pub fn f16_to_f32(h: u16) -> f32
Converts an IEEE 754 half-precision (FP16) 16-bit integer to a single-precision (FP32) float.