Real-Time Workshop User's Guide

Real-Time Workshop User's Guide
Real-Time Workshop
®
For Use with Simulink ®
Modeling
Simulation
Implementation
User’s Guide
Version 5
How to Contact The MathWorks:
www.mathworks.com
comp.soft-sys.matlab
Web
Newsgroup
[email protected]
Technical support
Product enhancement suggestions
Bug reports
Documentation error reports
Order status, license renewals, passcodes
Sales, pricing, and general information
508-647-7000
Phone
508-647-7001
Fax
The MathWorks, Inc.
3 Apple Hill Drive
Natick, MA 01760-2098
Mail
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
For contact information about worldwide offices, see the MathWorks Web site.
Real-Time Workshop User’s Guide
 COPYRIGHT 1994 - 2002 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used
or copied only under the terms of the license agreement. No part of this manual may be photocopied or reproduced in any form without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation by
or for the federal government of the United States. By accepting delivery of the Program, the government
hereby agrees that this software qualifies as "commercial" computer software within the meaning of FAR
Part 12.212, DFARS Part 227.7202-1, DFARS Part 227.7202-3, DFARS Part 252.227-7013, and DFARS Part
252.227-7014. The terms and conditions of The MathWorks, Inc. Software License Agreement shall pertain
to the government’s use and disclosure of the Program and Documentation, and shall supersede any
conflicting contractual terms or conditions. If this license fails to meet the government’s minimum needs or
is inconsistent in any respect with federal procurement law, the government agrees to return the Program
and Documentation, unused, to MathWorks.
MATLAB, Simulink, Stateflow, Handle Graphics, and Real-Time Workshop are registered trademarks, and
TargetBox is a trademark of The MathWorks, Inc.
Other product or brand names are trademarks or registered trademarks of their respective holders.
Printing History: May 1994
January 1998
January 1999
September 2000
June 2001
October 2001
July 2002
First printing
Second printing
Third printing
Fourth printing
Online only
Online only
Online only
Version 1
Version 2.1
Version 3.11 (Release 11)
Version 4 (Release 12)
Updated for Version 4.1 (Release 12.1)
Updated for Version 4.2 (Release 13)
Updated for Version 5.0 (Release 13)
Contents
About This Guide
Understanding Real-Time Workshop
1
Product Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2
Some Real-Time Workshop Capabilities . . . . . . . . . . . . . . . . . . 1-3
Software Design with Real-Time Workshop . . . . . . . . . . . . . . . 1-3
The Rapid Prototyping Process . . . . . . . . . . . . . . . . . . . . . . . .
Key Aspects of Rapid Prototyping . . . . . . . . . . . . . . . . . . . . . . .
Rapid Prototyping for Digital Signal Processing . . . . . . . . . . . .
Rapid Prototyping for Control Systems . . . . . . . . . . . . . . . . . . .
1-5
1-5
1-8
1-9
Open Architecture of Real-Time Workshop . . . . . . . . . . . . . 1-11
Where to Find Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-14
How Do I... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-14
Code Generation and the Build Process
2
The Real-Time Workshop User Interface . . . . . . . . . . . . . . . . 2-2
Using the Real-Time Workshop Pane . . . . . . . . . . . . . . . . . . . . 2-2
Target Configuration Options . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
General Code Generation Options . . . . . . . . . . . . . . . . . . . . . . . 2-7
General Code Generation Options (cont.) . . . . . . . . . . . . . . . . 2-12
General Code Appearance Options . . . . . . . . . . . . . . . . . . . . . . 2-13
Target-specific Code Generation Options . . . . . . . . . . . . . . . . 2-16
i
TLC Debugging Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-18
Real-Time Workshop Submenu . . . . . . . . . . . . . . . . . . . . . . . . . 2-20
Simulation Parameters and Code Generation . . . . . . . . . . .
Solver Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Workspace I/O Options and Data Logging . . . . . . . . . . . . . . . .
Diagnostics Pane Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Advanced Options Pane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tracing Generated Code Back to Your
Simulink Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Other Interactions Between Simulink
and Real-Time Workshop . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2-21
2-21
2-22
2-25
2-26
2-33
2-35
Selecting a Target Configuration . . . . . . . . . . . . . . . . . . . . . . 2-40
The System Target File Browser . . . . . . . . . . . . . . . . . . . . . . . . 2-40
Available Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-41
Making an Executable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-47
Generated Source Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-47
Compilation and Linking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-49
Choosing and Configuring Your Compiler . . . . . . . . . . . . . . 2-51
Template Makefiles and Make Options . . . . . . . . . . . . . . . . . 2-54
Compiler-Specific Template Makefiles . . . . . . . . . . . . . . . . . . . 2-54
Template Makefile Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 2-58
Configuring the Generated Code via TLC . . . . . . . . . . . . . . 2-59
Target Language Compiler Variables and Options . . . . . . . . . 2-59
Generated Code Formats
3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2
Choosing a Code Format for Your Application . . . . . . . . . . . 3-3
ii
Contents
Real-Time Code Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Unsupported Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
System Target Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Template Makefiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3-6
3-6
3-6
3-6
Real-Time malloc Code Format . . . . . . . . . . . . . . . . . . . . . . . . .
Unsupported Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
System Target Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Template Makefiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3-8
3-8
3-8
3-9
S-Function Code Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-10
Embedded C Code Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-11
Building Subsystems
4
Nonvirtual Subsystem Code Generation . . . . . . . . . . . . . . . . . 4-2
Nonvirtual Subsystem Code Generation Options . . . . . . . . . . . 4-2
Modularity of Subsystem Code . . . . . . . . . . . . . . . . . . . . . . . . . 4-13
Code Reuse Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-13
Generating Code and Executables from Subsystems . . . . . 4-15
Working with Data Structures
5
Parameters: Storage, Interfacing, and Tuning . . . . . . . . . . . 5-2
Storage of Nontunable Parameters . . . . . . . . . . . . . . . . . . . . . . . 5-2
Tunable Parameter Storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-4
Storage Classes of Tunable Parameters . . . . . . . . . . . . . . . . . . . 5-5
Using the Model Parameter Configuration Dialog . . . . . . . . . . 5-8
Tunable Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-12
Tunability of Linear Block Parameters . . . . . . . . . . . . . . . . . . 5-14
iii
Parameter Configuration Quick Reference Diagram . . . . . . . . 5-16
Signals: Storage, Optimization, and Interfacing . . . . . . . . .
Signal Storage Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Signals with Auto Storage Class . . . . . . . . . . . . . . . . . . . . . . . .
Declaring Test Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interfacing Signals to External Code . . . . . . . . . . . . . . . . . . . .
Symbolic Naming Conventions for Signals
in Generated Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary of Signal Storage Class Options . . . . . . . . . . . . . . . .
C API for Parameter Tuning and Signal Monitoring . . . . . . . .
Target Language Compiler API for Parameter
Tuning and Signal Monitoring . . . . . . . . . . . . . . . . . . . . . . . . .
Parameter Tuning via MATLAB Commands . . . . . . . . . . . . . .
5-17
5-17
5-20
5-24
5-25
Simulink Data Objects and Code Generation . . . . . . . . . . .
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Parameter Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Parameter Object Configuration Quick
Reference Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Signal Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Signal Object Configuration Quick
Reference Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Resolving Conflicts in Configuration of Parameter
and Signal Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Customizing Code for Parameter and Signal Objects . . . . . . .
Using Objects to Export ASAP2 Files . . . . . . . . . . . . . . . . . . . .
5-32
5-32
5-34
Block States: Storing and Interfacing . . . . . . . . . . . . . . . . . .
Storage of Block States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Block State Storage Classes . . . . . . . . . . . . . . . . . . . . . . . . . . .
Using the State Properties Dialog Box to Interface
States to External Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Symbolic Names for Block States . . . . . . . . . . . . . . . . . . . . . . .
Block States and Simulink Signal Objects . . . . . . . . . . . . . . . .
Summary of State Storage Class Options . . . . . . . . . . . . . . . .
5-49
5-49
5-50
5-27
5-29
5-30
5-30
5-30
5-38
5-39
5-42
5-43
5-47
5-47
5-51
5-52
5-54
5-56
Storage Classes for Data Store Memory Blocks . . . . . . . . . 5-57
Data Store Memory and Simulink Signal Objects . . . . . . . . . . 5-59
iv
Contents
External Mode
6
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2
Using the External Mode User Interface . . . . . . . . . . . . . . . . 6-3
External Mode Related Menu and Toolbar Items . . . . . . . . . . . 6-3
External Mode Control Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-8
Connection and Start/Stop Controls . . . . . . . . . . . . . . . . . . . . . . 6-9
Target Interface Dialog Box . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-10
External Signal & Triggering Dialog Box . . . . . . . . . . . . . . . . . 6-11
Data Archiving Dialog Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-15
Parameter Download Options . . . . . . . . . . . . . . . . . . . . . . . . . . 6-18
External Mode Compatible Blocks and Subsystems . . . . . 6-19
Compatible Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-19
Signal Viewing Subsystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-19
External Mode Communications Overview . . . . . . . . . . . . . 6-23
The Download Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-23
Inlined and Tunable Parameters . . . . . . . . . . . . . . . . . . . . . . . 6-24
The TCP/IP Implementation . . . . . . . . . . . . . . . . . . . . . . . . . .
Using the TCP/IP Implementation . . . . . . . . . . . . . . . . . . . . . .
The External Interface MEX-File . . . . . . . . . . . . . . . . . . . . . . .
External Mode Compatible Targets . . . . . . . . . . . . . . . . . . . . .
Running the External Program . . . . . . . . . . . . . . . . . . . . . . . . .
Error Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Implementing an External Mode Protocol Layer . . . . . . . . . . .
6-26
6-26
6-28
6-29
6-29
6-32
6-32
Limitations of External Mode . . . . . . . . . . . . . . . . . . . . . . . . . 6-33
Program Architecture
7
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-2
v
Model Execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-4
Program Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-12
Program Execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-13
External Mode Communication . . . . . . . . . . . . . . . . . . . . . . . . 7-13
Data Logging In Singletasking
and Multitasking Model Execution . . . . . . . . . . . . . . . . . . . . . . 7-13
Rapid Prototyping and Embedded
Model Execution Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-14
Rapid Prototyping Model Functions . . . . . . . . . . . . . . . . . . . . . 7-15
Embedded Model Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-21
Rapid Prototyping Program Framework . . . . . . . . . . . . . . .
Rapid Prototyping Program Architecture . . . . . . . . . . . . . . . . .
Rapid Prototyping System-Dependent Components . . . . . . . .
Rapid Prototyping System-Independent Components . . . . . . .
Rapid Prototyping Application Components . . . . . . . . . . . . . .
7-23
7-24
7-25
7-26
7-29
Embedded Program Framework . . . . . . . . . . . . . . . . . . . . . . . 7-34
Models with Multiple Sample Rates
8
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2
Singletasking vs. Multitasking Environments . . . . . . . . . . . . 8-3
Executing Multitasking Models . . . . . . . . . . . . . . . . . . . . . . . . . 8-5
Multitasking and Pseudomultitasking . . . . . . . . . . . . . . . . . . . . 8-5
Building the Program for Multitasking Execution . . . . . . . . . . 8-8
Singletasking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-8
Building the Program for Singletasking Execution . . . . . . . . . . 8-9
Model Execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-9
Simulating Models with Simulink . . . . . . . . . . . . . . . . . . . . . . . 8-9
Executing Models in Real Time . . . . . . . . . . . . . . . . . . . . . . . . . 8-10
Singletasking vs. Multitasking Operation . . . . . . . . . . . . . . . . 8-11
Sample Rate Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-12
Data Transfer Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-13
vi
Contents
Rate Transition Block Options . . . . . . . . . . . . . . . . . . . . . . . . .
Faster to Slower Transitions in Simulink . . . . . . . . . . . . . . . .
Faster to Slower Transitions in Real Time . . . . . . . . . . . . . . . .
Slower to Faster Transitions in Simulink . . . . . . . . . . . . . . . .
Slower to Faster Transitions in Real Time . . . . . . . . . . . . . . . .
8-14
8-16
8-16
8-18
8-19
Singletasking and Multitasking
Execution of a Model: an Example . . . . . . . . . . . . . . . . . . . . . 8-22
Singletasking Execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-23
Multitasking Execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-26
Optimizing the Model for Code Generation
9
General Modeling Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 9-2
Expression Folding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3
Expression Folding Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3
Using and Configuring Expression Folding . . . . . . . . . . . . . . . . 9-5
Supporting Expression Folding in S-Functions . . . . . . . . . . . . 9-10
Categories of Output Expressions . . . . . . . . . . . . . . . . . . . . . . . 9-11
Acceptance or Denial of Requests for
Input Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-16
Utilizing Expression Folding in Your TLC Block
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-19
Conditional Branch Execution . . . . . . . . . . . . . . . . . . . . . . . . 9-25
Block Diagram Performance Tuning . . . . . . . . . . . . . . . . . . .
Look-Up Tables and Polynomials . . . . . . . . . . . . . . . . . . . . . . .
Accumulators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Use of Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9-26
9-26
9-36
9-38
Stateflow Optimizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-43
Simulation Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-44
vii
Compiler Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-46
The S-Function Target
10
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-2
Intellectual Property Protection . . . . . . . . . . . . . . . . . . . . . . . . 10-2
Creating an S-Function Block from a Subsystem . . . . . . . . 10-3
Sample Time Propagation in Generated S-Functions . . . . . . . 10-8
Choice of Solver Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-8
Tunable Parameters in Generated S-Functions . . . . . . . . . 10-9
Automated S-Function Generation . . . . . . . . . . . . . . . . . . . 10-11
Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-15
Limitations on Use of Goto and From Blocks . . . . . . . . . . . . . 10-15
Other Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-16
Unsupported Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-17
System Target File and Template Makefiles . . . . . . . . . . . 10-18
System Target File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-18
Template Makefiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-18
Real-Time Workshop Rapid Simulation Target
11
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-2
Licensing Protocols for Simulink Solvers in Executables . . . . 11-3
Building for the Rapid Simulation Target . . . . . . . . . . . . . . 11-5
Running a Rapid Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-6
viii Contents
Simulation Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-15
Batch and Monte Carlo Simulations . . . . . . . . . . . . . . . . . . . . 11-15
Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-16
Targeting Tornado for Real-Time Applications
12
The Tornado Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-2
Confirming Your Tornado Setup Is Operational . . . . . . . . . . . 12-2
VxWorks Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-3
Run-Time Architecture Overview . . . . . . . . . . . . . . . . . . . . . . 12-5
Parameter Tuning and Monitoring . . . . . . . . . . . . . . . . . . . . . . 12-5
Implementation Overview . . . . . . . . . . . . . . . . . . . . . . . . . . .
Adding Device Driver Blocks . . . . . . . . . . . . . . . . . . . . . . . . . .
Configuring the Template Makefile . . . . . . . . . . . . . . . . . . . .
Tool Locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Building the Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Downloading and Running the Executable
Interactively . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12-11
12-13
12-13
12-14
12-14
12-18
Asynchronous Support
13
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2
Interrupt Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-5
Interrupt Control Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-5
Task Synchronization Block . . . . . . . . . . . . . . . . . . . . . . . . . . 13-12
Asynchronous Rate Transition Block . . . . . . . . . . . . . . . . . . . 13-16
Unprotected Asynchronous Rate Transition Block . . . . . . . . 13-18
Creating a Customized Asynchronous Library . . . . . . . . . 13-21
ix
Targeting Real-Time Systems
14
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-2
Components of a Custom Target Configuration . . . . . . . . .
Code Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
User-Written Run-Time Interface Code . . . . . . . . . . . . . . . . . .
Run-Time Interface for Rapid Prototyping . . . . . . . . . . . . . . . .
Run-Time Interface for Embedded Targets . . . . . . . . . . . . . . .
Control Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14-3
14-3
14-4
14-5
14-5
14-6
Tutorial: Creating a Custom Target Configuration . . . . . . 14-9
Customizing the Build Process . . . . . . . . . . . . . . . . . . . . . . .
System Target File Structure . . . . . . . . . . . . . . . . . . . . . . . . .
Adding a Custom Target to the System Target
File Browser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Template Makefiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14-16
14-16
Creating Device Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Inlined and Noninlined Drivers . . . . . . . . . . . . . . . . . . . . . . .
Device Driver Requirements and Limitations . . . . . . . . . . . .
Parameterizing Your Driver . . . . . . . . . . . . . . . . . . . . . . . . . .
Writing a Noninlined S-Function Device Driver . . . . . . . . . .
Writing an Inlined S-Function Device Driver . . . . . . . . . . . .
Building the MEX-File and the Driver Block . . . . . . . . . . . . .
Source Code for Inlined ADC Driver . . . . . . . . . . . . . . . . . . . .
14-39
14-40
14-42
14-43
14-44
14-53
14-59
14-60
Interfacing Parameters and Signals . . . . . . . . . . . . . . . . . .
Signal Monitoring via Block Outputs . . . . . . . . . . . . . . . . . . .
C API for Parameter Tuning . . . . . . . . . . . . . . . . . . . . . . . . . .
Target Language Compiler API for
Signals and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14-70
14-70
14-77
14-27
14-28
14-92
Creating an External Mode Communication Channel . . . 14-94
The Design of External Mode . . . . . . . . . . . . . . . . . . . . . . . . . 14-94
External Mode Communications Overview . . . . . . . . . . . . . . 14-95
External Mode Source Files . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-97
Guidelines for Implementing the Transport Layer . . . . . . . 14-100
x
Contents
Combining Multiple Models . . . . . . . . . . . . . . . . . . . . . . . . . 14-103
DSP Processor Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-107
For DSP Blockset Users . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-107
Glossary
A
Blocks That Depend on Absolute Time
B
Targeting DOS for Real-Time Applications
C
DOS Target Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-2
DOS Device Drivers Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-2
Implementation Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-4
System Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-5
Sample Rate Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-7
Device Driver Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Device Driver Block Library . . . . . . . . . . . . . . . . . . . . . . . . . . .
Configuring Device Driver Blocks . . . . . . . . . . . . . . . . . . . . . . .
Adding Device Driver Blocks to the Model . . . . . . . . . . . . . . . .
C-10
C-10
C-11
C-17
Building the Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-18
Running the Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-19
xi
The Real-Time Workshop Development Process
D
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D-2
A Next-Generation Development Tool . . . . . . . . . . . . . . . . . . .
Key Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Integration with Simulink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
D-3
D-4
D-7
D-9
How MathWorks Tools Streamline Development . . . . . . . . D-12
Code Formats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D-18
Target Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D-21
Code Generation Optimizations . . . . . . . . . . . . . . . . . . . . . . . . D-28
An Open and Extensible Environment . . . . . . . . . . . . . . . . . D-33
xii
Contents
About This Guide
About This Guide
If you are just beginning to use Real-Time Workshop, please see the overviews,
explanations and tutorials in either the online or printed version of the Getting
Started Guide to orient yourself. The following material picks up from there,
gradually introducing additional details about code generation, targeting,
optimizations, and other useful topics:
Understanding Real-Time Workshop describes concepts and terminology of
the Real-Time Workshop. It describes the rapid prototyping process that the
open architecture of the Real-Time Workshop facilitates, and points to
discussions of basic real-time development tasks elsewhere in this document.
Code Generation and the Build Process describes the automatic program
building process in detail. It discusses all code generation options controlled by
the Real-Time Workshop’s graphical user interface. Topics include data
logging, inlining and tuning parameters, and template makefiles. The chapter
also summarizes available target configurations.
Generated Code Formats compares and contrasts targets and their
associated code formats. This include the real-time, real-time malloc,
embedded C, and S-function code formats.
Building Subsystems describes how to control code generation for
conditionally executed and atomic subsystems.
Working with Data Structures teaches you how to generate storage
declarations to import and export parameters and block states, configure
storage for signals and data objects, and utilize custom storage classes.
External Mode contains information about external mode, a simulation
environment that supports on-the-fly parameter tuning, signal monitoring,
and data logging.
Program Architecture discusses the architecture of programs generated by
the Real-Time Workshop, and the run-time interface.
Models with Multiple Sample Rates describes how to handle multirate
systems.
Optimizing the Model for Code Generation discusses techniques for
optimizing your generated programs.
The S-Function Target explains how to generate S-Function blocks from
models and subsystems. This enables you to encapsulate models and
subsystems and protect your designs by distributing only binaries.
xiv
Real-Time Workshop Rapid Simulation Target discusses the rapid
simulation target (RSIM), which executes your model in nonreal-time on your
host computer. Use this feature to generate fast, stand-alone simulations that
allow batch parameter tuning and the loading of new simulation data (signals)
from MATLAB MAT-files without needing to recompile your model.
Targeting Tornado for Real-Time Applications contains information that is
specific to developing programs that target Tornado, and signal monitoring
using StethoScope.
Asynchronous Support describes the Interrupt Template library, which allow
you to model synchronous/asynchronous event handling.
Targeting Real-Time Systems discusses advanced techniques for developing
programs for custom targets, including device driver blocks, customizing
system target files and template makefiles, combining multiple models into a
single executable, and APIs for external mode communication, signal
monitoring, and parameter tuning.
Appendix A is a glossary that contains definitions of terminology associated
with the Real-Time Workshop and real-time software development.
Appendix B lists blocks whose use is restricted due to dependency on absolute
time.
Appendix C details the DOS target (now obsolete) and provides useful
guidance for working with device drivers.
Appendix D provides an overview that describes how using the Real-Time
Workshop development environment can dramatically accelerate the design,
refinement and deployment of real-time systems on a variety of target systems.
xv
About This Guide
xvi
1
Understanding Real-Time
Workshop
We begin by summarizing what Real-Time Workshop can do and how you can use it to accelerate
development of high-quality real-time software. This is followed by an overview of the software
components that Real-Time Workshop calls upon to generate source code from a Simulink model, and
shows how they work together in an extensible way. Information resources are provided to help you
understand where to look to answer some commonly asked questions.
Product Overview (p. 1-2)
Real-Time Workshop at a glance
The Rapid Prototyping Process (p. 1-5)
Key advantages of rapid prototyping, along with
descriptions of its application in two domains
Open Architecture of Real-Time
Workshop (p. 1-11)
Modules and files involved in code generation that you
can customize for your own targets and applications
Where to Find Help (p. 1-14)
Pointers to both basic descriptions and advanced
information on specific topics
1
Understanding Real-Time Workshop
Product Overview
Real-Time Workshop® generates optimized, portable, and customizable ANSI
C code from Simulink models to create stand-alone implementations of models
that operate in real-time and non-real-time in a variety of target environments.
Generated code can run on PC hardware, DSPs, microcontrollers on bare-board
environments, and with commercial or proprietary real-time operating
systems (RTOS). Real-Time Workshop lets you speed up simulations, build in
intellectual property protection, and operate across a wide variety of real-time
rapid prototyping targets. Figure 1-1 illustrates the role of Real-Time
Workshop (shaded elements) in the software design process.
.
MATLAB
and
Toolboxes
Design
and
Analysis
Real-Time
Workshop
components
Simulink,
Stateflow, and Blocksets
Modeling and simulation
Rapid simulations
External Mode
Monitoring and
parameter tuning
Simulink Code
Generator
Generates C
Customer defined
Monitoring and
parameter tuning
Make
process
Rapid Prototyping Target
Real-time test environment
Early rapid prototyping iterations
Production
Target
Final production
iteration
Figure 1-1: Software Design and Deployment Using MATLAB and Simulink
1-2
Product Overview
Some Real-Time Workshop Capabilities
With Real-Time Workshop, you can quickly generate code for discrete-time,
continuous-time (fixed-step), and hybrid systems, as well as for finite state
machines modeled in Stateflow® using the optional Stateflow Coder. The
optional Real-Time Workshop Embedded Coder works with Real-Time
Workshop to generate efficient, embeddable source code.
Using integrated makefile-based targeting support, Real-Time Workshop
builds programs that can help speed up your simulations, provide intellectual
property protection, and run on a wide variety of real-time rapid prototyping or
production targets. Simulink's external mode run-time monitor works
seamlessly with real-time targets, providing an elegant signal monitoring and
parameter tuning interface. Real-Time Workshop supports continuous-time,
discrete-time and hybrid systems, including conditionally executed and atomic
systems. Real-Time Workshop accelerates your development cycle, producing
higher quality results in less time.
Real-Time Workshop is a key link in the set of system design tools provided by
The MathWorks, providing a real-time development environment — a direct
path from system design to hardware implementation. You can streamline
application development and reduce costs with Real-Time Workshop by testing
design iterations with real-time hardware. Real-Time Workshop supports the
execution of dynamic system models on hardware by automatically converting
models to code and providing model-based debugging support. It is well suited
for accelerating simulations, rapid prototyping, turnkey solutions, and
production embedded real-time applications.
Software Design with Real-Time Workshop
A typical product cycle using the MathWorks toolset starts with modeling in
Simulink, followed by an analysis of the simulations in MATLAB. During the
simulation process, you use the rapid simulation features of Real-Time
Workshop to speed up your simulations.
After you are satisfied with the simulation results, you use Real-Time
Workshop in conjunction with a rapid prototyping target, such as xPC Target.
The rapid prototyping target is connected to your physical system. You test and
observe your system, using your Simulink model as the interface to your
physical target. Once your simulation is functioning properly, you use
Real-Time Workshop to transform your model to C code. An extensible make
process and download procedure creates an executable for your model and
1-3
1
Understanding Real-Time Workshop
places it on the target system. Finally, using external mode, you can monitor
and tune parameters in real-time as your model executes on the target
environment.
There are two broad classes of targets: rapid prototyping targets and the
embedded target. Code generated for the rapid prototyping targets supports
increased monitoring and tuning capabilities. Code generated for embedded
targets is highly optimized and suitable for deployment in production systems,
and can include application-specific entry points to monitor signals and tune
parameters.
To support embedded targets, The MathWorks distributes Real-Time
Workshop Embedded Coder as a separate product. Embedded Coder is an
extension of Real-Time Workshop designed to generate C code for embedded
discrete-time systems, where efficiency, configurability, readability, and
traceability of the generated code are extremely important. Real-Time
Workshop Embedded Coder enhances Real-Time Workshop code generation
technology to generate embeddable ANSI C code that compares favorably with
hand-optimized code in terms of performance, ROM code size, RAM
requirements, and readability. The Real-Time Workshop Embedded Coder
documentation contains information about optimization specifically for
embedded code.
For a more complete general overview of the key features, capabilities, and
benefits of Real-Time Workshop, please see Appendix D, “The Real-Time
Workshop Development Process.”
1-4
The Rapid Prototyping Process
The Rapid Prototyping Process
Real Time Workshop supports rapid prototyping, an application development
process that allows you to
• Conceptualize solutions graphically in a block diagram modeling
environment
• Evaluate system performance early on — before laying out hardware, coding
production software, or committing to a fixed design
• Refine your design by rapid iteration between algorithm design and
prototyping
• Tune parameters while your real-time model runs, using Simulink in
external mode as a graphical front end
Key Aspects of Rapid Prototyping
The figure below contrasts the rapid prototyping development process with the
traditional development process.
Algorithm
development
Hardware and
software design
Implementation of
production system
Rapid Prototyping Process
Rapid Iteration
Manual Iteration
Traditional Approach
Algorithm design
and prototyping
Implementation of
production system
Figure 1-2: Traditional vs. Rapid Prototyping Development Processes
1-5
1
Understanding Real-Time Workshop
The traditional approach to real-time design and implementation typically
involves multiple teams of engineers, including an algorithm design team,
software design team, hardware design team, and an implementation team.
When the algorithm design team has completed its specifications, the software
design team implements the algorithm in a simulation environment and then
specifies the hardware requirements. The hardware design team then creates
the production hardware. Finally, the implementation team integrates the
hardware into the larger overall system.
This traditional development process takes so much time because algorithm
designers often do not have access to the hardware that is actually deployed.
The rapid prototyping process combines the algorithm, software, and hardware
design phases, eliminating potential bottlenecks by allowing engineers to see
results and rapidly iterate solutions before building expensive hardware.
Automating Programming
Automatic program building allows you to make design changes directly to the
block diagram, puttting algorithm development (including coding, compiling,
linking, and downloading to target hardware) under control of a single process:
• Design a Model in Simulink
You begin the rapid prototyping process with the development of a model in
Simulink. In control engineering, you model plant dynamics and other
dynamic components that constitute a controller and/or an observer.
• Simulate your Model in Simulink
You use MATLAB, Simulink, and toolboxes to aid in the development of
algorithms and analysis of the results. If the results are not satisfactory, you
can iterate the modeling and analysis process until results are acceptable.
• Generate Source Code with Real-Time Workshop
Once simluation results are acceptable, you generate downloadable C code
that implements the appropriate portions of the model. You can use
Simulink in external mode to tune parameters and further refine your
model, quickly iterating through solutions.
• Implement a Production Prototype
At this stage, the rapid prototyping process is complete. You can begin the
final implementation for production with confidence that the underlying
algorithms work properly in your real-time production system.
1-6
The Rapid Prototyping Process
The next diagram illustrates the flow of this process.
Algorithm Design and Prototyping
Identify system
and/ or algorithm
requirements
Build/edit model in
Simulink
Run simulations and analyze results
using Simulink and MATLAB
Are
No
results
OK?
Yes
Invoke the Real-Time Workshop build procedure,
download and run on your target hardware
Analyze results and tune the model
using external mode
Are
results
OK?
No
Yes
Implement production system
Figure 1-3: The Rapid Prototyping Development Process
1-7
1
Understanding Real-Time Workshop
Highly productive development cycles are possible due to the integration of
Real-Time Workshop, MATLAB, and Simulink. Each component adds value to
your application design process:
• MATLAB: Provides design, analysis, and data visualization tools.
• Simulink: Provides system modeling, simulation, and validation.
• Real-Time Workshop: Generates C code from Simulink model; provides
framework for running generated code in real-time, tuning parameters, and
viewing real-time data.
Rapid Prototyping for Digital Signal Processing
The first step in the rapid prototyping process for digital signal processing is to
consider the kind and quality of the data to be worked on, and to relate it to the
system requirements. Typically this includes examining the signal-to-noise
ratio, distortion, and other characteristics of the incoming signal, and relating
them to algorithm and design choices.
System Simulation and Algorithm Design
In the rapid prototyping process, the block diagram plays two roles in
algorithm development. The block diagram helps to identify processing
bottlenecks, and to optimize the algorithm or system architecture. The block
diagram also functions as a high-level system description. That is, the diagram
provides a hierarchical framework for evaluating the behavior and accuracy of
alternative algorithms under a range of operating conditions.
Analyzing Results, Tuning Parameters, and Monitoring Signals
After creating an algorithm (or a set of candidate algorithms), the next stage is
to consider architectural and implementation issues. These include complexity,
speed, and accuracy. In a conventional development environment, this would
mean running the algorithm and recoding it in C or in a hardware design and
simulation package.
Simulink external mode allows you to change parameters interactively, while
your signal processing algorithms execute in real time on the target hardware.
After building the executable and downloading it to your hardware, you tune
(modify) block parameters in Simulink. Simulink automatically downloads the
new values to the hardware. You can monitor the effects of your parameter
changes by simply connecting Scope blocks to signals that you want to observe.
1-8
The Rapid Prototyping Process
Rapid Prototyping for Control Systems
Rapid prototyping for control systems is similar to digital signal processing,
with one major difference. In control systems design, you must model your
plant prior to developing algorithms in order to simulate closed-loop
performance. Once your plant model is sufficiently accurate, the rapid
prototyping process for control system design continues in much the same
manner as digital signal processing design.
Rapid prototyping begins with developing block diagram plant models of
sufficient fidelity for preliminary system design and simulation. Once
simulations indicate acceptable system performance levels, the controller block
diagram is separated from the plant model and I/O device driver blocks are
attached to it. Automatic code generation immediately converts the entire
system to real-time executable code, which can be automatically loaded onto
target hardware.
Modeling Systems in Simulink
The first step in the design process is development of a plant model. The
Simulink collection of linear and nonlinear components helps you to build
models involving plant, sensor, and actuator dynamics. Because Simulink is
customizable, you can further simplify modeling by creating custom blocks and
block libraries from continuous- and discrete-time components.
Using the System Identification Toolbox, you can analyze test data to develop
an empirical plant model; or you can use the Symbolic Math Toolbox to
translate the equations of the plant dynamics into state-variable form.
Analyzing Simulation Results
You can use MATLAB and Simulink to analyze the results produced from a
model developed in the first step of the rapid prototyping process. At this stage,
you can design and add a controller to your plant.
Deriving and Analyzing Algorithms
From the block diagrams developed during the modeling stage, you can extract
state-space models through linearization techniques. These matrices can be
used in control system design. You can use the following toolboxes to facilitate
control system design, and work with the matrices that you derived:
• Control System Toolbox
1-9
1
Understanding Real-Time Workshop
• LMI Control Toolbox
• Model Predictive Control Toolbox
• Robust Control Toolbox
• System Identification Toolbox
• SimMechanics
Once you have your controller designed, you can create a closed-loop system by
connecting it to the Simulink plant model. Closed-loop simulations allow you
to determine how well the initial design meets performance requirements.
Once you have a satisfactory model, it is a simple matter to generate C code
directly from the Simulink block diagram, compile it for the target processor,
and link it with supplied or user-written application modules.
Analyzing Results, Tuning Parameters, and Monitoring Signals
You can load output data from your program into MATLAB for analysis, or
display the data with third party monitoring tools. You can easily make design
changes to the Simulink model and then regenerate the C code.
1-10
Open Architecture of Real-Time Workshop
Open Architecture of Real-Time Workshop
Real-Time Workshop is an open system designed for use with a wide variety of
operating environments and hardware types. Figure 1-4 shows how you can
extend key elements of Real-Time Workshop.
You can configure the Real-Time Workshop program generation process to
your own needs by modifying the following components:
• Simulink and the model file (model.mdl)
Simulink provides a very high-level language (VHLL) development
environment. The language elements are blocks and subsystems that
visually embody your algorithms. You can think of Real-Time Workshop as
a compiler that processes a VHLL source program (model.mdl), and emits
code suitable for a traditional high-level language (HLL) compiler.
S-functions written in C let you extend the Simulink VHLL by adding new
general purpose blocks, or incorporating legacy code into a block.
• The intermediate model description (model.rtw)
The initial stage of the code generation process is to analyze the source
model. The resultant description file contains a hierarchical structure of
records describing systems and blocks and their connections.
The S-function API includes a special function, mdlRTW, that lets you
customize the code generation process by inserting parameter data from
your own blocks into the model.rtw file.
• The Target Language Compiler (TLC) program
The Target Language Compiler interprets a program that reads the
intermediate model description and generates code that implements the
model as a program.
You can customize the elements of the TLC program in two ways. First, you
can implement your own system target file, which controls overall code
generation parameters. Second, you can implement block target files, which
control how code is generated from individual blocks such as your own
S-function blocks.
1-11
1
Understanding Real-Time Workshop
C code S-functions
Simulink
MATLAB
model.mdl
Real-Time Workshop
Real-Time Workshop build
system.tmf
model.rtw
Target Language
Compiler (TLC) program:
• System target file
Target
Language
Compiler
• Block target files
model.c
model.h
model_private.h
• Function library
Run-time interface
support files
model.mk
make
model.exe
Download to target hardware
Start execution using Simulink external mode
Figure 1-4: Real-Time Workshop Architecture
1-12
Open Architecture of Real-Time Workshop
• Source code generated from the model; for descriptions of these files, see
“Summary of Files Created by the Build Procedure” in the Real-Time
Workshop Getting Started Guide.
There are several ways to customize generated code, or interface it to custom
code:
- Exported entry points let you interface your hand-written code to the
generated code. This makes it possible to develop your own timing and
execution engine, or to combine code generated from several models into a
single executable.
- You can automatically make signals, parameters, and other data
structures within generated code visible to your own code, facilitating
parameter tuning and signal monitoring.
- Prepare or modify Target Language Compiler script files to customize the
transformation of Simulink blocks into source code. See the Target
Language Compiler Reference Guide for further details.
• Run-time interface support files
The run-time interface consists of code interfacing to the generated model
code. You can create a custom set of run-time interface files, including:
- A harness (main) program
- Code to implement a custom external mode communication protocol
- Code that interfaces to parameters and signals defined in the generated
code
- Timer and other interrupt service routines
- Hardware I/O drivers
• The template makefile and model.mk
A makefile, model.mk, controls the compilation and linking of generated
code. Real-Time Workshop generates model.mk from a template makefile
during the code generation and build process. You can create a custom
template makefile to control compiler options and other variables of the make
process.
All of these components contribute to the process of transforming a Simulink
model into an executable program. The topics in the next section point you to
documentation describing each of them.
1-13
1
Understanding Real-Time Workshop
Where to Find Help
Documentation for Real-Time Workshop and related products from The
MathWorks covers many topics—some in considerable depth—and includes
many examples of use. Some of the major topics covered are summarized below,
enabling you to locate directly what you need to proceed.
If you are a less experienced user, you will benefit from reading the Getting
Started guide, which introduces the product and describes its capabilities,
applications, benefits, and general usage. Inside that guide are tutorials that
provide immediate hands-on experience to get you familiar with the look, feel,
and capabilities of Real-Time Workshop. That guide also discusses
• The role of Real-Time Workshop in your development cycle
• Basic real-time system concepts and terms
• General and platform-specific installation instructions
• Related product descriptions
• Simulink demos that illustrate code generation
How Do I...
If you need specific details about how to use Real-Time Workshop, scan the
topics and descriptions below to locate documentation relevant to your
development tasks and interests. You can also search the index to find
information not included in this list.
Operate the Real-Time Workshop User Interface
You control most aspects of code generation through the Real-Time Workshop
tab of the Simulation Parameters dialog, and the dialogs descending from it.
See “The Real-Time Workshop User Interface” on page 2-2 for full descriptions
of the options at your disposal.
Select Targets and Customize Compilation
Setting up targets for code generation is simple with the Target File Browser,
described in “Selecting a Target Configuration” on page 2-40. Look there also
for information on configuring compilers (“Choosing and Configuring Your
Compiler” on page 2-51) and modifying makefiles (“Template Makefiles and
Make Options” on page 2–54). For details on working with specific targets, see
“The S-Function Target” on page 10-1, “Real-Time Workshop Rapid Simulation
1-14
Where to Find Help
Target” on page 11-1, “Targeting Tornado for Real-Time Applications” on
page 12-1, Appendix C, “Targeting DOS for Real-Time Applications,” and the
Real-Time Workshop Embedded Coder documentation.
Generate Single- and Multitasking Code
Real-Time Workshop fully supports singletasking and multitasking code
generation. See See “Program Architecture” on page 7-1 and See “Models with
Multiple Sample Rates” on page 8-1 for a complete description.
Customize Generated Code
Real-Time Workshop supports customization of the generated code.
The principle approach to customizing generated code is to modify Target
Language Compiler (TLC) files. The Target Language Compiler is an
interpreted language that translates Simulink models into C code. Using the
Target Language Compiler, you can direct the code generation process.
There are two TLC files, hookslib.tlc and cachelib.tlc, that contain
functions you can use to customize Real-Time Workshop generated code. See
the Target Language Compiler documentation for details on these TLC files.
See also the source code, located in matlabroot/rtw/c/tlc/lib/cachelib.tlc
and matlabroot/rtw/c/tlc/mw/hookslib.tlc.
Optimize Generated Code
The default code generation settings are generic for flexible rapid prototyping
systems. The penalty for this flexibility is code that is less than optimal. There
are several optimization techniques that you can use to minimize the source
code size and memory usage once you have a model that meets your
requirements.
See “Code Generation and the Build Process” on page 2–1 and “Optimizing the
Model for Code Generation” on page 9-1 for details on code optimization
techniques available for all target configurations.
The Real-Time Workshop Embedded Coder documentation contains
information about optimization specifically for embedded code.
Make Subsystem Code Reuseable
If your models contain multiple references to the same atomic subsystem, you
can ask Real-Time Workshop to generate a single reentrant function to
1-15
1
Understanding Real-Time Workshop
represent the subsystem, rather than inlining it or generating multiple
functions that all do the same thing. “Building Subsystems” on page 4-1 tells
how to do this, and describes model characteristics that can limit or prevent
subsystem reuse.
Validate Generated Code
Using Real-Time Workshop data logging features, you can create an executable
that runs on your workstation and creates a data file. You can then compare
the results of your program with the results of running an equivalent Simulink
simulation.
For more information on how to validate Real-Time Workshop generated code,
see “Workspace I/O Options and Data Logging” on page 2-22. See also “Tutorial
2: Data Logging” on page 3-15 and “Tutorial 3: Code Validation” on page 3-19
of the Real-Time Workshop Getting Started Guide.
Incorporate Generated Code into Larger Systems
If your Real-Time Workshop generated code is intended to function within an
existing code base (for example, if you want to use the generated code as a
plug-in function), you should use Real-Time Workshop Embedded Coder. The
Real-Time Workshop Embedded Coder documentation describes the entry
points and header files you will need to interface your code to Real-Time
Workshop Embedded Coder generated code.
Incorporate Existing Code into Generated Code
To interface your hand-written code with Real-Time Workshop generated code,
you can use an S-function wrapper. See the Simulink Writing S-Functions
documentation and the Target Language Compiler documentation for more
information.
Create and Communicate with Device Drivers
S-functions provide a flexible method for communicating with device drivers.
See “Targeting Real-Time Systems” on page 14–1 for a description of how to
build device drivers. Also, for a complete discussion of S-functions, see the
Simulink Writing S-Functions documentation.
Trace Code back to Blocks
Real-Time Workshop includes special tags throughout the generated code that
make it easy to trace generated code back to your Simulink model. See “Tracing
1-16
Where to Find Help
Generated Code Back to Your Simulink Model” on page 2-33 of the Getting
Started Guide for more information about this feature.
Automate Builds
Using Real-Time Workshop, you can generate code with the push of a button.
The automatic build procedure, initiated by a single mouse click, generates
code, a makefile, and optionally compiles (or cross-compiles) and downloads a
program. See “Automatic Program Building” on page 2-2 of the Getting Started
guide for an overview, and “Code Generation and the Build Process” on
page 2-1 for complete details.
Tune Parameters During Execution
Parameter tuning enables you to change block parameters while a generated
program runs, thus avoiding recompiling the generated code. Real-Time
Workshop supports parameter tuning in four different environments:
• External mode: You can tune parameters from Simulink while running the
generated code on a target processor. See “External Mode” on page 6–1 for
information on this mode.
• External C application program interface (API): You can write your own C
API interface for parameter tuning using support files provided by The
MathWorks. See “Targeting Real-Time Systems” on page 14-1 for more
information.
• Rapid simulation: You can use the Rapid Simulation Target (rsim) in batch
mode to provide fast simulations for performing parametric studies.
Although this is not an on-the-fly application of parameter tuning, it is
nevertheless a useful way to evaluate a model. This mode is also useful for
Monte Carlo simulation. See “Real-Time Workshop Rapid Simulation
Target” on page 11-1 for further information.
• Simulink: Prior to generating real-time code, you can tune parameters
on-the-fly in your Simulink model.
See also “Interface with Signals and Parameters” on page 1-18.
Monitor Signals and Log Data
There are several ways to monitor signals and data in Real-Time Workshop:
1-17
1
Understanding Real-Time Workshop
• External mode: You can monitor and log signals from an externally
executing program via Scope blocks and several other types of external mode
compatible blocks. See “External Signal & Triggering Dialog Box” on
page 6-11 for a discussion of this method.
• External C application program interface (API): You can write your own C
API for signal monitoring using support files provided by The MathWorks.
See “Targeting Real-Time Systems” on page 14-1 for more information.
• MAT-file logging: You can use a MAT-file to log data from the generated
executable. See “Workspace I/O Options and Data Logging” on page 2-22 for
more information.
• Simulink: You can use any of the Simulink data logging capabilities.
Interface with Signals and Parameters
You can interface signals and parameters in your model to hand-written code
by specifying the storage declarations of signals and parameters. For more
information, see
• “Parameters: Storage, Interfacing, and Tuning” on page 5-2
• “Signals: Storage, Optimization, and Interfacing” on page 5-17
• “Interfacing Signals to External Code” on page 5-25
Learn from Sample Implementations
Real-Time Workshop provides sample implementations that illustrate the
development of real-time programs under DOS and Tornado, as well as generic
real-time programs under Windows and UNIX.
These sample implementations are located in the following directories:
• matlabroot/rtw/c/grt: Generic real-time examples
• matlabroot/rtw/c/dos: DOS examples
• matlabroot/rtw/c/tornado: Tornado examples
1-18
2
Code Generation and the
Build Process
This chapter continues the discussion of code generation and the build process, previously introduced
in Chapter 1, “Understanding Real-Time Workshop.” First we present the details of the Real-Time
Workshop user interface. The sections that follow concern the code generation phase of the build
process.
The Real-Time Workshop User
Interface (p. 2-2)
The features that you control via the Real-Time
Workshop tab of the Simulation Parameters dialog
Simulation Parameters and Code
Generation (p. 2-21)
Describes how options on the Simulink Solver,
Workspace I/O, Diagnostics, and Advanced panes interact
with code generation, and how to trace code back to the
blocks that generated it
Selecting a Target Configuration
(p. 2-40)
Describes how to use the System Target File Browser,
with summaries of target configurations that you can
access through the browser
Making an Executable (p. 2-47)
How to control generation of executables during the build
process
Choosing and Configuring Your
Compiler (p. 2-51)
Aspects of installing a compiler and choosing appropriate
template makefiles
Template Makefiles and Make Options Summarizes available template makefiles and make
command options
(p. 2-54)
Configuring the Generated Code via
TLC (p. 2-59)
Using the Target Language Compiler to generate source
code in specific ways or to possess specific characteristics
2
Code Generation and the Build Process
The Real-Time Workshop User Interface
Many parameters and options affect the way that Real-Time Workshop
generates code from your model and builds an executable. To set these
parameters and options, you interact with the panes of the Simulation
Parameters dialog box.
The Simulink Solver, Workspace I/O, Diagnostics, and Advanced panes affect
both the behavior of the model in simulation, and the code generated from the
model. “Simulation Parameters and Code Generation” on page 2-21 discusses
how Simulink settings affect the code generation process.
The Real-Time Workshop pane lets you set parameters that directly affect code
generation and optimization. You also initiate and control the build process
from the Real-Time Workshop pane.
Using the Real-Time Workshop Pane
There are two ways to open the Real-Time Workshop pane:
• From the Simulation menu, choose Simulation Parameters. When the
Simulation Parameters dialog box opens, click on the Real-Time
Workshop tab.
• Alternatively, select Options from the Real-Time Workshop submenu of
the Tools menu in the Simulink window.
The Real-Time Workshop pane is divided into two sections. The upper section
contains the Category menu and the Build button.
Category Menu
The Category menu lets you select and work with various groups of options
and controls. The currently-selected group of options is displayed in the lower
section of the pane. Figure 2-1 shows the Category menu in the Real-Time
Workshop pane.
2-2
The Real-Time Workshop User Interface
Category menu selects groups of code
generation options and controls.
Build button initiates code generation and
build process.
Figure 2-1: Category Menu and Build Button in Real-Time Workshop Pane
The categories of options available from the Category menu are:
• Target configuration: High-level options related to control of the code
generation and build process and selection of control files.
• TLC debugging: Target Language Compiler debugging and execution
profiling options.
• General code generation options: Code generation settings that are
common to all target configurations.
• General code appearance options: Code and identifier formatting settings
that are common to all target configurations.
• Target-specific code generation options: One or more groups of options
that are specific to the selected target configuration.
Build Button
Click on the Build button to initiate the code generation and build process.
The following methods of initiating a build are exactly equivalent to clicking
the Build button:
• Select Build Model from the Real-Time Workshop submenu of the Tools
menu in the Simulink window (or use the key sequence Ctrl+B).
2-3
2
Code Generation and the Build Process
• Invoke the rtwbuild command from the MATLAB command line. The syntax
of the rtwbuild command is
rtwbuild modelname
or
rtwbuild('modelname')
where modelname is the name of the source model. If the source model is not
loaded into Simulink, rtwbuild loads the model.
Note When Generate code only is selected on the Target Configuration
portion of the Real-Time Workshop pane, the Build button’s name changes to
Generate code.
Getting Context-sensitive Help with ToolTips
The Real-Time Workshop pane supports “ToolTip” online help. Place your
cursor over any edit field name or check box to display a message box that
briefly explains the option.
The following sections summarize each category of options or parameters
controlled by the Real-Time Workshop pane, with references to subsequent
sections that give details on each option or parameter.
2-4
The Real-Time Workshop User Interface
Target Configuration Options
Figure 2-2 shows the Target configuration options of the Real-Time
Workshop pane.
Name of your model
Target configuration category shows
current configuration of system target file,
template makefile, and make command for
your desired target.
Browse button opens System Target File
Browser for selection of a target
configuration.
System target file name is
displayed or entered here.
Specify TLC options after
filename.
Make command name is
displayed or entered here.
Specify make options after
make command name.
Figure 2-2: The Real-Time Workshop Pane: Target Configuration Options
Browse Button
The Browse button opens the System Target File Browser (See Figure 2-8 on
page 2-41). The browser lets you select a preset target configuration consisting
of a system target file, template makefile, and make command.
“Selecting a Target Configuration” on page 2-40 details the use of the browser
and includes a complete list of available target configurations.
System Target File Field
The System target file field has these functions:
2-5
2
Code Generation and the Build Process
• If you have selected a target configuration using the System Target File
Browser, this field displays the name of the chosen system target file
(target.tlc).
• If you are using a target configuration that does not appear in the System
Target File Browser, you must enter the name of the desired system target
file in this field.
• After the system target filename, you can enter code generation options and
variables for the Target Language Compiler. See “Target Language
Compiler Variables and Options” on page 2-59 for details.
Template Makefile Field
The Template makefile field has these functions:
• If you have selected a target configuration using the System Target File
Browser, this field displays the name of an M-file that selects an appropriate
template makefile for your development environment. For example, in
Figure 2-2, the Template makefile field displays grt_default_tmf,
indicating that the build process will invoke grt_default_tmf.m.
“Template Makefiles and Make Options” on page 2-54 gives a detailed
description of the logic by which Real-Time Workshop selects a template
makefile.
• Alternatively, you can explicitly enter the name of a specific template
makefile (including the extension) in this field. You must do this if you are
using a target configuration that does not appear in the System Target File
Browser. This is necessary if you have written your own template makefile
for a custom target environment.
If you specify your own template makefile, be careful to include the filename
extension. If a filename extension is not included in the Template makefile
field, Real-Time Workshop attempts to find and execute a file with the
extension .m (i.e., an M-file).
Make Command Field
A high-level M-file command, invoked when a build is initiated, controls the
Real-Time Workshop build process. Each target has an associated make
command. The Make command field displays this command.
2-6
The Real-Time Workshop User Interface
Almost all targets use the default command, make_rtw. “Targets Available from
the System Target File Browser” on page 2-42 lists the make command
associated with each target.
Third-party targets may supply another make command. See the vendor’s
documentation.
In addition to the name of the make command, you can supply arguments in the
Make command field. These arguments include compiler-specific options,
include paths, and other parameters. When the build process invokes the make
utility, these arguments are passed along in the make command line.
“Template Makefiles and Make Options” on page 2-54 lists the Make
command arguments you can use with each supported compiler.
Generate Code Only Option
When this option is selected, the build process generates code but does not
invoke the make command. The code is not compiled and an executable is not
built.
When this option is selected, the caption of the Build button changes to
Generate code.
Stateflow Options Button
If the model contains any Stateflow blocks, this button will launch the
Stateflow Options dialog box. Refer to the Stateflow documentation for
information.
General Code Generation Options
The general code generation options are common to all target configurations.
These options are organized into two groups, selected from the Category
menu, as shown in Figure 2-3 and Figure 2-4.
2-7
2
Code Generation and the Build Process
Figure 2-3: General Code Generation Options
Figure 2-4: General Code Generation Options (cont.)
Show Eliminated Statements Option
If this option is selected, statements that were eliminated as the result of
optimizations (such as parameter inlining) appear as comments in the
generated code. The default is not to include eliminated statements.
2-8
The Real-Time Workshop User Interface
Loop Rolling Threshold Field
The loop rolling threshold determines when a wide signal or parameter should
be wrapped into a for-loop and when it should be generated as a separate
statement for each element of the signal. The default threshold value is 5.
For example, consider the model below:
The gain parameter of the Gain block is the vector myGainVec.
Assume that the loop rolling threshold value is set to the default, 5.
If myGainVec is declared as
myGainVec = [1:10];
an array of 10 elements, rtP.Gain_Gain[] is declared within the Parameters
data structure, rtP. The size of the gain array exceeds the loop rolling
threshold. Therefore the code generated for the Gain block iterates over the
array in a for loop, as shown in the following code fragment:
/* Gain: '<Root>/Gain'
*
* Regarding '<Root>/Gain':
*
Gain value: myGainVec
*/
{
int_T i1;
2-9
2
Code Generation and the Build Process
real_T *y0 = &rtB.Gain[0];
const real_T *p_Gain_Gain = &rtP.Gain_Gain[0];
for (i1=0; i1 < 10; i1++) {
y0[i1] = rtb_foo * p_Gain_Gain[i1];
}
}
If myGainVec is declared as
myGainVec = [1:3];
an array of three elements, rtP.Gain_Gain[] is declared within the
Parameters data structure, rtP. The size of the gain array is below the loop
rolling threshold. The generated code consists of inline references to each
element of the array, as in the code fragment below.
rtB.Gain[0] = rtb_foo * (rtP.Gain_Gain[0]);
rtB.Gain[1] = rtb_foo * (rtP.Gain_Gain[1]);
rtB.Gain[2] = rtb_foo * (rtP.Gain_Gain[2]);
See the Target Language Compiler Reference Guide for more information on
loop rolling.
Verbose Builds Option
If this option is selected, the MATLAB command window displays progress
information during code generation; compiler output is also made visible.
Generate HTML Report Option
If this option is selected, Real-Time Workshop produces a code generation
report in HTML format and automatically opens it for viewing in the MATLAB
Help browser. The contents of the report vary from one target to another, but
all reports contain the following code generation details:
• The Summary section lists version and date information, TLC options used
in code generation, and Simulink model settings.
• The Generated Source Files section contains a table of source code files
generated from your model. You can view the source code in the MATLAB
Help browser. Hyperlinks within the displayed source code let you view the
blocks or subsystems from which the code was generated. Click on the
2-10
The Real-Time Workshop User Interface
hyperlinks to view the relevant blocks or subsystems in a Simulink model
window.
The Real-Time Workshop Embedded Coder code generation report produces
additional information, such as suggestions for code generation options, to help
you optimize what is output. For further information see the Real-Time
Workshop Embedded Coder documentation.
Inline Invariant Signals Option
An invariant signal is a block output signal that does not change during
Simulink simulation. For example, the signal S3 in this block diagram is an
invariant signal.
Note The Inline invariant signals option is unavailable unless the Inline
parameters option (on the Advanced pane) is selected.
Given the model above, if both Inline parameters and Inline invariant
signals are selected, Real-Time Workshop inlines the invariant signal S3 in the
generated code.
Note that an invariant signal is not the same as an invariant constant. (See the
Using Simulink manual for information on invariant constants.) In the above
example, the two constants (1 and 2) and the gain value of 3 are invariant
constants. To inline these invariant constants, select Inline parameters.
2-11
2
Code Generation and the Build Process
Local Block Outputs Option
When this option is selected, block signals will be declared locally in functions
instead of being declared globally (when possible).
Note This check box is disabled when the Signal storage reuse item on the
Advanced pane is turned off.
For further information on the use of the Local block outputs option, see
“Signals: Storage, Optimization, and Interfacing” on page 5-17. Also go
through “Tutorial 4: A First Look at Generated Code” on page 3-23 of the
Getting Started guide if you have not done so already.
Force Generation of Parameter Comments Option
The Force generation of parameter comments option controls the generation
of comments in the model parameter structure declaration in model_prm.h.
Parameter comments indicate parameter variable names and the names of
source blocks.
When this option is off (the default), parameter comments are generated if less
than 1000 parameters are declared. This reduces the size of the generated file
for models with a large number of parameters.
When this option is on, parameter comments are generated regardless of the
number of parameters.
General Code Generation Options (cont.)
Buffer Reuse Option
When the Buffer reuse option is on (the default) Real-Time Workshop reuses
signal memory whenever possible. When Buffer reuse is off, signals are
stored in unique locations.
Note that the Buffer reuse option is enabled only when the Signal storage
reuse option on the Advanced pane of the Simulation Parameters dialog box
is selected.
2-12
The Real-Time Workshop User Interface
See “Signals: Storage, Optimization, and Interfacing” on page 5-17 for further
information (including generated code example) on Buffer reuse and other
signal storage options.
Expression Folding Options
Expression folding is a code optimization technique that can dramatically
improve the efficiency of generated code by minimizing the computation of
intermediate results and the use of temporary buffers or variables.
Expression folding is enabled by default. We strongly recommended that you
use this option. See “Expression Folding” on page 9-3 for full details on this
feature and related options that you can control from the General code
generation options (cont.) pane.
General Code Appearance Options
The General code appearance options control formatting of source code and
construction of identifiers. This interface is shown below.
Maximium Identifier Length Option
The Maximium identifier length field allows you to limit the number of
characters in function, typedef, and variable names. The default is 31
characters, but Real-Time Workshop imposes no upper limit.You may choose
2-13
2
Code Generation and the Build Process
to increase this length for models with deep hierarchical structure, as well as
when exercising some of the mnemonic identifier options described below.
Include Data Type Acronym in Identifier Option
Selecting Include data type acronym in identifier enables you to prepend
acronyms such as i32 (for long integers) to signal and work vector identifiers
to make code more readable. The default is not to include datatype acronyms
in identifiers. For example, with this option selected, Real-Time Workshop
identifies a scalar double signal from a discrete pulse generator as follows:
{
/* local block i/o variables */
real_T rtb_r64_A_Pulse;
.
.
.
rtY.Out1 = (rtP.A_Gain_Gain * rtb_r64_A_Pulse);
}
Include System Hierarchy Number in Identifiers Option
When this option is selected, Real-Time Workshop inserts identification tags in
the generated code (in addition to tags included in comments). The tags are
designed to help you identify the nesting level, within your source model, of the
block that generated a given line of code.
When this option is ON, the tag format is either
• The string root_ for root-level blocks; or
• The string sN_ where N is a unique system number assigned by Simulink, for
blocks at the subsystem level.
By default, Include system hierarchy number in identifiers is OFF, in order
to generate more compact code.
As an example, consider hier.mdl, the model in this picture.
2-14
The Real-Time Workshop User Interface
The subsystem within hier.mdl is shown in the picture below.
With Include system hierarchy number in identifiers on, the following code
is generated for the Out1 block of hier.mdl. The code includes the tag s1_ in
the symbols generated for the subsystem, and the tag root_ in the symbol
generated for the root-level Out1 block.
/* Outport: <Root>/Out1 incorporates:
*
Gain: <S1>/A_Gain
*
* Regarding <S1>/A_Gain:
*
Gain value: hier_P.s1_A_Gain_Gain
*/
hier_Y.root_Out1 = (hier_P.s1_A_Gain_Gain * rtb_s1_A_Pulse);
This code, generated with Include system hierarchy number in identifiers
off, does not contain a subsystem tag in the generated symbols.
/* Outport: <Root>/Out1 incorporates:
*
Gain: <S1>/A_Gain
*
* Regarding <S1>/A_Gain:
*
Gain value: hier_P.A_Gain_Gain
*/
hier_Y.Out1 = (hier_P.A_Gain_Gain * rtb_A_Pulse);
See “Tracing Generated Code Back to Your Simulink Model” on page 2-33 for
further information on using system and block identification tags.
Prefix Model Name to Global Identifiers Option
When this option is selected, subsystem function names are prefixed with the
name of the model (model_) for all code formats. In addition, when appropriate
to the code format, the model name is also prefixed to the names of functions
and data structures at the model level. This is useful when you need to compile
and link code from two or more models into a single executable, as it avoids
2-15
2
Code Generation and the Build Process
potential name clashes. Prefix model name to global identifiers is ON by
default.
Generate Scalar Inlined Parameters as Option
When the Inline Parameters Option is selected and signals are scalars having
constant sample time, this pull-down menu enables you to control how
parameters are expressed in the code. There are two choices for this option:
• Literals — parameters are expressed as numeric constants
• Macros — parameters are expressed as variables (via #define macros)
The default is Literals. This provides backward compatibility to prior
versions of Real-Time Workshop, which lacked this option. It also may help in
debugging TLC code, as it makes the values of parameters easy to search for.
The Macros option, on the other hand, may make code more readable.
Generate Comments Option
By default, Generate comments is ON. If this option is OFF, generation of
comments in the code is completely suppressed. The Show eliminated
statements and Force generation of parameter comments options in the
General code generation category enable the inclusion of those specific types
of comments.
Target-specific Code Generation Options
Different target configurations support different code generation options that
are not supported by all available targets. For example, the grt, grt_malloc, ert,
rapid simulation, Tornado, xPC, TI DSP, and Real-Time Windows targets
support external mode, but other targets do not.
This section summarizes the options specific to the generic real-time (GRT)
target. For information on options specific to other targets, see the
documentation relevant to those targets. “Available Targets” on page 2-41 lists
targets and related chapters and manuals.
2-16
The Real-Time Workshop User Interface
Figure 2-5: GRT Code Generation Options
MAT-File Variable Name Modifier Menu
This menu selects a string to be added to the variable names used when logging
data to MAT-files. You can select a prefix (rt_), suffix (_rt), or choose to have
no modifier. Real-Time Workshop prepends or appends the string chosen to the
variable names for system outputs, states, and simulation time specified in the
Workspace I/O pane.
See “Workspace I/O Options and Data Logging” on page 2-22 for information
on MAT-file data logging.
External Mode Option
Selecting this option turns on generation of code to support external mode
communication between host and target systems. This option is available for
most targets. For information see “External Mode” on page 6-1.
Ignore Custom Storage Classes Option
Note This option is enabled only if your installation is licensed to use the
Real-Time Workshop Embedded Coder. If you do not have a license for
Embedded Coder, this option will be disabled (grayed out).
2-17
2
Code Generation and the Build Process
When this option is on, data objects with custom storage classes are treated as
if their storage class attribute is set to Auto.
This option is useful if you have defined data objects with custom storage
classes in your model (for use with the Real-Time Workshop Embedded Coder),
but also want to generate code from your model using other targets (such as
GRT or grt_malloc). In such a case, you can turn Ignore Custom Storage
Classes on to generate code that does not include custom storage definitions,
without reconfiguring the storage definitions of the model.
For the GRT and grt_malloc targets, this option is on by default. For the
Real-Time Workshop Embedded Coder, this option is off by default.
You can also enter the option directly into the System target file field in the
Target configuration category of the Real-Time Workshop pane. The
following example turns the option on
-aIgnoreCustomStorageClasses=1
See “Using Custom Storage Classes” in the Real-Time Workshop Embedded
Coder documentation for further information.
TLC Debugging Options
The TLC Debugging options are of interest to those who are writing TLC code
when customizing targets, integrating legacy code, or developing new blocks.
2-18
The Real-Time Workshop User Interface
These options are summarized here; refer to the Target Language Compiler
documentation for details. The TLC Debugging options are
• Retain .rtw file
Normally, the build process deletes the model.rtw file from the build
directory at the end of the build. When Retain .rtw file is selected,
model.rtw is not deleted. This option is useful if you are modifying the target
files, in which case you will need to look at the model.rtw file.
• Profile TLC
When this option is selected, the TLC profiler analyzes the performance of
TLC code executed during code generation, and generates a report. The
report is in HTML format and can be read by your Web browser.
• Start TLC debugger when generating code
This option starts the TLC debugger during code generation.
You can also invoke the TLC debugger by entering the -dc argument into the
System Target File field on the Real-Time Workshop pane.
To invoke the debugger and run a debugger script, enter -df filename into
the System Target File field on the Real-Time Workshop pane.
• Start TLC coverage when generating code
When this option is selected, the Target Language Compiler generates a
report containing statistics indicating how many times each line of TLC code
is hit during code generation.
This option is equivalent to entering the -dg argument into the System
Target File field on the Real-Time Workshop pane.
• Enable TLC Assertions
When this box is selected, Real-Time Workshop will halt building if any
user-supplied TLC file contain an %assert directive that evaluates to FALSE.
The box is not selected by default, meaning that TLC assertion code will be
ignored. You may also use these MATLAB commands to control TLC
assertion handling:
set_param(model, 'TLCAssertion', 'on|off') to set this flag on or off.
Default is Off.
get_param(model, 'TLCAssertion') to see the current setting.
2-19
2
Code Generation and the Build Process
Real-Time Workshop Submenu
The Tools menu of the Simulink window contains a Real-Time Workshop
submenu. The submenu items are:
• Options: Open the Real-Time Workshop pane of the Simulation
Parameters dialog.
• Build Model: Initiate code generation and build process; equivalent to
clicking the Build button in the Real-Time Workshop pane.
• Build Subsystem: Generate code and build an executable from a subsystem;
enabled only when a subsystem is selected. See “Generating Code and
Executables from Subsystems” on page 4-15.
