CUresult cuDeviceCanAccessPeer ( int *  canAccessPeer,
CUdevice  dev,
CUdevice  peerDev 
)

Returns in *canAccessPeer a value of 1 if contexts on dev are capable of directly accessing memory from contexts on peerDev and 0 otherwise. If direct access of peerDev from dev is possible, then access may be enabled on two specific contexts by calling cuCtxEnablePeerAccess().

Parameters:
canAccessPeer - Returned access capability
dev - Device from which allocations on peerDev are to be directly accessed.
peerDev - Device on which the allocations to be directly accessed by dev reside.
Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_ERROR_INVALID_DEVICE
Note:
Note that this function may also return error codes from previous, asynchronous launches.
See also:
cuCtxEnablePeerAccess, cuCtxDisablePeerAccess


Generated by Doxygen for NVIDIA CUDA Library  NVIDIA