Attach GrOptDrawState into shader building pipeline

Attach GrOptDrawState into shader building pipeline

The OptDrawState is now used for creating the actual gl shader. Current
optimizations dones in GrOptDrawState include:
All blend optimizations
Constant color/coverage stage optimizations

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/ee6206572b42fec11f83ad0c1e6d435903640518

Committed: https://skia.googlesource.com/skia/+/170f90b4576f291879371ecd6ae4bc2b1d85c64a

[egdaniel, 2014-08-26 18:01:43]

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

[egdaniel, 2014-09-08 19:13:24]

rebased up to recent gpu changes

[egdaniel, 2014-09-15 13:53:41]

ping. I've added removing the blendOpt cache here since with this CL getBlendOpt
is only called in createOptState and couldApplyCoverage. However the later sets
the force coverage flag so the cache wasn't used anyways

[bsalomon, 2014-09-15 14:04:35]

https://codereview.chromium.org/504203004/diff/100001/src/gpu/GrOptDrawState.h
File src/gpu/GrOptDrawState.h (right):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/GrOptDrawState....
src/gpu/GrOptDrawState.h:86: friend GrOptDrawState*
GrDrawState::createOptState() const;
Is it possible that GrOptState's cons just take a const GrDrawState& and be
public and all optflags/optblendcoeff stuff happens under GrOptState()?

https://codereview.chromium.org/504203004/diff/100001/src/gpu/GrRODrawState.h
File src/gpu/GrRODrawState.h (right):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/GrRODrawState.h...
src/gpu/GrRODrawState.h:357: enum BlendOptFlags {
w00t

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGLProgramD...
File src/gpu/gl/GrGLProgramDesc.cpp (left):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGLProgramD...
src/gpu/gl/GrGLProgramDesc.cpp:66: bool skipCoverage = SkToBool(blendOpts &
GrDrawState::kEmitTransBlack_BlendOptFlag);
love seeing this code go away.

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGpuGL_prog...
File src/gpu/gl/GrGpuGL_program.cpp (right):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGpuGL_prog...
src/gpu/gl/GrGpuGL_program.cpp:209: optState.unref();
This seems really dangerous... it is assuming an implementation detail about
GrDrawState, that it keeps a ref to GrODS and that it won't drop it before this
function ends.

Howabout:

SkAutoTUnref optState(this->getDrawState()->createOptState());

and adjust bewlow with ->/* as needed.

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGpuGL_prog...
src/gpu/gl/GrGpuGL_program.cpp:299: optState.unref();
ditto

[egdaniel, 2014-09-15 17:51:35]

https://codereview.chromium.org/504203004/diff/100001/src/gpu/GrOptDrawState.h
File src/gpu/GrOptDrawState.h (right):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/GrOptDrawState....
src/gpu/GrOptDrawState.h:86: friend GrOptDrawState*
GrDrawState::createOptState() const;
On 2014/09/15 14:04:34, bsalomon wrote:
> Is it possible that GrOptState's cons just take a const GrDrawState& and be
> public and all optflags/optblendcoeff stuff happens under GrOptState()?

In order to keep all the blendOpt flag stuff private, for now since some
friending will have to happen between GrOptState and GrDrawState. Once we are
able to get all the blend optimizations in GrOptDrawState (or a new
GrXferEffect!!) then all this can be cleaned up.

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGLProgramD...
File src/gpu/gl/GrGLProgramDesc.cpp (left):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGLProgramD...
src/gpu/gl/GrGLProgramDesc.cpp:66: bool skipCoverage = SkToBool(blendOpts &
GrDrawState::kEmitTransBlack_BlendOptFlag);
On 2014/09/15 14:04:34, bsalomon wrote:
> love seeing this code go away.

more code going away soon!

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGpuGL_prog...
File src/gpu/gl/GrGpuGL_program.cpp (right):

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGpuGL_prog...
src/gpu/gl/GrGpuGL_program.cpp:209: optState.unref();
On 2014/09/15 14:04:34, bsalomon wrote:
> This seems really dangerous... it is assuming an implementation detail about
> GrDrawState, that it keeps a ref to GrODS and that it won't drop it before
this
> function ends.
> 
> Howabout:
> 
> SkAutoTUnref optState(this->getDrawState()->createOptState());
> 
> and adjust bewlow with ->/* as needed.

Done.

https://codereview.chromium.org/504203004/diff/100001/src/gpu/gl/GrGpuGL_prog...
src/gpu/gl/GrGpuGL_program.cpp:299: optState.unref();
On 2014/09/15 14:04:34, bsalomon wrote:
> ditto

Done.

[bsalomon, 2014-09-15 18:44:29]

lgtm

[egdaniel, 2014-09-15 19:22:22]

egdaniel@google.com changed reviewers:
+ joshualitt@google.com

[egdaniel, 2014-09-15 19:22:23]

Adding Joshua to quickly confirm my removal of the need to map vertex attrib
indices

[egdaniel, 2014-09-15 21:13:10]

Patchset #10 (id:180001) has been deleted

[egdaniel, 2014-09-16 13:24:17]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2014-09-16 13:25:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patchset/504203004/220001

[commit-bot: I haz the power, 2014-09-16 13:34:47]

Committed patchset #11 (id:220001) as ee6206572b42fec11f83ad0c1e6d435903640518

[egdaniel, 2014-09-16 13:57:46]

A revert of this CL (patchset #11 id:220001) has been created in
https://codereview.chromium.org/560443004/ by egdaniel@google.com.

The reason for reverting is: Failing some msaa gm's.

[egdaniel, 2014-09-16 19:45:40]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2014-09-16 19:46:22]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patchset/504203004/280001

[commit-bot: I haz the power, 2014-09-16 19:54:45]

Committed patchset #14 (id:280001) as 170f90b4576f291879371ecd6ae4bc2b1d85c64a