[petsc-dev] tao4py and tao in PETSc master

[Barry Smith]

  Great, thanks. I have made these updates but note that I have 

Not put a Petsc in front of PetscTAOTRON etc

Removed the tao_ part of the string name version of the TaoTypes

Removed -tao_method since it is redundant to -tao_type (the PETSc style).

Updated in barry/tao-fixes-for-petsc4py and next

Please fix any of my fixes in barry/tao-fixes-for-petsc4py

   Barry

On Feb 16, 2014, at 1:37 PM, Lisandro Dalcin <dalcinl at gmail.com> wrote:

> This branch contains what is needed to expose TAO within petsc4py:
> https://bitbucket.org/petsc/petsc4py/branch/dalcinl/tao
> 
> Two comments/requests for TAO in petsc-dev:
> 
> * I would like to have TaoGetConstraintTolerances(), right now I'm
> defining it in petsc4py as shown in the link below:
> https://bitbucket.org/petsc/petsc4py/commits/ab01ebe336491190a2db85c179c48b1c64fdc673?at=master#chg-src/include/custom.h
> 
> * Look at all these definitions, they should be exposed in the C API
> to be able to call e.g. TaoSetType(tao,TAOLMVM);
> https://bitbucket.org/petsc/petsc4py/commits/ab01ebe336491190a2db85c179c48b1c64fdc673?at=master#Lsrc/PETSc/petsctao.pxiT4
> 
> Final comment: I've named the solver "TAO" (all uppercase) in
> petsc4py. This deviates from the current PETSc convention using "Tao",
> but I'm thinking that at some point you will get upset with the
> inconsistency and rename it to TAO.
> 
> Let me know what you think. After merging this branch, "tao4py" should
> die and the functionalities will be available directly from petsc4py.
> 
> 
> 
> On 15 February 2014 01:08, Lisandro Dalcin <dalcinl at gmail.com> wrote:
>> On 12 February 2014 18:21, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>> 
>>>  Lisandro,
>>> 
>>>    Now that tao is a subdirectory in the PETSc package (currently in master) and behaves “just like” other parts of PETSc will you be updating Tao4py? With the reorganization of Tao into PETSc I think this should even simplify Tao4py.
>>> 
>> 
>> I'm still not sure what to do. Perhaps I should just add the TAO
>> features to petsc4py directly? Anywat, I'll first update tao4py to
>> work with TAO in petsc-dev.
>> 
>>> 
>>>   Let me know if there is anything broken in my reorg of Tao into PETSc.
>>> 
>> 
>> I'll do once I have some free slots to work on it.
>> 
>> 
>> --
>> Lisandro Dalcin
>> ---------------
>> CIMEC (UNL/CONICET)
>> Predio CONICET-Santa Fe
>> Colectora RN 168 Km 472, Paraje El Pozo
>> 3000 Santa Fe, Argentina
>> Tel: +54-342-4511594 (ext 1016)
>> Tel/Fax: +54-342-4511169
> 
> 
> 
> -- 
> Lisandro Dalcin
> ---------------
> CIMEC (UNL/CONICET)
> Predio CONICET-Santa Fe
> Colectora RN 168 Km 472, Paraje El Pozo
> 3000 Santa Fe, Argentina
> Tel: +54-342-4511594 (ext 1016)
> Tel/Fax: +54-342-4511169