Commit | Line | Data |
da096611 |
1 | =head1 NAME |
2 | |
3 | perlperf - Perl Performance and Optimization Techniques |
4 | |
5 | =head1 DESCRIPTION |
6 | |
7 | This is an introduction to the use of performance and optimization techniques |
e4866e2e |
8 | which can be used with particular reference to perl programs. While many perl |
da096611 |
9 | developers have come from other languages, and can use their prior knowledge |
10 | where appropriate, there are many other people who might benefit from a few |
11 | perl specific pointers. If you want the condensed version, perhaps the best |
12 | advice comes from the renowned Japanese Samurai, Miyamoto Musashi, who said: |
13 | |
14 | "Do Not Engage in Useless Activity" |
15 | |
16 | in 1645. |
17 | |
18 | =head1 OVERVIEW |
19 | |
20 | Perhaps the most common mistake programmers make is to attempt to optimize |
21 | their code before a program actually does anything useful - this is a bad idea. |
22 | There's no point in having an extremely fast program that doesn't work. The |
23 | first job is to get a program to I<correctly> do something B<useful>, (not to |
24 | mention ensuring the test suite is fully functional), and only then to consider |
25 | optimizing it. Having decided to optimize existing working code, there are |
26 | several simple but essential steps to consider which are intrinsic to any |
27 | optimization process. |
28 | |
29 | =head2 ONE STEP SIDEWAYS |
30 | |
31 | Firstly, you need to establish a baseline time for the existing code, which |
32 | timing needs to be reliable and repeatable. You'll probably want to use the |
33 | C<Benchmark> or C<Devel::DProf> modules, or something similar, for this step, |
34 | or perhaps the unix system C<time> utility, whichever is appropriate. See the |
35 | base of this document for a longer list of benchmarking and profiling modules, |
36 | and recommended further reading. |
37 | |
38 | =head2 ONE STEP FORWARD |
39 | |
40 | Next, having examined the program for I<hot spots>, (places where the code |
41 | seems to run slowly), change the code with the intention of making it run |
42 | faster. Using version control software, like C<subversion>, will ensure no |
43 | changes are irreversible. It's too easy to fiddle here and fiddle there - |
44 | don't change too much at any one time or you might not discover which piece of |
45 | code B<really> was the slow bit. |
46 | |
47 | =head2 ANOTHER STEP SIDEWAYS |
48 | |
49 | It's not enough to say: "that will make it run faster", you have to check it. |
50 | Rerun the code under control of the benchmarking or profiling modules, from the |
51 | first step above, and check that the new code executed the B<same task> in |
52 | I<less time>. Save your work and repeat... |
53 | |
54 | =head1 GENERAL GUIDELINES |
55 | |
56 | The critical thing when considering performance is to remember there is no such |
57 | thing as a C<Golden Bullet>, which is why there are no rules, only guidelines. |
58 | |
59 | It is clear that inline code is going to be faster than subroutine or method |
60 | calls, because there is less overhead, but this approach has the disadvantage |
61 | of being less maintainable and comes at the cost of greater memory usage - |
62 | there is no such thing as a free lunch. If you are searching for an element in |
63 | a list, it can be more efficient to store the data in a hash structure, and |
64 | then simply look to see whether the key is defined, rather than to loop through |
65 | the entire array using grep() for instance. substr() may be (a lot) faster |
66 | than grep() but not as flexible, so you have another trade-off to access. Your |
67 | code may contain a line which takes 0.01 of a second to execute which if you |
68 | call it 1,000 times, quite likely in a program parsing even medium sized files |
69 | for instance, you already have a 10 second delay, in just one single code |
70 | location, and if you call that line 100,000 times, your entire program will |
71 | slow down to an unbearable crawl. |
72 | |
73 | Using a subroutine as part of your sort is a powerful way to get exactly what |
74 | you want, but will usually be slower than the built-in I<alphabetic> C<cmp> and |
75 | I<numeric> C<E<lt>=E<gt>> sort operators. It is possible to make multiple |
76 | passes over your data, building indices to make the upcoming sort more |
77 | efficient, and to use what is known as the C<OM> (Orcish Maneuver) to cache the |
78 | sort keys in advance. The cache lookup, while a good idea, can itself be a |
79 | source of slowdown by enforcing a double pass over the data - once to setup the |
80 | cache, and once to sort the data. Using C<pack()> to extract the required sort |
81 | key into a consistent string can be an efficient way to build a single string |
82 | to compare, instead of using multiple sort keys, which makes it possible to use |
83 | the standard, written in C<c> and fast, perl C<sort()> function on the output, |
84 | and is the basis of the C<GRT> (Guttman Rossler Transform). Some string |
85 | combinations can slow the C<GRT> down, by just being too plain complex for it's |
86 | own good. |
87 | |
88 | For applications using database backends, the standard C<DBIx> namespace has |
89 | tries to help with keeping things nippy, not least because it tries to I<not> |
90 | query the database until the latest possible moment, but always read the docs |
91 | which come with your choice of libraries. Among the many issues facing |
92 | developers dealing with databases should remain aware of is to always use |
93 | C<SQL> placeholders and to consider pre-fetching data sets when this might |
94 | prove advantageous. Splitting up a large file by assigning multiple processes |
95 | to parsing a single file, using say C<POE>, C<threads> or C<fork> can also be a |
96 | useful way of optimizing your usage of the available C<CPU> resources, though |
97 | this technique is fraught with concurrency issues and demands high attention to |
98 | detail. |
99 | |
100 | Every case has a specific application and one or more exceptions, and there is |
101 | no replacement for running a few tests and finding out which method works best |
102 | for your particular environment, this is why writing optimal code is not an |
103 | exact science, and why we love using Perl so much - TMTOWTDI. |
104 | |
105 | =head1 BENCHMARKS |
106 | |
107 | Here are a few examples to demonstrate usage of Perl's benchmarking tools. |
108 | |
109 | =head2 Assigning and Dereferencing Variables. |
110 | |
111 | I'm sure most of us have seen code which looks like, (or worse than), this: |
