This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] Fix PR target/13785
- From: Andrew Pinski <pinskia at physics dot uc dot edu>
- To: Gcc Patch List <gcc-patches at gcc dot gnu dot org>, fj Jahanian <fjahanian at apple dot com>
- Cc: Andrew Pinski <pinskia at physics dot uc dot edu>
- Date: Wed, 21 Jan 2004 12:42:03 -0800
- Subject: [PATCH] Fix PR target/13785
The problem here looks like a simple copy and paste error when David
fixed some code
generation problem.
This problem only exists when compiling 64bit in 32bit userland aka
-mpowerpc64.
The testcase is:
long long (*y)(int t);
long long get_alias_set (int t)
{
return y(t);
}
OK for mainline and 3.4? Bootstrapped on powerpc-apple-darwin7.2.0 with
no regressions.
This is just one step further on bootstrapping with -mcpu=G5 on darwin.
Thanks,
Andrew Pinski
ChangeLog:
* rs6000.md (call_value): Force operand 1 not operand 0
into a register.
Patch:
Index: rs6000.md
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/rs6000/rs6000.md,v
retrieving revision 1.284
diff -u -p -r1.284 rs6000.md
--- rs6000.md 12 Jan 2004 18:37:40 -0000 1.284
+++ rs6000.md 21 Jan 2004 20:31:44 -0000
@@ -10181,7 +10181,7 @@
if (DEFAULT_ABI == ABI_V4
|| DEFAULT_ABI == ABI_DARWIN)
- operands[0] = force_reg (Pmode, operands[0]);
+ operands[1] = force_reg (Pmode, operands[1]);
else if (DEFAULT_ABI == ABI_AIX)
{