Polynomial Homotopies on Multicore Workstations

Abstract:

Homotopy continuation methods to solve polynomial systems scale very well on parallel machines. In this talk we examine its parallel implementation on multiprocessor multicore workstations, using threads. Preliminary timings indicate good speedups for basic pleasingly parallel path tracking jobs. The use of multitasking will lead to more efficient parallel implementations in a multi-tiered approach. Developing parallel code for complicated homotopies on multicore workstation is more convenient with threads than with MPI.

14th SIAM Conference on Parallel Processing for Scientific Computing. MS46 High-Performance Symbolic Computing. Seattle, Washington, 26 February 2010.

slides of the talk