ArchSim:A System-Level Parallel Simulation Platform for the Architecture Design of High Performance Computer

High performance computer （HPC） is a complex huge system, of which the architecture design meets increasing difficulties and risks. Traditional methods, such as theoretical analysis, component-level simulation and sequential simulation, are not applicable to system-level simulations of HPC systems. Even the parallel simulation using large-scale parallel machines also have many difficulties in scalability, reliability, generality, as well as efficiency. According to the current needs of HPC architecture design, this paper proposes a system-level parallel simulation platform： ArchSim. We first introduce the architecture of ArchSim simulation platform which is composed of a global server （GS）, local server agents （LSA） and entities. Secondly, we emphasize some key techniques of ArchSim, including the synchronization protocol, the communication mechanism and the distributed checkpointing/restart mechanism. We then make a synthesized test of some main performance indices of ArchSim with the phold benchmark and analyze the extra overhead generated by ArchSim. Finally, based on ArchSim, we construct a parallel event-driven interconnection network simulator and a system-level simulator for a small scale HPC system with 256 processors. The results of the performance test and HPC system simulations demonstrate that ArchSim can achieve high speedup ratio and high scalability on parallel host machine and support system-level simulations for the architecture design of HPC systems.

An MPI parallel DEM-IMB-LBM framework for simulating fluid-solid interaction problems

The high-resolution DEM-IMB-LBM model can accurately describe pore-scale fluid-solid interactions,but its potential for use in geotechnical engineering analysis has not been fully unleashed due to its prohibitive computational costs.To overcome this limitation,a message passing interface(MPI)parallel DEM-IMB-LBM framework is proposed aimed at enhancing computation efficiency.This framework utilises a static domain decomposition scheme,with the entire computation domain being decomposed into multiple subdomains according to predefined processors.A detailed parallel strategy is employed for both contact detection and hydrodynamic force calculation.In particular,a particle ID re-numbering scheme is proposed to handle particle transitions across sub-domain interfaces.Two benchmarks are conducted to validate the accuracy and overall performance of the proposed framework.Subsequently,the framework is applied to simulate scenarios involving multi-particle sedimentation and submarine landslides.The numerical examples effectively demonstrate the robustness and applicability of the MPI parallel DEM-IMB-LBM framework.

Static Analysis Techniques for Fixing Software Defects in MPI-Based Parallel Programs

The Message Passing Interface (MPI) is a widely accepted standard for parallel computing on distributed memorysystems.However, MPI implementations can contain defects that impact the reliability and performance of parallelapplications. Detecting and correcting these defects is crucial, yet there is a lack of published models specificallydesigned for correctingMPI defects. To address this, we propose a model for detecting and correcting MPI defects(DC_MPI), which aims to detect and correct defects in various types of MPI communication, including blockingpoint-to-point (BPTP), nonblocking point-to-point (NBPTP), and collective communication (CC). The defectsaddressed by the DC_MPI model include illegal MPI calls, deadlocks (DL), race conditions (RC), and messagemismatches (MM). To assess the effectiveness of the DC_MPI model, we performed experiments on a datasetconsisting of 40 MPI codes. The results indicate that the model achieved a detection rate of 37 out of 40 codes,resulting in an overall detection accuracy of 92.5%. Additionally, the execution duration of the DC_MPI modelranged from 0.81 to 1.36 s. These findings show that the DC_MPI model is useful in detecting and correctingdefects in MPI implementations, thereby enhancing the reliability and performance of parallel applications. TheDC_MPImodel fills an important research gap and provides a valuable tool for improving the quality ofMPI-basedparallel computing systems.

