[llvm] [AMDGPU] Extend wave reduce intrinsics for i32 type (PR #126469)

[via llvm-commits]

================
@@ -4940,6 +4940,28 @@ static MachineBasicBlock *emitIndirectDst(MachineInstr &MI,
   return LoopBB;
 }
 
+static uint32_t getInitialValueForWaveReduction(unsigned Opc) {
+  switch (Opc) {
+  case AMDGPU::S_MIN_U32:
+    return std::numeric_limits<uint32_t>::max();
+  case AMDGPU::S_MIN_I32:
+    return std::numeric_limits<int32_t>::max();
+  case AMDGPU::S_MAX_U32:
+    return std::numeric_limits<uint32_t>::min();
+  case AMDGPU::S_MAX_I32:
+    return std::numeric_limits<int32_t>::min();
+  case AMDGPU::S_ADD_I32:
+  case AMDGPU::S_SUB_I32:
+  case AMDGPU::S_OR_B32:
+  case AMDGPU::S_XOR_B32:
+    return std::numeric_limits<uint32_t>::min();
----------------
easyonaadit wrote:

The identity element for XOR is 0.
The same value is also used in the atomic optimizer pass. 


https://github.com/llvm/llvm-project/pull/126469