Cover Image

Parallel scientific computation : a structured approach using BSP /

Show other versions (1)

Bisseling explains how to use the bulk synchronous parallel (BSP) model and the freely available BSPlib communication library in parallel algorithm design and parallel programming. An appendix on the message-passing interface (MPI) discusses how to program using the MPI communication library.

Bibliographic Details
Main Author: Bisseling, Rob H. (Author)
Format: eBook
Language:English
Published: Oxford : Oxford University Press, 2020.
Edition:Second edition.
Subjects:
Parallel processing (Electronic computers) > Scientific applications.
Supercomputers > Scientific applications.
Parallel computers.
Parallélisme (Informatique) > Applications scientifiques.
Superordinateurs > Applications scientifiques.
Ordinateurs parallèles.
Parallel computers
Supercomputers > Scientific applications
Electronic books.
Online Access:Connect to the full text of this electronic book
Table of Contents:
  • Cover
  • Parallel Scientific Computation: A Structured Approach Using BSP
  • Copyright
  • PREFACE
  • ACKNOWLEDGEMENTS
  • ABOUT THE AUTHOR
  • CONTENTS
  • 1 Introduction
  • 1.1 Parallel computing is everywhere
  • 1.2 The BSPmodel
  • 1.3 BSP algorithm for inner product computation
  • 1.4 Starting with BSPlib: example program bspinprod
  • 1.5 BSP benchmarking
  • 1.6 Example programbspbench
  • 1.7 Benchmark results
  • 1.8 Sorting
  • 1.9 Example function bspsort
  • 1.10 Experimental results for samplesort on a Cartesius node
  • 1.11 Bibliographic notes
  • 1.11.1 BSP-related models of parallel computation
  • 1.11.2 BSP libraries
  • 1.11.3 The non-BSP world: message passing and threads
  • 1.11.4 Benchmarking
  • 1.11.5 Sorting
  • 1.12 exercises
  • 2 LU decomposition
  • 2.1 The problem
  • 2.2 Sequential LU decomposition
  • 2.3 Basic parallel algorithm
  • 2.4 Two-phase broadcasting and other improvements
  • 2.5 High-performance LU decomposition
  • 2.6 Example function bsplu
  • 2.7 Experimental results on the Cori supercomputer
  • 2.8 Bibliographic notes
  • 2.8.1 Matrix distributions
  • 2.8.2 Collective communication
  • 2.8.3 Parallel matrix computations
  • 2.9 exercises
  • 3 The fast Fourier transform
  • 3.1 The problem
  • 3.2 Sequential recursive fast Fourier transform
  • 3.3 Sequential nonrecursive algorithm
  • 3.4 Parallel algorithm
  • 3.5 Weight reduction
  • 3.6 Example function bspfft
  • 3.7 Experimental results on the Cartesius supercomputer
  • 3.8 Bibliographic notes
  • 3.8.1 Sequential FFT algorithms
  • 3.8.2 Parallel FFT algorithms
  • 3.8.3 Applications
  • 3.9 exercises
  • 4 Sparse matrix-vector multiplication
  • 4.1 The problem
  • 4.2 Sparsematrices and their data structures
  • 4.3 Parallel algorithm
  • 4.4 Cartesianmatrix distribution
  • 4.5 Mondriaan distribution for general sparsematrices
  • 4.6 Fine-grain and medium-grainmatrix distribution
  • 4.7 Vector distribution
  • 4.8 Random sparsematrices
  • 4.9 Laplacian matrices
  • 4.10 Parallel algorithm for hybrid-BSP
  • 4.11 Example function bspmv
  • 4.12 Experimental results on the Cartesius supercomputer
  • 4.13 Bibliographic notes
  • 4.13.1 Sparse matrix computations
  • 4.13.2 Parallel sparse matrix-vector multiplication algorithms
  • 4.13.3 Partitioning methods
  • 4.14 exercises
  • 5 Graph matching
  • 5.1 The problem
  • 5.2 Sequential algorithm
  • 5.3 Suitors and sorting
  • 5.4 Parallel algorithm
  • 5.5 Correctness
  • 5.6 Tie-breaking
  • 5.7 Load balancing
  • 5.8 Further improvements
  • 5.9 Example function bspmatch
  • 5.10 Experimental results on the Cartesius supercomputer
  • 5.11 Bibliographic notes
  • 5.11.1 Sequential graph matching
  • 5.11.2 Parallel graph matching
  • 5.11.3 GraphBLAS
  • 5.12 exercises
  • APPENDIX A: AUXILIARY BSPEDUPACK FUNCTIONS
  • A.1 Header file bspedupack.h
  • A.2 Utility file bspedupack.c
  • APPENDIX B: A QUICK REFERENCE GUIDE TO BSPLIB
  • REFERENCES
  • INDEX