应用多GPU的可压缩湍流并行计算

利用CUDA Fortran语言发展了基于图形处理器(GPU)的计算流体力学可压缩湍流求解器。该求解器基于结构网格有限体积法,空间离散采用AUSMPW+格式,湍流模型为k-ωSST两方程模型,采用MPI实现并行计算。针对最新的GPU架构,讨论了通量计算的优化方法及GPU计算与PCIe数据传输、MPI通信重叠的多GPU并行算法。进行了超声速进气道及空天飞机等算例的数值模拟以验证GPU在大网格量情况下的加速性能。计算结果表明:相对于Intel Xeon E5-2670 CPU单一核心的计算时间,单块NVIDIA GTX Titan Black GPU可获得107~125倍的加速比。利用四块GPU实现了复杂外形1.34亿网格的快速计算,并行效率为91.6%。     

基于SHARC流水集束型多处理器板研制

结合某重点工程课题 ,设计和实现了基于 SHARC流水集束型多处理器板即研制出了具有 5个 SHARC的高速并行数字信号处理机。板上处理器数量可实现重配置以获得高的性价比。当板上前端 4个 SHARC用于 FFT时 ,研究和分析了板上前端 4个 SHARC用于 FFT时的三种不同缓冲数据模式下 FFT节点处理性能     

Asymptotically optimal component assembly plans in repairable systems and server allocation in parallel multiserver queues

T identical exponential lifetime components out of which G are initially functioning (and B are not) are to be allocated to N subsystems, which are connected either in parallel or in series. Subsystem i, i = 1,…, N, functions when at least K_i of its components function and the whole system is maintained by a single repairman. Component repair times are identical independent exponentials and repaired components are as good as new. The problem of the determination of the assembly plan that will maximize the system reliability at any (arbitrary) time instant t is solved when the component failure rate is sufficiently small. For the parallel configuration, the optimal assembly plan allocates as many components as possible to the subsystem with the smallest K_i and allocates functioning components to subsystems in increasing order of the K_i's. For the series configuration, the optimal assembly plan allocates both the surplus and the functioning components equally to all subsystems whenever possible, and when not possible it favors subsystems in decreasing order of the K_i's. The solution is interpreted in the context of the optimal allocation of processors and an initial number of jobs in a problem of routing time consuming jobs to parallel multiprocessor queues. © John Wiley & Sons, Inc. Naval Research Logistics 48: 732–746, 2001     

栅格地理数据并行格式转换引擎

如今大规模地理数据正在社会各个部门和组织中迅速积累,但是由于部门利益和历史沿袭等原因,大规模地理数据共享仍然极具挑战,相应共享技术需求仍然极其旺盛。作为地理数据共享的基础方式之一,传统单机地理数据格式转换技术,一方面受限于磁盘读写及带宽瓶颈,另一方面面对日趋庞大的数据规模,已很难满足实际应用需求。因此提出一种针对栅格地理数据的并行格式转换引擎,采用高性能计算集群环境支持大规模栅格地理数据转换共享,大幅降低了大规模栅格地理数据转换过程的时间成本。栅格地理数据并行格式转换引擎采用基于公共接口的设计理念,框架灵活、具有良好的扩展性,支持地理数据格式的读写自定义以及新数据格式添加,能够实现接入数据格式间的任意两两转换。为验证引擎框架及其处理效率,在Lustre并行集群环境下以格网数据交换格式(国家地理空间数据交换格式)向常见栅格地理格式的转换为示例进行了测试实验。结果表明,栅格地理数据并行格式转换引擎能够在8个节点Lustre集群中达到7.54的良好并行加速比。     

Optimal allocation of minimal repairs in parallel and series systems

In reliability engineering, the concept of minimal repair describes that the repair brings the failed unit (eg, system or component) to the situation which is same as it was just before the failure. With the help of the well‐known Gamma‐Poisson relationship, this paper investigates optimal allocation strategies of minimal repairs for parallel and series systems through implementing stochastic comparisons of various allocation policies in terms of the hazard rate, the reversed hazard rate, and the likelihood ratio orderings. Numerical examples are presented to illustrate these findings as well. These results not only strengthen and generalize some known ones in the seminal work of Shaked and Shanthikumar, but also solve the open problems proposed by Chahkandi et al.'s study and Arriaza et al.'s study.     

基于加密机制模式的无线路由器网络安全设计

在传统网络当中,各种密钥管理技术已经非常成熟,从密钥的生成、更新、维护、管理、销毁等各个方面均有成熟的技术。但无线路由器网络的节点无法使用传统网络的非对称密钥体制,因此,传统的密钥管理技术也同样不适用于无线路由器网络。从基于位置的随机密钥预分配模型出发,探讨如何对其进行加密设计。     

MPI并行程序中通信等待问题的诊断方法及其应用

随着并行规模的扩大,现有通信等待问题的诊断方法存在内存开销大、测量时间开销大等问题。通过对现有通信等待问题诊断方法的深入分析,同时考虑测量开销可控的实际需求,建立基于热点函数的通信等待问题诊断模型。基于上述模型,总结出一种更精简、更实用的通信等待问题诊断方法。将该诊断方法分别应用到二维LARED集成、LARED-S、LAP3D等大规模MPI并行程序的通信等待问题诊断过程,应用效果表明本诊断方法可精确定位导致通信等待问题的关键代码段,给出的优化方案及性能提升空间对于后续的程序改进具有参考价值,其中根据诊断结果优化后的LARED-S程序性能提升32%,通信等待时间减少44%。     

基于DSP的SMI方法快速实现研究

空时自适应处理(STAP)权值计算有数据域和均方域两种方法,分别以QR分解和样本协方差矩阵求逆(SMI)方法为代表.QR分解方法可以映射到脉动阵上并行实现,但实现复杂且设计成本较高;SMI方法实现则相对简单,但需要对样本协方差矩阵直接求逆.首先考察了不同矩阵求逆方法的内在并行性,基于DSP支持的片内并行技术,提出并实现了SMI方法的单DSP分块并行处理,进一步给出了数值稳定性分析和改善方法,实验结果证明了方法的有效性.     

Server scheduling on parallel dedicated machines with fixed job sequences

We consider server scheduling on parallel dedicated machines to minimize the makespan. Each job has a loading operation and a processing operation. The loading operation requires a server that serves all the jobs. Each machine has a given set of jobs to process, and the processing sequence is known and fixed. We design a polynomial‐time algorithm to solve the two‐machine case of the problem. When the number of machines is arbitrary, the problem becomes strongly NP‐hard even if all the jobs have the same processing length or all the loading operations require a unit time. We design two heuristic algorithms to treat the case where all the loading times are unit and analyze their performance.     

盲提取方法在柴油机活塞-缸套撞击信号提取中的应用

讨论了盲反卷积在机械振动信号处理中的应用以及盲反卷积中常用的模型和方法,利用信号和高斯信号之间的Kullback-Leibler距离构建目标函数,提出了一种以伪并行遗传算法为优化工具的盲反卷积算法,并通过仿真验证了算法的正确性.实验也证明,该算法能成功地从柴油机机体振动信号中提取出活塞-缸套撞击信号.