[petsc-dev] Recent changes to Chebyshev

[Garth N. Wells]

On Thu, Mar 5, 2015 at 7:59 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>
>   Run the program with -help
>

KSP Chebyshev Options
  -mg_levels_1_ksp_chebyshev_eststeps <10>: Number of est steps in Chebyshev ()
  -mg_levels_1_ksp_chebyshev_eigenvalues <0>: extreme eigenvalues
(KSPChebyshevSetEigenvalues)
  -mg_levels_1_ksp_chebyshev_estimate_eigenvalues <-1>: estimate
eigenvalues using a Krylov method, then use this transform for
Chebyshev eigenvalue bounds (KSPChebyshevSetEstimateEigenvalues)


>   I cannot explain why that doesn't work. Are you running with -mg_levels_ksp_type cheby  ?
>

Yes. Using

  -mg_levels_ksp_chebyshev_eststeps

does change the number of iterations used in the eigenvalue estimation.

Garth

>   Barry
>
>
>
>> On Mar 5, 2015, at 1:55 PM, Garth N. Wells <gnw20 at cam.ac.uk> wrote:
>>
>> On Thu, Mar 5, 2015 at 7:41 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>>
>>> _eigest
>>>
>>
>> I've tried (amongst other things):
>>
>>  -mg_levels_eigest_ksp_type
>>  -mg_levels_eigest_ksp_max_it
>>
>> but it isn't right. Could you give me the full option name? Thanks.
>>
>> Garth
>>
>>
>>>  I unified the names of the functions and the prefixes. The functions related to this all have EigEst  and the prefix _eigest  for Eigenvalue Estimator
>>>
>>>  Barry
>>>
>>>
>>>> On Mar 5, 2015, at 1:29 PM, Garth N. Wells <gnw20 at cam.ac.uk> wrote:
>>>>
>>>> I see there have been some recent changes relating to Chebyshev. What
>>>> should I now use in place of the old options
>>>>
>>>> -mg_est_ksp_type
>>>> -gamg_est_ksp_type
>>>>
>>>> ?
>>>>
>>>> Garth
>>>
>