[Mlir-commits] [mlir] [mlir][vector] Support multi-dimensional vectors in VectorFromElementsLowering (PR #151175)

[Yang Bai]

================
@@ -238,6 +239,44 @@ Value createReadOrMaskedRead(OpBuilder &builder, Location loc, Value source,
 ///      static sizes in `shape`.
 LogicalResult isValidMaskedInputVector(ArrayRef<int64_t> shape,
                                        ArrayRef<int64_t> inputVectorSizes);
+
+/// Generic utility for unrolling n-D vector operations to (n-1)-D operations.
+/// This handles the common pattern of:
+/// 1. Check if already 1-D. If so, return failure.
+/// 2. Check for scalable dimensions. If so, return failure.
+/// 3. Create poison initialized result.
+/// 4. Loop through the outermost dimension, execute the UnrollVectorOpFn to
+/// create sub vectors.
+/// 5. Insert the sub vectors back into the final vector.
+/// 6. Replace the original op with the new result.
+using UnrollVectorOpFn =
+    function_ref<Value(PatternRewriter &, Location, VectorType, int64_t)>;
+
+template <typename VectorOpType>
+LogicalResult unrollVectorOp(VectorOpType op, PatternRewriter &rewriter,
----------------
yangtetris wrote:

That makes sense. Let me update it.

https://github.com/llvm/llvm-project/pull/151175