the JIT hasn't been used for a while now, and I know it's being cleaned up currently, but has anyone considered the merits of using NanoJIT? It's small, has no real dependencies, and was made for flash... LLVM seems overly large and complex, and results so far haven't been good (even for similar projects like Unladen Swallow for python).
Before a serious effort is made into the LLVM jit we should consider NanoJIT
Or it can be developed on a branch regardless of what's going on with the LLVM one, so we can compare both and pick the better one :)
Exactly, it should be no problem (even in the same branch) to develop them alongside and compare them when both are implemented. Feel free to do that!
And personally I feel that there has been already some work into the LLVM JIT. It uses some good some optimization opportunities and already covers all opcodes. Additionally, I have the feeling that LLVM will see some very nice improvements with the upcoming version 3.0.