comppi



Richard Fateman wrote:
> Could you tell us about your timing results? On what computer and which
> lisp?

I would like to see timing results too, just out of curiosity.

> 
> I think that changes to code that works already should be held to extremely
> high standards.

I belong to the school of "if it ain't broke, don't fix it", but I think 
I might be in the minority. :-)
>> I changed float.lisp/comppi and also wrote a new function 
>> float.lisp/compe which computes 
>> E by some modified Taylor series.
>>
>> Volker van Nek

Looking at the commit message, I see you, Volker, have this comment:

+(defun fpe1 nil
+  (bcons (list (fpround (compe (+ fpprec 12))) (+ -12 *m))))
+;;
+;; compe is the bignum part of the bfloat(%e) computation
+;; (compe N)/(2.0^N) is an approximation to E
+;; The algorithm is based on the series
+;;
+;; %e = sum( 1/i! ,i,0,inf )
+;;
+;; but sums up 1001 terms to one.

Why only 1001 terms?  But a glance at the actual code seems to show that 
you keep summing until H is zero, so perhaps the comment is not quite 
right.  Or maybe I don't understand the comment at all.

Ray