Supercomputing has always aimed at much higher performance than that of the most powerful computers of the day. Parallel processing is one of the key technologies for the realization of such computational performance, as is required by highly advanced scientific problems. In contrast to the conventional numerical approaches based on the von Neumann architecture, which do not critically depend on the characteristics of their target machines, the numerical algorithms for parallel computational environments are closely connected with machine architectures, and must be reconsidered according to the analysis of their application.
Currently, our main research emphasis is on parallel numerical algorithms and their application, the target of which is large-scale simulation in wide fields, ranging from particle physics to biological science. One of our foci is on convection diffusion problems, which possess essential behaviors involved in broad area of scientific application. Another urgent topic of importance is the establishment of parallelization techniques, such as parallel processing languages and mapping or scheduling algorithms, to meet the need of applicational problems. Highly motivated and widely interested students who want to tackle these challenging problems are expected.
Theoretical researches on numerical algorithms are conducted in parallel with the above practical researches. Themes range from the computational linear algebra and discretization methods to theoretical studies on the performance, precision, stability, and applicability of several numerical methods, such as iterative linear solvers, eigen solvers, and random number generators for Monte Carlo methods. Our special interests are on the analysis of preconditioned conjugate gradient type methods and Krylov projection methods for eigenproblems, which are widely noticed, and the study of pseudorandom numbers, which is essential in discrete algorithms and optimizations.
We will never put up with the second to support for the freedom of study area.
Last modified: Sat Feb 2 23:46:01 2002