On Tue, Oct 19, 2021 at 6:12 AM Matthew Knepley <[email protected]> wrote:
On Mon, Oct 18, 2021 at 10:35 PM Adrian Croucher < [email protected]> wrote:
Any response on this?
This is a bit of a showstopper for me - I can't upgrade to PETSc 3.16 if it does not allow my users to read their HDF5 files created using earlier versions of PETSc.
So far I can't see a workaround. Possibly the timestepping functions need some kind of optional parameter to specify what the default timestepping attribute should be, if it's not present in the file (rather than just assuming it's false)?
I will fix it. I think I can do it tomorrow. Class just started this week do it is hectic :)
I think you are right. We should always write the attribute, but have it be false. We should interpret a missing attribute as an old file.
Okay, I think I have it. Can you look at this branch? https://gitlab.com/petsc/petsc/-/merge_requests/4483 There is now an option that lets you set the default timestepping behavior -viewer_hdf5_default_timestepping I think that is what you want. Thanks, Matt
Thanks,
Matt
Regards, Adrian
On 10/14/21 4:19 PM, Adrian Croucher wrote:
hi
I am just testing out PETSc 3.16 and making the necessary changes to my code. Amongst other things I now have to add a PetscViewerHDF5PushTimestepping() call before starting to output time-dependent results to HDF5 using a PetscViewer.
I now also have to add this call before reading in sets of previously computed time-dependent results (for restarting a simulation from the results of a previous run).
The problem with this is that if I try to read in the results of any previous run, computed with an earlier version of PETSc (< 3.16), an error is raised because the time-dependent datasets in the file do not have the 'timestepping' attribute.
Is there something else I need to do to make this work?
- Adrian
-- Dr Adrian Croucher Senior Research Fellow Department of Engineering Science University of Auckland, New Zealand email: [email protected] tel: +64 (0)9 923 4611
-- 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/>