Activating CUDA

[brsbilgin]

| Camera Model | D435 |
| Firmware Version | 05.12.03.00 |
| Operating System & Version | Linux (Ubuntu 18.04) |
| Kernel Version (Linux Only) | 4.9.140-tegra |
| Platform | NVIDIA Jetson |
| SDK Version | 2.32.1 |

For activating CUDA cores should I build from sources with correct flags or it is enough install from debian packages? How can I evaluate it is using CUDA cores or not?

[MartyG-RealSense]

CUDA support can be enabled with a CMake flag called BUILD_WITH_CUDA, which is set to False by default.

CUDA support can be enabled from within a CMake build instruction for the Librealsense SDK using a build instruction similar to the one below:

cmake ../ -DCMAKE_BUILD_TYPE=Release -DBUILD_EXAMPLES=true -DBUILD_GRAPHICAL_EXAMPLES=true -DBUILD_WITH_CUDA=true

An easy way to check whether the effect of CUDA support is being applied is to check CPU usage. There should be a clear reduction in CPU % usage with CUDA support enabled, as highlighted by the example chart below:

[brsbilgin]

Hi MartyG,
Thank you for quick reply. Actually my question is I already install with "sudo apt-get install librealsense2-utils librealsense2-dev -y". For CUDA support should I remove this packages and build according above your instructions.
Yesterday I build with -DBUILD_WITH_CUDA=true and today with new jetson image I just install with "sudo apt-get install librealsense2-utils librealsense2-dev -y" I couldn't see any dramatic change between yesterday and today.

[dorodnic]

Hi @brsbilgin
Debian packages are already built with CUDA.

[brsbilgin]

Thank you Dorodnic