This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: odd things in md files
Richard Henderson wrote:
> On Sun, Sep 12, 1999 at 12:17:54AM -0700, Zack Weinberg wrote:
> > 1999-09-11 23:26 -0700 Zack Weinberg <zack@bitmover.com>
> >
> > * i370.md (mulsi, divsi, modsi anonymous insns): Use DImode
> > operation.
>
> Ok.
Committed.
Now, this is my best guess how to fix gmicro.md:
Index: gmicro.md
===================================================================
RCS file: /cvs/egcs/egcs/gcc/config/gmicro/gmicro.md,v
retrieving revision 1.4
diff -u -p -r1.4 gmicro.md
--- gmicro.md 1999/09/07 05:48:30 1.4
+++ gmicro.md 1999/09/13 19:33:41
@@ -57,7 +57,7 @@
(define_insn ""
[(set (match_operand:DI 0 "push_operand" "=m")
- (match_operand:DF 1 "general_operand" "rmF"))]
+ (subreg:DI (match_operand:DF 1 "general_operand" "rmF") 0))]
""
"*
{
but I don't know if that is legal, if it means what the author wanted
it to mean, or if it agrees with the effect of the output block.
This is not a float-to-int conversion pattern; I'm fairly sure the
intent is to stash a DFmode value in a DImode register temporarily.
zw