PH7 vs PHP?

[staalmannen]

Has anyone tried to run PTS on the smaller PH7 engine instead of the standard PHP?

http://ph7.symisc.net/

I remember seeing a benchmark of normal PHP vs Facebook's PHP-to-C++ converter. It would be interesting to see how this one performs...

[Michael]

Has anyone tried to run PTS on the smaller PH7 engine instead of the standard PHP?

http://ph7.symisc.net/

I remember seeing a benchmark of normal PHP vs Facebook's PHP-to-C++ converter. It would be interesting to see how this one performs...

I've played around with HipHop and Roadsend PHP for PTS. Mostly this has just been a coding exercise to ensure PTS remains very portable and clean. It's also fun trying out the different engines. To some surprise, I don't think I've ever heard of PH7 previously. However, as soon as time allows I'll give PH7 for PTS a try. It looks like PH7 implements much of PHP 5.3 so I think it should work fine; PTS just depends upon PHP 5.2.0 functionality.

[staalmannen]

if interested, I just put up a PKGBUILD on Arch AUR
https://aur.archlinux.org/packages/ph7/
I will try to package it for other platforms via OBS later.

[staalmannen]

I have no idea if this is a real result or just php5 running twice:

php5:

### OVERALL DATA ###

PHP: 5.4.15
DETAILED_SYSTEM_INFO: 0.493 seconds
LIST_AVAILABLE_TESTS: 2.369 seconds
LIST_AVAILABLE_SUITES: 143.947 seconds
INFO: 0.121 seconds
CLONE_OPENBENCHMARKING_RESULT: 441.958 seconds
RESULT_FILE_TO_TEXT: 44.661 seconds
DIAGNOSTICS: 0.830 seconds
DUMP_POSSIBLE_OPTIONS: 0.043 seconds

ELAPSED TIME: 1903.279 seconds
PEAK MEMORY USAGE: 34.25 MB
PEAK MEMORY USAGE (emalloc): 32.703 MB

PHP_BIN="/usr/bin/php" phoronix-test-suite debug-self-test 1362,74s user 26,96s system 72% cpu 31:46,02 total


ph7:


### OVERALL DATA ###

PHP: 5.4.15
DETAILED_SYSTEM_INFO: 0.293 seconds
LIST_AVAILABLE_TESTS: 2.220 seconds
LIST_AVAILABLE_SUITES: 130.903 seconds
INFO: 0.083 seconds
CLONE_OPENBENCHMARKING_RESULT: 385.182 seconds
RESULT_FILE_TO_TEXT: 46.413 seconds
DIAGNOSTICS: 0.811 seconds
DUMP_POSSIBLE_OPTIONS: 0.054 seconds

ELAPSED TIME: 1697.875 seconds
PEAK MEMORY USAGE: 35 MB
PEAK MEMORY USAGE (emalloc): 32.679 MB

PHP_BIN="/usr/bin/ph7 -r" phoronix-test-suite debug-self-test 1354,52s user 24,74s system 81% cpu 28:19,57 total


I had to run ph7 with -r because otherwise it failed to execute phoronix-test-suite (a test in there to get version number perhaps where ph7 return null?)

The two results look suspiciously similar, so I wonder if both are actually the php5 and if there is another way to test it.

I have run some small php test scripts with ph7 so the binary works.

[mendieta]

Interesting! I think the main reason why results are so similar is that PHP is probably not the bottleneck, at all, and all the computational time is used in OS calls, etc

[Michael]

Interesting! I think the main reason why results are so similar is that PHP is probably not the bottleneck, at all, and all the computational time is used in OS calls, etc

HHVM can be much faster than PHP... PTS does lots of OS system calls, but also taking lots of time is lots of XML parsing/generation, in particular. There's also zip/tar.gz compression/decompression, in-PHP computational tasks, and other calculations. Graph rendering with pts_Graph is perhaps the most computationally intense inside PTS.