[llvm] [AMDGPU][SDAG] Legalise v2i32 or/xor/and instructions to make use of 64-bit wide instructions (PR #140694)

[Janek van Oirschot via llvm-commits]

================
@@ -5929,6 +5937,18 @@ SDValue SITargetLowering::splitUnaryVectorOp(SDValue Op,
   return DAG.getNode(ISD::CONCAT_VECTORS, SDLoc(Op), VT, OpLo, OpHi);
 }
 
+// Enable lowering of ROTR for vxi32 types. This is a workaround for a
+// regression caused by legalising v2i32 or.
+SDValue SITargetLowering::lowerROTR(SDValue Op, SelectionDAG &DAG) const {
+  EVT VT = Op.getValueType();
+
+  assert((VT == MVT::v2i32 || VT == MVT::v4i32 || VT == MVT::v8i32 ||
+          VT == MVT::v16i32) &&
+         "Unexpected ValueType.");
----------------
JanekvO wrote:

May warn on non-assert builds: `[[maybe_unused]]` or `(void)VT;`

https://github.com/llvm/llvm-project/pull/140694