This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] Work around for an unfortunate fold-const vs. tree-optimizer interaction
- From: law at redhat dot com
- To: Andrew MacLeod <amacleod at redhat dot com>
- Cc: Zack Weinberg <zack at codesourcery dot com>, Roger Sayle <roger at eyesopen dot com>, Diego Novillo <dnovillo at redhat dot com>, Steven Bosschner <stevenb at suse dot de>, gcc-patches <gcc-patches at gcc dot gnu dot org>
- Date: Mon, 10 Nov 2003 13:32:19 -0700
- Subject: Re: [tree-ssa] Work around for an unfortunate fold-const vs. tree-optimizer interaction
- Reply-to: law at redhat dot com
In message <1068154239.7012.40.camel@p4>, Andrew MacLeod writes:
>On Thu, 2003-11-06 at 16:05, Zack Weinberg wrote:
>Ona side note, Im about to turn GIMPLE back into GENERIC in
>SSA->normal, but only as far as expressions go... ie, allowing something
>more than a two operand expression. So we can get a string of
>assignments into temporaries that are only used once turned back into an
[ ... ]
>Which the expanders/backend generate far better code for right now.
>Now, down the road if we make the back end or the expanders smart enough
>to take care of this, then we could always go back to generating
>straight GIMPLE. Its going to be under flag control anyway.
>I do think the plan is to abolish a lot of those other hideous things
>the expanders have to deal with tho, so they ought to be a lot more
So what I think this is saying is that long term the expanders need to
handle generic and that they should not assume gimple.
Just wanted to be explicit! :-)