Issue during replacing ipp DCT function with MKL DCT function

[Ashwin_A_1]

Hi,

I want to replace my IPP based DCT function with mkl based DCT function .

I am getting different output data when I will cross check with the ipp DCT vs mkl DCT function output.

I used below functions to get the DCT by usng IPP.lib function calls :

ippsDCTFwdInitAlloc_32f
ippsDCTFwd_32f
ippsDCTFwdFree_32f

Below is my code :

//pfa of the fileinput.txt 

int main(int argc, char* argv[]){

    float *dpar;
    float *out;
    MKL_INT    *ipar;
    MKL_INT tt_type,stat,n_1,nn;
    FILE *fp,*fw,*fonce;
    fp = fopen( "D:\\dump\\fileinput.txt","r" );
    if(fp == NULL){
        cout<<"file not created properly"<<endl;
    }
    DFTI_DESCRIPTOR_HANDLE handle = 0;
    int n = 65; //Hardcoded to run for my code TODO:going to change after integrating into my main codebase
    nn = (MKL_INT)n;
    tt_type = MKL_STAGGERED_COSINE_TRANSFORM;

    n_1 = nn + 1 ;
    out = (float*)malloc((n+1)*sizeof(float));
    dpar= (float*)malloc((5*n_1/2+2)*sizeof(float));
    ipar= (MKL_INT*)malloc((128)*sizeof(int));
    s_init_trig_transform(&n_1,&tt_type,ipar,dpar,&stat);
    for (int srcSize =0 ;srcSize< n ; srcSize++)
    {
        fscanf(fp,"%f\n",&out[srcSize]);
    }
    fclose(fp);
    if (stat != 0)
    {
        printf("\n============================================================================\n");
        printf("FFTW2MKL FATAL ERROR: MKL TT initialization has failed with status=%d\n",(MKL_INT)stat);
        printf("Please refer to the Trigonometric Transform Routines Section of MKL Manual\n");
        printf("to find what went wrong...\n");
        printf("============================================================================\n");
        return NULL;
    }
    ipar[10] = 1;    //nx, that is, the number of intervals along the x-axis, in the Cartesian case.
    ipar[14] = n_1;  //specifies the internal partitioning of the dpar array.
    ipar[15] = 1;    //value of ipar[14]+1,Specifies the internal partitioning of the dpar array.
    s_commit_trig_transform(out,&handle,ipar,dpar,&stat);
    if (stat != 0)
    {
        printf("\n============================================================================\n");
        printf("FFTW2MKL FATAL ERROR: MKL TT commit step has failed with status=%d\n",(MKL_INT)stat);
        printf("Please refer to the Trigonometric Transform Routines Section of MKL Manual\n");
        printf("to find what went wrong...\n");
        printf("============================================================================\n");
        return NULL;
    }
    s_forward_trig_transform(out,&handle,ipar,dpar,&stat);
    if (stat != 0)
    {
        printf("\n============================================================================\n");
        printf("FFTW2MKL FATAL ERROR: MKL TT commit step has failed with status=%d\n",(MKL_INT)stat);
        printf("Please refer to the Trigonometric Transform Routines Section of MKL Manual\n");
        printf("to find what went wrong...\n");
        printf("============================================================================\n");
        return NULL;
    }
    free_trig_transform(&handle,ipar,&stat);
    printf("\n===== DCT GOT OVER ======== \n");

    return 0;

}

[Ashwin_A_1]

More update :

 I took 8x8 Matrix as input to to above mentioned code with tt_type = MKL_COSINE_TRANSFORM;  

255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255
255    255    255    255    255    255    255    255

DCT from  above code is wrongly came as -
                            
32385             -727.349243     -619.955444    -458.675903    -267.323547    -74.151535     92.318069       207.505844
764.404419     717.50531       602.316772      435.846893     242.6754          51.322887    -109.956779    -217.350159
-762.619629    -706.582764    -583.869446    -412.598663    -218.058334    -28.972321    126.74057       226.088898
759.648132     694.603516     564.651367      388.979492    193.525208      7.147152      -142.635284     -233.70639
-755.498901    -681.59967     -544.712891    -365.049927    -169.138          14.096443     157.598663     240.180206
750.180115     667.597656     524.094971     340.860809    144.948898    -34.714363    -171.600922     -245.499771
-743.705566    -652.633789    -502.85144     -316.473694    -121.019257    54.652824    184.604614     249.648224
736.088623     636.739502     481.026154     291.940552     97.400169    -73.870857    -196.583725     -252.620102

It is totally wrong .

Can anyone tell me the mistake in my above code .

Please make a note that I did Direct DCT by initializing tt_type = MKL_COSINE_TRANSFORM;  

[Ying_H_Intel]

Hi Ashwin, 

Just let you  know, I'm investigating the problem with our developer, but may be a little delay as new year holiday.  Sorry for the delay

Happy new year!

