Collection Of Processors Research Articles

A Dataflow Processor as the Basis of a Tiled Polymorphic Computing Architecture with Fine-Grain Instruction Migration

A dataflow processor is presented that is to be used as the basis of a tiled polymorphic computing architecture. The key contribution is a new scheme that allows a program's instructions to be migrated before and during runtime in a fine-grained manner across the collection of processors. The primary reason to perform this migration is to execute programs faster. Other reasons to perform instruction migration are to prioritize computational resources and to achieve thermal balancing. The act of performing instruction migration across of a collection of processors is logically equivalent to rearranging the computer architecture under the program (i.e., polymorphic computing). Additionally, a new dataflow instruction set which enables the migration is presented. This instruction set is built upon the concept of a single RISC-like dataflow instruction that can atomically execute, make decisions, and independently route results. Furthermore, the novel concept of an operation cell is presented. An operation cell holds a single instruction and its data. It also contains logic to independently determine when an instruction is executed and when to forward data to other operation cells in the collection of processors. In addition, the internal architecture of this processor is presented. This includes the arithmetic logic unit (ALU) that is used to execute the instructions and a series of buses that allow data movement to occur in parallel to instruction execution. Finally, an inter-processor migration strategy is defined.

A Comparative Study of Technologies Developed in Perspective of Distributed Operating Systems

The collection of processors that do not share memory and clock is called distributed operating system. It is an infrastructure which is programmed so that it allows the use of multiple workstations as a single integrated system. In distributed OS users can access remote resources as they access local ones. The advantages of distributed operating systems are performance through parallelism, reliability and availability through replication and in addition scalability, expansion and flexibility of resources. This research paper includes a detailed and comparative introduction of different state-of-art technologies and operating systems, their internal structure, design and key features which are developed in perspective of distributed operating systems. These operating systems includes Chorus, V-systems, Amoeba and Mach.

Simple Neural-Like P Systems for Maximal Independent Set Selection

Membrane systems (P systems) are distributed computing models inspired by living cells where a collection of processors jointly achieves a computing task. The problem of maximal independent set (MIS) selection in a graph is to choose a set of nonadjacent nodes to which no further nodes can be added. In this letter, we design a class of simple neural-like P systems to solve the MIS selection problem efficiently in a distributed way. This new class of systems possesses two features that are attractive for both distributed computing and membrane computing: first, the individual processors do not need any information about the overall size of the graph; second, they communicate using only one-bit messages.

Secure Hypergraphs: Privacy from Partial Broadcast

A broadcast channel enables one processor to send the same message---simultaneously and privately---to a fixed subset of processors. Suppose that a collection of processors are connected by an arbitrary network of partial broadcast channels (a hypergraph). We initiate the study of necessary and sufficient conditions, complexity bounds, and protocols for individual processors to exchange private messages across this network. Private message exchange, in turn, enables the realization of general secure computation primitives. The model (motivated by various environments such as multicast network architectures and group communication in distributed systems) is an intermediate setting between the private channels model and the full information model, both of which have been investigated extensively in the last few years. We assume a computationally unlimited adversary (i.e., the information theoretic notion of security), and our techniques are combinatorial. Both the possibility and the polynomial-time feasibility of private message exchange are investigated.

Generalized multipartitioning of multi-dimensional arrays for parallelizing line-sweep computations

Multipartitioning is a strategy for decomposing multi-dimensional arrays into tiles and mapping the resulting tiles onto a collection of processors. This class of partitionings enables efficient parallelization of “line-sweep” computations that solve one-dimensional recurrences along each dimension of a multi-dimensional array. Multipartitionings yield balanced parallelism for line sweeps by assigning each processor the same number of data tiles to compute at each step of a sweep along any array dimension. Also, they induce only coarse-grain communication. This paper considers the problem of computing generalized multipartitionings, which decompose d-dimensional arrays, d⩾2, onto an arbitrary number of processors. We describe an algorithm that computes an optimal multipartitioning onto all of the processors for this general case. We use a cost model to select the dimensionality of the best partitioning and the number of cuts to make along each array dimension; then, we show how to construct a mapping that assigns the resulting data tiles to each of the processors. The assignment of tiles to processors induced by this class of multipartitionings corresponds to an instance of a latin hyper-rectangle, a natural extension of latin squares, which have been widely studied in mathematics and statistics. Finally, we describe how we extended the Rice dHPF compiler for High Performance Fortran to generate code that employs our strategy for generalized multipartitioning and show that the compiler's generated code for the NAS SP computational fluid dynamics benchmark achieves scalable high performance.

Using destination-set prediction to improve the latency/bandwidth tradeoff in shared-memory multiprocessors

Destination-set prediction can improve the latency/bandwidth tradeoff in shared-memory multiprocessors. The destination set is the collection of processors that receive a particular coherence request. Snooping protocols send requests to the maximal destination set (i.e., all processors), reducing latency for cache-to-cache misses at the expense of increased traffic. Directory protocols send requests to the minimal destination set, reducing bandwidth at the expense of an indirection through the directory for cache-to-cache misses. Recently proposed hybrid protocols trade-off latency and bandwidth by directly sending requests to a predicted destination set.This paper explores the destination-set predictor design space, focusing on a collection of important commercial workloads. First, we analyze the sharing behavior of these workloads. Second, we propose predictors that exploit the observed sharing behavior to target different points in the latency/bandwidth tradeoff. Third, we illustrate the effectiveness of destination-set predictors in the context of a multicast snooping protocol. For example, one of our predictors obtains almost 90% of the performance of snooping while using only 15% more bandwidth than a directory protocol (and less than half the bandwidth of snooping).

