Bobby Holley (:bholley) Assignee
	Description • 7 years ago

We've been meaning to fix this for a while, and Julian's profiling raised this as an enormous source of transient allocations.

	Bobby Holley (:bholley) Assignee
	Comment 2 • 7 years ago

We could considering shrinking this to 8, since julian's testing found that we basically never write to the upper 8 slots on the obama testcase. On the other hand, it may depend a lot on the stylesheet.

	Emilio Cobos Álvarez (:emilio)
	Comment 3 • 7 years ago

Comment on attachment 8867769 [details] [diff] [review]
Use a SmallVec when gathering applicable declarations. v1

Review of attachment 8867769 [details] [diff] [review]:
-----------------------------------------------------------------

::: servo/components/style/matching.rs
@@ +865,5 @@
>      }
>  }
>  
>  fn compute_rule_node<E: TElement>(rule_tree: &RuleTree,
> +                                  applicable_declarations: &mut SmallVec<[ApplicableDeclarationBlock; 16]>,

Could we perhaps make this a typedef?

type ApplicableDeclarationsList = SmallVec<[ApplicableDeclarationBlock; 16]>;

(Perhaps with a comment or something re. the reasoning and the allocator measurements).

Also, perhaps it's worth to change the one in lazy_pseudo_rules, what do you think?

	Bobby Holley (:bholley) Assignee
	Comment 4 • 7 years ago

Sure, will do on both counts.

	Julian Seward [:jseward]
	Comment 6 • 7 years ago

Is there a way to get hold of the patch that is associated with the NEW->RESOLVED
transition above?  It doesn't seem to be in m-i right now.

	Bobby Holley (:bholley) Assignee
	Comment 7 • 7 years ago

Sorry, I realized I never pasted the PR in the bug. Normally I do that: https://github.com/servo/servo/pull/16874

(1): pull autoland
(2): git log
(3): search in your pager for 16874
(4) figure out whether that commit is in central by the same procedure.

Appears not to have merged yet, in this case.