Optimize Eq Poly Construction during Grand Products

[sragss]

During each layer of grand products we compute a progressively larger multilinear lagrange basis polynomial EqPolynomial_0(r_0), EqPolynomial_1(r_0, r_1), ... EqPolynomial_n(r_0, r_1, ... r_n).

The polynomial is computed fresh every time but can take as input the previous layer's EqPolynomial. EqPolynomial_n(EqPolynomial_{n-1}) to accelerate by O(n/2).

[sragss]

Looks like 50% of the largest layer's Eq evals is unsafe_alloc_zero_vec so probably not worth doing. Maybe worthwhile for Spartan's sumchecks.

[moodlezoup]

after discussion with Justin we decided that this is not actually possible