I am new in the field.
Please explain what are interfaces between CPU and GPU?
Is L4T running on CPU and accessing GPU over various APIs of supported libraries ?
Yes, the system is run on CPU, and access GPU, other hardware engines for tasks. For developing your usecase, please try gstreamer or jetson_multimedia_api. For running deep learning inference please try DeepStream SDK. Documents are in
Jetson Linux | NVIDIA Developer