[Mlir-commits] [clang] [llvm] [mlir] [IR] Replace alignment argument with attribute on masked intrinsics (PR #163802)

[Nikita Popov]

================
@@ -45,25 +45,20 @@ static bool tryToImproveAlign(
   switch (II->getIntrinsicID()) {
   case Intrinsic::masked_load:
   case Intrinsic::masked_store: {
-    int AlignOpIdx = II->getIntrinsicID() == Intrinsic::masked_load ? 1 : 2;
-    Value *PtrOp = II->getIntrinsicID() == Intrinsic::masked_load
-                       ? II->getArgOperand(0)
-                       : II->getArgOperand(1);
+    unsigned PtrOpIdx = II->getIntrinsicID() == Intrinsic::masked_load ? 0 : 1;
----------------
nikic wrote:

In the default pipeline, InferAlignment is that common pass. Though it would probably make sense to extend this code to handle all memory intrinsics instead of only masked.load/store now.

In the non-default pipeline, I expect that this will enable alignment inference in Attributor.

https://github.com/llvm/llvm-project/pull/163802