Fast Parallel Hermite Normal Form Computation of Matrices over F[x]

Clemens Wagner

Abstract
We present an algorithm for computing the Hermite normal form of a polynomial matrix and an unimoular transformation matrix on a distributed computer network. We provide an algorithm for reducing the off-diagonal entries which is a combination of the standard algorithm and the reduce off-diagonal algorithm given by Chou-Collins. We provide a technique for producing small multiplier matrices if the input matrix is not full-rank, and give an upper bound for the degrees of the entries in the multiplier matrix. All described algorithms have been implemented in C/C++, and we give run-time examples for HNF computation of matrices over $F_2[x]$
Contact
Clemens Wagner
Rechenzentrum,Universitaet Karlsruhe,Zirkel 2,76128 Karlsruhe/Germany
rz15@nic.de