@schmee it's close, but different in some rather critical ways. In Truffle you create an AST, and then Truffle merges the AST execute methods into a single large method. In PyPy you write an interpreter in a subset of Python, then the framework analyzes the interpreter and finds you interpreter dispatch loop (via a special hint). From there it constructs a JIT for your interpreter. So Truffle requires you to compile your code to an AST, PyPy figures out how your interpreter works and builds a JIT for however you chose to write your interpreter.

So PyPy is still a bytecode interpreter. While Zippy (Python on Truffle) is a python to AST compiler with a JIT backend.

The cool thing about PyPy's approach is that PyPy still looks like a mostly normal Python interpreter. You can still introspect stack frames, construct bytecode on-the-fly, etc.

A good overview of that whole process: https://morepypy.blogspot.com/2011/04/tutorial-part-2-adding-jit.html

@tbaldridge I see, thanks! so know I have to read up on tracing JITs, just when I thought I was getting some sort of understanding on this topic 😂

oh, this is just what I needed! http://stefan-marr.de/papers/oopsla-marr-ducasse-meta-tracing-vs-partial-evaluation

so PyPy = tracing, Truffle = partial evaluation

Hey Everyone , Another question - How do you discover open source projects ? what are your sources and is there any glitch or gap in open source discovery? Pls reply on thread, it's easy to track