面向国产异构众核系统的Parallel C语言设计与实现

异构众核架构具有超高的性能功耗比,已成为超级计算机体系结构的重要发展方向.但众核系统更为复杂的并行层次和存储层次,给编程和优化带来了极大的挑战.因此,研究面向众核系统的并行编程技术,对于降低国产众核系统并行应用的编程难度、提升并行程序的性能都具有重要的意义.提出统一架构的多模式并行编程模型,包括异构融合的加速运算模型和按同构方式编程的自主运算模型,根据编程模型设计了Parallel C语言,能够有效地描述国产众核系统的异构并行性.与其他众核系统上MPI+X的使用模式相比,编程和系统优化都具有全局视角,在多级局部性描述、单边消息、兼容已有多核应用等方面具有特色;基于Open64构建了Parallel C编译系统,全面支持加速运算模型和自主运算模型,提出并实现了数据布局与自动DMA、编译指导的线程代理和拓扑位置感知的集合通信等优化.Micro Benchmark和实际应用在神威太湖之光计算机系统上的测试数据结果表明:Parallel C语言和编译系统具有良好的性能和可扩展性,能够有效支撑大型应用.

一种基于Message Passing的并行程序设计技术

本文首先分析介绍构造并行计算机系统的两种方法：Ｍｕｌｔｉｐｒｏｃｅｓｓｏｒｓ和Ｍｕｌｔｉｃｏｍｐｕｔｅｒｓ，然后介绍一种Ｍｕｌｔｉｃｏｍｐｕｔｅｒｓ并行计算机系统中基于ｍｅｓｓａｇｅｐａｓｓｉｎｇ的并行程序设计环境以及并行程序设计方法，它具有程序设计简单、直观等特点，易于构造程序模型，并能获得较好的加速比。

基于Message Passing的并行编程环境

在分布式并行计算机系统中，由于处理机间无共事内存，因此采用了MessagePassins方式实现处理机间的通信.文中讨论了基于MessagePassing把的并行编程环境所应具备的特点，然后介绍几种被人们普遍接受的并行编程环境.

Parallel computation of unified finite-difference time-domain for underwater sound scattering

In this work, we treat scattering objects, water, surface and bottom in a truly unified manner in a parallel finitedifference time-domain （FDTD） scheme, which is suitable for distributed parallel computing in a message passing interface （MPI） programming environment. The algorithm is implemented on a cluster-based high performance computer system. Parallel computation is performed with different division methods in 2D and 3D situations. Based on analysis of main factors affecting the speedup rate and parallel efficiency, data communication is reduced by selecting a suitable scheme of task division. A desirable scheme is recommended, giving a higher speedup rate and better efficiency. The results indicate that the unified parallel FDTD algorithm provides a solution to the numerical computation of acoustic scattering.

Coupling analysis of transmission lines excited by space electromagnetic fields based on time domain hybrid method using parallel technique

We present a time domain hybrid method to realize the fast coupling analysis of transmission lines excited by space electromagnetic fields, in which parallel finite-difference time-domain (FDTD) method, interpolation scheme, and Agrawal model-based transmission line (TL) equations are organically integrated together. Specifically, the Agrawal model is employed to establish the TL equations to describe the coupling effects of space electromagnetic fields on transmission lines. Then, the excitation fields functioning as distribution sources in TL equations are calculated by the parallel FDTD method through using the message passing interface (MPI) library scheme and interpolation scheme. Finally, the TL equations are discretized by the central difference scheme of FDTD and assigned to multiple processors to obtain the transient responses on the terminal loads of these lines. The significant feature of the presented method is embodied in its parallel and synchronous calculations of the space electromagnetic fields and transient responses on the lines. Numerical simulations of ambient wave acting on multi-conductor transmission lines (MTLs), which are located on the PEC ground and in the shielded cavity respectively, are implemented to verify the accuracy and efficiency of the presented method.

Multi-Deme Parallel FGAs-Based Algorithm for Multitarget Tracking

For data association in multisensor and multitarget tracking, a novel parallel algorithm is developed to improve the efficiency and real-time performance of FGAs-based algorithm. One Cluster of Workstation （COW） with Message Passing Interface （MPI） is built. The proposed Multi-Deme Parallel FGA （MDPFGA） is run on the platform. A serial of special MDPFGAs are used to determine the static and the dynamic solutions of generalized m-best S-D assignment problem respectively, as well as target states estimation in track management. Such an assignment-based parallel algorithm is demonstrated on simulated passive sensor track formation and maintenance problem. While illustrating the feasibility of the proposed algorithm in multisensor multitarget tracking, simulation results indicate that the MDPFGAs-based algorithm has greater efficiency and speed than the FGAs-based algorithm.

