While trying to refactor away some annoying atom count parameter in the PME code, I ran into TPI logic again, and realized it's not covered by tests.
While trying to throw together a test, I ran into "TPI does not work (yet) with the Verlet cut-off scheme" message.
Without at least a basic TPI sanity test, I doubt this feature will survive much longer...
- Related to Task #1971: Removing buggy features vs. keeping workflows added
- Status changed from New to Accepted
Do you already have a test set up? We can add it with the group scheme and later convert it to Verlet.
I should fix up the nbnxn pair search for TPI, that's actually easier than the group scheme code.
No, I don't, I've only tried adding some more mdp parameters to the PmeTest.
Maybe you could provide the mdrun input/arguments?
Here's one test with files provided by Berk.
There's more to test here - tpic, PME..
Also available in: Atom