Design Of Integrated Heterogeneous Computing Node Framework For Signal Processing

In the current computing scenario,due to the sharp increase in the amount of data,the demand for computing power in the processing center is also increasing.In some general computing scenarios,such as cloud computing,deep learning and other business scenarios,the computing requirements have exceeded the affordability of a single general-purpose processor,so it has become a trend to use heterogeneous computing in existing scenarios.Among them,in the application scenario of satellite communication,higher requirements are put forward for the real-time performance and reliability of computing.Therefore,integrated infrastructure and edge computing solutions are proposed for heterogeneous computing to solve such problems.However,there are different needs and problems in the integrated infrastructure design,so it is necessary to design a special framework to deal with these problems.This thesis aims to meet the requirements of communication signal processing and real-time control signal processing by designing an integrated edge computing node and heterogeneous computing framework.The contributions of this thesis mainly includes the following aspects:First,the construction of the heterogeneous computing platform framework has been completed,including the basic operating environments such as processor boards in the processing center,Linux system migration in the ARM part of the heterogeneous computing node,and the integration of FPGA resources.Second,the integrated data interaction section is completed.According to different frameworks used in different scenarios,a unified Redis database is adopted for data interaction between the processing center and computing nodes,and it is applied to different computing frameworks.In addition,according to different computing frameworks,the corresponding parts of the processing center and computing nodes are virtualized and controlled by the processing center.Third,a scheme of using SSD to save communication data in satellite communication scenarios with high reliability is proposed.and based on this scenario,the use of LDPC coding as an error correction code in the reading process is proposed.Finally,a heterogeneous computing node framework is developed according to this scheme.The framework is constructed and designed,and its feasibility and performance are verified.Fourth,the control problem of satellite nodes is analyzed,and a set of control schemes based on heterogeneous computing is proposed for the real-time control signal processing.The corresponding node framework is designed on the basis of heterogeneous computing platforms for this scheme.And the components are developed,then a specific testbench is taken to verify the scheme's feasibility.This thesis completes the design of heterogeneous computing frameworks for different signal processing in satellite communication scenarios,and implements and verifies different framework designs respectively,confirming that they meet the edge computing requirements of integration under the premise of completing basic functions.The architecture can ensure real-time user control and data processing.It has significant prospect in engineering.