Had you read the log we post. It is complaining about header file not found , not the library path.
#include <cuda.h>
^~~~~~~~
compilation terminated.
error: command ‘aarch64-linux-gnu-gcc’ failed with exit status 1
Pls check it yourself on a clean jetson nano official image.
Again, unless you have set some other additional environment variables or you are using un-release image.
It does not works.
We finally able to build the pycuda, not only you need to set LIBRARY_PATH, there is another environment variable need to set too.
Otherwise it cannot locate the header during pip3 pycuda install.
And we were able to verify the result is actually correctly in python.
And we found the bug in our C++ tensorrt usage.
Closing this topic.
Hi,
Good to know it works now.
In general, we don’t need to manually set up a parameter for pyCUDA.
But we really found out that the environment parameter may vary even from the same JetPack installer.
(Ex. different setup due to authority in network, host, … )
Would you mind to share the environment variable need to be updated with us??
So we can help others who meet the similar issues.
Thanks.
I use this script to install “pycuda” from source on Jetson Nano.
[url]https://github.com/jkjung-avt/tensorrt_demos/blob/master/ssd/install.sh[/url]