Question about PCFieldSplit
Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae
The only example currently supports the splitting in dmsag is through PCFieldSplitSetDetectSaddlePoint. One approach might be writing a dummy solver, fill one block as diagonal 0, call PCFieldSplitSetDetectSaddlePoint and get its IC, and then use IS in the actual solver. I can see it should work but it would be good if there will be an easier solution. @Zakariae, I believe dofs get mixed together in dmstag Thanks, Qi -- Qi Tang T5@LANL On Apr 16, 2021, at 6:39 PM, Jorti, Zakariae via petsc-users <[email protected]<mailto:[email protected]>> wrote: Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users < [email protected]> wrote:
Hello,
I have a DMStag grid with one dof on each edge and face center.
I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields.
In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries.
Can I use that PCFieldSplitSetDetectSaddlePoint in this case?
If not, how should I do?
Should I do like this example ( https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html ):
const PetscInt Bfields[1] = {0},Efields[1] = {1};
KSPGetPC(ksp,&pc);
PCFieldSplitSetBlockSize(pc,2);
PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields);
where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA. Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes? Thanks, Matt
One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you.
Best regards,
Zakariae
-- 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 https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay! You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected] <mailto:[email protected]>> wrote: Hello,
I have a DMStag grid with one dof on each edge and face center.
I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields.
In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries.
Can I use that PCFieldSplitSetDetectSaddlePoint in this case?
If not, how should I do?
Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html>):
const PetscInt Bfields[1] = {0},Efields[1] = {1};
KSPGetPC(ksp,&pc);
PCFieldSplitSetBlockSize(pc,2);
PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields);
where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you.
Best regards,
Zakariae
-- 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
https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
Thanks a lot, Patrick. We appreciate your help. Qi On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay! You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly. Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected]<mailto:[email protected]>> wrote: Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html<https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? That will not work.That interface only works for colocated fields that you get from DMDA. Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes? Thanks, Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/<https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
Patrick, We were able to get field split working in DMStag with the dummy solver approach I described. Let us know when you add the capability and we will be happy to test it. Meanwhile, please take your time to merge it. Thanks, Qi On Apr 18, 2021, at 11:51 PM, Tang, Qi <[email protected]<mailto:[email protected]>> wrote: Thanks a lot, Patrick. We appreciate your help. Qi On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay! You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly. Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected]<mailto:[email protected]>> wrote: Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html<https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? That will not work.That interface only works for colocated fields that you get from DMDA. Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes? Thanks, Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/<https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
That's great - if you run into issues that I can help with in the meantime, let me know. Even if you can't share the code, maybe you could share a screen or something like that. This would undoubtedly be helpful for me in seeing how people use the code "in the wild" : )
Am 20.04.2021 um 08:29 schrieb Tang, Qi <[email protected]>:
Patrick, We were able to get field split working in DMStag with the dummy solver approach I described. Let us know when you add the capability and we will be happy to test it. Meanwhile, please take your time to merge it.
Thanks, Qi
On Apr 18, 2021, at 11:51 PM, Tang, Qi <[email protected] <mailto:[email protected]>> wrote:
Thanks a lot, Patrick. We appreciate your help.
Qi
On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected] <mailto:[email protected]>> wrote:
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay!
You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected] <mailto:[email protected]>>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected] <mailto:[email protected]>> wrote: Hello,
I have a DMStag grid with one dof on each edge and face center.
I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields.
In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries.
Can I use that PCFieldSplitSetDetectSaddlePoint in this case?
If not, how should I do?
Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>):
const PetscInt Bfields[1] = {0},Efields[1] = {1};
KSPGetPC(ksp,&pc);
PCFieldSplitSetBlockSize(pc,2);
PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields);
where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you.
