Adaptive smoothed aggregation multigrid for non-symmetric problems

Geoffrey Sanders
Dept. of Applied Mathematics, 526 UCB
University of Colorado, Boulder CO 80309-0526
sandersg@colorado.edu
T. Manteuffel, S. McCormick, M. Brezina, J. Ruge

The performance of the adaptive smooth aggregation multigrid ($ \alpha$SA) algorithm suffers for non-symmetric problems. I will present a non-symmetric version of the algorithm that uses Kaczmarz iteration as a relaxation step and smoothed aggregation of local right and left singular vectors to form a hierarchy of coarse grid operators. Testing has been done on one and two dimensional convection dominated convection-diffusion. I will present the current results and struggles of these tests. Expect some whining.