================
@@ -7558,6 +7628,18 @@ static void genMetadirective(lower::AbstractConverter
&converter,
}
}
+ if (!hasLoopAssociatedCandidate &&
+ llvm::any_of(reachableVariantSpecs, [](const auto *spec) {
+ return spec && hasDirectiveAssociation(spec->DirId(),
+ llvm::omp::Association::Block);
+ }))
+ associatedBlockEval = spliceAssociatedEval<parser::BlockConstruct>(
----------------
MattPD wrote:
An ignored compiler directive still prevents the following BLOCK from being
associated with the selected TARGET. The scan skips intervening evaluations
only when `isLoop` is true, so the BLOCK path returns null at the ignored
directive.
You can reproduce this with `flang -fopenmp -fopenmp-version=52 -S -emit-llvm
repro.f90 -o -`:
```fortran
subroutine selected()
!$omp metadirective when(implementation={vendor(llvm)}: target)
!dir$ ignored_comment
block
!$omp metadirective when(construct={target}: taskyield) otherwise(taskwait)
end block
end subroutine
```
At 9c59c04, Flang warns that the directive was ignored, emits an empty TARGET,
and places TASKWAIT outside it. Removing the ignored line or writing TARGET
directly places TASKYIELD inside TARGET. The [merge
base](https://github.com/llvm/llvm-project/commit/ae7f7ef515657eeb037e1cad045ade1a45cc7951)
also loses this association, so this is a remaining case in the BLOCK fix
rather than a new regression.
Could BLOCK association pass over this ignored directive without treating
compiler directives with actual effects the same way?
https://github.com/llvm/llvm-project/pull/224431
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
