In this paper, we present a new strongly polynomial time algorithm for the minimum cost flow problem, based on a refinement of the Edmonds-Karp scaling technique. Our algorithm solves the uncapacitated minimum cost flow problem as a sequence of O(n log n) shortest path problems on networks with n nodes and m arcs and runs in O(n log n (m + n log n)) time. Using a standard transformation, thjis approach yields an O(m log n (m + n log n)) algorithm for the capacitated minimum cost flow problem. This algorithm improves the best previous strongly polynomial time algorithm, due to Z. Galil and E. Tardos, by a factor of n 2 /m. Our algorithm for the capacitated minimum cost flow problem is even more efficient if the number of arcs with finite upper bounds, say n', is much less than m. In this case, the running time of the algorithm is O((m ' + n)log n(m + n log n)).

) Abstract In this paper we study the Local Memory P-RAM. This model allows unit cost communication but assumes that the shared memory is divided into modules. This model is motivated by a consideration of potential optical computers. We show that fundamental problems such as list-ranking and parallel tree contraction can be implemented on this model in O(log n) time using n= log n processors. To solve the list-ranking problem we introduce a general asynchronous technique which has relevance to a number of problems. 1 Introduction We consider a model of parallel computation that is especially suited to pointer based computation. We motivate this model by showing that basic problems, like list-ranking and parallel tree contraction, can be performed in O(log n) time using only n= log n processors. We also show that any step on this model can be simulated in unit time on this model by a machine with an optical communication architecture. Thus we contend that the basic problem of list-ra...

For an arbitrary fixed surface S, a linear time algorithm is presented that for a given graph G either finds an embedding of G in S or identifies a subgraph of G that is homeomorphic to a minimal forbidden subgraph for embeddability in S. A side result of the proof of the algorithm is that minimal forbidden subgraphs for embeddability in S cannot be arbitrarily large. This yields a constructive proof of the result of Robertson and Seymour that for each closed surface there are only finitely many minimal forbidden subgraphs. The results and methods of this paper can be used to solve more general embedding extension problems.

This paper considers the ultimate impact of fundamental physical limitations---notably, speed of light and device size---on parallel computing machines. Although we fully expect an innovative and very gradual evolution to the limiting situation, we take here the provocative view of exploring the consequences of the accomplished attainment of the physical bounds. The main result is that scalability holds only for neighborly interconnections, such as the square mesh, of bounded-size synchronous modules, presumably of the area-universal type. We also discuss the ultimate infeasibility of latencyhiding, the violation of intuitive maximal speedups, and the emerging novel processor-time tradeoffs.

PRAMs also approximate the situation where communication to and from shared memory is much more expensive than local operations, for example, where each processor is located on a separate chip and access to shared memory is through a combining network. Not surprisingly, abstract PRAMs can be much more powerful than restricted instruction set PRAMs. THEOREM 21.16 Any function of n variables can be computed by an abstract EROW PRAM in O(log n) steps using n= log 2 n processors and n=2 log 2 n shared memory cells. PROOF Each processor begins by reading log 2 n input values and combining them into one large value. The information known by processors are combined in a binary-tree-like fashion. In each round, the remaining processors are grouped into pairs. In each pair, one processor communicates the information it knows about the input to the other processor and then leaves the computation. After dlog 2 ne rounds, one processor knows all n input values. Then this processor computes th...

ix Chapter 1 Introduction 1 1.1 Algorithmic Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Technical Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.1 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2.2 Parallel Computational Complexity . . . . . . . . . . . . . . . . . . . . . 7 1.2.3 Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.2.4 Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1.2.5 Random Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.2.6 Group Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 1.3 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Chapter 2 Parallel Uniform Generation of Unlabelled Graphs 25 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2 Sampling O...

this paper, if T is time-constructible, then

We consider the problem of storing an n element subset S of a universe of size m, so that membership queries (is x 2 S?) can be answered efficiently. The model of computation is a random access machine with the standard instruction set (direct and indirect adressing, conditional branching, addition, subtraction, and multiplication). We show that if s memory registers are used to store S, where n s m=n , then query time \Omega\Gammame/ n) is necessary in the worst case. That is, under these conditions, the solution consisting of storing S as a sorted table and doing binary search is optimal. The condition s m=n is essentially optimal; we show that if n + m=n o(1) registers may be used, query time o(log n) is possible.

We extend Levin's definition of average polynomial time to arbitrary time-bounds in accordance with the following general principles: (1) It essentially agrees with Levin's notion when applied to polynomial time-bounds. (2) If a language L belongs to DTIME(T(n)), for some time-bound T(n), then every distributional problem (L;µ) is T on the µ-average. (3) If L does not belong to DTIME(T(n)) almost everywhere, then no distributional problem (L;µ) is T on the µ-average. We present hierarchy theorems for average-case complexity, for arbitrary timebounds, that are as tight as the well-known Hartmanis-Stearns [HS65] hierarchy theorem for deterministic complexity. As a consequence, for every time-bound T(n), there are distributional problems (L;µ) that can be solved using only a slight increase in time but that cannot be solved on the µ-average in time T(n). Keywords: computational complexity, average time complexity classes, hierarchy, Average-P, logarithmico-exponential ACM Computing R...

Upper bounds are derived for the processor-time tradeoffs of machines such as linear arrays and two-dimensional meshes, which are compatible with the physical limitation expressed by bounded-speed propagation of messages (due to the finiteness of the speed of light). It is shown that parallelism and locality combined may yield speedups superlinear in the number of processors. The speedups are inherent, due to the optimality of the obtained tradeoffs as established in a companion paper. Simulations are developed of multiprocessor machines by analogous machines with fewer processors. A crucial role is played by the hierarchical nature of the memory system. A divide-and-conquer technique for hierarchical memories is developed, based on the graph-theoretic notion of topological separator. For multiprocessors, this technique also requires a careful balance of memory access and interprocessor communication costs, which leads to non-intuitive orchestrations of the simulation process. Dipart...