HPC software is an essential part of the infrastructure for science, but in order to push science, it must efficiently use state-of-the-art hardware and has to be continually brought forward as the hardware landscape changes. Users often lack the time and skill to improve their codes, computing centers lack the personnel, and improvements through tools (e.g., autoparallelization) work, without accompanying code redesign, only in rare circumstances. To help users in this situation, this project pursues a two-pronged approach.
This project is part of the DFG Program “Performance Engineering for Scientific Software”. The first objective of this project is to push the partners’ expertise on GPU performance engineering, performance modeling for scalability analysis, algorithmic stability, performance, and reproducibility into performance engineering practices, thus empowering HPC users in areas critical for performance engineering.
The second objective of this project is to bundle the distributed expertise for HPC support and performance engineering within the Hessian HPC-Competence Center (HKHLR) and the Alliance for HPC in Rhineland-Palatinate (AHRP) under a new umbrella organization to support scientists to efficiently use the HPC resources in Hesse and Rhineland-Palatinate.
- Technische Universität Darmstadt (Coordinator)
- Johannes Gutenberg-Universität Mainz
- Technische Universität Kaiserslauten
- Goethe-Universität Frankfurt
01/2017 -- 12/2019