Tom Boothby: Implement Pippenger's Algorithm for multivariate polynomial evaluation

[http://cr.yp.to/papers/pippenger.pdf Paper]

Skip to page 13. I've already implemented:

Direct computation (option 1 in the paper)
Input partitioning (option 2)
Most of input clumping (option 3)

I need:

Output partitioning
Output clumping
Some insight as to why DJB thinks that one can "quickly compute the parameter sequence given p, q. I'll bring a print copy of the 3 papers by Pippenger.
The general case (not certain this is a reasonable goal for the amount of time we have)