Closed Bug 621581 Opened 14 years ago Closed 11 years ago

On the "MoonBat" Web Worker modification of SunSpider, JM is much slower than V8

Categories

(Core :: JavaScript Engine, defect)

x86
Linux
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE

People

(Reporter: pcwalton, Unassigned)

References

()

Details

Attachments

(1 file)

On this workerized version of SunSpider [1], I get 3193 ms on JM, but 1351 ms in Chrome. That's over 230% slower. On regular SunSpider, JM beats Chrome handily.

Apologies if this is misfiled; I'm not sure whether Web Worker issues should go here.

[1]: http://www.yafla.com/dforbes/resources/moonbat/moonbat-driver.html
Depends on: 584860, 615723
Per bug 615723, JM isn't running at all in the worker... so you're just measuring interp plus a bit of broken TM.
On different hardware (MacBook Pro), I now have:

Chrome:
Elapsed:               4492.0ms
Total:                 341.0ms +/- 9.0%

Firefox 4:
Elapsed:               2791.0ms
Total:                 461.6ms +/- 14.5%

So Fx4 is much faster at spawning Web Workers on the Mac, but is still 35% slower than Chrome.
No longer depends on: 584860
No longer depends on: 615723
Depends on: 584860
Depends on: 615723
Hmm.  So what are the "elapsed" and "total" times there?
Attached is sunspider, runnable in a worker and normally. run_the_benchmark.js just runs the benchmark, run_a_worker.js runs the former file in a worker.

I don't see a performance difference between the two. So the issue doesn't appear to happen in the shell, looks like it's a problem just in the browser.
With bug 649537 I see these results:

 - 5 threads running 10 cycles per test.

 - Chrome:
   Elapsed:               3185.0ms
   Total:                  948.7ms +/- 20.6%

 - Nightly:
   Elapsed:               2310.0ms
   Total:                 1048.4ms +/- 19.9%

What do they mean? I don't see how we can be faster overall but have a bigger number for 'Total'...
The benchmark is gone and the only copy I can find (http://nerget.com/bugs/worker-sunspider/moonbat-driver.html) doesn't work anymore.

We might or might not have solved this, but there probably are better ways to track our workers performance than dead benchmarks.
Status: NEW → RESOLVED
Closed: 11 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: