Move WakeUpGpu logic to GpuChannelManager

Move WakeUpGpu logic to GpuChannelManager

We don't need to tie the WakeUpGpu to a particular surface, so instead
make it a control message, and have the GpuChannelManager pick any
available context. We signal GPU access on AsyncFlush instead of
OnMakeCurrent, since that's both more accurate and simpler.

At that point, that means renderer surfaces don't need to be
ImageTransportSurfaces any more, so we can remove NullTransportSurface
and ImageTransportSurfaceAndroid, we can use the default offscreen
surface instead anyway.
DirectSurfaceAndroid can be replaced by PassThroughImageTransportSurface
directly, because it's only used to signal GPU access (which is done on
AsyncFlush anyway).

BUG=487471
Committed: https://crrev.com/fa4f254803e2952851400a35f9ec20605e9c3461
Cr-Commit-Position: refs/heads/master@{#350491}

[piman, 2015-09-23 00:31:46]

piman@chromium.org changed reviewers:
+ inferno@chromium.org, sievers@chromium.org

[piman, 2015-09-23 00:31:46]

sievers: please review
inferno: OWNERS for gpu_messages.h

[inferno, 2015-09-23 01:28:30]

inferno@chromium.org changed reviewers:
+ tsepez@chromium.org
- inferno@chromium.org

[piman, 2015-09-23 06:39:20]

On 2015/09/23 00:31:46, piman (slow to review) wrote:
> inferno: OWNERS for gpu_messages.h

Not sure what happened here. tsepez, please review gpu_messages.h instead of
inferno.

[no sievers, 2015-09-23 19:32:33]

lgtm, thanks!

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
File content/common/gpu/gpu_channel.cc (right):

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
content/common/gpu/gpu_channel.cc:917: if (stub->decoder())
should we also check stub->decoder()->WasContextLost() for the corner case where
this context was lost (and it could be some webgl context that is just not
getting recreated)?

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
File content/common/gpu/gpu_channel_manager.cc (right):

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
content/common/gpu/gpu_channel_manager.cc:34: // Amount of time the GPU is
allowed to idle before it powers down.
nit: while you're at it, can you rephrase it since it's really an assumption
about the GPU/driver power management. maybe 'Amount of time we expect the GPU
to stay powered up without being used.'?

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
content/common/gpu/gpu_channel_manager.cc:316: TRACE_EVENT2("gpu",
"ImageTransportSurfaceAndroid::ScheduleWakeUp",
nit: ImageTransportSurface->GpuChannelManager

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/imag...
File content/common/gpu/image_transport_surface.cc (right):

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/imag...
content/common/gpu/image_transport_surface.cc:99: return false;
NOTREACHED() even?

[Tom Sepez, 2015-09-23 20:54:29]

OWNERS stamp LGTM.

[piman, 2015-09-23 22:31:33]

The CQ bit was checked by piman@chromium.org

[piman, 2015-09-23 22:31:34]

The patchset sent to the CQ was uploaded after l-g-t-m from tsepez@chromium.org,
sievers@chromium.org
Link to the patchset: https://codereview.chromium.org/1366473002/#ps40001
(title: "review comments")

[piman, 2015-09-23 22:32:32]

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
File content/common/gpu/gpu_channel.cc (right):

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
content/common/gpu/gpu_channel.cc:917: if (stub->decoder())
On 2015/09/23 19:32:33, sievers wrote:
> should we also check stub->decoder()->WasContextLost() for the corner case
where
> this context was lost (and it could be some webgl context that is just not
> getting recreated)?

Done.

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
File content/common/gpu/gpu_channel_manager.cc (right):

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
content/common/gpu/gpu_channel_manager.cc:34: // Amount of time the GPU is
allowed to idle before it powers down.
On 2015/09/23 19:32:33, sievers wrote:
> nit: while you're at it, can you rephrase it since it's really an assumption
> about the GPU/driver power management. maybe 'Amount of time we expect the GPU
> to stay powered up without being used.'?

Done.

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/gpu_...
content/common/gpu/gpu_channel_manager.cc:316: TRACE_EVENT2("gpu",
"ImageTransportSurfaceAndroid::ScheduleWakeUp",
On 2015/09/23 19:32:33, sievers wrote:
> nit: ImageTransportSurface->GpuChannelManager

Done.

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/imag...
File content/common/gpu/image_transport_surface.cc (right):

https://codereview.chromium.org/1366473002/diff/20001/content/common/gpu/imag...
content/common/gpu/image_transport_surface.cc:99: return false;
On 2015/09/23 19:32:33, sievers wrote:
> NOTREACHED() even?

Good idea, done.

[piman, 2015-09-23 22:36:20]

The CQ bit was checked by piman@chromium.org

[piman, 2015-09-23 22:36:20]

The patchset sent to the CQ was uploaded after l-g-t-m from tsepez@chromium.org,
sievers@chromium.org
Link to the patchset: https://codereview.chromium.org/1366473002/#ps60001
(title: "fix typo")

[commit-bot: I haz the power, 2015-09-24 01:02:33]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1366473002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1366473002/60001

[commit-bot: I haz the power, 2015-09-24 07:45:24]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-09-24 07:46:15]

Patchset 4 (id:??) landed as
https://crrev.com/fa4f254803e2952851400a35f9ec20605e9c3461
Cr-Commit-Position: refs/heads/master@{#350491}