festifun/backend/tests
m3ultra 95b71788a3 Sync solver: cache per-signal FFTs across the pairwise GCC-PHAT stage
pairwise_offsets now computes each signal's rFFT once at a shared padded
length (2*max_len) and reuses it for every pair: O(K) forward FFTs +
cheap per-pair spectrum products instead of O(K^2) full FFTs. Each
pair's lag search stays capped at +/-(la+lb)/2 — the same window the
per-pair transform used — and gcc_phat() itself is unchanged for
callers (drift windows, tests). Measured 1.8x on 6 cams x 10 min;
the advantage grows with camera count.

New test locks the cached path against direct gcc_phat, including
unequal-length signals where the padded sizes genuinely differ.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-17 17:02:27 +10:00
..
conftest.py Foundation (M0 + M3 + frozen contracts): scaffold, synthetic fixtures, full API 2026-07-16 00:45:55 +10:00
test_api.py Live capture: USB webcams, laptop cams, and phones -> data/raw 2026-07-17 15:43:51 +10:00
test_audio_sync.py Sync solver: cache per-signal FFTs across the pairwise GCC-PHAT stage 2026-07-17 17:02:27 +10:00
test_capture.py Live capture: USB webcams, laptop cams, and phones -> data/raw 2026-07-17 15:43:51 +10:00
test_events_ai.py lane D (M7): audio moment detection + pluggable AI classification 2026-07-16 08:34:56 +10:00
test_geometry.py Lane B (M2 + M8 geometry): COLMAP reconstruction, geometry math, graceful degradation 2026-07-16 08:40:46 +10:00
test_ingest.py Lane A (M1): ingest + GCC-PHAT audio sync 2026-07-16 01:13:56 +10:00
test_resolve.py Phase 4 polish: anchor supersede + delete, docs/ideas, field-test scaffolding 2026-07-16 12:15:35 +10:00
test_sfm.py Lane B (M2 + M8 geometry): COLMAP reconstruction, geometry math, graceful degradation 2026-07-16 08:40:46 +10:00
test_synthetic.py Foundation (M0 + M3 + frozen contracts): scaffold, synthetic fixtures, full API 2026-07-16 00:45:55 +10:00