[Openmp-commits] [PATCH] D108528: [OpenMP][Offloading] Add support for event related interfaces

Jon Chesterfield via Phabricator via Openmp-commits openmp-commits at lists.llvm.org
Mon Aug 23 15:00:02 PDT 2021

JonChesterfield added inline comments.

Comment at: openmp/libomptarget/include/omptargetplugin.h:157
+// to be non-blocking, just set dependence and return.
+// 4) Call __tgt_rtl_sync_event to sync the event. It is expected to block the
+// thread calling the function. 5) Destroy the event (__tgt_rtl_destroy_event).
Is this a means of coordinating across different async objects? Across different targets, or just between a cpu thread and one target? Trying to guess what the sequence of record / wait / sync might map onto.

In particular it's difficult to map this onto the HSA operations without knowing the cuda terminology. This is presumably a way of making some kernels wait for others to complete before they start executing but I'd expect that to be a single API call, probably named 'barrier'.

  rG LLVM Github Monorepo



More information about the Openmp-commits mailing list