[Mlir-commits] [mlir] [mlir][linalg] Enable masked vectorisation for depthwise convolutions (PR #81625)

[Andrzej Warzyński]

banach-space wrote:

Thanks all for reviewing, I will be merging this shortly (it's quite critical for us - we can't leverage wider SVE vectors otherwise).

> (...)  we know that this is a short-term specialization and that all the convolution support in the vectorizer needs to be revisited.

I would be very keen to participate in any such refactor.  Lets continue brainstorming about this!

>  As such, it may happen that some of this functionality has to be disabled/revisited in the future if it hinders a more generic path forward.

I am more than happy to redesign/refactor this if need be 👍🏻 

https://github.com/llvm/llvm-project/pull/81625