Yu-Fai Fung

Heuristic algorithms for multiprocessor task scheduling in a two-stage hybrid flow-shop

Abstract The aim of this paper is to propose heuristic algorithms for a two-stage flow-shop scheduling problem with multiprocessor tasks to minimize the makespan. The heuristic algorithms are constructive in nature. They schedule the tasks one at a time from a given sequence. The rules considered for sequencing the tasks are based on simple priority rules from the literature. Some lower bounds for the problem are also derived to be used in the performance analysis of the heuristic algorithms. Next, the average performance of the proposed heuristic algorithms is analyzed by a computational experiment using randomly generated problem instances. The results suggest that these heuristic algorithms are both efficient and effective. The paper concludes with a discussion of the insights obtained from the experimental analysis about this type of scheduling problems.

The design and evaluation of a multiprocessor system for computer vision

Abstract This paper introduces a multi-layer MIMD system for computer vision applications. The system is provided with three hierarchical processing layers, each of which is dedicated to one level of processing to create a pipelining effect. Unlike the traditional approach, where a 2D mesh connected array is used, layers in this system are linear arrays. Simplicity and expandability are the main advantages. A prototype of the system is implemented using off-the-shelf components and the performance of the various vision operations is analyzed.

A parallel solution to linear systems

Streaming SIMD Extensions (SSE) is a unique feature embedded in the Pentium III and IV classes of microprocessors. By fully exploiting SSE, parallel algorithms can be implemented on a standard personal computer and a theoretical speedup of four can be achieved. In this paper, we demonstrate the implementation of a parallel LU matrix decomposition algorithm for solving linear systems with SSE and discuss advantages and disadvantages of this approach based on our experimental study.

Automatic device configuration and data validation through mobile communication

The introduction of personal computing and wireless communication technology provides an option for on site device software updating and data retrieving. This is especially true for any devices sitting in a remote site where computing network is not accessible. In many advanced computing systems, frequent software updating and configuration profiles refreshing are required. This is clumsy and error prone procedures when users are not familiar with the operating systems. Suppose all the necessary files and programs are predefined in a mobile computing device such as notebooks, PDAs, or even mobile phones. All necessary files and software can be transferred to the corresponding computing devices and PCs at remote sites through wireless communication links such as Bluetooth, infrared, general packet radio service (GPRS). This idea helps solve the initial installation cost of a communication network to a remote site.

An advanced evolutionary algorithm for parameter estimation of the discrete Kalman filter

Abstract In this work we design an advanced Evolutionary Algorithm (EA) for optimizing the discrete Kalman filter (KF) model. The EA employs parallel architecture and an advanced mutation operator called the “Selection Follower”. Its performance is benchmarked with that of the Expectation-Maximization algorithm (EM) in minimizing the mean-square-error of the KF prediction. Experimental results show that the EA consistently outperforms the EM and runs significantly faster under the same number of function evaluations.

Performance of particle swarm optimization in scheduling hybrid flow-shops with multiprocessor tasks

In many industrial and computing applications, proper scheduling of tasks can determine the overall efficiency of the system. The algorithm, presented in this paper, tackles the scheduling problem in a multi-layer multiprocessor environment, which exists in many computing and industrial applications. Based on the scheduling terminology, the problem can be defined as multiprocessor task scheduling in hybrid flow-shops. This paper presents a particle swarm optimization algorithm for the solution and reports its performance. The results are compared with other well known meta-heuristic techniques proposed for the solution of the same problem. Our results show that particle swarm optimization has merits in solving multiprocessor task scheduling in a hybrid flow-shop environment.

A PC based parallel LU decomposition algorithm for sparse matrices

Linear equations are used in the mathematical models of many engineering problems, including load-flow in power engineering and electrical railway simulation. A common approach for solving the linear equations is by LU decomposition, which is then followed by forward and backward substitution. The LU decomposition operation is a computation intensive algorithm and in this paper, we present a cost-effective SIMD parallel algorithm for the LU decomposition of sparse matrices. The algorithm can be implemented using a common personal computer and does not require other hardware support.

Scheduling image processing tasks in a multilayer system

Abstract Multilayer multiprocessor systems are generally employed in real-time applications such as robotics and computer vision. This paper introduces three heuristic algorithms for multiprocessor task scheduling in such systems. In our model, tasks with arbitrary processing times and arbitrary processor requirements are considered. The scheduling aims at minimising completion time of processes in a two-layer system. We employed an effective lower bound (LB) for the problem. Then, we analysed the average performance of the heuristic algorithms by computing the average percentage deviation of each heuristic solution from the LB on a set of randomly generated problems. We have also applied these algorithms for scheduling computer vision tasks running on prototype multilayer architecture. Our computational and empirical results showed that the proposed heuristic algorithms perform well.

Parallel image processing with one-dimensional DSP arrays

Abstract This paper describes a study of parallel image processing algorithms implemented on a one-dimensional DSP array. DSPs are developed for computationally intensive signal processing operations. Recently introduced parallel DSPs can be used for all levels of image processing operations and they provide easy development of a parallel system. In addition, due to the computing power delivered by these processors, we can employ coarse grain parallelism instead of the traditional fine-grain parallelism. Modularity, expandability and easy programming are other advantages of parallel DSPs. In this paper, parallel implementation of some selected image processing algorithms is described and performance results are presented.

A system to assist the learning of embedded microprocessor programming

When implementing final year projects, students may need to program a microprocessor in order to achieve the project objectives. Therefore, a system is designed in order to assist students who have limited background knowledge in computer engineering to learn the basics of microprocessor programming primarily using the C programming language. In addition to the software, various hardware components are provided so that students can develop simple systems so the learning outcomes can be enhanced by practice. From feedback collected, users found that they are able to learn simple C language programming for a microprocessor with the help of the system.