Introduction to OpenMP offload

[lucaparisi91]

Base material at OpenMPTargetOffload.ppt and IHPCSS___Introduction_to_OpenMP___GPUs.pdf

• OpenMP programming model ( generic model [ fork-join model ] vs CUDA model where threads are always active ) . See #20 .
• shared , firstprivate, lastprivate clause
• #pragma omp target : run on the gpu
• #pragma omp teams : spread across blocks , num_teams clause , num_threads clause
• # pragma omp parallel for
• device clause : selection for the device
• #pragma omp declare: run an external function
• collapse : for nested parallelism