论文信息 - Matrix multiplication on the connection machine

Matrix multiplication on the connection machine

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.

S. Lennart Johnsson | Tim Harris | Kapil K. Mathur

[1] Charles L. Lawson,et al. Basic Linear Algebra Subprograms for Fortran Usage , 1979, TOMS.

[2] W. Daniel Hillis,et al. The connection machine , 1985 .

[3] Lennart Johnsson. Computational Arrays for Band Matrix Equations , 1981 .

[4] S. Lennart Johnsson. Data parallel programming and basic linear algebra subroutines , 1988 .

[5] Lennart Johnsson. VLSI algorithms for Doolittle's, Crout's, and Cholesky's methods , 1982 .

[6] Jack J. Dongarra,et al. LAPACK working note No. 10: Installing and testing the initial release of LAPACK Unix and non-Unix versions , 1989 .

[7] A. Nijenhuis. Combinatorial algorithms , 1975 .

[8] Sartaj Sahni,et al. Parallel Matrix and Graph Algorithms , 1981, SIAM J. Comput..

[9] S. Lennart Johnsson,et al. Communication Efficient Basic Linear Algebra Computations on Hypercube Architectures , 1987, J. Parallel Distributed Comput..

[10] Lennart Johnsson. HIGHLY CONCURRENT ALGORITHMS FOR SOLVING LINEAR SYSTEMS OF EQUATIONS , 1984 .

[11] Lynn Elliot Cannon,et al. A cellular computer to implement the kalman filter algorithm , 1969 .