On Sat, Aug 18, 2012 at 9:28 AM, David Cournapeau cournape@gmail.com wrote:

Hi,

I have started toying with implementing a quad precision dtype for numpy on supported platforms, using the __float128 + quadmath lib from gcc. I have noticed invalid (and unexpected) downcast to long double in some cases, especially for ufuncs (e.g. when I don't define my own ufunc for a given operation).

Looking down in numpy ufunc machinery, I can see that the issue is coming from the assumption that long double is the highest precision possible for a float type, and the only way I can 'fix' this is to define kind to a value != 'f' in my dtype definition (in which case I get an expected invalid cast exception). Is there a way to still avoid those casts while keeping the 'f' kind ?

I never looked at that code, but why not change the ufunc to remove the current assumption? I suppose if you ask the questions is that this is not trivial to do?

Fred