112 | |
113 | if ( $obj->{_ref}->{_myscore} >= $obj->{_ref}->{_yourscore} ) { |
114 | ... |
115 | |
116 | This sort of code can be a real eyesore to read, as well as being very |
117 | sensitive to typos, and it's much clearer to dereference the variable |
118 | explicitly. We're side-stepping the issue of working with object-oriented |
119 | programming techniques to encapsulate variable access via methods, only |
120 | accessible through an object. Here we're just discussing the technical |
121 | implementation of choice, and whether this has an effect on performance. We |
122 | can see whether this dereferencing operation, has any overhead by putting |
123 | comparative code in a file and running a C<Benchmark> test. |
124 | |
125 | # dereference |
126 | |
127 | #!/usr/bin/perl |
128 | |
129 | use strict; |
130 | use warnings; |
131 | |
132 | use Benchmark; |
133 | |
134 | my $ref = { |
135 | 'ref' => { |
136 | _myscore => '100 + 1', |
137 | _yourscore => '102 - 1', |
138 | }, |
139 | }; |
140 | |
141 | timethese(1000000, { |
142 | 'direct' => sub { |
143 | my $x = $ref->{ref}->{_myscore} . $ref->{ref}->{_yourscore} ; |
144 | }, |
145 | 'dereference' => sub { |
146 | my $ref = $ref->{ref}; |
147 | my $myscore = $ref->{_myscore}; |
148 | my $yourscore = $ref->{_yourscore}; |
149 | my $x = $myscore . $yourscore; |
150 | }, |
151 | }); |
152 | |
153 | It's essential to run any timing measurements a sufficient number of times so |
154 | the numbers settle on a numerical average, otherwise each run will naturally |
155 | fluctuate due to variations in the environment, to reduce the effect of |
156 | contention for C<CPU> resources and network bandwidth for instance. Running |
157 | the above code for one million iterations, we can take a look at the report |
158 | output by the C<Benchmark> module, to see which approach is the most effective. |
159 | |
160 | $> perl dereference |
161 | |
162 | Benchmark: timing 1000000 iterations of dereference, direct... |
163 | dereference: 2 wallclock secs ( 1.59 usr + 0.00 sys = 1.59 CPU) @ 628930.82/s (n=1000000) |
164 | direct: 1 wallclock secs ( 1.20 usr + 0.00 sys = 1.20 CPU) @ 833333.33/s (n=1000000) |
165 | |
166 | The difference is clear to see and the dereferencing approach is slower. While |
167 | it managed to execute an average of 628,930 times a second during our test, the |
168 | direct approach managed to run an additional 204,403 times, unfortunately. |
169 | Unfortunately, because there are many examples of code written using the |
170 | multiple layer direct variable access, and it's usually horrible. It is, |
171 | however, miniscully faster. The question remains whether the minute gain is |
172 | actually worth the eyestrain, or the loss of maintainability. |
173 | |
174 | =head2 Search and replace or tr |
175 | |
176 | If we have a string which needs to be modified, while a regex will almost |
177 | always be much more flexible, C<tr>, an oft underused tool, can still be a |
178 | useful. One scenario might be replace all vowels with another character. The |
179 | regex solution might look like this: |
180 | |
181 | $str =~ s/[aeiou]/x/g |
182 | |
183 | The C<tr> alternative might look like this: |
184 | |
185 | $str =~ tr/aeiou/xxxxx/ |
186 | |
187 | We can put that into a test file which we can run to check which approach is |
188 | the fastest, using a global C<$STR> variable to assign to the C<my $str> |
189 | variable so as to avoid perl trying to optimize any of the work away by |
190 | noticing it's assigned only the once. |
191 | |
192 | # regex-transliterate |
193 | |
194 | #!/usr/bin/perl |
195 | |
196 | use strict; |
197 | use warnings; |
198 | |
199 | use Benchmark; |
200 | |
201 | my $STR = "$$-this and that"; |
202 | |
203 | timethese( 1000000, { |
204 | 'sr' => sub { my $str = $STR; $str =~ s/[aeiou]/x/g; return $str; }, |
205 | 'tr' => sub { my $str = $STR; $str =~ tr/aeiou/xxxxx/; return $str; }, |
206 | }); |
207 | |
208 | Running the code gives us our results: |
209 | |
210 | $> perl regex-transliterate |
211 | |
212 | Benchmark: timing 1000000 iterations of sr, tr... |
213 | sr: 2 wallclock secs ( 1.19 usr + 0.00 sys = 1.19 CPU) @ 840336.13/s (n=1000000) |
214 | tr: 0 wallclock secs ( 0.49 usr + 0.00 sys = 0.49 CPU) @ 2040816.33/s (n=1000000) |
215 | |
216 | The C<tr> version is a clear winner. One solution is flexible, the other is |
217 | fast - and it's appropriately the programmers choice which to use in the |
218 | circumstances. |
219 | |
220 | Check the C<Benchmark> docs for further useful techniques. |
221 | |
222 | =head1 PROFILING TOOLS |
223 | |
224 | A slightly larger piece of code will provide something on which a profiler can |
225 | produce more extensive reporting statistics. This example uses the simplistic |
226 | C<wordmatch> program which parses a given input file and spews out a short |
227 | report on the contents. |
228 | |
229 | # wordmatch |
230 | |
231 | #!/usr/bin/perl |
232 | |
233 | use strict; |
234 | use warnings; |
235 | |
236 | =head1 NAME |
237 | |
238 | filewords - word analysis of input file |
239 | |
240 | =head1 SYNOPSIS |
241 | |
242 | filewords -f inputfilename [-d] |
243 | |
244 | =head1 DESCRIPTION |
245 | |
246 | This program parses the given filename, specified with C<-f>, and displays a |
247 | simple analysis of the words found therein. Use the C<-d> switch to enable |
248 | debugging messages. |
249 | |
250 | =cut |
251 | |
252 | use FileHandle; |
253 | use Getopt::Long; |
254 | |
255 | my $debug = 0; |
256 | my $file = ''; |
257 | |
258 | my $result = GetOptions ( |
259 | 'debug' => \$debug, |
260 | 'file=s' => \$file, |
261 | ); |
262 | die("invalid args") unless $result; |
263 | |
264 | unless ( -f $file ) { |
265 | die("Usage: $0 -f filename [-d]"); |
266 | } |
267 | my $FH = FileHandle->new("< $file") or die("unable to open file($file): $!"); |
268 | |
269 | my $i_LINES = 0; |
270 | my $i_WORDS = 0; |
271 | my %count = (); |
272 | |
273 | my @lines = <$FH>; |
274 | foreach my $line ( @lines ) { |
275 | $i_LINES++; |
276 | $line =~ s/\n//; |
277 | my @words = split(/ +/, $line); |
278 | my $i_words = scalar(@words); |
279 | $i_WORDS = $i_WORDS + $i_words; |
280 | debug("line: $i_LINES supplying $i_words words: @words"); |
281 | my $i_word = 0; |
282 | foreach my $word ( @words ) { |
283 | $i_word++; |
284 | $count{$i_LINES}{spec} += matches($i_word, $word, '[^a-zA-Z0-9]'); |
285 | $count{$i_LINES}{only} += matches($i_word, $word, '^[^a-zA-Z0-9]+$'); |
286 | $count{$i_LINES}{cons} += matches($i_word, $word, '^[(?i:bcdfghjklmnpqrstvwxyz)]+$'); |
