kenjis
commented
9 years ago Thank you for your comment.
I did not. https://github.com/kenjis/php-framework-benchmark/blob/master/setup.sh is all I ran.
I didn't know phalcon.so overhead is so big, but you are probably true.
But It is not a problem to me, because I accept the cost. Because I may use phalcon class in other frameworks.
I ran with and without phalcon.so. The overhead is about 10%?
w/z phalcon.so
| framework | requests per second | relative | peak memory | relative |
|---|---|---|---|---|
| codeigniter-3.0 | 693.24 | 16.7 | 0.50 | 1.0 |
| slim-2.6 | 725.55 | 17.5 | 0.50 | 1.0 |
| yii-2.0 | 385.49 | 9.3 | 1.50 | 3.0 |
| fuel-1.8-dev | 294.12 | 7.1 | 0.75 | 1.5 |
| silex-1.2 | 349.25 | 8.4 | 1.00 | 2.0 |
| bear-1.0 | 328.41 | 7.9 | 1.00 | 2.0 |
| cake-3.0 | 249.54 | 6.0 | 1.00 | 2.0 |
| symfony-2.6 | 246.28 | 5.9 | 1.00 | 2.0 |
| laravel-5.0 | 80.08 | 1.9 | 2.75 | 5.5 |
| zf-2.4 | 69.69 | 1.7 | 2.25 | 4.5 |
| typo3-flow-2.3 | 41.46 | 1.0 | 5.25 | 10.5 |
w/o phalcon.so
| framework | requests per second | relative | peak memory | relative |
|---|---|---|---|---|
| codeigniter-3.0 | 728.09 | 16.4 | 0.50 | 0.0 |
| slim-2.6 | 794.40 | 17.9 | 0.50 | 0.0 |
| yii-2.0 | 413.65 | 9.3 | 1.50 | 0.0 |
| fuel-1.8-dev | 314.51 | 7.1 | 0.75 | 0.0 |
| silex-1.2 | 367.80 | 8.3 | 1.00 | 0.0 |
| bear-1.0 | 369.46 | 8.3 | 1.00 | 0.0 |
| cake-3.0 | 270.94 | 6.1 | 1.00 | 0.0 |
| symfony-2.6 | 303.39 | 6.8 | 1.00 | 0.0 |
| laravel-5.0 | 81.32 | 1.8 | 2.75 | 0.0 |
| zf-2.4 | 69.84 | 1.6 | 2.25 | 0.0 |
| typo3-flow-2.3 | 44.33 | 1.0 | 5.25 | 0.0 |
It's not clear from the code here whether or not you used a specific php.ini that loads phalcon.so for the non-phalcon benchmarks. Loading a blank page with phalcon.so has a non-trivial overhead over a blank page without phalcon.so being loaded. As such, the non-phalcon benchmarks should all make sure that phalcon is not being loaded when they're run.
You may have done this but it's not clear.