Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
/* cairo - a vector graphics library with display and print output
|
|
|
|
|
*
|
|
|
|
|
* Copyright © 2005 Red Hat, Inc
|
|
|
|
|
* Copyright © 2007 Adrian Johnson
|
|
|
|
|
* Copyright © 2009 Chris Wilson
|
|
|
|
|
*
|
|
|
|
|
* This library is free software; you can redistribute it and/or
|
|
|
|
|
* modify it either under the terms of the GNU Lesser General Public
|
|
|
|
|
* License version 2.1 as published by the Free Software Foundation
|
|
|
|
|
* (the "LGPL") or, at your option, under the terms of the Mozilla
|
|
|
|
|
* Public License Version 1.1 (the "MPL"). If you do not alter this
|
|
|
|
|
* notice, a recipient may use your version of this file under either
|
|
|
|
|
* the MPL or the LGPL.
|
|
|
|
|
*
|
|
|
|
|
* You should have received a copy of the LGPL along with this library
|
|
|
|
|
* in the file COPYING-LGPL-2.1; if not, write to the Free Software
|
2010-04-27 10:17:23 +02:00
|
|
|
* Foundation, Inc., 51 Franklin Street, Suite 500, Boston, MA 02110-1335, USA
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
* You should have received a copy of the MPL along with this library
|
|
|
|
|
* in the file COPYING-MPL-1.1
|
|
|
|
|
*
|
|
|
|
|
* The contents of this file are subject to the Mozilla Public License
|
|
|
|
|
* Version 1.1 (the "License"); you may not use this file except in
|
|
|
|
|
* compliance with the License. You may obtain a copy of the License at
|
|
|
|
|
* http://www.mozilla.org/MPL/
|
|
|
|
|
*
|
|
|
|
|
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY
|
|
|
|
|
* OF ANY KIND, either express or implied. See the LGPL or the MPL for
|
|
|
|
|
* the specific language governing rights and limitations.
|
|
|
|
|
*
|
|
|
|
|
* The Original Code is the cairo graphics library.
|
|
|
|
|
*
|
|
|
|
|
* The Initial Developer of the Original Code is Red Hat, Inc.
|
|
|
|
|
*
|
|
|
|
|
* Contributor(s):
|
|
|
|
|
* Chris Wilson <chris@chris-wilson.co.uk>
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include "cairoint.h"
|
|
|
|
|
|
2010-01-18 16:58:40 +00:00
|
|
|
#include "cairo-error-private.h"
|
2011-07-14 18:35:32 +01:00
|
|
|
#include "cairo-pattern-private.h"
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
#include "cairo-surface-wrapper-private.h"
|
|
|
|
|
|
|
|
|
|
/* A collection of routines to facilitate surface wrapping */
|
|
|
|
|
|
2010-01-18 22:47:34 +00:00
|
|
|
static void
|
|
|
|
|
_copy_transformed_pattern (cairo_pattern_t *pattern,
|
|
|
|
|
const cairo_pattern_t *original,
|
|
|
|
|
const cairo_matrix_t *ctm_inverse)
|
|
|
|
|
{
|
|
|
|
|
_cairo_pattern_init_static_copy (pattern, original);
|
|
|
|
|
|
|
|
|
|
if (! _cairo_matrix_is_identity (ctm_inverse))
|
|
|
|
|
_cairo_pattern_transform (pattern, ctm_inverse);
|
|
|
|
|
}
|
|
|
|
|
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_acquire_source_image (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_image_surface_t **image_out,
|
|
|
|
|
void **image_extra)
|
|
|
|
|
{
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
|
|
|
|
return _cairo_surface_acquire_source_image (wrapper->target,
|
|
|
|
|
image_out, image_extra);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
_cairo_surface_wrapper_release_source_image (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_image_surface_t *image,
|
|
|
|
|
void *image_extra)
|
|
|
|
|
{
|
|
|
|
|
_cairo_surface_release_source_image (wrapper->target, image, image_extra);
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
static void
|
|
|
|
|
_cairo_surface_wrapper_get_transform (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_matrix_t *m)
|
|
|
|
|
{
|
|
|
|
|
cairo_matrix_init_identity (m);
|
|
|
|
|
|
|
|
|
|
if (wrapper->has_extents && (wrapper->extents.x || wrapper->extents.y))
|
|
|
|
|
cairo_matrix_translate (m, -wrapper->extents.x, -wrapper->extents.y);
|
|
|
|
|
|
|
|
|
|
if (! _cairo_matrix_is_identity (&wrapper->transform))
|
|
|
|
|
cairo_matrix_multiply (m, &wrapper->transform, m);
|
|
|
|
|
|
|
|
|
|
if (! _cairo_matrix_is_identity (&wrapper->target->device_transform))
|
|
|
|
|
cairo_matrix_multiply (m, &wrapper->target->device_transform, m);
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-26 09:03:37 +01:00
|
|
|
static void
|
|
|
|
|
_cairo_surface_wrapper_get_inverse_transform (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_matrix_t *m)
|
|
|
|
|
{
