Auto merge of #6365 - pcwalton:sequential-display-list-construction, r=larsbergstrom

layout: Disable parallel display list building by default.

I've never see it result in a speedup. Actually, I don't think I've seen
it result in anything better than a 50% slowdown. The arithmetic
intensity is just too low, at least with the current algorithm.

Parallel DL building can still be enabled with a debug flag if the
algorithm is improved.

r? @metajack

<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/servo/servo/6365)
<!-- Reviewable:end -->
This commit is contained in:
bors-servo 2015-07-06 07:08:58 -06:00
commit b876a54dce
2 changed files with 12 additions and 6 deletions

View file

@ -836,18 +836,18 @@ impl LayoutTask {
flow::mut_base(&mut **layout_root).clip =
ClippingRegion::from_rect(&data.page_clip_rect);
match rw_data.parallel_traversal {
None => {
sequential::build_display_list_for_subtree(layout_root,
shared_layout_context);
}
Some(ref mut traversal) => {
match (&mut rw_data.parallel_traversal, opts::get().parallel_display_list_building) {
(&mut Some(ref mut traversal), true) => {
parallel::build_display_list_for_subtree(layout_root,
self.profiler_metadata(),
self.time_profiler_chan.clone(),
shared_layout_context,
traversal);
}
_ => {
sequential::build_display_list_for_subtree(layout_root,
shared_layout_context);
}
}
if data.goal == ReflowGoal::ForDisplay {