[Mlir-commits] [mlir] [MLIR] support dynamic indexing in `VectorEmulateNarrowTypes` (PR #114169)

[llvmlistbot at llvm.org]

================
@@ -75,9 +77,6 @@ static FailureOr<Operation *> getCompressedMaskOp(OpBuilder &rewriter,
   shape.back() = numElements;
   auto newMaskType = VectorType::get(shape, rewriter.getI1Type());
   if (createMaskOp) {
-    // TODO: handle the case with non-zero intraDataOffset for CreateMaskOp.
-    if (intraDataOffset != 0)
-      return failure();
----------------
lialan wrote:

I realized that here we can relax this assert because whether statical or dynamic indexing, the generated compressed mask is correct to cover the necessary bytes we need to load. This is because `intraDataOffset < scale` so the beginning index of the compressed mask will not change (but the size of the mask could change).

https://github.com/llvm/llvm-project/pull/114169