[RESOLVED] make -j fails on crosscompilation drivework sample

Autonomous Vehicles DRIVE PX2 DriveWorks
#1

Hello All,

After having installed without issues Driveworks throught the DriveInstall, next step was to rebuild the set of sample to check that the host is correcly setup ( ubuntu 14.04).

so the in /usr/local/driveworks-0.2.1/sample/build I ran the cmake command wihtout issues.

but when I do the ‘make -j’ right after the cmake, it fails linking the libcuda.so.1 lib:

[ 33%] Linking CXX executable sample_camera_tracker
[ 35%] Linking CXX executable sample_rig_configuration
[ 36%] Linking CXX executable sample_occupancy_grid
[ 37%] Linking CXX executable sample_renderer
/usr/bin/ld: warning: libcuda.so.1, needed by /usr/local/driveworks/targets/x86_64-linux/lib/libdriveworks.so, not found (try using -rpath or -rpath-link)
/usr/bin/ld: warning: libnvcuvid.so.1, needed by /usr/local/driveworks/targets/x86_64-linux/lib/libdriveworks.so, not found (try using -rpath or -rpath-link)

I’m suprised as I’ve added the path to libcuda.so.1 in LD_LIBRARY_PATH .

(in the .bashrc i’ve added: export LD_LIBRARY_PATH=/home/esi/Documents/nvidia/VibrantePDK/vibrante-t186ref-linux/lib-target:$LD_LIBRARY_PATH )

I’ve installed the out of box ubuntu 14.04 and used the latest cmake 3.8.

So i do not understand why it cannot find it, any ideas ?

thanks!

#2

Hi again,

I succedded in compiling correcly following the tip from the following topic on this forum

Driveworks-0.2.1 build sample but cannot run on PX2 : https://devtalk.nvidia.com/default/topic/997207/driveworks/driveworks-0-2-1-build-sample-but-cannot-run-on-px2/

I add to do :

sudo chown -R $(whoami):$(whoami) /usr/local/driveworks-0.2.1/samples

and that was the trick :-)

Maybe this would be nice to add in the starting guide pdf ?

Thanks anyway!

Good day,

Fx

#3

Hello fxjegaden,

Could you please update DriveWorks 0.3?
And please see below link, the file will give how to use DriveWorks and DPX2 with some examples.

#4

Thanks, that’s the file I was looking into.

By the way you mention Driveworks0.3 but when I login to https://partners.nvidia.com there is only driveworks0.2.1 available

can you please update it or tell me where to find this new release ? thanks

#5

Hello fxjegaden,

Could you download DriveInstall 4.1.6.1 Linux SDK via https://developer.nvidia.com/nvidia-drive-downloads?
You can download DriveWorks 0.3. Thanks.

#6

Hello,

I also have the same problem as fxjegaden. As your advice, I had update DriveWorks 0.3 on host PC but the cross-compile doesn’t succeed. The problem is the same as following:

/usr/bin/ld: warning: libcuda.so.1, needed by /usr/local/driveworks/targets/x86_64-linux/lib/libdriveworks.so, not found (try using -rpath or -rpath-link)
/usr/bin/ld: warning: libnvcuvid.so.1, needed by /usr/local/driveworks/targets/x86_64-linux/lib/libdriveworks.so, not found (try using -rpath or -rpath-link)

And I didn’t find the libnvcuvid.so.1 in any directory. So could you give me any advice?

Regards

#7

Hi,

@yaom : Were you able to solve this?

I am having the exact same issue. I have DriveWorks 0.3 as well.
I am getting the following warnings:

libcuda.so.1, needed by /usr/local/driveworks-0.3/targets/x86_64-linux/lib/libdriveworks.so, not found (try using -rpath or -rpath-link)
libnvcuvid.so.1, needed by /usr/local/driveworks-0.3/targets/x86_64-linux/lib/libdriveworks.so, not found (try using -rpath or -rpath-link)

I am unable to find libnvcuvid.so.1 anywhere on my system either.

P.S.: I am using nsight.

Thanks

#8

Dear yaom, ana9017,

Did you finish CUDA environment setup like below? Could you check it? Thanks.

======================
$gedit ~/.bashrc
export PATH=/usr/local/cuda-8.0/bin/:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/targets/aarch64-linux/lib:$LD_LIBRARY_PATH

source ~/.bashrc nvcc --version

$ /usr/local/cuda-8.0/bin/cuda-install-samples-8.0.sh ~/

$cd ~/NVIDIA_CUDA8.0_Samples/1_Utilities/deviceQuery
make ./deviceQuery

#9

Hi SteveNV,

I am getting following result after running ./deviceQuery

./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version Result = FAIL

All other setup is complete. .bashrc has required paths.

#10

Dear ana9017,

It’s weird, did you check CUDA version with “nvcc --version”? Thanks.

#11

Hi SteveNV,

Thanks for your help!

I was finally able to solve the issues.

So the path for libnvcuvid.so.1 file is

/usr/lib/<nvidia-driver-name>/libnvcuvid.so.1

For some reason, even after adding all the paths to bashrc, it was not able to find this file.

So, I created a symbolic link in /usr/local/driveworks/targets/x86_64-linux/lib/ for libnvcuvid.so.1

sudo ln -s /usr/lib/nvidia-352/libnvcuvid.so.1 /usr/local/driveworks/targets/x86_64-linux/lib/

After doing this, I just followed the steps provided by nvidia for building samples in the document: /usr/local/driveworks/doc/nvdwx_html/dwx_samples_building.html

Everything working perfectly well now!

Note: @SteveNV: my CUDA version is 8.0