Dear Jiang, thanks I got it to work by using my own version of `builtin_converters.hpp` locally together with the code that you recommended.
-Holger On Tue, Nov 6, 2012 at 3:25 AM, jiangzuo...@gmail.com <jiangzuo...@gmail.com> wrote: > how about return a wrap type of long double to cheating boost/python, and > register youself converter as > > PyObject * o = PyArrayScalar_New(LongDouble); > PyLongDoubleScalarObject * s = (PyLongDoubleScalarObject*)o; > s->obval = value; > > Don't forget to call import_array() in your model init function. > > BTW, I think float128 is not useful, it's slow. and in most case, change > algorithm to avoid precision problem is another *better* approach. > > > Changsheng Jiang > > > On Mon, Nov 5, 2012 at 6:27 PM, Holger Brandsmeier <brandsme...@gmx.de> > wrote: >> >> Dear list, >> >> the boost python buildin converter for long double: >> BOOST_PYTHON_TO_PYTHON_BY_VALUE(long double, >> ::PyFloat_FromDouble(x), &PyFloat_Type) >> converter/builtin_converters.hpp >> identifies long double, which on my system is a 128bit float, with a >> 64bit float in python. I would like to identify it with numpy`s >> float128. Can you give me any hings how I would be able to do that? >> >> -Holger >> _______________________________________________ >> Cplusplus-sig mailing list >> Cplusplus-sig@python.org >> http://mail.python.org/mailman/listinfo/cplusplus-sig > > > > _______________________________________________ > Cplusplus-sig mailing list > Cplusplus-sig@python.org > http://mail.python.org/mailman/listinfo/cplusplus-sig _______________________________________________ Cplusplus-sig mailing list Cplusplus-sig@python.org http://mail.python.org/mailman/listinfo/cplusplus-sig