[Openmp-commits] [PATCH] D120671: [OpenMP][libomp] omp_in_explicit_task() implementation.

[Andrey Churbanov via Phabricator via Openmp-commits]

AndreyChurbanov added inline comments.


================
Comment at: openmp/runtime/src/kmp_ftn_entry.h:1574
+#else
+  int gtid = __kmp_entry_gtid();
+  return __kmp_thread_from_gtid(gtid)->th.th_current_task->td_flags.tasktype;
----------------
protze.joachim wrote:
> Is this code sufficient for all scenarios of lazily initialized OpenMP threads?
> Should this function handle uninitialized 
The `__kmp_entry_gtid()` initializes the library if needed and initializes the calling thread including its implicit task.
Thus the structures cannot be uninitialized here.


================
Comment at: openmp/runtime/test/api/omp_in_explicit_task.c:32
+    }
+    #pragma omp task
+    {
----------------
protze.joachim wrote:
> Would it make sense to test different kinds of tasks, like undeferred (and detachable) tasks, instead of two simple tasks?
> At the moment it should not make a difference, but things might change in the future.
I doubt an explicit task can become implicit in future.  That would be very strange. And we can always adjust the test in future, or write a new one for some particular case.




Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D120671/new/

https://reviews.llvm.org/D120671