Parallel/vector integration methods for dynamical astronomy (Q1976095)

The author investigates the effect of parallelization in three typical numerical ODE integrators. First, in the Picard-Chebyshev method using a large number of Chebyshev polynomials, 100-1000 times acceleration is detected. Second, by the parallelization of a symplectic integrator (using fixed point iterations) 50 times acceleration is achieved. Third, an extrapolation method which uses trial integrations in parallel, is accelerated by the ``folding'' technique. The author also presents an all-purpose method for which 3.5 times acceleration is achieved, and gives a perspective of employing implicit integrators with multiple corrections.