On Dec 3, 2007 12:12 PM, "Martin v. Löwis" <[EMAIL PROTECTED]> wrote: > >> Sure. However, you could also use _testcapi.PY_SSIZE_T_MAX, or > >> ctypes.sizeof(ctypes.py_object). > > > > Much less intuitive though. > > Actually, I find the ctypes version a more direct answer to the > question "what is the address space?".
Also the more correct answer. :-) It's just that I didn't even know about it -- and I suspect most users would never think of it. OTOH most users will expect sys.maxint to be 2**63-1 on a 64-bit machine (never mind that they are wrong for Win64 :-). > >> I'd be in favor of the latter. I never meant the PyInt_CheckExact hack > >> to persist into the release. > > > > OK. What's needed? > > The remaining occurrences of PyInt_CheckExact must be replaced with > PyLong_Check, and the subsequent PyLong_AsLong calls must check for > overflow. There are 22 uses still left. The ones in ceval.c are > quite tricky: it's not clear that the speed-up they originally meant > to provide is still present. The options would be to either continue > special-casing small integers, or to remove the check altogether, > trusting that the binary operator is fast enough (which may actually > involve fewer function calls than the "inline" version). I would remove the optimizations for now; we should benchmark again later in the 3.0 beta release cycle. In general I'm all for removing clutter that was once added in the hope of speeding things up; for 3.0 my main priority is correctness. -- --Guido van Rossum (home page: http://www.python.org/~guido/) _______________________________________________ Python-3000 mailing list Python-3000@python.org http://mail.python.org/mailman/listinfo/python-3000 Unsubscribe: http://mail.python.org/mailman/options/python-3000/archive%40mail-archive.com
