Efficient simulation of large materials clusters using the jaguar quantum chemistry program: Parallelization and wavefunction initialization

An outline of the improvements to the pseudospectral electronic structure program Jaguar is presented, showing efficient and robust performance of hybrid‐DFT calculations for large systems with thousands of basis functions, focusing on materials applications. The improvements include re‐engineered parallelization, the design of a fragment‐based initial guess generation method, and the validation of small eigenvalue cutoff values. An OpenMP/MPI hybrid parallelization has been implemented for the pseudospectral algorithm, which extends Jaguar's scalability to up to 256 cores in tests of TiO₂ clusters with 1295‐4961 basis functions. In the largest test case, the code delivers 84.4× speedup for 128 cores in total calculation time. In addition, a fragment‐based initial guess method has been constructed for large systems containing many transition metals, where the conventional (atomic) approach often fails. Overall, Jaguar is now capable of efficiently and robustly performing hybrid‐DFT geometrical optimizations for large systems with more than 600 atoms in reasonable runtimes. © 2015 Wiley Periodicals, Inc.     

基于网络编码的机会网络高效路由算法

针对采用Epidemic机制的机会网络路由算法在数据分组传送阶段存在通信冗余的问题,提出了一种基于网络编码的高效路由算法--NCBER(network-coding-based epidemic routing）。NCBER在传送数据分组的过程中使用主动异或网络编码和多播,并取消了 Request（请求）控制分组,从而减少数据分组的转发次数和控制分组数量,降低网络开销,缩短分组传输时延。理论分析和仿真结果表明,NCBER 算法在网络开销和数据分组端到端时延性能方面优于经典的Epidemic路由算法及其改进算法MRRMR（message redundancy removal of multi-copy routing）,并且使数据传送成功率保持在100%。     

Shared-memory parallelization of the TURBOMOLE programs AOFORCE, ESCF, and EGRAD: how to quickly parallelize legacy code

The programs ESCF, EGRAD, and AOFORCE are parts of the TURBOMOLE program package and compute excited-state properties and ground-state geometric hessians, respectively, for Hartree-Fock and density functional methods. The range of applicability of these programs has been extended by allowing them to use all CPU cores on a given node in parallel. The parallelization strategy is not new and duplicates what is standard today in the calculation of ground-state energies and gradients. The focus is on how this can be achieved without needing extensive modifications of the existing serial code. The key ingredient is to fork off worker processes with separated address spaces as they are needed. Test calculations on a molecule with about 80 atoms and 1000 basis functions show good parallel speedup up to 32 CPU cores.     

Non-unitary precoder design for reducing feedback overhead in MIMO-OFDM systems

The precoder feedback overhead in MIMO-OFDM systems is heavy as the number of subcarriers is usually large. Most of the current limited feedback works focus on unitary precoding schemes. Based on the fact that the channel matrices of neighboring subcarriers in MIMO-OFDM systems are correlated, an alternative transceiver is proposed here. In this paper, non-unitary precoders shared by several adjacent subcarriers, and the corresponding individual equalizers are jointly designed under partial CSI. As the number of precoding matrices is smaller, the feedback overhead can be reduced. The transceivers are designed to minimize the detection mean-squared error (MSE) under the total transmit power constraint. A convergent iterative algorithm based on the Lagrange multipliers method is proposed. The necessary conditions for an optimal transceiver, Karush–Kuhn–Tucker (KKT) conditions, are satisfied in each iteration. Numerical results under the frequency-selective MIMO-OFDM channel, which is generated by the classical Zheng and Xiao's model, prove that the proposed transceiver can significantly reduce the feedback overhead without severe performance degradation.     

