If all atoms are totally constrained, do NOT add the corresponding bond/angle force. Toolkit does this in a semi-hardcoded manner, Interchange is forced to duplicate this behavior
JW – On the surface, I’d think that we want to START ADDING the bond/angle physics parameters, even if they’re constrained.
(more discussion/explanation)
MT – I’ll have the new code path go toolkit → interchange with BOTH constraints AND bond physics, and just drop bond parameters by default in Interchange.to_openmm. I may have a kwarg to keep the bond physics parameters around.
MT – In the future, I’ll raise an issue/PR about whether we should keep both the constraint info AND bond physics parameters in Interchange.to_openmm. This issue/PR may be the initial implementation of the kwarg, or be about a change in default behavior.
MT - FYI – There’s an emerging discussion about evaluator on OpenCL
MT – Discussions I had with OM got pushed forward into Issue #424. Some question about whether Evaluator-OpenMM interface works with OpenCL.
JW – I’d recommend reading it over and commenting, but not approving. Do tag SB to give him right of first refusal to be approver
MT – When should I open Evaluator #413 for review? Who should be approver?
JW – (commented) Let’s keep it as WIP until the OFFTK 0.11.0 RC package is out.
Pair work
MT – What is the finish line w/ VU?
Serving as advisors is fun, but not a great use of our effort (logistic & cognitive)
JW – JBG’s defense (in a month that begins with M)
Bespokefit conda-forge release
JW – Update bespokefit recipe - Update from test env. DOWNPIN toolkit to avoid biopolymer release breaks, all deps in test env are required, none are actually optional!!