Parallel Processing Method Research Articles

5404551 Parallel processor which processes instructions, after a branch instruction, in parallel prior to executing the branch instruction and its parallel processing method

In a typical operating system, one-third of a program consists of branch instructions. This means a performance of a processor of a typical operating system depends greatly on whether or not an instruction before and after a branch instruction can be executed in parallel. In order to provide a high performance processor with parallel processing, provided is a structure with a plurality of operating units and a plurality of registers where a set of registers are specified with the same address. A selection sequence of registers is stored by a plurality of selection sequence storages. Contents of registers are determined or not depending on the information stored in a plurality of determination identification storages. A register is specified by a register selector according to the contents of the selection sequence storages. This register selector is also used to update the contents of the selection sequence storages. The contents of the determination identification storages are rewritten by a determination identifier when the contents of a register proves to be a correct result.

5404472 Parallel processing apparatus and method capable of switching parallel and successive processing modes

When executing successive processing of conventional software, a parallel processing apparatus turns a processing state discrimination flag off, increases a program count by 1 at a time, reads out one instruction, and processes that instruction in an arithmetic unit. When executing parallel processing for new software, the parallel processing apparatus turns the processing state discrimination on, increases the program count by m at a time, reads out m instructions, and exercises parallel processing over m instructions in m arithmetic units. In order to select either of the above described two kinds of processing, a discrimination changeover instruction having function of changing over the processing state discrimination flag is added. Instructions are processed in arithmetic unit(s) in accordance with the processing state discrimination flag. In this way, successive processing and parallel processing are provided with compatibility and are selectively executed. Further, the parallel processing apparatus making great account of compatibility of a great part of software reads out m instructions without using the processing state flag, decodes the m instructions, checks whether a branch instruction exists in the k-th instruction, then executes the first to the (k+1)-th instructions in k+1 arithmetic units, and prevent execution of the (k+ 2)-th to m-th instructions. By executing the k-th branch instruction, the parallel processing apparatus calculates an address nm+h of its branch destination, performs calculation to check whether the condition is satisfied or not, then prevents execution of instructions of addresses nm to nm+h-1, and executes instructions of addresses nm+h to (n+1)m. In this way, the parallel processing apparatus executes a plurality of instructions and successively executes branch instructions.

Modeling second-order nonlinear effects in optical waveguides using a parallel-processing beam propagation method

In this work, we present a simple efficient numerical solution for the three-dimensional coupled wave equations containing a second-order nonlinearity, using an explicit finite difference beam propagation method (EFD BPM). The linear EFD-BPM is known to be very efficient and to gain large speed up when implemented on parallel computers. The new nonlinear version of the EFD-BPM has the same features of the linear counterpart in using two separate computational windows, one for the fundamental field and the other for the second-harmonic field. We demonstrate the implementation and discuss the application of this method to a nonlinear rib waveguide using the quasi-phase-matching technique.

Parallel Processings for Direct Simulation Monte Carlo Method.

Parallel processing for the DSMC method (direct simulation Monte Carlo method) using a multiprocessor computer is discussed in this paper. Rarefied gas flows through a finite length circular tube, over a flat plate, and in aclosed cavity were simulated by the DSMC method. The performance of parallel processing with a dynamic domain decomposition technique is evaluated in terms of parallel processing efficiencies and acceleration. An estimation method of parallel processing for evaluation of the DSMC method is proposed.

Parallel Processing of Boiling Water Reactor Core Analysis.

A parallel processing method for the analysis of a Boiling Water Reactor (BWR) core has been developed to drastically reduce the computation time. In the proposed method, a BWR core is divided into smaller segments, each of which is assigned to one of the processing elements (PE) working in parallel. The whole computing task is divided into smaller tasks that are distributed to the PEs as equally as possible. To solve the neutron diffusion equations in BWR neutronics calculations, the three-dimensional checker-board block iterative method was adopted. In the thermal-hydraulic calculation, the whole task can be divided into parallel tasks except for the coolant enthalpy distribution calculation along a flow channel. Parallelization efficiency of the proposed method was examined by measuring computing time on a hypercube type parallel processor with 64PEs. The computation speed gradually degrades with the number of segmentation, because of delay due to communications between PEs and to waiting time caused b...

