Closed Bug 1333701 Opened 9 years ago Closed 9 years ago

stylo: Tweak the memory orderings when accessing the rule tree

Categories

(Core :: CSS Parsing and Computation, defect, P2)

Product:
Core
Component:
CSS Parsing and Computation
Type:
defect

Tracking

()

Status:
RESOLVED FIXED

People

(Reporter: emilio, Assigned: emilio)

References

(Blocks 1 open bug)

Details

Just filling this because Bobby was looking for potential performance improvements and this is low-hanging fruit. When I landed the rule tree I intentionally used SeqConst atomic ordering almost everywhere. We should be able to improve on that a lot though (if only, refcount bumps and drops should be relaxed). This can be benchmarked with ./mach test-unit -p style --bench in the Servo repo.
Emilio has patches in https://github.com/servo/servo/pull/15562 but review comments still need addressing.
Priority: -- → P2
This landed. Thanks Emilio!
Assignee: nobody → emilio+bugs
On my machine, this improved the performance on bench_insertion_basic_per_element by about 10% (420 -> 380). The rest of the benchmarks didn't seem to move much. It probably depends a lot on the workload.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.