[petsc-users] Modify matrix nonzero structure

Barry Smith bsmith at petsc.dev
Tue May 28 20:27:32 CDT 2024


   There was a bug in my fix for parallel which I have fixed. You will need to 

   git fetch
   git checkout main
   git branch -D barry/2024-05-27/fix-bug-baij-setvaluesblocked/release
   git checkout barry/2024-05-27/fix-bug-baij-setvaluesblocked/release

    I get the same results with your example with 1,2, and 3 ranks


> On May 28, 2024, at 7:45 PM, Adrian Croucher <a.croucher at auckland.ac.nz> wrote:
> 
> hi Barry,
> 
> Thanks, that change has fixed the error on 2 ranks for me.
> 
> When I run on 3 ranks, there is no error, but it doesn't actually add the extra values in to the matrix. Do you see that behaviour too?
> 
> - Adrian
> 
> On 29/05/24 4:33 am, Barry Smith wrote:
>> 
>>    Adrian,
>> 
>>    I could reproduce with 3 MPI ranks.
>> 
>>    Another error I had to fix. I also added a test example 
>> 
>>    SInce I rebased the branch you will need to do something like
>> 
>>    git fetch
>>    git checkout main
>>    git branch -D barry/2024-05-27/fix-bug-baij-setvaluesblocked/release
>>   git checkout barry/2024-05-27/fix-bug-baij-setvaluesblocked/release
>> 
>>   Thanks for your patience 
>> 
>>    Barry
>> 
>> 
>>> On May 27, 2024, at 10:42 PM, Adrian Croucher <a.croucher at auckland.ac.nz> <mailto:a.croucher at auckland.ac.nz> wrote:
>>> 
>>> Hmm, that's a bit weird. I haven't modified the test code - I checked the file date to make sure. I also tried deleting my PETSc build dir and rebuilding it, then rebuilding the test code. I still get the error if I run on 2 ranks with -dm_mat_type (or -mat_type) baij or mpibaij, but it's fine on aij or mpiaij.
>>> 
>>> My actual (non-test) code is however working ok now, in serial or parallel. So I don't think this should hold up merging your bugfix.
>>> 
>>> - Adrian
>>> 
>>> On 28/05/24 2:13 pm, Barry Smith wrote:
>>>>  
>>>>   When I run the exact code you sent with two ranks and—mat_type mpibaij, it runs as expected. If you modified the code in any way to demonstrate the bug, please send the modified code.
>>>> 
>>>> 
>>>> 
>>>>> On May 27, 2024, at 9:37 PM, Adrian Croucher <a.croucher at auckland.ac.nz> <mailto:a.croucher at auckland.ac.nz> wrote:
>>>>> 
>>>>> hi Barry,
>>>>> 
>>>>> On 28/05/24 7:46 am, Barry Smith wrote:
>>>>>> 
>>>>>>    Thanks for reporting this. It is a bug. I have a fixed branch barry/2024-05-27/fix-bug-baij-setvaluesblocked/release  and associated merge request https://urldefense.us/v3/__https://gitlab.com/petsc/petsc/-/merge_requests/7578__;!!G_uCfscf7eWS!dngbcXmJfofRckcEMElrf4BvXnHLnRh4FePsIwherUCHmsPBqTWSGvW36-62EUO3uDDIffqW8q9rei8toRdYLUY$ 
>>>>>> 
>>>>> Thanks very much, that does appear to fix the bug when I run my test program in serial.
>>>>> 
>>>>> If I run it on 2 processes, it is OK with AIJ matrix type, but with BAIJ I get an error (see below). Is there another problem, or I am doing something else wrong?
>>>>> 
>>>>> - Adrian
>>>>> 
>>>>> [0]PETSC ERROR: --------------------- Error Message --------------------------------------------------------------
>>>>> [0]PETSC ERROR: Petsc has generated inconsistent data
>>>>> [0]PETSC ERROR: Incorrect colmap
>>>>> [0]PETSC ERROR: See https://urldefense.us/v3/__https://petsc.org/release/faq/__;!!G_uCfscf7eWS!dngbcXmJfofRckcEMElrf4BvXnHLnRh4FePsIwherUCHmsPBqTWSGvW36-62EUO3uDDIffqW8q9rei8tCH9MbY4$  for trouble shooting.
>>>>> [0]PETSC ERROR: Petsc Release Version 3.21.1, unknown 
>>>>> [0]PETSC ERROR: ./matmodify on a main-debug named EN438880 by acro018 Tue May 28 13:33:46 2024
