Myunghwan Kim

Optimal task assignment in linear array networks

The problem of assigning tasks to the processors of a distributed computing system such that the sum of execution and communication costs is minimized is discussed. This problem is known to be NP-complete in the general case, and thus intractable for systems with a large number of processors. H.S. Stones (1977) network flow approach for a two-processor system is extended to the case for a linear array of any number of processors. The task assignment problem for a linear array network is first transformed into the two-terminal network flow problem, and then solved by applying the Goldberg-Tarjan (1987) network flow algorithm in time no worse than O(n/sup 2/m/sup 3/ log n), where n and m are the number of processors and the number of tasks, respectively. >

An efficient k-way graph partitioning algorithm for task allocation in parallel computing systems

The k-way graph partitioning problem can be transformed into the maximum k-cut problem using a proposed technique of graph modification. It is possible to transform the graph partitioning problem into the max-cut problem by incorporating node size information into the edge weight. After transformation, a very simple cost function can be devised which makes the proposed algorithm more efficient than the Kernighan-Lin (K-L) algorithm (1970). The computing time per iteration of the algorithm is O(k*N/sup 2/), where N is the number of nodes in the given graph. Experimental results show that the proposed algorithm outperforms the K-L algorithm both in the quality of solutions and in the elapsed time. Also, as the difference between the sizes of the nodes increases, the performance gap between the proposed algorithm and the K-L algorithm becomes larger.<<ETX>>

A heuristic algorithm for ordering the columns in one-dimensional logica arrays

The authors focus on the ordering of the columns to minimize the necessary number of tracks in one-dimensional logic array. They use a column-orientation approach to this problem. Each net is converted into a complete graph (or clique). The weighted graph using such topological transformation is unique. The necessary number of tracks can be evaluated by the cut of the two seeds of orderings, the left seed vertex and the right seed vertex. The authors find the minimal track assignments by minimizing the cut of the seed vertex. Additionally, a useful concept called the overriding property is introduced. It determines the local optimal assignments for two or more columns with the above property. The authors algorithm works well with either a uni- or bidirectional approach to select the seed of ordering. Results obtained by the proposed algorithm and the simulated annealing approach are compared. >

A distributed synchronization scheme for fair multi-process handshakes

Abstract We propose a new scheme for synchronizing multiple processes in common handshakes, which can ensure weak and strong (conspiracy-safe) fairness among handshakes. The scheme is based on a transformation that chooses one among the processes involved in each handshake as the master of the others. On the simplified domain a basic synchronization algorithm is constructed, and two layered fair-scheduling policies which can be embedded in the algorithm are introduced.

Stepwise-overlapped parallel annealing and its application to floorplan designs

Abstract A new parallel annealing scheme, stepwise-overlapped parallel annealing, is proposed. This can provide a massive speedup using a multiprocessor system with a large number of processors. It decomposes the annealing process by means of Markov chains. The annealing scheme of the systolic algorithm proposed by Aart et al. is improved. The improved annealing schedule keeps a good temperature profile even when a large number of processors are used. The communication pattern is also enhanced. The new parallel annealing algorithm is applied to floorplan design problems. Experimental results show that the new parallel annealing algorithm is highly effective for a large number of processors. The experimental results of the proposed algorithm are compared with those of other parallel annealing algorithms and the distributed genetic algorithm.

An improved hardware implementation of the fault-tolerant clock synchronization algorithm for large multiprocessor systems

An improved implementation of clock synchronization of multiprocessor systems in the presence of malicious faults is proposed. The proposed hardware implementation for the reference clock selection has a lower gate complexity, smaller time delay, and greater flexibility than the previously published implementation. The improvement is achieved by replacing the sorter with a counting encoder and comparators and by introducing threshold generation logic with programmable registers. The scheme has a gate complexity of O(n) and a delay of O(log n), where n is the total number of inputs to a particular clock, and is programmable for different values of n and m, the maximum number of faults. >

A distributed scheme for dynamic deadlock detection and resolution

Abstract A distributed scheme for dynamically detecting resource deadlocks is presented. The proposed scheme dynamically partitions the processes in the system into a set of groups and defines dependency relations between any two processes within a group (group information). This process grouping enables the detection of the deadlock without chasing all the edges of the cycle since the group information remains valid until it is updated again. Since the group information is updated during run-time, there is no separate detection phase in the proposed scheme and the deadlock is detected dynamically during run-time. The proposed scheme ensures that only one process in the cycle detects deadlock, which simplifies the process of deadlock resolution. The deadlock resolution rules that make the group information remain valid after the resolution of the deadlock are also presented. Performance of the proposed scheme is discussed through simulation.

A model of distributed control: dependency and uncertainty

Abstract A scheme for coordination of the cooperating processes with less communication overhead is described. We propose weak copy consistency constraints, which correlate the outdated state information to the current state. We also derive sufficient conditions, under which each process can guarantee the consistency of the global state without communication.

Hybrid join: an improved sort-based join algorithm

The algorithm consists of completely sorting only the smaller relation and partitioning the other one into ranged buckets according to the order statistics of the sorted relation. The final join is performed on the sorted relation and the ranged buckets

Shape decomposition by collinearity

Abstract The analysis of collinearity with semantic considerations in shape decomposition is presented. The method is based on a lot of heuristics satisfying human intuition about collinearity. Man-made objects which often contain many collinear lines are analyzed to verify this approach.