ABataev added inline comments.
================
Comment at: clang/lib/Sema/SemaOpenMP.cpp:8653
? (DSAStack->hasMutipleLoops() ? OMPC_lastprivate :
OMPC_linear)
- : OMPC_private;
- if (((isOpenMPSimdDirective(DKind) && DVar.CKind != OMPC_unknown &&
- DVar.CKind != PredeterminedCKind && DVar.RefExpr &&
- (LangOpts.OpenMP <= 45 || (DVar.CKind != OMPC_lastprivate &&
- DVar.CKind != OMPC_private))) ||
- ((isOpenMPWorksharingDirective(DKind) || DKind == OMPD_taskloop ||
- DKind == OMPD_master_taskloop ||
- DKind == OMPD_parallel_master_taskloop ||
- isOpenMPDistributeDirective(DKind)) &&
- !isOpenMPSimdDirective(DKind) && DVar.CKind != OMPC_unknown &&
- DVar.CKind != OMPC_private && DVar.CKind != OMPC_lastprivate)) &&
- (DVar.CKind != OMPC_private || DVar.RefExpr)) {
+ : DVar.CKind == OMPC_lastprivate ? OMPC_lastprivate
+ : OMPC_private;
----------------
jyu2 wrote:
> ABataev wrote:
> > Why need this change?
> The rule:
> The loop iteration variable in any associated loop of a loop construct is
> lastprivate.
>
> So if there is lastprivate clause, should be predetermined as lastprivate.
The patch is for threadprivates, better to do it in a separate patch(or
patches).
================
Comment at: clang/lib/Sema/SemaOpenMP.cpp:8662-8667
+ if (LangOpts.OpenMP > 45 && isOpenMPSimdDirective(DKind)) {
+ if (DVar.CKind == OMPC_lastprivate)
+ PredeterminedCKind = OMPC_lastprivate;
+ else if (DVar.CKind == OMPC_private)
+ PredeterminedCKind = OMPC_private;
+ }
----------------
jyu2 wrote:
> ABataev wrote:
> > Again, why do we need this?
> In the comment: for OpenMP 5.0 and up:
>
> If a simd construct has just one associated loop, the variable may be listed
> in a private, lastprivate or linear.
>
> The linear part is set on line 8652, there just need to set private and
> lastprivate.
>
>
Same here, better to split this change
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D112292/new/
https://reviews.llvm.org/D112292
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
