[llvm] [NVPTX] support packed f32 instructions for sm_100+ (PR #126337)

[Alex MacLean via llvm-commits]

================
@@ -4986,9 +5017,11 @@ combineUnpackingMovIntoLoad(SDNode *N, TargetLowering::DAGCombinerInfo &DCI) {
     return SDValue();
 
   EVT ElemVT = N->getValueType(0);
-  if (!Isv2x16VT(ElemVT))
+  if (!isPackedVectorTy(ElemVT))
----------------
AlexMaclean wrote:

Same question here

https://github.com/llvm/llvm-project/pull/126337