Fast Forgiving Base64 Decoding with SIMD Instructions [r-libre/4275]

Lemire, Daniel ORCID: https://orcid.org/0000-0003-3306-6922 (sous presse). Fast Forgiving Base64 Decoding with SIMD Instructions. ACM Transactions on Architecture and Code Optimization.

Fichier(s) associé(s) à ce document :

[thumbnail of TACO-2026-292-final-noacmref.pdf]

PDF - TACO-2026-292-final-noacmref.pdf
Contenu du fichier : Document de travail
Licence : Creative Commons CC BY.

Télécharger

Catégorie de document :	Articles de revues
Évaluation par un comité de lecture :	Oui
Étape de publication :	Accepté (sous presse)
Résumé :	On the Internet, we use the Base64 format to represent binary data (e.g., images or cryptographic keys) as text. The ES2026 JavaScript standard, following the WHATWG forgiving-base64 rules, requires decoders to ignore ASCII whitespace while rejecting other non-alphabet characters. Prior SIMD Base64 decoders assume whitespace-free input. We present a block-oriented decoding algorithm that removes whitespace inside the SIMD loop: clean 64-character blocks are decoded directly, while blocks with ignorable characters are compacted into a small stack buffer. We provide implementations for x64 (SSSE3 through AVX-512 VBMI2), ARM NEON, and LoongArch. Our C++ decoder reaches up to 94 GB/s on recent AMD Zen 5 processors. In Node.js 25, integration of our decoder raises Uint8Array.fromBase64 throughput by up to 6x compared with Node.js 19. The implementation has been adopted by major browsers and JavaScript runtimes.
Déposant:	Lemire, Daniel
Responsable :	Daniel Lemire
Dépôt :	13 oct. 2026 18:03
Dernière modification :	13 oct. 2026 18:03

Actions (connexion requise)

RÉVISER