The int16 is to keep the bytes in the data type and not slip into 32bit

fixed-point.js

const scale = 8;

const d2f = (x) => Uint16Array.of(x * (1 << scale));
const f2d = (x) => Int16Array.of(x)[0] / (1 << scale);
const mul = (x, y) => Uint16Array.of(((x[0] >> 4) * (y[0] >> 2)) >> 2);

const a = d2f(-5); // ?
f2d(a) // ?
const b = d2f(0.8) // ? 
f2d(b) // ?
const c = mul(a, b) // ?

// should be -4
f2d(c) // ?