[Mlir-commits] [mlir] [MLIR] Allowing unsupported conv2d op to fail gracefully from vectorization (PR #130181)

[Diego Caballero]

================
@@ -1990,8 +1996,21 @@ static LogicalResult vectorizeLinalgOpPrecondition(
   // 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()))
+  if (isa<ConvolutionOpInterface>(linalgOp.getOperation())) {
+    // Create a dummy rewriter first, a rewriter is not required for
+    // validation
+    IRRewriter dummyBuilder(linalgOp.getContext());
----------------
dcaballe wrote:

It's a bit weird that we need an IRRewriter to check the vectorization preconditions of a convolution. We may want to decouple the preconditions from the vectorization itself to the extent possible.

This function is also growing too much so it'd be good to create a dedicated `vectorizeConvolutionPrecondition`.

https://github.com/llvm/llvm-project/pull/130181