From 9bc928ee16389ab0da232781fe45701c40cac195 Mon Sep 17 00:00:00 2001 From: Dev Ojha Date: Thu, 3 Sep 2026 19:30:37 +0200 Subject: [PATCH] Clarify public power optimization impact --- docs/changelog/unreleased/333.md | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/docs/changelog/unreleased/333.md b/docs/changelog/unreleased/333.md index 6f5b9dab..5ac9f0a0 100644 --- a/docs/changelog/unreleased/333.md +++ b/docs/changelog/unreleased/333.md @@ -12,8 +12,8 @@ - Changed the public-domain `x^(2^k)` calculation used by proof creation and verification to call Pasta's dedicated repeated-squaring implementation. - At `k = 11`, the isolated calculation was 63.6x faster on Apple M4 and - 64.9x faster on x86_64 Linux; proof bytes and transcripts are unchanged + At `k = 11`, the isolated calculation was 64.1–64.4x faster on Apple M4 and + 64.9–65.0x faster on x86_64 Linux; proof bytes and transcripts are unchanged ([#333](https://github.com/zakura-core/common/pull/333)). ## zakura-halo2-gadgets @@ -22,8 +22,8 @@ - Changed generic fixed-base window preprocessing to advance public window scales iteratively instead of recomputing each power with constant-time - exponentiation, reducing full-width Lagrange-coefficient generation by - 14.7% on Apple M4 and 16.3% on x86_64 Linux. Orchard uses precomputed - coefficient tables, so these measurements do not represent an Orchard - key-generation speedup + exponentiation. For an 85-window full-width table, this replaces 764 + constant-time exponentiations with 168 scalar multiplications. Orchard uses + precomputed coefficient tables, so this is not an Orchard key-generation + speedup ([#333](https://github.com/zakura-core/common/pull/333)).