Added profiling stubs to SkPicturePlayback

Originally these stubs passed in the command number but that proved problematic since the clips allow some commands to be skipped. This CL passes the stream offset to the stubs and it is up to the derived classes to figure it out.

[reed1, 2012-11-13 16:17:04]

1. What are some sample subclasses and their impls of pre and post?
2. What is the % overhead for a simple picture to call empty versions of these?
3. if #2 > epsilon, do we really need this on by default in all our builds?

[robertphillips, 2012-11-13 17:05:33]

1. What are some sample subclasses and their impls of pre and post?

The CL here (https://codereview.appspot.com/6817114/) contains the entire
change. In SkDebuggerGUI.cpp I create two SkPicturePlayback-derived classes
(SkOffsetPicturePlayback & SkTimedPicturePlayback). The former simply overrides
preDraw to cache the SkPicture's offsets. The latter overrides both preDraw and
postDraw to time each individual command.

2. What is the % overhead for a simple picture to call empty versions of these?

On Linux:

With the virtual functions:

bench_pictures: --repeat 500 --mode tile 256 256 ../../skp/simsimi.skp
running bench [1200 1859] simsimi.skp   tile_256x256: cmsecs =   4.30

Without the virtual functions:

bench_pictures: --repeat 500 --mode tile 256 256 ../../skp/simsimi.skp
running bench [1200 1859] simsimi.skp   tile_256x256: cmsecs =   4.34


3. if #2 > epsilon, do we really need this on by default in all our builds?

Ideally we would only have this turned on when someone wanted to profile in the
debugger. We could leave it turned off in the .gyp file and then a developer
would have to turn it on locally.

[DerekS, 2012-11-13 17:45:21]

https://codereview.appspot.com/6851044/diff/1/gyp/common_conditions.gypi
File gyp/common_conditions.gypi (right):

https://codereview.appspot.com/6851044/diff/1/gyp/common_conditions.gypi#newc...
gyp/common_conditions.gypi:8: 'SK_PICTURE_PROFILING_STUBS'
I'm all for making this only part of the debugger target.

[robertphillips, 2012-11-13 17:47:46]

https://codereview.appspot.com/6851044/diff/1/gyp/common_conditions.gypi
File gyp/common_conditions.gypi (right):

https://codereview.appspot.com/6851044/diff/1/gyp/common_conditions.gypi#newc...
gyp/common_conditions.gypi:8: 'SK_PICTURE_PROFILING_STUBS'
I don't think we can do that quite yet (i.e., have the debugger define it and
get a separate compilation tree just for the debugger). I was proposing
disabling it and requiring individual developers turn it on locally.

[DerekS, 2012-11-13 17:48:58]

On 2012/11/13 17:47:46, robertphillips wrote:
> https://codereview.appspot.com/6851044/diff/1/gyp/common_conditions.gypi
> File gyp/common_conditions.gypi (right):
> 
>
https://codereview.appspot.com/6851044/diff/1/gyp/common_conditions.gypi#newc...
> gyp/common_conditions.gypi:8: 'SK_PICTURE_PROFILING_STUBS'
> I don't think we can do that quite yet (i.e., have the debugger define it and
> get a separate compilation tree just for the debugger). I was proposing
> disabling it and requiring individual developers turn it on locally.

Gotcha.  I'm all for that too!

[robertphillips, 2012-11-13 17:57:32]

Updated w/o gyp change

[DerekS, 2012-11-13 18:11:00]

lgtm

[reed1, 2012-11-13 18:26:26]

lgtm

[robertphillips, 2012-11-13 18:28:27]

committed as r6397