I want to cross compile the nvmedia sample application “vid_play” from my host machine to the Drive Xavier. I copied the vid_play folder from the nvmedia samples to my cuda workspace folder. Then I import the project as New CUDA C/C++ Project (CUDA Runtime Project, CUDA Toolkit 10.0), choose /home/user/cuda-workspace/vid_play as source folder and choose PTX and GPU code 7.2. In the next step i choose the remote connection using IP address and username and use an empty folder on the Drive as Project Path, the Toolkit is detected and I choose AArch64 as CPU Architecture. Then I click on finish.
Now if I click on the arrow next to the build hammer and choose “Debug” on the target machine I get the error: “fatal error: nvmedia_core.h: No such file or directory”. If I copy the path to the project folder or specify the absolute path for all the .h files I get further errors. I tried different ways to add library and include paths in the project properties from different articles on your homepage but none of them worked. Could you tell me the correct way to add the header file paths to the project?
Dear david.frenud,
By default, Nvmedia samples are configured for cross complilation. Could you make changes in sample and build using make. Did you check that?
I am able to run the compiled executable on the Drive platform now.
But I do have another question: Is there any sample application that uses nvmedia for recording a video using the camera attached to the Drive and storing it into a video file?