High Execution Speed Research Articles

Artificial neural networks as a tool in urban storm drainage

The introduction of Artificial Neural Networks (ANNs) as a tool in the field of urban storm drainage is discussed. Besides some basic theory on the mechanics of ANNs and a general classification of the different types of ANNs, two ANN application examples are presented; The prediction of runoff coefficients and the restoration of rainfall data. From the results, it can be concluded that ANNs can deal with problems that are traditionally difficult for conventional modelling techniques to solve. Their advantages include good generalisation abilities, high fault tolerance, high execution speed, and the ability to adapt and learn. However, ANNs rely strongly on the quantity of data examples, their training is occasionally slow, and they are not transparent and obstruct any closer analysis and interpretation of their performance. Finally, it is expected that the future of ANNs will lie in its integration with other conventional and more advanced modelling techniques, creating so-called hybrid models.

High-speed parallel implementation of a modified PBR algorithm on DSP-based EH topology

Algebraic Reconstruction Technique (ART) is an age-old method used for solving the problem of three-dimensional (3-D) reconstruction from projections in electron microscopy and radiology. In medical applications, direct 3-D reconstruction is at the forefront of investigation. The simultaneous iterative reconstruction technique (SIRT) is an ART-type algorithm with the potential of generating in a few iterations tomographic images of a quality comparable to that of convolution backprojection (CBP) methods. Pixel-based reconstruction (PBR) is similar to SIRT reconstruction, and it has been shown that PBR algorithms give better quality pictures compared to those produced by SIRT algorithms. In this work, we propose a few modifications to the PBR algorithms. The modified algorithms are shown to give better quality pictures compared to PBR algorithms. The PBR algorithm and the modified PBR algorithms are highly compute intensive, Not many attempts have been made to reconstruct objects in the true 3-D sense because of the high computational overhead. In this study, we have developed parallel two-dimensional (2-D) and 3-D reconstruction algorithms based on modified PBR. We attempt to solve the two problems encountered by the PBR and modified PBR algorithms, i.e., the long computational time and the large memory requirements, by parallelizing the algorithm on a multiprocessor system. We investigate the possible task and data partitioning schemes by exploiting the potential parallelism in the PBR algorithm subject to minimizing the memory requirement. We have implemented an extended hypercube (EH) architecture for the high-speed execution of the 3-D reconstruction algorithm using the commercially available fast floating point digital signal processor (DSP) chips as the processing elements (PEs) and dual-port random access memories (DPR) as channels between the PEs. We discuss and compare the performances of the PBR algorithm on an IBM 6000 RISC workstation, on a Silicon Graphics Indigo 2 workstation, and on an EH system. The results show that an EH(3,1) using DSP chips as PEs executes the modified PBR algorithm about 100 times faster than an LBM 6000 RISC workstation. We have executed the algorithms on a 4-node IBM SP2 parallel computer. The results show that execution time of the algorithm on an EH(3,1) is better than that of a 4-node IBM SP2 system. The speed-up of an EH(3,1) system with eight PEs and one network controller is approximately 7.85.

Artificial neural networks as a tool in urban storm drainage

The introduction of Artificial Neural Networks (ANNs) as a tool in the field of urban storm drainage is discussed. Besides some basic theory on the mechanics of ANNs and a general classification of the different types of ANNs, two ANN application examples are presented: The prediction of runoff coefficients and the restoration of rainfall data. From the results, it can be concluded that ANNs can deal with problems that are traditionally difficult for conventional modelling techniques to solve. Their advantages include good generalisation abilities, high fault tolerance, high execution speed, and the ability to adapt and learn. However, ANNs rely strongly on the quantity of data examples, their training is occasionally slow, and they are not transparent and obstruct any closer analysis and interpretation of their performance. Finally, it is expected that the future of ANNs will lie in its integration with other conventional and more advanced modelling techniques, creating so-called hybrid models.

Minimum loss reconfiguration of unbalanced distribution networks

A heuristic method for determining the configuration with minimum resistive line losses for three-phase unbalanced distribution networks is described. This method is based on previous papers dealing with balanced networks. Using a fast and reliable load flow solution technique and also, efficient algorithms for network elements inspection, so that methodology becomes capable of solving the challenging problem while still maintaining a high execution speed. A developed algorithm is applied on a practical distribution network. Performed analysis highlights the influence of load imbalances on a solution of minimum loss reconfiguration problem.

