On 18/07/13 23:30, Lev Givon wrote:
> I'm trying to access the memory associated with a GPUArray from within a
> compiled extension built using Cython's memoryview feature. According to the
> Cython documentation, it is possible to access C arrays using this feature;
> however, when I attempt to do so using a GPUArray's pointer with a GPU (and
> version of CUDA) that supports UVA, the extension segfaults. Does anyone have
> any thoughts as to why this might be occurring?

Are you sure about this?  Just because CUDA pointers reside in a common
address space does not necessarily mean that they can be deferenced.
Indeed, the only valid combination I am aware of (and my knowledge may
well be incomplete/outdated) is that page-locked host allocations made
with the "DEVICEMAP" flag can be dereferenced inside CUDA kernels
without requiring an explicit copy.

Even if it is possible it is highly likely to result in crummy
performance and is almost certainly ill-advised.

Polemically yours, Freddie.


Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
PyCUDA mailing list
[email protected]
http://lists.tiker.net/listinfo/pycuda

Reply via email to