Javascript performance for QML is much worse than javascript performance within the WebView in QML.

The performance of the javascript version of fbench by John Walker in QML is significantly worse than the same benchmark in a WebView within QML. One would expect both to be roughly equivalent.

Also, the performance degrades very quickly with each run if line 283 of the QML version of fbench.js. This line declares a variable that is used in a loop and is marked with a comment.

The following are the results from my WindowsXP machine with a snapshot from gitorious from around 17 Feb 2010 compiled with Microsoft Visual Studio 2008 in release mode with 5000 iterations.

QML
------------------
Elapsed time in seconds: 10.110
Time for 1000 iterations: 2.0220
No errors in results.
Elapsed time in seconds: 10.207
Time for 1000 iterations: 2.0414
No errors in results.
Elapsed time in seconds: 10.268
Time for 1000 iterations: 2.0536
No errors in results.

WebView in QML
----------------------------
Elapsed time in seconds: 0.226
Time for 1000 iterations: 0.0452
No errors in results.
Elapsed time in seconds: 0.227
Time for 1000 iterations: 0.0454
No errors in results.
Elapsed time in seconds: 0.228
Time for 1000 iterations: 0.0456
No errors in results.

Also, one more datapoint is Safari 4.0.4 (531.21.10) on the same machine
-------------------------------------------------------------------------------------------------------
Elapsed time in seconds: 0.061
Time for 1000 iterations: 0.0122
No errors in results.
Elapsed time in seconds: 0.061
Time for 1000 iterations: 0.0122
No errors in results.
Elapsed time in seconds: 0.060
Time for 1000 iterations: 0.0120
No errors in results.