Design And Optimize Of Parallel Computing Model For NCS Radar Imaging Algorithm

Synthetic aperture radar imaging system is an advanced microwave observation system. After decades of development, it has been penetrated into various fields of science and engineering and its performance is now different from the first generation system. Synthetic aperture radar imaging algorithm based on frequency domain FFT fast Fourier transform algorithm solves the problems related to the frequency and orientation, but it is not conducive to high-resolution, low-frequency case of parallel computing. NCS algorithm is one of them.NCS imaging algorithm processes the source data processing including the FFT/ IFFT and complex calculations, and some transpositions, and so on. This paper introduces the multi-core parallel computing system architecture and algorithms used by NCS, also introduced the algorithm implementation process. This paper elaborates the overall system architecture of NCS algorithm and computing cluster, transpose clusters and FFT module in detail. Also introduces PRAM and other commonly used parallel computing models. Its performance analyzed in detail provides a theoretical basis for performance evaluation of NCS algorithm.In the detailed design, the paper describes the implementation of Linux-based NCS algorithm. The most important module is the FFT calculation module and the module matrix transpose. FFT module costs the most time in the calculation process of the entire program, so this paper focuses on the design and implementation of FFT module.NCS algorithm model is based on the characteristics of the memory accurate system. These models are a parallel computing algorithm for any problem and the tasks need to be divided before the beginning and put data in particular memory address. It is time-saving in subsequent modification, but the drawback is working more in the initial phase.Finally, optimization has been implemented in part of NCS model. Because the purpose of NCS algorithm is to build the model providing the hardware division of tasks among programs and provide data to support post-debug of hardware system, so the task division of NCS algorithm directly affect the ultimate realization of the process and achieve efficiencies. We simulate the multi-core parallel computing through the method of multi-threading technoLogy. Meanwhile, according to the time the program used to analyze the efficiency of the program, which also provides a reference that the hardware description in certain workload, the number of processor cores is not the more the better, reasonable division of tasks and adequate use of processor resources is essential for a multi-core system. Experimental results show that the running time before optimization TFFT=89.1s, TSTART=5.5s while after the optimization, the Tsub= 65.3s, TFFT=48.6s, Tstart=5.5s and the speed-up ratio is 1.39. System serial code section accounts for 25.6%, so the speedup limit of 3.9. Under the experimental conditions and does not consider the Logic and power requirements, the speedup ratio limit of 2.56, mainly due to the optimized under realistic conditions can not be completely ignore the part of parallel time.Currently the project has been completed based on FPGA prototype demo presentation system.