cc: optimize Skia image filter application via SkImage::makeWithFilter().

cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is
active/valid, and returns a rect (outSubset) which is the active region of
the returned image. This required supporting a non-zero texture
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

NOTRY=true
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/5e8ab54bd2b0ab10a85b943a10dc9ebea051bada
Cr-Commit-Position: refs/heads/master@{#395263}

[Stephen White, 2016-05-09 13:51:35]

Description was changed from

==========
cc: optimize Skia image filter application via filterImage().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImageFilter::filterImage() directly. This saves a texture allocation
and a blit.

WIP: this currently does not do caching of intermediate results.
Background filters are also likely broken, since they do not apply
the computed offset.

It also requires a change in Skia to move SkSpecialImage.h from
src/core to include/core.
==========

to

==========
cc: optimize Skia image filter application via filterImage().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImageFilter::filterImage() directly. This saves a texture allocation
and a blit.

WIP: this currently does not do caching of intermediate results.
Background filters are also likely broken, since they do not apply
the computed offset.

It also requires a change in Skia to move SkSpecialImage.h from
src/core to include/core.
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[Stephen White, 2016-05-17 17:36:31]

Description was changed from

==========
cc: optimize Skia image filter application via filterImage().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImageFilter::filterImage() directly. This saves a texture allocation
and a blit.

WIP: this currently does not do caching of intermediate results.
Background filters are also likely broken, since they do not apply
the computed offset.

It also requires a change in Skia to move SkSpecialImage.h from
src/core to include/core.
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWIthFilter() directly. This saves a texture allocation
and a blit.

This changes depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[Stephen White, 2016-05-17 17:41:38]

Patchset #6 (id:100001) has been deleted

[Stephen White, 2016-05-17 17:42:22]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWIthFilter() directly. This saves a texture allocation
and a blit.

This changes depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWIthFilter() directly. This saves a texture allocation
and a blit.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[esprehn, 2016-05-17 21:29:37]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWIthFilter() directly. This saves a texture allocation
and a blit.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[Stephen White, 2016-05-19 21:53:03]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[Stephen White, 2016-05-19 21:53:04]

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

[Stephen White, 2016-05-19 21:53:19]

enne@: PTAL. Thanks!

[Stephen White, 2016-05-19 21:57:53]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkSpecialImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is
active/valid, and returns a rect (outSubset) which is the active region of the
returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[Stephen White, 2016-05-19 21:58:47]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is
active/valid, and returns a rect (outSubset) which is the active region of the
returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is 
active/valid, and returns a rect (outSubset) which is the active region of 
the returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-05-20 00:53:45]

Is it possible to force your tests to use offsets at all?

https://codereview.chromium.org/1959033002/diff/220001/cc/output/gl_renderer.cc
File cc/output/gl_renderer.cc (right):

https://codereview.chromium.org/1959033002/diff/220001/cc/output/gl_renderer....
cc/output/gl_renderer.cc:653: *flip_texture = image->getTexture()->origin() ==
kBottomLeft_GrSurfaceOrigin;
Is this ever different? Also, it seems like if you're returning image, you don't
need this out param as well?

[Stephen White, 2016-05-20 03:31:39]

https://codereview.chromium.org/1959033002/diff/220001/cc/output/gl_renderer.cc
File cc/output/gl_renderer.cc (right):

https://codereview.chromium.org/1959033002/diff/220001/cc/output/gl_renderer....
cc/output/gl_renderer.cc:653: *flip_texture = image->getTexture()->origin() ==
kBottomLeft_GrSurfaceOrigin;
On 2016/05/20 00:53:45, enne wrote:
> Is this ever different?

Yep: with SVG filters, you can have an feImage which replaces the input. It gets
converted to a top-down SkImage in Skia. There's actually a layout test that
fails without this (css3/filters/effect-reference-image-hw.html).

> Also, it seems like if you're returning image, you don't
> need this out param as well?

Good idea. I'll just move it into the caller.

[Stephen White, 2016-05-20 03:41:02]

On 2016/05/20 00:53:45, enne wrote:
> Is it possible to force your tests to use offsets at all?

I assume you mean src_offset? (because the offset returned by makeWithFilter()
is non-zero in a number of tests, including
EnlargedTextureWithCropOffsetFilter).

