I guess I'm arguing for just using whatever comes out of the buffer interface. Seeing as all you can possibly get on the CUDA side is a pointer to the start of a bunch of bytes, it makes sense to me to have to know the layout of those bytes before sending them into CUDA, and for the API to not mess with it. But I've done a lot of low-level programming so I guess I'm more used to tools always giving the programmer the benefit of the doubt (i.e. always assuming the programmer knows what they're doing).
Off-topic again, that's why I put the "++" in brackets. Since templates, constructors and overloading seem to be fairly well-used in CUDA, it seems much more like C++ than C to me :) Cheers, Ian ________________________________________ From: [email protected] [[email protected]] On Behalf Of Nicholas Tung [[email protected]] Sent: Wednesday, 4 March 2009 6:00 AM To: [email protected] Subject: Re: [PyCuda] pycuda memcpy_htod the question is, what is "blindly copying bytes" when you have a numpy array? The builtin bytes() function seems to suggest supporting both F-contiguous and C-contiguous layouts (depending on what the current state of the numpy array is). btw c++ is not actually supported, just some features http://forums.nvidia.com/index.php?s=&showtopic=79933&view=findpost&p=453778 regards, Nicholas _______________________________________________ PyCuda mailing list [email protected] http://tiker.net/mailman/listinfo/pycuda_tiker.net
