Advanced OpenMP – tasks and GPU offloading

OpenMP is a directive-based programming model that can target both multicore CPUs and GPUs. This course extends the basic work-sharing constructs to task-based parallelism in CPUs and to GPU usage via offloading. The course starts with a brief review of the basics of OpenMP and continues then to the more advanced themes. The topics include task construct, GPU offload directives and optimizing data movement between CPU and GPU. Participants are expected to have a working knowledge of basic OpenMP, but no prior experience with GPUs is needed. The course consists of lectures and hands-on exercises.

Learning outcome

– Awareness of task based parallelism
– Ability to implement task parallel algorithms with OpenMP
– Ability to utilize GPUs efficiently with OpenMP offloading

Prerequisites

– Good knowledge of C/C++ or Fortran
– Basic knowledge of OpenMP for CPUs

The course is an onsite training held at the CSC – IT Centre for Science training facilities at Life Science Center, Keilaranta 14, Espoo, Finland.

Register for the course. Deadline for registrations is 4.10.2024