Every test created is public, and listed there. So anybody can test it, in any platform or browser and all test results are keept. So as more people runs each test, more analysis data we have.
To see how it works, and also to answer one question regarding the performance of drawing text with shadow on canvas, I have setup a test on jsperf.
As I only tested on Chrome and Firefox what I can say for now is:
On Chrome 18.0 the performance dorps by almost half when drawing texts with shadows.
On Firefox 12.0 is the same thing. Doesn't matter it keep the same performance, but comparing the results Chrome is almost 50% faster than Firefox when drawing texts without shadows.
Remember that it is only a test and I have only test it on my computer.