[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Matrix convergence.

  Run the parallel PETSc job with -ksp_type gmres -ksp_gmres_restart 200
-sub_pc_type lu


On Fri, 3 Mar 2006, billy@dem.uminho.pt wrote:


Hi,

I tried to solve the 18x18 matrix with laspack and the results are the following:

Solver: GMRES, Preconditioning: NULL, Iterations: 500, Residual: 6.16E-2
Solver: GMRES, Preconditioning: SSOR, Iterations: 500, Residual: 6.57E-2
Solver: GMRES, Preconditioning: ILU, ERROR

Solver: BiCGSTAB, Preconditioning: NULL, Iterations: 101, Residual: 8.88E-7
Solver: BiCGSTAB, Preconditioning: SSOR, Iterations: 36, Residual: 9.87E-7
Solver: BiCGSTAB, Preconditioning: ILU, ERROR


Billy.


Quoting Barry Smith <bsmith@mcs.anl.gov>:


   Billy,

    Most likely there is something wrong with the matrix being
generated. I would first look at the matrix entries and make
sure they make sense. Then I would solve the linear systems
with a direct solver -pc_type lu (on one processor and make
sure the solutions are reasonable and your simulation has
"correct" answers).

   It is possible the parallel matrix generation is wrong,

   Barry


On Thu, 2 Mar 2006, billy@dem.uminho.pt wrote:


Hi,


I have been testing my parallel implementation with a 2D case. I have used

the

following grids with 2 and 4 processes: 4x4, 10x10, 16x16 and 18x18.

It works for the first three (max. number of iterations to converge aprox.

70)

but when I increase to 18x18 the matrix does not converge. I have increased

the

maximum number of iterations and I have used GMRES, BiCGSTAB with
preconditioners ASM, JACOBI, etc.

With KSPType = GMRES and PCType = ASM:
Number of iterations: 500 Residual: +1.420471E-01

With KSPType = GMRES and PCType = JACOBI:
Number of iterations: 500 Residual: +2.351882E-02

With KSPType = BCGS and PCType = ASM:
Number of iterations: 136 Residual: +3.911446E+04

With KSPType = BCGS and PCType = NULL:
Number of iterations: 157 Residual: +3.254728E+03

Is it normal that the convergence detiorates so much with such a slight

increase

in the dimension of the matrix? Is there any optimization parameter that I

may

be missing?


Billy.