GitLab

Improve the performance of sending the AmoebaVdwLambda to Cuda using pinned host memory; Updated the AmoebaVdwForceProxy to version 3, and added backward compatibility to version 2; updated TestAPIUnits.py to handle the per particle lambda flag

changed the type of the isAlchemical Cuda array to float, which is needed for the CudaNonbondedUtilities.addParameter method to work; the TestCudaAmoebaVdwForce now passes for its current set of tests, but new tests are needed to check the softcore behavior

Rough draft AmoebaVdwForce Cuda code; next steps are to add CUDA kernal compilation support for USE_SOFTCORE, VDW_LAMBDA, VDW_SOFTCORE_POWER, VDW_SOFTCORE_ALPHA, and per particle isAlchemicalI / isAlchemicalJ flags

Fixed units in swigInputConfig.py for new Alchemical AMOEBA vdW methods exposed in the OpenMM API; implemented the getDefaultParameters() on the AmoebaVdwForceImpl to return 1.0 for AmoebaVdwAlpha; Use AmoebaVdwForce::Lambda() to refer to this parameter rather than the String AmoebaVdwAlpha

AMOEBA vdW softcore is propogated through the Reference platform; pending some feedback adding it to the CUDA code can begin

for the AmoebaVdwLambda global parameter was added that returns the parameter's name; added the value None to the AlchemicalMethod enumeration; provided default values for the isAlchemical flag

Improved robustness of custom GB forces

Fixed errors on Navi GPUs

Fixes for #2353

Updated some links for changed organization

Fixed Bondi radius for Cl

Fix 2360

Added CONTRIBUTING and SUPPORT files

Support for multiple paths in OPENMM_PLUGIN_DIR

Fixed the documentation. I should probably go add commentary in a README or something denoting that OPENMM_PLUGIN_DIR can take :'s.

when the number of molecules does not agree
with the number of holders.

Improved API for setting periodic box vectors in CharmmPsfFile