Open
Bug 1482834
Opened 6 years ago
Updated 2 years ago
Improve the Performance Counter in nsThread
Categories
(Toolkit :: Performance Monitoring, enhancement)
Toolkit
Performance Monitoring
Tracking
()
NEW
People
(Reporter: tarek, Unassigned)
Details
In bug 1477755 we have implemented an ad-hoc counter for timers because in the current design, we are unable to track it. The way we track counters is two-fold: - in DocGroup or WorkerThread, we increment the counter when the event is dispatched (so we know the TaskCategory) - in nsThread::ProcessNextEvent, we call GetPerformanceCounter() to get back to the counter, wheter it's located in a DocGroup or a Worker For timers, when a setTimeout or setInterval is called, a new entry is directly added in the ThrottledEventQueue and everytime the timeout occurs, a new nsTimer runnable is triggered via a TimeoutExecutor runnable that runs all the time. These level of indirections makes it imposible right now to cral back to the counter unless we make all layers aware of this. Since this part of the code will be refactored for Project Fission, we took the decision to make an ad-hoc counter and wait for implementing a better generic design, instead of changing many *Timer* classes.
Updated•2 years ago
|
Severity: normal → S3
You need to log in
before you can comment on or make changes to this bug.
Description
•