Rounding a half-integer Gaussian point moves it by at most (Lemma 2.1)
ProvedIntMul.HvdH.rho0C_sub_halfinteger-multiplication
Let be such that and are integers. Then rounding towards zero componentwise moves by at most :
This is the estimate in the proof of Lemma 2.1 (addition/subtraction) of Harvey–van der Hoeven. For fixed-point inputs , satisfies the hypothesis after scaling by , so has error .
Preamble
import Definitions.Def_IntMul_HvdH_FixedPoint
Formal statement
namespace IntMul.HvdH
theorem rho0C_sub_half (u : ℂ) (hre : ∃ n : ℤ, 2 * u.re = n) (him : ∃ n : ℤ, 2 * u.im = n) :
‖rho0C u - u‖ ≤ Real.sqrt 2 / 2 := 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, Section 2.2, proof of Lemma 2.1