Detect Overflow for full range and resolution

I am looking for overflow and underflow during fixed point precision

I have following code

iwl and ifl is 32 bit wih

integer wordlength = 32 bits

integer floatlength = 24 bits

I would like to know under which circumstance overflow and underflow will happen

DWORD : 1signed+32WL+24FL 
coefs = int64(-100:1:100)
scale = 2.^FL
max = (2.^WL-1)-1
min = -2.^WL-1
range = coefs*scale

How to derive condition for Overflow , underflow, range , resolution , slope and bias ?