[PATCH] D102107: [OpenMP] Codegen aggregate for outlined function captures

Giorgis Georgakoudis via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon May 24 08:44:59 PDT 2021


ggeorgakoudis added a comment.

In D102107#2776237 <https://reviews.llvm.org/D102107#2776237>, @jdoerfert wrote:

> This allows us to remove the switch in the device runtime, right?

Yes, with a complication: for combined directives of worksharing loops (distributed parallel for) clang emits the lower bound and upper bound as distinct parameters besides the global tid, bound tid and the captured variables (aggregated in the struct of this patch). We will need to have a flag in parallel_51 to unwrap the global args array for this case.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102107/new/

https://reviews.llvm.org/D102107



More information about the cfe-commits mailing list