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

[Jacob Yu]

pingshiyu wrote:

@joker-eph 
> I don't see any difference right now between the example at hand here for MLIR and most transformations in LLVM.

The difference is that for the majority of cases, there is only one unit test that can be written for the lowering pattern: the lowering pattern itself (i.e. the example shown). This is a problem for lowering passes only and not for other transformations.

Because of this, it results in insufficient testing for the lowering passes imo.

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