[PeCaN]

I sort of wonder if this approach to JITing is worth it over just writing a faster interpreter. This is basically like what V8's baseline JIT used to be and they switched to an interpreter without that much of a performance hit (and there's still a lot of potential optimizations for their interpreter). LuaJIT 1's compiler was similar, although somewhat more elaborate, and yet still routinely beaten by LuaJIT 2's interpreter (to be fair LuaJIT 2's interpreter is an insane feat of engineering).

[htgb]

They detail why they chose the current path rather than continuing improvements of the interpreter in the previous post [1]. See the last few paragraphs for a summary.

[1] http://blog.erlang.org/a-closer-look-at-the-interpreter/