nvprof complains 'Error: unable to locate profiling library libcuinj64.so"

[Dan Kegel]

Workarounds:

sudo ln -s ../nvidia-375/libnvidia-fatbinaryloader.so.375.66 /usr/lib/x86_64-linux-gnu/

or

export LD_LIBRARY_PATH=/usr/lib/nvidia-375:$LD_LIBRARY_PATH

That lets nvprof start (assuming you're using nvidia-375, I suppose).

Evidently this is all a bit fragile. See also

https://github.com/tensorflow/tensorflow/issues/9071
https://www.bountysource.com/issues/43938575-importerror-libnvidia-fatbinaryloader-so-375-51-cannot-open-shared-object-file

ProblemType: Bug
DistroRelease: Ubuntu 16.04
Package: nvidia-profiler 7.5.18-0ubuntu1
ProcVersionSignature: Ubuntu 4.4.0-81.104-generic 4.4.67
Uname: Linux 4.4.0-81-generic x86_64
ApportVersion: 2.20.1-0ubuntu2.6
Architecture: amd64
CurrentDesktop: Unity
Date: Wed Jun 21 12:21:22 2017
JournalErrors:
 Error: command ['journalctl', '-b', '--priority=warning', '--lines=1000'] failed with exit code 1: Hint: You are currently not seeing messages from other users and the system.
       Users in the 'systemd-journal' group can see all messages. Pass -q to
       turn off this notice.
 No journal files were opened due to insufficient permissions.
SourcePackage: nvidia-cuda-toolkit
UpgradeStatus: No upgrade log present (probably fresh install)

[Dan Kegel (dank)]

This may be an artifact of running on a machine with no nvidia card.

[Launchpad Janitor (janitor)]

[Expired for nvidia-cuda-toolkit (Ubuntu) because there has been no activity for 60 days.]