Aura/ÜC: Drop frames on background tabs

Aura/ÜC: Drop frames on background tabs

We want to make sure we release memory on background tabs to avoid scaling
memory usage with the number of open tabs.
This CL hooks into the same memory manager as for software frames (extracted
into RendererFrameManager), which mainly tries to keep a maximum of 2 to 5
(depending on available memory) tabs with an allocated frame, for faster tab
switching. The other ones will discard their frames and return the resources
to the renderer.

BUG=311353
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=231473

[piman, 2013-10-25 23:34:14]

danakj: please review
jbauman/ccameron: FYI for the software changes, though feel free to chime in.

[danakj, 2013-10-28 21:39:20]

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/render_widget_host_view_aura.cc (right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura.cc:1497:
window_->layer()->SetShowPaintedContent();
Should this just call EvictDelegatedFrame() rather than duplicating it? Or do
you see value in that?

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/render_widget_host_view_aura.h (right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura.h:400: class
DelegatedFrameEvictionObserver : public RendererFrameManagerClient {
What if this was a class in a standalone file in content/browser/aura/? Would
that be reasonable?

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/render_widget_host_view_aura_unittest.cc
(right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura_unittest.cc:912:
TEST_F(RenderWidgetHostViewAuraTest, DiscardDelegatedFrames) {
Thanks for test!

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura_unittest.cc:921: new
RenderWidgetHostImpl* [renderer_count]);
the extra spaces here seem weird? what would git cl format do?

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura_unittest.cc:978: //
Make all renderers but [0] visible as swap a frame on them, keep [0]
s/as/and/

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/renderer_frame_manager.cc (right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/renderer_frame_manager.cc:12:
////////////////////////////////////////////////////////////////////////////////
What's up with these comment things? Is this a c/b/renderer_host/ style?

[piman, 2013-10-28 22:08:38]

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/render_widget_host_view_aura.cc (right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura.cc:1497:
window_->layer()->SetShowPaintedContent();
On 2013/10/28 21:39:21, danakj wrote:
> Should this just call EvictDelegatedFrame() rather than duplicating it? Or do
> you see value in that?

Done, and above as well.

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/render_widget_host_view_aura.h (right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura.h:400: class
DelegatedFrameEvictionObserver : public RendererFrameManagerClient {
On 2013/10/28 21:39:21, danakj wrote:
> What if this was a class in a standalone file in content/browser/aura/? Would
> that be reasonable?

Done, though I left in c/b/renderer_host, because it could be used in more
things than Aura, and depends on c/b/renderer_host things anyway.

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/render_widget_host_view_aura_unittest.cc
(right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura_unittest.cc:921: new
RenderWidgetHostImpl* [renderer_count]);
On 2013/10/28 21:39:21, danakj wrote:
> the extra spaces here seem weird? what would git cl format do?

git cl format did it!
I left as is to avoid noise.

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/render_widget_host_view_aura_unittest.cc:978: //
Make all renderers but [0] visible as swap a frame on them, keep [0]
On 2013/10/28 21:39:21, danakj wrote:
> s/as/and/

Done.

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
File content/browser/renderer_host/renderer_frame_manager.cc (right):

https://codereview.chromium.org/43193002/diff/40001/content/browser/renderer_...
content/browser/renderer_host/renderer_frame_manager.cc:12:
////////////////////////////////////////////////////////////////////////////////
On 2013/10/28 21:39:21, danakj wrote:
> What's up with these comment things? Is this a c/b/renderer_host/ style?

Not particularly, but I removed from here where it has no value.

[danakj, 2013-10-28 22:29:41]

LGTM

https://codereview.chromium.org/43193002/diff/240001/content/browser/renderer...
File content/browser/renderer_host/delegated_frame_evictor.h (right):

https://codereview.chromium.org/43193002/diff/240001/content/browser/renderer...
content/browser/renderer_host/delegated_frame_evictor.h:29: //
RendererFrameManagerClient implementation
nit: .

[piman, 2013-10-28 22:40:46]

https://codereview.chromium.org/43193002/diff/240001/content/browser/renderer...
File content/browser/renderer_host/delegated_frame_evictor.h (right):

https://codereview.chromium.org/43193002/diff/240001/content/browser/renderer...
content/browser/renderer_host/delegated_frame_evictor.h:29: //
RendererFrameManagerClient implementation
On 2013/10/28 22:29:41, danakj wrote:
> nit: .

Done.

[commit-bot: I haz the power, 2013-10-28 22:43:21]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/piman@chromium.org/43193002/300001

[commit-bot: I haz the power, 2013-10-28 23:05:12]

Failed to trigger a try job on win_x64_rel
HTTP Error 400: Bad Request

[commit-bot: I haz the power, 2013-10-28 23:08:10]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/piman@chromium.org/43193002/350001

[commit-bot: I haz the power, 2013-10-29 01:20:04]

Change committed as 231473