[petsc-users] using petsc tools to solve isolated irregular domains with finite difference
Matthew Knepley
knepley at gmail.com
Thu Oct 31 17:34:13 CDT 2013
On Thu, Oct 31, 2013 at 5:02 PM, Bishesh Khanal <bisheshkh at gmail.com> wrote:
>
>
>
> On Thu, Oct 31, 2013 at 7:43 PM, Matthew Knepley <knepley at gmail.com>wrote:
>
>> On Thu, Oct 31, 2013 at 1:25 PM, Bishesh Khanal <bisheshkh at gmail.com>wrote:
>>
>>> I did not call DMCreateMatrix() before when using just dmda (and it is
>>> working). In any case, I added a call to DMCreateMatrix() now but still
>>> there are problems. Here are the code snippets and the error I get:
>>>
>>
>> Then you were allowing it to be called automatically by PETSc, and it
>> would have been this time as well.
>>
>>
>>> //Setting up the section and linking it to DM:
>>> ierr = PetscSectionCreate(PETSC_COMM_WORLD, &s);CHKERRQ(ierr);
>>> ierr = DMDAGetNumCells(dm, NULL, NULL, NULL, &nC);CHKERRQ(ierr);
>>> ierr = PetscSectionSetChart(s, 0, nC);CHKERRQ(ierr);
>>> for (PetscInt j = info.ys; j < info.ys+info.ym; ++j) {
>>> for (PetscInt i = info.xs; i < info.xs+info.xm; ++i) {
>>> PetscInt point;
>>> if(isPosInDomain(&testPoisson,i,j,0)) {
>>> ierr = DMDAGetCellPoint(dm, i, j, 0,
>>> &point);CHKERRQ(ierr);
>>> ierr = PetscSectionSetDof(s, point, testPoisson.mDof);
>>> // No. of dofs associated with the point.
>>> }
>>> }
>>> }
>>> ierr = PetscSectionSetUp(s);CHKERRQ(ierr);
>>> ierr = DMSetDefaultSection(dm, s);CHKERRQ(ierr);
>>> ierr = PetscSectionDestroy(&s);CHKERRQ(ierr);
>>> ierr = DMCreateMatrix(dm,&A);CHKERRQ(ierr);
>>>
>>> //Set up KSP:
>>> ierr = KSPCreate(PETSC_COMM_WORLD,&ksp);CHKERRQ(ierr);
>>> ierr = KSPSetDM(ksp,dm);CHKERRQ(ierr);
>>> ierr =
>>> KSPSetComputeOperators(ksp,computeMatrix2dSection,(void*)&testPoisson);CHKERRQ(ierr);
>>> ierr =
>>> KSPSetComputeRHS(ksp,computeRhs2dSection,(void*)&testPoisson);CHKERRQ(ierr);
>>> ierr = KSPSetFromOptions(ksp);CHKERRQ(ierr);
>>> ierr = KSPSolve(ksp,NULL,NULL);CHKERRQ(ierr);
>>>
>>> ------------------------------------------------------
>>> The computeMatrix2dSection function has:
>>>
>>> ierr = KSPGetDM(ksp,&da);CHKERRQ(ierr);
>>> ierr = DMDAGetLocalInfo(da,&info);CHKERRQ(ierr);
>>> ierr = DMGetDefaultGlobalSection(da,&gs);CHKERRQ(ierr);
>>> for(PetscInt j = info.ys; j<info.ys+info.ym; ++j) {
>>> for(PetscInt i = info.xs; i<info.xs+info.xm; ++i) {
>>> if (isPosInDomain(ctx,i,j)) {
>>> ierr = DMDAGetCellPoint(da,i,j,0,&point);CHKERRQ(ierr);
>>> ierr = PetscSectionGetOffset(gs,point,&row);
>>> ierr = PetscSectionGetDof(gs,point,&dof);
>>> for(PetscInt cDof = 0; cDof < dof; ++cDof) {
>>> num = 0;
>>> row+=cDof;
>>> col[num] = row; //(i,j) position
>>> v[num++] = -4;
>>> if(isPosInDomain(ctx,i+1,j)) {
