mac: Improve framerate on accelerated pages in HiDPI mode.

mac: Improve framerate on accelerated pages in HiDPI mode.

RenderWidgetHost asked RWHV for a backing store in view coordinates,
but RWHV matched in pixel coordinates. As a result, backing store
allocation was considered unsuccessful repeatedly every frame, and every
frame would retry until the frame timeout (50ms) was elapsed.

If the browser process is very busy, the fps meter tends to show NaN
fps: It renders FPS in the renderer, and since the renderer was almost
always doing nothing (since it waited on the browser), almost all frame
times were filtered out (very short frame times are filtered out), and
often there were 0 frame times to sample from for FPS computation.

BUG=132178
TEST=Open poster circle with FPS meter enabled. 60fps instead of NaN. 
TBR=avi

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=141675

[Nico, 2012-06-12 01:24:23]

The way I compute the compositing IO surface view unit size is a bit hacky at
the moment. I believe that will become cleaner once the runtime DPI switching
code is in.

[jbates, 2012-06-12 17:11:13]

LGTM with nit

https://chromiumcodereview.appspot.com/10533098/diff/1/content/browser/render...
File content/browser/renderer_host/compositing_iosurface_mac.h (right):

https://chromiumcodereview.appspot.com/10533098/diff/1/content/browser/render...
content/browser/renderer_host/compositing_iosurface_mac.h:57: const gfx::Size&
io_surface_size() const { return io_surface_size_; }
I think this should be renamed so that it's clear that it's not the pixel size.
Can it mention something about nsview coordinates? How about
nsview_io_surface_size() with a comment that it's the surface size mapped to
NSView coordinates.

[Nico, 2012-06-12 17:18:11]

done, thanks. avi: OWNERS

[Avi (use Gerrit), 2012-06-12 19:00:20]

owner rubberstamp lgtm