Resolves "Huge increase in audio latency on Windows following r192046"

Resolves "Huge increase in audio latency on Windows following r192046".

BUG=227043
TEST=content_unittests and mish audio tests in Chrome

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=194573

[henrika (OOO until Aug 14), 2013-04-15 09:02:03]

Dale, could you check this one out?

The key here is

"if (input_params.channel_layout() != channel_layout)"

where I avoid making the new CoreAudioUtil call unless the requested layout
differs from the hardware size.

[DaleCurtis, 2013-04-15 18:08:57]

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
File media/audio/win/audio_manager_win.cc (right):

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:353: AudioParameters
AudioManagerWin::GetPreferredOutputStreamParameters(
Ugh, not your fault, but this whole function is really hard to follow :(  Feel
free to take a stab at rewriting the whole damn thing :)

I think we should just have multiple return sections instead of trying to funnel
all the parameters down into a single return.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:394: // Check if it is possible to open up
at the specified input channel
Indent is wrong; should be 2 spaces.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:398: if (input_params.channel_layout() !=
channel_layout) {
In the Exclusive mode case, channel_layout won't be set yet, so you'll end up
making an extra check here.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:400: input_params.channel_layout())) {
Wrong indent; you can't partially wrap some of the parameters and not others. 
It's either all or nothing.

[henrika (OOO until Aug 14), 2013-04-15 20:28:28]

Comments and question only; no changes yet. PTAL.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
File media/audio/win/audio_manager_win.cc (right):

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:353: AudioParameters
AudioManagerWin::GetPreferredOutputStreamParameters(
I agree, it sucks. Wonder who has written most parts of it ;-)

I can give it a try but I felt that I might as well add stuff to avoid
introducing andy mistakes.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:394: // Check if it is possible to open up
at the specified input channel
Will fix.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:398: if (input_params.channel_layout() !=
channel_layout) {
Can you elaborate. Do you want me to add a flag-check and do what if the flag is
enabled? I assume you want to always open at input layout in exclusive mode.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:400: input_params.channel_layout())) {
Will fix.

[DaleCurtis, 2013-04-15 20:31:07]

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
File media/audio/win/audio_manager_win.cc (right):

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:353: AudioParameters
AudioManagerWin::GetPreferredOutputStreamParameters(
On 2013/04/15 20:28:28, henrika wrote:
> I agree, it sucks. Wonder who has written most parts of it ;-)
> 
> I can give it a try but I felt that I might as well add stuff to avoid
> introducing andy mistakes.

I don't know what you're talking about!! :p  I'll take a stab at it when I have
some time.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:398: if (input_params.channel_layout() !=
channel_layout) {
On 2013/04/15 20:28:28, henrika wrote:
> Can you elaborate. Do you want me to add a flag-check and do what if the flag
is
> enabled? I assume you want to always open at input layout in exclusive mode.

In the HasSwitch() block above, I'd just add a:

if (input_params.IsValid())
  channel_layout = input_params.channel_layout();

[henrika (OOO until Aug 14), 2013-04-16 09:04:35]

Finalized the easy fixes (no refactoring).

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
File media/audio/win/audio_manager_win.cc (right):

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:394: // Check if it is possible to open up
at the specified input channel
On 2013/04/15 20:28:28, henrika wrote:
> Will fix.

Done.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:398: if (input_params.channel_layout() !=
channel_layout) {
On 2013/04/15 20:31:07, DaleCurtis wrote:
> On 2013/04/15 20:28:28, henrika wrote:
> > Can you elaborate. Do you want me to add a flag-check and do what if the
flag
> is
> > enabled? I assume you want to always open at input layout in exclusive mode.
> 
> In the HasSwitch() block above, I'd just add a:
> 
> if (input_params.IsValid())
>   channel_layout = input_params.channel_layout();

Done.

https://codereview.chromium.org/14244005/diff/1/media/audio/win/audio_manager...
media/audio/win/audio_manager_win.cc:400: input_params.channel_layout())) {
On 2013/04/15 20:28:28, henrika wrote:
> Will fix.

Done.

[DaleCurtis, 2013-04-16 17:36:01]

lgtm, did you figure out why this regressed in the first place? I'm still not
sure why this should matter.

[henrika (OOO until Aug 14), 2013-04-16 20:34:28]

I did run WebRTC unit tests which does the complete setup (audio manager, output
proxy, get pref. params, opens etc.): one with the new code and one without it.
I did 1000 repetitions a could measure an increase of 10 seconds. Hence,
10/1000=1/100=10ms was added to each round. I don't know where the large
increase comes from. This fix is the best I can do for now.

[DaleCurtis, 2013-04-16 20:36:07]

Okay, bots probably see a larger hit, but it shouldn't be this bad, I'll see if
I can figure out what is going on here.

[henrika (OOO until Aug 14), 2013-04-17 12:00:08]

Committed patchset #3 manually as r194573 (presubmit successful).