On Dec 30, 2009, at 2:30 AM, Stefan Behnel wrote: > Robert Bradshaw, 30.12.2009 10:45: >> On Dec 27, 2009, at 12:41 AM, Stefan Behnel wrote: >>> Robert Bradshaw, 27.12.2009 05:38: >>>> On Dec 14, 2009, at 11:22 AM, Dag Sverre Seljebotn wrote: >>>>> Be advised however that there's quite a few features not >>>>> available in pure Python mode, >>>> Though it may be interesting how far we could get with ctypes. >>> Do you mean accelerating ctypes usage in Cython (somehow) or >>> reimplementing >>> (parts of) ctypes as compiler type declaration code? The latter >>> could >>> require flow control analysis to work at all, although it might be >>> possible >>> to find a subset of ctypes that would just enable C type >>> declarations, and >>> then basically support that in "cython.declare()", without >>> supporting >>> reference counted pointers etc. >> >> No, I meant leveraging ctypes to give the "pure" mode a more complete >> emulation of the compiled version (calling external functions, >> integer >> overflow semantics, etc.). > > Ok, but that means you'd loose a lot of performance through the > Python call > overhead of ctypes.
Yes. Of course, it's always a given (I hope) that the Python version will be slower than the compiled. It's more for providing compiled and non-compiled versions out of the same source. - Robert _______________________________________________ Cython-dev mailing list [email protected] http://codespeak.net/mailman/listinfo/cython-dev
