[llvm] Handle VECREDUCE intrinsics in NVPTX backend (PR #136253)

[Alex MacLean via llvm-commits]

================
@@ -83,6 +83,8 @@ class NVPTXTTIImpl : public BasicTTIImplBase<NVPTXTTIImpl> {
   }
   unsigned getMinVectorRegisterBitWidth() const { return 32; }
 
+  bool shouldExpandReduction(const IntrinsicInst *II) const { return false; }
----------------
AlexMaclean wrote:

Why have we completely disabled the generic expansion? It seems like for many of these operations we're not doing anything special and the existing SelectionDAG support would work just as well. 

https://github.com/llvm/llvm-project/pull/136253