[PATCH] D99888: [flang][OpenMP] Add semantic checks for occurrence of nested Barrier regions

[Arnamoy B via Phabricator via llvm-commits]

arnamoy10 added inline comments.


================
Comment at: flang/lib/Semantics/check-omp-structure.cpp:462
+    if (CurrentDirectiveIsNested() &&
+        llvm::omp::nestedBarrierErrSet.test(GetContextParent().directive)) {
+      context_.Say(parser::FindSourceLocation(x),
----------------
kiranchandramohan wrote:
> arnamoy10 wrote:
> > kiranchandramohan wrote:
> > > Definition of closely nested region is 
> > > "A region nested inside another region with no parallel region nested between them."
> > > 
> > > Does this check capture this definition correctly?
> > Thanks for pointing in out.
> > 
> > 1.  Just double checking -- is //parallel region// a region enclosed by the `!$omp parallel` construct?
> > 2. IIUC, [[ https://github.com/llvm/llvm-project/blob/0116d04d04f20e9ae62ba847075840c3cb298080/flang/lib/Semantics/check-omp-structure.cpp#L130 | this ]] check (by which my code was inspired) also does not cover the definition of "close nesting"?  Please confirm.
> > 
> 1. Yes. But it is a dynamic concept, i.e it is the body of code executed during the execution.
> 2. Yes that code also does not cover the definition. Will appreciate a fix.
I need one more clarification before implementing the fix.

>From the definition:

```A region nested inside another region with no parallel region nested between them```

What does nesting "between" mean?  I can think of two cases.

Case 1:

```
!$omp ordered
  !$omp parallel <-- Child of ordered
    !$omp barrier <-- Child of parallel, grandchild of ordered
```

Case 2:

```
!$omp ordered
  !$omp parallel <-- Child of ordered
  !$omp barrier <-- Child of ordered
```

Can you comment on which one of the above two cases has a `barrier` closely nested inside the `ordered` construct?



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99888/new/

https://reviews.llvm.org/D99888