I have an lengthy analytical function that has many sums of products divided by sums of products. The variables that make up the function consist of large numbers and very small numbers.

I would like to re-arrange the function to reduce numerical roundoff error. What rules or recommendations are there regarding this?

Hi,

plzz see "HP-UX Floating-Point Guide HP 9000 Computers
Edition 5",

its freely downloadable at http://docs.hp.com/en/B3906-90006/index.html


chapter 3, might be helpful to you.

regards
Zubair