There are discussed implementational aspects of the special-purpose computer algebra system FELIX designed for computations in constructive algebra. In particular, data types developed for the representation of and computation with commutative and non-commutative polynomials are described. Furthermore, comparisons of time and memory requirements of different polynomial representations are reported.

The special-purpose computer algebra system FELIX is designed for computations in constructive commutative and non-commutative algebra. In this paper we discuss some features of the system supporting the computation of rather complex problems, especially standard basis computations, using standard hardware. There is a frst aspect concerning the definition and implementation of the basic data types which should be a good compromise between space and time efficient representations of the algebraic objects. Usually, rather complex computations are very time consuming (up to weeks) and often require several attempts. So, there are included special session saving methods in FELIX which allows to backup the attained intermediate results in form of memory images into special session files and to restart later on. Finally, we describe our efforts crunching complex problems by parallelization. The implemented interface is based on stream sockets and includes a special protocol for the data exchange. It supports the distributed computation on heterogeneous, loosely coupled systems.