Damped Jacobi Preconditioning and Coarse Grid Deflation for Conjugate Gradient Iteration on Parallel Computers
Lois E. Mansfield · SIAM Journal on Scientific and Statistical Computing · 1991
A combination of damped Jacobi preconditioning and deflation is given as a way to improve the convergence of conjugate gradient iteration on local memory multiprocessor computers such as the hypercube. Experimental results are given for the linear system associated with the elastic bending of a cantilever beam on the iPSC/2 with both 4 and 16 processors.