The Cache Complexity of Multithreaded Cache Oblivious Algorithms

We present a technique for analyzing the number of cache misses incurred by multithreaded cache oblivious algorithms on an idealized parallel machine in which each processor has a private cache. We specialize this technique to computations executed by the Cilk work-stealing scheduler on a machine with dag-consistent shared memory. We show that a multithreaded cache oblivious matrix multiplication incurs cache misses when executed by the Cilk scheduler on a machine with P processors, each with a cache of size Z, with high probability. This bound is tighter than previously published bounds. We also present a new multithreaded cache oblivious algorithm for 1D stencil computations incurring cache misses with high probability, one for Gaussian elimination and back substitution, and one for the length computation part of the longest common subsequence problem incurring cache misses with high probability. This work was supported in part by the Defense Advanced Research Projects Agency (DARPA) under contract No. NBCH30390004.     

Scheduling with conflicts: online and offline algorithms

We consider the following problem of scheduling with conflicts (swc): Find a minimum makespan schedule on identical machines where conflicting jobs cannot be scheduled concurrently. We study the problem when conflicts between jobs are modeled by general graphs. Our first main positive result is an exact algorithm for two machines and job sizes in {1,2}. For jobs sizes in {1,2,3}, we can obtain a -approximation, which improves on the -approximation that was previously known for this case. Our main negative result is that for jobs sizes in {1,2,3,4}, the problem is APX-hard. Our second contribution is the initiation of the study of an online model for swc, where we present the first results in this model. Specifically, we prove a lower bound of on the competitive ratio of any deterministic online algorithm for m machines and unit jobs, and an upper bound of 2 when the algorithm is not restricted computationally. For three machines we can show that an efficient greedy algorithm achieves this bound. For two machines we present a more complex algorithm that achieves a competitive ratio of when the number of jobs is known in advance to the algorithm.     

Approximation Algorithms for Multi-Criteria Traveling Salesman Problems

We analyze approximation algorithms for several variants of the traveling salesman problem with multiple objective functions. First, we consider the symmetric TSP (STSP) with γ-triangle inequality. For this problem, we present a deterministic polynomial-time algorithm that achieves an approximation ratio of and a randomized approximation algorithm that achieves a ratio of . In particular, we obtain a 2+ε approximation for multi-criteria metric STSP. Then we show that multi-criteria cycle cover problems admit fully polynomial-time randomized approximation schemes. Based on these schemes, we present randomized approximation algorithms for STSP with γ-triangle inequality (ratio ), asymmetric TSP (ATSP) with γ-triangle inequality (ratio ), STSP with weights one and two (ratio 4/3) and ATSP with weights one and two (ratio 3/2). A preliminary version of this work has been presented at the 4th Workshop on Approximation and Online Algorithms (WAOA 2006) (Lecture Notes in Computer Science, vol. 4368, pp. 302–315, 2007). B. Manthey is supported by the Postdoc-Program of the German Academic Exchange Service (DAAD). He is on leave from Saarland University and has done part of the work at the Institute for Theoretical Computer Science of the University of Lübeck supported by DFG research grant RE 672/3 and at the Department of Computer Science at Saarland University.     

An Application of an Initialization Protocol to Permutation Routing in a Single-Hop Mobile Ad Hoc Networks

In 1999 Nakano, Olariu, and Schwing in [20], they showed that the permutation routing of n items pretitled on a mobile ad hoc network (MANET for short) of p stations (p known) and k channels (MANET{(n, p, k)) with k < p, can be carried out in broadcast rounds if k p and if each station has a -memory locations. And if k and if each station has a -memory locations, the permutations of these n pretitled items can be done also in broadcast rounds. They used two assumptions: first they suppose that each station of the mobile ad hoc network has an identifier beforehand. Secondly, the stations are partitioned into k groups such that each group has stations, but it was not shown how this partition can be obtained. In this paper, the stations have not identifiers beforehand and p is unknown. We develop a protocol which first names the stations, secondly gives the value of p, and partitions stations in groups of stations. Finally we show that the permutation routing problem can be solved on it in broadcast rounds in the worst case. It can be solved in broadcast rounds in the better case. Note that our approach does not impose any restriction on k.     

