The fish and sharks application motivated a separate study of the problem of computing large sparse histograms. By sparse we mean that the histogram table is large enough that the probability that two threads try to update the same cell at the same time is small.