DioCrafts b505a974b9 docs(thumbnails): record Phase 1.7 (rayon) as tested-and-reverted ...

Benchmark refuted the rayon-oversubscription hypothesis: throughput stayed flat
at the real operating point (14 permits) and PNG single-image latency regressed
66% when the 3-size resize was made sequential. The code was reverted; this
records the negative result in benches/BASELINE.md so it isn't retried (like the
dropped defer-Large task).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-21 16:03:06 +02:00

..

BASELINE.md

docs(thumbnails): record Phase 1.7 (rayon) as tested-and-reverted

2026-06-21 16:03:06 +02:00

thumbnails.rs

perf(thumbnails): shrink-on-load JPEG decode (1.8-2× faster, 5-15× less RAM)

2026-06-21 15:13:03 +02:00