High-speed 2-D convolution with a custom computing machine

Custom computing machines, a class of computational platforms consisting of reconfigurable functional units with reconfigurable interconnection networks, provide a middle-ground between special-purpose hardware, which provide high execution speed, and general-purpose computers, which offer flexibility. The Splash-2 system, one such custom computing machine, is an experimental platform for complex computations requiring the high speed of special-purpose hardware. The reconfigurability and modularity of Splash-2, along with automated synthesis tools, allow for rapid, staged development of applications. This paper demonstrates the applicability of Splash-2 to the image-processing area and gives an introduction to the programming environment used in developing applications. An example image-processing application based upon two-dimensional convolution is described and the operating procedures of custom computing machines are presented. Also presented are the details of directly implementing two-dimensional convolution in a straight-forward, systolic fashion in reconfigurable hardware.

Superfast autoconfiguring artificial neural networks and their application to power systems

Abstract This paper presents the new generation of artificial neural netoworks (ANNS) for solving the task of power system operation planning. Today the error back-propagation ANNs are used most because of their simplicity and the possibility of parallel implementation on neuro-computers for high-speed execution. In spite of their popularity they have two major drawbacks: the learning process is time consuming and there is no exact rule for setting the number of neurons to avoid overfitting or underfitting and to achieve, hopefully, a converging learning phase. To avoid these difficulties, a new generation of ANNs has been developed based on the theory of radial basis functions for approximations. A comparison test on an actual problem in power system operation was performed. The results show that this new algorithm is superior to back-propagation ANNs and optimal configured back-propagation ANNs achieved with genetic algorithms.

A three-phase power flow method for real-time distribution system analysis

This paper presents a three-phase power flow solution method for real-time analysis of primary distribution systems. This method is a direct extension of the compensation-based power flow method for weakly meshed distribution systems from single phase to three-phase, with the emphasis on modeling of dispersed generation (PV nodes), unbalanced and distributed loads, and voltage regulators and shunt capacitors with automatic local tap controls. The method proposed here is capable of addressing these modeling challenges while still maintaining a high execution speed required for real-time application in distribution automation systems. The paper also includes test results from the application of a computer program developed based on the proposed method to large primary electric distribution systems.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Development of a Propagation Model to Determine the Spread of Accidental Pollution in Rivers

Pollutant spread in rivers is affected by numerous complicated phenomena, and only through modelling such spread is it possible to predict the concentration of solutes at strategic points along the river in case of accidental water pollution and to set up appropriate procedures to limit undesirable effects. The BRGM (Bureau de Recherches Géologiques et Minières) has developed such a model to simulate the behaviour of any pollutant spilled into the surface water network and has integrated it in the “SeauS” (SOS Water) software package designed by the BRGM to help reach a decision in matters of water protection. The model is based on the “mixing cells” method which has a high execution speed and is suitable for simulating the various flow mechanisms occurring in rivers; i.e. advection, diffusion, connected networks, interactions between the main flow and areas of stagnant water or groundwater, etc. Once its numerical performance had been proved, the model was tested in schematic cases and its results compared with analytical solutions. In each case the model showed its ability to represent the natural phenomena very well.

表参照方式Ｈｏｕｇｈ曲線描画における表構成の効率化と高速化

The table look-up is a powerful method for high-speed execution of the Hough transform, and this paper describes how an efficient look-up table should be arranged. Investigating the constitution of the table from the viewpoint of size, number of execution steps, execution times, amount of error, etc reveals that the best way is to prepare a table based on the equationρ=r·sin (θ+φ), derived from p=x·cosθ+y·sinθ (which is the original equation of the Hough transform) A software experiment showed that the time of θ-ρ plane drawing is reduced 80-fold from that using the original Hough transform equation The proposed principle is implemented in experimental hardware in which the dimension of the plane in 512 (9bit) ×512, and only one ROM.The drawing time for one point on the θ-ρ plane is less than 100 ns.

