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

[Zhuoran Yin]

================
@@ -112,6 +112,25 @@ module attributes {transform.with_named_sequence} {
 
 // -----
 
+func.func @conv2d(%3: tensor<1x64x58x58xf32>, %4:  tensor<64x64x3x3xf32>) {
----------------
jerryyin wrote:

> none of the 2d and 3d convs would vectorize with this.

I can add an additional conv3d test

> [nit] Big input tensors are IMHO a distraction

Agreed, I copy-pasted the actual crashing op without looking thoroughly in the context of this test. Will align the sizes like other tests in this file.

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