This was seen from a numerical analysis text/reference.

There is a vague statement here: square the result repeatedly. The scaling is rather clear: divide by 2 until $|x|<1$. How about the squaring? I tried squaring the result by the number of times I scaled, and the result was way off.
Have you seen any proof or justification to this technique?
thanks
Seems to be working for me. Just make sure you follow the directions exactly. The first sixteen terms of the Taylor expansion are $e^x=1+x+\frac{x^2}{2!}+...+\frac{x^{15}}{15!}$. So write $e^{30}$ as $(e^{30/2^5})^{2^5}$, because $2^5$ is the smallest power of 2 that makes the exponent less in absolute value than 1. Do the Taylor expansion for $e^{30/2^5}$, and then raise the result to the power of $2^5$. It matches $e^{30}$. I also tried $\exp -3$ and it worked.
-out