cufftXtMemcpy device to device would change data value(simpleCUFFT_2d_MGPU)?

[xiuyangleiasp]

Hi, When I run the simpleCUFFT_2d_MGPU, after the call the result = cufftXtMemcpy(planComplex, d_d_f, d_f, CUFFT_COPY_DEVICE_TO_DEVICE);,I copyData d_f、d_d_f from gpus to host memory, like this,

In host Memory, d_f_out!=d_d_f_out. But according to the cuda docs cufftXtMemcpy Device to Device just change the order of data in gpus,

So Is there something I do wrong? Thanks.