Modifies size of ring buffer in shared memory on the audio capture side

Modifies size of ring buffer in shared memory on the audio capture side.

We have seen reports in Chrome where segments of repeated input audio has
damaged AEC performance in WebRTC clients. This patch is an attempt to limit
the number of possible places in Chrome where such a patter could be created.

It is a speculative patch and the idea is that it should live in Canary
for a while so we can see if there are any positive effects.

TBR=tommi
BUG=b/13976602
TEST=manual tests of different WebRTC clients running in loopback to ensure that no
audio artifacts can be detected

Committed: https://crrev.com/bda4fc7018c3ce93d88b74461a065f9eaf0b8b5f
Cr-Commit-Position: refs/heads/master@{#340261}

[henrika (OOO until Aug 14), 2015-07-24 09:17:06]

henrika@chromium.org changed reviewers:
+ tommi@chromium.org

[henrika (OOO until Aug 14), 2015-07-24 09:25:54]

TBR tommi since we would like to start monitoring any effects of this change and
tommi is currently OOO.

[henrika (OOO until Aug 14), 2015-07-24 09:26:00]

The CQ bit was checked by henrika@chromium.org

[commit-bot: I haz the power, 2015-07-24 09:26:32]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1254793002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1254793002/40001

[commit-bot: I haz the power, 2015-07-24 12:09:42]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-07-24 12:10:25]

Patchset 3 (id:??) landed as
https://crrev.com/bda4fc7018c3ce93d88b74461a065f9eaf0b8b5f
Cr-Commit-Position: refs/heads/master@{#340261}

[DaleCurtis, 2015-07-24 16:16:55]

dalecurtis@chromium.org changed reviewers:
+ dalecurtis@chromium.org

[DaleCurtis, 2015-07-24 16:16:56]

Hmm, this is a dangerous patch -- I suspect you're going to break a lot of our
input users as this was the only thing preventing shared memory from being
trampled by back to back delivery. Did you verify the original bugs around this
are no longer happening?

[henrika (OOO until Aug 14), 2015-07-24 19:39:08]

I did this as an experiment on Canary to see if we can see any improvements in
the WebRTC AEC performance which currently suffers from bad performance due to
segments of repeated audio. This was the only place in Chrome I could find which
could possibly cause these repetitions. I will revert if there is no positive
effect. Also, the initial issue was for Android "only" and I have not seen any
reports for other platforms.

Dale, if you have any other ideas regarding potential sources for repeated audio
segments in the capture audio path, do ping me.

[DaleCurtis, 2015-07-24 19:41:21]

CrOS had a recent bug which was causing problems like you describe,
https://code.google.com/p/chrome-os-partner/issues/detail?id=41120

No idea which platforms you're seeing this on.

[henrika (OOO until Aug 14), 2015-07-24 19:43:35]

We have mainly seen it on Windows and Mac.

On Fri, Jul 24, 2015 at 9:41 PM, <dalecurtis@chromium.org> wrote:

> CrOS had a recent bug which was causing problems like you describe,
> https://code.google.com/p/chrome-os-partner/issues/detail?id=41120
>
> No idea which platforms you're seeing this on.
>
> https://codereview.chromium.org/1254793002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[Henrik Grunell, 2015-08-07 11:41:36]

A revert of this CL (patchset #3 id:40001) has been created in
https://codereview.chromium.org/1277173002/ by grunell@chromium.org.

The reason for reverting is: This was an speculative change and should now be
reverted. There is no sign that this change would fix or improve the issue..