VSL Summary Statistics VSL_SS_SUM error

[Shigeo_Kobayashi]

Pls let me know why the code bellow shows invalid results(sum) when NR is greater than 1999.

#define NR  2000   // MEAN=1 but invalid SUM !
//#define NR  1999 // MEAN=1 and SUM==1999 of course
#define DIM 4
void Test()
{
    VSLSSTaskPtr task;
    int errcode,dim = DIM, n=NR,x_storage=VSL_SS_MATRIX_STORAGE_COLS;
    double x[NR*DIM],mean[DIM], sum[DIM],W[2];
    W[0]=0; W[1]=0;
    for(int j=0;j<DIM;++j)   {mean=0;sum=0;}
    for(int i=0;i<NR*DIM;++i) x=1;
    errcode = vsldSSNewTask( &task, &dim, &n, &x_storage, (double*)x, 0, 0 );
    errcode = vsldSSEditTask(task,VSL_SS_ED_ACCUM_WEIGHT,W);
    errcode = vsldSSEditTask(task,VSL_SS_ED_MEAN,mean);
    errcode = vsldSSEditTask(task,VSL_SS_ED_SUM,sum);
    errcode = vsldSSCompute( task, VSL_SS_MEAN | VSL_SS_SUM, VSL_SS_METHOD_FAST );
    for(int i=0;i<dim;++i)  printf("M[%d] %g  S[%d] %g\n",i,mean,i, sum);
    errcode = vslSSDeleteTask( &task );
}

The result when NR=2000 is:

M[0] 1  S[0] -2.65698e+303
M[1] 1  S[1] -2.65698e+303
M[2] 1  S[2] -2.65698e+303
M[3] 1  S[3] -2.65698e+303

I got this result under
  Windows 8.1
  Intel(R) Math Kernel Library Version 11.1.3 Product Build 20140416 for 32-bit applications
 

[mecej4]

I tried your program (after changing "Test" to "main") with the 14.0.2.176 32-bit compiler and MKL 11.1.3 on Windows 8.1 X64 , and I could not reproduce the problem. Please state the compiler version number and specify the compiler options used.

[Shigeo_Kobayashi]

Hi mecej4,
Thank you for your quick follow!
Ok, It may depend on some optimization done by the compiler used.
I am using MS Visual studio 2012 and the program is  in some other tesiting stuffs.
I will extract it alone and make a new simple solution.
I will upload whole solution files if I succeed to reproduce the error.
 

[Shigeo_Kobayashi]

MS Visual studio may have bug.
I got correct result in 'Release' mode,the error was in 'Debug' mode!!! 
I should have tested it in both modes even for such a simple example ...
But this is not a MKL issue anyway... thank you.