No CUDA Library in usr/lib after auto NVIDIA driver installation

Hi, everyone.
While trying to find a root of the “CUDA driver version is insufficient for CUDA runtime version”
I found that CUDA library is absent in the usr/lib (no /usr/lib/libcuda.so.384.59).
And other few required by NVIDIA Accelerated Linux Graphics Driver README and Installation Guide also.
I used package manager installation method and everything accomplished well.
My configuration:

  • ThinkPad-T460:~/NVIDIA_CUDA-9.0_Samples$ whereis nvidia
    nvidia: /usr/lib/nvidia /usr/share/nvidia /usr/src/nvidia-384-384.59/nvidia
  • 06:00.0 3D controller: NVIDIA Corporation GM108M [GeForce 940MX] (rev a2)
  • Ubuntu-17.04
  • Kernel 4.10.0-35-generic
  • gcc version is gcc (Ubuntu 6.3.0-12ubuntu2) 6.3.0 20170406
  • nvcc: NVIDIA ® Cuda compiler driver
    Copyright © 2005-2017 NVIDIA Corporation
    Built on Mon_Jun_26_16:13:28_CDT_2017
    Cuda compilation tools, release 9.0, V9.0.102

My question is why installing NVIDIA Drivers with distribution-specific installation method some necessary results of this process (i.e. few of libs required by NVIDIA driver) are missing in required place (usr/lib)? And might the fact be a root of the inconsistence of CUDA driver version and CUDA runtime version ?

Thanks in advance.

Because that isn’t necessarily the required place. Use the linux find command to locate every instance of the libcuda.so*

One of them will be the 384.59 one and it will be in the required place.