You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Some cuda.cooperative interfaces only accept Numba types, not NumPy dtypes (e.g. only numba.int32 works, not numpy.int32). This is quite annoying, as many of the host-side APIs we use to create and manage arrays only accept NumPy dtypes, not Numba types.
Everywhere that we accept a type, NumPy types should be supported.
The text was updated successfully, but these errors were encountered:
Some cuda.cooperative interfaces only accept Numba types, not NumPy dtypes (e.g. only
numba.int32
works, notnumpy.int32
). This is quite annoying, as many of the host-side APIs we use to create and manage arrays only accept NumPy dtypes, not Numba types.Everywhere that we accept a type, NumPy types should be supported.
The text was updated successfully, but these errors were encountered: