Benchmark of new VM View as Markdown

On 2004-09-24 01:42:48 +0200, Andi Gutmans andi@zend.com said:

Hi,

Take it at face value but I get the following results with the new VM
vs. 5.0.x on the bench.php script I commited to php5/Zend/tests.

PHP 5.0.x PHP 5.1.x
37.5secs vs 29secs

You know I'm not a big believer in synthetic benchmarks but it's still
quite nice.

Andi

on my box:

5.0.x 45,3
5.1.x 36.1
4.3.x(*) 27.1

i just could not resist...

re,
tc

*with fasterharderstronger patch

Hi Thies,

I'll see which parts of your patch can be adapted for 5.1.x. I think the
var lookup is probably the most substantial one but I will probably
implement it slightly different. I'll mail internals@ when I have something.
As I mailed you, I tried the zval_copy_ctor() and zval_dtor() optimization
of yours but didn't get any improvement. Can you try and run the patch I
sent you to see if you get different results?
I'm also attaching it again, in case others want to see if it changes anything.

Andi

At 06:36 PM 9/25/2004 +0200, Thies C. Arntzen wrote:

On 2004-09-24 01:42:48 +0200, Andi Gutmans andi@zend.com said:

Hi,
Take it at face value but I get the following results with the new VM vs.
5.0.x on the bench.php script I commited to php5/Zend/tests.
PHP 5.0.x PHP 5.1.x
37.5secs vs 29secs
You know I'm not a big believer in synthetic benchmarks but it's still
quite nice.
Andi

on my box:

5.0.x 45,3
5.1.x 36.1
4.3.x(*) 27.1

i just could not resist...

re,
tc

*with fasterharderstronger patch

The script gets nuked when buildconf is run (since it matches *.php in
the .cvsignore file).
Not sure how to fix that :-)

--Wez.

bench.php script I commited to php5/Zend/tests.