Fix SDF gradient calculation for non-uniform xforms

Fix SDF gradient calculation for non-uniform xforms

Also adds two non-uniform test cases to the dftext GM.

BUG=skia:3540
Committed: https://skia.googlesource.com/skia/+/354eba5cb61801130a84378356434d3cc0a4b71a

[jvanverth1, 2015-03-16 16:30:22]

jvanverth@google.com changed reviewers:
+ egdaniel@google.com, robertphillips@google.com

[robertphillips, 2015-03-16 16:56:44]

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:98: } else {
overlength ?

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:111: fsBuilder->codeAppend("}
else {");
Why not just call normalize here ?

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:384: // this gives us a smooth
step across approximately one fragment
*dx ?

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:386: } else {
overlength ?

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:402: } else {
ditch \n ?

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:671: } else {
overlength ?

[jvanverth1, 2015-03-16 17:55:37]

Addressed the comments and fixed some other things found after running on the
S3.

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:98: } else {
On 2015/03/16 16:56:44, robertphillips wrote:
> overlength ?

Done.

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:111: fsBuilder->codeAppend("}
else {");
On 2015/03/16 16:56:44, robertphillips wrote:
> Why not just call normalize here ?

I've already done part of the work that normalize would do -- I figured that
this would be faster.

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:384: // this gives us a smooth
step across approximately one fragment
On 2015/03/16 16:56:44, robertphillips wrote:
> *dx ?

Done.

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:386: } else {
On 2015/03/16 16:56:44, robertphillips wrote:
> overlength ?

Done.

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:402: } else {
On 2015/03/16 16:56:44, robertphillips wrote:
> ditch \n ?

Done.

https://codereview.chromium.org/1013773002/diff/40001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:671: } else {
On 2015/03/16 16:56:44, robertphillips wrote:
> overlength ?

Done.

[robertphillips, 2015-03-16 18:11:36]

lgtm

https://codereview.chromium.org/1013773002/diff/60001/src/gpu/effects/GrDista...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/1013773002/diff/60001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:101: // (which is the inverse
transform for this fragment) and take the length of the result.
Move this computation down to where it is used?

https://codereview.chromium.org/1013773002/diff/60001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:389: // (which is the inverse
transform for this fragment) and take the length of the result.
here too ?

[jvanverth1, 2015-03-16 18:17:41]

https://codereview.chromium.org/1013773002/diff/60001/src/gpu/effects/GrDista...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/1013773002/diff/60001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:101: // (which is the inverse
transform for this fragment) and take the length of the result.
On 2015/03/16 18:11:35, robertphillips wrote:
> Move this computation down to where it is used?

Done.

https://codereview.chromium.org/1013773002/diff/60001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:389: // (which is the inverse
transform for this fragment) and take the length of the result.
On 2015/03/16 18:11:35, robertphillips wrote:
> here too ?

Done.

[jvanverth1, 2015-03-16 18:17:47]

The CQ bit was checked by jvanverth@google.com

[jvanverth1, 2015-03-16 18:17:47]

The patchset sent to the CQ was uploaded after l-g-t-m from
robertphillips@google.com
Link to the patchset: https://codereview.chromium.org/1013773002/#ps80001
(title: "Move Jdx & Jdy calcs down to where they're used")

[commit-bot: I haz the power, 2015-03-16 18:17:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1013773002/80001

[commit-bot: I haz the power, 2015-03-16 18:32:54]

Committed patchset #5 (id:80001) as
https://skia.googlesource.com/skia/+/354eba5cb61801130a84378356434d3cc0a4b71a