Distance field fixes for Android

Distance field fixes for Android
- Expand glyph size by 2 on each side to compensate for bilerp lookup
- Correct for Adreno tendency to drop entire tile if any pixel has divide-by-0
- Fix blurriness on Adreno by using uv coords to compute gradient instead
  of st coords
- Add faster version for uniform scale

BUG=skia:2173
Committed: http://code.google.com/p/skia/source/detail?r=13955

Committed: http://code.google.com/p/skia/source/detail?r=13958

[bsalomon, 2014-03-26 13:18:58]

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:78: #ifdef SK_BUILD_FOR_ANDROID
Can we make this a runtime check based on Vendor or Renderer? Branches are
really expensive on some of the mobile GPUs and we shouldn't pay the penalty on
all devices if we don't have to.

Actually, I think we should capture the drops-tile-on-divide-by-zero in GrGLCaps
and query that here. It has come up before and likely will come up again. If we
ever refine the detection mechanism then all effects that care will benefit.

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
File src/gpu/effects/GrDistanceFieldTextureEffect.h (right):

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
src/gpu/effects/GrDistanceFieldTextureEffect.h:24: static GrEffectRef*
Create(GrTexture* tex, const GrTextureParams& p, const SkISize& s,
It seems a little weird that the public function has short param names and the
private cons has more explicit names.

What is the texture size for? Is it different than texture->width(),
texture->height()?

[jvanverth1, 2014-03-26 16:24:00]

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:78: #ifdef SK_BUILD_FOR_ANDROID
On 2014/03/26 13:18:58, bsalomon wrote:
> Can we make this a runtime check based on Vendor or Renderer? Branches are
> really expensive on some of the mobile GPUs and we shouldn't pay the penalty
on
> all devices if we don't have to.
> 
> Actually, I think we should capture the drops-tile-on-divide-by-zero in
GrGLCaps
> and query that here. It has come up before and likely will come up again. If
we
> ever refine the detection mechanism then all effects that care will benefit.

Done.

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
File src/gpu/effects/GrDistanceFieldTextureEffect.h (right):

https://codereview.chromium.org/205343008/diff/20001/src/gpu/effects/GrDistan...
src/gpu/effects/GrDistanceFieldTextureEffect.h:24: static GrEffectRef*
Create(GrTexture* tex, const GrTextureParams& p, const SkISize& s,
On 2014/03/26 13:18:58, bsalomon wrote:
> It seems a little weird that the public function has short param names and the
> private cons has more explicit names.
> 
> What is the texture size for? Is it different than texture->width(),
> texture->height()?

I've removed it -- it's not necessary. I also added code to key off of uniform
scale/nonUniform scale.

[bsalomon, 2014-03-26 17:27:46]

lgtm

[jvanverth1, 2014-03-26 17:44:25]

The CQ bit was checked by jvanverth@google.com

[commit-bot: I haz the power, 2014-03-26 17:44:29]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/jvanverth@google.com/205343008/100001

[commit-bot: I haz the power, 2014-03-26 18:03:39]

Change committed as 13955

[jvanverth1, 2014-03-26 18:20:54]

A revert of this CL has been created in
https://codereview.chromium.org/212953008/ by jvanverth@google.com.

The reason for reverting is: compile error at line 110 in
GrDistanceFieldTextureEffect.cpp.

[jvanverth1, 2014-03-26 19:13:56]

The CQ bit was checked by jvanverth@google.com

[commit-bot: I haz the power, 2014-03-26 19:14:04]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/jvanverth@google.com/205343008/120001

[commit-bot: I haz the power, 2014-03-26 19:49:20]

Change committed as 13958