Opticalcommunication is likely to significantly speed up parallel computation because the vast bandwidth of the optical medium can be divided to produce communication networks of very high degree. However, the problem...
详细信息
Opticalcommunication is likely to significantly speed up parallel computation because the vast bandwidth of the optical medium can be divided to produce communication networks of very high degree. However, the problem of contention in high-degree networks makes the routing problem in these networks theoretically (and practically) difficult. In this paper we examine Valiant's h-relation routing problem, which is a fundamental problem in the theory of parallel computing. The h-relation routing problem arises both in the direct implementation of specific parallel algorithms on distributed-memory machines and in the general simulation of shared-memory models such as the PRAM on distributed-memory machines. In an h-relation routing problem each processor has up to h messages that it wishes to send to other processors and each processor is the destination of at most h messages. We present a lower bound for routing an h-relation (for any h > 1) on a complete optical network of size n. Our lower bound applies to any randomized distributed algorithm for this task. Specifically, we show that the expected number of communication steps required to route an arbitrary h-relation is
In this paper we demonstrate the power of reconfiguration by presenting efficient randomized algorithms for both packet routing and sorting on a reconfigurable mesh connected computer. The run times of these algorithm...
详细信息
In this paper we demonstrate the power of reconfiguration by presenting efficient randomized algorithms for both packet routing and sorting on a reconfigurable mesh connected computer. The run times of these algorithms are better than the best achievable time bounds on a conventional mesh. Many variations of the reconfigurable mesh can be found in the literature. We define yet another variation which we call as M r . We also make use of the standard PARBUS model. We show that permutation routing problem can be solved on a linear array M r of size n in steps, whereas n-1 is the best possible run time without reconfiguration. A trivial lower bound for routing on M r will be . On the PARBUS linear array, n is a lower bound and hence any standard n-step routing algorithm will be optimal. We also show that permutation routing on an n×n reconfigurable mesh M r can be done in time n+o(n) using a randomized algorithm or in time 1.25n+o(n) deterministically. In contrast, 2n-2 is the diameter of a conventional mesh and hence routing and sorting will need at least 2n-2 steps on a conventional mesh. A lower bound of is in effect for routing on the 2D mesh M r as well. On the other hand, n is a lower bound for routing on the PARBUS and our algorithms have the same time bounds on the PARBUS as well. Thus our randomized routing algorithm is optimal upto a lower order term. In addition we show that the problem of sorting can be solved in randomized time n+o(n) on M r as well as on PARBUS. Clearly, this sorting algorithm will be optimal on the PARBUS model. The time bounds of our randomized algorithms hold with high probability.
In this paper, we discuss two variations of the two-dimensional post-office problem that arise when the post-offices are n postmen moving with constant velocities. The first variation addresses the question: given a p...
详细信息
In this paper, we discuss two variations of the two-dimensional post-office problem that arise when the post-offices are n postmen moving with constant velocities. The first variation addresses the question: given a point q(0) and time t(0) who is the nearest postman to q(0) at time t(0)? We present a randomized incremental data structure that answers the query in expected O(log(2) n) time. The second variation views a query point as a dog searching for a postman to bite and finds the postman that a dog running with speed v(d) could reach first. While it is quite straightforward to design a data structure for the first problem, designing one for the second appears more difficult. We show that if the dog is quicker than all of the postmen then there is a nice correspondence between the problems. This correspondence will permit us to use the data structure developed for the first problem to solve the second one in O(log(2) n) time as well. The proposed structure is semi-dynamic, that is the set of postmen can be modified by inserting new postmen. A fully dynamic structure that also supports deletions can be obtained, but in that case the query time becomes O(log(3) n).
A randomized (Las Vegas) algorithm is given for finding the Gallai-Edmonds decomposition of a graph. Let n denote the number of vertices, and let M(n) denote the number of arithmetic operations for multiplying two n X...
详细信息
A randomized (Las Vegas) algorithm is given for finding the Gallai-Edmonds decomposition of a graph. Let n denote the number of vertices, and let M(n) denote the number of arithmetic operations for multiplying two n X n matrices. The sequential running time (i.e., number of bit operations) is within a poly-logarithmic factor of M(n). The parallel complexity is O((log n)(2)) parallel time using a number of processors within a poly-logarithmic factor of M(n). The same complexity bounds suffice for solving several other problems: (i) finding a minimum vertex cover in a bipartite graph, (ii) finding a minimum X --> Y vertex separator in a directed graph, where X and Y are specified sets of vertices, (iii) finding the allowed edges (i.e., edges that occur in some maximum matching) of a graph, and (iv) finding the canonical partition of the vertex set of an elementary graph. The sequential algorithms for problems (i), (ii), and (iv) are Las Vegas, and the algorithm for problem (iii) is Monte Carlo. The new complexity bounds are significantly better than the best previous ones, e.g., using the best value of M(n) currently known, the new sequential running time is O(n(2.38)) versus the previous best O(n(2.5)/(log n)) or more.
A randomized algorithm for finding a hyperplane separating two finite point sets in the Euclidean space Rd and a randomized algorithm for solving linearly constrained general convex quadratic problems are proposed. Th...
