Small cleanup of cmake build for NVIDIA and AMD OpenCL
For AMD Stream SDK use:
INCLUDE_DIRECTORIES( ${AMD_OPENCL_INCLUDES} )
LINK_LIBRARIES( ${CMAKE_ATISTREAMSDK_LIBPATH}/OpenCL.lib )
For NVIDIA CUDA SDK:
INCLUDE_DIRECTORIES( ${NVIDIA_OPENCL_INCLUDES} )
LINK_LIBRARIES( ${NVIDIA_OPENCL_LIBRARIES})
Create a separate library for MiniCL (a rudimentary OpenCL wrapper to run kernels on multi-core CPU using the default C++ compiler, using Win32 Threads or Posix)
tweak vectormath/vmInclude.h for PS3 platforms
fix warning in btKinematicCharacterController
The kernel in the .cl file is also compiled by the native C++ cpu compiler, when using MiniCL.
When you want to debug the kernel using MiniCL, and want to put breakpoints, it is best to:
1) enabled the define #define DEBUG_MINICL_KERNELS 1 in Bullet/src/BulletMultiThreaded/MiniCL.cpp
2) temporarily remove the stringify lines in the .cl kernel, because it prevents the debugger from finding the right line.
