摘要
为实现软件化雷达在不同信号处理器上的实时信号处理,需要研究通用高性能处理器,如CPU和GPU信号处理算法的并行机制。论文以雷达脉冲压缩运算模块为例,重点研究了利用GPU信号处理的并行机制。首先给出雷达脉冲压缩数学模型,然后针对算法实现流程,分别从片上缓存、内核线程和数据并行等方面设计了三种GPU并行优化策略。仿真测试表明,所提出的GPU并行机制与典型多核CPU平台相比,具有更好的实时性能。
In order to achieve the real-time signal processing of software radar for various processors,it is essential to studythe parallel mechanism for the general high performance processor,i.e.,CPU and GPU. Focusing on the parallel realization of GPUprocessor,this paper takes the pulse compression processing as an example. Firstly the algorithm model is presented,and thenthree parallel ways are designed,including the on-chip cache,kernel threads and data in parallel. The simulations show that theproposal mechanism for GPU has a better real-time property comparing to the multiple-kernel CPU.
出处
《舰船电子工程》
2017年第10期30-32,107,共4页
Ship Electronic Engineering