• Generate S-Function: Generate code and build an S-function from a
subsystem; enabled only when a subsystem is selected. See “Automated
S-Function Generation” on page 10-11.
2-20
Simulation Parameters and Code Generation
Simulation Parameters and Code Generation
This section discusses how the simulation parameters of your model interact
with Real-Time Workshop code generation. Only simulation parameters that
affect code generation are mentioned here. For a full description of simulation
parameters, see the Simulink documentation.
This discussion is organized around the following panes of the Simulation
Parameters dialog box:
• Solver pane
• Workspace I/O pane
• Diagnostics pane
• Advanced pane
To view these panes, choose Simulation parameters from the Simulation
menu. When the dialog box opens, click the appropriate tab.
Solver Options
Solver Type. If you are using an S-function or Rapid Simulation (RSIM) target,
you can specify either a fixed-step or a variable-step solver. All other targets
require a fixed-step solver.
Mode. Real-Time Workshop supports both single- and multitasking modes. See
“Models with Multiple Sample Rates” on page 8-1 for full details.
Start and Stop Times. The stop time must be greater than or equal to the start
time. If the stop time is zero, or if the total simulation time (Stop - Start) is
less than zero, the generated program runs for one step. If the stop time is set
to inf, the generated program runs indefinitely.
Note that when using the GRT or Tornado targets, you can override the stop
time when running a generated program from the DOS or UNIX command line.
To override the stop time that was set during code generation, use the -tf
switch.
model -tf n
The program will run for n seconds. If n = inf, the program will run
indefinitely. See “Part 3: Running the External Mode Target Program” on
2-21
2
Code Generation and the Build Process
page 3-40 of the Real-Time Workshop Getting Started Guide for an example of
the use of this option.
Note Certain blocks have a dependency on absolute time. If you are
designing a program that is intended to run indefinitely (Stop time = inf), you
must not use these blocks. See Appendix B, “Blocks That Depend on Absolute
Time” for documentation on which blocks behave this way.
Workspace I/O Options and Data Logging
This section discusses several different methods by which a Real-Time
Workshop generated program can save data to a MAT-file for later analysis.
These methods include
• Using the Workspace I/O pane to define and log workspace return variables
• Logging data from Scope and To Workspace blocks
• Logging data using To File blocks
“Tutorial 2: Data Logging” on page 3-15 of the Real-Time Workshop Getting
Started Guide is an exercise designed to give you hands-on experience with
data logging features of Real-Time Workshop.
Note Data logging is available only for targets that have access to a file
system.
Logging States, Time, and Outputs via the Workspace I/O Pane
The Workspace I/O pane enables a generated program to save system states,
outputs, and simulation time at each model execution time step. The data is
written to a MAT-file, named (by default) model.mat.
Before using this data logging feature, you should learn how to configure a
Simulink model to return output to the MATLAB workspace. This is discussed
in the Simulink documentation.
For each workspace return variable that you define and enable, Real-Time
Workshop defines a MAT-file variable. For example, if your model saves
2-22
Simulation Parameters and Code Generation
simulation time to the workspace variable tout, your generated program will
log the same data to a variable named (by default) rt_tout.
Real-Time Workshop logs the following data:
• All root Outport blocks
The default MAT-file variable name for system outputs is rt_yout.
The sort order of the rt_yout array is based on the port number of the
Outport block, starting with 1.
• All continuous and discrete states in the model
The default MAT-file variable name for system states is rt_xout.
• Simulation time
The default MAT-file variable name for simulation time is rt_tout.
Real-Time Workshop data logging follows the Workspace I/O Save options:
(Limit data points, Decimation, and Format).
Overriding the Default MAT-File Name. The MAT-file name defaults to model.mat.
To specify a different filename:
1 Choose Simulation parameters from the Simulation menu. The dialog box
opens. Click the Real-Time Workshop tab.
2 Append the following option to the existing text in the Make command field.
OPTS="-DSAVEFILE=filename"
Overriding Default MAT-File Variable Names. By default, Real-Time Workshop
prepends the string rt_ to the variable names for system outputs, states, and
simulation time to form MAT-file variable names. To change this prefix:
1 Choose Simulation parameters from the Simulation menu. The dialog box
opens. Click the Real-Time Workshop tab.
2 Select the target-specific code generation options item from the Category
menu.
3 Select a prefix(rt_) or suffix (_rt) from the MAT-file variable name
modifier field, or choose none for no prefix.
2-23
2
Code Generation and the Build Process
Logging Data with Scope and To Workspace Blocks
Real-Time Workshop also logs data from these sources:
• All Scope blocks that have the save data to workspace option enabled
You must specify the variable name and data format in each Scope block’s
dialog box.
• All To Workspace blocks in the model
You must specify the variable name and data format in each To Workspace
block’s dialog box.
The variables are written to model.mat, along with any variables logged from
the Workspace I/O pane.
Logging Data with To File Blocks. You can also log data to a To File block. The
generated program creates a separate MAT-file (distinct from model.mat) for
each To File block in the model. The file contains the block’s time and input
variable(s). You must specify the filename, variable name(s), decimation, and
sample time in the To File block’s dialog box.
Note that the To File block cannot be used in DOS real-time targets because of
limitations of the DOS target.
Data Logging Differences in Single- and Multitasking Models
When logging data in singletasking and multitasking systems, you will notice
differences in the logging of
• Noncontinuous root Outport blocks
• Discrete states
In multitasking mode, the logging of states and outputs is done after the first
task execution (and not at the end of the first time step). In singletasking mode,
Real-Time Workshop logs states and outputs after the first time step.
See “Data Logging In Singletasking and Multitasking Model Execution” on
page 7–13 for more details on the differences between single- and multitasking
data logging.
2-24
Simulation Parameters and Code Generation
Note The rapid simulation target (rsim) provides enhanced logging options.
See “Real-Time Workshop Rapid Simulation Target” on page 11-1 for more
information.
Diagnostics Pane Options
The Diagnostics pane specifies what action should be taken when various
model conditions such as unconnected ports are encountered. You can specify
whether to ignore a given condition, issue a warning, or raise an error. If an
error condition is encountered during a build, the build is terminated. The
Diagnostics pane is fully described in the Simulink documentation.
2-25
2
Code Generation and the Build Process
Advanced Options Pane
The Advanced pane includes several options that affect the performance of
generated code. The Advanced pane has two sections. Options in the Model
parameter configuration section let you specify how block parameters are
represented in generated code, and how they are interfaced to externally
written code. Options in the Optimizations section help you to optimize both
memory usage and code size and efficiency.
Note that the Zero crossing detection option affects only simulations with
variable-step solvers. Therefore, this option is only applicable to code
generation when using the rapid simulation (rsim) target, which is the only
target that allows variable-step solvers. See the Simulink documentation for
further information on the Zero crossing detection option.
Inline Parameters Option
Selecting this option has two effects:
1 Real-Time Workshop uses the numerical values of model parameters,
instead of their symbolic names, in generated code.
If the value of a parameter is a workspace variable, or an expression
including one or more workspace variables, the variable or expression is
evaluated at code generation time. The hard-coded result value appears in
the generated code. An inlined parameter, since it has in effect been
2-26
Simulation Parameters and Code Generation
transformed into a constant, is no longer tunable. That is, it is not visible to
externally written code, and its value cannot be changed at run-time.
2 The Configure button becomes enabled. Clicking the Configure button
opens the Model Parameter Configuration dialog box.
The Model Parameter Configuration dialog box lets you remove individual
parameters from inlining and declare them to be tunable variables (or global
constants). When you declare a parameter tunable, Real-Time Workshop
generates a storage declaration that allows the parameter to be interfaced
to externally written code. This enables your hand-written code to change
the value of the parameter at run-time.
The Model Parameter Configuration dialog box lets you improve overall
efficiency by inlining most parameters, while at the same time retaining the
flexibility of run-time tuning for selected parameters.
See “Parameters: Storage, Interfacing, and Tuning” on page 5-2 for further
information on interfacing parameters to externally written code.
The Inline parameters option also instructs Simulink to propagate constant
sample times. Simulink computes the output signals of blocks that have
constant sample times once during model startup. This improves performance,
since such blocks do not compute their outputs at every time step of the model.
Selecting Inline parameters also interacts with other code generation
parameters as follows:
• When Inline parameters is selected, the Inline invariant signals code
generation option becomes available. See “Inline Invariant Signals Option”
on page 2-11.
• The Parameter pooling option is used only when Inline parameters is
selected. See “Parameter Pooling Option” on page 2-29.
Block Reduction Option
When this option is selected, Simulink collapses certain groups of blocks into a
single, more efficient block, or removes them entirely. This results in faster
model execution during simulation and in generated code. The appearance of
the source model does not change.
By default, the Block reduction option is on.
2-27
2
Code Generation and the Build Process
The types of block reduction optimizations currently supported are
Accumulator Folding. Simulink recognizes certain constructs as accumulators,
and reduces them to a single block. For a detailed example, see “Accumulators”
on page 9-36.
Removal of Redundant Type Conversions. Unnecessary type conversion blocks are
removed. For example, an int type conversion block whose input and output
are of type int is redundant and will be removed.
Dead Code Elimination. Any blocks or signals in an unused code path are
eliminated from the generated code the Block reduction option is on. There
are three conditions that all need to be met for a block to be considered part of
an unused code path:
1 The block is in a signal path that ends with a Terminator block or a disabled
Assertion block.
2 The block is not in any other signal path.
3 The block does not reference any tunable or global parameters or signal
storage.
Consider the model in the following block diagram.
Code is always generated for the signal path between In1 and Out1, because
this path does not meet condition 1 above. If Inline parameters is off, code is
also generated for the signal path between the In2 and Terminator blocks,
because condition 3 is not satisfied (Gain2 is tunable).
If Inline parameters is on, however, the terminated signal path meets all
three conditions, and is eliminated. The resultant MdlOutputs function is
shown in the following code excerpt.
2-28
Simulation Parameters and Code Generation
void MdlOutputs(int_T tid)
{
/* Outport: '/Out1' incorporates:
*
Gain: '/Gain1'
*
Inport: '/In1'
*
* Regarding '/Gain1':
*
Gain value: 2.0
*/
rtY.Out1 = (2.0 * rtU.In1);
}
Boolean Logic Signals Option
By default, Simulink does not signal an error when it detects that double
signals are connected to blocks that prefer Boolean input. This ensures
compatibility with models created by earlier versions of Simulink that support
only double data types. You can enable strict Boolean type checking by
selecting the Boolean logic signals option.
Selecting this option is recommended. Generated code will require less
memory, because a Boolean signal typically requires one byte of storage while
a double signal requires eight bytes of storage.
Parameter Pooling Option
Parameter pooling occurs when multiple block parameters refer to storage
locations that are separately defined but structurally identical. The
optimization is similar to that of a C compiler that encounters declarations
such as:
int a[] = {1,2,3};
int b[] = {1,2,3};
In such a case, an optimizing compiler would collapse a and b into a single text
location containing the values 1, 2, 3 and initialize a and b from the same
code.
2-29
2
Code Generation and the Build Process
To understand the effect of parameter pooling in Real-Time Workshop,
consider the following scenario.
Assume that the MATLAB workspace variables a and b are defined as follows:
a = [1:1000]; b = [1:1000];
Suppose that a and b are used as vectors of input and output values in two
Look-Up Table blocks in a model. Figure 2-6 shows the model.
Figure 2-6: Model with Pooled Storage for Look-Up Table Blocks
The figure below shows the use of a and b as a parameters of the Look-Up
Table1 and Look-Up Table2 blocks.
Figure 2-7: Pooled Storage in Look-Up Table Blocks
If Parameter pooling is on, pooled storage is used for the input/output data of
the Look-Up Table blocks. The following code fragment shows the definition of
2-30
Simulation Parameters and Code Generation
the global parameter structure of the model (rtP). The input data references
to a and b are pooled in the field rtP.p2. Likewise, while the output data
references (expressions including a and b) are pooled in the field rtP.p3.
typedef struct Parameters_tag {
real_T p2[1000]; /* Variable: p2
* External Mode Tunable: no
* Referenced by blocks:
* <Root>/Look-Up Table1
* <Root>/Look-Up Table2
*/
real_T p3[1000]; /* Expression: tanh(a)
* External Mode Tunable: no
* Referenced by blocks:
* <Root>/Look-Up Table1
* <Root>/Look-Up Table2
*/
} Parameters;
If Parameter pooling is off, separate arrays are declared for the input/output
data of the Look-Up Table blocks. Twice the amount of storage is used:
typedef struct Parameters_tag {
real_T root_Look_Up_Table1_XData[1000];
real_T root_Look_Up_Table1_YData[1000];
real_T root_Look_Up_Table2_XData[1000];
real_T root_Look_Up_Table2_YData[1000];
} Parameters;
The Parameter pooling option has the following advantages:
• Reduces ROM size
• Reduces RAM size for all compilers (rtP is a global vector)
• Speeds up code generation by reducing the size of model.rtw
• Can speed up execution
Note that the generated parameter names consist of the letter p followed by a
number generated by Real-Time Workshop. Comments indicate what
parameters are pooled.
2-31
2
Code Generation and the Build Process
Note The Parameter pooling option affects code generation only when
Inline parameters is on.
Signal Storage Reuse Option
This option instructs Real-Time Workshop to reuse signal memory. This
reduces the memory requirements of your real-time program. We recommend
selecting this option. Disabling Signal storage reuse makes all block outputs
global and unique, which in many cases significantly increases RAM and ROM
usage.
For further details on the Signal storage reuse option, see “Signals: Storage,
Optimization, and Interfacing” on page 5-17.
Note Selecting Signal storage reuse also enables the Local block outputs
option and the Buffer reuse option in the General code generation options
category of the Real-Time Workshop pane. See “Local Block Outputs Option”
on page 2-12 and “Buffer Reuse Option” on page 2-12.
Control over Assertion Block Behavior
The Advanced pane of the Simulation Parameters dialog shown above also
provides you with a contol to specify whether model verification blocks such as
Assert, Check Static Gap, and related range check blocks will be enabled, not
enabled, or default to their local settings. This Model Verification block
control popup menu has the same effect on code generated by Real-Time
Workshop as it does on simulation behavior.
For Assertion blocks that are not disabled, the generated code for a model will
include one of the following statements
utAssert(input_signal);
utAssert(input_signal != 0.0);
utAssert(input_signal != 0);
at appropriate locations, depending on the block’s input signal type (Boolean,
real, or integer, respectively).
2-32
Simulation Parameters and Code Generation
By default utAssert is a noop in generated code. For assertions to abort
execution you must enable them by including a parameter in the make_rtw
command. Specify the Make command field on the Target configuration
category pane as follows:
make_rtw OPTS=’-DDOASSERTS’
If you want triggered assertions to not abort execution and instead to print out
the assertion statement, use the following make_rtw variant:
make_rtw OPTS=’-DDOASSERTS -DPRINT_ASSERTS’
utAssert is defined as
#define utAssert(exp)
assert(exp)
You can provide your own definition of utAssert in a hand-coded header file if
you wish to customize assertion behavior in generated code. See
<matlabroot>/rtw/c/libsrc/rtlibsrc.h for implementation details.
Finally, when running a model in accelerator mode, Simulink will call back to
itself to execute assertion blocks instead of using generated code. Thus
user-defined callback will still be called when assertions fail.
Tracing Generated Code Back to Your
Simulink Model
Real-Time Workshop writes system/block identification tags in the generated
code. The tags are designed to help you identify the block, in your source model,
that generated a given line of code. Tags are located in comment lines above
each line of generated code, and are provided with hyperlinks in HTML codee
generation reports that you can optionally generate.
The tag format is <system>/block_name, where:
• system is either:
- the string 'root', or
- a unique system number assigned by Simulink
• block_name is the name of the block.
The following code fragment illustrates a tag comment adjacent to a line of code
generated by a Gain block at the root level of the source model.
2-33
2
Code Generation and the Build Process
/* Gain Block: <Root>/Gain */
rtb_temp3 *= (rtP.root_Gain_Gain);
The following code fragment illustrates a tag comment adjacent to a line of code
generated by a Gain block within a subsystem one level below the root level of
the source model:
/* Gain Block: <S1>/Gain */
rtB.temp0 *= (rtP.s1_Gain_Gain);
In addition to the tags, Real-Time Workshop documents the tags for each
model in comments in the generated header file model.h. The following
illustrates such a comment, generated from a source model, foo, which has a
subsystem Outer with a nested subsystem Inner:
/* Here is the system hierarchy for this model.
*
* <Root> : foo
* <S1>
: foo/Outer
* <S2>
: foo/Outer/Inner
*/
There are two ways to trace code back to subsystems, blocks and parameters in
your model:
• Through HTML code generation reports via the Help Browser, and
• By typing the appropriate hilite_system commands to MATLAB.
The HTML report for your model.c file displays hyperlinks in “Regarding,”
“Ouport,” and other comment lines such as are shown above. Clicking on such
links in comments will cause the associated block or subsystem to be
highlighted in the model diagram. For further information, see “HTML Code
Generation Reports” on page 3-31 of the Real-Time Workshop Getting Started
Guide.
Using HTML reports is generally the fastest way to trace code back to the
model, but when you know what you are looking for you may achieve the same
result by at the command line. To manually trace a tag back to the generating
block using the hilite_system command:
1 Open the source model.
2 Close any other model windows that are open.
2-34
Simulation Parameters and Code Generation
3 Use the MATLAB hilite_system command to view the desired system and
block.
As an example, consider the model foo mentioned above. If foo is open,
hilite_system('<S1>')
opens the subsystem Outer and
hilite_system('<S2>/Gain1')
opens the subsystem Outer and selects and highlights the Gain block Gain1
within that subsystem.
Other Interactions Between Simulink
and Real-Time Workshop
The Simulink engine propagates data from one block to the next along signal
lines. The data propagated are
• Data type
• Line widths
• Sample times
The first stage of code generation is compilation of the block diagram. This
compile stage is analogous to that of a C program. The C compiler carries out
type checking and preprocessing. Similarly, Simulink verifies that
input/output data types of block ports are consistent, line widths between
blocks are of the correct thickness, and the sample times of connecting blocks
are consistent.
2-35
2
Code Generation and the Build Process
The Simulink engine typically derives signal attributes from a source block.
For example, the Inport block’s parameters dialog box specifies the signal
attributes for the block.
In this example, the Inport block has a port width of 3, a sample time of .01
seconds, the data type is double, and the signal is complex.
This figure shows the propagation of the signal attributes associated with the
Inport block through a simple block diagram.
In this example, the Gain and Outport blocks inherit the attributes specified
for the Inport block.
Sample Time Propagation
Inherited sample times in source blocks (e.g., a root inport) can sometimes lead
to unexpected and unintended sample time assignments. Since a block may
specify an inherited sample time, information available at the outset is often
insufficient to compile a block diagram completely. In such cases, the Simulink
engine propagates the known or assigned sample times to those blocks that
have inherited sample times but which have not yet been assigned a sample
2-36
Simulation Parameters and Code Generation
time. Thus, Simulink continues to fill in the blanks (the unknown sample
times) until sample times have been assigned to as many blocks as possible.
Blocks that still do not have a sample time are assigned a default sample time
according to the following rules:
1 If the current system has at least one rate in it, the block is assigned the
fastest rate.
2 If no rate exists and the model is configured for a variable-step solver, the
block is assigned a continuous sample time (but fixed in minor time steps).
Note that Real-Time Workshop (with the exception of the S-function target)
does not currently support variable-step solvers.
3 If no rate exists and the model is configured for a fixed-step solver, the block
is assigned a discrete sample time of (Tf - Ti)/50, where Ti is the simulation
start time and Tf is the simulation stop time. If Tf is infinity, the default
sample time is set to 0.2.
To ensure a completely deterministic model (one where no sample times are set
using the above rules), you should explicitly specify the sample time of all your
source blocks. Source blocks include root inport blocks and any blocks without
input ports. You do not have to set subsystem input port sample times. You
may want to do so, however, when creating modular systems.
An unconnected input implicitly sources ground. For ground blocks and ground
connections, the default sample time is derived from destination blocks or the
default rule.
All blocks have an inherited sample time (Ts = -1). They will all be assigned a
sample time of (Tf - Ti)/50.
Block Execution Order
Once Simulink compiles the block diagram, it creates a model.rtw file
(analogous to an object file generated from a C file). The model.rtw file
contains all the connection information of the model, as well as the necessary
2-37
2
Code Generation and the Build Process
signal attributes. Thus, the timing engine in Real-Time Workshop can
determine when blocks with different rates should be executed.
You cannot override this execution order by directly calling a block (in
hand-written code) in a model. For example, the disconnected_trigger model
below will have its trigger port source to ground, which may lead to all blocks
inheriting a constant sample time. Calling the trigger function, f(), directly
from user code will not work correctly and should never be done. Instead, you
should use a function-call generator to properly specify the rate at which f()
should be executed, as shown in the connected_trigger model below.
Instead of the function-call generator, you could use any other block that can
drive the trigger port. Then, you should call the model’s main entry point to
execute the trigger function.
For multirate models, a common use of Real-Time Workshop is to build
individual models separately and then hand-code the I/O between the models.
This approach places the burden of data consistency between models on the
developer of the models. Another approach is to let Simulink and Real-Time
Workshop ensure data consistency between rates and generate multirate code
for use in a multitasking environment. The Real-Time Workshop interrupt
template and VxWorks support libraries provide blocks that support both
synchronous and asynchronous I/O via a double-buffering scheme. For a
description of the Real-Time Workshop libraries, see “Asynchronous Support”
on page 13-1 For more information on multirate code generation, see “Models
with Multiple Sample Rates” on page 8-1
Algebraic Loops Unsupported
Real-Time Workshop does not support models containing algebraic loops. An
algebraic loop exists whenever the output of a block having direct feedthrough
2-38
Simulation Parameters and Code Generation
(such as Gain, Sum, Product, and Transfer fcn) is fed back as an input to the
same block. Simulink is often able to solve models that contain algebraic loops,
such as the diagram shown below.
The code generator does not produce code that solves algebraic loops. This
restriction includes models that use Algebraic Constraint blocks in feedback
paths.
2-39
2
Code Generation and the Build Process
Selecting a Target Configuration
The process of generating target-specific code is controlled by three things:
• A system target file
• A template makefile
• A make command
The System Target File Browser lets you specify such a configuration in a
single step, choosing from a wide variety of ready-to-run configurations.
The System Target File Browser
To select a target configuration using the System Target File Browser:
1 Click the Real-Time Workshop tab of the Simulation Parameters dialog
box. The Real-Time Workshop pane appears.
2 Select Target configuration from the Category menu.
3 Click the Browse button next to the System target file field. This opens the
System Target File Browser. The browser displays a list of all currently
available target configurations. When you select a target configuration,
Real-Time Workshop automatically chooses the appropriate system target
file, template makefile, and make command.
Figure 2-8 shows the System Target File Browser with the generic real-time
target selected.
4 Double-click on the desired entry in the list of available configurations.
Alternatively, you can select the desired entry in the list and click OK.
2-40
Selecting a Target Configuration
Figure 2-8: The System Target File Browser
5 When you choose a target configuration, Real-Time Workshop automatically
chooses the appropriate system target file, template makefile, and make
command for the selected target, and displays them in the Real-Time
Workshop pane.
Available Targets
Table 2-1 lists all the supported system target files and their associated code
formats, and template makefiles. The table also gives references to relevant
manuals or chapters in this book. All of these targets are built using the
make_rtw make command.
2-41
2
Code Generation and the Build Process
Table 2-1: Targets Available from the System Target File Browser
2-42
Target/Code Format
System Target File
Template Makefile
Relevant
Chapters
Real-Time Workshop
Embedded Coder (PC
or UNIX)
ert.tlc
ert_default_tmf
Real-Time
Workshop
Embedded Coder
documentation
Real-Time Workshop
Embedded Coder for
Watcom
ert.tlc
ert_watc.tmf
Real-Time
Workshop
Embedded Coder
documentation
Real-Time Workshop
Embedded Coder for
Visual C/C++
ert.tlc
ert_vc.tmf
Real-Time
Workshop
Embedded Coder
documentation
Real-Time Workshop
Embedded Coder for
Visual C/C++ Project
Makefile
ert.tlc
ert_msvc.tmf
Real-Time
Workshop
Embedded Coder
documentation
Real-Time Workshop
Embedded Coder for
Borland
ert.tlc
ert_bc.tmf
Real-Time
Workshop
Embedded Coder
documentation
Real-Time Workshop
Embedded Coder for
LCC
ert.tlc
ert_lcc.tmf
Real-Time
Workshop
Embedded Coder
documentation
Real-Time Workshop
Embedded Coder for
UNIX
ert.tlc
ert_unix.tmf
Real-Time
Workshop
Embedded Coder
documentation
Selecting a Target Configuration
Table 2-1: Targets Available from the System Target File Browser (Continued)
Target/Code Format
System Target File
Template Makefile
Relevant
Chapters
Real-Time Workshop
Embedded Coder for
Tornado (VxWorks)
ert.tlc
ert_tornado.tmf
Real-Time
Workshop
Embedded Coder
documentation
Generic Real-Time
for PC/UNIX
grt.tlc
grt_default_tmf
3
Generic Real-Time
for Watcom
grt.tlc
grt_watc.tmf
3
Generic Real-Time
for Visual C/C++
grt.tlc
grt_vc.tmf
3
Generic Real-Time
for Visual C/C++
Project
Makefile
grt.tlc
grt_msvc.tmf
3
Generic Real-Time
for Borland
grt.tlc
grt_bc.tmf
3
Generic Real-Time
for LCC
grt.tlc
grt_lcc.tmf
3
Generic Real-Time
for UNIX
grt.tlc
grt_unix.tmf
3
Generic Real-Time
(dynamic) for
PC/UNIX
grt_malloc.tlc
grt_malloc_default_tmf
3
Generic Real-Time
(dynamic) for Watcom
grt_malloc.tlc
grt_malloc_watc.tmf
3
Generic Real-Time
(dynamic) for Visual
C/C++
grt_malloc.tlc
grt_malloc_vc.tmf
3
2-43
2
Code Generation and the Build Process
Table 2-1: Targets Available from the System Target File Browser (Continued)
2-44
Target/Code Format
System Target File
Template Makefile
Relevant
Chapters
Generic Real-Time
(dynamic) for Visual
C/C++ Project
Makefile
grt_malloc.tlc
grt_malloc_msvc.tmf
3
Generic Real-Time
(dynamic) for Borland
grt_malloc.tlc
grt_malloc_bc.tmf
3
Generic Real-Time
(dynamic) for LCC
grt_malloc.tlc
grt_malloc_lcc.tmf
3
Generic Real-Time
(dynamic) for UNIX
grt_malloc.tlc
grt_malloc_unix.tmf
3
Rapid Simulation
Target (default for PC
or UNIX)
rsim.tlc
rsim_default_tmf
11
Rapid Simulation
Target for Watcom
rsim.tlc
rsim_watc.tmf
11
Rapid Simulation
Target for Visual
C/C++
rsim.tlc
rsim_vc.tmf
11
Rapid Simulation
Target for Borland
rsim.tlc
rsim_bc.tmf
11
Rapid Simulation
Target for LCC
rsim.tlc
rsim_lcc.tmf
11
Rapid Simulation
Target for UNIX
rsim.tlc
rsim_unix.tmf
11
S-Function Target for
PC or UNIX
rtwsfcn.tlc
rtwsfcn_default_tmf
10
Selecting a Target Configuration
Table 2-1: Targets Available from the System Target File Browser (Continued)
Target/Code Format
System Target File
Template Makefile
Relevant
Chapters
S-Function Target for
Watcom
rtwsfcn.tlc
rtwsfcn_watc.tmf
10
S-Function Target for
Visual C/C++
rtwsfcn.tlc
rtwsfcn_vc.tmf
10
S-Function Target for
Borland
rtwsfcn.tlc
rtwsfcn_bc.tmf
10
S-Function Target for
LCC
rtwsfcn.tlc
rtwsfcn_lcc.tmf
rtwsfcn_unix.tmf
10
Tornado (VxWorks)
Real-Time Target
tornado.tlc
tornado.tmf
12
Windows Real-Time
Target for Watcom
rtwin.tlc
win_watc.tmf
Real-Time
Windows Target
documentation
Windows Real-Time
Target for Visual
C/C++
rtwin.tlc
win_vc.tmf
Real-Time
Windows Target
documentation
Embedded Target for
TIC6000 DSP
ti_c6000.tlc
ti_c6000.tmf
Developer's Kit
for Texas
Instruments DSP
documentation
xPC Target for
Watcom C/C++ or
Visual C/C++
xpctarget.tlc
xpc_default_tmf
xpc_vc.tmf
xpc_watc.tmf
xPC Target
documentation
DOS (4GW)
drt.tlc
drt_watc.tmf
11 and 3
LE/O (Lynx
embedded OSEK)
Real-Time Target
osek_leo.tlc
osek_leo.tmf
Readme file in
matlabroot/rtw/
c/osek_leo
2-45
2
Code Generation and the Build Process
Table 2-1: Targets Available from the System Target File Browser (Continued)
Target/Code Format
System Target File
Template Makefile
Relevant
Chapters
ASAM-ASAP2 Data
Definition Target
asap2.tlc
asap2_generic.tmf
Real-Time
Workshop
Embedded Coder
documentation
ECRobot Target
(ECRobot demo)
ECRobot.tlc
ECRobot.tmf
See demo in
Embedded Target for
Motorola MPC555
Developers Kit
mpc555exp.tlc
matlabroot/tool
box/rtw/targets
/ECRobot
mpc555pil.tlc
mpc555rt.tlc
mpc555exp.tmf
mpc555exp_diab.tmf
mpc555pil.tmf
mpc555pil_diab.tmf
mpc555rt.tmf
Embedded Target
for Motorola
MPC555
documentation
Note The LE/O, DOS, and ECRobot targets are included as examples only.
2-46
Making an Executable
Making an Executable
Real-Time Workshop generates code into a set of source files that vary little
among different targets. Not all possible files will be generated for every model.
Some files are only created when the model includes subsystems or particular
types of data.
The file packaging of the Real-Time Workshop Embedded Coder differs slightly
(but significantly) from the file packaging described below. See the “Data
Structures and Code Modules” section in the Real-Time Workshop Embedded
Coder documentation for further information.
Generated Source Files
The following table summarizes the structure of source code generated by the
Real-Time Workshop. All code modules described are written to the build
directory within your current working directory. Figure 2-9 on page 2-49
summarizes the dependencies among these files.
Table 2-2: Real-Time Workshop File Packaging
File
Description
model.c
Contains entry points for all code implementing the model algorithm
(MdlStart, MdlOutputs, MdlUpdate, MdlInitializeSizes,
MdlInitializeSampleTimes). Also contains model registration code.
model_private.h
Contains local defines and local data that are required by the model and
subsystems. This file is included by the genberated source files in the
model. You do not need to include model_private.h when interfacing
hand-written code to a model.
model.h
Defines model data structures and a public interface to the model entry
points and data structures. Also provides an interface to the real-time
model data structure (model_rtM) via accessor macros. model.h is
included by subsystem .c files in the model.
If you are interfacing your hand-written code to generated code for one
or more models, you should include model.h for each model to which you
want to interface.
2-47
2
Code Generation and the Build Process
Table 2-2: Real-Time Workshop File Packaging (Continued)
File
Description
model_data.c
model_data.c is conditionally generated. It contains the declarations for
(conditional)
the parameters data structure and the constant block I/O data structure.
If these data structures are not used in the model, model_data.c is not
generated. Note that these structures are declared extern in model.h.
model_types.h
Provides forward declarations for the real-time model data structure and
the parameters data structure. These may be needed by function
declarations of reusable functions. model_types.h is included by all the
generated header files in the model.
rtmodel.h
Contains #include directives required by static main program modules
such as grt_main.c and grt_malloc_main.c. Since these modules are
not created at code generation time, they include rt_model.h to access
model-specific data structures and entry points. If you create your own
main program module, take care to include rtmodel.h.
model_pt.c
(optional)
Provides data structures that enable a running program to access model
parameters without use of external mode. To learn how to generate and
use the model_pt.c file, see “C API for Parameter Tuning” on
page 14-77.
model_bio.c
(optional)
Provides data structures that enable your code to access block outputs.
To learn how to generate and use the model_bio.c file, see “Signal
Monitoring via Block Outputs” on page 14-70.
If you have interfaced hand-written code to code generated by previous releases
of the Real-Time Workshop, you may need to remove dependencies on header
files that are no longer generated. Use #include model.h directives, and
remove #include directives referencing any of the following:
• model_common.h (replaced by model_types.h and model_private.h)
• model_export.h (replaced by model.h)
• model_prm.h (replaced by model_data.c)
• model_reg.h (subsumed by model_.c)
Real-Time Workshop generated source file dependencies are depicted in
Figure 2-9 on page 2-49. Arrows emitting from a file indicate the files it
2-48
Making an Executable
includes. As the illustration notes, other dependencies exist, for example on
Simulink header files files tmw_types.h, simstruc_types.h, and optionally
on rtlibsrc.h, plus C library files. The diagram only maps inclusion relations
between files that are generated in the build directory.
The diagram shows that parent system header files (model.h) include all child
subsystem header files (subsystem.h). In more layered models, subsystems
similarly include their children’s header files, on down the model hierarchy. As
a consequence, subsystems are able to recursively “see” into all their
descendents’ subsystems, as well as to see into the root system (because every
subsystem.c includes model.h and model_private.h).
model_private.h
model.c
subsystem.c
model_data.c
subsystem.h
rtmodel.h is a
dummy include
file used only
for grt and
grt_malloc targets
model_types.h
model.h
rtmodel.h
NOTE Files model.h, model_private.h and subsystem.h also depend on Simulink
header files tmw_types.h, simstruct_types.h, and conditionally on rtlibsrc.h
Figure 2-9: Real-Time Workshop Generated File Dependencies
Compilation and Linking
After completing code generation, the build process determines whether or not
to continue and compile and link an executable program. This decision is
governed by the following parameters:
2-49
2
Code Generation and the Build Process
• Generate code only option
When this option is selected, the build process always omits the make phase.
• Makefile-only target
The Visual C/C++ Project Makefile versions of the grt, grt_malloc, and
Real-Time Workshop Embedded Coder target configurations generate a
Visual C/C++ project makefile (model.mak). To build an executable, you must
open model.mak in the Visual C/C++ IDE and compile and link the model
code.
• HOST template makefile variable
The template makefile variable HOST identifies the type of system upon
which your executable is intended to run. The HOST variable can take on one
of three possible values: PC, UNIX, or ANY.
By default, HOST is set to UNIX in template makefiles designed for use with
UNIX (such as grt_unix.tmf), and to PC in the template makefiles designed
for use with development systems for the PC (such as grt_vc.tmf).
If Simulink is running on the same type of system as that specified by the
HOST variable, then the executable is built. Otherwise:
- If HOST = ANY, an executable is still built. This option is useful when you
want to cross-compile a program for a system other than the one Simulink
is running on.
- Otherwise, processing stops after generating the model code and the
makefile; the following message is displayed on the MATLAB command
line.
### Make will not be invoked - template makefile is for a different
host
2-50
Choosing and Configuring Your Compiler
Choosing and Configuring Your Compiler
The Real-Time Workshop build process depends upon the correct installation
of one or more supported compilers. Note that compiler, in this context, refers
to a development environment containing a linker and make utility, in addition
to a high-level language compiler.
The build process also requires the selection of a template makefile. The
template makefile determines which compiler will be run, during the make
phase of the build, to compile the generated code.
This section discusses how to install a compiler and choose an appropriate
template makefile, on both Windows and UNIX systems.
Choosing and Configuring Your Compiler on Windows
On Windows, you must install one or more supported compilers, In addition,
you must define an environment variable associated with each compiler.Make
sure that your compiler is installed as described in “Third-Party Compiler
Installation on Windows” on page 1-11.
You can select a template makefile that is specific to your compiler. For
example, grt_bc.tmf designates the Borland C/C++ compiler, and grt_vc.tmf
designates the Visual C/C++ compiler.
Alternatively, you can choose a default template makefile that will select the
default compiler for your system. The default compiler is the compiler
MATLAB uses to build MEX-files. You can set up the default compiler by using
the MATLAB mex command as shown below.
mex –setup
See the “External Interfaces/API” in the MATLAB online documentation for
information on the mex command.
Default template makefiles are named target_default_tmf. For example, the
default template makefile for the generic real-time (GRT) target is
grt_default_tmf.
The build process uses the following logic to locate a compiler for the generated
code:
1 If a specific compiler is named in the template makefile, the build process
uses that compiler.
2-51
2
Code Generation and the Build Process
2 If the template makefile designates a default compiler (as in
grt_default_tmf), the build process uses the same compiler as those used
for building C MEX-files.
3 If no default compiler is established, the build process examines
environment variables which define the path to installed compilers, and
selects the first compiler located. The variables are searched in the following
order:
- MSDevDir or DEVSTUDIO (defining the path to the Microsoft Visual C/C++)
- WATCOM (defining the path to the Watcom C/C++ compiler)
- BORLAND (defining the path to the Borland C/C++ compiler)
4 If none of the above environment variables is defined, the build process
selects the lcc compiler, which is shipped and installed with MATLAB.
Compile/Build Options for Visual C/C++. Real-Time Workshop offers two sets of
template makefiles designed for use with Visual C/C++.
To compile under Visual C/C++ and build an executable within the Real-Time
Workshop build process, use one of the target_vc.tmf template makefiles:
• ert_vc.tmf
• grt_malloc_vc.tmf
• grt_vc.tmf
• rsim_vc.tmf
Alternatively, you can choose to create a project makefile (model.mak) suitable
for use with the Visual C/C++ IDE. In this case, you must compile and link your
code within the Visual C/C++ environment. To create a Visual C/C++ project
makefile, choose one of the Visual C/C++ Project Makefile versions of the grt,
ert, or grt_malloc target configurations. These configurations use the
target_msvc.tmf template makefiles:
• ert_msvc.tmf
• grt_malloc_msvc.tmf
• grt_msvc.tmf
Choosing and Configuring Your Compiler On UNIX
On UNIX, the Real-Time Workshop build process uses the default compiler. cc
is the default on all platforms except SunOS, where gcc is the default.
2-52
Choosing and Configuring Your Compiler
You should choose the UNIX-specific template makefile that is appropriate to
your target. For example, grt_unix.tmf is the correct template makefile to
build a generic real-time program under UNIX.
Available Compiler/Makefile/Target Configurations
To determine which template makefiles are appropriate for your compiler and
target, see Table 2-1, Targets Available from the System Target File Browser,
on page 2-42.
2-53
2
Code Generation and the Build Process
Template Makefiles and Make Options
Real-Time Workshop includes a set of built-in template makefiles that are
designed to build programs for specific targets.
There are two types of template makefiles:
• Compiler-specific template makefiles are designed for use with a particular
compiler or development system.
By convention, compiler-specific template makefiles are named according to
the target and compiler (or development system). For example, grt_vc.tmf
is the template makefile for building a generic real-time program under
Visual C/C++; ert_lcc.tmf is the template makefile for building a
Real-Time Workshop Embedded Coder program under the LCC compiler.
• Default template makefiles make your model designs more portable, by
choosing the correct compiler-specific makefile and compiler for your
installation. “Choosing and Configuring Your Compiler” on page 2-51
describes the operation of default template makefiles in detail.
Default template makefiles are named target_default_tmf. For example,
grt_default_tmf is the default template makefile for building a generic
real-time program; ert_default_tmf is the default template makefile
building a Real-Time Workshop Embedded Coder program.
You can supply options to makefiles via arguments to the Make command
field of the Target configuration category of the Real-Time Workshop tab of
the Simulation Parameters dialog. Append the arguments after make_rtw (or
make_xpc or other make command), as in the following example.
make_rtw OPTS="-DMYDEFINE=1"
The syntax for make command options differs slightly for different compilers.
Compiler-Specific Template Makefiles
This section documents the available compiler-specific template makefiles and
common options you can use with each.
Template Makefiles for UNIX
• ert_unix.tmf
2-54
Template Makefiles and Make Options
• grt_malloc_unix.tmf
• grt_unix.tmf
• rsim_unix.tmf
• rtwsfcn_unix.tmf
The template makefiles for UNIX platforms are designed to be used with GNU
Make. These makefile are set up to conform to the guidelines specified in the
IEEE Std 1003.2-1992 (POSIX) standard.
You can supply options via arguments to the make command.
• OPTS — User-specific options, for example,
make_rtw OPTS="-DMYDEFINE=1"
• OPT_OPTS — Optimization options. The default optimization option is -O. To
turn off optimization and add debugging symbols, specify the -g compiler
switch in the make command, for example,
make_rtw OPT_OPTS="-g"
For additional options, see the comments at the head of each template
makefile.
Template Makefiles for Visual C/C++
Real-Time Workshop offers two sets of template makefiles designed for use
with Visual C/C++.
To build an executable within Real-Time Workshop build process, use one of
the target_vc.tmf template makefiles:
• ert_vc.tmf
• grt_malloc_vc.tmf
• grt_vc.tmf
• rsim_vc.tmf
• rtwsfcn_vc.tmf
You can supply options via arguments to the make command.
• OPTS — User-specific options, for example,
make_rtw OPTS="-DMYDEFINE=1"
2-55
2
Code Generation and the Build Process
• OPT_OPTS — Optimization options. The default optimization option is -Ot. To
turn off optimization and add debugging symbols, specify the -Zd compiler
switch in the make command.
make_rtw OPT_OPTS="-Zd"
For additional options, see the comments at the head of each template
makefile.
To create a Visual C/C++ project makefile (model.mak) without building an
executable, use one of the target_msvc.tmf template makefiles:
• ert_msvc.tmf
• grt_malloc_msvc.tmf
• grt_msvc.tmf
These template makefiles are designed to be used with nmake, which is bundled
with Visual C/C++.
You can supply the following options via arguments to the nmake command:
• OPTS — User-specific options, for example,
make_rtw OPTS="/D MYDEFINE=1"
For additional options, see the comments at the head of each template
makefile.
Template Makefiles for Watcom C/C++
Note As of this printing, the Watcom C compiler is no longer available from
the manufacturer. Real-Time Workshop continues to ship with
Watcom-related template makefiles at this time. However, this policy may be
subject to change in the future.
• drt_watc.tmf
• ert_watc.tmf
• grt_malloc_watc.tmf
• grt_watc.tmf
• rsim_watc.tmf
• rtwsfcn_watc.tmf
• win_watc.tmf
2-56
Template Makefiles and Make Options
Real-Time Workshop provides template makefiles to create an executable for
Windows using Watcom C/C++. These template makefiles are designed to be
used with wmake, which is bundled with Watcom C/C++.
You can supply options via arguments to the make command. Note that the
location of the quotes is different from the other compilers and make utilities
discussed in this chapter:
• OPTS — User specific options, for example,
make_rtw "OPTS=-DMYDEFINE=1"
• OPT_OPTS — Optimization options. The default optimization option is -oxat.
To turn off optimization and add debugging symbols, specify the -d2
compiler switch in the make command, for example,
make_rtw "OPT_OPTS=-d2"
For additional options, see the comments at the head of each template
makefile.
Template Makefiles for Borland C/C++
• ert_bc.tmf
• grt_bc.tmf
• grt_malloc_bc.tmf
• rsim_bc.tmf
• rtwsfcn_bc.tmf
Real-Time Workshop provides template makefiles to create an executable for
Windows using Borland C/C++.
You can supply these options via arguments to the make command:
• OPTS — User-specific options, for example,
make_rtw OPTS="-DMYDEFINE=1"
• OPT_OPTS — Optimization options. Default is none. To turn off optimization
and add debugging symbols, specify the -v compiler switch in the make
command.
make_rtw OPT_OPTS="-v"
For additional options, see the comments at the head of each template
makefile.
2-57
2
Code Generation and the Build Process
Template Makefiles for LCC
• ert_lcc.tmf
• grt_lcc.tmf
• grt_malloc_lcc.tmf
• rsim_lcc.tmf
• rtwsfcn_lcc.tmf
Real-Time Workshop provides template makefiles to create an executable for
Windows using LCC compiler Version 2.4 and GNU Make (gmake).
You can supply options via arguments to the make command:
• OPTS — User-specific options, for example,
make_rtw OPTS="-DMYDEFINE=1"
• OPT_OPTS — Optimization options. Default is none. To enable debugging,
specify -g4 in the make command:
make_rtw OPT_OPTS="-g4"
For additional options, see the comments at the head of each template
makefile.
Template Makefile Structure
The detailed structure of template makefiles is documented in “Template
Makefiles” on page 14-28. This information is provided for those who want to
customize template makefiles.
2-58
Configuring the Generated Code via TLC
Configuring the Generated Code via TLC
This section covers features of the Real-Time Workshop Target Language
Compiler that help you to fine-tune your generated code. To learn more about
TLC, please see the Target Language Compiler Reference Guide.
Target Language Compiler Variables and Options
The Target Language Compiler supports extended code generation variables
and options in addition to those included in the code generation options
categories of the Real-Time Workshop pane.
There are two ways to set TLC variables and options:
• Assigning TLC variables in the system target file
• Entering TLC options or variables into the System Target File field on the
Real-Time Workshop tab of the Simulation Parameters dialog.
Assigning Target Language Compiler Variables
The %assign statement lets you assign a value to a TLC variable, as in
%assign MaxStackSize = 4096
This is also known as creating a parameter name/parameter value pair.
The %assign statement is described in the Target Language Compiler
Reference Guide. It is recommended that you write your %assign statements
in the Configure RTW code generation settings section of the system target
file.
2-59
2
Code Generation and the Build Process
The following table lists the code generation variables you can set with the
%assign statement.
Table 2-3: Target Language Compiler Optional Variables
Variable
MaxStackSize=N
Description
When Local block outputs is enabled,
the total allocation size of local variables
that are declared by all functions in the
entire model may not exceed
MaxStackSize (in bytes). N can be any
positive integer. The default value for
MaxStackSize is rtInf, i.e. unlimited
stack size.
MaxStackVariableSize=N
When Local block outputs is enabled, this
limits the size of any local variable declared
in a function to N bytes, where N>0. A
variable whose size exceeds
MaxStackVariableSize will be allocated in
global, rather than local, memory
WarnNonSaturatedBlocks=
Flag to control display of overflow warnings
for blocks that have saturation capability,
but have it turned off (unchecked) in their
dialog. These are the options:
value
• 0 — no warning is displayed
• 1 — displays one warning for the model
during code generation
• 2 — displays one warning that contains a
list of all offending blocks
2-60
Configuring the Generated Code via TLC
Table 2-3: Target Language Compiler Optional Variables (Continued)
Variable
Description
BlockIOSignals=value
Supports monitoring signals in a running
model. See “Signal Monitoring via Block
Outputs” on page 14-70. Setting the
variable causes the model_bio.c file to be
generated. These are the options:
• 0 — deactivates this feature
• 1 — creates model_bio.c
ParameterTuning=value
Setting the variable to 1 causes a
parameter tuning file (model_pt.c) to be
generated. model_pt.c contains data
structures that enable a running program
to access model parameters independent of
external mode. See “C API for Parameter
Tuning” on page 14-77.
Setting Target Language Compiler Options
You can enter TLC options directly into the System target file field in the
Target configuration category of the Real-Time Workshop tab of the
Simulation Parameters dialog, by appending the options and arguments to
the system target filename. This is equivalent to invoking the Target Language
Compiler with options on the MATLAB command line.
The common options are shown in the table below.
Table 2-4: Target Language Compiler Options
Option
Description
−Ipath
Adds path to the list of paths in which to search
for target files (.tlc files).
−m[N|a]
Maximum number of errors to report when an
error is encountered (default is 5). For example,
−m3 specifies that at most three errors will be
reported. To report all errors, specify −ma.
2-61
2
Code Generation and the Build Process
Table 2-4: Target Language Compiler Options
Option
Description
−d[g|n|o]
Specifies debug mode (generate, normal, or
off). Default is off. When −dg is specified,
a .log file is create for each of your TLC files.
When debug mode is enabled (i.e., generate or
normal), the Target Language Compiler displays
the number of times each line in a target file is
encountered.
−aVariable=val
Equivalent to the TLC statement
%assign Variable = val
Note: It is recommended that you use %assign
statements in the TLC files, rather than the -a
option.
2-62
3
Generated Code Formats
This chapter summarizes distinguishing characteristics of code formats that Real-Time Workshop
generates:
Introduction (p. 3-2)
Explains the concept of code formats and relationship to
targets.
Choosing a Code Format for Your
Application (p. 3-3)
Discusses the applicability and limitations of code
formats and targets with regard to types of applications
Real-Time Code Format (p. 3-6)
Describes code generation for building nonembedded
applications
Real-Time malloc Code Format (p. 3-8)
Describes code generation for building nonembedded
applications with dynamic allocation
S-Function Code Format (p. 3-10)
Describes code generation for building S-function targets
Embedded C Code Format (p. 3-11)
Describes code generation for building embedded
applications
3
Generated Code Formats
Introduction
Real-Time Workshop provides four different code formats. Each code format
specifies a framework for code generation suited for specific applications.
The four code formats and corresponding application areas are:
• Real-time: Rapid prototyping
• Real-time malloc: Rapid prototyping
• S-function: Creating proprietary S-function .dll or MEX-file objects, code
reuse, and speeding up your simulation
• Embedded C: Deeply embedded systems
This chapter discusses the relationship of code formats to the available target
configurations, and factors you should consider when choosing a code format
and target. This chapter also summarizes the real-time, real-time malloc,
S-function, and embedded C code formats.
3-2
Choosing a Code Format for Your Application
Choosing a Code Format for Your Application
Your choice of code format is the most important code generation option. The
code format specifies the overall framework of the generated code and
determines its style.
When you choose a target, you implicitly choose a code format. Typically, the
system target file will specify the code format by assigning the TLC variable
CodeFormat. The following example is from ert.tlc.
%assign CodeFormat = "Embedded-C"
If the system target file does not assign CodeFormat, the default is RealTime (as
in grt.tlc).
If you are developing a custom target, you must consider which code format is
best for your application and assign CodeFormat accordingly.
Choose the real-time or real-time malloc code format for rapid prototyping. If
your application does not have significant restrictions in code size, memory
usage, or stack usage, you may want to continue using the generic real-time
(GRT) target throughout development. The real-time format is the most
comprehensive code format and supports almost all the built-in blocks. It is
also capable of executing in “hard real time” (however, if the hard execution
time constraints are not satisfied, a catastrophic system failure occurs). For
further information on satisfying time constraints in both singletasking and
multitasking environments, see “Models with Multiple Sample Rates” on
page 8-1.
If your application demands that you limit source code size, memory usage, or
maintain a simple call structure, then you should choose the Real-Time
Workshop Embedded Coder target, which uses the embedded C format.
Finally, you should choose the S-function format if you are not concerned about
RAM and ROM usage and want to:
• Use a model as a component, for scalability
• Create a proprietary S-function .dll or MEX-file object
• Interface the generated code using the S-function C API
• Speed up your simulation
3-3
3
Generated Code Formats
Table 3-1 summarizes the various options available for each Real-Time
Workshop code format/target, noting exceptions below.
Table 3-1: Features Supported by Real-Time Workshop Targets and Code Formats
Feature
GRT
Static mem.
allocation
X
Dynamic
mem.
allocation
RealRTW
DOS
time
Embe
malloc dded
Coder
X
X
SFunc
RSIM
X
X
X
X
X
RT
Win
xPC
TI
DSP
MPC
555
X
X
X
X
X
X
Continuous
time
X
X
X
X
X
X
X
X
C MEX
S-functions
(noninlined)
X
X
X
X
X
X
X
X
Any
S-function
(inlined)
X
X
X
X
X
X
X
X
Minimize
RAM / ROM
usage
X
X
Supports
external
mode
X
X
Intended for
rapid
prototyping
X
X
Production
code
3-4
X
Tornado
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X3
Choosing a Code Format for Your Application
Table 3-1: Features Supported by Real-Time Workshop Targets and Code Formats (Continued)
Feature
GRT
RealRTW
DOS
time
Embe
malloc dded
Coder
Tornado
SFunc
Batch
parameter
tuning and
Monte Carlo
methods
RT
Win
xPC
TI
DSP
MPC
555
X
X
X
X2
X
Executes in
hard real
time
X1
X1
X1
Non
real-time
executable
included
X
X
X
Multiple
instances of a
model (no
Stateflow
blocks in
model)
RSIM
X
X
Supports
variable-step
solvers
X
X
X
X
X
X
X
1
The default GRT, GRT malloc, and ERT rt_main files emulate execution of hard real time, and
when explicitly connected to a real-time clock execute in hard real-time.
2
Except MPC555 (processor-in-the-loop) and MPC555 (algorithm export) targets
3
Exccept MPC555 (algorithm export) targets
3-5
3
Generated Code Formats
Real-Time Code Format
The real-time code format (corresponding to the generic real-time target) is
useful for rapid prototyping applications. If you want to generate real-time
code while iterating model parameters rapidly, you should begin the design
process with the generic real-time target. The real-time code format supports:
• Continuous time
• Continuous states
• C MEX S-functions (inlined and noninlined)
For more information on inlining S-functions, see the Target Language
Compiler Reference Guide.
The real-time code format declares memory statically, that is, at compile time.
Unsupported Blocks
The real-time format does not support the following built-in blocks:
• Functions & Tables
- MATLAB Fcn (note that Simulink Fcn blocks are supported)
- S-Function — M-file S-functions, Fortran S-functions, or C MEX
S-functions that call into MATLAB (Simulink Fcn calls are supported).
System Target Files
• drt.tlc — DOS real-time target
• grt.tlc — Generic real-time target
• osek_leo.tlc — Lynx-Embedded OSEK target (example only)
• rsim.tlc — Rapid simulation target
• tornado.tlc — Tornado (VxWorks) real-time target
Template Makefiles
• drt.tmf
• grt
- grt_bc.tmf — Borland C
3-6
Real-Time Code Format
- grt_vc.tmf — Visual C
- grt_watc.tmf — Watcom C
- grt_lcc.tmf — LCC compiler
- grt_unix.tmf — UNIX host
• osek_leo.tmf
• rsim
- rsim_bc.tmf — Borland C
- rsim_vc.tmf — Visual C
- rsim_watc.tmf — Watcom C
- rsim_lcc.tmf — LCC compiler
- rsim_unix.tmf — UNIX host
• tornado.tmf
• win_watc.tmf
3-7
3
Generated Code Formats
Real-Time malloc Code Format
The real-time malloc code format (corresponding to the generic real-time
malloc target) is very similar to the real-time code format. The differences are:
• Real-time malloc declares memory dynamically.
Note that for blocks provided by the MathWorks, malloc calls are limited to
the model initialization code. Generated code is designed to be free from
memory leaks, provided that the model termination function is called.
• Real-time malloc allows you to multiply instance the same model with each
instance maintaining its own unique data.
• Real-time malloc allows you to combine multiple models together in one
executable. For example, to integrate two models into one larger executable,
real-time malloc maintains a unique instance of each of the two models. If
you do not use the real-time malloc format, the Real-Time Workshop will not
necessarily create uniquely named data structures for each model,
potentially resulting in name clashes.
grt_malloc_main.c, the main routine for the generic real-time malloc
(grt_malloc) target, supports one model by default. See “Combining
Multiple Models” on page 14–103 for information on modifying
grt_malloc_main to support multiple models. grt_malloc_main.c is located
in the directory matlabroot/rtw/c/grt_malloc.
Unsupported Blocks
The real-time malloc format does not support the following built-in blocks, as
shown:
• Functions & Tables
- MATLAB Fcn (note that Simulink Fcn blocks are supported)
- S-Function — M-file S-functions, Fortran S-functions, or C MEX
S-functions that call into MATLAB (Simulink Fcn calls are supported).
System Target Files
• grt_malloc.tlc
• tornado.tlc — Tornado (VxWorks) real-time target
3-8
Real-Time malloc Code Format
Template Makefiles
• grt_malloc
- grt_malloc_bc.tmf — Borland C
- grt_malloc_vc.tmf — Visual C
- grt_malloc_watc.tmf — Watcom C
- grt_malloc_lcc.tmf — LCC compiler
- grt_malloc_unix.tmf — UNIX host
• tornado.tmf
3-9
3
Generated Code Formats
S-Function Code Format
The S-function code format (corresponding to the S-Function Target) generates
code that conforms to the Simulink C MEX S-function API. Using the
S-Function Target, you can build an S-function component and use it as an
S-Function block in another model.
The S-function code format is also used by the Simulink Accelerator to create
the Accelerator MEX-file.
In general you should not use the S-function code format in a system target file.
However, you may need to do special handling in your inlined TLC files to
account for this format. You can check the TLC variable CodeFormat to see if
the current target is a MEX-file. If CodeFormat = "S-Function” and the TLC
variable Accelerator is set to 1, the target is a Simulink Accelerator MEX-file.
See Chapter 10, “The S-Function Target” for further information.
3-10
Embedded C Code Format
Embedded C Code Format
The embedded C code format corresponds to the Real-Time Workshop
Embedded Coder target. This code format includes a number of memory-saving
and performance optimizations. See the Real-Time Workshop Embedded Coder
documentation for full details.
3-11
3
Generated Code Formats
3-12
4
Building Subsystems
This chapter describes how to generate code for atomic and conditionally executed subsystems. Topics
covered in detail include the following:
Nonvirtual Subsystem Code
Generation (p. 4-2)
Discusses ways to generate separate code modules from
nonvirtual subsystems
Generating Code and Executables from Describes how to generate and build a stand-alone
Subsystems (p. 4-15)
executable from a subsystem
4
Building Subsystems
Nonvirtual Subsystem Code Generation
Real-Time Workshop allows you to control how code is generated for any
nonvirtual subsystem. The categories of nonvirtual subsystems are:
• Conditionally executed subsystems: execution depends upon a control signal
or control block. These include triggered subsystems, enabled subsystems,
action and iterator subsystems, subsystems that are both triggered and
enabled, and function call subsystems. See Using Simulink for information
on conditionally executed subsystems.
• Atomic subsystems: Any virtual subsystem can be declared atomic (and
therefore nonvirtual) via the Treat as atomic unit option in the Block
Parameters dialog.
See Using Simulink, and run the sl_subsys_semantics demo for further
information on nonvirtual subsystems and atomic subsystems.
You can control the code generated from nonvirtual subsystems as follows:
• You can instruct Real-Time Workshop to generate separate functions, within
separate code files, for selected nonvirtual systems. You can control both the
names of the functions and of the code files generated from nonvirtual
subsystems.
• You can cause multiple instances of a subsystem to generate reusable code,
that is, as a single re-entrant function, instead of replicating the code for
each instance of a subsystem or each time it is called.
• You can generate inlined code from selected nonvirtual subsystems within
your model. When you inline a nonvirtual subsystem, a separate function
call is not generated for the subsystem.
Nonvirtual Subsystem Code Generation Options
For any nonvirtual subsystem, you can choose the following code generation
options from the RTW system code pop-up menu in the subsystem Block
parameters dialog:
• Auto: This is the default option, and provides the greatest flexibility in most
situations. See “Auto Option” below.
• Inline: This option explicitly directs Real-Time Workshop to inline the
subsystem unconditionally.
4-2
Nonvirtual Subsystem Code Generation
• Function: This option explicitly directs Real-Time Workshop to generate a
separate function with no arguments, and (optionally) place the subsystem
in a separate file. You can name the generated function and file. As functions
created with this option rely on global data, they are not re-entrant.
• Reusable function: Generates a function with arguments, that allows the
subsystem’s code to be shared by other instances of it in the model. To enable
sharing, Real-Time Workshop must be able to determine (via checksums)
that subsystems are identical. The generated function will have arguments
for block inputs and outputs, continuous states, parameters, etc.
The sections below further discuss the Auto, Inline, Function, and Reusable
function options.
Auto Option
The Auto option is the default, and is generally appropriate. Auto causes
Real-Time Workshop to inline the subsystem when there is only one instance
of it in the model. When multiple instances of a subsystem exist, the Auto
option will result in a single copy of the function whenever possible (as a
reusable function). Otherwise, the result will be as though you selected Inline
(except for function call subsystems with multiple callers, which will be
handled as if you specified Function). Choose Inline to always inline
subsystem code, or Function when you specifically want to generate a separate
function without arguments for each instance, optionally in a separate file.
Note When you want multiple instances of a subsystem to be represented as
one reusable function, you may designate each one of them as Auto or as
Reusable function. It is best to use one or the other, as using both will create
two reusable functions, one for each designation. The outcomes of these
choices will differ only when reuse is not possible.
To use the Auto option:
1 Select the subsystem block. Then select Subsystem parameters from the
Simulink Edit menu. The Block Parameters dialog opens, as shown in
Figure 4-1.
Alternatively, you can open the Block Parameters dialog by:
4-3
4
Building Subsystems
- Shift-double-clicking on the Subsystem block
- Right-clicking on the Subsystem block and selecting Block parameters
from the pop-up menu.
2 If the subsystem is virtual, select Treat as atomic unit as shown in
Figure 4-1. This makes the subsystem nonvirtual, and the RTW system
code option becomes enabled.
If the system is already nonvirtual, the RTW system code option is already
enabled.
3 Select Auto from the RTW system code pop-up menu as shown in
Figure 4-1.
4 Click Apply and close the dialog.
Figure 4-1: Auto Code Generation Option for a Nonvirtual Subsystem
Auto Optimization for Special Cases. Rather than reverting to Inline, the Auto
option will optimize code in special situations in which identical subsystems
contain other identical subsystems, by both reusing and inlining generated
4-4
Nonvirtual Subsystem Code Generation
code. Suppose a model, such as schematized in Figure 4-2, contains identical
subsystems A1 and A2. A1 contains subsystem B1, and A2 contains subsystem
B2, which are themselves identical. In such cases, the Auto option will cause
one function will be generated which will be called for both A1 and A2, and this
function will contain one piece of inlined code to execute B1 and B2, insuring
that the resulting code will run as efficiently as possible.
Special Case Optimization:
When B1=B2 and A1=A2, selecting the Auto
option inlines code for B within code for function A
B1
A1
B2
A2
Figure 4-2: Reuse of Identical Nested Subsystems with the Auto Option
Inline Option
As noted above, you can choose to inline subsystem code when the subsystem
is nonvirtual (virtual subsystems are always inlined).
Exceptions to Inlining. Note that there are certain cases in which Real-Time
Workshop will not inline a nonvirtual subsystem, even though the Inline
option is selected. These cases are:
• If the subsystem is a function-call subsystem that is called by a noninlined
S-function, the Inline option is ignored. Noninlined S-functions make such
calls via function pointers; therefore the function-call subsystem must
generate a function with all arguments present.
• In a feedback loop involving function-call subsystems, Real-Time Workshop
will force one of the subsystems to be generated as a function instead of
4-5
4
Building Subsystems
inlining it. Real-Time Workshop selects the subsystem to be generated as a
function based on the order in which the subsystems are sorted internally.
• If a subsystem is called from an S-function block that sets the option
SS_OPTION_FORCE_NONINLINED_FCNCALL to TRUE, it will not be inlined. This
may be the case when user-defined Asynchronous Interrupt blocks or Task
Synchronization blocks are required. Such blocks must be generated as
functions. The VxWorks Asynchronous Interrupt and Task Synchronization
blocks, shipped with Real-Time Workshop, use the
SS_OPTION_FORCE_NONINLINED_FCNCALL option.
To generate inlined subsystem code:
1 Select the subsystem block. Then select Subsystem parameters from the
Simulink Edit menu. The Block Parameters dialog opens, as shown in
Figure 4-3.
Alternatively, you can open the Block Parameters dialog by:
- Shift-double-clicking on the Subsystem block
- Right-clicking on the Subsystem block and selecting Block parameters
from the pop-up menu.
2 If the subsystem is virtual, select Treat as atomic unit as shown in
Figure 4-3. This makes the subsystem atomic, and the RTW system code
pop-up menu becomes enabled.
If the system is already nonvirtual, the RTW system code menu is already
enabled.
3 Select Inline from the RTW system code menu as shown in Figure 4-3.
4 Click Apply and close the dialog.
4-6
Nonvirtual Subsystem Code Generation
Figure 4-3: Inlined Code Generation for a Nonvirtual Subsystem
When you generate code from your model, Real-Time Workshop writes inline
code within model.c (or in its parent system’s source file) to perform subsystem
computations. You can identify this code by system/block identification tags,
such as the following.
/* Atomic SubSystem Block: <Root>/AtomicSubsys1 */
See“Tracing Generated Code Back to Your Simulink Model” in Chapter 2 for
further information on system/block identification tags.
Function Option
Choosing the function option (or Reusable function) lets you direct
Real-Time Workshop to generate a separate function and (optionally) a
separate file for the subsystem. When you select the Function option, two
additional options are enabled:
• The RTW function name options menu lets you control the naming of the
generated function.
4-7
4
Building Subsystems
• The RTW file name options menu lets you control the naming of the
generated file (if a separate file is generated).
Figure 4-4 shows the Block Parameters dialog with the Function option
selected.
