摘要
为了能够有效提高基于时域的SAR回波仿真的运行速度,提出了一种基于图形处理器(GPU)架构的SAR回波仿真优化实现方法。该方法结合GPU的计算密度高、高度并行的特点并利用CUDA流在GPU上同时执行多个任务,实现任务并行、指令并行和数据并行的三重并行,极大地挖掘了回波模拟全过程的并行性,缩短了回波仿真的运算时间。实验结果表明,该方法相对于传统的CPU上的串行算法平均加速比达到128倍,可用于实时信号处理。
To enhance the computational efficiency of SAR echo data simulation, an simulation method for SAR is presented. Combining the characteristics of high computation ism and CUDA stream, this method achieves tasks parallelism, instructions parallelism and largely exploited the parallelism in SAR echo simulation and method obtain the speed up about 128 times compared with processing. efficient raw echo data density, high paralleldata parallelism, which shortened computing time. The results show that this CPU. and the method is suitable for real-time signal
出处
《科学技术与工程》
北大核心
2014年第13期85-89,共5页
Science Technology and Engineering
基金
国家自然科学基金重大专项(GFZX04013905)资助
关键词
合成孔径雷达
图形处理器
CUDA
并行计算
回波仿真
synthetic aperture radar (SAR) architecture (CUDA) parallel computing graphic processing unit (GPU) computer unified device echo simulation