面向通讯同步的多处理器阵列重构

从多处理器阵列中获取所需大小并且同步通讯性能优良的子阵列,是高性能拓扑重构的核心问题之一。基于不同的逻辑列剔除策略提出了3种面向通讯同步的拓扑重构算法:基于分治思想剔除逻辑列的重构算法(SCA_01),该算法能够使被优化的逻辑列相对均匀地分布在物理阵列中;优先剔除长逻辑列的贪心重构算法(SCA_02),该算法能够使被优化的逻辑列的长链接总数最少;基于分治与长链接数的混成重构算法(SCA_03),该算法将某一区域内的最长逻辑列剔除,且尽可能将剩余逻辑列均匀分布在物理阵列中。同时,对逻辑阵列的最大通讯延时给出了下界的求解算法。实验结果表明,3种算法在故障率小于1%、逻辑列的剔除率超过20%时,算法重构出的逻辑阵列的通讯延时特别接近计算出的性能下界。在多数情况下SCA_01优于SCA_02和SCA_03,而后两者的性能相近。在小阵列上且故障率与剔除率较小时,SCA_02具有性能优势,但在大阵列上SCA_03具有优势。在32×32的阵列上,SCA_01构造的阵列产生的通讯延时较SCA_02和SCA_03产生的延时平均减少25%,并且运行速度也提升了19.4%。     

容错处理器阵列的多逻辑列并行重构算法

处理器阵列的容错重构技术是片上网络多核、众核高性能体系结构的可靠性技术之一。现有的最大逻辑阵列并行重构技术仅对单条逻辑列的构造实现了并行化,而对多条逻辑列的同步并行仍未见可行算法。依据处理器阵列的潜在并行性,在分治策略的基础上,提出了一种阵列分块的并行重构算法。算法对处理器阵列实施横向分块划分,对每个阵列块进行并行重构,并对所得逻辑子阵列进行归并,实现了多条逻辑列的同步并行重构。与现有的并行算法相比,新算法同样能够生成最大逻辑列,并且减少了通信开销与计算中的数据冗余,有效提高了运行速度。实验结果表明,在物理阵列大小为64×64的处理器阵列上,运行速度比现有并行算法提高39.55%,并且具有良好的可扩展性。     

容错处理器阵列的并行重构及VHDL实现

网格连接的处理器阵列是一种应用广泛的高性能体系结构,而容错处理器阵列的重构技术是近年来的研究热点之一.现有的研究多数集中在串行重构算法上,忽视了该结构重构时内在的可并行性.本文根据阵列结构的特点设计了一种基于VHDL语言的重构算法,该算法从第一行的各个无故障处理器单元同时向下选路,具有潜在的并行性,.实验结果表明,与现有的串行算法相比,本文提出的并行算法同样能够生成最大规模的目标阵列并且当物理阵列大小为48×48,本文提出的并行算法加速重构将近20倍.     

环网处理器阵列的容错重构技术

高效的容错技术对于提高多处理器系统的可靠性至关重要。环网(Torus)是连接多处理器阵列的重要网络结构,而环网处理器阵列上的容错重构技术目前尚属空白。针对环网阵列的特殊连接方式,将环网阵列重构问题转化为矛盾图上求解最大独立集问题。矛盾图上的结点表示故障处理器的替换方案,而边代表了不同替换方案之间的不可共存特性。主要是根据三种不同的冗余处理器分布方案,设计生成矛盾图算法,求解最大独立集算法,以及由独立集生成逻辑处理器阵列算法,取得了令人满意的结果。实验结果表明,当阵列规模较小或故障率较低时,一行一列和十字型的冗余单元分布的重构能力较好;而随着阵列规模或故障率的增大,三种冗余单元分布策略的重构成功率都随之下降,但可通过增加冗余单元以及调整冗余分布来改善容错效果。此外,从实验结果中还可以看出,环网处理器阵列的容错能力显然优于网格(Mesh)处理器阵列。     

流水线配置技术在可重构处理器中的应用

提出一种应用于可重构处理器中的流水线配置技术,能够有效减低配置时间,提高应用程序的执行速度。可重构处理器包括通用处理器和一个粗颗粒度的可重构阵列。可重构阵列将处理应用中占据大量执行时间的循环,这些循环将被分解为不同的行在阵列上以流水线的方式执行。该技术在FPGA验证系统上得到了验证。验证的应用包括H.264基准中的整数离散余弦变换和运动估计。相比传统的可重构处理器PipeRench, MorphoSys以及TI的DSP TMS320DM642有大约3.5倍的性能提升。     

