Shawn Walker wrote:

> On 12/08/11 14:46, Danek Duvall wrote:
> >Shawn Walker wrote:
> >
> >>Greetings,
> >>
> >>The following webrev contains changes to improve the performance of
> >>python 2.6 and python 2.7 by using profile-guided optimisation
> >>during compilation:
> >>
> >>   6771457 explore compiler performance improvements for python
> >>
> >>   https://cr.opensolaris.org/action/browse/pkg/swalker/pyopt/webrev/
> >
> >python26/Makefile:
> >
> >   - line 57: It'd be nice to have an explanation of why these values were
> >     used (minimum configuration supported on S11, presumably)
> 
> Right; I'll add a comment.
> 
> >   - line 67 (removed): why are we enabling the static library?
> 
> We're not; that's not actually a valid configure option for Python
> 2.6, only for Python 2.7.  The static is already enabled regardless.
> 
> >   - line 87, 88: we've been pretty consistent about putting one definition
> >     per line.
> 
> Oh, I didn't even notice that.  Changed.
> 
> >Python26-24-symbolic.patch:
> >
> >   - A header to the patch explaining why this is necessary (or desired)
> >     would be nice.  As far as I could tell from the cc and ld docs, it's
> >     not required for profile-guided optimization, but does have performance
> >     benefits of its own.  You also don't talk about it in the bug, so it'll
> >     be hard for future maintainers to tell why this is here.
> 
> I'll add a comment to the bug and a patch header.
> 
> >Python26-25-studio-profile.patch:
> >
> >   - line 40, 41: the comment in the makefiles said that you need to compile
> >     and link with the same flags, but here the optimization flags are
> >     different.  Why the specific -xO1 in CFLAGS?
> 
> This is the coverage testing target which isn't used for
> profile-guided builds; studio requires at least optimization level 1
> and keeping the optimisation level lower results in better coverage
> testing.  I just wanted to be certain the target worked.

Works for me.

Thanks,
Danek
_______________________________________________
userland-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/userland-discuss

Reply via email to