media/gpu/vaapi: batch submitting VABufferIDs for H264 decoding
This CL follows crrev.com/c/2393629 by batching up SubmitBuffer calls in the H264 decode accelerator. This is verified via chrome:tracing and codepen.io/miguelao/full/MMxGJw that plays 4 1280x546 H.264 videos at the same time. Tracing is captured for a few seconds, basically with this patch we have 2 SubmitBuffers() calls adding up to ~0.461ms/frame versus ToT individual SubmitBuffer() calls (4 of them) ~0.504ms in total -- which gives an aggregate of about ~40us improvement, or savings of about ~10%. Since decoding itself takes about the 0.390ms/frame, savings are OKish, so let's take them. This is all on my BSW device (reks), Improvements are of course extremely small, the advantages of this CL are in reducing lock/unlock churn and associated contention. This benefit grows with the amount of decodes (e.g. Meet grid scenarios). Bug: b/166646505 Change-Id: Id51bb6db39e4f7b9d67760f86b8f8963f20c2b2b Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2405586Reviewed-by:Jeffrey Kardatzke <jkardatzke@google.com> Commit-Queue: Miguel Casas <mcasas@chromium.org> Cr-Commit-Position: refs/heads/master@{#806264}
Showing
Please register or sign in to comment