Calculate stage stats in GrOptDrawState

Calculate stage stats in GrOptDrawState

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/a7dc0a83f0b59424a61e01bf3435cd7957e84e35

[egdaniel, 2014-09-05 14:20:31]

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

[egdaniel, 2014-09-05 14:20:31]

This is part 4 of add a OptDrawState. Follows
https://codereview.chromium.org/504203004/

I need to know info like readsDst to pull out a lot of the dual source blending
code in a follow up cl.

[bsalomon, 2014-09-05 15:27:55]

https://codereview.chromium.org/545693004/diff/1/src/gpu/GrOptDrawState.cpp
File src/gpu/GrOptDrawState.cpp (right):

https://codereview.chromium.org/545693004/diff/1/src/gpu/GrOptDrawState.cpp#n...
src/gpu/GrOptDrawState.cpp:225: if (stage.getEffect()->requiresVertexShader()) {
this is going away... reqVS will be equivalent to this->hasGeometryProcessor().

https://codereview.chromium.org/545693004/diff/1/src/gpu/GrOptDrawState.h
File src/gpu/GrOptDrawState.h (right):

https://codereview.chromium.org/545693004/diff/1/src/gpu/GrOptDrawState.h#new...
src/gpu/GrOptDrawState.h:91: // These flags give general info on the effect
stages that are used when building the programs
general info -> aggregated info?

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.cpp
File src/gpu/gl/GrGLProgramDesc.cpp (right):

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.c...
src/gpu/gl/GrGLProgramDesc.cpp:62: bool readsDst = optState.readsDst();
remove the vars?

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.c...
src/gpu/gl/GrGLProgramDesc.cpp:66: bool requiresVertexShader =
I wonder whether this should just be an assert... We better not be doing nvpr
and have a GP.

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.h
File src/gpu/gl/GrGLProgramDesc.h (right):

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.h...
src/gpu/gl/GrGLProgramDesc.h:197: static bool GetEffectKey(const GrEffectStage&
stage, const GrGLCaps& caps,
yay

[egdaniel, 2014-09-08 19:19:49]

https://codereview.chromium.org/545693004/diff/1/src/gpu/GrOptDrawState.h
File src/gpu/GrOptDrawState.h (right):

https://codereview.chromium.org/545693004/diff/1/src/gpu/GrOptDrawState.h#new...
src/gpu/GrOptDrawState.h:91: // These flags give general info on the effect
stages that are used when building the programs
On 2014/09/05 15:27:55, bsalomon wrote:
> general info -> aggregated info?

Done.

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.cpp
File src/gpu/gl/GrGLProgramDesc.cpp (right):

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.c...
src/gpu/gl/GrGLProgramDesc.cpp:62: bool readsDst = optState.readsDst();
On 2014/09/05 15:27:55, bsalomon wrote:
> remove the vars?
removed

https://codereview.chromium.org/545693004/diff/1/src/gpu/gl/GrGLProgramDesc.c...
src/gpu/gl/GrGLProgramDesc.cpp:66: bool requiresVertexShader =
Added assert below

On 2014/09/05 15:27:55, bsalomon wrote:
> I wonder whether this should just be an assert... We better not be doing nvpr
> and have a GP.

[egdaniel, 2014-09-16 21:04:03]

updated

[bsalomon, 2014-09-16 21:54:20]

https://codereview.chromium.org/545693004/diff/80001/src/gpu/gl/GrGLProgramDe...
File src/gpu/gl/GrGLProgramDesc.cpp (right):

https://codereview.chromium.org/545693004/diff/80001/src/gpu/gl/GrGLProgramDe...
src/gpu/gl/GrGLProgramDesc.cpp:141: // We will only use vertex shader if we have
a geometry processor and we should not have a
use->require ?

https://codereview.chromium.org/545693004/diff/80001/src/gpu/gl/GrGLProgramDe...
src/gpu/gl/GrGLProgramDesc.cpp:145:
SkASSERT(!GrGpu::IsPathRenderingDrawType(drawType) ||
!optState.requiresVertexShader());
can requriesLocalCoordAttrib be part of optState.requiresVertexShader()'s
result? Also wouldn't vertex color or coverage also cause us to require a VS?

Maybe in the future we say a path draw causes us to ignore vertex attribs and
GPs. The ODS could just drop them when computing an ODS from a DS for a path
draw. That might also mean a different, smaller desc for the shader since we
don't require a VS.

[egdaniel, 2014-09-17 15:00:14]

https://codereview.chromium.org/545693004/diff/80001/src/gpu/gl/GrGLProgramDe...
File src/gpu/gl/GrGLProgramDesc.cpp (right):

https://codereview.chromium.org/545693004/diff/80001/src/gpu/gl/GrGLProgramDe...
src/gpu/gl/GrGLProgramDesc.cpp:141: // We will only use vertex shader if we have
a geometry processor and we should not have a
On 2014/09/16 21:54:20, bsalomon wrote:
> use->require ?

Done.

https://codereview.chromium.org/545693004/diff/80001/src/gpu/gl/GrGLProgramDe...
src/gpu/gl/GrGLProgramDesc.cpp:145:
SkASSERT(!GrGpu::IsPathRenderingDrawType(drawType) ||
!optState.requiresVertexShader());
On 2014/09/16 21:54:20, bsalomon wrote:
> can requriesLocalCoordAttrib be part of optState.requiresVertexShader()'s
result?

Moved the checking of requires local coord attrib to ODS

> Also wouldn't vertex color or coverage also cause us to require a VS?

requiresVS now is just checking if fVACount > 1

> 
> Maybe in the future we say a path draw causes us to ignore vertex attribs and
> GPs. The ODS could just drop them when computing an ODS from a DS for a path
> draw. That might also mean a different, smaller desc for the shader since we
> don't require a VS.

[bsalomon, 2014-09-17 15:11:26]

lgtm

https://codereview.chromium.org/545693004/diff/120001/src/gpu/GrOptDrawState.cpp
File src/gpu/GrOptDrawState.cpp (right):

https://codereview.chromium.org/545693004/diff/120001/src/gpu/GrOptDrawState....
src/gpu/GrOptDrawState.cpp:197: // will requires a vertexShader
requires->require

[egdaniel, 2014-09-17 15:16:35]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2014-09-17 15:16:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patchset/545693004/140001

[commit-bot: I haz the power, 2014-09-17 15:25:09]

Committed patchset #8 (id:140001) as a7dc0a83f0b59424a61e01bf3435cd7957e84e35