This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] Fix gcc.dg/20020201-3.c
- From: Jakub Jelinek <jakub at redhat dot com>
- To: rth at redhat dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sat, 2 Feb 2002 22:55:35 +0100
- Subject: [PATCH] Fix gcc.dg/20020201-3.c
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
Hi!
The following patch fixes the IA-32 -mmmx ICE testcase I've commited
yesterday. I don't see why a SFmode value cannot be moved from or into an
MMX register similarly to movsi (though it certainly shouldn't be
encouraged).
Ok to commit if testing is ok?
2002-02-02 Jakub Jelinek <jakub@redhat.com>
* config/i386/i386.md (movsf_1): Allow moving SF values in MMX
registers.
--- gcc/config/i386/i386.md.jj Wed Jan 23 16:29:20 2002
+++ gcc/config/i386/i386.md Sat Feb 2 23:32:41 2002
@@ -2713,8 +2713,8 @@
(set (mem:SF (reg:DI 7)) (match_dup 1))])
(define_insn "*movsf_1"
- [(set (match_operand:SF 0 "nonimmediate_operand" "=f#xr,m,f#xr,r#xf,m,x#rf,x#rf,x#rf,m")
- (match_operand:SF 1 "general_operand" "fm#rx,f#rx,G,rmF#fx,Fr#fx,H,x,xm#rf,x#rf"))]
+ [(set (match_operand:SF 0 "nonimmediate_operand" "=f#xr,m,f#xr,r#xf,m,x#rf,x#rf,x#rf,m,!*y,!rm")
+ (match_operand:SF 1 "general_operand" "fm#rx,f#rx,G,rmF#fx,Fr#fx,H,x,xm#rf,x#rf,rm,*y"))]
"(GET_CODE (operands[0]) != MEM || GET_CODE (operands[1]) != MEM)
&& (reload_in_progress || reload_completed
|| (ix86_cmodel == CM_MEDIUM || ix86_cmodel == CM_LARGE)
@@ -2762,12 +2762,16 @@
case 8:
return "movss\t{%1, %0|%0, %1}";
+ case 9:
+ case 10:
+ return "movd\t{%1, %0|%0, %1}";
+
default:
abort();
}
}
- [(set_attr "type" "fmov,fmov,fmov,imov,imov,sse,sse,sse,sse")
- (set_attr "mode" "SF,SF,SF,SI,SI,TI,SF,SF,SF")])
+ [(set_attr "type" "fmov,fmov,fmov,imov,imov,sse,sse,sse,sse,mmx,mmx")
+ (set_attr "mode" "SF,SF,SF,SI,SI,TI,SF,SF,SF,SI,SI")])
(define_insn "*swapsf"
[(set (match_operand:SF 0 "register_operand" "+f")
Jakub