>>> ierr =
>>> DMDAGetCellPoint(da,i+1,j,0,&point);CHKERRQ(ierr);
>>> ierr = PetscSectionGetOffset(gs,point,&col[num]);
>>> col[num] += cDof;
>>> v[num++] = 1;
>>> }
>>> if(isPosInDomain(ctx,i-1,j)) {
>>> ierr =
>>> DMDAGetCellPoint(da,i-1,j,0,&point);CHKERRQ(ierr);
>>> ierr = PetscSectionGetOffset(gs,point,&col[num]);
>>> col[num] += cDof;
>>> v[num++] = 1;
>>> }
>>> if(isPosInDomain(ctx,i,j+1)) {
>>> ierr =
>>> DMDAGetCellPoint(da,i,j+1,0,&point);CHKERRQ(ierr);
>>> ierr = PetscSectionGetOffset(gs,point,&col[num]);
>>> col[num] += cDof;
>>> v[num++] = 1;
>>> }
>>> if(isPosInDomain(ctx,i,j-1)) {
>>> ierr =
>>> DMDAGetCellPoint(da,i,j-1,0,&point);CHKERRQ(ierr);
>>> ierr = PetscSectionGetOffset(gs,point,&col[num]);
>>> col[num] += cDof;
>>> v[num++] = 1;
>>> }
>>> ierr =
>>> MatSetValues(A,1,&row,num,col,v,INSERT_VALUES);CHKERRQ(ierr);
>>> }
>>> }
>>> }
>>> }
>>> ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
>>> ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
>>>
>>> But this is not working. For e.g. for a 6X6 grid size I get the
>>> following error:
>>>
>>
>> Okay, here is how to debug this:
>>
>> 0) Go to a single scalar equations to make things easier
>>
>> 1) Use MatView(A, PETSC_VIEWER_STDOUT_WORLD), which should output a 36
>> row matrix with
>> the preallocated nonzero pattern, filled with zeros
>>
>> 2) Make sure this is the pattern you want
>>
>> 3) Run in the debugger with -start_in_debugger
>>
>> 4) When you get the error, see
>>
>> a) If the (i, j) is one that should be setting a value
>>
>> b) Why this (i, j) was not preallocated
>>
>
> Up to 4 (a), it is correct. There is a problem in the way DMDAGetCellPoint
> and PetscSectionGetOffset works in this case scenario. I will try to
> explain it below for the case of 4X4 grid.
> First case:
> If I set the computational domain to be all the points of the dmda grid,
> (i.e. isPosInDomain(..,i,j,..) returns true for all i and j in the dmda
> grid), the program runs fine and does not give any error.
>
> Second case:
> I want the computational domain to be some part of the whole grid. There
> is a problem in this case.
> The following test is in a case where,
> isPosInDomain(..,i,j,..) returns true ONLY for (i,j) pairs of (2,1) (1,2)
> (2,2) (3,2) and (3,2). The grid with its corresponding point number in the
> petscsection is shown below:
>
> 12 13 14 15
> 8 9 10 11
> 4 5 6 7
> 0 1 2 3
>
> of which 6, 9, 10, 11 and 14 correspond to the (i,j) points that returns
> true for isPosInDomain(..,i,j,..)
> MatView gives me the 16-row matrix with the star stencil non-zero
> structure as expected.
> The error I get is: new non-zero at (0,2) caused a malloc.
>
> This error is for the value of (i,j) = (2,1), i.e. point 6.
