Research On Perfomance Optimization Of CPU-GPU Heterogeneous Platform And Its Application In Real-time Signal Simulation Technologies

With the increase of computing power and programmability of GPU,heterogeneous computing based on CPU-GPU gradually becomes a hotspot from the scientific community to the industry.Compared with homogeneous systems,heterogeneous systems have better performance,but also brings greater programming complexity and performance optimization difficulty.Signal simulator is widely used in the field of sensor and sensor engineering to assist its system design and verification.The signal simulation algorithm needs to deal with complex control logic and large amount of data calculation,which has a good match with the heterogeneous system.Therefore,it is of great practical value to study the realization of signal simulation on heterogeneous platform.The main work of this paper is as follows:1)Integrate the CPU-GPU heterogeneous platform on the CPU and GPU-side optimization technologies,the entire platform for a comprehensive performance optimization exploration.At the same time,new optimization techniques such as group packet and slice data transmission,computation and communication flow hydration and kernel fusion are proposed,and theoretical analysis and experimental verification are also carried out.2)In order to further verify the effect of various optimization techniques,and to introduce the CPU-GPU heterogeneous platform programming process,this paper takes solving complex electronic system's echo signal simulation problems as an example,and does relevant algorithm design,programming and performance optimization.The results show that the real-time signal simulator based on CPU-GPU heterogeneous platform can achieve the throughput of 5.79GB/s,which is 21 times faster than that of isomorphic platform,and the elastic support expansion Multi-GPU.