Go backward to Parallel Resultant Computation Go up to Top Go forward to References |
The parallelization of multivariant resultant computation in Distributed Maple
on a heterogenous network of PCs, workstations, and multiprocessors has shown
very promising results. This algorithm is however only one (if major)
component of the CASA function pacPlot
whose speedup is our ultimate
goal. We are therefore working on the parallelization of real root isolation
and the seamless integration of the different parallel algorithms and
parallelization layers such that the resulting code can be efficiently
executed in different system environments.
Our future parallelization efforts will extend to other parts of the CASA
library such as the function ssiPlot
for plotting intersections of
space surfaces [11] and the powerful CASA facilities for analyzing and
manipulating algebraic sets. This will also give helpful insights for further
improvements of the Distributed Maple system especially with respect to load
balancing in an heterogeneous network. Combining different parallel
algorithms and tuning them for different application contexts and system
environments will present an interesting challenge.