详细信息
A randomized algorithm for finding a hyperplane separating two finite point sets in the Euclidean space Rd and a randomized algorithm for solving linearly constrained general convex quadratic problems are proposed. The expected running time of the separating algorithm is O(dd! (m + n)), where m and n are cardinalities of sets to be separated. The expected running time of the algorithm for solving quadratic problems is O(dd! s) where s is the number of inequality constraints. These algorithms are based on the ideas of Seidel's linear programming algorithm [6]. They are closely related to algorithms of [8], [2], and [9] and belong to an abstract class of algorithms investigated in [1]. The algorithm for solving quadratic problems has some features of the one proposed in [7].
Suppose we are given two sets R and B, each of n points in the plane. Define the cost of a matching to be the teal distance of the edges in the, matching. The minimum matching problem on Euclidean space is to find a c...
详细信息
Suppose we are given two sets R and B, each of n points in the plane. Define the cost of a matching to be the teal distance of the edges in the, matching. The minimum matching problem on Euclidean space is to find a complete bipartite matching which has the minimum cost on Euclidean space. In this paper, we are interested in the on-line minimum matching problem on Euclidean space. We assume the set R is known, but the points of set B are revealed one by one. When a point of set B arrives, we must decide what match to make involving the point. None of the matches can be changed after they are made. The on-line minimum matching problem on Euclidean space tries to minimize the cost of the complete bipartite matching that we find. This paper proposes a family of randomized algorithms, Algorithm RM(m) (1 less than or equal to m less than or equal to n), for solving this problem. When a point in the set B arrived, Algorithm RM(m) randomly chooses at most m unmatched points in the set R, and adds the minimum edge between the arriving point and the chosen points to the matching. In each decision step, the algorithm only runs in O(m) time, which is superior to the known non-randomized algorithms for this problem. In this paper, we show that Algorithm RM(m) is not a competitive on-line algorithm for 1 less than or equal to m less than or equal to n-1. However, we further show that if 2n is large, the average cost incurred by Algorithm RM(m) (1 less than or equal to m less than or equal to n) is bounded by O(root n) times the average cost of the optimal Euclidean minimum matching.
Reliability of compute-intensive applications can be improved by introducing fault tolerance into the system. Algorithm-based fault tolerance (ABFT) is a low-cost scheme which provides the required fault tolerance to ...
详细信息
Reliability of compute-intensive applications can be improved by introducing fault tolerance into the system. Algorithm-based fault tolerance (ABFT) is a low-cost scheme which provides the required fault tolerance to the system through system level encoding. In this paper, we propose randomized construction techniques, under an extended model, for the design of ABFT systems with the required fault tolerance capability. The model considers failures in the processors performing the checking operations.
A massively parallel optimization approach based on simple neighbourhood search techniques is developed and applied to the problem of VLSI cell placement. Statistical models are developed to analyse the performance of...
详细信息
A massively parallel optimization approach based on simple neighbourhood search techniques is developed and applied to the problem of VLSI cell placement. Statistical models are developed to analyse the performance of the approach in general, and to derive statistical bounds on the quality of obtainable results. Specific questions addressed are: (1) Given a solution with a known cost, how can we measure its quality? (2) Given a target cost for the solution, how likely is the algorithm to generate a solution with that cost or better? (3) Are there any performance bounds for the solutions obtainable by neighbourhood search methods? (4) How can we measure or quantify the performance of different neighbourhood search methods? The results of these analyses suggest a simple framework for approximate solution of difficult problems. The approach is inherently parallel, and it can be implemented on any type of parallel computer. We implemented it on the PVM environment running on a network of workstations connected by Ethernet. The method is empirically verified by testing its performance on a number of sample problems and by comparing the results found to earlier results reported in the literature.
Given a pattern string of length m. for the string-matching problem, we design an algorithm that computes deterministic samples of a sufficiently long substring of the pattern in constant time. This problem used to be...
详细信息
Given a pattern string of length m. for the string-matching problem, we design an algorithm that computes deterministic samples of a sufficiently long substring of the pattern in constant time. This problem used to be the bottleneck in the pattern preprocessing for one- and two-dimensional pattern matching. The best previous time bound was O(log(2) m/log log m). We use this algorithm to obtain the following results (all algorithms below are optimal parallel algorithms on a CRCW PRAM): 1. a deterministic string-matching algorithm which takes O(log log m) time for preprocessing and constant time for text search, which are the best possible in both preprocessing and text search;2. a constant-time deterministic string-matching algorithm in the case where the text length n satisfies n = Omega(m(1+epsilon)) for a constant epsilon > 0;3. a simple string-matching algorithm that has constant time with high probability for random input;4. the main result: a constant-expected-time Las Vegas algorithm for computing the period of the pattern and all witnesses and thus for string matching itself;in both cases, an Omega(log log m) lower bound is known for deterministic algorithms.
We present a lower bound of 1 + e(-1/2) approximate to 1.6065 on the competitive ratio of randomized algorithms for the weighted 2-cache problem, which is a special case of the 2-server problem. This improves the Prev...
详细信息
We present a lower bound of 1 + e(-1/2) approximate to 1.6065 on the competitive ratio of randomized algorithms for the weighted 2-cache problem, which is a special case of the 2-server problem. This improves the Previously best known lower bound of e/(e-1) approximate to 1.582 for both problems. (C) 1997 Elsevier Science B.V.
暂无评论