- Prof. Mitsuhisa Sato, Leader of Programming Environment Research Team, RIKEN, Japan
Title: Challenges for Unified Parallel Programming Models for Accelerated Clusters
Abstract: For accelerated clusters with GPUs, programmers write programs for their applications by a combination of MPI and one of the available accelerator programming models. We have been developing a source-to-source Omni OpenACC compiler. XcalableACC (XACC) is an extension for a unified parallel programming for accelerated cluster by an “orthogonal” integration of a Partitioned Global Address Space (PGAS) language XcalableMP (XMP) and OpenACC. On the other hand, as OpenMP task parallel programming is getting attention, integration of task-parallelism and offloading to accelerators such as GPUs and FGPAs is an interesting topic for unified programming models. In this talk, our OpenACC and XACC project will be described as well as issues for task parallel programming for accelerated clusters.
Bio: Mitsuhisa Sato at the RIKEN Advanced Institute for Computational Science (AICS) he led the Programming Environment Research Team from 2010, and since 2014 has been appointed to deputy project leader for the Flagship 2020 Project at AICS (now R-CCS). He serves as Deputy Director of R-CCS since fiscal year 2018. He is also Professor (Cooperative Graduate School Program), Tsukuba University; Professor Emeritus, the University of Tsukuba; and Fellow of the Information Processing Society of Japan. His research interests include: parallel processing architecture; programming models, languages, and compilers; computer performance evaluation technology.
Prof. Simon McIntosh-Smith, Professor, University of Bristol
Title: Performance Portability across diverse computer architectures
Abstract: Achieving performance portability is challenging even with fairly simple codes and a small number of architectures. As the codes become more complex and the number and diversity of architectures increases, the performance portability challenge increases significantly. In this talk we’ll present some recent work performing one of the widest ever performance portability studies. In this work we have taken structured grid, unstructured grid and Monte Carlo codes from the particle transport domain and developed performance portable versions. These codes were then assessed across a wide range of CPUs and GPUs. The results should help advance our understanding of the fundamental issues in the performance portability space.
Invited Speakers (panel)
Will be updated shortly.