Research On Key Techniques Of Random Number Generation In Hardware Accelerated Monte Carlo Calculation

Monte Carlo algorithms have important applications in computational physics,filtering problems,reliability problems,biomedical science,financial engineering and scientific experimental simulation.However,with the development of science,the Monte Carlo calculation model becomes more and more complex,and the amount of calculation also increases,which will lead to a long time for Monte Carlo calculation.The generation of random Numbers in Monte Carlo calculation is the foundation,and the acceleration of random number generation can be realized in the whole Monte Carlo calculation.Therefore,accelerating the generation of random Numbers in Monte Carlo calculation has become an urgent problem.Compared with general computing platforms such as CPU and GPU,using FPGA to generate random Numbers performs better in terms of computing speed,hardware resource occupancy and power consumption.Firstly,this paper describes the current research status of monte carlo calculation based on FPGA,and proposes the key technology of hardware acceleration studied in this paper,aiming at the shortcomings of random number generation speed,quality and logical resource utilization.Secondly,the advantages and disadvantages of different random number generation algorithms are compared and analyzed.Then,this paper takes reducing hardware resource consumption and improving the generation speed and quality of random Numbers as the research objectives,and makes an in-depth study on the key technologies such as the generation of uniform and gaussian random Numbers in monte carlo calculation.In the uniformly distributed random number generation technology,the PMT19937 hardware structure was proposed,and the four-port BRAM was designed and realized to complete three reads and one writes in a single cycle.The four-port BRAM module was applied to the PMT19937 structure to accelerate the generation of uniformly distributed random Numbers.In the gaussian distribution random number generation technology,a combined gaussian random number generation scheme based on the boxmuller algorithm and the central limit theorem is proposed.The idea of greedy algorithm is applied to improve the Angle recoding CORDIC algorithm,and the algorithm is applied to calculate the trigonometric function in the box-muller algorithm.Finally,two kinds of random number generators designed in this paper are verified by experiments.The results show that the hardware structure of the two kinds of random number generators designed in this paper has better performance in the aspects of random number quality,hardware resource occupation and generation speed compared with the similar designs.In addition,compared with the general purpose processor,the uniformly distributed random number generator has obtained about 8 times the acceleration performance,and the gaussian random number generator has obtained about 6.5 times the acceleration performance.The research results of this paper have some practical significance for the development of the monte carlo computing hardware accelerator based on FPGA.