The Gmicro/500 superscalar microprocessor with branch buffers

The Gmicro/500, which features a RISC-like dual-pipeline structure for high-speed execution of basic instructions and represents a significant advance for the TRON architecture, is presented. Upwardly-object-compatible with earlier members of the Gmicro series, this microprocessor uses resident dedicated branch buffers to greatly enhance branch instruction execution speed. Its microprograms simultaneously use dual execution blocks to execute high-level language instructions effectively. Fabricated with a 0.6- mu m CMOS technology on a 10.9-mm*16-mm die, the chip operates at 50/66 MHz and achieves a processing rate of 100/132 MIPS.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Measurement of two-dimensional strain distribution by speckle correlation

This paper reports a parametrically amplified MEMS rate gyroscope. A practical parametric excitation scheme implemented using Digital Signal Processing (DSP) has been developed to enable either amplification of the primary mode of the gyroscope or amplification of the response to the applied angular velocity. The primary objective of this work is to improve the scale-factor and the signal to noise performance of the gyroscope. Parametric amplification of the primary mode of the gyroscope is achieved by frequency tracking and regulation of the amplitudes of the harmonic forcing and parametric excitation to maintain a desired parametric gain by closed loop PID control. Stable parametric amplification of the primary mode by a factor of 20 is demonstrated experimentally. This has important benefits regarding the minimisation of electrical feedthrough of the drive signal to the sense electrodes of the secondary mode. By taking advantage of the phase dependence of parametric amplification and the orthogonality of the Coriolis force and quadrature forcing, the response to the applied angular velocity may be parametrically amplified by applying excitation of a particular phase directly to the sensing mode. The major advantage of parametric amplification applied to MEMS gyroscopes is that it can mechanically amplify the Coriolis response before being picked off electrically. This is particularly advantageous for sensors where electronic noise is the major noise contributor. In this case parametric amplification can significantly improve the signal to noise ratio of the secondary mode by an amount approximately equal to the parametric amplification. Preliminary rate table tests performed in open loop demonstrate a magnification of the signal to noise ratio of the secondary mode by a factor of 9.5 and the scale-factor by 11. The excitation and control scheme have been implemented digitally using a DSP development board enabling very high sampling precision and execution speed necessary for gyroscopic applications.

Calculating short range forces on a massively parallel computer: SPH on the Connection Machine

Three implementations of SPH on a Connection Machine are presented. The first one is a highly parallel direct summation implementation which maximally utilizes CM computational power but suffers from N 2 scaling. The second one uses cells to group particles in space and hence reduce the number of inter-particle forces to be computed. Very high execution speeds are obtained for regular particle distributions by processing these cells in parallel, however, for uneven particle distributions, the approach degrades again into N 2 scaling behavior. The third approach uses multiple grids and the duplication of particles to combine linear scaling with good load-balancing. A full SPH time step takes 3.1 seconds for a regular particle distribution and 46 seconds for an irregular one for 128k particles on a 64k CM200.

Automated Generation of Solid Objects from Wire-Frame Data Using Operations on Graphs

Recently, mechanical CAD systems with solid model are used widely in the design process. However, it is not easy to build solid model with some kind of primitive objects and set operations. To solve this problem, we propose an algorithm of automatic conversion from wire-frame data to solid model. Although a large number of studies have been made on this area, previous methods couldn't handle solid objects with free-form surfaces and only treated polyhedra-approximated solid objects. Moreover, some of these approaches require lots of execution time. Because they are based on the combinatorial search algorithm which find actual faces of objects. Our approach solved above two problems on some conditions. It offers high-speed execution and can treat free-form surfaces.

A new DSP-oriented algorithm for calculation of the square root using a nonlinear digital filter

A high-speed algorithm for calculating the square root is proposed. This algorithm, which can be regarded as calculation of the step response of a kind of nonlinear IIR filter, requires no divisions. Therefore, it is suitable for a VLSI digital signal processor (DSP) which has a high-speed hardware multiplier but does not usually have a high-speed hardware divider. The convergence properties of the algorithm are analyzed and used to develop a practical implementation of the procedure. It is implemented on the commercially available DSP (TM320C25) and is compared with the Newton-Raphson method. The proposed algorithm has two advantages over the Newton-Raphson method: higher execution speed and smaller calculation error.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Implementing a DSP-based Petri-net simulation tool

