构造并行化系统交互环境的若干关键技术

交互式并行化系统通过提供友好的交互功能并引入用户知识来协助程序的并行化,是解决自动并行化能力不足的一条有效途径.描述了一个并行化系统交互环境TIPSIE(interactive en vironment of Tsinghua interactive parallelizing system),并就构造该环境的性能预测、增量编译和数据相关查询等关键技术进行了讨论.实验结果表明,这些技术能够有效地提高系统的并行化能力和效率.     

一个交互式的Fortran77并行化系统

并行化编译器可以把现有的串行程序自动或半自动地转换为并行程序.现有并行化系统的自动并行化效果与手工并行化的效果相比还有一定的差距,这是由于并行化工具的分析能力不足以及程序中所固有的语义信息无法被并行化工具所理解而造成的.TIPS(Tsinghua interactive parallelizing system)系统通过提供一些友好的交互式工具,使用户与编译器紧密协作,是提高并行化系统的能力和效率的一条有效途径.     

面向CFD的交互式并行化系统Paractive

该文介绍了面向计算流体力学(CFD)的交互式并行化系统Paractive以及Paractive系统实现时所采用的关键技术。该文根据CFD程序及其并行化的特点提出了CFD程序并行化的区域计算模型以及区域相关的概念。区域计算模型将CFD程序看成区域操作的有序组合,区域相关则以区域操作为基本单位,区域操作本身所具有的对大块数据进行整体操作的特点,使得区域计算模型和区域相关非常适合开发CFD程序中蕴含的数据并行性。另外,该文还介绍了基于程序对象树的增量分析技术以及CFD程序并行化的静态性能预测技术,并在最后给出了使用Paractive并行化CFD实际算例的测试数据。     

并行化洪水演进模拟研究综述

近年来,并行化洪水演进模拟技术发展迅速,在防汛减灾领域发挥重要作用。在考虑洪水演进模型的数值方法、并行模式和编程技术等因素后,选取一些有代表性的洪水演进模型,分析了同构并行和异构并行洪水演进模型涉及的技术细节,提出并行化模型开发的技术难点和解决方法。最后,提出将来并行化洪水演进模型研发的着力点：非结构网格模型的异构并行化;混合并行的洪水演进模型;适于GPU异构并行的网格形式;并行环境下的实时可视化和交互式计算;基于动态编程语言的模型开发;界面式开发及模型应用推广。     

数值天气预报后处理系统的作业级并行实现

对于具体问题采取何种并行方法,达到好的并行效果,是在并行化工作中需要考虑的.通过对数值天气预报模式T511后处理系统的特点分析,结合其运行的计算机环境特点,提出了一种"作业级的并行实现"方法,减少T511后处理系统所需的计算墙钟时间.该方法具有并行化工作量小、易于实现、并行效果好等特点.     

有限差分法的并行化计算实现

有限差分法是求解偏微分方程近似解的一种重要的数值方法。串行算法并不能高效的解决大规模复杂计算问题,并行化计算方法可提高复杂计算问题的效率．从而使并行机上计算有限差分问题成为可能。二维场中拉普拉斯方的差分程格式非常适合并行化方法的计算,将串行部分并行化以提高大规模计算的效率具有重要的现实意义。MPI（消息传递接口）是实现并行程序设计的标准之一。虚拟进程（MPI_PROC_NULL）的引用简化了MPI编程中的通信部分,串行算法可更改为并行化计算方法,最终实现有限差分方法的并行化计算。     

有限差分法的并行化计算实现

有限差分法是求解偏微分方程近似解的一种重要的数值方法。串行算法并不能高效的解决大规模复杂计算问题,并行化计算方法可提高复杂计算问题的效率,从而使并行机上计算有限差分问题成为可能。二维场中拉普拉斯方程的差分格式非常适合并行化方法的计算,将串行部分并行化以提高大规模计算的效率具有重要的现实意义。MPI(消息传递接口)是实现并行程序设计的标准之一。虚拟进程(MPI_PROC_NULL)的引用简化了MPI编程中的通信部分,串行算法可更改为并行化计算方法,最终实现有限差分方法的并行化计算。     

有限差分法的并行化计算实现

有限差分法是求解偏微分方程近似解的一种重要的数值方法。并行化计算可提高复杂计算问题的效率,二维场中拉普拉斯方程的差分格式非常适合并行化方法的计算。如何将串行部分并行化以提高大规模计算的效率,MPI（消息传递接口）是实现并行程序设计的标准之一。虚拟进程（MPI_PROC_NULL）是MPI中的假想进程,它的引用可简化MPI编程中的通信部分,引入虚拟进程编写代码,可实现有限差分方法的并行化计算。     

