Unable to FFT in place multiple times?

Accelerated Computing CUDA CUDA Programming and Performance
1

I’ve been facing an error when trying to do multiple FFTs in place using CUDA’s cuFFT API. Some sample code ahead that gives the same error I keep getting:

std::vector<Complex> host_signalx(N);
std::copy(host_signalx.begin(), host_signalx.end(), datablock);

checkCudaErrors(cufftXtMemcpy (plan_input, d_signalx, datablock,
            CUFFT_COPY_HOST_TO_DEVICE));

checkCudaErrors(cufftXtExecDescriptorC2C(plan_input, d_signalx, d_signalx,
                CUFFT_FORWARD));
checkCudaErrors(cufftXtExecDescriptorC2C(plan_input, d_signalx, d_signalx,
                CUFFT_INVERSE));

The error:

CUDA error at test2.cu:235 code=3(CUFFT_INVALID_TYPE) "cufftXtExecDescriptorC2C(plan_input, d_signalx, d_signalx, CUFFT_INVERSE)"

I’ve malloc’ed and it works when the FFT is computed once, but not twice- which stumps me, since I am looking to use it in a loop multiple times without having to free and allocate memory each time.

The error on the website is vague too, and says:

CUFFT_INVALID_TYPE   = 3,  //  No longer used

Any insights would be much appreciated.