Search

Abstract. External memory (EM) algorithms are designed for large-scale computational problems in which the size of the internal memory of the computer is only a small fraction of the problem size. Typical EM algorithms are specially crafted for the EM situation. In the past, several attempts have been made to relate the large body of work on parallel algorithms to EM, but with limited success. The combination of EM computing, on multiple disks, with multiprocessor parallelism has been posted as a challenge by the ACM Working Group on Storage I/ O for Large-Scale Computing. In this paper we provide a simulation technique which produces efficient parallel EM algorithms from efficient BSP-like parallel algorithms. The techniques obtained can accommodate one or multiple processors on the EM target machine, each with one or more disks, and they also adapt to the disk blocking factor of the target machine. When applied to existing BSP-like algorithms, our simulation technique produces improved parallel EM algorithms for a large number of problems. [ABSTRACT FROM AUTHOR]

In this paper, we study the time-dependent shortest paths problem for two types of time-dependent FIFO networks. First, we consider networks where the availability of links, given by a set of disjoint time intervals for each link, changes over time. Here, each interval is assigned a non-negative real value which represents the travel time on the link during the corresponding interval. The resulting shortest path problem is the time-dependent shortest path problem for availability intervals ( $\mathcal{TDSP}_{\mathrm{int}}$), which asks to compute all shortest paths to any (or all) destination node(s) d for all possible start times at a given source node s. Second, we study time-dependent networks where the cost of using a link is given by a non-decreasing piece-wise linear function of a real-valued argument. Here, each piece-wise linear function represents the travel time on the link based on the time when the link is used. The resulting shortest paths problem is the time-dependent shortest path problem for piece-wise linear functions ( $\mathcal{TDSP}_{\mathrm{lin}}$) which asks to compute, for a given source node s and destination d, the shortest paths from s to d, for all possible starting times. We present an algorithm for the $\mathcal{TDSP}_{\mathrm{lin}}$ problem that runs in time O(( F+ γ)(| E|+| V|log | V|)) where F is the output size (i.e., number of linear pieces needed to represent the earliest arrival time function to d) and γ is the input size (i.e., number of linear pieces needed to represent the local earliest arrival time functions for all links in the network). We then solve the $\mathcal{TDSP}_{\mathrm{int}}$ problem in O( λ(| E|+| V|log | V|)) time by reducing it to an instance of the $\mathcal{TDSP}_{\mathrm{lin}}$ problem. Here, λ denotes the total number of availability intervals in the entire network. Both methods improve significantly on the previously known algorithms. [ABSTRACT FROM AUTHOR]

Algorithmic methods based on the theory of fixed-parameter tractability are combined with powerful computational platforms to launch systematic attacks on combinatorial problems of significance. As a case study, optimal solutions to very large instances of the NP-hard vertex cover problem are computed. To accomplish this, an efficient sequential algorithm and various forms of parallel algorithms are devised, implemented, and compared. The importance of maintaining a balanced decomposition of the search space is shown to be critical to achieving scalability. Target problems need only be amenable to reduction and decomposition. Applications in high throughput computational biology are also discussed. [ABSTRACT FROM AUTHOR]

Focuses on a parallel list ranking algorithm for distributed memory multiprocessors. Description of a simple version; Improved version requiring high probability; Reid-Miller's algorithm as a list ranking implementation; Practical relevance of the results.

We describe the design and implementation of a workbench for computational geometry. We discuss issues arising from this implementation, including comparisons of different algorithms for constant factors, code size, and ease of implementation. The workbench is not just a library of computational geometry algorithms and data structures, but is designed as a geometrical programming environment, providing tools for: creating, editing, and manipulating geometric objects; demonstrating and animating geometric algorithms; and, most importantly, for implementing and maintaining complex geometric algorithms. [ABSTRACT FROM AUTHOR]