RTW Function Name Options Menu. This menu offers the following choices, but note
that the resulting identifiers are also affected by which General code
appearance options are in effect for the model:
• Auto: By default, Real-Time Workshop assigns a unique function name using
the default naming convention: model_subsystem(), where subsystem is the
name of the subsystem (or that of an identical one when code is being
reused). When the Include system hierarchy number in identifiers option
of the General code appearance options is selected, a sequential identifier
(s0, s1,...sn) assigned by Simulink and prefixed to the model name, e.g.
sn_model_subsystem(). When the Prefix model name to global identifiers
option of the General code appearance options is not selected, the above
generic function identifier will take the form of sn_subsystem().
• Use subsystem name: Real-Time Workshop uses the subsystem name as the
function name. The General code appearance options Prefix model name
to global identifiers option setting also affects the resulting identifiers.
Note When a subsystem is a library block, the Use subsystem name option
will cause its function identifier (and filename, see below) to be that of the
library block, regardless of the name(s) used for that subsystem in the model.
• User specified: When this option is selected, the RTW function name text
entry field is enabled. Enter any legal function name. Note that the function
name must be unique, and the General code appearance options settings
are ignored for the function.
RTW File Name Options Menu. This menu offers the following choices:
• Use subsystem name: Real-Time Workshop generates a separate file, using
the subsystem (or library block) name as the filename.
4-8
Nonvirtual Subsystem Code Generation
• Use function name: Real-Time Workshop generates a separate file, using
the function name (as specified by the RTW function name options) as the
filename.
• User specified: When this option is selected, the RTW file name (no
extension) text entry field is enabled. Real-Time Workshop generates a
separate file, using the name you enter as the filename. Enter any filename
desired, but do not include the .c (or any other) extension. This filename
need not be unique.
Note While a subsytem source filename need not be unique, you must avoid
giving nonunique names that result in cyclic dependencies (e.g, sys_a.h
includes sys_b.h, sys_b.h includes sys_c.h, and sys_c.h includes sys_a.h).
• Auto: Real-Time Workshop does not generate a separate file for the
subsystem. Code generated from the subsystem is generated within the code
module generated from the subsystem’s parent system. If the subsystem’s
parent is the model itself, code generated from the subsystem is generated
within model.c.
To generate both a separate subsystem function and a separate file:
1 Select the subsystem block. Then select Subsystem parameters from the
Simulink Edit menu, to open the Block Parameters dialog.
Alternatively, you can open the Block Parameters dialog by:
- Shift-double-clicking on the Subsystem block
- Right-clicking on the Subsystem block and selecting Block parameters
from the pop-up menu.
2 If the subsystem is virtual, select Treat as atomic unit as shown in
Figure 4-4. This makes the subsystem atomic, and the RTW system code
menu becomes enabled.
If the system is already nonvirtual, the RTW system code menu is already
enabled.
3 Select Function from the RTW system code menu as shown in Figure 4-4.
4-9
4
Building Subsystems
4 Set the function name, using the RTW function name options described in
“RTW Function Name Options Menu” on page 4-8.
5 Set the filename, using any RTW file name option other than Auto (options
are described in “RTW File Name Options Menu” on page 4-8).
Figure 4-4 shows the use of the UserSpecified filename option.
6 Click Apply and close the dialog.
Figure 4-4: Subsystem Function Code Generation
with Separate User-Defined File Name
Reusable Function Option
The difference between functions and reusable functions is that the latter have
data passed to them as arguments (enabling them to be re-entrant), while the
former communicate via global data. Choosing the Reusable function option
directs Real-Time Workshop to generate a single function (optionally in a
separate file) for the subsystem, and to call that code for each identical
subsystem in the model, if possible.
4-10
Nonvirtual Subsystem Code Generation
Note The Reusable function option will yield code that gets called from
multiple sites (hence reused) only when the Auto option would also do so. The
difference between these options’ behavior is that when reuse is not possible,
selecting Auto yields inlined code (or if circumstances prohibit inlining, create
a function without arguements), while choosing Reusable function yields a
separate function (with arguments) that is called from only one site.
Subsystems that are superfically identical still may not generate reusable code.
Specifically, Real-Time Workshop is not able to reuse subsystems having any
of the following characteristics:
• Input signals with differing sample times
• Input signals with differing dimensions
• Input signals with differing datatype or complexity
• Subsystem masks designating different run-time parameters
• Subsystems containing identical blocks with different names
• Subsystems containing identical blocks with different parameter settings
Some of these situations can arise even when subsystems are copied and pasted
within or between models or are manually constructed to be identical. If
Real-Time workshop determines that code for a subsystem cannot be reused, it
will output the subsystem as a function with arguments when Reusable
function is selected, but the function will not be reused. If you prefer that
subsystem code be inlined in such circumstances rather than deployed as
functions, you should choose Auto for the RTW system code option.
The presence of certain blocks in a subsystem can also prevent its code from
being reused. These are:
• Scope blocks (with data logging enabled)
• S-function blocks that fail to meet certain criteria
• To File blocks
• To Workspace blocks
Regarding S-function blocks, there are several requirements that need to be
met in order for subsystems containing them to be reused. See “Creating
Code-Reuse-Compatible S-Functions” in the Simulink documentation.
4-11
4
Building Subsystems
When you select the Reusable function option, two additional options are
enabled. See the explanation of “Function Option” on page 4-7 for descriptions
of these options and fields. If you enter names in these fields, you must specify
exactly the same function name and filename for each instance of identical
subsystems for Real-Time Workshop to be able to reuse the subsytem code.
Figure 4-5: Subsystem Reusable Function Code Generation Option
To request that Real-Time Workshop generate reusable subsystem code:
1 Select the subsystem block. Then select Subsystem parameters from the
Simulink Edit menu. The Block Parameters dialog opens, as shown in
Figure 4-3.
Alternatively, you can open the Block Parameters dialog by:
- Shift-double-clicking on the Subsystem block
- Right-clicking on the Subsystem block and selecting Block parameters
from the pop-up menu.
4-12
Nonvirtual Subsystem Code Generation
2 If the subsystem is virtual, select Treat as atomic unit as shown in
Figure 4-5. This makes the subsystem atomic, and the RTW system code
pop-up menu becomes enabled.
If the system is already nonvirtual, the RTW system code menu is already
enabled.
3 Select Resusable function from the RTW system code menu as shown in
Figure 4-5.
4 If you wish to give the function a specific name, set the function name, using
the RTW function name options described in “RTW Function Name Options
Menu” on page 4-8.
If you do not choose the RTW function name Auto option, and want code to
be reused, you must assign exactly the same function name to all other
subsystem blocks that you want to share this code.
5 If you wish to direct the generated code to a specific file, set the filename
using any RTW file name option other than Auto (options are described in
“RTW File Name Options Menu” on page 4-8).
In order for code to be reused, you must follow this step for all other
subsystem blocks that you want to share this code, using the same filename.
6 Click Apply and close the dialog.
Modularity of Subsystem Code
Note that code generated from nonvirtual subsystems, when written to
separate files, is not completely independent of the generating model. For
example, subsystem code may reference global data structures of the model.
Each subsystem code file contains appropriate include directives and
comments explaining the dependencies. Real-Time Workshop checks for cyclic
file dependencies and warns about them at build time. For descriptions of how
generated code is packaged, see “Generated Source Files” on page 2-47.
Code Reuse Diagnostics
HTML code generation reports (see “Generate HTML Report Option” on page 2-10)
contain a Subsystems link in their Contents section to a table that summarizes
4-13
4
Building Subsystems
how nonvirtual subsystems were converted to generated code. The Subsystems
section contains diagnostic information that helps to explain why the contents
of some subsystems were not generated as reusable code. In addition to
diagnosing exceptions, the HTML report’s Subsystems section also maps each
noninlined subsystem in the model to functions or reused functions in the
generated code.
4-14
Generating Code and Executables from Subsystems
Generating Code and Executables from Subsystems
Real-Time Workshop can generate code and build an executable from any
subsystem within a model. The code generation and build process uses the code
generation and build parameters of the root model.
To generate code and build an executable from a subsystem:
1 Set up the desired code generation and build parameters in the Simulation
Parameters dialog, just as you would for code generation from a model.
2 Select the desired subsystem block.
3 Right-click on the subsystem block and select Build Subsystem from the
Real-Time Workshop submenu of the subsystem block’s context menu.
Alternatively, you can select Build Subsystem from the Real-Time
Workshop submenu of the Tools menu. This menu item is enabled when a
subsystem is selected in the current model.
Note If the model is operating in external mode when you select Build
Subsystem, Real-Time Workshop automatically turns off external mode for
the duration of the build, then restores external mode upon its completion.
4 The Build Subsystem window opens. This window displays a list of the
subsystem parameters. The upper pane displays the name, class, and
storage class of each variable (or data object) that is referenced as a block
parameter in the subsystem.When you select a parameter in the upper pane,
the lower pane shows all the blocks that reference the parameter, and the
parent system of each such block.
The StorageClass column contains a popup menu for each row. The menu
lets you set the storage class of any parameter, or inline the parameter. To
inline a parameter, select the Inline option from the menu. To declare a
4-15
4
Building Subsystems
parameter to be tunable, set the storage class to any value other than
Inline.)
In the illustration above, the parameter K2 is inlined, while the other
parameters are tunable and have various storage classes.
See “Parameters: Storage, Interfacing, and Tuning” on page 5-2 and
“Simulink Data Objects and Code Generation” on page 5-32 for further
information on tunable and inlined parameters and storage classes.
5 After selecting tunable parameters, click the Proceed button. This initiates
the code generation and build process.
6 The build process displays status messages in the MATLAB command
window. When the build completes, the generated executable is in your
working directory. The name of the generated executable is subsystem.exe
(PC) or subsystem (UNIX), where subsystem is the name of the source
subsystem block.
The generated code is in a build subdirectory, named
subsystem_target_rtw, where subsystem is the name of the source
subsystem block and target is the name of the target configuration.
4-16
Generating Code and Executables from Subsystems
Note You can generate subsystem code using any target configuration
available in the System Target File Browser. However, if the S-function target
is selected, Build Subsystem behaves identically to Generate S-function.
(See “Automated S-Function Generation” on page 10-11.)
4-17
4
Building Subsystems
4-18
5
Working with Data
Structures
This chapter continues the discussion of code generation and the build process, introduced in Chapter
1, “Understanding Real-Time Workshop.” Topics covered in detail include the following :
Parameters: Storage, Interfacing, and
Tuning (p. 5-2)
How to generate storage declarations for communicating
model parameters to and from user-written code
Signals: Storage, Optimization, and
Interfacing (p. 5-17)
How signal storage optimizations work, and how to
generate storage declarations for communicating model
signals to and from user-written code
Simulink Data Objects and Code
Generation (p. 5-32)
How to represent and store signals and parameters in
Simulink data objects, and how code is generated from
these objects
Block States: Storing and Interfacing
(p. 5-49)
How to generate storage declarations for communicating
discrete block states to and from user-written code
Storage Classes for Data Store Memory How to control data structures which define and initialize
Blocks (p. 5-57)
named shared memory regions, used by the Data Store
Read and Data Store Write blocks
5
Working with Data Structures
Parameters: Storage, Interfacing, and Tuning
Simulink external mode (see Chapter 6, “External Mode”) offers a quick and
easy way to monitor signals and modify parameter values while generated
model code executes. However, external mode may not be appropriate for your
application in some cases. The S-function and DOS targets do not support
external mode, for example. For other targets, you may want your existing code
to access parameters and signals of a model directly, rather than using the
external mode communications mechanism.
This section discusses how Real-Time Workshop generates parameter storage
declarations, and how you can generate the storage declarations you need to
interface block parameters to your code. For guidance on implementing a
parameter tuning interface using a C-API, see “C API for Parameter Tuning”
on page 14-77.
Storage of Nontunable Parameters
By default, block parameters are not tunable in the generated code. In the
default case, Real-Time Workshop has control of parameter storage
declarations and the symbolic naming of parameters in the generated code.
Nontunable parameters are stored as fields within rtP, a model-specific global
parameter data structure. Real-Time Workshop initializes each field of rtP to
the value of the corresponding block parameter at code generation time.
When the Inline parameters option is on, block parameters are evaluated at
code generation time, and their values appear as constants in the generated
code. (A vector parameter, however, may be represented as an array of
constants within rtP.) Use the Generate scalar inline parameters as
pull-down menu on the General code appearance category pane to choose
whether to represent such parameters as literals (numeric constants) or as
macros (#define constants) in the generated code.
As an example of nontunable parameter storage, consider this model.
5-2
Parameters: Storage, Interfacing, and Tuning
The workspace variable Kp sets the gain of the Gain1 block.
Assume that Kp is nontunable, and has a value of 5.0. Table 5-1 shows the
variable declarations and the code generated for Kp in the noninlined and
inlined cases.
Notice that the generated code does not preserve the symbolic name Kp. The
noninlined code represents the gain of the Gain1 block as rtP.Gain1_Gain.
5-3
5
Working with Data Structures
Table 5-1: Nontunable Parameter Storage Declarations and Code
Inline
Parameters
Generated Variable Declaration and Code
Off
typedef struct Parameters_tag {
real_T Sine_Wave_Amp;
real_T Sine_Wave_Bias;
real_T Sine_Wave_Freq;
real_T Sine_Wave_Phase;
real_T Gain1_Gain;
} Parameters;
.
.
Parameters rtP = {
1.0 , /*Sine_Wave_Amp :'<Root>/Sine Wave' */
0.0 , /*Sine_Wave_Bias:'<Root>/Sine Wave' */
1.0 , /*Sine_Wave_Freq:'<Root>/Sine Wave' */
0.0 , /*Sine_Wave_Phase:'<Root>/Sine Wave'*/
5.0
/*Gain1_Gain : '<Root>/Gain1' */
};
.
.
rtY.Out1 = (rtP.Gain1_Gain * rtb_u);
On
rtY.Out1 = (5.0 * rtb_u);
Tunable Parameter Storage
A tunable parameter is a block parameter whose value can be changed at
run-time. A tunable parameter is inherently noninlined. A tunable expression
is an expression that contains one or more tunable parameters.
When you declare a parameter tunable, you control whether or not the
parameter is stored within rtP. You also control the symbolic name of the
parameter in the generated code.
When you declare a parameter tunable, you specify:
5-4
Parameters: Storage, Interfacing, and Tuning
• The storage class of the parameter.
In Real-Time Workshop, the storage class property of a parameter specifies
how Real-Time Workshop declares the parameter in generated code.
Note that the term “storage class,” as used in Real-Time Workshop, is not
synonymous with the term storage class specifier, as used in the C language.
• A storage type qualifier, such as const or volatile. This is simply an string
that is included in the variable declaration, without error checking.
• (Implicitly) the symbolic name of the variable or field in which the parameter
is stored. Real-Time Workshop derives variable and field names from the
names of tunable parameters.
Real-Time Workshop generates a variable or struct storage declaration for
each tunable parameter. Your choice of storage class controls whether the
parameter is declared as a member of rtP or as a separate global variable.
You can use the generated storage declaration to make the variable visible to
your code. You can also make variables declared in your code visible to the
generated code. You are responsible for properly linking your code to generated
code modules.
You can use tunable parameters or expressions in your root model and in
masked or unmasked subsystems, subject to certain restrictions (See “Tunable
Expressions” on page 5-12.)
To declare tunable parameters, you must first enable the Inline parameters
option. You then use the Model Parameter Configuration dialog to remove
individual parameters from inlining and declare them to be tunable. This
allows you to improve overall efficiency by inlining most parameters, while at
the same time retaining the flexibility of run-time tuning for selected
parameters.
The mechanics of declaring tunable parameters is discussed in “Using the
Model Parameter Configuration Dialog” on page 5-8.
Storage Classes of Tunable Parameters
Real-Time Workshop defines four storage classes for tunable parameters. You
must declare a tunable parameter to have one of the following storage classes:
• SimulinkGlobal(Auto): SimulinkGlobal(Auto) is the default storage class.
Real-Time Workshop stores the parameter as a member of rtP. Each
5-5
5
Working with Data Structures
member of rtP is initialized to the value of the corresponding workspace
variable at code generation time.
• ExportedGlobal: The generated code instantiates and initializes the
parameter and model_private.h exports it as a global variable. An exported
global variable is independent of the rtP data structure. Each exported
global variable is initialized to the value of the corresponding workspace
variable at code generation time.
• ImportedExtern: model_private.h declares the parameter as an extern
variable. Your code must supply the proper variable definition and
initializer, if any.
• ImportedExternPointer: model_private.h declares the variable as an
extern pointer. Your code must supply the proper pointer variable definition
and initializer, if any.
The generated code for model.h includes model_private.h in order to make
the extern declarations available to subsystem files.
As an example of how the storage class declaration affects the code generated
for a parameter, consider the model shown below.
The workspace variable Kp sets the gain of the Gain1 block. Assume that the
value of Kp is 5.0. Table 5-2 shows the variable declarations and the code
generated for the gain block when Kp is declared as a tunable parameter. An
example is shown for each storage class.
5-6
Parameters: Storage, Interfacing, and Tuning
Note Real-Time Workshop uses column-major ordering for two-dimensional
signal and parameter data. When interfacing your hand-written code
interfaces to such signals or parameters via ExportedGlobal,
ImportedExtern, or ImportedExternPointer declarations, make sure that
your code observes this ordering convention.
Note that the symbolic name Kp is preserved in the variable and field names in
the generated code.
Table 5-2: Tunable Parameter Storage Declarations and Code
Storage Class
Generated Variable Declaration and Code
SimulinkGlobal(Auto)
typedef struct Parameters_tag {
real_T Kp;
} Parameters;
.
.
Parameters rtP = {
5.0
};
.
.
rtY.Out1 = (rtP.Kp * rtb_u);
ExportedGlobal
extern real_T Kp;
.
.
real_T Kp = 5.0;
.
.
rtY.Out1 = (Kp * rtb_u);
5-7
5
Working with Data Structures
Table 5-2: Tunable Parameter Storage Declarations and Code
Storage Class
Generated Variable Declaration and Code
ImportedExtern
extern real_T Kp;
.
.
rtY.Out1 = (Kp * rtb_u);
ImportedExternPointer
extern real_T *Kp;
.
.
rtY.Out1 = ((*Kp) * rtb_u);
Using the Model Parameter Configuration Dialog
The Model Parameter Configuration dialog is available only when the Inline
parameters option on the Advanced page is selected. Selecting this option
activates the Configure button.
5-8
Parameters: Storage, Interfacing, and Tuning
Clicking on the Configure button opens the Model Parameter Configuration
dialog.
Figure 5-1: The Model Parameter Configuration Dialog
The Model Parameter Configuration dialog lets you select workspace
variables and declare them to be tunable parameters in the current model. The
dialog is divided into two panels:
• The Global (tunable) parameters panel displays and maintains a list of
tunable parameters associated with the model.
• The Source list panel displays a list of workspace variables and lets you add
them to the tunable parameters list.
To declare tunable parameters, you select one or more variables from the
source list, add them to the Global (tunable) parameters list, and set their
storage class and other attributes.
5-9
5
Working with Data Structures
Source List Panel. The Source list panel displays a menu and a scrolling table of
numerical workspace variables.
The menu lets you choose the source of the variables to be displayed in the list.
Currently there is only one choice: MATLAB workspace. The source list
displays names of the variables defined in the MATLAB base workspace.
Selecting one or more variables from the source list enables the Add to table
button. Clicking Add to table adds selected variables to the tunable
parameters list in the Global (tunable) parameters panel. This action is all
that is necessary to declare tunable parameters.
In the source list, the names of variables that have been added to the tunable
parameters list are displayed in italics (See Figure 5-1).
The Refresh list button updates the table of variables to reflect the current
state of the workspace. If you define or remove variables in the workspace
while the Model Parameter Configuration dialog is open, click the Refresh
list button when you return to the dialog. The new variables are added to the
source list.
Global (Tunable) Parameters Panel. The Global (tunable) parameters panel
displays a scrolling table of variables that have been declared tunable in the
current model, and lets you specify their attributes. The Global (tunable)
parameters panel also lets you remove entries from the list, or create new
tunable parameters.
You select individual variables and change their attributes directly in the
table. The attributes are:
• Storage class of the parameter in the generated code. Select one of
- SimulinkGlobal(Auto)
- ExportedGlobal
- ImportedExtern
- ImportedExternPointer
See “Storage Classes of Tunable Parameters” on page 5-5 for definitions.
• Storage type qualifier of the variable in the generated code. For variables
with any storage class except SimulinkGlobal(Auto), you can add a qualifier
(such as const or volatile) to the generated storage declaration. To do so,
you can select a predefined qualifier from the list, or add additional qualifiers
to the list. Note that the code generator does not check the storage type
5-10
Parameters: Storage, Interfacing, and Tuning
qualifier for validity. The code generator includes the qualifier string in the
generated code without syntax checking.
• Name of the parameter. This field is used only when creating a new tunable
variable.
The Remove button deletes selected variables from the Global (tunable)
parameters list.
The New button lets you create new tunable variables in the Global (tunable)
parameters list. At a later time, you can add references to these variables in
the model.
If the name you enter matches the name of an existing workspace variable in
the Source list, that variable is declared tunable, and is displayed in italics in
the Source list.
To define a new tunable variable, click the New button. This creates an empty
entry in the table. Then, enter the name and attributes of the variable and click
Apply.
Note If you edit the name of an existing variable in the list, you actually
create a new tunable variable with the new name. The previous variable is
removed from the list and loses its tunability (that is, it is inlined).
Declaring Tunable Variables
To declare an existing variable tunable:
1 Open the Model Parameter Configuration dialog.
2 In the Source list panel, click on the desired variable in the list to select it.
3 Click the Add to table button. The variable then appears in the table of
tunable variables in the Global (tunable) parameters panel.
4 Click on the variable in the Global (tunable) parameters panel.
5 Select the desired storage class from the Storage class menu.
6 Optionally, select (or enter) a storage type qualifier.
5-11
5
Working with Data Structures
7 Click Apply, or click OK to apply changes and close the dialog.
Tunable Expressions
Real-Time Workshop supports the use of tunable variables in expressions. An
expression that contains one or more tunable parameters is called a tunable
expression.
Currently, there are certain limitations on the use of tunable variables in
expressions. When an expression described below as not supported is
encountered during code generation, a warning is issued and a nontunable
expression is generated in the code. The limitations on tunable expressions are:
• Complex expressions are not supported, except where the expression is
simply the name of a complex variable.
• The use of certain operators or functions in expressions containing tunable
operands is restricted. Restrictions are applied to four categories of operators
or functions, classified in Table 5-3.
Table 5-3: Tunability Classification of Operators and Functions
Category
Operators or Functions
1
+ - .* ./ < > <= >= == ~= & |
2
* /
3
abs, acos, asin, atan, atan2, boolean, ceil, cos,
cosh, exp, floor, int8, int16, int32, log, log10,
rem, sign, sin, sinh, sqrt, tan, tanh, uint8,
uint16, uint32
4
: .^ ^ [] {} . \ .\ ' .' ; ,
The rules applying to each category are as follows:
• Category 1 is unrestricted. These operators can be used in tunable
expressions with any combination of scalar or vector operands.
• Category 2 operators can be used in tunable expressions where at least one
operand is a scalar. That is, scalar/scalar and scalar/matrix operand
combinations are supported, but not matrix/matrix.
5-12
Parameters: Storage, Interfacing, and Tuning
• Category 3 lists all functions that support tunable arguments. Tunable
arguments passed to these functions retain their tunability. Tunable
arguments passed to any other functions lose their tunability.
• Category 4 operators are not supported.
Note The “dot” (structure membership) operator is not supported. This
means that expressions that include a structure member are not tunable.
Tunable Expressions in Masked Subsystems
Tunable expressions are allowed in masked subsystems. You can use tunable
parameter names or tunable expressions in a masked subsystem dialog. When
referenced in lower-level subsystems, such parameters remain tunable.
As an example, consider the masked subsystem depicted below. The masked
dialog variable k sets the gain parameter of theGain.
Suppose that the base workspace variable b is declared tunable with
SimulinkGlobal(Auto) storage class. Figure 5-2 shows the tunable expression
b*3 in the subsystem’s mask dialog.
Figure 5-2: Tunable Expression in Subsystem Mask Dialog
5-13
5
Working with Data Structures
Real-Time Workshop produces the following output computation for theGain.
The variable b is represented as a member of the global parameters structure,
rtP. (Note that for clarity in showing the individual Gain block computation,
Expression folding was turned off in this example.)
/* Gain Block: <S1>/theGain */
rtb_temp0 *= (rtP.b * 3.0);
Limitations of Tunable Expressions in Masked Subsystems. Expressions that include
variables that were declared or modified in mask initialization code are not
tunable.
As an example, consider the subsystem above, modified as follows:
• The mask initialization code is
t = 3 * k;
• The parameter k of the myGain block is 4 + t.
• Workspace variable b = 2. The expression b * 3 is plugged into the mask
dialog as in Figure 5-2.
Since the mask initialization code can only run once, k is evaluated at code
generation time as
4 + (3 * (2 * 3) )
Real-Time Workshop inlines the result. Therefore, despite the fact that b was
declared tunable, the code generator produces the following output
computation for theGain. (Note that for clarity in showing the individual Gain
block computation, Expression folding was turned off in this example.)
/* Gain Block: <S1>/theGain */
rtb_temp0 *= (22.0);
Tunability of Linear Block Parameters
The following blocks have a Realization parameter that affects the tunability
of their parameters:
• Transfer Fcn
• State-Space
• Discrete Transfer Fcn
5-14
Parameters: Storage, Interfacing, and Tuning
• Discrete State-Space
• Discrete Filter
The Realization parameter must be set via the MATLAB set_param
command, as in the following example.
set_param(gcb,'Realization','auto')
The following values are defined for the Realization parameter:
• general: The block's parameters are preserved in the generated code,
permitting parameters to be tuned.
• sparse: The block's parameters are represented in the code by transformed
values that increase the computational efficiency. Because of the
transformation, the block’s parameters are no longer tunable.
• auto: This setting is the default. A general realization is used if one or more
of the block's parameters are tunable. Otherwise sparse, is used.
Note To tune the parameter values of a block of one of the above types
without restriction during an external mode simulation, you must use set
Realization to general.
5-15
5
Working with Data Structures
Parameter Configuration Quick Reference Diagram
Figure 5-3 diagrams the code generation and storage class options that control
the representation of parameters in generated code.
Kp = 5.0;
u
Kp
y
REAL-TIME WORKSHOP CONTROLS SYMBOL USED IN CODE
1 y = u* (rtP.<???>);
[OFF]
Include parameter fields in a
global structure (names may be
mangled)
REAL-TIME WORKSHOP CONTROLS SYMBOL USED IN CODE
[Auto]
(implicit)
Use numeric value of
parameter(if possible)
2
y = u* (5.0);
3
2
const *p_<???> = &rtP.<???>[0];
for (i=0; i<N; i++){
Otherwise, include in a
constant global structure
y[i] = u * (p_<???>[i]);
}
Inline
Parameters
INCLUDED IN LIST OF GLOBAL (TUNABLE) PARAMETERS
ON
[SimulinkGlobal(Auto)] 4
ExportedGlobal
5
y = u* (rtP.Kp);
Include in a
global structure
y = u* (Kp);
ImportedExtern
6
y = u* (Kp);
ImportedExternPointer
7
y = u* (*Kp);
Unstructured
storage
KEY:
[option] : default for code generation option
<???> : RTW generated symbol for parameter storage field
Figure 5-3: Parameter Configuration Quick Reference Diagram
5-16
Symbol preserved
(must be unique)
Signals: Storage, Optimization, and Interfacing
Signals: Storage, Optimization, and Interfacing
Real-Time Workshop offers a number of options that let you control how
signals in your model are stored and represented in the generated code. This
section discusses how you can use these options to:
• Control whether signal storage is declared in global memory space, or locally
in functions (i.e., in stack variables).
• Control the allocation of stack space when using local storage.
• Ensure that particular signals are stored in unique memory locations by
declaring them as test points.
• Reduce memory usage by instructing Real-Time Workshop to store signals
in reusable buffers.
• Control whether or not signals declared in generated code are interfaceable
(visible) to externally written code. You can also specify that signals are to
be stored in locations declared by externally written code.
• Preserve the symbolic names of signals in generated code by using signal
labels.
The discussion in the following sections refers to code generated from
Signals_examp, the model shown in the figure below.
Figure 5-4: Signals_examp Model
Signal Storage Concepts
This section discusses structures and concepts you must understand in order
to choose the best signal storage options for your application:
• The global block I/O data structure rtB
• The concept of signal storage classes as used in Real-Time Workshop
5-17
5
Working with Data Structures
rtB: the Global Block I/O Structure
By default, Real-Time Workshop attempts to optimize memory usage by
sharing signal memory and using local variables.
However, there are a number of circumstances in which it is desirable or
necessary to place signals in global memory. For example:
• You may want a signal to be stored in a structure that is visible to externally
written code.
• The number and/or size of signals in your model may exceed the stack space
available for local variables.
In such cases, it is possible to override the default behavior and store selected
(or all) signals in a model-specific global block I/O data structure. The global
block I/O structure is called rtB.
The following code fragment illustrates how rtB is defined and declared in code
generated (with signal storage optimizations off) from the Signals_examp
model shown in Figure 5-4.
typedef struct BlockIO_tag {
real_T SinSig;
real_T Gain1Sig;
} BlockIO;
.
.
.
/* Block I/O Structure */
BlockIO rtB;
/* <Root>/Sine Wave */
/* <Root>/Gain1 */
Field names for signals stored in rtB are generated according to the rules
described in “Symbolic Naming Conventions for Signals in Generated Code” on
page 5-27.
Storage Classes for Signals
In Real-Time Workshop, the storage class property of a signal specifies how
Real-Time Workshop declares and stores the signal. In some cases this
specification is qualified by further options.
Note that in the context of Real-Time Workshop, the term “storage class” is not
synonymous with the term storage class specifier, as used in the C language.
5-18
Signals: Storage, Optimization, and Interfacing
Default Storage Class. Auto is the default storage class. Auto is the appropriate
storage class for signals that you do not need to interface to external code.
Signals with Auto storage class may be stored in local and/or shared variables,
or in a global data structure. The form of storage depends on the Signal
storage reuse, Buffer reuse, and Local block outputs options, and on
available stack space. See “Signals with Auto Storage Class” on page 5-20 for a
full description of code generation options for signals with Auto storage class.
Explicitly Assigned Storage Classes. Signals with storage classes other than Auto
are stored either as members of rtB, or in unstructured global variables,
independent of rtB. These storage classes are appropriate for signals that you
want to monitor and/or interface to external code.
The Signal storage reuse and Local block outputs optimizations do not apply
to signals with storage classes other than Auto.
Use the Signal Properties dialog to assign these storage classes to signals:
• SimulinkGlobal(Test Point): Test points are stored as fields of the rtB
structure that are not shared or reused by any other signal. See “Declaring
Test Points” on page 5-24 for further information.
• ExportedGlobal: The signal is stored in a global variable, independent of the
rtB data structure. model_private.h exports the variable. Signals with
ExportedGlobal storage class must have unique signal names. See
“Interfacing Signals to External Code” on page 5-25 for further information.
• ImportedExtern: model_private.h declares the signal as an extern
variable. Your code must supply the proper variable definition. Signals with
ImportedExtern storage class must have unique signal names. See
“Interfacing Signals to External Code” on page 5-25 for further information.
• ImportedExternPointer: model_private.h declares the signal as an extern
pointer. Your code must supply a proper pointer variable definition. Signals
with ImportedExtern storage class must have unique signal names. See
“Interfacing Signals to External Code” on page 5-25 for further information.
5-19
5
Working with Data Structures
Signals with Auto Storage Class
This section discusses options that are available for signals with Auto storage
class. These options let you control signal memory reuse and choose local or
global (rtB) storage for signals.
The Signal storage reuse and Buffer reuse options control signal memory
reuse. The Signal storage reuse option is on the Advanced page of the
Simulation Parameters dialog box.
When Signal storage reuse is on, the Buffer reuse option becomes enabled.
The Buffer reuse option is located on the General Code Generation Options
(cont.) category of the Real-Time Workshop pane. When the Buffer reuse
option is selected, signal storage is reused whenever possible.
5-20
Signals: Storage, Optimization, and Interfacing
The Local block outputs option determines whether signals are stored as
members of rtB, or as local variables in functions. This option is in the General
code generation options category of the Real-Time Workshop pane.
By default, both Signal storage reuse and Local block outputs are on.
Note that these options interact. When the Signal storage reuse option is on:
• The Buffer reuse option is enabled. By default, Buffer reuse is on and
signal memory is reused whenever possible.
5-21
5
Working with Data Structures
• The Local block outputs option is enabled. This lets you choose whether
reusable signal variables are declared as local variables in functions, or as
members of rtB.
The following code examples illustrate the effects of the Signal storage reuse,
Buffer reuse, and Local block outputs options. The examples were generated
from the Signals_examp model (see Figure 5-4).
The first example illustrates maximal signal storage optimization, with Signal
storage reuse, Buffer reuse, and Local block outputs on (the default). The
output signals from the Sine Wave and Gain blocks reuse rtb_SinSig, a
variable local to the MdlOutputs function.
/* local block i/o variables */
real_T rtb_SinSig;
/* Sin Block: <Root>/Sine Wave */
rtb_SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq * rtmGetT(rtM_Signals_examp) + ...
rtP.Sine_Wave_Phase) + rtP.Sine_Wave_Bias;
/* Expression for <Root>/Out1 incorporates: */
/*
Gain Block: <Root>/Gain1 */
/* Outport Block: <Root>/Out1 */
rtY.Out1 = (rtP.Gain1_Gain * rtb_SinSig);
If you are constrained by limited stack space, you can turn Local block
outputs off and still benefit from memory reuse. The following example was
generated with Local block outputs off and Signal storage reuse and Buffer
reuse on. The output signals from the Sine Wave and Gain blocks reuse
rtB.temp0, a member of rtB.
rtB.temp0 = rtP.Sine_Wave_Amp * sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) + rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
/* Gain Block: <Root>/Gain1 */
rtB.temp0 *= rtP.Gain1_Gain;
5-22
Signals: Storage, Optimization, and Interfacing
When the Signal storage reuse option is off, Buffer reuse and Local block
outputs are disabled. This makes all block outputs global and unique, as in the
following code fragment.
/* Sin Block: <Root>/Sine Wave */
rtB.SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq * rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) + rtP.Sine_Wave_Bias;
/* Gain Block: <Root>/Gain1 */
rtB.Gain1Sig = rtB.SinSig * rtP.Gain1_Gain;
In large models, disabling Signal storage reuse can significantly increase
RAM and ROM usage. Therefore, this approach is not recommended.
Table 5-4 summarizes the possible combinations of the Signal storage reuse/
Buffer reuse and Local block outputs options.
Table 5-4: Global, Local, and Reusable Signal Storage Options
Signal storage reuse and
Buffer reuse ON
Signal storage reuse OFF
(Buffer reuse disabled)
Local Block
Outputs ON
Reuse signals in local
memory (fully optimized)
N/A
Local Block
Outputs OFF
Reuse signals in rtB
structure
Individual signal storage in
rtB structure
Controlling Stack Space Allocation
When the Local block outputs option is on, the use of stack space is
constrained by the following TLC variables:
• MaxStackSize: the total allocation size of local variables that are declared by
all functions in the entire model may not exceed MaxStackSize (in bytes).
MaxStackSize can be any positive integer. If the total size of local variables
exceeds this maximum, the Target Language Compiler will allocate the
remaining variables in global, rather than local, memory.
The default value for MaxStackSize is rtInf, i.e. unlimited stack size.
5-23
5
Working with Data Structures
• MaxStackVariableSize: limits the size of any local variable declared in a
function to N bytes, where N>0. A variable whose size exceeds
MaxStackVariableSize will be allocated in global, rather than local,
memory.
You can change the values of these variables in your system target file if
necessary. See“Target Language Compiler Variables and Options” on
page 2-59 for further information.
Declaring Test Points
A test point is a signal that is stored in a unique location that is not shared or
reused by any other signal. Test-pointing is the process of declaring a signal to
be a test point.
Test points are stored as members of the rtB structure, even when the Signal
storage reuse and Local block outputs option are selected. Test-pointing lets
you override these options for individual signals. Therefore, you can test-point
selected signals, without losing the benefits of optimized storage for the other
signals in your model.
5-24
Signals: Storage, Optimization, and Interfacing
To declare a test point, use the Simulink Signal Properties dialog box as
follows:
1 In your Simulink block diagram, select the line that carries the signal. Then
select Signal properties from the Edit menu of your model. This opens the
Signal properties dialog box.
Alternatively, you can right-click the line that carries the signal, and select
Signal properties from the pop-up menu.
2 Check the SimulinkGlobal (Test Point) option.
3 Click Apply.
For an example of storage declarations and code generated for a test point, see
Table 5-5, Signal Properties Options and Generated Code, on page 5-29.
Interfacing Signals to External Code
The Simulink Signal Properties dialog lets you interface selected signals to
externally written code. In this way, your hand-written code has access to such
5-25
5
Working with Data Structures
signals for monitoring or other purposes. To interface a signal to external code,
use the Signal Properties dialog box to assign one of the following storage
classes to the signal:
• ExportedGlobal
• ImportedExtern
• ImportedExternPointer
Set the storage class as follows:
1 In your Simulink block diagram, select the line that carries the signal.Then
select Signal Properties from the Edit menu of your model. This opens the
Signal Properties dialog box.
Alternatively, you can right-click the line that carries the signal, and select
Signal properties from the pull-down menu.
2 Deselect the SimulinkGlobal (Test Point) option if necessary. This enables
the RTW storage class field.
3 Select the desired storage class (ExportedGlobal, ImportedExtern, or
ImportedExternPointer) from the RTW storage class menu.
5-26
Signals: Storage, Optimization, and Interfacing
4 Optional: For storage classes other than Auto and SimulinkGlobal, you can
enter a storage type qualifier such as const or volatile in the RTW storage
type qualifier field. Note that Real-Time Workshop does not check this
string for errors; whatever you enter is included in the variable declaration.
5 Click Apply.
Note You can also interface test points and other signals that are stored as
members of rtB to your code. To do this, your code must know the address of
the rtB structure where the data is stored, and other information. This
information is not automatically exported. Real-Time Workshop provides C
and Target Language Compiler APIs that give your code access to rtB and
other data structures. See “Interfacing Parameters and Signals” on page 14-70
for further information.
Limitation on Stateflow Outputs. Note that a nonscalar output signal exiting a
Stateflow chart can not be assigned storage class ImportedExternPointer.
Symbolic Naming Conventions for Signals
in Generated Code
When signals have a storage class other than Auto, Real-Time Workshop
preserves symbolic information about the signals or their originating blocks in
the generated code.
For labelled signals, field names in rtB derive from the signal names. In the
following example, the field names rtB.SinSig and rtB.Gain1Sig derive from
the corresponding labeled signals in the Signals_examp model (shown in
Figure 5-4).
typedef struct BlockIO_tag {
real_T SinSig;
real_T Gain1Sig;
} BlockIO;
/* <Root>/Sine Wave */
/* <Root>/Gain1 */
For unlabeled signals, rtB field names derive from the name of the source block
or subsystem. The naming format is
rtB.system#_BlockName_outport#
5-27
5
Working with Data Structures
where system# is a unique system number assigned by Simulink, BlockName is
the name of the source block, and outport# is a port number. The port number
(outport#) is used only when the source block or subsystem has multiple
output ports.
When a signal has Auto storage class, Real-Time Workshop controls generation
of variable or field names without regard to signal labels.
5-28
Signals: Storage, Optimization, and Interfacing
Summary of Signal Storage Class Options
Table 5-5 shows, for each signal storage class option, the variable declaration
and the code generated for Sine Wave output (SinSig) of the model shown in
Figure 5-4.
Table 5-5: Signal Properties Options and Generated Code
Storage Class
Auto
Declaration
Code
real_T rtb_SinSig;
rtb_SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
Test point
typedef struct
BlockIO_tag {
real_T SinSig;
real_T Gain1Sig;
} BlockIO;
.
.
BlockIO rtB;
rtB.SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
Exported
Global
extern real_T SinSig;
rtB.SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
(with storage
optimizations
on)
(declared in
model_private.h
Imported
Extern
extern real_T SinSig;
(declared in
model_private.h)
Imported
Extern
Pointer
extern real_T *SinSig;
(declared in
model_private.h)
rtB.SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
*SinSig) = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
5-29
5
Working with Data Structures
C API for Parameter Tuning and Signal Monitoring
Real-Time Workshop includes support for development of a C application
program interface (API) for tuning parameters and monitoring signals
independent of external mode. See “Interfacing Parameters and Signals” on
page 14-70 for information.
Target Language Compiler API for Parameter
Tuning and Signal Monitoring
Real-Time Workshop includes support for development of a Target Language
Compiler API for tuning parameters and monitoring signals independent of
external mode. See “Target Language Compiler API for Signals and
Parameters” on page 14-92 for information.
Parameter Tuning via MATLAB Commands
The Model Parameter Configuration dialog is the recommended way to see
or set the attributes of tunable parameters. However, you can also use
MATLAB get_param and set_param commands.
The following commands return the tunable parameters and/or their
attributes:
• get_param(gcs, 'TunableVars')
• get_param(gcs, 'TunableVarsStorageClass')
• get_param(gcs, 'TunableVarsTypeQualifier')
The following commands declare tunable parameters or set their attributes:
• set_param(gcs, 'TunableVars', str)
The argument str (string) is a comma-separated list of variable names.
• set_param(gcs, 'TunableVarsStorageClass', str)
The argument str (string) is a comma-separated list of storage class
settings.
The valid storage class settings are:
- Auto
- ExportedGlobal
- ImportedExtern
- ImportedExternPointer
5-30
Signals: Storage, Optimization, and Interfacing
• set_param(gcs, 'TunableVarsTypeQualifier', str)
The argument str (string) is a comma-separated list of storage type
qualifiers.
The following example declares the variable k1 to be tunable, with storage class
ExportedGlobal and type qualifier const.
set_param(gcs, 'TunableVars', 'k1')
set_param(gcs, 'ExportedGlobal')
set_param(gcs, 'TunableVarsTypeQualifier','const')
5-31
5
Working with Data Structures
Simulink Data Objects and Code Generation
Before using Simulink data objects with Real-Time Workshop, please read the
following:
• The discussion of Simulink data objects in Using Simulink
• “Parameters: Storage, Interfacing, and Tuning” on page 5-2
• “Signals: Storage, Optimization, and Interfacing” on page 5-17
Overview
Within the class hierarchy of Simulink data objects, Simulink provides two
classes that are designed as base classes for signal and parameter storage.
These are:
• Simulink.Parameter: Objects that are instances of the Simulink.Parameter
class or any class derived from Simulink.Parameter are called parameter
objects.
• Simulink.Signal: Objects that are instances of the Simulink.Signal class
or any class derived from Simulink.Signal are called signal objects.
The RTWInfo properties of parameter and signal objects are used by Real-Time
Workshop during code generation. These properties let you assign storage
classes to the objects, thereby controlling how the generated code stores and
represents signals and parameters.
Real-Time Workshop also writes information about the properties of
parameter and signal objects to the model.rtw file. This information, formatted
as Object records, is accessible to Target Language Compiler programs. For
general information on Object records, see “Object information in the
model.rtw file” in the Target Language Compiler Reference Guide.
The general procedure for using Simulink data objects in code generation is as
follows:
1 Define a subclass of one of the built-in Simulink.Data classes.
- For parameters, define a subclass of Simulink.Parameter.
- For signals, define a subclass of Simulink.Signal.
5-32
Simulink Data Objects and Code Generation
2 Instantiate parameter or signal objects from your subclass and set their
properties appropriately, using the Simulink Data Explorer.
3 Use the objects as parameters or signals within your model.
4 Generate code and build your target executable.
The following sections describe the relationship between Simulink data objects
and code generation in Real-Time Workshop.
5-33
5
Working with Data Structures
Parameter Objects
This section discusses how to use parameter objects in code generation.
Configuring Parameter Objects for Code Generation
In configuring parameter objects for code generation, you use the following
code generation and parameter object properties:
• The Inline parameters option (see “Parameters: Storage, Interfacing, and
Tuning” on page 5-2).
• Parameter object properties:
- Value. This property is the numeric value of the object, used as an initial
(or inlined) parameter value in generated code.
- RTWInfo.StorageClass. This property controls the generated storage
declaration and code for the parameter object.
Other parameter object properties (such as user-defined properties of classes
derived from Simulink.Parameter) do not affect code generation.
Note If Inline parameters is off (the default), the RTWInfo.StorageClass
parameter object property is ignored in code generation.
Effect of Storage Classes on Code Generation for Parameter Objects
Real-Time Workshop generates code and storage declarations based on the
RTWInfo.StorageClass property of the parameter object. The logic is as
follows:
• If the storage class is 'Auto' (the default), the parameter object is inlined (if
possible), using the Value property.
• For storage classes other than 'Auto', the parameter object is handled as a
tunable parameter.
- A global storage declaration is generated. You can use the generated
storage declaration to make the variable visible to your hand-written code.
You can also make variables declared in your hand-written code visible to
the generated code.
5-34
Simulink Data Objects and Code Generation
- The symbolic name of the parameter object is preserved in the generated
code.
See Table 5-6 for examples of code generated for each possible setting of
RTWInfo.StorageClass.
Example of Parameter Object Code Generation
In this section, we use the Gain block computations of the model shown in the
figure below as an example of how Real-Time Workshop generates code for a
parameter object.
Figure 5-5: Model Using Parameter Object Kp As Block Parameter
In this model, Kp sets the gain of the Gain1 block.
To configure a parameter object such as Kp for code generation:
1 Define a subclass of Simulink.Parameter. In this example, the parameter
object is an instance of the example class SimulinkDemos.Parameter, which
is provided with Simulink. For the definition of SimulinkDemos.Parameter,
see the directory
matlabroot/toolbox/simulink/simdemos/@SimulinkDemos.
5-35
5
Working with Data Structures
2 Instantiate a parameter object from your subclass. The following example
instantiates Kp as a parameter object of class SimulinkDemos.Parameter.
Kp = SimulinkDemos.Parameter;
Make sure that the name of the parameter object matches the desired block
parameter in your model. This ensures that Simulink can associate the
parameter name with the correct object. For example, in the model of
Figure 5-5, the Gain block parameter Kp resolves to the parameter object Kp.
3 Set the object properties.You can do this via the Simulink Data Explorer.
Alternatively, you can assign properties via MATLAB commands, as follows:
- Set the Value property, for example:
Kp.Value = 5.0;
- Set the RTWInfo.StorageClass property, for example:
Kp.RTWInfo.StorageClass = 'ExportedGlobal';
Table 5-6 shows the variable declarations for Kp and the code generated for the
Gain block in the model shown in Figure 5-5, with Inline parameters on. (Due
to expression folding optimizations, the gain computation is included in the
output computation.) An example is shown for each possible setting of
RTWInfo.StorageClass.
5-36
Simulink Data Objects and Code Generation
Table 5-6: Code Generation from Parameter Objects (Inline Parameters ON)
StorageClass Property
Generated Variable Declaration
and Code
Auto
rtY.Out1 = (5.0 * rtb_u)
Simulink Global
typedef struct Parameters_tag {
real_T Kp;
.
.
Parameters rtP = {
5.0
};
.
.
rtY.Out1 = (rtP.Kp * rtb_u);
Exported Global
extern real_T Kp;
.
.
real_T Kp = 5.0;
.
.
rtY.Out1 = (Kp * rtb_u);
Imported Extern
extern real_T Kp;
.
.
rtY.Out1 = (Kp * rtb_u);
Imported Extern Pointer
extern real_T *Kp;
.
.
rtY.Out1 = ((*Kp) * rtb_u);
5-37
5
Working with Data Structures
Parameter Object Configuration Quick
Reference Diagram
The following figure diagrams the code generation and storage class options
that control the representation of parameter objects in generated code.
Kp = Simulink.Parameter; Kp.Value = 5.0;
u
Kp
y
REAL-TIME WORKSHOP CONTROLS SYMBOL USED IN CODE
1 y = u* (rtP.<???>);
[OFF]
Include parameter fields in a
global structure (names may be
mangled)
REAL-TIME WORKSHOP CONTROLS SYMBOL USED IN CODE
Use numeric value of
parameter(if possible)
2
y = u* (5.0);
3
2
const *p_<???> = &rtP.<???>[0];
for (i=0; i<N; i++){
Otherwise, include in a
constant global structure
y[i] = u * (p_<???>[i]);
}
SimulinkGlobal
4
y = u* (rtP.Kp);
ExportedGlobal
5
y = u* (Kp);
[Auto]
Inline
Parameters
ON
ImportedExtern
6
ImportedExternPointer 7
y = u* (Kp);
Include in a
global structure
Unstructured
storage
Symbol preserved
(must be unique)
y = u* (*Kp);
KEY:
[option] : default for code generation option
<???> : RTW generated symbol for parameter storage field
Figure 5-6: Parameter Object Configuration Quick Reference Diagram
5-38
Simulink Data Objects and Code Generation
Signal Objects
This section discusses how to use signal objects in code generation.
Configuring Signal Objects for Code Generation
In configuring signal objects for code generation, you use the following code
generation options and signal object properties:
• The Signal storage reuse code generation option (see “Signals: Storage,
Optimization, and Interfacing” on page 5-17).
• The Local block outputs code generation option (see “Signals: Storage,
Optimization, and Interfacing” on page 5-17).
• Signal object properties:
- RTWInfo.StorageClass. The storage classes defined for signal objects, and
their effect on code generation, are the same for model signals and signal
objects (see “Storage Classes for Signals” on page 5–18).
Other signal object properties (such as user-defined properties of classes
derived from Simulink.Signal) do not affect code generation.
Effect of Storage Classes on Code Generation for Signal Objects
The way in which Real-Time Workshop uses storage classes to determine how
signals are stored is the same with and without signal objects. However, if a
signal’s label resolves to a signal object, the object’s RTWInfo.StorageClass
property is used in place of the port configuration of the signal.
The default storage class is Auto. If the storage type is Auto, Real-Time
Workshop follows the Signal storage reuse, Buffer reuse, and Local block
outputs code generation options to determine whether signal objects are stored
in reusable and/or local variables. Make sure that these options are set
correctly for your application.
To generate a a test point or externally interfaceable signal storage
declaration, use an explicit RTWInfo.StorageClass assignment. For example,
setting the storage class to SimulinkGlobal, as in the following command, is
equivalent to declaring a signal as a test point.
SinSig.RTWInfo.StorageClass = 'SimulinkGlobal';
5-39
5
Working with Data Structures
Example of Signal Object Code Generation
The discussion and code examples in this section refers to the model shown in
Figure 5-7.
Figure 5-7: Example Model With Signal Object
To configure a signal object, you must first create it and associate it with a
labelled signal in your model. To do this:
1 Define a subclass of Simulink.Signal. In this example, the signal object is
an instance of the example class SimulinkDemos.Signal, which is provided
with Simulink. For the definition of SimulinkDemos.Signal, see the
directory
matlabroot/toolbox/simulink/simdemos/@SimulinkDemos.
2 Instantiate a signal object from your subclass. The following example
instantiates SinSig, a signal object of class SimulinkDemos.Signal.
SinSig = SimulinkDemos.Signal;
Make sure that the name of the signal object matches the label of the desired
signal in your model. This ensures that Simulink can resolve the signal label
to the correct object. For example, in the model shown in Figure 5-7, the
signal label SinSig would resolve to the signal object SinSig.
3 Set the object properties as required. You can do this via the Simulink Data
Explorer. Alternatively, you can assign properties via MATLAB commands.
For example, assign the signal object’s storage class by setting the
RTWInfo.StorageClass property as follows.
SinSig.RTWInfo.StorageClass = 'ExportedGlobal';
Table 5-7 shows, for each setting of RTWInfo.StorageClass, the variable
declaration and the code generated for Sine Wave output (SinSig) of the model
shown in Figure 5-7.
5-40
Simulink Data Objects and Code Generation
Table 5-7: Signal Properties Options and Generated Code
Storage Class
Auto
Declaration
Code
real_T rtb_SinSig;
rtb_SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
Simulink
Global
typedef struct
BlockIO_tag {
real_T SinSig;
real_T Gain1Sig;
} BlockIO;
.
.
BlockIO rtB;
rtb_SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
Exported
Global
extern real_T SinSig;
rtb_SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
Imported
Extern
extern real_T SinSig;
rtb_SinSig = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
Imported
Extern
Pointer
extern real_T *SinSig;
(*SinSig) = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq *
rtmGetT(rtM_Signals_examp) +
rtP.Sine_Wave_Phase) +
rtP.Sine_Wave_Bias;
(with storage
optimizations
on)
5-41
5
Working with Data Structures
Signal Object Configuration Quick
Reference Diagram
Figure 5-8 diagrams the code generation and storage class options that control
the representation of signal objects in generated code.
SIG = Simulink.Signal;
SIG
REAL-TIME WORKSHOP CONTROLS SYMBOL USED IN CODE
Can reuse
signal
[Signal
storage
reuse
ON]
[Auto]
Cannot
reuse signal
Signal
storage
reuse
OFF
[Local block
outputs ON]
Declare locally
1 rtb_temp# = ... (store on stack)
Local block
outputs OFF
Include in a
2 rtB.temp# = ...
global structure
Reuse signal
memory (if
possible)
Include in a
global structure
(see
SimulinkGlobal)
3 rtB.<???> = ...
Storage
Class
4 rtB.<???> = ...
Signal labelled
5
rtB.SIG = ...
Signal not labelled
6
rtB.<???> = ...
ExportedGlobal
7
SIG = ...
ImportedExtern
8
SIG = ...
ImportedExternPointer
9
(*SIG) = ...
SimulinkGlobal
(Test Point)
Include in a
global structure
Unstructured storage
declaration; symbol
preserved (symbol label
must be unique)
KEY:
[option] : default for code generation option
<???> : RTW generated symbol for signal storage field or variable
# : suffix (number) to variable name, appended by RTW
Figure 5-8: Signal Object Configuration Quick Reference Diagram
5-42
Simulink Data Objects and Code Generation
Resolving Conflicts in Configuration of Parameter
and Signal Objects
This section describes how to avoid and resolve certain conflicts that can arise
when using parameter and signal objects.
Parameters
Figure 5-9 and Figure 5-10 illustrate a case where both a tunable parameter Kp
(declared in the Model Parameter Configuration dialog box) and an
identically named parameter object Kp (defined in the Simulink Data
Explorer) exist. If Kp is used as a block parameter, there is a potential for
ambiguity when Simulink attempts to resolve the symbol Kp.
Figure 5-9: Parameter Kp Defined with SimulinkGlobal Storage Class
5-43
5
Working with Data Structures
Figure 5-10: Parameter Object Kp Defined with Auto Storage Class
An obvious solution would be to assign different names to the parameter and
the parameter object.
If this is not desirable, however, you should make sure that the storage class
properties of identically named parameters and parameter objects are
compatible in accordance with Figure 5-11, Compatible Parameter/Parameter
Object Storage Class Configurations. If they are not, an error message will be
displayed when the model is run, and/or when code generation is initiated.
5-44
Simulink Data Objects and Code Generation
In Figure 5-9 and Figure 5-10, the parameter Kp has SimulinkGlobal(auto)
storage class and the parameter object Kp has Auto storage class. Accordingly,
the symbol Kp would resolve to the parameter object Kp.
Figure 5-11: Compatible Parameter/Parameter Object Storage Class
Configurations
Signals and Block States
Figure 5-12 and Figure 5-13 illustrate a case where both a signal Sig (defined
in the Signal Properties dialog box) and a signal object Sig (defined in the
Simulink Data Explorer) exist. There is a potential for ambiguity when
Simulink attempts to resolve the symbol Sig.
5-45
5
Working with Data Structures
Figure 5-12: Signal Sig Defined as SimulinkGlobal (Test Point)
Figure 5-13: Signal Object Sig Defined with Auto Storage Class
An obvious solution would be to assign different names to the signal and the
signal object. If this is not desirable, however, you should make sure that the
storage class properties of identically named signals and signal objects are
compatible in accordance with Figure 5-14, Compatible Signal/Signal Object
5-46
Simulink Data Objects and Code Generation
Configurations. If they are not, an error message will be displayed when model
is run, and/or when code generation is initiated.
In Figure 5-12 and Figure 5-13, the signal and signal objects Sig both have
SimulinkGlobal storage class. Therefore no conflict would arise, and Sig would
resolve to the signal object Sig.
Note The rules for compatibility between block states/signal objects are
identical to those given for signals/signal objects.
Figure 5-14: Compatible Signal/Signal Object Configurations
Customizing Code for Parameter and Signal Objects
You can further influence the treatment of parameter and signal objects in
generated code by using TLC to access fields in object records in model.rtw
files. For details on doing this, please see “Object information in the model.rtw
file” in the Target Language Compiler Reference Guide.
Using Objects to Export ASAP2 Files
The ASAM-ASAP2 Data Definition Target provides special signal and
parameter subclasses that support exporting of signal and parameter object
information to ASAP2 data files. For information about the ASAP2 target and
5-47
5
Working with Data Structures
its associated classes and TLC files, see “Generating ASAP2 Files” in the
Real-Time Workshop Embedded Coder User’s Guide.
5-48
Block States: Storing and Interfacing
Block States: Storing and Interfacing
For certain block types, Real-Time Workshop lets you control how block states
in your model are stored and represented in the generated code. Using the
State Properties dialog, you can:
• Control whether or not states declared in generated code are interfaceable
(visible) to externally written code. You can also specify that states are to be
stored in locations declared by externally written code.
• Assign symbolic names to block states in generated code.
Storage of Block States
The discussion of block state storage in this section applies to the following
block types:
• Discrete Filter
• Discrete State-Space
• Discrete-Time Integrator
• Discrete Transfer Function
• Discrete Zero-Pole
• Memory
• Unit Delay
These block types require persistent memory to store values representing the
state of the block between consecutive time intervals. By default, such values
are stored in a data type work vector. This vector is usually referred to as the
DWork vector. It is represented in generated code as rtDWork, a global data
structure. For further information on the DWork vector, see the Target
Language Compiler Reference Guide.
If you want to interface a block state to your hand-written code, you can specify
that the state is to be stored in a location other than the DWork vector. You do
this by assigning a storage class to the block state.
You can also define a symbolic name, to be used in code generation, for a block
state.
5-49
5
Working with Data Structures
Block State Storage Classes
The storage class property of a block state specifies how Real-Time Workshop
declares and stores the state in a variable. Storage class options for block states
are similar to those for signals. The available storage classes are:
• Auto
• ExportedGlobal
• ImportedExtern
• ImportedExternPointer
Default Storage Class
Auto is the default storage class. Auto is the appropriate storage class for states
that you do not need to interface to external code. States with Auto storage
class are stored as members of the Dwork vector.
You can assign a symbolic name to states with Auto storage class. If you do not
supply a name, Real-Time Workshop generates one, as described in “Symbolic
Names for Block States” on page 5-52.
Explicitly Assigned Storage Classes
Block states with storage classes other than Auto are stored in unstructured
global variables, independent of the Dwork vector. These storage classes are
appropriate for states that you want to interface to external code. The following
storage classes are available for states:
• ExportedGlobal: The state is stored in a global variable. model_private.h
exports the variable. States with ExportedGlobal storage class must have
unique names.
• ImportedExtern: model_private.h declares the state as an extern variable.
Your code must supply the proper variable definition. States with
ImportedExtern storage class must have unique names.
• ImportedExternPointer: model_private.h declares the state as an extern
pointer. Your code must supply the proper pointer variable definition. States
with ImportedExternPointer storage class must have unique names.
Table 5-8, State Properties Options and Generated Code, gives examples of
variable declarations and the code generated for block states with each type of
storage class.
5-50
Block States: Storing and Interfacing
You can assign a symbolic name to states with any of the above storage classes.
If you do not supply a name, Real-Time Workshop generates one, as described
in “Symbolic Names for Block States” on page 5-52.
The next section describes how to use the State Properties dialog box to assign
storage classes to block states.
Using the State Properties Dialog Box to Interface
States to External Code
The State Properties dialog box lets you interface a block’s state to external
code by assigning a storage class other than Auto (i.e., ExportedGlobal,
ImportedExtern, or ImportedExternPointer) to the state.
Set the storage class as follows:
1 In your Simulink block diagram, select the desired block. Then select State
properties from the Edit menu of your model. This opens the State
Properties dialog box.
Alternatively, you can right-click the block, and select State properties
from the pull-down menu.
This picture shows the default settings of the State Properties dialog box.
2 Select the desired storage class (ExportedGlobal, ImportedExtern, or
ImportedExternPointer) from the RTW storage class menu.
3 Optional: For storage classes other than Auto, you can enter a storage type
qualifier such as const or volatile in the RTW storage type qualifier
5-51
5
Working with Data Structures
field. Note that Real-Time Workshop does not check this string for errors;
whatever you enter is included in the variable declaration.
4 Click Apply and close the dialog box.
Symbolic Names for Block States
To determine the variable or field name generated for a block’s state, you can
either:
• Use a default name generated by Real-Time Workshop.
or
• Define a symbolic name via the State Name field of the State Properties
dialog box.
Default Block State Naming Convention
If you do not define a symbolic name for a block state, Real-Time Workshop
uses the following default naming convention:
BlockType#_DSTATE
where
• BlockType is the name of the block type (e.g., Discrete_Filter).
• # is a unique ID number (#) assigned by Real-Time Workshop if multiple
instances of the same block type appear in the model. The ID number is
appended to BlockType.
• _DSTATE is a string that is always appended to the block type and ID number.
For example, consider the model shown in Figure 5-15.
5-52
Block States: Storing and Interfacing
Figure 5-15: Model with Two Discrete Filter Block States
We will examine code generated for the states of the two Discrete Filter blocks.
Assume that:
• Neither block’s state has a user-defined name.
• The upper Discrete Filter block has Auto storage class (and is therefore
stored in the DWork vector).
• The lower Discrete Filter block has ExportedGlobal storage class.
The initialization code for the states of the two Discrete Filter blocks would be
as shown in the following code fragment.
/* DiscreteFilter Block: <Root>/Discrete Filter */
rtDWork.Discrete_Filter_DSTATE = 0.0;
/* DiscreteFilter Block: <Root>/Discrete Filter1 */
Discrete_Filter1_DSTATE = 0.0;
User-Defined Block State Names
Using the State Properties dialog box, you can define your own symbolic name
for a block state. To do this:
1 Select the desired block. Then select State properties from the Edit menu
of your model. This opens the State Properties dialog box.
Alternatively, you can right-click on the block, and select State properties
from the pull-down menu.
2 Enter the symbolic name into the State name field of the State Properties
dialog box. In this picture, the state name Top_filter is entered.
5-53
5
Working with Data Structures
3 Click Apply and close the dialog box.
The following state initialization code was generated from the example model
shown in Figure 5-7, under the following conditions:
• The upper Discrete Filter block has the state name Top_filter, and Auto
storage class (and is therefore stored in the DWork vector.)
• The lower Discrete Filter block has the state name Lower_filter, and
ExportedGlobal storage class.
/* DiscreteFilter Block: <Root>/Discrete Filter */
rtDWork.Top_filter = 0.0;
/* DiscreteFilter Block: <Root>/Discrete Filter1 */
Lower_filter = 0.0;
Block States and Simulink Signal Objects
If you are not familiar with Simulink data objects and signal objects, you
should read “Simulink Data Objects and Code Generation” on page 5–32 before
reading this section.
You can associate a block state with a signal object, and control code generation
for the block state through the signal object. To do this:
1 Instantiate the desired signal object, and set its RTWInfo.StorageClass
property as you require.
5-54
Block States: Storing and Interfacing
2 Open the State Properties dialog box for the block whose state you want to
associate with the signal object. Enter the name of the signal object into the
State name field.
3 Make sure that the storage class and type qualifier settings of the block’s
State Properties dialog box are compatible with those of the signal object.
See “Resolving Conflicts in Configuration of Parameter and Signal Objects”
on page 5-43.
4 Click Apply and close the dialog box.
Note When associating a block state with a signal object, the mapping
between the block state and the signal object must be one-to-one. If two or
more identically named entities, such as a block state and a signal, map to the
same signal object, the name conflict will be flagged as an error at code
generation time.
5-55
5
Working with Data Structures
Summary of State Storage Class Options
Table 5-8 shows, for each state storage class option, the variable declaration
and MdlInitialize code generated for a Discrete Filter block state. The block
state has the user-defined state name filt_state.
Table 5-8: State Properties Options and Generated Code
Storage
Class
5-56
Declaration
Code
Auto
typedef struct D_Work_tag {
real_T filt_state;
struct {
int_T ClockTicksCounter;
} DiscPulse_IWORK;
} D_Work;
(declared in model.h)
.
.
