Parallel Computing For Reverse Monte Carlo Method Based On MPI+OpenMP

Reverse Monte Carlo method is a commonly used method to explain experimental phenomena and discover physical laws based on the partially determined physical laws as constraints and multiple simulations based on experimental data.It is a variant of the standard Metropolis-Hastings algorithm that can be used to solve the inverse problem and adjust the model to achieve greater consistency between the parameters and the experimental data.Inverse Monte Carlo method is widely used in the study of high-energy physics and solid chemistry,but due to the development of technology and the increase of simulation accuracy,the experimental scale and data amount of simulation become larger and larger,and the time-consuming process simulation becomes More and more,to get a better configuration to wait a long time,affecting the efficiency of scientific research.In response to the above problems,this paper studies the inverse Monte Carlo method based on MPI + OpenMP parallel optimization.Taking a reverse Monte Carlo Empirical Potential Structure Refinement(EPSR for short)as an example,a detailed parallel optimization study was carried out.A coupling algorithm of GROMACS and EPSR is proposed to reduce the energy consumption of the system effectively.The optimization of Message Passing Interface(MPI)between EPSR nodes is carried out in parallel and the OpenMP parallel optimization,Under the premise of ensuring that the simulation precision meets the experimental requirements,the simulation efficiency of the EPSR is enhanced.The contributions of this paper mainly include the following three aspects:First of all,a coupled algorithm of GRMACS and EPSR is proposed in this paper.The atomic coordinates,molecular model and potential energy of atoms in GROMACS and EPSR are coupled to make full use of the parallel advantages of GROMACS energy minimization to improve the system energy minimization effectiveness.Secondly,this paper designs and implements MPI parallelization method for inverse Monte Carlo using EPSR.MPI parallel programming design mode is adopted,in which the master / slave is mixed.The coordinate information of the particles in the EPSR system is distributed to different nodes for simulation calculation,then analyzed back and analyzed and iterated until the system reaches the equilibrium state.The speedup of simulation iteration module reaches 2.2,and the simulation process has good convergence.Thirdly,this paper designs and implements the OpenMP optimization method of inverse Monte Carlo application EPSR.Using OpenMP programming model,the program can be executed concurrently in parallel by multi-threading to realize multithreaded parallel computing in nodes.By merging parallel regions,increasing theparallel granularity and fully mining the computing power of multi-core processors,the simulation can be greatly reduced Simulation time.Finally,the simulation of parallelism is evaluated,the feasibility and effectiveness of the parallel algorithm are verified,and the performance bottleneck is analyzed.Future work in this paper will do more optimization around the inverse Monte-Carlo method and extend MPI to more nodes.