[petsc-dev] gurus what does this mean?

Jed Brown jed at jedbrown.org
Tue Apr 29 08:20:55 CDT 2014


Barry Smith <bsmith at mcs.anl.gov> writes:

> ex10 -f0 ~/Datafiles/Matrices/Foad -ksp_type gmres -pc_type none -ksp_gmres_restart 2000 -ksp_plot_eigenvalues -draw_pause -1  -ksp_monitor_true_residual -ksp_rtol 1.e-6 -ksp_max_it 1990 -geometry 0,0,600,600 -draw_save -pc_type lu -ksp_rtol 1.e-12 -pc_factor_mat_solver_package superlu
>   0 KSP preconditioned resid norm 1.752626139140e+05 true resid norm 1.860478790525e+07 ||r(i)||/||b|| 1.000000000000e+00
>   1 KSP preconditioned resid norm 3.174750685865e+04 true resid norm 1.816271753408e+07 ||r(i)||/||b|| 9.762388921916e-01
>   2 KSP preconditioned resid norm 5.903085310063e-07 true resid norm 1.783619436500e-05 ||r(i)||/||b|| 9.586884008478e-13
>
> superlu does lots of pivoting for stability. What does it mean that the first iteration barely affects the residual and then the second iteration wipes it out? 

Can you compute the error (rather than residual)?

> Right precondition wipes out the residual on one iteration 
>
> gmres_restart 2000 -ksp_plot_eigenvalues -draw_pause -1  -ksp_monitor_true_residual -ksp_rtol 1.e-6 -ksp_max_it 1990 -geometry 0,0,600,600 -draw_save -pc_type lu -ksp_rtol 1.e-12 -pc_factor_mat_solver_package superlu -ksp_pc_side right
>   0 KSP unpreconditioned resid norm 1.860478790525e+07 true resid norm 1.860478790525e+07 ||r(i)||/||b|| 1.000000000000e+00
>   1 KSP unpreconditioned resid norm 2.559955879548e-07 true resid norm 2.505789584416e-07 ||r(i)||/||b|| 1.346852002386e-14
> Number of iterations =   1
> Residual norm 2.50579e-07
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20140429/6f76a994/attachment.sig>


More information about the petsc-dev mailing list