Polynomial-Time Algorithms for the Ordered Maximum Agreement Subtree Problem

For a set of rooted, unordered, distinctly leaf-labeled trees, the NP-hard maximum agreement subtree problem (MAST) asks for a tree contained (up to isomorphism or homeomorphism) in all of the input trees with as many labeled leaves as possible. We study the ordered variants of MAST where the trees are uniformly or non-uniformly ordered. We provide the first known polynomial-time algorithms for the uniformly and non-uniformly ordered homeomorphic variants as well as the uniformly and non-uniformly ordered isomorphic variants of MAST. Our algorithms run in time , , , and , respectively, where n is the number of leaf labels and k is the number of input trees.     

Many-to-Many Communication in Radio Networks

Radio networks model wireless data communication when the bandwidth is limited to one wave frequency. The key restriction of such networks is mutual interference of packets arriving simultaneously at a node. The many-to-many (m2m) communication primitive involves p participant nodes from among n nodes in the network, where the distance between any pair of participants is at most d. The task is to have all the participants get to know all the input messages. We consider three cases of the m2m communication problem. In the ad-hoc case, each participant knows only its name and the values of n, p and d. In the partially centralized case, each participant knows the topology of the network and the values of p and d, but does not know the names of the other participants. In the centralized case, each participant knows the topology of the network and the names of all the participants. For the centralized m2m problem, we give deterministic protocols, for both undirected and directed networks, working in time, which is provably optimal. For the partially centralized m2m problem, we give a randomized protocol for undirected networks working in time with high probability (whp), and we show that any deterministic protocol requires time. For the ad-hoc m2m problem, we develop a randomized protocol for undirected networks that works in time whp. We show two lower bounds for the ad-hoc m2m problem. One lower bound states that any randomized protocol for the m2m ad hoc problem requires expected time. Another lower bound states that for any deterministic protocol for the m2m ad hoc problem, there is a network on which the protocol requires time when n−p(n)=Ω(n) and d>1, and that it requires Ω(n) time when n−p(n)=o(n). The results of this paper appeared in a preliminary form in “On many-to-many communication in packet radio networks” in Proceedings of the 10th Conference on Principles of Distributed Systems (OPODIS), Bordeaux, France, 2006, Lecture Notes in Computer Science 4305, Springer, Heidelberg, pp. 258–272. The work of B.S. Chlebus was supported by NSF Grant 0310503.     

Faster Deterministic Communication in Radio Networks

We study the communication primitives of broadcasting (one-to-all communication) and gossiping (all-to-all communication) in known topology radio networks, i.e., where for each primitive the schedule of transmissions is precomputed based on full knowledge about the size and the topology of the network. We show that gossiping can be completed in time units in any radio network of size n, diameter D, and maximum degree Δ=Ω(log n). This is an almost optimal schedule in the sense that there exists a radio network topology, specifically a Δ-regular tree, in which the radio gossiping cannot be completed in less than units of time. Moreover, we show a schedule for the broadcast task. Both our transmission schemes significantly improve upon the currently best known schedules by Gąsieniec, Peleg, and Xin (Proceedings of the 24th Annual ACM SIGACT-SIGOPS PODC, pp. 129–137, 2005), i.e., a O(D+Δlog n) time schedule for gossiping and a D+O(log ³ n) time schedule for broadcast. Our broadcasting schedule also improves, for large D, a very recent O(D+log ² n) time broadcasting schedule by Kowalski and Pelc. A preliminary version of this paper appeared in the proceedings of ISAAC’06. F. Cicalese supported by the Sofja Kovalevskaja Award 2004 of the Alexander von Humboldt Stiftung. F. Manne and Q. Xin supported by the Research Council of Norway through the SPECTRUM project.     

