docs update
This commit is contained in:
@@ -102,7 +102,7 @@ This work is currently split across:
|
||||
|
||||
- [DeckLinkSession::ConfigureInput](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/videoio/decklink/DeckLinkSession.cpp:221)
|
||||
- [CaptureDelegate::VideoInputFrameArrived](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/videoio/decklink/DeckLinkFrameTransfer.cpp:33)
|
||||
- [OpenGLVideoIOBridge::VideoFrameArrived](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp:50)
|
||||
- [OpenGLVideoIOBridge::UploadInputFrame](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp:11)
|
||||
|
||||
### 3. Output Lifecycle and Output Callback Handling
|
||||
|
||||
@@ -307,7 +307,7 @@ Today the callback path effectively does this:
|
||||
|
||||
That path is visible in:
|
||||
|
||||
- [OpenGLVideoIOBridge::PlayoutFrameCompleted](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp:83)
|
||||
- [OpenGLVideoIOBridge::RenderScheduledFrame](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp:18)
|
||||
|
||||
This couples output timing directly to render work.
|
||||
|
||||
@@ -354,7 +354,7 @@ Recommended default policy for live playout:
|
||||
|
||||
The current "skip upload if the GL bridge is busy" behavior is directionally correct for live timing:
|
||||
|
||||
- [OpenGLVideoIOBridge::VideoFrameArrived](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp:57)
|
||||
- [OpenGLVideoIOBridge::UploadInputFrame](/c:/Users/Aiden/Documents/GitHub/video-shader-toys/apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp:11)
|
||||
|
||||
But in the target architecture that decision should move out of GL lock acquisition and into an explicit backend-to-render handoff queue policy.
|
||||
|
||||
|
||||
Reference in New Issue
Block a user