Best regards,
Zakariae
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
https://www.cse.buffalo.edu/~knepley/ <https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
Hello, I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits. Many thanks. Best regards, Zakariae ________________________________ From: petsc-users <[email protected]> on behalf of Tang, Qi <[email protected]> Sent: Sunday, April 18, 2021 11:51:59 PM To: Patrick Sanan Cc: [email protected]; Tang, Xianzhu Subject: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit Thanks a lot, Patrick. We appreciate your help. Qi On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay! You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly. Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected]<mailto:[email protected]>> wrote: Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html<https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? That will not work.That interface only works for colocated fields that you get from DMDA. Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes? Thanks, Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/<https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
Hi Zakariae - The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] . There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch. I'll track it down and see if it can quickly be added to the main branch. [1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html
Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]>:
Hello,
I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits.
Many thanks.
Best regards,
Zakariae From: petsc-users <[email protected]> on behalf of Tang, Qi <[email protected]> Sent: Sunday, April 18, 2021 11:51:59 PM To: Patrick Sanan Cc: [email protected]; Tang, Xianzhu Subject: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Thanks a lot, Patrick. We appreciate your help.
Qi
On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected] <mailto:[email protected]>> wrote:
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay!
You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected] <mailto:[email protected]>>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected] <mailto:[email protected]>> wrote: Hello,
I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you. Best regards,
Zakariae
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
https://www.cse.buffalo.edu/~knepley/ <https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
On Wed, Jun 23, 2021 at 12:51 AM Patrick Sanan <[email protected]> wrote:
Hi Zakariae -
The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] .
There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch.
I'll track it down and see if it can quickly be added to the main branch.
Note that an easy way to keep track of this would be to create a section with the different locations as fields. This Section could then easily create the ISes, and could automatically interface with PCFIELDSPLIT. Thanks, Matt
[1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html
Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]>:
Hello,
I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits.
Many thanks.
Best regards,
Zakariae ------------------------------ *From:* petsc-users <[email protected]> on behalf of Tang, Qi <[email protected]> *Sent:* Sunday, April 18, 2021 11:51:59 PM *To:* Patrick Sanan *Cc:* [email protected]; Tang, Xianzhu *Subject:* [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Thanks a lot, Patrick. We appreciate your help.
Qi
On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]> wrote:
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay!
You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users < [email protected]> wrote:
Hello,
I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example ( https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$> ): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt
One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you. Best regards,
Zakariae
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
https://www.cse.buffalo.edu/~knepley/ <https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
-- 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 https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
Here is an MR which intends to introduce some logic to support DMCreateFieldDecomposition(). It doesn't use the PetscSection approach, which might be preferable, but nonetheless is a necessary component so It'd like to get it in, even if it has room for further optimization. Hopefully this can be followed fairly soon with some more examples and tests using PCFieldSplit itself. https://gitlab.com/petsc/petsc/-/merge_requests/4740 Am Mi., 23. Juni 2021 um 12:15 Uhr schrieb Matthew Knepley < [email protected]>:
On Wed, Jun 23, 2021 at 12:51 AM Patrick Sanan <[email protected]> wrote:
Hi Zakariae -
The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] .
There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch.
I'll track it down and see if it can quickly be added to the main branch.
Note that an easy way to keep track of this would be to create a section with the different locations as fields. This Section could then easily create the ISes, and could automatically interface with PCFIELDSPLIT.
Thanks,
Matt
[1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html
Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]>:
Hello,
I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits.
Many thanks.
Best regards,
Zakariae ------------------------------ *From:* petsc-users <[email protected]> on behalf of Tang, Qi <[email protected]> *Sent:* Sunday, April 18, 2021 11:51:59 PM *To:* Patrick Sanan *Cc:* [email protected]; Tang, Xianzhu *Subject:* [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Thanks a lot, Patrick. We appreciate your help.
Qi
On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]> wrote:
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay!