基于SimOS的并行模拟平台的设计与实现

设计一个大规模并行处理系统，要求研制人员在设计阶段就尽可能多地掌握未来系统运行期间的各种信息。系统模拟逐渐成为人们获取系统信息的一种常用手段。模拟并行机器需要耗费的计算量与存储量随着模拟机器的规模增大和精度增加而大得惊人，因此要想跟上实际机器规模的发展，必须采用并行化的模拟方法。本文对并行化模拟方法进行了探讨。根据目前大规模并行系统的发展方向，设计并实现了一个并行模拟平台。     

面向语义属性查询的动画场景图像检索系统

设计并实现了一个交互式检索动画场景图像的系统。通过从动画场景图像中自动提出三种语义属性信息（类别、时间以及复杂度）,用户可以输入例如“一幅不复杂的傍晚的天空”作为查询来找到满意的检索结果。该系统是全自动的,并且是可扩展的。所有的语义属性特征计算在后台完成,从而保证系统的实时检索。此外,还设计了针对用户的几种不同交互式检索界面,极大地增强了检索的灵活性及正确性。     

Advanced tools and techniques for automatic parallelization

This paper describes some of the tools and techniques that are being used in the interactive SUPRENUM parallelization system SUPERB. Emphasis is placed on specific problems arising from the interactive nature of the system, in particular the necessity to incrementally update data flow information that is used to determine the applicability and the effect of transformations.     

Toward an automatic parallelization of sparse matrix computations

In this paper, we propose a generic method of automatic parallelization for sparse matrix computation. This method is based on both a refinement of the data-dependence test proposed by Bernstein and an inspector–executor scheme which is specialized to each input program of the compiler. This analysis mixes compilation process and run-time process.The sparsity of underlying data-structure determines a specific parallelism which increases the degree of parallelism of an algorithm. Such a source of parallelism had already been applied to many numerical algorithms such as the usual Cholesky factorization or LU-decomposition algorithms considered as the gold standards of parallelization based on sparsity. The standard automatic parallelization method cannot tackle such source of parallelism because it is based on the value of cells arrays and not merely on the memory addressing function.Addressing the automatization of this parallelism requires to develop a mixed compile-time and runtime approach integrated in a inspector–executor process. The compilation step provides a dedicated inspector devoted to the analyzed program. The inspector computes the dependence graph at runtime which allows a dynamic parallelization of the execution.As expressed just before, the generic scheme developed in this paper follows the design principles which have been applied, but at each time in an ad hoc way, to many sparse parallelization of numerical algorithms such as Cholesky algorithm. As far as we know, no general formal framework has been proposed to automate such a method of sparse parallelization. In this paper, we propose a generic framework of sparse parallelization (i.e. numerical program independent) which can be applied to any numerical programs satisfying the usual syntactic constraints of parallelization.     

基于交互信息的投机并行化方法*

针对投机并行化中如何权衡策略并确定合适的执行模型来获取理想性能的问题,提出了一种基于交互信息的投机并行化方法,利用交互信息来确定投机并行化的执行模型,建立相关评价模型,并着重从线程抽取创建角度提出了相应的策略及对应的性能评价。通过实验表明,基于交互信息进行“按需”并行化,可以达到所需的性能要求。     

Performance analysis of a parallel multi-view rendering architecture using light fields

Multiple view rendering is a common problem for applications where multiple users visualize a common dataset, as in multi-player games and collaborative engineering tools. For a system to be able to render a large number of views at interactive rates efficiently, parallel processing is an attractive technique. In this work, we present the implementation of a pipelined multiview light field renderer using a cluster with GPUs and MPI. We discuss the parallelization model and the problem of partitioning the tasks of the pipeline among the cluster machines based on the pipeline model and the costs of the stages. Our solution achieves 83% efficiency with ten machines, against only 11% efficiency of a naive parallelization.     

Web应用前后端融合的遗传算法并行化测试用例生成

Web应用测试用例生成并行化，是提升Web应用测试生成效率的一个有效手段.Web应用的前后端分离、事件驱动等特性，导致传统的并行化技术难以直接应用于Web应用的测试用例自动生成中.因此，如何针对Web应用进行并行化测试用例生成，是一项具有挑战性的工作.将种群并行化计算引入到基于遗传算法的Web应用前后端融合的测试用例生成中，通过线程池及调度逻辑、多浏览器进程管理及后端覆盖路径的获取，实现种群个体在多浏览器上的并行化执行及基于后端路径覆盖的适应度值并行化计算，以更高效地生成Web应用的测试用例.实验结果表明：相较于Web应用的GA串行化测试用例生成方法，所提的并行化测试生成方法能更充分地利用系统资源，极大地提升Web应用测试用例的生成效率.     

