Sharing Large Memory Variable in a parfor

I have a code which uses a very large variable of size [3 10 61 50 50 50 60], which is about 55 Gb. The form of the code is roughly like the following:

 numthreads = 8;
 numvals = 10^6;
 load('dat.mat') % loads the big matrix "data"
 for i=1:numthreads 
    vals{i} = zeros(numvals,1);
 end

 parfor i=1:numthreads
    for j=1:numvals
      a  = randi(3);
      b  = randi(10);
      c  = randi(61);
      d  = randi(50);
      e  = randi(50);
      f  = randi(50);
      g  = randi(60);
      refind = sub2ind([3 10 61 50 50 50 60],a,b,c,d,e,f,g);
      vals{i}(j) = data(refind);
    end
 end

Unfortunately, because the matrix 'data is 55 Gb and i only have 64 Gb, I cannot run the above code.

Is there a way I can get around this? Is it possible, for example, to access the matrix directly from the hard drive? If so, perhaps I can load it on a RAM Disk and get fast access as well.

Thanks.