Nvidia-docker: [l4t-base] Is there any way to install cicc on l4t-base:r32.2?

Created on 24 Jul 2019  路  9Comments  路  Source: NVIDIA/nvidia-docker

1. Issue or feature description

When I tried to build a CUDA application in a Docker image which is based on l4t-base:r32.2, the build was terminated with cicc: not found error.
I ran a l4t-base container and I found that /usr/local/cuda/nvvm did not exist in l4t-base docker image.

Is there a recommended way to install nvvm/cicc on l4t-base or do you have any plan to release a cicc-preinstalled version of l4t-base?

2. Steps to reproduce the issue

On Jetson Xavier (w/ JetPack 4.2.1),

git clone https://github.com/seiya0412/docker-libSGM_jetson.git
cd docker-libSGM_jetson
docker build -t libsgm_jetson .

Most helpful comment

Same with the latest r32.4.2. @RenaudWasTaken, please reopen.

All 9 comments

Looking into this, hoping to release the fix soon :)
Thanks for bringing this up!

A new image is now up with the fix (same tag)!

Successfully built! Thank you!

I'm having the same issue in the r32.3.1 image @RenaudWasTaken

I am as well

I'm having the same problem on R32.3.1. Adding /usr/local/cuda-10.0/nvvm/bin/ to my path didn't solve the issue

Directory /usr/local/cuda/nvvm missing again in r32.3.1 (r32.2 did work)

Same with the latest r32.4.2. @RenaudWasTaken, please reopen.

I couldn't build darknet in docker based on l4t-base:r32.2 with cuda.
/dev/null || pkg-config --libs opencv` -L/usr/local/cuda/lib64 -lcuda -lcudart -lcublas -lcurand -L/usr/local/cudnn/lib64 -lcudnn -lstdc++
/usr/bin/ld: cannot find -lcuda
/usr/bin/ld: cannot find -lcudart
/usr/bin/ld: cannot find -lcublas
/usr/bin/ld: cannot find -lcurand
/usr/bin/ld: cannot find -lcudnn
collect2: error: ld returned 1 exit status
Makefile:168: recipe for target 'darknet' failed
make: * [darknet] Error 1

Any help on this please ?

Was this page helpful?
0 / 5 - 0 ratings