[Mlir-commits] [flang] [llvm] [mlir] [flang] run CFG conversion on omp reduction declare ops (PR #84953)

[Valentin Clement バレンタイン クレメン]

clementval wrote:

> > > > Wouldn't applying the patterns on the module in a single pass work here as well?
> > > 
> > > 
> > > Yes that would work but we would loose parallelism. I wanted to keep that because there was talk in the past about breaking up the existing Module passes so that different functions (or other container operations) can be processed in parallel
> > 
> > 
> > Ok that makes sense. It would be nice to have a solution to for top level operations since we have more than two. Have you discuss with MLIR folks yet?
> 
> I have created an RFC here https://discourse.llvm.org/t/rfc-add-an-interface-for-top-level-container-operations/77807
> 
> Would it be okay to merge this pass as it is for now (there is already precedent in how fir.global is handled) and then fix everything together once a solution has been agreed upon?

Thanks for the RFC. I'm ok if you merge this for the time being.

https://github.com/llvm/llvm-project/pull/84953