An E cient Algorithm for Concurrent Priority

We present a new algorithm for concurrent access to array-based priority queue heaps. Deletions proceed top-down as they do in a previous algorithm due to Rao and Kumar 6], but insertions proceed bottom-up, and consecutive insertions use a bit-reversal technique to scatter accesses across the fringe of the tree, to reduce contention. Because insertions do not have to traverse the entire height of the tree (as they do in previous work), as many as O(M) operations can proceed in parallel, rather than O(logM) on a heap of size M. Experimental results on a Silicon Graphics Challenge multiprocessor demonstrate good overall performance for the new algorithm on small heaps, and signiicant performance improvements over known alternatives on large heaps with mixed insertion/deletion workloads. 1 void insert(int priority, int data, heap_t *heap) { LOCK(heap->lock); i = ++heap->size; ip = &heap->itemi]; ip->pri = priority; ip->dat = data; while(i>1) { parent = i>>1; pp = &heap->itemmparent]; if(ip->pri < pp->pri) SWAP(ip,pp); /* swap items i and parent */ else break; i = parent; ip = pp; } UNLOCK(heap->lock); } int delete(int *priority, int *data, heap_t *heap) { LOCK(heap->lock); if(heap->size < 1){ UNLOCK(heap->lock); return 0; } i=1; ip = &heap->itemm1]; *priority = ip->pri; *data = ip->dat; lp = &heap->itemmheap->size-]; ip->pri = lp->pri; ip->dat = lp->dat; while(i <= heap->size>>1){ min = MIN_CHILD(i); /* index of child with higher priority */ mp = &heap->itemmmin]; if(ip->pri > mp->pri) SWAP(ip,mp); /* swap items i and min */ else break; i = min; ip = mp; } UNLOCK(heap->lock); return 1; } Figure 1: Single-lock heap operations.