorts an array) will be used to sort the symbol table. Also, the symbol lookup routine (`symtab.c:sym_lookup'), which finds symbols based on memory address, d1802 1 a1802 1 sorted array. Function symbols are indicated with an `is_func' flag. d1804 1 a1804 1 can have an `is_static' flag to indicate that it is a local symbol. d1807 11 a1817 11 Syms (`sym_ids.c:sym_id_parse'). Remember that a single symspec can match multiple symbols. An array of symbol tables (`syms') is created, each entry of which is a symbol table of Syms to be included or excluded from a particular listing. The master symbol table and the symspecs are examined by nested loops, and every symbol that matches a symspec is inserted into the appropriate syms table. This is done twice, once to count the size of each required symbol table, and again to build the tables, which have been malloced between passes. From now on, to determine whether a symbol is on an include or exclude symspec list, `gprof' simply uses its standard symbol lookup routine on the appropriate table in the `syms' array. d1820 2 a1821 2 (`gmon_io.c:gmon_out_read'), first by checking for a new-style `gmon.out' header, then assuming this is an old-style BSD `gmon.out' if d1824 1 a1824 1 New-style histogram records are read by `hist.c:hist_read_rec'. For d1826 6 a1831 6 bins, and read them in. When multiple profile data files (or files with multiple histogram records) are read, the memory ranges of each pair of histogram records must be either equal, or non-overlapping. For each pair of histogram records, the resolution (memory region size divided by the number of bins) must be the same. The time unit must be the same for all histogram records. If the above containts are met, all d1834 1 a1834 1 As each call graph record is read (`call_graph.c:cg_read_rec'), the d1836 2 a1837 2 call graph arc is created by `cg_arcs.c:arc_add', unless the arc fails a symspec check against INCL_ARCS/EXCL_ARCS. As each arc is added, a d1842 6 a1847 6 Basic-block records are read (`basic_blocks.c:bb_read_rec'), but only if line-by-line profiling has been selected. Each basic-block address is matched to a corresponding line symbol in the symbol table, and an entry made in the symbol's bb_addr and bb_calls arrays. Again, if multiple basic-block records are present for the same address, the call counts are cumulative. d1849 1 a1849 1 A gmon.sum file is dumped, if requested (`gmon_io.c:gmon_out_write'). d1852 10 a1861 11 (`hist.c:hist_assign_samples') by iterating over all the sample bins and assigning them to symbols. Since the symbol table is sorted in order of ascending memory addresses, we can simple follow along in the symbol table as we make our pass over the sample bins. This step includes a symspec check against INCL_FLAT/EXCL_FLAT. Depending on the histogram scale factor, a sample bin may span multiple symbols, in which case a fraction of the sample count is allocated to each symbol, proportional to the degree of overlap. This effect is rare for normal profiling, but overlaps are more common during line-by-line profiling, and can cause each of two adjacent lines to be credited with half a hit, for example. d1863 2 a1864 2 If call graph data is present, `cg_arcs.c:cg_assemble' is called. First, if `-c' was specified, a machine-dependent routine (`find_call') d1867 18 a1884 19 A topological sort is performed by depth-first numbering all the symbols (`cg_dfn.c:cg_dfn'), so that children are always numbered less than their parents, then making a array of pointers into the symbol table and sorting it into numerical order, which is reverse topological order (children appear before parents). Cycles are also detected at this point, all members of which are assigned the same topological number. Two passes are now made through this sorted array of symbol pointers. The first pass, from end to beginning (parents to children), computes the fraction of child time to propagate to each parent and a print flag. The print flag reflects symspec handling of INCL_GRAPH/EXCL_GRAPH, with a parent's include or exclude (print or no print) property being propagated to its children, unless they themselves explicitly appear in INCL_GRAPH or EXCL_GRAPH. A second pass, from beginning to end (children to parents) actually propagates the timings along the call graph, subject to a check against INCL_TIME/EXCL_TIME. With the print flag, fractions, and timings now stored in the symbol structures, the topological sort array is now discarded, and a new array of pointers is assembled, this time sorted by propagated time. d1887 2 a1888 2 fairly straightforward. The call graph (`cg_print.c:cg_print') and flat profile (`hist.c:hist_print') are regurgitations of values already d1890 1 a1890 1 (`basic_blocks.c:print_annotated_source') uses basic-block information, d1895 1 a1895 1 source code itself (`cg_print.c'). Basically, the functions with the d1903 1 a1903 1 9.4 Debugging `gprof' d1906 4 a1909 4 If `gprof' was compiled with debugging enabled, the `-d' option triggers debugging output (to stdout) which can be helpful in understanding its operation. The debugging number specified is interpreted as a sum of the following options: a1912 1 a1914 1 a1917 1 a1920 1 a1923 1 d1926 1 a1926 1 file. For line-by-line profiling (`-l' option), also shows line a1927 1 d1929 1 a1929 2 Trace operation of `-c' option a1931 1 a1933 1 d1936 1 a1936 2 meaningful with `-l' option) a1938 1 d1940 1 a1940 1 Tracks operation of `-A' option d1951 1 a1951 1 `http://fsf.org/' d1976 2 a1977 2 of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is d1983 2 a1984 2 that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice d1988 3 a1990 3 of the public is a licensee, and is addressed as "you". You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. d2008 6 a2013 6 titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. d2024 10 a2033 10 straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not "Transparent" is called "Opaque". d2037 8 a2044 8 SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. d2082 2 a2083 2 distribute a large enough number of copies you must also follow the conditions in section 3. d2097 5 a2101 6 front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects. d2109 11 a2119 12 numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public. d2122 3 a2124 3 the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document. d2130 5 a2134 5 release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version: d2137 5 a2141 5 distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. d2171 6 a2176 6 authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. d2181 5 a2185 5 previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. d2188 2 a2189 2 Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor d2192 3 a2194 4 L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. d2207 5 a2211 5 material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles. d2220 9 a2228 9 and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one. d2238 2 a2239 2 modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, d2266 2 a2267 2 rules of this License for verbatim copying of each of the documents in all other respects. d2271 3 a2273 3 a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document. d2278 2 a2279 2 separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an "aggregate" if the d2324 2 a2325 2 provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the d2337 4 a2340 4 the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it. d2342 1 a2342 1 10. FUTURE REVISIONS OF THIS LICENSE d2348 1 a2348 1 `http://www.gnu.org/copyleft/'. d2355 5 a2359 5 published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that d2363 1 a2363 1 11. RELICENSING a2392 1 d2409 1 a2409 1 Texts, replace the "with...Texts." line with this: d2420 3 a2422 3 recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software. d2427 18 a2444 18 Node: Top722 Node: Introduction2045 Node: Compiling4537 Node: Executing8593 Node: Invoking11381 Node: Output Options12796 Node: Analysis Options19885 Node: Miscellaneous Options23803 Node: Deprecated Options25058 Node: Symspecs27127 Node: Output28953 Node: Flat Profile29993 Node: Call Graph34946 Node: Primary38178 Node: Callers40766 Node: Subroutines42883 Node: Cycles44724 Node: Line-by-line51501 d2446 11 a2456 11 Node: Inaccuracy58573 Node: Sampling Error58831 Node: Assumptions61735 Node: How do I?63205 Node: Incompatibilities64759 Node: Details66253 Node: Implementation66646 Node: File Format72543 Node: Internals76833 Node: Debugging85328 Node: GNU Free Documentation License86929 @ 1.5.12.1 log @Sync with HEAD @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.3 from gprof.texi. d3 6 a8 1 This file documents the gprof profiler of the GNU system. d10 1 a10 1 Copyright (C) 1988-2018 Free Software Foundation, Inc. a18 5 INFO-DIR-SECTION Software development START-INFO-DIR-ENTRY * gprof: (gprof). Profiling your program's execution END-INFO-DIR-ENTRY d25 1 a25 1 This manual describes the GNU profiler, 'gprof', and how you can use it d28 1 a28 1 programs. GNU 'gprof' was written by Jay Fenlason. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.0. d33 2 a34 2 Documentation License version 1.3. A copy of the license is included in the section entitled "GNU Free Documentation License". d42 1 a42 1 * Invoking:: How to run 'gprof', and its options d44 1 a44 1 * Output:: Interpreting 'gprof''s output d48 1 a48 1 * Incompatibilities:: (between GNU 'gprof' and Unix 'gprof'.) d67 6 a72 5 execution of your program, it can be used on programs that are too large or too complex to analyze by reading the source. However, how your program is run will affect the information that shows up in the profile data. If you don't use some feature of your program while it is being profiled, no profile information will be generated for that feature. d82 1 a82 1 * You must run 'gprof' to analyze the profile data. *Note 'gprof' d94 6 a99 5 The "call graph" shows, for each function, which functions called it, which other functions it called, and how many times. There is also an estimate of how much time was spent in the subroutines of each function. This can suggest places where you might try to eliminate function calls that use a lot of time. *Note The Call Graph: Call Graph. d118 1 a118 1 To compile a source file for profiling, specify the '-pg' option when d122 3 a124 3 To link the program for profiling, if you use a compiler such as 'cc' to do the linking, simply specify '-pg' in addition to your usual options. The same option, '-pg', alters either compilation or linking d130 1 a130 1 The '-pg' option also works with a command that both compiles and d135 3 a137 3 Note: The '-pg' option must be part of your compilation options as well as your link options. If it is not then no call-graph data will be gathered and when you run 'gprof' you will get an error message like d142 2 a143 2 If you add the '-Q' switch to suppress the printing of the call graph data you will still be able to see the time samples: d154 8 a161 8 If you run the linker 'ld' directly instead of through a compiler such as 'cc', you may have to specify a profiling startup file 'gcrt0.o' as the first input file instead of the usual startup file 'crt0.o'. In addition, you would probably want to specify the profiling C library, 'libc_p.a', by writing '-lc_p' instead of the usual '-lc'. This is not absolutely necessary, but doing this gives you number-of-calls information for standard library functions such as 'read' and 'open'. For example: d166 2 a167 2 libraries you may run into problems with the profiling support code in a shared library being called before that library has been fully d170 3 a172 3 against a static version of the library containing the profiling support code, which for 'gcc' users can be done via the '-static' or '-static-libgcc' command line option. For example: d176 1 a176 1 If you compile only some of the modules of the program with '-pg', d178 6 a183 6 information about the modules that were compiled without '-pg'. The only information you get for the functions in those modules is the total time spent in them; there is no record of how many times they were called, or from where. This will not affect the flat profile (except that the 'calls' field for the functions will be blank), but will greatly reduce the usefulness of the call graph. d186 1 a186 1 'gcov' tool instead of 'gprof'. See that tool's manual or info pages d189 3 a191 3 Note, older versions of 'gcc' produce line-by-line profiling information that works with 'gprof' rather than 'gcov' so there is still support for displaying this kind of information in 'gprof'. *Note d194 5 a198 5 It also worth noting that 'gcc' implements a '-finstrument-functions' command line option which will insert calls to special user supplied instrumentation routines at the entry and exit of every function in their program. This can be used to implement an alternative profiling scheme. d207 1 a207 1 generate the information that 'gprof' needs. Simply run the program as d221 2 a222 2 'gmon.out' just before exiting. If there is already a file called 'gmon.out', its contents are overwritten. There is currently no way to d227 3 a229 3 In order to write the 'gmon.out' file properly, your program must exit normally: by returning from 'main' or by calling 'exit'. Calling the low-level function '_exit' does not write the profile data, and d232 1 a232 1 The 'gmon.out' file is written in the program's _current working d234 2 a235 2 'chdir', the 'gmon.out' file will be left in the last directory your program 'chdir''d to. If you don't have permission to write in this d239 1 a239 1 called 'bb.out'. This file, if present, contains an human-readable d241 4 a244 4 appearance of a human-readable 'bb.out' means the basic-block counts didn't get written into 'gmon.out'. The Perl script 'bbconv.pl', included with the 'gprof' source distribution, will convert a 'bb.out' file into a format readable by 'gprof'. Invoke it like this: d248 2 a249 2 This translates the information in 'bb.out' into a form that 'gprof' can understand. But you still need to tell 'gprof' about the existence d251 1 a251 1 'gprof' command line, _along with 'gmon.out'_, like this: d258 1 a258 1 4 'gprof' Command Summary d261 2 a262 2 After you have a profile data file 'gmon.out', you can run 'gprof' to interpret the information in it. The 'gprof' program prints a flat d264 1 a264 1 redirect the output of 'gprof' into a file with '>'. d266 1 a266 1 You run 'gprof' like this: d272 5 a276 4 If you omit the executable file name, the file 'a.out' is used. If you give no profile data file name, the file 'gmon.out' is used. If any file is not in the proper format, or if the profile data file does not appear to belong to the executable file, an error message is printed. d286 2 a287 2 * Output Options:: Controlling 'gprof''s output style * Analysis Options:: Controlling how 'gprof' analyzes its data d299 1 a299 1 These options specify which of several output formats 'gprof' should d302 4 a305 4 Many of these options take an optional "symspec" to specify functions to be included or excluded. These options can be specified multiple times, with different symspecs, to include or exclude sets of symbols. *Note Symspecs: Symspecs. d307 1 a307 1 Specifying any of these options overrides the default ('-p -q'), d310 3 a312 3 '-A[SYMSPEC]' '--annotated-source[=SYMSPEC]' The '-A' option causes 'gprof' to print annotated source code. If d316 3 a318 3 '-b' '--brief' If the '-b' option is given, 'gprof' doesn't print the verbose d323 3 a325 3 '-C[SYMSPEC]' '--exec-counts[=SYMSPEC]' The '-C' option causes 'gprof' to print a tally of functions and d330 1 a330 1 specifying the '-l' option, along with '-C', will cause basic-block d333 9 a341 9 '-i' '--file-info' The '-i' option causes 'gprof' to display summary information about the profile data file(s) and then exit. The number of histogram, call graph, and basic-block count records is displayed. '-I DIRS' '--directory-path=DIRS' The '-I' option specifies a list of search directories in which to d346 4 a349 4 '-J[SYMSPEC]' '--no-annotated-source[=SYMSPEC]' The '-J' option causes 'gprof' not to print annotated source code. If SYMSPEC is specified, 'gprof' prints annotated source, but d352 2 a353 2 '-L' '--print-path' d355 1 a355 1 suppressed. The '-L' option causes 'gprof' to print the full d360 10 a369 10 '-p[SYMSPEC]' '--flat-profile[=SYMSPEC]' The '-p' option causes 'gprof' to print a flat profile. If SYMSPEC is specified, print flat profile only for matching symbols. *Note The Flat Profile: Flat Profile. '-P[SYMSPEC]' '--no-flat-profile[=SYMSPEC]' The '-P' option causes 'gprof' to suppress printing a flat profile. If SYMSPEC is specified, 'gprof' prints a flat profile, but d372 3 a374 3 '-q[SYMSPEC]' '--graph[=SYMSPEC]' The '-q' option causes 'gprof' to print the call graph analysis. d378 4 a381 4 '-Q[SYMSPEC]' '--no-graph[=SYMSPEC]' The '-Q' option causes 'gprof' to suppress printing the call graph. If SYMSPEC is specified, 'gprof' prints a call graph, but excludes d384 3 a386 3 '-t' '--table-length=NUM' The '-t' option causes the NUM most active source lines in each d390 2 a391 2 '-y' '--separate-files' d393 1 a393 1 'gprof' prints annotated source files to standard-output. If this d395 9 a403 9 'path/FILENAME' is generated in the file 'FILENAME-ann'. If the underlying file system would truncate 'FILENAME-ann' so that it overwrites the original 'FILENAME', 'gprof' generates annotated source in the file 'FILENAME.ann' instead (if the original file name has an extension, that extension is _replaced_ with '.ann'). '-Z[SYMSPEC]' '--no-exec-counts[=SYMSPEC]' The '-Z' option causes 'gprof' not to print a tally of functions d407 3 a409 3 '-r' '--function-ordering' The '--function-ordering' option causes 'gprof' to print a d419 7 a425 7 '-R MAP_FILE' '--file-ordering MAP_FILE' The '--file-ordering' option causes 'gprof' to print a suggested .o link line ordering for the program based on profiling data. This option suggests an ordering which may improve paging, tlb and cache behavior for the program on systems which do not support arbitrary ordering of functions in an executable. d427 1 a427 1 Use of the '-a' argument is highly recommended with this option. d431 1 a431 1 similar to the output of the program 'nm'. d442 1 a442 2 To create a MAP_FILE with GNU 'nm', type a command like 'nm d445 4 a448 4 '-T' '--traditional' The '-T' option causes 'gprof' to print its output in "traditional" BSD style. d450 2 a451 2 '-w WIDTH' '--width=WIDTH' d455 7 a461 7 '-x' '--all-lines' This option affects annotated source output only. By default, only the lines at the beginning of a basic-block are annotated. If this option is specified, every line in a basic-block is annotated by repeating the annotation for the first line. This behavior is similar to 'tcov''s '-a'. d463 2 a464 2 '--demangle[=STYLE]' '--no-demangle' d467 1 a467 1 '--no-demangle' option may be used to turn off demangling. d478 3 a480 3 '-a' '--no-static' The '-a' option causes 'gprof' to suppress the printing of d486 2 a487 2 the executable file. This option affects both the flat profile and the call graph. d489 3 a491 3 '-c' '--static-call-graph' The '-c' option causes the call graph of the program to be d496 12 a507 12 have been called, but never were. Calls to functions that were not compiled with profiling enabled are also identified, but only if symbol table entries are present for them. Calls to dynamic library routines are typically _not_ found by this option. Parents or children identified via this heuristic are indicated in the call graph with call counts of '0'. '-D' '--ignore-non-functions' The '-D' option causes 'gprof' to ignore symbols which are not known to be functions. This option will give more accurate profile data on systems where it is supported (Solaris and HPUX for d510 2 a511 2 '-k FROM/TO' The '-k' option allows you to delete from the call graph any arcs d514 3 a516 3 '-l' '--line' The '-l' option enables line-by-line profiling, which causes d519 2 a520 2 compiled by older versions of the 'gcc' compiler. Newer versions of 'gcc' are designed to work with the 'gcov' tool instead. d522 5 a526 5 If the program was compiled with basic-block counting enabled, this option will also identify how many times each line of code was executed. While line-by-line profiling can help isolate where in a large function a program is spending its time, it also significantly increases the running time of 'gprof', and magnifies d530 4 a533 4 '--inline-file-names' This option causes 'gprof' to print the source file after each symbol in both the flat profile and the call graph. The full path to the file is printed if used with the '-L' option. d535 2 a536 2 '-m NUM' '--min-count=NUM' d540 8 a547 3 '-nSYMSPEC' '--time=SYMSPEC' The '-n' option causes 'gprof', in its call graph analysis, to only d550 13 a562 4 '-NSYMSPEC' '--no-time=SYMSPEC' The '-n' option causes 'gprof', in its call graph analysis, not to propagate times for symbols matching SYMSPEC. a563 13 '-SFILENAME' '--external-symbol-table=FILENAME' The '-S' option causes 'gprof' to read an external symbol table file, such as '/proc/kallsyms', rather than read the symbol table from the given object file (the default is 'a.out'). This is useful for profiling kernel modules. '-z' '--display-unused-functions' If you give the '-z' option, 'gprof' will mention all functions in the flat profile, even those that were never called, and that had no time spent in them. This is useful in conjunction with the '-c' option for discovering which routines were never called. d571 4 a574 4 '-d[NUM]' '--debug[=NUM]' The '-d NUM' option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging 'gprof': d577 3 a579 3 '-h' '--help' The '-h' option prints command line usage. d581 2 a582 2 '-ONAME' '--file-format=NAME' d584 2 a585 2 are 'auto' (the default), 'bsd', '4.4bsd', 'magic', and 'prof' (not yet supported). d587 3 a589 3 '-s' '--sum' The '-s' option causes 'gprof' to summarize the information in the d591 11 a601 11 called 'gmon.sum', which contains all the information from the profile data files that 'gprof' read in. The file 'gmon.sum' may be one of the specified input files; the effect of this is to merge the data in the other input files into 'gmon.sum'. Eventually you can run 'gprof' again without '-s' to analyze the cumulative data in the file 'gmon.sum'. '-v' '--version' The '-v' flag causes 'gprof' to print the current version number, d604 1 d613 2 a614 2 '-e FUNCTION_NAME' The '-e FUNCTION' option tells 'gprof' to not print information d617 11 a627 10 functions that call it, but its index number will be shown as '[not printed]'. More than one '-e' option may be given; only one FUNCTION_NAME may be indicated with each '-e' option. '-E FUNCTION_NAME' The '-E FUNCTION' option works like the '-e' option, but time spent in the function (and children who were not called from anywhere else), will not be used to compute the percentages-of-time for the call graph. More than one '-E' option may be given; only one FUNCTION_NAME may be indicated with each '-E' option. d629 2 a630 2 '-f FUNCTION_NAME' The '-f FUNCTION' option causes 'gprof' to limit the call graph to d632 10 a641 2 children...). More than one '-f' option may be given; only one FUNCTION_NAME may be indicated with each '-f' option. a642 7 '-F FUNCTION_NAME' The '-F FUNCTION' option works like the '-f' option, but only time spent in the function and its children (and their children...) will be used to determine total-time and percentages-of-time for the call graph. More than one '-F' option may be given; only one FUNCTION_NAME may be indicated with each '-F' option. The '-F' option overrides the '-E' option. d644 2 a645 2 Note that only one function can be specified with each '-e', '-E', '-f' or '-F' option. To specify more than one function, use multiple d651 1 a651 1 'foo' or 'bar' and were not reachable from 'boring'. d670 6 a675 6 'main.c' Selects everything in file 'main.c'--the dot in the string tells 'gprof' to interpret the string as a filename, rather than as a function name. To select a file whose name does not contain a dot, a trailing colon should be specified. For example, 'odd:' is interpreted as the file named 'odd'. d677 2 a678 2 'main' Selects all functions named 'main'. d687 2 a688 2 necessary to add a leading colon to the name. For example, ':.mul' selects function '.mul'. d691 2 a692 2 'gprof' will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as 'gprof' prints d694 2 a695 2 '_main' from your 'main' function, 'gprof' still prints it as 'main' in its output, so you should use 'main' in symspecs. d697 2 a698 2 'main.c:main' Selects function 'main' in file 'main.c'. d700 2 a701 2 'main.c:134' Selects line 134 in file 'main.c'. d706 1 a706 1 5 Interpreting 'gprof''s Output d709 1 a709 1 'gprof' can produce several different output styles, the most important d722 1 a722 1 * Line-by-line:: 'gprof' can analyze individual source code lines d733 1 a733 1 executing each function. Unless the '-z' option is given, functions d764 1 a764 1 functions 'mcount' and 'profil' are part of the profiling apparatus and d774 8 a781 8 'cumulative seconds' field. Since each sample counted for 0.01 seconds, this means only six samples were taken during the run. Two of the samples occurred while the program was in the 'open' function, as indicated by the 'self seconds' field. Each of the other four samples occurred one each in 'offtime', 'memccpy', 'write', and 'mcount'. Since only six samples were taken, none of these values can be regarded as particularly reliable. In another run, the 'self seconds' field for 'mcount' might well be '0.00' or '0.02'. *Note Statistical Sampling d784 2 a785 2 The remaining functions in the listing (those whose 'self seconds' field is '0.00') didn't appear in the histogram samples at all. d787 1 a787 1 they are listed, sorted in decreasing order by the 'calls' field. d794 1 a794 1 '% time' d798 1 a798 1 'cumulative seconds' d803 1 a803 1 'self seconds' d807 1 a807 1 'calls' d813 1 a813 1 'self ms/call' d818 1 a818 1 'total ms/call' d825 2 a826 2 'name' This is the name of the function. The flat profile is sorted by d838 2 a839 2 they themselves may not have used much time, called other functions that did use unusual amounts of time. d842 1 a842 1 same 'gprof' run as the flat profile example in the previous section. d877 4 a880 4 number in square brackets. The end of this line says which function the entry is for. The preceding lines in the entry describe the callers of this function and the following lines describe its subroutines (also called "children" when we speak of the call graph). d885 1 a885 1 The internal profiling function 'mcount' (*note The Flat Profile: d894 1 a894 1 such as 'a' calls 'b' calls 'a'... d907 1 a907 1 'report' in our main example, together with the heading line that shows d916 1 a916 1 'index' d926 1 a926 1 '% time' d935 1 a935 1 'self' d937 1 a937 1 should be identical to the number printed in the 'seconds' field d940 5 a944 5 'children' This is the total amount of time spent in the subroutine calls made by this function. This should be equal to the sum of all the 'self' and 'children' entries of the children listed directly below this function. d946 1 a946 1 'called' d950 1 a950 1 separated by a '+'. The first number counts non-recursive calls, d953 2 a954 2 In the example above, the function 'report' was called once from 'main'. d956 1 a956 1 'name' d963 1 a963 1 example, if function 'gnurr' is part of cycle number one, and has d974 3 a976 3 A function's entry has a line for each function it was called by. These lines' fields correspond to the fields of the primary line, but their meanings are different because of the difference in context. d979 1 a979 1 'report', the primary line and one caller-line preceding it, together d987 2 a988 2 Here are the meanings of the fields in the caller-line for 'report' called from 'main': d990 14 a1003 14 'self' An estimate of the amount of time spent in 'report' itself when it was called from 'main'. 