Development of high performance casting analysis software by coupled parallel computation

Up to now,so much casting analysis software has been continuing to develop the new access way to real casting processes. Those include the melt flow analysis,heat transfer analysis for solidification calculation,mechanical property predictions and microstructure predictions. These trials were successful to obtain the ideal results comparing with real situations,so that CAE technologies became inevitable to design or develop new casting processes. But for manufacturing fields,CAE technologies are not so frequently being used because of their difficulties in using the software or insufficient computing performances. To introduce CAE technologies to manufacturing field,the high performance analysis is essential to shorten the gap between product designing time and prototyping time. The software code optimization can be helpful,but it is not enough,because the codes developed by software experts are already optimized enough. As an alternative proposal for high performance computations,the parallel computation technologies are eagerly being applied to CAE technologies to make the analysis time shorter. In this research,SMP (Shared Memory Processing) and MPI (Message Passing Interface) (1) methods for parallelization were applied to commercial software "Z-Cast" to calculate the casting processes. In the code parallelizing processes,the network stabilization,core optimization were also carried out under Microsoft Windows platform and their performances and results were compared with those of normal linear analysis codes.

Efficient Data-parallel Computations on Distributed Systems

Task scheduling determines the performance of NOW computing to a large extent. However, the computer system architecture, computing capability and system load are rarely proposed together. In this paper, a biggest heterogeneous scheduling algorithm is presented. It fully considers the system characteristics (from application view), structure and state. So it always can utilize all processing resource under a reasonable premise. The results of experiment show the algorithm can significantly shorten the response time of jobs.

Parallelizing AT with open multi-processing and MPI

Simulating charged particle motion through the elements is necessary to understand modern particle accelerators. The particle numbers and the circling turns in a synchrotron are huge, and a simulation can be timeconsuming. Open multi-processing(Open MP) is a convenient method to speed up the computing of multi-cores for computers based on share memory model. Using message passing interface(MPI) which is based on nonuniform memory access architecture, a coarse grain parallel algorithm is set up for the Accelerator Toolbox(AT)for dynamic tracking processes. The computing speedup of the tracking process is 3.77 times with a quad-core CPU computer and the speed almost grows linearly with the number of CPU.

MPI:大有希望的massage_passing标准

message-passing是一个被广泛应用的并行计算模型,但至今,无论从语法或语义上message-Passing都还没有形成一个标准.并行计算机厂商在提供标准串行语言的同时,提供了各自不同的专有message一passing库,并行程序设计难免依赖于具体的系统和应用,这既限制了可移植性,也限制了听取用户对某些不易获得或对当前的应用不合适的系统的意见。没有一个标准,厂商们非常自然地把他们的专有库的优点视为有竟争力的优势,更加着力与众不同(因而不可移植)。为了应付可移植问题,研究团体已经开发了许多库来收集其它库之间的不同点.

Development of Ubiquitous Simulation Service Structure Based on High Performance Computing Technologies

The simulation field became essential in designing or developing new casting products and in improving manufacturing processes within limited time, because it can help us to simulate the nature of processing, so that developers can make ideal casting designs. To take the prior occupation at commercial simulation market, so many development groups in the world are doing their every effort. They already reported successful stories in manufacturing fields by developing and providing the high performance simulation technologies for multipurpose. But they all run at powerful desk-side computers by well-trained experts mainly, so that it is hard to diffuse the scientific designing concept to newcomers in casting field. To overcome upcoming problems in scientific casting designs, we utilized information technologies and full-matured hardware backbones to spread out the effective and scientific casting design mind, and they all were integrated into Simulation Portal on the web. It professes scientific casting design on the NET including ubiquitous access way represented by ＂Anyone, Anytime, Anywhere＂ concept for casting designs.

