How do I deploy the application which use "CUDA driver API"?

[hsucks86]

What’s the right way to deploy the application which use CUDA driver API?
I know the simplest way is installing CUDA into target system, but I don’t prefer this way…

What I did is including below files and setting up LD_LIBRARY_PATH, but it’s failed to run.

libcuda.so
libcuda.so.1
libcuda.so.375.39
libEGL_nvidia.so.0
libEGL_nvidia.so.375.39
libEGL.so.1
libGLdispatch.so.0
libGLESv1_CM_nvidia.so.1
libGLESv1_CM_nvidia.so.375.39
libGLESv1_CM.so.1
libGLESv2_nvidia.so.2
libGLESv2_nvidia.so.375.39
libGLESv2.so.2
libGL.so.1
libGL.so.1.0.0
libGLX_indirect.so.0
libGLX_nvidia.so.0
libGLX_nvidia.so.375.39
libGLX.so.0
libnvcuvid.so.1
libnvcuvid.so.375.39
libnvidia-compiler.so.375.39
libnvidia-eglcore.so.375.39
libnvidia-egl-wayland.so.375.39
libnvidia-encode.so.1
libnvidia-encode.so.375.39
libnvidia-fatbinaryloader.so.375.39
libnvidia-fbc.so.1
libnvidia-fbc.so.375.39
libnvidia-glcore.so.375.39
libnvidia-glsi.so.375.39
libnvidia-ifr.so.1
libnvidia-ifr.so.375.39
libnvidia-ml.so.1
libnvidia-ml.so.375.39
libnvidia-opencl.so.1
libnvidia-opencl.so.375.39
libnvidia-ptxjitcompiler.so.375.39
libnvidia-tls.so.375.39
libOpenGL.so.0
libvdpau_nvidia.so.1
libvdpau_nvidia.so.375.39

[Robert_Crovella]

Assuming the target machine has a NVIDIA GPU, with a properly installed recent driver (e.g. the latest driver for that GPU) the only thing that should be required is the application itself. Driver API applications only depend on libcuda.so, and that is installed by the driver.

And, yes, the LD_LIBRARY_PATH on the target machine has to include the path to libcuda.so

The 375.39 driver in your listing would support applications that were built against any CUDA version up to 8.0, but would not support a CUDA driver API application built against CUDA 9.0 or 9.1 You would need a newer driver on the target machine for that.

Again, you have to have a properly installed GPU driver. You don’t just copy these files to the target machine.