论文部分内容阅读
基于分布式平台开展一种新的时域有限差分(FDTD)并行算法研究,该算法基于VC++、CUDA5.0平台开发,调用Intel MPI 4.1.0库进行测试,在上海交通大学高性能计算中心图形处理单元(GPU)集群、上海超级计算机中心的“魔方”商用超级计算机以及国家超级计算济南中心的“神威蓝光”国产超级计算机等平台开展软件调试。通过对纯CPU、GPU以及CPU和GPU的混合测试,线程调度水平、核心函数处理速度得到明显提升,同时减少了通信执行时间比例,提高了加速比和并行效率,最后以2×2微带阵列为验证模型进行拓扑优化测试,结果证明该算法准确、有效。
Based on the distributed platform, a new FDTD parallel algorithm is developed. The algorithm is developed based on VC ++ and CUDA5.0 platform, and is tested by calling Intel MPI 4.1.0 library. In the graphics of Shanghai Jiaotong University High Performance Computing Center Processing unit (GPU) cluster, Shanghai Supercomputer Center “Rubik’s Cube ” commercial supercomputer and the National Supercomputing Jinan Center “Shenwei Blu-ray ” domestic supercomputer and other platforms to carry out software debugging. Through the pure CPU, GPU and CPU and GPU hybrid testing, the thread scheduling level, the core function processing speed has been significantly improved, while reducing the proportion of communication execution time, improve the speedup and parallel efficiency, and finally 2 × 2 microstrip array The topology optimization test for the verification model shows that the algorithm is accurate and effective.