[llvm] [GlobalISel] Migrate remaining generic wip_match_opcode combines to MIR-pattern. (PR #222851)

[Yaxun Liu via llvm-commits]

================
@@ -572,3 +572,14 @@ bool AMDGPUCombinerHelper::matchConstantIs32BitMask(Register Reg) const {
   // Check if low 32 bits or high 32 bits are all ones.
   return MaskLen >= 32 && ((MaskIdx == 0) || (MaskIdx == 64 - MaskLen));
 }
+
+bool AMDGPUCombinerHelper::matchBinopWith32BitMask(Register X,
+                                                   Register Y) const {
+  // If both operands are constants, constant_fold_binop can fold the whole
----------------
yxsamliu wrote:

Could you add `G_AND` and `G_OR` tests where both operands are constants? They should confirm that this bailout lets `constant_fold_binop` produce one `G_CONSTANT`.

https://github.com/llvm/llvm-project/pull/222851