Adding optimization to avoid image copy in SkSurface copy on write when content is discardable

Adding optimization to avoid image copy in SkSurface copy on write when content is discardable

This patch also adds code to SkDeferredCanvas to trigger the optimization.

TEST=DeferredSurfaceCopy bench, Surface unit test

Committed: http://code.google.com/p/skia/source/detail?r=8797

[Justin Novosad, 2013-04-17 14:38:04]

PTAL

The DeferredSurfaceCopy micro benchmark validates the performance gain that is
achieved by discarding discardable contents:
CPU: 10.1ms -> 8.1ms
GPU: 6.7ms -> 3.9ms

[reed1, 2013-04-18 14:48:33]

I like the idea of the new parameter, but find a boolean to be too opaque

https://codereview.chromium.org/14063015/diff/6002/include/core/SkSurface.h
File include/core/SkSurface.h (right):

https://codereview.chromium.org/14063015/diff/6002/include/core/SkSurface.h#n...
include/core/SkSurface.h:86: */
Perhaps an enum or bit-field, so at the call-site we don't have to
guess/remember what true/false mean?

https://codereview.chromium.org/14063015/diff/6002/src/image/SkSurface_Base.h
File src/image/SkSurface_Base.h (right):

https://codereview.chromium.org/14063015/diff/6002/src/image/SkSurface_Base.h...
src/image/SkSurface_Base.h:52: virtual void onCopyOnWrite(bool
canDiscardContents) = 0;
Lets use an enum or bit-flags for this and aboutToDraw(). Seeing a bool at the
call-sites is confusing

[Justin Novosad, 2013-04-18 18:55:36]

Done.

New Patch

[reed1, 2013-04-18 19:27:09]

I think we should use the new enum in the copyonwrite methods too.

Standard enum convention is to use the enum as a suffix in each value.

e.g.

kRetain_ContentChangeMode,
kDiscard_ContentChangeMode

https://codereview.chromium.org/14063015/diff/13001/src/image/SkSurface_Base.h
File src/image/SkSurface_Base.h (right):

https://codereview.chromium.org/14063015/diff/13001/src/image/SkSurface_Base....
src/image/SkSurface_Base.h:52: virtual void onCopyOnWrite(bool
canDiscardContents) = 0;
Seems like we should use the same enum here...

[Justin Novosad, 2013-04-18 19:35:49]

Whoops. I was counting on compile errors to make me propagate the enum to all
the right places, but I guess enums can cast implicitly to bool...

New Patch.

[reed1, 2013-04-19 12:09:17]

lgtm

[commit-bot: I haz the power, 2013-04-19 17:58:42]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/junov@chromium.org/14063015/21001

[commit-bot: I haz the power, 2013-04-21 20:56:46]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/junov@chromium.org/14063015/21001

[commit-bot: I haz the power, 2013-04-22 14:28:05]

Change committed as 8797