287 | $count{$i_LINES}{vows} += matches($i_word, $word, '^[(?i:aeiou)]+$'); |
288 | $count{$i_LINES}{caps} += matches($i_word, $word, '^[(A-Z)]+$'); |
289 | } |
290 | } |
291 | |
292 | print report( %count ); |
293 | |
294 | sub matches { |
295 | my $i_wd = shift; |
296 | my $word = shift; |
297 | my $regex = shift; |
298 | my $has = 0; |
299 | |
300 | if ( $word =~ /($regex)/ ) { |
301 | $has++ if $1; |
302 | } |
303 | |
304 | debug("word: $i_wd ".($has ? 'matches' : 'does not match')." chars: /$regex/"); |
305 | |
306 | return $has; |
307 | } |
308 | |
309 | sub report { |
310 | my %report = @_; |
311 | my %rep; |
312 | |
313 | foreach my $line ( keys %report ) { |
314 | foreach my $key ( keys %{ $report{$line} } ) { |
315 | $rep{$key} += $report{$line}{$key}; |
316 | } |
317 | } |
318 | |
319 | my $report = qq| |
320 | $0 report for $file: |
321 | lines in file: $i_LINES |
322 | words in file: $i_WORDS |
323 | words with special (non-word) characters: $i_spec |
324 | words with only special (non-word) characters: $i_only |
325 | words with only consonants: $i_cons |
326 | words with only capital letters: $i_caps |
327 | words with only vowels: $i_vows |
328 | |; |
329 | |
330 | return $report; |
331 | } |
332 | |
333 | sub debug { |
334 | my $message = shift; |
335 | |
336 | if ( $debug ) { |
337 | print STDERR "DBG: $message\n"; |
338 | } |
339 | } |
340 | |
341 | exit 0; |
342 | |
343 | =head2 Devel::DProf |
344 | |
345 | This venerable module has been the de-facto standard for Perl code profiling |
346 | for more than a decade, but has been replaced by a number of other modules |
347 | which have brought us back to the 21st century. Although you're recommended to |
348 | evaluate your tool from the several mentioned here and from the CPAN list at |
349 | the base of this document, (and currently L<Devel::NYTProf> seems to be the |
350 | weapon of choice - see below), we'll take a quick look at the output from |
351 | L<Devel::DProf> first, to set a baseline for Perl profiling tools. Run the |
352 | above program under the control of C<Devel::DProf> by using the C<-d> switch on |
353 | the command-line. |
354 | |
355 | $> perl -d:DProf wordmatch -f perl5db.pl |
356 | |
357 | <...multiple lines snipped...> |
358 | |
359 | wordmatch report for perl5db.pl: |
360 | lines in file: 9428 |
361 | words in file: 50243 |
362 | words with special (non-word) characters: 20480 |
363 | words with only special (non-word) characters: 7790 |
364 | words with only consonants: 4801 |
365 | words with only capital letters: 1316 |
366 | words with only vowels: 1701 |
367 | |
368 | C<Devel::DProf> produces a special file, called F<tmon.out> by default, and |
369 | this file is read by the C<dprofpp> program, which is already installed as part |
370 | of the C<Devel::DProf> distribution. If you call C<dprofpp> with no options, |
371 | it will read the F<tmon.out> file in the current directory and produce a human |
372 | readable statistics report of the run of your program. Note that this may take |
373 | a little time. |
374 | |
375 | $> dprofpp |
376 | |
377 | Total Elapsed Time = 2.951677 Seconds |
378 | User+System Time = 2.871677 Seconds |
379 | Exclusive Times |
380 | %Time ExclSec CumulS #Calls sec/call Csec/c Name |
381 | 102. 2.945 3.003 251215 0.0000 0.0000 main::matches |
382 | 2.40 0.069 0.069 260643 0.0000 0.0000 main::debug |
383 | 1.74 0.050 0.050 1 0.0500 0.0500 main::report |
384 | 1.04 0.030 0.049 4 0.0075 0.0123 main::BEGIN |
385 | 0.35 0.010 0.010 3 0.0033 0.0033 Exporter::as_heavy |
386 | 0.35 0.010 0.010 7 0.0014 0.0014 IO::File::BEGIN |
387 | 0.00 - -0.000 1 - - Getopt::Long::FindOption |
388 | 0.00 - -0.000 1 - - Symbol::BEGIN |
389 | 0.00 - -0.000 1 - - Fcntl::BEGIN |
390 | 0.00 - -0.000 1 - - Fcntl::bootstrap |
391 | 0.00 - -0.000 1 - - warnings::BEGIN |
392 | 0.00 - -0.000 1 - - IO::bootstrap |
393 | 0.00 - -0.000 1 - - Getopt::Long::ConfigDefaults |
394 | 0.00 - -0.000 1 - - Getopt::Long::Configure |
395 | 0.00 - -0.000 1 - - Symbol::gensym |
396 | |
397 | C<dprofpp> will produce some quite detailed reporting on the activity of the |
398 | C<wordmatch> program. The wallclock, user and system, times are at the top of |
399 | the analysis, and after this are the main columns defining which define the |
400 | report. Check the C<dprofpp> docs for details of the many options it supports. |
401 | |
402 | See also C<Apache::DProf> which hooks C<Devel::DProf> into C<mod_perl>. |
403 | |
404 | =head2 Devel::Profiler |
405 | |
406 | Let's take a look at the same program using a different profiler: |
407 | C<Devel::Profiler>, a drop-in Perl-only replacement for C<Devel::DProf>. The |
408 | usage is very slightly different in that instead of using the special C<-d:> |
409 | flag, you pull C<Devel::Profiler> in directly as a module using C<-M>. |
410 | |
411 | $> perl -MDevel::Profiler wordmatch -f perl5db.pl |
412 | |
413 | <...multiple lines snipped...> |
414 | |
415 | wordmatch report for perl5db.pl: |
416 | lines in file: 9428 |
417 | words in file: 50243 |
418 | words with special (non-word) characters: 20480 |
419 | words with only special (non-word) characters: 7790 |
420 | words with only consonants: 4801 |
421 | words with only capital letters: 1316 |
422 | words with only vowels: 1701 |
423 | |
424 | |
425 | C<Devel::Profiler> generates a tmon.out file which is compatible with the |
426 | C<dprofpp> program, thus saving the construction of a dedicated statistics |
427 | reader program. C<dprofpp> usage is therefore identical to the above example. |
428 | |
429 | $> dprofpp |
430 | |
431 | Total Elapsed Time = 20.984 Seconds |
432 | User+System Time = 19.981 Seconds |
433 | Exclusive Times |
434 | %Time ExclSec CumulS #Calls sec/call Csec/c Name |
435 | 49.0 9.792 14.509 251215 0.0000 0.0001 main::matches |
436 | 24.4 4.887 4.887 260643 0.0000 0.0000 main::debug |
437 | 0.25 0.049 0.049 1 0.0490 0.0490 main::report |
438 | 0.00 0.000 0.000 1 0.0000 0.0000 Getopt::Long::GetOptions |
439 | 0.00 0.000 0.000 2 0.0000 0.0000 Getopt::Long::ParseOptionSpec |
440 | 0.00 0.000 0.000 1 0.0000 0.0000 Getopt::Long::FindOption |
441 | 0.00 0.000 0.000 1 0.0000 0.0000 IO::File::new |
442 | 0.00 0.000 0.000 1 0.0000 0.0000 IO::Handle::new |
443 | 0.00 0.000 0.000 1 0.0000 0.0000 Symbol::gensym |
444 | 0.00 0.000 0.000 1 0.0000 0.0000 IO::File::open |
445 | |
446 | Interestingly we get slightly different results, which is mostly because the |
447 | algorithm which generates the report is different, even though the output file |
