The "update" signal on drawable or projection can actually be emitted
many times for a single painting event. Just add new signals ("painted"
on GimpDrawable and "rendered" on GimpProjection) which are emitted once
for a single update (from user point of view), at the end, after actual
rendering is done (i.e. after the various "update" signals).
Also better support the sample merge vs current drawable paths for
bucket fill.
... and G_TYPE_INSTANCE_GET_PRIVATE()
g_type_class_add_private() and G_TYPE_INSTANCE_GET_PRIVATE() were
deprecated in GLib 2.58. Instead, use
G_DEFINE_[ABSTRACT_]TYPE_WITH_PRIVATE(), and
G_ADD_PRIVATE[_DYNAMIC](), and the implictly-defined
foo_get_instance_private() functions, all of which are available in
the GLib versions we depend on.
This commit only covers types registered using one of the
G_DEFINE_FOO() macros (i.e., most types), but not types with a
custom registration function, of which we still have a few -- GLib
currently only provides a (non-deprecated) public API for adding a
private struct using the G_DEFINE_FOO() macros.
Note that this commit was 99% auto-generated (because I'm not
*that* crazy :), so if there are any style mismatches... we'll have
to live with them for now.
In GimpProjection, use an adaptive chunk size when rendering the
projection asynchronously, rather than using a fixed chunk size.
The chunk size is determined according to the number of pixels
processed during the last frame, and the time it took to process
them, aiming for some target frame-rate (currently, 15 FPS). In
other words, the chunks become bigger when processing is fast, and
smaller when processing is slow. We're currently aiming for
generally-square chunks, whose sides are powers of 2, within a
predefined range.
Note that the chunk size represents a trade off between throughput
and responsiveness: bigger chunks result in better throughput,
since each individual chunk incurs an overhead, in particular when
rendering area filters or multithreaded ops, while smaller chunks
result in better responsiveness, since the time each chunk
individual takes to render is smaller, allowing us to more
accurately meet the target frame rate. With this commit, we aim to
find a good compromise dynamically, rather than statically.
The use of adaptive chunk sizes can be disabled by defining the
environment variable GIMP_NO_ADAPTIVE_CHUNK_SIZE, in which case we
use a fixed chunk size, as before.
Add a boolean "chunk" parameter to
gimp_projection_chunk_render_iteration(), which determines whether
the work area should be sub-divided into chunks prior to rendering
(previously, the work area would always be sub-divided.) Only
pass TRUE when rendering the projection asynchronously, in the
render callback, and pass FALSE when rendering the projection
synchronously, in gimp_projection_finish_draw(), which is called
when flushing the projection through the GimpPickable interface.
Rendering the projection using as big chunks as possible improves
performance, while worsening responsiveness. Since responsiveness
doesn't matter when rendering synchronously, there's no reason to
render in chunks.
In GimpProjection, change gimp_projection_add_update_area() to take
coordinates in the projection's coordinate system, rather than the
image coordinate system, and move the offset adjustment to the
projectable invalidation handler.
Modify gimp_projection_projectable_structure_changed() to pass
projection-space coordinates to gimp_projection_add_update_area().
gimp_projection_get_buffer() and
gimp_projection_projectable_bounds_changed() already pass
projection-space coordinates to gimp_projection_add_update_area(),
which was wrong before, when the projection had a nontrivial
offset, but is correct now.
In gimp_projection_projectable_bounds_changed(), bail early by
calling gimp_projection_projectable_structure_changed() instead, if
the new bounds don't intersect the old bounds.
In gimp_projection_projectable_bounds_changed(), which is called by
GimpProjection in response to a GimpProjectable::bounds-changed
signal, invalidate all regions of the new projection that weren't
copied from the old projection, so that they get rendered upon
flushing, instead of remaining empty.
Additionally, fix preview invalidation -- in particular, don't
directly invalidate the projectable's preview, even if preview
invalidation is already queued and chunk rendering was finished by
the boundary change, and instead always queue a preview
invalidation.
