Re: cvs: php-benchmarks /benchcli/tests test_whirl.php /benchcli/tests/phpWhirl 99bottles.wr AUTHORS COPYING keymaker.wr whirl.php
/benchcli/tests/phpWhirl/classes WhirlMathRing.php WhirlMemory.php WhirlOpsRing.php WhirlParser.php WhirlRing.

On Tue, Jun 2, 2009 at 3:37 PM, Alexander Hjalmarsson <hjalle@sgh.se> wrote: >> I wouldn't worry too much about the micro-benchmarks. They wouldnt be >> much worse than they are now. > > > That might be correct. :) > When adding up the cachegrind time (or valgrind), the time for the benchmark > would be as you know, about 20x longe. How long is reasonable for a > benchmark to run? I'm mostly thinking of when enabling valgrind/cachegrind > or any other tool. Ah, of course. That is a real problem. That would probably make it impractical to run 1-minute long benchmarks in valgrind. A few options: - Allow it, but run only short things by default in valgrind - Seek out faster ways to do the things that valgrind gets you. Options include: For memory allocation: - There might be settings to ask valgrind to just record the memory allocated? - Does glibc allow you to set an environmental variable giving the allocated memory? - Maybe its simple to link in another memory allocator like jemalloc or Google's tcmalloc which might support that? - Google's unladen-swallow project measures memory usage by sampling (see http://code.google.com/p/unladen-swallow/source/browse/tests/perf.py). That may be a better approach. For cache/branch behaviour: - One option is to use PAPI, but that requires a patched kernel. Perhaps allow both options, and the cachegrind one would just be slow, and people wanting fast results should use PAPI? Not the best solution unfortunately. > Boh. I'm talking about the test for the j4p5 parser, found here: > http://j4p5.sourceforge.net/index.php > The indata that is used now is the example1, found here: > http://j4p5.sourceforge.net/demos/2.php > It's not really anything to work with and I think it would be good to have > something larger. Try something like this: http://www.koders.com/default.aspx?s=the&btn=&la=JavaScript&li=* Its a search for the word "the" in JS files on koders.com. There are a few choice samples in there. Paul >> >> > On Tue, Jun 2, 2009 at 3:19 PM, Alexander Hjalmarsson >> >> > <hjalle@php.net>wrote: >> >> > >> >> >> hjalle          Tue Jun  2 13:19:40 2009 UTC >> >> >> >> >> >>  Added files: >> >> >>    /php-benchmarks/benchcli/tests      test_whirl.php >> >> >>    /php-benchmarks/benchcli/tests/phpWhirl     99bottles.wr AUTHORS >> >> >>                                               >> >> >>  COPYING keymaker.wr >> >> >>                                               >> >> >>  whirl.php >> >> >>    /php-benchmarks/benchcli/tests/phpWhirl/classes >> >> >> WhirlMathRing.php >> >> >> >> >> >>  WhirlMemory.php >> >> >> >> >> >>  WhirlOpsRing.php >> >> >> >> >> >>  WhirlParser.php >> >> >>                                                 >> >> >>        WhirlRing.php >> >> >>  Log: >> >> >>  Whirl parser test >> >> >> >> >> >> >> >> >> -- >> >> >> PHP Quality Assurance Mailing List >> >> >> <http://www.php.net/> >> >> >> To unsubscribe, visit: >> >> >> http://www.php.net/unsub.php >> >> >> >> >> > >> >> >> >> >> >> >> >> -- >> >> Paul Biggar >> >> paul.biggar@gmail.com >> > >> > >> >> >> >> -- >> Paul Biggar >> paul.biggar@gmail.com > > -- Paul Biggar paul.biggar@gmail.com