A study on parallel processing of electro-magnetic transient simulation

This paper presents a parallel processing method for power system electromagnetic transient simulation. It describes an outline of the algorithm, improvement in combined calculation between generators and networks, the relationship of granularity of divided subsystem and execution time and evaluation of precision by comparison with the results of EMTP. >

A domain decomposition parallel processing algorithm for molecular dynamics simulations of polymers

We describe in this paper a domain decomposition molecular dynamics algorithm for use on distributed memory parallel computers which is capable of handling systems containing rigid bond constraints and three- and four-body potentials as well as non-bonded potentials. The algorithm has been successfully implemented on the Fujitsu 1024 processor element AP1000 machine. The performance has been compared with and benchmarked against the alternative cloning method of parallel processing [D. Brown, J.H.R. Clarke, M. Okuda and T. Yamazaki, J. Chem. Phys., 100 (1994) 1684] and results obtained using other scalar and vector machines. Two parallel versions of the SHAKE algorithm, which solves the bond length constraints problem, have been compared with regard to optimising the performance of this procedure.

A parallel processing method in finite element analysis using domain division

Current parallel processing approaches in finite element analysis can be roughly classified into two categories: In the domain method, an analysis region is divided into subdomains and one CPU assigned to each subdomain. Alternatively, one may calculate in parallel the matrix and vector products which arise in the process of solving the set of simultaneous equations. In this paper, we present a hybrid of the above two methods. Iteration to bring values on the subdomain boundaries coincide is not necessary, but the idea of domain division is used.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Parallel processing of incremental ray tracing on a shared-memory multiprocessor

This paper presents a novel parallel-processing method for image synthesis using incremental ray tracing on a shared-memory multiprocessor workstation. The most efficient technique for image synthesis is ray tracing, proposed by Whitted in 1980. Ray-tracing algorithms are simple and can generate realistic images. However, they are time-consuming, since calculations of the intersections between objects and ray increase exponentially as the complexity of scenes increases. Fast image synthesis for animation is one of the most important topics in computer graphics. As the area of computer applications has broadened, the complexity of images to be synthesized has increased. Parallel processing of computer graphics is one way of achieving fast image synthesis. This paper describes a parallel processing technique for incremental ray tracing, which recalculates only the rays changed by moving objects in successive scenes of continuous image synthesis. The performance of parallel ray tracing was evaluated on the multiprocessor workstation TOP-1. Strategies for allocating pixels to processes under a multiprocess operating system on this workstation are discussed.

Review of quantum computation

Digital computers are machines that can be programmed to perform logical and arithmetical operations. Contemporary digital computers are ``universal,`` in the sense that a program that runs on one computer can, if properly compiled, run on any other computer that has access to enough memory space and time. Any one universal computer can simulate the operation of any other; and the set of tasks that any such machine can perform is common to all universal machines. Since Bennett`s discovery that computation can be carried out in a non-dissipative fashion, a number of Hamiltonian quantum-mechanical systems have been proposed whose time-evolutions over discrete intervals are equivalent to those of specific universal computers. The first quantum-mechanical treatment of computers was given by Benioff, who exhibited a Hamiltonian system with a basis whose members corresponded to the logical states of a Turing machine. In order to make the Hamiltonian local, in the sense that its structure depended only on the part of the computation being performed at that time, Benioff found it necessary to make the Hamiltonian time-dependent. Feynman discovered a way to make the computational Hamiltonian both local and time-independent by incorporating the direction of computation in the initial condition. In Feynman`smore » quantum computer, the program is a carefully prepared wave packet that propagates through different computational states. Deutsch presented a quantum computer that exploits the possibility of existing in a superposition of computational states to perform tasks that a classical computer cannot, such as generating purely random numbers, and carrying out superpositions of computations as a method of parallel processing. In this paper, we show that such computers, by virtue of their common function, possess a common form for their quantum dynamics.« less

A parallel processing method for implementing the RSA cryptosystem

AbstractModular multiplication: R = A · B mod N (where A, B, N, and R are integers) is an essential operation used in public‐key cryptosystems such as RSA, El Gamal, and Rabin. Thus, if high‐speed modular multiplication circuits can be constructed, these cryptosystems can operate at high speeds. In this paper, we show how to construct, with a high‐speed modular multiplication circuit which employs parallel processing methods using systolic arrays, a circuit which performs modular exponentiation by repeated modular multiplication. In particular, we demonstrate its validity by using the RSA scheme as a first concrete example. Since a circuit using the method of this paper as its construction rule has simple and regular structure, it provides an optimal construction of a highspeed RSA chip in VLSI. The method is also applicable for a simple implementation of the RSA scheme with small‐sized chips. In this case, the number of chips needed by this method (the circuit size) allows the practical realization of proportionately higher speeds. Flexibility can be achieved for a change in the number of key bits by a corresponding increase or decrease in the number of chips.

