Re: [PyCUDA] different results (nan values)