GitLab

Decoding a bitstream in VA has two steps: one, submitting the parsed
parameters and encoded chunk as VABufferIDs, and two, executing the
decode. This CL transfers managing the lifetime of the VABufferIDs to
VP9VaapiVideoDecoderDelegate, avoiding vaCreateBuffer() calls.

To do that this CL reshuffles a bit the VaapiWrapper calls to be able
to call vaCreateBuffer(), MapAndCopy and Execute, in a piecemeal,
configurable fashion.

This is verified via chrome:tracing and codepen.io/full/RwarYvG that
plays 4 1280x572 VP9 videos at the same time. Tracing is captured for
a few seconds, and basically we avoid calling the vaCreateBuffer(),
which takes ~50us every time, reducing the overall decode time from
~3.740ms to ~3.425ms on my BSW (reks), ~10% reduction. (Also tried on
kohaku and numbers are similar albeit smaller bc the SoC is faster).
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).

This CL also removes the unused VaapiWrapper::Execute() method.

Bug: b/166646505
Change-Id: Id44dbb21825328dfb4fc47e1c7807bd725efd196
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2404740
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Reviewed-by: Andres Calderon Jaramillo <andrescj@chromium.org>
Cr-Commit-Position: refs/heads/master@{#808040}