Nvidia Jetson

The Ensenso SDK can be installed on Nvidia Jetson systems.

When running on a Jetson system CUDA is always enabled and used automatically.

Compilation of CUDA Kernels

With the first execution of CUDA-accelerated NxLib commands, the CUDA runtime automatically compiles the CUDA kernels for your specific GPU and Nvidia driver. This can take up to a minute due to the limited performance of the CPU on Jetson systems.

The compilation is only done once on the same system. The result is cached in ~/.nv/ComputeCache and stays valid until you update the Ensenso SDK or the Nvidia driver.

Limitations

Ensenso Engage is not available on Nvidia Jetson yet.

The following commands currently don’t have an implementation using CUDA and are not supported on Nvidia Jetson:

• EstimateDisparitySettings

• FitPrimitive

• PartFinder