how do i run this parallel programming using parfor?

i'm trying to run insertion sort as a parallel programming using parfor, but it gives me some kind of error which I cant fix, here is the code:-

clear A
t=cputime;

x = 1 : 100;

n = length(x);

parfor j = 2:n

      pivot = x(j);

      i = j;
      while ((i > 1) && (x(i - 1) > pivot))

          x(i) = x(i - 1);
          i = i - 1;
      end
      x(i) = pivot;
  end
  cputime-t

Error: The variable x in a parfor cannot be classified.
See Parallel for Loops in MATLAB, "Overview".

Answer 1

[Walter Roberson]

It is not allowed to access the distributed variable at two different indices.

The code you are doing that moves the pivot earlier and earlier would have different results depending upon the iterations for the earlier j had already been run or not. That is not allowed: every parfor iteration needs to be independent of the other iterations so that the order the iterations are run in does not matter (except for accumulated round-off error in "reduction variables".)

[mohamed Gabr]

I agree, got any ideas how to run it ?

[Walter Roberson]

I don't think I would try to run an insertion sort as a parallel routine.

If I did, I would work it on the idea of breaking up into blocks each done in parallel. Any sufficiently small block would be sorted "locally" (it is too expensive to go parallel below a good size.) Once you have blocks that are each partially sorted, you can start merging them, taking advantage of the partial ordering by keeping track of the maximum and minimum in the block to know whether a given value sorts before, after, or within the block. But getting this to work right would probably approach closely to just doing a quicksort.

[mohamed Gabr]

Thank you pretty much :D