>>>>> [0]PETSC ERROR: Configure options --with-x --download-hdf5 --download-zlib --download-netcdf --download-pnetcdf --download-exodusii --download-triangle --download-ptscotch --download-chaco --download-hypre
>>>>> [0]PETSC ERROR: #1 MatSetValuesBlocked_MPIBAIJ() at /home/acro018/software/PETSc/code/src/mat/impls/baij/mpi/mpibaij.c:448
>>>>> [0]PETSC ERROR: #2 MatSetValuesBlocked() at /home/acro018/software/PETSc/code/src/mat/interface/matrix.c:2030
>>>>> [0]PETSC ERROR: --------------------- Error Message --------------------------------------------------------------
>>>>> [0]PETSC ERROR:   It appears a new error in the code was triggered after a previous error, possibly because:
>>>>> [0]PETSC ERROR:   -  The first error was not properly handled via (for example) the use of
>>>>> [0]PETSC ERROR:      PetscCall(TheFunctionThatErrors()); or
>>>>> [0]PETSC ERROR:   -  The second error was triggered while handling the first error.
>>>>> [0]PETSC ERROR:   Above is the traceback for the previous unhandled error, below the traceback for the next error
>>>>> [0]PETSC ERROR:   ALL ERRORS in the PETSc libraries are fatal, you should add the appropriate error checking to the code
>>>>> [0]PETSC ERROR: Petsc has generated inconsistent data
>>>>> [0]PETSC ERROR: Incorrect colmap
>>>>> [0]PETSC ERROR: See https://urldefense.us/v3/__https://petsc.org/release/faq/__;!!G_uCfscf7eWS!dngbcXmJfofRckcEMElrf4BvXnHLnRh4FePsIwherUCHmsPBqTWSGvW36-62EUO3uDDIffqW8q9rei8tCH9MbY4$  for trouble shooting.
>>>>> [0]PETSC ERROR: Petsc Release Version 3.21.1, unknown 
>>>>> [0]PETSC ERROR: ./matmodify on a main-debug named EN438880 by acro018 Tue May 28 13:33:46 2024
>>>>> [0]PETSC ERROR: Configure options --with-x --download-hdf5 --download-zlib --download-netcdf --download-pnetcdf --download-exodusii --download-triangle --download-ptscotch --download-chaco --download-hypre
>>>>> [0]PETSC ERROR: #1 MatSetValuesBlocked_MPIBAIJ() at /home/acro018/software/PETSc/code/src/mat/impls/baij/mpi/mpibaij.c:448
>>>>> [0]PETSC ERROR: #2 MatAssemblyEnd_MPIBAIJ() at /home/acro018/software/PETSc/code/src/mat/impls/baij/mpi/mpibaij.c:943
>>>>> [0]PETSC ERROR: #3 MatAssemblyEnd() at /home/acro018/software/PETSc/code/src/mat/interface/matrix.c:5820
>>>>> [0]PETSC ERROR: #4 matmodify.F90:40
>>>>> 
>>>>> 
>>>>> 
>>>>>>   Barry
>>>>>> 
>>>>>> 
>>>>>>> On May 26, 2024, at 10:45 PM, Adrian Croucher <a.croucher at auckland.ac.nz> <mailto:a.croucher at auckland.ac.nz> wrote:
>>>>>>> 
>>>>>>> hi,
>>>>>>> 
>>>>>>> I've been trying creating a matrix with DMCreateMatrix() and then adding extra blocks of nonzeros into it using MatSetValuesBlocked(), but getting some unexpected results if I set the matrix type to BAIJ. It seems to behave as expected if I use matrix type AIJ.
>>>>>>> 
>>>>>>> I've attached a minimal example program. It reads in the DMPlex from file, sets up a section on it, creates a matrix (blocksize 2) and then inserts a single 2x2 block at global block indices (0,7). It views the matrix before and after the insertion.
>>>>>>> 
>>>>>>> If I run with "-dm_mat_type aij" it gives the expected results, but with "-dm_mat_type baij" it doesn't - e.g. if run in serial, it adds the new nonzeros in the right place but also adds a whole lot of other duplicated entries in block row 0.
>>>>>>> 
>>>>>>> Is there something I'm not understanding about BAIJ, or about MatSetValuesBlocked()? or possibly some other mistake?
>>>>>>> 
>>>>>>> - Adrian
>>>>>>> 
>>>>>>> On 20/05/24 12:24 pm, Barry Smith wrote:
>>>>>>>> 
