Odd performance problem

Hello,

While debugging a script that ran too slowly I came across something that I
can't explain.

It is inactive code that when removed doubles the run time of the script.

Specifically, the issue is a switch statement, where one of the cases is
never reached. If I remove the case from the code, the run time of the
script goes from 6 seconds to 12 seconds.

To see this problem in action, you can download the problematic script here:
http://sebrechts.net/files/2008/phpperfissue.zip

Anyone have any ideas on how this is even possible? I'd like to improve the
run time of the script, but if every part of code that I take out makes it
run longer ...

Thanks

On Wed, 2008-05-07 at 17:31 +0200, Joeri Sebrechts wrote:
> Hello,
>
> While debugging a script that ran too slowly I came across something that I
> can't explain.
>
> It is inactive code that when removed doubles the run time of the script.
>
> Specifically, the issue is a switch statement, where one of the cases is
> never reached. If I remove the case from the code, the run time of the
> script goes from 6 seconds to 12 seconds.
>
> To see this problem in action, you can download the problematic script here:
> http://sebrechts.net/files/2008/phpperfissue.zip
>
> Anyone have any ideas on how this is even possible? I'd like to improve the
> run time of the script, but if every part of code that I take out makes it
> run longer ...

Runs the same for me either way... under PHP4 and PHP5. You must have a
nice system, takes 22 seconds both ways on my Athlon 2400 :)

Cheers,
Rob.
--
http://www.interjinn.com
Application and Templating Framework for PHP

My tests on PHP 4.4.8

Default files:

Run 1)
done at 100045

Time: 12.5804 Seconds

Run 2)
done at 100045

Time: 12.6775 Seconds

Run 3)
done at 100045

Time: 13.0696 Seconds

Now editing the file, removed the switch case 'a' part.

Run 1)
done at 100045

Time: 12.6981 Seconds

Run 2)
done at 100045

Time: 12.8659 Seconds

Run 3)
done at 100045

Time: 12.7886 Seconds

As you can see, there's no really big difference here. It really does look
like it's slower, but .1 sec slower.


Hey Robert, this was just a semprom 3k... didn't know they were that dif



Thiago Henrique Pojda
Desenvolvimento Web
+55 41 3033-7676
thiago.pojda@softpartech.com.br
Excelência em Softwares Financeiros

-----Mensagem original-----
De: Robert Cummings [mailto:robert@interjinn.com]
Enviada em: sexta-feira, 9 de maio de 2008 04:44
Para: Joeri Sebrechts
Cc: php-general@lists.php.net
Assunto: Re: [php] Odd performance problem

On Wed, 2008-05-07 at 17:31 +0200, Joeri Sebrechts wrote:
> Hello,
>
> While debugging a script that ran too slowly I came across something that
I
> can't explain.
>
> It is inactive code that when removed doubles the run time of the script.
>
> Specifically, the issue is a switch statement, where one of the cases is
> never reached. If I remove the case from the code, the run time of the
> script goes from 6 seconds to 12 seconds.
>
> To see this problem in action, you can download the problematic script
here:
> http://sebrechts.net/files/2008/phpperfissue.zip
>
> Anyone have any ideas on how this is even possible? I'd like to improve
the
> run time of the script, but if every part of code that I take out makes it

> run longer ...

Runs the same for me either way... under PHP4 and PHP5. You must have a
nice system, takes 22 seconds both ways on my Athlon 2400 :)

Cheers,
Rob.
--
http://www.interjinn.com
Application and Templating Framework for PHP


--
PHP General Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php

Ha, I tried again today, and now there's no difference. I'm thinking now
that this was some sort of page boundary issue, or something of the sort. On
the one hand it's a relief I can move on, on the other hand it's frustrating
because I know for a fact that I saw those differences, and this problem
might be happening for some of our customers.

I'll just blame windows, that always works :)

By the way, a 22 second run-time is definitely abnormal. I only have a 3 ghz
P4, a 3 year old machine. When I disable xdebug it takes 4.5 seconds.
Another by the way, while measuring the performance of this script I noticed
that wincachegrind's time reports are incorrect. It reported a cumulative
run-time of about one second. Kcachegrind (the KDE4 for windows version) did
report correct times, so I'll be using that from now on.


""Thiago Pojda"" <thiago.pojda@softpartech.com.br> wrote in message
news:012501c8b1c9$bb172c20$0201a8c0@softpartech...
My tests on PHP 4.4.8

Default files:

Run 1)
done at 100045

Time: 12.5804 Seconds

Run 2)
done at 100045

Time: 12.6775 Seconds

Run 3)
done at 100045

Time: 13.0696 Seconds

Now editing the file, removed the switch case 'a' part.

Run 1)
done at 100045

Time: 12.6981 Seconds

Run 2)
done at 100045

Time: 12.8659 Seconds

Run 3)
done at 100045

Time: 12.7886 Seconds

As you can see, there's no really big difference here. It really does look
like it's slower, but .1 sec slower.


Hey Robert, this was just a semprom 3k... didn't know they were that dif



Thiago Henrique Pojda
Desenvolvimento Web
+55 41 3033-7676
thiago.pojda@softpartech.com.br
Excelência em Softwares Financeiros

-----Mensagem original-----
De: Robert Cummings [mailto:robert@interjinn.com]
Enviada em: sexta-feira, 9 de maio de 2008 04:44
Para: Joeri Sebrechts
Cc: php-general@lists.php.net
Assunto: Re: [php] Odd performance problem

On Wed, 2008-05-07 at 17:31 +0200, Joeri Sebrechts wrote:
> Hello,
>
> While debugging a script that ran too slowly I came across something that
I
> can't explain.
>
> It is inactive code that when removed doubles the run time of the script.
>
> Specifically, the issue is a switch statement, where one of the cases is
> never reached. If I remove the case from the code, the run time of the
> script goes from 6 seconds to 12 seconds.
>
> To see this problem in action, you can download the problematic script
here:
> http://sebrechts.net/files/2008/phpperfissue.zip
>
> Anyone have any ideas on how this is even possible? I'd like to improve
the
> run time of the script, but if every part of code that I take out makes it

> run longer ...

Runs the same for me either way... under PHP4 and PHP5. You must have a
nice system, takes 22 seconds both ways on my Athlon 2400 :)

Cheers,
Rob.
--
http://www.interjinn.com
Application and Templating Framework for PHP


--
PHP General Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php

On Fri, 2008-05-09 at 08:42 -0300, Thiago Pojda wrote:
>
> Runs the same for me either way... under PHP4 and PHP5. You must have
> a nice system, takes 22 seconds both ways on my Athlon 2400 :)

Maybe it's because I use Evolution for my mail client and it has
steadily increased it's CPU suckage over the years. I have no idea what
it does but these days I noticed it's eating a steady 8-12% of my
CPU :( Unfortunately I didn't like Thunderbird last time I checked.
Maybe I'll give Opera's mail client a go since I use the Opera browser
anyways.

Cheers,
Rob.
--
http://www.interjinn.com
Application and Templating Framework for PHP