-bit approximation of roots of unity
ProvedIntMul.HvdH.twiddle_boundinteger-multiplication
Let , and be integers, and let be the Gaussian integer computed by twiddle n k j (see IntMul_HvdH_Twiddle). Then
This is the content of Lemma 2.12 of Harvey–van der Hoeven (computing the twiddle factors of the FFT to precision ), for an explicit algorithm: reduction of the angle to by exact symmetries of , followed by a fixed-point Taylor series.
Preamble
import Definitions.Def_IntMul_HvdH_Twiddle import Theorems.Thm_IntMul_HvdH_piBits_bound import Theorems.Thm_IntMul_HvdH_expBits_bound
Formal statement
namespace IntMul.HvdH
theorem twiddle_bound (n k j : ℕ) (hk : 0 < k) :
‖((twiddle n k j).1 : ℂ) + ((twiddle n k j).2 : ℂ) * Complex.I -
2 ^ n * Complex.exp (2 * Real.pi * Complex.I * j / k)‖ ≤ 1 := by sorry
end IntMul.HvdHSource
Harvey, van der Hoeven, Integer multiplication in time O(n log n), Annals of Math. 193 (2021), https://hal.science/hal-02070778, Lemma 2.12 (complex exponentials e^{2 pi i j/k})