GTC Silicon Valley-2019: OpenMP 5.0 for Accelerators and What Comes Next
Note: This video may require joining the NVIDIA Developer Program or login
GTC Silicon Valley-2019 ID:S9353:OpenMP 5.0 for Accelerators and What Comes Next
Bronis deSupinski(Lawrence Livermore National Laboratory),Tom Scogland(Lawrence Livermore National Laboratory)
With the recent release of OpenMP 5.0, OpenMP has reached a new milestone in support of performance-efficient programming of accelerators. New features better support underlying hardware and programmer intent to address distributed-memory devices and offloading to accelerators. As we prepare for the next generation of supercomputers and GPUs, OpenMP is growing to meet the challenges of productively programming scientific applications in a world of accelerators, unified memory, and explicitly hierarchical memories. We'll discuss the major additions to OpenMP 5.0 for accelerator programming, provide a view of our plans for the specification, and explain how our plans are both shaped by and shape large-scale scientific applications.