A new parallel solver for ODEs implementing a ''parallelism across the
steps'' has been recently proposed in (Amodio and Brugnano, 1997; Bru
gnano and Trigiante, 1998), where it is shown that it is able to obtai
n an almost perfect speed-up on linear problems, and given mesh. A pos
sible way to adapt this algorithm to efficiently handle nonlinear init
ial value problems has been studied in the companion paper (Brugnano a
nd Trigiante, this issue). The corresponding algorithm is here analyze
d in details, in order to show its parallel efficiency. Numerical test
s on a distributed memory parallel computer are also included. (C) 199
8 Published by Elsevier Science B.V. and IMACS. All rights reserved.