Hamid R. Abachi

Network expandability and cost analysis of torus, hypercube and tree multiprocessor systems

This paper describes the network expandability and cost analysis of three important multi-processor system architectures. Included are the comparative network growth and network size of torus, hypercube and tree architectures by means of network modelling and graphical presentation. In this paper advantages and disadvantages of message-passing systems are highlighted. In addition, a new architecture called the augmented hypercube torus is proposed in order to eliminate limitations that apply to message passing architectures.

Analytical Modeling of Task Scheduling for Master-Slave Super Hypercube Architecture

This paper highlights the analytical modeling of task scheduling for a subclass of message passing architecture which is categorized as Master-Slave Super-Hypercube (MSSHP) topology. To establish this analytical modeling, its foundation based on Hypercube (HP) and Super-Hypercube architectures is first laid down and it is then expanded to include the MSSHP architecture.In order to highlight the advantages of the proposed modeling scheme, graphical presentations are introduced and compared. This comparison underlines the relationship between the time complexity for the HP and the new version for the MSSHP architecture. The outcome of this study reveals that the completion time for matrix product on the MSSHP is more favorable than Hypercube architecture.

Parallel processing computer architectures for process control

Modern, large-scale, high-speed process control automation is quite complex in nature. Continuous manufacturing processes currently in use require advanced computer technologies such as simultaneous parallel processing architecture for synchronous process operations that take place independent of each other. In order to optimise the performance of a process control system measured through its output variance, suitably structured robust computing structures are necessary for a multi-input/output process control. This paper describes: (1) a multi-ported-memory, star-ring parallel processing computer architecture for the process control of multiple inputs/outputs; and (2) a suitable parallel processing architecture for optimising the performance of an interconnected process control system. Both the architectures use a feedback control algorithm for multi-input/output process control by measuring its performance through its output product quality and for optimising the performance of the inter-connected process control system by using the product quality output as a performance measure.

An analysis of software engineering metrics in OO environment

This paper describes an automated software tool for computing a vector of software metrics. Such an automated tool helped perform a software engineering validation experiment. The preliminary results and the analysis of the experiment are presented.