GitLab

Added doxygenable comment that Langevin friction coefficient is in units of collisions per picosecond (as opposed to picoseconds), which Sherm and I deduced by reading reference implementation and testing extreme values.

(2) I disabled the block of code in the main CMakeLists file that tries to second guess VC++'s choices for runtime libraries. This was required for me to get CUDA to work in VC 2005 where the compiler always chooses multithreaded libraries. I verified that Brook will still build this way, but I haven't tested it.

Modified the API for lots of classes so that the number of particles/bonds/etc. need not be specified in the constructor.

Implemented the Ewald summation method in the reference platform. Preliminary cuda prototype is included.

Implemented setting random seed for reference platform, as well as adding it as an option to BrownianIntegrator and AndersenThermostat.

Checked in explicit solvent implementation for Cuda platform (cutoffs, periodic boundary conditions, SETTLE)

Optimization: introduced a kernel to clear forces at the start of each time step rather than relying on the stream

Implemented dynamic loading of platforms from DLLs.  Also fixed a lot of compilation errors and warnings under Windows.

Changed how kernels get initialized.  StandardMMForceField now has you set the 1-4 interactions explicitly.  CMMotionRemover allows the removal frequency to be set.

1) Added DLL export macros to many classes; enough to get tests to compile
2) Cosmetic changes to reduce warnings in Visual Studio:
   a) change many "for (int i ="... to "for (size_t i ="... to avoid unsigned mismatch warnings
   b) included many explicit static_cast<OpenMMReal>() casts to document places where double->float conversion warnings occur

Added INCLUDE(Dart) to OpenMM CMakeLists.txt