Cross-compile TensorRT / DeepStream applications for TX2

I’ve installed the gcc and cuda toolchain for aarch64 on the host using SDK Manager, but I don’t see the TensorRT / DeepStream aarch64 libraries for cross compiling on the host installed anywhere.

What’s the “official” process then to cross compile TensorRT applications for the TX2?

duplicate to topic 1061826:
https://devtalk.nvidia.com/default/topic/1061826/jetson-agx-xavier/tensorrt-5-1-cross-compile-for-jetson-agx-xavier/