H211b adaptive time-stepping

[RolfSander]

We already have the solver rosenbrock_posdef_h211b_qssa in the
user_contributed directory. Unfortunately, it has not been maintained
recently, and it hasn't worked with KPP3. Together with colleagues from
Jülich, I have now upgraded it so that it works out-of-the-box. We are
also writing a manuscript about the solver which should be ready soon.
The new code is in the branch
https://github.com/KineticPreProcessor/KPP/tree/feature/h211b. It
contains the new solver file int/rosenbrock_h211b_qssa.f90. You can
also find a simple test case in examples/mcm_h211b.

I hope that the code will be ready for a pull request soon. In the
meantime, please let me know if you have any comments or suggestions...

[yantosca]

Thanks @RolfSander. I pushed a fix to the feature/h211b branch. Also am working on adding C-I tests.

[RolfSander]

Oops. Thanks for the fix!

[yantosca]

I also pushed some fixes for the examples & C-I tests. We now have C-I tests for mcm and `mcm_h211b. Here is a log of the output: ci_log.txt

[RolfSander]

Thanks for adding the C-I tests!

I have a couple of comments/suggestions about the latest commit:

• Thanks for creating the xrun scripts in both sh and csh syntax! Is
  there a reason why the new scripts don't set the environment variable
  KPP_HOME anymore?

• The file constants_mcm.f90 has been downloaded from the MCM web page
  (https://mcm.york.ac.uk/MCM/export). For future compatibility, I'd
  like to change files from the MCM as little as possible. Specifically,
  I prefer to keep the name SUBROUTINE define_constants_mcm() instead
  of renaming it to SUBROUTINE define_constants().

• In driver_mcm.f90, you changed USE constants_mcm to USE constants_KPP_ROOT. I think this should be reverted because
  constants_mcm.f90 is not a KPP-generated file. After reverting, the
  files mcm_isoprene.eqn and mcm_h211b_isoprene.eqn will be
  identical again.

[yantosca]

Thanks @RolfSander. I can revert the changes. The problem that I was having is that the C-I tests were failing because the name of the folder is mcm_h211b but the KPP file is mcm.kpp. I can try to fix that so we don't change the MCM-related code that much.

[RolfSander]

If the only requirement is that the *.kpp file has the same basename
as the directory, maybe it is sufficient to rename mcm_h211b/mcm.kpp
to mcm_h211b/mcm_h211b.kpp and keep the other files unchanged?

ps: It seems you've reverted everything now, including all the other
(very helpful) edits you've made. I hope it will be possible to
re-introduce them...

[yantosca]

Thanks @RolfSander. I'll re-introduce what I did. Stay tuned.

[yantosca]

Hi @RolfSander. I pushed some fixes. I edited the C-I test scripts so that you can specify not only the name of the folder but the name of the mechanism. This allows us to use mcm.kpp for both the MCM and MCM_H211b examples.

I also restored the run scripts (now named run_example.csh and run_example.sh) and made sure that the C-I tests build. Should be OK now.

[RolfSander]

Thanks, this looks good now!