'children' An estimate of the amount of time spent in subroutines of 'report' when 'report' was called from 'main'. The sum of the 'self' and 'children' fields is an estimate of the amount of time spent within calls to 'report' from 'main'. 'called' Two numbers: the number of times 'report' was called from 'main', followed by the total number of non-recursive calls to 'report' d1006 2 a1007 2 'name and index number' The name of the caller of 'report' to which this line applies, d1011 1 a1011 1 'gprof' request the omission of certain functions. When a caller d1019 1 a1019 1 dummy caller-line is printed which has '' as the "caller's d1034 1 a1034 1 'main', the primary line and a line for a subroutine, together with the d1042 2 a1043 2 Here are the meanings of the fields in the subroutine-line for 'main' calling 'report': d1045 17 a1061 17 'self' An estimate of the amount of time spent directly within 'report' when 'report' was called from 'main'. 'children' An estimate of the amount of time spent in subroutines of 'report' when 'report' was called from 'main'. The sum of the 'self' and 'children' fields is an estimate of the total time spent in calls to 'report' from 'main'. 'called' Two numbers, the number of calls to 'report' from 'main' followed by the total number of non-recursive calls to 'report'. This ratio is used to determine how much of 'report''s 'self' and 'children' time gets credited to 'main'. *Note Estimating 'children' Times: Assumptions. d1063 2 a1064 2 'name' The name of the subroutine of 'main' to which this line applies, d1076 2 a1077 2 The graph may be complicated by the presence of "cycles of recursion" in the call graph. A cycle exists if a function calls another function d1079 2 a1080 2 function. For example: if 'a' calls 'b', and 'b' calls 'a', then 'a' and 'b' form a cycle. d1083 4 a1086 4 they belong to the same cycle. If 'a' and 'b' call each other and 'b' and 'c' call each other, all three make one cycle. Note that even if 'b' only calls 'a' if it was not called from 'a', 'gprof' cannot determine this, so 'a' and 'b' are still considered a cycle. d1090 1 a1090 1 graph it is followed by ''. d1093 4 a1096 4 call graph paradoxical. The "time spent in children" of 'a' should include the time spent in its subroutine 'b' and in 'b''s subroutines--but one of 'b''s subroutines is 'a'! How much of 'a''s time should be included in the children of 'a', when 'a' is indirectly d1099 1 a1099 1 The way 'gprof' resolves this paradox is by creating a single entry d1104 2 a1105 2 "callers" of the cycle are the functions, outside the cycle, that called functions in the cycle. d1108 2 a1109 2 containing functions 'a' and 'b'. The cycle was entered by a call to 'a' from 'main'; both 'a' and 'b' called 'c'. d1131 2 a1132 2 for 'main', which calls 'a', and an entry for 'c', with callers 'a' and 'b'.) d1165 2 a1166 2 The 'self' field of the cycle's primary line is the total time spent in all the functions of the cycle. It equals the sum of the 'self' d1170 6 a1175 6 The 'children' fields of the cycle's primary line and subroutine lines count only subroutines outside the cycle. Even though 'a' calls 'b', the time spent in those calls to 'b' is not counted in 'a''s 'children' time. Thus, we do not encounter the problem of what to do when the time in those calls to 'b' includes indirect recursive calls back to 'a'. d1177 1 a1177 1 The 'children' field of a caller-line in the cycle's entry estimates d1182 1 a1182 1 The 'called' field in the primary line for the cycle has two numbers: d1185 8 a1192 8 called by functions in the cycle (including times when a function in the cycle calls itself). This is a generalization of the usual split into non-recursive and recursive calls. The 'called' field of a subroutine-line for a cycle member in the cycle's entry says how many time that function was called from functions in the cycle. The total of all these is the second number in the primary line's 'called' field. d1196 4 a1199 4 These lines show how many times each function in the cycle called or was called from each other function in the cycle. The 'self' and 'children' fields in these lines are blank because of the difficulty of defining meanings for them when recursion is going on. d1207 1 a1207 1 'gprof''s '-l' option causes the program to perform "line-by-line" d1209 3 a1211 3 functions, but to individual lines of source code. This only works with programs compiled with older versions of the 'gcc' compiler. Newer versions of 'gcc' use a different program - 'gcov' - to display d1214 2 a1215 2 With the older versions of 'gcc' the program usually has to be compiled with a '-g' option, in addition to '-pg', in order to generate d1217 1 a1217 1 versions of 'gcc' the program had to be compiled with the '-a' command d1222 1 a1222 1 version of 'gprof' does not propagate call graph arcs from source code d1226 3 a1228 3 Here is a section of 'gprof''s output, without line-by-line profiling. Note that 'ct_init' accounted for four histogram hits, and 13327 calls to 'init_block'. d1251 2 a1252 3 Now let's look at some of 'gprof''s output from the same program run, this time with line-by-line profiling enabled. Note that 'ct_init''s d1255 1 a1255 1 note how 'ct_init''s 13327 calls to 'init_block' are broken down into a1290 1 d1297 1 a1297 1 'gprof''s '-A' option triggers an annotated source listing, which lists d1299 2 a1300 2 times it was called. You may also need to specify the '-I' option, if 'gprof' can't find the source code files. d1302 1 a1302 1 With older versions of 'gcc' compiling with 'gcc ... -g -pg -a' d1304 3 a1306 3 function counting code. This enables 'gprof' to determine how many times each line of code was executed. With newer versions of 'gcc' support for displaying basic-block counts is provided by the 'gcov' d1332 6 a1337 7 'updcrc' has at least five basic-blocks. One is the function itself. The 'if' statement on line 9 generates two more basic-blocks, one for each branch of the 'if'. A fourth basic-block results from the 'if' on line 13, and the contents of the 'do' loop form the fifth basic-block. The compiler may also generate additional basic-blocks to handle various special cases. d1340 3 a1342 3 'gprof -l -A'. The '-x' option is also helpful, to ensure that each line of code is labeled at least once. Here is 'updcrc''s annotated source listing for a sample 'gzip' run: d1365 2 a1366 2 each branch of the 'if' statement. The body of the 'do' loop was executed a total of 26312 times. Note how the 'while' statement is d1374 1 a1374 1 6 Inaccuracy of 'gprof' Output d1388 1 a1388 1 The run-time figures that 'gprof' gives you are based on a sampling d1401 2 a1402 2 beware that the mcount counting function in glibc is _not_ thread-safe). *Note Implementation of Profiling: Implementation. d1411 1 a1411 1 period is 0.01 seconds and 'foo''s run-time is 1 second, N is 100 d1413 1 a1413 1 error in 'foo''s run-time is 0.1 seconds (10*0.01 seconds), or ten d1415 5 a1419 5 seconds and 'bar''s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in 'bar''s run-time is 1 second, or one percent of the observed value. It is likely to vary this much _on the average_ from one profiling run to the next. (_Sometimes_ it will vary more.) d1429 1 a1429 1 the data from several runs, using the '-s' option of 'gprof'. Here is d1434 1 a1434 1 2. Issue the command 'mv gmon.out gmon.sum'. d1438 1 a1438 1 4. Merge the new data in 'gmon.out' into 'gmon.sum' with this command: d1451 1 a1451 1 6.2 Estimating 'children' Times d1455 2 a1456 2 'children' time values and all the time figures in caller and subroutine lines. d1459 1 a1459 1 profile data itself. Instead, 'gprof' estimates them by making an d1463 4 a1466 3 any function 'foo' is not correlated with who called 'foo'. If 'foo' used 5 seconds in all, and 2/5 of the calls to 'foo' came from 'a', then 'foo' contributes 2 seconds to 'a''s 'children' time, by assumption. d1469 6 a1474 6 far from true. Suppose that 'foo' returns very quickly when its argument is zero; suppose that 'a' always passes zero as an argument, while other callers of 'foo' pass other arguments. In this program, all the time spent in 'foo' is in the calls from callers other than 'a'. But 'gprof' has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in 'foo' to the children of 'a'. d1476 1 a1476 1 We hope some day to put more complete data into 'gmon.out', so that a1486 1 d1488 1 a1488 1 Because 'gprof' can only report call times and counts by function, d1492 1 a1492 1 artificial hot spots since compiling with '-pg' adds a significant d1497 1 a1497 2 Use the 'gcov' program. d1500 1 a1500 2 Use 'gprof -l' and lookup the function in the call graph. The a1503 1 d1519 1 d1523 1 a1523 1 8 Incompatibilities with Unix 'gprof' d1526 3 a1528 3 GNU 'gprof' and Berkeley Unix 'gprof' use the same data file 'gmon.out', and provide essentially the same information. But there are a few differences. d1530 1 a1530 1 * GNU 'gprof' uses a new, generalized file format with support for d1532 1 a1532 1 cookie and version number allows 'gprof' to easily identify new d1536 3 a1538 3 * For a recursive function, Unix 'gprof' lists the function as a parent and as a child, with a 'calls' field that lists the number of recursive calls. GNU 'gprof' omits these lines and puts the d1541 2 a1542 2 * When a function is suppressed from the call graph with '-e', GNU 'gprof' still lists it as a subroutine of functions that call it. d1544 2 a1545 2 * GNU 'gprof' accepts the '-k' with its argument in the form 'from/to', instead of 'from to'. d1547 3 a1549 3 * In the annotated source listing, if there are multiple basic blocks on the same line, GNU 'gprof' prints all of their counts, separated by commas. d1552 1 a1552 1 'gprof' prints blurbs after the tables, so that you can see the d1564 3 a1566 3 * File Format:: Format of 'gmon.out' files * Internals:: 'gprof''s internal operation * Debugging:: Using 'gprof''s '-d' option d1579 2 a1580 2 the '-pg' option, which causes every function to call 'mcount' (or '_mcount', or '__mcount', depending on the OS and compiler) as one of d1583 1 a1583 1 The 'mcount' routine, included in the profiling library, is d1586 3 a1588 3 done by examining the stack frame to find both the address of the child, and the return address in the original parent. Since this is a very machine-dependent operation, 'mcount' itself is typically a short d1590 4 a1593 4 and then calls '__mcount_internal' (a normal C function) with two arguments--'frompc' and 'selfpc'. '__mcount_internal' is responsible for maintaining the in-memory call graph, which records 'frompc', 'selfpc', and the number of times each of these call arcs was traversed. d1596 1 a1596 1 ('__builtin_return_address'), which allows a generic 'mcount' function d1600 1 a1600 1 'mcount' is used for performance reasons. d1604 2 a1605 2 same as in the usual C library, but they were compiled with '-pg'. If you link your program with 'gcc ... -pg', it automatically uses the d1608 5 a1612 5 Profiling also involves watching your program as it runs, and keeping a histogram of where the program counter happens to be every now and then. Typically the program counter is looked at around 100 times per second of run time, but the exact frequency may vary from system to system. d1615 1 a1615 1 provide a 'profil()' system call, which registers a memory array with d1617 7 a1623 7 address space maps into the array. Typical scaling values cause every 2 to 8 bytes of address space to map into a single array slot. On every tick of the system clock (assuming the profiled program is running), the value of the program counter is examined and the corresponding slot in the memory array is incremented. Since this is done in the kernel, which had to interrupt the process anyway to handle the clock interrupt, very little additional system overhead is required. d1626 9 a1634 8 earlier), do not provide a 'profil()' system call. On such a system, arrangements are made for the kernel to periodically deliver a signal to the process (typically via 'setitimer()'), which then performs the same operation of examining the program counter and incrementing a slot in the memory array. Since this method requires a signal to be delivered to user space every time a sample is taken, it uses considerably more overhead than kernel-based profiling. Also, due to the added delay required to deliver the signal, this method is less accurate as well. d1637 8 a1644 8 either calls 'profil()' or sets up a clock signal handler. This routine ('monstartup') can be invoked in several ways. On Linux systems, a special profiling startup file 'gcrt0.o', which invokes 'monstartup' before 'main', is used instead of the default 'crt0.o'. Use of this special startup file is one of the effects of using 'gcc ... -pg' to link. On SPARC systems, no special startup files are used. Rather, the 'mcount' routine, when it is invoked for the first time (typically when 'main' is called), calls 'monstartup'. d1646 1 a1646 1 If the compiler's '-a' option was used, basic-block counting is also d1649 1 a1649 1 basic-block begins (i.e., when an 'if' statement appears), an extra d1656 5 a1660 5 The profiling library also includes a function ('mcleanup') which is typically registered using 'atexit()' to be called as the program exits, and is responsible for writing the file 'gmon.out'. Profiling is turned off, various headers are output, and the histogram is written, followed by the call-graph arcs and the basic-block counts. d1662 1 a1662 1 The output from 'gprof' gives no indication of parts of your program d1664 8 a1671 8 of the program counter are taken at fixed intervals of the program's run time. Therefore, the time measurements in 'gprof' output say nothing about time that your program was not running. For example, a part of the program that creates so much data that it cannot all fit in physical memory at once may run very slowly due to thrashing, but 'gprof' will say it uses little time. On the other hand, sampling by run time has the advantage that the amount of load due to other users won't directly affect the output you get. d1681 2 a1682 2 'gprof' file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU 'gprof' d1684 1 a1684 1 compatibility, GNU 'gprof' continues to support the old BSD-derived d1689 1 a1689 1 The new file format is defined in header file 'gmon_out.h'. It d1693 1 a1693 1 the profile was collected. GNU 'gprof' adapts automatically to the d1700 1 a1700 1 reading a file, GNU 'gprof' will ensure records of the same type are d1710 15 a1724 15 spans, the size of the histogram in bytes (unlike in the old BSD format, this does not include the size of the header), the rate of the profiling clock, and the physical dimension that the bin counts represent after being scaled by the profiling clock rate. The physical dimension is specified in two parts: a long name of up to 15 characters and a single character abbreviation. For example, a histogram representing real-time would specify the long name as "seconds" and the abbreviation as "s". This feature is useful for architectures that support performance monitor hardware (which, fortunately, is becoming increasingly common). For example, under DEC OSF/1, the "uprofile" command can be used to produce a histogram of, say, instruction cache misses. In this case, the dimension in the histogram header could be set to "i-cache misses" and the abbreviation could be set to "1" (because it is simply a count, not a physical dimension). Also, the profiling rate would have to be set to 1 in this case. d1737 8 a1744 8 program execution. Arcs are specified by a pair of addresses: the first must be within caller's function and the second must be within the callee's function. When performing profiling at the function level, these addresses can point anywhere within the respective function. However, when profiling at the line-level, it is better if the addresses are as close to the call-site/entry-point as possible. This will ensure that the line-level call-graph is able to identify exactly which line of source code performed calls to a function. d1750 5 a1754 4 sequence of address/count pairs. The header simply specifies the length of the sequence. In an address/count pair, the address identifies a basic-block and the count specifies the number of times that basic-block was executed. Any address within the basic-address can be used. d1759 1 a1759 1 9.3 'gprof''s Internal Operation d1762 6 a1767 6 Like most programs, 'gprof' begins by processing its options. During this stage, it may building its symspec list ('sym_ids.c:sym_id_add'), if options are specified which use symspecs. 'gprof' maintains a single linked list of symspecs, which will eventually get turned into 12 symbol tables, organized into six include/exclude pairs--one pair each for the flat profile (INCL_FLAT/EXCL_FLAT), the call graph arcs d1774 2 a1775 2 After option processing, 'gprof' finishes building the symspec list by adding all the symspecs in 'default_excluded_list' to the exclude d1781 2 a1782 2 it is an object file, and read its symbol table ('core.c:core_init'), using 'bfd_canonicalize_symtab' after mallocing an appropriately sized d1784 2 a1785 2 '--file-ordering' option has been specified), and the core text space is read into memory (if the '-c' option was given). d1787 1 a1787 1 'gprof''s own symbol table, an array of Sym structures, is now built. d1789 1 a1789 1 whether line-by-line profiling ('-l' option) has been enabled. For d1795 4 a1798 4 the symbols. In between the two passes, a single array of type 'Sym' is created of the appropriate length. Finally, 'symtab.c:symtab_finalize' is called to sort the symbol table and remove duplicate entries (entries with the same memory address). d1801 3 a1803 3 the 'qsort' library function (which sorts an array) will be used to sort the symbol table. Also, the symbol lookup routine ('symtab.c:sym_lookup'), which finds symbols based on memory address, d1805 1 a1805 1 sorted array. Function symbols are indicated with an 'is_func' flag. d1807 1 a1807 1 can have an 'is_static' flag to indicate that it is a local symbol. d1810 11 a1820 11 Syms ('sym_ids.c:sym_id_parse'). Remember that a single symspec can match multiple symbols. An array of symbol tables ('syms') is created, each entry of which is a symbol table of Syms to be included or excluded from a particular listing. The master symbol table and the symspecs are examined by nested loops, and every symbol that matches a symspec is inserted into the appropriate syms table. This is done twice, once to count the size of each required symbol table, and again to build the tables, which have been malloced between passes. From now on, to determine whether a symbol is on an include or exclude symspec list, 'gprof' simply uses its standard symbol lookup routine on the appropriate table in the 'syms' array. d1823 2 a1824 2 ('gmon_io.c:gmon_out_read'), first by checking for a new-style 'gmon.out' header, then assuming this is an old-style BSD 'gmon.out' if d1827 1 a1827 1 New-style histogram records are read by 'hist.c:hist_read_rec'. For d1829 6 a1834 6 bins, and read them in. When multiple profile data files (or files with multiple histogram records) are read, the memory ranges of each pair of histogram records must be either equal, or non-overlapping. For each pair of histogram records, the resolution (memory region size divided by the number of bins) must be the same. The time unit must be the same for all histogram records. If the above containts are met, all d1837 1 a1837 1 As each call graph record is read ('call_graph.c:cg_read_rec'), the d1839 2 a1840 2 call graph arc is created by 'cg_arcs.c:arc_add', unless the arc fails a symspec check against INCL_ARCS/EXCL_ARCS. As each arc is added, a d1845 6 a1850 6 Basic-block records are read ('basic_blocks.c:bb_read_rec'), but only if line-by-line profiling has been selected. Each basic-block address is matched to a corresponding line symbol in the symbol table, and an entry made in the symbol's bb_addr and bb_calls arrays. Again, if multiple basic-block records are present for the same address, the call counts are cumulative. d1852 1 a1852 1 A gmon.sum file is dumped, if requested ('gmon_io.c:gmon_out_write'). d1855 11 a1865 10 ('hist.c:hist_assign_samples') by iterating over all the sample bins and assigning them to symbols. Since the symbol table is sorted in order of ascending memory addresses, we can simple follow along in the symbol table as we make our pass over the sample bins. This step includes a symspec check against INCL_FLAT/EXCL_FLAT. Depending on the histogram scale factor, a sample bin may span multiple symbols, in which case a fraction of the sample count is allocated to each symbol, proportional to the degree of overlap. This effect is rare for normal profiling, but overlaps are more common during line-by-line profiling, and can cause each of two adjacent lines to be credited with half a hit, for example. d1867 2 a1868 2 If call graph data is present, 'cg_arcs.c:cg_assemble' is called. First, if '-c' was specified, a machine-dependent routine ('find_call') d1871 19 a1889 18 A topological sort is performed by depth-first numbering all the symbols ('cg_dfn.c:cg_dfn'), so that children are always numbered less than their parents, then making a array of pointers into the symbol table and sorting it into numerical order, which is reverse topological order (children appear before parents). Cycles are also detected at this point, all members of which are assigned the same topological number. Two passes are now made through this sorted array of symbol pointers. The first pass, from end to beginning (parents to children), computes the fraction of child time to propagate to each parent and a print flag. The print flag reflects symspec handling of INCL_GRAPH/EXCL_GRAPH, with a parent's include or exclude (print or no print) property being propagated to its children, unless they themselves explicitly appear in INCL_GRAPH or EXCL_GRAPH. A second pass, from beginning to end (children to parents) actually propagates the timings along the call graph, subject to a check against INCL_TIME/EXCL_TIME. With the print flag, fractions, and timings now stored in the symbol structures, the topological sort array is now discarded, and a new array of pointers is assembled, this time sorted by propagated time. d1892 2 a1893 2 fairly straightforward. The call graph ('cg_print.c:cg_print') and flat profile ('hist.c:hist_print') are regurgitations of values already d1895 1 a1895 1 ('basic_blocks.c:print_annotated_source') uses basic-block information, d1900 1 a1900 1 source code itself ('cg_print.c'). Basically, the functions with the d1908 1 a1908 1 9.4 Debugging 'gprof' d1911 4 a1914 4 If 'gprof' was compiled with debugging enabled, the '-d' option triggers debugging output (to stdout) which can be helpful in understanding its operation. The debugging number specified is interpreted as a sum of the following options: d1918 1 d1921 1 d1925 1 d1929 1 d1933 1 d1936 1 a1936 1 file. For line-by-line profiling ('-l' option), also shows line d1938 1 d1940 2 a1941 1 Trace operation of '-c' option d1944 1 d1947 1 d1950 2 a1951 1 meaningful with '-l' option) d1954 1 d1956 1 a1956 1 Tracks operation of '-A' option d1967 1 a1967 1 d1992 2 a1993 2 of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is d1999 2 a2000 2 that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice d2004 3 a2006 3 of the public is a licensee, and is addressed as "you". You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. d2024 6 a2029 6 titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. d2040 10 a2049 10 straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not "Transparent" is called "Opaque". d2053 8 a2060 8 SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. d2098 2 a2099 2 distribute a large enough number of copies you must also follow the conditions in section 3. d2113 6 a2118 5 front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects. d2126 12 a2137 11 numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public. d2140 3 a2142 3 the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document. d2148 5 a2152 5 release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version: d2155 5 a2159 5 distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. d2189 6 a2194 6 authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. d2199 5 a2203 5 previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. d2206 2 a2207 2 Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor d2210 4 a2213 3 L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. d2226 5 a2230 5 material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles. d2239 9 a2247 9 and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one. d2257 2 a2258 2 modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, d2285 2 a2286 2 rules of this License for verbatim copying of each of the documents in all other respects. d2290 3 a2292 3 a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document. d2297 2 a2298 2 separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an "aggregate" if the d2343 2 a2344 2 provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the d2356 4 a2359 4 the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it. d2361 1 a2361 1 10. FUTURE REVISIONS OF THIS LICENSE d2367 1 a2367 1 . d2374 5 a2378 5 published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that d2382 1 a2382 1 11. RELICENSING d2412 1 d2429 1 a2429 1 Texts, replace the "with...Texts." line with this: d2440 3 a2442 3 recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software. d2447 18 a2464 18 Node: Top719 Node: Introduction2044 Node: Compiling4535 Node: Executing8591 Node: Invoking11379 Node: Output Options12794 Node: Analysis Options19886 Node: Miscellaneous Options23806 Node: Deprecated Options25060 Node: Symspecs27123 Node: Output28949 Node: Flat Profile29989 Node: Call Graph34942 Node: Primary38174 Node: Callers40762 Node: Subroutines42880 Node: Cycles44721 Node: Line-by-line51498 d2466 11 a2476 11 Node: Inaccuracy58572 Node: Sampling Error58830 Node: Assumptions61734 Node: How do I?63204 Node: Incompatibilities64761 Node: Details66255 Node: Implementation66648 Node: File Format72547 Node: Internals76835 Node: Debugging85325 Node: GNU Free Documentation License86915 @ 1.5.12.2 log @Sync with HEAD, resolve a couple of conflicts @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.4 from gprof.texi. