Floating point format

This came up in a thread that is a bit esoteric so I thought it might be worth repeating in a seperate thread...

Quote:

Original post by Dredge-Master Assuming Little Edian (Edit: I've put Big Edian next to it in comments, just swap the values - on an 0x8086 based processor you should be using little edian so just leave the code as is). typedef union { float f; struct float_bits { unsigned int mantissa : 23; // negative:1 unsigned int exponent : 8; // exponent:8 unsigned int negative : 1; // mantissa:23 } ieee_bits; } ieee754_float;

It provides a nice easy way to inspect how a floating point number is stored and play with them. If you're going to use floating point heavy then you should understand them along the same lines as you understand integer.