While working with PyCuda, I've noticed that several GPU array functions
seem to be giving errors, that should work fine.
*These are using one-dimensional arrays.*
__imul__ works fine for an array times a scalar, but has errors when an
array is multiplied by another array (this was recently fixed in PyOpenCL)
__pow__ seems to have problems with complex arrays. If I take a complex
array, a, and say a**2, I get an error. This works fine with floats though.
I'm fairly certain that these are just bugs, since both work fine in
PyOpenCL. Is that the case?
Show replies by date