how to build a GPU program with Dockefile such as darknet which needs nvcc compile environment

Technica-Corporation / Tegra-Docker

Instructions and key files to enable Docker support on NVIDIA Tegra devices, specifically the TX-2.

Apache License 2.0

234 stars 61 forks source link

how to build a GPU program with Dockefile such as darknet which needs nvcc compile environment #3

Closed dqj closed 6 years ago

dqj commented 6 years ago

Running a GPU program within a Docker container is right. While i want to build a GPU program, it occurs an error. The error message is: make: /usr/local/cuda-9.0/bin/nvcc: Command not found

kgdad commented 6 years ago

If you are trying to compile from within your container, you'll either need to install the TX-2 CUDA packages in your Docker container or if you have CUDA already installed on your device, you can share those files with your container by adding /usr/local/cuda as a volume.

Shrunoti commented 6 years ago

Hi, I wanted to ask if this worked for @dqj . Because I tried mounting /usr/local/cuda and /usr/local/cuda-9.0 , but I get following error after running deviceQuery -

CUDA driver version is insufficient for CUDA runtime version
Result = FAIL

Also running darknet gives me following error- error while loading shared libraries: libcudart.so.9.0: cannot open shared object file: No such file or directory.

Note that cuda on the host machine runs fine and is compatible with the nvidia drivers.