invalid resource handle for cudnnSetTensor4dDescriptorEx

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

cudnnSetTensor4dDescriptorEx in code below returns “invalid resource handle”.
Any hints to debug? Thank you.

Below is my spec:
board: jetson Tx1
nvcc --version: release 7.0, V7.0.74

Blob *b = 0;  
b = ins[0];	    ASSERT(0 == cudnnSetTensor4dDescriptorEx(  xs_desc, CUDNN_FLOAT, b->N,b->C,b->H,b->W, b->H*b->W*b->C, b->H*b->W, b->W, 1 ));
cudaError_t error = cudaGetLastError(); 
if(error != cudaSuccess) 
	{ 		
	printf("------ CUDA error: %s\n", cudaGetErrorString(error)); 
	}
2

cudnnGetErrorString should be used for cudnn.
cudnnGetErrorString should come before ASSERT().