摘要
针对机载MIMO雷达杂波模型运算量大的问题,建立了基于CUDA的模型,利用图形处理器的并行运算能力来加速运算。为了符合CUDA的通用计算架构特性,该模型在算法上优先设计并行运算,在编程上尽量引入并行线程,实现图形处理器的浮点运算、存储器以及带宽资源的有效利用。最后测试了加速效果与模型结构、数据规模的关系。实验结果表明,与CPUIntel i5 750比较,图形处理器GTX465的建模运算结果准确,效率提升数十倍,能支持更大的运算量。
Considering big operand problem of clutter model for airborne MIMO radar, a clutter based on Compute Unified Device Architecture （CUDA） is established and operation is speeded up by using parallel operational capa- bility of Graphic Processing Unit（GPU）. In order to accord with general computation architecture characteristic of CUPD, the new model prefers designing parallel computing in algorithm and introducing parallel threads in programming to implement effective usage of floating point operation, storage and wideband resources of GPU. And finally, relation between the speedup effect and model architecture and data size is tested. The tested results prove that modeling computation result of GPU based on GTX 465 is accurate in computing result, and efficiency is increased by tens of times and bigger operand can be supported comparing with CPUIntel i5 750.
出处
《火控雷达技术》
2011年第4期35-39,共5页
Fire Control Radar Technology
基金
电子信息控制重点实验室项目资助(20090102)
中央高校基本科研业务费专项资金资助(ZYGX2009J015)