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.90. d2428 29 a2456 29 Node: Introduction2045 Node: Compiling4536 Node: Executing8592 Node: Invoking11380 Node: Output Options12795 Node: Analysis Options19887 Node: Miscellaneous Options23807 Node: Deprecated Options25061 Node: Symspecs27124 Node: Output28950 Node: Flat Profile29990 Node: Call Graph34943 Node: Primary38175 Node: Callers40763 Node: Subroutines42881 Node: Cycles44722 Node: Line-by-line51499 Node: Annotated Source55575 Node: Inaccuracy58573 Node: Sampling Error58831 Node: Assumptions61735 Node: How do I?63205 Node: Incompatibilities64762 Node: Details66256 Node: Implementation66649 Node: File Format72548 Node: Internals76836 Node: Debugging85326 Node: GNU Free Documentation License86916 @ 1.4 log @merge conflicts, regen for x86. @ text @d10 1 a10 1 Copyright (C) 1988-2015 Free Software Foundation, Inc. d30 1 a30 1 This manual is for `gprof' (GNU Binutils) version 2.26. @ 1.4.2.1 log @Sync with HEAD @ text @d10 1 a10 1 Copyright (C) 1988-2016 Free Software Foundation, Inc. d30 1 a30 1 This manual is for `gprof' (GNU Binutils) version 2.27. @ 1.3 log @resolve conflicts @ text @d10 1 a10 2 Copyright (C) 1988, 1992, 1997, 1998, 1999, 2000, 2001, 2003, 2007, 2008, 2009 Free Software Foundation, Inc. d30 1 a30 1 This manual is for `gprof' (GNU Binutils) version 2.23.1. d530 5 d2447 30 a2476 30 Node: Top777 Node: Introduction2103 Node: Compiling4595 Node: Executing8651 Node: Invoking11439 Node: Output Options12854 Node: Analysis Options19943 Node: Miscellaneous Options23641 Node: Deprecated Options24896 Node: Symspecs26965 Node: Output28791 Node: Flat Profile29831 Node: Call Graph34784 Node: Primary38016 Node: Callers40604 Node: Subroutines42721 Node: Cycles44562 Node: Line-by-line51339 Node: Annotated Source55412 Node: Inaccuracy58411 Node: Sampling Error58669 Node: Assumptions61573 Node: How do I?63043 Node: Incompatibilities64597 Node: Details66091 Node: Implementation66484 Node: File Format72381 Node: Internals76671 Node: Debugging85166 Node: GNU Free Documentation License86767 @ 1.2 log @Merge new binutils. Only x86_64 works now; the rest will need to run mknative again (at least). @ text @d31 1 a31 1 This manual is for `gprof' (GNU Binutils)\ version 2.21.1. @ 1.2.8.1 log @Rebase to HEAD as of a few days ago. @ text @d31 1 a31 1 This manual is for `gprof' (GNU Binutils) version 2.23.1. @ 1.2.2.1 log @sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments") @ text @d31 1 a31 1 This manual is for `gprof' (GNU Binutils) version 2.23.1. @ 1.1 log @Initial revision @ text @d3 1 d10 2 a11 2 Copyright (C) 1988, 92, 97, 98, 99, 2000, 2001, 2003, 2007 Free Software Foundation, Inc. d14 1 a14 1 under the terms of the GNU Free Documentation License, Version 1.1 or d31 1 a31 1 This manual is for `gprof' (GNU Binutils) version 2.18.90. d34 2 a35 2 Documentation License. A copy of the license is included in the section entitled "GNU Free Documentation License". d166 11 d546 7 d607 1 a607 2 These options have been replaced with newer versions that use symspecs. d1393 6 a1398 1 vary from run to run if your program is deterministic. d1960 1 a1960 1 Version 1.1, March 2000 d1962 2 a1963 2 Copyright (C) 2000, 2003 Free Software Foundation, Inc. 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA a1967 1 d1971 6 a1976 6 written document "free" in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others. a1991 1 d1994 9 a2002 5 This License applies to any manual or other work that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. The "Document", below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "you." d2008 10 a2017 10 A "Secondary Section" is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document's overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (For example, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them. d2022 4 a2025 1 License. d2029 3 a2031 1 that says that the Document is released under this License. d2035 2 a2036 2 general public, whose contents can be viewed and edited directly and straightforwardly with generic text editors or (for images d2041 5 a2045 3 otherwise Transparent file format whose markup has been designed to thwart or discourage subsequent modification by readers is not Transparent. A copy that is not "Transparent" is called "Opaque." d2050 7 a2056 6 standard-conforming simple HTML designed for human modification. Opaque formats include PostScript, PDF, proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML produced by some word processors for output purposes only. d2065 19 d2102 5 a2106 4 If you publish printed copies of the Document numbering more than 100, and the Document's license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and d2124 4 a2127 5 state in or with each Opaque copy a publicly-accessible computer-network location containing a complete Transparent copy of the Document, free of added material, which the general network-using public has access to download anonymously at no charge using public-standard network protocols. If you use the d2150 69 a2218 51 A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. B. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors, if it has less than five). C. State on the Title page the name of the publisher of the Modified Version, as the publisher. D. Preserve all the copyright notices of the Document. E. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices. F. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below. G. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document's license notice. H. Include an unaltered copy of this License. I. Preserve the section entitled "History", and its title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. J. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. K. In any section entitled "Acknowledgements" or "Dedications", preserve the section's title, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein. L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. M. Delete any section entitled "Endorsements." Such a section may not be included in the Modified Version. N. Do not retitle any existing section as "Endorsements" or to conflict in title with any Invariant Section. d2228 1 a2228 1 You may add a section entitled "Endorsements", provided it contains d2230 3 a2232 3 parties-for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2256 2 a2257 1 combined work in its license notice. d2269 1 a2269 1 In the combination, you must combine any sections entitled d2271 3 a2273 3 entitled "History"; likewise combine any sections entitled "Acknowledgements", and any sections entitled "Dedications." You must delete all sections entitled "Endorsements." d2294 6 a2299 7 a storage or distribution medium, does not as a whole count as a Modified Version of the Document, provided no compilation copyright is claimed for the compilation. Such a compilation is called an "aggregate", and this License does not apply to the other self-contained works thus compiled with the Document, on account of their being thus compiled, if they are not themselves derivative works of the Document. d2302 6 a2307 5 copies of the Document, then if the Document is less than one quarter of the entire aggregate, the Document's Cover Texts may be placed on covers that surround only the Document within the aggregate. Otherwise they must appear on covers around the whole aggregate. d2317 12 a2328 4 translation of this License provided that you also include the original English version of this License. In case of a disagreement between the translation and the original English version of this License, the original English version will prevail. d2333 23 a2355 6 except as expressly provided for under this License. Any other attempt to copy, modify, sublicense or distribute the Document is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance. d2363 1 a2363 1 http://www.gnu.org/copyleft/. d2373 34 a2406 1 Free Software Foundation. d2416 18 a2433 13 Copyright (C) YEAR YOUR NAME. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.1 or any later version published by the Free Software Foundation; with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover Texts being LIST, and with the Back-Cover Texts being LIST. A copy of the license is included in the section entitled "GNU Free Documentation License." If you have no Invariant Sections, write "with no Invariant Sections" instead of saying which ones are invariant. If you have no Front-Cover Texts, write "no Front-Cover Texts" instead of "Front-Cover Texts being LIST"; likewise for Back-Cover Texts. d2443 30 a2472 30 Node: Top719 Node: Introduction2033 Node: Compiling4525 Node: Executing7996 Node: Invoking10784 Node: Output Options12199 Node: Analysis Options19288 Node: Miscellaneous Options22689 Node: Deprecated Options23944 Node: Symspecs26023 Node: Output27849 Node: Flat Profile28889 Node: Call Graph33842 Node: Primary37074 Node: Callers39662 Node: Subroutines41779 Node: Cycles43620 Node: Line-by-line50397 Node: Annotated Source54470 Node: Inaccuracy57469 Node: Sampling Error57727 Node: Assumptions60297 Node: How do I?61767 Node: Incompatibilities63321 Node: Details64815 Node: Implementation65208 Node: File Format71105 Node: Internals75395 Node: Debugging83890 Node: GNU Free Documentation License85491 @ 1.1.1.1 log @Import binutils 2.19.1 as requested by matt@@ OK'ed by board@@ @ text @@ 1.1.1.2 log @from ftp.gnu.org @ text @a2 1 INFO-DIR-SECTION Software development d9 2 a10 2 Copyright (C) 1988, 1992, 1997, 1998, 1999, 2000, 2001, 2003, 2007, 2008, 2009 Free Software Foundation, Inc. d13 1 a13 1 under the terms of the GNU Free Documentation License, Version 1.3 or d30 1 a30 1 This manual is for `gprof' (GNU Binutils) version 2.21.1. d33 2 a34 2 Documentation License version 1.3. A copy of the license is included in the section entitled "GNU Free Documentation License". a164 11 If you are running the program on a system which supports shared libraries you may run into problems with the profiling support code in a shared library being called before that library has been fully initialised. This is usually detected by the program encountering a segmentation fault as soon as it is run. The solution is to link against a static version of the library containing the profiling support code, which for `gcc' users can be done via the `-static' or `-static-libgcc' command line option. For example: gcc -g -pg -static-libgcc myprog.c utils.c -o myprog a533 7 `-SFILENAME' `--external-symbol-table=FILENAME' The `-S' option causes `gprof' to read an external symbol table file, such as `/proc/kallsyms', rather than read the symbol table from the given object file (the default is `a.out'). This is useful for profiling kernel modules. d588 2 a589 1 These options have been replaced with newer versions that use symspecs. d1375 1 a1375 6 vary from run to run if your program is deterministic and single threaded. In multi-threaded applications, or single threaded applications that link with multi-threaded libraries, the counts are only deterministic if the counting function is thread-safe. (Note: beware that the mcount counting function in glibc is _not_ thread-safe). *Note Implementation of Profiling: Implementation. d1937 1 a1937 1 Version 1.3, 3 November 2008 d1939 2 a1940 2 Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. `http://fsf.org/' d1945 1 d1949 6 a1954 6 functional and useful document "free" in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others. d1970 1 d1973 5 a1977 9 This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that work under the conditions stated herein. The "Document", below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "you". You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. d1983 10 a1992 10 A "Secondary Section" is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document's overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them. d1997 1 a1997 4 License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. d2001 1 a2001 3 that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most 25 words. d2005 2 a2006 2 general public, that is suitable for revising the document straightforwardly with generic text editors or (for images d2011 3 a2013 5 otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not "Transparent" is called "Opaque". d2018 6 a2023 7 standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. a2031 19 The "publisher" means any person or entity that distributes copies of the Document to the public. A section "Entitled XYZ" means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specific section name mentioned below, such as "Acknowledgements", "Dedications", "Endorsements", or "History".) To "Preserve the Title" of such a section when you modify the Document means that it remains a section "Entitled XYZ" according to this definition. The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License. d2050 4 a2053 5 If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Document's license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and d2071 5 a2075 4 state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the d2098 51 a2148 69 A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. B. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors, if it has fewer than five), unless they release you from this requirement. C. State on the Title page the name of the publisher of the Modified Version, as the publisher. D. Preserve all the copyright notices of the Document. E. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices. F. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below. G. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document's license notice. H. Include an unaltered copy of this License. I. Preserve the section Entitled "History", Preserve its Title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. J. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. K. For any section Entitled "Acknowledgements" or "Dedications", Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein. L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. M. Delete any section Entitled "Endorsements". Such a section may not be included in the Modified Version. N. Do not retitle any existing section to be Entitled "Endorsements" or to conflict in title with any Invariant Section. O. Preserve any Warranty Disclaimers. d2158 1 a2158 1 You may add a section Entitled "Endorsements", provided it contains d2160 3 a2162 3 parties--for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2186 1 a2186 2 combined work in its license notice, and that you preserve all their Warranty Disclaimers. d2198 1 a2198 1 In the combination, you must combine any sections Entitled d2200 3 a2202 3 Entitled "History"; likewise combine any sections Entitled "Acknowledgements", and any sections Entitled "Dedications". You must delete all sections Entitled "Endorsements." d2223 7 a2229 6 a storage or distribution medium, is called an "aggregate" if the copyright resulting from the compilation is not used to limit the legal rights of the compilation's users beyond what the individual works permit. When the Document is included in an aggregate, this License does not apply to the other works in the aggregate which are not themselves derivative works of the Document. d2232 5 a2236 6 copies of the Document, then if the Document is less than one half of the entire aggregate, the Document's Cover Texts may be placed on covers that bracket the Document within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate. d2246 4 a2249 12 translation of this License, and all the license notices in the Document, and any Warranty Disclaimers, provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer, the original version will prevail. If a section in the Document is Entitled "Acknowledgements", "Dedications", or "History", the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title. d2254 6 a2259 23 except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense, or distribute it is void, and will automatically terminate your rights under this License. However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation. Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice. Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it. d2267 1 a2267 1 `http://www.gnu.org/copyleft/'. d2277 1 a2277 34 Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Document. 11. RELICENSING "Massive Multiauthor Collaboration Site" (or "MMC Site") means any World Wide Web server that publishes copyrightable works and also provides prominent facilities for anybody to edit those works. A public wiki that anybody can edit is an example of such a server. A "Massive Multiauthor Collaboration" (or "MMC") contained in the site means any set of copyrightable works thus published on the MMC site. "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 license published by Creative Commons Corporation, a not-for-profit corporation with a principal place of business in San Francisco, California, as well as future copyleft versions of that license published by that same organization. "Incorporate" means to publish or republish a Document, in whole or in part, as part of another Document. An MMC is "eligible for relicensing" if it is licensed under this License, and if all works that were first published under this License somewhere other than this MMC, and subsequently incorporated in whole or in part into the MMC, (1) had no cover texts or invariant sections, and (2) were thus incorporated prior to November 1, 2008. The operator of an MMC Site may republish an MMC contained in the site under CC-BY-SA on the same site at any time before August 1, 2009, provided the MMC is eligible for relicensing. d2287 13 a2299 18 Copyright (C) YEAR YOUR NAME. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled ``GNU Free Documentation License''. If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace the "with...Texts." line with this: with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover Texts being LIST, and with the Back-Cover Texts being LIST. If you have Invariant Sections without Cover Texts, or some other combination of the three, merge those two alternatives to suit the situation. d2309 30 a2338 30 Node: Top777 Node: Introduction2102 Node: Compiling4594 Node: Executing8650 Node: Invoking11438 Node: Output Options12853 Node: Analysis Options19942 Node: Miscellaneous Options23640 Node: Deprecated Options24895 Node: Symspecs26964 Node: Output28790 Node: Flat Profile29830 Node: Call Graph34783 Node: Primary38015 Node: Callers40603 Node: Subroutines42720 Node: Cycles44561 Node: Line-by-line51338 Node: Annotated Source55411 Node: Inaccuracy58410 Node: Sampling Error58668 Node: Assumptions61572 Node: How do I?63042 Node: Incompatibilities64596 Node: Details66090 Node: Implementation66483 Node: File Format72380 Node: Internals76670 Node: Debugging85165 Node: GNU Free Documentation License86766 @ 1.1.1.3 log @from ftp.gnu.org @ text @d31 1 a31 1 This manual is for `gprof' (GNU Binutils) version 2.23.1. @ 1.1.1.4 log @Import binutils-2.26 @ text @d10 2 a11 1 Copyright (C) 1988-2015 Free Software Foundation, Inc. d31 1 a31 1 This manual is for `gprof' (GNU Binutils) version 2.26. a530 5 `--inline-file-names' This option causes `gprof' to print the source file after each symbol in both the flat profile and the call graph. The full path to the file is printed if used with the `-L' option. d2443 30 a2472 30 Node: Top722 Node: Introduction2045 Node: Compiling4537 Node: Executing8593 Node: Invoking11381 Node: Output Options12796 Node: Analysis Options19885 Node: Miscellaneous Options23803 Node: Deprecated Options25058 Node: Symspecs27127 Node: Output28953 Node: Flat Profile29993 Node: Call Graph34946 Node: Primary38178 Node: Callers40766 Node: Subroutines42883 Node: Cycles44724 Node: Line-by-line51501 Node: Annotated Source55574 Node: Inaccuracy58573 Node: Sampling Error58831 Node: Assumptions61735 Node: How do I?63205 Node: Incompatibilities64759 Node: Details66253 Node: Implementation66646 Node: File Format72543 Node: Internals76833 Node: Debugging85328 Node: GNU Free Documentation License86929 @ 1.1.1.5 log @Import binutils-2.27 minus the testsuites @ text @d10 1 a10 1 Copyright (C) 1988-2016 Free Software Foundation, Inc. d30 1 a30 1 This manual is for `gprof' (GNU Binutils) version 2.27. @ 1.1.1.6 log @2018-01-10 Nick Clifton * config-ml.in: Sync with gcc sources. * config.guess: Likewise. * config.sub: Likewise. * configure.ac: Likewise. * configure: Regenerate. 2017-12-14 Nick Clifton * COPYING.LIBGLOSS: Update address of FSF in copyright notice. 2017-12-12 Stafford Horne * configure.ac: Remove logic adding gdb to noconfigsdirs for or1k. * configure: Regenerate. 2017-09-15 Nick Clifton * src-release.sh (LZIPPROG): New define. Provides the name of the lzip program. (do_lz): New function. Compresses a tarball using the lzip program. (do_compress): Add support for lzip compression. (usage): Mention -l option. (build_release): Support -l option to invoke lzip compression. 2017-09-15 Alan Modra * src-release.sh (do_proto_toplev): Revert last patch. Enable or disable binutils, gas, gdb, gold, gprof, ld, libdecnumber, readline, and sim depending on $tool and $support_files. Echo configure line. 2017-09-14 Matt Rice * src-release.sh (do_proto_toplev): Enable gold during release process. 2017-04-13 Andrew Jenner * config.sub: Sync with master version in config project. 2017-04-03 Andrew Jenner * config.sub: Handle ia16 in $basic_machine. bfd/ * config.bfd: Handle ia16. gas/ * configure.tgt: Handle ia16. ld/ * configure.tgt: Handle ia16. 2017-03-22 Nick Clifton * config.sub: Sync with master version in config project. * config.guess: Likewise. 2017-01-23 Nick Clifton * configure.ac: Update year in copyright notice. Sync from FSF GCC mainline, bringing in the following patches. * Makefile.def: Likewise. * Makefile.tpl: Likewise. * configure: Regenerate. * Makefile.in: Regenerate. 2016-12-21 Jakub Jelinek * configure.ac: Don't bootstrap libmpx unless --with-build-config includes bootstrap-mpx. 2016-12-01 Matthias Klose * configure.ac: Don't use pkg-config to check for bdw-gc. 2016-11-30 Matthias Klose * Makefile.def: Remove reference to boehm-gc target module. * configure.ac: Include pkg.m4, check for --with-target-bdw-gc options and for the bdw-gc pkg-config module. 2016-11-15 Matthias Klose * config-ml.in: Remove references to GCJ. * configure.ac: Likewise. 2016-09-30 Jakub Jelinek * configure.ac: Add target-libffi to target_libraries. Readd libgcj target disablings, modified to only target-libffi. Readd target addition of go to unsupported languages. 2016-09-30 Andrew Haley * Makefile.def: Remove libjava. * Makefile.tpl: Likewise. * configure.ac: Likewise. 2016-09-26 Anton Kolesov * configure.ac: Disable "sim" directory for arc*-*-*. 2016-09-12 Maciej W. Rozycki * configure.ac: Check for the minimum in-tree MPFR version handled. 2016-12-31 Alan Modra * config.sub: Import from upstream. 2016-12-08 Alan Modra * configure: Regenerate. 2016-12-02 Josh Conner * configure.ac: Add fuchsia to targets that use ELF. * configure: Regenerated. 2016-11-07 Doug Evans * config.sub: Sync with upstream version 2016-11-03. git://git.sv.gnu.org/config.git * config.guess: Sync with upstream version 2016-10-02. 2016-09-27 Simon Marchi * .gitignore: Add archives and make stamps. 2016-07-20 Yan-Ting Lin * configure.ac (nds32*-*-*): Remove entry to enable gdb. * configure: Regenerated. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.3 from gprof.texi. d3 6 a8 1 This file documents the gprof profiler of the GNU system. d10 1 a10 1 Copyright (C) 1988-2018 Free Software Foundation, Inc. a18 5 INFO-DIR-SECTION Software development START-INFO-DIR-ENTRY * gprof: (gprof). Profiling your program's execution END-INFO-DIR-ENTRY d25 1 a25 1 This manual describes the GNU profiler, 'gprof', and how you can use it d28 1 a28 1 programs. GNU 'gprof' was written by Jay Fenlason. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.0. d33 2 a34 2 Documentation License version 1.3. A copy of the license is included in the section entitled "GNU Free Documentation License". d42 1 a42 1 * Invoking:: How to run 'gprof', and its options d44 1 a44 1 * Output:: Interpreting 'gprof''s output d48 1 a48 1 * Incompatibilities:: (between GNU 'gprof' and Unix 'gprof'.) d67 6 a72 5 execution of your program, it can be used on programs that are too large or too complex to analyze by reading the source. However, how your program is run will affect the information that shows up in the profile data. If you don't use some feature of your program while it is being profiled, no profile information will be generated for that feature. d82 1 a82 1 * You must run 'gprof' to analyze the profile data. *Note 'gprof' d94 6 a99 5 The "call graph" shows, for each function, which functions called it, which other functions it called, and how many times. There is also an estimate of how much time was spent in the subroutines of each function. This can suggest places where you might try to eliminate function calls that use a lot of time. *Note The Call Graph: Call Graph. d118 1 a118 1 To compile a source file for profiling, specify the '-pg' option when d122 3 a124 3 To link the program for profiling, if you use a compiler such as 'cc' to do the linking, simply specify '-pg' in addition to your usual options. The same option, '-pg', alters either compilation or linking d130 1 a130 1 The '-pg' option also works with a command that both compiles and d135 3 a137 3 Note: The '-pg' option must be part of your compilation options as well as your link options. If it is not then no call-graph data will be gathered and when you run 'gprof' you will get an error message like d142 2 a143 2 If you add the '-Q' switch to suppress the printing of the call graph data you will still be able to see the time samples: d154 8 a161 8 If you run the linker 'ld' directly instead of through a compiler such as 'cc', you may have to specify a profiling startup file 'gcrt0.o' as the first input file instead of the usual startup file 'crt0.o'. In addition, you would probably want to specify the profiling C library, 'libc_p.a', by writing '-lc_p' instead of the usual '-lc'. This is not absolutely necessary, but doing this gives you number-of-calls information for standard library functions such as 'read' and 'open'. For example: d166 2 a167 2 libraries you may run into problems with the profiling support code in a shared library being called before that library has been fully d170 3 a172 3 against a static version of the library containing the profiling support code, which for 'gcc' users can be done via the '-static' or '-static-libgcc' command line option. For example: d176 1 a176 1 If you compile only some of the modules of the program with '-pg', d178 6 a183 6 information about the modules that were compiled without '-pg'. The only information you get for the functions in those modules is the total time spent in them; there is no record of how many times they were called, or from where. This will not affect the flat profile (except that the 'calls' field for the functions will be blank), but will greatly reduce the usefulness of the call graph. d186 1 a186 1 'gcov' tool instead of 'gprof'. See that tool's manual or info pages d189 3 a191 3 Note, older versions of 'gcc' produce line-by-line profiling information that works with 'gprof' rather than 'gcov' so there is still support for displaying this kind of information in 'gprof'. *Note d194 5 a198 5 It also worth noting that 'gcc' implements a '-finstrument-functions' command line option which will insert calls to special user supplied instrumentation routines at the entry and exit of every function in their program. This can be used to implement an alternative profiling scheme. d207 1 a207 1 generate the information that 'gprof' needs. Simply run the program as d221 2 a222 2 'gmon.out' just before exiting. If there is already a file called 'gmon.out', its contents are overwritten. There is currently no way to d227 3 a229 3 In order to write the 'gmon.out' file properly, your program must exit normally: by returning from 'main' or by calling 'exit'. Calling the low-level function '_exit' does not write the profile data, and d232 1 a232 1 The 'gmon.out' file is written in the program's _current working d234 2 a235 2 'chdir', the 'gmon.out' file will be left in the last directory your program 'chdir''d to. If you don't have permission to write in this d239 1 a239 1 called 'bb.out'. This file, if present, contains an human-readable d241 4 a244 4 appearance of a human-readable 'bb.out' means the basic-block counts didn't get written into 'gmon.out'. The Perl script 'bbconv.pl', included with the 'gprof' source distribution, will convert a 'bb.out' file into a format readable by 'gprof'. Invoke it like this: d248 2 a249 2 This translates the information in 'bb.out' into a form that 'gprof' can understand. But you still need to tell 'gprof' about the existence d251 1 a251 1 'gprof' command line, _along with 'gmon.out'_, like this: d258 1 a258 1 4 'gprof' Command Summary d261 2 a262 2 After you have a profile data file 'gmon.out', you