[Cast] CastAudioSink audio reference_time adjustment and fixes/clean-up.

[Cast] CastAudioSink audio reference_time adjustment and fixes/clean-up.

This change adds a computation to adjust the reference_time based on the
amount of audio data backlogged in the resampler buffers.  In practice,
it seems to offset the timestamp by -20ms, and the performance test
shows this moves overall AV sync almost 20ms closer to zero (good).

A number of bugs and cleanups were also made in this change: 1) Calls to
AudioBus::FromInterleaved() were passing an invalid 3rd argument, which
happened to work because sizeof(int16) == number_of_channels.
2) Possible, yet improbable, FIFO buffer overruns are now checked.
3) Unused members of the CastAudioSink class were removed.  4) The fast,
non-resampling code path is now reachable.

BUG=335335, 437064
Committed: https://crrev.com/1cf451a6580a01a12a3c8bc95d3752e69eb31692
Cr-Commit-Position: refs/heads/master@{#306338}

[miu, 2014-11-27 00:39:36]

miu@chromium.org changed reviewers:
+ hubbe@chromium.org

[miu, 2014-11-27 00:39:37]

hubbe: PTAL.

[hubbe, 2014-12-01 22:49:18]

https://codereview.chromium.org/759383003/diff/20001/chrome/renderer/media/ca...
File chrome/renderer/media/cast_rtp_stream.cc (right):

https://codereview.chromium.org/759383003/diff/20001/chrome/renderer/media/ca...
chrome/renderer/media/cast_rtp_stream.cc:420: if (fifo_->frames() <
converter_->ChunkSize())
I wonder if we may need this to be:

while(fifo_frames() < converter_->ChunkSize()) {
  // allocate audio bus
  // resample
  // InsertAudio
}

I'm worried that the fifo could possibly grow otherwise if the number of samples
needed to generate one output frame is not an integer.

[miu, 2014-12-02 00:06:31]

PTAL.

https://codereview.chromium.org/759383003/diff/20001/chrome/renderer/media/ca...
File chrome/renderer/media/cast_rtp_stream.cc (right):

https://codereview.chromium.org/759383003/diff/20001/chrome/renderer/media/ca...
chrome/renderer/media/cast_rtp_stream.cc:420: if (fifo_->frames() <
converter_->ChunkSize())
On 2014/12/01 22:49:18, hubbe wrote:
> I wonder if we may need this to be:
> 
> while(fifo_frames() < converter_->ChunkSize()) {
>   // allocate audio bus
>   // resample
>   // InsertAudio
> }
> 
> I'm worried that the fifo could possibly grow otherwise if the number of
samples
> needed to generate one output frame is not an integer.

Done.

[hubbe, 2014-12-02 00:45:02]

lgtm

https://codereview.chromium.org/759383003/diff/60001/chrome/renderer/media/ca...
File chrome/renderer/media/cast_rtp_stream.cc (right):

https://codereview.chromium.org/759383003/diff/60001/chrome/renderer/media/ca...
chrome/renderer/media/cast_rtp_stream.cc:434: if (fifo_->frames() <
converter_->ChunkSize())
It seems like you're making the loop more complicated to save a few cycles, I
don't think it's worth it. (The compiler might be able to figure it out
anyways...)

[miu, 2014-12-02 02:36:19]

https://codereview.chromium.org/759383003/diff/60001/chrome/renderer/media/ca...
File chrome/renderer/media/cast_rtp_stream.cc (right):

https://codereview.chromium.org/759383003/diff/60001/chrome/renderer/media/ca...
chrome/renderer/media/cast_rtp_stream.cc:434: if (fifo_->frames() <
converter_->ChunkSize())
On 2014/12/02 00:45:02, hubbe wrote:
> It seems like you're making the loop more complicated to save a few cycles, I
> don't think it's worth it. (The compiler might be able to figure it out
> anyways...)

Good point.  Done.

[miu, 2014-12-02 02:36:22]

The CQ bit was checked by miu@chromium.org

[commit-bot: I haz the power, 2014-12-02 02:37:25]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/759383003/80001

[commit-bot: I haz the power, 2014-12-02 03:58:40]

Committed patchset #5 (id:80001)

[commit-bot: I haz the power, 2014-12-02 03:59:26]

Patchset 5 (id:??) landed as
https://crrev.com/1cf451a6580a01a12a3c8bc95d3752e69eb31692
Cr-Commit-Position: refs/heads/master@{#306338}