[petsc-users] Problem with snes: petsc 3.4.5

Barry Smith bsmith at petsc.dev
Wed Sep 20 09:59:20 CDT 2023 

10000 KSP preconditioned resid norm 1.027326121286e-13 true resid norm 4.122100161195e-13 ||r(i)||/||b|| 1.256880059386e-03

The residual norms are very small. From the final column we see that the relative decrease in the residual norm was only 1.e-3 meaning the initial linear system residual norm was already very small at the beginning of the linear solve.

I suspect you have a very small SNES tolerance? 

When using matrix-free SNES with complex systems you cannot expect to get really tight convergence.

I suggest -snes_rtol 1.e-8 -ksp_atol 1.e-11

But you can try out various values 

> On Sep 20, 2023, at 10:36 AM, Harry-Arthur Cousin <harry-arthur at hotmail.fr> wrote:
> 
> Will do. The output is very long, I'm sending you the last 20 iteration of KSP when the nonlinear solver did not converge. That's what i have:
> 
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9980 KSP preconditioned resid norm 7.911586848688e-14 true resid norm 3.571299151668e-13 ||r(i)||/||b|| 1.088933920648e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9981 KSP preconditioned resid norm 7.773677194223e-14 true resid norm 3.519061016604e-13 ||r(i)||/||b|| 1.073005857832e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9982 KSP preconditioned resid norm 7.142532169111e-14 true resid norm 3.536747918236e-13 ||r(i)||/||b|| 1.078398816059e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9983 KSP preconditioned resid norm 6.325819556307e-14 true resid norm 3.698647905235e-13 ||r(i)||/||b|| 1.127764153463e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9984 KSP preconditioned resid norm 6.119227152008e-14 true resid norm 3.563406240582e-13 ||r(i)||/||b|| 1.086527273025e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9985 KSP preconditioned resid norm 6.075666963630e-14 true resid norm 3.710895872549e-13 ||r(i)||/||b|| 1.131498712373e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9986 KSP preconditioned resid norm 5.985095889340e-14 true resid norm 3.509409502428e-13 ||r(i)||/||b|| 1.070062990062e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9987 KSP preconditioned resid norm 5.243339292054e-14 true resid norm 3.627159951643e-13 ||r(i)||/||b|| 1.105966579450e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9988 KSP preconditioned resid norm 4.491198445126e-14 true resid norm 4.006851442317e-13 ||r(i)||/||b|| 1.221739278968e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9989 KSP preconditioned resid norm 4.410245117574e-14 true resid norm 4.031793308619e-13 ||r(i)||/||b|| 1.229344366950e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9990 KSP preconditioned resid norm 4.072064997637e-13 true resid norm 4.072064997637e-13 ||r(i)||/||b|| 1.241623710224e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9991 KSP preconditioned resid norm 2.614566817566e-13 true resid norm 3.327810250174e-13 ||r(i)||/||b|| 1.014691099513e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9992 KSP preconditioned resid norm 2.214885215262e-13 true resid norm 3.122694737130e-13 ||r(i)||/||b|| 9.521488059891e-04
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9993 KSP preconditioned resid norm 1.937414464538e-13 true resid norm 2.949371229399e-13 ||r(i)||/||b|| 8.993002937814e-04
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9994 KSP preconditioned resid norm 1.683143904883e-13 true resid norm 3.116491153488e-13 ||r(i)||/||b|| 9.502572555001e-04
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9995 KSP preconditioned resid norm 1.455417378095e-13 true resid norm 3.848207009730e-13 ||r(i)||/||b|| 1.173366600951e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9996 KSP preconditioned resid norm 1.254552363609e-13 true resid norm 3.494786173247e-13 ||r(i)||/||b|| 1.065604153515e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9997 KSP preconditioned resid norm 1.187610795222e-13 true resid norm 3.364834320451e-13 ||r(i)||/||b|| 1.025980203084e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9998 KSP preconditioned resid norm 1.126887893465e-13 true resid norm 4.347573700680e-13 ||r(i)||/||b|| 1.325629770605e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   9999 KSP preconditioned resid norm 1.067379754742e-13 true resid norm 3.907245052617e-13 ||r(i)||/||b|| 1.191368040981e-03
> Linear solve converged due to CONVERGED_ITS iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
>   10000 KSP preconditioned resid norm 1.027326121286e-13 true resid norm 4.122100161195e-13 ||r(i)||/||b|| 1.256880059386e-03
>   Linear solve did not converge due to DIVERGED_ITS iterations 10000
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 1
> Linear solve converged due to CONVERGED_ITS iterations 1
> 
> De : Barry Smith <bsmith at petsc.dev <mailto:bsmith at petsc.dev>>
> Envoyé : mercredi 20 septembre 2023 16:26
> À : Harry-Arthur Cousin <harry-arthur at hotmail.fr <mailto:harry-arthur at hotmail.fr>>
> Cc : petsc-users at mcs.anl.gov <mailto:petsc-users at mcs.anl.gov> <petsc-users at mcs.anl.gov <mailto:petsc-users at mcs.anl.gov>>
> Objet : Re: [petsc-users] Problem with snes: petsc 3.4.5
>  
> 
>    Run with -snes_monitor and -ksp_monitor_true_residual and send the output
> 
>> On Sep 20, 2023, at 9:34 AM, Harry-Arthur Cousin <harry-arthur at hotmail.fr <mailto:harry-arthur at hotmail.fr>> wrote:
>> 
>> Hello,
>> 
>> I'm on petsc version 3.4.5.
>> 
>> My program has a time loop where I have to solve a non-linear system of the form F(X) = D*L*J + I + kappa(X) where X is an unknown vector, D and L are diagonal square matrices, J is a vector that can be calculated by solving a linear system with X and kappa is a non-linear application. I use the "-snes_mf" option to, as I understand it, approximate the Jacobian using finite differences.
>> 
>> A first version of my problem could be solved, so I made it more complex by changing the values of the matrix L, which used to have the same value on the whole diagonal and now has values of the same order of magnitude but different. This second version takes longer and doesn't converge after a while. I get the message: Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE. However, when I look at the faq (https://petsc.org/release/faq/#why-is-newton-s-method-snes-not-converging-or-converges-slowly), it indicates that it's probably a Jacobian problem. What should I do? Do I have to calculate the explicit Jacobian? The explicit Jacobian requires the inversion of a matrix (due to the linear system for solving J).
>> 
>> 
>> Kind regards,
>> 
>> Harry Cousin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20230920/eab84d77/attachment.html>

More information about the petsc-users mailing list