# GCC/CFLAGS Benchmark for HEAD and PHP_5_0 - Thread: https://externals.io/message/13146 - Original archive: https://news-web.php.net/php.internals/13146 - Started by Sebastian Bergmann on 2004-10-04T18:30:47+00:00 - Messages: 15 - 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 --- ## #13146 — Sebastian Bergmann — 2004-10-04T18:30:47+00:00 URL: https://externals.io/message/13146#13146 I ran the Zend/bench.php script with PHP 5.0 and PHP 5.1 compiled with GCC 3.3 and GCC 3.4 using different optimizations on my Intel Pentium-M laptop: HEAD (PHP 5.1.0-dev) GCC 3.3.4 (-march=pentium3) -O0: 48.602 -Os: 29.920 -O1: 31.349 -O2: 29.029 -O3: 29.644 GCC 3.4.2 (-march=pentium-m -mtune=pentium-m) -O0: 48.966 -Os: 26.286 -O1: 29.253 -O2: 29.100 -O3: 27.767 -fprofile-{generate|use} -Os: 24.216 -fprofile-{generate|use} -O1: 26.575 -fprofile-{generate|use} -O2: 26.339 -fprofile-{generate|use} -O3: 25.537 PHP_5_0 (PHP 5.0.3-dev) GCC 3.3.4 (-march=pentium3) -O0: 58.394 -Os: 42.570 -O1: 43.454 -O2: 42.092 -O3: 42.066 GCC 3.4.2 (-march=pentium-m -mtune=pentium-m) -O0: 59.272 -Os: 35.853 -O1: 38.275 -O2: 37.989 -O3: 41.020 -fprofile-{generate|use} -Os: 33.926 -fprofile-{generate|use} -O1: 36.853 -fprofile-{generate|use} -O2: 35.335 -fprofile-{generate|use} -O3: 38.897 For the -fprofile-{generate|use} builds I used the Zend/bench.php script, too, to generate the profile information. If we were to add Makefile target (like GCC itself ("make profiled-bootstrap")) to the PHP build that makes use of -fprofile-{generate|use} it would make sense to use "make test" to generate the profiling information. When I find the time I will repeat the benchmark on my AMD Athlon64 box because I think the small L1/L2 caches of the laptop CPU affect the performance of the bigger code generated through optimizations. You can clearly see that code optimized for size (-Os) is faster than the more aggressive -O1, -O2, and -O3 optimization levels. -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13147 — Marcus Boerger — 2004-10-04T18:37:51+00:00 URL: https://externals.io/message/13146#13147 · In reply to: #13146 Hello Sebastian, could you please repeat the tests for php 4.3 to see the interesting difference? Also how about providing a spreadsheet for a better overiew? How did you do the timeings, i assume time command? And what CPU did you us, especially how much cache arewe talking about? best regards marcus Monday, October 4, 2004, 8:30:47 PM, you wrote: > I ran the Zend/bench.php script with PHP 5.0 and PHP 5.1 compiled with > GCC 3.3 and GCC 3.4 using different optimizations on my Intel Pentium-M > laptop: > HEAD (PHP 5.1.0-dev) > GCC 3.3.4 (-march=pentium3) > -O0: 48.602 > -Os: 29.920 > -O1: 31.349 > -O2: 29.029 > -O3: 29.644 > GCC 3.4.2 (-march=pentium-m -mtune=pentium-m) > -O0: 48.966 > -Os: 26.286 > -O1: 29.253 > -O2: 29.100 > -O3: 27.767 > -fprofile-{generate|use} -Os: 24.216 > -fprofile-{generate|use} -O1: 26.575 > -fprofile-{generate|use} -O2: 26.339 > -fprofile-{generate|use} -O3: 25.537 > PHP_5_0 (PHP 5.0.3-dev) > GCC 3.3.4 (-march=pentium3) > -O0: 58.394 > -Os: 42.570 > -O1: 43.454 > -O2: 42.092 > -O3: 42.066 > GCC 3.4.2 (-march=pentium-m -mtune=pentium-m) > -O0: 59.272 > -Os: 35.853 > -O1: 38.275 > -O2: 37.989 > -O3: 41.020 > -fprofile-{generate|use} -Os: 33.926 > -fprofile-{generate|use} -O1: 36.853 > -fprofile-{generate|use} -O2: 35.335 > -fprofile-{generate|use} -O3: 38.897 > For the -fprofile-{generate|use} builds I used the Zend/bench.php > script, too, to generate the profile information. If we were to add > Makefile target (like GCC itself ("make profiled-bootstrap")) to the > PHP build that makes use of -fprofile-{generate|use} it would make > sense to use "make test" to generate the profiling information. > When I find the time I will repeat the benchmark on my AMD Athlon64 box > because I think the small L1/L2 caches of the laptop CPU affect the > performance of the bigger code generated through optimizations. You can > clearly see that code optimized for size (-Os) is faster than the more > aggressive -O1, -O2, and -O3 optimization levels. > -- > Sebastian