Closed Bug 894921 Opened 8 years ago Closed 7 years ago

dromaeo-dom regression from not immediately deoptimizing typesets on string SETELEM

Categories

(Core :: JavaScript Engine, defect)

x86
macOS
defect
Not set
normal

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: bzbarsky, Unassigned)

References

Details

Attachments

(3 files)

STEPS TO REPRODUCE:

1)  Load http://dromaeo.com/?dom-attr&numTests=1
2)  Click Run button
3)  Observe the score reported for the "element.expando = value" test.
4)  Back out bug 893897.
5)  Run test again, see that it's a good bit faster.
Oh, the actual source of the test is at http://dromaeo.com/tests/dom-attr.html
Brian, any idea why the change in bug 893897 caused issues here?
Flags: needinfo?(bhackett1024)
(In reply to Boris Zbarsky [:bz] from comment #2)
> Brian, any idea why the change in bug 893897 caused issues here?

Not offhand, does the JIT inspector show any difference in the code generated for that test?
Flags: needinfo?(bhackett1024)
No obvious difference is jumping out at me, though the old code (before bug 893897) seems to show two recompiles while the new code only shows one?

I'll attach JIT inspector dumps and a small standalone test.
Attached file Testcase
Note that sometimes I do get two older compilations here, not just 1....
Also, I can't seem to reproduce this in shell. In a browser, on the attached testcase I reliably see about 210 alerted on tip, and about 180 with bug 893897 backed out.

In shell I see about 130 either way, with the <a> replaced by just "var a = {};".
Assignee: general → nobody
Nightly x Chrome 42
On the attached testcase I get 350 x 240 
On Dromaeo I get 242 runs/s x 123 runs/s

I suppose this is fixed, so setting WFM.
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.