Commit | Line | Data |
a0d0e21e |
1 | package Benchmark; |
2 | |
53aa2791 |
3 | use strict; |
4 | |
5 | |
f06db76b |
6 | =head1 NAME |
7 | |
8a4f6ac2 |
8 | Benchmark - benchmark running times of Perl code |
431d98c2 |
9 | |
f06db76b |
10 | =head1 SYNOPSIS |
11 | |
f36484b0 |
12 | use Benchmark qw(:all) ; |
13 | |
f06db76b |
14 | timethis ($count, "code"); |
15 | |
523cc92b |
16 | # Use Perl code in strings... |
f06db76b |
17 | timethese($count, { |
18 | 'Name1' => '...code1...', |
19 | 'Name2' => '...code2...', |
20 | }); |
21 | |
523cc92b |
22 | # ... or use subroutine references. |
23 | timethese($count, { |
24 | 'Name1' => sub { ...code1... }, |
25 | 'Name2' => sub { ...code2... }, |
26 | }); |
27 | |
431d98c2 |
28 | # cmpthese can be used both ways as well |
29 | cmpthese($count, { |
30 | 'Name1' => '...code1...', |
31 | 'Name2' => '...code2...', |
32 | }); |
33 | |
34 | cmpthese($count, { |
35 | 'Name1' => sub { ...code1... }, |
36 | 'Name2' => sub { ...code2... }, |
37 | }); |
38 | |
39 | # ...or in two stages |
40 | $results = timethese($count, |
41 | { |
42 | 'Name1' => sub { ...code1... }, |
43 | 'Name2' => sub { ...code2... }, |
44 | }, |
45 | 'none' |
46 | ); |
47 | cmpthese( $results ) ; |
48 | |
f06db76b |
49 | $t = timeit($count, '...other code...') |
50 | print "$count loops of other code took:",timestr($t),"\n"; |
51 | |
431d98c2 |
52 | $t = countit($time, '...other code...') |
53 | $count = $t->iters ; |
54 | print "$count loops of other code took:",timestr($t),"\n"; |
55 | |
e3d6de9a |
56 | # enable hires wallclock timing if possible |
57 | use Benchmark ':hireswallclock'; |
58 | |
f06db76b |
59 | =head1 DESCRIPTION |
60 | |
61 | The Benchmark module encapsulates a number of routines to help you |
62 | figure out how long it takes to execute some code. |
63 | |
8a4f6ac2 |
64 | timethis - run a chunk of code several times |
65 | |
66 | timethese - run several chunks of code several times |
67 | |
68 | cmpthese - print results of timethese as a comparison chart |
69 | |
70 | timeit - run a chunk of code and see how long it goes |
71 | |
72 | countit - see how many times a chunk of code runs in a given time |
73 | |
74 | |
f06db76b |
75 | =head2 Methods |
76 | |
77 | =over 10 |
78 | |
79 | =item new |
80 | |
81 | Returns the current time. Example: |
82 | |
83 | use Benchmark; |
84 | $t0 = new Benchmark; |
85 | # ... your code here ... |
86 | $t1 = new Benchmark; |
87 | $td = timediff($t1, $t0); |
a24a9dfe |
88 | print "the code took:",timestr($td),"\n"; |
f06db76b |
89 | |
90 | =item debug |
91 | |
92 | Enables or disable debugging by setting the C<$Benchmark::Debug> flag: |
93 | |
523cc92b |
94 | debug Benchmark 1; |
f06db76b |
95 | $t = timeit(10, ' 5 ** $Global '); |
523cc92b |
96 | debug Benchmark 0; |
f06db76b |
97 | |
431d98c2 |
98 | =item iters |
99 | |
100 | Returns the number of iterations. |
101 | |
f06db76b |
102 | =back |
103 | |
104 | =head2 Standard Exports |
105 | |
523cc92b |
106 | The following routines will be exported into your namespace |
f06db76b |
107 | if you use the Benchmark module: |
108 | |
109 | =over 10 |
110 | |
111 | =item timeit(COUNT, CODE) |
112 | |
523cc92b |
113 | Arguments: COUNT is the number of times to run the loop, and CODE is |
114 | the code to run. CODE may be either a code reference or a string to |
115 | be eval'd; either way it will be run in the caller's package. |
116 | |
117 | Returns: a Benchmark object. |
118 | |
119 | =item timethis ( COUNT, CODE, [ TITLE, [ STYLE ]] ) |
120 | |
121 | Time COUNT iterations of CODE. CODE may be a string to eval or a |
122 | code reference; either way the CODE will run in the caller's package. |
123 | Results will be printed to STDOUT as TITLE followed by the times. |
124 | TITLE defaults to "timethis COUNT" if none is provided. STYLE |
125 | determines the format of the output, as described for timestr() below. |
126 | |
6ee623d5 |
127 | The COUNT can be zero or negative: this means the I<minimum number of |
128 | CPU seconds> to run. A zero signifies the default of 3 seconds. For |
129 | example to run at least for 10 seconds: |
130 | |
131 | timethis(-10, $code) |
132 | |
133 | or to run two pieces of code tests for at least 3 seconds: |
134 | |
135 | timethese(0, { test1 => '...', test2 => '...'}) |
136 | |
137 | CPU seconds is, in UNIX terms, the user time plus the system time of |
138 | the process itself, as opposed to the real (wallclock) time and the |
139 | time spent by the child processes. Less than 0.1 seconds is not |
140 | accepted (-0.01 as the count, for example, will cause a fatal runtime |
