Skip to main content
NASA Advanced Supercomputing (NAS) Division World-class modeling, simulation, and supercomputing in support of the NASA mission

Thinking OpenMP with NVIDIA HPC Compilers

AMS Seminar Series

Speaker: Guray Ozen, NVIDIA Corp

Presentation

Seminar Slide Deck (PDF-2.1MB)

Abstract

OpenMP is a directive-based programming model developed to solve productivity and performance as well as portability across target systems. In this talk, we will present NVIDIA's OpenMP target offload solution that provides performance portability across NVIDIA GPUs and multicore CPUs. We explain how to take advantage of our solution in C, C++, and Fortran using NVIDIA HPC compilers. We will also provide a list of concrete recommendations to programmers so that they, too, can obtain high performance on NVIDIA's GPUs. We will conclude our presentation by explaining interoperability among other models we implemented such as OpenACC, and standard language parallelism in C++ and Fortran.

Biography

Photo of Guray Ozen
Guray Ozen is a senior compiler software engineer in NVIDIA’s HPC compiler team. Before joining NVIDIA, he worked in the programming models team at Barcelona Supercomputing Center (BSC) while doing his Ph.D. in Universitat Politècnica de Catalunya – BarcelonaTech. His current focus is on how the C, C++, Fortran, OpenMP, and OpenACC programming models should evolve to take advantage of GPUs.