[resolved] peak memory usage |
Site Admin
|
Would you please compare non-accelerated php against php accelerated with phpexpress to see if it makes any difference in memory usage?
If it makes, please provide a code sample and name the tool(s) you used to measure memory usage. |
||||||||||||
_________________ The PHP IDE team |
|
Without any accelerators
I got times in the 0.36 to 0.59 range, 0.47 avg With PHPExpress loaded with extension=
I got times in the 0.31 to 0.39 range, 0.34 avg, second run 0.34 avg With PHPExpress loaded with zend_extension_ts=
I got times in the 0.3 to 0.38 range, 0.31 avg, second run 0.32 avg With Xcache loaded with zend_extension_ts= Options (without these options set Xcache isn't REALLY enabled) xcache.cacher = On xcache.size = 32M
I got times in the 0.33 to 0.41 range, 0.36 avg With APC (no ini options configured) loaded with extention=
I got times in the range of 0.3 to 0.39, 0.34 avg APC does not appear to let itself be loaded with zend_extension_ts= ?? With eAccelerator (no ini options configured) loaded with zend_extension_ts=
I got times in the range of 0.3 to 0.41, 0.32 avg Once more without any accelerators
I got times in the 0.38 to 0.56 range, 0.45 avg To try to make this unscientific test as fair as possible while run on a development desktop - I didn't open or close any other applications in between restarts of apache - I had all applications minimized save for firefox, notepad++, and the services window. - The dashboard displays are from the last run not the best one. - I did not include the initial page load after restarting apache and only averaged the 20 refreshes after that initial hit. I completely take back my concern that PHPExpress was doing something wacky to my memory usage since Xcache and eAccelerator are the ones that do something wacky when compared to runs without accelerators. I'll have to research this. Also interesting to me was the loading of PHPExpress with zend_extension_ts. I reloaded php_express.dll one more time with extension= to see if other processes on my machine caused the change in times and average. The results I got back were the same as before so I suppose something is gained by loading it that way. I then double checked and reran the zend_extension_ts test and got similar results. Memory usage for each second run matched the first. I guess this also serves as a simple quick and dirty 'ymmv' benchmark between PHP standalone, PhpExpress, Xcache, APC, and eAccelerator. The values of Peak and mem come from these php functions memory_get_peak_usage() and memory_get_usage(). I can't release the code to reproduce this page. |
||||||||||||||||||||||||||
|
Site Admin
|
Not needed. They run bytecodes in the shared memory directly and BTW they do this without any locks. Because they both are turk mmcache successors, there is no difference in the approach and certainly they inherited all pros and cons. Thanks for your test and feedback provided. Would be even better if you post some details about scripts you used and certain info about platform (CPU, its clock, memory size, memory clock) and software platform (OS, php version, web server version, php startup as FASTCGI/CGI or SAPI module). Let know if you're interested in further tests. I can provide you with next version of PhpExpress. |
||||||||||||||
_________________ The PHP IDE team |
[resolved] peak memory usage |
|
||
Content © NuSphere Corp., PHP IDE team
Powered by phpBB © phpBB Group, Design by phpBBStyles.com | Styles Database.
Powered by
Powered by phpBB © phpBB Group, Design by phpBBStyles.com | Styles Database.
Powered by