141 | exception). |
142 | |
143 | Note that the CPU seconds is the B<minimum> time: CPU scheduling and |
144 | other operating system factors may complicate the attempt so that a |
145 | little bit more time is spent. The benchmark output will, however, |
146 | also tell the number of C<$code> runs/second, which should be a more |
147 | interesting number than the actually spent seconds. |
148 | |
149 | Returns a Benchmark object. |
150 | |
523cc92b |
151 | =item timethese ( COUNT, CODEHASHREF, [ STYLE ] ) |
f06db76b |
152 | |
523cc92b |
153 | The CODEHASHREF is a reference to a hash containing names as keys |
154 | and either a string to eval or a code reference for each value. |
155 | For each (KEY, VALUE) pair in the CODEHASHREF, this routine will |
156 | call |
f06db76b |
157 | |
523cc92b |
158 | timethis(COUNT, VALUE, KEY, STYLE) |
f06db76b |
159 | |
1d2dff63 |
160 | The routines are called in string comparison order of KEY. |
161 | |
162 | The COUNT can be zero or negative, see timethis(). |
6ee623d5 |
163 | |
3c6312e9 |
164 | Returns a hash of Benchmark objects, keyed by name. |
165 | |
523cc92b |
166 | =item timediff ( T1, T2 ) |
f06db76b |
167 | |
523cc92b |
168 | Returns the difference between two Benchmark times as a Benchmark |
169 | object suitable for passing to timestr(). |
f06db76b |
170 | |
6ee623d5 |
171 | =item timestr ( TIMEDIFF, [ STYLE, [ FORMAT ] ] ) |
f06db76b |
172 | |
523cc92b |
173 | Returns a string that formats the times in the TIMEDIFF object in |
174 | the requested STYLE. TIMEDIFF is expected to be a Benchmark object |
175 | similar to that returned by timediff(). |
176 | |
3c6312e9 |
177 | STYLE can be any of 'all', 'none', 'noc', 'nop' or 'auto'. 'all' shows |
178 | each of the 5 times available ('wallclock' time, user time, system time, |
523cc92b |
179 | user time of children, and system time of children). 'noc' shows all |
180 | except the two children times. 'nop' shows only wallclock and the |
181 | two children times. 'auto' (the default) will act as 'all' unless |
182 | the children times are both zero, in which case it acts as 'noc'. |
3c6312e9 |
183 | 'none' prevents output. |
523cc92b |
184 | |
185 | FORMAT is the L<printf(3)>-style format specifier (without the |
186 | leading '%') to use to print the times. It defaults to '5.2f'. |
f06db76b |
187 | |
188 | =back |
189 | |
190 | =head2 Optional Exports |
191 | |
192 | The following routines will be exported into your namespace |
193 | if you specifically ask that they be imported: |
194 | |
195 | =over 10 |
196 | |
523cc92b |
197 | =item clearcache ( COUNT ) |
198 | |
199 | Clear the cached time for COUNT rounds of the null loop. |
200 | |
201 | =item clearallcache ( ) |
f06db76b |
202 | |
523cc92b |
203 | Clear all cached times. |
f06db76b |
204 | |
8962dfd6 |
205 | =item cmpthese ( COUNT, CODEHASHREF, [ STYLE ] ) |
ac8eabc1 |
206 | |
d1083c7a |
207 | =item cmpthese ( RESULTSHASHREF, [ STYLE ] ) |
ac8eabc1 |
208 | |
d1083c7a |
209 | Optionally calls timethese(), then outputs comparison chart. This: |
ac8eabc1 |
210 | |
d1083c7a |
211 | cmpthese( -1, { a => "++\$i", b => "\$i *= 2" } ) ; |
212 | |
213 | outputs a chart like: |
214 | |
215 | Rate b a |
216 | b 2831802/s -- -61% |
217 | a 7208959/s 155% -- |
218 | |
219 | This chart is sorted from slowest to fastest, and shows the percent speed |
220 | difference between each pair of tests. |
221 | |
222 | c<cmpthese> can also be passed the data structure that timethese() returns: |
223 | |
224 | $results = timethese( -1, { a => "++\$i", b => "\$i *= 2" } ) ; |
ac8eabc1 |
225 | cmpthese( $results ); |
226 | |
d1083c7a |
227 | in case you want to see both sets of results. |
228 | |
229 | Returns a reference to an ARRAY of rows, each row is an ARRAY of cells from the |
230 | above chart, including labels. This: |
231 | |
232 | my $rows = cmpthese( -1, { a => '++$i', b => '$i *= 2' }, "none" ); |
233 | |
234 | returns a data structure like: |
235 | |
236 | [ |
237 | [ '', 'Rate', 'b', 'a' ], |
238 | [ 'b', '2885232/s', '--', '-59%' ], |
239 | [ 'a', '7099126/s', '146%', '--' ], |
240 | ] |
241 | |
242 | B<NOTE>: This result value differs from previous versions, which returned |
243 | the C<timethese()> result structure. If you want that, just use the two |
244 | statement C<timethese>...C<cmpthese> idiom shown above. |
245 | |
246 | Incidently, note the variance in the result values between the two examples; |
247 | this is typical of benchmarking. If this were a real benchmark, you would |
248 | probably want to run a lot more iterations. |
