Mastering GPU-Acceleration on OpenPOWER Platform for Optimal Application Performance @ MdlS / Idris
Date: 2 - 4 November 2016
Presentation of the new Open Power - Nvidia prototype recently installed at Idris (see http://www.idris.fr/ouessant/ for details on the installed machine)
The training will introduce propspective users to this innovative architecture. It will present the programming models and tools available, as well as hilight the best practices, so as to obtain optimal performance when porting applications.
Preliminary program
Welcome
OpenPOWER IBM P8+ architecture
New Nvidia GPU Tesla P100 architecture
Programing models & Software stack
Scientific libraries and "GPU-Aware" runtimes
OpenACC on Ouessant2 [+ Hands-On]
OpenMP on Ouessant2 [+ Hands-On]
Conclusion
Event types:
- Workshops and courses
Activity log