Post from the WebKit team explaining how they've added in LLVM based code generation to their JIT.
Rather than use LLVM to do all code generation, it is used as a forth tier of their JIT architecture, turning the static type-inferred analysis from the lower tiers in to highly optimised, high performance code.
Rather than use LLVM to do all code generation, it is used as a forth tier of their JIT architecture, turning the static type-inferred analysis from the lower tiers in to highly optimised, high performance code.