ac8eabc1 |
249 | |
250 | =item countit(TIME, CODE) |
251 | |
252 | Arguments: TIME is the minimum length of time to run CODE for, and CODE is |
253 | the code to run. CODE may be either a code reference or a string to |
254 | be eval'd; either way it will be run in the caller's package. |
255 | |
256 | TIME is I<not> negative. countit() will run the loop many times to |
257 | calculate the speed of CODE before running it for TIME. The actual |
258 | time run for will usually be greater than TIME due to system clock |
259 | resolution, so it's best to look at the number of iterations divided |
260 | by the times that you are concerned with, not just the iterations. |
261 | |
262 | Returns: a Benchmark object. |
263 | |
523cc92b |
264 | =item disablecache ( ) |
f06db76b |
265 | |
523cc92b |
266 | Disable caching of timings for the null loop. This will force Benchmark |
267 | to recalculate these timings for each new piece of code timed. |
268 | |
269 | =item enablecache ( ) |
270 | |
271 | Enable caching of timings for the null loop. The time taken for COUNT |
272 | rounds of the null loop will be calculated only once for each |
273 | different COUNT used. |
f06db76b |
274 | |
ac8eabc1 |
275 | =item timesum ( T1, T2 ) |
276 | |
277 | Returns the sum of two Benchmark times as a Benchmark object suitable |
278 | for passing to timestr(). |
279 | |
f06db76b |
280 | =back |
281 | |
e3d6de9a |
282 | =head2 :hireswallclock |
283 | |
284 | If the Time::HiRes module has been installed, you can specify the |
285 | special tag C<:hireswallclock> for Benchmark (if Time::HiRes is not |
286 | available, the tag will be silently ignored). This tag will cause the |
287 | wallclock time to be measured in microseconds, instead of integer |
702fa71c |
288 | seconds. Note though that the speed computations are still conducted |
289 | in CPU time, not wallclock time. |
e3d6de9a |
290 | |
f06db76b |
291 | =head1 NOTES |
292 | |
293 | The data is stored as a list of values from the time and times |
523cc92b |
294 | functions: |
f06db76b |
295 | |
431d98c2 |
296 | ($real, $user, $system, $children_user, $children_system, $iters) |
f06db76b |
297 | |
298 | in seconds for the whole loop (not divided by the number of rounds). |
299 | |
300 | The timing is done using time(3) and times(3). |
301 | |
302 | Code is executed in the caller's package. |
303 | |
f06db76b |
304 | The time of the null loop (a loop with the same |
305 | number of rounds but empty loop body) is subtracted |
306 | from the time of the real loop. |
307 | |
3c6312e9 |
308 | The null loop times can be cached, the key being the |
f06db76b |
309 | number of rounds. The caching can be controlled using |
310 | calls like these: |
311 | |
523cc92b |
312 | clearcache($key); |
f06db76b |
313 | clearallcache(); |
314 | |
523cc92b |
315 | disablecache(); |
f06db76b |
316 | enablecache(); |
317 | |
3c6312e9 |
318 | Caching is off by default, as it can (usually slightly) decrease |
319 | accuracy and does not usually noticably affect runtimes. |
320 | |
54e82ce5 |
321 | =head1 EXAMPLES |
322 | |
323 | For example, |
324 | |
14393033 |
325 | use Benchmark qw( cmpthese ) ; |
326 | $x = 3; |
327 | cmpthese( -5, { |
328 | a => sub{$x*$x}, |
329 | b => sub{$x**2}, |
330 | } ); |
54e82ce5 |
331 | |
332 | outputs something like this: |
333 | |
334 | Benchmark: running a, b, each for at least 5 CPU seconds... |
14393033 |
335 | Rate b a |
336 | b 1559428/s -- -62% |
337 | a 4152037/s 166% -- |
338 | |
54e82ce5 |
339 | |
340 | while |
341 | |
14393033 |
342 | use Benchmark qw( timethese cmpthese ) ; |
343 | $x = 3; |
344 | $r = timethese( -5, { |
345 | a => sub{$x*$x}, |
346 | b => sub{$x**2}, |
347 | } ); |
348 | cmpthese $r; |
54e82ce5 |
349 | |
350 | outputs something like this: |
351 | |
14393033 |
352 | Benchmark: running a, b, each for at least 5 CPU seconds... |
353 | a: 10 wallclock secs ( 5.14 usr + 0.13 sys = 5.27 CPU) @ 3835055.60/s (n=20210743) |
354 | b: 5 wallclock secs ( 5.41 usr + 0.00 sys = 5.41 CPU) @ 1574944.92/s (n=8520452) |
355 | Rate b a |
356 | b 1574945/s -- -59% |
357 | a 3835056/s 144% -- |
54e82ce5 |
358 | |
359 | |
f06db76b |
360 | =head1 INHERITANCE |
361 | |
362 | Benchmark inherits from no other class, except of course |
363 | for Exporter. |
364 | |
365 | =head1 CAVEATS |
366 | |
80eab818 |
367 | Comparing eval'd strings with code references will give you |
431d98c2 |
368 | inaccurate results: a code reference will show a slightly slower |
80eab818 |
369 | execution time than the equivalent eval'd string. |
370 | |
f06db76b |