On the Competitive Ratio for Online Facility Location

We consider the problem of Online Facility Location, where the demand points arrive online and must be assigned irrevocably to an open facility upon arrival. The objective is to minimize the sum of facility and assignment costs. We prove that the competitive ratio for Online Facility Location is Θ . On the negative side, we show that no randomized algorithm can achieve a competitive ratio better than Ω against an oblivious adversary even if the demands lie on a line segment. On the positive side, we present a deterministic algorithm which achieves a competitive ratio of in every metric space. A preliminary version of this work appeared in the Proceedings of the 30th International Colloquium on Automata, Languages and Programming (ICALP 2003), Lecture Notes in Computer Science 2719. This work was done while the author was at the Max-Planck-Institut für Informatik, Saarbrücken, Germany, and was partially supported by the Future and Emerging Technologies programme of the EU under contract number IST-1999-14186 (ALCOM–FT).     

Competitive Analysis of Scheduling Algorithms for Aggregated Links

We study an online job scheduling problem arising in networks with aggregated links. The goal is to schedule n jobs, divided into k disjoint chains, on m identical machines, without preemption, so that the jobs within each chain complete in the order of release times and the maximum flow time is minimized. We present a deterministic online algorithm with competitive ratio , and show a matching lower bound, even for randomized algorithms. The performance bound for we derive in the paper is, in fact, more subtle than a standard competitive ratio bound, and it shows that in overload conditions (when many jobs are released in a short amount of time), ’s performance is close to the optimum. We also show how to compute an offline solution efficiently for k=1, and that minimizing the maximum flow time for k,m≥2 is -hard. As by-products of our method, we obtain two offline polynomial-time algorithms for minimizing makespan: an optimal algorithm for k=1, and a 2-approximation algorithm for any k. W. Jawor and M. Chrobak supported by NSF grants OISE-0340752 and CCR-0208856. Work of C. Dürr conducted while being affiliated with the Laboratoire de Recherche en Informatique, Université Paris-Sud, 91405 Orsay. Supported by the CNRS/NSF grant 17171 and ANR Alpage.     

Algorithms to Separate $\{0,\frac{1}{2}\}$ -Chvátal-Gomory Cuts

Chvátal-Gomory cuts are among the most well-known classes of cutting planes for general integer linear programs (ILPs). In case the constraint multipliers are either 0 or , such cuts are known as -cuts. It has been proven by Caprara and Fischetti (Math. Program. 74:221–235, 1996) that separation of -cuts is -hard. In this paper, we study ways to separate -cuts effectively in practice. We propose a range of preprocessing rules to reduce the size of the separation problem. The core of the preprocessing builds a Gaussian elimination-like procedure. To separate the most violated -cut, we formulate the (reduced) problem as integer linear program. Some simple heuristic separation routines complete the algorithmic framework. Computational experiments on benchmark instances show that the combination of preprocessing with exact and/or heuristic separation is a very vital idea to generate strong generic cutting planes for integer linear programs and to reduce the overall computation times of state-of-the-art ILP-solvers.     

Optimally Adaptive Integration of Univariate Lipschitz Functions

We consider the problem of approximately integrating a Lipschitz function f (with a known Lipschitz constant) over an interval. The goal is to achieve an additive error of at most ε using as few samples of f as possible. We use the adaptive framework: on all problem instances an adaptive algorithm should perform almost as well as the best possible algorithm tuned for the particular problem instance. We distinguish between and , the performances of the best possible deterministic and randomized algorithms, respectively. We give a deterministic algorithm that uses samples and show that an asymptotically better algorithm is impossible. However, any deterministic algorithm requires samples on some problem instance. By combining a deterministic adaptive algorithm and Monte Carlo sampling with variance reduction, we give an algorithm that uses at most samples. We also show that any algorithm requires samples in expectation on some problem instance (f,ε), which proves that our algorithm is optimal.     

