This partially reverts 2edbda2526.
Using cairo-groups increases performance on MacOS, and retains
retina-resolution.
However it adds a performance regression for MS Windows graphics
rendering. cairo-groups use a "similar" surface, not an image surface.
Empirically this adds significant overhead compared to rendering
using the CPU and using bitblt.
For MacOS/X this is equivalent, rendering happens using a
CGBitmapContext + image-surface. Windows and Linux needs profiling
for respective equivalent surfaces.
This is an intermediate commit, before replacing image surfaces with
cairo pattern groups.
The eventual goal is to reduce flickering and/or use
CPU + bitblt for specific widgets instead of cairo
graphics-cards accel.
This also removes excessive calls to getenv() for every rendering
operation.
Generated by tools/f2s. Some hand-editing will be required in a few places to fix up comments related to timecode
and video in order to keep the legible
The drawing itself should be unchanged but much of the rest of the
implementation has changed. The WaveViewThreads and WaveViewDrawingThread
classes were added and allow multiple drawing threads.
The Item::prepare_for_render interface is implemented by WaveView to enable
queuing draw requests for the drawing threads to process as soon as the state
change occurs during Editor::visual_changer, which often means the images will
be finished by the time they are needed in WaveView::render. This can
significantly reduce total render time and also flickering caused by images not
being ready for display.
If the drawing thread/s cannot finish the request by the time it is required in
WaveView::render then cancel it and draw the WaveViewImage in the GUI thread if
it is likely it can be completed in the current render pass/frame. This change
also helps reduce the flickering caused by images not being ready with threaded
rendering, but with several drawing threads, drawing in the GUI thread may not
often occur (unless explicitly requested).
Allow unfinished images to be returned from the cache in
WaveView::prepare_for_render so that new draw requests aren't queued for
duplicate images. This reduces the amount of drawing for instance in
compositions where there are many instances of the same sample/waveform
displayed on the canvas as only a single image should be drawn.
Use a random width within a certain range for
WaveView::optimal_image_width_samples so that image drawing is less likely to
occur at the same time (which will cause a spike in render/draw time and
increase the chance of flickering waveforms).
Move implementations of the private WaveView classes into wave_view_private.h
and wave_view_private.cc source files.
Incorporate a fix for limiting the waveview image size to the cairo image size
limit.
Should hopefully Resolve: #6478
Called when an item has requested a redraw and intersects with visible
canvas area.
Also add Canvas::prepare_for_render that will call Item::prepare_for_render for
items visible on the canvas.
With large sample-rates and high-zoom-level the minimal request-size
or at least 1/10sed can exceed the max image size supported by cairo.
e.g. @192kSPS; 100ms = 19200 samples.
* 1 sample/pixel, * 2 (left/right) = 38400px > 2^15px.
Various GdkEvents are not yet handled correctly, eg. unpacking a widget
from its container does not unmap it, nor are remaining widgets in the
contained re-positioned (size allocation does not change, nor does the
mapping). This affects eg. Mixbus Strips
This fixes an -Woverloaded-virtual ambiguity introduced in b5e613d45
void render (cairo_t*, cairo_rectagle*)
void render (Cairo::RefPtr<Cairo::Context> const&, cairo_rectangle_t*)
ArdourCanvas prefers cairomm and CairoWidget itself uses Cairo::Context,
this improves overall API consistency.
