[Openmp-commits] [openmp] [Flang][OpenMP][MLIR] Initial array section mapping MLIR -> LLVM-IR lowering utilising omp.bounds (PR #68689)

[via Openmp-commits]

agozillon wrote:

Updated the PR to tidy up the test a little using the recommended optimisation passes and utilising the MapInfoTy from the OMPIRBuilder, thank you both for your suggestions. 

I will try to split the PR up on Monday or Tuesday next week, but feel free to further review this PR until they become available. 

The plan for splitting up the PR will likely be into 3 patches:
   1) NFC changes that are refactoring for the follow up additions 
   2) The addition of ByRef/ByCopy handling for values on the device/host border (marshalling data on host and accessing data 
        on the device) which begins to replicate Clang a bit more closely and allows a wider range of map behaviour
   3) The array sectioning specific lowering changes

There will likely be a dependency for these patches based on ordering e.g.  Patch 3 needs 2+1, 2 needs 1 and 1 will ideally be self contained. I am not sure based on the wording of the original request for splinting up comment @kiranchandramohan if you'd prefer if I just pushed the NFC changes directly without review (patch 1 in this case) or if you'd prefer it to go through review still? 

https://github.com/llvm/llvm-project/pull/68689