Open Bug 807188 Opened 12 years ago Updated 2 years ago

IonMonkey: Fix v8-richards performance

Categories

(Core :: JavaScript Engine, defect)

defect

Tracking

()

People

(Reporter: dvander, Unassigned)

References

(Depends on 3 open bugs, Blocks 1 open bug)

Details

(Whiteboard: [ion:p1])

IonMonkey's score on v8-richards is roughly 9600 to Crankshaft's 14800. A performance improvement of ~%50 on this benchmark would raise our overall V8 score by ~5.5%.
Whiteboard: [ion:p1]
Depends on: 824249
Depends on: 824257
Depends on: 834614
No longer blocks: 864756
Depends on: 864756
Depends on: 894596
Depends on: 910960
Depends on: 911738
Depends on: 941760
According to the new Windows Machine on AWFY and my personal tests on my Win 7, Firefox Nightly is faster than Chrome 32 on Richards.
But on the Mac machines on AWFY, Chrome is faster by a big margin.
Maybe there is a Mac specific problem?
IIRC, Chrome is still 32-bit on OSX, while Firefox is 64-bit on OSX, which could be a factor.
But then the results from Machine 11 (32 bit) would be different from Machine 12 (64 bit), right?

Richards on Machine 11:
Chrome ~25400
Firefox GGC ~20400
Firefox noGGC ~22000

Richards on Machine 12:
Chrome ~24500
Firefox GGC ~17000
Firefox noGGC ~19500

Firefox is slower on both cases, though on Machine 12 the results are worse.
Assignee: general → nobody
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.