Closed Bug 1181074 Opened 9 years ago Closed 6 years ago

Sunspider is ~15% slower in e10s compared to e10s off on Linux

Categories

(Core :: JavaScript Engine, defect, P1)

x86_64
Windows 7
defect

Tracking

()

RESOLVED WONTFIX
Tracking Status
e10s + ---

People

(Reporter: elbart, Assigned: h4writer, NeedInfo)

References

(Blocks 1 open bug, )

Details

e10s on
688.2ms +/- 3.2%
Results: http://tinyurl.com/ss-e10s-on

e10s off (or "New non-e10s window")
591.0ms +/- 3.6%
Results: http://tinyurl.com/ss-e10s-off

kraken and octane are about the same.
Blocks: e10s-perf
On windows 7, I see us doing better now - 

non-e10s:
Total:                 243.9ms +/- 6.8%

e10s:
Total:                 211.4ms +/- 5.0%

Tracy, mind confirming? We should test on linux as well.
Flags: needinfo?(twalker)
Mac 10.9.5:

non-e10s:
Total:                 165.9ms +/- 7.4%

e10s:
Total:                 167.5ms +/- 2.9%

----------------------------------------

Linux (Ubuntu 64-bit 14.0.4.2 VM)

non-e10s:
Total:                 218.4ms +/- 4.7%

e10s:
Total:                 283.1ms +/- 5.5%

-----------------------------------------

still not so good on Linux.
Flags: needinfo?(twalker)
Summary: Sunspider is ~15% slower in e10s compared to e10s off → Sunspider is ~15% slower in e10s compared to e10s off on Linux
Status: UNCONFIRMED → NEW
Ever confirmed: true
Naveed, any idea why Sunspider could regress with e10s?
Flags: needinfo?(nihsanullah)
Priority: -- → P1
Hannes please take a look at this. I think we may need an E10S browser build on AWFY as well. I wouldn't have expected E10S to impact a micro benchmark.
Assignee: nobody → hv1989
Flags: needinfo?(nihsanullah)
Hi Hannes,
Any progress here? I think it would be useful to at least get a sense of what's going on before doing anything on AWFY. Maybe the GCs are happening at different times?
Flags: needinfo?(hv1989)
(In reply to Naveed Ihsanullah [:naveed] from comment #4)
> Hannes please take a look at this. I think we may need an E10S browser build
> on AWFY as well. I wouldn't have expected E10S to impact a micro benchmark.

Oh I was thinking I missed a needinfo, which I get mail about. Seems I don't get mail on getting assigned to a bug.
Bill thanks to needinfo me. For the mozbench/awfy integration I'm actually just in the progress of having non-e10s builds on AWFY. Which will be handy to view our progress.

I'll try to take a look next week into this. Might indeed be GC related.
Component: General → JavaScript Engine
Product: Firefox → Core
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.