371 | The real time timing is done using time(2) and |
372 | the granularity is therefore only one second. |
373 | |
374 | Short tests may produce negative figures because perl |
523cc92b |
375 | can appear to take longer to execute the empty loop |
376 | than a short test; try: |
f06db76b |
377 | |
378 | timethis(100,'1'); |
379 | |
380 | The system time of the null loop might be slightly |
381 | more than the system time of the loop with the actual |
a24a9dfe |
382 | code and therefore the difference might end up being E<lt> 0. |
f06db76b |
383 | |
8a4f6ac2 |
384 | =head1 SEE ALSO |
385 | |
386 | L<Devel::DProf> - a Perl code profiler |
387 | |
f06db76b |
388 | =head1 AUTHORS |
389 | |
5aabfad6 |
390 | Jarkko Hietaniemi <F<jhi@iki.fi>>, Tim Bunce <F<Tim.Bunce@ig.co.uk>> |
f06db76b |
391 | |
392 | =head1 MODIFICATION HISTORY |
393 | |
394 | September 8th, 1994; by Tim Bunce. |
395 | |
523cc92b |
396 | March 28th, 1997; by Hugo van der Sanden: added support for code |
397 | references and the already documented 'debug' method; revamped |
398 | documentation. |
f06db76b |
399 | |
6ee623d5 |
400 | April 04-07th, 1997: by Jarkko Hietaniemi, added the run-for-some-time |
401 | functionality. |
402 | |
3c6312e9 |
403 | September, 1999; by Barrie Slaymaker: math fixes and accuracy and |
404 | efficiency tweaks. Added cmpthese(). A result is now returned from |
431d98c2 |
405 | timethese(). Exposed countit() (was runfor()). |
3c6312e9 |
406 | |
0e74ff8e |
407 | December, 2001; by Nicholas Clark: make timestr() recognise the style 'none' |
408 | and return an empty string. If cmpthese is calling timethese, make it pass the |
409 | style in. (so that 'none' will suppress output). Make sub new dump its |
410 | debugging output to STDERR, to be consistent with everything else. |
411 | All bugs found while writing a regression test. |
412 | |
e3d6de9a |
413 | September, 2002; by Jarkko Hietaniemi: add ':hireswallclock' special tag. |
414 | |
523cc92b |
415 | =cut |
a0d0e21e |
416 | |
3f943bd9 |
417 | # evaluate something in a clean lexical environment |
53aa2791 |
418 | sub _doeval { no strict; eval shift } |
3f943bd9 |
419 | |
420 | # |
421 | # put any lexicals at file scope AFTER here |
422 | # |
423 | |
4aa0a1f7 |
424 | use Carp; |
a0d0e21e |
425 | use Exporter; |
53aa2791 |
426 | |
427 | our(@ISA, @EXPORT, @EXPORT_OK, %EXPORT_TAGS, $VERSION); |
428 | |
429 | @ISA=qw(Exporter); |
ac8eabc1 |
430 | @EXPORT=qw(timeit timethis timethese timediff timestr); |
431 | @EXPORT_OK=qw(timesum cmpthese countit |
432 | clearcache clearallcache disablecache enablecache); |
f36484b0 |
433 | %EXPORT_TAGS=( all => [ @EXPORT, @EXPORT_OK ] ) ; |
a0d0e21e |
434 | |
53aa2791 |
435 | $VERSION = 1.051; |
e3d6de9a |
436 | |
437 | # --- ':hireswallclock' special handling |
438 | |
439 | my $hirestime; |
440 | |
441 | sub mytime () { time } |
8a4f6ac2 |
442 | |
359218de |
443 | init(); |
a0d0e21e |
444 | |
e3d6de9a |
445 | sub BEGIN { |
446 | if (eval 'require Time::HiRes') { |
447 | import Time::HiRes qw(time); |
448 | $hirestime = \&Time::HiRes::time; |
449 | } |
450 | } |
451 | |
452 | sub import { |
453 | my $class = shift; |
454 | if (grep { $_ eq ":hireswallclock" } @_) { |
455 | @_ = grep { $_ ne ":hireswallclock" } @_; |
456 | *mytime = $hirestime if defined $hirestime; |
457 | } |
458 | Benchmark->export_to_level(1, $class, @_); |
459 | } |
460 | |
53aa2791 |
461 | our($Debug, $Min_Count, $Min_CPU, $Default_Format, $Default_Style, |
462 | %_Usage, %Cache, $Do_Cache); |
463 | |
a0d0e21e |
464 | sub init { |
53aa2791 |
465 | $Debug = 0; |
466 | $Min_Count = 4; |
467 | $Min_CPU = 0.4; |
468 | $Default_Format = '5.2f'; |
469 | $Default_Style = 'auto'; |
a0d0e21e |
470 | # The cache can cause a slight loss of sys time accuracy. If a |
471 | # user does many tests (>10) with *very* large counts (>10000) |
472 | # or works on a very slow machine the cache may be useful. |
359218de |
473 | disablecache(); |
474 | clearallcache(); |
a0d0e21e |
475 | } |
476 | |
53aa2791 |
477 | sub debug { $Debug = ($_[1] != 0); } |
478 | |
479 | sub usage { |
480 | my $calling_sub = (caller(1))[3]; |
481 | $calling_sub =~ s/^Benchmark:://; |
482 | return $_Usage{$calling_sub} || ''; |
483 | } |
484 | |
bba8fca5 |
485 | # The cache needs two branches: 's' for strings and 'c' for code. The |
359218de |
486 | # empty loop is different in these two cases. |
53aa2791 |
487 | |
f695f0e6 |
488 | $_Usage{clearcache} = <<'USAGE'; |
489 | usage: clearcache($count); |
490 | USAGE |
491 | |
492 | sub clearcache { |
493 | die usage unless @_ == 1; |
53aa2791 |
