Open dotsdl opened 3 months ago
Minimization performed on the work server proved to be a massive performance bottleneck. This is being addressed in the openmm-core
. We will need to add the new settings exposed by the core to this PR, and make use of them.
Also added #13 as an issue to solve later; would like to give users of this protocol defaults that don't generally result in really short FAH WUs.
@hmacdope raised a good point: we likely need to specify minimum core numbers/versions that we support for a given alchemiscale-fah
release. If possible, it would be even better to enforce this at runtime for the FahAsynchronousComputeService
, perhaps by checking each configured PROJECT's project.xml
.
Will use this PR to track observations and any changes needed to accommodate issues.