I'll give it a shot. Somehow it needs to get clipBounds that clip the top left
corner of the resulting image.

[Stephen White, 2016-05-20 15:21:51]

New patch up: added a pixel test to exercise src_offset.

[enne (OOO), 2016-05-20 17:41:55]

lgtm

Thanks for the extra test.  :)

[Stephen White, 2016-05-21 06:08:15]

The CQ bit was checked by senorblanco@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-05-21 06:08:20]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1959033002/260001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1959033002/260001

[commit-bot: I haz the power, 2016-05-21 06:53:06]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-05-21 06:53:07]

Dry run: Try jobs failed on following builders:
  mac_chromium_rel_ng on tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...)

[Stephen White, 2016-05-21 14:54:24]

The CQ bit was checked by senorblanco@chromium.org

[Stephen White, 2016-05-21 14:54:25]

The patchset sent to the CQ was uploaded after l-g-t-m from enne@chromium.org
Link to the patchset: https://codereview.chromium.org/1959033002/#ps280001
(title: "Mark more tests for rebaseline; tweak fuzzy test compares on Win")

[commit-bot: I haz the power, 2016-05-21 14:54:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1959033002/280001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1959033002/280001

[commit-bot: I haz the power, 2016-05-21 15:46:03]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-05-21 15:46:04]

Try jobs failed on following builders:
  linux_android_rel_ng on tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/linux_androi...)

[Stephen White, 2016-05-21 17:07:01]

The CQ bit was checked by senorblanco@chromium.org

[Stephen White, 2016-05-21 17:07:02]

The patchset sent to the CQ was uploaded after l-g-t-m from enne@chromium.org
Link to the patchset: https://codereview.chromium.org/1959033002/#ps300001
(title: "Win pixeltest tweakage")

[commit-bot: I haz the power, 2016-05-21 17:07:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1959033002/300001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1959033002/300001

[commit-bot: I haz the power, 2016-05-21 19:35:23]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-05-21 19:35:25]

Try jobs failed on following builders:
  linux_chromium_rel_ng on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[Stephen White, 2016-05-21 20:24:26]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is 
active/valid, and returns a rect (outSubset) which is the active region of 
the returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is 
active/valid, and returns a rect (outSubset) which is the active region of 
the returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

NOTRY=true
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[Stephen White, 2016-05-21 20:24:33]

The CQ bit was checked by senorblanco@chromium.org

[commit-bot: I haz the power, 2016-05-21 20:24:39]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1959033002/300001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1959033002/300001

[commit-bot: I haz the power, 2016-05-21 20:27:41]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is 
active/valid, and returns a rect (outSubset) which is the active region of 
the returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

NOTRY=true
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is 
active/valid, and returns a rect (outSubset) which is the active region of 
the returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

NOTRY=true
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[commit-bot: I haz the power, 2016-05-21 20:27:41]

Committed patchset #15 (id:300001)

[commit-bot: I haz the power, 2016-05-21 20:28:51]

Description was changed from

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is 
active/valid, and returns a rect (outSubset) which is the active region of 
the returned image. This required supporting a non-zero texture 
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

NOTRY=true
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
cc: optimize Skia image filter application via SkImage::makeWithFilter().

Instead of allocating a Skia SkSurface, and drawing the source bitmap to
its canvas, wrap the input in an SkImage, and call
SkImage::makeWithFilter() directly. This saves a texture allocation
and a blit.

In order to support approx-match textures, the makeWithFilter API also
takes a rect (subset) representing the subset of the image which is
active/valid, and returns a rect (outSubset) which is the active region of
the returned image. This required supporting a non-zero texture
coordinate origin in cc.

This change depends on https://codereview.chromium.org/1964043002/ in Skia.

NOTRY=true
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/5e8ab54bd2b0ab10a85b943a10dc9ebea051bada
Cr-Commit-Position: refs/heads/master@{#395263}
==========

[commit-bot: I haz the power, 2016-05-21 20:28:52]

Patchset 15 (id:??) landed as
https://crrev.com/5e8ab54bd2b0ab10a85b943a10dc9ebea051bada
Cr-Commit-Position: refs/heads/master@{#395263}