We present an experimental study of the single source shortest path problem with non-negative edge weights (NSSP) on large-scale graphs using the $\Delta$-stepping parallel algorithm. We report performance results on the Cray MTA-2, a multithreaded parallel computer. The MTA-2 is a high-end shared memory system offering two unique features that aid the efficient parallel implementation of irregular algorithms: the ability to exploit fine-grained parallelism, and low-overhead synchronization primitives. Our implementation exhibits remarkable parallel speedup when compared with competitive sequential algorithms, for low-diameter sparse graphs. For instance, $\Delta$-stepping on a directed scale-free graph of 100 million vertices and 1 billion edges takes less than ten seconds on 40 processors of the MTA-2, with a relative speedup of close to 30. To our knowledge, these are the first performance results of a shortest path problem on realistic graph instances in the order of billions of vertices and edges.

We study the average-case complexity of the parallel single-source shortest-path (SSSP) problem, assuming arbitrary directed graphs with n nodes, m edges, and independent random edge weights uniformly distributed in [0; 1]. We provide a new bucket-based parallel SSSP algorithm that runs in T = O(log 2 n min i f2 i L + jV i jg) average-case time using O(n+m+T ) work on a PRAM where L denotes the maximum shortest-path weight and jV i j is the number of graph vertices with in-degree at least 2 i . All previous algorithms either required more time or more work. The minimum performance gain is a logarithmic factor improvement; on certain graph classes, accelerations by factors of more than n 0:4 can be achieved. The algorithm allows adaptation to distributed memory machines, too.

In this paper we discuss the problem of determining a meeting point of a set of scattered robots R = {r1,r2,...,rs} in a weighted terrain P which has n> s triangular faces. Our algorithmic approach is to produce a discretization of P by producing a graph G = {V G,E G} which lies on the surface of P. For a chosen vertex p ′ ∈ V G, we define ‖Π(ri,p ′)‖ as the minimum weight cost of traveling from ri to p ′. We show that minp ′ ∈V G{max1≤i≤s{‖Π(ri,p ′)‖}} ≤ minp∗∈P{max1≤i≤s{‖Π(ri,p ∗)‖}} + W |L | where L is the longest edge of P, W is the maximum cost weight of a face of P, and p ∗ is the optimal solution. Our algorithm requires O(snmlog(snm)+snm 2) time to run, where m = n in the Euclidean metric and m = n2 in the weighted metric. However, we show through experimentation that only a constant value of m is required (e.g., m=8) in order to produce very accurate solutions (< 1 % error). Hence, for typical terrain data, the expected running time of our algorithm is O(snlog(sn)). Also, as part of our experiments we show that by using geometrical subsets (i.e., 2D/3D convex hulls, 2D/3D bounding boxes and random selection) of the robots we can improve the running time for finding p ′, with minimal or no additional accuracy error when comparing p ′ to p ∗. 1

Abstract. Two new ways of transforming a priority queue into a double-ended priority queue are introduced. These methods can be used to improve all known bounds for the comparison complexity of double-ended priority-queue operations. Using an efficient priority queue, the first transformation can produce a doubleended priority queue which guarantees the worst-case cost of O(1) for find-min, find-max, and insert; and the worst-case cost of O(lg n) including at most lg n + O(1) element comparisons for delete, but the data structure cannot support meld efficiently. Using a meldable priority queue that supports decrease efficiently, the second transformation can produce a meldable double-ended priority queue which guarantees the worst-case cost of O(1) for find-min, find-max, and insert; the worst-case cost of O(lg n) including at most lg n + O(lg lg n) element comparisons for delete; and the worst-case cost of O(min {lg m, lg n}) for meld. Here, m and n denote the number of elements stored in the data structures prior to the operation in question, and lg n is a shorthand for log 2 (max {2, n}). 1.

. The complexity of maintaining a set under the operations Insert, Delete and FindMin is considered. In the comparison model it is shown that any randomized algorithm with expected amortized cost t comparisons per Insert and Delete has expected cost at least n=(e2 2t ) \Gamma 1 comparisons for FindMin. If FindMin is replaced by a weaker operation, FindAny, then it is shown that a randomized algorithm with constant expected cost per operation exists; in contrast, it is shown that no deterministic algorithm can have constant cost per operation. Finally, a deterministic algorithm with constant amortized cost per operation for an offline version of the problem is given. CR Classification: F.2.2 1. Introduction We consider the complexity of maintaining a set S of elements from a totally ordered universe under the following operations: Insert(x): inserts the element x into S, Delete(x): removes from S the element x provided it is known where x is stored, and Supported by the Danish...

