[Mlir-commits] [mlir] [Linalg] Support i1 data type in matchConvolutionOpOfType utility (PR #176704)

[Abhishek Varma]

Abhishek-Varma wrote:

> Thanks!
> 
> Just double checking my understanding - this change applies to all conv Ops, not only to `linalg.conv_2d_nhwc_hwcf` and `linalg.pooling_nhwc_max` that are being tested here?

Yes, that's correct. This change applies to all conv ops. Eg: In the PR [Update Vectorization for named as well as generic LinalgOp](https://github.com/llvm/llvm-project/pull/176339) there's a `linalg.conv_1d_nwc_wcf` with `i1` input/output that's being tested and relies on this change.

https://github.com/llvm/llvm-project/pull/176704