An approach to the optimum design of computer graphics systems

Abstract

Display system designers are faced with the difficult task of selecting major subsystems in an intelligent way. Each subsystem is chosen from large numbers of alternatives; the selection is based on considerations such as system response time, system cost, and the distribution of data storage and processing between the graphics processor and its supporting data processing system. The work reported here develops an objective, quantitative design procedure and helps give a better understanding of how to configure display systems. This is accomplished by means of a mathematical model of a computer driven graphics system.The parameters of the model are functions of the capabilities of the graphics hardware and of the computational requirements of the graphics application. The model can be analyzed using numerical queueing analysis or simulation to obtain an average response time prediction. By combining the model with an optimization, the best graphics system configuration, subject to a cost constraint, is found for several applications. The optimum configurations are in turn used to find general display system design guidelines.

Configurable software for satellite graphics

“…People at the University of North Carolina produced a system for Configurable Applications for Graphics Employing Satellites (CAGES). Although Foley had previously done modeling studies of computer graphics systems in which performance was a function of the capabilities of the graphics hardware and the computational requirements of the application (Foley, 1971), the hardware was not as extraordinary as that assembled at Brown. The satellite computer is a PDP 11/45 and the host computer is an IBM 360/75.…”

Cost-benefit analysis of computer graphics systems

“…One of the first published results of a quantitative analysis of design, tradeoffs was Foley's study of satellite graphics systems [11].…”

“…Clearly, the matter of the responsiveness required of a graphics system is not a straightforward question, and it does require serious consideration on the part of the system designer. * The dilemma is exacerbated by the demonstrated sensitivity of responsiveness to additional investment [11],…”

Operating system performance

“…Once this situation has been quantified, together with a quantification of the objective functions, we may proceed to the formal optimization of the system in an easy step. For an example of sueh an approach see Foley [17] and Stimler and Bions [35]. I must point out that an adequate methodology for describing the parameters and relationships is not in the state of the art.…”