|
|
|
|
|
cairo_matrix_init_identity (m);
|
|
|
|
|
|
|
|
|
|
if (! _cairo_matrix_is_identity (&wrapper->target->device_transform_inverse))
|
|
|
|
|
cairo_matrix_multiply (m, &wrapper->target->device_transform_inverse, m);
|
|
|
|
|
|
|
|
|
|
if (! _cairo_matrix_is_identity (&wrapper->transform)) {
|
|
|
|
|
cairo_matrix_t inv;
|
|
|
|
|
cairo_status_t status;
|
|
|
|
|
|
|
|
|
|
inv = wrapper->transform;
|
|
|
|
|
status = cairo_matrix_invert (&inv);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
cairo_matrix_multiply (m, &inv, m);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (wrapper->has_extents && (wrapper->extents.x || wrapper->extents.y))
|
|
|
|
|
cairo_matrix_translate (m, wrapper->extents.x, wrapper->extents.y);
|
|
|
|
|
}
|
2011-07-23 12:19:17 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
static cairo_clip_t *
|
|
|
|
|
_cairo_surface_wrapper_get_clip (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
const cairo_clip_t *clip)
|
|
|
|
|
{
|
|
|
|
|
cairo_clip_t *copy;
|
|
|
|
|
|
2011-07-24 16:27:07 +01:00
|
|
|
copy = _cairo_clip_copy (clip);
|
2011-08-14 14:45:00 +01:00
|
|
|
if (wrapper->has_extents) {
|
2011-07-24 16:27:07 +01:00
|
|
|
copy = _cairo_clip_intersect_rectangle (copy, &wrapper->extents);
|
2011-08-14 14:45:00 +01:00
|
|
|
}
|
2011-09-20 14:04:29 +01:00
|
|
|
copy = _cairo_clip_transform (copy, &wrapper->transform);
|
2012-02-22 19:55:22 +00:00
|
|
|
if (! _cairo_matrix_is_identity (&wrapper->target->device_transform))
|
|
|
|
|
copy = _cairo_clip_transform (copy, &wrapper->target->device_transform);
|
2011-07-24 15:33:21 +01:00
|
|
|
if (wrapper->clip)
|
|
|
|
|
copy = _cairo_clip_intersect_clip (copy, wrapper->clip);
|
|
|
|
|
|
|
|
|
|
return copy;
|
|
|
|
|
}
|
|
|
|
|
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_paint (cairo_surface_wrapper_t *wrapper,
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_operator_t op,
|
|
|
|
|
const cairo_pattern_t *source,
|
2011-08-13 10:48:21 +01:00
|
|
|
const cairo_clip_t *clip)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
2011-07-24 15:33:21 +01:00
|
|
|
cairo_clip_t *dev_clip;
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_pattern_union_t source_copy;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
dev_clip = _cairo_surface_wrapper_get_clip (wrapper, clip);
|
|
|
|
|
if (_cairo_clip_is_all_clipped (dev_clip))
|
|
|
|
|
return CAIRO_INT_STATUS_NOTHING_TO_DO;
|
2010-04-28 14:26:21 +01:00
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
if (wrapper->needs_transform) {
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_t m;
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
_cairo_surface_wrapper_get_transform (wrapper, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
status = cairo_matrix_invert (&m);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
|
|
|
|
|
_copy_transformed_pattern (&source_copy.base, source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
source = &source_copy.base;
|
|
|
|
|
}
|
2009-11-27 18:13:45 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
status = _cairo_surface_paint (wrapper->target, op, source, dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
_cairo_clip_destroy (dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_mask (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_operator_t op,
|
|
|
|
|
const cairo_pattern_t *source,
|
|
|
|
|
const cairo_pattern_t *mask,
|
2011-07-14 21:19:54 +01:00
|
|
|
const cairo_clip_t *clip)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
2011-07-24 15:33:21 +01:00
|
|
|
cairo_clip_t *dev_clip;
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_pattern_union_t source_copy;
|
|
|
|
|
cairo_pattern_union_t mask_copy;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
dev_clip = _cairo_surface_wrapper_get_clip (wrapper, clip);
|
|
|
|
|
if (_cairo_clip_is_all_clipped (dev_clip))
|
|
|
|
|
return CAIRO_INT_STATUS_NOTHING_TO_DO;
|
2010-04-28 14:26:21 +01:00
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
if (wrapper->needs_transform) {
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_t m;
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
_cairo_surface_wrapper_get_transform (wrapper, &m);
|
2010-04-28 14:26:21 +01:00
|
|
|
|
|
|
|
|
status = cairo_matrix_invert (&m);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
|
|
|
|
|
