Function copy_peer 

pub fn copy_peer<T: Copy>(
    dst: &mut DeviceBuffer<T>,
    dst_device: &Device,
    src: &DeviceBuffer<T>,
    src_device: &Device,
) -> CudaResult<()>

Copies data between device buffers on different GPUs (synchronous).

Both buffers must have the same length. Peer access should be enabled between the source and destination devices before calling this function.

Errors

• CudaError::InvalidValue if buffer lengths do not match.
• CudaError::PeerAccessNotEnabled if peer access has not been enabled.