H.E. Plumblee · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1990
The architecture and performance of massively parallel computers and their applicability to aerospace problems are discussed with particular reference to results obtained with a parallel machine using 16,384 processors. The discussion focuses on work conducted in the field of computational fluid dynamics and electromagnetics. The cost/performance of a massively parallel computer is evaluated.
Cited by 2
Robert E. Benner, John L. Gustafson, G.R. Montry · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 2022
A parallel computing system and method having improved performance where a program is concurrently run on a plurality of nodes for reducing total processing time, each node having a processor, a memory, and a predetermined number of communication channels connected to the node and independently connected directly to other nodes. The present invention improves performance of the parallel computing system by providing a system which can provide efficient communication between the processors and between the system and input and output devices. A method is also disclosed which can locate defective nodes with the computing system.
Cited by 0Open access
Eric M. Aldrich · 2013
Cited by 4
Steve Furber · Institution of Engineering and Technology eBooks · 2019
Half a century of progress in computer technology has delivered machines of formidable capability and an expectation that similar advances will continue into the foreseeable future. However, much of the past progress has been driven by developments in semiconductor technology following Moore's Law, and there are strong grounds for believing that these cannot continue at the same rate. This, and related issues, suggest that there are huge challenges ahead in meeting the expectations of future progress, such as understanding how to exploit massive parallelism and how to deliver improvements in energy efficiency and reliability in the face of diminishing component reliability. Alongside these issues, recent advances in machine learning have created a demand for machines with cognitive capabilities, for example, to control autonomous vehicles, that we will struggle to deliver. Biological systems have, through evolution, found solutions to many of these problems, but we lack a fundamental understanding of how these solutions function. If we could advance our understanding of biological systems, we would open a rich source of ideas for unblocking progress in our engineered systems. An overview is given of SpiNNaker - a spiking neural network architecture. The SpiNNaker machine puts these principles together in the form of a massively parallel computer architecture designed both to model the biological brain, in order to accelerate our understanding of its principles of operation, and also to explore engineering applications of such machines.
Cited by 0
Jorge L. C. Sanz · 2003
Highly parallel systems play a central role in the future of computing. This role is increasing rapidly as the costs involved in improving semiconductor circuit speed and density become higher. The two major hurdles that prevent massive parallelism from becoming mainstream are the relatively slow progress of interconnection networks compared to that of uniprocessor technology and the immature status of parallel software. The paper gives a brief summary of some of the recent advances in the field of highly parallel computing, with an emphasis on message routing architectures.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Cited by 0
Robert Birke, Germán Rodríguez, Cyriel Minkenberg · 2012
The power of high-performance computing (HPC) is applied to simulate highly complex systems and processes in many scientific communities, e.g. in particle physics, weather and climate research, bio-sciences, materials science, pharmaceutics, astronomy, or finance. Current HPC systems are so comple
Cited by 10
Sudip S Dosanjh · University of North Texas Digital Library (University of North Texas) · 1999
The computing power available to scientists and engineers has increased dramatically in the past decade, due in part to progress in making massively parallel computing practical and available. The expectation for these machines has been great. The reality is that progress has been slower than expected. Nevertheless, massively parallel computing is beginning to realize its potential for enabling significant break-throughs in science and engineering. This paper provides a perspective on the state of the field, colored by the authors' experiences using large scale parallel machines at Sandia National Laboratories. We address trends in hardware, system software and algorithms, and we also offer our view of the forces shaping the parallel computing industry.
Cited by 0Open access
Litt, John, Parkinson, Dennis · CERN Document Server (European Organization for Nuclear Research) · 1990
Cited by 0
DAVID E WOMBLE · University of North Texas Digital Library (University of North Texas) · 1999
Since the mid-1980's, there have been a number of commercially available parallel computers with hundreds or thousands of processors. These machines have provided a new capability to the scientific community, and they been used successfully by scientists and engineers although with varying degrees of success. One of the reasons for the limited success is the difficulty, or perceived difficulty, in developing code for these machines. In this paper we discuss many of the issues and challenges in developing scalable hardware, system software and algorithms for machines comprising hundreds or thousands of processors.
Cited by 0Open access
Peter Brezány · Lecture notes in computer science · 1997
Cited by 8
V.L. Varscavsky · 2002
From the standpoint of hardware experts, asynchronism is connected with the concept of physical time as an independent physical variable and is determined by the variations of transient process durations in hardware circuits, modules and blocks that are physical objects by their nature. Software and architecture experts treat asynchronism as a partial order on events that are logical objects, i.e. they think in terms of logical time. In these terms, asynchronism is the variation of the process step quantity without respect to the real duration of these seeps in physical time. The measuring tool for time is a clock and the precision of the clock (along with the system of signal delivery) we can attain determines the area of its application (the allowed value of physical time step). The basic idea of self-timing is detecting the moments when transient processes in physical components are over and producing the corresponding logical signals that provide the transition to logical time (delay-insensitive design) in spite of the delay variation reasons. As all the logical signals invariant to the physical time and representing the events in the system are formed, self-timed methodology has a number of efficient hardware support methods to coordinate the events of the corresponding concurrent specification.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Cited by 1
Ron Brightwell, Lee Ann Fisk, David S. Greenberg et al. · Parallel Computing · 2000
Cited by 85
Santavy, Martin · eScholarship@McGill (McGill) · 1987
This thesis studies mz.ssively parallel synchronous processing models and algorithms.We survey the basic models, discuss their interrelationships and analyze properties of some feasible network models.A new definition of Gentleman's 0'-function is given.We present routines that simulate the CUBE ASCEND /DESCEND class algorithms on the CCC (cube-connected cycles) and the PERFECT SHUFFLE machines of any sufficient size.We give the exact (non-asymptotic) computation times and prove the correctness of the algorithms.This extends the previous results of Stone (1971) and Preparata-Vuillemin (1979).Borodin-von zur Gathen-Hopcroft (1984) laid out a program to obtain a "theory package for parallel algebraic manipulation".We continue work in this program and focus on the GCD of two polynomials, which is one of the basic problems of algebraic manipulation algorithms.B-G-H gave a GCD algorithm that works over arbitrary fields in 0 ( log 2 n) time and requires polynomial number of processors on a general type of parallel machine, such as P-RAM or algebraic circuits.The algorithm uses a system of ( n+m-2i) X ( n+m-2i) asymmetric matrices.If the result of Berkowitz ( 1984) is applied, the required number of processors is 0 ( n a+ 2 +E ).We modify the algorithm and present the following results: A new matrix formula for polynomial GCD is given that uses a system of upper left principal minors of a symmetric n X n matrix.The Berkowitz (1984) parallelization of the Samuelson (1942) method is modified and combined with the previous result.This allows us to decrease the required number of processors by the factor of 0 ( n ).Morover, the full strength of the general parallel models is not needed for the algorithm.We present a CUBE-feasible algorithm (composed of ASCEND /DESCEND subroutines) that computes the GCD of two polynomials over arbitrary field in 0 ( log 2 n ) time and requires n a+l+E processors, a = 3.A Hoare-style axiomatic verification system for CUBE-feasible algorithms is developed and used to prove correctness of the presented routines.
