Research On Communication And Task Scheduling Based For Multidimensional Heterogeneous Computing System

The arrival of 5G has promoted the widespread application of the Internet of Things,and also brought massive data.These data will occupy a lot of space and resources when they are stored,so they need to be processed in time.Heterogeneous computing has more advantages than traditional computing.It has a large amount of computing resources and can assign different tasks according to the characteristics of different computing units.Therefore,heterogeneous computing is an important method to solve the problem of computing power and energy consumption in future data centers.In order to solve the problems of complex programming and unreliable reliability in heterogeneous computing,I proposed a programming framework based on state transition matrix(STM).It uses the application program interface(API)of CUDA,Vivado and other applications to control the resources of computing unit such as GPU,FPGA through STM,and it can also automatically generate a framework of standard C code for heterogeneous computing.For the scheduling problem,I proposed the pheromone-optimized artificial ant colony algorithm(MACO),and proposed the optimization of heterogeneous architectures for the expected factors and information factors,and optimized pheromone volatilization mechanism.Then I implemented an effective heterogeneous task scheduling method.Finally I compared the MACO algorithm with other algorithms through key parameters such as scheduling length and iteration efficiency to prove the effectiveness of the algorithm proposed in this paper.On the other hand,for the communication problem,I proposed an efficient PCIe communication method that uses FPGA and GPU as the main controller respectively,and connects different computing units through a high-speed serial computer expansion bus(PCI-E)without CPU memory,it breaks through the short board of the read operation in PCIe communication and further improves the communication efficiency of heterogeneous platforms.Finally,I built a heterogeneous computing platform based on FPGA-GPU-CPU.Based on the above research content,I designed the communication verification experiment and task scheduling experiment respectively,and then conducted the actual application experiment through Fasthog + SVM.The experimental results are analyzed to verify the reliability and effectiveness of the method and platform built in this paper.