Thanks

Ying

[Ying_H_Intel]

Hi Ashwin, 

We find the problem: 

s_init_trig_transform(&n_1,&tt_type,ipar,dpar,&stat);

the n_1  in init functionality – number of intervals, for cosine transform.  So  it is equal number of points minus one.

if you try the below fix, 

int n = 64; //Hardcoded to run for my code TODO:going to change after integrating into my main codebase

    nn = (MKL_INT)n;
    tt_type = MKL_COSINE_TRANSFORM; ;

    n_1 = n - 1 ;

You will get the correct result. 

 Ipar[10] and other  are internal parameters, we didn’t allow to change them. 

===== DCT GOT OVER ========

 pDst 16065.000000

pDst 0.000000

pDst 0.000122

pDst 0.000052

pDst 0.000000

pDst 0.000052

pDst 0.000036

pDst 0.000145

pDst -0.000002

pDst 0.000220

But please notes  It is at 8 time greater than ipp one because ipp result is divided by sqrt N. 

Best Regards,

Ying 

and in case if  you want to use theMKL_STAGGERED_COSINE_TRANSFORM,  according the documents, 

/f[64] for staggered cosine transforms = 0; 

int n = 65;

tt_type =  MKL_STAGGERED_COSINE_TRANSFORM

//; // MKL_COSINE_TRANSFORM; //MKL_STAGGERED_COSINE_TRANSFORM;

float pSrc[65]={

255,    255 ,   255 ,   255 ,   255 ,   255,    255 ,   255,

255 ,   255 ,   255 ,   255  ,  255  ,  255 ,   255,    255,

255 ,   255 ,   255,    255 ,   255 ,   255 ,   255,    255,

255,    255 ,   255 ,   255,    255 ,   255  ,  255 ,   255,

255 ,   255,    255 ,   255 ,   255  ,  255 ,   255 ,   255,

255 ,   255,    255,    255,    255,   255  ,  255  ,  255,

255,    255 ,   255 ,   255 ,   255,    255 ,   255 ,   255,

255,    255,    255,    255,    255,    255,    255,    255,

0};

    n_1 = n - 1 ;

s_init_trig_transform(&n_1,&tt_type,ipar,spar,&stat);

[Palanivel_G_]

Hi Ying,

I tried your suggested fixes and below are the comparative results between IPP and MKL implementations.

Input : 8x8 with all elements 255

IPP Output:

2040    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0    
0    0    0    0    0    0    0    0

MKL Output:

510    0    -2.12484e-007    -8.95694e-008    2.02821e-006    1.98985e-006    -2.44188e-007    1.36767e-006    
1.25338e-008    8.65079e-007    -4.82064e-007    1.93344e-006    2.85854e-008    1.90692e-006    0    1.90692e-006    
-2.95537e-008    1.90544e-006    6.00573e-007    2.19466e-006    -1.37825e-007    2.03849e-006    -6.08452e-008    1.85739e-006    
-3.83681e-008    2.36938e-006    8.6844e-009    2.35244e-006    0    2.35244e-006    -1.61479e-008    2.35742e-006    
4.02007e-007    2.39765e-006    -1.16682e-007    2.33411e-006    -7.94424e-008    2.23449e-006    9.64763e-008    2.57738e-006    
1.70139e-009    2.56381e-006    0    2.56381e-006    -1.02961e-008    2.57161e-006    3.07488e-007    2.49331e-006    
-1.0582e-007    2.47736e-006    -8.96722e-008    2.42256e-006    1.75386e-007    2.66651e-006    -2.62343e-009    2.65502e-006    
0    2.65502e-006    -6.2848e-009    2.66475e-006    2.39635e-007    2.50137e-006    -9.76727e-008    2.52111e-006

The output is ~4 times lesser than the DCT output.

Could you please let me know what is going on wrong here?.

The code snippet used in MKL are as below,

    int n = 64; // 8x8 input, so 64 co-efficients
    nn = (MKL_INT)n;
    tt_type = MKL_COSINE_TRANSFORM;

    n_1 = n - 1 ; // no of intervals
    out = (float*)   malloc ((n + 1) * sizeof(float));
    dpar= (float*)   malloc (((5 * (n_1 / 2)) + 2) * sizeof(float));
    ipar= (MKL_INT*) malloc ((128) * sizeof(int));

    s_init_trig_transform(&n_1, &tt_type, ipar, dpar, &stat);

    /*
    ** Populate the input
    */
    for (int k = 0; k < n; k++) {
           out = 255.0;
    }

     s_commit_trig_transform(out,&handle,ipar,dpar,&stat);

     s_forward_trig_transform(out,&handle,ipar,dpar,&stat);

    Thanks in advance.

Best regards

Palanivel

[Ying_H_Intel]

Hi Palanivel, 

I upload the whole cpp file. You may check what is difference for the result.  (sorry, I upload again)

Regards,

Ying

[Palanivel_G_]

Hi Ying,

Thanks for the reply.

Request you to please upload the complete cpp file.

Thanks,

Palanivel

[Ying_H_Intel]

Hi Palanivel,

I attached the test code with MKL_COSINE_TRANSFORM  in # 6   and theMKL_STAGGERED_COSINE_TRANSFORM  in this reply.

Best Regards,

Ying