[petsc-dev] sor smoothers

[Barry Smith]

Jed did not like the _ in it. I liked it :-(



On Sep 12, 2013, at 5:19 PM, "Mark F. Adams" <mfadams at lbl.gov> wrote:

> And there is already:
> 
> PETSC_DEPRECATED("Use PCMGResidualDefault()") PETSC_STATIC_INLINE PetscErrorCode PCMGResidual_Default(Mat A,Vec b,Vec x,Vec r) {
>  return PCMGResidualDefault(A,b,x,r);
> }
> 
> Should I assume PCMGResidualDefault is dead and change this to:
> 
> PETSC_DEPRECATED("Use MatResidualDefault()") PETSC_STATIC_INLINE PetscErrorCode PCMGResidualDefault(Mat A,Vec b,Vec x,Vec r) {
>  return PCMGResidualDefault(A,b,x,r);
> }
> 
> ?
> 
> On Sep 12, 2013, at 5:46 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> 
>> 
>> On Sep 12, 2013, at 4:39 PM, "Mark F. Adams" <mfadams at lbl.gov> wrote:
>> 
>>> I'm not sure what to do with:
>>> 
>>> PetscErrorCode  PCMGSetResidual(PC pc,PetscInt l,PetscErrorCode (*residual)(Mat,Vec,Vec,Vec),Mat mat) ?
>>> 
>>> I'm thinking to have this in the code:
>>> 
>>> mglevels[l]->A = mat;
>>> mglevels[l]->residual = residual;
>>> 
>>> and change the calls from using PCMGResidualDefault to using MatResidualDefault.
>> 
>>  Seems ok to me.
>> 
>>> This will keep from breaking code that uses this to set the Mat also.  And should this be called PCMGSetResidualAndMat?
>> 
>>  No.
>> 
>>  Barry
>> 
>>> 
>>> Mark
>>> 
>>> On Sep 9, 2013, at 9:25 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>> 
>>>> 
>>>> On Sep 9, 2013, at 8:03 PM, "Mark F. Adams" <mfadams at lbl.gov> wrote:
>>>> 
>>>>> OK, so its ground hog day.  The plan?:
>>>>> 
>>>>> 1) I should add a residual method to Mat, make a MatResidualDefault, and have all (yuck) SetUp_MatXXX set this default.  I guess there is no base class SetUp ….  then make a MatResidual_SeqAIJ that just does a dumb residual for now, with testing logic if Jed's stuff is available.
>>>> 
>>>> Take a look at MatCreate_SeqAIJ. These fill up all the function points in one swoop with the big function table, this is why there is not "base class setup".
>>>> 
>>>>> 
>>>>> 2) Have PCMGResidualDefault call MatResidual instead MatMult & VecAYPX.  Or do we want to nuke PCMGResidualDefault?
>>>> 
>>>> You don't really need PCMGResidualDefault, just make the residual function pointer in PCMG point to MatResidual
>>>> 
>>>>> Jed's defect correction argument might sound like a typical Jed, e.g., what if a Martian gets hepatitis during mardi gras, but defect correction is really useful and I would like to see its support supported …
>>>> 
>>>> Don't need it to support defect correction  but anyways :-)
>>>> 
>>>>> 
>>>>> 3) Add fields in Mat: mat_res_state, vec_res_state, vec_res_id.  I need to know the types but I can do this in the next phase.
>>>> 
>>>> Look in _p_PetscObject for the types of state and id, both currently PetscInt.
>>>> 
>>>>> 
>>>> 
>>>>> Mark
>>>>> 
>>>>> On Sep 9, 2013, at 5:35 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>>>> 
>>>>>> 
>>>>>> On Sep 9, 2013, at 4:32 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
>>>>>> 
>>>>>>> Barry Smith <bsmith at mcs.anl.gov> writes:
>>>>>>>> I don't care if PCMGSetResidual stays around for now but I do not
>>>>>>>> think we should "point to it" as a standard way for people to do
>>>>>>>> things; I think we should improve how the two slots are used with
>>>>>>>> PCMG, it really has never been thought about at all and whenever
>>>>>>>> possible I think using the two slots appropriately is better than
>>>>>>>> PCMGSetResidual and will solve most of the "use" cases.
>>>>>>> 
>>>>>>> Fine, I was just not eager to _delete_ PCMGSetResidual until we had
>>>>>>> established a complete replacement, and even then, it might be enough to
>>>>>>> leave it as "advanced" and suggest the preferred alternative in the man
>>>>>>> page.  It just doesn't seem expensive to support and I don't think it
>>>>>>> causes people to tangle themselves up in bad design.
>>>>>> 
>>>>>> It doesn't have its own object which is kind of limiting.
>>>>>> 
>>>>>> 
>>>>>> 
>>>>> 
>>>> 
>>> 
>> 
>