Bergmann http://www.sebastian-bergmann.de/ > GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 -- Best regards, Marcus mailto:helly@php.net --- ## #13148 — Sebastian Bergmann — 2004-10-04T18:43:43+00:00 URL: https://externals.io/message/13146#13148 · In reply to: #13147 Marcus Boerger wrote: > could you please repeat the tests for php 4.3 to see the interesting > difference? Yes, I can do that. > How did you do the timeings, i assume time command? I used the "total" output of Zend/bench.php. > And what CPU did you us, especially how much cache arewe talking about? Intel Pentium-M 1500 MHz, Frequency Scaling has been disabled for the benchmark run. L1 I cache: 32K, L1 D cache: 32K, L2 cache: 1024K -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13149 — Sebastian Bergmann — 2004-10-04T18:55:07+00:00 URL: https://externals.io/message/13146#13149 · In reply to: #13147 Marcus Boerger wrote: > Also how about providing a spreadsheet for a better overiew? http://www.sebastian-bergmann.de/stuff/gcc-cflags.png -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13150 — Marcus Boerger — 2004-10-04T19:00:01+00:00 URL: https://externals.io/message/13146#13150 · In reply to: #13149 Hello Sebastian, Monday, October 4, 2004, 8:55:07 PM, you wrote: > Marcus Boerger wrote: > > […] > http://www.sebastian-bergmann.de/stuff/gcc-cflags.png Looks very nice so far but i am eagerly awaiting the 4.3 results now :-) -- Best regards, Marcus mailto:helly@php.net --- ## #13151 — Ilia Alshanetsky — 2004-10-04T19:05:04+00:00 URL: https://externals.io/message/13146#13151 · In reply to: #13146 Very interesting numbers, I'd like to second Marcus' request for a 4.3 benchmark. I was somewhat surprised by O2 and O1 being slower then Os, while O3 in some cases may end over optimizing which would it explain it's poor showing. However, it could be that it makes simple situations slower, while more complex operations that are generally more CPU intensive will in fact become faster. If you don't mind, could you please all include data for "time make test" as it seems to cover a much greater quantity of code. Ilia --- ## #13152 — Sterling Hughes — 2004-10-04T19:18:47+00:00 URL: https://externals.io/message/13146#13152 · In reply to: #13151 > Very interesting numbers, I'd like to second Marcus' request for a 4.3 > benchmark. > > I was somewhat surprised by O2 and O1 being slower then Os, while O3 in > some cases may end over optimizing which would it explain it's poor > showing. However, it could be that it makes simple situations slower, > while more complex operations that are generally more CPU intensive will > in fact become faster. If you don't mind, could you please all include > data for "time make test" as it seems to cover a much greater quantity > of code. > Thies and I found the same thing when we did our patch. It relates to the size of the executor loop that is generated, if you have too much inlining you end up blowing your instruction cache. -Sterling --- ## #13153 — Marcus Boerger — 2004-10-04T19:26:51+00:00 URL: https://externals.io/message/13146#13153 · In reply to: #13152 Hello Sterling, hello Sebastian, Monday, October 4, 2004, 9:18:47 PM, you wrote: > > > […] > Thies and I found the same thing when we did our patch. It relates to > the size of the executor loop that is generated, if you have too much > inlining you end up blowing your instruction cache. That means not using all opcodes could result in faster execution with the function based executor. Maybe sebastian could try that too? -- Best regards, Marcus mailto:helly@php.net --- ## #13154 — Sebastian Bergmann — 2004-10-04T20:00:05+00:00 URL: https://externals.io/message/13146#13154 · In reply to: #13150 Marcus Boerger wrote: > very nice so far but i am eagerly awaiting the 4.3 results now :-) You'll have to wait at least until tomorrow. -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13156 — Sebastian Bergmann — 2004-10-04T20:01:11+00:00 URL: https://externals.io/message/13146#13156 · In reply to: #13151 Ilia Alshanetsky wrote: > I was somewhat surprised by O2 and O1 being slower