Worst-Case Execution Time Analysis Martin Toft mt@cs.aau.dk PhD - - PowerPoint PPT Presentation

worst case execution time analysis
SMART_READER_LITE
LIVE PREVIEW

Worst-Case Execution Time Analysis Martin Toft mt@cs.aau.dk PhD - - PowerPoint PPT Presentation

Worst-Case Execution Time Analysis Martin Toft mt@cs.aau.dk PhD student Distributed and Embedded Systems Department of Computer Science Aalborg University October 22, 2009 Introduction The challenges My research Embedded software System


slide-1
SLIDE 1

Worst-Case Execution Time Analysis

Martin Toft mt@cs.aau.dk

PhD student Distributed and Embedded Systems Department of Computer Science Aalborg University

October 22, 2009

slide-2
SLIDE 2

1 / 10

Introduction The challenges My research Embedded software System types WCET analysis

Embedded software is everywhere

Consumer electronics Household appliances Home automation Transportation Industrial production plants Medical equipment Military equipment Trend: more software, less specialised hardware Easier to upgrade/change software than hardware

Picture: http://www.xdesignlabs.com/guide/xtachbig.jpg Martin Toft Worst-Case Execution Time Analysis

slide-3
SLIDE 3

2 / 10

Introduction The challenges My research Embedded software System types WCET analysis

Safety-critical systems

Many safety-critical systems are controlled by embedded software The Therac-25 accidents, 1985-1987, where patients in at least six cases were given massive overdoses of radiation The Ariane 5 Flight 501, 1996, where an incorrect number conversion caused the space ship to explode

Pictures: http://instantmedicalcare.com/images/galler/instantmedicalcare-image-gallery-xray-machine.jpg http://www.educnet.education.fr/orbito/lanc/enviro/images/explo501.jpg http://www.zeitgeistzephyr.com/wordpress/wp-content/uploads/2009/08/a_ariane5.gif Martin Toft Worst-Case Execution Time Analysis

slide-4
SLIDE 4

3 / 10

Introduction The challenges My research Embedded software System types WCET analysis

Software bugs are still a problem today

Martin Toft Worst-Case Execution Time Analysis

slide-5
SLIDE 5

4 / 10

Introduction The challenges My research Embedded software System types WCET analysis

Time-critical systems

Again, many safety-critical systems are time-critical Not enough that calculations are correct Must be carried out in a timely fashion There are strict deadlines For example: anti-lock brakes on a car, steering gear on an airplane Also known as “real-time systems”

Safety−critical systems embedded software Systems controlled by Time−critical systems

Martin Toft Worst-Case Execution Time Analysis

slide-6
SLIDE 6

5 / 10

Introduction The challenges My research Embedded software System types WCET analysis

Time-critical systems (cont.)

A real-time system consists of:

A number of processes A scheduler

The processes have a number of properties, e.g.:

Type (periodic/aperiodic) Period (if periodic) (T) Priority (P) Deadline Worst-case execution time (WCET)

Process P T WCET a 1 80 40 b 2 40 10 c 3 20 5

Time c 10 20 30 40 60 50 b a 70 80

Martin Toft Worst-Case Execution Time Analysis

slide-7
SLIDE 7

6 / 10

Introduction The challenges My research Embedded software System types WCET analysis

Worst-case execution time analysis

“Provide guarantees for proper timing behaviour by computing bounds for processes’ execution time on given hardware platforms” Measurement-based methods are unsafe All possible input values from the environment must be considered The way to go: analyse the processes’ code without executing it

  • WCET

possible execution times Time TIGHT SAFE UNSAFE upper bound Probability

Martin Toft Worst-Case Execution Time Analysis

slide-8
SLIDE 8

7 / 10

Introduction The challenges My research Modern hardware Modularity Ease of use

Modern hardware is complex

The WCET of a process depends on the hardware platform The effects of optimisation features in the platform, e.g. caching and pipelining, must be captured Caching: store frequently used data in a fast memory Pipelining: parallelise the steps involved in executing a process

Main memory Cache memory Processor core Cycle 1 2 3 4 5 Stage 1 Stage 2 Stage 3 Stage 4

  • Instr. 1
  • Instr. 2
  • Instr. 3
  • Instr. 4
  • Instr. 5
  • Instr. 1
  • Instr. 1
  • Instr. 1
  • Instr. 2
  • Instr. 2
  • Instr. 2
  • Instr. 3
  • Instr. 3
  • Instr. 4

Martin Toft Worst-Case Execution Time Analysis

slide-9
SLIDE 9

8 / 10

Introduction The challenges My research Modern hardware Modularity Ease of use

The approach must be modular

Substitution of sub-analyses in

  • rder to experiment with

different combinations Component-based model of the hardware platform Reuse of components Experimentation with components in order to reduce system costs

Picture: http://leax.dk/uploads/images/ETNA/puzzle.jpg Martin Toft Worst-Case Execution Time Analysis

slide-10
SLIDE 10

9 / 10

Introduction The challenges My research Modern hardware Modularity Ease of use

Ease of use

Many hardware platforms exist New platforms enter the market all the time It must be possible for non-computer-scientists to create support for new platforms Reuse of components due to modularity

Picture: http://www.sciencedaily.com/images/2009/02/090220102249-large.jpg Martin Toft Worst-Case Execution Time Analysis

slide-11
SLIDE 11

10 / 10

Introduction The challenges My research Overview

My research: WCET analysis using model checking and static analysis

Martin Toft Worst-Case Execution Time Analysis

slide-12
SLIDE 12

Thank you for your attention Questions?

These slides are available at http://martintoft.dk/slides/procom-course.pdf