TMS320F28375D: Problems with running CLA code

Hi!

Thanks to your previous help, I have fixed my linker file and all memory arrangement, and simple programs run flawlessly in my CLA.

But then I get to running an FFT code I've written, and thr problems begin. The code runs fine on the main core (tested the correctness via MATLAB).

When I transferred the code to CLA first time, it just didn't execute at all, showing no errors during building or debugging. Then I split the FFT into several smaller sub-functions, and the code miraculously started executing, but the output was just ve-e-ery slightry relevant to what should be, an in terms of dsp such small accuracy is unacceptable.

Here are my FFT tries: the united one and a split one. The CLA code is triggered from the main function by Cla1ForceTask1andWait();

#include "fft_test1_cla.h"
#include "F28x_Project.h"


//FFT variables:
int16_t n, nspan, submatrix, node;   
int16_t N1;// = 8;//1<<logN;       
double temp, angle, realtwiddle, imtwiddle;
int16_t span;

int16_t i;
//Rev bits variables:
int16_t forward, rev, toggle;
int16_t nodd, noddrev;               // to hold bitwise negated or odd values
int16_t  halfn, quartn, nmin1;




void innerloop();
void secondloop();
void outerloop();
void addandsub();
void rotations();

void reverse (double *real, double *output);
__interrupt void Cla1Task1 ( void )
{
   N1=8;
   span=8;
   n=0;
outerloop();
}


void innerloop()
{
    addandsub();
    rotations();
}

void addandsub()
{
    nspan = n+span;
    temp = real[n] + real[nspan];       // additions & subtractions
    real[nspan] = real[n]-real[nspan];
    real[n] = temp;
    temp = im[n] + im[nspan];
    im[nspan] = im[n] - im[nspan];
    im[n] = temp;
}

void rotations()
{
    angle = primitive_root * node;      // rotations
    realtwiddle = CLAcos(angle);
    imtwiddle = -CLAsin(angle);
    temp = realtwiddle * real[nspan] - imtwiddle * im[nspan];
    im[nspan] = realtwiddle * im[nspan] + imtwiddle * real[nspan];
    real[nspan] = temp;
    n++;
}

void secondloop()
{
    for(submatrix=0; submatrix<(N1>>1)/span; submatrix++)
    {
        for(node=0; node<span; node++){
         innerloop();
        }
    }
}

void outerloop()
{
    for(span=N1>>1; span; span>>=1)      // loop over the FFT stages
            {
               primitive_root = -(3.14159265359/span);
               secondloop();
            }
}

__interrupt void Cla1Task1 ( void )
{
   N1=8;
   span=8;
   n=0;
 for(span=N1>>1; span; span>>=1)      // loop over the FFT stages
        {
           primitive_root = -(3.14159265359/span);   // define MINPI in the header
           //secondloop();
           //__mdebugstop();
//#pragma MUST_ITERATE(loop2)
           for(submatrix=0; submatrix<(N1>>1)/span; submatrix++)
           {
              // __mdebugstop();

//#pragma MUST_ITERATE(loop3)
              for(node=0; node<span; node++)
              {
                  __mdebugstop();
                nspan = n+span;
                temp = real[n] + real[nspan];       // additions & subtractions
                real[nspan] = real[n]-real[nspan];
                real[n] = temp;
                temp = im[n] + im[nspan];
                im[nspan] = im[n] - im[nspan];
                im[n] = temp;

                angle = primitive_root * node;      // rotations
                realtwiddle = CLAcos(angle);
                imtwiddle = -CLAsin(angle);
                temp = realtwiddle * real[nspan] - imtwiddle * im[nspan];
                im[nspan] = realtwiddle * im[nspan] + imtwiddle * real[nspan];
                real[nspan] = temp;

                n++;   // not forget to increment n

              } // end of loop over nodes
              __mdebugstop();
              n = (n+span) & (N1-1);   // jump over the odd blocks

            } // end of loop over submatrices

         } // end of loop over FFT stages
    __mdebugstop();
//reverse(real,outputR);
//reverse(im, outputI);
}