[Mlir-commits] [mlir] [Linalg] Update Vectorization to work with both named as well as generic conv ops (PR #176339)

[Han-Chung Wang]

================
@@ -2431,10 +2432,11 @@ static LogicalResult vectorizeLinalgOpPrecondition(
   if (isElementwise(linalgOp))
     return success();
 
-  // TODO: isaConvolutionOpInterface that can also infer from generic
-  // features. But we will still need stride/dilation attributes that will be
-  // annoying to reverse-engineer...
-  if (isa<ConvolutionOpInterface>(linalgOp.getOperation()))
+  // Check for convolution ops - both named ops implementing
+  // ConvolutionOpInterface and generic ops that semantically match convolution
+  // patterns.
+  if (isa<ConvolutionOpInterface>(linalgOp.getOperation()) ||
----------------
hanhanW wrote:

Do we still need `isa<ConvolutionOpInterface>`? I thought the other one is more robust than `isa<>` method.

https://github.com/llvm/llvm-project/pull/176339