Matrix multiplication on the connection machine

Authors:
S. L. Johnsson;T. Harris;K. K. Mathur
Affiliations:
Thinking Machines Corp., 245 First Street, Cambridge, MA and Department of Computer Science, Yale University, New Haven, CT;Thinking Machines Corp., 245 First Street, Cambridge, MA;Thinking Machines Corp., 245 First Street, Cambridge, MA
Venue:
Proceedings of the 1989 ACM/IEEE conference on Supercomputing
Year:
1989

Citing 6
Cited 3

The connection machine

The connection machine
Communication effect basic linear algebra computations on hypercube architectures

Journal of Parallel and Distributed Computing
Data parallel programming and basic linear algebra subroutines

Mathematical aspects of scientific software
Basic Linear Algebra Subprograms for Fortran Usage

ACM Transactions on Mathematical Software (TOMS)
A cellular computer to implement the kalman filter algorithm

A cellular computer to implement the kalman filter algorithm
Combinatorial Algorithms: Theory and Practice

Combinatorial Algorithms: Theory and Practice

HPFBench: a high performance Fortran benchmark suite

ACM Transactions on Mathematical Software (TOMS)
An Evaluation of High Performance Fortran Compilers Using the HPFBench Benchmark Suite

Euro-Par '99 Proceedings of the 5th International Euro-Par Conference on Parallel Processing
An object-oriented parallel programming language for distributed-memory parallel computing platforms

Science of Computer Programming

Quantified Score

Hi-index	0.00

Visualization

Abstract

A data parallel implementation of the multiplication of matrices of arbitrary shapes and sizes is presented. A systolic algorithm based on a rectangular processor layout is used by the implementation. All processors contain submatrices of the same size for a given operand. Matrix-vector multiplication is used as a primitive for local matrix-matrix multiplication in the Connection Machine system CM-2 implementation. The peak performance of the local matrix-matrix multiplication is in excess of 20 Gflops s-1. The overall algorithm including all required data motion has a peak performance of 5.8 Gflops s-1.