众核处理架构在水下航行器相位编码脉冲回波检测中的应用

Application of many-core processing architecture to echo detection of phase-coded pulse for autonomous underwater vehicle

针对宽带编码脉冲、多输入多输出等新型目标探测体制发展带来的运算量和数据存储需求剧增的问题,根据水下航行器相位编码脉冲回波检测算法的数据级并行特点,提出应用图形处理器（Graphics Processing Unit,GPU）众核处理架构,并从任务分配策略、数据处理流程、GPU硬件资源利用率和存储器访问等角度考虑,设计了算法在GPU上的并行实现框架。利用湖试数据测试了桌面级GPU平台、嵌入式GPU平台与基于多核数字信号处理器（Digital Signal Processor,DSP）的传统航行器信号处理平台的性能,与多核DSP平台相比,嵌入式GPU平台在功耗、运算性能等方面更有优势。研究结果表明采用嵌入式GPU平台可大幅提升每瓦特性能指标并简化系统设计,能满足新型航行器探测系统大数据量、低功耗和实时性的应用需求。

With the development of new target detection technology such as wideband coded pulse and multipleinput multiple output, the computational complexity and memory space requirement increase seriously. To address this problem, basing on the characteristic of data parallelism of echo detection algorithm of phase-coded pulse for Autonomous Underwater Vehicle （AUV）, a strategy of using Graphics Processing Unit （GPU） with many-core processing architecture is proprosed. And a GPU parallelism implementation framework of the alogrithm is designed by considering a series of optimization measures including task allocation strategy, data processing procedure, GPU occupancy and memory access pattern. The desktop GPU platform, embedded GPU platform and traditional signal processing platform of AUV based on multicore Digital Signal Processor （DSP） are tested by experimental data, and the result shows that the embedded GPU platform is more dominant than multicore DSP platform in power consumption and computing performance. The research results indicate that the embedded GPU platform can greatly improve the performace per watt and simplify the system design, and it can meet the requirement of large dataset, low power consumption, and real time for AUV detection system.