448 | format was allegedly identical. The elapsed, user and system times are clearly |
449 | showing the time it took for C<Devel::Profiler> to execute it's own run, but |
450 | the column listings feel more accurate somehow than the ones we had earlier |
451 | from C<Devel::DProf>. The 102% figure has disappeared, for example. This is |
452 | where we have to use the tools at our disposal, and recognise their pros and |
453 | cons, before using them. Interestingly, the numbers of calls for each |
454 | subroutine are identical in the two reports, it's the percentages which differ. |
455 | As the author of C<Devel::Proviler> writes: |
456 | |
457 | ...running HTML::Template's test suite under Devel::DProf shows output() |
458 | taking NO time but Devel::Profiler shows around 10% of the time is in output(). |
459 | I don't know which to trust but my gut tells me something is wrong with |
460 | Devel::DProf. HTML::Template::output() is a big routine that's called for |
461 | every test. Either way, something needs fixing. |
462 | |
463 | YMMV. |
464 | |
465 | See also C<Devel::Apache::Profiler> which hooks C<Devel::Profiler> into C<mod_perl>. |
466 | |
467 | =head2 Devel::SmallProf |
468 | |
469 | The C<Devel::SmallProf> profiler examines the runtime of your Perl program and |
470 | produces a line-by-line listing to show how many times each line was called, |
471 | and how long each line took to execute. It is called by supplying the familiar |
472 | C<-d> flag to Perl at runtime. |
473 | |
474 | $> perl -d:SmallProf wordmatch -f perl5db.pl |
475 | |
476 | <...multiple lines snipped...> |
477 | |
478 | wordmatch report for perl5db.pl: |
479 | lines in file: 9428 |
480 | words in file: 50243 |
481 | words with special (non-word) characters: 20480 |
482 | words with only special (non-word) characters: 7790 |
483 | words with only consonants: 4801 |
484 | words with only capital letters: 1316 |
485 | words with only vowels: 1701 |
486 | |
487 | C<Devel::SmallProf> writes it's output into a file called F<smallprof.out>, by |
488 | default. The format of the file looks like this: |
489 | |
490 | <num> <time> <ctime> <line>:<text> |
491 | |
492 | When the program has terminated, the output may be examined and sorted using |
493 | any standard text filtering utilities. Something like the following may be |
494 | sufficient: |
495 | |
496 | $> cat smallprof.out | grep \d*: | sort -k3 | tac | head -n20 |
497 | |
498 | 251215 1.65674 7.68000 75: if ( $word =~ /($regex)/ ) { |
499 | 251215 0.03264 4.40000 79: debug("word: $i_wd ".($has ? 'matches' : |
500 | 251215 0.02693 4.10000 81: return $has; |
501 | 260643 0.02841 4.07000 128: if ( $debug ) { |
502 | 260643 0.02601 4.04000 126: my $message = shift; |
503 | 251215 0.02641 3.91000 73: my $has = 0; |
504 | 251215 0.03311 3.71000 70: my $i_wd = shift; |
505 | 251215 0.02699 3.69000 72: my $regex = shift; |
506 | 251215 0.02766 3.68000 71: my $word = shift; |
507 | 50243 0.59726 1.00000 59: $count{$i_LINES}{cons} = |
508 | 50243 0.48175 0.92000 61: $count{$i_LINES}{spec} = |
509 | 50243 0.00644 0.89000 56: my $i_cons = matches($i_word, $word, |
510 | 50243 0.48837 0.88000 63: $count{$i_LINES}{caps} = |
511 | 50243 0.00516 0.88000 58: my $i_caps = matches($i_word, $word, '^[(A- |
512 | 50243 0.00631 0.81000 54: my $i_spec = matches($i_word, $word, '[^a- |
513 | 50243 0.00496 0.80000 57: my $i_vows = matches($i_word, $word, |
514 | 50243 0.00688 0.80000 53: $i_word++; |
515 | 50243 0.48469 0.79000 62: $count{$i_LINES}{only} = |
516 | 50243 0.48928 0.77000 60: $count{$i_LINES}{vows} = |
517 | 50243 0.00683 0.75000 55: my $i_only = matches($i_word, $word, '^[^a- |
518 | |
519 | You can immediately see a slightly different focus to the subroutine profiling |
520 | modules, and we start to see exactly which line of code is taking the most |
521 | time. That regex line is looking a bit suspicious, for example. Remember that |
522 | these tools are supposed to be used together, there is no single best way to |
523 | profile your code, you need to use the best tools for the job. |
524 | |
525 | See also C<Apache::SmallProf> which hooks C<Devel::SmallProf> into C<mod_perl>. |
526 | |
527 | =head2 Devel::FastProf |
528 | |
529 | C<Devel::FastProf> is another Perl line profiler. This was written with a view |
530 | to getting a faster line profiler, than is possible with for example |
531 | C<Devel::SmallProf>, because it's written in C<C>. To use C<Devel::FastProf>, |
532 | supply the C<-d> argument to Perl: |
533 | |
534 | $> perl -d:FastProf wordmatch -f perl5db.pl |
535 | |
536 | <...multiple lines snipped...> |
537 | |
538 | wordmatch report for perl5db.pl: |
539 | lines in file: 9428 |
540 | words in file: 50243 |
541 | words with special (non-word) characters: 20480 |
542 | words with only special (non-word) characters: 7790 |
543 | words with only consonants: 4801 |
544 | words with only capital letters: 1316 |
545 | words with only vowels: 1701 |
546 | |
547 | C<Devel::FastProf> writes statistics to the file F<fastprof.out> in the current |
548 | directory. The output file, which can be specified, can be interpreted by using |
549 | the C<fprofpp> command-line program. |
550 | |
551 | $> fprofpp | head -n20 |
552 | |
553 | # fprofpp output format is: |
554 | # filename:line time count: source |
555 | wordmatch:75 3.93338 251215: if ( $word =~ /($regex)/ ) { |
556 | wordmatch:79 1.77774 251215: debug("word: $i_wd ".($has ? 'matches' : 'does not match')." chars: /$regex/"); |
557 | wordmatch:81 1.47604 251215: return $has; |
558 | wordmatch:126 1.43441 260643: my $message = shift; |
559 | wordmatch:128 1.42156 260643: if ( $debug ) { |
560 | wordmatch:70 1.36824 251215: my $i_wd = shift; |
561 | wordmatch:71 1.36739 251215: my $word = shift; |
562 | wordmatch:72 1.35939 251215: my $regex = shift; |
563 | |
564 | Straightaway we can see that the number of times each line has been called is |
565 | identical to the C<Devel::SmallProf> output, and the sequence is only very |
566 | slightly different based on the ordering of the amount of time each line took |
567 | to execute, C<if ( $debug ) { > and C<my $message = shift;>, for example. The |
568 | differences in the actual times recorded might be in the algorithm used |
569 | internally, or it could be due to system resource limitations or contention. |
570 | |
571 | See also the L<DBIx::Profiler> which will profile database queries running |
572 | under the C<DBIx::*> namespace. |
573 | |
574 | =head2 Devel::NYTProf |
575 | |
576 | C<Devel::NYTProf> is the B<next generation> of Perl code profiler, fixing many |
