Hardware encoding follow-ups after #857 #866

Open
opened 2026-11-18 05:24:32 +00:00 by charles · 0 comments
Owner

From the report of PR #865:

  • VAAPI: the I420 to NV12 upload is a per-pixel CPU loop in the encoder thread (5.9 s vs NVENC 2.7 s of encoder-thread CPU per minute at 1080p60); use libyuv I420ToNV12 or vaPutImage.
  • VAAPI has no H.265 or AV1 encoder in webrtc-sys, although radeonsi encodes HEVC.
  • VAAPI still sends an IDR every 5 s (NVENC uses an infinite GOP).
  • Two NVIDIA GPUs: NVENC uses CUDA device 0, which may not be the display GPU.
  • Untested: Intel VAAPI (no hardware), hybrid laptops (rule unit-tested only), Chromium/Electron receivers of the VAAPI stream, the Encoding GPU setting in the running app.
  • Windows: see #229.
From the report of PR #865: - VAAPI: the I420 to NV12 upload is a per-pixel CPU loop in the encoder thread (5.9 s vs NVENC 2.7 s of encoder-thread CPU per minute at 1080p60); use libyuv I420ToNV12 or vaPutImage. - VAAPI has no H.265 or AV1 encoder in webrtc-sys, although radeonsi encodes HEVC. - VAAPI still sends an IDR every 5 s (NVENC uses an infinite GOP). - Two NVIDIA GPUs: NVENC uses CUDA device 0, which may not be the display GPU. - Untested: Intel VAAPI (no hardware), hybrid laptops (rule unit-tested only), Chromium/Electron receivers of the VAAPI stream, the Encoding GPU setting in the running app. - Windows: see #229.
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
charles/fluxer-native#866
No description provided.