Hmmm... thats a pretty big range!

But if u normalize all values to a single fixed point e.g. multiply by 10E4 to get a range of 0-10E7, the maximum value (9999999) is a 3-byte value. So you should be able to manipulate everything in a 32-bit fixed point format (3 byte mantissa, 1 byte exponent in the range 0-7). Except for multi-byte division all other operations should be easy. Division can be done using the standard multibyte division libraries.

Alternately you could use a 2 byte variable to represent all numbers to the left of the decimal point and a 20-bit or 3-byte value to represent all numbers to the right of the decimal point, but that doesn't seem to be a flexible approach.

What format does the FP-51 library use? I'm sure its not the one intel uses... (1 sign bit, 7 exponent bits, 52 mantissa bits - following IEEE specification). Any comments on where it would be better to use the floating point library and where you would be better off normalizing everything to fixed point?

kundi