Gaussian process on GPU

Gaussian process regression project in C++, that leverage the GPU (cuda)\

General notes:

After installing cuda (in the following I am using cuda-9.0)

• Add the following in your ~/.bashrc file

export PATH=$PATH:/usr/local/cuda-9.0/bin
export LD_LIBRARY_PATH=LD_LIBRARY_PATH:/usr/local/cuda/lib64

Check if the following running in the terminal:

>> nvcc which
out: /usr/local/cuda-9.0/bin/nvcc
>> nvcc -V
out: information about  nvcc

• Change the GPU architecture inside the CMakeLists.txt file
  In the line 10:

set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode arch=compute_<num>,code=sm_<num>")

You can find the numbers that correspond with your GPU in this link

• Change include directories to match your installation:

if ($ENV{CLION_IDE})
    include_directories(/usr/local/cuda-9.0/include)
endif ()

notes when using Clion:

1. Add *.cu file extension to the C++ file type, the steps are:

File -> Setting -> Editor -> FileTypes -> C\C++ 
Click +
add *.cu

2. Build the project:

Build -> Build 'GP_GPU'

3. Edit configuration:

Run -> Edit Configuration
Change the executable to the file with the same as your project inside :
cmake-build-debug/CMakeFiles