can run 'gprof' to interpret the information in it. The 'gprof' program prints a flat d264 1 a264 1 redirect the output of 'gprof' into a file with '>'. d266 1 a266 1 You run 'gprof' like this: d272 5 a276 4 If you omit the executable file name, the file 'a.out' is used. If you give no profile data file name, the file 'gmon.out' is used. If any file is not in the proper format, or if the profile data file does not appear to belong to the executable file, an error message is printed. d286 2 a287 2 * Output Options:: Controlling 'gprof''s output style * Analysis Options:: Controlling how 'gprof' analyzes its data d299 1 a299 1 These options specify which of several output formats 'gprof' should d302 4 a305 4 Many of these options take an optional "symspec" to specify functions to be included or excluded. These options can be specified multiple times, with different symspecs, to include or exclude sets of symbols. *Note Symspecs: Symspecs. d307 1 a307 1 Specifying any of these options overrides the default ('-p -q'), d310 3 a312 3 '-A[SYMSPEC]' '--annotated-source[=SYMSPEC]' The '-A' option causes 'gprof' to print annotated source code. If d316 3 a318 3 '-b' '--brief' If the '-b' option is given, 'gprof' doesn't print the verbose d323 3 a325 3 '-C[SYMSPEC]' '--exec-counts[=SYMSPEC]' The '-C' option causes 'gprof' to print a tally of functions and d330 1 a330 1 specifying the '-l' option, along with '-C', will cause basic-block d333 9 a341 9 '-i' '--file-info' The '-i' option causes 'gprof' to display summary information about the profile data file(s) and then exit. The number of histogram, call graph, and basic-block count records is displayed. '-I DIRS' '--directory-path=DIRS' The '-I' option specifies a list of search directories in which to d346 4 a349 4 '-J[SYMSPEC]' '--no-annotated-source[=SYMSPEC]' The '-J' option causes 'gprof' not to print annotated source code. If SYMSPEC is specified, 'gprof' prints annotated source, but d352 2 a353 2 '-L' '--print-path' d355 1 a355 1 suppressed. The '-L' option causes 'gprof' to print the full d360 10 a369 10 '-p[SYMSPEC]' '--flat-profile[=SYMSPEC]' The '-p' option causes 'gprof' to print a flat profile. If SYMSPEC is specified, print flat profile only for matching symbols. *Note The Flat Profile: Flat Profile. '-P[SYMSPEC]' '--no-flat-profile[=SYMSPEC]' The '-P' option causes 'gprof' to suppress printing a flat profile. If SYMSPEC is specified, 'gprof' prints a flat profile, but d372 3 a374 3 '-q[SYMSPEC]' '--graph[=SYMSPEC]' The '-q' option causes 'gprof' to print the call graph analysis. d378 4 a381 4 '-Q[SYMSPEC]' '--no-graph[=SYMSPEC]' The '-Q' option causes 'gprof' to suppress printing the call graph. If SYMSPEC is specified, 'gprof' prints a call graph, but excludes d384 3 a386 3 '-t' '--table-length=NUM' The '-t' option causes the NUM most active source lines in each d390 2 a391 2 '-y' '--separate-files' d393 1 a393 1 'gprof' prints annotated source files to standard-output. If this d395 9 a403 9 'path/FILENAME' is generated in the file 'FILENAME-ann'. If the underlying file system would truncate 'FILENAME-ann' so that it overwrites the original 'FILENAME', 'gprof' generates annotated source in the file 'FILENAME.ann' instead (if the original file name has an extension, that extension is _replaced_ with '.ann'). '-Z[SYMSPEC]' '--no-exec-counts[=SYMSPEC]' The '-Z' option causes 'gprof' not to print a tally of functions d407 3 a409 3 '-r' '--function-ordering' The '--function-ordering' option causes 'gprof' to print a d419 7 a425 7 '-R MAP_FILE' '--file-ordering MAP_FILE' The '--file-ordering' option causes 'gprof' to print a suggested .o link line ordering for the program based on profiling data. This option suggests an ordering which may improve paging, tlb and cache behavior for the program on systems which do not support arbitrary ordering of functions in an executable. d427 1 a427 1 Use of the '-a' argument is highly recommended with this option. d431 1 a431 1 similar to the output of the program 'nm'. d442 1 a442 2 To create a MAP_FILE with GNU 'nm', type a command like 'nm d445 4 a448 4 '-T' '--traditional' The '-T' option causes 'gprof' to print its output in "traditional" BSD style. d450 2 a451 2 '-w WIDTH' '--width=WIDTH' d455 7 a461 7 '-x' '--all-lines' This option affects annotated source output only. By default, only the lines at the beginning of a basic-block are annotated. If this option is specified, every line in a basic-block is annotated by repeating the annotation for the first line. This behavior is similar to 'tcov''s '-a'. d463 2 a464 2 '--demangle[=STYLE]' '--no-demangle' d467 1 a467 1 '--no-demangle' option may be used to turn off demangling. d478 3 a480 3 '-a' '--no-static' The '-a' option causes 'gprof' to suppress the printing of d486 2 a487 2 the executable file. This option affects both the flat profile and the call graph. d489 3 a491 3 '-c' '--static-call-graph' The '-c' option causes the call graph of the program to be d496 12 a507 12 have been called, but never were. Calls to functions that were not compiled with profiling enabled are also identified, but only if symbol table entries are present for them. Calls to dynamic library routines are typically _not_ found by this option. Parents or children identified via this heuristic are indicated in the call graph with call counts of '0'. '-D' '--ignore-non-functions' The '-D' option causes 'gprof' to ignore symbols which are not known to be functions. This option will give more accurate profile data on systems where it is supported (Solaris and HPUX for d510 2 a511 2 '-k FROM/TO' The '-k' option allows you to delete from the call graph any arcs d514 3 a516 3 '-l' '--line' The '-l' option enables line-by-line profiling, which causes d519 2 a520 2 compiled by older versions of the 'gcc' compiler. Newer versions of 'gcc' are designed to work with the 'gcov' tool instead. d522 5 a526 5 If the program was compiled with basic-block counting enabled, this option will also identify how many times each line of code was executed. While line-by-line profiling can help isolate where in a large function a program is spending its time, it also significantly increases the running time of 'gprof', and magnifies d530 4 a533 4 '--inline-file-names' This option causes 'gprof' to print the source file after each symbol in both the flat profile and the call graph. The full path to the file is printed if used with the '-L' option. d535 2 a536 2 '-m NUM' '--min-count=NUM' d540 8 a547 3 '-nSYMSPEC' '--time=SYMSPEC' The '-n' option causes 'gprof', in its call graph analysis, to only d550 13 a562 4 '-NSYMSPEC' '--no-time=SYMSPEC' The '-n' option causes 'gprof', in its call graph analysis, not to propagate times for symbols matching SYMSPEC. a563 13 '-SFILENAME' '--external-symbol-table=FILENAME' The '-S' option causes 'gprof' to read an external symbol table file, such as '/proc/kallsyms', rather than read the symbol table from the given object file (the default is 'a.out'). This is useful for profiling kernel modules. '-z' '--display-unused-functions' If you give the '-z' option, 'gprof' will mention all functions in the flat profile, even those that were never called, and that had no time spent in them. This is useful in conjunction with the '-c' option for discovering which routines were never called. d571 4 a574 4 '-d[NUM]' '--debug[=NUM]' The '-d NUM' option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging 'gprof': d577 3 a579 3 '-h' '--help' The '-h' option prints command line usage. d581 2 a582 2 '-ONAME' '--file-format=NAME' d584 2 a585 2 are 'auto' (the default), 'bsd', '4.4bsd', 'magic', and 'prof' (not yet supported). d587 3 a589 3 '-s' '--sum' The '-s' option causes 'gprof' to summarize the information in the d591 11 a601 11 called 'gmon.sum', which contains all the information from the profile data files that 'gprof' read in. The file 'gmon.sum' may be one of the specified input files; the effect of this is to merge the data in the other input files into 'gmon.sum'. Eventually you can run 'gprof' again without '-s' to analyze the cumulative data in the file 'gmon.sum'. '-v' '--version' The '-v' flag causes 'gprof' to print the current version number, d604 1 d613 2 a614 2 '-e FUNCTION_NAME' The '-e FUNCTION' option tells 'gprof' to not print information d617 11 a627 10 functions that call it, but its index number will be shown as '[not printed]'. More than one '-e' option may be given; only one FUNCTION_NAME may be indicated with each '-e' option. '-E FUNCTION_NAME' The '-E FUNCTION' option works like the '-e' option, but time spent in the function (and children who were not called from anywhere else), will not be used to compute the percentages-of-time for the call graph. More than one '-E' option may be given; only one FUNCTION_NAME may be indicated with each '-E' option. d629 2 a630 2 '-f FUNCTION_NAME' The '-f FUNCTION' option causes 'gprof' to limit the call graph to d632 10 a641 2 children...). More than one '-f' option may be given; only one FUNCTION_NAME may be indicated with each '-f' option. a642 7 '-F FUNCTION_NAME' The '-F FUNCTION' option works like the '-f' option, but only time spent in the function and its children (and their children...) will be used to determine total-time and percentages-of-time for the call graph. More than one '-F' option may be given; only one FUNCTION_NAME may be indicated with each '-F' option. The '-F' option overrides the '-E' option. d644 2 a645 2 Note that only one function can be specified with each '-e', '-E', '-f' or '-F' option. To specify more than one function, use multiple d651 1 a651 1 'foo' or 'bar' and were not reachable from 'boring'. d670 6 a675 6 'main.c' Selects everything in file 'main.c'--the dot in the string tells 'gprof' to interpret the string as a filename, rather than as a function name. To select a file whose name does not contain a dot, a trailing colon should be specified. For example, 'odd:' is interpreted as the file named 'odd'. d677 2 a678 2 'main' Selects all functions named 'main'. d687 2 a688 2 necessary to add a leading colon to the name. For example, ':.mul' selects function '.mul'. d691 2 a692 2 'gprof' will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as 'gprof' prints d694 2 a695 2 '_main' from your 'main' function, 'gprof' still prints it as 'main' in its output, so you should use 'main' in symspecs. d697 2 a698 2 'main.c:main' Selects function 'main' in file 'main.c'. d700 2 a701 2 'main.c:134' Selects line 134 in file 'main.c'. d706 1 a706 1 5 Interpreting 'gprof''s Output d709 1 a709 1 'gprof' can produce several different output styles, the most important d722 1 a722 1 * Line-by-line:: 'gprof' can analyze individual source code lines d733 1 a733 1 executing each function. Unless the '-z' option is given, functions d764 1 a764 1 functions 'mcount' and 'profil' are part of the profiling apparatus and d774 8 a781 8 'cumulative seconds' field. Since each sample counted for 0.01 seconds, this means only six samples were taken during the run. Two of the samples occurred while the program was in the 'open' function, as indicated by the 'self seconds' field. Each of the other four samples occurred one each in 'offtime', 'memccpy', 'write', and 'mcount'. Since only six samples were taken, none of these values can be regarded as particularly reliable. In another run, the 'self seconds' field for 'mcount' might well be '0.00' or '0.02'. *Note Statistical Sampling d784 2 a785 2 The remaining functions in the listing (those whose 'self seconds' field is '0.00') didn't appear in the histogram samples at all. d787 1 a787 1 they are listed, sorted in decreasing order by the 'calls' field. d794 1 a794 1 '% time' d798 1 a798 1 'cumulative seconds' d803 1 a803 1 'self seconds' d807 1 a807 1 'calls' d813 1 a813 1 'self ms/call' d818 1 a818 1 'total ms/call' d825 2 a826 2 'name' This is the name of the function. The flat profile is sorted by d838 2 a839 2 they themselves may not have used much time, called other functions that did use unusual amounts of time. d842 1 a842 1 same 'gprof' run as the flat profile example in the previous section. d877 4 a880 4 number in square brackets. The end of this line says which function the entry is for. The preceding lines in the entry describe the callers of this function and the following lines describe its subroutines (also called "children" when we speak of the call graph). d885 1 a885 1 The internal profiling function 'mcount' (*note The Flat Profile: d894 1 a894 1 such as 'a' calls 'b' calls 'a'... d907 1 a907 1 'report' in our main example, together with the heading line that shows d916 1 a916 1 'index' d926 1 a926 1 '% time' d935 1 a935 1 'self' d937 1 a937 1 should be identical to the number printed in the 'seconds' field d940 5 a944 5 'children' This is the total amount of time spent in the subroutine calls made by this function. This should be equal to the sum of all the 'self' and 'children' entries of the children listed directly below this function. d946 1 a946 1 'called' d950 1 a950 1 separated by a '+'. The first number counts non-recursive calls, d953 2 a954 2 In the example above, the function 'report' was called once from 'main'. d956 1 a956 1 'name' d963 1 a963 1 example, if function 'gnurr' is part of cycle number one, and has d974 3 a976 3 A function's entry has a line for each function it was called by. These lines' fields correspond to the fields of the primary line, but their meanings are different because of the difference in context. d979 1 a979 1 'report', the primary line and one caller-line preceding it, together d987 2 a988 2 Here are the meanings of the fields in the caller-line for 'report' called from 'main': d990 14 a1003 14 'self' An estimate of the amount of time spent in 'report' itself when it was called from 'main'. 'children' An estimate of the amount of time spent in subroutines of 'report' when 'report' was called from 'main'. The sum of the 'self' and 'children' fields is an estimate of the amount of time spent within calls to 'report' from 'main'. 'called' Two numbers: the number of times 'report' was called from 'main', followed by the total number of non-recursive calls to 'report' d1006 2 a1007 2 'name and index number' The name of the caller of 'report' to which this line applies, d1011 1 a1011 1 'gprof' request the omission of certain functions. When a caller d1019 1 a1019 1 dummy caller-line is printed which has '' as the "caller's d1034 1 a1034 1 'main', the primary line and a line for a subroutine, together with the d1042 2 a1043 2 Here are the meanings of the fields in the subroutine-line for 'main' calling 'report': d1045 17 a1061 17 'self' An estimate of the amount of time spent directly within 'report' when 'report' was called from 'main'. 'children' An estimate of the amount of time spent in subroutines of 'report' when 'report' was called from 'main'. The sum of the 'self' and 'children' fields is an estimate of the total time spent in calls to 'report' from 'main'. 'called' Two numbers, the number of calls to 'report' from 'main' followed by the total number of non-recursive calls to 'report'. This ratio is used to determine how much of 'report''s 'self' and 'children' time gets credited to 'main'. *Note Estimating 'children' Times: Assumptions. d1063 2 a1064 2 'name' The name of the subroutine of 'main' to which this line applies, d1076 2 a1077 2 The graph may be complicated by the presence of "cycles of recursion" in the call graph. A cycle exists if a function calls another function d1079 2 a1080 2 function. For example: if 'a' calls 'b', and 'b' calls 'a', then 'a' and 'b' form a cycle. d1083 4 a1086 4 they belong to the same cycle. If 'a' and 'b' call each other and 'b' and 'c' call each other, all three make one cycle. Note that even if 'b' only calls 'a' if it was not called from 'a', 'gprof' cannot determine this, so 'a' and 'b' are still considered a cycle. d1090 1 a1090 1 graph it is followed by ''. d1093 4 a1096 4 call graph paradoxical. The "time spent in children" of 'a' should include the time spent in its subroutine 'b' and in 'b''s subroutines--but one of 'b''s subroutines is 'a'! How much of 'a''s time should be included in the children of 'a', when 'a' is indirectly d1099 1 a1099 1 The way 'gprof' resolves this paradox is by creating a single entry d1104 2 a1105 2 "callers" of the cycle are the functions, outside the cycle, that called functions in the cycle. d1108 2 a1109 2 containing functions 'a' and 'b'. The cycle was entered by a call to 'a' from 'main'; both 'a' and 'b' called 'c'. d1131 2 a1132 2 for 'main', which calls 'a', and an entry for 'c', with callers 'a' and 'b'.) d1165 2 a1166 2 The 'self' field of the cycle's primary line is the total time spent in all the functions of the cycle. It equals the sum of the 'self' d1170 6 a1175 6 The 'children' fields of the cycle's primary line and subroutine lines count only subroutines outside the cycle. Even though 'a' calls 'b', the time spent in those calls to 'b' is not counted in 'a''s 'children' time. Thus, we do not encounter the problem of what to do when the time in those calls to 'b' includes indirect recursive calls back to 'a'. d1177 1 a1177 1 The 'children' field of a caller-line in the cycle's entry estimates d1182 1 a1182 1 The 'called' field in the primary line for the cycle has two numbers: d1185 8 a1192 8 called by functions in the cycle (including times when a function in the cycle calls itself). This is a generalization of the usual split into non-recursive and recursive calls. The 'called' field of a subroutine-line for a cycle member in the cycle's entry says how many time that function was called from functions in the cycle. The total of all these is the second number in the primary line's 'called' field. d1196 4 a1199 4 These lines show how many times each function in the cycle called or was called from each other function in the cycle. The 'self' and 'children' fields in these lines are blank because of the difficulty of defining meanings for them when recursion is going on. d1207 1 a1207 1 'gprof''s '-l' option causes the program to perform "line-by-line" d1209 3 a1211 3 functions, but to individual lines of source code. This only works with programs compiled with older versions of the 'gcc' compiler. Newer versions of 'gcc' use a different program - 'gcov' - to display d1214 2 a1215 2 With the older versions of 'gcc' the program usually has to be compiled with a '-g' option, in addition to '-pg', in order to generate d1217 1 a1217 1 versions of 'gcc' the program had to be compiled with the '-a' command d1222 1 a1222 1 version of 'gprof' does not propagate call graph arcs from source code d1226 3 a1228 3 Here is a section of 'gprof''s output, without line-by-line profiling. Note that 'ct_init' accounted for four histogram hits, and 13327 calls to 'init_block'. d1251 2 a1252 3 Now let's look at some of 'gprof''s output from the same program run, this time with line-by-line profiling enabled. Note that 'ct_init''s d1255 1 a1255 1 note how 'ct_init''s 13327 calls to 'init_block' are broken down into a1290 1 d1297 1 a1297 1 'gprof''s '-A' option triggers an annotated source listing, which lists d1299 2 a1300 2 times it was called. You may also need to specify the '-I' option, if 'gprof' can't find the source code files. d1302 1 a1302 1 With older versions of 'gcc' compiling with 'gcc ... -g -pg -a' d1304 3 a1306 3 function counting code. This enables 'gprof' to determine how many times each line of code was executed. With newer versions of 'gcc' support for displaying basic-block counts is provided by the 'gcov' d1332 6 a1337 7 'updcrc' has at least five basic-blocks. One is the function itself. The 'if' statement on line 9 generates two more basic-blocks, one for each branch of the 'if'. A fourth basic-block results from the 'if' on line 13, and the contents of the 'do' loop form the fifth basic-block. The compiler may also generate additional basic-blocks to handle various special cases. d1340 3 a1342 3 'gprof -l -A'. The '-x' option is also helpful, to ensure that each line of code is labeled at least once. Here is 'updcrc''s annotated source listing for a sample 'gzip' run: d1365 2 a1366 2 each branch of the 'if' statement. The body of the 'do' loop was executed a total of 26312 times. Note how the 'while' statement is d1374 1 a1374 1 6 Inaccuracy of 'gprof' Output d1388 1 a1388 1 The run-time figures that 'gprof' gives you are based on a sampling d1401 2 a1402 2 beware that the mcount counting function in glibc is _not_ thread-safe). *Note Implementation of Profiling: Implementation. d1411 1 a1411 1 period is 0.01 seconds and 'foo''s run-time is 1 second, N is 100 d1413 1 a1413 1 error in 'foo''s run-time is 0.1 seconds (10*0.01 seconds), or ten d1415 5 a1419 5 seconds and 'bar''s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in 'bar''s run-time is 1 second, or one percent of the observed value. It is likely to vary this much _on the average_ from one profiling run to the next. (_Sometimes_ it will vary more.) d1429 1 a1429 1 the data from several runs, using the '-s' option of 'gprof'. Here is d1434 1 a1434 1 2. Issue the command 'mv gmon.out gmon.sum'. d1438 1 a1438 1 4. Merge the new data in 'gmon.out' into 'gmon.sum' with this command: d1451 1 a1451 1 6.2 Estimating 'children' Times d1455 2 a1456 2 'children' time values and all the time figures in caller and subroutine lines. d1459 1 a1459 1 profile data itself. Instead, 'gprof' estimates them by making an d1463 4 a1466 3 any function 'foo' is not correlated with who called 'foo'. If 'foo' used 5 seconds in all, and 2/5 of the calls to 'foo' came from 'a', then 'foo' contributes 2 seconds to 'a''s 'children' time, by assumption. d1469 6 a1474 6 far from true. Suppose that 'foo' returns very quickly when its argument is zero; suppose that 'a' always passes zero as an argument, while other callers of 'foo' pass other arguments. In this program, all the time spent in 'foo' is in the calls from callers other than 'a'. But 'gprof' has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in 'foo' to the children of 'a'. d1476 1 a1476 1 We hope some day to put more complete data into 'gmon.out', so that a1486 1 d1488 1 a1488 1 Because 'gprof' can only report call times and counts by function, d1492 1 a1492 1 artificial hot spots since compiling with '-pg' adds a significant d1497 1 a1497 2 Use the 'gcov' program. d1500 1 a1500 2 Use 'gprof -l' and lookup the function in the call graph. The a1503 1 d1519 1 d1523 1 a1523 1 8 Incompatibilities with Unix 'gprof' d1526 3 a1528 3 GNU 'gprof' and Berkeley Unix 'gprof' use the same data file 'gmon.out', and provide essentially the same information. But there are a few differences. d1530 1 a1530 1 * GNU 'gprof' uses a new, generalized file format with support for d1532 1 a1532 1 cookie and version number allows 'gprof' to easily identify new d1536 3 a1538 3 * For a recursive function, Unix 'gprof' lists the function as a parent and as a child, with a 'calls' field that lists the number of recursive calls. GNU 'gprof' omits these lines and puts the d1541 2 a1542 2 * When a function is suppressed from the call graph with '-e', GNU 'gprof' still lists it as a subroutine of functions that call it. d1544 2 a1545 2 * GNU 'gprof' accepts the '-k' with its argument in the form 'from/to', instead of 'from to'. d1547 3 a1549 3 * In the annotated source listing, if there are multiple basic blocks on the same line, GNU 'gprof' prints all of their counts, separated by commas. d1552 1 a1552 1 'gprof' prints blurbs after the tables, so that you can see the d1564 3 a1566 3 * File Format:: Format of 'gmon.out' files * Internals:: 'gprof''s internal operation * Debugging:: Using 'gprof''s '-d' option d1579 2 a1580 2 the '-pg' option, which causes every function to call 'mcount' (or '_mcount', or '__mcount', depending on the OS and compiler) as one of d1583 1 a1583 1 The 'mcount' routine, included in the profiling library, is d1586 3 a1588 3 done by examining the stack frame to find both the address of the child, and the return address in the original parent. Since this is a very machine-dependent operation, 'mcount' itself is typically a short d1590 4 a1593 4 and then calls '__mcount_internal' (a normal C function) with two arguments--'frompc' and 'selfpc'. '__mcount_internal' is responsible for maintaining the in-memory call graph, which records 'frompc', 'selfpc', and the number of times each of these call arcs was traversed. d1596 1 a1596 1 ('__builtin_return_address'), which allows a generic 'mcount' function d1600 1 a1600 1 'mcount' is used for performance reasons. d1604 2 a1605 2 same as in the usual C library, but they were compiled with '-pg'. If you link your program with 'gcc ... -pg', it automatically uses the d1608 5 a1612 5 Profiling also involves watching your program as it runs, and keeping a histogram of where the program counter happens to be every now and then. Typically the program counter is looked at around 100 times per second of run time, but the exact frequency may vary from system to system. d1615 1 a1615 1 provide a 'profil()' system call, which registers a memory array with d1617 7 a1623 7 address space maps into the array. Typical scaling values cause every 2 to 8 bytes of address space to map into a single array slot. On every tick of the system clock (assuming the profiled program is running), the value of the program counter is examined and the corresponding slot in the memory array is incremented. Since this is done in the kernel, which had to interrupt the process anyway to handle the clock interrupt, very little additional system overhead is required. d1626 9 a1634 8 earlier), do not provide a 'profil()' system call. On such a system, arrangements are made for the kernel to periodically deliver a signal to the process (typically via 'setitimer()'), which then performs the same operation of examining the program counter and incrementing a slot in the memory array. Since this method requires a signal to be delivered to user space every time a sample is taken, it uses considerably more overhead than kernel-based profiling. Also, due to the added delay required to deliver the signal, this method is less accurate as well. d1637 8 a1644 8 either calls 'profil()' or sets up a clock signal handler. This routine ('monstartup') can be invoked in several ways. On Linux systems, a special profiling startup file 'gcrt0.o', which invokes 'monstartup' before 'main', is used instead of the default 'crt0.o'. Use of this special startup file is one of the effects of using 'gcc ... -pg' to link. On SPARC systems, no special startup files are used. Rather, the 'mcount' routine, when it is invoked for the first time (typically when 'main' is called), calls 'monstartup'. d1646 1 a1646 1 If the compiler's '-a' option was used, basic-block counting is also d1649 1 a1649 1 basic-block begins (i.e., when an 'if' statement appears), an extra d1656 5 a1660 5 The profiling library also includes a function ('mcleanup') which is typically registered using 'atexit()' to be called as the program exits, and is responsible for writing the file 'gmon.out'. Profiling is turned off, various headers are output, and the histogram is written, followed by the call-graph arcs and the basic-block counts. d1662 1 a1662 1 The output from 'gprof' gives no indication of parts of your program d1664 8 a1671 8 of the program counter are taken at fixed intervals of the program's run time. Therefore, the time measurements in 'gprof' output say nothing about time that your program was not running. For example, a part of the program that creates so much data that it cannot all fit in physical memory at once may run very slowly due to thrashing, but 'gprof' will say it uses little time. On the other hand, sampling by run time has the advantage that the amount of load due to other users won't directly affect the output you get. d1681 2 a1682 2 'gprof' file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU 'gprof' d1684 1 a1684 1 compatibility, GNU 'gprof' continues to support the old BSD-derived d1689 1 a1689 1 The new file format is defined in header file 'gmon_out.h'. It d1693 1 a1693 1 the profile was collected. GNU 'gprof' adapts automatically to the d1700 1 a1700 1 reading a file, GNU 'gprof' will ensure records of the same type are d1710 15 a1724 15 spans, the size of the histogram in bytes (unlike in the old BSD format, this does not include the size of the header), the rate of the profiling clock, and the physical dimension that the bin counts represent after being scaled by the profiling clock rate. The physical dimension is specified in two parts: a long name of up to 15 characters and a single character abbreviation. For example, a histogram representing real-time would specify the long name as "seconds" and the abbreviation as "s". This feature is useful for architectures that support performance monitor hardware (which, fortunately, is becoming increasingly common). For example, under DEC OSF/1, the "uprofile" command can be used to produce a histogram of, say, instruction cache misses. In this case, the dimension in the histogram header could be set to "i-cache misses" and the abbreviation could be set to "1" (because it is simply a count, not a physical dimension). Also, the profiling rate would have to be set to 1 in this case. d1737 8 a1744 8 program execution. Arcs are specified by a pair of addresses: the first must be within caller's function and the second must be within the callee's function. When performing profiling at the function level, these addresses can point anywhere within the respective function. However, when profiling at the line-level, it is better if the addresses are as close to the call-site/entry-point as possible. This will ensure that the line-level call-graph is able to identify exactly which line of source code performed calls to a function. d1750 5 a1754 4 sequence of address/count pairs. The header simply specifies the length of the sequence. In an address/count pair, the address identifies a basic-block and the count specifies the number of times that basic-block was executed. Any address within the basic-address can be used. d1759 1 a1759 1 9.3 'gprof''s Internal Operation d1762 6 a1767 6 Like most programs, 'gprof' begins by processing its options. During this stage, it may building its symspec list ('sym_ids.c:sym_id_add'), if options are specified which use symspecs. 