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

[Han-Chung Wang]

================
@@ -129,26 +128,80 @@ static FailureOr<Operation *> getCompressedMaskOp(OpBuilder &rewriter,
   return newMask;
 }
 
+/// Extracts 1-D subvector from a 1-D vector. It is a wrapper function for
+/// emitting `vector.extract_strided_slice`.
 static Value extractSubvectorFrom(RewriterBase &rewriter, Location loc,
-                                  VectorType extractType, Value vector,
+                                  VectorType extractType, Value source,
                                   int64_t frontOffset, int64_t subvecSize) {
+  auto vectorType = dyn_cast<VectorType>(source.getType());
+  assert(vectorType && vectorType.getRank() == 1 &&
+         extractType.getRank() == 1 &&
+         "expected 1-D source and destination types");
----------------
hanhanW wrote:

nit1: I'd use `cast` which already asserts it for you.
nit2: I'd wrap all the conditions with a pair of brace. We've seen some bugs in assertions when there are **OR** (i.e., `||`) operations. Just in case to prevent a similar bug here.

```suggestion
  auto vectorType = cast<VectorType>(source.getType());
  assert((vectorType.getRank() == 1 &&
         extractType.getRank() == 1) &&
         "expected 1-D source and destination types");
```

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