[Mlir-commits] [mlir] [Mlir] decompose generic by unfolding projected permutation crash fix (PR #122449)

[Adam Siemieniuk]

adam-smnk wrote:

> fail feels a bit strong though. While it is indeed pattern application "failure" (in most cases), some folks might think it's some more fundamental failure, no?

That's my impression too.

> If we manage to reduce the number of options that people use to those two, that would already be a success 😅 I am happy to write a PR that will update MLIR recommendation for testing if you also feel that this would be beneficial.

The proposal sounds reasonable. In the end, it's matter of consistency and convention given the usage is widespread enough.
I'm a bit skeptical whether there's really an efficient way to communicate and/or steer such design choices.
That being said, it doesn't hurt to try with a recommendation and a PSA. A concrete passage always helps in spreading the word around.

https://github.com/llvm/llvm-project/pull/122449