Closed Bug 1162678 Opened 10 years ago Closed 10 years ago

Memory regression for Homescreen

Categories

(Firefox OS Graveyard :: Gaia::Homescreen, defect)

ARM
Gonk (Firefox OS)
defect
Not set
normal

Tracking

(blocking-b2g:2.5+)

RESOLVED WORKSFORME
blocking-b2g 2.5+

People

(Reporter: gwagner, Unassigned)

References

Details

(Whiteboard: [systemsfe])

NAME v2.1 v2.2 master b2g 53.3 61.2 64.6 (Nuwa) 2.3 5.5 8.9 Homescreen 10.6 13.8 24.1 (Preallocated) 5.1 5.2 -
Kevin, can you take this one?
Flags: needinfo?(kgrandon)
Whiteboard: [systemsfe]
Eli, maybe we are lucky and we see a spike in our raptor measurements?
blocking-b2g: --- → 3.0+
Flags: needinfo?(eperelman)
Right now all apps are regressing memory along with launch time because of the pre-allocated process bug 1151672. Also, memory measurements right now are only captured for launch-able apps as the reboot test is not yet up on Raptor.
Flags: needinfo?(eperelman)
ftr, I don't think there are any significant changes in homescreen between 2.2 and master. Not entirely surprised to see an increase in 2.2, but I wouldn't have expected it to be very large... If we're looking for savings in 2.2, might be worth looking into - I think for 3.0 we'd be better off investing in getting rid of gaia-grid (which I'm working on).
Lets get some new numbers with bug 1156611 fixed.
(In reply to Gregor Wagner [:gwagner] from comment #5) > Lets get some new numbers with bug 1156611 fixed. Looks like bug 1156611 has landed on inbound. Can we get some new numbers after it gets merged/uplifted?
Flags: needinfo?(kgrandon)
Gabriele, can we close this bug now?
Flags: needinfo?(gsvelto)
Just tested on today's nightly build, USS is slightly below ~20MiB. It's definitely better but a far cry from where it used to be. Is that considered acceptable?
Flags: needinfo?(gsvelto)
Nick, whats the latest and greatest that can help us finding the memory regression here? Is comparing about:memory our best option?
Flags: needinfo?(n.nethercote)
Yes, diffing memory reports (view them in about:memory, use "Load and diff...") will give the most detailed data.
Flags: needinfo?(n.nethercote)
I've re-tested today and now the homescreen's USS drops to 10.8 MiB when minimized which is as good as we can expect. I'm not sure what's changed today but I've noticed that there's been some activity related to bug 1151672 so maybe that's what fixed the issue. I'd close this as WORKSFORME now.
\o/ Thanks Gabriele!
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.