On the Average Running Time of Odd–Even Merge Sort

This paper gives an upper bound for the average running time of Batcher's odd–even merge sort when implemented on a collection of processors. We consider the case wheren, the size of the input, is an arbitrary multiple of the numberpof processors used. We show that Batcher's odd–even merge (for two sorted lists of lengthneach) can be implemented to run in timeO((n/p)(log(2+p2/n))) on the average,11In this paper,logxalways meanslog2x.and that odd–even merge sort can be implemented to run in timeO((n/p)(logn+logplog(2+p2/n))) on the average. In the case of merging (sorting), the average is taken over all possible outcomes of the merge (all possible permutations ofnelements). That means that odd–even merge and odd–even merge sort have an optimal average running time ifn≥p2. The constants involved are also quite small.

Scheduling for reactive real-time systems

In designing Chinook, a hardware-software cosynthesis system for reactive real-time controllers, the impact of timing constraints on software scheduling has been a central concern. By dividing constraints into two levels, corresponding to low-level interactions with device interfaces and high-level real-time response and rate requirements, we have developed solutions tailored to each aspect. These scheduling techniques enable Chinook to map a high-level specification onto a specified collection of processors and peripheral devices while respecting performance requirements.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

The design of an operating system for a scalable parallel computing engine

There are substantial benefits to be gained from building computing systems from a number of processors working in parallel. One of the frequently-stated advantages of parallel and distributed systems is that they may be scaled to the needs of the user. This paper discusses some of the problems associated with designing a general-purpose operating system for a scalable parallel computing engine and then describes the solutions adopted in our experimental parallel operating system. We explain why a parallel computing engine composed of a collection of processors communicating through point-to-point links provides a suitable vehicle in which to realize the advantages of scaling. We then introduce a parallel-processing abstraction which can be used as the basis of an operating system for such a computing engine. We consider how this abstraction can be implemented and retain the ability to scale. As a concrete example of the ideas presented here we describe our own experimental scalable parallel operating-system project, concentrating on the Wisdom nucleus and the Sage file system. Finally, after introducing related work, we describe some of the lessons learnt from our own project.

Strong Consistency of a Parallel Stochastic Approximation Algorithm with Non-additive Noise

In this paper, a parallel stochastic approximation algorithm is considered. The esseace of this algorithm is to bring out the potential of parallel processing and asynchronous communication into full play. Instead of using a single serial processor, a collection of processors is used in such a way that the processors operate in parallel, compute and communicate asynchronously and at random times. The scope of this investigation is to extend the strong convergence result to the non-additive noise case. Under rather broad conditions, strong consistency is obtained via the method of randomly varying truncations.

Reconfigurable cellular computers

When a collection of processors P 1 ,,P n< operates in parallel, it is desirable that at any given stage of the computation, each {ifP i should have a task of about the same size to perform, and each {ifP i should require about the same amount of information from the other P 's in order to perform its task. To the extent that these conditions are violated, parallelism is impaired, in the sense that the P 's are not all used with equal efficiency. In cellular computers, e.g., as they might be used for parallel image processing, these conditions are maintained by having the P 's all perform similar computations on different parts of the input data, and by allowing each P i to receive information from a fixed set of the others (its “neighbors≓), where these sets are all of bounded size. This paper discusses, on an abstract level, the concept of a reconfigurable cellular computer, in which each P i can receive information from a set S i of the other P 's, and the S i 's areall of bounded size, but they need not remain fixed throughout a computation. Requiring the S i 's to have bounded size implies that most P 's cannot communicate directly; the expected time required for two arbitrary P 's to communicate depends on the graph structure defined by the sets S i . The question of how to change the S i 's in parallel during the course of a computation is also discussed.

Minimal parallelism for associative computations under time constraints

The parallel evaluation ofAN=a1∘a2∘...∘aN, where ∘ is binary associative, is studied. Under an idealized model of parallel computation, the minimal number of parallel processors required to computeAN in at mostt steps is determined for ⌈log2N⌉≤t≤N−1. This indicates that it is not always desirable to reduce the running time to an absolute minimum, and provides a lower bound on the processing power required for time-constrained evaluation of general arithmetic expressions. Results for two-input processors, are generalized tob-input processors, and then to non-homogeneous collections of processors. The latter does not have a closed-form solution, so approximations are analyzed.

Periodic Job Scheduling in a Distributed Processor System

This investigation considers the problem of the nonpreemptive scheduling of a set of independent periodic jobs in a computer system characterized by a collection of processors, memories, and interconnection mechanisms. Such a collection of resources can be used to perform the functions of a process control environment as are found, for example, in the avionics of a military or commercial aircraft. The distribution of these resources permits weight, volume, power, cost, and reliability requirements to be satisfied. Each periodic job is characterized by its frequency and its execution time. The inverse of the frequency, the period, is assumed to be an integer, and the initiation time and the completion time of job must be strictly met. An optimal algorithm is developed for a job set whose members are related by a binary frequency distribution. That is, if the jobs in the set are ordered in terms of decreasing job execution frequency, then fi = 2fi+1. The optimal result is generalized to include conditions in which the jobs are related by fi = kfi+1, where k is an integer.