This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: PR2876 vs. reload
- To: Mark Mitchell <mark at codesourcery dot com>
- Subject: Re: PR2876 vs. reload
- From: Bernd Schmidt <bernds at redhat dot com>
- Date: Tue, 5 Jun 2001 18:04:01 +0100 (BST)
- cc: <gcc at gcc dot gnu dot org>, <rth at cygnus dot com>
On Tue, 5 Jun 2001, Mark Mitchell wrote:
> >>>>> "Bernd" == Bernd Schmidt <bernds@redhat.com> writes:
>
> First, your patch does fix the problem, which is not too surprising
> given:
>
> Bernd> It just gets rid of all USEs generated by reload. Since
>
> I think I asked the wrong question: I meant what is the point of
> creating the USE in the first place? There is no comment on the code
> that creates the USE, so it's very hard to understand why we're doing
> that.
It tells code in reload1.c that there's a reference to a pseudo somewhere
in the insn, but it's no longer obvious since the pseudo has been replaced
with something equivalent (probably a memory slot). If we don't know about
this we may do invalid transformations later on (e.g. deleting an output
reload that is storing into that pseudo).
Bernd