This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFA:] Fix PR rtl-optimization/20466, flow not invalidating knowledge for (mem (mem ...))
- From: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- To: dje at watson dot ibm dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 21 Mar 2005 23:35:11 +0100
- Subject: Re: [RFA:] Fix PR rtl-optimization/20466, flow not invalidating knowledge for (mem (mem ...))
> Date: Mon, 21 Mar 2005 23:29:17 +0100
> From: Hans-Peter Nilsson <hp@axis.com>
> > Date: Mon, 21 Mar 2005 16:31:28 -0500
> > From: David Edelsohn <dje@watson.ibm.com>
> > If this change catches an address that is in a generic alias set, it will
> > cause a lot of extra memory traffic.
>
> Only if invalidate_mems_from_set and/or reg_overlap_mentioned_p
> actually looked at alias sets, which they don't.
Or more to the point regarding the patch,
reg_overlap_mentioned_p (x, in) says a x=MEM does not overlap a
in=non-MEM expression (but all x=MEMs overlap all in=MEMs; but
"in" is never MEM for a machine that doesn't have (mem ...) in
an address.
brgds, H-P