[petsc-users] Using DMCOMPOSITE with TS

[Jed Brown]

It shouldn't have any affect.  This will need to be debugged.  There's
no chance I'll have time for at least a week; hopefully one of the other
TS contributors can look sooner.

"Ellen M. Price via petsc-users" <petsc-users at mcs.anl.gov> writes:

> Quick update: I found that changing TS_EXACTFINALTIME_INTERPOLATE to
> TS_EXACTFINALTIME_MATCHSTEP fixes the problem; is there a reason I can't
> use the interpolation when using a DMCOMPOSITE?
>
> Ellen Price
>
>
> On 2/20/19 1:42 PM, Ellen Price wrote:
>> Hi fellow PETSc users!
>> 
>> I am attempting to use a DMCOMPOSITE alongside TS and have run into some
>> trouble. I'm attaching a MWE demonstrating the problem. The goal is to
>> combine a DMDA3d for spatial data and a DMREDUNDANT for non-spatial,
>> time-dependent fields. In the attached example, this additional field is
>> temperature.
>> 
>> The DMDA data is currently just temperature-dependent, and the
>> temperature is supposed to increase linearly. Unfortunately, no matter
>> how I integrate (explicitly, using Euler or RK, or implicitly, using
>> backward Euler or BDF), I get the wrong final temperature. The
>> integration proceeds for 100 timesteps and then stops (verified by
>> -ts_monitor). At a heating rate of 1, and an initial temperature of 150,
>> I should get a final temperature of 250 very easily. However, I get
>> something closer to 200 (not exact, which makes me think this isn't a
>> simple missing-a-factor-of-2 error).
>> 
>> I'm currently using TSSetDM with the composite DM to let PETSc compute
>> the Jacobian. Having checked all the inputs and outputs as well as I
>> can, this is the only step that seems like it may be causing a problem;
>> yet even explicit integration doesn't work, and that shouldn't need the
>> Jacobian at all. I'm at a loss.
>> 
>> Run using:
>> ./mwe -implicit yes -ts_type beuler -ts_monitor OR ./mwe -implicit no
>> -ts_type euler -ts_monitor
>> 
>> Thanks in advance for any help,
>> Ellen Price