Two dimensional fft on DM6437

Hello,

I'm working on video stabilisation project on dm6437.

It's my MsC final  project.I progress my code example based on phase correlation.

I have  to do  two dimensional fft on image row by row than column by column. And I select 256 *256 region from image.

I defined global variables just like this:

# define N 256

#pragma DATA_ALIGN(x_16x16, 128);
 int  x_16x16[900000];

#pragma DATA_ALIGN(y_16x16, 128);
 int   y_16x16[900000];

#pragma DATA_ALIGN(y2_16x16, 128);
 int   y2_16x16[900000];

#pragma DATA_ALIGN(w_16x16, 128);

short   w_16x16 [2 * N * N];

#pragma DATA_ALIGN(w2_16x16, 8);
short   w2_16x16 [2 * N];



#pragma DATA_ALIGN(line, 128);
 int   line [N];


#pragma DATA_ALIGN(lineIn, 128);
int   lineIn [2 * N];

#pragma DATA_ALIGN(lineOut, 128);
 int   lineOut [2 * N];

#pragma DATA_ALIGN(lineIn2, 128);

int   lineIn2 [2 * N];

#pragma DATA_ALIGN(lineOut2, 128);
 int   lineOut2 [2 * N];

#pragma DATA_ALIGN(FFTOut, 128);
int   FFTOut [900000];

I write fft function like this:

void fft(void* currentFrame,int yRows, int xPixels)
{
int i = 0, j = 0;
    int y = 0,t=0;
    int *xp;
    int FFTSize = N * N;
    float scale = 8; 
    xPixels = 4;
    gen_twiddle_fft16x32(w_16x16, N);

    //fftrow
     for(i = 0; i < 2 * N * N ; i++) 
     {                                 
            if(i % 2 == 0)             
            {                          
             x_16x16[i] = *( ( ( int *)currentFrame ) + i ) ; ;          
                                 
             }                         
             else                      
             x_16x16[i] = 0;           
      
     }   

    for(i = 0; i < 2 * FFTSize; i += 2 * N)
    {
    DSP_blk_move((short *)(x_16x16 + i),(short *)(lineIn), 4 * N );

    DSP_fft16x32(w_16x16, N, lineIn, lineOut);

// Scaled fft out

/*    for(j=0;j<(2 * N);j++)

    {lineOut[j]=lineOut[j]/512;}*/
   

    DSP_blk_move((short *)(lineOut),(short *)(y_16x16 + i), 4 * N );

   
    }
 

 
 //fftcolumn

DSP_mat_trans2(y_16x16,256, 256 , y2_16x16);

 
    for(i = 0; i < 2 * FFTSize; i += 2 * N)
    {
    DSP_blk_move((short *)(y2_16x16 + i),(short *)(lineIn), 4 * N );

    DSP_fft16x32(w_16x16, N, lineIn, lineOut);

//Scaled fft out

/*    for(j=0;j<(2 * N);j++)

    {lineOut[j]=lineOut[j]/20;}*/

    DSP_blk_move((short *)(lineOut),(short *)(FFTOut + i), 4 * N );

  
    }


 
 DSP_mat_trans2(FFTOut, 256, 256 , y2_16x16);

 


}

 
// Transpose function
void DSP_mat_trans2(int *x, int rows, int columns, int *r)
{
int i , j, y = 0;
for(i = 0; i < 2 * N ; i++) 
     {         
     if(i % 2 == 0)
     {
             for(j = 0; j < 2 * N * N; j+= 2 * N)
             {                       
                        
             r[y] = *( ( (int *)x ) + i + j) ;                           
             r[y + 1] = *( ( (int *)x ) + i + j + 1);   
                    
        
            y+=2;
            }
           
   
       

     }
                        
     }   

}

I scaled lineOut and LineIn outs by a lot of number.But    it never work.How can I overcome of this problem?I spent a lot of time unfortunately. Or  I wrote more effective code.Could you help me please?Best Regards..