Big Perl performance problem in FC5

[SadGamerGeek]

Please note that the Perl issue only affects SlimServer before 6.5.

I'm going to have to beg to differ on this. I've just done some comparison timings. I selected "All Songs" after "Albums" from the web page 3 times for each version of Perl. The latest Perl consistently takes 4 times as long as version 4! This is using the latest 6.5b release.

Perl 5.8.8-4
============
2006-07-16 23:21:58.6197 Web Page Build > .5 : 4.80213212966919

2006-07-16 23:22:25.5047 Web Page Build > .5 : 4.74195599555969

2006-07-16 23:25:42.8866 Web Page Build > .5 : 4.69465899467468

Perl 5.8.8-5
============
2006-07-16 23:28:16.4340 Web Page Build > .5 : 19.551097869873

2006-07-16 23:29:17.3801 Web Page Build > .5 : 19.6564581394196

2006-07-16 23:31:33.5804 Select Task > .5 : 19.6658449172974

I'm getting pretty fast at swapping between them now, so let me know if there are any particular tests you'd like run.

[Dan Sully]

* SadGamerGeek shaped the electrons to say...

>> Please note that the Perl issue only affects SlimServer before 6.5.
>
>I'm going to have to beg to differ on this. I've just done some
>comparison timings. I selected "All Songs" after "Albums" from the web
>page 3 times for each version of Perl. The latest Perl consistently
>takes 4 times as long as version 4! This is using the latest 6.5b
>release.

Yes, it's slower - but not due to that Perl bug.

-D
--
dd if=/dev/sarcasm of=/dev/clue

[SadGamerGeek]

Yes, it's slower - but not due to that Perl bug.

Fair enough.

Does this mean there's likely to be another Perl 5.8.8-5 bug causing this?

[Dan Sully]

* SadGamerGeek shaped the electrons to say...

>Does this mean there's likely to be another Perl 5.8.8-5 bug causing this?

No.. the code just hasn't been optimized yet.

Though it's fast for me on a Macbook Pro

-D
--
Ya gotta love UNIX, where else do you wonder whether
you can kill a zombie spawned by a daemon's fork?

[stinkingpig]

....

> Though it's fast for me on a Macbook Pro


.....
<humor type=sarcastic style=strangelove>
Always thought you were a red Commie, first the rotating signature, then an
Apple product... what's next, an IDE with drag and drop?
</humor>
--
"I spent all me tin with the ladies drinking gin,
So across the Western ocean I must wander" -- traditional

[ChrisOwens]

Also see http://forums.slimdevices.com/showthread.php?p=122348

[SadGamerGeek]

* SadGamerGeek shaped the electrons to say...

>Does this mean there's likely to be another Perl 5.8.8-5 bug causing this?

No.. the code just hasn't been optimized yet.

I must have got the wrong end of the stick here, as I'm still a bit confused on this. Does the code have to be optimised for each Perl release then?

Richard

[Dan Sully]

* SadGamerGeek shaped the electrons to say...

>I must have got the wrong end of the stick here, as I'm still a bit
>confused on this. Does the code have to be optimised for each Perl
>release then?

No. There are some general performance issues with 6.5 in some circumstances.

-D
--
<faisal> my life is collapsing to what will soon be NEGATIVE INTEGER degrees of separation.

[SadGamerGeek]

* SadGamerGeek shaped the electrons to say...

>I must have got the wrong end of the stick here, as I'm still a bit
>confused on this. Does the code have to be optimised for each Perl
>release then?

No. There are some general performance issues with 6.5 in some circumstances.

OK, but doesn't this problem sound directly related to the version of Perl rather than SlimServer?

[Triode]

I can confirm there was a significant performance problem with FC5 perl5.8.8-5 for 6.5.

This should be resolved in the latest 6.5 versions - svn 8581 and the next nightly.

It turns out that although the DBIx library does not rebless, the way it blesses its object seemed to trigger the slow code added to the latest perl. Slightly altering how DBIx blesses the ResultSet object avoids this.

Thanks to Richard for letting me play with his machine for so long to find this!

Adrian