Cited by 0Open access
Thien Huu Nguyen, Daanish Arya, Marcus W. Doherty et al. · arXiv (Cornell University) · 2022
Quantum computing has the potential to offer substantial computational advantages over conventional computing. Recent advances in quantum computing hardware and algorithms have enabled a class of classically parallel quantum workloads, whereby individual quantum circuits can execute independently on many quantum processing units. Here, we present the full-stack software framework developed at Quantum Brilliance to enable multi-modal parallelism for hybrid quantum workloads. Our software provides the capability to distribute quantum workloads across multiple quantum accelerators hosted by nodes of a locally-networked cluster, via the industry-standard MPI (Message Passing Interface) protocol, or to distribute workloads across a large number of cloud-hosted quantum accelerators.
Cited by 0Open access
Jack B. Dennis · 2005
One important tool for building software is facilitzes for encapsulating abstractions. The cui-wntly popular programming paradigms for parallel computing do not provide such an encapsulation capability if parallelism is to be exploited within program modules. The function application operator of functional programming provides an interface for encapsulating abstractions that satisfi nts for modular softwuie constructio advantages of functional programming languages for programming purallel coniputers and survexthe areas wh.ere work ,is needed so functional programming principles may be applzed 211. eficient and practical parallel computing.
Cited by 0
R. Rinn, Robert A. Wagner · 2002
The authors discuss usability requirements for massively parallel processing (MPP) systems and present the fault tolerance techniques used in Parsytec's GC in order to achieve the criteria of usability. The activity undertaken by the FTMPS (Fault Tolerant Massively Parallel System) ESPRIT Project Hash 6731 is primarily concerned with the fault tolerance issues used in the GC and therefore covered in the present work.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Cited by 3
V. Varshavsky, V. Marakhovsky, R. Lashevsky · 2002
The problems are discussed that arise when designing massively parallel computer systems. The transition from globally synchronized working of such systems to globally asynchronous behavior resolves most of them. This transition implies that all the local processes in the system should interact between each other on the base of asynchronous interfaces. The problems of asynchronous interaction of local processes with the system of their global coordination on the base of handshake are considered as well as the problems of self-timed data transmission between processes. If the system modules that realize local processes are not asynchronous and implemented in CMOS-technology, then, to detect the moments of the transient processes completion in them, the idea of current indication is used. A circuit of a current sensor is suggested with wide range of permissible changes of the measured current and with admissible characteristics. Two ways of organizing the interaction between circuits with current sensors are developed. The principles of self-timed data exchange between local processes of the system and data transmission by means of a dual-rail code and binary code with handshake for every bit are considered. The possibility of organizing single-wire bit handshake is demonstrated and its self-timed implementation is developed with the transmission rate no worse than that of double-wire bit handshake.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Cited by 6
Hongjun Lu, Xiaohui Huang, Zhixian Li · 1997
To better support decision making, it was proposed to extend SQL to include data cube operations. Computation of data cube requires computing a number of interrelated group-bys, which is rather expensive operation when databases are large. In this paper, we propose to couple a relational database management system with massively parallel processors (MPP) to facilitate on-line analytic processing. Extended SQL queries involving complex data analysis such as data cube computation are decomposed. Data retrieved from the database are pipelined to the MPP machine where data cubes are computed in parallel. The system architecture and issues related to parallel computation of data cubes are described. A brute force parallel data cube processing algorithm was implemented and the results of some preliminary experiments are presented. 1. Introduction Aggregation is widely used in on-line analytical processing. In relational database systems, SQL supports a set of aggregate functions (SUM, COUN...
Cited by 0
K. K. Mathur, Zdeněk Johan, S. Lennart Johnsson et al. · Digital Access to Scholarship at Harvard (DASH) (Harvard University) · 1993
Massively parallel computing holds the promise of extreme performance. Critical for achieving high performance is the ability to exploit locality of reference and effective management of the communication resources. This article describes two communication primitives and associated mapping strategies that have been used for several different unstructured, three-dimensional, finite element applications in computational fluid dynamics and structural mechanics.
Cited by 0Open access
Dennis Jack B. · Advances in parallel computing · 2012
A program execution model (PXM) is a precise specification of the features of a computer system that application programs use to meet goals for users of the system. Program execution models have been studied since the late 1960s.
Cited by 0