/* Data Type Work (DWork)
Structure */
D_Work rtDWork;
(declared in model.c)
rtDWork.filt_state = 0.0;
Exported
Global
extern real_T filt_state;
(declared in model_private.h)
filt_state = 0.0;
Imported
Extern
extern real_T filt_state;
(declared in model_private.h)
filt_state = 0.0;
Imported
Extern
Pointer
extern real_T *filt_state;
(declared in model_private.h)
*(filt_state) = 0.0;
Storage Classes for Data Store Memory Blocks
Storage Classes for Data Store Memory Blocks
You can control how Data Store Memory blocks in your model are stored and
represented in the generated code by assigning storage classes and type
qualifiers. You do this in almost exactly the same way you assign storage
classes and type qualifiers for block states.
Data Store Memory blocks, like block states, have Auto storage class by
default, and their memory is stored within the DWork vector. The symbolic
name of the storage location is based on the block name.
Note that you can generate code from multiple Data Store Memory blocks that
have the same name, subject to the following restriction: at most one of the
identically-named blocks can have a storage class other than Auto. An error
will be reported if this condition is not met. For blocks with Auto storage class,
Real-Time Workshop generates a unique symbolic name for each block (if
necessary) to avoid name clashes. For blocks with non- Auto storage classes,
Real-Time Workshop simply uses the block name to generate the symbol.
To control the storage declaration for a Data Store Memory block, use the RTW
storage class and RTW storage type qualifier fields of the Data Store
Memory block parameters dialog.
In the following block diagram, a Data Store Write block writes to memory
declared by the Data Store Memory block myData.
Data Store Memory blocks are nonvirtual, as code is generated for their
initialization, and declarations in model header files. The Data Store Memory
block parameter dialog is shown next. Note that it documents which blocks
write to and read from it.
5-57
5
Working with Data Structures
Table 5-9 shows code generated for the Data Store Memory block in this model.
The table gives the variable declarations and MdlOutputs code generated for
the myData block.
5-58
Storage Classes for Data Store Memory Blocks
Table 5-9: Storage Class Options for Data Store Memory Blocks and Generated Code
Storage
Class
Declaration
Code
Auto
typedef struct D_Work_tag {
real_T myData;
} D_Work;
(declared in model.h)
.
.
/* Data Type Work (DWork)
Structure */
D_Work rtDWork;
(declared in model.c)
rtDWork.myData = rtb_Sine_Wave;
Exported
Global
extern real_T myData;
(declared in model_private.h)
myData = rtb_Sine_Wave;
Imported
Extern
extern real_T myData;
(declared in model_private.h)
myData = rtb_Sine_Wave;
Imported
Extern
Pointer
extern real_T *myData;
(declared in model_private.h)
*(myData) = rtb_Sine_Wave;
Data Store Memory and Simulink Signal Objects
If you are not familiar with Simulink data objects and signal objects, you
should read “Simulink Data Objects and Code Generation” on page 5–32 before
reading this section.
You can associate a Data Store Memory block with a signal object, and control
code generation for the block through the signal object. To do this:
1 Instantiate the desired signal object, and set its RTWInfo.StorageClass
property as you require.
5-59
5
Working with Data Structures
2 Open the block parameters dialog box for the Data Store Memory block
whose state you want to associate with the signal object. Enter the name of
the signal object into the Data store name field.
3 Make sure that the storage class and type qualifier settings of the block
parameters dialog box are compatible with those of the signal object. See
“Resolving Conflicts in Configuration of Parameter and Signal Objects” on
page 5-43.
4 Click Apply and close the dialog box.
Note When associating a Data Store Memory block with a signal object, the
mapping between the Data store name and the signal object name must be
one-to-one. If two or more identically named entities map to the same signal
object, the name conflict will be flagged as an error at code generation time.
5-60
6
External Mode
In external mode, Real-Time Workshop establishes a communications link between a model running
in Simulink and code executing on a target system. Further details on external mode are provided
elsewhere in this documentation: Chapter 14, “Creating an External Mode Communication Channel”
contains advanced information for those who want to implement their own external mode
communications layer. You may want to read it to gain increased insight into the architecture and
code structure of external mode communications. In addition, Chapter 12, “Targeting Tornado for
Real-Time Applications” discusses the use of external mode in the VxWorks Tornado environment.
The following discussion of external mode covers these major topics:
Introduction (p. 6-2)
Summary of external mode features and architecture
Using the External Mode User
Interface (p. 6-3)
Describes all elements of the external mode user interface
External Mode Compatible Blocks and
Subsystems (p. 6-19)
Types of blocks that receive and view signals in external
mode
External Mode Communications
Overview (p. 6-23)
Summary of the communications process between
Simulink and the target program
The TCP/IP Implementation (p. 6-26)
Features, bundled targets, and techniques for using
external mode protocol via TCP/IP
Limitations of External Mode (p. 6-33)
External mode restrictions imposed by the structure of a
model
6
External Mode
Introduction
External mode allows two separate systems — a host and a target — to
communicate. The host is the computer where MATLAB and Simulink are
executing. The target is the computer where the executable created by
Real-Time Workshop runs.
The host (Simulink) transmits messages requesting the target to accept
parameter changes or to upload signal data. The target responds by executing
the request. External mode communication is based on a client/server
architecture, in which Simulink is the client and the target is the server.
External mode lets you:
• Modify, or tune, block parameters in real time. In external mode, whenever
you change parameters in the block diagram, Simulink automatically
downloads them to the executing target program. This lets you tune your
program’s parameters without recompiling. In external mode, the Simulink
model becomes a graphical front end to the target program.
• View and log block outputs in many types of blocks and subsystems. You can
monitor and/or store signal data from the executing target program, without
writing special interface code. You can define the conditions under which
data is uploaded from target to host. For example, data uploading could be
triggered by a selected signal crossing zero in a positive direction.
Alternatively, you can manually trigger data uploading.
External mode works by establishing a communication channel between
Simulink and code generated by Real-Time Workshop. This channel is
implemented by a low-level transport layer that handles physical transmission
of messages. Both Simulink and the generated model code are independent of
this layer. The transport layer and the code directly interfacing to the transport
layer are isolated in separate modules that format, transmit, and receive
messages and data packets.
This design makes it possible for different targets to use different transport
layers. For example, the GRT, GRT malloc, ERT, and Tornado targets support
host/target communication via TCP/IP, whereas the xPC Target supports both
RS232 (serial) and TCP/IP communication. The Real-Time Windows Target
implements external mode communication via shared memory.
6-2
Using the External Mode User Interface
Using the External Mode User Interface
This section discusses the elements of the Simulink and Real-Time Workshop
user interface that control the operation of external mode. These elements
include:
• External mode related menu items in Simulation and Tools menus and in
the Simulink toolbar.
• External Mode Control Panel
• Target Interface Dialog Box
• External Signal & Triggering Dialog Box
• Data Archiving Dialog Box
External Mode Related Menu and Toolbar Items
To communicate with a target program, the model must be operating in
external mode. The Simulation menu and the toolbar provide two ways to
enable external mode:
• Select External from the Simulation menu.
• Select External from the simulation mode menu in the toolbar. The
simulation mode menu is shown in this picture.
Simulation mode menu
Once external mode is enabled, you can use the Simulation menu or the
toolbar to connect to and control the target program.
6-3
6
External Mode
Note You can enable external mode, and simultaneously connect to the
target system, by using the External Mode Control Panel. See “External
Mode Control Panel” on page 6-8.
Simulation Menu
When Simulink is in external mode, the upper section of the Simulation menu
contains external mode options. Initially, Simulink is disconnected from the
target program, and the menu displays the options shown in this picture.
Figure 6-1: Simulation Menu External Mode Options
(Host Disconnected from Target)
The Connect to target option establishes communication with the target
program. When a connection is established, the target program may be
executing model code, or it may be awaiting a command from the host to start
executing model code.
If the target program is executing model code, the Simulation menu contents
change, as shown in this picture.
Figure 6-2: Simulation Menu External Mode Options
(Target Executing Model Code)
6-4
Using the External Mode User Interface
The Disconnect from target option disconnects Simulink from the target
program, which continues to run. The Stop real-time code option terminates
execution of the target program and disconnects Simulink from the target
system.
If the target program is in a wait state, the Start real-time code option is
enabled, as shown in this picture. The Start real-time code option instructs
the target program to begin executing the model code.
Figure 6-3: Simulation Menu External Mode Options
(Target Awaiting Start Command)
Toolbar Controls
The Simulink toolbar controls, shown in Figure 6-4, let you control the same
external mode functions as the Simulation menu. Simulink displays external
mode icons to the left of the Simulation mode menu. Initially, the toolbar
displays a Connect to target icon and a disabled Start real-time code button
(shown in Figure 6-4). Click on the Connect to target icon to connect Simulink
to the target program.
6-5
6
External Mode
Simulation mode menu
Connect to target icon
Start real-time code button
(disabled)
Figure 6-4: External Mode Toolbar Controls (Host Disconnected from Target)
When a connection is established, the target program may be executing model
code, or it may be awaiting a command from the host to start executing model
code.
If the target program is executing model code, the toolbar displays a Stop
real-time code button and a Disconnect from target icon (shown in
Figure 6-5). Click on the Stop real-time code button to command the target
program to stop executing model code and disconnect Simulink from the target
system. Click on the Disconnect from target icon to disconnect Simulink from
the target program while leaving the target program running.
6-6
Using the External Mode User Interface
Disconnect from target icon
Stop real-time code button
Figure 6-5: External Mode Toolbar Controls (Target Executing Model Code)
If the target program is in a wait state, the toolbar displays a Start real-time
code button and a Disconnect from target icon (shown in Figure 6-6). Click
on the Start real-time code button to instruct the target program to start
executing model code. Click on the Disconnect from target icon to disconnect
Simulink from the target program.
6-7
6
External Mode
Disconnect from target icon
Start real-time code button
Figure 6-6: External Mode Toolbar Controls (Target in Wait State)
External Mode Control Panel
The External Mode Control Panel provides centralized control of all external
mode features, including:
• Host/target connection, disconnection, and target program start/stop
functions, and enabling of external mode
• Arming and disarming the data upload trigger
• External mode communications configuration
• Timing of parameter downloads
• Selection of signals from the target program to be viewed and monitored on
the host
• Configuration of data archiving features
6-8
Using the External Mode User Interface
Select External mode control panel from the Simulink Tools menu to open
the External Mode Control Panel.
These buttons control the connection between
host and manual arming of the data uploading
trigger.
This check box and button control the timing of
parameter downloads.
These buttons open dialog boxes that configure
external mode target interface, signal
properties, and data archiving.
The following sections describe the features supported by the External Mode
Control Panel.
Connection and Start/Stop Controls
The External Mode Control Panel performs the same connect/disconnect and
start/stop functions found in the Simulation menu and the Simulink toolbar
(see “External Mode Related Menu and Toolbar Items” on page 6-3.)
The Connect/Disconnect button connects to or disconnects from the target
program. The button text changes in accordance with the state of the
connection.
Note that if external mode is not enabled at the time the Connect button is
clicked, the External Mode Control Panel enables external mode
automatically.
The Start/Stop real-time code button commands the target to start or
terminate model code execution. The button is disabled until a connection to
the target is established. The button text changes in accordance with the state
of the target program.
6-9
6
External Mode
Target Interface Dialog Box
Pressing the Target Interface button activates the External Target
Interface dialog box.
Specify name of external interface MEX-file here.
Default is ext_comm.
Enter optional arguments to the external interface MEX-file here.
The External Target Interface dialog box lets you specify the name of a
MEX-file that implements host/target communications. This is known as the
external interface MEX-file. The fields of the External Target Interface dialog
box are:
• MEX-file for external interface: Name of the external interface MEX-file.
The default is ext_comm, the TCP/IP-based external interface file provided
for use with the GRT, GRT malloc, ERT, and Tornado targets
Custom or third-party targets may use a different external interface
MEX-file.
• MEX-file arguments: Arguments for the external interface MEX-file. For
example, ext_comm allows three optional arguments: the network name of
your target, the verbosity level, and a TCP/IP server port number.
See “The External Interface MEX-File” on page 6-28 for details on ext_comm
and its arguments.
6-10
Using the External Mode User Interface
External Signal & Triggering Dialog Box
Clicking the Signal & triggering button activates the External Signal &
Triggering dialog box.
Figure 6-7: Default Settings of the External Signal & Triggering Dialog Box
The External Signal & Triggering dialog box displays a list of all blocks and
subsystems in your model that support external mode signal uploading. See
“External Mode Compatible Blocks and Subsystems” on page 6-19 for
information on which types of blocks are external mode compatible.
The External Signal & Triggering dialog box lets you select which signals are
collected from the target system and viewed in external mode. It also lets you
select a signal that triggers uploading of data when certain signal conditions
are met, and define the triggering conditions.
6-11
6
External Mode
Default Operation
Figure 6-7 shows the default settings of the External Signal and Triggering
dialog box. The default operation of the External Signal and Triggering
dialog box is designed to simplify monitoring the target program. If you use the
default settings, you do not need to preconfigure signals and triggers. Simply
start the target program and connect the Simulink model to it. All external
mode compatible blocks will be selected and the trigger will be armed. Signal
uploading will begin immediately upon connection to the target program.
The default configuration is:
• Arm when connect to target: on
• Trigger Mode: normal
• Trigger Source: manual
• Select all: on
Signal Selection
All external mode compatible blocks in your model appear in the Signal
selection list of the External Signal & Triggering dialog box. You use this list
to select signals to be viewed. An X appears to the left of each selected block’s
name.
The Select all check box selects all signals. By default, Select all is on.
If Select all is off, you can select or deselect individual signals using the on and
off radio buttons. To select a signal, click on the desired list entry and click the
on radio button. To deselect a signal, click on the desired list entry and click
the off radio button. Alternatively, you can double-click a signal in the list to
toggle between selection and deselection.
The Clear all button deselects all signals.
Trigger Options
The Trigger panel located at the bottom left of the External Signal &
Triggering dialog box contains options that control when and how signal data
is collected (uploaded) from the target system. These options are:
6-12
Using the External Mode User Interface
• Source: manual or signal. Selecting manual directs external mode to start
logging data when the Arm trigger button on the External Mode Control
Panel is clicked.
Selecting signal tells external mode to start logging data when a selected
trigger signal satisfies trigger conditions specified in the Trigger signal
panel. When the trigger conditions are satisfied (that is, the signal crosses
the trigger level in the specified direction) a trigger event occurs. If the
trigger is armed, external mode monitors for the occurrence of a trigger
event. When a trigger event occurs, data logging begins.
• Arm when connect to target: If this option is selected, external mode arms
the trigger automatically when Simulink has connected to the target. If the
trigger source is manual, uploading begins immediately. If the trigger mode
is signal, monitoring of the trigger signal begins immediately, and
uploading begins upon the occurrence of a trigger event.
If Arm when connect to target is not selected, you must manually arm the
trigger by clicking the Arm trigger button in the External Mode Control
Panel.
• Duration: The number of base rate steps for which external mode logs data
after a trigger event. For example, if the fastest rate in the model is 1 second
and a signal sampled at 1 Hz is being logged for a duration of 10 seconds,
then external mode will collect 10 samples. If a signal sampled at 2 Hz is
logged, only 5 samples will be collected.
• Mode: normal or one-shot. In normal mode, external mode automatically
rearms the trigger after each trigger event. In one-shot mode, external mode
collects only one buffer of data each time you arm the trigger. See “Data
Archiving Dialog Box” on page 6-15 for further details on the effect of the
Mode setting.
• Delay: The delay represents the amount of time that elapses between a
trigger occurrence and the start of data collection. The delay is expressed in
base rate steps, and can be positive or negative. A negative delay corresponds
to pretriggering. When the delay is negative, data from the time preceding
the trigger is collected and uploaded.
Trigger Signal Selection
You can designate one signal as a trigger signal. To select a trigger signal,
select signal from the Trigger Source menu. This activates the Trigger
6-13
6
External Mode
signal panel (see Figure 6-8). Then, click on the desired entry in the Signal
selection list, and click the Trigger signal button.
When a signal is selected as a trigger, a T appears to the left of the block’s name
in the Signal selection list. In Figure 6-8, the Pilot G force Scope signal is
the trigger. Pilot G force Scope is also selected for viewing, as indicated by
the X to the left of the block name.
The Trigger Signal panel
Figure 6-8: Signals & Triggering Window with Trigger Selected
After selecting the trigger signal, you can define the trigger conditions in the
Trigger signal panel, and set the Port and Element fields located on the right
side of the Trigger panel.
6-14
Using the External Mode User Interface
Setting Trigger Conditions
Note The Trigger signal panel and the Port and Element fields of the
External Signal & Trigger dialog box are enabled only when Trigger source
is set to signal.
By default, any element of the first input port of the specified trigger block can
cause the trigger to fire (i.e., Port 1, any element). You can modify this behavior
by adjusting the Port and Element fields located on the right side of the
Trigger panel. The Port field accepts a number or the keyword last. The
Element field accepts a number or the keywords any and last.
The Trigger Signal panel defines the conditions under which a trigger event
will occur. These are:
• Level: Specifies a threshold value. The trigger signal must cross this value
in a designated direction to fire the trigger. By default, the level is 0.
• Direction: rising, falling, or either. This specifies the direction in which
the signal must be travelling when it crosses the threshold value. The default
is rising.
• Hold-off: Applies only to normal mode. Expressed in base rate steps,
Hold-off is the time between the termination of one trigger event and the
rearming of the trigger.
Data Archiving Dialog Box
Pressing the Data Archiving button of the External Mode Control Panel
opens the External Data Archiving dialog box.
6-15
6
External Mode
This panel supports the following features:
Directory Notes. Use this option to add annotations that pertain to a collection of
related data files in a directory.
Pressing the Edit directory note button opens the MATLAB editor. Place
comments that you want saved to a file in the specified directory in this
window. By default, the comments are saved to the directory last written to by
data archiving.
File Notes. Pressing Edit file note opens a file finder window that is, by default,
set to the last file to which you have written. Selecting any MAT-file opens an
edit window. Add or edit comments in this window that you want saved with
your individual MAT-file.
Data Archiving. Clicking the Enable Archiving check box activates the
automated data archiving features of external mode. To understand how the
archiving features work, it is necessary to consider the handling of data when
archiving is not enabled. There are two cases, one-shot and normal mode.
In one-shot mode, after a trigger event occurs, each selected block writes its
data to the workspace just as it would at the end of a simulation. If another
one-shot is triggered, the existing workspace data will be overwritten.
In normal mode, external mode automatically rearms the trigger after each
trigger event. Consequently, you can think of normal mode as a series of
one-shots. Each one-shot in this series, except for the last, is referred to as an
intermediate result. Since the trigger can fire at any time, writing intermediate
results to the workspace generally results in unpredictable overwriting of the
workspace variables. For this reason, the default behavior is to write only the
results from the final one-shot to the workspace. The intermediate results are
discarded. If you know that sufficient time exists between triggers for
inspection of the intermediate results, then you can override the default
behavior by checking the Write intermediate results to workspace check box.
Note that this option does not protect the workspace data from being
overwritten by subsequent triggers.
The options in the External Data Archiving dialog box support automatic
writing of logging results, including intermediate results, to disk. Data
archiving provides the following settings:
6-16
Using the External Mode User Interface
• Directory: Specifies the directory in which data is saved. External mode
appends a suffix if you select Increment directory when trigger armed.
• File: Specifies the filename in which data is saved. External mode appends
a suffix if you select Increment file after one-shot.
• Increment directory when trigger armed: External mode uses a different
directory for writing log files each time that you press the Arm trigger
button. The directories are named incrementally; for example: dirname1,
dirname2, and so on.
• Increment file after one-shot: New data buffers are saved in incremental
files: filename1, filename2, etc. Note that this happens automatically in
normal mode.
• Append file suffix to variable names: Whenever external mode increments
filenames, each file contains variables with identical names. Choosing
Append file suffix to variable name results in each file containing unique
variable names. For example, external mode will save a variable named
xdata in incremental files (file_1, file_2, etc.) as xdata_1, xdata_2, and so
on. This is useful if you want to load the MAT-files into the workspace and
compare variables in MATLAB. Without the unique names, each instance of
xdata would overwrite the previous one in the MATLAB workspace.
This picture shows the External Data Archiving dialog box with archiving
enabled.
Unless you select Enable archiving, entries for the Directory and File fields
are not accepted.
6-17
6
External Mode
Parameter Download Options
The batch download check box on the External Mode Control Panel enables
or disables batch parameter changes.
By default, batch download is not enabled. When batch download is not
enabled, changes made directly to block parameters are sent immediately to
the target. Changes to MATLAB workspace variables are sent when an
Update diagram is performed.
When batch download is enabled, the Download button is enabled. Changes
made to block parameters are stored locally until you click the Download
button. When you click the Download button, the changes are sent in a single
transmission.
When parameter changes have been made and are awaiting batch download,
the External Mode Control Panel displays the message Parameter changes
pending... to the right of the download button. (See Figure 6-9.) This message
disappears after Simulink receives notification from the target that the new
parameters have been installed into the parameter vector of the target system.
Figure 6-9 shows the External Mode Control Panel with the batch download
option activated.
Parameter changes pendin g... message
appears if unsent parameter value changes
are awaiting download.
Figure 6-9: External Mode Control Panel in Batch Download Mode
6-18
External Mode Compatible Blocks and Subsystems
External Mode Compatible Blocks and Subsystems
Compatible Blocks
In external mode, you can use the following types of blocks to receive and view
signals uploaded from the target program:
• Scope blocks
• Blocks in the Dials & Gauges Blockset
• Display blocks
• To Workspace blocks
• User-written S-Function blocks
An external mode method has been added to the S-function API. This method
allows user-written blocks to support external mode. See
matlabroot/simulink/simstruc.h.
• XY Graph blocks
In addition to these types of blocks, you can designate certain subsystems as
Signal Viewing Subsystems and use them to receive and view signals uploaded
from the target program. See “Signal Viewing Subsystems” on page 6-19 for
further information.
External mode compatible blocks and subsystems are selected, and the trigger
is armed, via the External Signal and Triggering dialog box. For example,
Figure 6-7 shows two Scope blocks, a Display block, and a Signal Viewing
Subsystem (theSink). All of these are selected and the trigger is set to be armed
when connected to the target program.
Signal Viewing Subsystems
A Signal Viewing Subsystem is an atomic subsystem that encapsulates
processing and viewing of signals received from the target system. A Signal
Viewing Subsystem runs only on the host, generating no code in the target
system. Signal Viewing Subsystems run in all simulation modes — normal,
accelerated, and external.
Signal Viewing Subsystems are useful in situations where you want to process
or condition signals before viewing or logging them, but you do not want to
perform these tasks on the target system. By using a Signal Viewing
6-19
6
External Mode
Subsystem, you can generate smaller and more efficient code on the target
system.
Like other external mode compatible blocks, Signal Viewing Subsystems are
displayed in the External Signal and Triggering dialog box.
To declare a subsystem to be a Signal Viewing Subsystem:
1 Select the Treat as atomic unit option in the Block Parameters dialog box.
See “Nonvirtual Subsystem Code Generation” on page 4-2 for further
information on atomic subsystems.
2 Use the following set_param command to turn the SimViewingDevice
property on.
set_param('blockname', 'SimViewingDevice','on')
where 'blockname' is the name of the subsystem.
3 Make sure the subsystem meets the following requirements:
- It must be a pure sink block. That is, it must contain no Outport blocks or
Data Store blocks. It may contain Goto blocks only if the corresponding
from blocks are contained within the subsystem boundaries.
- It must have no continuous states.
The model shown below, sink_examp, contains an atomic subsystem, theSink.
The subsystem theSink, shown below, applies a gain and an offset to its input
signal, and displays it on a Scope block.
6-20
External Mode Compatible Blocks and Subsystems
If theSink is declared as a Signal Viewing Subsystem, the generated target
program includes only the code for the Sine Wave block. If theSink is selected
and armed in the External Signal and Triggering dialog box (as shown in
Figure 6-10), the target program uploads the sine wave signal to theSink
during simulation.You can then modify the parameters of the blocks within
theSink and observe their effect upon the uploaded signal.
Figure 6-10: Signal Viewing Subsystem Selected in External
Signals & Triggering Dialog Box
6-21
6
External Mode
Note that if theSink were not declared as a Signal Viewing Subsystem, its
Gain, Constant, and Sum blocks would run as subsystem code on the target
system. The Sine Wave signal would be uploaded to Simulink after being
processed by these blocks, and viewed on sink_examp/theSink/Scope2.
Processing demands on the target system would be increased by the additional
signal processing, and by the downloading of block parameter changes from the
host.
6-22
External Mode Communications Overview
External Mode Communications Overview
This section describes how Simulink and the target program communicate, and
how and when they transmit parameter updates and signal data to each other.
Depending on the setting of the Inline parameters option when the target
program is generated, there are differences in the way parameter updates are
handled. “The Download Mechanism” on page 6-23 describes the operation of
external mode communications with Inline parameters off. “Inlined and
Tunable Parameters” on page 6-24 describes the operation of external mode
with Inline parameters on.
The Download Mechanism
In external mode, Simulink does not simulate the system represented by the
block diagram. By default, when external mode is enabled, Simulink
downloads current values of all parameters to the target system. After the
initial download, Simulink remains in a waiting mode until you change
parameters in the block diagram or until Simulink receives data from the
target.
When you change a parameter in the block diagram, Simulink calls the
external interface MEX-file, passing new parameter values (along with other
information) as arguments.
The external interface MEX-file contains code that implements one side of the
interprocess communication (IPC) channel. This channel connects the
Simulink process (where the MEX-file executes) to the process that is executing
the external program. The MEX-file transfers the new parameter values via
this channel to the external program.
The other side of the communication channel is implemented within the
external program. This side writes the new parameter values into target’s
parameter structure (rtP).
The Simulink side initiates the parameter download operation by sending a
message containing parameter information to the external program. In the
terminology of client/server computing, the Simulink side is the client and the
external program is the server. The two processes can be remote, or they can
be local. Where the client and server are remote, a protocol such as TCP/IP is
used to transfer data. Where the client and server are local, shared memory can
be used to transfer data.
6-23
6
External Mode
The following diagram illustrates this relationship
Simulink calls the external interface MEX-file whenever you change
parameters in the block diagram. The MEX-file then downloads the
parameters to the external program via the communication channel.
.
Simulink Process
External
Program
Process
mexFunction
External Program
Client
IPC Code
External Interface
MEX-file (ext_comm)
Server
IPC Code
ext_svr
Interprocess Communication Channel
Figure 6-11: External Mode Architecture
Inlined and Tunable Parameters
By default, all parameters (except those listed in “Limitations of External
Mode” on page 6-33) in an external mode program are tunable; that is, you can
change them via the download mechanism described in this section.
If you select the Inline parameters option (on the Advanced page of the
Simulation Parameters dialog box), Real-Time Workshop embeds the
numerical values of model parameters (constants), instead of symbolic
6-24
External Mode Communications Overview
parameter names, in the generated code. Inlining parameters generates
smaller and more efficient code. However, inlined parameters, since they are
effectively transformed into constants, are not tunable.
Real-Time Workshop lets you improve overall efficiency by inlining most
parameters, while at the same time retaining the flexibility of run-time tuning
for selected parameters that are important to your application. When you
inline parameters, you can use the Model Parameter Configuration dialog to
remove individual parameters from inlining and declare them to be tunable. In
addition, the Model Parameter Configuration dialog offers you options for
controlling how parameters are represented in the generated code.
For further information on tunable parameters please see “Parameters:
Storage, Interfacing, and Tuning” on page 5-2.
Automatic Parameter Uploading on Host/Target Connection
Each time Simulink connects to a target program that was generated with
Inline parameters on, the target program uploads the current value of its
tunable parameters (if any) to the host. These values are assigned to the
corresponding MATLAB workspace variables. This procedure ensures that the
host and target are synchronized with respect to parameter values.
All workspace variables required by the model must be defined to an initial
value at the time of host/target connection. Otherwise the uploading cannot
proceed and an error will result. Once the connection is made, these variables
are updated to reflect the current parameter values on the target system.
Note that automatic parameter uploading takes place only if the target
program was generated with Inline parameters on. “The Download
Mechanism” on page 6-23 describes the operation of external mode
communications with Inline parameters off.
6-25
6
External Mode
The TCP/IP Implementation
Real-Time Workshop provides code to implement both the client and server
side based on TCP/IP. You can use the socket-based external mode
implementation provided by Real-Time Workshop with the generated code,
provided that your target system supports TCP/IP.
A low-level transport layer handles physical transmission of messages. Both
Simulink and the model code are independent of this layer. Both the transport
layer and code directly interfacing to the transport layer are isolated in
separate modules that format, transmit, and receive messages and data
packets.
This design makes it possible for different targets to use different transport
layers. For example, the GRT, GRT malloc, ERT, and Tornado targets support
host/target communication via TCP/IP, whereas the xPC target supports both
RS232 (serial) and TCP/IP communication.
Using the TCP/IP Implementation
This section discusses how to use the TCP/IP-based client/server
implementation of external mode with real-time programs on a UNIX or PC
system. Chapter 12, “Targeting Tornado for Real-Time Applications”
illustrates the use of external mode in the Tornado environment.
In order to use Simulink external mode, you must:
• Specify the name of the external interface MEX-file in the External Target
Interface dialog box. By default, this is ext_comm.
• Configure the template makefile so that it links the proper source files for
the TCP/IP server code and defines the necessary compiler flags when
building the generated code.
• Build the external program.
• Run the external program.
• Set Simulink to external mode and connect to the target.
6-26
The TCP/IP Implementation
This figure shows the structure of the TCP/IP-based implementation.
UNIX or PC Host
Target
Simulink in External Mode
Target Code
Process block
parameter
ext_svr.c
Update block
ext_comm
TCP/IP on Ethernet
External Mode Message Format
header data in target format
Figure 6-12: TCP/IP-Based Client/Server Implementation for External Mode
The following sections discuss the details of how to use the external mode of
Simulink.
6-27
6
External Mode
The External Interface MEX-File
You must specify the name of the external interface MEX-file in the External
Target Interface dialog box.
Enter the name of the external interface MEX-file in
the box (you do not need to enter the .mex
extension). This file must be in the current
directory or in a directory that is on your MATLAB
path.
The default external interface MEX-file is ext_comm. ext_comm implements
TCP/IP-based communications. ext_comm has three optional arguments,
discussed in the next section.
MEX-File Optional Arguments
In the External Target Interface dialog box, you can specify optional
comma-delimited arguments that are passed to the MEX-file. These are:
• Target network name: the network name of the computer running the
external program. By default, this is the computer on which Simulink is
running. The name can be:
- a string delimited by single quotes, such as 'myPuter'
- an IP address delimited by single quotes, such as '148.27.151.12'
• Verbosity level: controls the level of detail of the information displayed
during the data transfer. The value is either 0 or 1 and has the following
meaning:
0 — no information
1 — detailed information
• TCP/IP server port number: The default value is 17725. You can change the
port number to a value between 256 and 65535 to avoid a port conflict if
necessary.
6-28
The TCP/IP Implementation
You must specify these options in order. For example, if you want to specify the
verbosity level (the second argument), then you must also specify the target
host name (the first argument).
Note that you can specify command line options to the external program. See
“Running the External Program” on page 6-29 for more information.
External Mode Compatible Targets
The ERT, GRT, GRT malloc, and Tornado targets support external mode. To
enable external mode code generation, check External mode in the
target-specific code generation options section of the Real-Time Workshop
pane. The following illustration shows the GRT code generation options with
external mode enabled.
Running the External Program
The external program must be running before you can use Simulink in external
mode. To run the external program, you type a command of the form
model -opt1 ... -optN
where model is the name of the external program and -opt1 ... -optN are
options. (See “Command Line Options for the External Program” on page 6–
6-29
6
External Mode
31). In the examples in this section, we assume the name of the external
program to be ext_example.
Running the External Program Under Windows
In the Windows environment, you can run the external programs in either of
the following ways:
• Open a Command Prompt window. At the command prompt, type the name
of the target executable, followed by any options, as in the following example.
ext_example -tf inf -w
• Alternatively, you can launch the target executable from the MATLAB
command prompt. In this case the command must be preceded by an
exclamation point (!) and followed by an ampersand (&) , as in the following
example.
!ext_example -tf inf -w &
Note that the ampersand (&) causes the operating system to spawn another
process to run the target executable.
Running the External Program Under UNIX
In the UNIX environment, you can run the external programs in either of the
following ways:
• Open an an Xterm window. At the command prompt, type the name of the
target executable, followed by any options, as in the following example.
ext_example -tf inf -w
• Alternatively, you can launch the target executable from the MATLAB
command prompt. In the UNIX environment, if you start the external
program from MATLAB, you must run it in the background so that you can
still access Simulink. The command must be preceded by an exclamation
point (!) and followed by an ampersand (&) , as in the following example.
!ext_example -tf inf -w &
runs the executable from MATLAB by spawning another process to run it.
6-30
The TCP/IP Implementation
Command Line Options for the External Program
External mode target executables generated by Real-Time Workshop support
the following command line options:
• -tf n option
The -tf option overrides the stop time set for the model in Simulink. The
argument n specifies the number of seconds the program will run. The value
inf directs the model to run indefinitely. In this case, the model code will run
until the target program receives a stop message from Simulink.
The following example sets the stop time to 10 seconds.
ext_example -tf 10
Note You may use the -tf option with GRT, GRT malloc, ERT, and Tornado
targets. If you are implementing a custom target and want to support the -tf
option, you must implement the option yourself. See “Creating an External
Mode Communication Channel” on page 14–94 for further information.
• -w option
The -w option instructs the target program to enter a wait state until it
receives a message from the host. At this point, the target is running, but not
executing the model code. The start message is sent when you select Start
real-time code from the Simulation menu or click the Start real-time code
button in the External Mode Control Panel.
Use the -w option if you want to view data from time step 0 of the target
program execution, or if you want to modify parameters before the target
program begins execution of model code.
• -port n option
the -port option specifies the TCP/IP port number, n, for the target program.
The port number of the target program must match that of the host. The
default port number is 17725. The port number must be a value between 256
and 65535.
6-31
6
External Mode
Note The -w and -port options are supported by the TCP/IP transport layer
modules shipped with Real-Time Workshop. By default, these modules are
linked into external mode target executables. If you are implementing a
custom external mode transport layer and want to support these options, you
must implement them in your code. See “Creating an External Mode
Communication Channel” on page 14–94 for further information. See
matlabroot/rtw/c/src/ext_transport.c for example code.
Error Conditions
If the Simulink block diagram does not match the external program, Simulink
displays an error box informing you that the checksums do not match (i.e., the
model has changed since you generated code). This means you must rebuild the
program from the new block diagram (or reload the correct one) in order to use
external mode.
If the external program is not running, Simulink displays an error informing
you that it cannot connect to the external program.
Implementing an External Mode Protocol Layer
If you want to implement your own transport layer for external mode
communication, you must modify certain code modules provided by Real-Time
Workshop, and rebuild ext_comm, the external interface MEX-file. This
advanced topic is described in detail in “Creating an External Mode
Communication Channel” on page 14–94.
6-32
Limitations of External Mode
Limitations of External Mode
In general, you cannot change a parameter if doing so results in a change in the
structure of the model. For example, you cannot change:
• The number of states, inputs, or outputs of any block
• The sample time or the number of sample times
• The integration algorithm for continuous systems
• The name of the model or of any block
• The parameters to the Fcn block
If you cause any of these changes to the block diagram, then you must rebuild
the program with newly generated code.
However, parameters in transfer function and state space representation
blocks can be changed in specific ways:
• The parameters (numerator and denominator polynomials) for the Transfer
Fcn (continuous and discrete) and Discrete Filter blocks can be changed (as
long as the number of states does not change).
• Zero entries in the State Space and Zero Pole (both continuous and discrete)
blocks in the user-specified or computed parameters (i.e., the A, B, C, and D
matrices obtained by a zero-pole to state-space transformation) cannot be
changed once external simulation is started.
• In the State Space blocks, if you specify the matrices in the controllable
canonical realization, then all changes to the A, B, C, D matrices that
preserve this realization and the dimensions of the matrices are allowed.
6-33
6
External Mode
6-34
7
Program Architecture
Code is generated by Real-Time Workshop in two styles, depending whether a target is embedded or
not. In addition, the structure of code is affected by whether a multitasking environment is available
for execution, and on what system and applications modules must be incorporated. The following
sections describe these architectural distinctions:
Introduction (p. 7-2)
Code styles and targets appropriate for development of
rapid prototyping and embedded systems
Model Execution (p. 7-4)
How code generated from models executes, including
singletasking and multitasking execution, timing, data
structures, entry points, and differences between rapid
prototyping and embedded code
Rapid Prototyping Program
Framework (p. 7-23)
Overal architecture and individual components of
programs generated by rapid prototyping targets
Embedded Program Framework
(p. 7-34)
Overview of the architecture of programs generated by
the Real-Time Workshop Embedded Coder.
For a detailed discussion of the structure of embedded real-time code, see the Real-Time Workshop
Embedded Coder documentation.
7
Program Architecture
Introduction
Real-Time Workshop generates two styles of code. One code style is suitable for
rapid prototyping (and simulation via code generation). The other style is
suitable for embedded applications. This chapter discusses the program
architecture, that is, the structure of code generated by Real-Time Workshop,
associated with these two styles of code. The table below classifies the targets
shipped with Real-Time Workshop. For related details about code style and
target characteristics, see “Choosing a Code Format for Your Application” on
page 3-3.
Table 7-1: Code Styles Listed By Target
7-2
Target
Code Style (using C unless noted)
Real-Time Workshop
Embedded Coder target
Embedded — useful as a starting point
when using the generated C code in an
embedded application.
Generic real-time (GRT)
target
Rapid prototyping — nonreal-time
simulation on your workstation. Useful
as a starting point for creating a rapid
prototyping real-time target that does not
use real-time operating system tasking
primitives. Also useful for validating the
generated code on your workstation.
Real-time malloc target
Rapid prototyping — very similar to the
generic real-time (GRT) target except
that this target allocates all model
working memory dynamically rather
than statically declaring it in advance.
Rapid simulation target
Rapid prototyping — nonreal-time
simulation of your model on your
workstation. Useful as a high-speed or
batch simulation tool.
S-function target
Rapid prototyping — creates a C-MEX
S-function for simulation of your model
within another Simulink model.
Introduction
Table 7-1: Code Styles Listed By Target (Continued)
Target
Code Style (using C unless noted)
Tornado (VxWorks) real-time
target
Rapid prototyping — runs model in real
time using the VxWorks real-time
operating system tasking primitives. Also
useful as a starting point for targeting a
real-time operating system.
Real-Time Windows target
Rapid prototyping — runs model in
real-time at interrupt level while your PC
is running Microsoft Windows in the
background.
xPC target
Rapid prototyping — runs model in real
time on target PC running xPC kernel.
DOS real-time target
Rapid prototyping — runs model in real
time at interrupt level under DOS.
Third-party vendors supply additional targets for Real-Time Workshop.
Generally, these can be classified as rapid prototyping targets. For more
information about third-party products, see the MATLAB Connections Web
page: http://www.mathworks.com/products/connections.
This chapter is divided into three sections. The first section discusses model
execution; the second section discusses the rapid prototyping style of code; and
the third section discusses the embedded style of code.
7-3
7
Program Architecture
Model Execution
Before looking at the two styles of generated code, you need to have a high-level
understanding of how the generated model code is executed. Real-Time
Workshop generates algorithmic code as defined by your model. You may
include your own code into your model via S-functions. S-functions can range
from high-level signal manipulation algorithms to low-level device drivers.
Real-Time Workshop also provides a run-time interface that executes the
generated model code. The run-time interface and model code are compiled
together to create the model executable. The diagram below shows a high-level
object-oriented view of the executable.
Execution driver for model code,
operating system interface routines,
I/O dependent routines,
solver and data logging routines.
Model code
and S-functions
Run-Time Interface
Figure 7-1: The Object-Oriented View of a Real-Time Program
In general, the conceptual design of the model execution driver does not change
between the rapid prototyping and embedded style of generated code. The
following sections describe model execution for singletasking and multitasking
environments both for simulation (nonreal-time) and for real-time. For most
models, the multitasking environment will provide the most efficient model
execution (i.e., fastest sample rate).
The following concepts are useful in describing how models execute:
• Initialization — Initializing the run-time interface code and the model
code.
• ModelOutputs — Calling all blocks in your model that have a time hit at the
current point in time and having them produce their output. ModelOutputs
can be done in major or minor time steps. In major time steps, the output is
7-4
Model Execution
a given simulation time step. In minor time steps, the run-time interface
integrates the derivatives to update the continuous states.
• ModelUpdate — Calling all blocks in your model that have a sample hit at the
current point in time and having them update their discrete states or similar
type objects.
• ModelDerivatives — Calling all blocks in your model that have continuous
states and having them update their derivatives. ModelDerivatives is only
called in minor time steps.
The pseudocode below shows the execution of a model for a singletasking
simulation (nonreal-time).
main()
{
Initialization
While (time < final time)
ModelOutputs
-- Major time step.
LogTXY
-- Log time, states and root outports.
ModelUpdate
-- Major time step.
Integrate:
-- Integration in minor time step for
-- models with continuous states.
ModelDerivatives
Do 0 or more:
ModelOutputs
ModelDerivatives
EndDo (Number of iterations depends upon the solver.)
Integrate derivatives to update continuous states.
EndIntegrate
EndWhile
Shutdown
}
The initialization phase begins first. This consists of initializing model states
and setting up the execution engine. The model then executes, one step at a
time. First ModelOutputs executes at time t, then the workspace I/O data is
logged, and then ModelUpdate updates the discrete states. Next, if your model
has any continuous states, ModelDerivatives integrates the continuous states’
derivatives to generate the states for time t new = t + h , where h is the step
size. Time then moves forward to t new and the process repeats.
7-5
7
Program Architecture
During the ModelOutputs and ModelUpdate phases of model execution, only
blocks that have hit the current point in time execute. They determine if they
have hit by using a macro (ssIsSampleHit, or ssIsSpecialSampleHit) that
checks for a sample hit.
The pseudocode below shows the execution of a model for a multitasking
simulation (nonreal-time).
main()
{
Initialization
While (time < final time)
ModelOutputs(tid=0)
-- Major time step.
LogTXY
-- Log time, states, and root
-- outports.
ModelUpdate(tid=1)
-- Major time step.
Integrate
-- Integration in minor time step for
-- models with continuous states.
ModelDerivatives
Do 0 or more:
ModelOutputs(tid=0)
ModelDerivatives
EndDo (Number of iterations depends upon the solver.)
Integrate derivatives to update continuous states.
EndIntegrate
For i=1:NumTids
ModelOutputs(tid=i) -- Major time step.
ModelUpdate(tid=i) -- Major time step.
EndFor
EndWhile
Shutdown
}
The multitasking operation is more complex when compared with the
singletasking execution because the output and update functions are
subdivided by the task identifier (tid) that is passed into these functions. This
allows for multiple invocations of these functions with different task identifiers
using overlapped interrupts, or for multiple tasks when using a real-time
operating system. In simulation, multiple tasks are emulated by executing the
code in the order that would occur if there were no preemption in a real-time
system.
7-6
Model Execution
Note that the multitasking execution assumes that all tasks are multiples of
the base rate. Simulink enforces this when you have created a fixed-step
multitasking model.
The multitasking execution loop is very similar to that of singletasking, except
for the use of the task identifier (tid) argument to ModelOutputs and
ModelUpdate. The ssIsSampleHit or ssIsSpecialSampleHit macros use the
tid to determine when blocks have a hit. For example, ModelOutputs (tid=5)
will execute only the blocks that have a sample time corresponding to task
identifier 5.
The pseudocode below shows the execution of a model in a real-time
singletasking system where the model is run at interrupt level.
rtOneStep()
{
Check for interrupt overflow
Enable "rtOneStep" interrupt
ModelOutputs
-- Major time step.
LogTXY
-- Log time, states and root outports.
ModelUpdate
-- Major time step.
Integrate
-- Integration in minor time step for models
-- with continuous states.
ModelDerivatives
Do 0 or more
ModelOutputs
ModelDerivatives
EndDo (Number of iterations depends upon the solver.)
Integrate derivatives to update continuous states.
EndIntegrate
}
main()
{
Initialization (including installation of rtOneStep as an
interrupt service routine, ISR, for a real-time clock).
While(time < final time)
Background task.
EndWhile
Mask interrupts (Disable rtOneStep from executing.)
Complete any background tasks.
7-7
7
Program Architecture
Shutdown
}
Real-time singletasking execution is very similar to the nonreal-time single
tasking execution, except that the execution of the model code is done at
interrupt level.
At the interval specified by the program’s base sample rate, the interrupt
service routine (ISR) preempts the background task to execute the model code.
The base sample rate is the fastest rate in the model. If the model has
continuous blocks, then the integration step size determines the base sample
rate.
For example, if the model code is a controller operating at 100 Hz, then every
0.01 seconds the background task is interrupted. During this interrupt, the
controller reads its inputs from the analog-to-digital converter (ADC),
calculates its outputs, writes these outputs to the digital-to-analog converter
(DAC), and updates its states. Program control then returns to the background
task. All of these steps must occur before the next interrupt.
The following pseudocode shows how a model executes in a real-time
multitasking system (where the model is run at interrupt level).
rtOneStep()
{
Check for interrupt overflow
Enable "rtOneStep" interrupt
ModelOutputs(tid=0)
-- Major time step.
LogTXY
-- Log time, states and root outports.
ModelUpdate(tid=0)
-- Major time step.
Integrate
-- Integration in minor time step for
-- models with continuous states.
ModelDerivatives
Do 0 or more:
ModelOutputs(tid=0)
ModelDerivatives
EndDo (Number of iterations depends upon the solver.)
Integrate derivatives and update continuous states.
EndIntegrate
For i=1:NumTasks
If (hit in task i)
ModelOutputs(tid=i)
7-8
Model Execution
ModelUpdate(tid=i)
EndIf
EndFor
}
main()
{
Initialization (including installation of rtOneStep as an
interrupt service routine, ISR, for a real-time clock).
While(time < final time)
Background task.
EndWhile
Mask interrupts (Disable rtOneStep from executing.)
Complete any background tasks.
Shutdown
}
Running models at interrupt level in real-time multitasking environment is
very similar to the previous singletasking environment, except that overlapped
interrupts are employed for concurrent execution of the tasks.
The execution of a model in a singletasking or multitasking environment when
using real-time operating system tasking primitives is very similar to the
interrupt-level examples discussed above. The pseudocode below is for a
singletasking model using real-time tasking primitives.
tSingleRate()
{
MainLoop:
If clockSem already "given", then error out due to overflow.
Wait on clockSem
ModelOutputs
-- Major time step.
LogTXY
-- Log time, states and root
--outports
ModelUpdate
-- Major time step
Integrate
-- Integration in minor time step
-- for models with continuous
-- states.
ModelDeriviatives
Do 0 or more:
ModelOutputs
7-9
7
Program Architecture
ModelDerivatives
EndDo (Number of iterations depends upon the solver.)
Integrate derivatives to update continuous states.
EndIntegrate
EndMainLoop
}
main()
{
Initialization
Start/spawn task "tSingleRate".
Start clock that does a "semGive" on a clockSem semaphore.
Wait on "model-running" semaphore.
Shutdown
}
In this singletasking environment, the model is executed using real-time
operating system tasking primitives. In this environment, we create a single
task (tSingleRate) to run the model code. This task is invoked when a clock
tick occurs. The clock tick gives a clockSem (clock semaphore) to the model task
(tSingleRate). The model task will wait for the semaphore before executing.
The clock ticks are configured to occur at the fundamental step size (base rate)
for your model.
The pseudocode below is for a multitasking model using real-time tasking
primitives.
tSubRate(subTaskSem,i)
{
Loop:
Wait on semaphore subTaskSem.
ModelOutputs(tid=i)
ModelUpdate(tid=i)
EndLoop
}
tBaseRate()
{
MainLoop:
If clockSem already "given", then error out due to overflow.
Wait on clockSem
7-10
Model Execution
For i=1:NumTasks
If (hit in task i)
If task i is currently executing, then error out due to
overflow.
Do a "semGive" on subTaskSem for task i.
EndIf
EndFor
ModelOutputs(tid=0)
-- major time step.
LogTXY
-- Log time, states and root outports.
ModelUpdate(tid=0)
-- major time step.
Loop:
-- Integration in minor time step for
-- models with continuous states.
ModelDeriviatives
Do 0 or more:
ModelOutputs(tid=0)
ModelDerivatives
EndDo (number of iterations depends upon the solver).
Integrate derivatives to update continuous states.
EndLoop
EndMainLoop
}
main()
{
Initialization
Start/spawn task "tSubRate".
Start/spawn task "tBaseRate".
Start clock that does a "semGive" on a clockSem semaphore.
Wait on "model-running" semaphore.
Shutdown
}
In this multitasking environment, the model is executed using real-time
operating system tasking primitives. In this environment, it is necessary to
create several model tasks (tBaseRate and several tSubRate tasks) to run the
model code. The base rate task (tBaseRate) has a higher priority than the
subrate tasks. The subrate task for tid=1 has a higher priority than the
subrate task for tid=2, and so on. The base rate task is invoked when a clock
tick occurs. The clock tick gives a clockSem to tBaseRate. The first thing
7-11
7
Program Architecture
tBaseRate does is give semaphores to the subtasks that have a hit at the
current point in time. Since the base rate task has a higher priority, it
continues to execute. Next it executes the fastest task (tid=0) consisting of
blocks in your model that have the fastest sample time. After this execution, it
resumes waiting for the clock semaphore. The clock ticks are configured to
occur at executing at the fundamental step size for your model.
Program Timing
Real-time programs require careful timing of the task invocations (either via
an interrupt or a real-time operating system tasking primitive) to ensure that
the model code executes to completion before another task invocation occurs.
This includes time to read and write data to and from external hardware.
The following diagram illustrates interrupt timing.
Sample interval is appropriate for this model code execution.
time
Time to execute
the model code
Time available to process background tasks
Sample interval is too short for this model code execution.
time
Time to execute the model code
Figure 7-2: Task Timing
The sample interval must be long enough to allow model code execution
between task invocations.
7-12
Model Execution
In the figure above, the time between two adjacent vertical arrows is the
sample interval. The empty boxes in the upper diagram show an example of a
program that can complete one step within the interval and still allow time for
the background task. The gray box in the lower diagram indicates what
happens if the sample interval is too short. Another task invocation occurs
before the task is complete. Such timing results in an execution error.
Note also that, if Real-Time program is designed to run forever (i.e., the final
time is 0 or infinite so the while loop never exits), then the shutdown code never
executes.
Program Execution
As the previous section indicates, a real-time program may not require 100% of
the CPU’s time. This provides an opportunity to run background tasks during
the free time.
Background tasks include operations like writing data to a buffer or file,
allowing access to program data by third-party data monitoring tools, or using
Simulink external mode to update program parameters.
It is important, however, that the program be able to preempt the background
task at the appropriate time to ensure real-time execution of the model code.
The way the program manages tasks depends on capabilities of the
environment in which it operates.
External Mode Communication
External mode allows communication between the Simulink block diagram
and the stand-alone program that is built from the generated code. In this
mode, the real-time program functions as an interprocess communication
server, responding to requests from Simulink.
Data Logging In Singletasking
and Multitasking Model Execution
The Real-Time Workshop data-logging features, described in “Workspace I/O
Options and Data Logging” on page 2-22, enable you to save system states,
outputs, and time to a MAT-file at the completion of the model execution. The
LogTXY function, which performs data logging, operates differently in
singletasking and multitasking environments.
7-13
7
Program Architecture
If you examine how LogTXY is called in the singletasking and multitasking
environments, you will notice that for singletasking LogTXY is called after
ModelOutputs. During this ModelOutputs call, all blocks that have a hit at time
t are executed, whereas in multitasking, LogTXY is called after
ModelOutputs(tid=0) that executes only the blocks that have a hit at time t
and that have a task identifier of 0. This results in differences in the logged
values between singletasking and multitasking logging. Specifically, consider
a model with two sample times, the faster sample time having a period of 1.0
second and the slower sample time having a period of 10.0 seconds. At time t =
k*10, k=0,1,2... both the fast (tid=0) and slow (tid=1) blocks have a hit. When
executing in multitasking mode, when LogTXY is called, the slow blocks will
have a hit, but the previous value will be logged, whereas in singletasking the
current value will be logged.
Another difference occurs when logging data in an enabled subsystem.
Consider an enabled subsystem that has a slow signal driving the enable port
and fast blocks within the enabled subsystem. In this case, the evaluation of
the enable signal occurs in a slow task and the fast blocks will see a delay of
one sample period, thus the logged values will show these differences.
To summarize differences in logged data between singletasking and
multitasking, differences will be seen when:
• Any root outport block has a sample time that is slower than the fastest
sample time
• Any block with states has a sample time that is slower than the fastest
sample time
• Any block in an enabled subsystem where the signal driving the enable port
is slower than the rate of the blocks in the enabled subsystem
For the first two cases, even though the logged values are different between
singletasking and multitasking, the model results are not different. The only
real difference is where (at what point in time) the logging is done. The third
(enabled subsystem) case results in a delay that can be seen in a real-time
environment.
Rapid Prototyping and Embedded
Model Execution Differences
The rapid prototyping program framework provides a common application
programming interface (API) that does not change between model definitions.
7-14
Model Execution
The Real-Time Workshop Embedded Coder provides a different framework
that we will refer to as the embedded program framework. The embedded
program framework provides a optimized API that is tailored to your model. It
is intended that when you use the embedded style of generated code, you are
modeling how you would like your code to execute in your embedded system.
Therefore, the definitions defined in your model should be specific to your
embedded targets. Items such as the model name, parameter, and signal
storage class are included as part of the API for the embedded style of code.
One major difference between the rapid prototyping and embedded style of
generated code is that the latter contains fewer entry-point functions. The
embedded style of code can be configured to have only one run-time function
model_step. You can define a single run-time function because the embedded
target:
• Can only be used with models that do not have continuous sample time (and
therefore no continuous states)
• Requires that all S-functions must be inlined with the Target Language
Compiler, which means that they do not access the SimStruct data structure
Thus, when looking at the model execution pseudocode presented earlier in this
chapter, you can eliminate the Loop...EndLoop statements, and group the
ModelOutputs, LogTXY, and ModelUpdate into a single statement, model_step.
For a detailed discussion of how generated embedded code executes, see the
Real-Time Workshop Embedded Coder documentation.
Rapid Prototyping Model Functions
The rapid prototyping code defines the following functions that interface with
the run-time interface:
• Model() — The model registration function. This function for initializes the
work areas (e.g., allocating and setting pointers to various data structures)
needed by the model. The model registration function calls the
MdlInitializeSizes and MdlInitializeSampleTimes functions. These two
functions are very similar to the S-function mdlInitializeSizes and
mdlInitializeSampleTimes methods.
• MdlStart(void) — After the model registration functions,
MdlInitializeSizes and MdlInitializeSampleTimes execute, the run-time
7-15
7
Program Architecture
interface starts execution by calling MdlStart. This routine is called once at
startup.
The function MdlStart has four basic sections:
- Code to initialize the states for each block in the root model that has states.
A subroutine call is made to the “initialize states” routine of conditionally
executed subsystems.
- Code generated by the one-time initialization (start) function for each
block in the model.
- Code to enable the blocks in the root model that have enable methods, and
the blocks inside triggered or function-call subsystems residing in the root
model. Simulink blocks can have enable and disable methods. An enable
method is called just before a block starts executing, and the disable
method is called just after the block stops executing.
- Code for each block in the model that has a constant sample time.
• MdlOutputs(int_T tid) — MdlOutputs updates the output of blocks at
appropriate times. The tid (task identifier) parameter identifies the task
that in turn maps when to execute blocks based upon their sample time. This
routine is invoked by the run-time interface during major and minor time
steps. The major time steps are when the run-time interface is taking an
actual time step (i.e., it is time to execute a specific task). If your model
contains continuous states, the minor time steps will be taken. The minor
time steps are when the solver is generating integration stages, which are
points between major outputs. These integration stages are used to compute
the derivatives used in advancing the continuous states.
• MdlUpdate(int_T tid) — MdlUpdate updates the discrete states and work
vector state information (i.e., states that are neither continuous nor discrete)
saved in work vectors. The tid (task identifier) parameter identifies the task
that in turn indicates which sample times are active allowing you to
conditionally update states of only active blocks. This routine is invoked by
the run-time interface after the major MdlOutputs has been executed.
• MdlDerivatives(void) — MdlDerivatives returns the block derivatives.
This routine is called in minor steps by the solver during its integration
stages. All blocks that have continuous states have an identical number of
derivatives. These blocks are required to compute the derivatives so that the
solvers can integrate the states.
7-16
Model Execution
• MdlTerminate(void) — MdlTerminate contains any block shutdown code.
MdlTerminate is called by the run-time interface, as part of the termination
of the real-time program.
The contents of the above functions are directly related to the blocks in your
model. A Simulink block can be generalized to the following set of equations.
y = f 0 ( t, x c, x d, u )
Output, y, is a function of continuous state, xc, discrete state, xd, and input, u.
Each block writes its specific equation in the appropriate section of MdlOutput.
x d + 1 = f u ( t, x d, u )
The discrete states, xd, are a function of the current state and input. Each block
that has a discrete state updates its state in MdlUpdate.
x· = f d ( t, x c, u )
The derivatives, x, are a function of the current input. Each block that has
continuous states provides its derivatives to the solver (e.g., ode5) in
MdlDerivatives. The derivatives are used by the solver to integrate the
continuous state to produce the next value.
The output, y, is generally written to the block I/O structure. Root-level
Outport blocks write to the external outputs structure. The continuous and
discrete states are stored in the states structure. The input, u, can originate
from another block’s output, which is located in the block I/O structure, an
external input (located in the external inputs structure), or a state. These
structures are defined in the model.h file that Real-Time Workshop generates.
7-17
7
Program Architecture
Figure 7-3 shows the general content of the rapid prototyping style of C code.
/*
* Version, Model options, TLC options,
* and code generation information are placed here.
*/
<includes>
void MdlStart(void)
{
/*
* State initialization code.
* Model start-up code - one time initialization code.
* Execute any block enable methods.
* Initialize output of any blocks with constant sample times.
*/
}
void MdlOutputs(int_T tid)
{
/* Compute: y = f0(t,xc,xd,u) for each block as needed. */
}
void MdlUpdate(int_T tid)
{
/* Compute: xd+1 = fu(t,xd,u) for each block as needed. */
}
void MdlDerivatives(void)
{
/* Compute: dxc = fd(t,xc,u) for each block as needed. */
}
void MdlTerminate(void)
{
/* Perform shutdown code for any blocks that
have a termination action */
}
Figure 7-3: Content of model.c for the Rapid Prototyping Code Style
7-18
Model Execution
Figure 7-4 shows a flow chart describing the execution of the rapid prototyping
generated code.
Start Execution
MdlStart
MdlOutput
MdlDerivatives
MdlOutput
MdlDerivatives
Integration
Execution Loop
MdlUpdate
MdlTerminate
End
Figure 7-4: Rapid Prototyping Execution Flow Chart
Each block places code into specific Mdl routines according to the algorithm
that it is implementing. Blocks have input, output, parameters, and states, as
well as other general items. For example, in general, block inputs and outputs
are written to a block I/O structure (rtB). Block inputs can also come from the
external input structure (rtU) or the state structure when connected to a state
port of an integrator (rtX), or ground (rtGround) if unconnected or grounded.
7-19
7
Program Architecture
Block outputs can also go to the external output structure (rtY). The following
figure shows the general mapping between these items.
External
Inputs
Struct,
rtU
rtGround
External
Outputs
Struct,
Block I/O
Struct,
rtB
rtY
Block
Work
Structs,
States
Struct,
Parameter
Struct,
rtX
rtP
rtRWork,
rtIWork,
rtPWork,
....
Figure 7-5: Data View of the Generated Code
Structure definitions:
• Block I/O Structure (rtB) — This structure consists of all block output
signals. The number of block output signals is the sum of the widths of the
data output ports of all nonvirtual blocks in your model. If you activate block
I/O optimizations, Simulink and Real-Time Workshop reduce the size of the
rtB structure by:
- Reusing the entries in the rtB structure
- Making other entries local variables
See “Signals: Storage, Optimization, and Interfacing” on page 5-17 for
further information on these optimizations.
Structure field names are determined by either the block’s output signal
name (when present) or by the block name and port number when the output
signal is left unlabeled.
• Block States Structures — The continuous states structure (rtX) contains
the continuous state information for any blocks in your model that have
7-20
Model Execution
continuous states. Discrete states are stored in a data structure called the
DWork vector (rtDWork).
• Block Parameters Structure (rtP) — The parameters structure contains all
block parameters that can be changed during execution (e.g., the parameter
of a Gain block).
• External Inputs Structure (rtU) —The external inputs structure consists of
all root-level Inport block signals. Field names are determined by either the
block’s output signal name, when present, or by the Inport block’s name
when the output signal is left unlabeled.
• External Outputs Structure (rtY) —The external outputs structure consists
of all root-level Outport blocks. Field names are determined by the root-level
Outport block names in your model.
• Real Work, Integer Work, and Pointer Work Structures (rtRWork, rtIWork,
rtPWork) — Blocks may have a need for real, integer, or pointer work areas.
For example, the Memory block uses a real work element for each signal.
These areas are used to save internal states or similar information.
Embedded Model Functions
The Real-Time Workshop Embedded Coder Coder target generates the
following functions:
• model_intialize — Performs all model initialization and should be called
once before you start executing your model.
• If the Single output/update function code generation option is selected,
then you will see:
- model_step(int_T tid) — Contains the output and update code for all
blocks in your model.
Otherwise you will see:
- model_output(int_T tid) — Contains the output code for all blocks in
your model.
- model_update(int_T tid) — This contains the update code for all blocks
in your model.
• If the Terminate function required code generation option is selected, then
you will see:
7-21
7
Program Architecture
- model_terminate — This contains all model shutdown code and should be
called as part of system shutdown.
See the Real-Time Workshop Embedded Coder documentation for complete
descriptions of these functions in the context of the Real-Time Workshop
Embedded Coder.
7-22
Rapid Prototyping Program Framework
Rapid Prototyping Program Framework
The code modules generated from a a Simulink model — model.c, model.h, and
other files — implement the model’s system equations, contain block
parameters, and perform initialization.
The Real-Time Workshop program framework provides the additional source
code necessary to build the model code into a complete, stand-alone program.
The program framework consists of application modules (files containing
source code to implement required functions) designed for a number of
different programming environments.
The automatic program builder ensures the program is created with the proper
modules once you have configured your template makefile. The application
modules and the code generated for a Simulink model are implemented using
a common API. This API defines a data structure (called a real-time model,
sometimes abbreviated as rtM) that encapsulates all data for your model.
This API is similar to that of S-functions, with one major exception: the API
assumes that there is only one instance of the model, whereas S-functions can
have multiple instances. The function prototypes also differ from S-functions.
7-23
7
Program Architecture
Rapid Prototyping Program Architecture
The structure of a real-time program consists of three components. Each
component has a dependency on a different part of the environment in which
the program executes. The following diagram illustrates this structure.
Rapid Prototyping Real-Time Program Architecture
SystemDependent
Components
SystemIndependent
Components
Main Program
Timing
Interrupt handling
I/O drivers
Data logging
External mode
communication
Integration solvers: ode1.c — ode5.c
Model execution scheduler: rt_sim.c
Run-Time Interface
Application
Components
Generated (Model) Code
MdlOutputs, etc.
Real-time Model data str.
Inlined S-functions
Model parameters
Noninlined
S-functions
mysfun.c
Figure 7-6: The Rapid Prototyping Program Architecture
7-24
Rapid Prototyping Program Framework
The Real-Time Workshop architecture consists of three parts. The first two
components, system dependent and independent, together form the run-time
interface.
This architecture readily adapts to a wide variety of environments by isolating
the dependencies of each program component. The following sections discuss
each component in more detail and include descriptions of the application
modules that implement the functions carried out by the system dependent,
system independent, and application components.
Rapid Prototyping System-Dependent Components
These components contain the program’s main function, which controls
program timing, creates tasks, installs interrupt handlers, enables data
logging, and performs error checking.
The way in which application modules implement these operations depends on
the type of computer. This means that, for example, the components used for a
DOS-based program perform the same operations, but differ in method of
implementation from components designed to run under Tornado on a VME
target.
The main Function
The main function in a C program is the point where execution begins. In
Real-Time Workshop application programs, the main function must perform
certain operations. These operations can be grouped into three categories:
initialization, model execution, and program termination.
Initialization
• Initialize special numeric parameters: rtInf, rtMinusInf, and rtNaN. These
are variables that the model code can use.
• Call the model registration function to get a pointer to the real-time model.
The model registration function has the same name as your model. It is
responsible for initializing real-time model fields and any S-functions in your
model.
• Initialize the model size information in the real-time model. This is done by
calling MdlInitializeSizes.
• Initialize a vector of sample times and offsets (for systems with multiple
sample rates). This is done by calling MdlInitializeSampleTimes.
7-25
7
Program Architecture
• Get the model ready for execution by calling MdlStart, which initializes
states and similar items.
• Set up the timer to control execution of the model.
• Define background tasks and enable data logging, if selected.
Model Execution
• Execute a background task, for example, communicate with the host during
external mode simulation or introduce a wait state until the next sample
interval.
