testing backward euler with most of the analytical tests #576
Conversation
K20shores
requested review from
sjsprecious,
mattldawson,
boulderdaze and
mwaxmonsky
| micm::RosenbrockSolverParameters::ThreeStageRosenbrockParameters()); | ||
| test_analytical_surface_rxn(builder); | ||
| test_analytical_surface_rxn(rosenbrock, 1e-5); | ||
| test_analytical_surface_rxn(backward_euler, 1e-1); |
There was a problem hiding this comment.
1e-1 seems large to me. This makes me think whether we should write a function of calculating the relative error instead of using expect_near, which compares the absolute difference.
There was a problem hiding this comment.
yeah these values are on the small side. I'll make that change
There was a problem hiding this comment.
done, only was able to use 0.05 for the relative error though
it was super duper scientific. I set a value of A more intelligent way would be to play with the solver tolerances to choose values that minimize the error |
Haha I see. Yeah that makes sense but would require a good amount of work to investigate that. I guess we don't need to worry about it for now |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #576 +/- ##
=======================================
Coverage ? 92.89%
=======================================
Files ? 49
Lines ? 3477
Branches ? 0
=======================================
Hits ? 3230
Misses ? 247
Partials ? 0 ☔ View full report in Codecov by Sentry. |
K20shores
deleted the
499-refactor-tests-to-use-both-solvers-for-all-analytical-tests
branch
Closes #499
Backward euler passes all tests, but the robertson tests (the solver fails to converge and seems to get stuck, will be investigated by #498). I don't know how to test backward euler against HIRES, Oregonator, E5, but hopefully that can be fixed with #575