why passing DFTI_SINGLE into DftiCreateDescriptor gives garbage values?

raavi — Fri, 07 Jun 2024 10:59:21 GMT

I have two questions.

1) why, when passing DFTI_SINGLE into DftiCreateDescriptor gives garbage values?

2) Though passing DFTI_DOUBLE into DftiCreateDescriptor gives sane values but the output does not match with Octave output.

MKL_LONG status; DFTI_DESCRIPTOR_HANDLE backward_handle_2d = NULL; MKL_Complex16 *fourier_2d_mkl = NULL; MKL_LONG dims[2]; dims[0] = 2; dims[1] = 23; int count_2d = dims[0]*dims[1]; status = DftiCreateDescriptor(&backward_handle_2d, DFTI_DOUBLE, DFTI_COMPLEX, 2, dims); status = DftiCommitDescriptor(backward_handle_2d); fourier_2d_mkl = (MKL_Complex16*)mkl_malloc(count_2d * sizeof(MKL_Complex16), 64); for (int i=0; i<count_2d; i++) { fourier_2d_mkl[i].real = 0.0; fourier_2d_mkl[i].imag = 0.0; } fourier_2d_mkl[0].real = -2.642292567870791e-17; fourier_2d_mkl[0].imag = 2.457213459869541e-17; fourier_2d_mkl[1].real = 3.394204453993388e-15; fourier_2d_mkl[1].imag = 6.465292243826133e-14; fourier_2d_mkl[2].real = 7.250014158956356e-13; fourier_2d_mkl[2].imag = -4.363287473914971e-12; fourier_2d_mkl[3].real = -1.428607589482790e-10; fourier_2d_mkl[3].imag = 1.188079301204293e-10; fourier_2d_mkl[4].real = 6.952990293826207e-09; fourier_2d_mkl[4].imag = 7.124671255671636e-10; fourier_2d_mkl[5].real = -1.247732021241605e-07; fourier_2d_mkl[5].imag = -1.525550389367994e-07; fourier_2d_mkl[6].real = -3.680801028326798e-07; fourier_2d_mkl[6].imag = 3.686621541467887e-06; fourier_2d_mkl[7].real = 2.677103698936326e-05; fourier_2d_mkl[7].imag = -2.208311063252299e-05; fourier_2d_mkl[8].real = 3.310982970348360e-06; fourier_2d_mkl[8].imag = 4.060528725376024e-06; fourier_2d_mkl[15].real = 2.701945410474277e-05;fourier_2d_mkl[15].imag = -2.177772108929361e-05; fourier_2d_mkl[16].real = 3.685365224055048e-06; fourier_2d_mkl[16].imag = 3.817322139893435e-07; fourier_2d_mkl[17].real = 1.243931716409880e-07; fourier_2d_mkl[17].imag = 1.528359352830827e-07; fourier_2d_mkl[18].real = -7.210215362767208e-10; fourier_2d_mkl[18].imag = 6.961508659626094e-09; fourier_2d_mkl[19].real = -1.449335800029190e-10; fourier_2d_mkl[19].imag = 1.184786074831728e-10; fourier_2d_mkl[20].real = -4.345677649234239e-12; fourier_2d_mkl[20].imag = -8.023045583080176e-13; fourier_2d_mkl[21].real = -2.174238868343064e-15; fourier_2d_mkl[21].imag = -6.468023321689881e-14; fourier_2d_mkl[22].real = -2.648066946353768e-17; fourier_2d_mkl[22].imag = -2.943784428508000e-17; fourier_2d_mkl[23].real = -1.647039786289660e-14; fourier_2d_mkl[23].imag = 1.531673055823963e-14; fourier_2d_mkl[24].real = 9.328254025265415e-13; fourier_2d_mkl[24].imag = 1.660593228562294e-11; fourier_2d_mkl[25].real = 9.970878341165177e-11; fourier_2d_mkl[25].imag = -4.852415762764660e-10; fourier_2d_mkl[26].real = -7.587349566415179e-09; fourier_2d_mkl[26].imag = 5.641756309256826e-09; fourier_2d_mkl[27].real = 1.569897070996509e-07; fourier_2d_mkl[27].imag = 2.418811983812866e-08; fourier_2d_mkl[28].real = -1.184957370080801e-06; fourier_2d_mkl[28].imag = -1.607459905410108e-06; fourier_2d_mkl[29].real = -2.422333576942743e-06; fourier_2d_mkl[29].imag = 1.649413356788869e-05; fourier_2d_mkl[30].real = 5.397110486561100e-05; fourier_2d_mkl[30].imag = -4.182910800968701e-05; fourier_2d_mkl[31].real = 1.389085535845420e-05; fourier_2d_mkl[31].imag = 1.904852960287619e-05; fourier_2d_mkl[38].real = 5.632425140135355e-05; fourier_2d_mkl[38].imag = -3.860245232056857e-05; fourier_2d_mkl[39].real = 1.644525380084373e-05; fourier_2d_mkl[39].imag = 2.735362516264706e-06; fourier_2d_mkl[40].real = 1.168718667543190e-06; fourier_2d_mkl[40].imag = 1.619167661329656e-06; fourier_2d_mkl[41].real = -2.520504976062189e-08; fourier_2d_mkl[41].imag = 1.569498873744191e-07; fourier_2d_mkl[42].real = -7.685758628322245e-09; fourier_2d_mkl[42].imag = 5.594175098567900e-09; fourier_2d_mkl[43].real = -4.829524388709853e-10; fourier_2d_mkl[43].imag = -1.083259323876055e-10; fourier_2d_mkl[44].real = -6.194460301224245e-13; fourier_2d_mkl[44].imag = -1.661409550202508e-11; fourier_2d_mkl[45].real = -1.650639172375137e-14; fourier_2d_mkl[45].imag = -1.834970939618425e-14; status = DftiComputeBackward(backward_handle_2d, fourier_2d_mkl); status = DftiFreeDescriptor(&backward_handle_2d); int K = 8; for (int j=0; j<dims[0]; j++) { for (int i=0; i<dims[1]; i++) { int index = i + ( j * dims[1] ); double phi = (2.0 * pi * i) / static_cast<double>(dims[1]); std::complex<double> signal(fourier_2d_mkl[index].real, fourier_2d_mkl[index].imag); std::complex<double> output = Internal::complex_multiply(signal, Internal::expi(-K * phi)); std::cout <<std::left <<output.real() << "\t" <<output.imag()<<std::endl; } std::cout << std::endl; } mkl_free(fourier_2d_mkl);

Re: why passing DFTI_SINGLE into DftiCreateDescriptor gives garbage values?

raavi — Sun, 09 Jun 2024 17:47:36 GMT

Please close this post.

topic Re: why passing DFTI_SINGLE into DftiCreateDescriptor gives garbage values? in Intel® oneAPI Math Kernel Library

why passing DFTI_SINGLE into DftiCreateDescriptor gives garbage values?

Re: why passing DFTI_SINGLE into DftiCreateDescriptor gives garbage values?