History log of /petsc/src/ts/tests/ex24.c (Results 26 – 28 of 28)
Revision Date Author Comments
# 4e278199 16-May-2021 Barry Smith <bsmith@mcs.anl.gov>

Remove all double blank lines from source

Commit-type: petsc-style
/2h


# b9588c04 28-Sep-2020 Satish Balay <balay@mcs.anl.gov>

Merge branch 'hongzh/2020-09-05/fix-computejacobian' into 'master'

Reworked: Move TS shift processing from TSComputeRHSJacobian

See merge request petsc/petsc!3237


# 5591c571 01-May-2019 Hong Zhang <hongzhang@anl.gov>

ex24: add a test for TSComputeIJacobian

- TSComputeIJacobian is called repeatedly with different shift values; this is used in TSAdjoint
- The test would fail for old versions of PETSc (before the 1

ex24: add a test for TSComputeIJacobian

- TSComputeIJacobian is called repeatedly with different shift values; this is used in TSAdjoint
- The test would fail for old versions of PETSc (before the 11-29-2018 commit https://bitbucket.org/petsc/petsc/commits/971015bcbf92bfb5359474763467c3e4753de490)
- TSComputeIJacobian() may call TSComputeRHSJacobian() if users provide only RHSJacobian. But there was a bug in the old versions. See the following example:
shift = 1;
TSComputeIJacobian();
shift = 2;
TSComputeIJacobian();
The second call to TSComputeIJacobian() does not update the RHSJacobian matrix, however, the matrix needs to be reshifted. This caused adjoint solvers giving incorrect results.

Funded-by:
Project: FASTMath adjoint
Time: 1.5 hours
Reported-by:
Thanks-to:

show more ...


12