_copy_transformed_pattern (&source_copy.base, source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
source = &source_copy.base;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
_copy_transformed_pattern (&mask_copy.base, mask, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
mask = &mask_copy.base;
|
|
|
|
|
}
|
2009-11-27 18:13:45 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
status = _cairo_surface_mask (wrapper->target, op, source, mask, dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
_cairo_clip_destroy (dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_stroke (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_operator_t op,
|
|
|
|
|
const cairo_pattern_t *source,
|
2011-07-14 21:19:54 +01:00
|
|
|
const cairo_path_fixed_t *path,
|
2010-01-18 18:20:16 +00:00
|
|
|
const cairo_stroke_style_t *stroke_style,
|
|
|
|
|
const cairo_matrix_t *ctm,
|
|
|
|
|
const cairo_matrix_t *ctm_inverse,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
double tolerance,
|
|
|
|
|
cairo_antialias_t antialias,
|
2011-07-14 21:19:54 +01:00
|
|
|
const cairo_clip_t *clip)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
2011-07-14 21:19:54 +01:00
|
|
|
cairo_path_fixed_t path_copy, *dev_path = (cairo_path_fixed_t *) path;
|
2011-07-24 15:33:21 +01:00
|
|
|
cairo_clip_t *dev_clip;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_matrix_t dev_ctm = *ctm;
|
|
|
|
|
cairo_matrix_t dev_ctm_inverse = *ctm_inverse;
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_pattern_union_t source_copy;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
dev_clip = _cairo_surface_wrapper_get_clip (wrapper, clip);
|
|
|
|
|
if (_cairo_clip_is_all_clipped (dev_clip))
|
|
|
|
|
return CAIRO_INT_STATUS_NOTHING_TO_DO;
|
2010-04-28 14:26:21 +01:00
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
if (wrapper->needs_transform) {
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_t m;
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
_cairo_surface_wrapper_get_transform (wrapper, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
status = _cairo_path_fixed_init_copy (&path_copy, dev_path);
|
|
|
|
|
if (unlikely (status))
|
|
|
|
|
goto FINISH;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
_cairo_path_fixed_transform (&path_copy, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
dev_path = &path_copy;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_multiply (&dev_ctm, &dev_ctm, &m);
|
|
|
|
|
|
|
|
|
|
status = cairo_matrix_invert (&m);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
2010-01-18 22:47:34 +00:00
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_multiply (&dev_ctm_inverse, &m, &dev_ctm_inverse);
|
|
|
|
|
|
|
|
|
|
_copy_transformed_pattern (&source_copy.base, source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
source = &source_copy.base;
|
2010-04-28 14:26:21 +01:00
|
|
|
}
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
status = _cairo_surface_stroke (wrapper->target, op, source,
|
|
|
|
|
dev_path, stroke_style,
|
|
|
|
|
&dev_ctm, &dev_ctm_inverse,
|
|
|
|
|
tolerance, antialias,
|
|
|
|
|
dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
FINISH:
|
|
|
|
|
if (dev_path != path)
|
|
|
|
|
_cairo_path_fixed_fini (dev_path);
|
2011-07-24 15:33:21 +01:00
|
|
|
_cairo_clip_destroy (dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_fill_stroke (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_operator_t fill_op,
|
|
|
|
|
const cairo_pattern_t *fill_source,
|
|
|
|
|
cairo_fill_rule_t fill_rule,
|
|
|
|
|
double fill_tolerance,
|
|
|
|
|
cairo_antialias_t fill_antialias,
|
2011-07-24 18:10:43 +01:00
|
|
|
const cairo_path_fixed_t*path,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_operator_t stroke_op,
|
|
|
|
|
const cairo_pattern_t *stroke_source,
|
2010-01-18 18:20:16 +00:00
|
|
|
const cairo_stroke_style_t *stroke_style,
|
|
|
|
|
const cairo_matrix_t *stroke_ctm,
|
|
|
|
|
const cairo_matrix_t *stroke_ctm_inverse,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
double stroke_tolerance,
|
|
|
|
|
cairo_antialias_t stroke_antialias,
|
2011-07-14 21:19:54 +01:00
|
|
|
const cairo_clip_t *clip)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
2011-07-24 18:10:43 +01:00
|
|
|
cairo_path_fixed_t path_copy, *dev_path = (cairo_path_fixed_t *)path;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_matrix_t dev_ctm = *stroke_ctm;
|
|
|
|
|
cairo_matrix_t dev_ctm_inverse = *stroke_ctm_inverse;
