Use GrallocBuffer in MediaPipeline for B2G directly can save 1 buffer copy per frame, which can reduce memory bandwidth consumption and save CPU power.
Created attachment 753588 [details] [diff] [review]
Use GrallocYCbCrImage directly to reduce memcpy
Current implementation for WebRTC decoding path will cause texture upload before the frame can be composited. Use GrallocPlanarYCbCrImage can share the decoded frame with compositor process directly and avoid texture upload.