This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Compiler/TMDSEVM6678: how to build fftlib 2 unit tests/examples ? ... here is how I did

Part Number: TMDSEVM6678
Other Parts Discussed in Thread: FFTLIB

Tool/software: TI C/C++ Compiler

Hi All,

Previously, I have asked how to build fftlib unit test examples; below I described the configs I used to achieve this task.

First and foremost, fftlib is no longer maintained by TI as discussed in another post here; however, you are can still download and compile it if you need to ... 

Steps I followed are as below (which worked for me) although the accuracy and correctness of the fttlib code is still not clear for me because as far as I could run the unit test examples I can see some differences between fftlib vs. dsplib implementations.

Here are the steps to build fftlib unit tests

  1. Download and install fftlib_c66x_2_0_0_2
Import the examples from C:\ti\fftlib_c66x_2_0_0_2\packages\ti\fftlib\src\fft_dp_1d_c2c in CCS (Code Composer 9.2 ) as you see below:
  1. Remember the example itself is in the k1 for (eg. k1\fft_dp_1d_c2c_k1_66_LE_ELF) 
  • Then open project properties and match screens as you see below:
  • Added the SOC_C6678 to defined symbols as you can see below:
  • Rebuild the project and it should compile all well.

Hope it helps sb ... 

Cheers,

Mike

  • I have one question too ... why twiddles coefficients are computed as below in the fft_dp_1d_c2c_d.c code and also all other places. 

    /* Function for generating Specialized sequence of twiddle factors */
    void tw_gen_cn (double *w, int n)
    {
        int i, j, k;
        const double PI = 3.141592654;
    
        for (j = 1, k = 0; j <= n >> 2; j = j << 2)
        {
            for (i = 0; i < n >> 2; i += j)
            {
                w[k]     = cos (2 * PI * i / n);
                w[k + 1] = sin (2 * PI * i / n);
                w[k + 2] = cos (4 * PI * i / n);
                w[k + 3] = sin (4 * PI * i / n);
                w[k + 4] = cos (6 * PI * i / n);
                w[k + 5] = sin (6 * PI * i / n);
                k += 6;
            }
        }
    }
    

    but the matlab code looks very different !!?? see below

    % Program for generating n-length FFT's twiddle factor
    % By: Denny Hermawanto
    % Puslit Metrologi LIPI, INDONESIA
    % Copyright 2015
    function [real_twiddle, im_twiddle] = tw_gen_cn(fft_length)
        if nargin < 1
            [real_twiddle, im_twiddle] = tw_gen_cn(8);
            return;
        end   
    % fft_length = input('Enter FFT length:');
    for mm = 0:1:(fft_length-1)
        theta = (-2*pi*mm*1/fft_length);
    %   Twiddle factor equation
    %   twiddle = exp(1i*theta);
    %   Euler equation for complex exponential 
    %   e^(j*theta) = cos(theta) + j(sin(theta)) 
        
        twiddle(mm+1) = cos(theta) + (1i*(sin(theta)));
        real_twiddle = real(twiddle);
        real_twiddle = real_twiddle';
        im_twiddle = imag(twiddle);
        im_twiddle = im_twiddle';
        
    end;
    end

  • Mike,

    Thanks for your contribution on how to compile the FFTLIB. This is helpful info and will benefit forum users.

    Rex

  • ok, I found that the codes are mathematically equivalent ... look here

    But now I have another question. From the code, I could see there are two ways to take FFT (1.Direct 2.ECPY); however the later doesn't seem to work:

    1. What ECPY do exactly?

    2. How to get it to work ?

    Below you can see the code ... let me know if you any idea ...

    Regards,

    int main ()
    {
        int     i, j, N, rad_cn;
        int     rad3 = 1, rad5 = 1, rad15 = 1;
        clock_t t_start, t_stop, t_overhead, t_opt;
        double  diff, max_diff = 0;
        fft_plan_t p;
        fft_callout_t plan_fxns;
    
        /* --------------------------------------------------------------------- */
        /* intialize hardware timers                                             */
        /* --------------------------------------------------------------------- */
        TSCL=0;TSCH=0;
    
        /* initalize callout functions */
        plan_fxns.memoryRequest   = fft_memory_request;
        plan_fxns.memoryRelease   = fft_memory_release;
    
        /* initialize ECPY */
        fft_assert( (EdmaMgr_init(DNUM, NULL) == EdmaMgr_SUCCESS), DNUM, "EdmaMgr_init() return error!");
        fftEdmaState = FFT_EDMA_STATE_INIT;
    
        /* radix 2&4 testing */
        for (N = 10; N <= 10; N = N*2)
        {
            memset (x_i,  0x55, sizeof (x_i) );
            memset (x_cn, 0x55, sizeof (x_cn));
            memset (w_i,  0x00, sizeof (w_i) );
    
            /* ---------------------------------------------------------------- */
            /* Initialize input vector temporarily.                             */
            /* ---------------------------------------------------------------- */
    
            for (i = 0; i < N; i++) {
              x_cn[PAD + 2*i  ] = i+1; // sin (2 * 3.1415 * 50 * i / (double) N);
              x_cn[PAD + 2*i+1] = 0; // sin (2 * 3.1415 * 60 * i / (double) N);
            }
            for (j = 0; j < 2*N; j++) {
              x_i[PAD + j] = x_cn[PAD + j];
            }
    
