alexei.dolgolyov fbf597dc29 Optimize streaming pipeline and capture hot paths ...

- Replace asyncio.to_thread with dedicated ThreadPoolExecutor (skip
  per-frame context copy overhead)
- Move brightness scaling into _process_frame thread (avoid extra
  numpy array copies on event loop)
- Remove PIL intermediate in MSS capture (direct bytes→numpy)
- Unify median/dominant pixel mapping to numpy arrays (eliminate
  Python list-of-tuples path and duplicate Phase 2/3 code)
- Cache CalibrationConfig.segments property (avoid ~240 rebuilds/sec)
- Make KC WebSocket broadcasts concurrent via asyncio.gather
- Fix fps_samples list.pop(0) → deque(maxlen=10) in both processors
- Cache time.time() calls to reduce redundant syscalls per frame
- Log event queue drops instead of silently discarding

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-02-19 22:55:21 +03:00

..

capture

Optimize streaming pipeline and capture hot paths

2026-02-19 22:55:21 +03:00

capture_engines

Optimize streaming pipeline and capture hot paths

2026-02-19 22:55:21 +03:00

devices

Add static color support, HAOS light entity, and real-time profile updates

2026-02-19 14:23:47 +03:00

filters

Refactor core/ into logical sub-packages and split filter files

2026-02-18 12:03:29 +03:00

processing

Optimize streaming pipeline and capture hot paths

2026-02-19 22:55:21 +03:00

profiles

Replace WMI process enumeration with Win32 EnumProcesses (350x faster)

2026-02-19 22:30:22 +03:00

__init__.py

Refactor core/ into logical sub-packages and split filter files

2026-02-18 12:03:29 +03:00