Has anyone tried implementing a half precision floating point number (16bits) on an arduino?
If I tried doing this directly in assembly, how would I generate the assembly code to edit it?
What would be the (perceived) benefit?
To expand a little on Paul's question, what advantage of 16 bit floating point number do you perceive compared to fixed comma arithmetic? Do you really have number of different magnitudes you need to work on in the same operation?
Korman
Personally I think 32 bit floating point is a bit on the thin side..... 