SEP 4, 2026 · PREPRINT
From 80x to 385x: A Best-Matching-Unit Search at the L2 Roof, Measured Against a Symmetrically Tuned Baseline
arXiv
A computational optimization report on GPU algorithm tuning with no clinical, biological, or medical application stated; this is a software engineering benchmark and does not constitute evidence for any clinical or scientific question.
Insufficient InformationAndrew James Amos
Reported
Speedup per epoch (SparseBin tune…5.6–10.1×
Speedup over earlier CUDA impleme…~80× to ~385×
Speedup of tuned cuSPARSE baseline2–3×