Hisashi Nakamura
Director, Computational Science and Engineering Division
Research Organization for Information Science and Technology
Optimization has been carried out for high-end simulation software, in order to promote effective use of petaflops-class computing capability in future and advanced computational science and engineering, as the related activities of the innovative high-end simulation software project. In this frame, it is planned that high-end simulation software is optimized through tuning by parallelization and vectorization for high-end computers, and run with a high sustained speed. In addition, the optimization leads to reduce time for solution of problems. These basic activities as the software technology will be conducted through the Earth Simulator that is well capable of high-end computing.
(1) Optimization of software for parallelization and vectorization:
For increasing a sustained running speed of simulation software, targeted simulation software is optimized to be more than 99% of parallelization and 95% of vectorization through restructuring the software within 10 nodes of the Erath Simulator.
(2) Optimization of algorithm for increasing a sustained speed:
Algorithms of a core part of targeted simulation software are evaluated and modified for more increase of its running speed through medium or large- scale simulations utilizing computing resource of up to several hundreds nodes of the Earth Simulator.
(3) Verification of Optimization:
Optimized simulation software through (1) and (2 ) is tested for the running speed-up and verified the simulation results by a typical problem in application fields as users are interested for their purposes.In this frame, a nano simulation code PHASE is now under optimization. This is the reason why large-scale simulation is required for predicting accurate atoms and molecules movements with increase of number of atoms, and thus simulation needs to reduce time for solution by increasing a running speed on high-end computers.
