hadoop1

QR算法是目前最广泛被用于求一般矩阵特征值的方法，但是由于串行算法的时间过于长，大大的降低了工作效率，所以我们将其在基于cuda架构上对其进行GPU加速，将原有的求矩阵特征值得串行算法进行并行化，大大地提高了工作效率

利用Python实现特征的MinMax标准化和Z-Score标准化两种特征归一化方法。并用原特征、两种归一化特征分别进行模型训练，比较模型性能。(Two normalization methods of MinMax normalization and Z-Score standardization are implemented by Python. The model is trained by the original feature and two normalization features respectively, and the performance of the model is compared.)

说明：  N体问题，它是指在已知N个物体的初始位置、速度和质量的情况下，在牛顿经典力学情况下研究它们的运动，包括轨迹预测等。(Back to the specific N-body problem, it refers to the study of the motion of n-bodies in Newtonian classical mechanics, including trajectory prediction, when the initial position, velocity and mass of n-bodies are known.)

说明：  MPI的工具 MPI_tuning_ClusterTools Selective_Tracing_and_Filtering.zip(MPI tools MPI_tuning_ClusterTools Selective_Tracing_and_Filtering.zip)

设计MPI程序模拟广播（MPI_Bcast）操作：将每个MPI进程按照所在节点名称建立node通信子域分组；再将各个node子通信域的0号进程再次组成一个名为head的通信域；在进行广播时，首先，由root进程将消息在head通信子域内广播，然后，再由head子域内各进程在其所在的node子域内进行广播。(Design the MPI program MPI_Bcast operation: Each MPI process establishes a node communication sub-domain grouping according to the node name; then, each process of the 0 node of each node sub-communication domain is again formed into a communication domain named head; At first, the root process broadcasts the message in the head communication sub-domain, and then the processes in the head sub-domain broadcast in the node sub-domain where it is)

Copyright 1997, Regents of the University of Minnesota. METIS was written by George Karypis (karypis@cs.umn.edu) Introduction ------------------------------------------------------------ METIS is a software package for partitioning unstructured graphs, partitioning meshes, and computing fill-reducing orderings of sparse matrices. The documentation of METIS can be found in the Doc/manual.ps file. METIS is written in ANSI C and should compile on Unix systems that have a ANSI C compiler. It has been extensively tested on the following architectures:    AIX    SunOS    Solaris    IRIX    Unicos    Linux    HP-UX    FreeBSD

secuential归并排序

Mapreduce实现的TF-IDF词频统计分析，可以直接运行于HADOOP环境下，适合初学者。(Realization Mapreduce TF-IDF, word frequency statistics, can be run directly in the the under HADOOP environment, suitable for beginners.)

消息传递并行编程环境MPI_北京应用物理与计算数学研究所_莫则尧(message passing parallel programming environments MPI Beijing applied physics and computational mathematics Institute _ were Mo Yao)

OpenCL1.2版本的扩展功能用户手册。(OpenCL1.2 version of the Expansion Function User' s Manual.)