Attempt to somewhat simplify GrContext::readSurfacePixels interaction with GrGpu.

Attempt to somewhat simplify GrContext::readSurfacePixels interaction with GrGpu.

Committed: https://skia.googlesource.com/skia/+/398260262f8508687614ec496a23b8152bcdc967

[bsalomon, 2015-07-22 21:41:17]

bsalomon@google.com changed reviewers:
+ robertphillips@google.com

[bsalomon, 2015-07-22 21:41:18]

Even though this is still very complex, I'm hoping it will make it easier to add
additional cases that should trigger draws (srg<->linear) without causing too
many minds to melt.

[robertphillips, 2015-07-23 13:04:23]

lgtm

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/GrGpu.h
File src/gpu/GrGpu.h (right):

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/GrGpu.h#newcode158
src/gpu/GrGpu.h:158: kNoDraw_DrawPreference,
space between 'client' and 'would' ?

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/GrGpu.h#newcode167
src/gpu/GrGpu.h:167: kGpuPrefersDraw_DrawPreference,
Can we also disallow kRequireDraw as input?

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/gl/GrGLGpu.cpp
File src/gpu/gl/GrGLGpu.cpp (right):

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/gl/GrGLGpu.cpp#...
src/gpu/gl/GrGLGpu.cpp:1725: 
Mv this to the top ?

[bsalomon, 2015-07-23 13:43:59]

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/GrGpu.h
File src/gpu/GrGpu.h (right):

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/GrGpu.h#newcode158
src/gpu/GrGpu.h:158: kNoDraw_DrawPreference,
On 2015/07/23 13:04:23, robertphillips wrote:
> space between 'client' and 'would' ?

Done.

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/GrGpu.h#newcode167
src/gpu/GrGpu.h:167: kGpuPrefersDraw_DrawPreference,
On 2015/07/23 13:04:23, robertphillips wrote:
> Can we also disallow kRequireDraw as input?

It isn't currently being used by the caller, but I can imagine how it would be
(we wanted to do a data transformation before reading that doesn't have a sw
implementation). If restricting it would simplify the code at all I'd be for it,
but I think it'd be an added assertion. The restriction on kGpu was just because
it's nonsensical for the caller to declare it, other than the assertion it
doesn't cause a correctness problem.

The reason I split kCallerPrefers and kGpuPrefers was so that when the unpremul
step couldn't be done on the GPU, the caller can know whether there was any
additional reason to do the draw or that it should skip the draw and do a
straight read.

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/gl/GrGLGpu.cpp
File src/gpu/gl/GrGLGpu.cpp (right):

https://codereview.chromium.org/1255483005/diff/20001/src/gpu/gl/GrGLGpu.cpp#...
src/gpu/gl/GrGLGpu.cpp:1725: 
On 2015/07/23 13:04:23, robertphillips wrote:
> Mv this to the top ?

Done.

[bsalomon, 2015-07-23 13:44:05]

The CQ bit was checked by bsalomon@google.com

[bsalomon, 2015-07-23 13:44:06]

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/1255483005/#ps40001
(title: "Address comments")

[commit-bot: I haz the power, 2015-07-23 13:44:19]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1255483005/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1255483005/40001

[commit-bot: I haz the power, 2015-07-23 15:07:24]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/398260262f8508687614ec496a23b8152bcdc967