Speaker:

Prof. Mitsuhisa Sato, Team Leader Exascale computing project, RIKEN AICS, Japan

Abstract:

Europe, the US, Japan and China are racing to develop the next generation of supercomputer – exascale machines – capable of a million trillion calculations a second by around 2020. To realize exascale systems, there are many challenges and issues including architectures and programming models to exploit billions of parallelism and the limitation of power consumption. Toward a post-petascale system as the successor to Japan’s petascale facility – the K computer – the FLAGSHIP 2020 project was launched last year to develop and deploy the post-K computer. While there are two major different approaches for exascale, manycore-based and accelerator-based, we take the manycore-based approach for our post-K computer. OpenMP is a key to make use of manycore efficiently, and is to evolve for exploiting large parallelism and integration with communication layers. In this talk, projects for post-petascale and exascale computing in Japan will be described, and challenges for parallel programming models and languages in these projects will be addressed.

Speaker:

Hitoshi Murai, RIKEN AICS, Japan

Abstract:

XcalableMP (XMP) is a directive-based PGAS language for distributed memory systems designed under PC cluster consortium Japan. XcalableMP supports two styles of parallelization: global- and local-view. The former is based on HPF-like data/work mapping directives, and the latter is a lower-level and more flexible one with coarrays. In this tutorial, I will explain using some examples how to write XMP programs in each of the two styles.

Date:

Tuesday, 1 December 2015, 14:00-16:30

Venue:

Jülich Supercomputing Centre, Rotunda, building 16.4, room 301

Announcement as pdf file: