1
play

1 Introduction to ES Architectures Components and Systems Embedded - PDF document

Oct 16, 2022 •299 likes •362 views

Lectures 5 & 6 Overview Processors and Architectures General Purpose vs. Application Specific for Embedded Systems Processors General Purpose Core (IP)-based design vs. Application Specific Processors Reconfigurable Systems

  1. Lectures 5 & 6 Overview Processors and Architectures • General Purpose vs. Application Specific for Embedded Systems Processors General Purpose • Core (IP)-based design vs. Application Specific Processors • Reconfigurable Systems General Purpose vs. Application Specific Processors Slide 1 General Purpose vs. Application Specific Processors Slide 2 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Agenda Embedded Microprocessors Overview Trends in Embedded Microprocessor design • Desktop vs. Embedded processors • Microcontrollers and Microprocessors • Trends in Embedded-Microprocessor design • New applications drive requirements • Embedded System Architectures – game consoles • Customized Instruction Sets for Embedded – handheld, palmtop, automobile and network PCs Processors – cellular phones and other mobile communicators • Selected Co-design problems – modems, fax machines, printers, etc. • Conclusions – digital cameras General Purpose vs. Application Specific Processors Slide 3 General Purpose vs. Application Specific Processors Slide 4 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Comparing Embedded Processors Standardization Trends in Embedded Trends in Embedded Microprocessor design Microprocessor design • Power consumption • The heterogenity of embedded architectures • Code density • A need to unify the embedded processor market? • Peripheral integration and chipsets • Windows CE and JAVA as examples • Multimedia acceleration • Will management influence standardization? • Price/performance ratio General Purpose vs. Application Specific Processors Slide 5 General Purpose vs. Application Specific Processors Slide 6 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 1

  2. Introduction to ES Architectures Components and Systems Embedded System Architectures Embedded System Architectures Introduction Architecture Specialization Techniques • ES Architectures are determined by: • Components as function blocks – circuit technology; • Systems as large entities of integrated – application requirements; components – market constraints: • Single control thread vs. multiple control threads • strict cost margins • architecture specialization • time-to-market and predictable design time – component specialization techniques • (hard) time constraints • power dissipation – system specialization techniques • safety • physical constraints General Purpose vs. Application Specific Processors Slide 7 General Purpose vs. Application Specific Processors Slide 8 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Component Specialization Techniques System Specialization Techniques Embedded System Architectures Embedded System Architectures Architecture Specialization Techniques Architectures Specialization Techniques • Instruction set specialization • Load distribution – Control decomposition (control clustering) • Function unit and data path specialization – Data decomposition (data clustering) • Memory specialization • Component interaction • Interconnect specialization • Control specialization General Purpose vs. Application Specific Processors Slide 9 General Purpose vs. Application Specific Processors Slide 10 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 System Control Specialization Dependent Coprocessors Embedded System Architectures Embedded System Architectures Architecture Specialization Techniques Architecture Specialization Techniques • Asymmetric control relationships require global control flow • The following global control mechanisms can be distinguished: – Independently controlled components – Dependent coprocessors – Incrementally controlled coprocessors – Partially dependent coprocessors General Purpose vs. Application Specific Processors Slide 11 General Purpose vs. Application Specific Processors Slide 12 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 2

  3. Incrementally Controlled Coprocessors Partially Dependent Coprocessors Embedded System Architectures Embedded System Architectures Architecture Specialization Techniques Architecture Specialization Techniques General Purpose vs. Application Specific Processors Slide 13 General Purpose vs. Application Specific Processors Slide 14 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Application System Classes Control Dominated Systems Architectures Embedded System Architectures Embedded System Architectures Embedded System Architectures Target Architectures and Application System Classes Architectures for Control Dominated Systems • Computation oriented systems • Input MOCs are coupled FSM or Petri-Nets • Control dominated systems • Co-design problems: – execution of concurent FSMs/Petri-Nets reacting to • Data dominated systems asynchronous input events • Mixed systems – FSM transition synchronization – event scheduling – correctness General Purpose vs. Application Specific Processors Slide 15 General Purpose vs. Application Specific Processors Slide 16 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 8051 - an 8bit Microcontroller Architecture Examples - Philips 80C552 Embedded System Architectures Embedded System Architectures Architectures for Control Dominated Systems Architectures for Control Dominated Systems General Purpose vs. Application Specific Processors Slide 17 General Purpose vs. Application Specific Processors Slide 18 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 3

  4. Data Dominated Systems Architectures Examples - MMX Embedded System Architectures Embedded System Architectures Architectures for Data Dominated Systems Architectures for Data Dominated Systems • Data transport or processing is dominant • Flow graph languages are used to describe such systems • Opportunities for specialization: – periodic execution often corresponds to a input data independent system data flow – input data is mostly generated with a fixed period (the sample rate) General Purpose vs. Application Specific Processors Slide 19 General Purpose vs. Application Specific Processors Slide 20 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Examples - Hitachi SH-DSP Examples - Philips Trimedia TM1000 Embedded System Architectures Embedded System Architectures Architectures for Data Dominated Systems Architectures for Data Dominated Systems General Purpose vs. Application Specific Processors Slide 21 General Purpose vs. Application Specific Processors Slide 22 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Examples - TI TMS320C80 Customized Instruction-Sets Embedded System Architectures Customized Instruction Sets Architectures for Data Dominated Systems Introduction • Instruction-Set Architectures (ISAs) are the visible instructions of a processor • Is there a strong motivation for customizing processors’ instruction-sets? General Purpose vs. Application Specific Processors Slide 23 General Purpose vs. Application Specific Processors Slide 24 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 4

  5. Barriers Co-design Problems Customized Instruction Sets Barriers • Existing binaries • Instruction-set definition • Toolchain costs and user familiarity • Instruction encoding for code compression • Lost savings/higher chip cost due to lower • Instruction encoding for memory optimization volumes • Global control and data flow optimization • Hardware development costs • Communication channel selection • The product development cycle for embedded • Component interface synthesis products • Component selection/reuse General Purpose vs. Application Specific Processors Slide 25 General Purpose vs. Application Specific Processors Slide 26 Eduard Turcan October 24, 2001 Eduard Turcan October 24, 2001 Conclusions ? General Purpose vs. Application Specific Processors Slide 27 Eduard Turcan October 24, 2001 5

Recommend

More recommend