This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] fix PR39715 part 2
On Mon, 2009-08-03 at 22:41 +0200, Paolo Bonzini wrote:
> On 08/03/2009 02:43 PM, Richard Earnshaw wrote:
> > On Sat, 2009-08-01 at 23:24 +0200, Paolo Bonzini wrote:
> >> + "TARGET_THUMB1&& !(reload_in_progress || reload_completed)"
> >>
> >
> > I don't follow this. That causes the pattern to stop matching at the
> > time that reload starts. Surely that can't be what you mean.
>
> Since it's splitted, by the time reload is started you shouldn't have it
> anymore. I can remove the clause though and just make it TARGET_THUMB1,
> it shouldn't make any difference.
I think removing it would be wise. I'm fairly certain that in the past
we didn't always split before reload (for example, when not scheduling).
R.