You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users < [email protected]> wrote:
Hello,
I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example ( https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$> ): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt
One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you. Best regards,
Zakariae
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
https://www.cse.buffalo.edu/~knepley/ <https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
-- 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
https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
Hi Patrick, Thanks for your recent updates on DMStag. After getting the Finite Difference Coloring to work with our solver, I was testing that DMCreateFieldDecomposition routine that you added last week. It seems to work fine when there is only one unknown per location (i.e. one unknown on vertices and/or one unknown on edges and/or one unknown on faces and/or one unknown on elements). That being said, when there is more than one unknown in some location (let's say 2 unknowns on vertices for instance), I could not get the ISs for those two unknowns with that routine. Should I still rely on PCFieldSplitSetDetectSaddlePoint in this case? Many thanks once again. Kind regards, Zakariae ________________________________ From: Patrick Sanan <[email protected]> Sent: Tuesday, January 25, 2022 5:36:17 AM To: Matthew Knepley Cc: Jorti, Zakariae; [email protected]; Tang, Xianzhu; Dave May Subject: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit Here is an MR which intends to introduce some logic to support DMCreateFieldDecomposition(). It doesn't use the PetscSection approach, which might be preferable, but nonetheless is a necessary component so It'd like to get it in, even if it has room for further optimization. Hopefully this can be followed fairly soon with some more examples and tests using PCFieldSplit itself. https://gitlab.com/petsc/petsc/-/merge_requests/4740 Am Mi., 23. Juni 2021 um 12:15 Uhr schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Wed, Jun 23, 2021 at 12:51 AM Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: Hi Zakariae - The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] . There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch. I'll track it down and see if it can quickly be added to the main branch. Note that an easy way to keep track of this would be to create a section with the different locations as fields. This Section could then easily create the ISes, and could automatically interface with PCFIELDSPLIT. Thanks, Matt [1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]<mailto:[email protected]>>: Hello, I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits. Many thanks. Best regards, Zakariae ________________________________ From: petsc-users <[email protected]<mailto:[email protected]>> on behalf of Tang, Qi <[email protected]<mailto:[email protected]>> Sent: Sunday, April 18, 2021 11:51:59 PM To: Patrick Sanan Cc: [email protected]<mailto:[email protected]>; Tang, Xianzhu Subject: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit Thanks a lot, Patrick. We appreciate your help. Qi On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay! You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly. Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected]<mailto:[email protected]>> wrote: Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html<https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? That will not work.That interface only works for colocated fields that you get from DMDA. Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes? Thanks, Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/<https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$> -- 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 https://www.cse.buffalo.edu/~knepley/<http://www.cse.buffalo.edu/~knepley/>
The current behavior is that a single IS is returned for each stratum, so if you have 2 unknowns on vertices, it should still return a single IS including both of those unknowns per vertex. Am I understanding that that's working as expected but you need *two* ISs in that case? Am Mo., 31. Jan. 2022 um 16:29 Uhr schrieb Jorti, Zakariae <[email protected]
:
Hi Patrick,
Thanks for your recent updates on DMStag. After getting the Finite Difference Coloring to work with our solver, I was testing that DMCreateFieldDecomposition routine that you added last week. It seems to work fine when there is only one unknown per location (i.e. one unknown on vertices and/or one unknown on edges and/or one unknown on faces and/or one unknown on elements). That being said, when there is more than one unknown in some location (let's say 2 unknowns on vertices for instance), I could not get the ISs for those two unknowns with that routine. Should I still rely on PCFieldSplitSetDetectSaddlePoint in this case? Many thanks once again.
Kind regards,
Zakariae ------------------------------ *From:* Patrick Sanan <[email protected]> *Sent:* Tuesday, January 25, 2022 5:36:17 AM *To:* Matthew Knepley *Cc:* Jorti, Zakariae; [email protected]; Tang, Xianzhu; Dave May *Subject:* [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Here is an MR which intends to introduce some logic to support DMCreateFieldDecomposition(). It doesn't use the PetscSection approach, which might be preferable, but nonetheless is a necessary component so It'd like to get it in, even if it has room for further optimization. Hopefully this can be followed fairly soon with some more examples and tests using PCFieldSplit itself.
https://gitlab.com/petsc/petsc/-/merge_requests/4740
Am Mi., 23. Juni 2021 um 12:15 Uhr schrieb Matthew Knepley < [email protected]>:
On Wed, Jun 23, 2021 at 12:51 AM Patrick Sanan <[email protected]> wrote:
Hi Zakariae -
The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] .
