This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix code quality regression on UltraSPARC
- From: Eric Botcazou <ebotcazou at libertysurf dot fr>
- To: Roger Sayle <roger at eyesopen dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 13 Dec 2004 07:59:04 +0100
- Subject: Re: [PATCH] Fix code quality regression on UltraSPARC
- References: <Pine.LNX.4.44.0412121428390.7027-100000@www.eyesopen.com>
> Agreed. In fact, with your proposed fallthruness test of the latch
> edge, I don't think we need to be much cleverer than we were before.
I actually was under the same impression.
> Can you check whether the following subset of your patch is sufficient
> to improve the UltraSPARC code we generate for gzip?
Yes, the fallthruness test of the latch edge is enough in this case (verified
on t28.ch and the final assembly code).
> I'm just trying to figure out which of your changes are necessary
> and which are sufficient. The "one_succ_pred" is a nice refinement,
> but I no longer think the FOR_EACH_EDGE/best_pred->next_bb helps
> any.
Agreed.
> This also finesses your question "??? Can this ever happen?",
> to which the answer is yes it can.
Do you mean the latch is allowed to have more than one edge to the header?
--
Eric Botcazou