Manual
Programming Reference
r16_float
Ordinary format with one component stored as 16-bit (half-precision) floating-point number.