Hi,
When possible, we translate VEC_PERM_EXPRs into vector splat operations.
A previous patch fixed this up for little endian, but there was an
oversight for V16QI mode. This patch fixes that oversight. This fixes
the failing test gcc.dg/torture/vshuf-v16qi.c.
Bootstrapped and tested on powerpc64{,le}-unknown-linux-gnu using
--with-cpu=power7 with no regressions. Is this ok for trunk?
Thanks,
Bill
2013-11-19 Bill Schmidt <[email protected]>
* config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Adjust
V16QI vector splat case for little endian.
Index: gcc/config/rs6000/rs6000.c
===================================================================
--- gcc/config/rs6000/rs6000.c (revision 205002)
+++ gcc/config/rs6000/rs6000.c (working copy)
@@ -29849,6 +29849,8 @@ altivec_expand_vec_perm_const (rtx operands[4])
break;
if (i == 16)
{
+ if (!BYTES_BIG_ENDIAN)
+ elt = 15 - elt;
emit_insn (gen_altivec_vspltb (target, op0, GEN_INT (elt)));
return true;
}