'gprof' maintains a single linked list of symspecs, which will eventually get turned into 12 symbol tables, organized into six include/exclude pairs--one pair each for the flat profile (INCL_FLAT/EXCL_FLAT), the call graph arcs d1774 2 a1775 2 After option processing, 'gprof' finishes building the symspec list by adding all the symspecs in 'default_excluded_list' to the exclude d1781 2 a1782 2 it is an object file, and read its symbol table ('core.c:core_init'), using 'bfd_canonicalize_symtab' after mallocing an appropriately sized d1784 2 a1785 2 '--file-ordering' option has been specified), and the core text space is read into memory (if the '-c' option was given). d1787 1 a1787 1 'gprof''s own symbol table, an array of Sym structures, is now built. d1789 1 a1789 1 whether line-by-line profiling ('-l' option) has been enabled. For d1795 4 a1798 4 the symbols. In between the two passes, a single array of type 'Sym' is created of the appropriate length. Finally, 'symtab.c:symtab_finalize' is called to sort the symbol table and remove duplicate entries (entries with the same memory address). d1801 3 a1803 3 the 'qsort' library function (which sorts an array) will be used to sort the symbol table. Also, the symbol lookup routine ('symtab.c:sym_lookup'), which finds symbols based on memory address, d1805 1 a1805 1 sorted array. Function symbols are indicated with an 'is_func' flag. d1807 1 a1807 1 can have an 'is_static' flag to indicate that it is a local symbol. d1810 11 a1820 11 Syms ('sym_ids.c:sym_id_parse'). Remember that a single symspec can match multiple symbols. An array of symbol tables ('syms') is created, each entry of which is a symbol table of Syms to be included or excluded from a particular listing. The master symbol table and the symspecs are examined by nested loops, and every symbol that matches a symspec is inserted into the appropriate syms table. This is done twice, once to count the size of each required symbol table, and again to build the tables, which have been malloced between passes. From now on, to determine whether a symbol is on an include or exclude symspec list, 'gprof' simply uses its standard symbol lookup routine on the appropriate table in the 'syms' array. d1823 2 a1824 2 ('gmon_io.c:gmon_out_read'), first by checking for a new-style 'gmon.out' header, then assuming this is an old-style BSD 'gmon.out' if d1827 1 a1827 1 New-style histogram records are read by 'hist.c:hist_read_rec'. For d1829 6 a1834 6 bins, and read them in. When multiple profile data files (or files with multiple histogram records) are read, the memory ranges of each pair of histogram records must be either equal, or non-overlapping. For each pair of histogram records, the resolution (memory region size divided by the number of bins) must be the same. The time unit must be the same for all histogram records. If the above containts are met, all d1837 1 a1837 1 As each call graph record is read ('call_graph.c:cg_read_rec'), the d1839 2 a1840 2 call graph arc is created by 'cg_arcs.c:arc_add', unless the arc fails a symspec check against INCL_ARCS/EXCL_ARCS. As each arc is added, a d1845 6 a1850 6 Basic-block records are read ('basic_blocks.c:bb_read_rec'), but only if line-by-line profiling has been selected. Each basic-block address is matched to a corresponding line symbol in the symbol table, and an entry made in the symbol's bb_addr and bb_calls arrays. Again, if multiple basic-block records are present for the same address, the call counts are cumulative. d1852 1 a1852 1 A gmon.sum file is dumped, if requested ('gmon_io.c:gmon_out_write'). d1855 11 a1865 10 ('hist.c:hist_assign_samples') by iterating over all the sample bins and assigning them to symbols. Since the symbol table is sorted in order of ascending memory addresses, we can simple follow along in the symbol table as we make our pass over the sample bins. This step includes a symspec check against INCL_FLAT/EXCL_FLAT. Depending on the histogram scale factor, a sample bin may span multiple symbols, in which case a fraction of the sample count is allocated to each symbol, proportional to the degree of overlap. This effect is rare for normal profiling, but overlaps are more common during line-by-line profiling, and can cause each of two adjacent lines to be credited with half a hit, for example. d1867 2 a1868 2 If call graph data is present, 'cg_arcs.c:cg_assemble' is called. First, if '-c' was specified, a machine-dependent routine ('find_call') d1871 19 a1889 18 A topological sort is performed by depth-first numbering all the symbols ('cg_dfn.c:cg_dfn'), so that children are always numbered less than their parents, then making a array of pointers into the symbol table and sorting it into numerical order, which is reverse topological order (children appear before parents). Cycles are also detected at this point, all members of which are assigned the same topological number. Two passes are now made through this sorted array of symbol pointers. The first pass, from end to beginning (parents to children), computes the fraction of child time to propagate to each parent and a print flag. The print flag reflects symspec handling of INCL_GRAPH/EXCL_GRAPH, with a parent's include or exclude (print or no print) property being propagated to its children, unless they themselves explicitly appear in INCL_GRAPH or EXCL_GRAPH. A second pass, from beginning to end (children to parents) actually propagates the timings along the call graph, subject to a check against INCL_TIME/EXCL_TIME. With the print flag, fractions, and timings now stored in the symbol structures, the topological sort array is now discarded, and a new array of pointers is assembled, this time sorted by propagated time. d1892 2 a1893 2 fairly straightforward. The call graph ('cg_print.c:cg_print') and flat profile ('hist.c:hist_print') are regurgitations of values already d1895 1 a1895 1 ('basic_blocks.c:print_annotated_source') uses basic-block information, d1900 1 a1900 1 source code itself ('cg_print.c'). Basically, the functions with the d1908 1 a1908 1 9.4 Debugging 'gprof' d1911 4 a1914 4 If 'gprof' was compiled with debugging enabled, the '-d' option triggers debugging output (to stdout) which can be helpful in understanding its operation. The debugging number specified is interpreted as a sum of the following options: d1918 1 d1921 1 d1925 1 d1929 1 d1933 1 d1936 1 a1936 1 file. For line-by-line profiling ('-l' option), also shows line d1938 1 d1940 2 a1941 1 Trace operation of '-c' option d1944 1 d1947 1 d1950 2 a1951 1 meaningful with '-l' option) d1954 1 d1956 1 a1956 1 Tracks operation of '-A' option d1967 1 a1967 1 d1992 2 a1993 2 of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is d1999 2 a2000 2 that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice d2004 3 a2006 3 of the public is a licensee, and is addressed as "you". You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. d2024 6 a2029 6 titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. d2040 10 a2049 10 straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not "Transparent" is called "Opaque". d2053 8 a2060 8 SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. d2098 2 a2099 2 distribute a large enough number of copies you must also follow the conditions in section 3. d2113 6 a2118 5 front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects. d2126 12 a2137 11 numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public. d2140 3 a2142 3 the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document. d2148 5 a2152 5 release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version: d2155 5 a2159 5 distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. d2189 6 a2194 6 authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. d2199 5 a2203 5 previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. d2206 2 a2207 2 Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor d2210 4 a2213 3 L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. d2226 5 a2230 5 material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles. d2239 9 a2247 9 and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one. d2257 2 a2258 2 modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, d2285 2 a2286 2 rules of this License for verbatim copying of each of the documents in all other respects. d2290 3 a2292 3 a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document. d2297 2 a2298 2 separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an "aggregate" if the d2343 2 a2344 2 provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the d2356 4 a2359 4 the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it. d2361 1 a2361 1 10. FUTURE REVISIONS OF THIS LICENSE d2367 1 a2367 1 . d2374 5 a2378 5 published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that d2382 1 a2382 1 11. RELICENSING d2412 1 d2429 1 a2429 1 Texts, replace the "with...Texts." line with this: d2440 3 a2442 3 recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software. d2447 18 a2464 18 Node: Top719 Node: Introduction2044 Node: Compiling4535 Node: Executing8591 Node: Invoking11379 Node: Output Options12794 Node: Analysis Options19886 Node: Miscellaneous Options23806 Node: Deprecated Options25060 Node: Symspecs27123 Node: Output28949 Node: Flat Profile29989 Node: Call Graph34942 Node: Primary38174 Node: Callers40762 Node: Subroutines42880 Node: Cycles44721 Node: Line-by-line51498 d2466 11 a2476 11 Node: Inaccuracy58572 Node: Sampling Error58830 Node: Assumptions61734 Node: How do I?63204 Node: Incompatibilities64761 Node: Details66255 Node: Implementation66648 Node: File Format72547 Node: Internals76835 Node: Debugging85325 Node: GNU Free Documentation License86915 @ 1.1.1.7 log @2018-07-18 Nick Clifton 2.31.1 Release point. 2018-07-18 Nick Clifton * (DEVO_SUPPORT): Fix typo in previous delta. (do_proto_toplev): Add --quiet option to configure command line. 2018-07-16 Nick Clifton * src-release.sh (DEVO_SUPPORT): Add test-driver and ar-lib. 2018-07-06 Sebastian Huber * config.sub: Sync with upstream version 2018-07-03. 2018-07-05 Sebastian Huber * config.guess: Sync with upstream version 2018-06-26. * config.sub: Sync with upstream version 2018-07-02. 2018-06-24 Nick Clifton * configure: Regenerate. 2018-06-19 Simon Marchi * libtool.m4: Use AC_LANG_SOURCE. * configure.ac: Remove AC_PREREQ, use AC_LANG_SOURCE. * README-maintainer-mode: Update version requirements. * ar-lib: New file. * test-driver: New file. * configure: Re-generate. 2018-06-18 Eric Botcazou * Makefile.def (fortran): Add check-target-libgomp-fortran. * Makefile.tpl (check-target-libgomp-fortran): New phony target. * Makefile.in: Regenerate. * configure: Regenerate. 2018-06-18 Simon Marchi * configure.ac: Sync with GCC, remove MPX-related things. 2018-05-01 Nick Clifton * config.guess: Synchronize with config project master sources. * config.sub: Likewise. 2018-05-01 Francois H. Theron * configure.ac: Added "nfp" target. * configure: Regenerate. 2018-02-13 Maciej W. Rozycki * configure.ac (noconfigdirs): Add `ld'. * configure: Regenerate. 2018-01-30 Nick Clifton * src-release.sh (do_proto_toplev): Add patterns for more junk files to delete before creating the tarball. 2018-01-29 Nick Clifton * src-release.sh (do_proto_toplev): Strip patch remnant files from the sources before creating the tarball. 2018-01-13 Nick Clifton * src-release.sh: Update copyright notice. Change reference to devo to be a reference to root. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.4 from gprof.texi. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.30.90. d2428 29 a2456 29 Node: Introduction2045 Node: Compiling4536 Node: Executing8592 Node: Invoking11380 Node: Output Options12795 Node: Analysis Options19887 Node: Miscellaneous Options23807 Node: Deprecated Options25061 Node: Symspecs27124 Node: Output28950 Node: Flat Profile29990 Node: Call Graph34943 Node: Primary38175 Node: Callers40763 Node: Subroutines42881 Node: Cycles44722 Node: Line-by-line51499 Node: Annotated Source55575 Node: Inaccuracy58573 Node: Sampling Error58831 Node: Assumptions61735 Node: How do I?63205 Node: Incompatibilities64762 Node: Details66256 Node: Implementation66649 Node: File Format72548 Node: Internals76836 Node: Debugging85326 Node: GNU Free Documentation License86916 @ 1.1.1.8 log @Import binutils-2.34 2020-01-18 Nick Clifton Binutils 2.34 branch created. 2020-01-18 Nick Clifton Synchronize top level configure files with master version: 2020-01-01 Ben Elliston * config.guess: Update copyright years. * config.sub: Likewise. 2019-12-21 Ben Elliston * config.guess (set_cc_for_build): Prevent multiple calls by checking if $tmp is already set. We can't check CC_FOR_BUILD as the user may set it externally. Thanks to Torbjörn Granlund for the bug report. 2019-12-21 Torbjörn Granlund * config.guess (alpha:Linux:*:*): Guard against missing /proc/cpuinfo by redirecting standard error to /dev/null. 2019-09-12 Daniel Bittman * config.guess (*:Twizzler:*:*): New. * config.sub (-twizzler*): New. 2019-07-24 Ben Elliston * config.guess (mips:OSF1:*.*): Whitespace cleanup. 2019-06-30 Ben Elliston * config.sub (case $os): Match nsk* and powerunix. Don't later match nsk* and set os=nsk which removes the OS version number. 2019-06-30 Ben Elliston * config.sub: Recognise os108*. 2019-06-26 Ben Elliston * config.sub (hp300): Set $os to hpux. 2019-06-26 Ben Elliston * config.sub (vsta): Move into alphabetical order. 2019-06-10 Ben Elliston * config.guess (*:OS108:*:*): Recognise new OS. 2019-05-28 Ben Elliston * config.guess (*:Darwin:*:*): Run xcode-select to determine if a system compiler is installed. If not, do not run set_cc_for_build, as the default cc will open a dialog box asking to install Xcode. If no C compiler is available, guess based on uname -p and uname -m. 2019-05-28 Ben Elliston * config.guess (*:Darwin:*:*): Simplify UNAME_PROCESSOR. 2020-01-17 Simon Marchi * Makefile.def: Add dependencies of all-gdbsupport on all-bfd. * Makefile.in: Re-generate. 2020-01-14 Tom Tromey * src-release.sh (GDB_SUPPORT_DIRS): Add gdbsupport. * MAINTAINERS: Add gdbsupport. * configure: Rebuild. * configure.ac (configdirs): Add gdbsupport. * gdbsupport: New directory, move from gdb/gdbsupport. * Makefile.def (host_modules, dependencies): Add gnulib. * Makefile.in: Rebuild. 2020-01-09 Aaron Merey * config/debuginfod.m4: New file. Add macro AC_DEBUGINFOD. Adds new configure option --with-debuginfod. * configure: Regenerate. * configure.ac: Call AC_DEBUGINFOD. 2019-12-26 Christian Biesinger * .gitignore: Add perf.data and perf.data.old. 2019-10-17 Sergio Durigan Junior * src-release.sh (GDB_SUPPORT_DIRS): Add libctf. 2019-10-17 Alan Modra PR 29 * src-release.sh (getver): Replace "head -1" with "head -n 1". 2019-07-30 Nick Alcock * Makefile.def (host_modules): libctf is no longer no_install. * Makefile.in: Regenerated. 2019-07-13 Nick Alcock * Makefile.def (dependencies): all-ld depends on all-libctf. * Makefile.in: Regenerated. 2019-09-09 Phil Blundell binutils 2.33 branch created 2019-08-19 Tom Tromey * configure: Rebuild. * configure.ac: Add --with-static-standard-libraries. 2019-08-09 Nick Clifton * libiberty: Sync with gcc. Bring in: 2019-08-08 Martin Liska PR bootstrap/91352 * lrealpath.c (is_valid_fd): New function. 2019-07-24 Martin Liska PR lto/91228 * simple-object-elf.c (simple_object_elf_copy_lto_debug_sections): Find first '\0' starting from gnu_lto + 1. 2019-07-12 Ren Kimura * simple-object-elf.c (simple_object_elf_match): Check zero value shstrndx. This fixes a Bug 90924. 2019-07-22 Martin Liska * simple-object-elf.c (simple_object_elf_copy_lto_debug_sections): Do not search for gnu_lto_v1, but search for first '\0'. 2019-07-18 Eduard-Mihai Burtescu * cplus-dem.c: Include rust-demangle.h. * rust-demangle.c: Include rust-demangle.h. * rust-demangle.h: New file. 2019-05-31 Michael Forney * cp-demangle.c: Don't define CP_DYNAMIC_ARRAYS if __STDC_NO_VLA__ is non-zero. 2019-04-30 Ben L * d-demangle.c (dlang_parse_assocarray): Correctly handle error result. * testsuite/d-demangle-expected: Add testcase. * d-demangle.c (dlang_parse_tuple): Correctly handle error result. * testsuite/d-demangle-expected: Add testcase. * d-demangle.c (dlang_parse_structlit): Correctly handle error result. * testsuite/d-demangle-expected: Add testcase. * d-demangle.c (dlang_parse_arrayliteral): Correctly handle error result. * testsuite/d-demangle-expected: Add testcase. * d-demangle.c (dlang_parse_integer): Fix stack underflow. * testsuite/d-demangle-expected: Add testcase. * cp-demangle (d_print_comp_inner): Guard against a NULL 'typed_name'. * testsuite/demangle-expected: Add testcase. * cp-demangle.c (d_encoding): Guard against NULL return values from d_right (dc). * testsuite/demangle-expected: Add testcase. 2019-04-29 Ben L * cp-demangle.c (d_expression_1): Don't peek ahead unless the current char is valid. * testsuite/demangle-expected: Add testcase. 2019-04-10 Nick Clifton PR 89394 * cp-demangle.c (cplus_demangle_fill_name): Reject negative lengths. (d_count_templates_scopes): Replace num_templates and num_scopes parameters with a struct d_print_info pointer parameter. Adjust body of the function accordingly. Add recursion counter and check that the recursion limit is not reached. (d_print_init): Pass dpi parameter to d_count_templates_scopes. Reset recursion counter afterwards, unless the recursion limit was reached. 2019-07-13 Joel Brobecker * src-release (getver): If $tool/gdbsupport/create-version.sh exists, use that to determine the version number. 2019-06-21 Andreas Schwab * src-release.sh (GDB_SUPPORT_DIRS): Add gnulib. 2019-06-14 Tom Tromey * MAINTAINERS: Add gnulib. * gnulib: New directory, move from gdb/gnulib. * configure.ac (host_libs): Add gnulib. * configure: Rebuild. * Makefile.def (host_modules, dependencies): Add gnulib. * Makefile.in: Rebuild. 2019-06-03 Nick Clifton Revert: 2019-05-29 Nick Clifton * configure.ac (noconfigdirs): Add libctf if the target does not use the ELF file format. * configure: Regenerate. 2019-05-29 Nick Clifton * src-release.sh (do_proto_toplev): Add libctf to list of directories that can be disabled. 2019-05-29 Nick Clifton * configure.ac (noconfigdirs): Add libctf if the target does not use the ELF file format. * configure: Regenerate. 2019-05-28 Nick Alcock * Makefile.def (dependencies): configure-libctf depends on all-bfd and all its deps. * Makefile.in: Regenerated. 2019-05-28 Nick Alcock * MAINTAINERS: Add libctf. 2019-05-28 Nick Alcock * Makefile.def (host_modules): Add libctf. * Makefile.def (dependencies): Likewise. libctf depends on zlib, libiberty, and bfd. * Makefile.in: Regenerated. * configure.ac (host_libs): Add libctf. * configure: Regenerated. 2019-05-23 Jose E. Marchesi * config.guess: Synchronize with config project master sources. * config.sub: Likewise. * readline/support/config.guess: Likewise. * readline/support/config.sub: Likewise. 2019-04-10 Nick Clifton * libiberty: Sync with gcc. Bring in: 2019-04-10 Nick Clifton PR 89394 * cp-demangle.c (cplus_demangle_fill_name): Reject negative lengths. (d_count_templates_scopes): Replace num_templates and num_scopes parameters with a struct d_print_info pointer parameter. Adjust body of the function accordingly. Add recursion counter and check that the recursion limit is not reached. (d_print_init): Pass dpi parameter to d_count_templates_scopes. Reset recursion counter afterwards, unless the recursion limit was reached. 2018-06-24 Nick Clifton 2.32 branch created. 2019-01-14 Rainer Orth Merge from GCC: PR target/88535 * config.guess: Import upstream version 2019-01-03. * config.sub: Import upstream version 2019-01-01. 2019-01-10 Nick Clifton * libiberty: Sync with gcc. Bring in: 2019-01-09 Sandra Loosemore PR other/16615 * cp-demangle.c: Mechanically replace "can not" with "cannot". * floatformat.c: Likewise. * strerror.c: Likewise. 2018-12-22 Jason Merrill Remove support for demangling GCC 2.x era mangling schemes. * cplus-dem.c: Remove cplus_mangle_opname, cplus_demangle_opname, internal_cplus_demangle, and all subroutines. (libiberty_demanglers): Remove entries for ancient GNU (pre-3.0), Lucid, ARM, HP, and EDG demangling styles. (cplus_demangle): Remove 'work' variable. Don't call internal_cplus_demangle. 2019-01-03 Дилян Палаузов * configure.ac: Don't configure readline if --with-system-readline is used. * configure: Re-generate. 2018-10-31 Joseph Myers Merge from GCC: PR bootstrap/82856 * multilib.am: New file. From automake. 2018-09-12 Sergio Durigan Junior * src-release.sh (GDB_SUPPORT_DIRS): Add "contrib". 2018-07-16 Nick Clifton @@ -20,6 +338,18 @@@@ * config.guess: Sync with upstream version 2018-06-26. * config.sub: Sync with upstream version 2018-07-02. 2018-06-29 Alexandre Oliva * configure.ac: Introduce support for @@unless/@@endunless. * Makefile.tpl (dep-kind): Rewrite with cond; return postbootstrap in some cases. (make-postboot-dep, postboot-targets): New. (dependencies): Do not output postbootstrap dependencies at first. Output non-target ones changed for configure to depend on stage_last @@if gcc-bootstrap, and the original deps @@unless gcc-bootstrap. * configure.in, Makefile.in: Rebuilt. 2018-06-24 Nick Clifton * configure: Regenerate. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.5 from gprof.texi. d5 1 a5 1 Copyright (C) 1988-2020 Free Software Foundation, Inc. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.34. d170 1 a170 1 '-static-libgcc' command-line option. For example: d193 1 a193 1 command-line option which will insert calls to special user supplied d1211 2 a1212 2 versions of 'gcc' the program had to be compiled with the '-a' command-line option as well. d2428 29 a2456 29 Node: Introduction2042 Node: Compiling4533 Node: Executing8589 Node: Invoking11377 Node: Output Options12792 Node: Analysis Options19884 Node: Miscellaneous Options23804 Node: Deprecated Options25058 Node: Symspecs27121 Node: Output28947 Node: Flat Profile29987 Node: Call Graph34940 Node: Primary38172 Node: Callers40760 Node: Subroutines42878 Node: Cycles44719 Node: Line-by-line51496 Node: Annotated Source55572 Node: Inaccuracy58570 Node: Sampling Error58828 Node: Assumptions61732 Node: How do I?63202 Node: Incompatibilities64759 Node: Details66253 Node: Implementation66646 Node: File Format72545 Node: Internals76833 Node: Debugging85323 Node: GNU Free Documentation License86913 @ 1.1.1.9 log @Update binutils from 2.34 to 2.39 2022-07-08 Nick Clifton * 2.39 branch created. 2022-07-04 Nick Clifton * libiberty: Synchronize with GCC. Bring in: 2022-07-01 Nick Clifton PR demangler/105039 * rust-demangle.c (demangle_const): Add recursion limit. 2022-06-26 Simon Marchi * configure.ac: Add AC_CONFIG_MACRO_DIRS call. * configure: Re-generate. 2022-04-12 Nick Clifton * zlib: Rebase to the 1.2.12 release. 2022-04-08 Simon Marchi * configure.ac: Add AC_SUBST(PKG_CONFIG_PATH). * configure: Re-generate. * Makefile.tpl (HOST_EXPORTS): Pass PKG_CONFIG_PATH. (PKG_CONFIG_PATH): New. * Makefile.in: Re-generate. 2022-03-15 Jose E. Marchesi * gprofng/src/gp-collect-app.cc (collect::check_args): Use fallthrough comment instead of attribute. 2022-03-11 Vladimir Mezentsev * Makefile.def: Add gprofng module. * configure.ac: Add --enable-gprofng option. * src-release.sh: Add gprofng. * Makefile.in: Regenerate. * configure: Regenerate. * gprofng: New directory. 2022-01-22 Nick Clifton * 2.38 release branch created. 2022-01-17 Nick Clifton Update config.[guess|sub] from upstream: 2022-01-09 Idan Horowitz config.guess: recognize SerenityOS * config.guess (*:SerenityOS:*:*): Recognize. (timestamp): Update. 2022-01-03 Bernhard Voelker Fix GPLv3 license headers to use a comma instead of semicolon See: https://www.gnu.org/licenses/gpl-3.0.html#howto Update license headers automatically using the following script: $ git grep -l 'Foundation; either version 3' \ | xargs sed -i '/Foundation; either version 3/ s/n; e/n, e/' * config.guess: Adjust via the above command. (timestamp): Update. * config.sub: Likewise. * doc/config.guess.1: Regenerate. * doc/config.sub.1: Likewise. 2022-01-01 Dmitry V. Levin Update copyright years * config.guess: Update copyright years. * config.sub: Likewise. 2021-12-25 Dmitry V. Levin config.sub: alias armh to armv7l ALT uses armh as an alias for armv7l-alt-linux-gnueabihf since 2012. * config.sub (armh-unknown|armh-alt): Set cpu, vendor, and basic_os. (timestamp): Update. 2021-12-24 Dmitry V. Levin config.sub: alias aarch64le to aarch64 Apparently, QNX reports aarch64 as aarch64le on little-endian machines. * config.sub (aarch64le-*): Set cpu to aarch64. (timestamp): Update. 2021-12-13 Dmitry V. Levin config.sub: fix typo in timestamp * config.sub: Fix timestamp. 2021-11-30 Andreas F. Borchert config.guess: x86_64-pc-solaris2.11 is not properly recognized config.guess guesses Solaris 11 to run on a 32-bit platform despite Solaris 11 no longer supporting any 32-bit platform. See the following code at lines 434 to 445: | SUN_ARCH=i386 | # If there is a compiler, see if it is configured for 64-bit objects. | # Note that the Sun cc does not turn __LP64__ into 1 like gcc does. | # This test works for both compilers. | if test "$CC_FOR_BUILD" != no_compiler_found; then | if (echo '#ifdef __amd64'; echo IS_64BIT_ARCH; echo '#endif') | \ | (CCOPTS="" $CC_FOR_BUILD -E - 2>/dev/null) | \ | grep IS_64BIT_ARCH >/dev/null | then | SUN_ARCH=x86_64 | fi | fi If "cc" is installed, i.e. the Oracle Studio compiler, this one is chosen for $CC_FOR_BUILD. This compiler, the gcc provided by Oracle and also gcc bootstrapped from sources on that platform with a default configuration will by default generate 32-bit binaries -- even on a 64-bit platform. And __amd64 will not be defined for compilations targeting a 32-bit platform. This is different from the corresponding behaviour on GNU/Linux systems where the local platform is targeted by default. Thus, as long as you do not add "-m64" or if you have a custom-built gcc which defaults to 64 bit, you will get 32-bit binaries on Solaris despite living on a 64-bit platform. * config.guess (i86pc:SunOS:5.