figwig: auto-optimized multithreaded bigWig reader/writer, 20x faster FIMO
jmschreiber91 · x · 2026-10-11
The author applied his auto-optimization loop to genomics I/O tools: bam2bw and tangermeme's extractloci got massive speedups, and the result is figwig, an open-source multithreaded bigWig reader/writer that moves data directly into and out of numpy (MIT, built on work by Nezar Abdennur and Jack Huey). Instead of one GIL-holding call per window, figwig reads tens of thousands of windows in a single call into a float32 array and writes predictions back the same way. FIMO was also optimized to run 20x faster, with changes already merged into memesuite-lite.
More from Infra
- Defending Musk's 'no custom equipment': how standardization inertia slows chip fab innovation — jwt0625 · 2026-10-11
- JPMorgan, Morgan Stanley struggle 6 months to offload $38B Oracle data center debt — SumitGup · 2026-10-11
- Morgan Stanley: AMD's Venice CPU may get full CoWoS-like packaging at ASE, later FOPLP at Powertech — zephyr_z9 · 2026-10-11
- Booking.com's 100M-embedding test shows purpose-built vector DBs outperform general-purpose search engines — CShorten30 · 2026-10-11
- Power Is Only ~7% of a 1GW Data Center's Cost — Yet the Hardest to Get — AccBalanced · 2026-10-11
- EIA forecasts record US electricity generation in 2026-27 as data center demand soars — emmanuelvivier · 2026-10-11