Fast Evaluation of Generalized Todd Polynomials: Applications to MacMahon's Partition Analysis and Integer Programming

Guoce Xin, Yingrui Zhang, ZiHao Zhang

Published 2023-04-26Version 1

The Todd polynomials $td_k=td_k(b_1,b_2,\dots,b_m)$ are defined by their generating functions $$\sum_{k\ge 0} td_k s^k = \prod_{i=1}^m \frac{b_i s}{e^{b_i s}-1}.$$ It appears as a basic block in Todd class of a toric variety, which is important in the theory of lattice polytopes and in number theory. We find generalized Todd polynomials arise naturally in MacMahon's partition analysis, especially in Erhart series computation.We give fast evaluation of generalized Todd polynomials for numerical $b_i$'s. In order to do so, we develop fast operations in the quotient ring $\mathbb{Z}_p[[x]]$ modulo $s^d$ for large prime $p$. As applications, i) we recompute the Ehrhart series of magic squares of order 6, which was first solved by the first named author. The running time is reduced from 70 days to about 1 day; ii) we give a polynomial time algorithm for Integer Linear Programming when the dimension is fixed, with a good performance.