Solving initial value problems by the Picard-Chebyshev method with NVIDIA GPUS

Xiaoli Bai, John L. Junkins

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Scopus citations

Abstract

Combining the Picard method with Chebyshev polynomials, we can solve a large number of initial value problems using parallel computation. We develop a matrixvector form of Picard-Chebyshev method and implement it using a NVIDIA graphics card. The algorithm is developed in the NVIDIA CUDA environment and utilize the CUBLAS toolbox. Compared with ODE45 (the Runge-Kutta 45 algorithm implemented in MATLAB), the speedup from using Picard-Chebyshev method for the exemplar problem is about a factor of 30 to 120. Compared with a CPU version of the Picard-Chebyshev algorithm implemented in MATLAB, the speedup from using one NVIDIA graphics card is about a factor of 2 to 3. These results show that even with the simplest utilization of graphical processing units presently available, the speedup with parallel implementations of the Picard-Chebyshev algorithm is up to two order of magnitude.

Original languageEnglish (US)
Title of host publicationSpaceflight Mechanics 2010 - Advances in the Astronautical Sciences
Subtitle of host publicationProceedings of the AAS/AIAA Space Flight Mechanics Meeting
Pages1459-1476
Number of pages18
StatePublished - 2010
Externally publishedYes
EventAAS/AIAA Space Flight Mechanics Meeting - San Diego, CA, United States
Duration: Feb 14 2010Feb 17 2010

Publication series

NameAdvances in the Astronautical Sciences
Volume136

Other

OtherAAS/AIAA Space Flight Mechanics Meeting
Country/TerritoryUnited States
CitySan Diego, CA
Period2/14/102/17/10

All Science Journal Classification (ASJC) codes

  • Aerospace Engineering
  • Space and Planetary Science

Fingerprint

Dive into the research topics of 'Solving initial value problems by the Picard-Chebyshev method with NVIDIA GPUS'. Together they form a unique fingerprint.

Cite this