unbalanced mul/div

[Paul Zimmermann]

       Hi,

I've done a similar comparison for total size 1000 limbs:

http://www.loria.fr/~zimmerma/tmuldiv1k.ps

(the previous http://www.loria.fr/~zimmerma/tmuldiv10k.ps was for 10k limbs)

In this case the improvement for division is marginal, and there is still
a negative spike for division of 1000 by 500 limbs. The improvement between
4.3.2 and 5.0.0 for multiplication is more visible, and there is now a "flat"
region from 250 to 750 limbs.

Paul