Allow gradient optimization with perspective

Allow gradient optimization with perspective

Before, gradients would only interpolate the linear portion of the
quadratic equation if there was no perspective. This updates them to
do so even in the case that there is perspective. The rearrangement
of math causes noise differences in the following gm tests:

gradients_no_texture_gpu
gradients_view_perspective_gpu
gradients_local_perspective_gpu
gradients_gpu

Committed: http://code.google.com/p/skia/source/detail?r=11595

[Chris Dalton, 2013-10-03 20:54:04]

The diff tool didn't notice that all those changes to the gradient files are
just an un-tab. The only change I made to the fragment-emitting code is this:

    // We interpolate the linear component in coords[1].
    SkASSERT(coords[0].type() == coords[1].type());
    const char* coords2D;
    SkString bVar;
    if (kVec3f_GrSLType == coords[0].type()) {
        builder->fsCodeAppendf("\tvec3 interpolants = vec3(%s.xy, %s.x) /
%s.z;\n",
                               coords[0].c_str(), coords[1].c_str(),
coords[0].c_str());
        coords2D = "interpolants.xy";
        bVar = "interpolants.z";
    } else {
        coords2D = coords[0].c_str();
        bVar.printf("%s.x", coords[1].c_str());
    }

Instead of finding the coords2D/bVar values the old way.

[bsalomon, 2013-10-03 21:01:36]

On 2013/10/03 20:54:04, Chris Dalton wrote:
> The diff tool didn't notice that all those changes to the gradient files are
> just an un-tab. The only change I made to the fragment-emitting code is this:
> 
>     // We interpolate the linear component in coords[1].
>     SkASSERT(coords[0].type() == coords[1].type());
>     const char* coords2D;
>     SkString bVar;
>     if (kVec3f_GrSLType == coords[0].type()) {
>         builder->fsCodeAppendf("\tvec3 interpolants = vec3(%s.xy, %s.x) /
> %s.z;\n",
>                                coords[0].c_str(), coords[1].c_str(),
> coords[0].c_str());
>         coords2D = "interpolants.xy";
>         bVar = "interpolants.z";
>     } else {
>         coords2D = coords[0].c_str();
>         bVar.printf("%s.x", coords[1].c_str());
>     }
> 
> Instead of finding the coords2D/bVar values the old way.

lgtm, I sent it off to a try bot just to look for GM changes on at least one
other configuration. The file to add the test names to is:
expectations/gm/ignored-tests.txt

That should be done as part of the change. It'll keep the bots from all going
red and then we can rebaseline when the new images are ready and remove from the
txt file.

[bsalomon, 2013-10-03 21:43:35]

On 2013/10/03 21:01:36, bsalomon wrote:
> On 2013/10/03 20:54:04, Chris Dalton wrote:
> > The diff tool didn't notice that all those changes to the gradient files are
> > just an un-tab. The only change I made to the fragment-emitting code is
this:
> > 
> >     // We interpolate the linear component in coords[1].
> >     SkASSERT(coords[0].type() == coords[1].type());
> >     const char* coords2D;
> >     SkString bVar;
> >     if (kVec3f_GrSLType == coords[0].type()) {
> >         builder->fsCodeAppendf("\tvec3 interpolants = vec3(%s.xy, %s.x) /
> > %s.z;\n",
> >                                coords[0].c_str(), coords[1].c_str(),
> > coords[0].c_str());
> >         coords2D = "interpolants.xy";
> >         bVar = "interpolants.z";
> >     } else {
> >         coords2D = coords[0].c_str();
> >         bVar.printf("%s.x", coords[1].c_str());
> >     }
> > 
> > Instead of finding the coords2D/bVar values the old way.
> 
> lgtm, I sent it off to a try bot just to look for GM changes on at least one
> other configuration. The file to add the test names to is:
> expectations/gm/ignored-tests.txt
> 
> That should be done as part of the change. It'll keep the bots from all going
> red and then we can rebaseline when the new images are ready and remove from
the
> txt file.

Ok no other images changed on the try bot. If you update with test names in the
CL I'll land it later this evening or first thing in the AM.

[commit-bot: I haz the power, 2013-10-04 01:11:01]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/cdalton@nvidia.com/25645006/7001

[commit-bot: I haz the power, 2013-10-04 01:20:13]

Change committed as 11595