Optimizing a Parallel Video Encoder with Message Passing and a Shared Memory Architecture

Implementing video applications on emerging multi-core processors is a promising technique for personal, real-time multi-media applications. However, when porting the legacy parallel video encoders developed for clusters to shared-memory multi-cores, the existing parallel algorithms result in workload imbalances on different cores and communication inefficiencies. This paper describes a strip-wise parallel scheme to balance workloads and a hybrid communication mechanism to reduce communication overhead. The implementation of the H.264 parallel encoder on an eight CPU Intel Xeon system achieves 5x to 6x speed-up over a single thread encoder and achieves a 29% performance improvement over the commonly used master-slave schemes on clusters. The paper also gives further analysis on scalability, parallel efficiency, workload balance, and communication overhead as the number of cores varies.

高分三号数据分布式负载均衡并行转换算法

针对PolSARpro软件原有高分三号数据转换模块无法适应分布式环境及数据转换效率相对较低的问题,提出一种基于MPI、MapReduce和OpenMP并按照该软件的数据格式要求进行分布式负载均衡的并行转换算法。该算法外层采用MPI按任务分布式计算,中间层采用MapReduce按景并行处理,内层采用OpenMP按极化方式并行计算,并采用“Z”字形数据负载均衡策略,将高分三号不同成像模式下的多极化数据快速精确定标及格式转换,以抗数据偏斜。实际数据的测试结果表明,该算法的整体性能提升了约50%,验证了算法的可行性、高效性和正确性。

基于MPI和MapReduce混合编程的高分三号多模数据分布式并行转换算法

我国高分三号形成三星陆海雷达卫星星座。为了提升PolSARpro极化SAR图像处理的开源软件在高分三号数据转换环节的处理能力,提出了基于OpenMP、MapReduce和MPI并按照PolSARpro软件的数据格式要求进行分布式并行转换处理算法,将不同成像模式下的多极化数据实现快速定标和格式转换,并且通过KingMap V10.0 GIS平台实现了算法并在实际数据中进行测试,验证了算法的高效性、可靠性、易扩展性和正确性。

基于模式记忆并行蚁群算法的输电网规划

该文讨论了目前大规模输电网规划求解中常见的无法完成对解空间的充分搜索,从而难于求得全局最优解的问题;给出了泛函形式的输电网规划模型,并重点分析了输电网规划解的模式。在分析了传统蚁群算法易产生未成熟收敛现象及其原因的基础上,设计了一种基于模式记忆的并行蚁群算法,该算法通过模式记忆实现了解空间分解,能够有效地识别、记忆和跳出局部最优解;通过局部细化搜索进一步加强了局部搜索能力;通过并行计算提高了计算速度。某实际77 节点的算例分析表明了该算法具有很高的计算效率和优秀的局部、全局收敛性,有效克服了现代启发式算法在求解输电网规划问题时存在的效率不高及未成熟收敛等现象。

梯级水电站群长期优化调度云计算随机动态规划算法

为了应对随机动态规划算法在解决梯级水电站群长期发电优化调度时的"维数灾"问题,并行化方法得到了广泛研究。单机多核并行算法扩展性不强;传统的分布式并行算法编程复杂,缺少负载均衡和容错机制。云计算平台作为新的分布式计算平台能够充分利用资源,具有诸多优势。为了探索云平台下的分布式并行随机动态规划模型,该文基于消息传递接口(message passing interface,MPI)和Spark框架分别实现了传统集群计算和云计算分布式随机动态规划算法,后者将计算模型转换为数据处理模型进行计算,并通过三库优化调度实验对算法进行了比较。算法分析及实验结果表明,基于云计算的分布式并行随机动态规划算法则可以有效利用云平台的优势,同时拥有完善的容错以及负载均衡机制,具有广阔的应用前景。

地理栅格数据的并行访问方法研究

在海量地理栅格数据处理中,数据I/O性能是影响处理算法程序整体性能的关键。目前针对地理栅格数据I/O优化问题的研究成果还很有限,通过对并行程序中的数据I/O模式进行深入分析,结合栅格数据逻辑模型和物理模型的特点,提出了面向地理栅格数据的并行I/O框架;基于消息传递模型,实现了4种并行访问方法。实验证明,并行访问方法优于传统的串行访问方法和分时多进程访问方法。该研究成果可以提高并行栅格处理程序的I/O访问效率,进而提高其整体并行性能。