基于GMPLS光网络的分布式多层故障定位方法

在GMPLS光网络中,为了在故障定位时减少定位数障据链路故障的信令开销,避免不必要的网络资源浪费,降低网络资源的阻塞率,提出了一种分布式多层故障定位方法。该方法在现有的单层故障定位方案的基础上,通过双向数据链路故障通知的方法,避免了一些不必要的故障相关操作,减少了网络节点的负担,提高了网络资源的利用率。     

光码标签交换网络边缘节点FEC的研究

构建了光码标签交换网边缘节点的结构,结合光码分复用技术提出了转发等价类(FEC)分组结构的设计方案。分组结构适合变长数据分组的传输,在核心节点能够实现对标签信息的可并行处理,提高了处理效率。在此基础上,针对开销比对网络传输性能的影响,对FEC结构中的各参量与开销比的关系进行了仿真研究;采用了固定时隙算法对数据分组进行封装,对到达率和封装时延与分组丢失率的关系进行仿真分析。     

密度矩阵重正化群的异构并行优化

密度矩阵重正化群方法（DMRG）在求解一维强关联格点模型的基态时可以获得较高的精度,在应用于二维或准二维问题时,要达到类似的精度通常需要较大的计算量与存储空间.本文提出一种新的DMRG异构并行策略,可以同时发挥计算机中央处理器（CPU）和图形处理器（GPU）的计算性能.针对最耗时的哈密顿量对角化部分,实现了数据的分布式存储,并且给出了CPU和GPU之间的负载平衡策略.以费米Hubbard模型为例,测试了异构并行程序在不同DMRG保留状态数下的运行表现,并给出了相应的性能基准.应用于4腿梯子时,观测到了高温超导中常见的电荷密度条纹,此时保留状态数达到10⁴,使用的GPU显存小于12 GB.     

ARM架构下计算机图像并行化处理技术研究

为在图像处理与分析时具备良好的视觉效果,提高图像处理的速度,需要对ARM架构下计算机图像并行化处理技术进行研究。当前采用的方法是对各种变换频域图像特征提取与计算机图像集合特征的提取进行相结合,克服了当前对图像进行提取时存在图像形状描述的缺陷,提取图像特征向量维数相对较低。实验表明,通过对图像进行特征提取能很好的对图像效果进行展示,将图像的纹理特征进行详细的表述,将该方法应用到图像处理技术当中,具有良好的去噪效果及扩展性,该方法过程简单,但存在图像视觉效果较差的问题。为此,提出一种ARM架构下计算机图像并行化处理技术研究方法。该方法首先利用非局部均值去噪算法对图像进行去噪处理,然后结合图像去噪的结果利用小波变换对去噪图像进行边缘检测,最后采用非线性增强算法对图像进行增强完成对ARM架构下计算机图像并行化处理技术研究。实验结果表明,所提方法不仅提高图像处理速度,还提高图像视觉效果,具有广泛的应用价值。     

Adaptive Distributed Channel Assignment in Wireless Mesh Networks

From the channel assignment schemes in time division multiple access (TDMA) slot assignment protocols developed in previous studies, we have found that these protocols do not have a convenient frame length reduction scheme after the expansion of the frame length. As the network size grows, the frame length expands rapidly, particularly when we set it as a power of two. A very long frame may result in poor channel utilization when it contains many unused slots. In this paper, we consider a wireless mesh network with a TDMA structure. We divide each code into time slots to form the channels. We propose a dynamic frame length expansion and recovery method termed adaptive distributed channel assignment (ADCA). This strategy is designed to enhance the utilization of the available channels by taking advantage of the spatial reuse concept. In ADCA, an increase in spatial reuse is achieved by adding a certain amount of control overhead. We show that the bandwidth saved due to the channel spatial reuse is larger than the additional bandwidth spent on the control overhead.

密度矩阵重正化群的异构并行优化

密度矩阵重正化群方法（DMRG）在求解一维强关联格点模型的基态时可以获得较高的精度,在应用于二维或准二维问题时,要达到类似的精度通常需要较大的计算量与存储空间.本文提出一种新的DMRG异构并行策略,可以同时发挥计算机中央处理器（CPU）和图形处理器（GPU）的计算性能.针对最耗时的哈密顿量对角化部分,实现了数据的分布式存储,并且给出了CPU和GPU之间的负载平衡策略.以费米Hubbard模型为例,测试了异构并行程序在不同DMRG保留状态数下的运行表现,并给出了相应的性能基准.应用于4腿梯子时,观测到了高温超导中常见的电荷密度条纹,此时保留状态数达到10⁴,使用的GPU显存小于12 GB.