GitLab

based on the FindCUDA.cmake that was in the platforms/cuda/cuda-make directory.
Setting of CUDA_NVCC_FLAGS is now in top-level CMakeLists.txt, in case I ever figure out how
to get this working with the cmake 2.8 installed system version of FindCUDA.cmake.
Before this change, there were three separate FindCUDA.cmake areas.  Now there is just one.

Downgraded recent cmake changes to ghastly compliance with cmake 2.4, for the benefit of the simtk dart server.

Modified Cuda FreeEnergy plugin to use a different windows export macro than main openmm library and cuda library.
Modified CUDA_NVCC_FLAGS to be consistent with Peter's latest version in all three locations they are defined.
Added "-Xcompiler '/MD'" to CUDA_NVCC_FLAGS on Windows to avoid link error in Cuda FreeEnergy plugin.

Export CudaFreeEnergyKernelFactory symbols to avoid link error in test program.

Turn default to ON for nightly building Cuda FreeEnergy plugin, since it is desired for pyopenmm release.

Still doesn't build on windows, but might help with linux (todo next)

Cast exponent in pow() function to RealOpenMM, to avoid overload ambiguity error on Windows when RealOpenMM is defined as double.

Associated code will be removed once confirmed deletions do not break builds

Fixed uninitialized value problem with setting box vectors.  This caused sporadic multithreading issues on Linux platforms.

Use a consistent (and more accurate) value for electric constant everywhere.  Also cleaned up handling of constants and units in various places.

Removed static variables in CudaFreeEnergyKernels.cpp
Moved declaration of global feSim to within method where actually used  in kCalculateNonbondedSoftcore.cu

windows export macro on ReferenceFreeEnergyKernelFactory, to avoid link error on TestReferenceGBVISoftcoreForce

Replace Platform::getPlatform by Platform::getPlatformByName where necessary, due to recently changed interface.

Removed cruft from Cmake file
Modified the way platform is specified in TstFreeEnergyCudaUsingParameterFile.cpp

(It still *fails*, but at least it is found)

Reorganized platform analysis in NML init method.
Failed attempt to get OPENMM_PLUGIN_DIR working for NML test.

Incorporate Normal Mode Langevin openmm code from Chris Sweet as a plugin.  (compiles on windows, but no actual testing done)

removed some cruft

Fixed bug: was calling kReduceObcGbsaSoftcoreBornForces instead of kReduceGBVISoftcoreBornForces in GB/VI force calculation