>>>>>>>>    You can call MatSetOption(mat,MAT_NEW_NONZERO_LOCATION_ERR) then insert the new values. If it is just a handful of new insertions the extra time should be small.
>>>>>>>> 
>>>>>>>>     Making a copy of the matrix won't give you a new matrix that is any faster to insert into so best to just use the same matrix.
>>>>>>>> 
>>>>>>>>   Barry
>>>>>>>> 
>>>>>>>> 
>>>>>>>>> On May 19, 2024, at 7:44 PM, Adrian Croucher <a.croucher at auckland.ac.nz> <mailto:a.croucher at auckland.ac.nz> wrote:
>>>>>>>>> 
>>>>>>>>> This Message Is From an External Sender
>>>>>>>>> This message came from outside your organization.
>>>>>>>>> hi,
>>>>>>>>> 
>>>>>>>>> I have a Jacobian matrix created using DMCreateMatrix(). What would be 
>>>>>>>>> the best way to add extra nonzero entries into it?
>>>>>>>>> 
>>>>>>>>> I'm guessing that DMCreateMatrix() allocates the storage so the nonzero 
>>>>>>>>> structure can't really be easily modified. Would it be a case of 
>>>>>>>>> creating a new matrix, copying the nonzero entries from the original one 
>>>>>>>>> and then adding the extra ones, before calling MatSetUp() or similar? If 
>>>>>>>>> so, how exactly would you copy the nonzero structure from the original 
>>>>>>>>> matrix?
>>>>>>>>> 
>>>>>>>>> Background: the flow problem I'm solving (on a DMPlex with finite volume 
>>>>>>>>> method) has complex source terms that depend on the solution (e.g. 
>>>>>>>>> pressure), and can also depend on other source terms. A simple example 
>>>>>>>>> is when fluid is extracted from one location, with a pressure-dependent 
>>>>>>>>> flow rate, and some of it is then reinjected in another location. This 
>>>>>>>>> can result in poor nonlinear solver convergence. I think the reason is 
>>>>>>>>> that there are effectively missing Jacobian entries in the row for the 
>>>>>>>>> reinjection cell, which should have an additional dependence on the 
>>>>>>>>> solution in the cell where fluid is extracted.
>>>>>>>>> 
>>>>>>>>> - Adrian
>>>>>>>>> 
>>>>>>>>> -- 
>>>>>>>>> Dr Adrian Croucher
>>>>>>>>> Senior Research Fellow
>>>>>>>>> Department of Engineering Science
>>>>>>>>> Waipapa Taumata Rau / University of Auckland, New Zealand
>>>>>>>>> email: a.croucher at auckland.ac.nz <mailto:a.croucher at auckland.ac.nz>
>>>>>>>>> tel: +64 (0)9 923 4611
>>>>>>>>> 
>>>>>>>>> 
>>>>>>>> 
>>>>>>> -- 
>>>>>>> Dr Adrian Croucher
>>>>>>> Senior Research Fellow
>>>>>>> Department of Engineering Science
>>>>>>> Waipapa Taumata Rau / University of Auckland, New Zealand
>>>>>>> email: a.croucher at auckland.ac.nz <mailto:a.croucher at auckland.ac.nz>
>>>>>>> tel: +64 (0)9 923 4611
>>>>>>> <3x3grid.exo><matmodify.F90>
>>>>>> 
>>>>> -- 
>>>>> Dr Adrian Croucher
>>>>> Senior Research Fellow
>>>>> Department of Engineering Science
>>>>> Waipapa Taumata Rau / University of Auckland, New Zealand
>>>>> email: a.croucher at auckland.ac.nz <mailto:a.croucher at auckland.ac.nz>
>>>>> tel: +64 (0)9 923 4611
>>>> 
>>> -- 
>>> Dr Adrian Croucher
>>> Senior Research Fellow
>>> Department of Engineering Science
>>> Waipapa Taumata Rau / University of Auckland, New Zealand
>>> email: a.croucher at auckland.ac.nz <mailto:a.croucher at auckland.ac.nz>
>>> tel: +64 (0)9 923 4611
>> 
> -- 
> Dr Adrian Croucher
> Senior Research Fellow
> Department of Engineering Science
> Waipapa Taumata Rau / University of Auckland, New Zealand
> email: a.croucher at auckland.ac.nz <mailto:a.croucher at auckland.ac.nz>
> tel: +64 (0)9 923 4611

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20240528/2fee2847/attachment-0001.html>


More information about the petsc-users mailing list