• Execute model (initiated by interrupt).
• Log data to buffer (if data logging is used).
• Return from interrupt.
Program Termination
• Call a function to terminate the program if it is designed to run for a finite
time — destroy the real-time model data structure, deallocate memory, and
write data to a file.
Rapid Prototyping Application Modules
for System Dependent Components
The application modules contained in the system dependent components
generally include a main module such as rt_main.c containing the main entry
point for C. There may also be additional application modules for such things
as I/O support and timer handling.
Rapid Prototyping System-Independent
Components
These components are collectively called system independent because all
environments use the same application modules to implement these
operations. This section steps through the model code (and if the model has
continuous states, calls one of the numerical integration routines). This section
also includes the code that defines, creates, and destroys the real-time model
data structure (rtM). The model code and all S-functions included in the
program define their own SimStruct.
7-26
Rapid Prototyping Program Framework
The model code execution driver calls the functions in the model code to
compute the model outputs, update the discrete states, integrate the
continuous states (if applicable), and update time. These functions then write
their calculated data to the real-time model.
Model Execution
At each sample interval, the main program passes control to the model
execution function, which executes one step though the model. This step reads
inputs from the external hardware, calculates the model outputs, writes
outputs to the external hardware, and then updates the states.
The following diagram illustrates these steps.
Read system inputs
from A/D
Calculate system outputs
Write system outputs
to D/A
Execute Model
Calculate and update
discrete states
Calculate and update
continuous states
Integration
Algorithm
Increment time
Figure 7-7: Executing the Model
7-27
7
Program Architecture
Note that this scheme writes the system outputs to the hardware before the
states are updated. Separating the state update from the output calculation
minimizes the time between the input and output operations.
Integration of Continuous States
The real-time program calculates the next values for the continuous states
based on the derivative vector, dx/dt, for the current values of the inputs and
the state vector.
These derivatives are then used to calculate the next value of the states using
a state-update equation. This is the state-update equation for the first order
Euler method (ode1)
dx
x = x + ------- h
dt
where h is the step size of the simulation, x represents the state vector, and
dx/dt is the vector of derivatives. Other algorithms may make several calls to
the output and derivative routines to produce more accurate estimates.
Note, however, that real-time programs use a fixed-step size since it is
necessary to guarantee the completion of all tasks within a given amount of
time. This means that, while you should use higher order integration methods
for models with widely varying dynamics, the higher order methods require
additional computation time. In turn, the additional computation time may
force you to use a larger step size, which can diminish the accuracy increase
initially sought from the higher order integration method.
Generally, the stiffer the equations, (i.e., the more dynamics in the system with
widely varying time constants), the higher the order of the method that you
must use.
In practice, the simulation of very stiff equations is impractical for real-time
purposes except at very low sample rates. You should test fixed-step size
integration in Simulink to check stability and accuracy before implementing
the model for use in real-time programs.
For linear systems, it is more practical to convert the model that you are
simulating to a discrete time version, for instance, using the c2d function in the
Control System Toolbox.
7-28
Rapid Prototyping Program Framework
Application Modules for System-Independent Components
The system independent components include these modules:
• ode1.c, ode2.c, ode3.c, ode4.c, ode5.c — These modules implement the
integration algorithms supported for real-time applications. See the
Simulink documentation for more information about these fixed-step
solvers.
• rt_sim.c — Performs the activities necessary for one time step of the model.
It calls the model function to calculate system outputs and then updates the
discrete and continuous states.
• simstruc_types.h — Contains definitions of various events, including
subsystem enable/disable and zero crossings. It also defines data logging
variables.
The system independent components also include code that defines, creates,
and destroys the real-time model data structure. All S-functions included in
the program define their own SimStruct.
The SimStruct data structure encapsulates all the data relating to
anS-function, including block parameters and outputs. See Writing
S-Functions for more information about the SimStruct.
Rapid Prototyping Application Components
The application components contain the generated code for the Simulink
model, including the code for any S-functions in the model. This code is referred
to as the model code because these functions implement the Simulink model.
However, the generated code contains more than just functions to execute the
model (as described in the previous section). There are also functions to
perform initialization, facilitate data access, and complete tasks before
program termination. To perform these operations, the generated code must
define functions that:
• Create the real-time model.
• Initialize model size information in the real-time model.
• Initialize a vector of sample times and sample time offsets and store this
vector in the real-time model.
• Store the values of the block initial conditions and program parameters in
the real-time model.
7-29
7
Program Architecture
• Compute the block and system outputs.
• Update the discrete state vector.
• Compute derivatives for continuous models.
• Perform an orderly termination at the end of the program (when the current
time equals the final time, if a final time is specified).
• Collect block and scope data for data logging (either with Real-Time
Workshop or third-party tools).
The Real-Time Model Data Structure
The real-time model data structure encapsulates model data and associated
information necessary to fully describe the model. Its contents include:
• Model parameters, inputs, and outputs
• Storage areas, such as dWork
• Timing information
• Solver identification
• Data logging information
• Simstructs for all child S-functions
• External mode information
The real-time model data structure is used for all targets. In previous releases,
the ERT target used the rtObject data structure, and other targets used the
simstruct data structure for encapsulating model data. Now all targets are
treated the same, except for the fact that the real-time model data structure is
pruned for ERT targets to save space in executables. Even when not pruned,
the real-time model data structure is more space-efficient than the root
simstruct used by earlier releases for non-ERT targets, as it only contains
fields for child (S-function) simstructs that are actually used in a model.
Rapid Prototyping Model Code Functions
The functions defined by the model code are called at various stages of program
execution (i.e., initialization, model execution, or program termination).
7-30
Rapid Prototyping Program Framework
The following diagram illustrates the functions defined in the generated code
and shows what part of the program executes each function.
Model Code
Main Program Initialization
Model registration function — model
Initialize sizes in the rtM — MdlInitializeSizes
Initialize sample times and offsets — MdlInitializeSampleTimes
Start model (initialize conditions, etc.) — MdlStart
Model Execution
Compute block and system outputs — MdlOutputs
Update discrete state vector — MdlUpdate
Compute derivatives for continuous models — MdlDerivatives
Main Program Termination
Orderly termination at end of the program — MdlTerminate
Figure 7-8: Execution of the Model Code
The Model Registration Function
The model registration function has the same name as the Simulink model
from which it is generated. It is called directly by the main program during
initialization. Its purpose is to initialize and return a pointer to the real-time
model data structure.
Models Containing S-Functions
A noninlined S-function is any C MEX S-function that is not implemented
using a customized TLC file. If you create a C MEX S-function as part of a
Simulink model, it is by default noninlined unless you write your own TLC file
7-31
7
Program Architecture
that inlines it within the body of the model.c code. Real-Time Workshop
automatically incorporates your non-inlined C code S-functions into the
program if they adhere to the S-function API described in the Simulink
documentation.
This format defines functions and a SimStruct that are local to the S-function.
This allows you to have multiple instances of the S-function in the model. The
model’s real-time model data structure contains a pointer to each S-function’s
SimStruct.
Code Generation and S-Functions
If a model contains S-functions, the source code for the S-function must be on
the search path the make utility uses to find other source files. The directories
that are searched are specified in the template makefile that is used to build
the program.
S-functions are implemented in a way that is directly analogous to the model
code. They contain their own public registration function (which is called by the
top-level model code) that initializes static function pointers in its SimStruct.
When the top-level model needs to execute the S-function, it does so via the
function pointers in the S-function’s SimStruct. There can be more than one
S-function with the same name in your model. This is accomplished by having
function pointers to static functions.
Inlining S-Functions
You can incorporate C MEX S-functions, along with the generated code, into
the program executable. You can also write a target file for your C MEX
S-function to inline the S-function, thus improving performance by eliminating
function calls to the S-function itself. For more information on inlining
S-functions, see the Target Language Compiler Reference Guide.
7-32
Rapid Prototyping Program Framework
Application Modules for Application Components
When Real-Time Workshop generates code, it produces the following files:
•model.c — The C code generated from the Simulink block diagram. This code
implements the block diagram’s system equations as well as performing
initialization and updating outputs.
• model_data.c — An optional file containing data for parameters and
constant block i/o, which are also declared as extern in model.h. Only
generated when rtP and rtC structures are populated.
• model.h — Header file containing the block diagram’s simulation
parameters, I/O structures, work structures, and other declarations. It
includes model_private.h.
• model_private.h — Header file containing declarations of exported signals
and parameters.
These files are named for the Simulink model from which they are generated.
In addition, a dummy include file always named rtmodel.h is generated, which
includes the above model-specific data structures and entry points. This
enables the (static) target-specific main programs to reference files generated
by Real-Time Workshop without needing to know the names of the models
involved.
If you have created custom blocks using C MEX S-functions, you need the
source code for these S-functions available during the build process.
7-33
7
Program Architecture
Embedded Program Framework
The Real-Time Workshop Embedded Coder provides a framework for
embedded programs. Its architecture is outlined by the following figure.
Embedded Program Architecture
System
Dependent
Components
Main Program
Timing
Interrupt handling
I/O drivers
Data logging
System
Independent
Components
Integration solvers: ode1.c — ode5.c
Model execution scheduler: rt_sim.c
Run-time Interface
Application
Components
Generated (Model) Code
MdlOutputs, etc.
Inlined S-functions
Model parameters
Figure 7-9: Embedded Program Architecture
7-34
Embedded Program Framework
Note the similarity between this architecture and the rapid prototyping
architecture in Figure 7-6. The main difference is the lack of the SimStruct
data structure and the removal of the noninlined S-functions.
Using this figure, you can compare the embedded style of generated code, used
in the Real-Time Workshop Embedded Coder, with the rapid prototyping style
of generated code of the previous section. Most of the rapid prototyping
explanations in the previous section hold for the Real-Time Workshop
Embedded Coder target. The Real-Time Workshop Embedded Coder target
simplifies the process of using the generated code in your custom-embedded
applications by providing a model- specific API and eliminating the SimStruct.
This target contains the same conceptual layering as the rapid prototyping
target, but each layer has been simplified.
For a discussion of the structure of embedded real-time code, see the Real-Time
Workshop Embedded Coder documentation.
7-35
7
Program Architecture
7-36
8
Models with Multiple
Sample Rates
This section discusses how and why real-time execution of code generated from models having
multiple sample rates differs from the simulation behavior of the models. Solutions to problems
arising from multirate model execution are also described. The topics covered are:
Introduction (p. 8-2)
Describes types of sample times and issues to consider
regarding execution of multirate models
Singletasking vs. Multitasking
Environments (p. 8-3)
Discusses how Real-Time Workshop handles execution of
multirate systems, in both multitasking and
pseudo-multitasking environments
Sample Rate Transitions (p. 8-12)
Shows how to handle transitions between blocks with
unequal sample rates using Rate Transition blocks
Singletasking and Multitasking
Execution of a Model: an Example
(p. 8-22)
Discusses how an example model executes in both
singletasking and multitasking solver modes, with timing
diagrams
8
Models with Multiple Sample Rates
Introduction
A Simulink block can be classified, according to its sample time, as constant,
continuous-time, discrete-time, inherited, or variable. Examples of each type
include:
• Constant — Constant block, Width
• Continuous-time — Integrator, Derivative, Transfer Function
• Discrete-time — Unit Delay, Digital Filter
• Inherited — Gain, Sum, Lookup Table
• Variable — These are blocks that set their time of next hit based upon
current information. These blocks work only with variable step solvers.
Examples of variable sample time blocks include the Pulse Generator and
some S-Function blocks.
Blocks in the inherited category assume the sample time of the blocks that are
driving them. Continuous blocks have a nominal sample time of zero. Every
Simulink block therefore has a sample time, whether it is explicit, as in the
case of continuous or discrete blocks, or implicit, as in the case of inherited
blocks.
Simulink allows you to create models without any restrictions on connections
between blocks with different sample times. It is therefore possible to have
blocks with differing sample times in a model (a mixed-rate system). A possible
advantage of employing multiple sample times is improved efficiency when
executing in a multitasking real-time environment.
Simulink provides considerable flexibility in building these mixed-rate
systems. However, the same flexibility also allows you to construct models for
which the code generator cannot generate correct real-time code for execution
in a multitasking environment. To make these models operate correctly in real
time (i.e., to give the right answers), you must modify your model. In general,
the modifications involve placing Rate Transition blocks between blocks that
have unequal sample rates. The sections that follow discuss the issues you
must address to use a mixed-rate model successfully in a multitasking
environment.
8-2
Singletasking vs. Multitasking Environments
Singletasking vs. Multitasking Environments
There are two execution modes for a fixed-step Simulink model: singletasking
and multitasking. You use the Mode pull-down menu on the Solver page of the
Simulation Parameters dialog box to specify how to execute your model. Auto
mode (the default) selects multitasking execution for a mixed-rate model, and
otherwise selects singletasking execution. You can also select SingleTasking
or MultiTasking execution explicitly.
Execution of models in a real-time system can be done with the aid of a
real-time operating system, or it can be done on a bare-board target, where the
model runs in the context of an interrupt service routine (ISR).
Note that the fact that a system (such as UNIX or Microsoft Windows) is
multitasking does not guarantee that the program can execute in real time.
This is because it is not guaranteed that the program can preempt other
processes when required.
In DOS, where only one process can exist at any given time, an interrupt
service routine (ISR) must perform the steps of saving the processor context,
executing the model code, collecting data, and restoring the processor context.
Tornado, on the other hand, provides automatic context switching and task
scheduling. This simplifies the operations performed by the ISR. In this case,
the ISR simply enables the model execution task, which is normally blocked.
8-3
8
Models with Multiple Sample Rates
Figure 8-1 illustrates this difference.
Real-Time Clock
Hardware
Interrupt Service
Routine
Interrupt
Save Context
Execute Model
Program execution using an
interrupt service routine (bareboard, with no real-time operating
system). See the grt target for an
example.
Collect Data
Restore Context
Real-Time Clock
Hardware
Interrupt Service
Routine
Context
Switch
Model Execution
Task
Interrupt
semGive
semTake
Execute Model
Program execution using a real-time
operating system primitive. See the
Tornado target for an example.
Figure 8-1: Real-Time Program Execution
8-4
Collect Data
Singletasking vs. Multitasking Environments
This chapter focuses on when and how the run-time interface executes your
model. See “Program Execution” on page 7-13 for a description of what happens
during model execution.
Executing Multitasking Models
In cases where the continuous part of a model executes at a rate that is
different from the discrete part, or a model has blocks with different sample
rates, the code assigns each block a task identifier (tid) to associate the block
with the task that executes at the block’s sample rate.
Certain restrictions apply to the sample rates that you can use:
• The sample rate of any block must be an integer multiple of the base (i.e., the
fastest) sample period. The base sample period is determined by the Fixed
step size specified on the Solver page of the Simulation parameters dialog
box (if a model has continuous blocks) or by the fastest sample time specified
in the model (if the model is purely discrete). Continuous blocks always
execute via an integration algorithm that runs at the base sample rate.
• The continuous and discrete parts of the model can execute at different rates
only if the discrete part is executed at the same or a slower rate than the
continuous part (and is an integer multiple of the base sample rate).
Multitasking and Pseudomultitasking
In a multitasking environment, the blocks with the fastest sample rates are
executed by the task with the highest priority, the next slowest blocks are
executed by a task with the next lower priority, and so on. Time available in
between the processing of high priority tasks is used for processing lower
priority tasks. This results in efficient program execution.
See “Multitasking System Execution” on page 8-7 for a graphical
representation of task timing.
In multitasking environments (i.e., a real-time operating system), you can
define separate tasks and assign them priorities. In a bare-board target (i.e.,
no real-time operating system present), you cannot create separate tasks.
However, Real-Time Workshop application modules implement what is
effectively a multitasking execution scheme using overlapped interrupts,
accompanied by manual context switching.
8-5
8
Models with Multiple Sample Rates
This means an interrupt can occur while another interrupt is currently in
progress. When this happens, the current interrupt is preempted, the
floating-point unit (FPU) context is saved, and the higher priority interrupt
executes its higher priority (i.e., faster sample rate) code. Once complete,
control is returned to the preempted ISR.
The following diagrams illustrate how mixed-rate systems are handled by
Real-Time Workshop in these two environments.
8-6
Singletasking vs. Multitasking Environments
t0
t1
t2
t3
t4
rate 1
Highest Priority
rate 2
rate 3
Lowest Priority
Vertical arrows indicate sample times.
Dotted lines with downward pointing
arrows indicate the release of control
to a lower priority task.
Dark gray areas indicate task execution.
Hashed areas indicate task preemption by a
higher priority task.
Dotted lines with upward pointing
arrows indicate preemption by a
higher priority task.
Light gray areas indicate task execution
is pending.
Figure 8-2: Multitasking System Execution
Figure 8-3 illustrates how overlapped interrupts are used to implement
pseudomultitasking. Note that in this case, Interrupt 0 does not return until
after Interrupts 1, 2, and 3.
8-7
8
Models with Multiple Sample Rates
.
Interrupt 0
Begins
t0
Interrupt 1
t1
Interrupt 2
Begins
t2
Interrupt 3
t3
t4
Highest Priority
Interrupt 2
Ends
Interrupt 0
Ends
Lowest Priority
Figure 8-3: Pseudomultitasking Using Overlapped Interrupts
Building the Program for Multitasking Execution
To use multitasking execution, select Auto (the default) or MultiTasking as the
mode on the Solver page of the Simulation Parameters dialog box. The Mode
menu is only active if you have selected Fixed-step as the Solver options type.
Auto solver mode will result in a multitasking environment if your model has
two or more different sample times. In particular, a model with a continuous
and a discrete sample time will run in singletasking mode if the fixed-step size
is equal to the discrete sample time.
Singletasking
It is possible to execute the model code in a strictly singletasking manner.
While this method is less efficient with regard to execution speed, in certain
situations it may allow you to simplify your model.
8-8
Singletasking vs. Multitasking Environments
In a singletasking environment, the base sample rate must define a time
interval that is long enough to allow the execution of all blocks within that
interval.
The following diagram illustrates the inefficiency inherent in singletasking
execution.
t0
t1
t2
t3
t4
Figure 8-4: Singletasking System Execution
Singletasking system execution requires a sample interval that is long enough
to execute one step through the entire model.
Building the Program for Singletasking Execution
To use singletasking execution, select the singletasking mode on the Solver
page of the Simulation Parameters dialog box. If the solver mode is Auto,
singletasking is used in the following cases:
• If your model contains one sample time
• If your model contains a continuous and a discrete sample time and the fixed
step size is equal to the discrete sample time
Model Execution
To generate code that executes correctly in real time, you may need to modify
sample rate transitions within the model before generating code. To
understand this process, first consider how Simulink simulations differ from
real-time programs.
Simulating Models with Simulink
Before Simulink simulates a model, it orders all of the blocks based upon their
topological dependencies. This includes expanding subsystems into the
individual blocks they contain and flattening the entire model into a single list.
Once this step is complete, each block is executed in order.
8-9
8
Models with Multiple Sample Rates
The key to this process is the proper ordering of blocks. Any block whose output
is directly dependent on its input (i.e., any block with direct feedthrough)
cannot execute until the block driving its input has executed.
Some blocks set their outputs based on values acquired in a previous time step
or from initial conditions specified as a block parameter. The output of such a
block is determined by a value stored in memory, which can be updated
independently of its input. During simulation, all necessary computations are
performed prior to advancing the variable corresponding to time. In essence,
this results in all computations occurring instantaneously (i.e., no
computational delay).
Executing Models in Real Time
A real-time program differs from a Simulink simulation in that the program
must execute the model code synchronously with real time. Every calculation
results in some computational delay. This means the sample intervals cannot
be shortened or lengthened (as they can be in Simulink), which leads to less
efficient execution.
t0
t1
t2
Time
Figure 8-5: Unused Time in Sample Interval
Sample interval t1 cannot be compressed to increase execution speed because
by definition, sample times are clocked in real time.
Real-Time Workshop application programs are designed to circumvent this
potential inefficiency by using a multitasking scheme. This technique defines
tasks with different priorities to execute parts of the model code that have
different sample rates.
See “Multitasking and Pseudomultitasking” on page 8–5 for a description of
how this works. It is important to understand that section before proceeding
here.
8-10
Singletasking vs. Multitasking Environments
Singletasking vs. Multitasking Operation
Singletasking programs require longer sample intervals, because all
computations must be executed within each clock period. This can result in
inefficient use of available CPU time, as shown in Figure 8-5.
The use of multitasking can improve the efficiency of your program if the model
is large and has many blocks executing at each rate.
However, if your model is dominated by a single rate, and only a few blocks
execute at a slower rate, multitasking can actually degrade performance. In
such a model, the overhead incurred in task switching can be greater than the
time required to execute the slower blocks. In this case, it is more efficient to
execute all blocks at the dominant rate.
If you have a model that can benefit from multitasking execution, you may
need to modify your Simulink model by adding Rate Transition blocks to
generate correct results. The next section, “Sample Rate Transitions” on
page 8-12, discusses issues related to rate transition blocks.
8-11
8
Models with Multiple Sample Rates
Sample Rate Transitions
There are two possible sample rate transitions that can exist within a model:
• A faster block driving a slower block
• A slower block driving a faster block
In singletasking systems, there are no issues involving multiple sample rates.
In multitasking and pseudomultitasking systems, however, differing sample
rates can cause problems. To prevent possible errors in calculated data, you
must control model execution at these transitions. In transitioning from faster
to slower blocks, you must add Rate Transition blocks between the faster and
slower blocks.
This diagram
T = 1 sec
Faster
Block
T = 2 sec
Slower
Block
becomes
T = 1 sec
Faster
Block
T = 2 sec
Rate
Transition
T = 2 sec
Slower
Block
Figure 8-6: Transitioning from Faster to Slower Blocks (T = sample period)
8-12
Sample Rate Transitions
In transitioning from slower to faster blocks, you must add Rate Transition
blocks between the slower and faster blocks.
This diagram
T = 2 sec
Slower
Block
T = 1 sec
Faster
Block
becomes
T = 2 sec
Slower
Block
T = 2 sec
Rate
Transition
T = 1 sec
Faster
Block
Figure 8-7: Transitioning from Slower to Faster Blocks (T = Sample Period)
Data Transfer Problems
Rate Transition blocks are designed to deal with the following problems that
occur in data transfer between blocks running at different rates:
• Data integrity: A problem of data integrity exists when the input to a block
changes during the execution of that block. Data integrity problems can be
caused by preemption.
Consider the following scenario: a faster block supplies the input to a slower
block. The slower block reads an input value V1 from the faster block and
begins computations using that value. These computations are preempted by
another execution of the faster block, which computes a new output value V2.
8-13
8
Models with Multiple Sample Rates
A data integrity problem now arises: when the slower block resumes
execution, it continues its computations, now using the “new” input value V2.
We will refer to such a data transfer as unprotected. Figure 8-8 illustrates an
unprotected data transfer.
In a protected data transfer, the output V1 of the faster block would be held
until the slower block finished executing.
• Deterministic vs. non-deterministic data transfer: In a deterministic data
transfer, the timing of the data transfer is completely predictable, as
determined by the sample rates of the blocks.
The timing of a non-deterministic data transfer depends on the availability
of data, the sample rates of the blocks, and the time at which the receiving
block begins to execute relative to the driving block.
You can use the Rate Transition block to ensure that data transfers in your
application are both protected and deterministic. These characteristics are
considered desirable in most applications. However, the Rate Transition block
supports flexible options that allow you to compromise data integrity and
determinism in favor of lower latency. The next section summarizes these
options.
Rate Transition Block Options
Several parameters of the Rate Transition block are relevant to its use in code
generation for real-time execution. These are discussed below. For full
documentation of the Rate Transition block and its block parameters, see the
“Simulink Blocks” section of Using Simulink.
The Rate Transition block handles both types of transitions (fast to slow, and
slow to fast). When inserted between two blocks of differing sample rates, the
Rate Transition block detects the two rates and automatically configures its
input and output sample rates for the appropriate type of transition.
The most critical decision you must make in configuring a Rate Transition
block is the choice of data transfer mechanism to be used between the two
rates. Your choice will be dictated by considerations of safety, memory usage,
and performance. The data transfer mechanism is controlled by two options:
• Ensure data integrity during data transfer: When this option is on, the
integrity of data transferred between rates is guaranteed (the data transfer
is protected). When this option is off, data integrity is not guaranteed (the
8-14
Sample Rate Transitions
data transfer is unprotected). By default, Ensure data integrity during
data transfer is on.
• Ensure deterministic data transfer (maximum delay): This option is
enabled only for protected data transfer (when Ensure data integrity
during data transfer is on). When this option is on, the Rate Transition
block behaves like a Zero-Order Hold block (for fast to slow transitions) or a
Unit Delay block (for slow to fast transitions). The Rate Transition block
controls the timing of data transfer in a completely predictable way. When
this option is off, the data transfer is non-deterministic. By default, Ensure
deterministic data transfer (maximum delay) is on.
Thus the Rate Transition block offers three modes of operation with respect to
data transfer. In order safety, from safest to least safe, these are:
• Protected/Deterministic (default): This is the safest mode. The drawback of
this mode is that it introduces latency into the system:
- Fast to slow transition: maximum latency is 1 sample period of the slower
task.
- Slow to fast transition: maximum latency is 2 sample periods of the slower
task.
• Protected/Non-Deterministic: In this mode, data integrity is protected by
double-buffering data transferred between rates. The blocks downstream
from the Rate Transition block always use the latest available data from the
block that drives the Rate Transition block. Maximum latency is less than or
equal to 1 sample period of the faster task.
The drawbacks of this mode are its non-deterministic timing and its use of
extra memory buffers. The advantage of this mode is its low latency.
• Unprotected/Non-Deterministic: This mode is the least safe, and is not
recommended for mission-critical applications. The latency of this mode is
the same as for Protected/Non-Deterministic mode, but memory
requirements are reduced since there is no double-buffering.
Note In unprotected mode (Ensure data integrity during data transfer
option off), the Rate Transition block does nothing other than allow the rate
transition to exist in the model.
8-15
8
Models with Multiple Sample Rates
The next four sections describe cases in which Rate Transition blocks are
necessary for sample rate transitions. The discussion and timing diagrams in
these sections are based on the assumption that the Rate Transition block is
used in its default (Protected/Deterministic) mode, with the Ensure data
integrity during data transfer and Ensure deterministic data transfer
(maximum delay) options on.
Faster to Slower Transitions in Simulink
In a model where a faster block drives a slower block having direct
feedthrough, the outputs of the faster block are always computed first. In
simulation intervals where the slower block does not execute, the simulation
progresses more rapidly because there are fewer blocks to execute.
The following diagram illustrates this situation.
t1
t0
T = 1 sec
T = 2 sec
Faster
Block
Slower
Block
T=1s
T=2s
t2
T=1s
t3
T=1s
T=2s
T=1s
Time
Simulink does not execute in real time, which means that it is not bound by
real-time constraints. Simulink waits for, or moves ahead to, whatever tasks
are necessary to complete simulation flow. The actual time interval between
sample time steps can vary.
Faster to Slower Transitions in Real Time
In models where a faster block drives a slower block, you must compensate for
the fact that execution of the slower block may span more than one execution
period of the faster block. This means that the outputs of the faster block may
change before the slower block has finished computing its outputs. The
following diagram illustrates a situation where this problem arises. The
8-16
Sample Rate Transitions
hashed area indicates times when tasks are preempted by higher priority
before completion.
T = 1 sec
T = 2 sec
Faster
Block
Slower
Block
2 Sec
Task
1 Sec
Task
1
T=1s
T=2s
2
T=2s
1
T=1s
3
T=1s
1
Time
The faster task (T=1s) completes.
2
Higher priority preemption occurs.
3
The slower task (T=2s) resumes and its inputs
have changed. This leads to unpredictable results.
2
T=1s
3
Figure 8-8: Time Overlaps in Faster to Slower Transitions (T = Sample Time)
In Figure 8-8, the faster block executes a second time before the slower block
has completed execution. This can cause unpredictable results because the
input data to the slow task is changing. Data integrity is not guaranteed in this
situation.
To avoid this situation, you must hold the outputs of the 1 second (faster) block
until the 2 second (slower) block finishes executing. The way to accomplish this
is by inserting a Rate Transition block between the 1 second and 2 second
blocks. This guarantees that the input to the slower block does not change
during its execution, ensuring data integrity..
T = 1 sec
Faster
Block
T = 2 sec
Rate
Transition
T = 2 sec
Slower
Block
We assume that the Rate Transition block is used in its default
(Protected/Deterministic) mode.
8-17
8
Models with Multiple Sample Rates
The Rate Transition block executes at the sample rate of the slower block, but
with the priority of the faster block.
t0
t2
2 Sec
Task
T=0.5
T=2s
t1
t0
1 Sec
Task
T=0.5
T=2s
T=1s
RT
t2
T=1s
t3
ZOH
RT
T=1s
T=1s
Time
This ensures that the Rate Transition block executes before the 1 second block
(its priority is higher) and that its output value is held constant while the 2
second block executes (it executes at the slower sample rate).
Slower to Faster Transitions in Simulink
In a model where a slower block drives a faster block, Simulink again computes
the output of the driving block first. During sample intervals where only the
faster block executes, the simulation progresses more rapidly.
The following diagram illustrates the execution sequence.
t0
T = 2 sec
T = 1 sec
Slower
Block
Faster
Block
t1
T=2s
t2
T=1s T=1s
t3
T=2s
T=1s T=1s
Time
As you can see from the preceding diagrams, Simulink can simulate models
with multiple sample rates in an efficient manner. However, Simulink does not
operate in real time.
8-18
Sample Rate Transitions
Slower to Faster Transitions in Real Time
In models where a slower block drives a faster block, the generated code
assigns the faster block a higher priority than the slower block. This means the
faster block is executed before the slower block, which requires special care to
avoid incorrect results.
t0
t2
2 Sec
Task
T = 2 sec
T = 1 sec
Block
Faster
Block
T=2s
1
t0
1 Sec
Task
T=1s
T=2s
2
t1
T=1s
1
t2
t3
T=1s
2
T=1s
t4
T=1s
Time
1 The faster block executes a second time prior to the completion
of the slower block.
2
The faster block executes before the slower block.
Figure 8-9: Time Overlaps in Slower to Faster Transitions
This timing diagram illustrates two problems:
• Execution of the slower block is split over more than one faster block
interval. In this case the faster task executes a second time before the slower
task has completed execution. This means the inputs to the slower task can
change.
• The faster block executes before the slower block (which is backwards from
the way Simulink operates). In this case, the 1 second block executes first;
but the inputs to the faster task have not been computed. This can cause
unpredictable results.
8-19
8
Models with Multiple Sample Rates
To eliminate these problems, you must insert a Rate Transition block between
the slower and faster blocks..
T = 2 sec
T = 2 sec
T = 1 sec
Rate
Transition
Slower
Block
Faster
Block
We assume that the Rate Transition block is used in its default
(Protected/Deterministic) mode.
The picture below shows the timing sequence that results with the added Rate
Transition block.
2
2 Sec
Task
RT
update
T=2s
t1
t0
1 Sec
Task
RT
T=1s
output
t2
1
1
T=1s
t3
1
3
RT
update
T=2s
RT
T=1s
output
1
T=1s
Time
Three key points about this diagram:
• The Rate Transition block output runs in the 1 second task, but only at its
rate (2 seconds). The output of the Rate Transition block feeds the 1 second
task blocks.
• The Rate Transition update uses the output of the 2 second task in its update
of its internal state.
8-20
Sample Rate Transitions
• The Rate Transition update uses the state of the Rate Transition in the 1
second task.
The output portion of a Rate Transition block is executed at the sample rate of
the slower block, but with the priority of the faster block. Since the Rate
Transition block drives the faster block and has effectively the same priority,
it is executed before the faster block. This solves the first problem.
The second problem is alleviated because the Rate Transition block executes at
a slower rate and its output does not change during the computation of the
faster block it is driving.
Note This use of the Rate Transition block changes the model. The output of
the slower block is now delayed by one time step compared to the output
without a Rate Transition block.
8-21
8
Models with Multiple Sample Rates
Singletasking and Multitasking
Execution of a Model: an Example
In this section we will examine how a simple multirate model executes in both
real time and simulation, using a fixed-step solver. We will consider the
operation of both SingleTasking and MultiTasking solver modes.
The example model is shown in Figure 8-10. We will refer to the six blocks of
the model as A through F, as labelled in the block diagram.
Note that the execution order of the blocks (indicated in the upper right of each
block) has been forced into the order shown by assigning higher priorities to
blocks F, E, and D. The ordering shown is one possible valid execution ordering
for this model. (See “Determining Block Update Order” in Using Simulink.)
The execution order is determined by data dependencies between blocks. In a
real-time system, the execution order determines the order in which blocks
execute, within a given time interval or task. In this discussion we will treat
the model’s execution order as a given, since we are concerned with the
allocation of block computations to tasks, and to the scheduling of task
execution.
Figure 8-10: Example Model with Multiple Rates and Transition Blocks
8-22
Singletasking and Multitasking Execution of a Model: an Example
Note The discussion and timing diagrams in this section is based on the
assumption that the Rate Transition blocks are used in the default
(Protected/Deterministic) mode, with the Ensure data integrity during data
transfer and Ensure deterministic data transfer (maximum delay) options
on.
Singletasking Execution
In this section, we will consider the execution of the model when the solver
mode is SingleTasking.
Note that in a singletasking system, if the Block reduction option is on,
fast-to-slow Rate Transition blocks are optimized out of the model. We show
the default case (Block reduction on); therefore block B does not appear in the
timing diagrams in this section.
Table 8-1 shows, for each block in the model, the execution order, sample time,
and whether the block has an output or update computation. Block A does not
have discrete states, and accordingly does not have an update computation.
8-23
8
Models with Multiple Sample Rates
Table 8-1: Execution Order and Sample Times (Singletasking)
Blocks
(in Execution Order)
Sample Time
(in seconds)
Output
Update
F
0.1
Y
Y
E
0.1
Y
Y
D
1
Y
Y
A
0.1
Y
N
C
1
Y
Y
Real-Time Singletasking Execution
Figure 8-11 shows the scheduling of computations when the generated code is
deployed in a real-time system. The generated program is shown running in
real time, under control of interrupts from a 10 Hz timer.
Output: F E D A C
Update:
Time: 0.0
FEA
(wait)
FEDC
0.1
FEDAC
FEA
(wait)
FE
FE
0.2
...
F E D C...
...
...
1.0
Figure 8-11: Singletasking Execution of Model in a Real-Time System
8-24
Singletasking and Multitasking Execution of a Model: an Example
At time 0.0, 1.0, and every second thereafter, both the slow and fast blocks
execute their output computations; this is followed by update computations for
blocks that have states. Within a given time interval, output and update
computations are sequenced in block execution order.
The fast blocks execute on every tick, at intervals of 0.1 sec. Output
computations are followed by update computations.
Note that the system spends some portion of each time interval (labelled
“wait”) idling. During the intervals when only the fast blocks execute, a larger
portion of the interval is spent idling. This illustrates an inherent inefficiency
of SingleTasking mode.
Simulated Singletasking Execution
Figure 8-12 shows the execution of the model in Simulink via the simulation
loop.
Output: F E D A C
Update:
Time: 0.0
FEA
FEDC
0.1
FEA
FE
0.2
FEDAC
FE
...
F E D C ...
...
...
1.0
Figure 8-12: Singletasking Execution of Model in Simulink
Since time is simulated, the placement of ticks represents the iterations of the
simulation loop. Blocks execute in exactly the same order as in Figure 8-11, but
without the constraint of a real-time clock. Therefore there is no idle time
between simulated sample periods.
8-25
8
Models with Multiple Sample Rates
Multitasking Execution
In this section, we will consider the execution of the model when the solver
mode is MultiTasking. Block computations are executed under two tasks,
prioritized by rate:
• The slower task, which gets lower priority, is scheduled to run every second.
We will refer to this as the 1 second task.
• The faster task, which gets higher priority, is scheduled to run 10 times per
second. We will refer to this as the 0.1 second task. The 0.1 second task can
preempt the 1 second task.
Table 8-2 shows, for each block in the model, the execution order, the task
under which the block runs, and whether the block has an output or update
computation.Blocks A and B do not have discrete states, and accordingly do not
have an update computation.
Table 8-2: Task Allocation of Blocks in Multitasking Execution
8-26
Blocks
(in Execution Order)
Task
Output
Update
F
0.1 second task
Y
Y
E
0.1 second task
Y
Y
D
Output promoted to run
under 0.1 second task (see
“Block Priority
Promotions”)
Update runs under 1
second task
Y
Y
A
0.1 second task
Y
N
B
Promoted to run under 0.1
second task (see “Block
Priority Promotions”)
Y
N
C
1 second task
Y
Y
Singletasking and Multitasking Execution of a Model: an Example
Real-Time Multitasking Execution
Figure 8-13 shows the scheduling of computations in MultiTasking solver
mode when the generated code is deployed in a real-time system.The generated
program is shown running in real time, as two tasks under control of interrupts
from a 10 Hz timer.
1 SECOND TASK
Output:
C
C
preemption
preemption
Update:
C
...
DC
...
Time: 0.0
1.0
0.1 SECOND TASK
Output: F E D A B
FEA
FEA
FEDAB
FEA
(wait)
Update:
FE
FE
FE
...
FE
...
Time: 0.0
0.1
0.2
1.0
1.1
Figure 8-13: Multitasking Execution of Model in a Real-Time System
8-27
8
Models with Multiple Sample Rates
Block Priority Promotions. Notice following block “promotions”:
• The rate-transition block B has been promoted to run at higher task priority,
under the 0.1 second task. However, B still executes only at 1-second
intervals, (that is, at every 10th tick of the 1-second task). In other words, B
runs at the higher priority but at the slower rate.
This promotion is required because C requires input from B. Running B at
higher task priority ensures that the output computation of B is always
completed before C needs it.
• The output computation for rate-transition block D has also been promoted
to run at higher task priority, under the 0.1 second task. Like B, D’s output
still executes only at 1-second intervals.
• The update computation for block D runs under the lower-priority 1 second
task, at the same priority as C. This is because the state of D is dependent
upon the output of C.
On each tick, all the outputs and updates for the faster blocks must run before
the lower-priority block (C) gets any run time. Only block C runs entirely in the
1 second task. In Figure 8-13, C does not complete its output computation
within the first 0.1 second tick, so it is preempted by the higher-priority task
at time 0.1. C then resumes and completes, at which point the update function
for D is executed. There is then some idle time before the next tick.
If the computations for block C were to take longer than 1 second, an interrupt
overflow error condition would exist.
Notice that in multitasking mode, the program makes more efficient use of
time than in singletasking mode, as it spends less time in an idle state.
8-28
Singletasking and Multitasking Execution of a Model: an Example
Simulated Multitasking Execution
Figure 8-14 shows the execution of the same model in Simulink, in
MultiTasking solver mode. In this case, Simulink runs all blocks in one thread
of execution, simulating multitasking. No preemption occurs.
1 SECOND
BLOCKS
Output:
C
C
Update:
...
DC
DC
...
Time: 0.0
1.0
0.1 SECOND
BLOCKS
Output: F E D A B
Update:
FEA
FE
FE
FEA
FEDAB
FEA
FE
...
FE
...
Time: 0.0
0.1
0.2
1.0
1.1
Figure 8-14: Multitasking Execution of Model in Simulink
8-29
8
Models with Multiple Sample Rates
8-30
9
Optimizing the Model for
Code Generation
You can optimize memory usage and performance of code generated from your model by Real-Time
Workshop a number of ways. Here we discuss optimization techniques that are common to all target
configurations and code formats. For optimizations specific to a particular target configuration, see
the chapter relevant to that target. Topics covered here include the following:
General Modeling Techniques (p. 9-2)
Optimizations that you can use with any target
configuration
Expression Folding (p. 9-3)
A default optimization that significantly reduces the need
to compute and store temporary results
Conditional Branch Execution (p. 9-25) A default optimization for executing inputs to switch
blocks only as often as required
Block Diagram Performance Tuning
(p. 9-26)
How to efficiently use look-up tables, accumulator
constructs, and data types
Stateflow Optimizations (p. 9-43)
Ways to optimize models containing Stateflow blocks
Simulation Parameters (p. 9-44)
Options on the Simulation Parameters dialog box that
affect code optimization
Compiler Options (p. 9-46)
Hints for helping your compiler build more efficient
executables
9
Optimizing the Model for Code Generation
General Modeling Techniques
The following are techniques that you can use with any code format:
• The slupdate command automatically converts older models to use current
features. Run slupdate on old models.
• Directly inline C code S-functions into the generated code by writing a TLC
file for the S-function. See the Target Language Compiler documentation for
more information on inlining S-functions. Also see “Creating Device Drivers”
on page 14-39 for information on inlining device driver S-functions.
• Use a Simulink data type other than double when possible. The available
data types are Boolean, signed and unsigned 8-, 16-, and 32-bit integers, and
32- and 64-bit floats. A double is a 64-bit float. See Using Simulink for more
information on data types.
• Remove repeated values in lookup table data.
• Use the Merge block to merge the output of function-call subsystems. This
block is particularly helpful when controlling the execution of function-call
subsystems with Stateflow.
This diagram is an example of how to use the Merge block.
9-2
Expression Folding
Expression Folding
Expression folding is a code optimization technique that minimizes the
computation of intermediate results at block outputs and the storage of such
results in temporary buffers or variables. When expression folding is on,
Real-Time Workshop collapses, or “folds,” block computations into single
expressions, instead of generating separate code statements and storage
declarations for each block in the model.
Expression folding can dramatically improve the efficiency of generated code,
frequently achieving results that compare favorably to hand-optimized code. In
many cases, entire groups of model computations fold into a single highly
optimized line of code.
By default, expression folding is on. The Real-Time Workshop code generation
options are configured to use expression folding wherever possible. Most
Simulink blocks support expression folding.
You can also take advantage of expression folding in your own inlined
S-function blocks. See “Supporting Expression Folding in S-Functions” on
page 9-10 for information on how to do this.
In the code generation examples that follow, note that signal storage
optimizations (Signal storage reuse, Buffer reuse and Local block outputs)
are turned on.
Expression Folding Example
As a simple example of how expression folding affects the code generated from
a model, consider the model shown in Figure 9-1.
Figure 9-1: Expression Folding Example Model
9-3
9
Optimizing the Model for Code Generation
With expression folding on, this model generates a single-line output
computation, as shown in this MdlOutputs function.
void MdlOutputs(int_T tid)
{
/* tid is required for a uniform function interface. This system
* is single rate, and in this case, tid is not accessed. */
UNUSED_PARAMETER(tid);
/* Outport: '<Root>/Out1' incorporates:
*
Product: '<Root>/Product'
*
Gain: '<Root>/k1'
*
Inport: '<Root>/In1'
*
Gain: '<Root>/k2'
*
Inport: '<Root>/In2'
*
* Regarding '<Root>/k1':
*
Gain value: rtP.k1_Gain
*
* Regarding '<Root>/k2':
*
Gain value: rtP.k2_Gain
*/
rtY.Out1 = ((rtP.k1_Gain * rtU.i1) * (rtP.k2_Gain * rtU.i2));
}
The generated comments indicate the block computations that were combined
into a single expression. The comments also document the block parameters
that appear in the expression.
With expression folding off, the same model computes temporary results for
both Gain blocks and the Product block before the final output, as shown in this
MdlOutputs function.
void MdlOutputs(int_T tid)
{
/* local block i/o variables */
real_T rtb_s2;
real_T rtb_temp1;
/* tid is required for a uniform function interface. This system
* is single rate, and in this case, tid is not accessed. */
UNUSED_PARAMETER(tid);
9-4
Expression Folding
/* Gain Block: '<Root>/k1'
*
Gain value: rtP.k1_Gain
*/
rtb_temp1 = rtU.i1 * rtP.k1_Gain;
/* Gain Block: '<Root>/k2'
*
Gain value: rtP.k2_Gain
*/
rtb_s2 = rtU.i2 * rtP.k2_Gain;
/* Product Block: '<Root>/Product' */
rtb_temp1 = rtb_temp1 * rtb_s2;
/* Outport Block: '<Root>/Out1' */
rtY.Out1 = rtb_temp1;
}
For a example of expression folding in the context of a more complex model,
link to the exprfolding demo, or type the following command at the MATLAB
prompt.
exprfolding
Using and Configuring Expression Folding
The options described in this section let you control the operation of expression
folding.
Enabling Expression Folding
Expression folding operates only on expressions involving local variables.
Expression folding is therefore available only when both the Signal storage
reuse and Local block outputs code generation options are on.
For a new model, default code generation options are set to use expression
folding. If you are configuring an existing model, you can ensure that
expression folding is turned on as follows:
9-5
9
Optimizing the Model for Code Generation
1 Select the Signal storage reuse option on the Advanced page of the
Simulation Parameters dialog box.
2 Select the Local block outputs option in the General code generation
options category of the Real-Time Workshop pane of the Simulation
Parameters dialog box.
3 Access the expression folding related options by selecting General code
generation options (cont.) from the Category menu of the Real-Time
Workshop pane.
The expression folding options are shown in Figure 9-2. By default, all
expression folding related options are selected, as shown. These options are
detailed in “Expression Folding Options” on page 9-6.
4 If necessary, select the Expression folding option and click Apply.
Figure 9-2: Expression Folding Options
Expression Folding Options
This section discusses the available code generation options related to
expression folding.
9-6
Expression Folding
Expression Folding. This option turns the expression folding feature on or off.
When Expression folding is selected, the Fold unrolled vectors and Enforce
integer downcast options are available.
Alternatively, you can turn expression folding on or off from the MATLAB
command line via the command
set_param(gcs, 'RTWExpressionDepthLimit', val)
If val = 1, expression folding is turned on. If val = 0, expression folding is
turned off.
Fold Unrolled Vectors. We recommend that you leave this option on, as it will
decrease the generated code (ROM) size.
Turning Fold unrolled vectors off will speed up code generation for vector
signals whose widths are less than the Loop rolling threshold (See“Loop
Rolling Threshold Field” on page 2-9). You may want to consider turning Fold
unrolled vectors off if:
• You are concerned with code generation speed.
• You mostly work with scalar signals.
• You mostly work with signals above the loop rolling threshold.
To understand the effect of Fold unrolled vectors, consider the model shown
in this diagram.
The input signals i1 and i2 are vectors of width 3. The input signal elements
are represented in the generated code as members of the rtU structure
(rtU.i1[n] and rtU.i2[n]).
Assuming the model’s loop rolling threshold is greater than 3, (the default
threshold is 5) computations on i1 are not rolled into a for loop. If Fold
9-7
9
Optimizing the Model for Code Generation
unrolled vectors is on, the gain computations for elements of i1 and i2 are
folded into the Outport block computations, as shown in this MdlOutputs
function.
void MdlOutputs(int_T tid)
/* tid is required for a uniform function interface. This system
* is single rate, and in this case, tid is not accessed. */
UNUSED_PARAMETER(tid);
{
/* Outport: <Root>/Out1 incorporates:
*
Product: <Root>/Product
*
Gain: <Root>/k1
*
Inport: <Root>/In1
*
Gain: <Root>/k2
*
Inport: <Root>/In2
*
* Regarding <Root>/k1:
*
Gain value: rtP.k1_Gain
*
* Regarding <Root>/k2:
*
Gain value: rtP.k2_Gain
*/
rtY.Out1[0] = ((rtP.k1_Gain * rtU.i1[0]) * (rtP.k2_Gain * rtU.i2[0]));
rtY.Out1[1] = ((rtP.k1_Gain * rtU.i1[1]) * (rtP.k2_Gain * rtU.i2[1]));
rtY.Out1[2] = ((rtP.k1_Gain * rtU.i1[2]) * (rtP.k2_Gain * rtU.i2[2]));
}
If Fold unrolled Vectors is off, computations for elements of i1 and i2 are
implemented as separate code statements, with intermediate results stored in
temporary variables, as shown in this MdlOutputs function.
void MdlOutputs(int_T tid)
{
/* local block i/o variables */
real_T rtb_s2[3];
real_T rtb_temp1[3];
/* tid is required for a uniform function interface. This system
* is single rate, and in this case, tid is not accessed. */
UNUSED_PARAMETER(tid);
/* Gain: '<Root>/k1' incorporates:
*
Inport: '<Root>/In1'
*
* Regarding '<Root>/k1':
*
Gain value: rtP.k1_Gain
*/
rtb_temp1[0] = rtU.i1[0] * rtP.k1_Gain;
rtb_temp1[1] = rtU.i1[1] * rtP.k1_Gain;
rtb_temp1[2] = rtU.i1[2] * rtP.k1_Gain;
9-8
Expression Folding
/* Gain: '<Root>/k2' incorporates:
*
Inport: '<Root>/In2'
*
* Regarding '<Root>/k2':
*
Gain value: rtP.k2_Gain
*/
rtb_s2[0] = rtU.i2[0] * rtP.k2_Gain;
rtb_s2[1] = rtU.i2[1] * rtP.k2_Gain;
rtb_s2[2] = rtU.i2[2] * rtP.k2_Gain;
/* Product: '<Root>/Product' */
rtb_temp1[0] = rtb_temp1[0] * rtb_s2[0];
rtb_temp1[1] = rtb_temp1[1] * rtb_s2[1];
rtb_temp1[2] = rtb_temp1[2] * rtb_s2[2];
/* Outport:
rtY.Out1[0]
rtY.Out1[1]
rtY.Out1[2]
'<Root>/Out1' */
= rtb_temp1[0];
= rtb_temp1[1];
= rtb_temp1[2];
}
Enforce Integer Downcast . This option refers to 8-bit operations on 16-bit
microprocessors and 8 and 16-bit operations on 32-bit microprocessors. To
ensure consistency between simulation and code generation, the results of 8
and 16-bit integer expressions must be explicitly downcast.
Deselecting this option improves code efficiency. However, the primary effect
of deselecting this option is that expressions involving 8 and 16-bit arithmetic
are less likely to overflow in code than they are in simulation. We recommend
that you turn on Enforce integer downcast for safety. Turn the option off only
if you are concerned with generating the smallest possible code, and you know
that 8 and 16-bit signals will not overflow.
As an example, consider this model.
The following code fragment shows the output computation (within the
MdlOutputs function) when Enforce integer downcast is on. The Gain blocks
are folded into a single expression. In addition to the typecasts generated by
the Type Conversion blocks, each Gain block output is cast to int8_T.
9-9
9
Optimizing the Model for Code Generation
int8_T rtb_Data_Type_Conversion;
.
.
.
rtY.Out1 = (int16_T)(int8_T)(rtP.Gain2_Gain * (int8_T)(rtP.Gain1_Gain *
(int8_T)(rtP.Gain_Gain * rtb_Data_Type_Conversion)));
If Enforce integer downcast is off, the code contains only the typecasts
generated by the Type Conversion blocks, as shown in the following code
fragment.
int8_T rtb_Data_Type_Conversion;
.
.
.
rtY.Out1 = (int16_T)(rtP.Gain2_Gain * (rtP.Gain1_Gain * (rtP.Gain_Gain *
rtb_Data_Type_Conversion)));
Supporting Expression Folding in S-Functions
This section describes how you can take advantage of expression folding to
increase the efficiency of code generated by your own inlined S-function blocks
by calling macros provided in the S-Function API.
This section assumes that you are familiar with:
• Writing inlined S-functions (see “Writing S-Functions” in the Simulink
documentation).
• The Target Language Compiler (see the Target Language Compiler
documentation).
The S-Function API lets you specify whether a given S-Function block should
nominally accept expressions at a given input port. A block should not always
accept expressions. For example, if the address of the signal at the input is
used, expressions should not be accepted at that input, because it is not
possible to take the address of an expression.
The S-Function API also lets you specify whether an expression can represent
the computations associated with a given output port. When you request an
expression at a block’s input or output port, Simulink determines whether or
not it can honor that request, given the block’s context. For example, Simulink
may deny a block’s request to output an expression if the destination block does
not accept expressions at its input; if the destination block has an update
function; or if there are multiple output destinations.
9-10
Expression Folding
The decision to honor or deny a request to output an expression can also depend
on the category of output expression the block uses (see “Categories of Output
Expressions” on page 9–11).
In the sections that follow, we explain:
• When and how you can request that a block accept expressions at an input
port.
• When and how you can request that a block generate expressions at an
outport.
• The conditions under which Simulink will honor or deny such requests.
To take advantage of expression folding in your S-functions, you need to
understand when it is appropriate to request acceptance and generation of
expressions for specific blocks. It is not necessary for you to understand the
algorithm by which Simulink chooses to accept or deny these requests.
However, if you want to trace between the model and the generated code, it will
be helpful to understand some of the more common situations which lead to
denial of a request.
Categories of Output Expressions
When you implement a C-MEX S-function, you can specify whether the code
corresponding to a block’s output is to be generated as an expression. If the
block generates an expression, you must specify that the expression is constant,
trivial, or generic.
A constant output expression is a direct access to one of the block’s parameters.
For example, the output of a Constant block is defined as a constant expression,
because the output expression is simply a direct access to the block’s Value
parameter.
A trivial output expression is an expression that may be repeated, without any
performance penalty, when the output port has multiple output destinations.
For example, the output of a Unit Delay block is defined as a trivial expression,
because the output expression is simply a direct access to the block’s state.
Since the output expression involves no computations, it may be repeated more
than once without degrading the performance of the generated code.
A generic output expression is an expression that should be assumed to have a
performance penalty if repeated. As such, a generic output expression is not
suitable for repeating when the output port has multiple output destinations.
9-11
9
Optimizing the Model for Code Generation
For instance, the output of a Sum block is a generic rather than a trivial
expression because, it is costly to recompute a Sum block output expression as
an input to multiple blocks.
Examples of Trivial and Generic Output Expressions
Consider the block diagram of Figure 9-3. The Delay block has multiple
destinations, yet its output is designated as a trivial output expression, so that
it can be used more than once without degrading the efficiency of the code.
Figure 9-3: Diagram With Delay Block Routed to Multiple Destinations
The following code excerpt shows code generated from the Unit Delay block in
this block diagram. Note that the three root outputs are directly assigned from
the state of the Unit Delay block, which is stored in a field of the global data
structure rtDWork. Since the assignment is direct, involving no expressions,
there is no performance penalty associated with using the trivial expression for
multiple destinations.
void MdlOutputs(int_T tid)
{
...
/* Outport: <Root>/Out1 incorporates:
*
UnitDelay: <Root>/Unit Delay */
rtY.Out1 = rtDWork.Unit_Delay_DSTATE;
/* Outport: <Root>/Out2 incorporates:
*
UnitDelay: <Root>/Unit Delay */
rtY.Out2 = rtDWork.Unit_Delay_DSTATE;
9-12
Expression Folding
/* Outport: <Root>/Out3 incorporates:
*
UnitDelay: <Root>/Unit Delay */
rtY.Out3 = rtDWork.Unit_Delay_DSTATE;
...
}
On the other hand, consider the Sum blocks in Figure 9-4.
Figure 9-4: Diagram With Sum Block Routed to Multiple Destinations
The upper Sum block in Figure 9-4 generates the signal labelled non_triv.
Computation of this output signal involves two multiplications and an
addition. If the Sum block’s output were permitted to generate an expression
even when the block had multiple destinations, the block’s operations would be
duplicated in the generated code. In the case illustrated, the generated
expressions would proliferate to four multiplications and two additions. This
would degrade the efficiency of the program. Accordingly the output of the Sum
block is not allowed to be an expression since it has multiple destinations
The code generated for the block diagram of Figure 9-4 illustrates how code is
generated for Sum blocks with single and multiple destinations.
The Simulink engine does not permit the output of the upper Sum block to be
an expression, since the signal non_triv is routed to two output destinations.
Instead, the result of the multiplication and addition operations is stored in a
temporary variable (rtb_non_triv) that is referenced twice in the statements
that follow, as seen in the code excerpt below.
In contrast, the lower Sum block, which has only a single output destination
(Out2), does generate an expression.
9-13
9
Optimizing the Model for Code Generation
void MdlOutputs(int_T tid)
{
/* local block i/o variables */
real_T rtb_non_triv;
real_T rtb_Sine_Wave;
/* Sum: <Root>/Sum incorporates:
*
Gain: <Root>/Gain
*
Inport: <Root>/u1
*
Gain: <Root>/Gain1
*
Inport: <Root>/u2
*
* Regarding <Root>/Gain:
*
Gain value: rtP.Gain_Gain
*
* Regarding <Root>/Gain1:
*
Gain value: rtP.Gain1_Gain
*/
rtb_non_triv = (rtP.Gain_Gain * rtU.u1) + (rtP.Gain1_Gain *
rtU.u2);
/* Outport: <Root>/Out1 */
rtY.Out1 = rtb_non_triv;
/* Sin Block: <Root>/Sine Wave */
rtb_Sine_Wave = rtP.Sine_Wave_Amp *
sin(rtP.Sine_Wave_Freq * rtmGetT(rtM_model) +
rtP.Sine_Wave_Phase) + rtP.Sine_Wave_Bias;
/* Outport: <Root>/Out2 incorporates:
*
Sum: <Root>/Sum1
*/
rtY.Out2 = (rtb_non_triv + rtb_Sine_Wave);
}
Specifying the Category of an Output Expression
The S-Function API provides macros that let you declare whether an output of
a block should be an expression, and if so, to specify the category of the
9-14
Expression Folding
expression. Table 9-1 specifies when to declare a block output to be a constant,
trivial, or generic output expression.
Table 9-1: Types of Output Expressions
Category of
Expression
When to Use
Constant
Use only if block output is a direct memory access to a
block parameter
Trivial
Use only if block output is an expression that may
appear multiple times in the code without reducing
efficiency (for example, a direct memory access to a
field of the DWork vector, or a literal)
Generic
Use if output is an expression, but not constant or
trivial
You must declare outputs as expressions in the mdlSetWorkWidths function,
using macros defined in the S-Function API. The macros have the following
arguments:
• SimStruct *S: pointer to the block’s SimStruct.
• int idx: zero-based index of the output port.
• bool value: pass in TRUE if the port generates output expressions.
The following macros are available for setting an output to be a constant,
trivial, or generic expression:
• void
• void
• void
ssSetOutputPortConstantOutputExprInRTW(SimStruct *S, int idx, bool value)
ssSetOutputPortTrivialOutputExprInRTW(SimStruct *S, int idx, bool value)
ssSetOutputPortOutputExprInRTW(SimStruct *S, int idx, bool value)
The following macros are available for querying the status set by any prior calls
to the macros above:
• bool
• bool
• bool
ssGetOutputPortConstantOutputExprInRTW(SimStruct *S, int idx)
ssGetOutputPortTrivialOutputExprInRTW(SimStruct *S, int idx)
ssGetOutputPortOutputExprInRTW(SimStruct *S, int idx)
9-15
9
Optimizing the Model for Code Generation
Note that the set of generic expressions is a superset of the set of trivial
expressions, and the set of trivial expressions is a superset of the set of constant
expressions.
Therefore, when you query an output that has been set to be a constant
expression with ssGetOutputPortTrivialOutputExprInRTW, it will return
True. A constant expression is considered a trivial expression, because it is a
direct memory access that may be repeated without degrading the efficiency of
the generated code.
Similarly, an output that has been configured to be a constant or trivial
expression will return true when queried for its status as a generic expression.
Acceptance or Denial of Requests for
Input Expressions
A block can request that its output be represented in code as an expression.
Such a request may be denied if the destination block cannot accept
expressions at its input port. Furthermore, conditions independent of the
requesting block and its destination block(s) can prevent acceptance of
expressions.
In this section, we will discuss block-specific conditions under which requests
for input expressions are denied. For information on other conditions that
prevent acceptance of expressions, see “Generic Conditions for Denial of
Requests to Output Expressions” on page 9-19.
A block should not be configured to accept expressions at its input port under
the following conditions:
• The block must take the address of its input data. It is not possible to take
the address of most types of input expressions.
• The code generated for the block will reference the input more than once (e.g.
the Abs or Max blocks). This would lead to duplication of a potentially
complex expression and a subsequent degradation of code efficiency.
If a block refuses to accept expressions at an input port, then no block that is
connected to that input port is permitted to output a generic or trivial
expression.
9-16
Expression Folding
A request to output a constant expression is never denied, because there is no
performance penalty for a constant expression, and it is always possible to take
the parameter’s address.
Example: Acceptance and Denial of Expressions at Block Inputs
This example illustrates how various built-in blocks handle requests to accept
different categories of expressions at their inputs.
The sample model of Figure 9-5 contains:
• Two Gain blocks. Gain blocks request their destination blocks to accept
generic expressions.
• An Abs block. This block always denies expressions at its input port. The Abs
block code uses the macro rt_ABS(u), which evaluates the input u twice. (see
the TLC implementation of the Abs block in
matlabroot/rtw/c/tlc/blocks/absval.tlc.)
• A Trigonometric Function block. This block accepts expressions at its input
port.
Figure 9-5: Two Gain Blocks Requesting to Output an Expression
The Gain1 block’s request to output an expression is denied by the Abs block.
The Gain2 block's request to output an expression is accepted by the
Trigonometric Function block.
The generated code is shown in the code excerpt below. Note that the output of
the Gain1 block is stored in the temporary variable rtb_Gain1, rather than
generating an input expression to the Abs block.
void MdlOutputs(int_T tid)
{
/* local block i/o variables */
real_T rtb_Gain1;
9-17
9
Optimizing the Model for Code Generation
/* Gain: '<Root>/Gain1' incorporates:
*
Inport: '<Root>/In1'
*
* Regarding '<Root>/Gain':
*
Gain value: 2.0
*/
rtb_Gain1 = rtU.In1 * 2.0;
/* Outport: '<Root>/Out1' incorporates:
*
Abs: '<Root>/Abs'
*/
rtY.Out1 = rt_ABS(rtb_Gain1);
/* Outport: '<Root>/Out2' incorporates:
*
Trigonometry: '<Root>/Trigonometric Function'
*
Gain: '<Root>/Gain2'
*
Inport: '<Root>/In2'
*
* Regarding '<Root>/Gain2':
*
Gain value: 2.0
*/
rtY.Out2 = sin((2.0 * rtU.In2));
}
Using the S-Function API to Specify Input Expression Acceptance
The S-Function API provides macros that let you:
• Specify whether a block input should accept non-constant expressions (i.e.
trivial or generic expressions).
• Query whether a block input accepts non-constant expressions.
By default, block inputs do not accept non-constant expressions.
You should call the macros in your mdlSetWorkWidths function. The macros
have the following arguments:
• SimStruct *S: pointer to the block’s SimStruct.
• int idx: zero-based index of the input port.
• bool value: pass in TRUE if the port accepts input expressions; otherwise
pass in FALSE.
9-18
Expression Folding
The macro available for specifying whether or not a block input should accept
a non-constant expression is as follows:
void ssSetInputPortAcceptExprInRTW(SimStruct *S, int portIdx, bool value)
The corresponding macro available for querying the status set by any prior
calls to ssSetInputPortAcceptExprInRTW is as follows:
bool ssGetInputPortAcceptExprInRTW(SimStruct *S, int portIdx)
Generic Conditions for Denial of Requests to Output Expressions
Even after a specific block requests that it be allowed to generate an output
expression, that request may be denied, for generic reasons. These reasons
include, but are not limited to:
• The output expression is non-trivial, and the output has multiple
destinations
• The output expression is non-constant, and the output is connected to at
least one destination that does not accept expressions at its input port
• The output is a test point
• The output has been assigned an external storage class
• The output must be stored using global data (e.g. is an input to a merge
block, or a block with states)
• The output signal is complex
You do not need to consider these generic factors when deciding whether or not
to utilize expression folding for a particular block. However, these rules may be
helpful when examining generated code, and analyzing cases where the
expression folding optimization is suppressed.
Utilizing Expression Folding in Your TLC Block
Implementation
To take advantage of expression folding, an inlined S-Function must be
modified in two ways:
• It must tell Simulink whether it generates or accepts expressions at its input
ports, as described in “Using the S-Function API to Specify Input Expression
Acceptance” on page 9-18.
9-19
9
Optimizing the Model for Code Generation
• It must tell Simulink whether it generates or accepts expressions at its
output ports, as described in “Categories of Output Expressions” on
page 9-11.
• The TLC implementation of the block must be modified.
In this section, we discuss required modifications to the TLC implementation.
Expression Folding Compliance
In the BlockInstanceSetup function of your S-function, you must ensure that
your block registers that it is compliant with expression folding. If you fail to
do this, any expression folding requested or allowed at the block’s outputs or
inputs will be disabled, and temporary variables will be utilized. To register
expression folding compliance, call the TLC library function
%LibBlockSetIsExpressionCompliant (block)
Note that you can conditionally disable expression folding at the inputs and
outputs of a block by making the call to this function conditionally.
If you have overridden one of the TLC block implementations provided by
Real-Time Workshop with your own implementation, you should not make the
above call until you have updated your implementation, as described by the
guidelines for expression folding in the following sections.
Outputting Expressions
The BlockOutputSignal function is used to generate code for a scalar output
expression, or one element of a non-scalar output expression. If your block
outputs an expression, you should add a BlockOutputSignal function. The
prototype of the BlockOutputSignal is
%function BlockOutputSignal(block,system,portIdx,ucv,lcv,idx,retType) void
The arguments to BlockOutputSignal are as follows:
• block: the record for the block for which an output expression is being
generated.