494 | delete $Cache{"$_[0]c"}; delete $Cache{"$_[0]s"}; |
495 | } |
496 | |
f695f0e6 |
497 | $_Usage{clearallcache} = <<'USAGE'; |
498 | usage: clearallcache(); |
499 | USAGE |
500 | |
501 | sub clearallcache { |
502 | die usage if @_; |
53aa2791 |
503 | %Cache = (); |
504 | } |
505 | |
f695f0e6 |
506 | $_Usage{enablecache} = <<'USAGE'; |
507 | usage: enablecache(); |
508 | USAGE |
509 | |
510 | sub enablecache { |
511 | die usage if @_; |
53aa2791 |
512 | $Do_Cache = 1; |
513 | } |
514 | |
f695f0e6 |
515 | $_Usage{disablecache} = <<'USAGE'; |
516 | usage: disablecache(); |
517 | USAGE |
518 | |
519 | sub disablecache { |
520 | die usage if @_; |
53aa2791 |
521 | $Do_Cache = 0; |
522 | } |
523 | |
a0d0e21e |
524 | |
a0d0e21e |
525 | # --- Functions to process the 'time' data type |
526 | |
e3d6de9a |
527 | sub new { my @t = (mytime, times, @_ == 2 ? $_[1] : 0); |
53aa2791 |
528 | print STDERR "new=@t\n" if $Debug; |
6ee623d5 |
529 | bless \@t; } |
a0d0e21e |
530 | |
531 | sub cpu_p { my($r,$pu,$ps,$cu,$cs) = @{$_[0]}; $pu+$ps ; } |
532 | sub cpu_c { my($r,$pu,$ps,$cu,$cs) = @{$_[0]}; $cu+$cs ; } |
533 | sub cpu_a { my($r,$pu,$ps,$cu,$cs) = @{$_[0]}; $pu+$ps+$cu+$cs ; } |
534 | sub real { my($r,$pu,$ps,$cu,$cs) = @{$_[0]}; $r ; } |
431d98c2 |
535 | sub iters { $_[0]->[5] ; } |
a0d0e21e |
536 | |
53aa2791 |
537 | |
538 | $_Usage{timediff} = <<'USAGE'; |
539 | usage: $result_diff = timediff($result1, $result2); |
540 | USAGE |
541 | |
523cc92b |
542 | sub timediff { |
a0d0e21e |
543 | my($a, $b) = @_; |
53aa2791 |
544 | |
545 | die usage unless ref $a and ref $b; |
546 | |
523cc92b |
547 | my @r; |
3f943bd9 |
548 | for (my $i=0; $i < @$a; ++$i) { |
a0d0e21e |
549 | push(@r, $a->[$i] - $b->[$i]); |
550 | } |
551 | bless \@r; |
552 | } |
553 | |
53aa2791 |
554 | $_Usage{timesum} = <<'USAGE'; |
555 | usage: $sum = timesum($result1, $result2); |
556 | USAGE |
557 | |
705cc255 |
558 | sub timesum { |
53aa2791 |
559 | my($a, $b) = @_; |
560 | |
561 | die usage unless ref $a and ref $b; |
562 | |
563 | my @r; |
564 | for (my $i=0; $i < @$a; ++$i) { |
705cc255 |
565 | push(@r, $a->[$i] + $b->[$i]); |
53aa2791 |
566 | } |
567 | bless \@r; |
705cc255 |
568 | } |
569 | |
53aa2791 |
570 | |
571 | $_Usage{timestr} = <<'USAGE'; |
572 | usage: $formatted_result = timestr($result1); |
573 | USAGE |
574 | |
523cc92b |
575 | sub timestr { |
a0d0e21e |
576 | my($tr, $style, $f) = @_; |
53aa2791 |
577 | |
578 | die usage unless ref $tr; |
579 | |
523cc92b |
580 | my @t = @$tr; |
6ee623d5 |
581 | warn "bad time value (@t)" unless @t==6; |
582 | my($r, $pu, $ps, $cu, $cs, $n) = @t; |
ce9550df |
583 | my($pt, $ct, $tt) = ($tr->cpu_p, $tr->cpu_c, $tr->cpu_a); |
53aa2791 |
584 | $f = $Default_Format unless defined $f; |
a0d0e21e |
585 | # format a time in the required style, other formats may be added here |
53aa2791 |
586 | $style ||= $Default_Style; |
0e74ff8e |
587 | return '' if $style eq 'none'; |
523cc92b |
588 | $style = ($ct>0) ? 'all' : 'noc' if $style eq 'auto'; |
589 | my $s = "@t $style"; # default for unknown style |
e3d6de9a |
590 | my $w = $hirestime ? "%2g" : "%2d"; |
591 | $s=sprintf("$w wallclock secs (%$f usr %$f sys + %$f cusr %$f csys = %$f CPU)", |
ce9550df |
592 | $r,$pu,$ps,$cu,$cs,$tt) if $style eq 'all'; |
e3d6de9a |
593 | $s=sprintf("$w wallclock secs (%$f usr + %$f sys = %$f CPU)", |
7be077a2 |
594 | $r,$pu,$ps,$pt) if $style eq 'noc'; |
e3d6de9a |
595 | $s=sprintf("$w wallclock secs (%$f cusr + %$f csys = %$f CPU)", |
7be077a2 |
596 | $r,$cu,$cs,$ct) if $style eq 'nop'; |
cc31225e |
597 | $s .= sprintf(" @ %$f/s (n=$n)", $n / ( $pu + $ps )) if $n && $pu+$ps; |
a0d0e21e |
598 | $s; |
599 | } |
523cc92b |
600 | |
601 | sub timedebug { |
a0d0e21e |
602 | my($msg, $t) = @_; |
53aa2791 |
603 | print STDERR "$msg",timestr($t),"\n" if $Debug; |
a0d0e21e |
604 | } |
605 | |
a0d0e21e |
606 | # --- Functions implementing low-level support for timing loops |
607 | |
53aa2791 |
608 | $_Usage{runloop} = <<'USAGE'; |
609 | usage: runloop($number, [$string | $coderef]) |
610 | USAGE |
611 | |
a0d0e21e |
612 | sub runloop { |
613 | my($n, $c) = @_; |
4aa0a1f7 |
614 | |
615 | $n+=0; # force numeric now, so garbage won't creep into the eval |
523cc92b |
616 | croak "negative loopcount $n" if $n<0; |
53aa2791 |
617 | confess usage unless defined $c; |
a0d0e21e |
618 | my($t0, $t1, $td); # before, after, difference |
619 | |
620 | # find package of caller so we can execute code there |
523cc92b |
621 | my($curpack) = caller(0); |
622 | my($i, $pack)= 0; |
a0d0e21e |
623 | while (($pack) = caller(++$i)) { |
624 | last if $pack ne $curpack; |
