1
Full Unicode Search at 50× ICU Speed with AVX‑512
ICU gets Unicode right and pays for it. This post shows a different approach: fold-safe windows, SIMD probes, and verifiers for fast UTF‑8 search.
Ash Vardanian
About Building Unum Cloud since 2015. Investing in AI & Infra companies at AAL VC. Computer Science & AI researcher (unpublished, by choice). Twice an Astrophysics dropout, lifelong Bioinformatics fan. Fluent in English, Russian & Armenian. Lived in 🇺🇸🇬🇧🇷🇺🇦🇲 & 🇲🇽🇵🇦🇦🇷🇩🇪🇦🇪🇹🇭🇲🇾🇻🇳🇮🇩. Hosting “Systems” meetups in SF…
- Popular posts
- 5
- Total points
- 873
- HN comments
- 212
- Rank
- #2,577 personal
Visit ashvardanian.com → RSS feedOn Hacker News May 2023 – Dec 2025
Systems & Performance 4Software Engineering 3AI & Machine Learning 2Databases & Data 1
2
Python, C, Assembly – Faster Cosine Similarity
From pure Python to AVX-512 assembly, optimizing cosine similarity reveals a 2,500x speedup through SIMD, FP16, and VNNI instructions on modern Intel CPUs.
3
Beyond OpenMP in C++ and Rust: Taskflow, Rayon, Fork Union
Most C++ and Rust thread pools run 10x slower than OpenMP on fork-join workloads. Fork Union closes the gap to 20% with 300 lines, dodging mutexes and CAS.