Just click "start revive" to start the animation. It performs very bad. On the page the author says: "The animation is done using a simple sine table, and using a interval the layers are placed at the next position from the sine table."

Interesting. On Linux, I see the following in a jprof profile: 88% of time is spent in poll() 7% of time is spent in select() That would seem to point to bug 86632 or something along those lines. That should not be an issue on Windows, however. Is there a bigger issue here with timers in general?

Assuming the moveTo function sets style.left and style.top, I'm going to guess it's the same problem as Bug 64516, Bug 40988, and others. Removing the first moveTo line results in 33 - 60% CPU Usage. Removing both moveTo lines makes CPU Usage go down to 0. (Win2k PIII-733, Build 2001111303) The speed of style setting is something that's going make people stay with IE.

The testcase link is dead.

*** This bug has been marked as a duplicate of 21762 ***

if it's a dupe of bug #21762 the dependency on *that* bug should be removed. (I'd like to do that but have no rights).