[For Help]"illegal memory access was encountered" when using cudaMemcpyAsync() on multi-GP

for(int ss=0; ss< Num_Devies, ss++)

CUDA_CALL(cudaMemcpyAsync(devPrjPage[ss], CPUPage[ss], PageDataSize* sizeof(float), cudaMemcpyHostToDevice, stream[ss]));
Kernel<<<…>>>(…)

there seems no syntax error. each time i run the .exe on Debug platform, it fails and gives an error propup!

“illegal memory access was encountered” in the cudaMemcpyAsync line

I have been mad for this!!!

Looking forward to someone to help me here~

perhaps post some code, particularly pertaining to the preceding memory allocations