[patch] ppc-e500: fix evfscfsi
Aldy Hernandez
aldyh@redhat.com
Thu Jun 19 14:13:00 GMT 2003
Instruction converts an integer to a float, not the other way around.
Committed to mainline and e500 branch.
2003-06-19 Aldy Hernandez <aldyh@redhat.com>
* config/rs6000/spe.md ("spe_evfscfsi"): Change operand types.
Change "fix" to "float".
Index: config/rs6000/spe.md
===================================================================
RCS file: /cvs/cvsfiles/devo/gcc/config/rs6000/spe.md,v
retrieving revision 1.8.4.2
diff -c -p -r1.8.4.2 spe.md
*** spe.md 2003/06/12 16:38:36 1.8.4.2
--- spe.md 2003/06/19 13:09:20
***************
*** 639,646 ****
(set_attr "length" "4")])
(define_insn "spe_evfscfsi"
! [(set (match_operand:V2SI 0 "gpc_reg_operand" "=r")
! (fix:V2SI (match_operand:V2SF 1 "gpc_reg_operand" "r")))]
"TARGET_SPE"
"evfscfsi %0,%1"
[(set_attr "type" "vecfloat")
--- 639,646 ----
(set_attr "length" "4")])
(define_insn "spe_evfscfsi"
! [(set (match_operand:V2SF 0 "gpc_reg_operand" "=r")
! (float:V2SF (match_operand:V2SI 1 "gpc_reg_operand" "r")))]
"TARGET_SPE"
"evfscfsi %0,%1"
[(set_attr "type" "vecfloat")
More information about the Gcc-patches
mailing list