Implementing linear algebra algorithms on a Meiko Computing Surface (Q1181509)

The paper deals with implementation of some linear algebra algorithms on a transputer-based parallel computer system of Meiko. The vector plus scalar time vector (SAXPY) scalar product of two vectors (SDOT) and elementwise division of two vectors (VECDIV) operations are tested from the point of view of basic performance parameters. From the algorithms for solving linear systems of algebraic equations, the problems with respectively bidiagonal and tridiagonal matrices are chosen. Further, the block variant of the matrix-vector and transposed matrix-vector product is examined for a square mesh of processors. The contribution of the paper is to present detailed performance results of some concrete linear algebra computations on a distributive memory computer architecture consisting of 64 T800 transputers.