This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[patch] target/21927
- From: Eric Christopher <echristo at redhat dot com>
- To: gcc-patches at gcc dot gnu dot org
- Date: Sun, 05 Jun 2005 22:12:07 -0700
- Subject: [patch] target/21927
This was a warning that didn't make a whole lot of sense to me when I
saw it before, but I hadn't been convinced I'd put it in on my target.
Apparently I had when I decided that BRANCH_COST should be unsigned on
my target :)
I don't see the point of a negative branch cost on any target,
especially when the doc says that the default is 1 and everything is
based relative to that. I mean, you may want zero to make it cheaper
than anything else, but negative just seems ridiculous.
Tested by building cc1 for mipsisa64-elf. Bootstrap in process.
OK?
-eric
2005-06-05 Eric Christopher <echristo@redhat.com>
target/21927
* expr.c (do_store_flag): Remove check for non-negative BRANCH_COST.
Index: expr.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/expr.c,v
retrieving revision 1.792
diff -u -p -w -r1.792 expr.c
--- expr.c 4 Jun 2005 17:22:17 -0000 1.792
+++ expr.c 6 Jun 2005 05:11:53 -0000
@@ -8729,8 +8729,7 @@ do_store_flag (tree exp, rtx target, enu
if ((code == LT && integer_zerop (arg1))
|| (! only_cheap && code == GE && integer_zerop (arg1)))
;
- else if (BRANCH_COST >= 0
- && ! only_cheap && (code == NE || code == EQ)
+ else if (! only_cheap && (code == NE || code == EQ)
&& TREE_CODE (type) != REAL_TYPE
&& ((abs_optab->handlers[(int) operand_mode].insn_code
!= CODE_FOR_nothing)