This is the mail archive of the
mailing list for the GCC project.
Re: [patch] Loop distribution for single nested loops
On Dec 26, 2007 12:38 PM, Mark Mitchell <email@example.com> wrote:
> I believe you've now fixed the ICE in apsi.f; I'm going to read further
> to see if I can find that message.
Yes, I've fixed that bug just before sending the patch to Toon. I
committed it to the graphite branch:
> I don't think documentation is sufficient. We know that users ignore
> it. If we provide a feature, it needs to work. I think that the fact
> that we've historically been casual about this kind of thing leads to
> the perception that GCC is never really as stable as the proprietary
One of the reasons I didn't proposed the loop distribution before is
that I knew it was buggy, so I preferred to fix the bugs first and
then ask for inclusion when I was confident enough of its stability.
I believe it is now stable enough, as Toon shown with his experiment
on HIRLAM and with my tests of SPEC Cpu2000 and Cpu2006.
AMD - GNU Tools