Cache the return values of getBlendOpts in GrDrawState

Cache the return values of getBlendOpts in GrDrawState

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/9514d24fa512c9cd73548715ac36ad02f34151e8

[egdaniel, 2014-07-17 18:28:28]

Please take at look at where I invalidate the cache. Do you notice any obvious
places that are missing or places where I don't need to be doing it?

[bsalomon, 2014-07-17 18:41:03]

I think the cache has to respect whether forceCoverage was set or not. Is it
ever being passed as true? Maybe it can be killed.

Don't we need to invalidate in setColor? Also, AutoRestoreEffects::set when it
removes stages.

We should probably invalidate it in setBlendConstant(). We may not do any
optimizations based on that presently, but we easily could. It's low-frequency
enough that I'd rather just be safe.

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.cpp
File src/gpu/GrDrawState.cpp (right):

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.cpp#newc...
src/gpu/GrDrawState.cpp:320: srcCoeff = &bogusSrcCoeff;
Seems like we don't need this anymore. Also if none of the callers pass NULL for
srcCoeff or dstCoeff we could probably just require the params to be non-NULL
and do away with this entirely.

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.h
File src/gpu/GrDrawState.h (right):

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.h#newcod...
src/gpu/GrDrawState.h:948: fOptSrcBlend = kOne_GrBlendCoeff;
Do we even need to set the blend flags here?

[egdaniel, 2014-07-17 19:35:50]

Addressed the comments. If forceCoverage is set to true we will simply calculate
the BlendOpts and return the values without reading or altering the cache.

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.cpp
File src/gpu/GrDrawState.cpp (right):

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.cpp#newc...
src/gpu/GrDrawState.cpp:320: srcCoeff = &bogusSrcCoeff;
On 2014/07/17 18:41:03, bsalomon wrote:
> Seems like we don't need this anymore. Also if none of the callers pass NULL
for
> srcCoeff or dstCoeff we could probably just require the params to be non-NULL
> and do away with this entirely.

I've actually already removed this in a follow up patch for the cl, but I want
to keep the ability to pass in NULL for src and dst since I will be using this
for some of my follow up cl's where all I care about is the blend flags.

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.h
File src/gpu/GrDrawState.h (right):

https://codereview.chromium.org/404473007/diff/1/src/gpu/GrDrawState.h#newcod...
src/gpu/GrDrawState.h:948: fOptSrcBlend = kOne_GrBlendCoeff;
On 2014/07/17 18:41:03, bsalomon wrote:
> Do we even need to set the blend flags here?

By blend flags are you referring to the Src and Dst? Those we don't need, but we
definitely need to set fBlendOptFlags to be invalid.

[bsalomon, 2014-07-17 19:53:50]

lgtm

https://codereview.chromium.org/404473007/diff/20001/src/gpu/GrDrawState.h
File src/gpu/GrDrawState.h (right):

https://codereview.chromium.org/404473007/diff/20001/src/gpu/GrDrawState.h#ne...
src/gpu/GrDrawState.h:407: n = fDrawState->fCoverageStages.count() -
fCoverageEffectCnt;
int m =

if (n + m > 0)

?

[egdaniel, 2014-07-18 13:04:16]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2014-07-18 13:05:06]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/egdaniel@google.com/404473007/60001

[commit-bot: I haz the power, 2014-07-18 13:15:48]

Change committed as 9514d24fa512c9cd73548715ac36ad02f34151e8