Use dst copies in porter duffer XP to correctly render certain blends.

Use dst copies in porter duffer XP to correctly render certain blends.

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/997c6358d94e188b1a7b89a4f86e24cbe0f5a164

Committed: https://skia.googlesource.com/skia/+/3ad6570e94965d54871eaea8b314f560b82ca792

[egdaniel, 2015-02-11 14:57:10]

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

[egdaniel, 2015-02-11 14:57:10]

This cl is dependent on https://codereview.chromium.org/912413002/ landing first

[bsalomon, 2015-02-11 15:53:31]

lgtm, some comments/qs inline

https://codereview.chromium.org/914003003/diff/40001/src/gpu/effects/GrPorter...
File src/gpu/effects/GrPorterDuffXferProcessor.cpp (right):

https://codereview.chromium.org/914003003/diff/40001/src/gpu/effects/GrPorter...
src/gpu/effects/GrPorterDuffXferProcessor.cpp:210: if (xp.hasSecondaryOutput())
{
can we skip the vcall/extra branch here since secondaryOutputType() has a kNone?
(maybe assert for consistency)

https://codereview.chromium.org/914003003/diff/40001/src/gpu/effects/GrPorter...
src/gpu/effects/GrPorterDuffXferProcessor.cpp:255: bool didAppend =
append_porterduff_blend(fsBuilder, xp.getSrcBlend(),
maybe append_porterduff_term?

https://codereview.chromium.org/914003003/diff/40001/src/gpu/effects/GrPorter...
src/gpu/effects/GrPorterDuffXferProcessor.cpp:264: fsBuilder->codeAppendf("%s =
%s * colorBlend + (vec4(1.0) - %s) * %s;",
do we know if cov is 4 chan here? can we treat it as scalar?

[egdaniel, 2015-02-13 14:19:14]

https://codereview.chromium.org/914003003/diff/40001/src/gpu/effects/GrPorter...
File src/gpu/effects/GrPorterDuffXferProcessor.cpp (right):

https://codereview.chromium.org/914003003/diff/40001/src/gpu/effects/GrPorter...
src/gpu/effects/GrPorterDuffXferProcessor.cpp:264: fsBuilder->codeAppendf("%s =
%s * colorBlend + (vec4(1.0) - %s) * %s;",
On 2015/02/11 15:53:31, bsalomon wrote:
> do we know if cov is 4 chan here? can we treat it as scalar?

In our current system we know that cov will be 4 chan here since we force it to
4 chan at start of pipeline. We still have the idea (though I feel low priority
right now) to be able to pass down scalars through coverage and only go to 4 if
the shader itself specifically requires.

[egdaniel, 2015-02-13 15:36:46]

New patchsets have been uploaded after l-g-t-m from bsalomon@google.com

[egdaniel, 2015-02-13 18:15:36]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2015-02-13 18:16:21]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/914003003/60001

[commit-bot: I haz the power, 2015-02-13 18:23:23]

Committed patchset #4 (id:60001) as
https://skia.googlesource.com/skia/+/997c6358d94e188b1a7b89a4f86e24cbe0f5a164

[egdaniel, 2015-02-13 19:07:04]

A revert of this CL (patchset #4 id:60001) has been created in
https://codereview.chromium.org/923153003/ by egdaniel@google.com.

The reason for reverting is: Failing GLProgramTest passing in stupid coeffs.

[egdaniel, 2015-02-17 19:09:19]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2015-02-17 19:09:29]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/914003003/80001

[commit-bot: I haz the power, 2015-02-17 19:15:52]

Committed patchset #5 (id:80001) as
https://skia.googlesource.com/skia/+/3ad6570e94965d54871eaea8b314f560b82ca792