Hello Dave, Thanks for your input. We took some time to check that and we can not confirm your findings.
> FYI to the devs: this method is now consuming a nontrivial amount of time > under IE8. If I create 180 objects in IE8, the profiler shows toHashCode > using 1200 ms. That would be pretty bad and we would have realized that already in out nightly testing system. > Yes, that is a lot of objects. 180 Objects are not that much at all. > But assigning hash codes > shouldn't take longer than creating the objects themselves. :) I see a > __postId suffix was added in QooxDoo 1.5, which makes this even slower. Yes, that was added because the hash code needs to be unique among all frames. The postfix is an id for the current frame. > Would it not suffice to just use a continuously incrementing number as the > object hash code? That what we had before that and that wasn't unique in all frames. > The current code uses an array cache; arrays are not only > slow in IE8, but they are notorious memory hogs. Can you maybe show us how you tested it? Here are our results which are we tested in IE8: Trunk: 007905 qx.test.performance.Object[42-0]: create qx.core.Object; Iterations: 100000; Time: 4422ms; Render time: 15ms 019216 qx.test.performance.Object[42-0]: toHashCode qx.core.Object; Iterations: 100000; Time: 1827ms; Render time: 16ms 1.4 021669 qx.test.performance.Object[42]: create qx.core.Object; 100000; 12530; 15 036089 qx.test.performance.Object[42]: toHashCode qx.core.Object; 100000; 1812; 16 As you can see, we tested 100000 objects and the method is not that much slower than what you describe. Compared between the two releases, its rather quite the opposite. Regards, Martin ------------------------------------------------------------------------------ BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA http://p.sf.net/sfu/rim-devcon-copy2 _______________________________________________ qooxdoo-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/qooxdoo-devel
