Pull the Voice Activity Detector out from the AGC

Pull the Voice Activity Detector out from the AGC

This change generates bit-exact values when running through audioproc_f than before.

This change was originally uploaded here:
* https://codereview.webrtc.org/1181933002/
* https://codereview.webrtc.org/1177043017/

And reverted because of an ASAN problem in Chrome here:
* https://codereview.webrtc.org/1192863006/
* https://codereview.webrtc.org/1194963003/

TBR=andrew@webrtc.org

Committed: https://crrev.com/ecf6b81644af9823dbff5c24a3d5b9bb596c0d5b
Cr-Commit-Position: refs/heads/master@{#9505}

[aluebs-webrtc, 2015-06-25 17:05:51]

The CQ bit was checked by aluebs@webrtc.org

[commit-bot: I haz the power, 2015-06-25 17:06:01]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1212543002/1

[commit-bot: I haz the power, 2015-06-25 19:28:53]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-06-25 19:29:02]

Patchset 1 (id:??) landed as
https://crrev.com/ecf6b81644af9823dbff5c24a3d5b9bb596c0d5b
Cr-Commit-Position: refs/heads/master@{#9505}

[Andrew MacDonald, 2015-06-29 03:35:40]

lgtm

Just curious, what was the ASAN fix?

[aluebs-webrtc, 2015-06-29 23:34:32]

https://codereview.webrtc.org/1212543002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/vad/voice_activity_detector.cc (right):

https://codereview.webrtc.org/1212543002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/vad/voice_activity_detector.cc:58:
CHECK_EQ(standalone_vad_->AddAudio(resampled_ptr, length), 0);
The bug that ASAN caught was that I was passing in |audio| instead of
|resampled_ptr|. I guess the test cases for which I tested bit exactness didn't
exercise the resampler at all.

[Andrew MacDonald, 2015-06-29 23:38:23]

https://codereview.webrtc.org/1212543002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/vad/voice_activity_detector.cc (right):

https://codereview.webrtc.org/1212543002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/vad/voice_activity_detector.cc:58:
CHECK_EQ(standalone_vad_->AddAudio(resampled_ptr, length), 0);
On 2015/06/29 23:34:32, aluebs-webrtc wrote:
> The bug that ASAN caught was that I was passing in |audio| instead of
> |resampled_ptr|. I guess the test cases for which I tested bit exactness
didn't
> exercise the resampler at all.

Thanks goodness for ASAN.

[aluebs-webrtc, 2015-06-29 23:40:56]

https://codereview.webrtc.org/1212543002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/vad/voice_activity_detector.cc (right):

https://codereview.webrtc.org/1212543002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/vad/voice_activity_detector.cc:58:
CHECK_EQ(standalone_vad_->AddAudio(resampled_ptr, length), 0);
On 2015/06/29 23:38:23, andrew wrote:
> On 2015/06/29 23:34:32, aluebs-webrtc wrote:
> > The bug that ASAN caught was that I was passing in |audio| instead of
> > |resampled_ptr|. I guess the test cases for which I tested bit exactness
> didn't
> > exercise the resampler at all.
> 
> Thanks goodness for ASAN.

Certainly :)