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.

move the plugin to its own project.
This change might also simplify pyopenmm nightly test issues.

day when we try it on Windows.
Added "-isysroot /Developer/SDKs/MacOSX10.5.sdk -mmacosx-version-min=10.5" to Mac compile flags,
to possibly help run on Leopard.

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

Set NVCC_FLAGS to particular value in new cmake_modules/FindCUDA.cmake

Turn on free energy plugin by default.
Better rule for when to default build OpenMM wrappers.
Several tweaks to advanced vs. non-advanced cmake variables.

So nightly builds can build CUDA and API wrappers when prerequisites are present.
While still being gentle on third party cmake builders.
New cmake_modules directory contains FindCUDA.cmake from cmake 2.8.0, and
Find OpenCL.cmake from platforms/OpenCL directory.

Free energy plugin added
Plugin will not run w/ Obc or GB/VI forces unless line 2004 of gpu.cpp (gpu->sim.totalNonbondOutputBuffers  = 2*gpu->sim.nonbondOutputBuffers;) is commented in -- working on removing this constraint
Also unit tests for GB/VI currently fail 

Better search for svnversion.exe on Windows

Deleted all the old hand-generated wrappers.

Created CMake script for generating API wrappers.  Also updated the wrapper scripts to handle vector<double>.

Examples: now builds CWrappers as a library; modified C builds to use it; added visual studio project to build Argon example in Fortran.
System.h: addForce() wasn't returning the index.

I'm checking in very early attempts at OpenMM Hello World examples so Chris and I can work on them together.

(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.