|
2011-07-24 15:33:21 +01:00
|
|
|
cairo_clip_t *dev_clip;
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_pattern_union_t stroke_source_copy;
|
|
|
|
|
cairo_pattern_union_t fill_source_copy;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
dev_clip = _cairo_surface_wrapper_get_clip (wrapper, clip);
|
|
|
|
|
if (_cairo_clip_is_all_clipped (dev_clip))
|
|
|
|
|
return CAIRO_INT_STATUS_NOTHING_TO_DO;
|
2010-04-28 14:26:21 +01:00
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
if (wrapper->needs_transform) {
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_t m;
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
_cairo_surface_wrapper_get_transform (wrapper, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
status = _cairo_path_fixed_init_copy (&path_copy, dev_path);
|
|
|
|
|
if (unlikely (status))
|
|
|
|
|
goto FINISH;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
_cairo_path_fixed_transform (&path_copy, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
dev_path = &path_copy;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_multiply (&dev_ctm, &dev_ctm, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
status = cairo_matrix_invert (&m);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
|
|
|
|
|
cairo_matrix_multiply (&dev_ctm_inverse, &m, &dev_ctm_inverse);
|
|
|
|
|
|
|
|
|
|
_copy_transformed_pattern (&stroke_source_copy.base, stroke_source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
stroke_source = &stroke_source_copy.base;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
_copy_transformed_pattern (&fill_source_copy.base, fill_source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
fill_source = &fill_source_copy.base;
|
2010-04-28 14:26:21 +01:00
|
|
|
}
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
status = _cairo_surface_fill_stroke (wrapper->target,
|
|
|
|
|
fill_op, fill_source, fill_rule,
|
|
|
|
|
fill_tolerance, fill_antialias,
|
|
|
|
|
dev_path,
|
|
|
|
|
stroke_op, stroke_source,
|
|
|
|
|
stroke_style,
|
|
|
|
|
&dev_ctm, &dev_ctm_inverse,
|
|
|
|
|
stroke_tolerance, stroke_antialias,
|
|
|
|
|
dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
FINISH:
|
|
|
|
|
if (dev_path != path)
|
|
|
|
|
_cairo_path_fixed_fini (dev_path);
|
2011-07-24 15:33:21 +01:00
|
|
|
_cairo_clip_destroy (dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_fill (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_operator_t op,
|
|
|
|
|
const cairo_pattern_t *source,
|
2011-07-14 21:19:54 +01:00
|
|
|
const cairo_path_fixed_t *path,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_fill_rule_t fill_rule,
|
|
|
|
|
double tolerance,
|
|
|
|
|
cairo_antialias_t antialias,
|
2011-07-14 21:19:54 +01:00
|
|
|
const cairo_clip_t *clip)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
2011-07-14 21:19:54 +01:00
|
|
|
cairo_path_fixed_t path_copy, *dev_path = (cairo_path_fixed_t *) path;
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_pattern_union_t source_copy;
|
2011-07-24 15:33:21 +01:00
|
|
|
cairo_clip_t *dev_clip;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
dev_clip = _cairo_surface_wrapper_get_clip (wrapper, clip);
|
|
|
|
|
if (_cairo_clip_is_all_clipped (dev_clip))
|
|
|
|
|
return CAIRO_INT_STATUS_NOTHING_TO_DO;
|
2010-04-28 14:26:21 +01:00
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
if (wrapper->needs_transform) {
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_t m;
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
_cairo_surface_wrapper_get_transform (wrapper, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
status = _cairo_path_fixed_init_copy (&path_copy, dev_path);
|
|
|
|
|
if (unlikely (status))
|
|
|
|
|
goto FINISH;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
_cairo_path_fixed_transform (&path_copy, &m);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
dev_path = &path_copy;
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
status = cairo_matrix_invert (&m);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
|
|
|
|
|
_copy_transformed_pattern (&source_copy.base, source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
source = &source_copy.base;
|
2010-04-28 14:26:21 +01:00
|
|
|
}
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
status = _cairo_surface_fill (wrapper->target, op, source,
|
|
|
|
|
dev_path, fill_rule,
|
|
|
|
|
