Research On Parallel Processing Of The Charge Distribution In Biomacromolecules By ABEEMσπ Model

In recent years, with the rapid development of computer technology, molecular dynamic simulations of biomacromolecules has become a very important theoretical research tool used to study a large number of biological molecules and their complex, dynamic and thermodynamic process. For large molecules the charge distribution has a great effect on the results of simulations, so it is very important to develop a method of calculating the charge distribution which has a smaller amount of calculation and higher accuracy. The atom-bond electronegativity equalization methodσπmodel (ABEEMσπ) which is a method of calculating the charge distribution in large molecules rapidly and accurately has the great advantages of being applied to large molecular systems, and has aroused a certain amount of concern and accepted by all in recent years. However, with the increasing of the size of molecular system, the time of calculating the charge distribution by the serial program of ABEEMσπmodel gets longer and longer. In order to expand the molecular system and improve the computing speed by ABEEMσπmodel, this paper studies parallel processing of calculating the charge distribution by ABEEMσπmodel, and the main contents include:1. The serial process of calculating the charge distribution by ABEEMσπmodel is analyzed, and an optimized program is given which reduces redundant codes; the use of dynamic storage technology is used to reduce the life cycle of the overall arrays and local arrays and saves memory space.2. Based on the optimized serial procedure of calculating the charge distribution by ABEEMσπa parallel project of calculating the charge distribution by ABEEMσπis designed and implemented successfully, and the scale of molecular system which can be calculated by the serial program is expanded greatly.3. According to the characteristics of calculation of charge distribution by ABEEMσπ, two parallel task partition strategies are proposed and compared, one of which is multiple rows wrapped interleaved division strategy that assigns multi-lines matrix to each processor, reduces the number of communication, reduces communication overhead, and gets a higher parallel speedup in computing the charge distribution in large scale of molecular system.4. The communication of Cholesky decomposition parallel algorithm of the parallel program of the calculation of charge distribution by ABEEMσπis optimized, and a method of overall communication is used to reduce the communication time of application-layer and improves further the parallel performance of the optimized parallel program of calculating the charge distribution by ABEEMσπmodel.5. The parallel computing of the charge distribution in biomacromolecules is realized, which has solved the problem of computing the actual charge distribution of the sites of more than 50,000 and has accessed to a high parallel speedup and efficiency.