Allocating tree structured programs in a distributed system withuniform communication costs

Studies the complexity of the problem of allocating m modules to n processors in a distributed system to minimize total communication and execution costs. When the communication graph is a tree, Bokhari has shown that the optimum allocation can be determined in O(mn²) time. Recently, this result has been generalized by Fernandez-Baca, who has proposed an allocation algorithm in O(mn^k+1) when the communication graph is a partial k-tree. The author shows that in the case where communication costs are uniform, the module allocation problem can be solved in O(mn) time if the communication graph is a tree. This algorithm is asymptotically optimum     

Scheduling DAG's for asynchronous multiprocessor execution

A new approach is given for scheduling a sequential instruction stream for execution “in parallel” on asynchronous multiprocessors. The key idea in our approach is to exploit the fine grained parallelism present in the instruction stream. In this context, schedules are constructed by a careful balancing of execution and communication costs at the level of individual instructions, and their data dependencies. Three methods are used to evaluate our approach. First, several existing methods are extended to the fine grained situation. Our approach is then compared to these methods using both static schedule length analyses, and simulated executions of the scheduled code. In each instance, our method is found to provide significantly shorter schedules. Second, by varying parameters such as the speed of the instruction set, and the speed/parallelism in the interconnection structure, simulation techniques are used to examine the effects of various architectural considerations on the executions of the schedules. These results show that our approach provides significant speedups in a wide-range of situations. Third, schedules produced by our approach are executed on a two-processor Data General shared memory multiprocessor system. These experiments show that there is a strong correlation between our simulation results, and these actual executions, and thereby serve to validate the simulation studies. Together, our results establish that fine grained parallelism can be exploited in a substantial manner when scheduling a sequential instruction stream for execution “in parallel” on asynchronous multiprocessors     

Pipelining and bypassing in a VLIW processor

This short note describes issues involved in the bypassing mechanism for a very long instruction word (VLIW) processor and its relation to the pipeline structure of the processor. The authors first describe the pipeline structure of their processor and analyze its performance and compare it to typical RISC-style pipeline structures given the context of a processor with multiple functional units. Next they study the performance effects of various bypassing schemes in terms of their effectiveness in resolving pipeline data hazards and their effect on the processor cycle time     

On the performance of synchronized programs in distributed networkswith random processing times and transmission delays

A synchronizer is a compiler that transforms a program designed to run in a synchronous network into a program that runs in an asynchronous network. The behavior of a simple synchronizer, which also represents a basic mechanism for distributed computing and for the analysis of marked graphs, was studied by S. Even and S. Rajsbaum (1990) under the assumption that message transmission delays and processing times are constant. We study the behavior of the simple synchronizer when processing times and transmission delays are random. The main performance measure is the rate of a network, i.e., the average number of computational steps executed by a processor in the network per unit time. We analyze the effect of the topology and the probability distributions of the random variables on the behavior of the network. For random variables with exponential distribution, we provide tight (i.e., attainable) bounds and study the effect of a bottleneck processor on the rate     

E-kernel: an embedding kernel on the IBM victor V256 multiprocessorfor program mapping and network reconfiguration

We present the design of E-kernel, an embedding kernel on the Victor V256 message-passing partitionable multiprocessor, developed for the support of program mapping and network reconfiguration. E-kernel supports the embedding of a new network topology onto Victor's 2D mesh and also the embedding of a task graph onto the 2D mesh network or the reconfigured network. In the current implementation, the reconfigured network can be a line or an even-size ring, and the task graphs meshes or tori of a variety of dimensions and shapes or graphs with similar topologies. For application programs having these task graph topologies and that are designed according to the communication model of E-kernel, they can be run without any change on partitions connected by the 2D mesh, line, or ring. Further, E-kernel attempts the communication optimization of these programs on the different networks automatically, thus making both the network topology and the communication optimization attempt completely transparent to the application programs. Many of the embeddings used in E-kernel are optimal or asymptotically optimal (with respect to minimum dilation cost). The implementation of E-kernel translated some of the many theoretical results in graph embeddings into practical tools for program mapping and network reconfiguration in a parallel system. E-kernel is functional on Victor V256. Measurements of E-kernel's performance on V256 are also included     

On the design of dynamic associative neural memories

We consider the design problem for a class of discrete-time and continuous-time neural networks. We obtain a characterization of all connection weights that store a given set of vectors into the network, that is, each given vector becomes an equilibrium point of the network. We also give sufficient conditions that guarantee the asymptotic stability of these equilibrium points.     

Using recurrent neural networks for adaptive communication channelequalization

Nonlinear adaptive filters based on a variety of neural network models have been used successfully for system identification and noise-cancellation in a wide class of applications. An important problem in data communications is that of channel equalization, i.e., the removal of interferences introduced by linear or nonlinear message corrupting mechanisms, so that the originally transmitted symbols can be recovered correctly at the receiver. In this paper we introduce an adaptive recurrent neural network (RNN) based equalizer whose small size and high performance makes it suitable for high-speed channel equalization. We propose RNN based structures for both trained adaptation and blind equalization, and we evaluate their performance via extensive simulations for a variety of signal modulations and communication channel models. It is shown that the RNN equalizers have comparable performance with traditional linear filter based equalizers when the channel interferences are relatively mild, and that they outperform them by several orders of magnitude when either the channel's transfer function has spectral nulls or severe nonlinear distortion is present. In addition, the small-size RNN equalizers, being essentially generalized IIR filters, are shown to outperform multilayer perceptron equalizers of larger computational complexity in linear and nonlinear channel equalization cases.     

Genetic evolution of the topology and weight distribution of neuralnetworks

This paper proposes a system based on a parallel genetic algorithm with enhanced encoding and operational abilities. The system, used to evolve feedforward artificial neural networks, has been applied to two widely different problem areas: Boolean function learning and robot control. It is shown that the good results obtained in both cases are due to two factors: first, the enhanced exploration abilities provided by the search-space reducing evolution of both coding granularity and network topology, and, second, the enhanced exploitational abilities due to a recently proposed cooperative local optimizing genetic operator.     

Second-order bounds on the domain of attraction and the rate ofconvergence of nonlinear dynamical systems and neural networks

We present a method for analyzing the convergence properties of nonlinear dynamical systems yielding second-order bounds on the domain of attraction of an asymptotically stable equilibrium point and on the time of convergence in the estimated domain. We show that under certain conditions on the system, there exists an analytic solution to the corresponding optimization problem. The method is applied in analyzing the dynamics of a neural network model.     

Connectivity and performance tradeoffs in the cascade correlationlearning architecture

The cascade correlation is a very flexible, efficient and fast algorithm for supervised learning. It incrementally builds the network by adding hidden units one at a time, until the desired input/output mapping is achieved. It connects all the previously installed units to the new unit being added. Consequently, each new unit in effect adds a new layer and the fan-in of the hidden and output units keeps on increasing as more units get added. The resulting structure could be hard to implement in VLSI, because the connections are irregular and the fan-in is unbounded. Moreover, the depth or the propagation delay through the resulting network is directly proportional to the number of units and can be excessive. We have modified the algorithm to generate networks with restricted fan-in and small depth (propagation delay) by controlling the connectivity. Our results reveal that there is a tradeoff between connectivity and other performance attributes like depth, total number of independent parameters, and learning time.