[petsc-dev] KSP rtol

Barry Smith bsmith at mcs.anl.gov
Wed Dec 14 14:44:26 CST 2011


On Dec 14, 2011, at 12:45 PM, Mark F. Adams wrote:

> I have a nonlinear problem and am finding the the convergence test for the interior PETSc linear solver is apparently testing against the norm of the first linear solve (see appended).

   No it is || r = b - Ax ||/|| b||   with a B in there if left preconditioned

   It has nothing to do with the "first linear solve". Each linear solve is self contained and it does not use anything from a previous linear solve to determine convergence of a later linear solve.

>  This is clearly not what anyone would want,

    Why not? 

   Barry


> is this the intended semantics?


>  Is there perhaps a method to reset rtol?
> 
> Mark
> 
> Picard iteration 0 max(resid) = 781020
>  0 KSP Residual norm 2.499264167662e+07 
>  1 KSP Residual norm 6.330915810095e+06 
>  2 KSP Residual norm 1.856999435466e+06 
>  3 KSP Residual norm 3.967325725053e+05 
>  4 KSP Residual norm 9.037332775808e+04 
>  5 KSP Residual norm 1.826960320400e+04 
> Linear solve converged due to CONVERGED_RTOL iterations 5
> Picard iteration 1 max(resid) = 727577 ------- Rate = 1.07345
>  0 KSP Residual norm 6.885328126772e+06 
>  1 KSP Residual norm 2.150319244570e+06 
>  2 KSP Residual norm 2.297759417371e+05 
>  3 KSP Residual norm 3.321819214451e+04 
>  4 KSP Residual norm 4.155970071625e+03 
> Linear solve converged due to CONVERGED_RTOL iterations 4
> Picard iteration 2 max(resid) = 486788 ------- Rate = 1.49465
>  0 KSP Residual norm 2.987417661953e+06 
>  1 KSP Residual norm 5.046119985008e+05 
>  2 KSP Residual norm 3.882757664605e+04 
>  3 KSP Residual norm 4.106091678345e+03 
> Linear solve converged due to CONVERGED_RTOL iterations 3
> Picard iteration 3 max(resid) = 137348 ------- Rate = 3.54418
>  0 KSP Residual norm 7.285359985607e+05 
>  1 KSP Residual norm 9.132700545850e+04 
>  2 KSP Residual norm 6.834608381922e+03 
> Linear solve converged due to CONVERGED_RTOL iterations 2
> Picard iteration 4 max(resid) = 25082.7 ------- Rate = 5.47582
>  0 KSP Residual norm 1.207523847817e+05 
>  1 KSP Residual norm 1.286174706329e+04 
> Linear solve converged due to CONVERGED_RTOL iterations 1
> Picard iteration 5 max(resid) = 5723.21 ------- Rate = 4.38264
>  0 KSP Residual norm 2.211945822133e+04 
> Linear solve converged due to CONVERGED_RTOL iterations 0
> 




More information about the petsc-dev mailing list