{ "content": [ { "type": "text", "text": "# Benchmark (man)\n\n## NAME\n\nBenchmark - benchmark running times of Perl code\n\n## SYNOPSIS\n\nuse Benchmark qw(:all) ;\ntimethis ($count, \"code\");\n# Use Perl code in strings...\ntimethese($count, {\n'Name1' => '...code1...',\n'Name2' => '...code2...',\n});\n# ... or use subroutine references.\ntimethese($count, {\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n});\n# cmpthese can be used both ways as well\ncmpthese($count, {\n'Name1' => '...code1...',\n'Name2' => '...code2...',\n});\ncmpthese($count, {\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n});\n# ...or in two stages\n$results = timethese($count,\n{\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n},\n'none'\n);\ncmpthese( $results ) ;\n$t = timeit($count, '...other code...')\nprint \"$count loops of other code took:\",timestr($t),\"\\n\";\n$t = countit($time, '...other code...')\n$count = $t->iters ;\nprint \"$count loops of other code took:\",timestr($t),\"\\n\";\n# enable hires wallclock timing if possible\nuse Benchmark ':hireswallclock';\n\n## DESCRIPTION\n\nThe Benchmark module encapsulates a number of routines to help you figure out how long it\ntakes to execute some code.\n\n## Sections\n\n- **NAME**\n- **SYNOPSIS**\n- **DESCRIPTION** (5 subsections)\n- **NOTES**\n- **EXAMPLES**\n- **INHERITANCE**\n- **CAVEATS**\n- **SEE ALSO**\n- **AUTHORS**\n- **MODIFICATION HISTORY**\n\nUse structuredContent.sections for detailed options, examples, and full documentation.\n" } ], "structuredContent": { "command": "Benchmark", "section": "", "mode": "man", "summary": "Benchmark - benchmark running times of Perl code", "synopsis": "use Benchmark qw(:all) ;\ntimethis ($count, \"code\");\n# Use Perl code in strings...\ntimethese($count, {\n'Name1' => '...code1...',\n'Name2' => '...code2...',\n});\n# ... or use subroutine references.\ntimethese($count, {\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n});\n# cmpthese can be used both ways as well\ncmpthese($count, {\n'Name1' => '...code1...',\n'Name2' => '...code2...',\n});\ncmpthese($count, {\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n});\n# ...or in two stages\n$results = timethese($count,\n{\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n},\n'none'\n);\ncmpthese( $results ) ;\n$t = timeit($count, '...other code...')\nprint \"$count loops of other code took:\",timestr($t),\"\\n\";\n$t = countit($time, '...other code...')\n$count = $t->iters ;\nprint \"$count loops of other code took:\",timestr($t),\"\\n\";\n# enable hires wallclock timing if possible\nuse Benchmark ':hireswallclock';", "tldr_summary": null, "tldr_examples": [], "tldr_source": null, "flags": [], "examples": [ "For example,", "use Benchmark qw( cmpthese ) ;", "$x = 3;", "cmpthese( -5, {", "a => sub{$x*$x},", "b => sub{$x2},", "} );", "outputs something like this:", "Benchmark: running a, b, each for at least 5 CPU seconds...", "Rate b a", "b 1559428/s -- -62%", "a 4152037/s 166% --", "while", "use Benchmark qw( timethese cmpthese ) ;", "$x = 3;", "$r = timethese( -5, {", "a => sub{$x*$x},", "b => sub{$x2},", "} );", "cmpthese $r;", "outputs something like this:", "Benchmark: running a, b, each for at least 5 CPU seconds...", "a: 10 wallclock secs ( 5.14 usr + 0.13 sys = 5.27 CPU) @ 3835055.60/s (n=20210743)", "b: 5 wallclock secs ( 5.41 usr + 0.00 sys = 5.41 CPU) @ 1574944.92/s (n=8520452)", "Rate b a", "b 1574945/s -- -59%", "a 3835056/s 144% --" ], "see_also": [], "section_outline": [ { "name": "NAME", "lines": 2, "subsections": [] }, { "name": "SYNOPSIS", "lines": 47, "subsections": [] }, { "name": "DESCRIPTION", "lines": 13, "subsections": [ { "name": "Methods", "lines": 17 }, { "name": "Standard Exports", "lines": 70 }, { "name": "Optional Exports", "lines": 80 }, { "name": ":hireswallclock", "lines": 6 }, { "name": "Benchmark Object", "lines": 27 } ] }, { "name": "NOTES", "lines": 25, "subsections": [] }, { "name": "EXAMPLES", "lines": 35, "subsections": [] }, { "name": "INHERITANCE", "lines": 2, "subsections": [] }, { "name": "CAVEATS", "lines": 13, "subsections": [] }, { "name": "SEE ALSO", "lines": 2, "subsections": [] }, { "name": "AUTHORS", "lines": 2, "subsections": [] }, { "name": "MODIFICATION HISTORY", "lines": 26, "subsections": [] } ], "sections": { "NAME": { "content": "Benchmark - benchmark running times of Perl code\n", "subsections": [] }, "SYNOPSIS": { "content": "use Benchmark qw(:all) ;\n\ntimethis ($count, \"code\");\n\n# Use Perl code in strings...