Dear GROMACS developers,
Our next biweekly meeting will be on 2023-10-18T15:00:00Z. At the moment we have two open slots, so if there’s anything you wish to discuss please comment below (or let me or Berk know)!
Berk: general issues (bugs, CI, maintenance)
Magnus: included force fields in GROMACS
Note that we have a new Zoom room since June:
Join our Zoom Meeting (magic link)
Meeting ID: 692 7708 9234
Join by SIP:
See you next week!
These are my notes from the meeting. Feel free to comment if I missed something important, or if something should be removed:
Minutes, biweekly developers meeting, 231018
- Reports from Stockholm:
Magnus: Included force fields. We want to update the force fields we distribute. We might need to find a balance if we want to include all updates.
Dmitry: OPLS v3 is popular.
- Since it is proprietary we cannot ship it.
CHARMM36, Amber14 and Amber99 seem to be the most used. Amber14 is available in PMX. Amber19 is still not supported due to intraresidue CMAP corrections.
- CHARMM36 will be a natural first step. We will discuss it with Justin.
Berk: 2023.3 release tomorrow (231019). 2024-beta in three weeks (milestone 231110, release within a week after that).
Increased focus on user-driven development - still difficult to decide what to focus on and not.
Dmitry said that there was a request for (selective) non-periodic non-bonded interactions, e.g., for some machine learning force fields.
Vedran: How is NB-LIB meant to be used in GROMACS?
Berk: We aim at using it from the md loop.
Szilárd: Large risk of losing performance.
Berk: We want to mainly use their stronger-typed (bonded?) force calculation functions.
- We will have to see what makes sense to use and how.
Vedran: System installed Boost (1.8.3) cannot be used with GROMACS (with OpenCL), issue System-wide installed version of Boost gets picked up during build instead of GROMACS bundled one and results in build failure with GMX_GPU=OpenCL (#4893) · Issues · GROMACS / GROMACS · GitLab. We mainly use iterator functionality. It is possible that C++20 will support that.
Alexey wants a discussion slot for next meeting.
Hubert: Has re-generated Colvars test reference data in double precision. How is the tolerance in tests set?
Berk: Needs to be set based on error propagation.