            /* ---------------------------------------------------------------- */
            /* Force uninitialized arrays to fixed values.                      */
            /* ---------------------------------------------------------------- */
            memset (y_i,  0xA5, sizeof (y_i) );
            memset (y_cn, 0xA5, sizeof (y_cn));
    
            /* ---------------------------------------------------------------- */
            /* Generate twiddle factors.                                        */
            /* ---------------------------------------------------------------- */
            j = 0;
            for (i = 0; i <= 31; i++)
                if ((N & (1 << i)) == 0)
                    j++;
                else
                    break;
    
            if (j % 2 == 0) {
                rad_cn = 4;
            }
            else {
                rad_cn = 2;
            }
    
            tw_gen_cn (ptr_w_cn, N);
    
            DSPF_dp_fftDPxDP (N, ptr_x_cn, ptr_w_cn, ptr_y_cn, rad_cn, 0, N);
    
            /* ---------------------------------------------------------------- */
            /* Compute the overhead of allocating and freeing EDMA              */
            /* ---------------------------------------------------------------- */
            p.edmaState = fft_assign_edma_resources();
            fft_free_edma_resources(p.edmaState);
            t_start = _itoll(TSCH, TSCL);
            p.edmaState = fft_assign_edma_resources();
            fft_free_edma_resources(p.edmaState);
            t_stop  = _itoll(TSCH, TSCL);
            t_overhead = t_stop - t_start;
    
            /***************************************
             * ecpy fft test
             ***************************************/
            plan_fxns.ecpyRequest = fft_assign_edma_resources;
            plan_fxns.ecpyRelease = fft_free_edma_resources;
    
            p = fft_dp_plan_1d_c2c (N, FFT_ECPY, plan_fxns);
            t_start = _itoll(TSCH, TSCL);
            fft_execute (p); // this doesn't work
            t_stop = _itoll(TSCH, TSCL);
            fft_destroy_plan (p);
            t_opt  = (t_stop - t_start) - t_overhead;
    
            /* ---------------------------------------------------------------- */
            /* compute difference and track max difference                      */
            /* ---------------------------------------------------------------- */
            diff = 0; max_diff = 0;
            for(i=0; i<2*N; i++) {
                printf("FFT_ECPY ptr_y_i[%d] = %f\n", i, ptr_y_i[i]);
              diff = _fabs(ptr_y_cn[i] - ptr_x_i[i]); // 
              if (diff > max_diff) max_diff = diff;
            }
            printf("fft_dp_1d_c2c_ecpy\tsize= %d\n", N);
            printf("max_diff = %f", max_diff);
            printf("\tN = %d\tCycle: %d\n\n", N, t_opt);
    
    //        /********************************************************
    //         * direct form fft test, up to 2048 due to twiddle size
    //         *******************************************************/
            if (N <= 2048) {
    
              /* ---------------------------------------------------------------- */
              /* Initialize input vector temporarily.                             */
              /* ---------------------------------------------------------------- */
    
              for (i = 0; i < N; i++) {
                x_i[PAD + 2*i  ] = i+1; //sin (2 * 3.1415 * 50 * i / (double) N);
                x_i[PAD + 2*i+1] = 0; sin (2 * 3.1415 * 60 * i / (double) N);
              }
    
              /* ---------------------------------------------------------------- */
              /* Compute the overhead of EDMA assigment                           */
              /* ---------------------------------------------------------------- */
              t_start = _itoll(TSCH, TSCL);
              t_stop  = _itoll(TSCH, TSCL);
              t_overhead = t_stop - t_start;
    
              plan_fxns.ecpyRequest = NULL;
              plan_fxns.ecpyRelease = NULL;
    
    
              p = fft_dp_plan_1d_c2c (N, FFT_DIRECT, plan_fxns);
              t_start = _itoll(TSCH, TSCL);
              fft_execute (p); // this works
              t_stop = _itoll(TSCH, TSCL);
              fft_destroy_plan (p);
              t_opt  = (t_stop - t_start) - t_overhead;
    
    
              /* ---------------------------------------------------------------- */
              /* compute difference and track max difference                      */
              /* ---------------------------------------------------------------- */
              diff = 0; max_diff = 0;
              for(i=0; i<2*N; i++) {
                  printf("FFT_DIRECT ptr_y_i[%d] = %f\n", i, ptr_y_i[i]);
                diff = _fabs(ptr_y_cn[i] - ptr_y_i[i]);
                if (diff > max_diff)
                  max_diff = diff;
              }
    
              printf("fft_dp_1d_c2c_direct\tsize= %d\n", N);
              printf("max_diff = %f", max_diff);
              printf("\tN = %d\tCycle: %d\n\n", N, t_opt);
    
              for(i=0; i<2*N; i+=2) {
                    printf("FFT_DIRECT fft[%d] = %f + i(%f) \n", i+1, ptr_y_i[i], ptr_y_i[i+1]);
              }
            }
        }
    
    }
    

  • Rex Chang said:

    Mike,

    Thanks for your contribution on how to compile the FFTLIB. This is helpful info and will benefit forum users.

    Rex

    No problem ... That's all we do hear :)