An efficient simulator of Petri nets, suitable for both loosely and tightly coupled parallel systems and featuring high-execution speed, is proposed. This speed is obtained through an optimized software structure on an original hardware architecture that is based on digital signal processors of the TMS320 family. The Petri net class outlined compactly represents a wide range of processes by restricting the model size and the relevant simulation times. The simulation algorithm and simulator implementation are described. >

A digital architecture employing stochasticism for the simulation of Hopfield neural nets

A digital architecture which uses stochastic logic for simulating the behavior of Hopfield neural networks is described. This stochastic architecture provides massive parallelism (since stochastic logic is very space-efficient), reprogrammability (since synaptic weights are stored in digital shift registers), large dynamic range (by using either fixed- or floating-point weights), annealing (by coupling variable neuron gains with noise from stochastic arithmetic), high execution speed ( approximately=N*10/sup 8/ connections per second), expandability (by cascading of multiple chips to host large networks), and practicality (by building with very conservative MOS device technologies). Results of simulations are given which show the stochastic architecture gives results similar to those found using standard analog neural networks or simulated annealing.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

High-speed implementations of rule-based systems

Rule-based systems are widely used in artificial intelligence for modeling intelligent behavior and building expert systems. Most rule-based programs, however, are extremely computation intensive and run quite slowly. The slow speed of execution has prohibited the use of rule-based systems in domains requiring high performance and real-time response. In this paper we explore various methods for speeding up the execution of rule-based systems. In particular, we examine the role of parallelism in the high-speed execution of rule-based systems and study the architectural issues in the design of computers for rule-based systems. Our results show that contrary to initial expectations, the speed-up that can be obtained from parallelism is quite limited, only about tenfold. The reasons for the small speed-up are: (1) the small number of rules relevant to each change to data memory; (2) the large variation in the processing requirements of relevant rules; and (3) the small number of changes made to data memory between synchronization steps. Furthermore, we observe that to obtain this limited factor of tenfold speed-up, it is necessary to exploit parallelism at a very fine granularity. We propose that a suitable architecture to exploit such fine-grain parallelism is a shared-memory multiprocessor with 32-64 processors. Using such a multiprocessor, it is possible to obtain execution speeds of about 3800 rule-firings/set. This speed is significantly higher than that obtained by other proposed parallel implementations of rule-based systems.

An adaptive seam tracker for welding heavy-section aluminum

Robotic welding systems which cannot adapt in real time to changes in the joint geometry along the weld seam and to the seam position itself have only limited success in many welding applications, due to heat distortion of the weldment or to an overall shift in the weld seam caused by fixturing errors. A solution to this problem requires the application of some form of adaptive joint geometry sensing system for guidance of the welding torch. Research is described that addresses the general aluminum welding problem by integrating a laser-based part-profiling sensor with a welding robot. An adaptive system is being developed that is referred to as ARTIST (Adaptive Real-Time Intelligent Seam Tracker). An IBM PC/AT-compatible microcomputer serves as the system supervisor, commands the motion of the robot, directs and processes the sensor scans, and establishes the required weld parameters. The programming language used in the microcomputers is C, chosen primarily for its flexibility and high execution speed. The host robot is a Unimate 760 Series electric six-axis articulated arm, using a VAL II controller.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Minimum-diameter cyclic arrangements in mapping data-flow graphs onto VLSI arrays

Regular arrays of processing elements in VLSI have proved to be suitable for high-speed execution of many matrix operations. To execute an arbitrary computational algorithm on such processing arrays, it has been suggested mapping the given algorithm directly onto a regular array. The computational algorithm is represented by a data-flow graph whose nodes are to be mapped onto processors in the VLSI array.This study examines the complexity of mapping data-flow graphs onto square and hexagonal arrays of processors. We specifically consider the problem of routing data from processors in a given (source) sequence to another (target) sequence.We show that under certain conditions, the above problem is equivalent to the one of finding a minimum-diameter cyclic arrangement. The complexity of the latter problem is analyzed and upper and lower bounds on the number of intermediate rows of processors (between the source and target rows) are derived.

New books

Of interest to AI system designers and researchers in expert systems, this monograph explores the role of parallelism in the high-speed execution of production systems. It presents simulation results for a large set of production systems exploiting different sources of parallelism. The thesis points out the features of existing programs that limit the speed-up obtainable from parallelism and suggests solutions for some of the bottlenecks.