I wish to share this find with galfar and the rest here, it's related to processing 16 bit floating point values:
http://www.fox-toolkit.org/ftp/fasthalffloatconversion.pdf
There is already a pascal implementation of this that i have found:
https://github.com/SAmeis/xplib/blob/master/DeHL/Library/src/Math/DeHL.Math.Half.pas
I hope you find it useful, surely a lookup table based solution with no cpu code branching is a big improvement if you process 16 bit floating point values regularly.
Thanks for the tip, looks pretty nice.