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

[Chris Jackson via llvm-commits]

================
@@ -4189,6 +4234,53 @@ SDValue AMDGPUTargetLowering::performSrlCombine(SDNode *N,
   SDLoc SL(N);
   unsigned RHSVal;
 
+  // When the shl64_reduce optimisation code is passed through vector
----------------
chrisjbris wrote:

Yes, it's the same combine correction so that ISel sees (shl (and )) rather rather than (shl (extract_element (and))).

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