[Mlir-commits] [mlir] [mlir][linalg] Implement Conv2D using Winograd Conv2D algorithm (PR #96181)

[Oleksandr Alex Zinenko]

================
@@ -2734,6 +2734,84 @@ FailureOr<SmallVector<Value>> SoftmaxOp::decomposeOperation(OpBuilder &b) {
   return SmallVector<Value>{result};
 }
 
+//===----------------------------------------------------------------------===//
+// WinogradFilterTransformOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult WinogradFilterTransformOp::verify() {
+  auto filterType = cast<ShapedType>(getFilter().getType());
+  auto outputType = cast<ShapedType>(getOutput().getType());
+  auto filterElemType = filterType.getElementType();
+  auto outputElemType = outputType.getElementType();
+  if (filterElemType != outputElemType) {
+    return emitOpError() << "expected element type of input " << filterElemType
+                         << " to match element type of output "
+                         << outputElemType;
+  }
+
+  unsigned filterRank = filterType.getRank();
+  if (filterRank != 4)
+    return emitOpError() << "expected rank of input is 4";
----------------
ftynse wrote:

This can be encoded in ODS by using `TensorRankOf<[AnyType], [4]>` instead of `AnyRankedTensor` as type constraint.

https://github.com/llvm/llvm-project/pull/96181