• The software is commonly valid for various discretized numerical schemes
• Several modification methods have been prepared to meet user's needs
• An element-based data structure has been selected as the user's interface in matrix solvers
• Modification for parallel computation can be done simply by a copy & paste process
• Several kinds of state-of-the-art matrix solvers are available.
• The software is compatible with standard tools (MPI, METIS)

The Research Institute for Computational Science's Continuum Modeling Research Group at the National Institute of Advanced Industrial Science and Technology (AIST), an independent administrative institution, and the Center for Computational Science and Engineering at FUJI Research Institute Corporation (FUJI RIC) have developed a parallel software general platform for various numerical schemes such as the finite element method, the finite difference method, and the finite volume method, as well as other numerical schemes, to assist a smooth shift to the world of parallel computation for non-specialists, which only requires ordinary input data and the subroutine to construct the user's stiffness matrix. The source code and manual have been released* to general users free of charge.

A major characteristic of this software platform is that it greatly assists users to parallelize their current numerical code in non-parallel computers with a high degree of parallel efficiency, regardless of the type of discretized numerical analysis technique used, and without requiring advanced knowledge or experience in parallel computation.

In the near future, eigenvalue analysis matrix solvers and other functions will be added to meet further requirements of users. More demo programs are supposed to be added to this platform, hopefully partly by user's collaborations.

The software was developed jointly by AIST and FUJI RIC for the New Energy and Industrial Technology Development Organization (NEDO) research and development project "Fundamental Research and Development for Research Information: Development of Parallel Computation Software Technology for Discretized Numerical Schemes" (1999), and for the "Development of a Common Platform on Parallel Computations for Discretized Numerical Schemes" research project at the AIST's Tsukuba Advanced Computing Center (TACC).

* Planned for general release after presentation:

Time/Date: 9:30 a.m., March 4, 2002 (Sunday)
Location: AIST Tsukuba Central #2 OSL

With this platform, users can enjoy large-scale, high-speed, parallel numerical computations in a few days, which may greatly activate computer-aided engineering (CAE) in industry. Because of severe competition in the worldwide business, most companies are trying to cut costs in the design section, by introducing commercial software developed in the United States and Europe instead of using original in-house numerical analysis software. However, "black box" use of commercial software results in the "death" of advanced technology in numerical simulation in this country, and too much dependency on imported software leads to lack of initiative in advanced fields and produces serious damage to industrial technology in Japan. By installing this platform on parallel computers, in-house numerical software can be easily reinforced by parallel computations, and engineers can enjoy the advantages of vastly greater speed and scale in a short developing time.

Since the prices of PC clusters, whose performance is almost equivalent to some supercomputers, have drastically decreased in recent years, even small-size companies can purchase them. This platform now provides the same CAE technology, previously available only in big businesses, to small and medium-sized enterprises, which might significantly contribute to reinforce the competitiveness of Japan's manufacturing industries.