Meinersbur wrote:

> I have a newbie (not familiar with the semantics of the `cancel` construct) 
> question specially given the changes in this PR, what is the point of 
> emitting the `__kmpc_cancel_barrier` specially that we now emit it at the 
> very end of the parallel region?

If I recall from #130135 correctly, a `parallel` region has an implicit barrier 
at that end which is not handled any different than a `#pragma omp barrier`. 
That is, it must skip the remaining code in the region (of which there is none 
for the implicit end barrier). but also reset the "has this region been 
cancelled" flag so the next execution of the region does not immediately cancel 
again.

https://github.com/llvm/llvm-project/pull/164586
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to