[Mlir-commits] [mlir] [mlir][arith] Add more canonicalization and integration tests coverage (PR #92272)

[Jacob Yu]

pingshiyu wrote:

@banach-space @kuhar Sorry for dropping the ball on this one last few weeks! I have been dealing with some personal issues, followed by a very busy period at work, but now I've found time to come back to this and get it merged! :)

Since I dropped the ball for so long I cannot ask for urgency :) Please review when convenient for you!

> Review process - multiple PRs
I think multiple PRs for this is a good idea since there are quite a few tests. I'll be splitting this one up into smaller ones which will soon follow.

> Why not split them along these axis (i.e. check 1., 2. and 3.) and either:
    write a function for every op and pass different arguments depending on the case being checked (e.g. "sane" vs "overflow" vs "poison"), or
    write a separate function for 1., 2. and 3.

I don't understand what this means fully - would you be able to give an example to clarify a bit further? I understand this suggestion as a refactoring of test code, although I don't see a common pattern that would work for each class of tests.

https://github.com/llvm/llvm-project/pull/92272