About CG
 

Hi all,

my question is about the behaviour of conjugate gradient method. My CG solver (with Jacobi Preconditioning) has been tested and works well. Its a part of my fem code. I took a small system [A]{x} = {b} [A] = 5X5 SPD and diagonally dominant very well conditioned matrix. I multiplied all the equations by a huge number say 1e6. CG solver actually diverged :( however Gauss-Siedel and Jacobi (SSOR) solvers just zipped through.... it took more iterations but they converged to the known solution. I think diagonal dominance is all that GS/Jacobi needs. I cant explain behaviour of CG (a possible bug ???) anyone out there try anything stupid like me ???

Thanks in Advance

Abhijit Tilak