Simple patch to reverse the order of the input operands when
concatenating for little endian code generation. Bootstrapped and
tested on powerpc64-unknown-linux-gnu and powerpc64le-unknown-linux-gnu
with no regressions. Fixes two tests in the testsuite for the latter.
Ok for trunk?
Thanks,
Bill
2013-10-15 Bill Schmidt <wschm...@linux.vnet.ibm.com>
* config/rs6000/vsx.md (vsx_concat_<mode>): Adjust output for LE.
(vsx_concat_v2sf): Likewise.
Index: gcc/config/rs6000/vsx.md
===================================================================
--- gcc/config/rs6000/vsx.md (revision 203508)
+++ gcc/config/rs6000/vsx.md (working copy)
@@ -1194,7 +1194,12 @@
(match_operand:<VS_scalar> 1 "vsx_register_operand" "ws,wa")
(match_operand:<VS_scalar> 2 "vsx_register_operand" "ws,wa")))]
"VECTOR_MEM_VSX_P (<MODE>mode)"
- "xxpermdi %x0,%x1,%x2,0"
+{
+ if (BYTES_BIG_ENDIAN)
+ return "xxpermdi %x0,%x1,%x2,0";
+ else
+ return "xxpermdi %x0,%x2,%x1,0";
+}
[(set_attr "type" "vecperm")])
;; Special purpose concat using xxpermdi to glue two single precision values
@@ -1207,7 +1212,12 @@
(match_operand:SF 2 "vsx_register_operand" "f,f")]
UNSPEC_VSX_CONCAT))]
"VECTOR_MEM_VSX_P (V2DFmode)"
- "xxpermdi %x0,%x1,%x2,0"
+{
+ if (BYTES_BIG_ENDIAN)
+ return "xxpermdi %x0,%x1,%x2,0";
+ else
+ return "xxpermdi %x0,%x2,%x1,0";
+}
[(set_attr "type" "vecperm")])
;; xxpermdi for little endian loads and stores. We need several of
