# PHP vs PHP NTS benackmark (win32) - Thread: https://externals.io/message/39043 - Original archive: https://news-web.php.net/php.internals/39043 - Started by Mario Brandt on 2008-07-17T09:36:11+00:00 - Messages: 7 - Reply by email to internals@lists.php.net - More for agents: search with https://externals.io/search.md?q={query} · JSON API at https://externals.io/api · documentation at https://externals.io/llms.txt --- ## #39043 — Mario Brandt — 2008-07-17T09:36:11+00:00 URL: https://externals.io/message/39043#39043 I made a test on my console (cmd.exe) ENV: WinXP SP 3 all updates, PHP 5.2.6 / PHP 5.2.6 non thread safe (NTS) Intel Celeron 2.4 GHz 1 GB DDR Ram It showed that the non thread safe is faster than the thread safe version. NTS Benchmark Time Peak Memory Peak Memory (Real) ------------------------------------------------------------------- ./benchmarks\ackermann 1.744 428704 524288 ./benchmarks\array 0.067 6583320 7340032 ./benchmarks\array2 0.059 6589168 7340032 ./benchmarks\array3 0.841 364848 524288 ./benchmarks\fibonacci 3.622 77976 262144 ./benchmarks\hash 0.150 3921712 4194304 ./benchmarks\hash2 0.130 144912 262144 ./benchmarks\heapsort 0.410 1639960 1835008 ./benchmarks\mandel 1.445 83352 262144 ./benchmarks\mandel2 1.716 80312 262144 ./benchmarks\matrix 0.410 357432 524288 ./benchmarks\nestedloop 0.742 80368 262144 ./benchmarks\sieve 0.532 1403920 1572864 ./benchmarks\simple 0.425 77768 262144 ./benchmarks\simplecall 0.553 76656 262144 ./benchmarks\simpleucall 0.881 77344 262144 ./benchmarks\simpleudcall 1.188 77440 262144 ./benchmarks\strcat 0.060 857848 1310720 TS Benchmark Time Peak Memory Peak Memory (Real) ------------------------------------------------------------------- ./benchmarks\ackermann 2.406 674272 786432 ./benchmarks\array 0.078 6583664 7340032 ./benchmarks\array2 0.065 6591032 7340032 ./benchmarks\array3 0.861 365352 524288 ./benchmarks\fibonacci 4.835 77720 262144 ./benchmarks\hash 0.234 3922136 4194304 ./benchmarks\hash2 0.155 145416 262144 ./benchmarks\heapsort 0.476 1640912 1835008 ./benchmarks\mandel 1.531 84552 262144 ./benchmarks\mandel2 1.849 80840 262144 ./benchmarks\matrix 0.365 357984 524288 ./benchmarks\nestedloop 0.811 80624 262144 ./benchmarks\sieve 0.531 1404096 1572864 ./benchmarks\simple 0.454 77736 262144 ./benchmarks\simplecall 0.743 76472 262144 ./benchmarks\simpleucall 1.285 77160 262144 ./benchmarks\simpleudcall 1.545 77256 262144 ./benchmarks\strcat 0.061 858000 1310720 with PHP6 the test failed regards Mario --- ## #39045 — david.zuelke@bitextender.com — 2008-07-17T10:03:53+00:00 URL: https://externals.io/message/39043#39045 · In reply to: #39043 Am 17.07.2008 um 11:36 schrieb Mario Brandt: > I made a test on my console (cmd.exe) > ENV: WinXP SP 3 all updates, PHP 5.2.6 / PHP 5.2.6 non thread safe > (NTS) > Intel Celeron 2.4 GHz 1 GB DDR Ram > > It showed that the non thread safe is faster than the thread safe > version. And your point/question is...? David --- ## #39046 — Runar Olsen — 2008-07-17T10:11:26+00:00 URL: https://externals.io/message/39043#39046 · In reply to: #39043 Hi, That should be correct as there is no time spent keeping internal variables thread safe. Mario Brandt skrev: > I made a test on my console (cmd.exe) > ENV: WinXP SP 3 all updates, PHP 5.2.6 / PHP 5.2.6 non thread safe (NTS) > Intel Celeron 2.4 GHz 1 GB DDR Ram > > It showed that the non thread safe is faster than the thread safe version. > > > NTS > > Benchmark Time Peak Memory Peak Memory (Real) > ------------------------------------------------------------------- > ./benchmarks\ackermann 1.744 428704 524288 > ./benchmarks\array 0.067 6583320 7340032 > ./benchmarks\array2 0.059 6589168 7340032 > ./benchmarks\array3 0.841 364848 524288 > ./benchmarks\fibonacci 3.622 77976 262144 > ./benchmarks\hash 0.150 3921712 4194304 > ./benchmarks\hash2 0.130 144912 262144 > ./benchmarks\heapsort 0.410 1639960 1835008 > ./benchmarks\mandel 1.445 83352 262144 > ./benchmarks\mandel2 1.716 80312 262144 > ./benchmarks\matrix 0.410 357432 524288 > ./benchmarks\nestedloop 0.742 80368 262144 > ./benchmarks\sieve 0.532 1403920 1572864 > ./benchmarks\simple 0.425 77768 262144 > ./benchmarks\simplecall 0.553 76656 262144 > ./benchmarks\simpleucall 0.881 77344 262144 > ./benchmarks\simpleudcall 1.188 77440 262144 > ./benchmarks\strcat 0.060 857848 1310720 > > > TS > > Benchmark Time Peak Memory Peak Memory (Real) > ------------------------------------------------------------------- > ./benchmarks\ackermann 2.406 674272 786432 > ./benchmarks\array 0.078 6583664 7340032 > ./benchmarks\array2 0.065 6591032 7340032 > ./benchmarks\array3 0.861 365352 524288 > ./benchmarks\fibonacci 4.835 77720 262144 > ./benchmarks\hash 0.234 3922136 4194304 > ./benchmarks\hash2 0.155 145416 262144 > ./benchmarks\heapsort 0.476 1640912 1835008 > ./benchmarks\mandel 1.531 84552 262144 > ./benchmarks\mandel2 1.849 80840 262144 > ./benchmarks\matrix 0.365 357984 524288 > ./benchmarks\nestedloop 0.811 80624 262144 > ./benchmarks\sieve 0.531 1404096 1572864 > ./benchmarks\simple 0.454 77736 262144 > ./benchmarks\simplecall 0.743 76472 262144 > ./benchmarks\simpleucall 1.285 77160 262144 > ./benchmarks\simpleudcall 1.545 77256 262144 > ./benchmarks\strcat 0.061 858000 1310720 > > > with PHP6 the test failed > > regards > Mario --- ## #39047 — Pierre Joye — 2008-07-17T10:24:31+00:00 URL: https://externals.io/message/39043#39047 · In reply to: #39043 hi, > I made a test on my console (cmd.exe) > ENV: WinXP SP 3 all updates, PHP 5.2.6 / PHP 5.2.6 non thread safe (NTS) > Intel Celeron 2.4 GHz 1 GB DDR Ram > > It showed that the non thread safe is faster than the thread safe version. It is expected and it is also why we provide TS and non TS builds for windows (fcgi usage). Cheers, -- Pierre http://blog.thepimp.net | http://www.libgd.org --- ## #39057 — Stanislav Malyshev — 2008-07-17T15:42:11+00:00 URL: https://externals.io/message/39043#39057 Hi! > a, what are the ramifications of having to run php without threading > when we live in a world that's increasingly moving towards > multi-core/multi- CPU systems? I'd say next to none, since multi-CPU systems are perfectly capable of running multiple processes and PHP lives in the context of the web server where multiple processes are just fine. > b, Is php not missing out here big time? Not really, even though some cool stuff may be done, but that's probably outside standard threading model. > d, Will threads ever become recommended in/for php? Unless some serious change happens, I don't see it likely. "Shared nothing" works not too bad. > e, If threads become the norm, will php code have to be written > differently to account for threads? too hypothetical for me :) > f, are there things I can/should be doing to my [production] > systems/software-stacks to leverage extra performance for my php > code from CPUs with multiple cores (and/or machines with multiple > CPUs)? Have Apache with bigger MaxClients? ;) -- Stanislav Malyshev, Zend Software Architect stas@zend.com http://www.zend.com/ (408)253-8829 MSN: stas@zend.com --- ## #39058 — Andi Gutmans — 2008-07-17T15:48:40+00:00 URL: https://externals.io/message/39043#39058 · In reply to: #39057 There are some things I mention in this blog entry re: why I think multi-process architecture is not a bad thing even with where CPUs are going: http://tinyurl.com/3eyfth This was obviously a heated discussion and I'd appreciate if you have flames/comments to send them to me directly and/or on the blog and not to this list. Just sending it though as it answers some of these concerns. Andi > -----Original Message----- > From: Stanislav Malyshev [mailto:stas@zend.com] > Sent: Thursday, July 17, 2008 8:42 AM > To: Jochem Maas > Cc: 'internals@lists.php.net' > Subject: Re: [PHP-DEV] PHP vs PHP NTS benackmark (win32) > > Hi! > > > […] > > I'd say next to none, since multi-CPU systems are perfectly capable of > running multiple processes and PHP lives in the context of the web > server where multiple processes are just fine. > > > […] > > Not really, even though some cool stuff may be done, but that's > probably > outside standard threading model. > > > […] > > Unless some serious change happens, I don't see it likely. "Shared > nothing" works not too bad. > > > […] > > too hypothetical for me :) > > > […] > > Have Apache with bigger MaxClients? ;) > -- > Stanislav Malyshev, Zend Software Architect > stas@zend.com http://www.zend.com/ > (408)253-8829 MSN: stas@zend.com --- ## #39061 — Jochem Maas — 2008-07-17T17:41:34+00:00 URL: https://externals.io/message/39043#39061 · In reply to: #39058 Thank you everyone for you replies, apparently I know a little more than I thought in general and also a little more than this morning. :-) rgds, Jochem Andi Gutmans schreef: > There are some things I mention in this blog entry re: why I think ...