[Openmp-commits] [PATCH] D57615: [AST][OpenMP] OpenMP master Construct contains Structured block
Roman Lebedev via Phabricator via Openmp-commits
openmp-commits at lists.llvm.org
Fri Feb 1 13:38:08 PST 2019
lebedev.ri added a comment.
In D57615#1381428 <https://reviews.llvm.org/D57615#1381428>, @ABataev wrote:
> In D57615#1381427 <https://reviews.llvm.org/D57615#1381427>, @lebedev.ri wrote:
> > In D57615#1381418 <https://reviews.llvm.org/D57615#1381418>, @ABataev wrote:
> > > In D57615#1381416 <https://reviews.llvm.org/D57615#1381416>, @lebedev.ri wrote:
> > >
> > > > @ABataev i'm not sure i have fully followed the https://bugs.llvm.org/show_bug.cgi?id=40563#c1
> > > >
> > > > > The outlined function is not generated for the loop, so there is no problem with the standard compatibility.
> > > >
> > > > Are you saying that in these cases of `master`, `critical`, `single` directives, `CapturedDecl` should **not** have `nothrow` bit set too?
> > >
> > >
> > > This flag just does not matter for them.
> > I'll rephrase:
> > Are you opposed to providing the correct (as per the specification) knowledge that
> > no exception will escape out of these `CapturedDecl`'s, because that knowledge
> > does not matter for the existing sema/codegen, and does not affect produced IR?
> Again, CapturedDecl and CapturedStmt is not the representation of the structured block. It is a helper structure for the codegen only. This flag does not help with anything, because for such OpenMP constructs the codegen bypasses CapturedStmt/CapturedDecl.
Okay, enough is enough :)
I think a very important phrase was repeated a number of times, and it highlights the **actual** problem here:
> is not the representation of the structured block
> I'm not opposed. I'm saying, that it is not required and not used. If it is not used, why do you want to set it?
Then could you please revoke LG from D57594 <https://reviews.llvm.org/D57594>.
CHANGES SINCE LAST ACTION
More information about the Openmp-commits