As we already blogged, in the last half-year or so, Michael Leuschel, Armin and me did a lot of JIT generator work on a Prolog prototype. The idea was to experiment more quickly with some techniques than what would have been possible with RPython. These experiments were quite successful in themselves. With very little code we managed to get a JIT that is not doing too badly when compared to existing projects for Prolog.
This Prolog work was also the subject of my Master's thesis. I finished the thesis about two weeks ago (and since then have been mostly sleeping and then sprinting). The thesis should be self-contained when it comes to explaining the JIT concepts but needs knowledge of Prolog to be understandable.