577 | shortcomings in other tools and implementing many cool features. First of all it |
578 | can be used as either a I<line> profiler, a I<block> or a I<subroutine> |
579 | profiler, all at once. It can also use sub-microsecond (100ns) resolution on |
580 | systems which provide C<clock_gettime()>. It can be started and stopped even |
581 | by the program being profiled. It's a one-line entry to profile C<mod_perl> |
582 | applications. It's written in C<c> and is probably the fastest profiler |
583 | available for Perl. The list of coolness just goes on. Enough of that, let's |
584 | see how to it works - just use the familiar C<-d> switch to plug it in and run |
585 | the code. |
586 | |
587 | $> perl -d:NYTProf wordmatch -f perl5db.pl |
588 | |
589 | wordmatch report for perl5db.pl: |
590 | lines in file: 9427 |
591 | words in file: 50243 |
592 | words with special (non-word) characters: 20480 |
593 | words with only special (non-word) characters: 7790 |
594 | words with only consonants: 4801 |
595 | words with only capital letters: 1316 |
596 | words with only vowels: 1701 |
597 | |
598 | C<NYTProf> will generate a report database into the file F<nytprof.out> by |
599 | default. Human readable reports can be generated from here by using the |
600 | supplied C<nytprofhtml> (HTML output) and C<nytprofcsv> (CSV output) programs. |
601 | We've used the unix sytem C<html2text> utility to convert the |
602 | F<nytprof/index.html> file for convenience here. |
603 | |
604 | $> html2text nytprof/index.html |
605 | |
606 | Performance Profile Index |
607 | For wordmatch |
608 | Run on Fri Sep 26 13:46:39 2008 |
609 | Reported on Fri Sep 26 13:47:23 2008 |
610 | |
611 | Top 15 Subroutines -- ordered by exclusive time |
612 | |Calls |P |F |Inclusive|Exclusive|Subroutine | |
613 | | | | |Time |Time | | |
614 | |251215|5 |1 |13.09263 |10.47692 |main:: |matches | |
615 | |260642|2 |1 |2.71199 |2.71199 |main:: |debug | |
616 | |1 |1 |1 |0.21404 |0.21404 |main:: |report | |
617 | |2 |2 |2 |0.00511 |0.00511 |XSLoader:: |load (xsub) | |
618 | |14 |14|7 |0.00304 |0.00298 |Exporter:: |import | |
619 | |3 |1 |1 |0.00265 |0.00254 |Exporter:: |as_heavy | |
620 | |10 |10|4 |0.00140 |0.00140 |vars:: |import | |
621 | |13 |13|1 |0.00129 |0.00109 |constant:: |import | |
622 | |1 |1 |1 |0.00360 |0.00096 |FileHandle:: |import | |
623 | |3 |3 |3 |0.00086 |0.00074 |warnings::register::|import | |
624 | |9 |3 |1 |0.00036 |0.00036 |strict:: |bits | |
625 | |13 |13|13|0.00032 |0.00029 |strict:: |import | |
626 | |2 |2 |2 |0.00020 |0.00020 |warnings:: |import | |
627 | |2 |1 |1 |0.00020 |0.00020 |Getopt::Long:: |ParseOptionSpec| |
628 | |7 |7 |6 |0.00043 |0.00020 |strict:: |unimport | |
629 | |
630 | For more information see the full list of 189 subroutines. |
631 | |
632 | The first part of the report already shows the critical information regarding |
633 | which subroutines are using the most time. The next gives some statistics |
634 | about the source files profiled. |
635 | |
636 | Source Code Files -- ordered by exclusive time then name |
637 | |Stmts |Exclusive|Avg. |Reports |Source File | |
638 | | |Time | | | | |
639 | |2699761|15.66654 |6e-06 |line . block . sub|wordmatch | |
640 | |35 |0.02187 |0.00062|line . block . sub|IO/Handle.pm | |
641 | |274 |0.01525 |0.00006|line . block . sub|Getopt/Long.pm | |
642 | |20 |0.00585 |0.00029|line . block . sub|Fcntl.pm | |
643 | |128 |0.00340 |0.00003|line . block . sub|Exporter/Heavy.pm | |
644 | |42 |0.00332 |0.00008|line . block . sub|IO/File.pm | |
645 | |261 |0.00308 |0.00001|line . block . sub|Exporter.pm | |
646 | |323 |0.00248 |8e-06 |line . block . sub|constant.pm | |
647 | |12 |0.00246 |0.00021|line . block . sub|File/Spec/Unix.pm | |
648 | |191 |0.00240 |0.00001|line . block . sub|vars.pm | |
649 | |77 |0.00201 |0.00003|line . block . sub|FileHandle.pm | |
650 | |12 |0.00198 |0.00016|line . block . sub|Carp.pm | |
651 | |14 |0.00175 |0.00013|line . block . sub|Symbol.pm | |
652 | |15 |0.00130 |0.00009|line . block . sub|IO.pm | |
653 | |22 |0.00120 |0.00005|line . block . sub|IO/Seekable.pm | |
654 | |198 |0.00085 |4e-06 |line . block . sub|warnings/register.pm| |
655 | |114 |0.00080 |7e-06 |line . block . sub|strict.pm | |
656 | |47 |0.00068 |0.00001|line . block . sub|warnings.pm | |
657 | |27 |0.00054 |0.00002|line . block . sub|overload.pm | |
658 | |9 |0.00047 |0.00005|line . block . sub|SelectSaver.pm | |
659 | |13 |0.00045 |0.00003|line . block . sub|File/Spec.pm | |
660 | |2701595|15.73869 | |Total | |
661 | |128647 |0.74946 | |Average | |
662 | | |0.00201 |0.00003|Median | |
663 | | |0.00121 |0.00003|Deviation | |
664 | |
665 | Report produced by the NYTProf 2.03 Perl profiler, developed by Tim Bunce and |
666 | Adam Kaplan. |
667 | |
668 | At this point, if you're using the I<html> report, you can click through the |
669 | various links to bore down into each subroutine and each line of code. Because |
670 | we're using the text reporting here, and there's a whole directory full of |
671 | reports built for each source file, we'll just display a part of the |
672 | corresponding F<wordmatch-line.html> file, sufficient to give an idea of the |
673 | sort of output you can expect from this cool tool. |
674 | |
675 | $> html2text nytprof/wordmatch-line.html |
676 | |
677 | Performance Profile -- -block view-.-line view-.-sub view- |
678 | For wordmatch |
679 | Run on Fri Sep 26 13:46:39 2008 |
680 | Reported on Fri Sep 26 13:47:22 2008 |
681 | |
682 | File wordmatch |
683 | |
684 | Subroutines -- ordered by exclusive time |
685 | |Calls |P|F|Inclusive|Exclusive|Subroutine | |
686 | | | | |Time |Time | | |
687 | |251215|5|1|13.09263 |10.47692 |main::|matches| |
688 | |260642|2|1|2.71199 |2.71199 |main::|debug | |
689 | |1 |1|1|0.21404 |0.21404 |main::|report | |
690 | |0 |0|0|0 |0 |main::|BEGIN | |
691 | |
692 | |
693 | |Line|Stmts.|Exclusive|Avg. |Code | |
694 | | | |Time | | | |
695 | |1 | | | |#!/usr/bin/perl | |
696 | |2 | | | | | |
697 | | | | | |use strict; | |
698 | |3 |3 |0.00086 |0.00029|# spent 0.00003s making 1 calls to strict:: | |
699 | | | | | |import | |
700 | | | | | |use warnings; | |
701 | |4 |3 |0.01563 |0.00521|# spent 0.00012s making 1 calls to warnings:: | |
702 | | | | | |import | |
703 | |5 | | | | | |
704 | |6 | | | |=head1 NAME | |
705 | |7 | | | | | |
706 | |8 | | | |filewords - word analysis of input file | |
707 | <...snip...> |
708 | |62 |1 |0.00445 |0.00445|print report( %count ); | |
709 | | | | | |# spent 0.21404s making 1 calls to main::report| |
710 | |63 | | | | | |
711 | | | | | |# spent 23.56955s (10.47692+2.61571) within | |
