OSelymesi/USM-Core

Universal Streaming Mechanism: A lightweight GPU paradigm for irregular, ragged, and nested workloads. 2.5x faster than naive baselines.

2 best points · 1 comments · 1 Hacker News post · 4 GitHub stars · View on GitHub

Hacker News discussions

• Show HN: USM-Core Header-only CUDA library for ragged reductions(2.5x baseline) (2 points, 1 comments, by ottoselymesi, 2026-01-04)