> The loop to set values in the matrix starts as:
> for(PetscInt j = info.ys; j<info.ys+info.ym; ++j) {
> for(PetscInt i = info.xs; i<info.xs+info.xm; ++i) {
> if (isPosInDomain(ctx,i,j)) {
> ierr = DMDAGetCellPoint(da,i,j,0,&point);CHKERRQ(ierr);
> ierr = PetscSectionGetOffset(gs,point,&row);
>
> Now, what I wanted from the beginning is to create the matrix containing
> the rows corresponding to only those points (i,j) which has true values for
> isPosInDomain(ctx,i,j), that means in this case 5 row matrix. In the
> current test, the first (i,j) that reaches DMDAGetCellPoint is (2,1), which
> gives sets point variable to 6.
> The line PetscSectionGetOffset(gs,point,&row) sets the value of row to 0.
> So I think this where the inconsistency lies.
> MatSetValues(A,1,&row,num,col,v,INSERT_VALUES) expects row variable and
> col[] array to correspond to (i,j) based on DMDA, but PetsSectionGetOffset
> gives result based on how we masked away some of the points from dmda grid.
> Or am I missing something very basic here ?
>
You are missing something basic. The row is correctly identified as 0, and
that means the 2 is
point 10. The problem must be preallocation. First, you should send the
result of MatView().
Matt
>
>> Thanks,
>>
>> Matt
>>
>>
>>> [0]PETSC ERROR: --------------------- Error Message
>>> ------------------------------------
>>> [0]PETSC ERROR: Argument out of range!
>>> [0]PETSC ERROR: New nonzero at (0,6) caused a malloc!
>>> [0]PETSC ERROR:
>>> ------------------------------------------------------------------------
>>> [0]PETSC ERROR: Petsc Development GIT revision:
>>> 61e5e40bb2c5bf2423e94b71a15fef47e411b0da GIT Date: 2013-10-25 21:47:45
>>> -0500
>>> [0]PETSC ERROR: See docs/changes/index.html for recent updates.
>>> [0]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
>>> [0]PETSC ERROR: See docs/index.html for manual pages.
>>> [0]PETSC ERROR:
>>> ------------------------------------------------------------------------
>>> [0]PETSC ERROR: build/poissonIrregular on a arch-linux2-cxx-debug named
>>> edwards by bkhanal Thu Oct 31 19:23:58 2013
>>> [0]PETSC ERROR: Libraries linked from
>>> /home/bkhanal/Documents/softwares/petsc/arch-linux2-cxx-debug/lib
>>> [0]PETSC ERROR: Configure run at Sat Oct 26 16:35:15 2013
>>> [0]PETSC ERROR: Configure options --download-mpich
>>> -download-f-blas-lapack=1 --download-metis --download-parmetis
>>> --download-superlu_dist --download-scalapack --download-mumps
>>> --download-hypre --with-clanguage=cxx
>>> [0]PETSC ERROR:
>>> ------------------------------------------------------------------------
>>> [0]PETSC ERROR: MatSetValues_SeqAIJ() line 413 in
>>> src/mat/impls/aij/seq/aij.c
>>> [0]PETSC ERROR: MatSetValues() line 1130 in src/mat/interface/matrix.c
>>> [0]PETSC ERROR: computeMatrix2dSection() line 318 in
>>> /user/bkhanal/home/works/poissonIrregular/poissonIrregular.cxx
>>> [0]PETSC ERROR: KSPSetUp() line 228 in src/ksp/ksp/interface/itfunc.c
>>> [0]PETSC ERROR: KSPSolve() line 399 in src/ksp/ksp/interface/itfunc.c
>>> [0]PETSC ERROR: main() line 598 in
>>> /user/bkhanal/home/works/poissonIrregular/poissonIrregular.cxx
>>> application called MPI_Abort(MPI_COMM_WORLD, 63) - process 0
>>> [unset]: aborting job:
>>> application called MPI_Abort(MPI_COMM_WORLD, 63) - process 0
>>>
>>>
>>>> Matt
>>>>
>>>>
>>>>> However, now since I do not want the rows in the matrix corresponding
>>>>> to the points in the grid which do not lie in the computational domain.
