[Mlir-commits] [mlir] [mlir] Add bufferization option for parallel region check (PR #94645)

[Matthias Springer]

matthias-springer wrote:

Generally speaking, the simpler the analysis the better.

> Handling parallel region RaW conflicts should usually be the responsibility of the source program, rather than bufferization analysis.

I'm wondering how you're going to do that. You need to know how tensor ops are going to bufferize. (E.g., by querying the `BufferizableOpInterface`.) And you also need to track alias sets, which are only decided during bufferization. So I think you will basically have to reimplement the bufferization RaW analysis.

I'm skeptical whether this is useful in practice. But it could certainly be useful as a testing flag (to debug bugs in the analysis), so feel free to merge.


https://github.com/llvm/llvm-project/pull/94645