712 | | | | | |main::matches which was called 251215 times, | |
713 | | | | | |avg 0.00005s/call: # 50243 times | |
714 | | | | | |(2.12134+0.51939s) at line 57 of wordmatch, avg| |
715 | | | | | |0.00005s/call # 50243 times (2.17735+0.54550s) | |
716 | |64 | | | |at line 56 of wordmatch, avg 0.00005s/call # | |
717 | | | | | |50243 times (2.10992+0.51797s) at line 58 of | |
718 | | | | | |wordmatch, avg 0.00005s/call # 50243 times | |
719 | | | | | |(2.12696+0.51598s) at line 55 of wordmatch, avg| |
720 | | | | | |0.00005s/call # 50243 times (1.94134+0.51687s) | |
721 | | | | | |at line 54 of wordmatch, avg 0.00005s/call | |
722 | | | | | |sub matches { | |
723 | <...snip...> |
724 | |102 | | | | | |
725 | | | | | |# spent 2.71199s within main::debug which was | |
726 | | | | | |called 260642 times, avg 0.00001s/call: # | |
727 | | | | | |251215 times (2.61571+0s) by main::matches at | |
728 | |103 | | | |line 74 of wordmatch, avg 0.00001s/call # 9427 | |
729 | | | | | |times (0.09628+0s) at line 50 of wordmatch, avg| |
730 | | | | | |0.00001s/call | |
731 | | | | | |sub debug { | |
732 | |104 |260642|0.58496 |2e-06 |my $message = shift; | |
733 | |105 | | | | | |
734 | |106 |260642|1.09917 |4e-06 |if ( $debug ) { | |
735 | |107 | | | |print STDERR "DBG: $message\n"; | |
736 | |108 | | | |} | |
737 | |109 | | | |} | |
738 | |110 | | | | | |
739 | |111 |1 |0.01501 |0.01501|exit 0; | |
740 | |112 | | | | | |
741 | |
742 | Oodles of very useful information in there - this seems to be the way forward. |
743 | |
744 | See also C<Devel::NYTProf::Apache> which hooks C<Devel::NYTProf> into C<mod_perl>. |
745 | |
746 | =head1 SORTING |
747 | |
748 | Perl modules are not the only tools a performance analyst has at their |
749 | disposal, system tools like C<time> should not be overlooked as the next |
750 | example shows, where we take a quick look at sorting. Many books, theses and |
751 | articles, have been written about efficient sorting algorithms, and this is not |
752 | the place to repeat such work, there's several good sorting modules which |
753 | deserve taking a look at too: C<Sort::Maker>, C<Sort::Key> spring to mind. |
754 | However, it's still possible to make some observations on certain Perl specific |
755 | interpretations on issues relating to sorting data sets and give an example or |
756 | two with regard to how sorting large data volumes can effect performance. |
757 | Firstly, an often overlooked point when sorting large amounts of data, one can |
758 | attempt to reduce the data set to be dealt with and in many cases C<grep()> can |
759 | be quite useful as a simple filter: |
760 | |
761 | @data = sort grep { /$filter/ } @incoming |
762 | |
763 | A command such as this can vastly reduce the volume of material to actually |
764 | sort through in the first place, and should not be too lightly disregarded |
765 | purely on the basis of it's simplicity. The C<KISS> principle is too often |
766 | overlooked - the next example uses the simple system C<time> utility to |
767 | demonstrate. Let's take a look at an actual example of sorting the contents of |
768 | a large file, an apache logfile would do. This one has over a quarter of a |
769 | million lines, is 50M in size, and a snippet of it looks like this: |
770 | |
771 | # logfile |
772 | |
773 | 188.209-65-87.adsl-dyn.isp.belgacom.be - - [08/Feb/2007:12:57:16 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" |
774 | 188.209-65-87.adsl-dyn.isp.belgacom.be - - [08/Feb/2007:12:57:16 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" |
775 | 151.56.71.198 - - [08/Feb/2007:12:57:41 +0000] "GET /suse-on-vaio.html HTTP/1.1" 200 2858 "http://www.linux-on-laptops.com/sony.html" "Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1" |
776 | 151.56.71.198 - - [08/Feb/2007:12:57:42 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/suse-on-vaio.html" "Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1" |
777 | 151.56.71.198 - - [08/Feb/2007:12:57:43 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1" |
778 | 217.113.68.60 - - [08/Feb/2007:13:02:15 +0000] "GET / HTTP/1.1" 304 - "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" |
779 | 217.113.68.60 - - [08/Feb/2007:13:02:16 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" |
780 | debora.to.isac.cnr.it - - [08/Feb/2007:13:03:58 +0000] "GET /suse-on-vaio.html HTTP/1.1" 200 2858 "http://www.linux-on-laptops.com/sony.html" "Mozilla/5.0 (compatible; Konqueror/3.4; Linux) KHTML/3.4.0 (like Gecko)" |
781 | debora.to.isac.cnr.it - - [08/Feb/2007:13:03:58 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/suse-on-vaio.html" "Mozilla/5.0 (compatible; Konqueror/3.4; Linux) KHTML/3.4.0 (like Gecko)" |
782 | debora.to.isac.cnr.it - - [08/Feb/2007:13:03:58 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/5.0 (compatible; Konqueror/3.4; Linux) KHTML/3.4.0 (like Gecko)" |
783 | 195.24.196.99 - - [08/Feb/2007:13:26:48 +0000] "GET / HTTP/1.0" 200 3309 "-" "Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.9) Gecko/20061206 Firefox/1.5.0.9" |
784 | 195.24.196.99 - - [08/Feb/2007:13:26:58 +0000] "GET /data/css HTTP/1.0" 404 206 "http://www.rfi.net/" "Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.9) Gecko/20061206 Firefox/1.5.0.9" |
785 | 195.24.196.99 - - [08/Feb/2007:13:26:59 +0000] "GET /favicon.ico HTTP/1.0" 404 209 "-" "Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.9) Gecko/20061206 Firefox/1.5.0.9" |
786 | crawl1.cosmixcorp.com - - [08/Feb/2007:13:27:57 +0000] "GET /robots.txt HTTP/1.0" 200 179 "-" "voyager/1.0" |
787 | crawl1.cosmixcorp.com - - [08/Feb/2007:13:28:25 +0000] "GET /links.html HTTP/1.0" 200 3413 "-" "voyager/1.0" |
788 | fhm226.internetdsl.tpnet.pl - - [08/Feb/2007:13:37:32 +0000] "GET /suse-on-vaio.html HTTP/1.1" 200 2858 "http://www.linux-on-laptops.com/sony.html" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" |
789 | fhm226.internetdsl.tpnet.pl - - [08/Feb/2007:13:37:34 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/suse-on-vaio.html" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" |
790 | 80.247.140.134 - - [08/Feb/2007:13:57:35 +0000] "GET / HTTP/1.1" 200 3309 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" |
791 | 80.247.140.134 - - [08/Feb/2007:13:57:37 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" |
792 | pop.compuscan.co.za - - [08/Feb/2007:14:10:43 +0000] "GET / HTTP/1.1" 200 3309 "-" "www.clamav.net" |
793 | livebot-207-46-98-57.search.live.com - - [08/Feb/2007:14:12:04 +0000] "GET /robots.txt HTTP/1.0" 200 179 "-" "msnbot/1.0 (+http://search.msn.com/msnbot.htm)" |
