[Openmp-commits] [PATCH] D68100: [OpenMP 5.0] declare mapper runtime implementation

[Lingda Li via Phabricator via Openmp-commits]

lildmh added a comment.

In D68100#2173949 <https://reviews.llvm.org/D68100#2173949>, @protze.joachim wrote:

> From my perspective, the declare_mapper_target.cpp code is semantically equivalent to:
>
>   #pragma omp target data map(tofrom: c)
>   #pragma omp target data map(tofrom: c.a[0:NUM])
>   #pragma omp target teams distribute parallel for
>   for (int i = 0; i < NUM; i++) {
>     ++c.a[i];
>   }
>   
>
> and
>
>   #pragma omp target enter data map(to: c)
>   #pragma omp target enter data map(to: c.a[0:NUM])
>  
>   #pragma omp target teams distribute parallel for
>   for (int i = 0; i < NUM; i++) {
>     ++c.a[i];
>   }
>   #pragma omp target exit data map(from: c.a[0:NUM])
>   #pragma omp target exit data map(from: c)
>   
>
> Can you express the behavior of your mapping implementation in terms of OpenMP target enter/exit data primitives?


You are basically right. In implementation, a function is generated for every mapper to do all internal mapping. More details can be found at https://github.com/lingda-li/public-sharing/blob/master/mapper_runtime_design.pptx


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68100/new/

https://reviews.llvm.org/D68100