then Os This is because the L1/L2 caches are very small. I will re-run the benchmark tomorrow (if I find the time) on my Athlon64 box. -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13157 — Sebastian Bergmann — 2004-10-04T20:01:36+00:00 URL: https://externals.io/message/13146#13157 · In reply to: #13153 Marcus Boerger wrote: > That means not using all opcodes could result in faster execution with > the function based executor. Maybe sebastian could try that too? How would I do that? -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13159 — Marcus Boerger — 2004-10-04T21:14:32+00:00 URL: https://externals.io/message/13146#13159 · In reply to: #13157 Hello Sebastian, Monday, October 4, 2004, 10:01:36 PM, you wrote: > Marcus Boerger wrote: > > […] > How would I do that? 3.*, 4.*: switch only 5.0: call only 5.1: ./configure --with-zend-vm=[CALL|SWITCH|GOTO] regards marcus --- ## #13160 — Andi Gutmans — 2004-10-04T23:59:28+00:00 URL: https://externals.io/message/13146#13160 · In reply to: #13146 I've done lots of testing with these kind of compiler flags. It's really a tough call because you might change 2 lines in the PHP code and then your results might differ completely because of CPU cache hits/misses. Also very much depends on the CPU. But in any case, interesting results. Andi At 08:30 PM 10/4/2004 +0200, Sebastian Bergmann wrote: > I ran the Zend/bench.php script with PHP 5.0 and PHP 5.1 compiled with > GCC 3.3 and GCC 3.4 using different optimizations on my Intel Pentium-M > laptop: > > HEAD (PHP 5.1.0-dev) > > GCC 3.3.4 (-march=pentium3) > > -O0: 48.602 > -Os: 29.920 > -O1: 31.349 > -O2: 29.029 > -O3: 29.644 > > GCC 3.4.2 (-march=pentium-m -mtune=pentium-m) > > -O0: 48.966 > -Os: 26.286 > -O1: 29.253 > -O2: 29.100 > -O3: 27.767 > > -fprofile-{generate|use} -Os: 24.216 > -fprofile-{generate|use} -O1: 26.575 > -fprofile-{generate|use} -O2: 26.339 > -fprofile-{generate|use} -O3: 25.537 > > > PHP_5_0 (PHP 5.0.3-dev) > > GCC 3.3.4 (-march=pentium3) > > -O0: 58.394 > -Os: 42.570 > -O1: 43.454 > -O2: 42.092 > -O3: 42.066 > > GCC 3.4.2 (-march=pentium-m -mtune=pentium-m) > > -O0: 59.272 > -Os: 35.853 > -O1: 38.275 > -O2: 37.989 > -O3: 41.020 > > -fprofile-{generate|use} -Os: 33.926 > -fprofile-{generate|use} -O1: 36.853 > -fprofile-{generate|use} -O2: 35.335 > -fprofile-{generate|use} -O3: 38.897 > > For the -fprofile-{generate|use} builds I used the Zend/bench.php > script, too, to generate the profile information. If we were to add > Makefile target (like GCC itself ("make profiled-bootstrap")) to the > PHP build that makes use of -fprofile-{generate|use} it would make > sense to use "make test" to generate the profiling information. > > When I find the time I will repeat the benchmark on my AMD Athlon64 box > because I think the small L1/L2 caches of the laptop CPU affect the > performance of the bigger code generated through optimizations. You can > clearly see that code optimized for size (-Os) is faster than the more > aggressive -O1, -O2, and -O3 optimization levels. > >-- >Sebastian Bergmann http://www.sebastian-bergmann.de/ >GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13163 — Sebastian Bergmann — 2004-10-05T05:14:39+00:00 URL: https://externals.io/message/13146#13163 · In reply to: #13159 Marcus Boerger wrote: > 5.1: ./configure --with-zend-vm=[CALL|SWITCH|GOTO] I do not seem to have that configure switch (current HEAD): sb@wopr-mobile php-5.1 % ./configure --help|grep vm --enable-sysvmsg Enable sysvmsg support -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69 --- ## #13164 — Sebastian Bergmann — 2004-10-05T06:08:31+00:00 URL: https://externals.io/message/13146#13164 · In reply to: #13150 Marcus Boerger wrote: > very nice so far but i am eagerly awaiting the 4.3 results now :-) I have uploaded the results for PHP 4.3.10-dev, PHP 5.0.3-dev, and PHP 5.1.0-dev to http://www.sebastian-bergmann.de/stuff/gcc-cflags.pdf The machine used was http://www.sebastian-bergmann.de/gallery/wopr -- Sebastian Bergmann http://www.sebastian-bergmann.de/ GnuPG Key: 0xB85B5D69 / 27A7 2B14 09E4 98CD 6277 0E5B 6867 C514 B85B 5D69