Support mesa config in render_pictures.

Support mesa config in render_pictures.

R=epoger@google.com, robertphillips@google.com

Committed: https://code.google.com/p/skia/source/detail?r=12916

[robertphillips, 2014-01-06 15:57:59]

lgtm + 1 suggestion & 1 question

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderer.h
File tools/PictureRenderer.h (right):

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderer.h#newc...
tools/PictureRenderer.h:285: // fall through
Move to after angle for consistency?

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderingFlags.cpp
File tools/PictureRenderingFlags.cpp (right):

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderingFlags....
tools/PictureRenderingFlags.cpp:30: DEFINE_string(config, "8888",
"[8888|gpu|msaa4|msaa16|angle]: Use the corresponding config.");
What if we have both angle and mesa defined?

[epoger, 2014-01-06 16:01:46]

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderingFlags.cpp
File tools/PictureRenderingFlags.cpp (right):

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderingFlags....
tools/PictureRenderingFlags.cpp:30: DEFINE_string(config, "8888",
"[8888|gpu|msaa4|msaa16|angle]: Use the corresponding config.");
On 2014/01/06 15:57:59, robertphillips wrote:
> What if we have both angle and mesa defined?

Right.  I guess we need something like this:

SkString allowedConfigs("8888|gpu|msaa4|msaa16");
#if SK_ANGLE
allowedConfigs.append("|angle");
#endif
#if SK_MESA
allowedConfigs.append("|mesa");
#endif
...

[bsalomon, 2014-01-06 16:30:36]

Curious what the motivation for this is. cloud instances w/out gpus?

[epoger, 2014-01-06 16:45:11]

On 2014/01/06 16:30:36, bsalomon wrote:
> Curious what the motivation for this is. cloud instances w/out gpus?

reed wrote (in email, didn't show up here):
render_pictures en-mass
compare (fuzzily) cpu -vs- gpu en-mass

robertphillips wrote (in email, didn't show up here):
It is to compare the GPU vs. raster images across the large picture set:
http://skia-tree-status.appspot.com/skia-telemetry/skia_try

In other words, yes. :-)  Exercising some of the GPU code on cloud instances
without gpus.

(I would like to know how much of the GPU code mesa allows us to exercise on
GPU-less platforms.  If it's a majority, maybe we should try to go this way for
ever-more of our testing?)

[rmistry, 2014-01-06 18:03:55]

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderer.h
File tools/PictureRenderer.h (right):

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderer.h#newc...
tools/PictureRenderer.h:285: // fall through
On 2014/01/06 15:57:59, robertphillips wrote:
> Move to after angle for consistency?

Done.

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderingFlags.cpp
File tools/PictureRenderingFlags.cpp (right):

https://codereview.chromium.org/125083004/diff/1/tools/PictureRenderingFlags....
tools/PictureRenderingFlags.cpp:30: DEFINE_string(config, "8888",
"[8888|gpu|msaa4|msaa16|angle]: Use the corresponding config.");
On 2014/01/06 16:01:47, epoger wrote:
> On 2014/01/06 15:57:59, robertphillips wrote:
> > What if we have both angle and mesa defined?
> 
> Right.  I guess we need something like this:
> 
> SkString allowedConfigs("8888|gpu|msaa4|msaa16");
> #if SK_ANGLE
> allowedConfigs.append("|angle");
> #endif
> #if SK_MESA
> allowedConfigs.append("|mesa");
> #endif
> ...

Done with something similar to the above suggestion.

[epoger, 2014-01-06 18:27:41]

lgtm

https://codereview.chromium.org/125083004/diff/100001/tools/PictureRenderingF...
File tools/PictureRenderingFlags.cpp (right):

https://codereview.chromium.org/125083004/diff/100001/tools/PictureRenderingF...
tools/PictureRenderingFlags.cpp:34: // ANGLE assumes GPU
you don't need the "ANGLE assumes GPU" comment anymore; they are added to the
string separately now.

[rmistry, 2014-01-06 18:33:25]

https://codereview.chromium.org/125083004/diff/100001/tools/PictureRenderingF...
File tools/PictureRenderingFlags.cpp (right):

https://codereview.chromium.org/125083004/diff/100001/tools/PictureRenderingF...
tools/PictureRenderingFlags.cpp:34: // ANGLE assumes GPU
On 2014/01/06 18:27:41, epoger wrote:
> you don't need the "ANGLE assumes GPU" comment anymore; they are added to the
> string separately now.

Done.

[rmistry, 2014-01-06 18:37:28]

Committed patchset #4 manually as r12916 (presubmit successful).