>>>>> This means the size of the matrix will not necessarily equal the total
>>>>> number of cells in DMDA. So in the following code:
>>>>>
>>>>> ierr = DMDAGetLocalInfo(dm,&info);CHKERRQ(ierr);
>>>>> ierr = PetscSectionCreate(PETSC_COMM_WORLD, &s);CHKERRQ(ierr);
>>>>> ierr = DMDAGetNumCells(dm, NULL, NULL, NULL, &nC);CHKERRQ(ierr);
>>>>> ierr = PetscSectionSetChart(s, 0, nC);CHKERRQ(ierr);
>>>>> for (PetscInt j = info.ys; j < info.ys+info.ym; ++j) {
>>>>> for (PetscInt i = info.xs; i < info.xs+info.xm; ++i) {
>>>>> PetscInt point;
>>>>> if(isPosInDomain(&testPoisson,i,j,0)) {
>>>>> ierr = DMDAGetCellPoint(dm, i, j, 0,
>>>>> &point);CHKERRQ(ierr);
>>>>> ierr = PetscSectionSetDof(s, point, testPoisson.mDof);
>>>>> // No. of dofs associated with the point.
>>>>> }
>>>>>
>>>>> }
>>>>> }
>>>>> ierr = PetscSectionSetUp(s);CHKERRQ(ierr);
>>>>> ierr = DMSetDefaultSection(dm, s);CHKERRQ(ierr);
>>>>> ierr = PetscSectionDestroy(&s);CHKERRQ(ierr);
>>>>>
>>>>> should I myself compute proper nC (i.e. total no. of points for which
>>>>> I want the matrix to have a corresponding row) before calling
>>>>> PetscSectionSetChart() or is the masking out of the points inside the loop
>>>>> with if(isPosInDomain(&testPoisson,i,j,0)) sufficient ?
>>>>> And, when you write:
>>>>> PetscSectionGetOffset(gs, p, &ind);
>>>>> row = ind < 0 ? -(ind+1) : ind;
>>>>>
>>>>> it seems you allow the possibility of getting a -ve ind when using
>>>>> PetscSectionGetOffset. When would an offset to a particular dof and point?
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>
>>>>>>> for(PetscInt j = info.ys; j<info.ys+info.ym; ++j) {
>>>>>>> for(PetscInt i = info.xs; i<info.xs+info.xm; ++i) {
>>>>>>> num = 0;
>>>>>>> /*ierr =
>>>>>>> DMDAGetCellPoint(da,i,j,0,&point);CHKERRQ(ierr);*/ /*Get the PetscPoint*/
>>>>>>> /*But I did now understand how I would emulate the
>>>>>>> row.c and col.c info with PetscSection*/
>>>>>>> row.i = i; row.j = j;
>>>>>>>
>>>>>>
>>>>>> Here you would call
>>>>>>
>>>>>> ierr = DMDAGetCellPoint(da, i, j, 0, &cell);CHKERRQ(ierr);
>>>>>> ierr = PetscSectionGetOffset(da, cell, &row);CHKERRQ(ierr);
>>>>>> ierr = PetscSectionGetDof(da, cell, &dof);CHKERRQ(ierr);
>>>>>>
>>>>>> This means that this cell has global rows = [row, row+dof).
