We research in pursuit for the form of next supercomputers to come, and future methods in scientific calculation to use them, while simultaneously providing supercomputing services, as the core computing platform for universities and research facilities in Japan.
The Foundation for Grand Challenge projects
“Supercomputers”, which have processing power ranging from several hundred times to several hundred thousand times of that compared to conventional PCs are being used to solve a large variety of problems requiring huge amounts of numerical computations, such as structural tolerance engineering in architecture, air resistance coefficient and drag simulation in aviation, modelling proteins and medicine interaction at molecular levels in biochemistry. We can even examine theories to the origin of our universe, by mapping real observatory results with simulations of galaxies consisting of several hundred billions of stars, colliding into each other. Supercomputers have become experimentation fields to perform experiments impossible in our real world, and “computational simulation” has become the third category of science following “theory” and “experiments”.
We extend our research into taking supercomputers to the next level, with designs based on repeated verifications of fundamental technology, and studying numerical processing program development techniques, along with technology to maintain and operate the complex supercomputer architectures.
The University of Tokyo Information Technology Center takes on the core role of Japan’s academic research computing infrastructure, and the supercomputers here provide service to over 1,500 users from numerous universities and research facilities. The current supercomputers are configured from hundreds to thousands of computers connected together, which require specialized programming knowledge to produce their full potential. We take part in projects that make such skills widely available, too.
Application Development Environment ppOpen-HPC
This is an open sourced application development platform environment, with numerical processing libraries and auto tuning feature sets, which allow programmers to easily benefit from the power of large scale supercomputers.
Next Generation Supercomputers
This topic involves verification of elemental technologies to investigate feature, performance, and feasibility for the next and succeeding generations of supercomputers, Experimental deployments of system software optimized to many-core architecture processors targeting Exa-FLOPS class performance, (100 times the performance of the “K-computer”) is one example of such research.