Beyond the worst-case bisection bound: Fast sorting and ranking on meshes

Sorting is an important subroutine in many parallel algorithms and has been studied extensively on meshes and related networks. If every processor of an n × n mesh is the source and destination of at most k elements, then sorting requires at least k k; n/2 steps in the worst-case, and simple algorithms have recently been proposed that nearly match this bound. However, this lower bound does not extend to non-worst-case inputs, or weaker definitions of sorting that are sufficient in many applications. In this paper, we give algorithms and lower bounds for several such problems. We first present a very simple scheme for k-k routing that performs optimally under both average-case and worst-case inputs. As an application of this scheme, we describe a simple k-k sorting algorithm based on sample sort that nearly matches this bound. The main part of the paper considers several 'sorting-like' problems. In the ranking problem, the ranks of all elements have to be determined, but there is no requirement about their final positions. We describe an algorithm running in time (1 +o(l)) k n/4 steps, which is nearly optimal under the considered model of the mesh. We show that integer versions of the sorting and ranking problems, where keys are drawn from {0,…, m — 1}, can be solved asymptotically faster than the general problems for small values of m. A related problem, the excess counting problem, can be solved in O(n) steps in many interesting cases.