Parallel Computing In Molecular Simulation

With the development of computer science and technology, computer simulation has been becoming a powerful tool in studying both structures and properties of real fluids. It has also been drawn great attention in chemical engineering field. Compared with the traditional method, computer simulation has its special advantages. By performing computer simulation on model fluids, microscopic properties can be revealed and many physical phenomena can also be predicted from the interaction between molecular.But the traditional computer simulation do not meet the need of high-speed computing. So, the parallel computing is imperative under the situation. In this work, I study the parallel computing in molecular simulation.First, the cluster of PC and its development environment-Linux OS and its platform MPI were introduced and how to install MPI on Linux OS is the main feature in the first section.Second, three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter-atomic force to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models can be difficult to parallelize efficiently - those with short-range forces where the neighbors of each atom change rapidly. They can be implemented on PC cluster installed at the first section. Each of them has advantages and disadvantages. The atom-decomposition algorithm is the simplest to implement and load-balances automatically. But because it performs all-to-all communication, its communication costs begin to dominate its run time on large number of processors. The force-decomposition algorithm is also relatively simple, though it often requires some pre-processors to assure load-balance. It also works wellindependent of the physical problem's geometry. Its O(N/P) scaling is better than that of AD algorithm, but is not optimal for large problems. The spatial-decomposition algorithm does exhibit optimal O(N/P) scaling for large simulation. However it suffersmore easily form load-imbalance and it more difficult to implement efficiently.In summery, this work will work well on the improving of computer simulation level in the thermodynamics of chemical engineering, the connecting to the international in this domain and the speeding-up applications of parallel computing inchemical engineering of our country.