To run this right out of the box:

export SHADER_MODEL XX
make 
./a.out

where xx is the compute capability of the GPU on that system (can be found by using devicequery)

The following need to be done in order to integrate into cpptraj

• makefile should use nvcc -arch=sm_xx instead of gcc/g++
• The following code/def needs to be added to action_closest.cpp/action_closest.h

   void Action_NoImage(Frame& frmIn,double maxD);
   void Action_ImageOrtho(Frame& frmIn, double maxD);
   void Action_ImageNonOrtho(Frame& frmIn, double maxD, Matrix_3x3 ucell, Matrix_3x3 recip);

   bool cuda_action_center(Frame& frmIn, double maxD, Matrix_3x3 ucell, Matrix_3x3 recip,int type, float &time_gpu);
   bool cuda_action_no_center(Frame& frmIn, double maxD, Matrix_3x3 ucell, Matrix_3x3 recip,int type, float &time_gpu);

• The folder Cuda_kernels needs to be copied