Command-line control over scalar-valued set routines #662
Conversation
drreynolds
requested review from
balos1,
gardner48 and
cswoodward
as code owners
Co-authored-by: Steven Roberts <roberts115@llnl.gov>
… since that causes compilation warnings/errors because main() uses char* argv[].
…uld be matched to process command-line arguments, which should support MRI and operator-splitting
|
I think I've finished revising the previous implementation according to the initial feedback. It would be great if folks who looked at this previously would give it another look. Once you're happy with the structure, I'll get started on the ARKODE stepper-specific functions, other integrators, linear solvers, nonlinear solvers, etc. |
…ber of options in each category. Co-authored-by: Steven Roberts <roberts115@llnl.gov>
…n prefix and keys for SetOption routines
…ms, and added control over ARKodeUseCompensatedSums
|
I believe I've addressed all PR comments except for moving command-line processing to base classes from implementations. As I noted in my responses, I have concerns with making that move, since I worry that it decreases robustness for the sake of saving a few lines of code from implementations (but adding around the same number of lines elsewhere). If others feel differently then I recommend we debate that later in a subsequent PR. I see no reason such a minor implementation detail should hold up this release, since even if we decide to eventually re-engineer the internals to make the move, it will not change the functionality or API. I'm still updating answer files due to changes in what is printed in some examples, but otherwise this PR should be good for re-review. |
Add functions to set options from command line args --------- Co-authored-by: Steven Roberts <roberts115@llnl.gov> Co-authored-by: David J. Gardner <gardner48@llnl.gov>
I believe this PR is now complete, although we should hold off on reviews until the current release is out.