*:* || i86xen:SunOS:5.*:*): Adapt the test by adding the "-m64" flag. This will work properly for Solaris 10 as well (the last Solaris release that supported x86 32-bit platforms). 2021-10-27 Jordi Sanfeliu Recognize Fiwix $ make check cd testsuite && bash config-guess.sh && rm uname PASS: config.guess checks (137 tests) cd testsuite && bash config-sub.sh PASS: config.sub checks (882 tests) PASS: config.sub idempotency checks (819 tests) PASS: config.sub canonicalise each config.guess testcase (137 tests) * config.guess (i*86:Fiwix:*:*): Recognize. * config.sub (fiwix*): Likewise. 2021-10-18 Kinshuk Dua config.sub: Fix typo in comment Fixes: 5e531d391852a54e7fab2d8ff55625fca514b305 2021-08-14 Nick Bowler config.sub: work around command assignment bug in some shells When combining variable assignments with a shell command, some older shells (notably heirloom-sh and presumably also Solaris 10 /bin/sh) have a bug which causes the assignment to alter the current execution environment whenever the command is a shell built-in. For example: % dash -c 'x=good; x=bad echo >/dev/null; echo $x' good % jsh -c 'x=good; x=bad echo >/dev/null; echo $x' bad The config.sub script contains a few commands of the form: IFS=- read ... which triggers this bug, causing the IFS assignment to persist for the remainder of the script. This can cause misbehaviour in certain cases, for example: % jsh config.sub i386-linux-gnu config.sub: test: unknown operator gnu % jsh config.sub i386-gnu/linux sed: can't read s|gnu/linux|gnu|: No such file or directory Invalid configuration `i386-gnu/linux': OS `' not recognized * config.sub: Save and restore IFS explicitly to avoid shell bugs. * doc/config.sub.1: Regenerate. 2021-08-04 Jeremy Soller config.sub: add Linux Relibc Target $ make check cd testsuite && bash config-guess.sh && rm uname PASS: config.guess checks (136 tests) cd testsuite && bash config-sub.sh PASS: config.sub checks (881 tests) PASS: config.sub idempotency checks (818 tests) PASS: config.sub canonicalise each config.guess testcase (136 tests) * config.sub (relibc*): Recognize. * doc/config.sub.1: Regenerate. * testsuite/config-sub.data (x86_64-linux-relibc): New test. 2021-07-06 Stephanos Ioannidis config.sub: add Zephyr RTOS support This adds the Zephyr RTOS targets in preparation for implementing the Zephyr RTOS-specific toolchain support. $ make check cd testsuite && bash config-guess.sh && rm uname PASS: config.guess checks (136 tests) cd testsuite && bash config-sub.sh PASS: config.sub checks (880 tests) PASS: config.sub idempotency checks (817 tests) PASS: config.sub canonicalise each config.guess testcase (136 tests) * config.sub (zephyr*): Recognize. * doc/config.sub.1: Regenerate. * testsuite/config-sub.data: Add testcases for *-zephyr. 2021-07-03 Ozkan Sezer config.sub: disable shellcheck SC2006 / SC2268 warnings This is in line with the recent config.guess change in commit 12fcf67c9108f4c4b581eaa302088782f0ee40ea * config.sub (shellcheck disable): Add SC2006,SC2268. Suggested-by: Jacob Bachmeyer 2021-07-03 Ozkan Sezer config.sub: normalize the quoting in the `echo FOO | sed ...` Some cases quote the argument to echo and some do not. At runtime it probably does not matter because the substituted values will never contain whitespace, but quoting them all would make shellcheck more useful. * config.sub: Consistently quote the argument of echo. * doc/config.sub.1: Regenerate. Suggested-by: Jacob Bachmeyer 2021-07-02 Ozkan Sezer config.sub: replace POSIX $( ) with classic ` ` throughout This is in line with the recent config.guess change in commit d70c4fa934de164178054c3a60aaa0024ed07c91. The patch was generated using patch-6.gawk script introduced in that commit. * config.sub: Revert POSIX command substitutions to classic form. 2021-06-04 Vineet Gupta Recognize arc32 This is the 32-bit variant of ARCv3 ISA (which is not compatible with the 32-bit ARCv2 ISA) | make check | cd testsuite && bash config-guess.sh && rm uname | PASS: config.guess checks (136 tests) | cd testsuite && bash config-sub.sh | PASS: config.sub checks (864 tests) | PASS: config.sub idempotency checks (801 tests) | PASS: config.sub canonicalise each config.guess testcase (136 tests) * config.guess (arc32:Linux:*:*): Recognize. * config.sub (arc32): Likewise. 2021-05-27 Jacob Bachmeyer Remove automatic patch generators These tools have served their purposes and need not be kept outside of the repository history any longer. This patch as a diff also collects the contents of the various tools in one convenient place. * patch-1.gawk: Remove. * patch-3.gawk: Likewise. * patch-6.gawk: Likewise. 2021-05-26 Jacob Bachmeyer config.guess: manual fixups after previous automatic patch The tool could not handle command substitutions that span lines, but fortunately there were only two such substitutions in the script. The test for which universe is active on Pyramid is rewritten into a case block because it was the only use of a command substitution as an argument to the test command, which would require quoting. * config.guess: Rewrite "if" for Pyramid systems to "case". 2021-05-26 Jacob Bachmeyer config.guess: replace POSIX $( ) with classic ` ` throughout The previous replacement of backticks with POSIX command substitutions was ill-considered and illogical: this script recognizes many archaic machine types that probably never had POSIX shells, therefore it needs to be able to run successfully under pre-POSIX shells. This patch was generated using the included GNU Awk program. * config.guess: Revert POSIX command substitutions to classic form. * patch-6.gawk: Store the tool that produced the automated patch. 2021-05-26 Jacob Bachmeyer config.guess: manual fixup after previous automated patches This patch provides the special handling for the GNU system. As these were two small and unique edits, they were not included in the scripts. This patch also cleans up other minor issues that must be addressed before reverting to classic command substitutions and updates "shellcheck" directives to account for changes in this script and the change in "shellcheck" towards reporting individual portability issues. 2021-05-26 Jacob Bachmeyer config.guess: automatic fixups after previous automated patch This patch was generated using the following command: sed -i config.guess \ -e '/="[^"]\+"\(-\|$\)/s/="\([^"([:space:])]\+\)"/=\1/' \ -e '/="[^"]\+"[[:alnum:]]/s/="\$\([^([:space:])]\+\)"/=${\1}/' \ -e \ '/\$(echo[^|]\+|/s/\([^[:space:]]\)[[:space:]]*|[[:space:]]*sed/\1 | sed/g' * config.guess: Remove unneeded quotes in other variable assignments, standardize spacing for "echo ... | sed" substitutions. 2021-05-26 Jacob Bachmeyer config.guess: remove unneeded quotes and factor command substitutions This is further cleanup and simplifies some constructs that can confuse Emacs' syntax highlighting while generally reducing required quoting. This patch was generated using the included GNU Awk program. * config.guess: Remove unneeded variable quotes and factor out command substitutions when setting GUESS. * patch-3.gawk: Store the tool that produced the automated patch. 2021-05-25 Jacob Bachmeyer config.guess: manual fixups after previous automatic patch * config.guess: Adjust a few "leftover" cases that the tool could not easily recognize and fixes comment indentation in a few other special cases. 2021-05-25 Jacob Bachmeyer config.guess: use intermediate variable with uname results This will allow quoting to be significantly simplified in another pass through the file. This patch was generated using the included GNU Awk program. * config.guess: Use GUESS variable to hold results of uname analysis. * patch-1.gawk: Store the tool that produced the automated patch. 2021-05-25 Jacob Bachmeyer config.guess: introduce intermediate variable with uname results This will allow quoting to be significantly simplified in another pass through the file. * config.guess: Introduce GUESS variable to hold results of uname analysis. 2021-05-24 Dmitry V. Levin config.guess: fix shellcheck warning SC2154 While, according to Plan 9 documentation, the environment variable $cputype is set to the name of the kernel's CPU's architecture, shellcheck warns that cputype is referenced but not assigned. Be on the safe side and do not use cputype if it is not defined or empty. * config.guess (*:Plan9:*:*): Fix shellcheck warning SC2154. 2021-05-24 Dmitry V. Levin config.guess: remove redundant quotes in case commands According to the GNU Autoconf Portable Shell Programming manual, the Bourne shell does not systematically split variables and back-quoted expressions, in particular on the right-hand side of assignments and in the argument of 'case'. The change is made automatically using the following command: $ sed -E -i 's/(\)/\1\2\3/' config.guess * config.guess: Simplify case commands by removing quotes around the argument. Suggested-by: Jacob Bachmeyer 2021-05-24 Dmitry V. Levin config.guess: simplify exit status workaround on alphaev67-dec-osf5.1 Commit 29865ea8a5622cdd80b7a69a0afa78004b4cd311 introduced an exit trap reset before exiting to avoid a spurious non-zero exit status on alphaev67-dec-osf5.1. Simplify that code a bit by moving the exit trap reset around. * config.guess (alpha:OSF1:*:*): Reset exit trap earlier. * doc/config.guess.1: Regenerate. 2021-10-29 Eli Zaretskii * gdb/doc/gdb.texinfo (Command Options): (Data): Document '-memory-tag-violations'. Update the example. 2021-09-28 Andrew Burgess * src-release.sh (GDB_SUPPPORT_DIRS): Add libbacktrace. 2021-09-27 Nick Alcock PR libctf/27967 * libtool.m4 (LT_PATH_NM): Try BSDization flags with a user-provided NM, if there is one. Run nm on itself, not on /dev/null, to avoid errors from nms that refuse to work on non-regular files. Remove other workarounds for this problem. Strip out blank lines from the nm output. 2021-09-27 Nick Alcock PR libctf/27967 * libtool.m4 (lt_cv_sys_global_symbol_pipe): Augment symcode for Solaris 11. 2021-07-03 Nick Clifton * 2.37 release branch created. 2021-07-03 Nick Clifton * libiberty: Sync with gcc. Bring in: 2021-06-30 Gerald Pfeifer * make-temp-file.c (usrtmp): Remove. (choose_tmpdir): Remove use of usrtmp. 2021-06-28 Indu Bhagat * simple-object.c (handle_lto_debug_sections): Copy over .BTF section. 2021-06-28 Indu Bhagat David Faust Jose E. Marchesi Weimin Pan * simple-object.c (handle_lto_debug_sections): Copy over .ctf sections. 2021-06-05 John David Anglin PR target/100734 * configure.ac: Use libiberty snprintf and vsnprintf on hppa*-*-hpux*. * configure: Regenerate. 2021-05-06 Tom Tromey * hashtab.c (htab_eq_string): New function. 2021-05-04 Eric Botcazou * configure.ac: Make test for variables more robust. * configure: Regenerate. 2021-05-03 H.J. Lu PR bootstrap/99703 * configure: Regenerated. 2021-04-21 Andreas Schwab PR demangler/100177 * rust-demangle.c (demangle_const_char): Properly print the character value. 2021-03-31 Patrick Palka PR c++/88115 * cp-demangle.c (d_dump, d_make_comp, d_expression_1) (d_count_templates_scopes): Handle DEMANGLE_COMPONENT_VENDOR_EXPR. (d_print_comp_inner): Likewise. : Revert r11-4926 change. : Likewise. * testsuite/demangle-expected: Adjust __alignof__ tests. 2021-03-16 Nick Clifton * sha1.c (sha1_process_bytes): Use memmove in place of memcpy. 2021-02-20 Mike Frysinger * Makefile.in (ACLOCAL, ACLOCAL_AMFLAGS, $(srcdir)/aclocal.m4): Define. (configure_deps): Rename to ... (aclocal_deps): ... this. Replace aclocal.m4 with acinclude.m4. ($(srcdir)/configure): Replace $(configure_deps) with $(srcdir)/aclocal.m4. * aclocal.m4: Move libiberty macros to acinclude.m4, then regenerate. * acinclude.m4: New file. * configure: Regenerate. 2021-02-19 Ayush Mittal * argv.c (expandargv): free allocated buffer if read fails. 2021-02-01 Martin Sebor * dyn-string.c (dyn_string_insert_cstr): Use memcpy instead of strncpy to avoid -Wstringop-truncation. 2021-05-29 Mike Frysinger * configure.ac: Add gnulib to configdirs for sim. * configure: Regenerate. 2021-05-24 Maciej W. Rozycki * MAINTAINERS: Update path to readline config.{sub,guess} files. 2021-05-24 Maciej W. Rozycki * config.guess: Import from upstream. * config.sub: Likewise. 2021-05-18 Mike Frysinger * Makefile.def: Add configure-sim dependency on all-gnulib. * Makefile.in: Regenerated. 2021-05-04 Nick Clifton * configure.ac (AC_PROG_CC): Replace with AC_PROG_CC_C99. * configure: Regenerate. 2021-03-18 Nick Alcock PR libctf/27482 * Makefile.def: Add install-bfd dependencies for install-libctf and install-ld, and install-strip-bfd dependencies for install-strip-libctf and install-strip-ld; move the install-ld dependency on install-libctf to join it. * Makefile.in: Regenerated. 2021-03-12 Mike Frysinger * Makefile.def: Remove all-sim dependency on configure-gdb. * Makefile.in: Regenerated. 2021-02-28 H.J. Lu PR binutils/26766 * Makefile.tpl (PGO_BUILD_TRAINING_FLAGS_TO_PASS): Add PGO_BUILD_TRAINING=yes. (PGO_BUILD_TRAINING_MFLAGS): New. (all): Pass $(PGO_BUILD_TRAINING_MFLAGS) to the PGO build. 2021-02-09 Alan Modra * configure.ac: Delete arm*-*-symbianelf* entry. * configure: Regenerate. 2021-01-26 Nick Alcock * Makefile.def: Add install-libctf dependency to install-ld. * Makefile.in: Regenerated. 2021-01-12 Mike Frysinger * src-release.sh (do_proto_toplev): Rewrite indentation. 2021-01-11 H.J. Lu PR binutils/26766 * configure.ac: * configure: Regenerated. 2021-01-11 H.J. Lu PR ld/27173 * configure: Regenerated. * libtool.m4 (_LT_CMD_OLD_ARCHIVE): Check if AR works with --plugin and rc before enabling --plugin. 2021-01-09 H.J. Lu PR binutils/26766 * Makefile.tpl (BUILD_CFLAGS): New. (CFLAGS): Append $(BUILD_CFLAGS). (CXXFLAGS): Likewise. (PGO_BUILD_GEN_FLAGS_TO_PASS): New. (PGO_BUILD_TRAINING_CFLAGS): Likewise. (PGO_BUILD_TRAINING_CXXFLAGS): Likewise. (PGO_BUILD_TRAINING_FLAGS_TO_PASS): Likewise. (PGO_BUILD_TRAINING_MFLAGS): Likewise. (PGO_BUILD_USE_FLAGS_TO_PASS): Likewise. (PGO-TRAINING-TARGETS): Likewise. (PGO_BUILD_TRAINING): Likewise. (all): Add '+' to the command line for recursive make. Support the PGO build. * configure.ac: Add --enable-pgo-build[=lto]. AC_SUBST PGO_BUILD_GEN_CFLAGS, PGO_BUILD_USE_CFLAGS and PGO_BUILD_LTO_CFLAGS. Enable the PGO build in Makefile. * Makefile.in: Regenerated. * configure: Likewise. 2021-01-09 H.J. Lu * Makefile.tpl (AR): Add @@AR_PLUGIN_OPTION@@ (RANLIB): Add @@RANLIB_PLUGIN_OPTION@@. * configure.ac: Include config/gcc-plugin.m4. AC_SUBST AR_PLUGIN_OPTION and RANLIB_PLUGIN_OPTION. * libtool.m4 (_LT_CMD_OLD_ARCHIVE): Pass --plugin to AR and RANLIB if possible. * Makefile.in: Regenerated. * configure: Likewise. 2021-01-09 Nick Clifton * 2.36 release branch crated. 2021-01-07 Samuel Thibault * libtool.m4: Match gnu* along with other GNU systems. 2021-01-07 Alan Modra * config.sub: Accept OS of eabi* and gnueabi*. 2021-01-05 Nick Alcock * Makefile.def (libctf): No longer no_check. Checking depends on all-ld. * Makefile.in: Regenerated. 2021-01-05 Nick Clifton * libiberty: Sync with gcc. Bring in: 2021-01-04 Martin Liska * strverscmp.c: Convert to utf8 from iso8859. 2020-12-22 Jason Merrill PR c++/67343 * cp-demangle.h (struct d_info): Add unresolved_name_state. * cp-demangle.c (d_prefix): Add subst parm. (d_nested_name): Pass it. (d_unresolved_name): Split out from... (d_expression_1): ...here. (d_demangle_callback): Maybe retry with old sr mangling. * testsuite/demangle-expected: Add test. 2020-12-21 Jason Merrill * cp-demangle.c (d_expression_1): Recognize qualified-id on RHS of dt/pt. * testsuite/demangle-expected: Add test. 2020-12-21 Jason Merrill * cp-demangle.c (d_unqualified_name): Clear is_expression. * testsuite/demangle-expected: Add tests. 2020-11-25 Matthew Malcomson * configure: Regenerate. * configure.ac: Avoid using sanitizer. 2020-11-13 Eduard-Mihai Burtescu * rust-demangle.c (struct rust_demangler): Add skipping_printing and bound_lifetime_depth fields. (eat): Add (v0-only). (parse_integer_62): Add (v0-only). (parse_opt_integer_62): Add (v0-only). (parse_disambiguator): Add (v0-only). (struct rust_mangled_ident): Add punycode{,_len} fields. (parse_ident): Support v0 identifiers. (print_str): Respect skipping_printing. (print_uint64): Add (v0-only). (print_uint64_hex): Add (v0-only). (print_ident): Respect skipping_printing, Support v0 identifiers. (print_lifetime_from_index): Add (v0-only). (demangle_binder): Add (v0-only). (demangle_path): Add (v0-only). (demangle_generic_arg): Add (v0-only). (demangle_type): Add (v0-only). (demangle_path_maybe_open_generics): Add (v0-only). (demangle_dyn_trait): Add (v0-only). (demangle_const): Add (v0-only). (demangle_const_uint): Add (v0-only). (basic_type): Add (v0-only). (rust_demangle_callback): Support v0 symbols. * testsuite/rust-demangle-expected: Add v0 testcases. 2020-11-13 Seija Kijin * strstr.c (strstr): Make implementation ANSI/POSIX compliant. 2020-11-11 Patrick Palka PR c++/88115 * cp-demangle.c (d_print_comp_inner) : Don't print the "operator " prefix for __alignof__. : Always print parens around the operand of __alignof__. * testsuite/demangle-expected: Test demangling for __alignof__. 2020-11-09 Christophe Lyon * pex-win32.c (pex_win32_exec_child): Initialize orig_err. 2020-10-06 Martin Liska PR lto/97290 * simple-object-elf.c (simple_object_elf_copy_lto_debug_sections): Use sh_link of a .symtab_shndx section. 2021-01-05 Alan Modra * config.guess: Import from upstream. * config.sub: Likewise. 2020-12-16 Martin Liska Tom de Vries * gdb/debuginfod-support.c (struct user_data): Remove has_printed field. Add meter field. (progressfn): Print progress using meter. 2020-12-02 Enze Li * .gitignore: Add gnu global outputs. 2020-12-02 Simon Marchi * .gitignore: Sync with gcc. 2020-10-26 Andreas Rammhold * src-release.sh: Use sha256sum instead of md5sum. 2020-10-14 Andrew Burgess * Makefile.in: Rebuild. * Makefile.def: Make distclean-gnulib depend on distclean-gdb and distclean-gdbserver. 2020-07-24 Aaron Merey * configure: Rebuild. * configure.ac: Remove AC_DEBUGINFOD. 2020-07-04 Nick Clifton Binutils 2.35 branch created. 2020-04-21 Stephen Casner PR 25830 * configure.ac (noconfigdirs): Exclude gdb & gprof for pdp11. * configure: Rebuild. 2020-03-12 Tom Tromey * Makefile.in: Rebuild. * Makefile.def (gdbserver): Depend on gdbsupport. 2020-03-12 Tom Tromey * Makefile.in: Rebuild. * Makefile.def (gdbsupport): Don't depend on bfd. 2020-03-12 Tom Tromey * Makefile.in: Rebuild. * Makefile.def (gdbsupport): Depend on intl. 2020-02-17 Tom Tromey * configure: Rebuild. * configure.ac (configdirs): Add gnulib and gdbsupport when building gdbserver. 2020-02-14 Tom Tromey * Makefile.in: Rebuild. * Makefile.def: Make gdbserver require gnulib and libiberty. 2020-02-07 Tom Tromey Pedro Alves * src-release.sh (GDB_SUPPORT_DIRS): Add gdbserver. * gdbserver: New directory, moved from gdb/gdbserver. * configure.ac (host_tools): Add gdbserver. Only build gdbserver on certain systems. * Makefile.in, configure: Rebuild. * Makefile.def (host_modules, dependencies): Add gdbserver. * MAINTAINERS: Add gdbserver. 2020-01-28 Sergio Durigan Junior * src-release.sh (getver): Look for gdbsupport's create-version.sh script at the current directory if tool is "gdb". 2020-01-19 Simon Marchi * remote-sim.c (gdbsim_target::wait): Return sim_data->remote_sim_ptid instead of inferior_ptid. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 6.8 from gprof.texi. d5 1 a5 1 Copyright (C) 1988-2022 Free Software Foundation, Inc. d30 1 a30 1 This manual is for 'gprof' (GNU Binutils) version 2.39. d220 4 a223 5 'gmon.out', its contents are overwritten. You can rename the file afterwards if you are concerned that it may be overwritten. If your system libc allows you may be able to write the profile data under a different name. Set the GMON_OUT_PREFIX environment variable; this name will be appended with the PID of the running program. d1678 1 a1678 1 magic cookie that allows one to check whether a data file really is a d2431 26 a2456 26 Node: Invoking11482 Node: Output Options12897 Node: Analysis Options19989 Node: Miscellaneous Options23909 Node: Deprecated Options25163 Node: Symspecs27226 Node: Output29052 Node: Flat Profile30092 Node: Call Graph35045 Node: Primary38277 Node: Callers40865 Node: Subroutines42983 Node: Cycles44824 Node: Line-by-line51601 Node: Annotated Source55677 Node: Inaccuracy58675 Node: Sampling Error58933 Node: Assumptions61837 Node: How do I?63307 Node: Incompatibilities64864 Node: Details66358 Node: Implementation66751 Node: File Format72650 Node: Internals76942 Node: Debugging85432 Node: GNU Free Documentation License87022 a2458 5  Local Variables: coding: utf-8 End: @ 1.1.1.10 log @Import binutils-2.42 (last was 2.39) 2024-01-15 Nick Clifton * 2.42 branch point. 2023-11-15 Arsen Arsenović * intl: Remove directory. Replaced with out-of-tree GNU gettext. * .gitignore: Add '/gettext*'. * configure.ac (host_libs): Replace intl with gettext. (hbaseargs, bbaseargs, baseargs): Split baseargs into {h,b}baseargs. (skip_barg): New flag. Skips appending current flag to bbaseargs. : Exempt --with-libintl-{type,prefix} from target and build machine argument passing. * configure: Regenerate. * Makefile.def (host_modules): Replace intl module with gettext module. (configure-ld): Depend on configure-gettext. * Makefile.in: Regenerate. * src-release.sh: Remove references to the intl/ directory. 2023-07-03 Nick Clifton 2.41 Branch Point. 2023-06-26 Nick Clifton * Import these updates to the config scripts commit 4ad4bb7c30aca1e705448ba8d51a210bbd47bb52 Author: Paul Eggert Date: Fri Jun 23 09:55:10 2023 -0700 Quote 'like this', not `like this'. commit 63acb96f92473ceb5e21d873d7c0aee266b3d6d3 Author: Paul Eggert Date: Sat Jan 21 00:15:01 2023 -0600 Fix config.sub spelling typo for "athlon" commit 4ce12a5c9125cedc0d0ba584444a6865396923ec Author: Dmitry V. Levin Date: Sun Jan 1 08:00:00 2023 +0000 Update copyright years commit c397e2c040bce50bcdccb131f90115ba7e8bfc19 Author: Arsen Arsenovi Date: Sat Sep 17 23:34:48 2022 +0200 config.sub: add linux-mlibc targets commit 9f9f9b0b13197269848c76e3e057a3ed0680b4bf Author: Arsen Arsenovi Date: Sat Sep 17 23:34:47 2022 +0200 config.guess: support running on Managarm systems commit 87e6687749da7bb2ab158a79fa83721c19ed9246 Author: Arsen Arsenovi Date: Sat Sep 17 23:34:46 2022 +0200 config.sub: add managarm-{mlibc,kernel} targets commit 20403c5701973a4cbd7e0b4bbeb627fcd424a0f1 Author: Xiaotian Wu Date: Mon Aug 1 16:05:29 2022 +0800 Remove loongarchx32 commit 02ba26b218d3d3db6c56e014655faf463cefa983 Author: Alexander von Gluck IV Date: Wed May 25 15:43:13 2022 -0500 config.guess: Update Haiku guesses commit f56a7140386d08a531bcfd444d632b28c61a6329 Author: Bruno Haible Date: Sun May 8 19:08:08 2022 +0200 config.guess (x86_64:Linux:*:*): Detect 32-bit ABI. 2023-04-20 Nick Clifton * SECURITY.txt: New file. * src-release.sh (DEVO_SUPPORT): Add SECURITY.txt. 2022-12-31 Nick Clifton * 2.40 binutils branch created. 2022-10-10 Nick Clifton * src-release.sh: Add "-r " option to create reproducible tarballs based upon a fixed timestamp of . * binutils/README-how-to-make-a-release: Add a line showing how to use -r when creating a binutils release. 2022-10-04 Nick Clifton * README-maintainer-mode: Add a minimum version of dejagnu requirement. 2022-09-08 Nick Clifton * README-maintainer-mode: Update minimum version of gettext required. