Connie U. Smith

An Architecture Design and Assessment System for Software/Hardware Codesign

Codesign of hardware and software for high performance signal processing systems is important if the potential benefits of VLSI are to be realized. This article describes a CAD system developed to support the codesign of hardware and software architectures for high performance digital signal processors which is based on a directed graph methodology. A comprehensive example is developed to demonstrate the use of the system, the fundamentals of the modeling and analysis methodology are discussed, and an overview of the tools in the system is presented along with a discussion of the enhancements currently being developed.

Aspects of software design analysis: Concurrency and blocking

This paper extends previous work on development of a methodology for the prediction of the performance of computer software systems from design level specifications and continuing through implementation. The effects of synchronized behavior, such as results from data reservation in multi-thread executions of data base systems, and competition for host system resources are incorporated. The previous methodology uses hierarchical graphs to represent the execution of software on some host computer system (or on some abstract machine). Performance metrics such as response time were obtained from analysis of these graphs assuming execution of a single copy on a dedicated host. This paper discusses the mapping of these execution graphs upon queueing network models of the host computing environment to yield performance metric estimates for more complex and realistic processing environments.

Independent general principles for constructing responsive software systems

Three general principles are presented that can be applied in early software life cycle stages for the definition of software requirements and designs with acceptable performance. They are genuine high-level considerations for meeting responsiveness goals without sacrificing understandability and maintainability, and without increasing development time and cost. The principles are derived from the interrelationships of two performance models: a queueing network based on computer system model and an execution graph software model. The performance effect of each of the principles is quantified using the models. Examples are given that illustrate how they can be applied to software systems.

Performance engineering of software systems: a case study

This paper summarizes the concepts of performance engineering in large software systems and illustrates the application of performance engineering techniques to the early design phase of a large database system.n Performance engineering is a methodology for evaluating the performance of software systems throughout their life cycles. The case study given here demonstrates that it is possible to predict resource usage patterns of complex software systems even in early design phases of the system, although detailed predictions of resource usage are not likely to be validated. The results presented here show the leverage of considering performance implications in the early design phases of a software project.

Performance analysis of software for an MIMD computer

This paper presents a technique for modeling and analyzing the performance of software for an MIMD (Multiple Instruction Multiple Data) computer. The models can be used as an alternative to experimentation for the evaluation of various algorithms and different degrees of parallelism. They can also be used to study the tradeoffs involved in increasing the amount of parallel computation at the expense of increased overhead for synchronization and communication. The detection and alleviation of performance bottlenecks is facilitated.

Software Performance Engineering

Performance is critical to the success of todays software systems. However, many software products fail to meet their performance objectives when they are initially constructed. Fixing these problems is costly and causes schedule delays, cost overruns, lost productivity, damaged customer relations, missed market windows, lost revenues, and a host of other difficulties. This chapter presents software performance engineering (SPE), a systematic, quantitative approach to constructing software systems that meet performance objectives. SPE begins early in the software development process to model the performance of the proposed architecture and high-level design. The models help to identify potential performance problems when they can be fixed quickly and economically.

Future Directions for VLSI and Software Engineering

A structure is developed and used for the comparison of the VLSI and software design processes. The comparison reveals both similarities and differences in the two disciplines and in their evolution. Several possible implications are presented which contribute to an understanding of the processes and provide insights into appropriate directions for future research. Finally, some areas with high potential gains through technology transfer are discussed.