Progress on gamma-correctness in the GPU backend. Fixed conversion of color and profile type to pix…

Progress on gamma-correctness in the GPU backend. Fixed conversion of color and profile type to pixel config, which makes many things "just work".

Added (color=8888|f16|srgb) option to gpu configurations, along with gpuf16, gpusrgb, and anglesrgb predefined configs. Runs the gpu backend in gamma-correct mode (with either FP16 linear or sRGB 8888 frambuffers).


BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1746253002

Committed: https://skia.googlesource.com/skia/+/eef980270d3385fee340eb1633962fe3ba8b7132

[Brian Osman, 2016-02-29 20:12:34]

Description was changed from

==========
Progress on gamma-correctness in the GPU backend. Fixed conversion of color and
profile type to pixel config, which makes many things "just work".

Added (color=8888|f16|srgb) option to gpu configurations, along with gpuf16,
gpusrgb, and anglesrgb predefined configs. Runs the gpu backend in gamma-correct
mode (with either FP16 linear or sRGB 8888 frambuffers).


BUG=skia:
==========

to

==========
Progress on gamma-correctness in the GPU backend. Fixed conversion of color and
profile type to pixel config, which makes many things "just work".

Added (color=8888|f16|srgb) option to gpu configurations, along with gpuf16,
gpusrgb, and anglesrgb predefined configs. Runs the gpu backend in gamma-correct
mode (with either FP16 linear or sRGB 8888 frambuffers).


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[Brian Osman, 2016-02-29 20:17:42]

brianosman@google.com changed reviewers:
+ bsalomon@google.com, egdaniel@google.com, mtklein@google.com

[mtklein, 2016-02-29 20:23:15]

lgtm

[bsalomon, 2016-02-29 20:27:46]

https://codereview.chromium.org/1746253002/diff/20001/src/gpu/SkGr.cpp
File src/gpu/SkGr.cpp (right):

https://codereview.chromium.org/1746253002/diff/20001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:362: GrPixelConfig SkImageInfo2GrPixelConfig(SkColorType ct,
SkAlphaType, SkColorProfileType pt) {
Not necessarily in this change, but wonder if we should make this able to fail
(or return kUnknown) for ct + pt that don't map to a GrPC (e.g. 4444 + sRGB or
even BGRA + sRGB).

https://codereview.chromium.org/1746253002/diff/20001/tools/flags/SkCommonFla...
File tools/flags/SkCommonFlagsConfig.cpp (right):

https://codereview.chromium.org/1746253002/diff/20001/tools/flags/SkCommonFla...
tools/flags/SkCommonFlagsConfig.cpp:253: *outColorType = kN32_SkColorType;
Do we need to say kN32 or can we just say kRGBA?

[Brian Osman, 2016-02-29 20:34:49]

https://codereview.chromium.org/1746253002/diff/20001/tools/flags/SkCommonFla...
File tools/flags/SkCommonFlagsConfig.cpp (right):

https://codereview.chromium.org/1746253002/diff/20001/tools/flags/SkCommonFla...
tools/flags/SkCommonFlagsConfig.cpp:253: *outColorType = kN32_SkColorType;
On 2016/02/29 20:27:46, bsalomon wrote:
> Do we need to say kN32 or can we just say kRGBA?

Maybe? I was trying to match previous behavior (where we were always specifying
kN32 as the color type in the eventual call to make the ImageInfo). Switching
this would alter behavior on Windows (or anywhere else that kN32 is BGRA?), but
I suspect that RGBA is always a legal format.

[bsalomon, 2016-02-29 20:50:47]

On 2016/02/29 20:34:49, Brian Osman wrote:
>
https://codereview.chromium.org/1746253002/diff/20001/tools/flags/SkCommonFla...
> File tools/flags/SkCommonFlagsConfig.cpp (right):
> 
>
https://codereview.chromium.org/1746253002/diff/20001/tools/flags/SkCommonFla...
> tools/flags/SkCommonFlagsConfig.cpp:253: *outColorType = kN32_SkColorType;
> On 2016/02/29 20:27:46, bsalomon wrote:
> > Do we need to say kN32 or can we just say kRGBA?
> 
> Maybe? I was trying to match previous behavior (where we were always
specifying
> kN32 as the color type in the eventual call to make the ImageInfo). Switching
> this would alter behavior on Windows (or anywhere else that kN32 is BGRA?),
but
> I suspect that RGBA is always a legal format.

That was my thinking. The only place where RGBA8 is not a legal format without
extension is ES2 (but neither is BGRA8). However, I think every ES2 device we
could care about has GL_OES_rgb8_rgba8 or GL_ARM_rgba8. I'd be OK with changing
the behavior. I don't think there was a good reason for using kN32 here and by
using RGBA we don't have to lie about mapping
kBGRA_8888_ColorType/kSRGB_ProfileType to ksRGBA_8888_GrPixelConfig.

[Brian Osman, 2016-02-29 21:21:39]

Switched to always test RGBA rather than N32 (in both 8888 and srgb sub-configs
of GPU).

[Brian Osman, 2016-02-29 21:26:34]

The CQ bit was checked by brianosman@google.com

[Brian Osman, 2016-02-29 21:26:35]

The patchset sent to the CQ was uploaded after l-g-t-m from mtklein@google.com
Link to the patchset: https://codereview.chromium.org/1746253002/#ps40001
(title: "Switch to RGBA rather than N32 when testing the GPU backend.")

[commit-bot: I haz the power, 2016-02-29 21:26:45]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1746253002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1746253002/40001

[commit-bot: I haz the power, 2016-02-29 21:39:12]

Description was changed from

==========
Progress on gamma-correctness in the GPU backend. Fixed conversion of color and
profile type to pixel config, which makes many things "just work".

Added (color=8888|f16|srgb) option to gpu configurations, along with gpuf16,
gpusrgb, and anglesrgb predefined configs. Runs the gpu backend in gamma-correct
mode (with either FP16 linear or sRGB 8888 frambuffers).


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
Progress on gamma-correctness in the GPU backend. Fixed conversion of color and
profile type to pixel config, which makes many things "just work".

Added (color=8888|f16|srgb) option to gpu configurations, along with gpuf16,
gpusrgb, and anglesrgb predefined configs. Runs the gpu backend in gamma-correct
mode (with either FP16 linear or sRGB 8888 frambuffers).


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/eef980270d3385fee340eb1633962fe3ba8b7132
==========

[commit-bot: I haz the power, 2016-02-29 21:39:12]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/eef980270d3385fee340eb1633962fe3ba8b7132

[Brian Osman, 2016-02-29 21:55:23]

A revert of this CL (patchset #3 id:40001) has been created in
https://codereview.chromium.org/1748823002/ by brianosman@google.com.

The reason for reverting is: Fixing the build..

[Brian Osman, 2016-03-01 14:55:22]

Trying again. Forcing kRGBA led to DM failures...