On Enumerating Minimal Dicuts and Strongly Connected Subgraphs

We consider the problems of enumerating all minimal strongly connected subgraphs and all minimal dicuts of a given strongly connected directed graph G=(V,E). We show that the first of these problems can be solved in incremental polynomial time, while the second problem is NP-hard: given a collection of minimal dicuts for G, it is NP-hard to tell whether it can be extended. The latter result implies, in particular, that for a given set of points , it is NP-hard to generate all maximal subsets of contained in a closed half-space through the origin. We also discuss the enumeration of all minimal subsets of whose convex hull contains the origin as an interior point, and show that this problem includes as a special case the well-known hypergraph transversal problem. This research was supported by the National Science Foundation (Grant IIS-0118635). The third and fourth authors are also grateful for the partial support by DIMACS, the National Science Foundation’s Center for Discrete Mathematics and Theoretical Computer Science. Our friend and co-author, Leonid Khachiyan tragically passed away on April 29, 2005.     

Optimality in External Memory Hashing

Hash tables on external memory are commonly used for indexing in database management systems. In this paper we present an algorithm that, in an asymptotic sense, achieves the best possible I/O and space complexities. Let B denote the number of records that fit in a block, and let N denote the total number of records. Our hash table uses I/Os, expected, for looking up a record (no matter if it is present or not). To insert, delete or change a record that has just been looked up requires I/Os, amortized expected, including I/Os for reorganizing the hash table when the size of the database changes. The expected external space usage is times the optimum of N/B blocks, and just O(1) blocks of internal memory are needed.     

Finding Pathway Structures in Protein Interaction Networks

The increased availability of data describing biological interactions provides important clues on how complex chains of genes and proteins interact with each other. Most previous approaches either restrict their attention to analyzing simple substructures such as paths or trees in these graphs, or use heuristics that do not provide performance guarantees when general substructures are analyzed. We investigate a formulation to model pathway structures directly and give a probabilistic algorithm to find an optimal path structure in time and space, where n and m are respectively the number of vertices and the number of edges in the given network, k is the number of vertices in the path structure, and t is the maximum number of vertices (i.e., "width") at each level of the structure. Even for the case t = 1 which corresponds to finding simple paths of length k, our time complexity is a significant improvement over previous probabilistic approaches. To allow for the analysis of multiple pathway structures, we further consider a variant of the algorithm that provides probabilistic guarantees for the top suboptimal path structures with a slight increase in time and space. We show that our algorithm can identify pathway structures with high sensitivity by applying it to protein interaction networks in the DIP database.     

Communication-Aware Processor Allocation for Supercomputers: Finding Point Sets of Small Average Distance

We give processor-allocation algorithms for grid architectures, where the objective is to select processors from a set of available processors to minimize the average number of communication hops. The associated clustering problem is as follows: Given n points in ℜ ^d , find a size-k subset with minimum average pairwise L ₁ distance. We present a natural approximation algorithm and show that it is a -approximation for two-dimensional grids; in d dimensions, the approximation guarantee is , which is tight. We also give a polynomial-time approximation scheme (PTAS) for constant dimension d, and we report on experimental results.     

Quantum Algorithms for Matching Problems

We present quantum algorithms for the following matching problems in unweighted and weighted graphs with n vertices and m edges:

A new PAC bound for intersection-closed concept classes

For hyper-rectangles in $\mathbb{R}^{d}$ Auer (1997) proved a PAC bound of $O(\frac{1}{\varepsilon}(d+\log \frac{1}{\delta}))$ , where $\varepsilon$ and $\delta$ are the accuracy and confidence parameters. It is still an open question whether one can obtain the same bound for intersection-closed concept classes of VC-dimension $d$ in general. We present a step towards a solution of this problem showing on one hand a new PAC bound of $O(\frac{1}{\varepsilon}(d\log d + \log \frac{1}{\delta}))$ for arbitrary intersection-closed concept classes, complementing the well-known bounds $O(\frac{1}{\varepsilon}(\log \frac{1}{\delta}+d\log \frac{1}{\varepsilon}))$ and $O(\frac{d}{\varepsilon}\log \frac{1}{\delta})$ of Blumer et al. and (1989) and Haussler, Littlestone and Warmuth (1994). Our bound is established using the closure algorithm, that generates as its hypothesis the intersection of all concepts that are consistent with the positive training examples. On the other hand, we show that many intersection-closed concept classes including e.g. maximum intersection-closed classes satisfy an additional combinatorial property that allows a proof of the optimal bound of $O(\frac{1}{\varepsilon}(d+\log \frac{1}{\delta}))$ . For such improved bounds the choice of the learning algorithm is crucial, as there are consistent learning algorithms that need $\Omega(\frac{1}{\varepsilon}(d\log\frac{1}{\varepsilon} +\log\frac{1}{\delta}))$ examples to learn some particular maximum intersection-closed concept classes.     

