first-class.
This implements the scheme described here:
https://groups.google.com/forum/#!topic/mozilla.dev.servo/sZVPSfPVfkg
This commit changes Servo to generate one display list per stacking
context instead of one display list per layer. This is purely a
refactoring; there are no functional changes. Performance is essentially
the same as before. However, there should be numerous future benefits
that this is intended to allow for:
* It makes the code simpler to understand because the "new layer needed"
vs. "no new layer needed" code paths are more consolidated.
* It makes it easy to support CSS properties that did not fit into our
previous flat display list model (without unconditionally layerizing
them):
o `opacity` should be easy to support because the stacking context
provides the higher-level grouping of display items to which opacity
is to be applied.
o `transform` can be easily supported because the stacking context
provides a place to stash the transformation matrix. This has the side
benefit of nicely separating the transformation matrix from the
clipping regions.
* The `flatten` logic is now O(1) instead of O(n) and now only needs to
be invoked for pseudo-stacking contexts (right now: just floats),
instead of for every stacking context.
* Layers are now a proper tree instead of a flat list as far as layout
is concerned, bringing us closer to a production-quality
compositing/layers framework.
* This commit opens the door to incremental display list construction at
the level of stacking contexts.
Future performance improvements could come from optimizing allocation of
display list items, and, of course, incremental display list
construction.
I'm sad to say that this improved performance significantly. A lot of
this win is due to the Rust compiler not being smart about not zeroing
objects out if it doesn't need to.
One part (of 8!) of css font family disambiguation is that font families should
be matched case-insensitively.
This patch implements that. Once it lands, a bug needs to be filed to do lowercasing
properly (as a string, instead of char-by-char -- it's a unicode thing).
r? @gw
During debugging, I found it useful to hook all task creation in a
central location, and util::task was the perfect place for it.
r? @pcwalton (or maybe someone else, I'm kinda sending you a bunch of
reviews today because I don't know who better to give them to)
We've discussed this some and I think there's consensus to do it as a
pragmatic decision for now. CPU painting is more stable, especially with
buggy drivers, and faster (because we aren't caching the necessary
OpenGL objects yet and possibly for other reasons), so it provides a
better "out of the box" experience for newcomers to Servo who don't know
to pass the `-c` option. This patch continues to reftest both Skia and
Skia-GL out of a desire to keep options open. Skia-GL remains a
first-class citizen.
r? @metajack
We've discussed this some and I think there's consensus to do it as a
pragmatic decision for now. CPU painting is more stable, especially with
buggy drivers, and faster (because we aren't caching the necessary
OpenGL objects yet and possibly for other reasons), so it provides a
better "out of the box" experience for newcomers to Servo who don't know
to pass the `-c` option. This patch continues to reftest both Skia and
Skia-GL out of a desire to keep options open. Skia-GL remains a
first-class citizen.
This is a grab bag of performance improvements that significantly improve style recalculation, layout, and painting on a few static pages.
Let me know if you'd like me to split this PR up.
r? @glennw
construction to avoid cloning and moving flows so much.
Besides amounting to a 5%-10% win on a page with a lot of text, this
simplifies and refactors the text layout code.
a separate `ClipDisplayItem`.
We push down clipping areas during absolute position calculation. This
makes display items into a flat list, improving cache locality. It
dramatically simplifies the code all around.
Because we need to push down clip rects even for absolutely-positioned
children of non-absolutely-positioned flows, this patch alters the
parallel traversal to compute absolute positions for
absolutely-positioned children at the same time it computes absolute
positions for other children. This doesn't seem to break anything either
in theory (since the overall order remains correct) or in practice. It
simplifies the parallel traversal code quite a bit.
See the relevant Gecko bug:
https://bugzilla.mozilla.org/show_bug.cgi?id=615734
Previously, any font template that could not be found would return
an item from the last resort font family. Now, the last resort font
family is only searched if the entire list of supplied font families
is empty.
This patch puts in the initial framework for incremental reflow. Nodes' styles
are no longer recalculated unless the node has changed.
I've been hacking on the general problem of incremental reflow for the past
couple weeks, and I've yet to get a full implementation that actually passes all
the reftests + wikipedia + cnn. Therefore, I'm going to try to land the different
parts of it one by one.
This patch only does incremental style recalc, without incremental flow
construction, inline-size bubbling, reflow, or display lists. Those will be coming
in that order as I finish them.
At least with this strategy, I can land a working version of incremental reflow,
even if not yet complete.
r? @pcwalton
The page_rect passed to DisplayListOptimizer is relative to the
RenderLayer origin, but the display list components are relative to the
page origin. Before passing the page rect to the display list, we
translate it by the RenderLayer position.
The page_rect passed to DisplayListOptimizer is relative to the
RenderLayer origin, but the display list components are relative to the
page origin. Before passing the page rect to the display list, we
translate it by the RenderLayer position.