• system: the record for the system containing the block.
• portIdx: zero-based index of the output port for which an expression is being
generated.
• ucv: user control variable defining the output element for which code is being
generated.
9-20
Expression Folding
• lcv: loop control variable defining the output element for which code is being
generated
• idx: signal index defining the output element for which code is being
generated
• retType: string defining the type of signal access desired:
"Signal" specifies the contents or address of the output signal.
"SignalAddr" specifies the address of the output signal.)
The BlockOutputSignal function returns an appropriate text string for the
output signal or address. The string should enforce the precedence of the
expression by utilizing opening and terminating parentheses, unless the
expression consists of a function call. The address of an expression may only be
returned for a constant expression; it is the address of the parameter whose
memory is being accessed. The code implementing the BlockOutputSignal
function for the Constant block is shown below.
%% Function: BlockOutputSignal =================================================
%% Abstract:
%%
Return the appropriate reference to the parameter. This function *may*
%%
be used by Simulink when optimizing the Block IO data structure.
%%
%function BlockOutputSignal(block,system,portIdx,ucv,lcv,idx,retType) void
%switch retType
%case "Signal"
%return LibBlockParameter(Value,ucv,lcv,idx)
%case "SignalAddr"
%return LibBlockParameterAddr(Value,ucv,lcv,idx)
%default
%assign errTxt = "Unsupported return type: %<retType>"
%<LibBlockReportError(block,errTxt)>
%endswitch
%endfunction
The code implementing the BlockOutputSignal function for the Relational
Operator block is shown below.
9-21
9
Optimizing the Model for Code Generation
%% Function: BlockOutputSignal =================================================
%% Abstract:
%%
Return an output expression. This function *may*
%%
be used by Simulink when optimizing the Block IO data structure.
%%
%function BlockOutputSignal(block,system,portIdx,ucv,lcv,idx,retType) void
%switch retType
%case "Signal"
%assign logicOperator = ParamSettings.Operator
%if ISEQUAL(logicOperator, "~=")
%assign op = "!="
%elseif ISEQUAL(logicOperator, "==")
%assign op = "=="
%else
%assign op = logicOperator
%endif
%assign u0 = LibBlockInputSignal(0, ucv, lcv, idx)
%assign u1 = LibBlockInputSignal(1, ucv, lcv, idx)
%return "(%<u0> %<op> %<u1>)"
%default
%assign errTxt = "Unsupported return type: %<retType>"
%<LibBlockReportError(block,errTxt)>
%endswitch
%endfunction
Expression Folding for Blocks with Multiple Outputs
When a block has a single output, the Outputs function in the block’s TLC file
is called only if the output is not an expression. Otherwise, the
BlockOutputSignal function is called.
If a block has multiple outputs, the Outputs function will be called if any output
port is not an expression. The Outputs function should guard against
generating code for output ports that are expressions. This is achieved by
guarding sections of code corresponding to individual output ports with calls to
LibBlockOutputSignalIsExpr().
For example, consider an S-Function with two inputs and two outputs, where:
• The first output, y0, is equal to two times the first input
• The second output, y1, is equal to four times the second input.
9-22
Expression Folding
The Outputs and BlockOutputSignal functions for the S-function are shown in
the following code excerpt.
%% Function: BlockOutputSignal =================================================
%% Abstract:
%%
Return an output expression. This function *may*
%%
be used by Simulink when optimizing the Block IO data structure.
%%
%function BlockOutputSignal(block,system,portIdx,ucv,lcv,idx,retType) void
%switch retType
%assign u = LibBlockInputSignal(portIdx, ucv, lcv, idx)
%case "Signal"
%if portIdx == 0
%return "(2 * %<u>)"
%elseif portIdx == 1
%return "(4 * %<u>)"
%endif
%default
%assign errTxt = "Unsupported return type: %<retType>"
%<LibBlockReportError(block,errTxt)>
%endswitch
%endfunction
%% Function: Outputs =================================================
%% Abstract:
%%
Compute output signals of block
%%
%function Outputs(block,system) Output
%roll sigIdx = RollRegions, lcv = RollThreshold, block, "Roller", rollVars
%assign u0 = LibBlockInputSignal(0, ucv, lcv, idx)
%assign u1 = LibBlockInputSignal(1, ucv, lcv, idx)
%assign y0 = LibBlockOutputSignal(0, ucv, lcv, idx)
%assign y1 = LibBlockOutputSignal(1, ucv, lcv, idx)
if !LibBlockOutputSignalIsExpr(0)
%<y0> = 2 * %<u0>;
%endif
%if !LibBlockOutputSignalIsExpr(1)
%<y1> = 4 * %<u1>;
%endif
%endroll
%endfunction
Comments for Blocks That Are Expression Folding Compliant
In the past, all blocks preceded their outputs code with comments of the form
/* %<Type> Block: %<Name> */
When a block is expression folding compliant, the initial line shown above is
generated automatically. You should not include the comment as part of the
9-23
9
Optimizing the Model for Code Generation
block’s TLC implementation. Additional information should be registered
using the LibCacheBlockComment function.
The LibCacheBlockComment function takes a string as an input, defining the
body of the comment, except for the opening header, the final newline of a
single or multi-line comment, and the closing trailer.
The following TLC code illustrates registering a block comment. Note the use
of the function LibBlockParameterForComment, which returns a string,
suitable for a block comment, specifying the value of the block parameter.
%openfile commentBuf
$c(*) Gain value: %<LibBlockParameterForComment(Gain)>
%closefile commentBuf
%<LibCacheBlockComment(block, commentBuf)>
9-24
Conditional Branch Execution
Conditional Branch Execution
Conditional input branch execution is a Simulation and code generation
optimization technique that improves model execution when the model
contains Switch and Multiport Switch blocks. By default, the Real-Time
Workshop code generation options are configured to use the conditional input
branch optimization.
When Conditional input branch optimization is on, instead of executing all
blocks driving the Switch block input ports at each time step, only the blocks
required to compute the control input and the data input selected by the control
input are executed.
You can turn conditional input branch optimization on or off by selecting the
Conditional input branch option on the Advanced pane of the Simulation
Parameters dialog box.
For a example of conditional input branch optimization demo, use this link to
the condinputexec demo, or type the following command at the MATLAB
prompt.
condinputexec
9-25
9
Optimizing the Model for Code Generation
Block Diagram Performance Tuning
Certain block constructs in Simulink will run faster, or require less code or
data memory, than other seemingly equivalent constructs. Knowing the
trade-offs between similar blocks and block parameter options will enable you
to create Simulink models that have intuitive diagrams, and to produce the
tight code that you want from Real-Time Workshop. Many of the options and
constructs discussed in this section will improve the simulation speed of the
model itself, even without code generation.
Look-Up Tables and Polynomials
Simulink provides several blocks that allow approximation of functions. These
include blocks that perform direct, interpolated and cubic spline lookup table
operations, and a polynomial evaluation block.
There are currently six different blocks in Simulink that perform lookup table
operations:
• Look-Up Table
• Look-Up Table (2-D)
• Look-Up Table (n-D)
• Direct Look-Up Table (n-D)
• PreLook-Up Index Search
• Interpolation (n-D) Using PreLook-Up Index Search
In addition, the Repeating Sequence block uses a lookup table operation, the
output of which is a function of the real-time (or simulation-time) clock.
To get the most out of the following discussion, you should familiarize yourself
with the features of these blocks, as documented in Using Simulink.
Each type of lookup table block has its own set of options and associated
trade-offs. The examples in this section show how to use lookup tables
effectively. The techniques demonstrated here will help you achieve maximal
performance with minimal code and data sizes.
Multi-Channel Nonlinear Signal Conditioning
Figure 9-6 shows a Simulink model that reads input from two 8-channel,
high-speed 8-bit analog/digital converters (ADCs). The ADCs are connected to
9-26
Block Diagram Performance Tuning
Type K thermocouples through a gain circuit with an amplification of 250.
Since the popular Type K thermocouples are highly nonlinear, there is an
international standard for converting their voltages to temperature. In the
range of 0 to 500 degrees Celsius, this conversion is a tenth-order polynomial.
One way to perform the conversion from ADC readings (0-255) into
temperature (in degrees Celsius) is to evaluate this polynomial. In the best
case, the polynomial evaluation requires 9 multiplications and 10 additions per
channel.
A polynomial evaluation is not the fastest way to convert these 8-bit ADC
readings into measured temperature. Instead, the model uses a Direct Look-Up
(n-D) Table block (named TypeK_TC) to map 8-bit values to temperature
values. This block performs one array reference per channel.
Figure 9-6: Direct Look-Up Table (n-D) Block Conditions ADC Input
The block’s table parameter is populated with 256 values that correspond to
the temperature at an ADC reading of 0, 1, 2, … up to 255. The table data,
calculated in MATLAB, is stored in the workspace variable TypeK_0_500. The
block’s Table data parameter field references this variable, as shown in
Figure 9-7.
9-27
9
Optimizing the Model for Code Generation
Figure 9-7: Parameters of Direct Look-Up Table (n-D) Block
The model uses a Mux block to collect all similar signals (e.g., Type K
thermocouple readings) and feed them into a single Direct Look-Up Table
block. This is more efficient than using one Direct Look-Up Table block per
device. If multiple blocks share a common parameter (such as the table in this
example), Real-Time Workshop creates only one copy of that parameter in the
generated code.
This is the recommended approach for signal conditioning when the size of the
table can fit within your memory constraints. In this example, the table stores
256 double (8-byte) values, utilizing 2 KB of memory.
Note that the TypeK_TC block processes 16 channels of data sequentially.
Real-Time Workshop generates the following code for the TypeK_TC block
shown in Figure 9-6.
/* (LookupNDDirect) Block: <Root>/TypeK_TC */
/* 1-dimensional Direct Look-Up Table returning 16 Scalars */
{
int_T i1;
const uint8_T *u0 = &rtb_s1_Data_Type_Conversion[0];
real_T *y0 = &rtb_root_TypeK_TC[0];
for (i1=0; i1 < 8; i1++) {
9-28
Block Diagram Performance Tuning
y0[i1] = (rtP.root_TypeK_TC_table[(uint8_T)u0[i1]]);
}
u0 = &rtb_s2_Data_Type_Conversion[0];
y0 = &rtb_root_TypeK_TC[8];
for (i1=0; i1 < 8; i1++) {
y0[i1] = (rtP.root_TypeK_TC_table[(uint8_T)u0[i1]]);
}
}
Notice that the core of each loop is one line of code that directly retrieves a table
element from the table and places it in the block output variable. There are two
loops in the generated code because the two simulated ADCs are not merged
into a contiguous memory array in the Mux block. Instead, to avoid a copy
operation, the Direct Look-Up Table block performs the lookup on two sets of
data using a single table array (rtP.root_TypeK_TC_table[]).
If the input accuracy for your application (not to be confused with the number
of I/O bits) is 24 bits or less, you can use a single precision table for signal
conditioning. Then, cast the lookup table output to double precision for use in
the rest of the block diagram. This technique, shown in Figure 9-8, causes no
loss of precision.
Figure 9-8: Single Precision Lookup Table Output Is Cast to Double Precision
Note that a direct lookup table covering 24 bits of accuracy would require 64
megabytes of memory, which is typically not practical. To create a single
precision table, use the MATLAB single() cast function in your table
calculations. Alternatively, you can perform the type cast directly in the Table
data parameter, as shown in Figure 9-9.
9-29
9
Optimizing the Model for Code Generation
Figure 9-9: Type Casting Table Data in a Direct Look-Up Block
When table size becomes impractical, you must use other nonlinear techniques,
such as interpolation or polynomial techniques. The Look-Up Table (n-D) block
supports linear interpolation and cubic spline interpolation.The Polynomial
block supports evaluation of noncomplex polynomials.
Compute-Intensive Equations
The blocks described in this section are useful for simplifying fixed, complex
relationships that are normally too time consuming to compute in real time.
The only practical way to implement some compute-intensive functions or
arbitrary nonlinear relationships in real time is to use some form of lookup
table. On processors that do not have floating-point instructions, even
functions like sqrt() can become too expensive to evaluate in real time.
An approximation to the nonlinear relationship in a known range will work in
most cases. For example, your application might require a square root
calculation that your target processor’s instruction set does not support. The
illustration below shows how you can use a Look-Up Table block to calculate
an approximation of the square root function that covers a given range of the
function.
9-30
Block Diagram Performance Tuning
The interpolated values are plotted on the block icon.
For more accuracy on widely spaced points, use a cubic spline interpolation in
the Look-Up Table (n-D) block, as shown below.
9-31
9
Optimizing the Model for Code Generation
Techniques available in Simulink include n-dimensional support for direct
lookup, linear interpolations in a table, cubic spline interpolations in a table,
and 1-D real polynomial evaluation.
The Look-Up Table (n-D) block supports flat interval lookup, linear
interpolation and cubic spline interpolation. Extrapolation for the Look-Up
Table (n-D) block can either be disabled (clipping) or enabled for linear or
spline extrapolations.
The icons for the Direct Look-Up Table (n-D) and Look-Up Table (n-D) blocks
change depending on the type of interpolation selected and the number of
dimensions in the table, as illustrated below.
Tables with Repeated Points
The Look-Up Table and Look-Up Table (2-D) blocks, shown below, support
linear interpolation with linear extrapolation. In these blocks, the row and
column parameters can have repeated points, allowing pure step behavior to be
mixed in with the linear interpolations. Note that this capability is not
supported by the Look-Up Table (n-D) block.
9-32
Block Diagram Performance Tuning
Slowly vs. Rapidly Changing
Look-Up Table Block Inputs
You can optimize lookup table operations using the Look-Up Table (n-D) block
for efficiency if you know the input signal’s normal rate of change. Figure 9-10
shows the parameters for the Look-Up Table (n-D) block.
Figure 9-10: Parameter Dialog for the Look-Up Table (n-D) Block
If you do not know the input signal’s normal rate of change in advance, it would
be better to choose the Binary Search option for the index search in the
Look-Up Table (n-D) block and the PreLook-Up Index Search block.
Regardless of signal behavior, if the table’s breakpoints are evenly spaced, it is
best to select the Evenly Spaced Points option from the Look-Up Table (n-D)
block’s parameter dialog.
9-33
9
Optimizing the Model for Code Generation
If the breakpoints are not evenly spaced, first decide which of the following best
describes the input signal behavior.
• Behavior 1: The signal stays in a given breakpoint interval from one time
step to the next. When the signal moves to a new interval, it tends to move
to an adjacent interval.
• Behavior 2: The signal has many discontinuities. It jumps around in the
table from one time step to the next, often moving three or more intervals per
time step.
Given behavior 1, the best optimization for a given lookup table is to use the
Linear search option and Begin index searches using previous index
results options, as shown below.
Given behavior 2, the Begin index searches using previous index results
option does not necessarily improve performance. Choose the Binary Search
option, as shown below.
The choice of an index search method can be more complicated for lookup table
operations of two or more dimensions with linear interpolation. In this case,
several signals are input to the table. Some inputs may have evenly spaced
points, while others may exhibit behavior 1 or behavior 2.
Here it may be best to use PreLook-Up Index Search blocks with different
search methods (evenly spaced, linear search or binary search) chosen
according to the input signal characteristics. The outputs of these search blocks
9-34
Block Diagram Performance Tuning
are then connected to an Interpolation (n-D) Using PreLook-Up Index Search
block, as shown in the block diagram below.
You can configure each PreLook-Up Index Search block independently to use
the best search algorithm for the breakpoints and input time variation cases.
Multiple Tables with Common Inputs
The index search can be the most time consuming part of flat or linear
interpolation calculations. In large block diagrams, lookup table blocks often
have the same input values as other lookup table blocks. If this is the case in
your block diagram, you can obtain a large savings in computation time by
making the breakpoints common to all tables. This savings is obtained by using
one set of PreLook-Up Index Search blocks to perform the searches once for all
tables, so that only the interpolation remains to be calculated. Figure 9-11 is
an example of a block diagram that can be optimized by this method.
Figure 9-11: Before Optimization
9-35
9
Optimizing the Model for Code Generation
Assume that Table A’s breakpoints are the same as Table B’s first input
breakpoints, and that Table C’s breakpoints are the same as Table B’s second
input breakpoints.
A 50% reduction in index search time is obtained by pulling these common
breakpoints out into a pair of PreLook-Up Index Search blocks, and using
Interpolation (n-D) Using PreLook-Up Index Search blocks to perform the
interpolation. Figure 9-12 shows the optimized block diagram.
Figure 9-12: After Optimization
In Figure 9-12, the Look-Up Table (n-D) blocks have been replaced with
Interpolation (n-D) Using PreLook-Up blocks.The PreLook-Up Index Search
blocks have been added to perform the index searches separately from the
interpolations, in order to realize the savings in computation time.
In large controllers and simulations, it is not uncommon for hundreds of
multidimensional tables to rely on a dozen or so breakpoint sets. Using the
optimization technique shown in this example, you can greatly increase the
efficiency of your application.
Accumulators
Simulink recognizes the block diagram shown in Figure 9-13 as an
accumulator. An accumulator construct — comprising a Constant block, a Sum
block, and feedback through a Unit Delay block — is recognized anywhere
across a block diagram, or within subsystems at lower levels.
9-36
Block Diagram Performance Tuning
Figure 9-13: An Accumulator Algorithm
By using the Block reduction option, you can significantly optimize code
generated from an accumulator. Turn this option on in the Advanced page of
the Simulink Simulation parameters dialog, as shown in Figure 9-14.
Figure 9-14: Block Reduction Option
With the Block reduction option on, Simulink creates a synthesized block,
Sum_synth_accum. This synthesized block replaces the block diagram of
Figure 9-13, resulting in a simple increment calculation.
void MdlOutputs(int_T tid)
{
/* UnadornAccum Block: <Root>/Sum_synth_accum */
rtB.Sum_synth_accum++;
9-37
9
Optimizing the Model for Code Generation
/* Outport Block: <Root>/Out1 */
rtY.Out1 = rtB.Sum_synth_accum;
}
With Block reduction turned off, the generated code reflects the block
diagram more literally, but less efficiently.
void MdlOutputs(int_T tid)
{
/* Expression for <Root>/Sum incorporates: */
/*
Constant Block: <Root>/Constant */
/*
UnitDelay Block: <Root>/Unit Delay */
/* Sum Block: <Root>/Sum */
rtB.Sum = 1.0 + rtDWork.Unit_Delay_DSTATE;
/* Outport Block: <Root>/Out1 */
rtY.Out1 = rtB.Sum;
}
Use of Data Types
In most processors, the use of integer data types can result in a significant
reduction in data storage requirements, as well as a large increase in the speed
of operation. You can achieve large performance gains on most processors by
identifying those portions of your block diagram that are really integer
calculations (such as accumulators), and implementing them with integer data
types.
Floating-point DSP targets are an obvious exception to this rule.
The accumulator from the previous example used 64-bit floating-point
calculations by default. The block diagram in Figure 9-14 implements the
accumulator with 16-bit integer operations.
9-38
Block Diagram Performance Tuning
Figure 9-15: Accumulator Implemented with 16-bit Integers
If the Saturate on integer overflow option of the Sum block is turned off, the
code generated from the integer implementation looks the same as code
generated from the floating-point block diagram. However, since
Sum_synth_accum is performing integer arithmetic internally, the accumulator
executes more efficiently.
Note that, by default, the Saturate on integer overflow option is on. This
option generates extra error-checking code from the integer implementation,
as in the following example.
void MdlOutputs(int_T tid)
{
/* UnadornAccum Block: <Root>/Sum_synth_accum */
{
int16_T tmpVar = rtB.Sum_synth_accum;
rtB.Sum_synth_accum = tmpVar + (1);
if ((tmpVar >= 0) && ((1) >= 0) && (rtB.Sum_synth_accum < 0)) {
rtB.Sum_synth_accum = MAX_int16_T;
} else if ((tmpVar < 0) && ((1) < 0) && (rtB.Sum_synth_accum >= 0)) {
rtB.Sum_synth_accum = MIN_int16_T;
}
}
/* Outport Block: <Root>/Out1 */
rtY.Out1 = rtB.Sum_synth_accum;
}
The floating-point implementation would not have generated the saturation
error checks, which apply only to integers. When using integer data types,
consider whether or not you need to generate saturation checking code.
Figure 9-16 shows an efficient way to add reset capability to the accumulator.
When resetSig is greater than or equal to the threshold of the Switch block,
the Switch block passes the reset value (0) back into the accumulator.
9-39
9
Optimizing the Model for Code Generation
Figure 9-16: Integer Accumulator with Reset via External Input
The size of the resultant code is minimal. The code uses no floating-point
operations.
void MdlOutputs(int_T tid)
{
/* local block i/o variables */
int16_T rtb_temp3;
/* UnitDelay Block: <Root>/accumState */
rtb_temp3 = rtDWork.accumState_DSTATE;
/* Expression for <Root>/Sum incorporates: */
/*
Constant Block: <Root>/Increment */
/* Sum Block: <Root>/Sum */
{
int16_T tmpVar1 = 0;
int16_T tmpVar2;
/* port 0 */
tmpVar1 = (1);
/* port 1 */
tmpVar2 = tmpVar1 + rtb_temp3;
if ((tmpVar1 >= 0) && (rtb_temp3 >= 0) && (tmpVar2 < 0)) {
tmpVar2 = MAX_int16_T;
} else if ((tmpVar1 < 0) && (rtb_temp3 < 0) && (tmpVar2 >= 0)) {
tmpVar2 = MIN_int16_T;
}
rtb_temp3 = tmpVar2;
}
/* Outport Block: <Root>/accumVal */
rtY.accumVal = rtb_temp3;
9-40
Block Diagram Performance Tuning
/* Expression for <Root>/Switch incorporates: */
/*
Inport Block: <Root>/resetSig */
/*
Constant Block: <Root>/ResetValue */
/* Switch Block: <Root>/Switch */
if (rtU.resetSig) {
rtB.Switch = (0);
} else {
rtB.Switch = rtb_temp3;
}
}
In this example, it would be easy to use an input to the system as the reset
value, rather than a constant.
Generating Pure Integer Code
The Real-Time Workshop Embedded Coder target provides the Integer code
only option to ensure that generated code contains no floating-point data or
operations. When this option is selected, an error is raised if any noninteger
data or expressions are encountered during compilation of the model. The error
message reports the offending blocks and parameters.
If pure integer code generation is important to your design, you should consider
using the Real-Time Workshop Embedded Coder target (or a target of your
own, based on the Real-Time Workshop Embedded Coder target).
To generate pure integer code, select ERT code generation options (1) from
the Category menu in the Real-Time Workshop pane. Then select the Integer
code only option, as shown below.
9-41
9
Optimizing the Model for Code Generation
The Real-Time Workshop Embedded Coder target offers many other
optimizations. See the Real-Time Workshop Embedded Coder documentation
for further information.
Data Type Optimizations with Fixed-Point Blockset
and Stateflow
The Fixed-Point Blockset (a separate product) is designed to deliver the highest
levels of performance for noninteger algorithms on processors lacking
floating-point hardware. The Fixed-Point Blockset’s code generation in
Real-Time Workshop implements calculations using a processor’s integer
operations. The code generation strategy maps the integer value set to a range
of expected real world values to achieve the high efficiency.
Finite-state machine or flowchart constructs can often represent decision logic
(or mode logic) efficiently. Stateflow (a separate product) provides these
capabilities. Stateflow, which is fully integrated into Simulink, supports
integer data-typed code generation.
9-42
Stateflow Optimizations
Stateflow Optimizations
If your model contains Stateflow blocks, select the Use Strong Data Typing
with Simulink I/O check box (on the Chart Properties dialog box) on a
chart-by-chart basis.
See the Stateflow User’s Guide for more information about the Chart
Properties dialog box.
9-43
9
Optimizing the Model for Code Generation
Simulation Parameters
Options on each page of the Simulation Parameters dialog box affect the
generated code.
Advanced Page
• Turn on the Signal storage reuse option. The directs Real-Time Workshop
to store signals in reusable memory locations. It also enables the Local block
outputs option (see “General Code Generation Options” on page 9-45).
Disabling Signal storage reuse makes all block outputs global and unique,
which in many cases significantly increases RAM and ROM usage.
• Enable strict Boolean type checking by selecting the Boolean logic signals
option.
Selecting this check box is recommended. Generated code will require less
memory, because a Boolean signal typically requires one byte of storage
while a double signal requires eight bytes of storage.
• Select the Inline parameters check box. Inlining parameters reduces global
RAM usage, since parameters are not declared in the global parameters
structure. Note that you can override the inlining of individual parameters
by using the Model Parameter Configuration dialog box.
9-44
Simulation Parameters
• Consider using the Parameter pooling option if you have multiple block
parameters referring to workspace locations that are separately defined but
structurally identical. See “Parameter Pooling Option” on page 2-29 for
further information.
General Code Generation Options
To access these options, select General code generation options or General
code generation options (cont.) from the Category menu on the Real-Time
Workshop pane.
• Set an appropriate Loop rolling threshold. The loop rolling threshold
determines when a wide signal should be wrapped into a for loop and when
it should be generated as a separate statement for each element of the signal
See “Loop Rolling Threshold Field” on page 2-9 for details on loop rolling.
• Select the Inline invariant signals option. Real-Time Workshop will not
generate code for blocks with a constant (invariant) sample time.
• Select the Local block outputs option. Block signals will be declared locally
in functions instead of being declared globally (when possible). You must
turn on the Signal storage reuse option in the Advanced page to enable the
Local block outputs check box.
• Select the Expression folding option, discussed in “Expression Folding” on
page 9-3.
• Select the Buffer reuse option. This option can reduce stack size. See “Buffer
Reuse Option” on page 2-12.
9-45
9
Optimizing the Model for Code Generation
Compiler Options
• If you do not require double precision for your application, define real_T as
float in your template make file, or you can simply specify -DREAL_T=float
after make_rtw in the Make command field.
• Turn on the optimizations for the compiler (e.g., -O2 for gcc, -Ot for Microsoft
Visual C).
9-46
10
The S-Function Target
S-functions are an important class of target for which Real-Time Workshop can generate code. The
ability to encapsulate a subsystem into an S-function allows you to increase its execution efficiency
and shield its internal logic from inspection and modification. Here we describe the properties of
S-function targets and demonstrate how to generate them. For further details on the structure of
S-functions, see Writing S-Functions in the Simulink documentation.
Introduction (p. 10-2)
Overview of the S-function target and its applications
Creating an S-Function Block from a
Subsystem (p. 10-3)
How to extract a subsystem from a model and use it to
generate a reusable S-function component; a step-by-step
demonstration
Tunable Parameters in Generated
S-Functions (p. 10-9)
How to declare tunable parameters in generated
S-functions and how they differ from those in other
targets
Automated S-Function Generation
(p. 10-11)
Step-by-step instructions for automatically generating an
S-function from a subsystem
Restrictions (p. 10-15)
Limitations constraining the use of the S-function target
Unsupported Blocks (p. 10-17)
Blocks not supported by the S-function target
System Target File and Template
Makefiles (p. 10-18)
Control files used by the S-function target
10
The S-Function Target
Introduction
Using the S-function target, you can build an S-function component and use it
as an S-Function block in another model. The S-function code format used by
the S-function target generates code that conforms to the Simulink C MEX
S-function application programming interface (API). Applications of this
format include:
• Conversion of a model to a component. You can generate an S-Function block
for a model, m1. Then, you can place the generated S-Function block in
another model, m2. Regenerating code for m2 does not require regenerating
code for m1.
• Conversion of a subsystem to a component. By extracting a subsystem to a
separate model, and generating an S-Function block from that model, you
can create a reusable component from the subsystem. See “Creating an
S-Function Block from a Subsystem” on page 10-3 for an example of this
procedure.
• Speeding up simulation. In many cases, an S-function generated from a
model performs more efficiently than the original model.
• Code reuse. You can incorporate multiple instances of one model inside
another without replicating the code for each instance. Each instance will
continue to maintain its own unique data.
The S-function target generates noninlined S-functions. You can generate an
executable from a model that contains generated S-functions by using the
generic real-time or real-time malloc targets. You cannot use the Real-Time
Workshop Embedded Coder target for this purpose, since it requires inlined
S-functions.
You can place a generated S-Function block into another model from which you
can generate another S-function format. This allows any level of nested
S-functions.
Intellectual Property Protection
In addition to the technical applications of the S-function target listed above,
you can use the S-function target to protect your designs and algorithms. By
generating an S-function from a proprietary model or algorithm, you can share
the model’s functionality without providing the source code. You need only
provide the binary .dll or MEX-file object to users.
10-2
Creating an S-Function Block from a Subsystem
Creating an S-Function Block from a Subsystem
This section demonstrates how to extract a subsystem from a model and
generate a reusable S-function component from it.
Figure 10-1 illustrates SourceModel, a simple model that inputs signals to a
subsystem. Figure 10-2 illustrates the subsystem, SourceSubsys. The signals,
which have different widths and sample times, are:
• A Step block with sample time 1
• A Sine Wave block with sample time 0.5
• A Constant block whose value is the vector [-2 3]
Figure 10-1: SourceModel
Figure 10-2: SourceSubsys
10-3
10
The S-Function Target
Our objective is to extract SourceSubsys from the model and build an
S-Function block from it, using the S-function target. We want the S-Function
block to perform identically to the subsystem from which it was generated.
Note that in this model, SourceSubsys inherits sample times and signal widths
from its input signals. However, S-function blocks created from a model using
the S-function target will have all signal attributes (such as signal widths or
sample times) hardwired. (The sole exception to this rule concerns samples
times, as described in “Sample Time Propagation in Generated S-Functions” on
page 10-8.)
In this example, we want the S-Function block to retain the properties of
SourceSubsys as it exists in SourceModel. Therefore, before building the
subsystem as a separate S-function component, the inport sample times and
widths must be set explicitly. In addition, the solver parameters of the
S-function component must be the same as those of the original model. This
ensures that the generated S-function component will operate identically to the
original subsystem (see “Choice of Solver Type” on page 10-8 for an exception
to this rule).
To build SourceSubsys as an S-function component:
1 Create a new model and copy/paste SourceSubsys into the empty window.
2 Set the signal widths and sample times of inports inside SourceSubsys such
that they match those of the signals in the original model. Inport 1, Filter,
has a width of 1 and a sample time of 1. Inport 2, Xferfcn, has a width of 1
and a sample time of 0.5. Inport 3, offsets, has a width of 2 and a sample
time of 0.5.
3 The generated S-Function block should have three inports and one outport.
Connect inports and an outport to SourceSubsys, as shown below.
10-4
Creating an S-Function Block from a Subsystem
Note that the correct signal widths and sample times propagate to these
ports.
4 Set the solver type, mode, and other solver parameters such that they are
identical to those of the source model.
5 Save the new model.
6 Open the Simulation Parameters dialog and click the Real-Time
Workshop tab. On the Real-Time-Workshop pane, select Target
configuration from the Category menu.
7 Click the Browse button to open the System Target Browser. Select the
S-function target in the System Target Browser, and click OK. The
Real-Time-Workshop pane parameters should appear as below.
10-5
10
The S-Function Target
8 Select RTW S-function code generation options from the Category menu.
Make sure that Create New Model is selected.
When this option is selected, the build process creates a new model after it
builds the S-function component. The new model contains an S-Function
block, linked to the S-function component.
9 Click Apply if necessary.
10 Click Build.
11 Real-Time Workshop builds the S-function component in the working
directory. After the build, a new model window displays.
10-6
Creating an S-Function Block from a Subsystem
12 You can now copy the Real-Time Workshop S-Function block from the new
model and use it in other models or in a library. Figure 10-3 shows the
S-Function block plugged in to the original model. Given identical input
signals, the S-Function block will perform identically to the original
subsystem.
Figure 10-3: Generated S-Function Plugged into SourceModel
Note that the speed at which the S-Function block executes is typically faster
than the original model. This difference in speed is more pronounced for larger
10-7
10
The S-Function Target
and more complicated models. By using generated S-functions, you can
increase the efficiency of your modeling process.
Sample Time Propagation in Generated S-Functions
Note that sample time propagation for the S-function code format is slightly
different from the other code formats. A generated S-Function block will
inherit its sample time from the model in which it is placed if (and only if) no
blocks in the original model specify their sample times.
Choice of Solver Type
If the model containing the subsystem from which you generate an S-function
uses a variable step solver, the generated S-function will contain zero crossing
functions. Therefore, the generated S-function will work properly in models
with either variable step or fixed step solvers.
On the other hand, if the model containing the subsystem from which you
generate an S-function uses a fixed step solver, the generated S-function
contains no zero crossing functions. In this case, you can use the generated
S-function only within models that use fixed-step solvers.
10-8
Tunable Parameters in Generated S-Functions
Tunable Parameters in Generated S-Functions
You can utilize tunable parameters in generated S-functions in two ways:
• Use the Generate S-function feature (see “Automated S-Function
Generation” on page 10-11).
or
• Use the Model Parameter Configuration dialog (see “Parameters: Storage,
Interfacing, and Tuning” on page 5-2) to declare desired block parameters
tunable.
Block parameters that are declared tunable with the auto storage class in
the source model become tunable parameters of the generated S-function.
Note that these parameters do not become part of a generated rtP parameter
data structure, as they would in code generated from other targets. Instead,
the generated code accesses these parameters via MEX API calls such as
mxGetPr or mxGetData. Your code should access these parameters in the
same way.
For further information on MEX API calls, see Writing S-Functions and
“External Interfaces/API” in the MATLAB online documentation.
S-Function blocks created via the S-function target are automatically masked.
The mask displays each tunable parameter in an edit field. By default, the edit
field displays the parameter by variable name, as in the following example.
You can choose to display the value of the parameter rather than its variable
name. To do this, select Use Value for Tunable Parameters in the Options
section.
10-9
10
The S-Function Target
When this option is chosen, the value of the variable (at code generation time)
is displayed in the edit field, as in the following example.
10-10
Automated S-Function Generation
Automated S-Function Generation
The Generate S-function feature automates the process of generating an
S-function from a subsystem. In addition, the Generate S-function feature
presents a display of parameters used within the subsystem, and lets you
declare selected parameters tunable.
As an example, consider SourceSubsys, the subsystem illustrated in
Figure 10-2. Our objective is to automatically extract SourceSubsys from the
model and build an S-Function block from it, as in the previous example. In
addition, we want to set the gain factor of the Gain block within SourceSubsys
to the workspace variable K (as illustrated below) and declare K as a tunable
parameter.
To auto-generate an S-function from SourceSubsys with tunable parameter K:
1 Click on the subsystem to select it.
2 Select Generate S-function from the Real-Time Workshop submenu of the
Tools menu. This menu item is enabled when a subsystem is selected in the
current model.
Alternatively, you can choose Generate S-function from the Real-Time
Workshop submenu of the subsystem block's context menu.
3 The Generate S-function window is diplayed (see Figure 10-4). This
window shows all variables (or data objects) that are referenced as block
parameters in the subsystem, and lets you declare them as tunable.
The upper pane of the window displays three columns:
10-11
10
The S-Function Target
- Variable name: name of the parameter.
- Class: If the parameter is a workspace variable, its data type is shown. I
the parameter is a data object, its and class is shown
- Tunable: Lets you select tunable parameters. To declare a parameter
tunable, select the check box. In Figure 10-4, the parameter K is declared
tunable.
When you select a parameter in the upper pane, the lower pane shows all
the blocks that reference the parameter, and the parent system of each such
block.
Figure 10-4: The Generate S-Function Window
4 If you have licensed and installed the Real-Time Workshop Embedded
Coder, the Use Embedded Coder check box is available, as in Figure 10-4.
Otherwise, it is grayed out. When Use Embedded Coder is selected, the
build process generates a wrapper S-Function via the Real-Time Workshop
Embedded Coder. See the Real-Time Workshop Embedded Coder
documentation for further information.
10-12
Automated S-Function Generation
5 After selecting tunable parameters, click the Build button. This initiates
code generation and compilation of the S-function, using the S-function
target. The Create New Model option is automatically enabled.
6 The build process displays status messages in the MATLAB command
window. When the build completes, the tunable parameters window closes,
and a new untitled model window opens.
7 The model window contains an S-Function block, subsys_blk, where subsys
is the name of the subsystem from which the block was generated.
The generated S-function component, subsys, is stored in the working
directory. The generated source code for the S-function is written to a build
directory, subsys_sfcn_rtw. Additionally a stub file, subsys_sf.c, is
written to the working directory. This file simply contains an include
directive that you can use to interface other C code to the generated code.
Note that if the Use Embedded Coder option was selected, the build
directory is named subsys_ert_rtw.
8 Note that the untitled generated model does not persist, unless you save it
via the File menu.
10-13
10
The S-Function Target
9 Note that the generated S-Function block has inports and outports whose
widths and sample times correspond to those of the original model.
The following code fragment, from the mdlOutputs routine of the generated
S-function code (in SourceSubsys_sf.c), illustrates how the tunable variable K
is referenced via calls to the MEX API.
static void mdlOutputs(SimStruct *S, int_T tid)
...
/* Expression for <Root>/Out1 incorporates: */
/*
Gain Block: <S1>/Gain */
/*
Sum Block: <S1>/Sum */
/*
Inport Block: <Root>/offsets */
/* Outport Block: <Root>/Out1 */
((real_T *)ssGetOutputPortSignal(S,0))[0] = ((*(real_T *)(mxGetData(K(S)))) *
(rtb_Product + *(((real_T**)ssGetInputPortSignalPtrs(S, 2))[0])));
((real_T *)ssGetOutputPortSignal(S,0))[1] = ((*(real_T *)(mxGetData(K(S)))) *
(rtb_Product + *(((real_T**)ssGetInputPortSignalPtrs(S, 2))[1])));
Note In automatic S-function generation, the Use Value for Tunable
Parameters option is always set to its default value (off).
10-14
Restrictions
Restrictions
Limitations on Use of Goto and From Blocks
When using the S-function target, Real-Time Workshop restricts I/O to
correspond to the root model's Inport and Outport blocks (or the Inport and
Outport blocks of the Subsystem block from which the S-function target was
generated). No code is generated for Goto or From blocks.
To work around this restriction, you should create your model and subsystem
with the required Inport and Outport blocks, instead of using Goto and From
blocks to pass data between the root model and subsystem. In the model that
incorporates the generated S-function, you would then add needed Goto and
From blocks.
As an example of this restriction, consider the model shown in Figure 10-5 and
its subsystem, Subsystem1, shown in Figure 10-6. The Goto block in
Subsystem1, which has global visibility, passes its input to the From block in
the root model.
Figure 10-5: Root Model With From Block
Figure 10-6: Subsystem1 With Goto Block
If SubSystem1 is built as an S-Function using the S-Function target, and
plugged into the original model (as shown in Figure 10-7), a warning is issued
when the model is run, because the generated S-function does not implement
the Goto block.
10-15
10
The S-Function Target
Figure 10-7: Generated S-Function Replaces Subsystem1
A workaround is shown in Figure 10-8. A conventional Outport is used in
Subsystem1.When the generated S-function is plugged into the root model, its
output is connected to the To Workspace block.
Figure 10-8: Use of Outport in Generated S-Function
Other Restrictions
• Hand-written S-functions without corresponding TLC files must contain
exception-free code. For more information on exception-free code, refer to
“Exception-Free Code” in Writing S-Functions.
• If you modify the source model that generated an S-Function block,
Real-Time Workshop does not automatically rebuild models containing the
generated S-Function block.
10-16
Unsupported Blocks
Unsupported Blocks
The S-function format does not support the following built-in blocks:
• MATLAB Fcn Block
• S-Function blocks containing any of the following:
- M-file S-functions
- Fortran S-functions
- C MEX S-functions that call into MATLAB
• Scope block
• To Workspace block
10-17
10
The S-Function Target
System Target File and Template Makefiles
The following system target file and template makefiles are provided for use
with the S-function target.
System Target File
• rtwsfcn.tlc
Template Makefiles
• rtwsfcn_bc.tmf — Borland C
• rtwsfcn_lcc.tmf — LCC compiler
• rtwsfc_unix.tmf — UNIX host
• rtwsfcn_vc.tmf — Visual C
• rtwsfcn_watc.tmf — Watcom C
10-18
11
Real-Time Workshop
Rapid Simulation Target
The rapid simulation (rsim) target provides a fast and flexible platform on your own host computer
for testing code generated for models, tuning parameters, and varying inputs to compile statistics
describing the behavior of your model across a range of initial conditions. In this chapter we discuss
the following topics:
Introduction (p. 11-2)
Overview of the Rapid Simulation (rsim) target, its
applications, and dependencies on Simulink
Building for the Rapid Simulation
Target (p. 11-5)
Generating and building an rsim executable
11
Real-Time Workshop Rapid Simulation Target
Introduction
The Real-Time Workshop rapid simulation target (rsim) consists of a set of
target files for nonreal-time execution on your host computer. You can use rsim
to generate fast, stand-alone simulations that allow batch parameter tuning
and loading of new simulation data (signals) from a standard MATLAB
MAT-file without needing to recompile your model.
The C code generated from Real-Time Workshop is highly optimized to provide
fast execution of Simulink models of hybrid, dynamic systems. This includes
models using variable step solvers and zero crossing detection.
The speed of the generated code makes the rsim target ideal for batch or Monte
Carlo simulation. The generated executable (model.exe) created using the
rsim target has the necessary run-time interface to read and write data to
standard MATLAB MAT-files. Using this interface model.exe can reads new
signals and parameters from input MAT-files at the start of the simulation and
write the simulation results to output MAT-files.
Having built an rsim executable with Real-Time Workshop and an appropriate
C compiler for your host computer, you can perform any combination of the
following by using command line options. Without recompiling, the rapid
simulation target allows you to:
• Specify a new file(s) that provides input signals for From File blocks
• Specify a new file that provides input signals with any Simulink data type
(double, float, int32, uint32, int16, uint16, int8, uint8, and complex data
types) by using the From Workspace block
• Replace the entire block diagram parameter vector and run a simulation
• Specify a new stop time for ending the stand-alone simulation
• Specify a new name of the MAT-file used to save model output data
• Specify name(s) of the MAT-files used to save data connected to To File
blocks
You can run these options:
• Directly from your operating system command line (for example, DOS box or
UNIX shell) or
• By using the bang (!) command with a command string at the MATLAB
prompt
11-2
Introduction
Therefore, you can easily write simple scripts that will run a set of simulations
in sequence while using new data sets. These scripts can be written to provide
unique filenames for both input parameters and input signals, as well as
output filenames for the entire model or for To File blocks.
The rsim target can be configured to either access all solvers available with
Simulink (which is the default configuration) or use only the fixed step solvers
packaged with Real-Time Workshop.
In the default configuration, the standalone executable (model.exe) created by
the rsim target links with the Simulink solver module (a shared library) if the
model uses a variable-step solver. When model.exe uses the Simulink solver
module, running model.exe will check out a Simulink license (see details
below). In such cases, model.exe requires read access to installed location of
MATLAB and Simulink in order to locate the license.dat file and the shared
libraries.
Licensing Protocols for Simulink Solvers in
Executables
The Rapid Simulation target supports variable step solvers by linking the
generated code with the Simulink solver module (a shared library). When this
rsim executable is run, it accesses proprietary Simulink variable step solver
technology. In order to do so, the executable needs to check out a Simulink
license for the duration of its execution.
Rapid Simulation executables that do not use Simulink solver module (for
example, rsim executable built for a fixed-step model using the Real-Time
Workshop fixed-step solvers) do not require any license when they run.
Note The default setting of auto for the Solver selection option in the rsim
code generation options page configures rsim to use the Simulink solver
module only when needed (i.e., when the model uses a variable step solver).
The rsim executable will look in the default locations for the license file
• Unix: matlabroot/etc/license.dat
• PC: matlabroot/bin/win32/license.dat,
11-3
11
Real-Time Workshop Rapid Simulation Target
where matlabroot is the one use when building the rsim executable. If the rsim
executable is unable to locate the license file (this may happen, for example,
if you run this executable on another machine, where matlabroot is no longer
valid), it will print the following error message and exit:
Error checking out SIMULINK license.
Cannot find license file
The license files (or server network addresses) attempted are
listed below. Use LM_LICENSE_FILE to use a different license
file, or contact your software provider for a license file.
Feature:
SIMULINK
Filename:
/apps/matlab/etc/license.dat
License path: /abbs/matlab/etc/license.dat
FLEXlm error: -1,359. System Error: 2 "No such file or directory"
For further information, refer to the FLEXlm End User Manual,
available at "www.globetrotter.com".
Error: Unable to checkout Simulink license
Error terminating RSIM Engine: License check failed
Note You can point the rsim executable to a different license file by setting
the environment variable LM_LICENSE_FILE. The location pointed to by that
variable will override the default location compiled into the rsim executable.
If the rsim executable is unable to check out a Simulink license (this would
happen, for example, if all Simulink licenses are currently checked out), or has
other errors when checking out a Simulink license it will display a detailed
error message (similar to the one above) returned by the FLEXlm API and exit.
11-4
Building for the Rapid Simulation Target
Building for the Rapid Simulation Target
To generate and build an rsim executable, press the Browse button on the
Real-Time Workshop pane of the Simulation Parameters dialog box, and
select the rapid simulation target from the System Target File Browser. This
picture shows the dialog box settings for the rapid simulation target.
Press the Browse button and select the rapid
simulation target from the System Target File
Browser. This automatically selects the correct
settings for the system target file, the template
makefile, and the make command.
Figure 11-1: Specifying Target and Make Files for rsim
After specifying system target and make files as noted above, select any desired
Workspace I/O settings, and press Build. Real-Time Workshop will
automatically generate C code and build the executable for your host machine
using your host machine C compiler. See “Choosing and Configuring Your
Compiler” on page 2-51 and “Template Makefiles and Make Options” on
page 2-54 for additional information on compilers that are compatible with
Simulink and Real-Time Workshop. The picture below shows rsim-specific code
generation options that allow you to avoid using the Simulink solver module
11-5
11
Real-Time Workshop Rapid Simulation Target
(i.e., use only the fixed step solvers packaged with Real-Time Workshop) and
enable the rsim executable to communicate with Simulink via external mode.
Choose whether to generate code for a fixed-step
or a variable-step solverwith this popup menu.
The auto option invokes the Simulimk solver
module only when the model requires it.
Select this checkbox to create an rsim executable
that communicates with Simulink via external mode
Note Rapid Simulation executables created without using the Simulink
solver module can be transferred and run on computers that do not have
MATLAB installed. When running an rsim executable on such a machine, it is
necessary to have the following dlls in your working directory: libmx.dll,
libut.dll, and libmat.dll. These dlls are required for the rsim executable to
write and read data from a .mat file. This deployment option is not available
for rsim executables that rely upon the Simulink solver module.
Running a Rapid Simulation
The rapid simulation target lets you run a simulation similar to the generic
real-time (GRT) target provided by Real-Time Workshop. This simulation does
not use timer interrupts, and therefore is a nonreal-time simulation
environment. The difference between GRT and rsim simulations is that
• rsim supports variable step solvers, and
• rsim allows you to change parameter values or input signals at the start of a
simulation without the need to generate code or recompile.
11-6
Building for the Rapid Simulation Target
The GRT target, on the other hand, is a starting point for targeting a new
processor.
A single build of your model can be used to study effects from varying
parameters or input signals. Command line arguments provide the necessary
mechanism to specify new data for your simulation. This table lists all
available command line options.
Table 11-1: rsim Command Line Options
Command Line Option
Description
model -f old.mat=new.mat
Read From File block input signal data from
a replacement MAT-file.
model -o newlogfile.mat
Write MAT-file logging data to a file named
newlogfile.mat.
model -p filename.mat
Read a new (replacement) parameter vector
from a file named filename.mat.
model -tf <stoptime>
Run the simulation until the time value
<stoptime> is reached.
model -t old.mat=new.mat
The original model specified saving signals
to the output file old.mat. For this run use
the file new.mat for saving signal data.
model -v
Run in verbose mode.
model -h
Display a help message listing options.
Note On Solaris platforms, to run the rsim executable created for a model
that uses variable step solvers in a seperate shell, the LD_LIBRARY_PATH
environment variable is needed to indicate the path to the MATLAB
installation directory, as follows:
% setenv LD_LIBRARY_PATH /apps/matlab/bin/sol2:$LD_LIBRARY_PATH
11-7
11
Real-Time Workshop Rapid Simulation Target
Obtaining the Parameter Structure from Your Model
To obtain a parameter structure for the current model settings you may use the
rsimgetrtp function, with the following syntax:
rtP = rsimgetrtp(‘model’, options)
The rtP structure is designed to be used with the Rapid Simulation target.
Getting it via rsimgetrtp forces an update diagram action. In addition to the
current model tunable block parameter settings, the rtP structure contains a
structural checksum. This checksum is used to ensure that the model structure
hasn’t changed since the rsim executable was generated.
Options to rsimgetrtp are passed as parameter-value pairs. Currently there is
one option, AddTunableParamInfo, which has two states, on and off:
rtP = rsimgetrtp(‘model’,‘AddTunableParamInfo’,’on’)
rtP = rsimgetrtp(‘model’,‘AddTunableParamInfo’,’on’)
The AddTunableParamInfo option causes Real-Time Workshop to generate
code that extract tunable parameter information from your model and places it
in the return argument (rtP). This information gives you a mapping between
the parameter structure and the tunable parameters.
To use the AddTunableParamInfo option, you must have selected the Inline
Parameters checkbox in the Advanced pane of the Simulation Parameters
dialog box. Exercising this option also creates, then deletes a model.rtw file in
your current working directory.
Tunable Fixed-Point parameters are reported according to their stored value.
For example, an sfix(16) parameter value of 1.4 with a scaling of 2^-8 will
have a value of 358 as an int16.
Example 1. Create an rsim executable and pass a different parameter structure:
1 Set the Real-Time Workshop target configuration to Rapid Simulation
Target using the Target File Browser
2 Create an rsim executable for the model by clicking the Build button or by
typing rtwbuild('model').
3 Modify parameters in your model and save the rtP structure:
rtP = rsimgetrtp('model')
save myrtp.mat rtP
11-8
Building for the Rapid Simulation Target
4 Run the generate executable with the new parameter set:
!model -p myrtp.mat
5 Load the results in to Matlab
load model.mat
Example 2. Create an rtP with the tunable parameter mapping information:
1 Create rtP with the tunable parameter information:
rtP = rsimgetrtp('model','AddTunableParamInfo','on')
2 The rtP structure contains:
modelChecksum: 1x4 vector that encodes the structure of the model
parameters: A structure of the tunable parameters in the model
3 The parameters structure contains the following member fields:
dataTypeName: The data type name, e.g., 'double'
dataTypeId: Internal data type identifier for use by Real-Time Workshop
complex: 0 if real, 1 if complex
Specifying a New Signal Data File for a From File Block
To understand how to specify new signal data for a From File block, create a
working directory and connect to that directory. Open the model rsimtfdemo
by typing
rsimtfdemo
at the MATLAB prompt. Type
w = 100;
zeta = 0.5;
to set parameters. rsimtfdemo requires a data file, rsim_tfdata.mat. Make a
local copy of matlabroot/toolbox/rtw/rtwdemos/rsim_tfdata.mat in your
working directory.
Be sure to specify rsim.tlc as the system target file and rsim_default_tmf as
the template makefile. Then press the Build button on the Real-Time
Workshop pane to create the rsim executable.
11-9
11
Real-Time Workshop Rapid Simulation Target
!rsimtfdemo
load rsimtfdemo
plot(rt_yout)
The resulting plot shows simulation results using the default input data.
Replacing Input Signal Data. New data for a From File block can be placed in a
standard MATLAB MAT-file. As in Simulink, the From File block data must
be stored in a matrix with the first row containing the time vector while
subsequent rows contain u vectors as input signals. After generating and
compiling your code, you can type the model name rsimtfdemo at a DOS
prompt to run the simulation. In this case, the file rsim_tfdata.mat provides
the input data for your simulation.
For the next simulation, create a new data file called newfrom.mat and use this
to replace the original file (rsim_tfdat.mat) and run an rsim simulation with
this new data. This is done by typing
t=[0:.001:1];
u=sin(100*t.*t);
tu=[t;u];
11-10
Building for the Rapid Simulation Target
save newfrom.mat tu;
!rsimtfdemo -f rsim_tfdata.mat=newfrom.mat
at the MATLAB prompt. Now you can load the data and plot the new results
by typing
load rsimtfdemo
plot(rt_yout)
This picture shows the resulting plot.
As a result the new data file is read and the simulation progresses to the stop
time specified in the Solver page of the Simulation Parameters dialog box.
It is possible to have multiple instances of From File blocks in your Simulink
model.
Since rsim does not place signal data into generated code, it reduces code size
and compile time for systems with large numbers of data points that originate
in From File blocks. The From File block requires the time vector and signals
to be data of type double. If you need to import signal data of a data type other
11-11
11
Real-Time Workshop Rapid Simulation Target
than double, use a From Workspace block with the data specified as a
structure.
The workspace data must be in the format
variable.time
variable.signals.values
If you have more than one signal, the format must be
variable.time
variable.signals(1).values
variable.signals(2).values
Specifying a New Output Filename for the Simulation
If you have specified Save to Workspace options (that is, checked Time,
States, Outputs, or Final States check boxes on the Workspace I/O page of the
Simulation Parameters dialog box), the default is to save simulation logging
results to the file model.mat. You can now specify a replacement filename for
subsequent simulations. In the case of the model rsimtfdemo, by typing
!rsimtfdemo
at the MATLAB prompt, a simulation runs and data is normally saved to
rsimtfdemo.mat.
!rsimtfdemo
created rsimtfdemo.mat
You can specify a new output filename for data logging by typing
!rsimtfdemo -o sim1.mat
In this case, the set of parameters provided at the time of code generation,
including any From File block data, is run. You can combine a variety of rsim
flags to provide new data, parameters, and output files to your simulation.
Note that the MAT-file containing data for the From File blocks is required.
This differs from the grt operation, which inserts MAT-file data directly into
the generated C code that is then compiled and linked as an executable. In
contrast, rsim allows you to provide new or replacement data sets for each
successive simulation. A MAT-file containing From File or From Workspace
data must be present, if any From File or From Workspace blocks exist in your
model.
11-12
Building for the Rapid Simulation Target
Changing Block Parameters for an rsim Simulation
Once you have altered one or more parameter in the Simulink block diagram,
you can extract the parameter vector, rtP, for the entire model. The rtP vector,
along with a model checksum, can then be saved to a MATLAB MAT-file. This
MAT-file can be read in directly by the stand-alone rsim executable, allowing
you to replace the entire parameter vector quickly, for running studies of
variations of parameter values where you are adjusting model parameters or
coefficients or importing new data for use as input signals.
The model checksum provides a safety check to ensure that any parameter
changes are only applied to rsim models that have the same model structure.
If any block is deleted, or a new block added, then when generating a new rtP
vector, the new checksum will no longer match the original checksum. The rsim
executable will detect this incompatibility in parameter vectors and exit to
avoid returning incorrect simulation results. In this case, where model
structure has changed, you must regenerate the code for the model.
The rsim target allows you to alter any model parameter, including parameters
that include side-effects functions. An example of a side-effects function is a
simple Gain block that includes the following parameter entry in a dialog box.
gain value:
2 * a
In general, Real-Time Workshop evaluates side-effects functions prior to
generating code. The generated code for this example retains only one memory
location entry, and the dependence on parameter a is no longer visible in the
generated code. The rsim target overcomes the problem of handling side-effects
functions by replacing the entire parameter structure, rtP. You must create
this new structure by using rsimgetrtp.m. and then save it in a MAT-file. For
the rsimtfdemo example, type
zeta = .2;
myrtp = rsimgetrtp('modelname');
save myparamfile myrtp;
at the MATLAB prompt.
In turn, rsim can read the MAT-file and replace the entire rtP structure
whenever you need to change one or more parameters — without recompiling
the entire model.
For example, assume that you have changed one or more parameters in your
model, generated the new rtP vector, and saved rtP to a new MAT-file called
11-13
11
Real-Time Workshop Rapid Simulation Target
myparamfile.mat. In order to run the same rsimtfdemo model and use these
new parameter values, execute the model by typing
!rsimtfdemo -p myparamfile.mat
load rsimtfdemo
plot(rt_yout)
Note that the p is lower-case and represents “Parameter file.”
Specifying a New Stop Time for an rsim Simulation
If a new stop time is not provided, the simulation will run until reaching the
value specified in the Solver page at the time of code generation. You can
specify a new stop time value as follows.
!rsimtfdemo -tf 6.0
In this case, the simulation will run until it reaches 6.0 seconds. At this point
it will stop and log the data according to the MAT-file data logging rules as
described above.
If your model includes From File blocks that also include a time vector in the
first row of the time and signal matrix, the end of the simulation is still
regulated by the original setting in the Solver page of the Simulation
Parameters dialog box or from the -s option as described above. However, if
the simulation time exceeds the end points of the time and signal matrix (that
is, if the final time is greater than the final time value of the data matrix), then
the signal data will be extrapolated out to the final time value as specified
above.
Specifying New Output Filenames for To File Blocks
In much the same way as you can specify a new system output filename, you
can also provide new output filenames for data saved from one or more To File
blocks. This is done by specifying the original filename at the time of code
generation with a new name as follows.
!mymodel -t original.mat=replacement.mat
In this case, assume that the original model wrote data to the output file called
original.mat. Specifying a new filename forces rsim to write to the file
replacement.mat. This technique allows you to avoid over-writing an existing
simulation run.
11-14
Building for the Rapid Simulation Target
Simulation Performance
It is not possible to predict accurately the simulation speedup of an rsim
simulation compared to a standard Simulink simulation. Performance will
vary. Larger simulations have achieved speed improvements of up to 10 times
faster than standard Simulink simulations. Some models may not show any
noticeable improvement in simulation speed. The only way to determine
speedup is to time your standard Simulink simulation and then compare its
speed with the associated rsim simulation.
Batch and Monte Carlo Simulations
The rsim target is intended to be used for batch simulations in which
parameters and/or input signals are varied for each new simulation. New
output filenames allow you run new simulations without over-writing prior
simulation results. A simple example of such a set of batch simulations can be
run by creating a .bat file for use under Microsoft Windows.
This simple file for Windows is created with any text editor and executed by
typing the filename, for example, mybatch, where the name of the text file is
mybatch.bat.
rsimtfdemo
rsimtfdemo
rsimtfdemo
rsimtfdemo
-f
-f
-f
-f
rsimtfdemo.mat=run1.mat
rsimtfdemo.mat=run2.mat
rsimtfdemo.mat=run3.mat
rsimtfdemo.mat=run4.mat
-o
-o
-o
-o
results1.mat
results2.mat
results3.mat
results4.mat
-s
-s
-s
-s
10.0
10.0
10.0
10.0
In this case, batch simulations are run using the four sets of input data in files
run1.mat, run2.mat, and so on. The rsim executable saves the data to the
corresponding files specified after the -o option.
The variable names containing simulation results in each of these files are
identical. Therefore, loading consecutive sets of data without renaming the
data once it is in the MATLAB workspace will result in over-writing the prior
workspace variable with new data. If you want to avoid over-writing, you can
copy the result to a new MATLAB variable prior to loading the next set of data.
You can also write M-file scripts to create new signals, and new parameter
structures, as well as to save data and perform batch runs using the bang
command (!).
For additional insight into the rapid simulation target, explore rsimdemo1 and
rsimdemo2, located in matlabroot/toolbox/rtw/rtwdemos/rsimdemos. These
11-15
11
Real-Time Workshop Rapid Simulation Target
examples demonstrate how rsim can be called repeatedly within an M-file for
Monte Carlo simulations.
Limitations
The rapid simulation target is subject to the following limitations:
• The rsim target does not support algebraic loops
• The rsim target does not support MATLAB function blocks.
• The rsim target does not support non-inlined M-file, FORTRAN and Ada
S-functions.
• In certain cases, changing block parameters may result in structural
changes to your model that change the model checksum. An example of such
a change would be changing the number of delays in a DSP simulation. In
such cases, you must regenerate the code for the model.
• Variable-step solver support for rsim is not available on HP700, on IBM_RS
platforms, or on PCWIN platforms using the following compiler versions:
- Watcom C/C++ compiler version 10.6
- Borland C/C++ compiler version 5.3.
11-16
12
Targeting Tornado for
Real-Time Applications
Tornado, a target supported by Real-Time Workshop, describes an integrated set of tools for creating
real-time applications to run under theVxWorks operating system, which has many Unix-like
features and runs on a variety of host systems and target processors. This chapter contains the
following topics:
The Tornado Environment (p. 12-2)
Overview of the Tornado (VxWorks) Real-Time Target
and the VxWorks Support library
Run-Time Architecture Overview
(p. 12-5)
Singletasking and multitasking architecture and
host/target communications
Implementation Overview (p. 12-11)
Design, implementation, and execution of a VxWorksbased real-time program using Real-Time Workshop
12
Targeting Tornado for Real-Time Applications
The Tornado Environment
This chapter describes how to create real-time programs for execution under
VxWorks, which is part of the Tornado environment.
The VxWorks real-time operating system is available from Wind River
Systems, Inc. It provides many UNIX-like features and comes bundled with a
complete set of development tools.
Note Tornado is an integrated environment consisting of VxWorks (a
high-performance real-time operating system), application building tools
(compiler, linker, make, and archiver utilities), and interactive development
tools (editor, debugger, configuration tool, command shell, and browser).
This chapter discusses the run-time architecture of VxWorks-based real-time
programs generated by Real-Time Workshop and provides specific information
on program implementation. Topics covered include:
• Configuring device driver blocks and makefile templates
• Building the program
• Downloading the object file to the VxWorks target
• Executing the program on the VxWorks target
• Using the StethoScope data acquisition and graphical monitoring tool, which
is available as an option with VxWorks. It allows you to access the output of
any block in the model (in the real-time program) and display the data on the
host.
• Using Simulink external mode to change model parameters and download
them to the executing program on the VxWorks target. Note that you cannot
use both external mode and StethoScope at the same time.
Confirming Your Tornado Setup Is Operational
Before beginning, you must install and configure Tornado on your host and
target hardware, as discussed in the Tornado documentation. You should then
run one of the VxWorks demonstration programs to ensure you can boot your
VxWorks target and download object files to it. See the Tornado User’s Guide
12-2
The Tornado Environment
for additional information about installation and operation of VxWorks and
Tornado products.
VxWorks Library
Selecting VxWorks Support under the Real-Time Workshop library in the
Simulink Library Browser opens the VxWorks Support library.
The blocks discussed in this chapter are located in the Asynchronous Support
library, a sublibrary of the VxWorks Support library:
• Interrupt Control
• Rate Transition
12-3
12
Targeting Tornado for Real-Time Applications
• Read Side
• Task Synchronization
• Write Side
A second sublibrary, the I/O Devices library, contains support for these drivers:
• Matrix MS-AD12
• Matrix MS-DA12
• VME Microsystems VMIVME-3115-110
• Xycom XVME-500/590
• Xycom XVME-505/595
Each of these blocks has online help available through the Help button on the
block’s dialog box. Refer to the Tornado User’s Guide for detailed information
on these blocks.
12-4
Run-Time Architecture Overview
Run-Time Architecture Overview
In a typical VxWorks-based real-time system, the hardware consists of a UNIX
or PC host running Simulink and Real-Time Workshop, connected to a
VxWorks target CPU via Ethernet. In addition, the target chassis may contain
I/O boards with A/D and D/A converters to communicate with external
hardware. The following diagram shows the arrangement.
Host
VxWorks Target
Simulink
Real-Time Workshop
Target
CPU
Tornado Compiler
ADC/DAC
Boards
Ethernet
Port
model.lo
Ethernet
Figure 12-1: Typical Hardware Setup for a VxWorks Application
The real-time code is compiled on the UNIX or PC host using the cross compiler
supplied with the VxWorks package. The object file (model.lo) output from the
Real-Time Workshop program builder is downloaded, using WindSh (the
command shell) in Tornado, to the VxWorks target CPU via an Ethernet
connection.
The real-time program executes on the VxWorks target and interfaces with
external hardware via the I/O devices installed on the target.
Parameter Tuning and Monitoring
You can change program parameters from the host and monitor data with
Scope blocks while the program executes using Simulink external mode. You
can also monitor program outputs using the StethoScope data analysis tool.
Using Simulink external mode or StethoScope allows you to change model
parameters in your program, and to analyze the results of these changes, in
real time.
12-5
12
Targeting Tornado for Real-Time Applications
External Mode
Simulink external mode provides a mechanism to download new parameter
values to the executing program and to monitor signals in your model. In this
mode, the external link MEX-file sends a vector of new parameter values to the
real-time program via the network connection. These new parameter values
are sent to the program whenever you make a parameter change without
requiring a new code generation or build iteration.
You can use the BlockIOSignals code generation option to monitor signals in
VxWorks. See “Interfacing Parameters and Signals” on page 14-70 for further
information and example code.
The real-time program (executing on the VxWorks target) runs a low priority
task that communicates with the external link MEX-file and accepts the new
parameters as they are passed into the program.
