[perl #45529] Performance problem with object creation in 64 bit perl 5.8.8 on linux

On Sun Sep 23 05:57:38 2007, philip@gladstonefamily.net wrote:
> After realizing that the current code doesn't correspond with the built
> code, I just pulled the redhat RPM, and there was the culprit.
> 
> There was a diff that adds the function S_reset_amagic that is called in
> exactly the right place. I attach the culprit patch to show what (I
> think) is causing the problem. It is certainly where all the CPU time is
> being spent. The original patch posting was in
> http://www.nntp.perl.org/group/perl.perl5.changes/2006/03/msg15320.html
> 
> This also suggests that the problem is not limited to 64-bit either!
> 
> Maybe someone more knowledgeable on perl internals can figure out why it
> has such pathological behavior in this case.
> 
> Thanks
> 
> Philip
> 
> On Fri Sep 21 19:54:48 2007, philip@gladstonefamily.net wrote:
> > Since this doesn't correspond to the current source code, and, when I 
> > build the current source code, I don't experience the problem, I am at 
> > a loss as to how to continue. It makes me a little nervous that there 
> > are multiple versions of 5.8.8 around which behave so differently.
> > 
> > Philip (more puzzled in Massachusetts)
> > 
> > 
> 
> 


As you suspected, this problem was caused by RedHat shipping a Perl with patches pulled in 
from the development version of Perl.  There is a reason why this was not yet in a production 
version.  Anyways, the fully fixed version was included in Perl 5.10.0 and in Perl 5.8.9-RC1.  
The best solution currently on RedHat is to compile a Perl from source and use that instead.  
Additional discussion of the problem is available at http://blog.vipul.net/2008/08/24/redhat-perl-what-a-tragedy/

Steve Peters

• [perl #45529] Performance problem with object creation in 64 bit perl 5.8.8 on linux by Steve Peters via RT