DOC PREVIEW
UT EE 382C - Characterization of Embedded Workloads

This preview shows page 1-2-3 out of 9 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 9 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 9 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 9 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 9 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

EE 382C EMBEDDED SOFTWARE SYSTEMS Literature Survey Report Characterization of Embedded Workloads Ajay Joshi March 30, 2004 ABSTRACT Security applications are a class of emerging workloads that will play a central role in determining the performance of next generation embedded microprocessors. The objective of this research work is to understand the inherent workload characteristics of security applications, and analyze their impact on microprocessor architecture design. The outcome of this work will provide an insight into the performance bottlenecks in existing architectures that challenge security applications, and enable us to propose architectural enhancements required to boost the performance of embedded and digital processors running security applications. In this report we summarize the popular methodologies for characterizing workloads, survey classic studies that have been performed to measure the degree of parallelism in applications, and present a proposal to study the micro-architecture independent characteristics of security applications. I. Introduction The phenomenal growth in the Information Technology industry has resulted in the emergence of several new computer applications. The computer systems that run these applications were designed before the advent of these workloads, and their2architectural features may not match the application characteristics; possibly resulting in loss of performance. In order to ensure that microprocessors and computer systems of tomorrow deliver highest performance, it is extremely important for computer architects to identify these emerging workloads and understand their idiosyncrasies that challenge the existing architectures. In this work, we identify security applications as a key emerging embedded workload that will play a central role in determining the performance of tomorrow’s embedded microprocessors. The proposed research comprises of understanding the characteristics of these applications, accounting for the cycles spent during their execution, and identifying possible sources for the loss of performance. This report is organized as follows: Section II explains the motivation behind characterizing the behavior of security applications. Section III describes the previous research in workload characterization, and techniques that have been used to measure parallelism in an application. Section IV outlines the objectives of this project and proposes a methodology for carrying out the research work. Section V summarizes the key findings and draws conclusions from the literature review that influence the methodology in our research work. II. SECURITY APPLICATIONS – AN EMERGING WORKLOAD Until a few years ago, general-purpose processors and computers have been the driving force in shaping the digital economy. However, recently we have seen the proliferation of embedded systems in our daily lives through telecommunications, consumer, automotive, and office automation applications. Many of these embedded applications like cell phones, text message pagers, wireless hand held devices, pay-TV,3DSL modems, audio-video consumer products, telephony systems, and network routers, heavily rely on security mechanisms. Data security will play a central role in the design of these embedded systems. Security applications use cryptography algorithms to encrypt and decrypt messages sent over an insecure medium. It is therefore important to understand the characteristics of cryptography algorithms in order to design efficient next generation embedded processor and digital signal processor architectures. II. PREVIOUS WORK A. WORKLOAD CHARACTERIZATION APPROACHES Several early studies in characterizing program behavior are available in literature [3][4]. Workload characteristics have been measured at three different levels: source code, micro-architecture independent, and system level [7]. Source code level characterization yields information about inherent nature of the application. However, this technique has not gained much popularity because of the difficulty in standardizing the measured characteristics across different programming language implementations. Micro-architecture independent attributes, although not biased by a particular machine implementation, are influenced by the programming language, Instruction Set Architecture (ISA), compiler, and operating system. System level characteristics are micro-architecture dependent, and are widely used because they provide an insight into the match between an application and architecture. B. WORKLOADS STUDIED During the seventies, eighties, and early nineties, researchers mainly focused on studying the characteristics of scientific and high-performance applications written in Fortran. In the last ten years, general-purpose, e-commerce, web, graphics, and4multimedia workloads, developed in C, C++, Java, and other web technologies have gained popularity. It is only during the last few years, due the growing market segment of embedded systems, embedded workloads have been the focus of study of computer architecture researchers. However, characterization of embedded workloads has mainly concentrated on network, telecommunication, signal processing, and multimedia applications. C. MEASURING PARALLELISM IN PROGRAMS Parallelism in a program can be classified into three different types: Instruction Level Parallelism (ILP), Data Level Parallelism (DLP), and Thread Level Parallelism (TLP) [6]. These workload characteristics respectively measure the number of instructions, data, and threads in the program that can be concurrently executed - assuming that infinite machine resources are available. Understanding the parallelism in programs is an important in selecting the type of computer system (uniprocessor or multiprocessor), and the microprocessor architecture philosophy (superscalar, VLIW, Vector etc.). The possible limits of parallelism have been investigated for almost three decades. Numerous experiments have been performed that yield widely varying results on the limits of parallelism; primarily due to the differences in machine models assumed [12]. In this section we review three such studies that used different approaches and methodologies. Wall’s [2] study on the limits of ILP is considered to be the most thorough limit study to date, accounting for speculative execution, memory disambiguation, and other factors. Wall used the instruction trace


View Full Document

UT EE 382C - Characterization of Embedded Workloads

Documents in this Course
Load more
Download Characterization of Embedded Workloads
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Characterization of Embedded Workloads and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Characterization of Embedded Workloads 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?