Mac Overlays: Allow SolidColor and Tile quads to be candidates for overlays.

Mac Overlays: Allow SolidColor and Tile quads to be candidates for overlays.

This is needed for the AllOrNothing overlay strategy that we will be using on Mac.

BUG=533687
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

[Andre, 2015-09-30 16:34:22]

andresantoso@chromium.org changed reviewers:
+ ccameron@chromium.org

[Andre, 2015-09-30 16:34:22]

ccameron PTAL.

[ccameron, 2015-09-30 18:52:09]

I think this needs some plumbing first.

https://codereview.chromium.org/1380653003/diff/1/cc/output/overlay_candidate.cc
File cc/output/overlay_candidate.cc (right):

https://codereview.chromium.org/1380653003/diff/1/cc/output/overlay_candidate...
cc/output/overlay_candidate.cc:222: candidate);
We're also going to want to make sure that things that derive from
OverlayStrategyCommon don't attempt these quad types (because if they do,
they'll hit the O(n^2) list modification, along with unknown performance
characteristics, etc).

https://codereview.chromium.org/1380653003/diff/1/cc/output/overlay_candidate...
cc/output/overlay_candidate.cc:310: OverlayCandidate* candidate) {
We're going to need to plumb through a "allow_overlay" flag here, because not
all resources are overlay-able (e.g, for now, CPU raster is possible, but GPU
raster isn't).

[Andre, 2015-09-30 20:57:17]

https://chromiumcodereview.appspot.com/1380653003/diff/1/cc/output/overlay_ca...
File cc/output/overlay_candidate.cc (right):

https://chromiumcodereview.appspot.com/1380653003/diff/1/cc/output/overlay_ca...
cc/output/overlay_candidate.cc:310: OverlayCandidate* candidate) {
On 2015/09/30 18:52:09, ccameron wrote:
> We're going to need to plumb through a "allow_overlay" flag here, because not
> all resources are overlay-able (e.g, for now, CPU raster is possible, but GPU
> raster isn't). 

I'm not sure how to do this.
I think we want to TileDrawQuad to allow_overlay if its resource has a GLImage.
The tile quads are created in PictureLayerImpl::AppendQuads(). There it has
access to a TileDrawInfo and resource_id.
reveman, how do you think this should be done?

[Andre, 2015-09-30 21:15:15]

https://chromiumcodereview.appspot.com/1380653003/diff/1/cc/output/overlay_ca...
File cc/output/overlay_candidate.cc (right):

https://chromiumcodereview.appspot.com/1380653003/diff/1/cc/output/overlay_ca...
cc/output/overlay_candidate.cc:222: candidate);
On 2015/09/30 18:52:09, ccameron wrote:
> We're also going to want to make sure that things that derive from
> OverlayStrategyCommon don't attempt these quad types (because if they do,
> they'll hit the O(n^2) list modification, along with unknown performance
> characteristics, etc).
> 

It's currently using OverlayCandidate::IsAllowedQuad() to filter it out in
OverlayStrategCommon.
But will need a better way or name if we add allow_overlay to TileDrawQuad.

[Andre, 2015-10-02 18:10:15]

https://codereview.chromium.org/1380653003/diff/20001/cc/layers/picture_layer...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/1380653003/diff/20001/cc/layers/picture_layer...
cc/layers/picture_layer_impl.cc:307: layer_tree_impl()->resource_provider(),
draw_info.resource_id());
Here we want allow_overlay to be set if the resource has a GLImage.
I created ScopedReadLockGpuMemoryBuffer but not sure if it's right.
Seems like it should call ResourceProvider:;LockForRead(), but if I do that I
hit this exported_count assert.
https://code.google.com/p/chromium/codesearch#chromium/src/cc/resources/resou...

Daniele, WDYT?

[Andre, 2015-10-05 21:33:25]

Patchset #3 (id:40001) has been deleted

[Andre, 2015-10-09 20:27:53]

Patchset #6 (id:120001) has been deleted