基于GReP通用可重构处理器的密码算子优化设计

以提升通用可重构处理器在信息安全应用领域的处理能力与执行效率为目的,对序列、分组、公钥及哈希函数等四大类32种密码算法进行深入分析。通过各算法的热点、频度分析,对基本处理单元进行同构、同态、参数化设计,提出了对密码算法性能影响大、复用性好且具有可重构特性的密码算子。以公钥密码算法中计算量大、实现复杂且被广泛使用的模乘算法为代表,提出了将密码算子映射到GReP（General-Purpose Reconfigurable Processor）通用可重构处理器上的方法。实验表明基于GReP通用可重构处理器架构实现的模乘算法其运行效率比Intel CORETM i7平台上实现效率提高60%左右。实验充分说明了GReP通用可重构处理器在提高处理能力与执行效率方面有明显优势。     

基于可重构处理器的并行优化算法

为挖掘可重构处理器的内在并行性,需要编译器通过分析程序的并行性来决定可重构处理器硬件最好的执行模式。为此,提出一种基于可重构处理器的并行优化算法。将有向无环图的并行计算部分映射到可重构处理器上,对任务实现3个不同层次的并行性(指令级并行、循环级并行、线程级并行)。测试结果表明,该算法使得可重构处理器在处理任务时比未用并行优化算法的性能提升1.2倍左右。     

一种利用部分重构技术实现DES算法的方法

针对当前利用可重构计算技术实现DES算法的方法中存在重构性能低和资源占用量大等缺陷,提出了一种利用基于模块的部分重构技术实现DES算法的方法.该方法利用DES具有的对合结构特性进行算法的模块划分,解决了算法部分重构时的模块间通讯和时序调整等关键问题.通过对DES算法的不同实现方案进行对比,验证了该方法的可行性和有效性.     

阵列众核结构上的一种多层分区Hash连接算法

连接是数据查询处理中最耗时、使用最频繁的操作之一,对提高连接操作的速率具有重要意义。阵列众核处理器是一类重要的众核处理器,具有强大的并行能力,可用来加速并行计算。基于阵列众核处理器的结构,设计和优化了一种高效的多层分区Hash连接算法。该算法通过多层划分的策略大大降低了主存访问次数,通过分区重排方法有效消除了数据倾斜的影响,获得了很高的性能。在异构融合阵列众核处理器DFMC(Deeply-Fused Many Core)原型系统上的实验结果表明,DFMC上多层分区Hash连接算法的性能是CPU-GPU耦合结构上最快的连接算法的8.0倍,表明利用阵列众核处理器加速数据查询应用具有优势。     

RAPWBN的矩阵乘法并行算法

在介绍带有宽总线网络的可重构计算阵列(RAPWBN)的基本结构及其二进制值的前缀和操作的基础上,提出了 RAPWBN 阵列上的整数求和算法,并由此得到了 RAPWBN 阵列上的两种快速高效的矩阵乘法运算并行算法。在具有 N3个处理器和 N2条行总线的 RAPWBN 阵列上,若总线带宽ω>logN 字节,矩阵乘法可以在 O(1)时间完成;在具有 N2个处理器和 N 条行总线的 RAPWBN 阵列上,矩阵乘法可以在 O(N)时间完成。它们的效率都为 O(N3),达到了最优。     

Efficient reconfiguration algorithms for communication-aware three-dimensional processor arrays

Homogeneous processor arrays are emerging in tera-scale computation and effective fault tolerance techniques are essential to improving the reliability of such complex integrated circuits. We study the degradable processor arrays to achieve fault tolerance by employing reconfiguration. Three bypass schemes and three rerouting schemes are proposed to reconfigure three-dimensional processor arrays with defective processors to achieve target arrays without faults. A heuristic algorithm is proposed to construct a target array on the selected rows and columns. It is also proved that the proposed greedy plane rerouting algorithm (GPR) produces maximum target array. In addition, the problem of constructing the communication efficient array is considered in this paper. An algorithm is proposed to refine the communication among processors within the target array constructed by GPR. Experimental study shows that the proposed algorithm GPR produces target arrays with higher harvest and lower degradation on the host arrays with fault density no more than 5%. In addition, the communication performance is significantly optimized by reducing the number of long interconnects, and the average improvement is about 34% for all cases considered in this paper.     

