This title appears in the Scientific Report :
2019
Application Porting and Optimization on GPU-Accelerated POWER Architectures
Application Porting and Optimization on GPU-Accelerated POWER Architectures
The POWER processor has re-emerged as a technology for supercomputer architectures. One major reason is the tight integration of processor and GPU accelerator through the NVLink technology. Two major sites in the US, ORNL and LLNL, deployed their pre-exascale systems based on this new architecture (...
Saved in:
Personal Name(s): | Herten, Andreas (Corresponding author) |
---|---|
Pleiter, Dirk / Wagner, Mathias / Hagleitner, Christoph / Ravindar, Archana / Papatheodore, Tom | |
Contributing Institute: |
Jülich Supercomputing Center; JSC |
Imprint: |
2019
|
Conference: | International Conference for High Performance Computing, Networking, Storage and Analysis (The Supercomputing Conference), Denver, CO (USA), 2019-11-18 - 2019-11-18 |
Document Type: |
Lecture |
Research Program: |
Supercomputer Facility Computational Science and Mathematical Methods |
Publikationsportal JuSER |
The POWER processor has re-emerged as a technology for supercomputer architectures. One major reason is the tight integration of processor and GPU accelerator through the NVLink technology. Two major sites in the US, ORNL and LLNL, deployed their pre-exascale systems based on this new architecture (Summit and Sierra, respectively).This tutorial gives an opportunity to obtain in-depth knowledge and experience with GPU-accelerated POWER nodes. It focuses on porting applications to a single node and covers the topics architecture, compilers, performance analysis and tuning, and multi-GPU programming. The tutorial includes an overview of the NVLink-based node architectures, lectures on first-hand experience in porting to this architecture, and exercises using tools to focus on performance. |