Emulate gamma fix by making glyphs thicker or thinner

Emulate distance field gamma fix by making glyphs thicker or thinner

The idea here is that we determine the 0.5 crossover for each
row in the gamma table, then invert the mapping to determine
which point that maps to in the original range [-.65, 
.65]. That gives us a change in the apparent width of the 
glyph that closely corresponds to the change produced by the gamma fix.

BUG=skia:2933
Committed: https://skia.googlesource.com/skia/+/21deace8efc8e167d8626187ef0e6b4c241324b6

[jvanverth1, 2015-03-31 18:58:21]

jvanverth@google.com changed reviewers:
+ bungeman@google.com, egdaniel@google.com, joshualitt@google.com

[bungeman-skia, 2015-04-01 15:38:56]

https://codereview.chromium.org/1042373002/diff/60001/src/gpu/effects/GrDista...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/1042373002/diff/60001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:15: #include
"../utils/SkCamera.h"
Seems like you can remove this?

[jvanverth1, 2015-04-01 15:42:13]

https://codereview.chromium.org/1042373002/diff/60001/src/gpu/effects/GrDista...
File src/gpu/effects/GrDistanceFieldTextureEffect.cpp (right):

https://codereview.chromium.org/1042373002/diff/60001/src/gpu/effects/GrDista...
src/gpu/effects/GrDistanceFieldTextureEffect.cpp:15: #include
"../utils/SkCamera.h"
On 2015/04/01 15:38:56, bungeman1 wrote:
> Seems like you can remove this?

Done.

[egdaniel, 2015-04-01 15:55:13]

https://codereview.chromium.org/1042373002/diff/100001/src/gpu/effects/GrDist...
File src/gpu/effects/GrDistanceFieldTextureEffect.h (right):

https://codereview.chromium.org/1042373002/diff/100001/src/gpu/effects/GrDist...
src/gpu/effects/GrDistanceFieldTextureEffect.h:192: bool operator!=(const
WidthAdjust& wa) const {
cleaner to do not of equals? or vice versa

[jvanverth1, 2015-04-01 16:21:01]

https://codereview.chromium.org/1042373002/diff/100001/src/gpu/effects/GrDist...
File src/gpu/effects/GrDistanceFieldTextureEffect.h (right):

https://codereview.chromium.org/1042373002/diff/100001/src/gpu/effects/GrDist...
src/gpu/effects/GrDistanceFieldTextureEffect.h:192: bool operator!=(const
WidthAdjust& wa) const {
On 2015/04/01 15:55:13, egdaniel wrote:
> cleaner to do not of equals? or vice versa

Done.

[bungeman-skia, 2015-04-01 17:38:23]

I'm mostly writing this as an explanation to myself... maybe this will help with
a general comment of what's going on.

The mask gamma hack is based off of guessing what the blend color is going to
be, and adjusting the mask so that when run through the linear blend will
produce the value closest to the desired result. However, in practice this means
that the 'adjusted' mask is just increasing or decreasing the coverage of the
mask depending on what it is thought it will blit against. For black (on assumed
white) this means that coverages are decreased (on a curve). For white (on
assumed black) this means that coverages are increased (on a a curve). At middle
(perceptual) gray (which could be blit against anything) the coverages remain
the same.

The idea here is that instead of determining the initial (real) coverage and
then adjusting that coverage, determine an adjusted coverage directly by
essentially manipulating the geometry (in this case, the distance field
gradient). So for black (on assumed white) this thins a bit; for white (on
assumed black) this fake bolds the geometry a bit. The diameter of this fake
bolding has a maximum of sqrt(2), but this is unobtainable unless using a huge
gamma. For sane values of gamma the diameter is more on the order of 0.5.

This has several implications:

For non-gray lcd smoothed text, each subpixel essentially is using a slightly
different geometry.

For black (on assumed white) this may not cover some pixels which were
previously covered; however those pixels would have been only slightly covered
and that slight coverage would have been decreased anyway. Also, some pixels
which were previously fully covered may no longer be fully covered.

For white (on assumed black) this may cover some pixels which weren't previously
covered at all.

[jvanverth1, 2015-04-01 19:14:15]

Updated to include an edited version of bungeman's comments, plus some other
clean up.

[bungeman-skia, 2015-04-01 19:21:57]

It's all approximations. If it's good enough, lgtm, at least for the idea.

[jvanverth1, 2015-04-01 19:23:51]

The CQ bit was checked by jvanverth@google.com

[commit-bot: I haz the power, 2015-04-01 19:24:02]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1042373002/140001

[commit-bot: I haz the power, 2015-04-01 19:43:56]

Committed patchset #8 (id:140001) as
https://skia.googlesource.com/skia/+/21deace8efc8e167d8626187ef0e6b4c241324b6