Minimizing interconnect length on reconfigurable meshes

Shorter total interconnect and fewer switches in a processor array definitely lead to less capacitance, power dissipation and dynamic communication cost between the processing elements. This paper presents an algorithm to find a maximum logical array (MLA) that has shorter interconnect and fewer switches in a reconfigurable VLSI array with hard/soft faults. The proposed algorithm initially generates the middle ([k/2]^th) logical column and then makes it nearly straight for the MLA with k logical columns. A dynamic programming approach is presented to compact other logical columns toward the middle logical column, resulting in a tightly-coupled MLA. In addition, the lower bound of the interconnect length of the MLA is proposed. Experimental results show that the resultant logical array is nearly optimal for the host array with large fault size, according to the proposed lower bound.     

Minimizing interconnect length on reconfigurable meshes

Shorter total interconnect and fewer switches in a processor array definitely lead to less capacitance, power dissipation and dynamic communication cost between the processing elements. This paper presents an algorithm to find a maximum logical array (MLA) that has shorter interconnect and fewer switches in a reconfigurable VLSI array with hard/soft faults. The proposed algorithm initially generates the middle (⌊k/2⌋ ^th ) logical column and then makes it nearly straight for the MLA with k logical columns. A dynamic programming approach is presented to compact other logical columns toward the middle logical column, resulting in a tightly-coupled MLA. In addition, the lower bound of the interconnect length of the MLA is proposed. Experimental results show that the resultant logical array is nearly optimal for the host array with large fault size, according to the proposed lower bound.     

基于杂交遗传算法的多处理器硬实时容错调度算法

传统的硬实时容错调度算法获得了较好的容错性能,但其任务拒绝率、处理器分配偏差比例以及最早完成时间等性能参数不佳,对此提出一种基于杂交遗传算法的优化方案,并对传统的硬实时容错算法进行优化。采用了中心型调度模型,并采用了任务备份方案来实现容错能力。将任务拒绝率、处理器分配偏差比例以及最早完成时间三个优化参数作为遗传算法适应度目标函数的三个带权分量,对其进行优化,通过遗传算法的杂交与迭代计算获得了优化的结果。最终使用不同的任务数量与处理器数量的组合对本算法与传统算法进行对比试验,结果可看出本算法的3个优化参数明显优于传统算法,且总适应度值亦比传统算法有明显改进。     

An Efficient Communication Scheduling Method for the Processor Mapping Technique Applied Data Redistribution

Array redistribution is usually required for more efficiently executing a data-parallel program on distributed memory multi-computers. In performing array redistribution using synchronous communication mode, data communications among the processors should be properly arranged to avoid incurring higher data transfer cost. Some efficient communication scheduling methods for the Block-Cyclic redistribution have been proposed. On the other hand, the processor mapping technique can help reduce the data transfer cost of redistribution. To avoid degrading the benefit of data transfer cost reduction, it is needed to construct optimal communication schedules for the redistribution in which the processor mapping technique is applied. In this paper, we present a unified approach to constructing optimal communication schedules for the processor mapping technique applied Block-Cyclic redistribution. The proposed method is founded on the processor mapping technique and can more efficiently construct the required communication schedules than other optimal scheduling methods.     

New Processor Array Architectures for the Longest Common Subsequence Problem

A longest common subsequence (LCS) of two strings is a common subsequence of two strings of maximal length. The LCS problem is to find an LCS of two given strings and the length of the LCS (LLCS). In this paper, we present a new linear processor array for solving the LCS problem. The array is based on parallelization of a recent LCS algorithm which consists of two phases, i.e. preprocessing and computation. The computation phase is based on bit-level dynamic programming approach. Implementations of the preprocessing and computation phases are discussed on the same processor array architecture for the LCS problem. Further, we propose a block processor array architecture which reduces the overall communication and time requirements. Finally, we develop a performance model for estimating the performance of the processor array architecture on Pentium processors.     

一种Malek模型下的系统故障诊断算法

在多机系统中,各个结点（处理器）在通信过程中极易发生故障,因此选择有效的诊断算法,快速、准确地判断出系统故障集十分重要。传统的PMC模型以结点相互测试的结果为基础,而故障结点的测试结果不唯一,导致该模型诊断结果相对不稳定。针对这种情况,采用Malek诊断模型代替传统的PMC模型,借助遗传算法特性,将复杂的网络拓扑图简化为二进制编码,并按照适应度函数值确定种群搜索方向,提高搜索效率。该算法根据Malek模型设计约束方程,提出新的适应度函数,优化变异算子。实验表明,算法改进后,缩短了判断故障集所需的CPU时间,同时,算法根据故障症候判断出目标故障集的概率更高,从而证明了用Malek模型代替PMC模型的高效性。