Antialias matrix-image-filter to get smooth diagonals

Antialias matrix-image-filter to get smooth diagonals

BUG=skia:

[reed1, 2015-05-15 18:26:22]

reed@google.com changed reviewers:
+ bsalomon@google.com, robertphillips@google.com, senorblanco@google.com

[reed1, 2015-05-15 18:26:23]

chrome guard https://codereview.chromium.org/1133743003/#

[Stephen White, 2015-05-15 18:37:21]

senorblanco@chromium.org changed reviewers:
+ senorblanco@chromium.org

[Stephen White, 2015-05-15 18:37:22]

https://codereview.chromium.org/1140573004/diff/20001/src/core/SkCanvas.cpp
File src/core/SkCanvas.cpp (right):

https://codereview.chromium.org/1140573004/diff/20001/src/core/SkCanvas.cpp#n...
src/core/SkCanvas.cpp:1037: usage = SkBaseDevice::kPossible_TileUsage;
This seems like it'll work for the case where we have only a matrix image filter
in the filter chain, and it's drawing the layer device contents rotated. But
what it about a more complex chain, e.g., blur -> matrix. The blur will be
creating a scratch texture, which the matrix will now draw incorrectly.

It seems like we'd have to propagate this flag somehow to
SkDeviceImageFilterProxy::createDevice(), to ensure that all scratch textures
are created with exact dimensions.

[reed1, 2015-05-15 19:22:38]

That's probably right.

Since I don't have a test for that right now, and it will likely involve a
(nearly) disjoint set of changes (adding a param to the proxy API) I think I
will make that change (and write its test) separate from this CL.

[reed1, 2015-05-15 19:24:28]

The CQ bit was checked by reed@google.com to run a CQ dry run

[commit-bot: I haz the power, 2015-05-15 19:24:50]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1140573004/20001

[Stephen White, 2015-05-15 19:30:36]

On 2015/05/15 19:22:38, reed1 wrote:
> That's probably right.
> 
> Since I don't have a test for that right now, and it will likely involve a
> (nearly) disjoint set of changes (adding a param to the proxy API) I think I
> will make that change (and write its test) separate from this CL.

I feel this is adding a fair bit of complexity to image filters. I recently 
worked hard to remove the use of this flag from the majority of image 
filter code, and now we're going back down that road to support a relatively 
minor feature.

Let's have a chat with (along with Brian) at your convenience.

[reed1, 2015-05-15 19:34:58]

On 2015/05/15 19:30:36, Stephen White wrote:
> On 2015/05/15 19:22:38, reed1 wrote:
> > That's probably right.
> > 
> > Since I don't have a test for that right now, and it will likely involve a
> > (nearly) disjoint set of changes (adding a param to the proxy API) I think I
> > will make that change (and write its test) separate from this CL.
> 
> I feel this is adding a fair bit of complexity to image filters. I recently 
> worked hard to remove the use of this flag from the majority of image 
> filter code, and now we're going back down that road to support a relatively 
> minor feature.
> 
> Let's have a chat with (along with Brian) at your convenience.

Happy to chat. I'll see if I can arrange one today.

[reed1, 2015-05-15 19:35:37]

The CQ bit was unchecked by reed@google.com