This is the mail archive of the
mailing list for the GCC project.
Re: PR 11319
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: Mark Mitchell <mark at codesourcery dot com>, Jim Wilson <wilson at tuliptree dot org>
- Cc: gcc at gcc dot gnu dot org
- Date: Sun, 07 Sep 2003 17:32:16 -0400
- Subject: Re: PR 11319
>>>>> Jim Wilson writes:
> Yes, it is fixed. However, David Edelsohn registered a strong complaint
> on performance grounds, without providing much solid evidence, for
> instance, SPEC results. In order to investigate this complaint, I had
> to buy software (SPEC) and hardware (PowerPC), and along the way I
> caught the flu, so I haven't completed everything I wanted to do. I did
> run SPEC on a Pentium4 and there was no significant performance change
> resulting from the patch. I bought a PowerMac G5 tonight, so I should
> have PowerPC SPEC results soon. Meanwhile, I will enter the pending
> info I have into the PR.
The automated SPEC testers showed a drop in performance after the
patch. I guess all of the loads hoisted out of loops had no effect on
performance. If you want to close the PRs, go ahead.