A schedulable utilization bound for the multiprocessor \mathsf{EPDF} Pfair algorithm

The earliest-pseudo-deadline-first ( ) Pfair scheduling algorithm is less expensive than some other known Pfair algorithms, but is not optimal for scheduling recurrent real-time tasks on more than two processors. In prior work, sufficient per-task weight (i.e., utilization) restrictions were established for ensuring that tasks either do not miss their deadlines or have bounded tardiness when scheduled under . Implicit in these restrictions is the assumption that the total system utilization may equal the total available processing capacity (i.e., the total number of processors). This paper considers an orthogonal issue, namely, determining a sufficient restriction on the total utilization of a task set for it to be schedulable (i.e., a schedulable utilization bound) under , assuming that there are no per-task weight restrictions. We prove that a task set with total utilization at most is correctly scheduled under on M processors, regardless of how large each task’s weight is. At present, we do not know whether this value represents the worst-case for , but we do provide a counterexample that shows that it cannot be improved to exceed 86% of the total processing capacity. The schedulable utilization bound we derive is expressed in terms of the maximum weight of any task, and hence, if this value is known, may be used to schedule task sets with total utilization greater than .

Efficient Prüfer-Like Coding and Counting Labelled Hypertrees

We consider labelled r-uniform hypertrees, 2≤r≤n, where n is the number of vertices in the hypertree. Any two hyperedges in a hypertree share at most one vertex and each hyperedge in an r-uniform hypertree contains exactly r vertices. We show that r-uniform hypertrees can be encoded in linear time using as little as n−2 integers in the range [1,n]. The decoding algorithm also runs in linear time. For general hypertrees, we require codes of length n+p−2 where p is the number of vertices belonging to more than one hyperedge in the given hypertree. Based on our coding technique, we show that there are at most distinct labelled r-uniform hypertrees, where f(n,r) is a lower bound on the number of labelled trees with maximum (vertex) degree exceeding . We suggest a counting scheme for determining such a lower bound f(n,r).     

An $\mathsf{EDF}$ -based restricted-migration scheduling algorithm for multiprocessor soft real-time systems

There has been much recent interest in the use of the earliest-deadline-first ( ) algorithm for scheduling soft real-time sporadic task systems on identical multiprocessors. In hard real-time systems, a significant disparity exists between -based schemes and Pfair scheduling: on M processors, the worst-case schedulable utilization for all known variants is approximately M/2, whereas it is M for optimal Pfair algorithms. This is unfortunate because -based algorithms entail lower scheduling and task-migration overheads. However, such a disparity in schedulability can be alleviated by easing the requirement that all deadlines be met, which may be sufficient for soft real-time systems. In particular, in recent work, we have shown that if task migrations are not restricted, then (i.e. , global ) can ensure bounded tardiness for a sporadic task system with no restrictions on total utilization. Unrestricted task migrations in global may be unappealing for some systems, but if migrations are forbidden entirely, then bounded tardiness cannot be guaranteed. In this paper, we address the issue of striking a balance between task migrations and system utilization by proposing an algorithm called , which is based upon and treads a middle path, by restricting, but not eliminating, task migrations. Specifically, under , the ability to migrate is required for at most M−1 tasks, and it is sufficient that every such task migrate between two processors and at job boundaries only. , like global , can ensure bounded tardiness to a sporadic task system as long as the available processing capacity is not exceeded, but, unlike global , may require that per-task utilizations be capped. The required cap is quite liberal, hence, should enable a wide range of soft real-time applications to be scheduled with no constraints on total utilization.