[Webkit-unassigned] [Bug 43903] Speeding up SVG filters with multicore (SMP) support

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=43903





--- Comment #20 from Gabor Loki <loki at webkit.org>  2011-03-02 00:55:30 PST ---
> You can find another version of the benchmarks on gitorious: http://gitorious.org/qtwebkit/performance.

It looks like this performance tool has some issue running as a benchmark. I used several tools from this project, but they provided strange values to me.

I create a simple svg with only one filter (on a BIG area). It takes about 5 sec to compute the filter with QtWebKit. And here is a sample result of cycler test:

********* Start testing of tst_Cycler *********
Config: Using QTest library 4.7.0, Qt 4.7.0
PASS   : tst_Cycler::initTestCase()
PASS   : tst_Cycler::load()
PASS   : tst_Cycler::cleanupTestCase()
Totals: 3 passed, 0 failed, 0 skipped
********* Finished testing of tst_Cycler *********
Printing result for: total
    benchmark: cycler::load    http://127.0.0.1/~smp/svg/turbulence.svg
        mean: 3 msecs +/- 3239 msecs, +/- 107967 %
        avg:  1685 msecs
            5741 3 3 3 2 2 2 5464 5636 3 
Total mean: 3 msecs per test, avg: 1685


And another one:
        mean: 3 msecs +/- 3850 msecs, +/- 128333 %
        avg:  2312 msecs
            5875 5682 5875 5679 3 2 2 2 2 2 

Btw, using WebKit's threading, the results:
        mean: 2 msecs +/- 1588 msecs, +/- 79400 %
        avg:  676 msecs
            3408 3336 3 3 2 2 2 2 2 2 

I think there is some kind of caching issue. I feel this environment cannot be used as a benchmark now.

Benjamin, do you have any idea what is happening here? Why did the environment produce this strange numbers?

Btw, the derivation of the valid numbers are also too high.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.