Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:22367
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
To: "'Rasmus Lerdorf'" <rasmus@lerdorf.com>,
	"'internals'" <internals@lists.php.net>
Date: Mon, 13 Mar 2006 15:12:01 +0300
Message-ID: <000001c64697$5936afc0$6e02a8c0@thinkpad>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Importance: Normal
In-Reply-To: <4414F63F.5030406@lerdorf.com>
Subject: RE: [PHP-DEV] Calling performance geeks
From: dmitry@zend.com ("Dmitry Stogov")
References: <4414F63F.5030406@lerdorf.com>

Hi Rasmus,

I made two improvements in 5.1 and run the same bechmarks on Intel =
Pentium M
1.5GHz 2M cache.

top/top5/top10
	=09
php-5.1		740	550	430 req/sec
php-4.4		680	440	290 req/sec

May be the problem is AMD chip? :)

Thanks. Dmitry.

> -----Original Message-----
> From: Rasmus Lerdorf [mailto:rasmus@lerdorf.com]=20
> Sent: Monday, March 13, 2006 7:34 AM
> To: internals
> Subject: [PHP-DEV] Calling performance geeks
>=20
>=20
> We have a bit of a performance disconnect between 4.4 and 5.1=20
> still.  I=20
> was doing some benchmarking today just as a sanity check on some APC=20
> work I have been doing lately and came up with this:
>=20
>    http://lerdorf.com/php/bm.html
>=20
> You can ignore the apc/eaccelerator stuff.  Those numbers are not=20
> surprising.  The surprising number to me is how much faster=20
> 4.4 still is.
>=20
> The graph labels are slightly off.  The 0, 5 and 10 includes should=20
> really be 1, 6 and 11.  The actual benchmark code is here:
>=20
>    http://www.php.net/~rasmus/bm.tar.gz
>=20
> Tested on a Linux 2.6 Ubuntu box on an AMD chip (syscalls are cheap=20
> there) with current PHP_4_4 and PHP_5_1 checkouts.  Was also testing=20
> 5.1.2 to see the effect of getting rid of that uncached realpath call.
>=20
> As far as I can tell auto_globals_jit isn't working at all, but I=20
> eliminated that by doing variables_order =3D GP for these benchmarks.=20
> Even so, the request_startup is significantly more expensive in 5.1.
>=20
> Here are callgrind dumps for each.  Load them up with kcachegrind and=20
> browse around:
>=20
> PHP 4.4  http://www.php.net/~rasmus/callgrind.out.1528.gz
> PHP 5.1  http://www.php.net/~rasmus/callgrind.out.1488.gz
>=20
> Each of these is 1000 requests against the top.php and=20
> 4top.php scripts.=20
>   from bm.tar.gz.  If you start at the
>=20
> The script is trivial and looks like this:
>=20
> <html>
> <?php
> $base_dir =3D '/var/www/bm/';
> include $base_dir . 'config.inc';
>=20
> function top_func($arg) {
>    $b =3D $arg.$arg;
>    echo $b;
> }
> class top_class {
>    private $prop;
>    function __construct($arg) {
>      $this->prop =3D $arg;
>    }
>    function getProp() {
>      return $this->prop;
>    }
>    function setProp($arg) {
>      $this->prop =3D strtolower($arg);
>    }
> }
>=20
> top_func('foo');
> $a =3D new top_class('bar');
> echo $a->getProp();
> $a->setProp("AbCdEfG");
> echo $a->getProp();
> echo <<<EOB
> The database is {$config['db']}
> and the user is {$config['db_user']}
>=20
> EOB;
> ?>
> </html>
>=20
> and config.inc is:
>=20
> <?php
> $config =3D array(
>    'db'      =3D> 'mysql',
>    'db_user' =3D> 'www',
>    'db_pwd'  =3D> 'foobar',
>    'config1' =3D> 123,
>    'config2' =3D> 456,
>    'config3' =3D> 789,
>    'sub1'    =3D> array(1,2,3,4,5,6,7,8,9,10),
>    'sub2'    =3D>=20
> array("abc","def","ghi","jkl","mno","pqr","stu","vwx","yz")
> );
> ?>
>=20
> 4top.php is identical except for the class definition being=20
> PHP 4-style=20
> instead.  As in no private and a PHP 4 constructor.  Otherwise it is=20
> identical.
>=20
> I have some ideas for things we can speed up in 5.1.  Like,=20
> for example,=20
> we should add the ap_add_common_vars() and ap_add_cgi_vars()=20
> to the jit=20
> mechanism.  There isn't much point filling these in unless the script=20
> tries to get them.  the ap_add_common_vars() call is=20
> extremely expensive=20
> since it does a qsort with a comparison function that uses=20
> strcasecmp.=20
> Of course, this same optimization can be done in 4.4.
>=20
> If you know your way around kcachegrind, load up the two=20
> callgrind files=20
> and see what stands out for you.  As far as I can tell, while=20
> we can do=20
> some tricks to speed up various helper bits, the slowdown is=20
> coming from=20
> the executor trashing its cache lines.
>=20
> -Rasmus
>=20
> --=20
> PHP Internals - PHP Runtime Development Mailing List
> To unsubscribe, visit: http://www.php.net/unsub.php
>=20
>=20
>=20