Research and development on computafional chemistry technologies by using "multiple molecular objective"approach adapted for large scale distdbuted processing

点击翻译按钮获取中文摘要

This project consists of three major parts : (1) the construction of the molecular calculation platform for the multi-molecular objective method,(2) the development of molecular calculation techniques adopted for the high-accurate hierarchized molecule simulation,(3) the development of load distribution algorithm optimized for hetero-computational task resources. Outline of the each research result is shown in the following : (1) With the aim of the improvement in the parallelization efficiency and computation accuracy by the multi-molecular objective method, crystal simulation technique of the large scale system was developed. As a result of the "molecular objective" approach, we have reached to large-scale finite crystal calculation of 0.1 μm diameter composed of four hundred million atoms within accuracy guarantee to the 14 digits of total intermolecular interaction energy. The practical application of thermodynamic analysis between crystal polymorphisms and their phase transition a … More lso can available. (2) Solution methods for TDHF/TDDFT time dependence Schrodinger equation and hyperpolarizability by using the multi-wavelet basis set have been successfully developed. At this stage, these results are not so easy to apply into the large scale system, it can greatly expected anytime soon that extended to the development to density functional theory for the heterogeneous system. (3) In the hetero distributed computing environment, we have successfully developed a new fundamental algorithm for efficient load distribution, and the performance of HPL(High Performance Linpack)and various benchmark programs are better than those of the other established model. On the other hand, it has noted that the performance for calculation task of the molecular simulation is insufficient for the practical usage. Therefore, although the development of the practical load distribution algorithm has still faced with some difficult problems, we have started to study some efficient load distribution algorithms adapted to the recent multi-core/multithread system. Less