Question

I have seen a lot of parallel programming code like finding the maximum of array, matrix multiplication, etc. use pointers. I don’t understand why it is used. Example:(shseg+(offset*sizeof(float))) = sum;

The code for matrix multiplication:

shseg = shmat(handle,NULL,0);
for(row=SIZE/2;row<SIZE;row++){

  for(column=0;column<SIZE;column++){
    sum = 0;

    for(tindex=0;tindex<SIZE;tindex++){ 
      sum+=a[row][tindex]*b[tindex][column];
    }

    *(shseg+(offset*sizeof(float))) = sum;
    offset++;
  }
}

Can anyone explain why a pointer is used?

Answer 1

This is because the example you show uses shared memory API, which provides you a flat chunk of memory, not an array of, say, floats. Therefore, you need to do all your pointer manipulations manually.

You could also cast your shared pointer to float* and use an index, like this:

shseg = shmat(handle,NULL,0);
float *fshseg = (float*)shseg;

...

fshseg[index++] = sum;