794 | livebot-207-46-98-57.search.live.com - - [08/Feb/2007:14:12:04 +0000] "GET /html/oracle.html HTTP/1.0" 404 214 "-" "msnbot/1.0 (+http://search.msn.com/msnbot.htm)" |
795 | dslb-088-064-005-154.pools.arcor-ip.net - - [08/Feb/2007:14:12:15 +0000] "GET / HTTP/1.1" 200 3309 "-" "www.clamav.net" |
796 | 196.201.92.41 - - [08/Feb/2007:14:15:01 +0000] "GET / HTTP/1.1" 200 3309 "-" "MOT-L7/08.B7.DCR MIB/2.2.1 Profile/MIDP-2.0 Configuration/CLDC-1.1" |
797 | |
798 | The specific task here is to sort the 286,525 lines of this file by Response |
799 | Code, Query, Browser, Referring Url, and lastly Date. One solution might be to |
800 | use the following code, which iterates over the files given on the |
801 | command-line. |
802 | |
803 | # sort-apache-log |
804 | |
805 | #!/usr/bin/perl -n |
806 | |
807 | use strict; |
808 | use warnings; |
809 | |
810 | my @data; |
811 | |
812 | LINE: |
813 | while ( <> ) { |
814 | my $line = $_; |
815 | if ( |
816 | $line =~ m/^( |
817 | ([\w\.\-]+) # client |
818 | \s*-\s*-\s*\[ |
819 | ([^]]+) # date |
820 | \]\s*"\w+\s* |
821 | (\S+) # query |
822 | [^"]+"\s* |
823 | (\d+) # status |
824 | \s+\S+\s+"[^"]*"\s+" |
825 | ([^"]*) # browser |
826 | " |
827 | .* |
828 | )$/x |
829 | ) { |
830 | my @chunks = split(/ +/, $line); |
831 | my $ip = $1; |
832 | my $date = $2; |
833 | my $query = $3; |
834 | my $status = $4; |
835 | my $browser = $5; |
836 | |
837 | push(@data, [$ip, $date, $query, $status, $browser, $line]); |
838 | } |
839 | } |
840 | |
841 | my @sorted = sort { |
842 | $a->[3] cmp $b->[3] |
843 | || |
844 | $a->[2] cmp $b->[2] |
845 | || |
846 | $a->[0] cmp $b->[0] |
847 | || |
848 | $a->[1] cmp $b->[1] |
849 | || |
850 | $a->[4] cmp $b->[4] |
851 | } @data; |
852 | |
853 | foreach my $data ( @sorted ) { |
854 | print $data->[5]; |
855 | } |
856 | |
857 | exit 0; |
858 | |
859 | When running this program, redirect C<STDOUT> so it is possible to check the |
860 | output is correct from following test runs and use the system C<time> utility |
861 | to check the overall runtime. |
862 | |
863 | $> time ./sort-apache-log logfile > out-sort |
864 | |
865 | real 0m17.371s |
866 | user 0m15.757s |
867 | sys 0m0.592s |
868 | |
869 | The program took just over 17 wallclock seconds to run. Note the different |
870 | values C<time> outputs, it's important to always use the same one, and to not |
871 | confuse what each one means. |
872 | |
873 | =over 4 |
874 | |
875 | =item Elapsed Real Time |
876 | |
877 | The overall, or wallclock, time between when C<time> was called, and when it |
878 | terminates. The elapsed time includes both user and system times, and time |
879 | spent waiting for other users and processes on the system. Inevitably, this is |
880 | the most approximate of the measurements given. |
881 | |
882 | =item User CPU Time |
883 | |
884 | The user time is the amount of time the entire process spent on behalf of the |
885 | user on this system executing this program. |
886 | |
887 | =item System CPU Time |
888 | |
889 | The system time is the amount of time the kernel itself spent executing |
890 | routines, or system calls, on behalf of this process user. |
891 | |
892 | =back |
893 | |
894 | Running this same process as a C<Schwarzian Transform> it is possible to |
895 | eliminate the input and output arrays for storing all the data, and work on the |
896 | input directly as it arrives too. Otherwise, the code looks fairly similar: |
897 | |
898 | # sort-apache-log-schwarzian |
899 | |
900 | #!/usr/bin/perl -n |
901 | |
902 | use strict; |
903 | use warnings; |
904 | |
905 | print |
906 | |
907 | map $_->[0] => |
908 | |
909 | sort { |
910 | $a->[4] cmp $b->[4] |
911 | || |
912 | $a->[3] cmp $b->[3] |
913 | || |
914 | $a->[1] cmp $b->[1] |
915 | || |
916 | $a->[2] cmp $b->[2] |
917 | || |
918 | $a->[5] cmp $b->[5] |
919 | } |
920 | map [ $_, m/^( |
921 | ([\w\.\-]+) # client |
922 | \s*-\s*-\s*\[ |
923 | ([^]]+) # date |
924 | \]\s*"\w+\s* |
925 | (\S+) # query |
926 | [^"]+"\s* |
927 | (\d+) # status |
928 | \s+\S+\s+"[^"]*"\s+" |
929 | ([^"]*) # browser |
930 | " |
931 | .* |
932 | )$/xo ] |
933 | |
934 | => <>; |
935 | |
936 | exit 0; |
937 | |
938 | Run the new code against the same logfile, as above, to check the new time. |
939 | |
940 | $> time ./sort-apache-log-schwarzian logfile > out-schwarz |
941 | |
942 | real 0m9.664s |
943 | user 0m8.873s |
944 | sys 0m0.704s |
945 | |
946 | The time has been cut in half, which is a respectable speed improvement by any |
947 | standard. Naturally, it is important to check the output is consistent with |
948 | the first program run, this is where the unix system C<cksum> utility comes in. |
949 | |
950 | $> cksum out-sort out-schwarz |
951 | 3044173777 52029194 out-sort |
952 | 3044173777 52029194 out-schwarz |
953 | |
954 | BTW. Beware too of pressure from managers who see you speed a program up by 50% |
955 | of the runtime once, only to get a request one month later to do the same again |
956 | (true story) - you'll just have to point out your only human, even if you are a |
957 | Perl programmer, and you'll see what you can do... |
958 | |
959 | =head1 LOGGING |
960 | |
961 | An essential part of any good development process is appropriate error handling |
962 | with appropriately informative messages, however there exists a school of |
963 | thought which suggests that log files should be I<chatty>, as if the chain of |
964 | unbroken output somehow ensures the survival of the program. If speed is in |
965 | any way an issue, this approach is wrong. |
966 | |
967 | A common sight is code which looks something like this: |
968 | |
969 | logger->debug( "A logging message via process-id: $$ INC: " . Dumper(\%INC) ) |
970 | |
971 | The problem is that this code will always be parsed and executed, even when the |
972 | debug level set in the logging configuration file is zero. Once the debug() |
973 | subroutine has been entered, and the internal C<$debug> variable confirmed to |
974 | be zero, for example, the message which has been sent in will be discarded and |
975 | the program will continue. In the example given though, the \%INC hash will |
976 | already have been dumped, and the message string constructed, all of which work |
977 | could be bypassed by a debug variable at the statement level, like this: |
978 | |
979 | logger->debug( "A logging message via process-id: $$ INC: " . Dumper(\%INC) ) if $DEBUG; |
980 | |
981 | This effect can be demonstrated by setting up a test script with both forms, |
982 | including a C<debug()> subroutine to emulate typical C<logger()> functionality. |