In the last three decades a considerable amount of research has been pursued in the efficient implementation of the pending event set (PES) associated with discrete-event simulation. The reason is simple: a fast event management has a very crucial impact in the total running time of both sequential and parallel simulations. This report focuses on this problem by studying the empirical performance of a number of solutions to the PES implementation in which we include a complete binary tree described in [26], 1 Introduction The PES is defined as the set of all the events generated during a discrete-event simulation and whose occurrence have not been simulated yet. In order to determine the next event to take place, it is necessary to extract the event with the least time from the PES. We call this operation extract-min. On the other hand, the occurrence of any event during the simulation can produce the insertion of new pending or future events in the PES; insert operation. These two b...

We address the question of theoretical vs. practical behavior of algorithms for the minimum spanning tree problem. We review the factors that influence the actual running time of an algorithm, from choice of language, machine, and compiler, through low-level implementation choices, to purely algorithmic issues. We discuss how to design a careful experimental comparison between various alternatives. Finally, we present some results from an ongoing study in which we are using: multiple languages, compilers, and machines; all the major variants of the comparison-based algorithms; and eight varieties of graphs with sizes of up to 130,000 vertices (in sparse graphs) or 750,000 edges (in dense graphs). 1 Introduction Finding spanning trees of minimum weight (minimum spanning trees or MSTs) is one of the best known graph problems; algorithms for this problem have a long history, for which see the article of Graham and Hell [6]. The best comparison-based algorithm to date, due to Gabow...

We give a simple implementation of meldable priority queues, achieving Insert , Find min, and Meld in O(1) worst case time, and Delete min and Delete in O(log n) worst case time. 1 Introduction The implementation of priority queues is a classic problem in computer science. The fundamental operations are Insert and Delete min, but various extra operations, such as Find min, Meld, Delete, and Decrease key , have been considered. Fibonacci heaps [6] support all of these, in O(log n) time for the Delete min and Delete operations, and O(1) for the rest. These bounds are, however, only amortized. Some earlier proposals [4, 5, 8] achieve such bounds in the worst case sense for various subsets of the operations supported by Fibonacci heaps. None of these subsets includes the meld operation. This has been remedied recently by Brodal, who has given worst case solutions, first [2] for the set Insert, Delete min, Find min, Meld, and Delete, and later [3] for the full set of Fibonacci heap opera...

Abstract. A simplification of a run-relaxed heap, called a relaxed weak queue, is presented. This new priority-queue implementation supports all operations as efficiently as the original: find-min, insert, and decrease (also called decrease-key) in O(1) worst-case time, and delete in O(lg n) worst-case time, n denoting the number of elements stored prior to the operation. These time bounds are valid on a pointer machine as well as on a random-access machine. A relaxed weak queue is a collection of at most ⌊lg n ⌋ + 1 perfect weak heaps, where there are in total at most ⌊lg n ⌋ + 1 nodes that may violate weak-heap order. In a pointer-based representation of a perfect weak heap, which is a binary tree, it is enough to use two pointers per node to record parent-child relationships. Due to decrease, each node must store one additional pointer. The auxiliary data structures maintained to keep track of perfect weak heaps and potential violation nodes only require O(lg n) words of storage. That is, excluding the space used by the elements themselves, the total space usage of a relaxed weak queue can be as low as 3n + O(lg n) words. ACM CCS Categories and Subject Descriptors. E.1 [Data Structures]: Lists, stacks, and queues; E.2 [Data Storage Representations]: Linked representations;

this paper we study the performance of the very first tournament based complete binary tree. We focus on discrete-event simulation and our results show that this unknown predecessor of heaps can be a more efficient alternative to the fastest pending event set implementations reported in the literature. We also extend the idea of binary tournaments to a (2; L)-tournament structure which exhibits the property of delaying the processing of events with larger timestamps whilst it keeps similar theoretical performance bounds to the native (2; 1)-structure or CBT. This property can be certainly useful in systems where many pending events are expected to be deleted or rescheduled during the simulation. 2 Tournament trees