[PATCH] D18286: [OPENMP] private and firstprivate clauses of teams code generation for nvptx

Thu Mar 24 10:45:53 PDT 2016

Hahnfeld added a comment.

In http://reviews.llvm.org/D18286#382698, @carlo.bertolli wrote:

> Hi Jonas
>
> Thanks for your comment. That scheme would not work for #parallel on the gpu in case we do not want to inline the parallel region.
>  There are patches coming in which that is going to happen so I think that we really need to distinguish between different pragmas.

Is a `#pragma omp parallel` changing the way code is generated for `#pragma omp teams`? How would that look like in `CodeGenFunction::EmitOMPTeamsDirective`?

My point is that specialization should happen in `CGOpenMPRuntimeNVPTX` (by means of overriding) without duplicating code, as Alexey said.

Repository:
  rL LLVM

http://reviews.llvm.org/D18286