Closed Bug 1689694 Opened 5 years ago Closed 5 years ago

Fission visual metrics regression on Mac for Reddit

Categories

(Core :: Performance: General, defect, P2)

defect

Tracking

()

RESOLVED FIXED
Fission Milestone M7

People

(Reporter: jesup, Assigned: jesup)

References

Details

(Whiteboard: fission-perf)

You can see a ~15-20% visualmetric regression for reddit on MacOS here: https://vismet.arewefastyet.com/mac/cold-page-load/overview?numDays=60

Profiles (2018 MBP):
e10s: https://share.firefox.dev/3iDIQq0
fission: https://share.firefox.dev/3o0iWhe
These are synced and each a 400ms slice of data at the start of the load.

You can see the visual difference (e10s starts painting the page; fission doesn't even start until about 50-100ms after this). The question is why is fission taking longer here? this is reddit.com (the TP6 reddit link, though loaded live); reddit shows an about 15-20% perf regression vs e10s on mac; on linux we see only 3-5% regression. (Separate point of interest: CSI shows a far smaller regression than PSI and SI on mac; not sure as to why)

Blocks: 1689695
Blocks: 1689699
Blocks: 1689702
No longer blocks: 1689695, 1689699, 1689702

Let's target fixing this in M7.

Assignee: nobody → rjesup
Severity: -- → S3
Status: NEW → ASSIGNED
Fission Milestone: ? → M7
Priority: -- → P2

This is tracked as part of Fission, so I don't think tracking this also as a qf bug is particularly useful.

Whiteboard: [qf] fission-perf → fission-perf

With the non-native theming turned on for mac, and the preallocation delay patch (bug 1694036), we're now apparently within 0-5% (depending on CSI (~0%), PSI & SI ~5%) of E10s. We may want to close this bug, though perhaps there is still something to look at.

Status: ASSIGNED → RESOLVED
Closed: 5 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.