| 摘要: |
| 传统离线信号分析软件处理的数据量庞大且采用单一串行结构，其处理速度通常比较慢。为此，提出了基于图形处理单元（GPU）+中央处理器（CPU）异构方式实现离线信号分析软件的并行化优化思路。首先将离线信号分析中的单一信号串行流程改进为多信号同时处理方式以实现并行效果，然后重点分析了流程中几个比较费时的典型侦察信号处理算法的特点并进行了算法的优化实现和验证，最后提出了离线信号处理优化准则和资源分配策略。试验结果表明，离线信号分析处理软件算法优化改进后，大多数算法处理加速比达到20以上，具有较好的工程实用性，同时该优化准则和策略对于软件化实时信号处理也具有一定的参考价值。 |
| 关键词: 侦察信号处理 软件并行化 离线信号分析 PN码估计 CPU+GPU异构 |
| DOI： |
|
| 基金项目: |
|
| Parallel analysis and processing technique for off-line signals |
| LIANG Xianming,CHEN Wenjie |
| (Southwest China Institute of Electronic Technology,Chengdu 610036,China) |
| Abstract: |
| Because of a large amount of processing data and the serial structure,the processing speed of the traditional off-line signal analysis software is low.To optimize the softwares,a parallelization idea based on CPU+GPU heterogeneous platform to implement the off-line signal analysis software is proposed.Firstly,a single serial process is modified to multiple parallel process in order to increase the processing speed.Then the structures of several typical algorithms executing a lot of time are analyzed and the algorithms are optimized and verified.Finally the optimization principles and resource allocation strategies for off-line software are proposed.The test results show that the speed-up ratio of most optimized algorithms is over 20,and they has good engineering practicability.The resource optimization criterion and strategy can also be applied in real time signal processing. |
| Key words: reconnaissance signal processing software parallel implementation off-line signal analysis PN code estimation CPU+GPU heterogeneous |
《电讯技术》编辑部 