I haven't had time to do a more detailed tests, his box was really cpu
bound, from the numbers I was playing with but would really need to take
some time to graph out to be sure it looks like solar runs better under
pressure, but Zend runs faster if its got room..
Of course, it could be disk related, since solar only includes 18 files,
and zend includes 34 files his mac mini could of had some serious IO
issues, so the framework with the least files has a large advantage.
My tests are all on a Dell dual cpu with raid 5 scsi drives.
Richard Thomas - Code Monkey
Cyberlot Technologies Group Inc.
507.398.4124 - Voice
Martel Valgoerad wrote:
Richard Thomas wrote:
Possibly a stupid idea, but since all Zend classes are required to
follow a specific standard instead of doing
>
require_once 'Zend/Exception.php';
Wouldn't
if(!class_exists('Zend_Exception')) {
require 'Zend/Exception.php';
}
Work and be a lot faster, considering the number of files loaded for a
simple application ( 32 ) it might make sense.
I think the overhead is actually caused by including the files and not
by checking if the files were previously loaded (since these were
optimized in recent php updates). And on top of this, all of those files
are in fact required so how much will you save? Like 5-10% of all
require_once calls?
It's always worth to try but I wouldn't expect any miracles out of this.
Now, on the Paul's benchmarks. I don't know how Solar is like 4 times
faster but I'll tell you why Symphony is significantly leaner. As far as
I know, it's based on Mojavi code base and Mojavi used to gather all of
it's classes in one big cached file on the first run. And while doing so
it avoided loading of tons of the files in a single request. Loading one
big file plus cached user configs instead.
Top of config_compile.conf.php:
<?php
// auth-generated by CompileConfigHandler
// date: 10/12/2006 15:17:25
abstract class Action extends MojaviObject
{