625 | } |
626 | |
3f943bd9 |
627 | my ($subcode, $subref); |
628 | if (ref $c eq 'CODE') { |
629 | $subcode = "sub { for (1 .. $n) { local \$_; package $pack; &\$c; } }"; |
630 | $subref = eval $subcode; |
631 | } |
632 | else { |
633 | $subcode = "sub { for (1 .. $n) { local \$_; package $pack; $c;} }"; |
634 | $subref = _doeval($subcode); |
635 | } |
4aa0a1f7 |
636 | croak "runloop unable to compile '$c': $@\ncode: $subcode\n" if $@; |
53aa2791 |
637 | print STDERR "runloop $n '$subcode'\n" if $Debug; |
a0d0e21e |
638 | |
3c6312e9 |
639 | # Wait for the user timer to tick. This makes the error range more like |
640 | # -0.01, +0. If we don't wait, then it's more like -0.01, +0.01. This |
641 | # may not seem important, but it significantly reduces the chances of |
642 | # getting a too low initial $n in the initial, 'find the minimum' loop |
431d98c2 |
643 | # in &countit. This, in turn, can reduce the number of calls to |
bba8fca5 |
644 | # &runloop a lot, and thus reduce additive errors. |
645 | my $tbase = Benchmark->new(0)->[1]; |
277427cf |
646 | while ( ( $t0 = Benchmark->new(0) )->[1] == $tbase ) {} ; |
a0d0e21e |
647 | &$subref; |
6ee623d5 |
648 | $t1 = Benchmark->new($n); |
a0d0e21e |
649 | $td = &timediff($t1, $t0); |
a0d0e21e |
650 | timedebug("runloop:",$td); |
651 | $td; |
652 | } |
653 | |
53aa2791 |
654 | $_Usage{timeit} = <<'USAGE'; |
655 | usage: $result = timeit($count, 'code' ); or |
656 | $result = timeit($count, sub { code } ); |
657 | USAGE |
a0d0e21e |
658 | |
659 | sub timeit { |
660 | my($n, $code) = @_; |
661 | my($wn, $wc, $wd); |
662 | |
53aa2791 |
663 | die usage unless defined $code and |
664 | (!ref $code or ref $code eq 'CODE'); |
665 | |
666 | printf STDERR "timeit $n $code\n" if $Debug; |
3c6312e9 |
667 | my $cache_key = $n . ( ref( $code ) ? 'c' : 's' ); |
53aa2791 |
668 | if ($Do_Cache && exists $Cache{$cache_key} ) { |
669 | $wn = $Cache{$cache_key}; |
523cc92b |
670 | } else { |
6bf773bc |
671 | $wn = &runloop($n, ref( $code ) ? sub { } : '' ); |
3c6312e9 |
672 | # Can't let our baseline have any iterations, or they get subtracted |
673 | # out of the result. |
674 | $wn->[5] = 0; |
53aa2791 |
675 | $Cache{$cache_key} = $wn; |
a0d0e21e |
676 | } |
677 | |
678 | $wc = &runloop($n, $code); |
679 | |
680 | $wd = timediff($wc, $wn); |
a0d0e21e |
681 | timedebug("timeit: ",$wc); |
682 | timedebug(" - ",$wn); |
683 | timedebug(" = ",$wd); |
684 | |
685 | $wd; |
686 | } |
687 | |
6ee623d5 |
688 | |
689 | my $default_for = 3; |
690 | my $min_for = 0.1; |
691 | |
3c6312e9 |
692 | |
53aa2791 |
693 | $_Usage{countit} = <<'USAGE'; |
694 | usage: $result = countit($time, 'code' ); or |
695 | $result = countit($time, sub { code } ); |
696 | USAGE |
697 | |
431d98c2 |
698 | sub countit { |
699 | my ( $tmax, $code ) = @_; |
6ee623d5 |
700 | |
53aa2791 |
701 | die usage unless @_; |
702 | |
6ee623d5 |
703 | if ( not defined $tmax or $tmax == 0 ) { |
704 | $tmax = $default_for; |
705 | } elsif ( $tmax < 0 ) { |
706 | $tmax = -$tmax; |
707 | } |
708 | |
431d98c2 |
709 | die "countit($tmax, ...): timelimit cannot be less than $min_for.\n" |
6ee623d5 |
710 | if $tmax < $min_for; |
711 | |
3c6312e9 |
712 | my ($n, $tc); |
6ee623d5 |
713 | |
bba8fca5 |
714 | # First find the minimum $n that gives a significant timing. |
3c6312e9 |
715 | for ($n = 1; ; $n *= 2 ) { |
716 | my $td = timeit($n, $code); |
717 | $tc = $td->[1] + $td->[2]; |
718 | last if $tc > 0.1; |
719 | } |
6ee623d5 |
720 | |
3c6312e9 |
721 | my $nmin = $n; |
722 | |
723 | # Get $n high enough that we can guess the final $n with some accuracy. |
724 | my $tpra = 0.1 * $tmax; # Target/time practice. |
725 | while ( $tc < $tpra ) { |
726 | # The 5% fudge is to keep us from iterating again all |
727 | # that often (this speeds overall responsiveness when $tmax is big |
728 | # and we guess a little low). This does not noticably affect |
729 | # accuracy since we're not couting these times. |
730 | $n = int( $tpra * 1.05 * $n / $tc ); # Linear approximation. |
731 | my $td = timeit($n, $code); |
c5d57293 |
732 | my $new_tc = $td->[1] + $td->[2]; |
733 | # Make sure we are making progress. |
734 | $tc = $new_tc > 1.2 * $tc ? $new_tc : 1.2 * $tc; |
6ee623d5 |
735 | } |
736 | |
3c6312e9 |
737 | # Now, do the 'for real' timing(s), repeating until we exceed |
738 | # the max. |
739 | my $ntot = 0; |
740 | my $rtot = 0; |
741 | my $utot = 0.0; |
742 | my $stot = 0.0; |
743 | my $cutot = 0.0; |
744 | my $cstot = 0.0; |
745 | my $ttot = 0.0; |
746 | |
747 | # The 5% fudge is because $n is often a few % low even for routines |
