Apparently his work has made it into GHC (the main Haskell compiler) as well, though he didn't elaborate on it. But at the beginning of his talk he said there were two versions of the talk: one in Haskell, and one in a more Java-like language, and so he chose the latter. I was sad.
I really ought to study concurrency in Haskell because it ought to be interesting. With functions without effects you never need a write lock; but at the same time, the reason you'd want to lock something is only when you mutate it, which means you do need some sort of locking, even in effect-free functions.
In any case, there are a number of interesting-looking papers on his site. In particular, "A Practical Multi-Word Compare-and-Swap Operation" at a glance looks like the seed of the implementation of the effect-log technique I described above.