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 7.0.2 from gprof.texi. d5 1 a5 1 Copyright © 1988-2024 Free Software Foundation, Inc. d11 2 a12 2 Texts. A copy of the license is included in the section entitled “GNU Free Documentation License”. d16 1 a16 1 * gprof: (gprof). Profiling your program’s execution d25 1 a25 1 This manual describes the GNU profiler, ‘gprof’, and how you can use it d28 1 a28 1 programs. GNU ‘gprof’ was written by Jay Fenlason. d30 1 a30 1 This manual is for ‘gprof’ (GNU Binutils) version 2.42. d34 1 a34 1 the section entitled “GNU Free Documentation License”. d42 1 a42 1 * Invoking:: How to run ‘gprof’, and its options d44 1 a44 1 * Output:: Interpreting ‘gprof’’s output d48 1 a48 1 * Incompatibilities:: (between GNU ‘gprof’ and Unix ‘gprof’.) d70 1 a70 1 data. If you don’t use some feature of your program while it is being d75 1 a75 1 • You must compile and link your program with profiling enabled. d78 1 a78 1 • You must execute your program to generate a profile data file. d81 1 a81 1 • You must run ‘gprof’ to analyze the profile data. *Note ‘gprof’ d88 1 a88 1 The “flat profile” shows how much time your program spent in each d93 1 a93 1 The “call graph” shows, for each function, which functions called it, d99 1 a99 1 The “annotated source” listing is a copy of the program’s source d116 1 a116 1 To compile a source file for profiling, specify the ‘-pg’ option when d120 3 a122 3 To link the program for profiling, if you use a compiler such as ‘cc’ to do the linking, simply specify ‘-pg’ in addition to your usual options. The same option, ‘-pg’, alters either compilation or linking d128 1 a128 1 The ‘-pg’ option also works with a command that both compiles and d133 1 a133 1 Note: The ‘-pg’ option must be part of your compilation options as d135 1 a135 1 gathered and when you run ‘gprof’ you will get an error message like d140 1 a140 1 If you add the ‘-Q’ switch to suppress the printing of the call graph d152 3 a154 3 If you run the linker ‘ld’ directly instead of through a compiler such as ‘cc’, you may have to specify a profiling startup file ‘gcrt0.o’ as the first input file instead of the usual startup file ‘crt0.o’. In d156 1 a156 1 ‘libc_p.a’, by writing ‘-lc_p’ instead of the usual ‘-lc’. This is not d158 1 a158 1 information for standard library functions such as ‘read’ and ‘open’. d169 2 a170 2 code, which for ‘gcc’ users can be done via the ‘-static’ or ‘-static-libgcc’ command-line option. For example: d174 3 a176 3 If you compile only some of the modules of the program with ‘-pg’, you can still profile the program, but you won’t get complete information about the modules that were compiled without ‘-pg’. The d180 1 a180 1 that the ‘calls’ field for the functions will be blank), but will d184 1 a184 1 ‘gcov’ tool instead of ‘gprof’. See that tool’s manual or info pages d187 3 a189 3 Note, older versions of ‘gcc’ produce line-by-line profiling information that works with ‘gprof’ rather than ‘gcov’ so there is still support for displaying this kind of information in ‘gprof’. *Note d192 1 a192 1 It also worth noting that ‘gcc’ implements a ‘-finstrument-functions’ d205 1 a205 1 generate the information that ‘gprof’ needs. Simply run the program as d211 2 a212 2 The way you run the program—the arguments and input that you give it—may have a dramatic effect on what the profile information shows. d219 2 a220 2 ‘gmon.out’ just before exiting. If there is already a file called ‘gmon.out’, its contents are overwritten. You can rename the file d226 3 a228 3 In order to write the ‘gmon.out’ file properly, your program must exit normally: by returning from ‘main’ or by calling ‘exit’. Calling the low-level function ‘_exit’ does not write the profile data, and d231 1 a231 1 The ‘gmon.out’ file is written in the program’s _current working d233 2 a234 2 ‘chdir’, the ‘gmon.out’ file will be left in the last directory your program ‘chdir’’d to. If you don’t have permission to write in this d238 1 a238 1 called ‘bb.out’. This file, if present, contains an human-readable d240 4 a243 4 appearance of a human-readable ‘bb.out’ means the basic-block counts didn’t get written into ‘gmon.out’. The Perl script ‘bbconv.pl’, included with the ‘gprof’ source distribution, will convert a ‘bb.out’ file into a format readable by ‘gprof’. Invoke it like this: d247 2 a248 2 This translates the information in ‘bb.out’ into a form that ‘gprof’ can understand. But you still need to tell ‘gprof’ about the existence d250 1 a250 1 ‘gprof’ command line, _along with ‘gmon.out’_, like this: d257 1 a257 1 4 ‘gprof’ Command Summary d260 2 a261 2 After you have a profile data file ‘gmon.out’, you can run ‘gprof’ to interpret the information in it. The ‘gprof’ program prints a flat d263 1 a263 1 redirect the output of ‘gprof’ into a file with ‘>’. d265 1 a265 1 You run ‘gprof’ like this: d271 2 a272 2 If you omit the executable file name, the file ‘a.out’ is used. If you give no profile data file name, the file ‘gmon.out’ is used. If any d284 2 a285 2 * Output Options:: Controlling ‘gprof’’s output style * Analysis Options:: Controlling how ‘gprof’ analyzes its data d297 1 a297 1 These options specify which of several output formats ‘gprof’ should d300 1 a300 1 Many of these options take an optional “symspec” to specify functions d305 1 a305 1 Specifying any of these options overrides the default (‘-p -q’), d308 3 a310 3 ‘-A[SYMSPEC]’ ‘--annotated-source[=SYMSPEC]’ The ‘-A’ option causes ‘gprof’ to print annotated source code. If d314 3 a316 3 ‘-b’ ‘--brief’ If the ‘-b’ option is given, ‘gprof’ doesn’t print the verbose d321 3 a323 6 ‘-B’ The ‘-B’ option causes ‘gprof’ to print the call graph analysis. ‘-C[SYMSPEC]’ ‘--exec-counts[=SYMSPEC]’ The ‘-C’ option causes ‘gprof’ to print a tally of functions and d328 1 a328 1 specifying the ‘-l’ option, along with ‘-C’, will cause basic-block d331 3 a333 3 ‘-i’ ‘--file-info’ The ‘-i’ option causes ‘gprof’ to display summary information about d337 3 a339 3 ‘-I DIRS’ ‘--directory-path=DIRS’ The ‘-I’ option specifies a list of search directories in which to d344 4 a347 4 ‘-J[SYMSPEC]’ ‘--no-annotated-source[=SYMSPEC]’ The ‘-J’ option causes ‘gprof’ not to print annotated source code. If SYMSPEC is specified, ‘gprof’ prints annotated source, but d350 2 a351 2 ‘-L’ ‘--print-path’ d353 1 a353 1 suppressed. The ‘-L’ option causes ‘gprof’ to print the full d358 3 a360 3 ‘-p[SYMSPEC]’ ‘--flat-profile[=SYMSPEC]’ The ‘-p’ option causes ‘gprof’ to print a flat profile. If SYMSPEC d364 4 a367 4 ‘-P[SYMSPEC]’ ‘--no-flat-profile[=SYMSPEC]’ The ‘-P’ option causes ‘gprof’ to suppress printing a flat profile. If SYMSPEC is specified, ‘gprof’ prints a flat profile, but d370 3 a372 3 ‘-q[SYMSPEC]’ ‘--graph[=SYMSPEC]’ The ‘-q’ option causes ‘gprof’ to print the call graph analysis. d376 4 a379 4 ‘-Q[SYMSPEC]’ ‘--no-graph[=SYMSPEC]’ The ‘-Q’ option causes ‘gprof’ to suppress printing the call graph. If SYMSPEC is specified, ‘gprof’ prints a call graph, but excludes d382 3 a384 3 ‘-t’ ‘--table-length=NUM’ The ‘-t’ option causes the NUM most active source lines in each d388 2 a389 2 ‘-y’ ‘--separate-files’ d391 1 a391 1 ‘gprof’ prints annotated source files to standard-output. If this d393 9 a401 9 ‘path/FILENAME’ is generated in the file ‘FILENAME-ann’. If the underlying file system would truncate ‘FILENAME-ann’ so that it overwrites the original ‘FILENAME’, ‘gprof’ generates annotated source in the file ‘FILENAME.ann’ instead (if the original file name has an extension, that extension is _replaced_ with ‘.ann’). ‘-Z[SYMSPEC]’ ‘--no-exec-counts[=SYMSPEC]’ The ‘-Z’ option causes ‘gprof’ not to print a tally of functions d405 3 a407 3 ‘-r’ ‘--function-ordering’ The ‘--function-ordering’ option causes ‘gprof’ to print a d417 3 a419 3 ‘-R MAP_FILE’ ‘--file-ordering MAP_FILE’ The ‘--file-ordering’ option causes ‘gprof’ to print a suggested .o d425 1 a425 1 Use of the ‘-a’ argument is highly recommended with this option. d429 1 a429 1 similar to the output of the program ‘nm’. d441 2 a442 2 To create a MAP_FILE with GNU ‘nm’, type a command like ‘nm --extern-only --defined-only -v --print-file-name program-name’. d444 3 a446 3 ‘-T’ ‘--traditional’ The ‘-T’ option causes ‘gprof’ to print its output in “traditional” d449 2 a450 2 ‘-w WIDTH’ ‘--width=WIDTH’ d454 2 a455 2 ‘-x’ ‘--all-lines’ d460 1 a460 1 similar to ‘tcov’’s ‘-a’. d462 2 a463 2 ‘--demangle[=STYLE]’ ‘--no-demangle’ d466 1 a466 1 ‘--no-demangle’ option may be used to turn off demangling. d477 3 a479 3 ‘-a’ ‘--no-static’ The ‘-a’ option causes ‘gprof’ to suppress the printing of d488 3 a490 3 ‘-c’ ‘--static-call-graph’ The ‘-c’ option causes the call graph of the program to be d500 1 a500 1 graph with call counts of ‘0’. d502 3 a504 3 ‘-D’ ‘--ignore-non-functions’ The ‘-D’ option causes ‘gprof’ to ignore symbols which are not d509 2 a510 2 ‘-k FROM/TO’ The ‘-k’ option allows you to delete from the call graph any arcs d513 3 a515 3 ‘-l’ ‘--line’ The ‘-l’ option enables line-by-line profiling, which causes d518 2 a519 2 compiled by older versions of the ‘gcc’ compiler. Newer versions of ‘gcc’ are designed to work with the ‘gcov’ tool instead. d525 1 a525 1 significantly increases the running time of ‘gprof’, and magnifies d529 2 a530 2 ‘--inline-file-names’ This option causes ‘gprof’ to print the source file after each d532 1 a532 1 to the file is printed if used with the ‘-L’ option. d534 2 a535 2 ‘-m NUM’ ‘--min-count=NUM’ d539 3 a541 3 ‘-nSYMSPEC’ ‘--time=SYMSPEC’ The ‘-n’ option causes ‘gprof’, in its call graph analysis, to only d544 3 a546 3 ‘-NSYMSPEC’ ‘--no-time=SYMSPEC’ The ‘-n’ option causes ‘gprof’, in its call graph analysis, not to d549 5 a553 5 ‘-SFILENAME’ ‘--external-symbol-table=FILENAME’ The ‘-S’ option causes ‘gprof’ to read an external symbol table file, such as ‘/proc/kallsyms’, rather than read the symbol table from the given object file (the default is ‘a.out’). This is d556 3 a558 3 ‘-z’ ‘--display-unused-functions’ If you give the ‘-z’ option, ‘gprof’ will mention all functions in d560 1 a560 1 no time spent in them. This is useful in conjunction with the ‘-c’ d569 4 a572 4 ‘-d[NUM]’ ‘--debug[=NUM]’ The ‘-d NUM’ option specifies debugging options. If NUM is not specified, enable all debugging. *Note Debugging ‘gprof’: d575 3 a577 3 ‘-h’ ‘--help’ The ‘-h’ option prints command line usage. d579 2 a580 2 ‘-ONAME’ ‘--file-format=NAME’ d582 1 a582 1 are ‘auto’ (the default), ‘bsd’, ‘4.4bsd’, ‘magic’, and ‘prof’ (not d585 3 a587 3 ‘-s’ ‘--sum’ The ‘-s’ option causes ‘gprof’ to summarize the information in the d589 2 a590 2 called ‘gmon.sum’, which contains all the information from the profile data files that ‘gprof’ read in. The file ‘gmon.sum’ may d592 1 a592 1 the data in the other input files into ‘gmon.sum’. d594 2 a595 2 Eventually you can run ‘gprof’ again without ‘-s’ to analyze the cumulative data in the file ‘gmon.sum’. d597 3 a599 3 ‘-v’ ‘--version’ The ‘-v’ flag causes ‘gprof’ to print the current version number, d610 2 a611 2 ‘-e FUNCTION_NAME’ The ‘-e FUNCTION’ option tells ‘gprof’ to not print information d614 3 a616 3 functions that call it, but its index number will be shown as ‘[not printed]’. More than one ‘-e’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-e’ option. d618 2 a619 2 ‘-E FUNCTION_NAME’ The ‘-E FUNCTION’ option works like the ‘-e’ option, but time spent d622 2 a623 2 call graph. More than one ‘-E’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-E’ option. d625 2 a626 2 ‘-f FUNCTION_NAME’ The ‘-f FUNCTION’ option causes ‘gprof’ to limit the call graph to d628 2 a629 2 children...). More than one ‘-f’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-f’ option. d631 2 a632 2 ‘-F FUNCTION_NAME’ The ‘-F FUNCTION’ option works like the ‘-f’ option, but only time d635 3 a637 3 call graph. More than one ‘-F’ option may be given; only one FUNCTION_NAME may be indicated with each ‘-F’ option. The ‘-F’ option overrides the ‘-E’ option. d639 2 a640 2 Note that only one function can be specified with each ‘-e’, ‘-E’, ‘-f’ or ‘-F’ option. To specify more than one function, use multiple d646 1 a646 1 ‘foo’ or ‘bar’ and were not reachable from ‘boring’. d655 1 a655 1 using “symspecs” (symbol specifications), which observe the following d665 3 a667 3 ‘main.c’ Selects everything in file ‘main.c’—the dot in the string tells ‘gprof’ to interpret the string as a filename, rather than as a d669 2 a670 2 a trailing colon should be specified. For example, ‘odd:’ is interpreted as the file named ‘odd’. d672 2 a673 2 ‘main’ Selects all functions named ‘main’. d682 2 a683 2 necessary to add a leading colon to the name. For example, ‘:.mul’ selects function ‘.mul’. d686 2 a687 2 ‘gprof’ will normally not print these underscores. When you name a symbol in a symspec, you should type it exactly as ‘gprof’ prints d689 2 a690 2 ‘_main’ from your ‘main’ function, ‘gprof’ still prints it as ‘main’ in its output, so you should use ‘main’ in symspecs. d692 2 a693 2 ‘main.c:main’ Selects function ‘main’ in file ‘main.c’. d695 2 a696 2 ‘main.c:134’ Selects line 134 in file ‘main.c’. d701 1 a701 1 5 Interpreting ‘gprof’’s Output d704 1 a704 1 ‘gprof’ can produce several different output styles, the most important d717 1 a717 1 * Line-by-line:: ‘gprof’ can analyze individual source code lines d727 2 a728 2 The “flat profile” shows the total amount of time your program spent executing each function. Unless the ‘-z’ option is given, functions d731 1 a731 1 and didn’t run long enough to show up on the program counter histogram, d759 1 a759 1 functions ‘mcount’ and ‘profil’ are part of the profiling apparatus and d764 1 a764 1 much time each sample counted as. This “sampling period” estimates the d768 2 a769 2 program’s total execution time was 0.06 seconds, as indicated by the ‘cumulative seconds’ field. Since each sample counted for 0.01 seconds, d771 3 a773 3 samples occurred while the program was in the ‘open’ function, as indicated by the ‘self seconds’ field. Each of the other four samples occurred one each in ‘offtime’, ‘memccpy’, ‘write’, and ‘mcount’. Since d775 2 a776 2 particularly reliable. In another run, the ‘self seconds’ field for ‘mcount’ might well be ‘0.00’ or ‘0.02’. *Note Statistical Sampling d779 2 a780 2 The remaining functions in the listing (those whose ‘self seconds’ field is ‘0.00’) didn’t appear in the histogram samples at all. d782 1 a782 1 they are listed, sorted in decreasing order by the ‘calls’ field. d789 1 a789 1 ‘% time’ d793 1 a793 1 ‘cumulative seconds’ d798 1 a798 1 ‘self seconds’ d802 1 a802 1 ‘calls’ d806 1 a806 1 compiled with profiling enabled), the “calls” field is blank. d808 1 a808 1 ‘self ms/call’ d813 1 a813 1 ‘total ms/call’ d820 1 a820 1 ‘name’ d822 1 a822 1 this field alphabetically after the “self seconds” and “calls” d831 1 a831 1 The “call graph” shows how much time was spent in each function and its d837 1 a837 1 same ‘gprof’ run as the flat profile example in the previous section. d868 1 a868 1 The lines full of dashes divide this table into “entries”, one for d875 1 a875 1 called “children” when we speak of the call graph). d880 1 a880 1 The internal profiling function ‘mcount’ (*note The Flat Profile: d885 3 a887 3 * Primary:: Details of the primary line’s contents. * Callers:: Details of caller-lines’ contents. * Subroutines:: Details of subroutine-lines’ contents. d889 1 a889 1 such as ‘a’ calls ‘b’ calls ‘a’... d897 1 a897 1 The “primary line” in a call graph entry is the line that describes the d902 1 a902 1 ‘report’ in our main example, together with the heading line that shows d911 1 a911 1 ‘index’ d921 1 a921 1 ‘% time’ d930 1 a930 1 ‘self’ d932 1 a932 1 should be identical to the number printed in the ‘seconds’ field d935 1 a935 1 ‘children’ d938 1 a938 1 ‘self’ and ‘children’ entries of the children listed directly below d941 1 a941 1 ‘called’ d945 1 a945 1 separated by a ‘+’. The first number counts non-recursive calls, d948 2 a949 2 In the example above, the function ‘report’ was called once from ‘main’. d951 1 a951 1 ‘name’ d956 1 a956 1 is printed between the function’s name and the index number (*note d958 1 a958 1 example, if function ‘gnurr’ is part of cycle number one, and has d966 1 a966 1 5.2.2 Lines for a Function’s Callers d969 2 a970 2 A function’s entry has a line for each function it was called by. These lines’ fields correspond to the fields of the primary line, but their d974 1 a974 1 ‘report’, the primary line and one caller-line preceding it, together d982 2 a983 2 Here are the meanings of the fields in the caller-line for ‘report’ called from ‘main’: d985 14 a998 14 ‘self’ An estimate of the amount of time spent in ‘report’ itself when it was called from ‘main’. ‘children’ An estimate of the amount of time spent in subroutines of ‘report’ when ‘report’ was called from ‘main’. The sum of the ‘self’ and ‘children’ fields is an estimate of the amount of time spent within calls to ‘report’ from ‘main’. ‘called’ Two numbers: the number of times ‘report’ was called from ‘main’, followed by the total number of non-recursive calls to ‘report’ d1001 3 a1003 3 ‘name and index number’ The name of the caller of ‘report’ to which this line applies, followed by the caller’s index number. d1006 1 a1006 1 ‘gprof’ request the omission of certain functions. When a caller d1014 2 a1015 2 dummy caller-line is printed which has ‘’ as the “caller’s name” and all other fields blank. This can happen for signal handlers. d1020 1 a1020 1 5.2.3 Lines for a Function’s Subroutines d1023 2 a1024 2 A function’s entry has a line for each of its subroutines—in other words, a line for each other function that it called. These lines’ d1029 1 a1029 1 ‘main’, the primary line and a line for a subroutine, together with the d1037 2 a1038 2 Here are the meanings of the fields in the subroutine-line for ‘main’ calling ‘report’: d1040 16 a1055 16 ‘self’ An estimate of the amount of time spent directly within ‘report’ when ‘report’ was called from ‘main’. ‘children’ An estimate of the amount of time spent in subroutines of ‘report’ when ‘report’ was called from ‘main’. The sum of the ‘self’ and ‘children’ fields is an estimate of the total time spent in calls to ‘report’ from ‘main’. ‘called’ Two numbers, the number of calls to ‘report’ from ‘main’ followed by the total number of non-recursive calls to ‘report’. This ratio is used to determine how much of ‘report’’s ‘self’ and ‘children’ time gets credited to ‘main’. *Note Estimating ‘children’ Times: d1058 3 a1060 3 ‘name’ The name of the subroutine of ‘main’ to which this line applies, followed by the subroutine’s index number. d1071 1 a1071 1 The graph may be complicated by the presence of “cycles of recursion” in d1074 2 a1075 2 function. For example: if ‘a’ calls ‘b’, and ‘b’ calls ‘a’, then ‘a’ and ‘b’ form a cycle. d1078 4 a1081 4 they belong to the same cycle. If ‘a’ and ‘b’ call each other and ‘b’ and ‘c’ call each other, all three make one cycle. Note that even if ‘b’ only calls ‘a’ if it was not called from ‘a’, ‘gprof’ cannot determine this, so ‘a’ and ‘b’ are still considered a cycle. d1085 1 a1085 1 graph it is followed by ‘’. d1088 4 a1091 4 call graph paradoxical. The “time spent in children” of ‘a’ should include the time spent in its subroutine ‘b’ and in ‘b’’s subroutines—but one of ‘b’’s subroutines is ‘a’! How much of ‘a’’s time should be included in the children of ‘a’, when ‘a’ is indirectly d1094 1 a1094 1 The way ‘gprof’ resolves this paradox is by creating a single entry d1097 1 a1097 1 “subroutines” of the cycle are the individual functions of the cycle, d1099 1 a1099 1 “callers” of the cycle are the functions, outside the cycle, that called d1103 2 a1104 2 containing functions ‘a’ and ‘b’. The cycle was entered by a call to ‘a’ from ‘main’; both ‘a’ and ‘b’ called ‘c’. d1126 2 a1127 2 for ‘main’, which calls ‘a’, and an entry for ‘c’, with callers ‘a’ and ‘b’.) d1160 2 a1161 2 The ‘self’ field of the cycle’s primary line is the total time spent in all the functions of the cycle. It equals the sum of the ‘self’ d1165 6 a1170 6 The ‘children’ fields of the cycle’s primary line and subroutine lines count only subroutines outside the cycle. Even though ‘a’ calls ‘b’, the time spent in those calls to ‘b’ is not counted in ‘a’’s ‘children’ time. Thus, we do not encounter the problem of what to do when the time in those calls to ‘b’ includes indirect recursive calls back to ‘a’. d1172 1 a1172 1 The ‘children’ field of a caller-line in the cycle’s entry estimates d1177 1 a1177 1 The ‘called’ field in the primary line for the cycle has two numbers: d1184 2 a1185 2 The ‘called’ field of a subroutine-line for a cycle member in the cycle’s entry says how many time that function was called from functions d1187 1 a1187 1 primary line’s ‘called’ field. d1192 1 a1192 1 called from each other function in the cycle. The ‘self’ and ‘children’ d1202 1 a1202 1 ‘gprof’’s ‘-l’ option causes the program to perform “line-by-line” d1205 2 a1206 2 programs compiled with older versions of the ‘gcc’ compiler. Newer versions of ‘gcc’ use a different program - ‘gcov’ - to display d1209 2 a1210 2 With the older versions of ‘gcc’ the program usually has to be compiled with a ‘-g’ option, in addition to ‘-pg’, in order to generate d1212 1 a1212 1 versions of ‘gcc’ the program had to be compiled with the ‘-a’ d1216 2 a1217 2 mode. The call graph isn’t as useful as normal, since the current version of ‘gprof’ does not propagate call graph arcs from source code d1221 3 a1223 3 Here is a section of ‘gprof’’s output, without line-by-line profiling. Note that ‘ct_init’ accounted for four histogram hits, and 13327 calls to ‘init_block’. d1247 3 a1249 3 Now let’s look at some of ‘gprof’’s output from the same program run, this time with line-by-line profiling enabled. Note that ‘ct_init’’s four histogram hits are broken down into four lines of source code—one d1251 1 a1251 1 note how ‘ct_init’’s 13327 calls to ‘init_block’ are broken down into d1294 4 a1297 4 ‘gprof’’s ‘-A’ option triggers an annotated source listing, which lists the program’s source code, each function labeled with the number of times it was called. You may also need to specify the ‘-I’ option, if ‘gprof’ can’t find the source code files. d1299 1 a1299 1 With older versions of ‘gcc’ compiling with ‘gcc ... -g -pg -a’ d1301 3 a1303 3 function counting code. This enables ‘gprof’ to determine how many times each line of code was executed. With newer versions of ‘gcc’ support for displaying basic-block counts is provided by the ‘gcov’ d1330 4 a1333 4 ‘updcrc’ has at least five basic-blocks. One is the function itself. The ‘if’ statement on line 9 generates two more basic-blocks, one for each branch of the ‘if’. A fourth basic-block results from the ‘if’ on line 13, and the contents of the ‘do’ loop form the fifth basic-block. d1338 3 a1340 3 ‘gprof -l -A’. The ‘-x’ option is also helpful, to ensure that each line of code is labeled at least once. Here is ‘updcrc’’s annotated source listing for a sample ‘gzip’ run: d1363 2 a1364 2 each branch of the ‘if’ statement. The body of the ‘do’ loop was executed a total of 26312 times. Note how the ‘while’ statement is d1372 1 a1372 1 6 Inaccuracy of ‘gprof’ Output d1386 1 a1386 1 The run-time figures that ‘gprof’ gives you are based on a sampling d1402 1 a1402 1 The “sampling period” that is printed at the beginning of the flat d1409 1 a1409 1 period is 0.01 seconds and ‘foo’’s run-time is 1 second, N is 100 d1411 1 a1411 1 error in ‘foo’’s run-time is 0.1 seconds (10*0.01 seconds), or ten d1413 2 a1414 2 seconds and ‘bar’’s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in ‘bar’’s run-time is 1 second, d1420 1 a1420 1 information. If the program’s _total_ run-time is large, a small d1422 1 a1422 1 insignificant fraction of the whole program’s time. Usually this means d1427 1 a1427 1 the data from several runs, using the ‘-s’ option of ‘gprof’. Here is d1432 1 a1432 1 2. Issue the command ‘mv gmon.out gmon.sum’. d1436 1 a1436 1 4. Merge the new data in ‘gmon.out’ into ‘gmon.sum’ with this command: d1449 1 a1449 1 6.2 Estimating ‘children’ Times d1452 2 a1453 2 Some of the figures in the call graph are estimates—for example, the ‘children’ time values and all the time figures in caller and subroutine d1457 1 a1457 1 profile data itself. Instead, ‘gprof’ estimates them by making an d1461 3 a1463 3 any function ‘foo’ is not correlated with who called ‘foo’. If ‘foo’ used 5 seconds in all, and 2/5 of the calls to ‘foo’ came from ‘a’, then ‘foo’ contributes 2 seconds to ‘a’’s ‘children’ time, by assumption. d1466 6 a1471 6 far from true. Suppose that ‘foo’ returns very quickly when its argument is zero; suppose that ‘a’ always passes zero as an argument, while other callers of ‘foo’ pass other arguments. In this program, all the time spent in ‘foo’ is in the calls from callers other than ‘a’. But ‘gprof’ has no way of knowing this; it will blindly and incorrectly charge 2 seconds of time in ‘foo’ to the children of ‘a’. d1473 1 a1473 1 We hope some day to put more complete data into ‘gmon.out’, so that d1486 1 a1486 1 Because ‘gprof’ can only report call times and counts by function, d1490 1 a1490 1 artificial hot spots since compiling with ‘-pg’ adds a significant d1496 1 a1496 1 Use the ‘gcov’ program. d1500 1 a1500 1 Use ‘gprof -l’ and lookup the function in the call graph. The d1523 1 a1523 1 8 Incompatibilities with Unix ‘gprof’ d1526 1 a1526 1 GNU ‘gprof’ and Berkeley Unix ‘gprof’ use the same data file ‘gmon.out’, d1530 1 a1530 1 • GNU ‘gprof’ uses a new, generalized file format with support for d1532 1 a1532 1 cookie and version number allows ‘gprof’ to easily identify new d1536 3 a1538 3 • For a recursive function, Unix ‘gprof’ lists the function as a parent and as a child, with a ‘calls’ field that lists the number of recursive calls. GNU ‘gprof’ omits these lines and puts the d1541 2 a1542 2 • When a function is suppressed from the call graph with ‘-e’, GNU ‘gprof’ still lists it as a subroutine of functions that call it. d1544 2 a1545 2 • GNU ‘gprof’ accepts the ‘-k’ with its argument in the form ‘from/to’, instead of ‘from to’. d1547 2 a1548 2 • In the annotated source listing, if there are multiple basic blocks on the same line, GNU ‘gprof’ prints all of their counts, separated d1551 2 a1552 2 • The blurbs, field widths, and output formats are different. GNU ‘gprof’ prints blurbs after the tables, so that you can see the d1564 3 a1566 3 * File Format:: Format of ‘gmon.out’ files * Internals:: ‘gprof’’s internal operation * Debugging:: Using ‘gprof’’s ‘-d’ option d1579 2 a1580 2 the ‘-pg’ option, which causes every function to call ‘mcount’ (or ‘_mcount’, or ‘__mcount’, depending on the OS and compiler) as one of d1583 1 a1583 1 The ‘mcount’ routine, included in the profiling library, is d1585 1 a1585 1 parent routine (the child) and its parent’s parent. This is typically d1588 1 a1588 1 machine-dependent operation, ‘mcount’ itself is typically a short d1590 4 a1593 4 and then calls ‘__mcount_internal’ (a normal C function) with two arguments—‘frompc’ and ‘selfpc’. ‘__mcount_internal’ is responsible for maintaining the in-memory call graph, which records ‘frompc’, ‘selfpc’, and the number of times each of these call arcs was traversed. d1596 1 a1596 1 (‘__builtin_return_address’), which allows a generic ‘mcount’ function d1600 1 a1600 1 ‘mcount’ is used for performance reasons. d1604 2 a1605 2 same as in the usual C library, but they were compiled with ‘-pg’. If you link your program with ‘gcc ... -pg’, it automatically uses the d1615 2 a1616 2 provide a ‘profil()’ system call, which registers a memory array with the kernel, along with a scale factor that determines how the program’s d1626 1 a1626 1 earlier), do not provide a ‘profil()’ system call. On such a system, d1628 1 a1628 1 the process (typically via ‘setitimer()’), which then performs the same d1636 5 a1640 5 either calls ‘profil()’ or sets up a clock signal handler. This routine (‘monstartup’) can be invoked in several ways. On Linux systems, a special profiling startup file ‘gcrt0.o’, which invokes ‘monstartup’ before ‘main’, is used instead of the default ‘crt0.o’. Use of this special startup file is one of the effects of using ‘gcc ... -pg’ to d1642 2 a1643 2 ‘mcount’ routine, when it is invoked for the first time (typically when ‘main’ is called), calls ‘monstartup’. d1645 1 a1645 1 If the compiler’s ‘-a’ option was used, basic-block counting is also d1648 1 a1648 1 basic-block begins (i.e., when an ‘if’ statement appears), an extra d1655 3 a1657 3 The profiling library also includes a function (‘mcleanup’) which is typically registered using ‘atexit()’ to be called as the program exits, and is responsible for writing the file ‘gmon.out’. Profiling is turned d1661 1 a1661 1 The output from ‘gprof’ gives no indication of parts of your program d1663 2 a1664 2 of the program counter are taken at fixed intervals of the program’s run time. Therefore, the time measurements in ‘gprof’ output say nothing d1667 1 a1667 1 memory at once may run very slowly due to thrashing, but ‘gprof’ will d1669 1 a1669 1 the advantage that the amount of load due to other users won’t directly d1680 2 a1681 2 ‘gprof’ file. Furthermore, it does not provide a version number, thus rendering changes to the file format almost impossible. GNU ‘gprof’ d1683 1 a1683 1 compatibility, GNU ‘gprof’ continues to support the old BSD-derived d1688 1 a1688 1 The new file format is defined in header file ‘gmon_out.h’. It d1692 1 a1692 1 the profile was collected. GNU ‘gprof’ adapts automatically to the d1699 1 a1699 1 reading a file, GNU ‘gprof’ will ensure records of the same type are d1715 1 a1715 1 would specify the long name as “seconds” and the abbreviation as “s”. d1718 1 a1718 1 For example, under DEC OSF/1, the “uprofile” command can be used to d1720 2 a1721 2 the dimension in the histogram header could be set to “i-cache misses” and the abbreviation could be set to “1” (because it is simply a count, d1737 2 a1738 2 must be within caller’s function and the second must be within the callee’s function. When performing profiling at the function level, d1757 1 a1757 1 9.3 ‘gprof’’s Internal Operation d1760 3 a1762 3 Like most programs, ‘gprof’ begins by processing its options. During this stage, it may building its symspec list (‘sym_ids.c:sym_id_add’), if options are specified which use symspecs. ‘gprof’ maintains a single d1764 1 a1764 1 tables, organized into six include/exclude pairs—one pair each for the d1772 2 a1773 2 After option processing, ‘gprof’ finishes building the symspec list by adding all the symspecs in ‘default_excluded_list’ to the exclude d1779 2 a1780 2 it is an object file, and read its symbol table (‘core.c:core_init’), using ‘bfd_canonicalize_symtab’ after mallocing an appropriately sized d1782 2 a1783 2 ‘--file-ordering’ option has been specified), and the core text space is read into memory (if the ‘-c’ option was given). d1785 1 a1785 1 ‘gprof’’s own symbol table, an array of Sym structures, is now built. d1787 1 a1787 1 whether line-by-line profiling (‘-l’ option) has been enabled. For d1791 1 a1791 1 either case, two passes are made through the symbol table—one to count d1793 2 a1794 2 the symbols. In between the two passes, a single array of type ‘Sym’ is created of the appropriate length. Finally, ‘symtab.c:symtab_finalize’ d1799 1 a1799 1 the ‘qsort’ library function (which sorts an array) will be used to sort d1801 1 a1801 1 (‘symtab.c:sym_lookup’), which finds symbols based on memory address, d1803 1 a1803 1 sorted array. Function symbols are indicated with an ‘is_func’ flag. d1805 1 a1805 1 can have an ‘is_static’ flag to indicate that it is a local symbol. d1808 2 a1809 2 Syms (‘sym_ids.c:sym_id_parse’). Remember that a single symspec can match multiple symbols. An array of symbol tables (‘syms’) is created, d1817 2 a1818 2 ‘gprof’ simply uses its standard symbol lookup routine on the appropriate table in the ‘syms’ array. d1821 2 a1822 2 (‘gmon_io.c:gmon_out_read’), first by checking for a new-style ‘gmon.out’ header, then assuming this is an old-style BSD ‘gmon.out’ if d1825 1 a1825 1 New-style histogram records are read by ‘hist.c:hist_read_rec’. For d1835 1 a1835 1 As each call graph record is read (‘call_graph.c:cg_read_rec’), the d1837 1 a1837 1 call graph arc is created by ‘cg_arcs.c:arc_add’, unless the arc fails a d1839 3 a1841 3 linked list is maintained of the parent’s child arcs, and of the child’s parent arcs. Both the child’s call count and the arc’s call count are incremented by the record’s call count. d1843 1 a1843 1 Basic-block records are read (‘basic_blocks.c:bb_read_rec’), but only d1846 1 a1846 1 entry made in the symbol’s bb_addr and bb_calls arrays. Again, if d1850 1 a1850 1 A gmon.sum file is dumped, if requested (‘gmon_io.c:gmon_out_write’). d1853 1 a1853 1 (‘hist.c:hist_assign_samples’) by iterating over all the sample bins and d1864 3 a1866 3 If call graph data is present, ‘cg_arcs.c:cg_assemble’ is called. First, if ‘-c’ was specified, a machine-dependent routine (‘find_call’) scans through each symbol’s machine code, looking for subroutine call d1869 1 a1869 1 (‘cg_dfn.c:cg_dfn’), so that children are always numbered less than d1878 1 a1878 1 a parent’s include or exclude (print or no print) property being d1888 2 a1889 2 fairly straightforward. The call graph (‘cg_print.c:cg_print’) and flat profile (‘hist.c:hist_print’) are regurgitations of values already d1891 1 a1891 1 (‘basic_blocks.c:print_annotated_source’) uses basic-block information, d1896 1 a1896 1 source code itself (‘cg_print.c’). Basically, the functions with the d1904 1 a1904 1 9.4 Debugging ‘gprof’ d1907 1 a1907 1 If ‘gprof’ was compiled with debugging enabled, the ‘-d’ option triggers d1927 1 a1927 1 file. For line-by-line profiling (‘-l’ option), also shows line d1930 1 a1930 1 Trace operation of ‘-c’ option d1937 1 a1937 1 meaningful with ‘-l’ option) d1941 1 a1941 1 Tracks operation of ‘-A’ option d1951 1 a1951 1 Copyright © 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. d1960 1 a1960 1 functional and useful document “free” in the sense of freedom: to d1967 1 a1967 1 This License is a kind of “copyleft”, which means that derivative d1988 2 a1989 2 “Document”, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as “you”. You accept d1993 1 a1993 1 A “Modified Version” of the Document means any work containing the d1997 1 a1997 1 A “Secondary Section” is a named appendix or a front-matter section d1999 1 a1999 1 publishers or authors of the Document to the Document’s overall d2008 1 a2008 1 The “Invariant Sections” are certain Secondary Sections whose d2016 1 a2016 1 The “Cover Texts” are certain short passages of text that are d2022 1 a2022 1 A “Transparent” copy of the Document means a machine-readable copy, d2034 1 a2034 1 “Transparent” is called “Opaque”. d2047 1 a2047 1 The “Title Page” means, for a printed book, the title page itself, d2050 3 a2052 3 works in formats which do not have any title page as such, “Title Page” means the text near the most prominent appearance of the work’s title, preceding the beginning of the body of the text. d2054 1 a2054 1 The “publisher” means any person or entity that distributes copies d2057 1 a2057 1 A section “Entitled XYZ” means a named subunit of the Document d2061 3 a2063 3 “Acknowledgements”, “Dedications”, “Endorsements”, or “History”.) To “Preserve the Title” of such a section when you modify the Document means that it remains a section “Entitled XYZ” according d2093 1 a2093 1 the Document’s license notice requires Cover Texts, you must d2165 1 a2165 1 Sections and required Cover Texts given in the Document’s d2170 1 a2170 1 I. Preserve the section Entitled “History”, Preserve its Title, d2173 1 a2173 1 Title Page. If there is no section Entitled “History” in the d2183 1 a2183 1 “History” section. You may omit a network location for a work d2188 1 a2188 1 K. For any section Entitled “Acknowledgements” or “Dedications”, d2197 1 a2197 1 M. Delete any section Entitled “Endorsements”. Such a section d2201 1 a2201 1 “Endorsements” or to conflict in title with any Invariant d2210 1 a2210 1 titles to the list of Invariant Sections in the Modified Version’s d2214 1 a2214 1 You may add a section Entitled “Endorsements”, provided it contains d2216 3 a2218 3 parties—for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2256 4 a2259 4 “History” in the various original documents, forming one section Entitled “History”; likewise combine any sections Entitled “Acknowledgements”, and any sections Entitled “Dedications”. You must delete all sections Entitled “Endorsements.” d2280 1 a2280 1 storage or distribution medium, is called an “aggregate” if the d2282 1 a2282 1 legal rights of the compilation’s users beyond what the individual d2289 1 a2289 1 of the entire aggregate, the Document’s Cover Texts may be placed d2311 2 a2312 2 If a section in the Document is Entitled “Acknowledgements”, “Dedications”, or “History”, the requirement (section 4) to d2353 1 a2353 1 version of this License “or any later version” applies to it, you d2361 1 a2361 1 proxy’s public statement of acceptance of a version permanently d2366 1 a2366 1 “Massive Multiauthor Collaboration Site” (or “MMC Site”) means any d2370 1 a2370 1 A “Massive Multiauthor Collaboration” (or “MMC”) contained in the d2374 1 a2374 1 “CC-BY-SA” means the Creative Commons Attribution-Share Alike 3.0 d2380 1 a2380 1 “Incorporate” means to publish or republish a Document, in whole or d2383 1 a2383 1 An MMC is “eligible for relicensing” if it is licensed under this d2410 1 a2410 1 Texts, replace the “with...Texts.” line with this: d2428 30 a2457 30 Node: Top726 Node: Introduction2083 Node: Compiling4604 Node: Executing8788 Node: Invoking11783 Node: Output Options13244 Node: Analysis Options20804 Node: Miscellaneous Options24904 Node: Deprecated Options26274 Node: Symspecs28457 Node: Output30372 Node: Flat Profile31426 Node: Call Graph36501 Node: Primary39771 Node: Callers42421 Node: Subroutines44641 Node: Cycles46597 Node: Line-by-line53625 Node: Annotated Source57792 Node: Inaccuracy60882 Node: Sampling Error61144 Node: Assumptions64104 Node: How do I?65657 Node: Incompatibilities67230 Node: Details68804 Node: Implementation69217 Node: File Format75279 Node: Internals79619 Node: Debugging88287 Node: GNU Free Documentation License89905 @ 1.1.1.11 log @Import binutils 2.45 (previous was 2.42) 2.45 Release Notes ------------------ Assembler: All sframe information generated by the assembler is now in compliance with the SFrame V2 specification. The assembler now supports .errif and .warnif directives, permitting user-controlled diagnostics with conditionals that are evaluated only at the end of assembly. The assembler predefines the symbol "GAS(version)". The assembler now supports the generation of SFrame stack trace information (.sframe) from CFI directives on s390 64-bit (s390x). For RISC-V, the ".option arch, -ext" format is deprecated due to its controversial use. For RISC-V, stop generating mapping symbols $x and replace with $x. The $x was defined to have the same ISA as previous $x, but now is defined to have the same ISA as elf architecture attribute. Once both used .option arch/rvc/norvc/push/pop directives (some code have different architectures with file attribute) and data directives in text, then the file need to be rebuilt since 2.45. The assembler supports the latest architecture extensions for the RISC-V, LoongArch and AArch64 architectures. Linker: The linker's --stats option can take an optional argument which if used is interpreted as a filename into which resource usage information should be stored. As an alternative mechanism the LD_STATS environment variable can also be used to achieve the same results. Resource usage information for various phases of the linking operation is now included in the report. If a map file is being produced then the information is also included there. The --no-stats option can be used to disable stat reporting, should it have been enabled. On s390 64-bit (s390x), generate SFrame stack trace information (.sframe) for the linker generated .plt section. On s390 32-bit, generate ".eh_frame" unwind information for the linker generated .plt section. Both features are enabled by default and can be disabled using linker option --no-ld-generated-unwind-info. On RISC-V, add new PLT formats, and GNU property merge rules for zicfiss and zicfilp extensions. On AVR, the default linker scripts now assert that the .progmem sections don't extend past 0xffff since they are accessed by means of LPM. For data in program memory that may be located past 0xffff, a .progmemx section should be used. On LoongArch, linker relaxation time complexity is no longer quadratic with respect to relocation counts. Linking time of large software should be improved. In addition R_LARCH_32_PCREL records are now checked for overflow. Other binary utilities: New versioned release of libsframe: libsframe.so.2. This release introduces versioned symbols with version node name LIBSFRAME_2.0. Some new symbols have been added to support the new flag SFRAME_F_FDE_FUNC_START_PCREL and retrieving flags from SFrame decoder and encoder objects: - Addition of sframe_decoder_get_flags, sframe_decoder_get_offsetof_fde_start_addr, sframe_encoder_get_flags, sframe_encoder_get_offsetof_fde_start_addr. This release also includes backward-incompatible ABI changes: - Removal of sframe_get_funcdesc_with_addr. - Change in the behavior of sframe_decoder_get_funcdesc_v2, sframe_encoder_add_funcdesc_v2 and sframe_encoder_write. For SFrame stack trace format, the function start address in each SFrame FDE has a changed encoding: The 32-bit signed integer now holds the offset of the start PC of the associated function from the sfde_func_start_address field itself (instead of the earlier where it was the offset from the start of the SFrame section itself). All SFrame sections generated by gas and ld now default to this new encoding, setting the (new) SFRAME_F_FDE_FUNC_START_PCREL flag. Relocatable SFrame links are now fixed. Readelf now recognizes RISC-V GNU_PROPERTY_RISCV_FEATURE_1_CFI_SS and GNU_PROPERTY_RISCV_FEATURE_1_CFI_LP_UNLABELED for zicfiss and zicfilp extensions. The LoongArch disassembler now properly accepts multiple disassembly options given by -M, such as "-M no-aliases,numeric". (Previously only the first option took effect.) 2.44 Release Notes ------------------ Assembler: Support for new architecture extensions for AArch64, Risc-V and x86. Linker: The default maximum page size was changed from 16KiB to 64KiB for LoongArch. This now supports mixed LTO and non-LTO object files in relocatable output. The ELF forms of the linker support a --image-base= option for compatibility with LLD. The --build-id= option now accepts an argument of "xx" which enables the use of the xxhash library. This produces a 128-bit hash and is 2-4x faster than md5 or sha1. The ELF linker option --package-metadata supports percent-encoded and %[string] encoded JSON payloads. Disassembler: The RISC-V disassembler now supports -M,max option like QEMU to dump instruction without checking architecture support as usual. GprofNG: Support added for hardware event counters for Neoverse-N1, Ampere-1, and Appliedmicro processors. Support for the Nios II target has been removed. 2.43 Release Notes ------------------ Assembler: * Add a .base64 directive to the assembler which allows base64 encoded binary data to be provided as strings. * Assembler macros as well as the bodies of .irp / .irpc / .rept can now use the syntax \+ to access the number of times a given macro has been executed. This is similar to the already existing \@@ syntax, except that the count is maintained on a per-macro basis. * References to FB and dollar labels, when supported, are no longer permitted in a radix other than 10. (Note that definitions of such labels were already thus restricted, except that leading zeroes were permitted). Linker: * Add support for DT_RELR type compressed runtime relocations for the AArch64 and LoongArch architectures. * Add --rosegment option which changes the -z separate-code option so that only one read-only segment is created (instead of two). * Add --section-ordering-file option to add extra mapping of input sections to output sections. * Add -plugin-save-temps to store plugin intermediate files permanently. Utilities: * Readelf will now display DT_RELR relocations in full detail. * Readelf now has a -j/--display-section option which takes the name or index of a section and displays its contents according to its type. The option can be used multiple times on the command line to display multiple sections. * When objdump or readelf are used to display the contents of a .eh_frame section they will now also display the contents of the .eh_frame_hdr section, if present. Gprofng: * Improved the support for hardware event counters: - Re-designed and streamlined the implementation. - Defined a common set of events for ARM processors. - Added specific events for AMD ZEN3 / ZEN4, and Intel Ice Lake processors. * Added a minimal support for RISC-V. General: * Target specific improvements for many architectures. @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 7.1.1 from gprof.texi. d5 1 a5 1 Copyright © 1988-2025 Free Software Foundation, Inc. d11 2 a12 2 Texts. A copy of the license is included in the section entitled "GNU Free Documentation License". d16 1 a16 1 * gprof: (gprof). Profiling your program's execution d30 1 a30 1 This manual is for ‘gprof’ (GNU Binutils) version 2.45. d34 1 a34 1 the section entitled "GNU Free Documentation License". d44 1 a44 1 * Output:: Interpreting ‘gprof’'s output d70 1 a70 1 data. If you don't use some feature of your program while it is being d99 1 a99 1 The “annotated source” listing is a copy of the program's source d175 1 a175 1 you can still profile the program, but you won't get complete d184 1 a184 1 ‘gcov’ tool instead of ‘gprof’. See that tool's manual or info pages d211 2 a212 2 The way you run the program--the arguments and input that you give it--may have a dramatic effect on what the profile information shows. d231 1 a231 1 The ‘gmon.out’ file is written in the program's _current working d234 1 a234 1 program ‘chdir’'d to. If you don't have permission to write in this d241 1 a241 1 didn't get written into ‘gmon.out’. The Perl script ‘bbconv.pl’, d284 1 a284 1 * Output Options:: Controlling ‘gprof’'s output style d316 1 a316 1 If the ‘-b’ option is given, ‘gprof’ doesn't print the verbose d449 1 a449 1 The ‘-T’ option causes ‘gprof’ to print its output in "traditional" d463 1 a463 1 similar to ‘tcov’'s ‘-a’. d669 1 a669 1 Selects everything in file ‘main.c’--the dot in the string tells d704 1 a704 1 5 Interpreting ‘gprof’'s Output d734 1 a734 1 and didn't run long enough to show up on the program counter histogram, d771 1 a771 1 program's total execution time was 0.06 seconds, as indicated by the d783 1 a783 1 field is ‘0.00’) didn't appear in the histogram samples at all. d888 3 a890 3 * Primary:: Details of the primary line's contents. * Callers:: Details of caller-lines' contents. * Subroutines:: Details of subroutine-lines' contents. d959 1 a959 1 is printed between the function's name and the index number (*note d969 1 a969 1 5.2.2 Lines for a Function's Callers d972 2 a973 2 A function's entry has a line for each function it was called by. These lines' fields correspond to the fields of the primary line, but their d1006 1 a1006 1 followed by the caller's index number. d1017 2 a1018 2 dummy caller-line is printed which has ‘’ as the "caller's name" and all other fields blank. This can happen for signal handlers. d1023 1 a1023 1 5.2.3 Lines for a Function's Subroutines d1026 2 a1027 2 A function's entry has a line for each of its subroutines--in other words, a line for each other function that it called. These lines' d1057 1 a1057 1 is used to determine how much of ‘report’'s ‘self’ and ‘children’ d1063 1 a1063 1 followed by the subroutine's index number. d1091 4 a1094 4 call graph paradoxical. The "time spent in children" of ‘a’ should include the time spent in its subroutine ‘b’ and in ‘b’'s subroutines--but one of ‘b’'s subroutines is ‘a’! How much of ‘a’'s time should be included in the children of ‘a’, when ‘a’ is indirectly d1100 1 a1100 1 "subroutines" of the cycle are the individual functions of the cycle, d1102 1 a1102 1 "callers" of the cycle are the functions, outside the cycle, that called d1163 1 a1163 1 The ‘self’ field of the cycle's primary line is the total time spent d1168 1 a1168 1 The ‘children’ fields of the cycle's primary line and subroutine d1170 1 a1170 1 ‘b’, the time spent in those calls to ‘b’ is not counted in ‘a’'s d1175 1 a1175 1 The ‘children’ field of a caller-line in the cycle's entry estimates d1188 1 a1188 1 cycle's entry says how many time that function was called from functions d1190 1 a1190 1 primary line's ‘called’ field. d1205 1 a1205 1 ‘gprof’'s ‘-l’ option causes the program to perform “line-by-line” d1219 1 a1219 1 mode. The call graph isn't as useful as normal, since the current d1224 1 a1224 1 Here is a section of ‘gprof’'s output, without line-by-line d1250 3 a1252 3 Now let's look at some of ‘gprof’'s output from the same program run, this time with line-by-line profiling enabled. Note that ‘ct_init’'s four histogram hits are broken down into four lines of source code--one d1254 1 a1254 1 note how ‘ct_init’'s 13327 calls to ‘init_block’ are broken down into d1297 2 a1298 2 ‘gprof’'s ‘-A’ option triggers an annotated source listing, which lists the program's source code, each function labeled with the number of d1300 1 a1300 1 ‘gprof’ can't find the source code files. d1342 1 a1342 1 line of code is labeled at least once. Here is ‘updcrc’'s annotated d1412 1 a1412 1 period is 0.01 seconds and ‘foo’'s run-time is 1 second, N is 100 d1414 1 a1414 1 error in ‘foo’'s run-time is 0.1 seconds (10*0.01 seconds), or ten d1416 2 a1417 2 seconds and ‘bar’'s run-time is 100 seconds, N is 10000 samples, sqrt(N) is 100 samples, so the expected error in ‘bar’'s run-time is 1 second, d1423 1 a1423 1 information. If the program's _total_ run-time is large, a small d1425 1 a1425 1 insignificant fraction of the whole program's time. Usually this means d1455 1 a1455 1 Some of the figures in the call graph are estimates--for example, the d1466 1 a1466 1 ‘foo’ contributes 2 seconds to ‘a’'s ‘children’ time, by assumption. d1568 2 a1569 2 * Internals:: ‘gprof’'s internal operation * Debugging:: Using ‘gprof’'s ‘-d’ option d1588 1 a1588 1 parent routine (the child) and its parent's parent. This is typically d1594 3 a1596 3 arguments--‘frompc’ and ‘selfpc’. ‘__mcount_internal’ is responsible for maintaining the in-memory call graph, which records ‘frompc’, ‘selfpc’, and the number of times each of these call arcs was traversed. d1619 1 a1619 1 the kernel, along with a scale factor that determines how the program's d1648 1 a1648 1 If the compiler's ‘-a’ option was used, basic-block counting is also d1666 1 a1666 1 of the program counter are taken at fixed intervals of the program's run d1672 1 a1672 1 the advantage that the amount of load due to other users won't directly d1718 1 a1718 1 would specify the long name as "seconds" and the abbreviation as "s". d1721 1 a1721 1 For example, under DEC OSF/1, the "uprofile" command can be used to d1723 2 a1724 2 the dimension in the histogram header could be set to "i-cache misses" and the abbreviation could be set to "1" (because it is simply a count, d1740 2 a1741 2 must be within caller's function and the second must be within the callee's function. When performing profiling at the function level, d1760 1 a1760 1 9.3 ‘gprof’'s Internal Operation d1767 1 a1767 1 tables, organized into six include/exclude pairs--one pair each for the d1788 1 a1788 1 ‘gprof’'s own symbol table, an array of Sym structures, is now built. d1794 1 a1794 1 either case, two passes are made through the symbol table--one to count d1842 3 a1844 3 linked list is maintained of the parent's child arcs, and of the child's parent arcs. Both the child's call count and the arc's call count are incremented by the record's call count. d1849 1 a1849 1 entry made in the symbol's bb_addr and bb_calls arrays. Again, if d1869 1 a1869 1 scans through each symbol's machine code, looking for subroutine call d1881 1 a1881 1 a parent's include or exclude (print or no print) property being d1970 1 a1970 1 This License is a kind of "copyleft", which means that derivative d1991 2 a1992 2 "Document", below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "you". You accept d1996 1 a1996 1 A "Modified Version" of the Document means any work containing the d2000 1 a2000 1 A "Secondary Section" is a named appendix or a front-matter section d2002 1 a2002 1 publishers or authors of the Document to the Document's overall d2011 1 a2011 1 The "Invariant Sections" are certain Secondary Sections whose d2019 1 a2019 1 The "Cover Texts" are certain short passages of text that are d2025 1 a2025 1 A "Transparent" copy of the Document means a machine-readable copy, d2037 1 a2037 1 "Transparent" is called "Opaque". d2050 1 a2050 1 The "Title Page" means, for a printed book, the title page itself, d2053 3 a2055 3 works in formats which do not have any title page as such, "Title Page" means the text near the most prominent appearance of the work's title, preceding the beginning of the body of the text. d2057 1 a2057 1 The "publisher" means any person or entity that distributes copies d2060 1 a2060 1 A section "Entitled XYZ" means a named subunit of the Document d2064 3 a2066 3 "Acknowledgements", "Dedications", "Endorsements", or "History".) To "Preserve the Title" of such a section when you modify the Document means that it remains a section "Entitled XYZ" according d2096 1 a2096 1 the Document's license notice requires Cover Texts, you must d2168 1 a2168 1 Sections and required Cover Texts given in the Document's d2173 1 a2173 1 I. Preserve the section Entitled "History", Preserve its Title, d2176 1 a2176 1 Title Page. If there is no section Entitled "History" in the d2186 1 a2186 1 "History" section. You may omit a network location for a work d2191 1 a2191 1 K. For any section Entitled "Acknowledgements" or "Dedications", d2200 1 a2200 1 M. Delete any section Entitled "Endorsements". Such a section d2204 1 a2204 1 "Endorsements" or to conflict in title with any Invariant d2213 1 a2213 1 titles to the list of Invariant Sections in the Modified Version's d2217 1 a2217 1 You may add a section Entitled "Endorsements", provided it contains d2219 3 a2221 3 parties--for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard. d2259 4 a2262 4 "History" in the various original documents, forming one section Entitled "History"; likewise combine any sections Entitled "Acknowledgements", and any sections Entitled "Dedications". You must delete all sections Entitled "Endorsements." d2283 1 a2283 1 storage or distribution medium, is called an "aggregate" if the d2285 1 a2285 1 legal rights of the compilation's users beyond what the individual d2292 1 a2292 1 of the entire aggregate, the Document's Cover Texts may be placed d2314 2 a2315 2 If a section in the Document is Entitled "Acknowledgements", "Dedications", or "History", the requirement (section 4) to d2356 1 a2356 1 version of this License "or any later version" applies to it, you d2364 1 a2364 1 proxy's public statement of acceptance of a version permanently d2369 1 a2369 1 "Massive Multiauthor Collaboration Site" (or "MMC Site") means any d2373 1 a2373 1 A "Massive Multiauthor Collaboration" (or "MMC") contained in the d2377 1 a2377 1 "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 d2383 1 a2383 1 "Incorporate" means to publish or republish a Document, in whole or d2386 1 a2386 1 An MMC is "eligible for relicensing" if it is licensed under this d2413 1 a2413 1 Texts, replace the "with...Texts." line with this: d2431 30 a2460 30 Node: Top720 Node: Introduction2071 Node: Compiling4588 Node: Executing8768 Node: Invoking11753 Node: Output Options13212 Node: Analysis Options20764 Node: Miscellaneous Options24864 Node: Deprecated Options26234 Node: Symspecs28417 Node: Output30331 Node: Flat Profile31383 Node: Call Graph36452 Node: Primary39716 Node: Callers42364 Node: Subroutines44570 Node: Cycles46515 Node: Line-by-line53512 Node: Annotated Source57664 Node: Inaccuracy60746 Node: Sampling Error61008 Node: Assumptions63956 Node: How do I?65506 Node: Incompatibilities67079 Node: Details68653 Node: Implementation69062 Node: File Format75113 Node: Internals79429 Node: Debugging88075 Node: GNU Free Documentation License89693 @ 1.1.1.12 log @Import binutils-2.46.0 (previous was 2.45) Changes in 2.45: * New versioned release of libsframe: libsframe.so.2. This release introduces versioned symbols with version node name LIBSFRAME_2.0. Some new symbols have been added to support the new flag SFRAME_F_FDE_FUNC_START_PCREL and retrieving flags from SFrame decoder and encoder objects: - Addition of sframe_decoder_get_flags, sframe_decoder_get_offsetof_fde_start_addr, sframe_encoder_get_flags, sframe_encoder_get_offsetof_fde_start_addr. This release also includes backward-incompatible ABI changes: - Removal of sframe_get_funcdesc_with_addr. - Change in the behavior of sframe_decoder_get_funcdesc_v2, sframe_encoder_add_funcdesc_v2 and sframe_encoder_write. * On s390 64-bit (s390x), gas, ld, objdump, and readelf now support generating and processing SFrame V2 stack trace information (.sframe). The assembler generates SFrame info from CFI directives with option "--gsframe". The linker generates SFrame info for the linker-generated .plt section and merges all .sframe sections. Both objdump and readelf dump SFrame info with option "--sframe[=]". * For SFrame stack trace format, the function start address in each SFrame FDE has a changed encoding: The 32-bit signed integer now holds the offset of the start PC of the associated function from the sfde_func_start_address field itself (instead of the earlier where it was the offset from the start of the SFrame section itself). All SFrame sections generated by gas and ld now default to this new encoding, setting the (new) SFRAME_F_FDE_FUNC_START_PCREL flag. Relocatable SFrame links are now fixed. * Readelf now recognizes RISC-V GNU_PROPERTY_RISCV_FEATURE_1_CFI_SS and GNU_PROPERTY_RISCV_FEATURE_1_CFI_LP_UNLABELED for zicfiss and zicfilp extensions. * For RISC-V dis-assembler, the definition of mapping symbol $x is changed, so the file needs to be rebuilt since 2.45 once used .option arch directives. * The LoongArch disassembler now properly accepts multiple disassembly options given by -M, such as "-M no-aliases,numeric". (Previously only the first option took effect.) @ text @d1 1 a1 1 This is gprof.info, produced by makeinfo version 7.2 from gprof.texi. d5 1 a5 1 Copyright © 1988-2026 Free Software Foundation, Inc. d30 1 a30 1 This manual is for ‘gprof’ (GNU Binutils) version 2.46.0. a264 4 Note that ‘gprofng’, a more modern profiler which is also part of binutils, can also read gmon files and provides more functionality than ‘gprof’. d443 1 d1249 1 d1290 1 d1332 1 d2428 1 d2431 1 a2431 1 Node: Top718 d2436 25 a2460 25 Node: Output Options13371 Node: Analysis Options20922 Node: Miscellaneous Options25022 Node: Deprecated Options26392 Node: Symspecs28575 Node: Output30489 Node: Flat Profile31541 Node: Call Graph36610 Node: Primary39874 Node: Callers42522 Node: Subroutines44728 Node: Cycles46673 Node: Line-by-line53670 Node: Annotated Source57820 Node: Inaccuracy60901 Node: Sampling Error61163 Node: Assumptions64111 Node: How do I?65661 Node: Incompatibilities67234 Node: Details68808 Node: Implementation69217 Node: File Format75268 Node: Internals79584 Node: Debugging88230 Node: GNU Free Documentation License89848 @