Half-precision floating point processing

Hi everybody,

I want to do some half-precision computations on the Jetson AGX Xavier’s CPU which is based on armv8.2 architecture.
I am wondering if there is someone who has experience in this area.

Actually, I know _Float16 is an arithmetic data type in half-precision. Can I use this data type in the C++ source code? Should I use a specific compiler for this purpose?

Thanks in advance