This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[new-ra] more enable-checking fallout
- From: Michael Matz <matz at suse dot de>
- To: gcc-patches at gcc dot gnu dot org
- Date: Mon, 13 Oct 2003 13:23:53 +0200 (CEST)
- Subject: [new-ra] more enable-checking fallout
Hi,
another thing noticed by --enable-checking. Bootstrapping is fairly slow
with that btw. ;-)
Ciao,
Michael.
--
* pre-reload.c (prefer_swapped): Change exit test to not
apply BLOCK_FOR_INSN on barriers.
Index: pre-reload.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/Attic/pre-reload.c,v
retrieving revision 1.1.2.18
retrieving revision 1.1.2.19
diff -u -p -r1.1.2.18 -r1.1.2.19
--- pre-reload.c 10 Oct 2003 18:20:57 -0000 1.1.2.18
+++ pre-reload.c 13 Oct 2003 11:19:17 -0000 1.1.2.19
@@ -2133,7 +2133,7 @@ prefer_swapped (insn, op0, op1)
op1 = SUBREG_REG (op1);
if (!bb || !REG_P (op0) || !REG_P (op1))
return 0;
- while (insn && bb == BLOCK_FOR_INSN (insn))
+ while (insn)
{
/* The way it's written (first testing OP0, then OP1) ensures,
that we do not prefer swapping if both ops die in the same insn. */
@@ -2142,6 +2142,8 @@ prefer_swapped (insn, op0, op1)
if (find_reg_note (insn, REG_DEAD, op1))
return 1;
insn = NEXT_INSN (insn);
+ if (insn == bb->end)
+ break;
}
return 0;
}