\ntimethese($count, {\n'Name1' => '...code1...',\n'Name2' => '...code2...',\n});\n\n# ... or use subroutine references.\ntimethese($count, {\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n});\n\n# cmpthese can be used both ways as well\ncmpthese($count, {\n'Name1' => '...code1...',\n'Name2' => '...code2...',\n});\n\ncmpthese($count, {\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n});\n\n# ...or in two stages\n$results = timethese($count,\n{\n'Name1' => sub { ...code1... },\n'Name2' => sub { ...code2... },\n},\n'none'\n);\ncmpthese( $results ) ;\n\n$t = timeit($count, '...other code...')\nprint \"$count loops of other code took:\",timestr($t),\"\\n\";\n\n$t = countit($time, '...other code...')\n$count = $t->iters ;\nprint \"$count loops of other code took:\",timestr($t),\"\\n\";\n\n# enable hires wallclock timing if possible\nuse Benchmark ':hireswallclock';\n", "subsections": [] }, "DESCRIPTION": { "content": "The Benchmark module encapsulates a number of routines to help you figure out how long it\ntakes to execute some code.\n\ntimethis - run a chunk of code several times\n\ntimethese - run several chunks of code several times\n\ncmpthese - print results of timethese as a comparison chart\n\ntimeit - run a chunk of code and see how long it goes\n\ncountit - see how many times a chunk of code runs in a given time\n", "subsections": [ { "name": "Methods", "content": "new Returns the current time. Example:\n\nuse Benchmark;\n$t0 = Benchmark->new;\n# ... your code here ...\n$t1 = Benchmark->new;\n$td = timediff($t1, $t0);\nprint \"the code took:\",timestr($td),\"\\n\";\n\ndebug Enables or disable debugging by setting the $Benchmark::Debug flag:\n\nBenchmark->debug(1);\n$t = timeit(10, ' 5 $Global ');\nBenchmark->debug(0);\n\niters Returns the number of iterations.\n" }, { "name": "Standard Exports", "content": "The following routines will be exported into your namespace if you use the Benchmark module:\n\ntimeit(COUNT, CODE)\nArguments: COUNT is the number of times to run the loop, and CODE is the code to\nrun. CODE may be either a code reference or a string to be eval'd; either way it\nwill be run in the caller's package.\n\nReturns: a Benchmark object.\n\ntimethis ( COUNT, CODE, [ TITLE, [ STYLE ]] )\nTime COUNT iterations of CODE. CODE may be a string to eval or a code reference;\neither way the CODE will run in the caller's package. Results will be printed to\nSTDOUT as TITLE followed by the times. TITLE defaults to \"timethis COUNT\" if none\nis provided. STYLE determines the format of the output, as described for timestr()\nbelow.\n\nThe COUNT can be zero or negative: this means the minimum number of CPU seconds to\nrun. A zero signifies the default of 3 seconds. For example to run at least for\n10 seconds:\n\ntimethis(-10, $code)\n\nor to run two pieces of code tests for at least 3 seconds:\n\ntimethese(0, { test1 => '...', test2 => '...'})\n\nCPU seconds is, in UNIX terms, the user time plus the system time of the process\nitself, as opposed to the real (wallclock) time and the time spent by the child\nprocesses. Less than 0.1 seconds is not accepted (-0.01 as the count, for example,\nwill cause a fatal runtime exception).\n\nNote that the CPU seconds is the minimum time: CPU scheduling and other operating\nsystem factors may complicate the attempt so that a little bit more time is spent.\nThe benchmark output will, however, also tell the number of $code runs/second,\nwhich should be a more interesting number than the actually spent seconds.\n\nReturns a Benchmark object.\n\ntimethese ( COUNT, CODEHASHREF, [ STYLE ] )\nThe CODEHASHREF is a reference to a hash containing names as keys and either a\nstring to eval or a code reference for each value. For each (KEY, VALUE) pair in\nthe CODEHASHREF, this routine will call\n\ntimethis(COUNT, VALUE, KEY, STYLE)\n\nThe routines are called in string comparison order of KEY.\n\nThe COUNT can be zero or negative, see timethis().\n\nReturns a hash reference of Benchmark objects, keyed by name.\n\ntimediff ( T1, T2 )\nReturns the difference between two Benchmark times as a Benchmark object suitable\nfor passing to timestr().