AU EIWRICAL COIUPARISON OF PRIORITY-QUEUE AUD EVEUFSET lMPLElWEUTd?lONS

During the last decade, a number of new priorityqueue implementations have evolved, including pagodas [6, 261, skew heaps [21, 221, splay trees [21, 231, pairing heaps 1183, and binomial queues [4, 5, 241. In addition, a number of special priority-queue implementations for representing the pending event set in discrete event simulation Shave been developed, including the two-list implementation [3] and Henriksen’s implementation [lo, 151. The resulting variety of implementations is reminiscent of the variety of sorting algorithms. Comparison between the newer priority-queue implementations and the older ones suggests that :many of the former are nearly optimal, but that no single implementation is best for all applications. The basic operations on the priority-queue abstract data type are enqueue and dequeue (sometimes called insert and delete-min [l]). Enqueue places an item in a priority queue, and dequeue removes and returns the highest priority item from a priority queue. By convention, increasing priorities correspond to decnaasing numerical values. As with all abstract data types, there are many ways to implement a priority queue; each implementation involves a data structure used to represent the queue, and an algorithm that implements each abstract operation in terms of this data structure.