[Mlir-commits] [mlir] [MLIR][OpenMP] Normalize lowering of omp.loop_nest (PR #127217)

[Sergio Afonso]

https://github.com/skatrak commented:

> I still se a lot of code moved into `convertOmpSimd`? Why is that?

I think the diff is messing up the view and making it more confusing. What was basically done was moving the `omp.loop_nest` lowering, which was duplicated in `convertOmpWsloop` and `convertOmpSimd`, into a new `convertOmpLoopNest` function. Everything that was wrapper-specific remained in the respective wrapper translation function, and the only thing that got moved from `convertOmpWsloop` into `convertOmpSimd` was the composite check, which results in SIMD information be ignored (same behavior as before, but checked where I think makes more sense).

https://github.com/llvm/llvm-project/pull/127217