A parallel complex zero finder

A recent paper 7] describes a variable precision interval arithmetic algorithm for the computation of the zeros of an analytic function inside a given rectangle and to a user-speciied accuracy. The algorithm is based on the argument principle in the set of complex numbers C and carries much potential for parallelization at various levels of granularity. Here we explain how to modify the sequential algorithm to take advantage of parallelism at four levels ranging from coarse to medium grain. The algorithm is tested in a distributed environment consisting of eight SPARC workstations. The underlying software environment is also discussed.