Musings on Tracing in PyPy

I feel like LuaJIT probably deserves a mention here. It doesn't always get astounding performance, but it very often does, and you get all the inlining and dynamic specialization goodness that tracing JITs make cheap. And it demonstrates extremely convincingly that PyPy's difficulty with invoking native code isn't essential to tracing JIT.

Nice article. I'd like to also know PyPy developers thoughts on the copy-and-patch approach chosen to implement the new JIT under development for CPython.