Communication between Simulink and the real-time program is accomplished
using the sockets network API. This implementation requires an Ethernet
network that supports TCP/IP. See Chapter 6, “External Mode” for more
information on external mode.
Changes to the block diagram structure (for example, adding or removing
blocks) require generation of model and execution of the build process.
Configuring VxWorks to Use Sockets
If you want to use Simulink external mode with your VxWorks program, you
must configure your VxWorks kernel to support sockets by including the
INCLUDE_NET_INIT, INCLUDE_NET_SHOW, and INCLUDE_NETWORK options in your
VxWorks image. For more information on configuring your kernel, see the
VxWorks Programmer’s Guide.
Before using external mode, you must ensure that VxWorks can properly
respond to your host over the network. You can test this by using the host
command
ping <target_name>
12-6
Run-Time Architecture Overview
Note You may need to enter a routing table entry into VxWorks if your host
is not on the same local network (subnet) as the VxWorks system. See
routeAdd in the VxWorks Reference Guide for more information.
Configuring Simulink to Use Sockets
Simulink external mode uses a MEX-file to communicate with the VxWorks
system. The MEX-file is
matlabroot/toolbox/rtw/rtw/ext_comm.*
where * is a host-dependent MEX-file extension. See Chapter 6, “External
Mode” for more information.
To use external mode with VxWorks, specify ext_comm as the MEX-file for
external interface in the External Target Interface dialog box (accessed
from the External Mode Control Panel). In the MEX-file arguments field
you must specify the name of the VxWorks target system and, optionally, the
verbosity and TCP port number. Verbosity can be 0 (the default) or 1 if extra
information is desired. The TCP port number ranges from 256 to 65535 (the
default is 17725). If there is a conflict with other software using TCP port
17725, you can change the port that you use by editing the third argument of
the MEX-file for external interface on the External Target Interface dialog
box. The format for the MEX-file arguments field is
'target_network_name' [verbosity] [TCP port number]
For example, this picture shows the External Target Interface dialog box
configured for a target system called halebopp with default verbosity and the
port assigned to 18000.
12-7
12
Targeting Tornado for Real-Time Applications
StethoScope
With StethoScope, you can access the output of any block in the model (in the
real-time program) and display this data on a host. Signals are installed in
StethoScope by the real-time program using the BlockIOSignals data
structure (See “Interfacing Parameters and Signals” on page 14-70 for
information on BlockIOSignals), or interactively from the WindSh while the
real-time program is running. To use StethoScope interactively, see the
StethoScope User’s Manual.
To use StethoScope you must specify certain options with the build command.
See “Code Generation Options” on page 12-16 for information on these options.
Run-Time Structure
The real-time program executes on the VxWorks target while Simulink and
StethoScope execute on the same or different host workstations. Simulink and
StethoScope require tasks on the VxWorks target to handle communication.
This diagram illustrates the structure of a VxWorks application using
Simulink external mode and StethoScope.
UNIX or PC Host
Simulink in
external mode
VxWorks Target
StethoScope
tRaten
tRate2
tRate1
Process GUI Events
tExtern
ext_comm
Ethernet
Figure 12-2: The Run-Time Structure
12-8
tBaseRate
tScope
Run-Time Architecture Overview
The program creates VxWorks tasks to run on the real-time system: one
communicates with Simulink, the others execute the model. StethoScope
creates its own tasks to collect data.
Host Processes
There are two processes running on the host side that communicate with the
real-time program:
• Simulink running in external mode. Whenever you change a parameter in
the block diagram, Simulink calls the external link MEX-file to download
any new parameter values to the VxWorks target.
• The StethoScope user interface module. This program communicates with
the StethoScope real-time module running on the VxWorks target to retrieve
model data and plot time histories.
VxWorks Tasks
You can run the real-time program in either singletasking or multitasking
mode. The code for both modes is located in
matlabroot/rtw/c/tornado/rt_main.c
Real-Time Workshop compiles and links rt_main.c with the model code during
the build process.
Singletasking. By default, the model is run as one task, tSingleRate. This may
actually provide the best performance (highest base sample rate) depending on
the model.
The tSingleRate task runs at the base rate of the model and executes all
necessary code for the slower sample rates. Execution of the tSingleRate task
is normally blocked by a call to the VxWorks semTake routine. When a clock
interrupt occurs, the interrupt service routine calls the semGive routine, which
causes the semTake call to return. Once enabled, the tSingleRate task
executes the model code for one time step. The loop then waits at the top by
again calling semTake. For more information about the semTake and semGive
routines, refer to the VxWorks Reference Manual. By default, it runs at a
relatively high priority (30), which allows it to execute without interruption
from background system activity.
12-9
12
Targeting Tornado for Real-Time Applications
Multitasking. Optionally, the model can run as multiple tasks, one for each
sample rate in the model:
• tBaseRate — This task executes the components of the model code run at the
base (highest) sample rate. By default, it runs at a relatively high priority
(30), which allows it to execute without interruption from background
system activity.
• tRaten — The program also spawns a separate task for each additional
sample rate in the system. These additional tasks are named tRate1,
tRate2, …, tRaten, where n is slowest sample rate in the system. The
priority of each additional task is one lower than its predecessor (tRate1 has
a lower priority than tBaseRate).
Supporting Tasks. If you select external mode and/or StethoScope during the
build process, these tasks will also be created:
• tExtern — This task implements the server side of a socket stream
connection that accepts data transferred from Simulink to the real-time
program. In this implementation, tExtern waits for a message to arrive from
Simulink. When a message arrives, tExtern retrieves it and modifies the
specified parameters accordingly.
tExtern runs at a lower priority than tRaten, the lowest priority model task.
The source code for tExtern is located in matlabroot/rtw/c/src/ext_svr.c.
• tScopeDaemon and tScopeLink — StethoScope provides its own VxWorks
tasks to enable real-time data collection and display. In singletasking mode,
tSingleRate collects signals; in multitasking mode, tBaseRate collects them.
Both perform the collection on every base time step. The StethoScope tasks
then send the data to the host for display when there is idle time, that is,
when the model is waiting for the next time step to occur. rt_main.c starts
these tasks if they are not already running.
12-10
Implementation Overview
Implementation Overview
To implement and run a VxWorks-based real-time program using Real-Time
Workshop, you must:
• Design a Simulink model for your particular application.
• Add the appropriate device driver blocks to the Simulink model, if desired.
• Configure the tornado.tmf template makefile for your particular setup.
• Establish a connection between the host running Simulink and the VxWorks
target via Ethernet.
• Use the automatic program builder to generate the code and the custom
makefile, invoke the make command to compile and link the generated code,
and load and activate the tasks required.
The figure below shows the Real-Time Workshop Tornado run-time interface
modules and the generated code for the f14 example model.
12-11
12
Targeting Tornado for Real-Time Applications
Main Program
rt_main.c
Generated Code
f14.c
f14_data.c
f14.h
f14_private.h
f14_types.h
rtmodel.h
Model
Execution
rt_sim.c
Template
Makefile
tornado.tmf
Integration
Module
ode5.c
Simulink
Data Structure
Makefile
f14.mk
Executable File
f14.lo
simstruc_types.h
External mode
ext_svr.c
ext_svr.h
ext_msg.h
ext_share.h
Figure 12-3: Source Modules Used to Build the VxWorks Real-Time Program
This diagram illustrates the code modules used to build a VxWorks real-time
program. Dashed boxes indicate optional modules.
12-12
Implementation Overview
Adding Device Driver Blocks
The real-time program communicates with the I/O devices installed in the
VxWorks target chassis via a set of device drivers. These device drivers contain
the necessary code that runs on the target processor for interfacing to specific
I/O devices.
To make device drivers easy to use, they are implemented as Simulink
S-functions using C code MEX-files. This means you can connect them to your
model like any other block and the code generator automatically includes a call
to the block’s C code in the generated code.
You can also inline S-functions via the Target Language Compiler. Inlining
allows you to restrict function calls to only those that are necessary for the
S-function. This can greatly increase the efficiency of the S-function. For more
information about inlining S-functions, see Writing S-Functions and the Target
Language Compiler Reference Guide.
You can have multiple instances of device driver blocks in your model. See
Targeting Real-Time Systems for more information about creating device
drivers.
Configuring the Template Makefile
To configure the VxWorks template, tornado.tmf, you must specify
information about the environment in which you are using VxWorks. This
section lists the lines in the file that you must edit.
VxWorks Configuration
To provide information used by VxWorks, you must specify the type of target
and the specific CPU on the target. The target type is then used to locate the
correct cross compiler and linker for your system.
The CPU type is used to define the CPU macro which is in turn used by many of
the VxWorks header files. Refer to the VxWorks Programmer’s Guide for
information on the appropriate values to use.
This information is in the section labeled
#-------------- VxWorks Configuration --------------
Edit the following lines to reflect your setup.
VX_TARGET_TYPE = 68k
12-13
12
Targeting Tornado for Real-Time Applications
CPU_TYPE = MC68040
Downloading Configuration
In order to perform automatic downloading during the build process, the target
name and host name that the Tornado target server will run on must be
specified. Modify these macros to reflect your setup.
#-------------- Macros for Downloading to Target-------------TARGET = targetname
TGTSVR_HOST = hostname
Tool Locations
In order to locate the Tornado tools used in the build process, the following
three macros must either be defined in the environment or specified in the
template makefile. Modify these macros to reflect your setup.
#-------------- Tool Locations -------------WIND_BASE = c:/Tornado
WIND_REGISTRY = $(COMPUTERNAME)
WIND_HOST_TYPE = x86–win32
Building the Program
Once you have created the Simulink block diagram, added the device drivers,
and configured the makefile template, you are ready to set the build options
and initiate the build process.
Specifying the Real-Time Build Options
Set the real-time build options using the Solver and Real-Time Workshop
pages of the Simulation Parameters dialog box. To access this dialog box,
select Simulation Parameters from the Simulink Simulation menu.
In the Solver pane, for models with continuous blocks, set the Type to
Fixed-step, the Step Size to the desired integration step size, and select the
integration algorithm. For models that are purely discrete, set the integration
algorithm to discrete.
Next, use the System Target File Browser to select the correct Real-Time
Workshop pane settings for Tornado. To access the browser, open the
Real-Time Workshop pane of the Simulation Parameters dialog box and
12-14
Implementation Overview
select Target configuration from the Category menu. Then click the Browse
button. The System Target Browser opens.
Select Tornado (VxWorks) Real-Time Target and click OK. This sets the
Target configuration options correctly:
• System target file — tornado.tlc
12-15
12
Targeting Tornado for Real-Time Applications
• Template makefile — tornado.tmf template, which you must configure
according to the instructions in “Configuring the Template Makefile” on
page 12-13. (You can rename this file; simply change the dialog box
accordingly.)
• Make command — make_rtw
You can optionally inline parameters for the blocks in the C code, which can
improve performance. Inlining parameters is allowed when using external
mode.
Code Generation Options. To specify code generation options specific to Tornado,
open the Real-Time Workshop pane and select Tornado code generation
options from the Category menu.
Real-Time Workshop provides flags that set the appropriate macros in the
template makefile, causing any necessary additional steps to be performed
during the build process.
The flags and switches are as follows:
• MAT-file logging: Select this option to enable data logging during program
execution. The program will create a file named model.mat at the end of
12-16
Implementation Overview
program execution; this file will contain the variables that you specified in
the Workspace I/O pane of the Simulation Parameters dialog box.
Real-Time Workshop adds a prefix or suffix to the names of the Workspace
I/O pane variables that you select for logging. The MAT-file variable name
modifier menu lets you select the prefix or suffix.
By default, the MAT-file is created in the root directory of the current default
device in VxWorks. This is typically the host file system that VxWorks was
booted from. Other remote file systems can be used as a destination for the
MAT-file using rsh or ftp network devices or NFS. See the VxWorks
Programmer’s Guide for more information. If a device or filename other than
the default is desired, add "-DSAVEFILE=filename" to the OPTS flag to the
make command. For example,
make_rtw OPTS="-DSAVEFILE=filename"
• External mode: Select this option to enable the use of external mode in the
generated executable. You can optionally enable a verbose mode of external
mode by appending -DVERBOSE to the OPTS flag in the make command. For
example,
make_rtw OPTS="-DVERBOSE"
causes parameter download information to be printed to the console of the
VxWorks system.
If you enable External mode, you cannot enable the StethoScope option.
• Code format: Selects RealTime or RealTimeMalloc code generation format.
• StethoScope: Select this option to enable the use of StethoScope with the
generated executable. When starting rt_main, there are two command line
arguments that control the block names used by StethoScope; you can use
them when starting the program on VxWorks. See the section, “Running the
Program” on page 12-19 for more information on these arguments.
If you enable StethoScope, you cannot enable the External mode option.
• Download to VxWorks target: Enables automatic downloading of the
generated program.
Additional options are available on the Real-Time Workshop pane. See “Using
the Real-Time Workshop Pane” on page 2-2 for information.
12-17
12
Targeting Tornado for Real-Time Applications
Initiating the Build
To build the program, click on the Build button in the Real-Time Workshop
pane of the Simulation parameters dialog. The resulting object file is named
with the .lo extension (which stands for loadable object). This file has been
compiled for the target processor using the cross compiler specified in the
makefile. If automatic downloading (Download to VxWorks target) is enabled
in the Tornado code generation options, the target server is started and the
object file is downloaded and started on the target. If StethoScope was checked
on the Tornado code generation options, you can now start StethoScope on
the host. The StethoScope object files, libxdr.so, libutilstssip.so, and
libscope.so, will be loaded on the VxWorks target by the automatic download.
See the StethoScope User’s Manual for more information.
Downloading and Running the Executable
Interactively
If automatic downloading is disabled, you must use the Tornado tools to
complete the process. This involves three steps:
1 Establishing a communication link to transfer files between the host and the
VxWorks target
2 Transferring the object file from the host to the VxWorks target
3 Running the program
Connecting to the VxWorks Target
After completing the build process, you are ready to connect the host
workstation to the VxWorks target. The first step is starting the target server
that is used for communication between the Tornado tools on the host and the
target agent on the target. This is done either from the DOS command line or
from within the Tornado development environment. From the DOS command
line use
tgtsvr target_network_name
Downloading the Real-Time Program
To download the real-time program, use the VxWorks ld routine from within
WindSh. WindSh (wind shell) can also be run from the command line or from
within the Tornado development environment. (For example, if you want to
12-18
Implementation Overview
download the file vx_equal.lo, which is in the /home/my_working_dir
directory, use the following commands at the WindSh prompt.
cd "/home/my_working_dir"
ld <vx_equal.lo
You will also need to load the StethoScope libraries if the StethoScope option
was selected during the build. The Tornado User’s Guide describes the ld
library routine.
Running the Program
The real-time program defines a function, rt_main(), that spawns the tasks to
execute the model code and communicate with Simulink (if you selected
external mode during the build procedure.) It also initializes StethoScope if you
selected this option during the build procedure.
The rt_main function is defined in the rt_main.c application module. This
module is located in the matlabroot/rtw/c/tornado directory.
The rt_main function takes six arguments, and is defined by the following
ANSI C function prototype.
RT_MODEL * (*model_name)(void),
char_T
*optStr,
char_T
*scopeInstallString,
int_T
scopeFullNames,
int_T
priority,
int_T
port
Table 12-1 lists the arguments to this function.
12-19
12
Targeting Tornado for Real-Time Applications
Table 12-1: Arguments to the rt_main RT_MODEL
Argument
Description
model_name
A pointer to the entry point function in the generated code. This
function has the same name as the Simulink model. It registers the
local functions that implement the model code by adding function
pointers to the model’s rtM. See Chapter 7, “Program Architecture” for
more information.
optStr
The options string used to specify a stop time (-tf) and whether to wait
(-w) in external mode for a message from Simulink before starting the
simulation. An example options string is
"-tf 20 -w"
The -tf option overrides the stop time that was set during code
generation. If the value of the -tf option is inf, the program runs
indefinitely.
scopeInstallString
A character string that determines which signals are installed to
StethoScope. Possible values are:
• NULL — Install no signals. This is the default value.
• "*" — Install all signals.
• "[A-Z]*" — Install signals from blocks whose names start with an
uppercase letter.
Specifying any other string installs signals from blocks whose names
start with that string.
scopeFullNames
This argument determines whether StethoScope uses full hierarchical
block names for the signals it accesses or just the individual block
name. Possible values are:
• 1 Use full block names.
• 0 Use individual block names. This is the default value.
It is important to use full block names if your program has multiple
instances of a model or S-function.
12-20
Implementation Overview
Table 12-1: Arguments to the rt_main RT_MODEL (Continued)
Argument
Description
priority
The priority of the program’s highest priority task (tBaseRate). Not
specifying any value (or specifying a value of zero) assigns tBaseRate
to the default priority, 30.
port
The port number that the external mode sockets connection should
use. The valid range is 256 to 65535. When nothing is specified, the
port number defaults to 17725.
Passing optStr Via the Template Makefile. You can also pass the -w and -tf options
(see optStr in Table 12-1) to rt_main by using the PROGRAM_OPTS macro in
tornado.tmf. PROGRAM_OPTS passes a string of the form
-opt1 val1 -opt2 val2
In the following examples, the PROGRAM_OPTS directive sets an infinite stop
time and instructs the program to wait for a message from Simulink before
starting the simulation. Note that the argument string must be delimited by
single quotes nested within double quotes:
PROGRAM_OPTS = "'-tf inf -w'"
Including the extra single quotes ensures that the argument string will be
passed to the target program correctly, under both Windows and UNIX.
Calling rt_main. To begin program execution, call rt_main from WindSh. For
example,
sp(rt_main, vx_equal, "-tf 20 -w", "∗", 0, 30, 17725)
• Begins execution of the vx_equal model
• Specifies a stop time of 20 seconds
• Provides access to all signals (block outputs) in the model by StethoScope
• Uses only individual block names for signal access (instead of the
hierarchical name)
• Uses the default priority (30) for the tBaseRate task
• Uses TCP port 17725, the default
12-21
12
Targeting Tornado for Real-Time Applications
12-22
13
Asynchronous Support
The Interrupt Templates are blocks that you can use as templates for building your own
asynchronous interrupts. This chapter include the following topics:
Introduction (p. 13-2)
Accessing asynchronous templates from Real-Time
Workshop libraries
Interrupt Handling (p. 13-5)
Blocks that let you model synchronous/asynchronous
event handling, including interrupt service routines
Creating a Customized Asynchronous
Library (p. 13-21)
Guidelines for creating your own asynchronous blocks,
using templates provided
13
Asynchronous Support
Introduction
The Interrupt Templates library is part of the Real-Time Workshop library,
which you access via the Simulink Library Browser, as shown in Figure 13-1.
Do this by typing the MATLAB command
simulink
then by clicking the plus sign to the left of the Real-Time Workshop entry, and
clicking on Interrupt Templates.
13-2
Introduction
Figure 13-1: Interrupt Templates in Simulink Library Browser
Note that, depending on which MathWorks products you have installed, your
browser may show a different collection of libraries.
13-3
13
Asynchronous Support
Other sublibraries in the Real-Time Workshop library are:
• DOS Device Drivers: Blocks for use with DOS. See Appendix C, “Targeting
DOS for Real-Time Applications” for information.
• S-Function Target: The S-Function Target sublibrary contains only one
block type, the RTW S-Function block. This block is intended for use with
generated S-functions. See Chapter 10, “The S-Function Target” for more
information.
• VxWorks Support: A collection of blocks that support VxWorks (Tornado).
See Chapter 12, “Targeting Tornado for Real-Time Applications” for
information on VxWorks.
13-4
Interrupt Handling
Interrupt Handling
The blocks in the Interrupt Templates library allow you to model
synchronous/asynchronous event handling, including interrupt service
routines (ISRs). These blocks include:
• Asynchronous Rate Transition (reader)
• Asynchronous Buffer block (write)
• Interrupt Control block
• Unprotected Asynchronous Rate Transition block
• Task Synchronization block
Using these blocks, you can create models that handle asynchronous events,
such as hardware generated interrupts and asynchronous read and write
operations. The following sections discuss each of these blocks in the context of
VxWorks Tornado operating system.
Interrupt Control Block
Interrupt service routines (ISR) are realized by connecting the outputs of the
VxWorks Interrupt Control block to the control input of a function-call
subsystem, the input of a VxWorks Task Synchronization block, or the input to
a Stateflow chart configured for a function-call input event.
The Interrupt Control block installs the downstream (destination) function-call
subsystem as an ISR and enables the specified interrupt level. The current
implementation of the VxWorks Interrupt Control block supports VME
interrupts 1-7 and uses the VxWorks system calls sysIntEnable,
sysIntDisable, intConnect, intLock and intUnlock. Ensure that your target
architecture (BSP) for VxWorks supports these functions.
When a function-call subsystem is connected to an Interrupt Control block
output, the generated code for that subsystem becomes the ISR. For large
subsystems, this can have a large impact on interrupt response time for
interrupts of equal and lower priority in the system. As a general rule, it is best
to keep ISRs as short as possible. To do this, you should only connect
function-call subsystems that contain few blocks.
A better solution for large systems is to use the Task Synchronization block to
synchronize the execution of the function-call subsystem to an event. The Task
Synchronization block is placed between the Interrupt Control block and the
13-5
13
Asynchronous Support
function-call subsystem (or Stateflow chart). The Interrupt Control block then
installs the Task Synchronization block as the ISR, which releases a
synchronization semaphore (performs a semGive) to the function-call
subsystem and then returns. See the VxWorks Task Synchronization block for
more information.
Using the Interrupt Control Block
The Interrupt Control block has two modes that help support rapid
prototyping:
• RTW mode. In RTW mode, the Interrupt Control block configures the
downstream system as an ISR and enables interrupts during model startup.
You can select this mode using the Interrupt Control block dialog box when
generating code.
• Simulation mode. In Simulation mode, simulated Interrupt Request (IRQ)
signals are routed through the Interrupt Control block’s trigger port. Upon
receiving a simulated interrupt, the block calls the associated system.
You should select this mode when simulating, in Simulink, the effects of an
interrupt signal. Note that there can only be one VxWorks Interrupt Control
block in a model and all desired interrupts should be configured by it.
In both RTW and Simulation mode, in the event that two IRQ signals occur
simultaneously, the Interrupt Control block executes the downstream systems
according to their priority interrupt level.
The Interrupt Control block provides these two modes to make the
development and implementation of real-time systems that include ISRs easier
and quicker. You can develop two models, one that includes a plant and a
controller for simulation, and one that only includes the controller for code
generation.
Using the Library feature of Simulink, you can implement changes to both
models simultaneously. Figure 13-2 illustrates how changes made to the plant
or controller, both of which are in a library, propagate to the models.
13-6
Interrupt Handling
Plant
Controller
Library: Changes made here
affect both models.
Plant
Interrupt
Block
(Simulation
mode)
Interrupt
Block
Real-Time Workshop library
Interrupt
Block
(RTW mode)
Model
(for simulation)
Controller
Model
(for code generation)
Controller
Figure 13-2: Using the Interrupt Control Block with Simulink Library Feature
in Rapid Prototyping Process
Real-Time Workshop models normally run from a periodic interrupt. All blocks
in a model run at their desired rate by executing them in multiples of the timer
interrupt rate. Asynchronous blocks, on the other hand, execute based on other
interrupt(s) that may or may not be periodic.
The hardware that generates the interrupt is not configured by the Interrupt
Control block. Typically, the interrupt source is a VME I/O board, which
generates interrupts for specific events (e.g., end of A/D conversion). The VME
interrupt level and vector are set up in registers or by using jumpers on the
board. You can use the mdlStart routine of a user-written device driver
(S-function) to set up the registers and enable interrupt generation on the
board. You must match the interrupt level and vector specified in the Interrupt
Control block dialog to the level and vector setup on the I/O board.
13-7
13
Asynchronous Support
Interrupt Control Block Parameters
The picture below shows the VxWorks Interrupt Control block dialog box.
Parameters associated with the Interrupt Control block are:
• Mode: In Simulation mode, the ISRs are executed nonpreemptively. If they
occur simultaneously, signals are executed in the order specified by their
number (1 being the highest priority). Interrupt mapping during simulation
is left to right, top to bottom. That is, the first control input signal maps to
the topmost ISR. The last control input signal maps to the bottom most ISR.
In RTW mode, Real-Time Workshop uses vxinterrupt.tlc to realize
asynchronous interrupts in the generated code. The ISR is passed one
argument, the root SimStruct, and the Simulink definition of the
function-call subsystem is remapped to conform with the information in the
SimStruct.
• VME Interrupt Number(s): Specify the VME interrupt numbers for the
interrupts to be installed. The valid range is 1-7; for example: [4 2 5]).
• VME Interrupt Vector Offset Number(s): Real-Time Workshop uses this
number in the call to intConnect(INUM_TO_IVEC(#),...). You should
specify a unique vector offset number for each interrupt number.
• Preemption Flag(s): By default, higher priority interrupts can preempt lower
priority interrupts in VxWorks. If desired, you can lock out interrupts during
the execution of a ISR by setting the preemption flag to 0. This causes
13-8
Interrupt Handling
intLock() and intUnlock() calls to be inserted at the beginning and end of
the ISR respectively. This should be used carefully since it increases the
system’s interrupt response time for all interrupts at the
intLockLevelSet() level and below.
• IRQ Direction: In simulation mode, a scalar IRQ direction is applied to all
control inputs, and is specified as 1 (rising), -1 (falling), or 0 (either).
Configuring inputs separately in simulation is done prior to the control
input. For example, a Gain block set to -1 prior to a specific IRQ input will
change the behavior of one control input relative to another. In RTW mode
the IRQ direction parameter is ignored.
Interrupt Control Block Example - Simulation Mode
This example shows how the Interrupt Control block works in simulation
mode.
Simulated Interrupt Signals
13-9
13
Asynchronous Support
The Interrupt Control block works as a “handler” that routes signals and sets
priority. If two interrupts occur simultaneously, the rule for handling which
signal is sent to which port is left to right and top to bottom. This means that
IRQ2 receives the signal from Plant 1 and IRQ1 receives the signal from Plant
2 simultaneously. IRQ1 still has priority over IRQ2 in this situation.
Note that the Interrupt Control block executes during simulation by processing
incoming signals and executing downstream functions. Also, interrupt
preemption cannot be simulated.
Interrupt Control Block Example - RTW Mode
This example shows the Interrupt Control block in RTW mode.
(Note that Plant is removed.)
Stand-alone functions are
installed as ISR’s.
Interrupt Vector Table
Offset
192
&f()
In this example, the simulated plant signals that were included in the previous
example have been removed. In RTW mode, the Interrupt Control block
receives interrupts directly from the hardware.
During the Target Language Compiler phase of code generation, the Interrupt
Control block installs the code in the Stateflow chart and the Subsystem block
as interrupt service routines. Configuring a function-call subsystem as an ISR
requires two function calls, int_connect and int_enable. For example, the
function f(u) in the Function block requires that the Interrupt Control block
inserts a call to int_connect and sysIntEnable in the mdlStart function, as
shown below.
13-10
Interrupt Handling
/* model start function */
MdlStart()
{
. . .
int_connect(f,192,1);
. . .
sysIntEnable(1);
. . .
}
Locking and Unlocking ISRs. It is possible to lock ISRs so that they are not
preempted by a higher priority interrupt. Configuring the interrupt as
nonpreemptive has this effect. The following code fragment shows where
Real-Time Workshop places the int_lock and int_unlock functions to
configure the interrupt as nonpreemptive.
f()
{
lock = int_lock();
. . .
. . .
. . .
int_unlock(lock);
Real-Time Workshop code
}
Finally, the model’s terminate function disables the interrupt:
/* model terminate function */
MdlTerminate()
{
...
int_disable(1);
...
}
13-11
13
Asynchronous Support
Task Synchronization Block
The VxWorks Task Synchronization block is a function-call subsystem that
spawns, as an independent VxWorks task, the function-call subsystem
connected to its output. Typically it would be placed between the VxWorks
Interrupt Control block and a function-call subsystem block or a Stateflow
chart. Another example would be to place the Task Synchronization block at
the output of a Stateflow diagram that has an Event, “Output to Simulink,”
configured as a function-call.
The VxWorks Task Synchronization block performs the following functions:
• The downstream function-call subsystem is spawned as an independent task
using the VxWorks system call taskSpawn(). The task is deleted using
taskDelete() during model termination.
• A semaphore is created to synchronize the downstream system to the
execution of the Task Synchronization block.
• Code is added to this spawned function-call subsystem to wrap it in an
infinite while loop.
• Code is added to the top of the infinite while loop of the spawned task to wait
on the semaphore, using semTake(). When semTake() is first called, NO_WAIT
is specified. This allows the task to determine if a second semGive() has
occurred prior to the completion of the function-call subsystem. This would
indicate the interrupt rate is too fast or the task priority is too low.
• Synchronization code, i.e., semgive(), is generated for the Task
Synchronization block (a masked function-call subsystem). This allows the
output function-call subsystem to run. As an example, if you connect the
Task Synchronization block to the output of a VxWorks Interrupt Control
block, only a semGive() would occur inside an ISR.
13-12
Interrupt Handling
Task Synchronization Parameters
The picture below shows the VxWorks Task Synchronization block dialog box.
Parameters associated with the Task Synchronization block are:
• Task Name — An optional name, which if provided, is used as the first
argument to the taskSpawn() system call. This name is used by VxWorks
routines to identify the task they are called from to aid in debugging.
• Task Priority — The task priority is the VxWorks priority that the
function-call subsystem task is given when it is spawned. The priority can be
a very important consideration in relation to other tasks priorities in the
VxWorks system. In particular, the default priority of the model code is 30
and, when multitasking is enabled, the priority of the each subrate task
increases by one from the default model base rate. Other task priorities in
the system should also be considered when choosing a task priority.
VxWorks priorities range from 0 to 255 where a lower number is a higher
priority.
• Stack Size — The function-call subsystem is spawned with the stack size
specified. This is maximum size to which the task’s stack can grow. The value
should be chosen based on the number of local variables in the task.
By default, Real-Time Workshop limits the number of bytes for local
variables in all of the generated code to 8192 bytes (see assignment of
MaxStackSize in
matlabroot/rtw/c/tornado/tornado.tlc). As a rule, providing twice 8192
bytes (16384) for the one function that is being spawned as a task should be
sufficient.
13-13
13
Asynchronous Support
Task Synchronization Block Example
This example shows a Task Synchronization block as a simple ISR.
The Task Synchronization block inserts this code during the Target Language
Compiler phase of code generation:
• In MdlStart, the Task Synchronization block is registered by the Interrupt
Control block as an ISR. The Task Synchronization block creates and
initializes the synchronization semaphore. It also spawns the function-call
subsystem as an independent task.
/* Create and spawn task: <Root>/Faster Rate(.015) */
if ((*(SEM_ID *)rtPWork.s6_S_Function.SemID =
semBCreate(SEM_Q_PRIORITY, SEM_EMPTY)) == NULL)
ssSetErrorStatus(rtS,"semBCreate call failed "
"for block <Root>/Faster Rate(.015).\n ");
}
if ((rtIWork.s6_S_Function.TaskID = taskSpawn("root_Faster_", 20,
VX_FP_TASK,
1024, (FUNCPTR)Sys_root_Faster__OutputUpdate,
(int_T)rtS, 0, 0, 0, 0, 0, 0, 0, 0, 0)) == ERROR) {
ssSetErrorStatus(rtS,"taskSpawn call failed for block
<Root>/ Faster Rate "
"(.015).\n");
}
• The Task Synchronization block modifies the downstream function-call
subsystem by wrapping it inside an infinite loop and adding semaphore
synchronization code.
/* Output and update for function-call system: <Root>/Faster
Rate(.015) */
void Sys_root_Faster__OutputUpdate(void *reserved, int_T
controlPortIdx, int_T tid)
13-14
Interrupt Handling
{
/* Wait for semaphore to be released by system: <Root>/Task
Synchronization */
for(;;) {
if (semTake(*(SEM_ID *)rtPWork.s6_S_Function.SemID,NO_WAIT)
!= ERROR) {
logMsg("Rate for function-call subsystem"
"Sys_root_Faster__OutputUpdate()
fast.\n",0,0,0,0,0,0);
#if STOPONOVERRUN
logMsg("Aborting real-time simulation.\n",0,0,0,0,0,0);
semGive(stopSem);
return(ERROR);
#endif
} else {
semTake(*(SEM_ID
*)rtPWork.s6_S_Function.SemID, WAIT_FOREVER);
}
/* UniformRandomNumber Block: <S3>/Uniform Random Number */
rtB.s3_Uniform_Random_Number =
rtRWork.s3_Uniform_Random_Number.NextOutput;
.
.
.
}
13-15
13
Asynchronous Support
Asynchronous Rate Transition Block
The VxWorks Asynchronous Rate Transition blocks are meant to be used to
interface signals to asynchronous function-call subsystems in a model. This is
needed whenever a function-call subsystem has input or output signals and its
control input ultimately connects (sources) to the VxWorks Interrupt Control
block or Task Synchronization block.
Because an asynchronous function-call subsystem can preempt or be
preempted by other model code, an inconsistency arises when more than one
signal element is connected to it. The issue is that signals passed to and/or from
the function-call subsystem can be in the process of being written or read when
the preemption occurs. Thus, partial old and partial new data will be used.
This situation can also occur with scalar signals in some cases. For example, if
a signal is a double (8 bytes), the read or write operation may require two
assembly instructions.
The Asynchronous Rate Transition blocks can be used to guarantee the data
passed to and/or from the function-call subsystem is all from the same
iteration.
The Asynchronous Rate Transition blocks are used in pairs, with a write side
driving the read side. To ensure the data integrity, no other connections are
allowed between the two Asynchronous Rate Transition blocks. The pair works
by using two buffers (“double buffering”) to pass the signal and, by using
mutually exclusive control, allow only exclusive access to each buffer. For
example, if the write side is currently writing into one buffer, the read side can
only read from the other buffer.
The initial buffer is filled with zeros so that if the read side executes before the
write side has had time to fill the other buffer, the read side will collect zeros
from the initial buffer.
13-16
Interrupt Handling
Asynchronous Rate Transition Block Parameters
There are two kinds of Asynchronous Rate Transition blocks, a reader and a
writer. The picture below shows the Asynchronous Rate Transition block’s
dialog boxes.
Both blocks require the Sample Time parameter. The sample time should be
set to -1 inside a function call and to the desired time otherwise.
13-17
13
Asynchronous Support
Asynchronous Rate Transition Block Example
This example shows how you might use the Asynchronous Rate Transition
block to control the execution of an interrupt service routine.
The ISR0 subsystem block, which is configured as a function-call subsystem,
contains another set of Asynchronous Rate Transition blocks.
Unprotected Asynchronous Rate Transition Block
The VxWorks Unprotected Asynchronous Rate Transition block provides a
sample time for blocks connected to an asynchronous function-call subsystem
when double buffering is not required. There are two options for connecting I/O
to an asynchronous function-call subsystem:
• Use the Unprotected Asynchronous Rate Transition block, or some other
block that requires a sample time to be set, at the input or output of the
13-18
Interrupt Handling
asynchronous function-call subsystem. This will cause blocks up- or
downstream from it, which would otherwise inherit from the function-call
subsystem, to use the sample time specified. Note that if the signal width is
greater than 1, data consistency is not guaranteed, which may or may not an
issue. See next option.
The Unprotected Asynchronous Rate Transition block does not introduce any
system delay. It only specifies the sample time of the downstream blocks. It
also informs Simlink to allow a non-buffered asynchronous connection. This
block is typically used for scalar signals that do not require double buffering.
• Use the Asynchronous Rate Transition block pair. This not only will set the
sample time of the blocks up or downstream that would otherwise inherit
from the function-call subsystem, and also guarantees consistency of the
data on the signal. See the Asynchronous Rate Transition block for more
information on data consistency.
Unprotected Asynchronous Rate Transition Block Parameters
This picture shows the VxWorks Unprotected Asynchronous Rate Transition
block’s dialog box.
The Sample time parameter sets the sample time to the desired rate.
13-19
13
Asynchronous Support
Unprotected Asynchronous Rate Transition Block Example
This picture shows a sample application of the Rate Transition block in an ISR.
In this example, the Rate Transition block on the input to the function-call
subsystem causes both the In and Gain1 blocks to run at the 0.1 second rate.
The Rate Transition block on the output of the function-call subsystem causes
both the Gain2 and Out blocks to run at the 0.2 second rate. Using this scheme
informs Simlink to allow non-buffered connections to an asynchronous
function-call subsystem.
13-20
Creating a Customized Asynchronous Library
Creating a Customized Asynchronous Library
You can use the Real-Time Workshop VxWorks asynchronous blocks as
templates that provide a starting point for creating your own asynchronous
blocks. Templates are provided for these blocks:
• Asynchronous Rate Transition block
• Interrupt Control block
• Unprotected Asynchronous Rate Transition block
• Task Synchronization block
You can customize each of these blocks by implementing a set of modifications
to files associated with each template. These files are:
• The block’s underlying S-function C MEX-file
Note that SS_OPTION_ASYNCHRONOUS_INTERRUPT should be used when a
function-call subsystem is attached to an interrupt. For further information,
see documentation for SS_OPTION and SS_OPTION_ASYNCHRONOUS in
matlabroot/simulink/include/simstuc.h
• The block’s mask and the associated mask M-file
Note that the strings 'read' and 'write' must appear in the mask types for
rate transition blocks.
• The TLC files that control code generation of the block
At a minimum, you must rename the system calls generated by the TLC files
to the correct names for the new real-time operating system (RTOS) and supply
the correct arguments for each file. There is a collection of files that you must
copy (and rename) from matlabroot/rtw/c/tornado/devices into a new
directory, for example, matlabroot/rtw/c/my_os/devices. These files are:
• Asynchronous Rate Transition block — vxdbuffer.tlc, vxdbuffer.c
• Interrupt Control block — vxinterrupt.tlc, vxinterrupt.c, vxintbuild.m
• O/S include file — vxlib.tlc
• Task Synchronization block — vxtask.tlc, vxtask.c
13-21
13
Asynchronous Support
13-22
14
Targeting Real-Time
Systems
This chapter provides information necessary to implement a custom target configuration, and covers
the followng topics:
Introduction (p. 14-2)
Motivation for implementing a custom target
configuration
Components of a Custom Target
Configuration (p. 14-3)
Overview of the code and control files that make up a
custom target configuration
Tutorial: Creating a Custom Target
Configuration (p. 14-9)
A hands-on exercise in building a custom rapid
prototyping target
Customizing the Build Process
(p. 14-16)
Information on System Target File Structure and
Template Makefiles, and how to modify them
Creating Device Drivers (p. 14-39)
Implementation of device drivers as S-Function blocks,
including both inlined and noninlined drivers
Interfacing Parameters and Signals
(p. 14-70)
Guidelines for use of the Real-Time Workshop signal
monitoring and parameter tuning APIs
Creating an External Mode
Communication Channel (p. 14-94)
How to support external mode on your custom target,
using your own low-level communications layer
Combining Multiple Models (p. 14-103) Strategies for combining several models (or several
instances of the same model) into a single executable
DSP Processor Support (p. 14-107)
How to emulate register sizes smaller than 32 bits
14
Targeting Real-Time Systems
Introduction
The target configurations bundled with Real-Time Workshop are suitable for
many different applications and development environments. Third-party
targets provide additional versatility. However, a number of users find that
they require a custom target configuration.You may want to implement a
custom target configuration for any of the following reasons:
• To support custom hardware and incorporate custom device driver blocks
into your models.
• To customize a bundled target configuration — such as the generic real-time
(GRT) or Real-Time Workshop Embedded Coder targets — to your needs.
• To configure the build process for a special compiler (such as a compiler for
an embedded microcontroller or DSP board).
As part of your custom target implementation, you may also need to:
• Interface generated model code with existing supervisory or supporting code
that calls the generated code.
• Interface signals and parameters within generated code to your own code.
• Combine code generated from multiple models into a single system.
• Implement external mode communication via your own low-level protocol
layer.
14-2
Components of a Custom Target Configuration
Components of a Custom Target Configuration
The components of a custom target configuration are:
• Code to supervise and support execution of generated model code
• Control files:
- A system target file to control the code generation process
- A template makefile to build the real-time executable
This section summarizes key concepts and terminology you will need to know
to begin developing each component. References to more detailed information
sources are provided, in case any of these topics are unfamiliar to you.
Code Components
A Real-Time Workshop program containing code generated from a Simulink
model consists of a number of code modules and data structures. These fall into
two categories.
Application Components
Application components are those which are specific to a particular model; they
implement the functions represented by the blocks in the model. Application
components are not specific to the target. Application components include:
• Modules generated from the model
• User-written blocks (S-functions)
• Parameters of the model that are visible, and can be interfaced to, external
code
Run-Time Interface Components
A number of code modules and data structures, referred to collectively as the
run-time interface, are responsible for managing and supporting the execution
of the generated program. The run-time interface modules are not
automatically generated. To develop a custom target, you must implement
14-3
14
Targeting Real-Time Systems
certain parts of the run-time interface. Table 14-1 summarizes the run-time
interface components.
Table 14-1: Run-Time Interface Components
User Provides:
Real-Time Workshop Provides:
Customized main program
Generic main program
Timer interrupt handler to
run model
Execution engine and integration
solver (called by timer interrupt
handler)
Other interrupt handlers
Example interrupt handlers
(Asynchronous Interrupt Blocks)
Device drivers
Example device drivers
Data logging and signal
monitoring user interface
Data logging, parameter tuning,
signal monitoring, and external mode
support
The components of the run-time interface vary, depending upon whether the
target is an embedded system or a rapid prototyping environment.
User-Written Run-Time Interface Code
Most of the run-time interface is provided by Real-Time Workshop. You must
implement the following elements:
• A timer interrupt service routine (ISR). The timer runs at the program’s base
sample rate. The timer ISR is responsible for operations that must be
completed within a single clock period, such as computing the current output
sample. The timer ISR usually calls the Real-Time Workshop-supplied
function, rt_OneStep.
• The main program. Your main program initializes the blocks in the model,
installs the ISR, and executes a background task or loop. The timer
periodically interrupts the main loop. If the main program is designed to run
for a finite amount of time, it is also responsible for cleanup operations - such
as memory deallocation and masking the timer interrupt - before
terminating the program.
14-4
Components of a Custom Target Configuration
• Device drivers to communicate with your target hardware.
Run-Time Interface for Rapid Prototyping
The run-time interface for a rapid prototyping target includes:
• Supervisory logic
- The main program
- Execution engine and integration solver
• Supporting logic
- I/O drivers
- Code to handle timing, and interrupts
• Monitoring, tuning, and debugging support
- Data logging code
- Signal monitoring
- Real-time parameter tuning
- External mode communications
The structure of the rapid prototyping run-time interface, and the execution of
rapid prototyping code, are detailed in Chapter 7, “Program Architecture” and
Chapter 8, “Models with Multiple Sample Rates.”
Development of a custom rapid prototyping target generally begins with
customization of one of the generic main programs, grt_main.c or
grt_malloc_main.c. As described in “User-Written Run-Time Interface Code”
above, you must modify the main program for real-time interrupt-driven
execution. You must also supply device drivers (optionally inlined).
Run-Time Interface for Embedded Targets
The run-time interface for an embedded (production) target includes:
• Supervisory logic
- The main program
- Execution engine and integration solver
• Supporting logic
- I/O drivers
- Code to handle timing, and interrupts
14-5
14
Targeting Real-Time Systems
• Monitoring and debugging support
- Data logging code
- Access to tunable parameters and external signals
Development of a custom embedded target generally begins with customization
of the Real-Time Workshop Embedded Coder main program, ert_main.c. The
Real-Time Workshop Embedded Coder documentation details the structure of
the Real-Time Workshop Embedded Coder run-time interface and the
execution of Real-Time Workshop Embedded Coder code, and provides
guidelines for customizing ert_main.c.
Control Files
System Target Files
The Target Language Compiler (TLC) generates target-specific C code from an
intermediate description of your Simulink block diagram (model.rtw). The
Target Language Compiler reads model.rtw and executes a program
consisting of several target files (.tlc files.) The output of this process is a
number of source files, which are fed to your development system’s make
utility.
The system target file controls the code generation process. You will need to
create a customized system target file to set code generation parameters for
your target. We recommend that you copy, rename, and modify one of the
standard system target files:
• The generic real-time (GRT) target file, matlabroot/rtw/c/grt/grt.tlc, for
rapid prototyping targets
• The Real-Time Workshop Embedded Coder target file,
matlabroot/rtw/c/ert/ert.tlc, for embedded (production) targets
Chapter 2, “Building an Application” of the Getting Started Guide and Chapter
2, “Code Generation and the Build Process” describe the role of the system
target file in the code generation and build process. Guidelines for creating a
custom system target file are given in “Customizing the Build Process” on
page 14-16.
14-6
Components of a Custom Target Configuration
Template Makefiles
A template makefile (.tmf file) provides information about your model and
your development system. Real-Time Workshop uses this information to create
an appropriate makefile (.mk file) to build an executable program. Real-Time
Workshop provides a large number of template makefiles suitable for different
types of targets and development systems. The standard template makefiles
are described in “Template Makefiles and Make Options” on page 2-54.
If one of the standard template makefiles meets your requirements, you can
simply copy and rename it in accordance with the conventions of your project.
If you need to make more extensive modifications, see “Template Makefiles” on
page 14-28 for a full description of the structure of template makefiles.
Hook Files for Communicating Target-specific Word Characteristics
In order to communicate details about target hardware characteristics, such as
word lengths and overflow behavior, you need to supply an M-file named
<target>_rtw_info_hook.m. Each system target file needs to implement a
hook file. Those provided for built-in targets are placed in the respective target
directories under toolbox/rtw/targets/.
Hook files provide an API to describe two essential aspects of hardware
characteristics:
• Word lengths (number of bits), specified via
- CharNumBits
Size of C char type
- ShortNumBits
Size of C short type
- IntNumBits
Size of C int type
- LongNumBits
Size of C long type
• Implementation-specific properties, specified as logical values
- ShiftRightIntArith
Set true if shift right on a signed integer is
implemented as arithmetic shift, and false
otherwise.
- Float2IntSaturates
Conversion from float to integer automatically
saturates, thus do not generate software
saturation code.
- IntPlusIntSaturates
Integer addition automatically saturates, thus
do not generate software saturation code.
14-7
14
Targeting Real-Time Systems
- IntTimesIntSaturates
Integer multiply automatically saturates, thus
do not generate software saturation code.
To supply a hookfile for the GRT target (grt.tlc), for example, you must name
the file grt_rtw_info_hook.m, and place it somewhere on the MATLAB path.
If the hook file is present, the target-specific information is extracted via the
API found in this file. If the hookfile is not provided, default values based on
the host’s characteristics will be used, which may not be appropriate.
For an example, see toolbox/rtw/rtwdemos/example_rtw_info_hook.m.
Note The TLC directive %assign DSP = 1 no longer has any effect. You need
to provide a hook file instead.
14-8
Tutorial: Creating a Custom Target Configuration
Tutorial: Creating a Custom Target Configuration
This tutorial walks through the task of creating a skeletal rapid prototyping
target. This exercise illustrates several tasks that are usually required when
creating a custom target:
• Incorporating a noninlined S-function into a model for use in simulation.
• Inlining the S-function in the generated code, using a corresponding TLC
file.
In a real-world application, you would incorporate inlined and noninlined
device driver S-functions into the model and the generated code. In this
tutorial, we inline a simple S-function that multiplies its input by two.
• Making minor modifications to a standard system target file and template
makefile.
• Generating code from the model by invoking your customized system target
file and template makefile.
You can use this process as a starting point for your own projects.
This example uses the LCC compiler under Windows. LCC is distributed with
Real-Time Workshop. If you use a different compiler, you can set up LCC
temporarily as your default compiler by typing the MATLAB command
mex -setup
A command prompt window will open; follow the prompts and select LCC.
Note On UNIX systems, make sure that you have a C compiler installed. You
can then do this exercise substituting appropriate UNIX directory syntax.
In this example, the code is generated from targetModel.mdl, a very simple
fixed-step model (see Figure 14-1). The resultant program behaves exactly as
if it had been built for the generic real-time target.
14-9
14
Targeting Real-Time Systems
Figure 14-1: targetModel.mdl
The S-Function block will use the source code from the timestwo example. See
the Writing S-Functions manual for a complete discussion of this S-function.
The Target Language Compiler documentation discusses timestwo.tlc, the
inlined version of timestwo.
To create the skeletal target system:
1 Create a directory to store your C source files and .tlc and .tmf files. We
refer to this directory as d:/work/mytarget.
2 Add d:/work/mytarget to your MATLAB path.
addpath d:/work/mytarget
3 Make d:/work/mytarget your working directory. Real-Time Workshop
writes the output files of the code generation process into a build directory
within the working directory.
4 Copy the timestwo S-function C source code from
matlabroot/toolbox/rtw/rtwdemos/tlctutorial/timestwo/timestwo.c
to
d:/work/mytarget.
5 Build the timestwo MEX-file in d:/work/mytarget.
mex timestwo.c
6 Create the model as illustrated in Figure 14-1. Use an S-Function block from
the Simulink Functions & Tables library in the Library Browser. Set the
solver options to fixed-step and ode4.
7 Double-click the S-Function block to open the Block Parameters dialog.
Enter the S-function name timestwo. The block is now bound to the
timestwo MEX-file. Click OK.
14-10
Tutorial: Creating a Custom Target Configuration
8 Open the Scope and run the simulation. Verify that the timestwo S-function
multiplies its input by 2.0.
9 In order to generate inlined code from the timestwo S-Function block, you
must have a corresponding TLC file in the working directory. If the Target
Language Compiler detects a C-code S-function and a TLC file with the
same name in the working directory, it generates inline code from the TLC
file. Otherwise, it generates a function call to the external S-function.
To ensure that the build process generates inlined code from the timestwo
block, copy the timestwo TLC source code from
matlabroot/toolbox/rtw/rtwdemos/tlctutorial/timestwo/timestwo.tl
c to
d:/work/mytarget.
10 Make local copies of the main program and system target files.
matlabroot/rtw/c/grt contains the main program (grt_main.c) and the
system target file (grt.tlc) for the generic real-time target. Copy
grt_main.c and grt.tlc to d:/work/mytarget. Rename them to
mytarget_main.c and mytarget.tlc.
11 Remove the initial comment lines from mytarget.tlc. The lines to remove
are shown below.
%% SYSTLC: Generic Real-Time Target \
%%
TMF: grt_default_tmf MAKE: make_rtw EXTMODE: ext_comm
%% SYSTLC: Visual C/C++ Project Makefile only for the "grt" target
\
%%
TMF: grt_msvc.tmf MAKE: make_rtw EXTMODE: ext_comm
The initial comment lines have significance only if you want to add
my_target to the System Target File Browser. For now you should remove
them.
14-11
14
Targeting Real-Time Systems
12 Real-Time Workshop creates a build directory in your working directory to
store files created during the code generation process. The build directory is
given the name of the model, followed by a suffix. This suffix is specified in
the rtwgensettings structure in the system target file.
To set the suffix to a more appropriate string, change the line
rtwgensettings.BuildDirSuffix = '_grt_rtw'
to
rtwgensettings.BuildDirSuffix = '_mytarget_rtw'
Your build directory will be named targetModel__mytarget_rtw.
13 Make a local copy of the template makefile. matlabroot/rtw/c/grt contains
several compiler-specific template makefiles for the generic real-time target.
The appropriate template makefile for the LCC compiler is grt_lcc.tmf.
Copy grt_lcc.tmf to d:/work/mytarget, and rename it to mytarget.tmf.
Note Some of the template makefile modifications described in the next step
are specific to the LCC template makefile. If you are using a different compiler
and template makefile, the rules for the source (REQ_SRCS) and object file
(%.obj :) lists may differ slightly.
14-12
Tutorial: Creating a Custom Target Configuration
14 Modify mytarget.tmf. The SYS_TARGET FILE parameter must be changed so
that the correct file reference is generated in the make file. Change the line
SYS_TARGET FILE = grt.tlc
to
SYS_TARGET FILE = mytarget.tlc
Also, change the source file list to include mytarget_main.c instead of
grt_main.c.
REQ_SRCS = $(MODEL).c $(MODULES) mytarget_main.c...
Finally, change the line
%.obj : $(MATLAB_ROOT)/rtw/c/grt/%.c
to
%.obj : d:/work/mytarget/%.c
15 This exercise requires no changes to mytarget_main.c. In an actual
application, you would modify mytarget_main.c to execute your model code
under the control of a timer interrupt, and make other changes.
14-13
14
Targeting Real-Time Systems
16 Open the Real-Time Workshop pane in the Simulation Parameters dialog.
Select Target configuration from the Category menu. Enter the system
target file, template makefile, and Make command parameters as below.
Be sure to explicitly specify the full name and extension of the template
makefile (mytarget.tmf) in the Make command field, as shown.
17 Click the Apply button.
18 Click the Build button. If the build is successful, MATLAB will display the
message below.
### Created executable: targetModel.exe
### Successful completion of Real-Time Workshop build procedure
for model: targetModel
Your working directory will contain the targetModel.exe file and the build
directory, targetModel_mytarget_rtw.
14-14
Tutorial: Creating a Custom Target Configuration
19 Edit the generated file
d:/work/mytarget/targetModel_mytarget_rtw/targetModel.c and locate
the MdlOutputs function. Observe the inlined code.
/* S-Function Block: <Root>/S-Function (timestwo) */
rtB.S_Function = 2.0 * rtB.Sine_Wave;
Because the working directory contained a TLC file (timestwo.tlc) with
the same name as the timestwo S-Function block, the Target Language
Compiler generated inline code instead of a function call to the external Ccode S-function.
20 As an optional final step to this exercise, you may want to add your custom
target configuration to the System Target File Browser. See “Adding a
Custom Target to the System Target File Browser” on page 14-27 to learn
how to do this.
14-15
14
Targeting Real-Time Systems
Customizing the Build Process
The Real-Time Workshop build process proceeds in two stages. The first stage
is code generation. The system target file exerts overall control of the code
generation stage. In the second stage, the template makefile generates a .mk
file, which compiles and links code modules into an executable.
In developing your custom target, you may need to create a customized system
target file and/or template makefile. This section provides information on the
structure of these files, and guidelines for modifying them.
System Target File Structure
This section is a guide to the structure and contents of a system target file. You
may want to refer to the system target files provided with Real-Time Workshop
while reading this section. Most of these files are stored in the target-specific
directories under matlabroot/rtw/c. Additional system target files are stored
in matlabroot/toolbox/rtw/targets/rtwin/rtwin and
matlabroot/toolbox/rtw/targets/xpc/xpc.
Before creating or modifying a system target file, you should acquire a working
knowledge of the Target Language Compiler. The Target Language Compiler
documentation documents the features and syntax of the language.
Figure 14-2 shows the general structure of a system target file.
14-16
Customizing the Build Process
%% SYSTLC: Example Real-Time Target
%%
TMF: example.tmf MAKE: make_rtw EXTMODE: ext_comm
%% Inital comments contain directives for System Target File Browser.
%% Documentation, date, copyright, and other info may follow.
%%
%% TLC Configuration Variables Section -----------------------------%% Assign code format, language, target type.
%%
%assign CodeFormat = "Embedded-C"
%assign TargetType = "RT"
%assign Language
= "C"
%%
%% TLC Program Entry Point -----------------------------------------%% Call entry point function.
%include "codegenentry.tlc"
%%
%% RTW Options Section ---------------------------------------------/%
BEGIN_RTW_OPTIONS
%% Define rtwoptions structure array. This array defines target-specific
%% code generation variables, and controls how they are displayed.
rtwoptions(1).prompt = 'example code generation options';
.
.
rtwoptions(6).prompt = 'Show eliminated statements';
rtwoptions(6).type = 'Checkbox';
.
.
%% Define additional TLC variables here.
.
.
%% Define suffix string for naming build directory here.
%%
rtwgensettings.BuildDirSuffix = '_mytarget_rtw'
END_RTW_OPTIONS
%/
Browser
Comments
TLC Configuration
Variables
TLC Program Entry
Point
rtwoptions Array
and Other TLC
Variables
Build
Directory
Name
Figure 14-2: Structure of a System Target File
Browser Comments
This section is optional. You can place comment lines at the head of the file to
identify your system target file to the System Target File Browser. These lines
have significance to the browser only. During code generation, the Target
Language Compiler treats them as comments.
14-17
14
Targeting Real-Time Systems
Note that you must place the browser comments at the head of the file, before
any other comments or TLC statements.
The comments contain the following directives:
• SYSTLC: This string is a descriptor that appears in the browser.
• TMF: Name of the template makefile to use during build process. When the
target is selected, this filename is displayed in the Template makefile field
of the Target configuration section of the Real-Time Workshop pane.
• MAKE: make command to use during build process. When the target is selected,
this command is displayed in the Make command field of the Target
configuration section of the Real-Time Workshop pane.
• EXTMODE: Name of external mode interface file (if any) associated with your
target. If your target does not support external mode, use no_ext_comm.
The following browser information comments are from
matlabroot/rtw/c/grt/grt.tlc.
%% SYSTLC: Generic Real-Time Target
%%
TMF: grt_default_tmf MAKE: make_rtw EXTMODE: ext_comm
See “Adding a Custom Target to the System Target File Browser” on
page 14-27 for further information.
Target Language Compiler Configuration Variables
This section assigns global TLC variables that affect the overall code
generation process. The following variables must be assigned:
• CodeFormat: The CodeFormat variable selects one of the available code
formats:
- RealTime: Designed for rapid prototyping, with static memory allocation.
- RealTimeMalloc: Similar to RealTime, but with dynamic memory
allocation.
- Embedded-C: Designed for production code, minimal memory usage,
simplified interface to generated code.
- S-Function: For use by S-function and Accelerator targets only.
The default CodeFormat value is RealTime.
Chapter 3, “Generated Code Formats” summarizes available code formats
and provides pointers to further details.
14-18
Customizing the Build Process
• Language: Selects code generation (currently C only).
It is possible to generate code in a language other than C. To do this would
require considerable development effort, including reimplementation of all
block target files to generate the desired target language code. See the
Target Language Compiler documentation for a discussion of the issues.
• TargetType: Real-Time Workshop defines the preprocessor symbols RT and
NRT to distinguish simulation code from real-time code. These symbols are
used in conditional compilation. The TargetType variable determines
whether RT or NRT is defined.
Most targets are intended to generate real-time code. They assign
TargetType as follows.
%assign TargetType = "RT"
Some targets, such as the Simulink Accelerator, generate code for use in non
real-time only. Such targets assign TargetType as follows.
%assign TargetType = "NRT"
See “Conditional Compilation for Simulink and Real-Time” on page 14–45
for further information on the use of these symbols.
Target Language Compiler Program Entry Point
The code generation process normally begins with codegenentry.tlc. The
system target file invokes codegenentry.tlc as follows.
%include "codegenentry.tlc"
14-19
14
Targeting Real-Time Systems
codegenentry.tlc in turn invokes other TLC files:
• genmap.tlc maps the block names to corresponding language-specific block
target files.
• commonsetup.tlc sets up global variables.
• commonentry.tlc starts the process of generating code in the format
specified by CodeFormat.
To customize the code generation process, you can call the lower-level TLC files
explicitly and include your own TLC functions at each stage of the process. See
the Target Language Compiler documentation for guidelines.
Note codegenentry.tlc and the lower-level TLC files assume that
CodeFormat, TargetType, and Language have been correctly assigned. Set
these variables before including codegenentry.tlc.
RTW_OPTIONS Section
The RTW_OPTIONS section (see Figure 14-2) is bounded by the directives:
%/
BEGIN_RTW_OPTIONS
.
.
END_RTW_OPTIONS
/%
The first part of the RTW_OPTIONS section defines an array of rtwoptions
structures. The rtwoptions structure is discussed in this section.
The second part of the RTW_OPTIONS section defines rtwgensettings, a
structure defining the build directory name and other settings for the code
generation process. See “Build Directory Name” on page 14-26 for information
about rtwgensettings.
The rtwoptions Structure. The fields of the rtwoptions structure define variables
and associated user interface elements to be displayed in the Real-Time
Workshop pane. Using the rtwoptions structure array, you can customize the
14-20
Customizing the Build Process
Category menu in the Real-Time Workshop pane, define the options displayed
in each category, and specify how these options are processed.
When the Real-Time Workshop pane opens, the rtwoptions structure array is
scanned and the listed options are displayed. Each option is represented by an
assigned user interface element (check box, edit field, pop-up menu, or
pushbutton), which displays the current option value.
The user interface elements can be in an enabled or disabled (grayed-out) state.
If the option is enabled, the user can change the option value.
You can also use the rtwoptions structure array to define special NonUI
elements that cause callback functions to be executed, but that are not
displayed in the Real-Time Workshop pane. See “NonUI Elements” on page 14–
24 for details.
The elements of the rtwoptions structure array are organized into groups that
correspond to items in the Category menu in the Real-Time Workshop pane.
Each group of items begins with a header element of type Category. The
default field of a Category header must contain a count of the remaining
elements in the category.
The header is followed by options to be displayed on the Real-Time Workshop
pane. The header in each category is followed by a maximum of seven elements.
Table 14-2 summarizes the fields of the rtwoptions structure.
The following example is excerpted from
matlabroot/rtw/c/rtwsfcn/rtwsfcn.tlc, the system target file for the
S-Function target. The code defines an rtwoptions structure array of three
elements. The default field of the first (header) element is set to 2, indicating
the number of elements that follow the header.
rtwoptions(1).prompt = 'RTW S-function code generation options';
rtwoptions(1).type = 'Category';
rtwoptions(1).enable = 'on';
rtwoptions(1).default = 2; % Number of items under this category
% excluding this one.
rtwoptions(1).popupstrings = '';
rtwoptions(1).tlcvariable
= '';
rtwoptions(1).tooltip
= '';
rtwoptions(1).callback
= '';
rtwoptions(1).opencallback = '';
rtwoptions(1).closecallback = '';
rtwoptions(1).makevariable = '';
rtwoptions(2).prompt = 'Create New Model';
14-21
14
Targeting Real-Time Systems
rtwoptions(2).type = 'Checkbox';
rtwoptions(2).default = 'on';
rtwoptions(2).tlcvariable = 'CreateModel';
rtwoptions(2).makevariable = 'CREATEMODEL';
rtwoptions(2).tooltip = ...
['Create a new model containing the generated RTW S-Function block inside it'];
rtwoptions(3).prompt = 'Use Value for Tunable Parameters';
rtwoptions(3).type = 'Checkbox';
rtwoptions(3).default = 'off';
rtwoptions(3).tlcvariable = 'UseParamValues';
rtwoptions(3).makevariable = 'USEPARAMVALUES';
rtwoptions(3).tooltip = ...
['Use value instead of variable name in generated block mask edit fields'];
The first element adds the RTW S-function code generation options item to
the Category menu of the Real-Time Workshop pane. The options defined in
rtwoptions(2) and rtwoptions(3) display as shown in Figure 14-3.
Figure 14-3: Code Generation Options for S-Function Target
If you want to define more than seven options, you can define multiple
Category menu items within a single system target file. For an example, see
the Tornado system target file, matlabroot/rtw/c/tornado/tornado.tlc.
Note that to verify the syntax of your rtwoptions definitions, you can execute
the commands in MATLAB by copying and pasting them to the MATLAB
command window.
14-22
Customizing the Build Process
For further examples of target-specific rtwoptions definitions, see “Using
rtwoptions: the Real-Time Workshop Options Example Target” on page 14-25.
The following table lists the fields of the rtwoptions structure.
Table 14-2: rtwoptions Structure Fields Summary
Field Name
Description
callback
Name of M-code function to call when value of option
changes. To access objects such as your Simulation
Parameters dialog custom option fields, pass in a
handle to the Simulation Parameters dialog. To do this,
use the reserved keyword DialogFig.
Note that DialogFig is a reserved keyword that should
be used with extreme caution. For an example of
callback usage, see “Using rtwoptions: the Real-Time
Workshop Options Example Target” on page 14-25.
closecallback
Name of M-code function to call when be executed when
dialog closes. To access objects such as your Simulation
Parameters dialog custom option fields, pass in a
handle to the Simulation Parameters dialog. To do this,
use the reserved keyword DialogFig.
Note that DialogFig is a reserved keyword that should
be used with extreme caution. For an example of
closecallback usage, see “Using rtwoptions: the
Real-Time Workshop Options Example Target” on
page 14-25.
default
Default value of the option (empty if the type is
Pushbutton).
enable
Must be on or off. If on, the option is displayed as an
enabled item; otherwise, as a disabled item.
