cudaMemcpyPeer and cudaMemcpyPeerAsync with no offset?

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

I have tested in fedora 16 and with cuda 4.1 and 5.0 and and cannot use offsets in the pointer devices…
Is this a limitation?
If no offset is set then there is no problems.

cudaError_t cudaMemcpyPeer ( void * dst,
int dstDevice,
const void * src,
int srcDevice,
size_t count
)

Anyone have this problem?

Thanks

2

I have to use…
type *ptr = reinterpret_cast( device_ptr + offset );

3

had to put in code block…

type *ptr = reinterpret_cast( device_ptr + offset );
4

it deleted the template part…