A processor architecture for 3D graphics

The DLX/3DCP architecture that uses a method of parallel processing on 3-D vectors to overcome the problem of the large number of floating-point operations required in 3-D graphics which limits the performance of graphics systems is described. The architecture's design offers general-purpose programmability from the high-level object-oriented language C++ and generates performance expected only from dedicated special-purpose hardware. Results that show the architecture's performance on graphics operations are presented and compared to the performance of other RISC processors.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Signal processing, image processing, and pattern recognition

A detailed exposition of the main areas of signal processing, this book is divided into three sections: one-dimensional signal processing and digital filters; two-dimensional signal processing and image processing; and pattern recognition. Among the more specific topics covered are: analog filters; discrete systems and signals; non-recursive filters; FFT; IIR design; quantization effects; and hardware and software design. There is also material on system stability, picture enhancement and restoration, and parallel processing methods, as well as a comprehensive treatment of syntactic methods, parsing and neural networks.

Continuous path planning via a non-inverting parallel algorithm

SUMMARYThe problem of path tracking in robotic manipulators applications is studied in this paper. The path is generated as a sequence of elementary motions. The characteristic feature of our algorithm is that it avoids singularities, because there is no need to use inverse kinematics. Direction and proximity criteria are introduced.The application of parallel processing methods to path tracking according to the previous algorithm is presented. The algorithm is implemented in the Alliant FX/80 parallel machine.

A noninverting algorithm for path tracking of two cooperating robot arms and its parallel implementation

This paper presents an algorithm for path tracking of two robot arms with end-effectors gripping a common inertial load. The path is generated as a sequence of elementary motions. The most important feature of the present algorithm is that it avoids singularities, because there is no need of using the inverse kinematics. Direction and proximity criteria are introduced. Holonomic constraints are formulated for the position and orientation of the two end-effectors. The application of parallel processing methods to path tracking according to the previous algorithm is presented. The algorithm is implemented in the Alliant FX/80 parallel machine.

Effective use of SIMD parallelism in low- and intermediate-level vision

The authors consider two examples from image understanding-focus-of-attention vision and contour image analysis-and present new parallel-processing methods that effectively support these types of computations. The research is a blend of theory and practice. On the one hand, the aim is to develop algorithms whose properties are well understood and can be formally related to key aspects of machine models. On the other hand, it is desired that the algorithms be easy to implement and practical in terms of their actual processing times on existing parallel machines. The experimental research was conducted on a 16384-processor Connection Machine CM2, and results of algorithm implementations on that machine are presented. >

パラレルリンク式微動ステージの６自由度位置決め制御

This paper describes a position control method and a control system configuration for newly developed fine motion stage using parallel linkage mechanism. A new static decoupling control method is proposed to compensate interference characteristics of 6-axes stage, including compensation of machine and actuator accuracy. Its decoupling parameters (Jacobian matrix) are calibrated by iterative process, and as a result total decoupling accuracy of 5% is obtained. The control algorithm, which requires complicated matrix calculation, could be executed within 250μs by parallel processing method using multi transputers. By experimental results of step response for each degree of freedom, it is confirmed that proposed decoupling method is more effective than conventional one. Consequently positioning resolution of 0.01μm (x, y), 0.123μrad (θz), response of 30-40 Hz, and dynamic rigidity of more than 12.3 N/μm (x-direction, frequency <40 Hz) are obtained.

Method and apparatus for vector quantizer parallel processing

Method and apparatus for vector quantizer parallel processing

Group explicit complete method vs. standard explicit method on mimd parallel systems

The explicit methods although very suitable for parallel processing always deny us reasonable accuracy and some stability; on the other hand, the implicit schemes offer stability, but the exploitation of these methods for parallel processing may be difficult and possibly inefficient. This paper investigates the applicability of a new class of group explicit methods to parallel processing, the experimental vehicle being a non-linear parabolic p.d.e. of second order. The method is briefly presented and then work is concentrated on the selection of a suitable algorithm which can effectively use the processing power of a MIMD parallel architecture and at the same time is stable and can produce highly accurate results.

A feasibility study of a direct-mapping parallel processing method to solve linear equations in load-flow calculation

With the growing size and complexity of power systems, system analysis—such as transients calculation—takes much time. Hence, fast calculation methods are required. Although parallel processing is a hopeful method, there have been difficulties in the parallel solution of linear equations which appear in power-flow calculations by the Newton-Raphson method. This paper aims at the fast calculation of the power-flow problem by means of parallel processing. In order to improve the suitability to the parallel solution of the differential equation in transients calculation, we assume the use of a direct-mapping parallel processing machine to map directly the network of a power system onto a network of processors. Under this assumption, we propose a new parallel-processing-oriented method in which the linear equation is solved by linear iterations between nodes with Aitken acceleration. We simulate the method on three model power systems and compare this Parallel Iterative Method (PIN) with a Parallel Direct Method (PDM) which uses the banded matrix according to the number of operations required. As a result, we can expect that PIM may solve linear equations faster than PDM with m processors, although the PIM might be inferior to the PDM with m × m processors, where m denotes the half-band width of the banded matrix.