983 | |
984 | # ifdebug |
985 | |
986 | #!/usr/bin/perl |
987 | |
988 | use strict; |
989 | use warnings; |
990 | |
991 | use Benchmark; |
992 | use Data::Dumper; |
993 | my $DEBUG = 0; |
994 | |
995 | sub debug { |
996 | my $msg = shift; |
997 | |
998 | if ( $DEBUG ) { |
999 | print "DEBUG: $msg\n"; |
1000 | } |
1001 | }; |
1002 | |
1003 | timethese(100000, { |
1004 | 'debug' => sub { |
1005 | debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) ) |
1006 | }, |
1007 | 'ifdebug' => sub { |
1008 | debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) ) if $DEBUG |
1009 | }, |
1010 | }); |
1011 | |
1012 | Let's see what C<Benchmark> makes of this: |
1013 | |
1014 | $> perl ifdebug |
1015 | Benchmark: timing 100000 iterations of constant, sub... |
1016 | ifdebug: 0 wallclock secs ( 0.01 usr + 0.00 sys = 0.01 CPU) @ 10000000.00/s (n=100000) |
1017 | (warning: too few iterations for a reliable count) |
1018 | debug: 14 wallclock secs (13.18 usr + 0.04 sys = 13.22 CPU) @ 7564.30/s (n=100000) |
1019 | |
1020 | In the one case the code, which does exactly the same thing as far as |
1021 | outputting any debugging information is concerned, in other words nothing, |
1022 | takes 14 seconds, and in the other case the code takes one hundredth of a |
1023 | second. Looks fairly definitive. Use a C<$DEBUG> variable BEFORE you call the |
1024 | subroutine, rather than relying on the smart functionality inside it. |
1025 | |
1026 | =head2 Logging if DEBUG (constant) |
1027 | |
1028 | It's possible to take the previous idea a little further, by using a compile |
1029 | time C<DEBUG> constant. |
1030 | |
1031 | # ifdebug-constant |
1032 | |
1033 | #!/usr/bin/perl |
1034 | |
1035 | use strict; |
1036 | use warnings; |
1037 | |
1038 | use Benchmark; |
1039 | use Data::Dumper; |
1040 | use constant |
1041 | DEBUG => 0 |
1042 | ; |
1043 | |
1044 | sub debug { |
1045 | if ( DEBUG ) { |
1046 | my $msg = shift; |
1047 | print "DEBUG: $msg\n"; |
1048 | } |
1049 | }; |
1050 | |
1051 | timethese(100000, { |
1052 | 'debug' => sub { |
1053 | debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) ) |
1054 | }, |
1055 | 'constant' => sub { |
1056 | debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) ) if DEBUG |
1057 | }, |
1058 | }); |
1059 | |
1060 | Running this program produces the following output: |
1061 | |
1062 | $> perl ifdebug-constant |
1063 | Benchmark: timing 100000 iterations of constant, sub... |
1064 | constant: 0 wallclock secs (-0.00 usr + 0.00 sys = -0.00 CPU) @ -7205759403792793600000.00/s (n=100000) |
1065 | (warning: too few iterations for a reliable count) |
1066 | sub: 14 wallclock secs (13.09 usr + 0.00 sys = 13.09 CPU) @ 7639.42/s (n=100000) |
1067 | |
1068 | The C<DEBUG> constant wipes the floor with even the C<$debug> variable, |
1069 | clocking in at minus zero seconds, and generates a "warning: too few iterations |
1070 | for a reliable count" message into the bargain. To see what is really going |
1071 | on, and why we had too few iterations when we thought we asked for 100000, we |
1072 | can use the very useful C<B::Deparse> to inspect the new code: |
1073 | |
1074 | $> perl -MO=Deparse ifdebug-constant |
1075 | |
1076 | use Benchmark; |
1077 | use Data::Dumper; |
1078 | use constant ('DEBUG', 0); |
1079 | sub debug { |
1080 | use warnings; |
1081 | use strict 'refs'; |
1082 | 0; |
1083 | } |
1084 | use warnings; |
1085 | use strict 'refs'; |
1086 | timethese(100000, {'sub', sub { |
1087 | debug "A $0 logging message via process-id: $$" . Dumper(\%INC); |
1088 | } |
1089 | , 'constant', sub { |
1090 | 0; |
1091 | } |
1092 | }); |
1093 | ifdebug-constant syntax OK |
1094 | |
1095 | The output shows the constant() subroutine we're testing being replaced with |
1096 | the value of the C<DEBUG> constant: zero. The line to be tested has been |
1097 | completely optimized away, and you can't get much more efficient than that. |
1098 | |
1099 | =head1 POSTSCRIPT |
1100 | |
1101 | This document has provided several way to go about identifying hot-spots, and |
1102 | checking whether any modifications have improved the runtime of the code. |
1103 | |
1104 | As a final thought, remember that it's not (at the time of writing) possible to |
1105 | produce a useful program which will run in zero or negative time and this basic |
1106 | principle can be written as: I<useful programs are slow> by their very |
1107 | definition. It is of course possible to write a nearly instantaneous program, |
1108 | but it's not going to do very much, here's a very efficient one: |
1109 | |
1110 | $> perl -e 0 |
1111 | |
1112 | Optimizing that any further is a job for C<p5p>. |
1113 | |
1114 | =head1 SEE ALSO |
1115 | |
1116 | Further reading can be found using the modules and links below. |
1117 | |
1118 | =head2 PERLDOCS |
1119 | |
ea8b8ad2 |
1120 | For example: C<perldoc -f sort>. |
da096611 |
1121 | |
ea8b8ad2 |
1122 | L<perlfaq4>. |
1123 | |
1124 | L<perlfork>, L<perlfunc>, L<perlretut>, L<perlthrtut>. |
1125 | |
1126 | L<threads>. |
da096611 |
1127 | |
1128 | =head2 MAN PAGES |
1129 | |
ea8b8ad2 |
1130 | C<time>. |
da096611 |
1131 | |
1132 | =head2 MODULES |
1133 | |
1134 | It's not possible to individually showcase all the performance related code for |
1135 | Perl here, naturally, but here's a short list of modules from the CPAN which |
1136 | deserve further attention. |
1137 | |
ea8b8ad2 |
1138 | Apache::DProf |
1139 | Apache::SmallProf |
1140 | Benchmark |
1141 | DBIx::Profiler |
1142 | Devel::AutoProfiler |
1143 | Devel::DProf |
1144 | Devel::DProfLB |
1145 | Devel::FastProf |
1146 | Devel::GraphVizProf |
1147 | Devel::NYTProf |
1148 | Devel::NYTProf::Apache |
1149 | Devel::Profiler |
1150 | Devel::Profile |
1151 | Devel::Profit |
1152 | Devel::SmallProf |
1153 | Devel::WxProf |
1154 | POE::Devel::Profiler |
1155 | Sort::Key |
1156 | Sort::Maker |
da096611 |
1157 | |
1158 | =head2 URLS |
1159 | |
1160 | Very useful online reference material: |
1161 | |
1162 | http://www.ccl4.org/~nick/P/Fast_Enough/ |
1163 | |
1164 | http://www-128.ibm.com/developerworks/library/l-optperl.html |
1165 | |
1166 | http://perlbuzz.com/2007/11/bind-output-variables-in-dbi-for-speed-and-safety.html |
1167 | |
1168 | http://en.wikipedia.org/wiki/Performance_analysis |
1169 | |
1170 | http://apache.perl.org/docs/1.0/guide/performance.html |
1171 | |
1172 | http://perlgolf.sourceforge.net/ |
1173 | |
1174 | http://www.sysarch.com/Perl/sort_paper.html |
1175 | |
da096611 |
1176 | =head1 AUTHOR |
1177 | |
1178 | Richard Foley <richard.foley@rfi.net> Copyright (c) 2008 |
1179 | |
1180 | =cut |