rs6000.c (rs6000_expand_vec_perm_const_1): Correct for little endian.

2013-11-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

	* config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Correct
	for little endian.

From-SVN: r205333

gcc/ChangeLog

+2013-11-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
+
+	* config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Correct
+	for little endian.
+
 2013-11-24  H.J. Lu  <hongjiu.lu@intel.com>
 
 	* graphite-sese-to-poly.c: Don't include extra "expr.h".

gcc/config/rs6000/rs6000.c

@@ -30046,6 +30046,21 @@ rs6000_expand_vec_perm_const_1 (rtx target, rtx op0, rtx op1,
       gcc_assert (GET_MODE_NUNITS (vmode) == 2);
       dmode = mode_for_vector (GET_MODE_INNER (vmode), 4);
 
+      /* For little endian, swap operands and invert/swap selectors
+	 to get the correct xxpermdi.  The operand swap sets up the
+	 inputs as a little endian array.  The selectors are swapped
+	 because they are defined to use big endian ordering.  The
+	 selectors are inverted to get the correct doublewords for
+	 little endian ordering.  */
+      if (!BYTES_BIG_ENDIAN)
+	{
+	  int n;
+	  perm0 = 3 - perm0;
+	  perm1 = 3 - perm1;
+	  n = perm0, perm0 = perm1, perm1 = n;
+	  x = op0, op0 = op1, op1 = x;
+	}
+
       x = gen_rtx_VEC_CONCAT (dmode, op0, op1);
       v = gen_rtvec (2, GEN_INT (perm0), GEN_INT (perm1));
       x = gen_rtx_VEC_SELECT (vmode, x, gen_rtx_PARALLEL (VOIDmode, v));