Zero-Point Computation

Problem

Compute the asymmetric zero point: step = (max - min) / (2^bits - 1), z = round(qmin - min/step) clamped to [qmin, qmax], where qmin/qmax are -2^(bits-1) and 2^(bits-1)-1 for signed mode, or 0 and 2^bits-1 otherwise. A constant range returns 0. Return an integer.

The signature is zero_point_computation(min_val, max_val, bits=8, signed=True).

Hint

The zero point shifts the real zero into the integer code grid.

Test cases

case 1

in: [-1,2,8,true]

expected: -43

case 2

in: [-0.5,1.5,8,true]

expected: -64

case 3

in: [-1.5,0.5,8,true]

expected: 63

case 4

in: [0,0,8,true]

expected: 0

case 5

in: [-2,3,4,true]

expected: -2

Starter codePython

def zero_point_computation(min_val, max_val, bits=8, signed=True):
    # Your code here
    pass

5 test cases · run your code against them in the editor