There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch.
I'll track it down and see if it can quickly be added to the main branch.
Note that an easy way to keep track of this would be to create a section with the different locations as fields. This Section could then easily create the ISes, and could automatically interface with PCFIELDSPLIT.
Thanks,
Matt
[1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html
Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]>:
Hello,
I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits.
Many thanks.
Best regards,
Zakariae ------------------------------ *From:* petsc-users <[email protected]> on behalf of Tang, Qi <[email protected]> *Sent:* Sunday, April 18, 2021 11:51:59 PM *To:* Patrick Sanan *Cc:* [email protected]; Tang, Xianzhu *Subject:* [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Thanks a lot, Patrick. We appreciate your help.
Qi
On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]> wrote:
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay!
You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users < [email protected]> wrote:
Hello,
I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example ( https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$> ): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt
One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you. Best regards,
Zakariae
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
https://www.cse.buffalo.edu/~knepley/ <https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
-- 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
https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
Yes, exactly. ________________________________ From: Patrick Sanan <[email protected]> Sent: Monday, January 31, 2022 8:47:23 AM To: Jorti, Zakariae Cc: [email protected]; Tang, Xianzhu; Tang, Qi Subject: Re: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit The current behavior is that a single IS is returned for each stratum, so if you have 2 unknowns on vertices, it should still return a single IS including both of those unknowns per vertex. Am I understanding that that's working as expected but you need *two* ISs in that case? Am Mo., 31. Jan. 2022 um 16:29 Uhr schrieb Jorti, Zakariae <[email protected]<mailto:[email protected]>>: Hi Patrick, Thanks for your recent updates on DMStag. After getting the Finite Difference Coloring to work with our solver, I was testing that DMCreateFieldDecomposition routine that you added last week. It seems to work fine when there is only one unknown per location (i.e. one unknown on vertices and/or one unknown on edges and/or one unknown on faces and/or one unknown on elements). That being said, when there is more than one unknown in some location (let's say 2 unknowns on vertices for instance), I could not get the ISs for those two unknowns with that routine. Should I still rely on PCFieldSplitSetDetectSaddlePoint in this case? Many thanks once again. Kind regards, Zakariae ________________________________ From: Patrick Sanan <[email protected]<mailto:[email protected]>> Sent: Tuesday, January 25, 2022 5:36:17 AM To: Matthew Knepley Cc: Jorti, Zakariae; [email protected]<mailto:[email protected]>; Tang, Xianzhu; Dave May Subject: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit Here is an MR which intends to introduce some logic to support DMCreateFieldDecomposition(). It doesn't use the PetscSection approach, which might be preferable, but nonetheless is a necessary component so It'd like to get it in, even if it has room for further optimization. Hopefully this can be followed fairly soon with some more examples and tests using PCFieldSplit itself. https://gitlab.com/petsc/petsc/-/merge_requests/4740 Am Mi., 23. Juni 2021 um 12:15 Uhr schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Wed, Jun 23, 2021 at 12:51 AM Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: Hi Zakariae - The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] . There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch. I'll track it down and see if it can quickly be added to the main branch. Note that an easy way to keep track of this would be to create a section with the different locations as fields. This Section could then easily create the ISes, and could automatically interface with PCFIELDSPLIT. Thanks, Matt [1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]<mailto:[email protected]>>: Hello, I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits. Many thanks. Best regards, Zakariae ________________________________ From: petsc-users <[email protected]<mailto:[email protected]>> on behalf of Tang, Qi <[email protected]<mailto:[email protected]>> Sent: Sunday, April 18, 2021 11:51:59 PM To: Patrick Sanan Cc: [email protected]<mailto:[email protected]>; Tang, Xianzhu Subject: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit Thanks a lot, Patrick. We appreciate your help. Qi On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]<mailto:[email protected]>> wrote: We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay! You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly. Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]<mailto:[email protected]>>: On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users <[email protected]<mailto:[email protected]>> wrote: Hello, I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example (https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html<https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$>): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers? That will not work.That interface only works for colocated fields that you get from DMDA. Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes? Thanks, Matt One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs. Thank you. Best regards, Zakariae -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/<https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$> -- 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 https://www.cse.buffalo.edu/~knepley/<http://www.cse.buffalo.edu/~knepley/>
On Mon, Jan 31, 2022 at 10:48 AM Jorti, Zakariae via petsc-users < [email protected]> wrote:
Yes, exactly.
