On Thu, 05 Sep 2013 14:14:34 +1000, Chris Angelico wrote: > ...Python's object model tends to be less well suited to massive > scaling;
That's probably true. > I don't know about PyPy and other Pythons, but certainly > CPython isn't designed to run on arbitrary numbers of cores (once you go > to multiprocessing, you then need to worry about IPC; if you work in a > lower level language like C, you can use threads and directly access > each other's memory). I think that's an exaggeration. CPython *is* designed to run on an arbitrary number of cores, but you need to approach it via techniques that you might not use in other languages. It would only be valid to say that "CPython is not designed to use multiple cores" if threads were the *only* valid way to use multiple cores. "Use multiprocessing" is just as much a valid way to use multiple cores as "use threads" might be in another language, and by some accounts, better than threads. Or you can use IronPython or Jython, neither of which have the GIL. Or use Stackless: http://entitycrisis.blogspot.com.au/2009/06/stackless-vs-gil-its-draw.html -- Steven -- https://mail.python.org/mailman/listinfo/python-list
