extended precision path tracking in parallel
Jan Verschelde and
Genady Yoffe
Abstract:
To compensate for the overhead of extended precision
(in particular: of double double and quad double arithmetic),
we investigated the use of multiple cores on regular processors
and general purpose graphics processing units.
Because the cost of evaluating and differentiating polynomials
often dominates the computational work in path tracking,
we developed and implemented
multithreaded algorithms on multicore processors
and massively parallel algorithms for general purpose GPUs
for multivariate polynomial evaluation and differentiation.
2012 SIAM Annual Meeting.
Minisymposium MS40: Numerical Methods for Polynomial Systems.
9-13 July 2012, Minneapolis, Minnesota.
slides of the talk