OpenCV 3.3.1 (self compiled vs. preinstalled) hardware acceleration and so on

Hello,

the aim is just to show a raw stream from a camera on the jetson xavier and i tested different opencv versions one preinstalled version via the jetpack 4.2.2 and one self compiled version, and i figured out that the preinstalled version from nvidia is slower and has more latency but takes less hardware resources on the cpu in comparison to the self compiled version from opencv. My question is just does the preinstalled version use hardware acceleration or something like that and is it possible to rebuild the preinstalled version from nvidia on the jetson xavier ? I also have to say that really less build options are used for the preinstalled opencv version.

Best regards. javascript:void(0);

Ok i found the problem there is a “problem” with tbb-threads and openmp, openmp is quite faster but takes a lot more ressources whereas compiling against tbb is a little bit slower but also takes less resources so using tbb is quite fair in terms of resources and speed.