>>>>>>
>>>>>>
>>>>>>> col[num].i = i; col[num].j = j;
>>>>>>> if (isPosInDomain(ctx,i,j)) { /*put the
>>>>>>> operator for only the values for only the points lying in the domain*/
>>>>>>> v[num++] = -4;
>>>>>>> if(isPosInDomain(ctx,i+1,j)) {
>>>>>>> col[num].i = i+1; col[num].j = j;
>>>>>>> v[num++] = 1;
>>>>>>> }
>>>>>>> if(isPosInDomain(ctx,i-1,j)) {
>>>>>>> col[num].i = i-1; col[num].j = j;
>>>>>>> v[num++] = 1;
>>>>>>> }
>>>>>>> if(isPosInDomain(ctx,i,j+1)) {
>>>>>>> col[num].i = i; col[num].j = j+1;
>>>>>>> v[num++] = 1;
>>>>>>> }
>>>>>>> if(isPosInDomain(ctx,i,j-1)) {
>>>>>>> col[num].i = i; col[num].j = j-1;
>>>>>>> v[num++] = 1;
>>>>>>> }
>>>>>>> } else {
>>>>>>> v[num++] = dScale; /*set Dirichlet identity
>>>>>>> rows for points in the rectangle but outside the computational domain*/
>>>>>>> }
>>>>>>>
>>>>>>
>>>>>> You do the same thing for the columns, and then call
>>>>>>
>>>>>> ierr = MatSetValues(A, dof, rows, num*dof, cols, v,
>>>>>> INSERT_VALUES);CHKERRQ(ierr);
>>>>>>
>>>>>>
>>>>>>> ierr =
>>>>>>> MatSetValuesStencil(A,1,&row,num,col,v,INSERT_VALUES);CHKERRQ(ierr);
>>>>>>> }
>>>>>>> }
>>>>>>> }
>>>>>>>
>>>>>>>
>>>>>>>> Thanks,
>>>>>>>>
>>>>>>>> Matt
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>>> Matt
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>> Matt
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Matt
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> If not, just put the identity into
>>>>>>>>>>>>>>>>>> the rows you do not use on the full cube. It will not
>>>>>>>>>>>>>>>>>> hurt scalability or convergence.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> In the case of Poisson with Dirichlet condition this might
>>>>>>>>>>>>>>>>> be the case. But is it always true that having identity rows in the system
>>>>>>>>>>>>>>>>> matrix will not hurt convergence ? I thought otherwise for the following
>>>>>>>>>>>>>>>>> reasons:
>>>>>>>>>>>>>>>>> 1) Having read Jed's answer here :
>>>>>>>>>>>>>>>>> http://scicomp.stackexchange.com/questions/3426/why-is-pinning-a-point-to-remove-a-null-space-bad/3427#3427
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Jed is talking about a constraint on a the pressure at a
>>>>>>>>>>>>>>>> point. This is just decoupling these unknowns from the rest
>>>>>>>>>>>>>>>> of the problem.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 2) Some observation I am getting (but I am still doing
>>>>>>>>>>>>>>>>> more experiments to confirm) while solving my staggered-grid 3D stokes flow
>>>>>>>>>>>>>>>>> with schur complement and using -pc_type gamg for A00 matrix. Putting the
>>>>>>>>>>>>>>>>> identity rows for dirichlet boundaries and for ghost cells seemed to have
>>>>>>>>>>>>>>>>> effects on its convergence. I'm hoping once I know how to use PetscSection,
>>>>>>>>>>>>>>>>> I can get rid of using ghost cells method for the staggered grid and get
>>>>>>>>>>>>>>>>> rid of the identity rows too.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> It can change the exact iteration, but it does not make the
>>>>>>>>>>>>>>>> matrix conditioning worse.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Matt
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Anyway please provide me with some pointers so that I
>>>>>>>>>>>>>>>>> can start trying with petscsection on top of a dmda, in the beginning for
>>>>>>>>>>>>>>>>> non-staggered case.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>>>>>> Bishesh
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Matt
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>>>>>>>> Bishesh
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> --
>>>>>>>>>>>>>>>>>> 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
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> --
>>>>>>>>>>>>>>>> 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
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> --
>>>>>>>>>>>>>> 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
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> --
>>>>>>>>>>>> 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
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> --
>>>>>>>>>> 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
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> 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
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> 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
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> 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
>>>>
>>>
>>>
>>
>>
>> --
>> 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
>>
>
>
--
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20131031/e4f37919/attachment-0001.html>
More information about the petsc-users
mailing list