Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Distributed memory implementation of multigrid methods for groundwater flow problems with rough coefficients

Conference ·
OSTI ID:390773
 [1];  [2]
  1. Oak Ridge National Lab., TN (United States)
  2. Univ. of Illinois, Urbana-Champaign, IL (United States)

In this paper the authors present parallel solvers for large linear systems arising from the finite-element discretization of three-dimensional groundwater flow problems. They have tested their parallel implementations on the Intel Paragon XP/S 150 supercomputer using up to 1024 parallel processors. Their solvers are based on multigrid and Krylov subspace methods. Their goal is to combine powerful algorithms and current generation high performance computers to enhance the capabilities of computer models for groundwater modeling. The authors demonstrate that multigrid can be a scalable algorithm on distributed memory machines. They demonstrate the effectiveness of parallel multigrid based solvers by solving problems requiring more than 64 million finite-elements in less than a minute. Their results show that multigrid as a stand alone solver works best for problems with smooth coefficients, but for rough coefficients it is best used as preconditioner for Krylov method.

DOE Contract Number:
AC05-96OR22464
OSTI ID:
390773
Report Number(s):
CONF-960482--
Country of Publication:
United States
Language:
English