Make audio and video in sync while playback rate != 1.0...

Make audio and video in sync while playback rate != 1.0

BUG=20290
TEST=play a video with rate 0.5, audio and video should be in sync

This change use the playback rate as a scaling factor for
delay introduced in the audio hardware buffer. After we
scaled the amount of audio buffer not played, we'll be
able to make audio / video in sync again.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=24568

[Alpha Left Google, 2009-08-26 23:44:09]

After doing some experiments, by scaling the unplayed bytes in audio hardware
buffer, we can estimate the current time closer to the correct value.

Audio and video will still be out of sync in about 1 seconds after playback rate
changes, but they will be in sync after that.

[scherkus (not reviewing), 2009-08-26 23:59:00]

nits on floating point math but LGTM

http://codereview.chromium.org/174584/diff/1/3
File chrome/renderer/media/audio_renderer_impl.cc (right):

http://codereview.chromium.org/174584/diff/1/3#newcode297
Line 297: static_cast<int64>(request_delay.InMicroseconds() *
you may want to be explicit in how the value is rounded i.e. ceil()

http://codereview.chromium.org/174584/diff/1/2
File media/filters/audio_renderer_impl.cc (right):

http://codereview.chromium.org/174584/diff/1/2#newcode60
Line 60: pending_bytes = static_cast<int>(pending_bytes * GetPlaybackRate());
same here.. ceil() perhaps?

[fbarchard, 2009-08-27 03:21:17]

build hasnt finished to confirm the fix... will have to test in the morning. 
just minor concerns

http://codereview.chromium.org/174584/diff/1/3
File chrome/renderer/media/audio_renderer_impl.cc (right):

http://codereview.chromium.org/174584/diff/1/3#newcode295
Line 295: if (GetPlaybackRate() != 1.0f) {
shouldnt you do this code within the 
if (current_time > request_timestamp_) {
above?

http://codereview.chromium.org/174584/diff/1/2
File media/filters/audio_renderer_impl.cc (right):

http://codereview.chromium.org/174584/diff/1/2#newcode59
Line 59: //                  without clicking.
we can probably remove this todo, now that we have triple buffering.  a single
zero packet wont cause a click.

http://codereview.chromium.org/174584/diff/1/2#newcode62
Line 62: base::Time::kMicrosecondsPerSecond * pending_bytes /
bytes_per_second_);
this can come close to overflowing an int can't it?