[Openmp-commits] [PATCH] D104418: [OpenMP][Offloading] Fixed data race in libomptarget caused by async data movement

[Ye Luo via Phabricator via Openmp-commits]

ye-luo added inline comments.


================
Comment at: openmp/libomptarget/src/device.cpp:279
+
+    void *Event = Entry->getEvent();
+    bool NeedNewEvent = Event == nullptr;
----------------
One more question. Should the event only protect IsNew or both IsNew and HasFlagAlways? I feel in the case of HasFlagAlways not IsNew, it is user's responsible to handle the ordering of multiple transfers.


================
Comment at: openmp/libomptarget/src/device.cpp:311
+      if (Event) {
+        int Ret = waitEvent(Event, AsyncInfo);
+        Entry->unlock();
----------------
ye-luo wrote:
> ye-luo wrote:
> > After the first transfer, every look up calls waitEvent which is still costly.
> > It is better to return the even back to the pool and avoid waitEvent cost.
> I misunderstood waitEvent. It just enqueues the event to the stream. Need to find a better place to reduce the necessary waitEvent.
I think waitEvent can be bypassed if the present modifier is true.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D104418/new/

https://reviews.llvm.org/D104418