If you already have the FieldSplit for the 2 dofs on vertices, you can easily recursively split this into 1 dof per block, since those are laid out like a colocated field. You just give another PCFIELDSPLIT of size 2 and the split will be automatic. Thanks, Matt
------------------------------ *From:* Patrick Sanan <[email protected]> *Sent:* Monday, January 31, 2022 8:47:23 AM *To:* Jorti, Zakariae *Cc:* [email protected]; Tang, Xianzhu; Tang, Qi *Subject:* Re: [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
The current behavior is that a single IS is returned for each stratum, so if you have 2 unknowns on vertices, it should still return a single IS including both of those unknowns per vertex. Am I understanding that that's working as expected but you need *two* ISs in that case?
Am Mo., 31. Jan. 2022 um 16:29 Uhr schrieb Jorti, Zakariae < [email protected]>:
Hi Patrick,
Thanks for your recent updates on DMStag. After getting the Finite Difference Coloring to work with our solver, I was testing that DMCreateFieldDecomposition routine that you added last week. It seems to work fine when there is only one unknown per location (i.e. one unknown on vertices and/or one unknown on edges and/or one unknown on faces and/or one unknown on elements). That being said, when there is more than one unknown in some location (let's say 2 unknowns on vertices for instance), I could not get the ISs for those two unknowns with that routine. Should I still rely on PCFieldSplitSetDetectSaddlePoint in this case? Many thanks once again.
Kind regards,
Zakariae ------------------------------ *From:* Patrick Sanan <[email protected]> *Sent:* Tuesday, January 25, 2022 5:36:17 AM *To:* Matthew Knepley *Cc:* Jorti, Zakariae; [email protected]; Tang, Xianzhu; Dave May *Subject:* [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Here is an MR which intends to introduce some logic to support DMCreateFieldDecomposition(). It doesn't use the PetscSection approach, which might be preferable, but nonetheless is a necessary component so It'd like to get it in, even if it has room for further optimization. Hopefully this can be followed fairly soon with some more examples and tests using PCFieldSplit itself.
https://gitlab.com/petsc/petsc/-/merge_requests/4740
Am Mi., 23. Juni 2021 um 12:15 Uhr schrieb Matthew Knepley < [email protected]>:
On Wed, Jun 23, 2021 at 12:51 AM Patrick Sanan <[email protected]> wrote:
Hi Zakariae -
The usual way to do this is to define an IS (index set) with the degrees of freedom of interest for the rows, and another one for the columns, and then use MatCreateSubmatrix [1] .
There's not a particularly convenient way to create an IS with the degrees of freedom corresponding to a particular "stratum" (i.e. elements, faces, edges, or vertices) of a DMStag, but fortunately I believe we have some code to do exactly this in a development branch.
I'll track it down and see if it can quickly be added to the main branch.
Note that an easy way to keep track of this would be to create a section with the different locations as fields. This Section could then easily create the ISes, and could automatically interface with PCFIELDSPLIT.
