Parallel Calculation And Optimization Research Of Magnetic Shell Parameter L Value Based On OpenMP

In the solar-terrestrial space environment,the magnetosphere is the main area where solar activity interacts with the earth and is an important place for the exchange of material energy.Radiation belt is one of the core areas of magnetosphere,and it is the main gathering place of high-energy particles captured by magnetosphere.These high-energy particles are carriers of material energy in the process of solar-terrestrial interaction.It is an important factor to threaten the safety and efficiency of spacecraft that the particles of radiation belt move in cyclotron,bounce and drift under the action of magnetic field.The(L,B)coordinate system in the study of radiation belt can reduce the threedimensional problem to a two-dimensional problem,which greatly improves the utilization of limited observation data in the three-dimensional space of the radiation belt.The magnetic shell parameter L is one of the important parameters,which is widely used in the research fields of particle characteristic analysis,spacecraft orbit radiation environment description,and radiation belt modeling.The L value is calculated by the integral of the magnetic field line tracing.The extremely complex geomagnetic fields configuration result in the step size of the magnetic field line tracking is strictly limited,resulting more steps of single tracking calculation.In addition,the tracing path and distance of the magnetic field line vary significantly with time and space at different spatial positions,which results in multiple L-value tracking calculations that are independent of each other,of varying complexity,and cannot be degenerate,further resulting in a slow and low-efficiency L-value calculation.With the rapid development of computer software and hardware technology,parallel computing provides a solution for such large and complex computing problems.In this paper,by analyzing different modes of parallel computing,we choose OpenMP as a parallel tool.In view of the parallel advantages of OpenMP,it has been widely used in different research fields,but it has not been found in the field of magnetic L-shell.Because L value is widely used in science and application fields,but its computing speed is slow and efficiency is low,so it is of great significance to improve the calculation efficiency of L value.In order to improve the calculation efficiency of L value,this paper uses OpenMP to realize the single-machine multithreading parallelization of the L value calculation program widely used at present,and analyzes the accuracy,speedup,efficiency and load balance of the calculation results of the parallel program.The results show that OpenMP can successfully achieve program parallelization and significantly improve the efficiency of L value calculation.When the number of threads is set equal to the maximum number of threads in the system(the maximum number of threads in the computer system used in this paper is 8),The parallel effect is the best,the speedup range is 1.57 ? 2.81,and the parallel efficiency range is 19.63% ? 35.13%.It is found that load imbalance between threads is the main factor affecting parallel performance.In order to further improve the parallel speedup and efficiency,the load imbalance is optimized.This paper proposes a sort factor,which represents the complexity of L-value calculation.Based on this,an algorithm of sorting and redistribution(sort,“s”)is proposed to optimize the load.The dynamic,sort + “S”,sort + dynamic,and sort + “S” + dynamic methods are used to optimize parallel programs,and the optimization effects of different methods are compared and analyzed.The results show that the optimization effect of sort + dynamic is the best and speeds up.The specific range is 2.30 ? 3.30,the efficiency is 28.75% ? 41.25%,and the parallel efficiency is increased by 6.12% ? 9.12% after optimization.The research in this paper shows that OpenMP can successfully achieve the parallelization of L-value calculation programs and significantly improve the L-value calculation efficiency.Aiming at the problem of load imbalance,the best parallel acceleration ratio and efficiency can be obtained by using sort + dynamic adjustment.This research can provide useful exploration and important reference for similar big data analysis research in space science and other fields.