[flang-commits] [flang] [flang][CUDA] Preserve data attrs on assignment RHS (PR #207849)

[via flang-commits]

https://github.com/jeanPerier commented:

Makes some sense to me, I just want to check to things:

1. Is it also OK/the expected behavior for regular calls with parenthesis in the source: i.e. in `call foo((x))`, is it always OK to propagate x cuda attribute to the actual arguments (maybe it is not allowed to write such code anyway)?

2. Can you add some lowering test, both in cases where we end-up making a temp and in cases where no temp is created, I want to make sure the temp creation for the device RHS is handled correctly on the host.

https://github.com/llvm/llvm-project/pull/207849