[Mlir-commits] [mlir] [MLIR] VectorEmulateNarrowType to support loading of unaligned vectors (PR #113411)

[Han-Chung Wang]

================
@@ -102,6 +129,36 @@ static FailureOr<Operation *> getCompressedMaskOp(OpBuilder &rewriter,
   return newMask;
 }
 
+static std::optional<int64_t>
+getIntraVectorOffset(ConversionPatternRewriter &rewriter, Location loc,
+                     const memref::LinearizedMemRefInfo linearizedInfo) {
+  auto foldedIntraVectorOffset = getValueOrCreateConstantIndexOp(
+      rewriter, loc, linearizedInfo.intraVectorOffset);
+  return getConstantIntValue(foldedIntraVectorOffset);
+}
----------------
hanhanW wrote:

I think we don't need the method. You can just replace all the usages with `getConstantIntValue(linearizedInfo.intraVectorOffset)`.

https://github.com/llvm/llvm-project/pull/113411