Message ID | 57743F20.6060604@linux.vnet.ibm.com |
---|---|
State | New |
Headers | show |
On Wed, Jun 29, 2016 at 03:35:28PM -0600, Kelvin Nilsen wrote: > Testing on a Power9 simulator revealed an error in the code emitted for > the *altivec_vpermr_<mode>_internal define_insn pattern. Two of the > operands of the vpermr operand had been emitted in wrong order. This > patch corrects the error. "Oops" :-) > The patch has bootstrapped and regression tested on > powerpc64le-unknown-linux with no regressions. Is this ok > for the trunk and for backporting to GCC6 after burn-in > on the trunk? Yes, thank you! Segher
Index: gcc/config/rs6000/altivec.md =================================================================== --- gcc/config/rs6000/altivec.md (revision 237799) +++ gcc/config/rs6000/altivec.md (working copy) @@ -2072,7 +2072,7 @@ UNSPEC_VPERMR))] "TARGET_P9_VECTOR" "@ - vpermr %0,%1,%2,%3 + vpermr %0,%2,%1,%3 xxpermr %x0,%x2,%x3" [(set_attr "type" "vecperm") (set_attr "length" "4")])