748 | # with stable times and avoiding extra timeit()s is nice for |
749 | # accuracy's sake. |
750 | $n = int( $n * ( 1.05 * $tmax / $tc ) ); |
751 | |
752 | while () { |
753 | my $td = timeit($n, $code); |
754 | $ntot += $n; |
755 | $rtot += $td->[0]; |
756 | $utot += $td->[1]; |
757 | $stot += $td->[2]; |
6ee623d5 |
758 | $cutot += $td->[3]; |
759 | $cstot += $td->[4]; |
3c6312e9 |
760 | $ttot = $utot + $stot; |
761 | last if $ttot >= $tmax; |
6ee623d5 |
762 | |
c5d57293 |
763 | $ttot = 0.01 if $ttot < 0.01; |
3c6312e9 |
764 | my $r = $tmax / $ttot - 1; # Linear approximation. |
bba8fca5 |
765 | $n = int( $r * $ntot ); |
6ee623d5 |
766 | $n = $nmin if $n < $nmin; |
6ee623d5 |
767 | } |
768 | |
769 | return bless [ $rtot, $utot, $stot, $cutot, $cstot, $ntot ]; |
770 | } |
771 | |
a0d0e21e |
772 | # --- Functions implementing high-level time-then-print utilities |
773 | |
6ee623d5 |
774 | sub n_to_for { |
775 | my $n = shift; |
776 | return $n == 0 ? $default_for : $n < 0 ? -$n : undef; |
777 | } |
778 | |
53aa2791 |
779 | $_Usage{timethis} = <<'USAGE'; |
780 | usage: $result = timethis($time, 'code' ); or |
781 | $result = timethis($time, sub { code } ); |
782 | USAGE |
783 | |
a0d0e21e |
784 | sub timethis{ |
785 | my($n, $code, $title, $style) = @_; |
53aa2791 |
786 | my($t, $forn); |
787 | |
788 | die usage unless defined $code and |
789 | (!ref $code or ref $code eq 'CODE'); |
6ee623d5 |
790 | |
791 | if ( $n > 0 ) { |
792 | croak "non-integer loopcount $n, stopped" if int($n)<$n; |
793 | $t = timeit($n, $code); |
794 | $title = "timethis $n" unless defined $title; |
795 | } else { |
53aa2791 |
796 | my $fort = n_to_for( $n ); |
431d98c2 |
797 | $t = countit( $fort, $code ); |
6ee623d5 |
798 | $title = "timethis for $fort" unless defined $title; |
799 | $forn = $t->[-1]; |
800 | } |
523cc92b |
801 | local $| = 1; |
523cc92b |
802 | $style = "" unless defined $style; |
3c6312e9 |
803 | printf("%10s: ", $title) unless $style eq 'none'; |
53aa2791 |
804 | print timestr($t, $style, $Default_Format),"\n" unless $style eq 'none'; |
6ee623d5 |
805 | |
806 | $n = $forn if defined $forn; |
523cc92b |
807 | |
a0d0e21e |
808 | # A conservative warning to spot very silly tests. |
809 | # Don't assume that your benchmark is ok simply because |
810 | # you don't get this warning! |
811 | print " (warning: too few iterations for a reliable count)\n" |
53aa2791 |
812 | if $n < $Min_Count |
a0d0e21e |
813 | || ($t->real < 1 && $n < 1000) |
53aa2791 |
814 | || $t->cpu_a < $Min_CPU; |
a0d0e21e |
815 | $t; |
816 | } |
817 | |
53aa2791 |
818 | |
819 | $_Usage{timethese} = <<'USAGE'; |
820 | usage: timethese($count, { Name1 => 'code1', ... }); or |
821 | timethese($count, { Name1 => sub { code1 }, ... }); |
822 | USAGE |
823 | |
a0d0e21e |
824 | sub timethese{ |
825 | my($n, $alt, $style) = @_; |
53aa2791 |
826 | die usage unless ref $alt eq 'HASH'; |
827 | |
523cc92b |
828 | my @names = sort keys %$alt; |
829 | $style = "" unless defined $style; |
3c6312e9 |
830 | print "Benchmark: " unless $style eq 'none'; |
6ee623d5 |
831 | if ( $n > 0 ) { |
832 | croak "non-integer loopcount $n, stopped" if int($n)<$n; |
3c6312e9 |
833 | print "timing $n iterations of" unless $style eq 'none'; |
6ee623d5 |
834 | } else { |
3c6312e9 |
835 | print "running" unless $style eq 'none'; |
6ee623d5 |
836 | } |
3c6312e9 |
837 | print " ", join(', ',@names) unless $style eq 'none'; |
6ee623d5 |
838 | unless ( $n > 0 ) { |
839 | my $for = n_to_for( $n ); |
df7779cf |
840 | print ", each" if $n > 1 && $style ne 'none'; |
841 | print " for at least $for CPU seconds" unless $style eq 'none'; |
6ee623d5 |
842 | } |
3c6312e9 |
843 | print "...\n" unless $style eq 'none'; |
523cc92b |
844 | |
845 | # we could save the results in an array and produce a summary here |
a0d0e21e |
846 | # sum, min, max, avg etc etc |
3c6312e9 |
847 | my %results; |
4dbb2df9 |
848 | foreach my $name (@names) { |
3c6312e9 |
849 | $results{$name} = timethis ($n, $alt -> {$name}, $name, $style); |
4dbb2df9 |
850 | } |
3c6312e9 |
851 | |
852 | return \%results; |
a0d0e21e |
853 | } |
854 | |
53aa2791 |
855 | |
856 | $_Usage{cmpthese} = <<'USAGE'; |
857 | usage: cmpthese($count, { Name1 => 'code1', ... }); or |
858 | cmpthese($count, { Name1 => sub { code1 }, ... }); or |
859 | cmpthese($result, $style); |
860 | USAGE |
861 | |
3c6312e9 |
862 | sub cmpthese{ |
53aa2791 |
863 | my ($results, $style); |
864 | |
865 | if( ref $_[0] ) { |
866 | ($results, $style) = @_; |
867 | } |
868 | else { |
869 | my($count, $code) = @_[0,1]; |
870 | $style = $_[2] if defined $_[2]; |
