Time Evaluation And Optimization Of GPU Discrete Computing Model

The great computing power of the GPU makes CUDA-GPU system applied widely in the research of the deological disasters. However, when we use the DEM to simulate the deological disasters with gradual changes, the large-scale computing and excessive output steps make the program run too long time. So the existing performance evaluation model needs two new requirement, firstly, the the model can accurately predict the overall running time of the computing program, which allows the users know the time invested in time, secondly, the model also can help the users to propose the optimization strategy to optimizate the computing program. Existing performance evaluation model can’t predict the running time of the program precisely. What’s more, in terms of performance optimization, they must evaluate the performance of the system at the end of the simulation, then put forward the optimization tactics.This paper presents a new performance evaluation model which is based on the memory access time and processing time. We consider that the main factors affecting the process of computing time are the memory access time and processing time.So we need to refine the change model of the memory access time and processing time.While training the model, the model use the early running of program as specimen to analyze and find the change model of the memory access time and processing time by using interpolation. To improve accuracy of the model, we find out the revise factor after comparing predicting time and actual time to fix the model, which can help the model to predict the running time of the program precisely. We can also use the model to locate the bottleneck of the program and propose the optimization strategy.This paper takes a test to examine the model, the results prove that in terms of predicting time, the model can control the deviation in the ideal range, and in terms of performance optimization, the model can locate the bottleneck at the beginning of the program, which can help the user to propose the optimization strategy to improve the computational efficiency of the computing program.