Commit ID f96eb02a55c0a6ae2fdcae16ce1542fd5801741b
|Date:||05-July-2012 at 20:32||Lines Added:||601|
|Repository:||http://git.chromium.org/git/chromium.git trunk||Lines Removed:||520|
|Commit Comment:||Make VideoDecodeAcceleratorTest work for VAVDA and fixed a bunch of bugs the test found - The bulk of the CL turns RenderingHelperEGL into RenderingHelperGL (by making it EGL-vs-GLX-agnostic) - VaapiH264Decoder::AssignPictureBuffer: it's fine to restart decode after a Reset(), so allow that. - VaapiH264Decoder::FillVARefFramesFromDPB: faster/clearer error condition, and small cleanup. - VaapiH264Decoder::Flush: emit the final in-progress picture. - VaapiH264Decoder::DecodeOneFrame: insist on having at least two output surfaces before attempting a decode to avoid running out. - VaapiVideoDecodeAccelerator::client_ is now a WeakPtr instead of a raw pointer, to do correct lifetime accounting. This also allows dropping the helper methods on VAVDA whose only purpose was to triple-check client_'s presence before dispatching calls on it. - VaapiVideoDecodeAccelerator::InitialDecodeTask: account for the possibility of multiple in-flight Decode()s, and remove unnecessary task-posting.
Two remaining issues are:
- the test-25fps.h264 testdata file triggers bug#135548
- the pure-decode performance (at least in Debug) of VAVDA is not as good as the reference that vdatest was originally built for, so some tests "fail" on lower-than-expected fps rates. This probably just needs to be a per-platform expectation.
Review URL: https://chromiumcodereview.appspot.com/10701071
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@145556 0039d316-1c4b-4281-b951-d872f2087c98
|Language||Code Added||Code Removed||Comments Added||Comment Removed||Blanks Added||Blanks Removed|
|File||Language||Code Added||Code Removed||Comments Added||Comment Removed||Blanks Added||Blanks Removed|
|content/content_common.gypi||No source code was detected in this file.|
|content/content_tests.gypi||No source code was detected in this file.|