In GimpProjection, respond to the projectable's "bounds-changed"
signal, by reallocating the buffer, and copying the corresponding
region of the old buffer (using
gimp_tile_handler_validate_buffer_copy(), added a few commits back,
so that the relevant portion of the validate handler's dirty region
is also copied). Additionally, shift and clip all outstanding
update regions as necessary (actually, we avoid copying the buffer
when a shift is necessary, and simply reconstruct the projection;
see FIXME comment in the code.)
... which should be used to properly remove a
GimpTileHandlerValidate from a buffer, instead of using
gegl_buffer_remove_handler() directly.
Use gimp_tile_handler_validate_unassign(), instead of
gegl_buffer_remove_handler(), in gimp_projection_free_buffer().
All babl formats now have a space equivalent to a color profile,
determining the format's primaries and TRCs. This commit makes GIMP
aware of this.
libgimp:
- enum GimpPrecision: rename GAMMA values to NON_LINEAR and keep GAMMA
as deprecated aliases, add PERCEPTUAL values so we now have LINEAR,
NON_LINEAR and PERCPTUAL for each encoding, matching the babl
encoding variants RGB, R'G'B' and R~G~B~.
- gimp_color_transform_can_gegl_copy() now returns TRUE if both
profiles can return a babl space, increasing the amount of fast babl
color conversions significantly.
- TODO: no solution yet for getting libgimp drawable proxy buffers in
the right format with space.
plug-ins:
- follow the GimpPrecision change.
- TODO: everything else unchanged and partly broken or sub-optimal,
like setting a new image's color profile too late.
app:
- add enum GimpTRCType { LINEAR, NON_LINEAR, PERCEPTUAL } as
replacement for all "linear" booleans.
- change gimp-babl functions to take babl spaces and GimpTRCType
parameters and support all sorts of new perceptual ~ formats.
- a lot of places changed in the early days of goat invasion didn't
take advantage of gimp-babl utility functions and constructed
formats manually. They all needed revisiting and many now use much
simpler code calling gimp-babl API.
- change gimp_babl_format_get_color_profile() to really extract a
newly allocated color profile from the format, and add
gimp_babl_get_builtin_color_profile() which does the same as
gimp_babl_format_get_color_profile() did before. Visited all callers
to decide whether they are looking for the format's actual profile,
or for one of the builtin profiles, simplifying code that only needs
builtin profiles.
- drawables have a new get_space_api(), get_linear() is now get_trc().
- images now have a "layer space" and an API to get it,
gimp_image_get_layer_format() returns formats in that space.
- an image's layer space is created from the image's color profile,
change gimpimage-color-profile to deal with that correctly
- change many babl_format() calls to babl_format_with_space() and take
the space from passed formats or drawables
- add function gimp_layer_fix_format_space() which replaces the
layer's buffer with one that has the image's layer format, but
doesn't change pixel values
- use gimp_layer_fix_format_space() to make sure layers loaded from
XCF and created by plug-ins have the right space when added to the
image, because it's impossible to always assign the right space upon
layer creation
- "assign color profile" and "discard color profile" now require use
of gimp_layer_fix_format_space() too because the profile is now
embedded in all formats via the space. Add
gimp_image_assign_color_profile() which does all that and call it
instead of a simple gimp_image_set_color_profile(), also from the
PDB set-color-profile functions, which are essentially "assign" and
"discard" calls.
- generally, make sure a new image's color profile is set before
adding layers to it, gimp_image_set_color_profile() is more than
before considered know-what-you-are-doing API.
- take special precaution in all places that call
gimp_drawable_convert_type(), we now must pass a new_profile from
all callers that convert layers within the same image (such as
image_convert_type, image_convert_precision), because the layer's
new space can't be determined from the image's layer format during
the call.
- change all "linear" properties to "trc", in all config objects like
for levels and curves, in the histogram, in the widgets. This results
in some GUI that now has three choices instead of two.
TODO: we might want to reduce that back to two later.
- keep "linear" boolean properties around as compat if needed for file
pasring, but always convert the parsed parsed boolean to
GimpTRCType.
