numerical integration



On 03/16/2010 03:25 PM, Richard Hennessy wrote:
> This is pretty fun.  How many digits of accuracy do you want?  I think
> that needs to be specified or else there is no point to it.
>
> Rich

Perhaps Ted Woollett's tanh-sinh code would help here:

http://www.csulb.edu/~woollett/quad_ts.mac

It works in arbitrary big float precision, it will convert the 
integration interval from a finite one to an infinite one, and then do 
simple Riemann sums over a finite range of the new variable... this 
might "stretch out" the excessive growth enough to capture the result we 
want.