ashtonsix/perf-portfolio

HPC research and demonstrations

132 best points · 63 comments · 5 Hacker News posts · 117 GitHub stars · View on GitHub

Hacker News discussions

• 86 GB/s bitpacking with ARM SIMD (single thread) (132 points, 29 comments, by ashtonsix, 2025-10-05)
• Prefix sum: 20 GB/s (2.6x baseline) (89 points, 32 comments, by ashtonsix, 2025-10-14)
• 46% smaller OLAP indexes with BSP-style bitsets (3 points, 0 comments, by ashtonsix, 2025-10-22)
• BSP-inspired bitsets: 46% smaller than Roaring (but probably not faster) (2 points, 0 comments, by ashtonsix, 2025-10-20)
• P4Synth: 1000x faster function synthesis vs. SAT (1 points, 2 comments, by ashtonsix, 2025-12-17)