\n\ntimestr ( TIMEDIFF, [ STYLE, [ FORMAT ] ] )\nReturns a string that formats the times in the TIMEDIFF object in the requested\nSTYLE. TIMEDIFF is expected to be a Benchmark object similar to that returned by\ntimediff().\n\nSTYLE can be any of 'all', 'none', 'noc', 'nop' or 'auto'. 'all' shows each of the\n5 times available ('wallclock' time, user time, system time, user time of children,\nand system time of children). 'noc' shows all except the two children times. 'nop'\nshows only wallclock and the two children times. 'auto' (the default) will act as\n'all' unless the children times are both zero, in which case it acts as 'noc'.\n'none' prevents output.\n\nFORMAT is the printf(3)-style format specifier (without the leading '%') to use to\nprint the times. It defaults to '5.2f'.\n" }, { "name": "Optional Exports", "content": "The following routines will be exported into your namespace if you specifically ask that they\nbe imported:\n\nclearcache ( COUNT )\nClear the cached time for COUNT rounds of the null loop.\n\nclearallcache ( )\nClear all cached times.\n\ncmpthese ( COUNT, CODEHASHREF, [ STYLE ] )\ncmpthese ( RESULTSHASHREF, [ STYLE ] )\nOptionally calls timethese(), then outputs comparison chart. This:\n\ncmpthese( -1, { a => \"++\\$i\", b => \"\\$i *= 2\" } ) ;\n\noutputs a chart like:\n\nRate b a\nb 2831802/s -- -61%\na 7208959/s 155% --\n\nThis chart is sorted from slowest to fastest, and shows the percent speed\ndifference between each pair of tests.\n\n\"cmpthese\" can also be passed the data structure that timethese() returns:\n\n$results = timethese( -1,\n{ a => \"++\\$i\", b => \"\\$i *= 2\" } ) ;\ncmpthese( $results );\n\nin case you want to see both sets of results. If the first argument is an\nunblessed hash reference, that is RESULTSHASHREF; otherwise that is COUNT.\n\nReturns a reference to an ARRAY of rows, each row is an ARRAY of cells from the\nabove chart, including labels. This:\n\nmy $rows = cmpthese( -1,\n{ a => '++$i', b => '$i *= 2' }, \"none\" );\n\nreturns a data structure like:\n\n[\n[ '', 'Rate', 'b', 'a' ],\n[ 'b', '2885232/s', '--', '-59%' ],\n[ 'a', '7099126/s', '146%', '--' ],\n]\n\nNOTE: This result value differs from previous versions, which returned the\n\"timethese()\" result structure. If you want that, just use the two statement\n\"timethese\"...\"cmpthese\" idiom shown above.\n\nIncidentally, note the variance in the result values between the two examples; this\nis typical of benchmarking. If this were a real benchmark, you would probably want\nto run a lot more iterations.\n\ncountit(TIME, CODE)\nArguments: TIME is the minimum length of time to run CODE for, and CODE is the code\nto run. CODE may be either a code reference or a string to be eval'd; either way\nit will be run in the caller's package.\n\nTIME is not negative. countit() will run the loop many times to calculate the\nspeed of CODE before running it for TIME. The actual time run for will usually be\ngreater than TIME due to system clock resolution, so it's best to look at the\nnumber of iterations divided by the times that you are concerned with, not just the\niterations.\n\nReturns: a Benchmark object.\n\ndisablecache ( )\nDisable caching of timings for the null loop. This will force Benchmark to\nrecalculate these timings for each new piece of code timed.\n\nenablecache ( )\nEnable caching of timings for the null loop. The time taken for COUNT rounds of the\nnull loop will be calculated only once for each different COUNT used.\n\ntimesum ( T1, T2 )\nReturns the sum of two Benchmark times as a Benchmark object suitable for passing\nto timestr().\n" }, { "name": ":hireswallclock", "content": "If the Time::HiRes module has been installed, you can specify the special tag\n\":hireswallclock\" for Benchmark (if Time::HiRes is not available, the tag will be silently\nignored). This tag will cause the wallclock time to be measured in microseconds, instead of\ninteger seconds. Note though that the speed computations are still conducted in CPU time,\nnot wallclock time.\n" }, { "name": "Benchmark Object", "content": "Many of the functions in this module return a Benchmark object, or in the case of\n\"timethese()\", a reference to a hash, the values of which are Benchmark objects. This is\nuseful if you want to store or further process results from Benchmark functions.