tolerance, antialias,
|
|
|
|
|
dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
FINISH:
|
|
|
|
|
if (dev_path != path)
|
|
|
|
|
_cairo_path_fixed_fini (dev_path);
|
2011-07-24 15:33:21 +01:00
|
|
|
_cairo_clip_destroy (dev_clip);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cairo_status_t
|
|
|
|
|
_cairo_surface_wrapper_show_text_glyphs (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_operator_t op,
|
|
|
|
|
const cairo_pattern_t *source,
|
|
|
|
|
const char *utf8,
|
|
|
|
|
int utf8_len,
|
2011-07-24 15:48:27 +01:00
|
|
|
const cairo_glyph_t *glyphs,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
int num_glyphs,
|
|
|
|
|
const cairo_text_cluster_t *clusters,
|
|
|
|
|
int num_clusters,
|
|
|
|
|
cairo_text_cluster_flags_t cluster_flags,
|
|
|
|
|
cairo_scaled_font_t *scaled_font,
|
2011-07-24 15:48:27 +01:00
|
|
|
const cairo_clip_t *clip)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
2011-07-24 15:33:21 +01:00
|
|
|
cairo_clip_t *dev_clip;
|
2011-07-24 15:45:57 +01:00
|
|
|
cairo_glyph_t stack_glyphs [CAIRO_STACK_ARRAY_LENGTH(cairo_glyph_t)];
|
|
|
|
|
cairo_glyph_t *dev_glyphs = stack_glyphs;
|
2012-02-27 14:25:39 +00:00
|
|
|
cairo_scaled_font_t *dev_scaled_font = scaled_font;
|
2010-01-18 22:47:34 +00:00
|
|
|
cairo_pattern_union_t source_copy;
|
2012-02-28 18:11:22 +00:00
|
|
|
cairo_font_options_t options;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
|
|
|
|
if (unlikely (wrapper->target->status))
|
|
|
|
|
return wrapper->target->status;
|
|
|
|
|
|
2011-07-24 15:33:21 +01:00
|
|
|
dev_clip = _cairo_surface_wrapper_get_clip (wrapper, clip);
|
2011-07-24 15:10:18 +01:00
|
|
|
if (_cairo_clip_is_all_clipped (dev_clip))
|
|
|
|
|
return CAIRO_INT_STATUS_NOTHING_TO_DO;
|
|
|
|
|
|
2012-02-28 18:11:22 +00:00
|
|
|
cairo_surface_get_font_options (wrapper->target, &options);
|
|
|
|
|
cairo_font_options_merge (&options, &scaled_font->options);
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
if (wrapper->needs_transform) {
|
2010-04-28 14:26:21 +01:00
|
|
|
cairo_matrix_t m;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
int i;
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
_cairo_surface_wrapper_get_transform (wrapper, &m);
|
2010-04-28 14:26:21 +01:00
|
|
|
|
2012-03-06 14:11:49 +00:00
|
|
|
if (! _cairo_matrix_is_translation (&wrapper->transform)) {
|
2012-02-28 18:11:22 +00:00
|
|
|
cairo_matrix_t ctm;
|
|
|
|
|
|
2012-03-06 14:11:49 +00:00
|
|
|
/* XXX No device-transform? A bug in the tangle of layers? */
|
|
|
|
|
_cairo_matrix_multiply (&ctm,
|
|
|
|
|
&wrapper->transform,
|
|
|
|
|
&scaled_font->ctm);
|
2012-02-27 14:25:39 +00:00
|
|
|
dev_scaled_font = cairo_scaled_font_create (scaled_font->font_face,
|
|
|
|
|
&scaled_font->font_matrix,
|
2012-02-28 18:11:22 +00:00
|
|
|
&ctm, &options);
|
2012-02-27 14:25:39 +00:00
|
|
|
}
|
|
|
|
|
|
2011-07-24 15:45:57 +01:00
|
|
|
if (num_glyphs > ARRAY_LENGTH (stack_glyphs)) {
|
|
|
|
|
dev_glyphs = _cairo_malloc_ab (num_glyphs, sizeof (cairo_glyph_t));
|
2011-07-30 17:28:21 +01:00
|
|
|
if (unlikely (dev_glyphs == NULL)) {
|
2011-07-24 15:45:57 +01:00
|
|
|
status = _cairo_error (CAIRO_STATUS_NO_MEMORY);
|
|
|
|
|
goto FINISH;
|
|
|
|
|
}
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < num_glyphs; i++) {
|
|
|
|
|
dev_glyphs[i] = glyphs[i];
|
2011-07-30 17:28:21 +01:00
|
|
|
cairo_matrix_transform_point (&m,
|
|
|
|
|
&dev_glyphs[i].x,
|
|
|
|
|
&dev_glyphs[i].y);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
}
|
2010-01-18 22:47:34 +00:00
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
status = cairo_matrix_invert (&m);
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
|
|
|
|
|
_copy_transformed_pattern (&source_copy.base, source, &m);
|
2010-01-18 22:47:34 +00:00
|
|
|
source = &source_copy.base;
|
2011-07-24 15:10:18 +01:00
|
|
|
} else {
|
2012-02-28 18:11:22 +00:00
|
|
|
if (! cairo_font_options_equal (&options, &scaled_font->options)) {
|
|
|
|
|
dev_scaled_font = cairo_scaled_font_create (scaled_font->font_face,
|
|
|
|
|
&scaled_font->font_matrix,
|
|
|
|
|
&scaled_font->ctm,
|
|
|
|
|
&options);
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-24 15:10:18 +01:00
|
|
|
/* show_text_glyphs is special because _cairo_surface_show_text_glyphs is allowed
|
|
|
|
|
* to modify the glyph array that's passed in. We must always
|
|
|
|
|
* copy the array before handing it to the backend.
