[Mlir-commits] [mlir] [Linalg] Rename `isaConvolutionOpOfType` to `matchConvolutionOpOfType` along with few more updates (PR #174722)

[llvmlistbot at llvm.org]

MaheshRavishankar wrote:

I might be missing the context here, but why are we trying to return the dilation/strides in the matcher itself? I'd separate the two (even at the cost of some redundancy). The matcher still returns true/false based on whether it thinks its a convolution or not. A separate utility method that returns the strides/dilations (it can assert that the op is a convolution) would make more sense to me and less disruptive.

https://github.com/llvm/llvm-project/pull/174722