June 24 2019¶
Completed since last meeting (June 10 2019)¶
- Merge 2019 to 2018 is not possible because failing test case with walls - Berk, Paul
- Update testing matrix versions for GROMACS 2020 release: How can we reduce the amount of testing to be done before the releases? Supporting the newest releases of,e.g., CUDA and gcc with all combinations takes large amounts of time. Solution: support new releases of compilers, etc., during patches over the year, not with the new release. Relax what we support upon a new release; cut-off 1st September for infrastructure to be supported.
- How to incorporate new patches? Edit larger patch sets with the first patch first, try to avoid rebasing, start out with technical review. Use topics.
- Tables support for 2020? Unclear, some minor issues.
- Technical guidelines for templated kernels - Erik to line out so that we can have a technical review style discussion.
- Half year plan meeting to discuss what should be attempted to merge into 2020 - June 25th 11:00-12:00. Set priorities before.
- Lots of spamming accounts are created in Redmine - can we move this offsite(hosted redmine, version upgrades, customization, price, number of users, cleaning of database)? - Christian to check options
- Paul submitted a series of patches for in-memory tpr reading and writing. Review needed. Can we ran out of memory? - put on hold for gromacs2020, because making sure that in all cases we would like to support that enough memory is available is hard
Progress since last meeting¶
- Post-submit failing in complex.nbnxn-ljpme-LB-geometric (issue #2939) - Fix submitted, Berk
- CUDA update/constrain unification patches waiting for review - Artem
- CUDA buffer ops patches - Szilard, Artem
- Constant pH code - working branch, code not very clean, nothing new for now - Berk, Paul
- Topology data structure cleanup - review needed - Paul, Mark
- Trajectory analysis output filters - Paul, Mark
- Paul started to split up EnergyAnalysis framework - David to write tests