Matrix-vector multiplication: parallel algorithms and architectures

We present an efficient parallel implementation of matrix-vector multiplication on a binary tree, whose leaves are connected to local memories, each containing one column of the matrix. The performance attained can be favourably compared with the one of the mesh of trees and the linear array. Further, we analyze the case of architectures with a fixed number of processors. Finally, some results concerning the parallel implementation of iterative methods for the solution of linear systems and for eigenvalue computations are described.