2*np.float32 gives float64

[Andreas Kloeckner]

Numpy:

>>> import numpy as np
>>> a = np.array([1,2,3], dtype=np.float32)
>>> (2*a).dtype
dtype('float32')
>>> 

Numpypy: (450881e, with Pypy 2.6.0-2 from Debian)

>>>> import numpy as np
>>>> a = np.array([1,2,3], dtype=np.float32)
>>>> (2*a).dtype
dtype('float64')
>>>> 

[Ronan Lamy]

This has already been fixed in PyPy nightlies.