This is exactly what I wanted to post. My links cover all the spectrum of tests employed, not just tests that are favorable to Flash. GuiMark3, for example, found that HTML5 rendering of video in high resolutions on mobile devices tend to be faster than Flash video too, depending on version. I'm a defender of pragmatism, not of Flash, and it drives me mad to see people saying things that are just not true, like the previous poster.

My point is that if one analyzes the data, one will find that Flash tend to win over HTML5/canvas performance more often than not. The canvas and visual performance tests are very telling.

As a side note, there's definitely a gap in code execution - that's the link you posted - but it's just one of the factors, as Adobe is well aware of that and is addressing it on the next release. The AS3 VM was much faster than JS when introduced, then they spent a lot of time not doing any optimization at all. From what I've seen from their presentations, they're ready to take that back. Time will tell.