|
|
|
|
|
*/
|
2011-07-24 15:45:57 +01:00
|
|
|
if (num_glyphs > ARRAY_LENGTH (stack_glyphs)) {
|
|
|
|
|
dev_glyphs = _cairo_malloc_ab (num_glyphs, sizeof (cairo_glyph_t));
|
|
|
|
|
if (unlikely (dev_glyphs == NULL)) {
|
|
|
|
|
status = _cairo_error (CAIRO_STATUS_NO_MEMORY);
|
|
|
|
|
goto FINISH;
|
|
|
|
|
}
|
2011-07-24 15:10:18 +01:00
|
|
|
}
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
|
2011-07-24 15:10:18 +01:00
|
|
|
memcpy (dev_glyphs, glyphs, sizeof (cairo_glyph_t) * num_glyphs);
|
|
|
|
|
}
|
2011-07-23 12:19:17 +01:00
|
|
|
|
2010-01-18 22:47:34 +00:00
|
|
|
status = _cairo_surface_show_text_glyphs (wrapper->target, op, source,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
utf8, utf8_len,
|
|
|
|
|
dev_glyphs, num_glyphs,
|
|
|
|
|
clusters, num_clusters,
|
|
|
|
|
cluster_flags,
|
2012-02-27 14:25:39 +00:00
|
|
|
dev_scaled_font,
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
dev_clip);
|
|
|
|
|
FINISH:
|
2011-07-24 15:33:21 +01:00
|
|
|
_cairo_clip_destroy (dev_clip);
|
2011-07-24 15:45:57 +01:00
|
|
|
if (dev_glyphs != stack_glyphs)
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
free (dev_glyphs);
|
2012-02-27 14:25:39 +00:00
|
|
|
if (dev_scaled_font != scaled_font)
|
|
|
|
|
cairo_scaled_font_destroy (dev_scaled_font);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cairo_surface_t *
|
|
|
|
|
_cairo_surface_wrapper_create_similar (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_content_t content,
|
|
|
|
|
int width,
|
|
|
|
|
int height)
|
|
|
|
|
{
|
2010-01-18 22:47:34 +00:00
|
|
|
return _cairo_surface_create_similar_scratch (wrapper->target,
|
|
|
|
|
content, width, height);
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cairo_bool_t
|
|
|
|
|
_cairo_surface_wrapper_get_extents (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_rectangle_int_t *extents)
|
|
|
|
|
{
|
2010-04-28 14:26:21 +01:00
|
|
|
if (wrapper->has_extents) {
|
|
|
|
|
if (_cairo_surface_get_extents (wrapper->target, extents))
|
|
|
|
|
_cairo_rectangle_intersect (extents, &wrapper->extents);
|
|
|
|
|
else
|
|
|
|
|
*extents = wrapper->extents;
|
|
|
|
|
|
|
|
|
|
return TRUE;
|
|
|
|
|
} else {
|
|
|
|
|
return _cairo_surface_get_extents (wrapper->target, extents);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-23 12:19:17 +01:00
|
|
|
static cairo_bool_t
|
|
|
|
|
_cairo_surface_wrapper_needs_device_transform (cairo_surface_wrapper_t *wrapper)
|
|
|
|
|
{
|
|
|
|
|
return
|
|
|
|
|
(wrapper->has_extents && (wrapper->extents.x | wrapper->extents.y)) ||
|
|
|
|
|
! _cairo_matrix_is_identity (&wrapper->transform) ||
|
|
|
|
|
! _cairo_matrix_is_identity (&wrapper->target->device_transform);
|
|
|
|
|
}
|
|
|
|
|
|
2010-04-28 14:26:21 +01:00
|
|
|
void
|
2011-07-24 18:10:43 +01:00
|
|
|
_cairo_surface_wrapper_intersect_extents (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
const cairo_rectangle_int_t *extents)
|
2010-04-28 14:26:21 +01:00
|
|
|
{
|
2011-07-24 18:10:43 +01:00
|
|
|
if (! wrapper->has_extents) {
|
2010-04-28 14:26:21 +01:00
|
|
|
wrapper->extents = *extents;
|
|
|
|
|
wrapper->has_extents = TRUE;
|
2011-07-24 18:10:43 +01:00
|
|
|
} else
|
|
|
|
|
_cairo_rectangle_intersect (&wrapper->extents, extents);
|
2011-07-23 12:19:17 +01:00
|
|
|
|
|
|
|
|
wrapper->needs_transform =
|
|
|
|
|
_cairo_surface_wrapper_needs_device_transform (wrapper);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
