Computer and Mathematical Sciences
High Performance Computing A15
KeywordsHigh Performance Computing, System Software, Performance-aware Programming, Machine Learning
Creating future supercomputing technologies
Our research focus is on establishing high-performance computing technologies to fully exploit future supercomputers. In pursuit of this goal, we operate a production supercomputer and collaborate with many researchers for identifying shortcomings in current supercomputers and researching solutions to be implemented in next-generation technologies.
(1) System Software for Productive High-Performance Computing
To achieve both performance and portability, we develop system software technologies. One example of our research is to adopt data-driven approaches such as machine learning to enhance programming and runtime environments for future supercomputers.
(2) New Ways of Using Supercomputers
Recently, supercomputers are being used for an increasing variety of fields, including machine learning, big data, and urgent computing. We are engaged in research on developing effective ways of using supercomputers for those new purposes without degrading the performance for the original purpose of academic use.
(3) Operation Technologies for Future Supercomputers
We develop operation technologies to enhance fault tolerance and power efficiency while achieving high utilization and throughput in future supercomputers, which will not only be larger in scale, but also be more complicated.
-
Making full use of supercomputers!
-
AI for HPC & HPC for AI