pub fn f32_to_f16(val: f32) -> u16
Converts a single-precision float (FP32) into an IEEE 754 half-precision (FP16) integer.