_cairo_surface_wrapper_set_inverse_transform (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
const cairo_matrix_t *transform)
|
|
|
|
|
{
|
|
|
|
|
cairo_status_t status;
|
|
|
|
|
|
|
|
|
|
if (transform == NULL || _cairo_matrix_is_identity (transform)) {
|
|
|
|
|
cairo_matrix_init_identity (&wrapper->transform);
|
|
|
|
|
|
|
|
|
|
wrapper->needs_transform =
|
|
|
|
|
_cairo_surface_wrapper_needs_device_transform (wrapper);
|
|
|
|
|
} else {
|
|
|
|
|
wrapper->transform = *transform;
|
|
|
|
|
status = cairo_matrix_invert (&wrapper->transform);
|
|
|
|
|
/* should always be invertible unless given pathological input */
|
|
|
|
|
assert (status == CAIRO_STATUS_SUCCESS);
|
|
|
|
|
|
|
|
|
|
wrapper->needs_transform = TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
_cairo_surface_wrapper_set_clip (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
const cairo_clip_t *clip)
|
|
|
|
|
{
|
|
|
|
|
wrapper->clip = clip;
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
}
|
|
|
|
|
|
2009-08-16 18:04:54 +01:00
|
|
|
void
|
|
|
|
|
_cairo_surface_wrapper_get_font_options (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_font_options_t *options)
|
|
|
|
|
{
|
2009-08-31 23:57:20 +01:00
|
|
|
cairo_surface_get_font_options (wrapper->target, options);
|
2009-08-16 18:04:54 +01:00
|
|
|
}
|
|
|
|
|
|
2009-08-13 18:37:01 +01:00
|
|
|
cairo_surface_t *
|
|
|
|
|
_cairo_surface_wrapper_snapshot (cairo_surface_wrapper_t *wrapper)
|
|
|
|
|
{
|
2011-08-14 09:47:04 +01:00
|
|
|
if (wrapper->target->backend->snapshot)
|
|
|
|
|
return wrapper->target->backend->snapshot (wrapper->target);
|
|
|
|
|
|
|
|
|
|
return NULL;
|
2009-08-13 18:37:01 +01:00
|
|
|
}
|
|
|
|
|
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
cairo_bool_t
|
|
|
|
|
_cairo_surface_wrapper_has_show_text_glyphs (cairo_surface_wrapper_t *wrapper)
|
|
|
|
|
{
|
|
|
|
|
return cairo_surface_has_show_text_glyphs (wrapper->target);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
_cairo_surface_wrapper_init (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_surface_t *target)
|
|
|
|
|
{
|
|
|
|
|
wrapper->target = cairo_surface_reference (target);
|
2011-07-23 22:57:48 +01:00
|
|
|
cairo_matrix_init_identity (&wrapper->transform);
|
2010-04-28 14:26:21 +01:00
|
|
|
wrapper->has_extents = FALSE;
|
2011-07-24 16:27:07 +01:00
|
|
|
wrapper->extents.x = wrapper->extents.y = 0;
|
2011-08-13 10:48:21 +01:00
|
|
|
wrapper->clip = NULL;
|
2011-07-23 22:57:48 +01:00
|
|
|
|
2011-08-10 13:22:58 +01:00
|
|
|
wrapper->needs_transform = FALSE;
|
|
|
|
|
if (target) {
|
|
|
|
|
wrapper->needs_transform =
|
|
|
|
|
! _cairo_matrix_is_identity (&target->device_transform);
|
|
|
|
|
}
|
Remove clip handling from generic surface layer.
Handling clip as part of the surface state, as opposed to being part of
the operation state, is cumbersome and a hindrance to providing true proxy
surface support. For example, the clip must be copied from the surface
onto the fallback image, but this was forgotten causing undue hassle in
each backend. Another example is the contortion the meta surface
endures to ensure the clip is correctly recorded. By contrast passing the
clip along with the operation is quite simple and enables us to write
generic handlers for providing surface wrappers. (And in the future, we
should be able to write more esoteric wrappers, e.g. automatic 2x FSAA,
trivially.)
