This is the mail archive of the
mailing list for the GCC project.
Re: PR 6212
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: mark at codesourcery dot com
- Cc: gcc at gcc dot gnu dot org
- Date: Tue, 7 May 02 08:12:57 EDT
- Subject: Re: PR 6212
Kenner, you're going to have to come up with a better patch.
Yesterday would be a good time.
I don't know of any, short of the major surgery to add the
NOP_WITH_ASSERT_EXPR, which we clearly can't do for 3.1.
Right now, what we have is a peformance issue, with no broken code.
With this patch, we have obscure cases (which had to be generated
specifically to make it fail) where the alignment is set wrong, but
where it actually doesn't matter in that particular case.
For 3.1, I think we're going to be stuck with this choice. My
recommendation is to go ahead with the patch since the cases where it
sets the wrong alignment are far less frequent than the cases when the
current code does so.
(The only other way is to disable fold's merging of MULT and DIV, but that
will cause far worse performance problems than this bug.)