GitLab

Reduced memory footprint

Add logging guards 

Set nonbondThreadsPerBlock to gpu->sim.nonbond_threads_per_block

Modified cudaLoadCudaFloatArray() call
Minor optimizations

Converted CUDAStream SysStream[0] and DevStream[0] references to SysData & Devdata

Added diagnostics for handling reorder of atoms

An exception is now thrown if the multipole axis type is not Bisector or Z-then-X
Redid velocity initialization in test code

Modified out-of-plane bend force to improve stability
Redid arguments to kCalculateAmoebaCudaKirkwood to reduce lmem