871 | |
872 | die usage unless ref $code eq 'HASH'; |
873 | |
874 | $results = timethese($count, $code, ($style || "none")); |
875 | } |
3c6312e9 |
876 | |
d1083c7a |
877 | $style = "" unless defined $style; |
3c6312e9 |
878 | |
879 | # Flatten in to an array of arrays with the name as the first field |
880 | my @vals = map{ [ $_, @{$results->{$_}} ] } keys %$results; |
881 | |
882 | for (@vals) { |
883 | # The epsilon fudge here is to prevent div by 0. Since clock |
884 | # resolutions are much larger, it's below the noise floor. |
885 | my $rate = $_->[6] / ( $_->[2] + $_->[3] + 0.000000000000001 ); |
886 | $_->[7] = $rate; |
887 | } |
888 | |
889 | # Sort by rate |
890 | @vals = sort { $a->[7] <=> $b->[7] } @vals; |
891 | |
892 | # If more than half of the rates are greater than one... |
893 | my $display_as_rate = $vals[$#vals>>1]->[7] > 1; |
894 | |
895 | my @rows; |
896 | my @col_widths; |
897 | |
898 | my @top_row = ( |
899 | '', |
900 | $display_as_rate ? 'Rate' : 's/iter', |
901 | map { $_->[0] } @vals |
902 | ); |
903 | |
904 | push @rows, \@top_row; |
905 | @col_widths = map { length( $_ ) } @top_row; |
906 | |
907 | # Build the data rows |
908 | # We leave the last column in even though it never has any data. Perhaps |
909 | # it should go away. Also, perhaps a style for a single column of |
910 | # percentages might be nice. |
911 | for my $row_val ( @vals ) { |
912 | my @row; |
913 | |
914 | # Column 0 = test name |
915 | push @row, $row_val->[0]; |
916 | $col_widths[0] = length( $row_val->[0] ) |
917 | if length( $row_val->[0] ) > $col_widths[0]; |
918 | |
919 | # Column 1 = performance |
920 | my $row_rate = $row_val->[7]; |
921 | |
922 | # We assume that we'll never get a 0 rate. |
53aa2791 |
923 | my $rate = $display_as_rate ? $row_rate : 1 / $row_rate; |
3c6312e9 |
924 | |
925 | # Only give a few decimal places before switching to sci. notation, |
926 | # since the results aren't usually that accurate anyway. |
927 | my $format = |
53aa2791 |
928 | $rate >= 100 ? |
3c6312e9 |
929 | "%0.0f" : |
53aa2791 |
930 | $rate >= 10 ? |
3c6312e9 |
931 | "%0.1f" : |
53aa2791 |
932 | $rate >= 1 ? |
3c6312e9 |
933 | "%0.2f" : |
53aa2791 |
934 | $rate >= 0.1 ? |
3c6312e9 |
935 | "%0.3f" : |
936 | "%0.2e"; |
937 | |
938 | $format .= "/s" |
939 | if $display_as_rate; |
53aa2791 |
940 | |
941 | my $formatted_rate = sprintf( $format, $rate ); |
942 | push @row, $formatted_rate; |
943 | $col_widths[1] = length( $formatted_rate ) |
944 | if length( $formatted_rate ) > $col_widths[1]; |
3c6312e9 |
945 | |
946 | # Columns 2..N = performance ratios |
947 | my $skip_rest = 0; |
948 | for ( my $col_num = 0 ; $col_num < @vals ; ++$col_num ) { |
949 | my $col_val = $vals[$col_num]; |
950 | my $out; |
951 | if ( $skip_rest ) { |
952 | $out = ''; |
953 | } |
954 | elsif ( $col_val->[0] eq $row_val->[0] ) { |
955 | $out = "--"; |
956 | # $skip_rest = 1; |
957 | } |
958 | else { |
959 | my $col_rate = $col_val->[7]; |
960 | $out = sprintf( "%.0f%%", 100*$row_rate/$col_rate - 100 ); |
961 | } |
962 | push @row, $out; |
963 | $col_widths[$col_num+2] = length( $out ) |
964 | if length( $out ) > $col_widths[$col_num+2]; |
965 | |
966 | # A little wierdness to set the first column width properly |
967 | $col_widths[$col_num+2] = length( $col_val->[0] ) |
968 | if length( $col_val->[0] ) > $col_widths[$col_num+2]; |
969 | } |
970 | push @rows, \@row; |
971 | } |
972 | |
d1083c7a |
973 | return \@rows if $style eq "none"; |
974 | |
3c6312e9 |
975 | # Equalize column widths in the chart as much as possible without |
976 | # exceeding 80 characters. This does not use or affect cols 0 or 1. |
977 | my @sorted_width_refs = |
978 | sort { $$a <=> $$b } map { \$_ } @col_widths[2..$#col_widths]; |
979 | my $max_width = ${$sorted_width_refs[-1]}; |
980 | |
277427cf |
981 | my $total = @col_widths - 1 ; |
3c6312e9 |
982 | for ( @col_widths ) { $total += $_ } |
983 | |
984 | STRETCHER: |
985 | while ( $total < 80 ) { |
986 | my $min_width = ${$sorted_width_refs[0]}; |
987 | last |
988 | if $min_width == $max_width; |
989 | for ( @sorted_width_refs ) { |
990 | last |
991 | if $$_ > $min_width; |
992 | ++$$_; |
993 | ++$total; |
994 | last STRETCHER |
995 | if $total >= 80; |
996 | } |
997 | } |
998 | |
999 | # Dump the output |
1000 | my $format = join( ' ', map { "%${_}s" } @col_widths ) . "\n"; |
1001 | substr( $format, 1, 0 ) = '-'; |
1002 | for ( @rows ) { |
1003 | printf $format, @$_; |
1004 | } |
1005 | |
d1083c7a |
1006 | return \@rows ; |
3c6312e9 |
1007 | } |
1008 | |
1009 | |
a0d0e21e |
1010 | 1; |