Closed Bug 492694 Opened 16 years ago Closed 16 years ago

TM: Major discrepancy in v8's richards module when run in browser vs. shell

Categories

(Core :: JavaScript Engine, defect)

x86
Windows Vista
defect
Not set
normal

Tracking

()

RESOLVED FIXED

People

(Reporter: Natch, Unassigned)

References

Details

(Keywords: fixed1.9.1, Whiteboard: [fixed by bug 492673])

When I clear the browser cache and visit the v8 benchmarks suite online occassionally I'll get a 1500+ score on Richards, thereafter I consistently get mid-100s. In the js shell I always get above 1500. Is this maybe because of DOM function calls falling off trace or something? $ $TR/dist/bin/js -j run.js Richards: 1593 DeltaBlue: 155 Crypto: 149 RayTrace: 245 EarleyBoyer: 334
Flags: wanted1.9.1?
Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20090512 Minefield/3.6a1pre With the following STR I can reproduce 100% of the time in the browser: 1) Clear cache. 2) Restart browser. 3) Run benchmark suite. 4) Reload the suite. The results for the Richards module in the 3 step for me are mid-1500s while on the second and subsequent runs it will be about 150.
(In reply to comment #2) Yup, you're right on target, if I run Richards after DeltaBlue I get 150, since by default it runs first I was always getting 1500 in the shell: $ $TR/dist/bin/js -j run-deltaFirst.js DeltaBlue: 119 Richards: 181 Crypto: 152 RayTrace: 249 EarleyBoyer: 337
Just running the shell, Richards seems to be bimodal for me -- I usually get either 1600-ish or 1000-ish.
This was fixed by bug 492673, I get a very strong score on Richards now.
Status: NEW → RESOLVED
Closed: 16 years ago
Depends on: 492673
Flags: wanted1.9.1?
Keywords: fixed1.9.1
Resolution: --- → FIXED
Whiteboard: [fixed by bug 492673]
You need to log in before you can comment on or make changes to this bug.