use fftw_plan_dft_r2c_1d get inf

[ShuoShenDe]

hello,guys~ I try to used fftw_plan_dft_r2c_1d for a 2D vector. So, vector shape is (NumFrames, WindowOutHalfLength)

So ,I used fftw_plan_dft_r2c_1d in following codes: ` float fftwInForward; fftwf_complex fftwOutForward; vector test(halfNFFT); fftwInForward = (float)fftwf_malloc(sizeof(float) 2 (stft.WindowOutLength / 2.0 + 1)); fftwOutForward = (fftwf_complex)fftwf_malloc(sizeof(fftwf_complex) * stft.WindowOutHalfLength); fftwf_plan forward = fftwf_plan_dft_r2c_1d(stft.WindowOutLength, fftwInForward, fftwOutForward, FFTW_DESTROY_INPUT);

for (int frame = 0; frame < stft.NumFrames; frame++)
{
    fftwInForward[0] = filteredCepstrum[frame][0];
    for (int bin = 1; bin < nCoeff + 1; bin++) {
        fftwInForward[bin] = filteredCepstrum[frame][bin] * window[int(halfNFFT / 2) + bin - 1];
    }
    for (int bin = halfNFFT - nCoeff; bin < halfNFFT; bin++) {
        fftwInForward[bin] = filteredCepstrum[frame][bin] * window[int(halfNFFT / 2) + bin - halfNFFT];
    }
    for (int bin = nCoeff; bin < halfNFFT - nCoeff; bin++)
    {
        fftwInForward[bin] = 0;
    }

    fftw_execute(forward);

    for (int bin = 0; bin < halfNFFT; bin++) {
        filteredCepstrum[frame][bin] = exp(fftwOutForward[bin][0]);
    }

}`

However, almost all data become "nan" or "-nan". like follows:

When I change the data type to double, some rows have several "inf" inside.

I checked the input data(fftwInForward), there is not "inf" or "nan" inside. Does anyone know what happened?