Performance Analysis of Real-Time Embedded Software

by Benjamin Ehrenberg (Joint Author), Yau-Tsun Steven Li, Sharad Malik

Overview

Embedded systems are characterized by the presence of processors running application-specific software. Recent years have seen a large growth of such systems, and this trend is projected to continue with the growth of systems on a chip. Many of these systems have strict performance and cost requirements. To design these systems, sophisticated timing analysis tools are needed to accurately determine the extreme case (best case and worst case) performance of the software components.
Existing techniques for this analysis have one or more of the following limitations:

• they cannot model complicated programs
• they cannot model advanced micro-architectural features of the processor, such as cache memories and pipelines
• they cannot be easily retargeted for new hardware platforms.

In Performance Analysis of Real-Time Embedded Software, a new timing analysis technique is presented to overcome the above limitations. The technique determines the bounds on the extreme case (best case and worst case) execution time of a program when running on a given hardware system. It partitions the problem into two sub-problems: program path analysis and microarchitecture modeling.
Performance Analysis of Real-Time Embedded Software will be of interest to Design Automation professionals as well as designers of circuits and systems.

Product Details

ISBN-13:	9780792383826
Publisher:	Springer US
Publication date:	11/30/1998
Edition description:	1999
Pages:	146
Product dimensions:	6.10(w) x 9.25(h) x 0.02(d)

Table of Contents

1. Introduction.- 1.1 The Emergence of Embedded Systems.- 1.2 Performance Constraints of Embedded Systems.- 1.3 Challenges in Designing Embedded Systems.- 1.4 Research Goals.- 1.5 Summary.- 1.6 Organization of this Book.- 2. Related Work in Timing Analysis for Embedded Software.- 2.1 Introduction.- 2.2 Program Path Analysis.- 2.3 Microarchitecture Modeling.- 2.4 Retargetability Issues.- 2.5 Summary.- 3. Program Path Analysis.- 3.1 Introduction.- 3.2 Problems with Program Path Analysis.- 3.3 Execution Count Analysis.- 3.4 Program Control Flow and Logical Flow.- 3.5 Integer Linear Programming Formulation.- 3.6 Solving ILP Problems.- 3.7 Experimental Validation.- 3.8 Chapter Conclusions.- 4. Microarchitecture Modeling.- 4.1 Introduction.- 4.2 Simple Microarchitectures.- 4.3 Advanced Microarchitectures and Memory Systems.- 4.4 Cache Modeling.- 4.5 Instruction Cache Modeling.- 4.6 Direct Mapped Instruction Cache Analysis.- 4.7 Set Associative Instruction Cache Analysis.- 4.8 Interprocedural Calls.- 4.9 Data Cache Modeling.- 4.10 Pipeline Modeling.- 4.11 Experiments.- 4.12 Chapter Conclusions.- 5. A Retargetable Timing Analysis Tool — Cinderella.- 5.1 Introduction.- 5.2 Issues in Timing Analysis.- 5.3 Classification of Retargeting Information.- 5.4 Implementation of Retargetable Modules.- 5.5 Operations.- 5.6 Chapter Conclusions.- 6. Conclusions.- 6.1 Contributions.- 6.2 Future Research Directions.- Appendices.- A — Practical Complexity of the ILP Problems.- References.

Editorial Reviews

Presents a new timing analysis technique to overcome limitations of current methods for determining the fastest and slowest that a program would take to run on a particular computer. The inability to make accurate and reliable estimates means that hardware must often be more powerful and expensive to make sure the software meets all timing deadlines. The new technique partitions the problem into two sub- problems: program path analysis, which looks at the structure of the program and determines the set of paths that result in the extreme case execution time; and microarchitecture modeling, which looks at the timing properties of the hardware, including pipelined processors and caches, and determines the execution time of a known sequence of instructions. It can be easily adapted to new hardware platforms. Annotation c. Book News, Inc., Portland, OR (booknews.com)

Booknews