This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
No Subject
- To: dje at watson dot ibm dot com, egcs at cygnus dot com, law at cygnus dot com, mark at markmitchell dot com, meissner at cygnus dot com
- From: "David S. Miller" <davem at pierdol dot cobaltmicro dot com>
- Date: Fri, 18 Sep 1998 13:33:37 -0700
From: "David S. Miller" <davem@pierdol.cobaltmicro.com>
To: dje@watson.ibm.com
CC: mark@markmitchell.com, law@cygnus.com, meissner@cygnus.com,
egcs@cygnus.com
In-reply-to: <9809182120.AA35006@marc.watson.ibm.com> (message from David
Edelsohn on Fri, 18 Sep 1998 17:20:45 -0400)
Subject: Re: PowerPC build breakage caused by 8/25/98 loop.c change
References: <9809182120.AA35006@marc.watson.ibm.com>
Date: Fri, 18 Sep 1998 17:20:45 -0400
From: David Edelsohn <dje@watson.ibm.com>
I tried building a recent CVS checkout with loop.c reverted
to version 1.65 on 8/25/98 prior to Mark's loop.c changes.
Bootstrapping is successful. Something about those changes
(possibly the AVOID_CC_MODE_COPIES tweak?) causes PowerPC failures.
If this is your theory, can you test it by reverting the
AVOID_CC_MODE_COPIES parts of the changes?
I believe we've addressed most of the issues wrt. the loop load/store
hoisting. And this 8/25 change fixes real bugs on any platform which
can create new pseudo registers during move_movables in loop.c
My more recent change to loop.c fixes a bug which existed long before
Mark's load store hoisting code went in, in particular it was
introduced by Rth's loop enhancements some time ago.
Later,
David S. Miller
davem@dm.cobaltmicro.com