Thanks,
Matt
[1]: https://petsc.org/release/docs/manualpages/Mat/MatCreateSubMatrix.html
Am 22.06.2021 um 22:29 schrieb Jorti, Zakariae <[email protected]>:
Hello,
I am working on DMStag and I have one dof on vertices (let us call it V), one dof on edges (let us call it E), one dof on faces ((let us call it F)) and one dof on cells (let us call it C). I build a matrix on this DM, and I was wondering if there was a way to get blocks (or sub matrices) of this matrix corresponding to specific degrees of freedom, for example rows corresponding to V dofs and columns corresponding to E dofs. I already asked this question before and the answer I got was I could call PCFieldSplitSetDetectSaddlePoint with the diagonal entries being of the matrix being zero or nonzero. That worked well. Nonetheless, I am curious to know if there was another alternative that does not require creating a dummy matrix with appropriate diagonal entries and solving a dummy linear system with this matrix to define the splits.
Many thanks.
Best regards,
Zakariae ------------------------------ *From:* petsc-users <[email protected]> on behalf of Tang, Qi <[email protected]> *Sent:* Sunday, April 18, 2021 11:51:59 PM *To:* Patrick Sanan *Cc:* [email protected]; Tang, Xianzhu *Subject:* [EXTERNAL] Re: [petsc-users] Question about PCFieldSplit
Thanks a lot, Patrick. We appreciate your help.
Qi
On Apr 18, 2021, at 11:30 PM, Patrick Sanan <[email protected]> wrote:
We have this functionality in a branch, which I'm working on cleaning up to get to master. It doesn't use PETScSection. Sorry about the delay!
You can only use PCFieldSplitSetDetectSaddlePoint when your diagonal entries being zero or non-zero defines the splits correctly.
Am 17.04.2021 um 21:09 schrieb Matthew Knepley <[email protected]>:
On Fri, Apr 16, 2021 at 8:39 PM Jorti, Zakariae via petsc-users < [email protected]> wrote:
Hello,
I have a DMStag grid with one dof on each edge and face center. I want to use a PCFieldSplit preconditioner on a Jacobian matrix that I assume is already split but I am not sure how to determine the fields. In the DMStag examples (ex2.c and ex3.c), the function PCFieldSplitSetDetectSaddlePoint is used to determine those fields based on zero diagonal entries. In my case, I have a Jacobian matrix that does not have zero diagonal entries. Can I use that PCFieldSplitSetDetectSaddlePoint in this case? If not, how should I do? Should I do like this example ( https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html <https://urldefense.com/v3/__https://www.mcs.anl.gov/petsc/petsc-master/src/ksp/ksp/tutorials/ex43.c.html__;!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcnksFSO3Q$> ): const PetscInt Bfields[1] = {0},Efields[1] = {1}; KSPGetPC(ksp,&pc); PCFieldSplitSetBlockSize(pc,2); PCFieldSplitSetFields(pc,"B",1,Bfields,Bfields); PCFieldSplitSetFields(pc,"E",1,Efields,Efields); where my B unknowns are defined on face centers and E unknowns are defined on edge centers?
That will not work.That interface only works for colocated fields that you get from DMDA.
Patrick, does DMSTAG use PetscSection? Then the field split would be automatically calculated. If not, does it maintain the field division so that it could be given to PCFIELDSPLIT as ISes?
Thanks,
Matt
One last thing, I do not know which field comes first. Is it the one defined for face dofs or edge dofs.
Thank you. Best regards,
Zakariae
-- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
https://www.cse.buffalo.edu/~knepley/ <https://urldefense.com/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!HXCxUKc!jbBwV2h9luOW4dtBcNh6n_W1ULQnSVeXpxl0Ef1752s4Hlef-nC2JcmGgSwfag$>
-- 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
https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
-- 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 https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
participants (4)
-
Jorti, Zakariae -
Matthew Knepley -
Patrick Sanan -
Tang, Qi