Specify the reduced-interval accuracy of trigonometric polynomials

Aperta
  • Qualcosa non va?
#1,041 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
38/100
Tipo di issue
Funzionalità
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Ambito
compilers, performance, testing-qa

Direzione di ricerca

Start with std/math/trig.wave:69, the reported small-angle probes, and the related large-angle work in #600. Before changing code, discuss the proposed absolute, relative, and ULP contract with @LunaStev, including behavior near zeros and poles. Done means the bounds are agreed, independent high-precision fixtures cover interval and quadrant boundaries, the reduced-interval approximation meets the contract, and the #600 regressions plus separate f32 wrapper checks still pass.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

discussion enhancement help wanted

Motivation

Small-angle testing avoids the large-angle reduction problem fixed in #600. At x=0.75, sin_f64 is 92 ULP from a 100-digit nearest reference; near pi/4 it is 184 ULP. cos differs by 4/10 ULP for those inputs. No correct-rounding contract is currently promised.

Starting point: std/math/trig.wave:69.

Validation

Audited canonical Wave 540728c2eb6cc12f5c69c3fb3fdb96e07ba059a4 (0.0.1-alpha-dev), whose tree matches the local checkout. Validation: Executed exact-bit small-angle probes and high-precision oracle; independent of huge-argument phase loss. Executed compiler probes use the matching checkout-built compiler and std on Linux amd64 (Rust 1.89.0 / LLVM 21.1.8). Source/header or mocked checks are not native execution on another OS.

Scope

Set an explicit error envelope for the reduced-interval sin/cos polynomial and resulting tan behavior. Improve the approximation only to meet that agreed goal; do not reopen the already replaced large-angle reduction algorithm as the same defect.

Related: #600. These links do not claim those broader or completed scopes are unfinished.

Design discussion required

This issue records a proposal or an unresolved contract; it is not approval to choose new semantics. Please mention @LunaStev to discuss the design before starting implementation.

Completion criteria

  • Define small/reduced-interval absolute/relative/ULP bounds and behavior near zeros/poles.
  • Use independent high-precision fixtures at interval/quadrant boundaries.
  • Retain #600 large-angle and non-finite regressions; verify f32 wrappers separately.
Lingua principale
Rust
Stelle
56
Fork
21
Merge medio
20h 49m
PR unite (30g)
45

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di wavefnd/Wave

Tutte le issue di wavefnd/Wave

Issue simili

Altre issue su Rust

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.