Parallel Symbolic Computation on the Desk-Top
The efficient parallelization of common software still poses a considerable challenge in Computer Science. Symbolic software is particularly demanding with highly data-dependent and irregular parallel loads. However, parallelism may be the only area which offers much needed hardware support to symbolic computation, such as 10--100 times larger memories and greater processing power. Software which can effectively utilize modern parallel and networked workstations will be significantly faster in everyday use. In this talk we give an overview of the parallel system environment and some parallel algorithms in the PARSAC Computer Algebra library. In particular we cover shared memory multi-threading, high-level distributed memory computation and parallel Gröbner Basis completion techniques.
Wolfgang Küchlin
Universität Tübingen, Germany
kuechlin@informatik.uni-tuebingen.de