摘要
针对目前飞行器航迹规划中航迹计算量大、耗时长的问题,结合图形图像处理器(GPU)大规模并行计算特性,提出了一种基于GPU的航迹快速计算方法。该方法以4阶Runge-Kutta(R-K)法为基础求解航迹微分方程组,通过计算资源的分配、数据流分段将航迹计算任务映射到GPU线程模型,利用CPU+GPU异构模型的数据流控制能力,实现多条航迹的并行计算。试验表明,该方法的计算精度满足要求,并获得了几十倍的理想加速效果,为航迹规划系统其他大规模并行计算提供了新的解决思路。
With reference to the problem that the current aircraft track planning requires much computing and time,a rapid track computing method based on the graphic processing unit（GPU） is proposed.The method uses the GPU ability to perform massive parallel computing.By the method,the fourth-order Runge-Kutta（R-K） method is used to solve the differential equations of the track,the task of track computing is mapped to the GPU thread model by allocating the computing resources and dividing the data stream,and the CPU＋GPU heterogeneous model for controlling the data stream is used to achieve parallel computing of multiple tracks.Experimental results show that the computing accuracy of the method is satisfactory and the computing speed increases by dozens of times.The method provides new solutions to massive parallel computing in an aircraft track planning system.
出处
《指挥信息系统与技术》
2011年第4期55-59,共5页
Command Information System and Technology
关键词
图形图像处理器
统一计算设备架构
并行计算
航迹计算
graphic processing unit（GPU）
compute unified device architecture（CUDA）
parallel computing
track computing
