[llvm] [PowerPC] Fix vector_shuffle combines when inputs are scalar_to_vector of differing types. (PR #80784)

[Amy Kwan via llvm-commits]

================
@@ -102,42 +102,42 @@ define <2 x i64> @test_none_v16i8(i8 %arg1, ptr nocapture noundef readonly %b) {
 ; CHECK-LE-P8-NEXT:    lxvd2x vs0, 0, r4
 ; CHECK-LE-P8-NEXT:    xxswapd v2, vs0
 ; CHECK-LE-P8-NEXT:    mtfprd f0, r3
-; CHECK-LE-P8-NEXT:    xxmrgld v2, vs0, v2
+; CHECK-LE-P8-NEXT:    xxpermdi v2, vs0, v2, 1
----------------
amy-kwan wrote:

I think this should be resolved now from our offline discussion, but there was an issue in the previous codegen where we produced a `SCALAR_TO_VECTOR_PERMUTED` but we did not fix up the shuffle mask. This patch updates the shuffle mask, so this codegen should be correct.

https://github.com/llvm/llvm-project/pull/80784