This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: question regarding regrename and failure of 950704-1.c on main branch
- From: law at redhat dot com
- To: Richard Henderson <rth at redhat dot com>
- Cc: John David Anglin <dave at hiauly1 dot hia dot nrc dot ca>, gcc at gcc dot gnu dot org
- Date: Thu, 09 May 2002 15:20:00 -0600
- Subject: Re: question regarding regrename and failure of 950704-1.c on main branch
- Reply-to: law at redhat dot com
In message <20020509140441.A15717@redhat.com>, Richard Henderson writes:
> On Thu, May 09, 2002 at 04:49:55PM -0400, John David Anglin wrote:
> > DI 22 overlaps DI 23. There is no early clobber in the contraints for
> > the pattern *pa.md:3828.
> >
> > The failure started on or around April 30. Is it now necessary to mark
> > the output of DI mode arithmetic operations on 32-bit machines early
> > clobber to prevent overlaps such as this? Or, is this a regrename bug?
>
> It has _always_ been necessary to mark registers that are
> clobbered before all inputs have been consumed as earlyclobber.
>
> How you've gotten away with it until now, I have no idea.
But why did we end up with DImode value in reg23? That's the
underlying problem -- HARD_REGNO_MODE_OK should have rejected
putting a DImode value in reg23.
jeff