D7.5: HPC Programming Techniques
Cevdet Aykanat, Antun Balaž, Iris Christadler, Ivan Girotto, José Gracia, Vladimir Slavnić, Andy Sunderland, Ata Türk · Zenodo (CERN European Organization for Nuclear Research) · 2012
Work Package 7 “Enabling Petascale Applications: Efficient Use of Tier-0 Systems” (WP7) ensures the effective exploitation of the PRACE Tier-0 systems by increasing scalability and performance of applications. Codes are either successful applicants for the preparatory access Type C calls (Task 7.1 “Applications Enabling for Capability Science”) or they are part of an established collaboration between PRACE and application communities (Task 7.2 “Applications Enabling with Communities”). While the focus in WP7 is on enabling applications for Tier-0 systems, the tasks should also benefit application performance on Tier-1 systems. Task 7.5 is called “Programming Techniques for High Performance Applications”. This task worked with users to implement new programming techniques, paradigms and algorithms for Tier-1 and Tier-0 systems, which have the potential to facilitate significant improvements in their applications performance. The task worked in close collaboration with Task 7.1 and Task 7.2 to ensure that the research communities benefit. Task 7.5 worked on five different areas to increase performance and scalability of user codes: Scalable algorithms Scalable libraries Multi-core/many core systems Accelerators Novel HPC languages Overall, more than 30 projects were carried out by 25 different partners and third parties. They ranged from the introduction of new algorithms for sparse matrix operations to the assessment of new languages like StarSs, Chapel, Cilk and ArBB; from the comparison of mathematical libraries to the hybridization of important user codes to test the mixed OpenMP and MPI programming model. Several projects were dedicated to porting applications to GPUs, one outcome of this activity is a freely available Quantum Espresso CUDA port. Task 7.5 covered a plethora of different approaches and codes. The following deliverable is a summary of all projects performed within Task 7.5. It consists of chapters for each of the five topics, containing high-level summaries of all projects. In most of the summaries, links to PRACE white papers or scientific publications are given for those readers that are interested in more detailed information. Each chapter contains a short introduction highlighting the different projects and a summary with conclusions and outlook.