- TODO: the image's "enable color management" switch is currently
broken, will fix that in another commit.
Implement GimpPickable::get_pixel_average(), added in the previous
commit, in GimpDrawable, GimpImage, and GimpProjection, using
gimp_gegl_average_color(), added in the commit before last. This
is significantly faster than the default implementation.
When adding a rectangle to a projection's update area, align the
rectangle to a coarse grid, to reduce the complexity of the overall
area. We currently align the rectangle to a 32x32 grid, which
seems to be a good tradeoff between the overhead of processing a
complex area, and the overhead of processing a large area.
... which control the priority of the projection's idle source.
The projection's priority is specified relatively to
GIMP_PRIORITY_PROJECTION_IDLE (i.e., a priority of 1 results in an
idle source with priority GIMP_PRIORITY_PROJECTION_IDLE + 1, etc.)
More than 2000 lines of code less in app/, instead of
if (instance->member)
{
g_object_unref/g_free/g_whatever (instance->member);
instance->member = NULL;
}
we now simply use
g_clear_object/pointer (&instance->member);
which isn't really for "picking", but it just fits too nicely into
GimpPickable to not put it there.
Also add utility function gimp_pickable_srgb_to_image_color() which
takes a "real" (sRGB) GimpRGB value, transforms it to the pickable's
colorspace and puts it into an "image color" GimpRGB.
...when a color profile is active
Add GimpPickable::pixel_to_srgb() which puts a picked raw image
pixel into a GimpRGB. Default to gimp_rgba_set_pixel() but implement
pixel_to_srgb() in GimpLayer, GimpProjection and GimpImage and
run the pixel through gimp_image_color_profile_pixel_to_srgb().
which currently all end in a call to gimp_color_managed_get_color_profile()
except for channels and masks. This is currently unused infrastructure but
will be used for things like layer previews, and return NULL if called
on a mask or channel, or if color management is disabled, or whatever.
Add virtual function validate() so subclasses can construct arbitrary
buffers on-the-fly. The default implementation blits from the
projection graph like before. Add boolean property "whole-tile" which
allows for switching between always validating entire tiles, and
validating the parts of the tile that are actually dirty.
Allow for more tiles to update, making it possible to see more of the updated
image when tuning paremters / dragging items interactively. Lower frame rate
for more image content is better than 60fps for a small rectangle in the upper
left.
In GimpProjection, use gimp_rectangle_intersect() instead of ad-hoc
CLAMP() constructs to sanitize rectangles, and don't let rectangles
enter the projection if they don't intersect with the projectable.
(apparently I have killed some check that did the same when porting
to cairo regions).
Pull freeing the update regions into gimp_projection_free_buffer()
because freeing the buffer definitely obsoletes them. Get rid of
the last x1, y1, x2, y2 and use x, y, w, h inatead.
Remove member "running" it was always identical to "idle_id != 0".
Rename and reorder variables to make more sense. Simplify
gimp_projection_chunk_render_iteration() work area calculation.
This may or may not remove some logic that avoids drawing tiny update
regions, and may or may not improve things or make them worse. Will
add code that actually tile-aligns update areas later.
- don't include <gdk-pixbuf/gdk-pixbuf.h> in headers in app/
- instead, include it in many .c files instead of <glib-object.h>,
finally acknowledging the fact that app/ depends on gdk-pixbuf almost
globally
- fix up includes as if libgimpbase depended in GIO, which it soon will
removing the area that is going to be rendered from the dirty area
before rendering it, prevents tile-handler-projection to render it
again.
This was the reason why caching showed an improvement also during
the construction of the projection. And it was a regression that
I introduced in a previous commit
when toggling the visibility of a single layer image with
alpha channel, zoomed at 50%, the projection wasn't
properly redrawn. It seems there's a GEGL bug exposed
by normal layer mode operation fast path.
measure the time it takes to render projection chunks and continue
rendering until 0.01 seconds have passed. This ways we avoid excessive
expose roundtrips.