Interactive parallel programming using the ParaScope Editor

The ParaScope Editor, an intelligent interactive editor for parallel Fortran programs, which is the centerpiece of the ParaScope project, an integrated collection of tools to help scientific programmers implement correct and efficient parallel programs, is discussed. ParaScope Editor reveals to users potential hazards of a proposed parallelization in a program. It provides a variety of powerful interactive program transformations that have been shown useful in converting programs to parallel form. ParaScope Editor supports general user editing through a hybrid text and structure editing facility that incrementally analyzes the modified program for potential hazards. It is shown that ParaScope Editor supports an exploratory programming style in which users get immediate feedback on their various strategies for parallelization     

GPU-accelerated DXT and JPEG compression schemes for low-latency network transmissions of HD, 2K,and 4K video

Low-latency transmissions of high resolution video such as HD, 2K, or 4K over both Internet and private IP networks have found a foothold in many interactive applications, ranging from collaborative environments in science and medicine to the arts and entertainment industry. In this paper we demonstrate how the power of commodity graphics processing units can be used for efficient implementation of JPEG and DXT compression. We propose an approach to fine-grained parallelization of JPEG compression and the use of auxiliary indexes for efficient decompression, which are backward compatible with the JPEG standard. In-depth performance analysis is provided to show various aspects of the proposed parallelization including the dependency on image content and on various settings of the compression algorithm, as well as the impact of compression for interactive applications in terms of end-to-end latency. The applicability of these compression schemes in medicine and cinematography has also been evaluated using double-blind ABX tests compared with uncompressed video. We describe selected successful real-world deployment use cases based on our open-source implementation within the UltraGrid framework, such as trans-Atlantic 4K interactive video streaming during the CineGrid 2011 workshop. As discussed in the paper, the proposed approaches to parallelization provide sufficient performance even for the future generation of 8K video processing systems, currently limited by availability of camera and display hardware.     

渐进式智能回溯向量化代码调优方法

为了充分发挥高性能计算机的计算能力,缓解程序员设计和编写并行程序的压力,扩充可用软件集合,设计并实现了利用交互界面深入挖掘程序中的可向量化语句,优化生成代码中的向量化语句,提高生成代码的执行效率.该方法对充分发挥高性能计算机的计算能力,增强系统可用性和扩展应用范围具有重要的意义,同时能够提供有效的辅助手段和工具支持.渐进式智能回溯向量化代码调优架构通过对用户提交的串行程序进行程序分析和变换,采用串行程序分析、数据依赖分析、向量化分析等技术手段,根据分析结果对程序进行变换和优化,自动生成最终的向量化代码.该方法通过分析串行程序中潜在的并行性,将其自动变换为等价的向量化代码形式,大大简化了程序员的工作.     

Physics-based character skinning using multidomain subspace deformations

In this extended version of our Symposium on Computer Animation paper, we describe a domain-decomposition method to simulate articulated deformable characters entirely within a subspace framework. We have added a parallelization and eigendecomposition performance analysis, and several additional examples to the original symposium version. The method supports quasistatic and dynamic deformations, nonlinear kinematics and materials, and can achieve interactive time-stepping rates. To avoid artificial rigidity, or “locking,” associated with coupling low-rank domain models together with hard constraints, we employ penaltybased coupling forces. The multidomain subspace integrator can simulate deformations efficiently, and exploits efficient subspace-only evaluation of constraint forces between rotated domains using a novel Fast Sandwich Transform (FST). Examples are presented for articulated characters with quasistatic and dynamic deformations, and interactive performance with hundreds of fully coupled modes. Using our method, we have observed speedups of between 3 and 4 orders of magnitude over full-rank, unreduced simulations.     

Visual simulation of thermal fluid dynamics in a pressurized water reactor

We present a simulation and visualization system for a critical application—analysis of the thermal fluid dynamics inside a pressurized water reactor of a nuclear power plant when cold water is injected into the reactor vessel. We employ a hybrid thermal lattice Boltzmann method (HTLBM), which has the advantages of ease of parallelization and ease of handling complex simulation boundaries. For efficient computation and storage of the irregular-shaped simulation domain, we classify the domain into nonempty and empty cells and apply a novel packing technique to organize the nonempty cells. This method is implemented on a GPU cluster for acceleration. We demonstrate the formation of cold-water plumes in the reactor vessel. A set of interactive visualization tools, such as side-view slices, 3D volume rendering, thermal layers rendering, and panorama rendering, are provided to collectively visualize the structure and dynamics of the temperature field in the vessel. To the best of our knowledge, this is the first system that combines 3D simulation and visualization for analyzing thermal shock risk in a pressurized water reactor.