[petsc-users] Help with SLEPc eigenvectors convergence.

[Eric Chamberland]

Hi,

wow, it was only that, it's now working perfectly!

Curiously, I still have 2 times the message about relative errors 
followed by a new one with converged values, see:

options_eps_gen Linear eigensolve converged (13 eigenpairs) due to 
CONVERGED_TOL; iterations 2
Problem: some of the first 10 relative errors are higher than the tolerance

Problem: some of the first 10 relative errors are higher than the tolerance

All requested eigenvalues computed up to the required tolerance:
     0.00000, 0.00000, 0.00000, 199.28609, 415.65289, 570.21994, 
1294.72406, 1295.67360,
     3769.78800, 3771.15894


So is it intended to write these 2 warnings?

Anyway thanks a lot! :)

Eric


On 2024-03-07 02:24, Jose E. Roman wrote:
> Pierre's diagnostic is right, but the suggested option is wrong: it should be -eps_target
> Also, I would suggest using a larger value, such as -eps_target 0.1, otherwise the tolerance 1e-14 might not be attained.
>
> Jose
>
>
>> El 7 mar 2024, a las 8:01, Pierre Jolivet<pierre at joliv.et>  escribió:
>>
>> This Message Is From an External Sender
>> This message came from outside your organization.
>> It seems your A is rank-deficient.
>> If you slightly regularize the GEVP, e.g., -st_target 1.0E-6, you’ll get errors closer to 0.
>>
>> Thanks,
>> Pierre
>>
>>> On 6 Mar 2024, at 8:57 PM, Eric Chamberland via petsc-users<petsc-users at mcs.anl.gov>  wrote:
>>>
>>> This Message Is From an External Sender
>>> This message came from outside your organization.
>>> Hi,
>>>
>>> we have a simple generalized Hermitian problem  (Kirchhoff plate
>>> vibration) for which we are comparing SLEPc results with Matlab results.
>>>
>>> SLEPc computes eigenvalues correctly, as Matlab does.
>>>
>>> However, the output eigenvectors are not fully converged and we are
>>> trying to understand where we have missed a convergence parameter or
>>> anything else about eigenvectors.
>>>
>>> SLEPc warns us at the end of EPSSolve with this message:
>>>
>>> ---
>>>    Problem: some of the first 10 relative errors are higher than the
>>> tolerance
>>> ---
>>>
>>> And in fact, when we import the resulting vectors into Matlab,
>>> "A*x-B*Lambda*x" isn't close to 0.
>>>
>>> Here are attached the EPS view output as the A and B matrices used.
>>>
>>> Any help or insights will be appreciated! :)
>>>
>>> Thanks,
>>> Eric
>>>
>>> <epsview.txt><matrice0_gen.m><matrice1_gen.m>

-- 
Eric Chamberland, ing., M. Ing
Professionnel de recherche
GIREF/Université Laval
(418) 656-2131 poste 41 22 42
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20240307/5e5fde60/attachment.html>