In brief, instead of the surface automatically applying the clip before
calling the backend, the backend can call into a generic helper to apply
clipping. For raster surfaces, clip regions are handled automatically as
part of the composite interface. For vector surfaces, a clip helper is
introduced to replay and callback into an intersect_clip_path() function
as necessary.
Whilst this is not primarily a performance related change (the change
should just move the computation of the clip from the moment it is applied
by the user to the moment it is required by the backend), it is important
to track any potential regression:
ppc:
Speedups
========
image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup
▌
image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup
▎
image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup
▏
image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup
▏
Slowdowns
=========
image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown
▏
2009-07-23 15:32:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
_cairo_surface_wrapper_fini (cairo_surface_wrapper_t *wrapper)
|
|
|
|
|
{
|
|
|
|
|
cairo_surface_destroy (wrapper->target);
|
|
|
|
|
}
|
2011-07-26 08:33:09 +01:00
|
|
|
|
|
|
|
|
cairo_bool_t
|
|
|
|
|
_cairo_surface_wrapper_get_target_extents (cairo_surface_wrapper_t *wrapper,
|
|
|
|
|
cairo_rectangle_int_t *extents)
|
|
|
|
|
{
|
2011-09-17 12:00:49 +01:00
|
|
|
cairo_rectangle_int_t clip;
|
|
|
|
|
cairo_bool_t has_clip;
|
2011-07-26 08:33:09 +01:00
|
|
|
|
2011-09-17 12:00:49 +01:00
|
|
|
has_clip = _cairo_surface_get_extents (wrapper->target, &clip);
|
2011-07-26 08:33:09 +01:00
|
|
|
if (wrapper->clip) {
|
2011-09-17 12:00:49 +01:00
|
|
|
if (has_clip) {
|
|
|
|
|
if (! _cairo_rectangle_intersect (&clip,
|
|
|
|
|
_cairo_clip_get_extents (wrapper->clip)))
|
|
|
|
|
return FALSE;
|
|
|
|
|
} else {
|
|
|
|
|
has_clip = TRUE;
|
|
|
|
|
clip = *_cairo_clip_get_extents (wrapper->clip);
|
|
|
|
|
}
|
|
|
|
|
}
|
2011-07-26 08:33:09 +01:00
|
|
|
|
2011-09-17 12:00:49 +01:00
|
|
|
if (has_clip && wrapper->needs_transform) {
|
|
|
|
|
cairo_matrix_t m;
|
|
|
|
|
double x1, y1, x2, y2;
|
2011-07-26 08:33:09 +01:00
|
|
|
|
2011-09-17 12:00:49 +01:00
|
|
|
_cairo_surface_wrapper_get_inverse_transform (wrapper, &m);
|
2011-07-26 08:33:09 +01:00
|
|
|
|
2011-09-17 12:00:49 +01:00
|
|
|
x1 = clip.x;
|
|
|
|
|
y1 = clip.y;
|
|
|
|
|
x2 = clip.x + clip.width;
|
|
|
|
|
y2 = clip.y + clip.height;
|
2011-07-26 08:33:09 +01:00
|
|
|
|
2011-09-17 12:00:49 +01:00
|
|
|
_cairo_matrix_transform_bounding_box (&m, &x1, &y1, &x2, &y2, NULL);
|
|
|
|
|
|
|
|
|
|
clip.x = floor (x1);
|
|
|
|
|
clip.y = floor (y1);
|
|
|
|
|
clip.width = ceil (x2) - clip.x;
|
|
|
|
|
clip.height = ceil (y2) - clip.y;
|
|
|
|
|
}
|
2011-07-26 08:33:09 +01:00
|
|
|
|
2011-09-17 12:00:49 +01:00
|
|
|
if (has_clip) {
|
2011-07-26 08:33:09 +01:00
|
|
|
if (wrapper->has_extents) {
|
|
|
|
|
*extents = wrapper->extents;
|
2011-09-17 12:00:49 +01:00
|
|
|
return _cairo_rectangle_intersect (extents, &clip);
|
2011-07-26 08:33:09 +01:00
|
|
|
} else {
|
2011-09-17 12:00:49 +01:00
|
|
|
*extents = clip;
|
2011-07-26 08:33:09 +01:00
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
} else if (wrapper->has_extents) {
|
|
|
|
|
*extents = wrapper->extents;
|
|
|
|
|
return TRUE;
|
|
|
|
|
} else {
|
|
|
|
|
_cairo_unbounded_rectangle_init (extents);
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|