\n\nInternally the Benchmark object holds timing values, described in \"NOTES\" below. The\nfollowing methods can be used to access them:\n\ncpup\nTotal CPU (User + System) of the main (parent) process.\n\ncpuc\nTotal CPU (User + System) of any children processes.\n\ncpua\nTotal CPU of parent and any children processes.\n\nreal\nReal elapsed time \"wallclock seconds\".\n\niters\nNumber of iterations run.\n\nThe following illustrates use of the Benchmark object:\n\n$result = timethis(100000, sub { ... });\nprint \"total CPU = \", $result->cpua, \"\\n\";\n" } ] }, "NOTES": { "content": "The data is stored as a list of values from the time and times functions:\n\n($real, $user, $system, $childrenuser, $childrensystem, $iters)\n\nin seconds for the whole loop (not divided by the number of rounds).\n\nThe timing is done using time(3) and times(3).\n\nCode is executed in the caller's package.\n\nThe time of the null loop (a loop with the same number of rounds but empty loop body) is\nsubtracted from the time of the real loop.\n\nThe null loop times can be cached, the key being the number of rounds. The caching can be\ncontrolled using calls like these:\n\nclearcache($key);\nclearallcache();\n\ndisablecache();\nenablecache();\n\nCaching is off by default, as it can (usually slightly) decrease accuracy and does not\nusually noticeably affect runtimes.\n", "subsections": [] }, "EXAMPLES": { "content": "For example,\n\nuse Benchmark qw( cmpthese ) ;\n$x = 3;\ncmpthese( -5, {\na => sub{$x*$x},\nb => sub{$x2},\n} );\n\noutputs something like this:\n\nBenchmark: running a, b, each for at least 5 CPU seconds...\nRate b a\nb 1559428/s -- -62%\na 4152037/s 166% --\n\nwhile\n\nuse Benchmark qw( timethese cmpthese ) ;\n$x = 3;\n$r = timethese( -5, {\na => sub{$x*$x},\nb => sub{$x2},\n} );\ncmpthese $r;\n\noutputs something like this:\n\nBenchmark: running a, b, each for at least 5 CPU seconds...\na: 10 wallclock secs ( 5.14 usr + 0.13 sys = 5.27 CPU) @ 3835055.60/s (n=20210743)\nb: 5 wallclock secs ( 5.41 usr + 0.00 sys = 5.41 CPU) @ 1574944.92/s (n=8520452)\nRate b a\nb 1574945/s -- -59%\na 3835056/s 144% --\n", "subsections": [] }, "INHERITANCE": { "content": "Benchmark inherits from no other class, except of course from Exporter.\n", "subsections": [] }, "CAVEATS": { "content": "Comparing eval'd strings with code references will give you inaccurate results: a code\nreference will show a slightly slower execution time than the equivalent eval'd string.\n\nThe real time timing is done using time(2) and the granularity is therefore only one second.\n\nShort tests may produce negative figures because perl can appear to take longer to execute\nthe empty loop than a short test; try:\n\ntimethis(100,'1');\n\nThe system time of the null loop might be slightly more than the system time of the loop with\nthe actual code and therefore the difference might end up being < 0.\n", "subsections": [] }, "SEE ALSO": { "content": "Devel::NYTProf - a Perl code profiler\n", "subsections": [] }, "AUTHORS": { "content": "Jarkko Hietaniemi , Tim Bunce \n", "subsections": [] }, "MODIFICATION HISTORY": { "content": "September 8th, 1994; by Tim Bunce.\n\nMarch 28th, 1997; by Hugo van der Sanden: added support for code references and the already\ndocumented 'debug' method; revamped documentation.\n\nApril 04-07th, 1997: by Jarkko Hietaniemi, added the run-for-some-time functionality.\n\nSeptember, 1999; by Barrie Slaymaker: math fixes and accuracy and efficiency tweaks. Added\ncmpthese(). A result is now returned from timethese(). Exposed countit() (was runfor()).\n\nDecember, 2001; by Nicholas Clark: make timestr() recognise the style 'none' and return an\nempty string. If cmpthese is calling timethese, make it pass the style in. (so that 'none'\nwill suppress output). Make sub new dump its debugging output to STDERR, to be consistent\nwith everything else. All bugs found while writing a regression test.\n\nSeptember, 2002; by Jarkko Hietaniemi: add ':hireswallclock' special tag.\n\nFebruary, 2004; by Chia-liang Kao: make cmpthese and timestr use time statistics for children\ninstead of parent when the style is 'nop'.\n\nNovember, 2007; by Christophe Grosjean: make cmpthese and timestr compute time consistently\nwith style argument, default is 'all' not 'noc' any more.\n\n\n\nperl v5.34.0 2025-07-25 Benchmark(3perl)", "subsections": [] } } } }