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

[Andrzej Warzyński]

================
@@ -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());
+    // Check if we can successfully construct a 1d convolution generator.
+    // For example, if it is 2d+ convolution, return failure because we don't
+    // support it. To use this pass on a 2d+ convolution, it should have already
+    // been decomposed to 1d convolution via
+    // DecomposeConvolutionToLowerDimOpsPass.
----------------
banach-space wrote:

I couldn't find such Pass in-tree.

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