[petsc-users] query DMPlexCreateSection

[Dharmendar Reddy]

I think, the issue is with petscSectionSetUp which assumes all fields
to have same chart. This is ok if the fields are defined at every
point of the chart.

On Fri, Jan 3, 2014 at 1:37 AM, Dharmendar Reddy
<dharmareddy84 at gmail.com> wrote:
> Hello,
>          I did not fully understand the flow. Need more help.
> If i look at the flow in DMPlexCreateSectionInitial
> (http://www.mcs.anl.gov/petsc/petsc-current/src/dm/impls/plex/plex.c.html#DMPlexCreateSection)
>
> 6023:   for (d = 0; d <= dim; ++d) {
> 6024:     numDofTot[d] = 0;
> 6025:     for (f = 0; f < numFields; ++f) numDofTot[d] += numDof[f*(dim+1)+d];
> 6026:   }
> 6027:   PetscSectionCreate(PetscObjectComm((PetscObject)dm), section);
> 6028:   if (numFields > 0) {
> 6029:     PetscSectionSetNumFields(*section, numFields);
> 6030:     if (numComp) {
> 6031:       for (f = 0; f < numFields; ++f) {
> 6032:         PetscSectionSetFieldComponents(*section, f, numComp[f]);
> 6033:       }
> 6034:     }
> 6035:   }
> 6036:   DMPlexGetChart(dm, &pStart, &pEnd);
> 6037:   PetscSectionSetChart(*section, pStart, pEnd);
> 6038:   for (d = 0; d <= dim; ++d) {
> 6039:     DMPlexGetDepthStratum(dm, d, &pStart, &pEnd);
> 6040:     for (p = pStart; p < pEnd; ++p) {
> 6041:       for (f = 0; f < numFields; ++f) {
> 6042:         PetscSectionSetFieldDof(*section, p, f, numDof[f*(dim+1)+d]);
> 6043:       }
> 6044:       PetscSectionSetDof(*section, p, numDofTot[d]);
> 6045:     }
> 6046:   }
>
> In line 6027, number of fields for entire section is defined. Now,
> dose the flow in your earlier replies requires me to do that ? I.e,
> define numFields to 2. Then
>
> Inside the loop 6040, i need to check if the point belongs to region 1 or 2.
>
> If it belongs to region 1 then numFields is 1 and if it belongs to
> region 2, numFields is 2.
>
> Where do i set the number of fields in the flow you mentioned in the
> earlier email ?
>
> Thanks
> Reddy
>
> On Thu, Jan 2, 2014 at 9:56 AM, Matthew Knepley <knepley at gmail.com> wrote:
>> On Thu, Jan 2, 2014 at 4:11 AM, Dharmendar Reddy <dharmareddy84 at gmail.com>
>> wrote:
>>>
>>> Hello,
>>>          I am trying to use DMPlexCreateSection from fortran. I was
>>> able to solve a Poisson equation earlier using DMPlex for handling
>>> mesh data.
>>>
>>> Now i need to solve a system of equations: (poisson + continuity
>>> equations)
>>>
>>> - div (grad phi) = (C + n)  --- (1)
>>>   div (J ) = 0  ---------------  (2)
>>>   J = n grad(phi)
>>> ( C is constant)
>>> Simulation domain is defined as, rectangular regions 1 and 2 shown below.
>>>
>>> -------------
>>> |     1        |
>>> -------------
>>> |               |
>>> |       2      |
>>> |               |
>>> --------------
>>>
>>> Now, equation 1 is defined in region 1 and 2
>>> and equation 2 is defined only for region 2.
>>>
>>> How do i setup the section  ? DMPlexCreateSection applies the given
>>> DOF layout  per cell to all cells in the mesh.
>>>
>>> I am not sure if all the function calls inside
>>> DMPlexCreateSectionIntial and DMPlexCreateSectionBCDof  are accessible
>>> from Fotran.
>>
>>
>> You don't want them anyway since they apply to the whole domain. The control
>> flow could be:
>>
>>   DMPlexSetChart()
>>   <loop over region 1>
>>     DMPlexSetDof() and DMPlexSetFieldDof()
>>   <loop over region 2>
>>     DMPlexSetDof() and DMPlexSetFieldDof()
>>   <loop over BC>
>>     DMPlexSetConstraintDof() and DMPlexSetFieldConstraintDof()
>>   DMPlexSetUp()
>>   <loop over BC>
>>     DMPlexSetConstraintIndices() and DMPlexSetFieldConstraintIndices()
>>
>> We could try and package some of this up if it looks generic.
>>
>>   Thanks,
>>
>>      Matt
>>
>>>
>>> Thanks
>>> Reddy
>>>
>>> --
>>> -----------------------------------------------------
>>> Dharmendar Reddy Palle
>>
>>
>>
>>
>> --
>> What most experimenters take for granted before they begin their experiments
>> is infinitely more interesting than any results to which their experiments
>> lead.
>> -- Norbert Wiener
>
>
>
> --
> -----------------------------------------------------



-- 
-----------------------------------------------------
Dharmendar Reddy Palle
Graduate Student
Microelectronics Research center,
University of Texas at Austin,
10100 Burnet Road, Bldg. 160
MER 2.608F, TX 78758-4445
e-mail: dharmareddy84 at gmail.com
Phone: +1-512-350-9082
United States of America.
Homepage: https://webspace.utexas.edu/~dpr342