14-23
14
Targeting Real-Time Systems
Table 14-2: rtwoptions Structure Fields Summary
Field Name
Description
makevariable
Template makefile token (if any) associated with
option. The makevariable will be expanded during
processing of the template makefile. See “Template
Makefile Tokens” on page 14-29.
opencallback
M-code to be executed when dialog opens. The purpose
of the code is to synchronize the displayed value of the
option with its previous setting. For an example of
opencallback usage, see “Using rtwoptions: the
Real-Time Workshop Options Example Target” on
page 14-25.
popupstrings
If type is Popup, popupstrings defines the items in the
pop-up menu. Items are delimited by the “|” (vertical
bar) character. The following example defines the items
of the MAT-file variable name modifier menu used by
the GRT target:
'rt_|_rt|none'
prompt
Label for the option.
tlcvariable
Name of TLC variable associated with the option.
tooltip
Help string displayed when mouse is over the item.
type
Type of element: Checkbox, Edit, NonUI, Popup,
Pushbutton, or Category.
NonUI Elements
Elements of the rtwoptions array that have type NonUI exist solely to invoke
callbacks. A NonUI element is not displayed in the Simulation Parameters
dialog. You can use a NonUI element if you wish to execute a callback that is not
associated with any user interface element, when the dialog opens or closes.
Only the opencallback and closecallback fields of a NonUI element have
significance. See the next section,“Using rtwoptions: the Real-Time Workshop
Options Example Target” for an example.
14-24
Customizing the Build Process
Using rtwoptions: the Real-Time Workshop Options Example Target
A working system target file, with M-file callback functions, has been provided
as an example of how to use the rtwoptions structure to display and process
custom options on the Real-Time Workshop pane. The example files are in the
directory
matlabroot/toolbox/rtw/rtwdemos/rtwoptions_demo. The example target
files are:
• usertarget.tlc: the example system target file. This file defines several
popups, checkboxes, an edit field, and a nonUI item. The file demonstrates
the use of callbacks, open callbacks, and close callbacks.
• usertargetcallback.m: an M-file callback invoked by a popup.
• usertargetclosecallback.m: an M-file callback invoked by an edit field.
Please refer to the example files while reading this section. The example
system target file, usertarget.tlc: demonstrates the use of callbacks
associated with the following UI elements:
• The Execution Mode popup executes an open callback that is coded inline
within the system target file. This callback displays a message and sets a
model property via a set_param().
• The Real-Time Interrupt Source popup executes a callback defined in an
external M-file, usertargetcallback.m. A handle to the popup object is
passed in to the callback, which displays the popup’s current value.
• The edit field Signal Logging Buffer Size in Doubles executes a close
callback defined in an external M-file, usertargetclosecallback.m. The
callback obtains a handle to the edit field object and displays the current
value of the edit field.
• The External Mode checkbox executes an open callback that is coded inline
within the system target file. The callback obtains a handle to the checkbox
object and sets its value to 1.
• The NonUi item defined in rtwoptions(8) executes open and close callbacks
that are coded inline within the system target file. Each callback simply
prints a status message.
We suggest that you study the example code while interacting with the
example target options in the Simulation Parameters dialog. To interact with
the example target file:
14-25
14
Targeting Real-Time Systems
1 Make matlabroot/toolbox/rtw/rtwdemos/rtwoptions_demo your working
directory.
2 Open any model of your choice.
3 Open the Real-Time Workshop pane in the Simulation Parameters dialog.
Select Target Configuration from the Category menu.
4 Click the Browse button. The System Target File Browser opens. Select
Real-Time Workshop Options Example Target. Then click OK.
5 Observe that the Category menu of the Real-Time Workshop pane contains
two custom items: userPreferred target options (I) and userPreferred
target options (II).
6 As you interact with the options in these two categories and open and close
the Simulation Parameters dialog, observe the messages displayed in the
MATLAB window. These messages are printed from code in the system
target file, or from callbacks invoked from the system target file.
Additional Code Generation Options
“Target Language Compiler Variables and Options” on page 2-59 describes
additional code generation variables. For readability, it is recommended that
you assign these variables in the Configure RTW code generation settings
section of the system target file.
Alternatively, you can append statements of the form
-aVariable=val
to the System target filename field on the Real-Time Workshop pane.
Build Directory Name
The final part of the system target file defines the BuildDirSuffix field of the
rtwgensettings structure. The build process appends the BuildDirSuffix
string to the model name to form the name of the build directory. For example,
if you define BuildDirSuffix as follows
rtwgensettings.BuildDirSuffix = '_mytarget_rtw'
the build directories are named model_mytarget_rtw.
14-26
Customizing the Build Process
See the Target Language Compiler documentation for further information on
the rtwgensettings structure.
Adding a Custom Target to the System Target
File Browser
As a convenience to end users of your custom target configuration, you can add
a custom target configuration to the System Target File Browser. To do this:
1 Modify (or add) browser comments at the head of your custom system target
file. For example,
%% SYSTLC: John’s Real-Time Target \
%%
TMF: mytarget.tmf MAKE: make_rtw EXTMODE: no_ext_comm
2 Create a directory <targetname> (e.g., /mytarget). Move your custom system
target file, custom template makefile, and run-time interface files (such as
your main program and S-functions) into the <targetname> subdirectory.
Note Your <targetname> subdirectory should not be located anywhere in the
MATLAB directory tree (that is, in or under the matlabroot directory). The
reason for this restriction is that if you install a new version of MATLAB, (or
reinstall your current version) the MATLAB directories will be recreated. This
process deletes any custom target directories existing within the MATLAB
tree.
3 Add your target directory to the MATLAB path.
addpath <targetname>
If you want <targetname> included in the MATLAB path each time
MATLAB starts up, include this addpath command in your startup.m file.
4 When the System Target File Browser opens, Real-Time Workshop detects
system target files that are on the MATLAB path, and displays the target
filenames and target description comments. Figure 14-4 shows how the
target file mytarget.tlc, which contains the browser comments above,
appears in the System Target File Browser.
14-27
14
Targeting Real-Time Systems
Figure 14-4: Custom System Target File Displayed in Browser
Template Makefiles
To configure or customize template makefiles, you should be familiar with how
the make command works and how the make command processes makefiles. You
should also understand makefile build rules. For information of these topics,
please refer to the documentation provided with the make utility you use.
There are also several good books on the make utility.
Template makefiles are made up of statements containing tokens. The
Real-Time Workshop build process expands tokens and creates a makefile,
model.mk. Template makefiles are designed to generate makefiles for specific
compilers on specific platforms. The generated model.mk file is specifically
tailored to compile and link code generated from your model, using commands
specific to your development system.
14-28
Customizing the Build Process
Template
Makefile
system.tmf
Makefile:
model.mk
Figure 14-5: Creation of model.mk
Template Makefile Tokens
The make_rtw M-file command (or a different command provided with some
targets) directs the process of generating model.mk. The make_rtw command
processes the template makefile specified on the Target configuration section
of the Real-Time Workshop pane of the Simulation Parameters dialog.
make_rtw copies the template makefile, line by line, expanding each token
encountered. Table 14-3 lists the tokens and their expansions.
Table 14-3: Template Makefile Tokens Expanded by make_rtw
Token
Expansion
|>COMPUTER<|
Computer type. See the MATLAB
computer command.
|>MAKEFILE_NAME<|
model.mk — The name of the makefile
that was created from the template
makefile.
|>MATLAB_ROOT<|
Path to where MATLAB is installed.
|>MATLAB_BIN<|
Location of the MATLAB executable.
|>MEM_ALLOC<|
Either RT_MALLOC or RT_STATIC.
Indicates how memory is to be allocated.
|>MEXEXT<|
MEX-file extension. See the MATLAB
mexext command.
|>MODEL_NAME<|
Name of the Simulink block diagram
currently being built.
14-29
14
Targeting Real-Time Systems
Table 14-3: Template Makefile Tokens Expanded by make_rtw (Continued)
14-30
Token
Expansion
|>MODEL_MODULES<|
Any additional generated source (.c)
modules. For example, you can split a
large model into two files, model.c and
model1.c. In this case, this token
expands to model1.c.
|>MODEL_MODULES_OBJ<|
Object filenames (.obj) corresponding
to any additional generated source (.c)
modules.
|>MULTITASKING<|
True (1) if solver mode is multitasking,
otherwise False (0).
|>NUMST<|
Number of sample times in the model.
|>RELEASE_VERSION<|
The release version of MATLAB.
|>S_FUNCTIONS<|
List of noninlined S-function (.c)
sources.
|>S_FUNCTIONS_LIB<|
List of S-function libraries available for
linking.
|>S_FUNCTIONS_OBJ<|
Object (.obj) file list corresponding to
noninlined S-function sources.
|>SOLVER<|
Solver source filename, e.g., ode3.c.
|>SOLVER_OBJ<|
Solver object (.obj) filename, e.g.,
ode3.obj.
|>TID01EQ<|
True (1) if sampling rates of the
continuous task and the first discrete
task are equal, otherwise False (0).
|>NCSTATES<|
Number of continuous states.
Customizing the Build Process
Table 14-3: Template Makefile Tokens Expanded by make_rtw (Continued)
Token
Expansion
|>BUILDARGS<|
Options passed to make_rtw. This token
is provided so that the contents of your
model.mk file will change when you
change the build arguments, thus
forcing an update of all modules when
your build options change.
|>EXT_MODE<|
True (1) to enable generation of
external mode support code, otherwise
False (0).
These tokens are expanded by substitution of parameter values known to the
build process. For example, if the source model contains blocks with two
different sample times, the template makefile statement
NUMST = |>NUMST<|
expands to the following in model.mk.
NUMST = 2
In addition to the above, make_rtw expands tokens from other sources:
• Target-specific tokens defined via the Target configuration section of the
Real-Time Workshop pane of the Simulation Parameters dialog box.
• Structures in the RTW Options section of the system target file. Any
structures in the rtwoptions structure array that contain the field
makevariable are expanded.
The following example is extracted from matlabroot/rtw/c/grt/grt.tlc.
The section starting with BEGIN_RTW_OPTIONS contains M-file code that sets
up rtwoptions. The directive
rtwoptions(2).makevariable = 'EXT_MODE'
causes the |>EXT_MODE<| token to be expanded into 1 (on) or 0 (off),
depending on how you set the External mode option in the Code generation
options section of the Real-Time Workshop pane.
14-31
14
Targeting Real-Time Systems
The Make Command
After creating model.mk from your template makefile, Real-Time Workshop
invokes a make command. To invoke make, Real-Time Workshop issues this
command.
makecommand -f model.mk
makecommand is defined by the MAKE macro in your system’s template makefile
(see Figure 14-6 on page 14-35). You can specify additional options to make in
the Make command field of the Real-Time Workshop pane. (see “Make
Command Field” on page 2-6 and “Template Makefiles and Make Options” on
page 2-54.)
For example, specifying OPT_OPTS=-O2 in the Make command field causes
make_rtw to generate the following make command.
makecommand -f model.mk OPT_OPTS=-O2
A comment at the top of the template makefile specifies the available make
command options. If these options do not provide you with enough flexibility,
you can configure your own template makefile.
Make Utilities
The make utility lets you control nearly every aspect of building your real-time
program. There are several different versions of make available. Real-Time
Workshop provides the Free Software Foundation’s GNU Make for both UNIX
and PC platforms in the platform-specific subdirectories below
matlabroot/rtw/bin.
It is possible to use other versions of make with Real-Time Workshop, although
GNU Make is recommended. To ensure compatibility with Real-Time
Workshop, make sure that your version of make supports the following
command format.
makecommand −f model.mk
Structure of the Template Makefile
A template makefile has four sections:
• The first section contains initial comments that describe what this makefile
targets.
14-32
Customizing the Build Process
• The second section defines macros that tell make_rtw how to process the
template makefile. The macros are:
- MAKE — This is the command used to invoke the make utility. For example,
if MAKE = mymake, then the make command invoked is
mymake −f model.mk
- HOST — What platform this template makefile is targeted for. This can be
HOST=PC, UNIX, computer_name (see the MATLAB computer command), or
ANY.
- BUILD — This tells make_rtw whether or not (BUILD=yes or no) it should
invoke make from the Real-Time Workshop build procedure.
- SYS_TARGET_FILE — Name of the system target file. This is used for
consistency checking by make_rtw to verify that the correct system target
file was specified in the Target configuration section of the Real-Time
Workshop pane of the Simulation Parameters dialog box.
- BUILD_SUCCESS — An optional macro that specifies the build success string
to be displayed on successful make completion on the PC. For example,
BUILD_SUCCESS = ### Successful creation of
The BUILD_SUCCESS macro, if used, replaces the standard build success
string found in the template makefiles distributed with the bundled
Real-Time Workshop targets (such as GRT):
@echo ### Created executable $(MODEL).exe
Your template makefile must include either the standard build success
string, or use the BUILD_SUCCESS macro. For an example of the use of
BUILD_SUCCESS, see
matlabroot/toolbox/rtw/c/grt/grt_bc.tmf
- BUILD_ERROR — An optional macro that specifies the build error message
to be displayed when an error is encountered during the make procedure.
For example,
BUILD_ERROR = ['Error while building ', modelName]
The following DOWNLOAD options apply only to the Tornado target:
- DOWNLOAD — An optional macro that you can specify as yes or no. If
specified as yes (and BUILD=yes), then make is invoked a second time with
the download target.
make -f model.mk download
14-33
14
Targeting Real-Time Systems
- DOWNLOAD_SUCCESS — An optional macro that you can use to specify the
download success string to be used when looking for a successful
download. For example,
DOWNLOAD_SUCCESS = ### Downloaded
- DOWNLOAD_ERROR — An optional macro that you can use to specify the
download error message to be displayed when an error is encountered
during the download. For example,
DOWNLOAD_ERROR = ['Error while downloading ', modelName]
• The third section defines the tokens make_rtw expands (see Table 14-3).
• The fourth section contains the make rules used in building an executable
from the generated source code. The build rules are typically specific to your
version of make.
14-34
Customizing the Build Process
Figure 14-6 shows the general structure of a template makefile.
#-- Section 1: Comments ------------------------------------------------------#
# Description of target type and version of make for which
# this template makefile is intended.
# Also documents any optional build arguments.
#-- Section 2: Macros read by make_rtw ---------------------------------------#
# The following macros are read by the Real-Time Workshop build procedure:
#
# MAKE
- This is the command used to invoke the make utility.
# HOST
- Platform this template makefile is designed
#
(i.e., PC or UNIX)
# BUILD
- Invoke make from the Real-Time Workshop build procedure
#
(yes/no)?
# SYS_TARGET_FILE - Name of system target file.
MAKE
= make
HOST
= UNIX
BUILD
= yes
SYS_TARGET_FILE = system.tlc
#-- Section 3: Tokens expanded by make_rtw -----------------------------------#
MODEL
MODULES
MAKEFILE
MATLAB_ROOT
...
COMPUTER
BUILDARGS
=
=
=
=
Comments
make_rtw
macros
make_rtw
tokens
|>MODEL_NAME<|
|>MODEL_MODULES<|
|>MAKEFILE_NAME<|
|>MATLAB_ROOT<|
= |>COMPUTER<|
= |>BUILDARGS<|
Build rules
#-- Section 4: Build rules ---------------------------------------------------#
# The build rules are specific to your target and version of make.
Figure 14-6: Structure of aTemplate Makefile
Customizing and Creating Template Makefiles
To customize or create a new template makefile, we recommend that you copy
an existing template makefile to your local working directory and modify it.
This section shows, through an example, how to use macros and
file-pattern-matching expressions in a template makefile to generate
commands in the model.mk file.
14-35
14
Targeting Real-Time Systems
The make utility processes the model.mk makefile and generates a set of
commands based upon dependency rules defined in model.mk. After make
generates the set of commands needed to build or rebuild test, make executes
them.
For example, to build a program called test, make must link the object files.
However, if the object files don’t exist or are out of date, make must compile the
C code. Thus there is a dependency between source and object files.
Each version of make differs slightly in its features and how rules are defined.
For example, consider a program called test that gets created from two
sources, file1.c and file2.c. Using most versions of make, the dependency
rules would be
test: file1.o file2.o
cc −o test file1.o file2.o
file1.o: file1.c
cc −c file1.c
file2.o: file2.c
cc −c file2.c
In this example, we assumed a UNIX environment. In a PC environment the
file extensions and compile and link commands will be different.
In processing the first rule
test: file1.o file2.o
make sees that to build test, it needs to build file1.o and file2.o. To build
file1.o, make processes the rule
file1.o: file1.c
If file1.o doesn’t exist, or if file1.o is older than file1.c, make compiles
file1.c.
The format of Real-Time Workshop template makefiles follows the above
example. Our template makefiles use additional features of make such as
macros and file-pattern-matching expressions. In most versions of make, a
macro is defined via
MACRO_NAME = value
14-36
Customizing the Build Process
References to macros are made via $(MACRO_NAME). When make sees this form
of expression, it substitutes value for $(MACRO_NAME).
You can use pattern matching expressions to make the dependency rules more
general. For example, using GNU Make you could replace the two "file1.o:
file1.c" and "file2.o: file2.c" rules with the single rule
%.o : %.c
cc −c $<
Note that $< above is a special macro that equates to the dependency file (i.e.,
file1.c or file2.c). Thus, using macros and the “%” pattern matching
character, the above example can be reduced to
SRCS = file1.c file2.c
OBJS = $(SRCS:.c=.o)
test: $(OBJS)
cc −o [email protected] $(OBJS)
%.o : %.c
cc −c $<
Note that the [email protected] macro above is another special macro that equates to the name
of the current dependency target, in this case test.
This example generates the list of objects (OBJS) from the list of sources (SRCS)
by using the string substitution feature for macro expansion. It replaces the
source file extension (.c) with the object file extension (.o). This example also
generalized the build rule for the program, test, to use the special “[email protected]” macro.
Customizing the Makefile Include Path
Real-Time Workshop template makefiles provide rules and macros that let you
add source directories, include directories, and libraries to generated makefiles
without having to modify the template makefiles themselves. This feature is
useful if you need to include your code when building S-functions.
To include a directory needed for a S-Function, you must create an M-function,
rtwmakecfg, in a file rtwmakecfg.m. This file must reside in the same directory
as your S-function component (.dll on Windows, .mex on UNIX). The
rtwmakecfg function is called during the build process. The rtwmakecfg
function must return a structured array with following elements:
14-37
14
Targeting Real-Time Systems
• makeInfo.includePath: a cell array containing additional include directory
names, which must be organized as row vector. These directory names will
be expanded into include instructions in the generated makefile.
• makeInfo.sourcePath: a cell array containing additional source directory
names, which must be organized as a row vector. These directory names will
be expanded into make rules in the generated makefile.
• makeInfo.library: a structure containing additional runtime library names
and module objects, which must be organized as a row vector. This
information will be expanded into make rules in the generated makefile.
- makeInfo.library(n).Name: String. Specifies the name of the library
(without extension).
- makeInfo.library(n).Location: String. Directory in which the library is
located.
- makeInfo.library(n).Modules: Cell array. Specifies the C files in the
library.
14-38
Creating Device Drivers
Creating Device Drivers
Device drivers that communicate with target hardware are essential to many
real-time development projects. This section describes how to integrate device
drivers into your target system. This includes incorporating drivers into your
Simulink model and into the code generated from that model.
Device drivers are implemented as Simulink device driver blocks. A device
driver block is an S-Function block that is bound to user-written driver code.
To implement device drivers, you should be familiar with the Simulink C MEX
S-function format and API. The following documents contain more information
about C MEX S-functions:
• Writing S-Functions describes S-functions, including how to write both
inlined and noninlined S-functions and how to access parameters from a
masked S-function. Writing S-Functions also describes how to use the special
mdlRTW function to parameterize an inlined S-function.
• “External Interfaces/API” in the MATLAB online documentation explains
how to write C and other programs that interact with MATLAB via the MEX
API. The Simulink S-function API is built on top of this API. To pass
parameters to your device driver block from MATLAB/Simulink you must
use the MEX API. “External Interfaces/API Reference” in the MATLAB
online documentation contains reference descriptions for the required
MATLAB mx* routines.
• The Target Language Compiler documentation describes the Target
Language Compiler. Knowledge of the Target Language Compiler is
required in order to inline S-functions. The Target Language Compiler
Reference Guide also describes the structure of the model.rtw file.
• “Using Masks to Customize Blocks” in Using Simulink describes how to
create a mask for an S-function.
Note Device driver blocks must be implemented as C MEX S-functions, not
as M-file S-functions. C MEX S-functions are limited to a subset of the
features available in M-file S-functions. See “Limitations of Device Driver
Blocks” on page 14-42 for information.
14-39
14
Targeting Real-Time Systems
This section covers the following topics:
• Inlined and noninlined device drivers
• General requirements and limitations for device drivers
• Obtaining S-function parameter values from a dialog box
• Writing noninlined device drivers
• Writing inlined device drivers
• Building the device driver MEX-file
Inlined and Noninlined Drivers
In your target system, a device driver has a dual function. First, it functions as
a code module that you compile and link with other code generated from your
model by Real-Time Workshop. In addition, the driver must interact with
Simulink during simulation. To meet both these requirements, you must
incorporate your driver code into a Simulink device driver block.
You can build your driver S-function in several ways:
• As a MEX-file component, bound to an S-Function block, for use in a
Simulink model. In this case, the Simulink engine calls driver routines in the
MEX-file during execution of the model.
• As a module within a stand-alone real-time program that is generated from
a model by Real-Time Workshop. The driver routines are called from within
the application in essentially the same way that Simulink calls them.
In many cases, the code generated from driver blocks for real-time execution
must run differently from the code used by the blocks in simulation. For
example, an output driver may write to hard device addresses in real time;
but these write operations could cause errors in simulation.
Real-Time Workshop provides standard compilation conditionals and
include files to let you build the drivers for both cases. (See “Conditional
Compilation for Simulink and Real-Time” on page 14-45.)
• As inlined code. The Target Language Compiler enables you to generate the
explicit code from your routines (instead of calls to these routines) in the body
of the application. Inlined code eliminates calling overhead, and reduces
memory usage.
14-40
Creating Device Drivers
Inlining an S-function can improve its performance significantly. However,
there is a tradeoff in increased development and maintenance effort. To inline
a device driver block, you must implement the block twice: first, as a C
MEX-file, and second, as a TLC program.
The C MEX-file version is for use in simulation. Since a simulation normally
does not have access to I/O boards or other target hardware, the C MEX-file
version often acts as a “dummy” block within a model. For example, a
digital-to-analog converter (DAC) device driver block is often implemented as
a stub for simulation.
Alternatively, the C MEX-file version can simulate the behavior of the
hardware. For example, an analog-to-digital converter (ADC) device driver
block might read sample values from a data file or from the MATLAB
workspace.
The TLC version generates actual working code that accesses the target
hardware in a production system.
Inlined device drivers are an appropriate design choice when:
• You are using the Real-Time Workshop Embedded Coder target. Inlined
S-functions are required when building code from the Real-Time Workshop
Embedded Coder target. S-functions for other targets can be either inlined
or noninlined.
• You need production code generated from the S-function to behave
differently than code used during simulation. For example, an output device
block may write to an actual hardware address in generated code, but
perform no output during simulation.
• You want to avoid overhead associated with calling the S-function API.
• You want to reduce memory usage. Note that each noninlined S-function
creates its own Simstruct. Each Simstruct uses over 1K of memory. Inlined
S-functions do not allocate any Simstruct. For optimal memory usage,
consider using inlined S-functions with the Real-Time Workshop Embedded
Coder target.
• You want to avoid making calls to routines that are required by Simulink,
but which are empty, in your generated code.
14-41
14
Targeting Real-Time Systems
Device Driver Requirements and Limitations
In order to create a device driver block, the following components are required:
• Hardware-specific driver code, which handles communication between a
real-time program and an I/O device. See your I/O device documentation for
information on hardware requirements.
• S-function code, which implements the model initialization, output, and
other functions required by the S-function API. The S-function code calls
your driver code.
Your S-function code and the hardware-specific driver code are compiled and
linked into a component that is bound to an S-Function block in your
Simulink model. The MATLAB mex utility builds this component (a DLL
under Windows, or a shared library under UNIX).
We recommend that you use the S-function template provided by Real-Time
Workshop as a starting point for developing your driver S-functions. The
template file is
matlabroot/simulink/src/sfuntmpl_basic.c
An extensively commented version of the S-function template is also available.
See matlabroot/simulink/src/sfuntmpl_doc.c.
The following components are optional:
• A TLC file that generates inline code for the S-function.
• A mask for the device driver block to create a customized user interface.
Limitations of Device Driver Blocks
The following limitations apply to noninlined driver blocks:
• Only a subset of MATLAB API functions are supported. See the “Noninlined
S-functions” section of Writing S-Functions for a complete list of supported
calls.
• Parameters must be doubles or characters contained in scalars, vectors, or
2-D matrices.
The following applies to inlined driver blocks:
14-42
Creating Device Drivers
• If the driver does not have a mdlRTW function, parameter restrictions are the
same as for noninlined drivers.
• If the driver has a mdlRTW function, any parameter type is supported.
Preemption
Consider preemption issues in the design of your drivers. In a typical real-time
program, a timer interrupt invokes rtOneStep, which in turn calls MdlOutputs,
which in turn calls your input (ADC) and /or output (DAC) drivers. In this
situation, your drivers are interruptible.
Parameterizing Your Driver
You can add a custom icon, dialog box, and initialization commands to an
S-Function block by masking it. This provides an easy-to-use graphical user
interface for your device driver in the Simulink environment.
You can parameterize your driver by letting the user enter hardware-related
variables. Figure 14-7 shows the dialog box of a masked device driver block for
an input (ADC) device. The Simulink user can enter the device address, the
number of channels, and other operational parameters.
Figure 14-7: Dialog Box for a Masked ADC Driver Block
14-43
14
Targeting Real-Time Systems
A masked S-Function block obtains parameter data from its dialog box using
macros and functions provided for the purpose.
To obtain a parameter value from the dialog:
1 Access the parameter from the dialog box using the ssGetSFcnParam macro.
The arguments to ssGetSFcnParam are a pointer to the block’s Simstruct,
and the index (0-based) to the desired parameter. For example, use the
following call to access the Number of Channels parameter from the dialog
above.
ssGetSFcnParam(S,3); /* S points to block’s Simstruct */
2 Parameters are stored in arrays of type mxArray, even if there is only a
single value. Get a particular value from the input mxArray using the
mxGetPr function. The following code fragment extracts the first (and only)
element in the Number of Channels parameter.
#define NUM_CHANNELS_PARAM
(ssGetSFcnParam(S,3))
#define NUM_CHANNELS ((uint_T) mxGetPr(NUM_CHANNELS_PARAM)[0])
uint_T num_channels;
num_channels = NUM_CHANNELS;
It is typical for a device driver block to read and validate input parameters in
its mdlInitializeSizes function. See the listing “adc.c” on page 14-60 for an
example.
By default, S-function parameters are tunable. To make a parameter
nontunable, use the ssSetSFcParamNotTunable macro in the
mdlInitializeSizes routine. Nontunable S-function parameters become
constants in the generated code, improving performance.
For further information on creation and use of masked blocks, see the Using
Simulink and Writing S-Functions manuals.
Writing a Noninlined S-Function Device Driver
Device driver S-functions are relatively simple to implement because they
perform only a few operations. These operations include:
• Initializing the SimStruct.
• Initializing the I/O device.
14-44
Creating Device Drivers
• Calculating the block outputs. How this is done depends upon the type of
driver being implemented:
- An input driver for a device such as an ADC reads values from an I/O
device and assigns these values to the block’s output vector y.
- An output driver for a device such as a DAC writes values from the block’s
input vector u to an I/O device.
• Terminating the program. This may require setting hardware to a “neutral”
state; for example, zeroing DAC outputs.
Your driver performs these operations by implementing certain specific
functions required by the S-function API.
Since these functions are private to the source file, you can incorporate
multiple instances of the same S-function into a model. Note that each such
noninlined S-function also instantiates a SimStruct.
Conditional Compilation for Simulink and Real-Time
Noninlined S-functions must function in both Simulink and in real-time
environments. Real-Time Workshop defines the preprocessor symbols
MATLAB_MEX_FILE, RT, and NRT to distinguish simulation code from real-time
code. Use these symbols as follows:
• MATLAB_MEX_FILE
Conditionally include code that is intended only for use in simulation under
this symbol. When you build your S-function as a MEX-file via the mex
command, MATLAB_MEX_FILE is automatically defined.
• RT
Conditionally include code that is intended to run only in a real-time
program under this symbol. When you generate code via the Real-Time
Workshop build command, RT is automatically defined.
• NRT
Conditionally include code that is intended only for use with a variable-step
solver, in a non-real-time standalone simulation or in a MEX-file for use with
Simulink, under this symbol.
Real-Time Workshop provides these conditionals to help ensure that your
driver S-functions access hardware only when it is appropriate to do so. Since
your target I/O hardware is not present during simulation, writing to
14-45
14
Targeting Real-Time Systems
addresses in the target environment can result in illegal memory references,
overwriting system memory, and other severe errors. Similarly, read
operations from nonexistent hardware registers can cause model execution
errors.
In the following code fragment, a hardware initialization call is compiled in
generated real-time code. During simulation, a message is printed to the
MATLAB command window.
#if defined(RT)
/* generated code calls function to initialize an A/D device */
INIT_AD();
#elif defined(MATLAB_MEX_FILE)
/* during simulation, just print a message */
if (ssGetSimMode(S) == SS_SIMMODE_NORMAL) {
mexPrintf("\n adc.c: Simulating initialization\n”);
}
#endif
The MATLAB_MEX_FILE and RT conditionals also control the use of certain
required include files. See “Required Defines and Include Files” below.
You may prefer to control execution of real-time and simulation code by some
other means. For an example, see the use of the variable ACCESS_HW in
matlabroot/rtw/c/dos/devices/das16ad.c
Required Defines and Include Files
Your driver S-function must begin with the following three statements, in the
following order:
1 #define S_FUNCTION_NAME name
This defines the name of the entry point for the S-function code. name must
be the name of the S-function source file, without the .c extension. For
example, if the S-function source file is das16ad.c:
#define S_FUNCTION_NAME das16ad
14-46
Creating Device Drivers
2 #define S_FUNCTION_LEVEL 2
This statement defines the file as a level 2 S-function. This allows you to
take advantage of the full feature set included with S-functions. Level-1
S-functions are currently used only to maintain backwards compatibility.
3 #include “simstruc.h”
The file simstruc.h defines the SimStruct (the Simulink data structure)
and associated accessor macros. It also defines access methods for the mx*
functions from the MATLAB MEX API.
Depending upon whether you intend to build your S-function as a MEX file or
as real-time code, you must include one of the following files at the end of your
S-function:
• simulink.c provides required functions interfacing to Simulink.
• cg_sfun.h provides the required S-function entry point for generated code.
A noninlined S-function should conditionally include both these files, as in the
following code from sfuntmpl_basic.c:
#ifdef MATLAB_MEX_FILE
/* File being compiled as a MEX-file? */
#include "simulink.c"
/* MEX-file interface mechanism */
#else
#include "cg_sfun.h" /* Code generation registration function */
#endif
Required Functions
The S-function API requires you to implement several functions in your driver:
• mdlInitializeSizes specifies the sizes of various parameters in the
SimStruct, such as the number of output ports for the block.
• mdlInitializeSampleTimes specifies the sample time(s) of the block.
If your device driver block is masked, your initialization functions can obtain
the sample time and other parameters entered by the user in the block’s
dialog box.
• mdlOutputs: for an input device, reads values from the hardware and sets
these values in the output vector y. For an output device, reads the input u
from the upstream block and outputs the value(s) to the hardware.
14-47
14
Targeting Real-Time Systems
• mdlTerminate resets hardware devices to a desired state, if any. This
function may be implemented as a stub.
In addition to the above, you may want to implement the mdlStart function.
mdlStart, which is called once at the start of model execution, is useful for
operations such as setting I/O hardware to some desired initial state.
This following sections provide guidelines for implementing these functions.
mdlInitializeSizes
In this function you specify the sizes of various parameters in the SimStruct.
This information may depend upon the parameters passed to the S-function.
“Parameterizing Your Driver” on page 14-43 describes how to access parameter
values specified in S-function dialog boxes.
Initializing Sizes - Input Devices. The mdlInitializeSizes function sets size
information in the SimStruct. The following implementation of
mdlInitializeSizes initializes a typical ADC driver block.
static void mdlInitializeSizes(SimStruct *S)
{
uint_T num_channels;
ssSetNumSFcnParams(S, 3); /* Number of expected parameters */
if (ssGetNumSFcnParams(S) != ssGetSFcnParamsCount(S)){
/*Return if number of expected != number of actual params */
return;
}
num_channels = mxGetPr(NUM_CHANNELS_PARAM)[0];
ssSetNumInputPorts(S, 0);
ssSetNumOutputPorts(S, num_channels);
ssSetNumSampleTimes(S,1);
}
This routine first validates that the number of input parameters is equal to the
number of parameters in the block’s dialog box. Next, it obtains the Number of
Channels parameter from the dialog.
ssSetNumInputPorts sets the number of input ports to 0 because an ADC is a
source block, having only outputs.
14-48
Creating Device Drivers
ssSetNumOutputPorts sets the number of output ports equal to the number of
I/O channels obtained from the dialog box.
ssSetNumSampleTimes sets the number of sample times to 1. This would be the
case where all ADC channels run at the same rate. Note that the actual sample
period is set in mdlInitializeSampleTimes.
Note that by default, the ADC block has no direct feedthrough. The ADC output
is calculated based on values read from hardware, not from data obtained from
another block.
Initializing Sizes - Output Devices. Initializing size information for an output
device, such as a DAC, differs in several important ways from initializing sizes
for an ADC:
• Since the DAC obtains its inputs from other blocks, the number of channels
is equal to the number of inputs.
• The DAC is a sink block. That is, it has input ports but no output ports. Its
output is written to a hardware device.
• The DAC block has direct feedthrough. The DAC block cannot execute until
the block feeding it updates its outputs.
The following example is an implementation of mdlInitializeSizes for a DAC
driver block.
static void mdlInitializeSizes(SimStruct *S)
{
uint_T num_channels;
ssSetNumSFcnParams(S, 3); /* Number of expected parameters */
if (ssGetNumSFcnParams(S) != ssGetSFcnParamsCount(S)){
/* Return if number of expected != number of actual params */
return;
}
num_channels = mxGetPr(NUM_CHANNELS_PARAM)[0];
ssSetNumInputPorts(S, num_channels);
/* Number of inputs is now the number of channels. */
ssSetNumOutputPorts(S, 0);
/* Set direct feedthrough for all ports */
{
uint_T i;
14-49
14
Targeting Real-Time Systems
for(i=0, i < num_channels, i++) {
ssSetInputPortDirectFeedThrough(S,i,1);
}
}
ssSetNumSampleTimes(S, 1);
}
mdlInitializeSampleTimes
Device driver blocks are discrete blocks, requiring you to set a sample time. The
procedure for setting sample times is the same for both input and output device
drivers. Assuming that all channels of the device run at the same rate, the
S-function has only one sample time.
The following implementation of mdlInitializeSampleTimes reads the sample
time from a block’s dialog box. In this case, sample time is the fifth parameter
in the dialog box. The sample time offset is set to 0.
static void mdlInitializeSampleTimes(SimStruct *S)
{
ssSetSampleTime(S, 0, mxGetPr(ssGetSFcnParams(S,4))[0]);
ssSetOffsetTime(S, 0, 0.0);
}
mdlStart
mdlStart is an optional function. It is called once at the start of model
execution, and is often used to initialize hardware. Since it accesses hardware,
you should compile it conditionally for use in real-time code or simulation, as
in this example:
static void mdlStart(SimStruct *S)
{
#if defined(RT)
/* Generated code calls function to initialize an A/D device */
INIT_AD(); /* This call accesses hardware */
#elif defined(MATLAB_MEX_FILE)
/* During simulation, just print a message */
if (ssGetSimMode(S) == SS_SIMMODE_NORMAL) {
mexPrintf("\n adc.c: Simulating initialization\n");
}
#endif
}
14-50
Creating Device Drivers
mdlOutputs
The basic purpose of a device driver block is to allow your program to
communicate with I/O hardware. Typically, you accomplish this by using low
level hardware calls that are part of your compiler’s C library, or by using
C-callable functions provided with your I/O hardware.
All S-functions implement a mdlOutputs function to calculate block outputs.
For a device driver block, mdlOutputs contains the code that reads from or
writes to the hardware.
mdlOutputs - Input Devices. In a driver for an input device (such as an ADC),
mdlOutputs must:
• Initiate a conversion for each channel.
• Read the board’s ADC output for each channel (and perhaps apply scaling to
the values read).
• Set these values in the output vector y for use by the model.
The following code is the mdlOutputs function from the ADC driver
matlabroot/rtw/c/dos/devices/das16ad.c. The function uses macros
defined in matlabroot/rtw/c/dos/devices/das16ad.h to perform low-level
hardware access. Note that the Boolean variable ACCESS_HW (rather than
conditional compilation) controls execution of simulation and real-time code.
The real-time code reads values from the hardware and stores them to the
output vector. The simulation code simply outputs 0 on all channels.
static void mdlOutputs(SimStruct *S, int_T tid)
{
real_T *y = ssGetOutputPortRealSignal(S,0);
uint_T i;
if (ACCESS_HW) {
/* Real-time code reads hardware*/
ADCInfo
*adcInfo
= ssGetUserData(S);
uint_T
baseAddr
= adcInfo->baseAddr;
real_T
offset
= adcInfo->offset;
real_T
resolution = adcInfo->resolution;
/* For each ADC channel initiate conversion,*/
/* then read channel value, scale and offset it and store */
/* it to output y */
for (i = 0; i < NUM_CHANNELS; i++) {
uint_T adcValue;
14-51
14
Targeting Real-Time Systems
adcStartConversion(baseAddr);
for ( ; ; ){
if (!adcIsBusy(baseAddr)) break;
}
adcValue = adcGetValue(baseAddr);
y[i] = offset + resolution*adcValue;
}
}
else {
/* simulation code just zeroes the output for all channels*/
for (i = 0; i < NUM_CHANNELS; i++){
y[i] = 0.0;
}
}
}
mdlOutputs - Output Devices. In a driver for an output device (such as a DAC),
mdlOutputs must:
• Read the input u from the upstream block.
• Set the board’s DAC output for each channel (and apply scaling to the input
values if necessary).
• Initiate a conversion for each channel.
The following code is the mdlOutputs function from the DAC driver
matlabroot/rtw/c/dos/devices/das16da.c. The function uses macros
defined in matlabroot/rtw/c/dos/devices/das16ad.h to perform low-level
hardware access. This function iterates over all channels, obtaining and
scaling a block input value. It then range-checks and (if necessary) trims each
value. Finally it writes the value to the hardware.
In simulation, this function is a stub.
static void mdlOutputs(SimStruct *S, int_T tid)
{
if (ACCESS_HW) {
int_T
i;
DACInfo
*dacInfo
= ssGetUserData(S);
uint_T
baseAddr
= dacInfo->baseAddr;
real_T
resolution = dacInfo->resolution;
InputRealPtrsType uPtrs = ssGetInputPortRealSignalPtrs(S,0);
14-52
Creating Device Drivers
for (i = 0; i < NUM_CHANNELS; i++) {
uint_T codeValue;
/* Get and scale input for channel i. */
real_T value = (*uPtrs[i] - MIN_OUTPUT)*resolution;
/* Range check value */
value = (value < DAC_MIN_OUTPUT) ? DAC_MIN_OUTPUT : value;
value = (value > DAC_MAX_OUTPUT) ? DAC_MAX_OUTPUT : value;
codeValue = (uint_T) value;
/* Output to hardware */
switch (i) {
case 0:
dac0SetValue(baseAddr, codeValue);
break;
case 1:
dac1SetValue(baseAddr, codeValue);
break; }
}
}
}
mdlTerminate
This final required function is typically needed only in DAC drivers. The
following routine sets the output of each DAC channel to zero:
static void mdlTerminate(SimStruct *S)
{
uint_T num_channels;
uint_T i;
num_channels = (uint_t)mxGetPr(ssGetSFcnParams(S,0)[0]);
for (i = 0; i < num_channels; i++){
ds1102_da(i + 1, 0.0); /* Hardware-specific DAC output */
}
}
ADC drivers usually implement mdlTerminate as an empty stub.
Writing an Inlined S-Function Device Driver
To inline a device driver, you must provide:
14-53
14
Targeting Real-Time Systems
• driver.c: C MEX S-function source code, implementing the functions
required by the S-function API. These are the same functions required for
noninlined drivers, as described in “Required Functions” on page 14-47. For
these functions, only the code for simulation in Simulink simulation is
required.
It is important to ensure that driver.c does not attempt to read or write
memory locations that are intended to be used in the target hardware
environment. The real-time driver implementation, generated via a
driver.tlc file, should access the target hardware.
• Any hardware support files such as header files, macro definitions, or code
libraries that are provided with your I/O devices.
• Optionally, a mdlRTW function within driver.c. The sole purpose of this
function is to evaluate and format parameter data during code generation.
The parameter data is output to the model.rtw file. If your driver block does
not need to pass information to the code generation process, you do not need
to write a mdlRTW function. See “mdlRTW and Code Generation” on
page 14-57 .
• driver.dll (PC) or driver (UNIX): MEX-file built from your C MEX
S-function source code. This component is used:
- In simulation: Simulink calls the simulation versions of the required
functions
- During code generation: if a mdlRTW function exists in the MEX-file, the
code generator executes it to write parameter data to the model.rtw file.
• driver.tlc: TLC functions that generate real-time implementations of the
functions required by the S-function API.
Example: An Inlined ADC Driver
As an aid to understanding the process of inlining a device driver, this section
describes an example driver block for an ADC device. “Source Code for Inlined
ADC Driver” on page 14-60 lists code for:
• adc.c, the C MEX S-function
• adc.tlc, the corresponding TLC file
• device.h, a hardware-specific header file included in both the simulation
and real-time generated code
14-54
Creating Device Drivers
The driver S-Function block is masked and has an icon. Figure 14-8 shows a
model using the driver S-Function block. Figure 14-9 shows the block’s dialog
box.
Figure 14-8: ADC S-function Driver Block in a Model
The dialog box lets the user enter:
• The ADC base address
• An array defining its signal range
• Its gain factor
• The block’s sample time
14-55
14
Targeting Real-Time Systems
Figure 14-9: ADC Driver Dialog Box
Simulation Code. adc.c consists almost entirely of functions to be executed
during simulation. (The sole exception is mdlRTW, which executes during code
generation.) Most of these functions are similar to the examples of
non-real-time code given in “Writing a Noninlined S-Function Device Driver”
on page 14-44. The S-function implements the following functions:
• mdlInitializeSizes validates input parameters (via mdlCheckParameters)
and declares all parameters nontunable. This function also initializes ports
and sets the number of sample times.
• mdlInitializeSampleTimes sets the sample time using the user-entered
value.
• mdlStart prints a message to the MATLAB command window.
• mdlOutputs outputs zero on all channels.
• mdlTerminate is a stub routine.
Since adc.c contains only simulation code, it uses a single test of
MATLAB_MEX_FILE to ensure that it is compiled as a C MEX-file.
#ifndef MATLAB_MEX_FILE
#error "Fatal Error: adc.c can only be used to create C-MEX
S-Function"
#endif
14-56
Creating Device Drivers
For the same reason, adc.c unconditionally includes simulink.c.
mdlRTW and Code Generation. mdlRTW is a mechanism by which an S-function can
generate and write data structures to the model.rtw file. The Target Language
Compiler, in turn, uses these data structures when generating code. Unlike the
other functions in the driver, mdlRTW executes at code generation time.
In this example, mdlRTW calls the ssWriteRTWParamSettings function to
generate a structure that contains both user-entered parameter values (base
address, hardware gain) and values computed from user-entered values
(resolution, offset).
static void mdlRTW(SimStruct *S)
{
boolean_T polarity
= adcIsUnipolar(MIN_SIGNAL_VALUE, MAX_SIGNAL_VALUE);
real_T
offset
= polarity ? 0.0 : MIN_SIGNAL_VALUE/HARDWARE_GAIN;
real_T
resolution = (((MAX_SIGNAL_VALUE-MIN_SIGNAL_VALUE)/HARDWARE_GAIN)/
ADC_NUM_LEVELS);
char_T
baseAddrStr[128];
if ( mxGetString(BASE_ADDRESS_PARAM, baseAddrStr, 128) ) {
ssSetErrorStatus(S, "Error reading Base Address parameter, "
"need to increase string buffer size.");
return;
}
if ( !ssWriteRTWParamSettings(S, 4,
SSWRITE_VALUE_QSTR,
SSWRITE_VALUE_NUM,
SSWRITE_VALUE_NUM,
SSWRITE_VALUE_NUM,
"BaseAddress",
"HardwareGain",
"Resolution",
"Offset",
baseAddrStr,
HARDWARE_GAIN,
resolution,
offset) ) {
return; /* An error occured, which will be reported by Simulink. */
}
} /* end: mdlRTW */
14-57
14
Targeting Real-Time Systems
The structure defined in model.rtw is
SFcnParamSettings {
BaseAddress "0x300"
HardwareGain 1.0
Resolution 0.0048828125
Offset -10.0
}
(The actual values of SFcnParamSettings derive from data entered by the
user.)
Values stored in the SFcnParamSettings structure are referenced in
driver.tlc, as in the following assignment statement.
%assign baseAddr = SFcnParamSettings.BaseAddress
The Target Language Compiler uses variables such as baseAddr to generate
parameters in real-time code files such as model.c and model.h. This is
discussed in the next section.
Note During code generation, RTW writes all runtime parameters
automatically to the model.rtw file, eliminating the need for the device driver
S-function to perform this task via a mdlRTW method. See the discussion of
runtime parameters in Writing S-Functions for further information.
The TLC File
adc.tlc contains three TLC functions. The BlockTypeSetup function
generates the statement
#include "device.h"
in the model.h file. The other two functions, Start and Outputs, generate code
within the MdlStart and MdlOutputs functions of model.c.
Statements in adc.tlc, and in the generated code, employ macros and symbols
defined in device.h, and parameter values in the SFcnParamSettings
structure. The following code uses the values from the SFcnParamSettings
structure above to generate code containing constant values:
%assign baseAddr = SFcnParamSettings.BaseAddress
14-58
Creating Device Drivers
%assign hwGain = SFcnParamSettings.HardwareGain
...
adcSetHardwareGain(%<baseAddr>, adcGetGainMask(%<hwGain>));
The TLC code above generates this statement in the MdlOutputs function of
model.c.
adcSetHardwareGain(0x300, adcGetGainMask(1.0));
adcSetHardwareGain and adcGetGainMask are macros that expand to low-level
hardware calls.
S-Function Wrappers
Another technique for integrating driver code into your target system is to use
S-function wrappers. In this approach, you write:
• An S-function (the wrapper) that calls your driver code as an external
module
• A TLC file that generates a call to the same driver code that was called in the
wrapper
See Writing S-Functions for a full description of how to use wrapper
S-functions.
Building the MEX-File and the Driver Block
This section outlines how to build a MEX-file from your driver source code for
use in Simulink. For full details on how to use mex to compile the device driver
S-function into an executable MEX-file, see “External Interfaces/API” in the
MATLAB online documentation. For details on masking the device driver
block, see “Using Masks to Customize Blocks” in Using Simulink.
1 Your C S-function source code should be in your working directory. To build
a MEX-file from mydriver.c type
mex mydriver.c
mex builds mydriver.dll (PC) or mydriver (UNIX).
2 Add an S-Function block (from the Simulink Functions & Tables library in
the Library Browser) to your model.
14-59
14
Targeting Real-Time Systems
3 Double-click the S-Function block to open the Block Parameters dialog.
Enter the S-function name mydriver. The block is now bound to the
mydriver MEX-file.
4 Create a mask for the block if you want to use a custom icon or dialog.
Source Code for Inlined ADC Driver
These files are described in “Example: An Inlined ADC Driver” on page 14-54.
adc.c
/*
* File
: adc.c
* Abstract:
* Example S-function device driver (analog to digital convertor) for use
* with Simulink and Real-Time Workshop.
* This S-function contains simulation code only (except mdlRTW, used
* only during code generation.) An error will be generated if
* this code is compiled without MATLAB_MEX_FILE defined. That
* is,it must be compiled via the MATLAB mex utility.
*
* DEPENDENCIES:
* (1) This S-function is intended for use in conjunction with adc.tlc,
* a Target Language Compiler program that generates inlined, real-time code that
* implements the real-time I/O functions required by mdlOutputs, etc.
*
* (2) device.h defines hardware-specific macros, etc. that implement
* actual I/O to the board
*
* (3) This file contains a mdlRTW function that writes parameters to
* the model.rtw file during code generation.
*
* Copyright (c) 1994-2000 by The MathWorks, Inc. All Rights Reserved.
*
*/
/*********************
* Required defines *
*********************/
#define S_FUNCTION_NAME adc
#define S_FUNCTION_LEVEL 2
/*********************
* Required includes *
*********************/
#include "simstruc.h"
14-60
/* The Simstruct API, definitions and macros */
Creating Device Drivers
/*
* Generate a fatal error if this file is (by mistake) used by Real-Time
* Workshop. There is a target file corresponding to this S-function: adc.tlc,
* which should be used to generate inlined code for this S-funciton.
*/
#ifndef MATLAB_MEX_FILE
# error "Fatal Error: adc.c can only be used to create C-MEX S-Function"
#endif
/*
* Define the number of S-function parameters and set up convenient macros to
* access the parameter values.
*/
#define NUM_S_FUNCTION_PARAMS
(4)
#define N_CHANNELS (2) /* For this example, num. of channels is fixed */
/* 1. Base Address */
#define BASE_ADDRESS_PARAM
(ssGetSFcnParam(S,0))
/* 2. Analog Signal Range */
#define SIGNAL_RANGE_PARAM
#define MIN_SIGNAL_VALUE
#define MAX_SIGNAL_VALUE
(ssGetSFcnParam(S,1))
((real_T) (mxGetPr(SIGNAL_RANGE_PARAM)[0]))
((real_T) (mxGetPr(SIGNAL_RANGE_PARAM)[1]))
/* 3. Hardware Gain */
#define HARDWARE_GAIN_PARAM
#define HARDWARE_GAIN
(ssGetSFcnParam(S,2))
((real_T) (mxGetPr(HARDWARE_GAIN_PARAM)[0]))
/* 4. Sample Time */
#define SAMPLE_TIME_PARAM
#define SAMPLE_TIME
(ssGetSFcnParam(S,3))
((real_T) (mxGetPr(SAMPLE_TIME_PARAM)[0]))
/*
* Hardware specific information pertaining to the A/D board. This information
* should be provided with the documentation that comes with the board.
*/
#include "device.h"
/*====================*
* S-function methods *
*====================*/
/* Function: mdlCheckParameters ================================================
* Abstract:
*
Check that the parameters passed to this S-function are valid.
*/
#define MDL_CHECK_PARAMETERS
static void mdlCheckParameters(SimStruct *S)
{
static char_T errMsg[256];
boolean_T allParamsOK = 1;
14-61
14
Targeting Real-Time Systems
/*
* Base I/O Address
*/
if (!mxIsChar(BASE_ADDRESS_PARAM)) {
sprintf(errMsg, "Base address parameter must be a string.\n");
allParamsOK = 0;
goto EXIT_POINT;
}
/*
* Signal Range
*/
if (mxGetNumberOfElements(SIGNAL_RANGE_PARAM) != 2) {
sprintf(errMsg,
"Signal Range must be a two element vector [minInp maxInp]\n");
allParamsOK = 0;
goto EXIT_POINT;
}
if ( !adcIsSignalRangeParamOK(MIN_SIGNAL_VALUE, MAX_SIGNAL_VALUE) ) {
sprintf(errMsg,
"The specified Signal Range is not supported by I/O board.\n");
allParamsOK = 0;
goto EXIT_POINT;
}
/*
* Hardware Gain
*/
if (mxGetNumberOfElements(HARDWARE_GAIN_PARAM) != 1) {
sprintf(errMsg, "Hardware Gain must be a scalar valued real number\n");
allParamsOK = 0;
goto EXIT_POINT;
}
if (!adcIsHardwareGainParamOK(HARDWARE_GAIN)) {
sprintf(errMsg, "The specified hardware gain is not supported.\n");
allParamsOK = 0;
goto EXIT_POINT;
}
/*
* Sample Time
*/
if (mxGetNumberOfElements(SAMPLE_TIME_PARAM) != 1) {
sprintf(errMsg, "Sample Time must be a positive scalar.\n");
allParamsOK = 0;
goto EXIT_POINT;
}
EXIT_POINT:
if ( !allParamsOK ) {
ssSetErrorStatus(S, errMsg);
}
} /* end: mdlCheckParameters */
14-62
Creating Device Drivers
/* Function: mdlInitializeSizes ================================================
* Abstract:
* Validate parameters,set number and width of ports.
*/
static void mdlInitializeSizes(SimStruct *S)
{
/* Set the number of parameters expected. */
ssSetNumSFcnParams(S, NUM_S_FUNCTION_PARAMS);
if ( ssGetNumSFcnParams(S) == ssGetSFcnParamsCount(S) ) {
/*
* If the number of parameter passed in is equal to the number of
* parameters expected, then check that the specified parameters
* are valid.
*/
mdlCheckParameters(S);
if ( ssGetErrorStatus(S) != NULL ) {
return; /* Error was reported in mdlCheckParameters. */
}
} else {
return; /* Parameter mismatch. Error will be reported by Simulink. */
}
/*
* This S-functions's parameters cannot be changed in the middle of a
* simulation, hence set them to be nontunable.
*/
{
int_T i;
for (i=0; i < NUM_S_FUNCTION_PARAMS; i++) {
ssSetSFcnParamNotTunable(S, i);
}
}
/* Has no input ports */
if ( !ssSetNumInputPorts(S, 0) ) return;
/* Number of output ports = number of channels specified */
if ( !ssSetNumOutputPorts(S, N_CHANNELS) ) return;
/* Set the width of each output ports to be one. */
{
int_T oPort;
for (oPort = 0; oPort < ssGetNumOutputPorts(S); oPort++) {
ssSetOutputPortWidth(S, oPort, 1);
}
}
ssSetNumSampleTimes( S, 1);
} /* end: mdlInitializeSizes */
14-63
14
Targeting Real-Time Systems
/* Function: mdlInitializeSampleTimes ==========================================
* Abstract:
*
Set the sample time of this block as specified via the sample time
*
parameter.
*/
static void mdlInitializeSampleTimes(SimStruct *S)
{
ssSetSampleTime(S, 0, SAMPLE_TIME);
ssSetOffsetTime(S, 0, 0.0);
} /* end: mdlInitializeSampleTimes */
/* Function: mdlStart ==========================================================
* Abstract:
*
At the start of simulation in Simulink, print a message to the MATLAB
*
command window indicating that output of this block will be zero during
*
simulation.
*/
#define MDL_START
static void mdlStart(SimStruct *S)
{
if (ssGetSimMode(S) == SS_SIMMODE_NORMAL) {
mexPrintf("\n adc.c: The output of the A/D block '%s' will be set "
"to zero during simulation in Simulink.\n", ssGetPath(S));
}
} /* end: mdlStart */
/* Function: mdlOutputs ========================================================
* Abstract:
*
Set the output to zero.
*/
static void mdlOutputs(SimStruct *S, int_T tid)
{
int oPort;
for (oPort = 0; oPort < ssGetNumOutputPorts(S); oPort++) {
real_T *y = ssGetOutputPortRealSignal(S, oPort);
y[0] = 0.0;
}
} /* end: mdlOutputs */
/* Function: mdlTerminate ======================================================
* Abstract:
*
Required S-function method that gets called at the end of simulation
*
and code generation. Nothing to do in simulation.
*/
static void mdlTerminate(SimStruct *S)
14-64
Creating Device Drivers
{
} /* end: mdlTerminate */
/* Function: mdlRTW ============================================================
* Abstract:
*
Evaluate parameter data and write it to the model.rtw file.
*/
#define MDL_RTW
static void mdlRTW(SimStruct *S)
{
boolean_T polarity
= adcIsUnipolar(MIN_SIGNAL_VALUE, MAX_SIGNAL_VALUE);
real_T
offset
= polarity ? 0.0 : MIN_SIGNAL_VALUE/HARDWARE_GAIN;
real_T
resolution = (((MAX_SIGNAL_VALUE-MIN_SIGNAL_VALUE)/HARDWARE_GAIN)/
ADC_NUM_LEVELS);
char_T
baseAddrStr[128];
if ( mxGetString(BASE_ADDRESS_PARAM, baseAddrStr, 128) ) {
ssSetErrorStatus(S, "Error reading Base Address parameter, "
"need to increase string buffer size.");
return;
}
if ( !ssWriteRTWParamSettings(S, 4,
SSWRITE_VALUE_QSTR,
SSWRITE_VALUE_NUM,
SSWRITE_VALUE_NUM,
SSWRITE_VALUE_NUM,
"BaseAddress",
"HardwareGain",
"Resolution",
"Offset",
baseAddrStr,
HARDWARE_GAIN,
resolution,
offset) ) {
return; /* An error occured, which will be reported by Simulink. */
}
} /* end: mdlRTW */
/*
* Required include for Simulink-MEX interface mechanism
*/
#include "simulink.c"
/* EOF: adc.c */
adc.tlc
%% File
: adc.tlc
%% Abstract:
%%
Target file for the C-Mex S-function adc.c
%%
%% Copyright (c) 1994-2000 by The MathWorks, Inc. All Rights Reserved.
%%
%implements "adc" "C"
14-65
14
Targeting Real-Time Systems
%% Function: BlockTypeSetup ===========================================
%% Abstract:
%%
This function is called once for all instance of the S-function
%% "dac" in the model. Since this block requires hardware specific
%% information about the I/O board, we generate code to include
%% "device.h" in the generated model.h file.
%%
%function BlockTypeSetup(block, system) void
%%
%% Use the Target Language Ccompiler global variable INCLUDE_DEVICE_H to make
sure that
%% the line "#include device.h" gets generated into the model.h
%%file only once.
%%
%if !EXISTS("INCLUDE_DEVICE_H")
%assign ::INCLUDE_DEVICE_H = 1
%openfile buffer
/* Include information about the I/O board */
#include "device.h"
%closefile buffer
%<LibCacheIncludes(buffer)>
%endif
%endfunction %% BlockTypeSetup
%% Function: Start ====================================================
%% Abstract:
%%
Generate code to set the number of channels and the hardware gain
%% mask in the start function.
%%
%function Start(block, system) Output
/* %<Type> Block: %<Name> (%<ParamSettings.FunctionName>) */
%%
%assign numChannels = block.NumDataOutputPorts
%assign baseAddr
= SFcnParamSettings.BaseAddress
%assign hwGain
= SFcnParamSettings.HardwareGain
%%
%% Initialize the Mux Scan Register to scan from 0 to NumChannels-1.
%% Also set the Gain Select Register to the appropriate value.
%%
adcSetLastChannel(%<baseAddr>, %<numChannels-1>);
adcSetHardwareGain(%<baseAddr>, adcGetGainMask(%<hwGain>));
%endfunction %% Start
%% Function: Outputs =================================================
%% Abstract:
%%
Generate inlined code to perform one A/D conversion on the enabled
%%
channels.
%%
%function Outputs(block, system) Output
14-66
Creating Device Drivers
%%
%assign offset
= SFcnParamSettings.Offset
%assign resolution = SFcnParamSettings.Resolution
%assign baseAddr
= SFcnParamSettings.BaseAddress
%%
/* %<Type> Block: %<Name> (%<ParamSettings.FunctionName>) */
{
int_T chIdx;
uint_T adcValues[%<NumDataOutputPorts>];
for (chIdx = 0; chIdx < %<NumDataOutputPorts>; chIdx++) {
adcStartConversion(%<baseAddr>);
while (adcIsBusy(%<baseAddr>)) {
/* wait for conversion */
}
adcValues[chIdx] = adcGetValue(%<baseAddr>);
}
%foreach oPort = NumDataOutputPorts
%assign y = LibBlockOutputSignal(oPort, "", "", 0)
%<y> = %<offset> + %<resolution>*adcValues[%<oPort>];
%endforeach
}
%endfunction %% Outputs
%% EOF: adc.tlc
device.h
/*
* File
: device.h
*
* Copyright (c) 1994-2000 by The MathWorks, Inc. All Rights
* Reserved.
*
*/
/*
* Operating system utilities to read and write to hardware
* registers.
*/
#define ReadByte(addr)
inp(addr)
#define WriteByte(addr,val)
outp(addr,val)
/*=============================================================*
* Specification of the Analog Input Section of the I/O board
* (used in the ADC device driver S-function, adc.c and *adc.tlc)
*=======================================================*/
/*
* Define macros for the attributes of the A/D board, such as the
14-67
14
Targeting Real-Time Systems
* number
*/
#define
#define
#define
of A/D channels and bits per channel.
ADC_MAX_CHANNELS
(16)
ADC_BITS_PER_CHANNEL
(12)
ADC_NUM_LEVELS ((uint_T) (1 << ADC_BITS_PER_CHANNEL))
/*
* Macros to check if the specified parameters are valid.
* These macros are used by the C-Mex S-function, adc.c
*/
#define adcIsUnipolar(lo,hi)
(lo
== 0.0 && 0.0 < hi)
#define adcIsBipolar(lo,hi)
(lo + hi == 0.0 && 0.0 < hi)
#define adcIsSignalRangeParamOK(l,h) (adcIsUnipolar(l,h) || adcIsBipolar(l,h))
#define
#define
#define
adcGetGainMask(g) ( (g==1.0) ? 0x0 : \
( (g==10.0) ? 0x1 : \
( (g==100.0) ? 0x2 : \
( (g==500.0) ? 0x3 : 0x4 ) ) ) )
adcIsHardwareGainParamOK(g)
(adcGetGainMask(g) != 0x4)
adcIsNumChannelsParamOK(n)
(1 <= n && n <= ADC_MAX_CHANNELS)
/* Hardware registers used by the A/D section of the I/O board */
#define
#define
#define
#define
#define
#define
/*
* Macros
*/
#define
#define
#define
#define
#define
#define
#define
ADC_START_CONV_REG(bA)
ADC_LO_BYTE_REG(bA)
ADC_HI_BYTE_REG(bA)
ADC_MUX_SCAN_REG(bA)
ADC_STATUS_REG(bA)
ADC_GAIN_SELECT_REG(bA)
(bA)
(bA)
(bA +
(bA +
(bA +
(bA +
0x1)
0x2)
0x8)
0xB)
for the A/D section of the I/O board
adcSetLastChannel(bA,n) WriteByte(ADC_MUX_SCAN_REG(bA), n<<4)
adcSetHardwareGain(bA,gM) WriteByte(ADC_GAIN_SELECT_REG(bA), gM)
adcStartConversion(bA) WriteByte(ADC_START_CONV_REG(bA), 0x00)
adcIsBusy(bA) (ReadByte(ADC_STATUS_REG(bA)) & 0x80)
adcGetLoByte(bA) ReadByte(ADC_LO_BYTE_REG(bA))
adcGetHiByte(bA) ReadByte(ADC_HI_BYTE_REG(bA))
adcGetValue(bA) ((adcGetLoByte(bA)>>4) | (adcGetHiByte(bA)<<4))
/*============================================================*
* Specification of the Analog Output Section of the I/O board
* (used in the DAC device driver S-function, adc.c and adc.tlc)
*============================================================*/
#define
#define
#define
#define
#define
#define
14-68
DAC_BITS_PER_CHANNEL
DAC_UNIPOLAR_ZERO
DAC_BIPOLAR_ZERO
DAC_MIN_OUTPUT
DAC_MAX_OUTPUT
DAC_NUM_LEVELS
(12)
( 0)
(1 << (DAC_BITS_PER_CHANNEL-1))
(0.0)
((real_T) ((1 << DAC_BITS_PER_CHANNEL)-1))
((uint_T) (1 << DAC_BITS_PER_CHANNEL))
Creating Device Drivers
/*
* Macros to check if the specified parameters are valid.
* These macros are used by the C-Mex S-function,dac.c.
*/
#define dacIsUnipolar(lo,hi)
(lo
== 0.0 && 0.0 < hi)
#define dacIsBipolar(lo,hi)
(lo+hi == 0.0 && 0.0 < hi)
#define dacIsSignalRangeParamOK(l,h) (dacIsUnipolar(l,h) || dacIsBipolar(l,h))
/* Hardware registers */
#define DAC_LO_BYTE_REG(bA)
#define DAC_HI_BYTE_REG(bA)
(bA + 0x4)
(bA + 0x5)
#define dacSetLoByte(bA,c) WriteByte(DAC_LO_BYTE_REG(bA),(c & 0x00f)<<4)
#define dacSetHiByte(bA,c) WriteByte(DAC_HI_BYTE_REG(bA),(c & 0xff0)>>4)
#define dacSetValue(bA,c) dacSetLoByte(bA,c); dacSetHiByte(bA,c)
/* EOF: device.h */
14-69
14
Targeting Real-Time Systems
Interfacing Parameters and Signals
Simulink external mode (see Chapter 6, “External Mode”) offers a quick and
easy way to monitor signals and modify parameter values while generated
model code executes. However, external mode may not be appropriate for your
target or optimal for your application. S-function targets do not support
external mode, nor do DOS targets. In other cases, you may prefer to use
existing code to access parameters and signals of a model directly, rather than
using the external mode mechanism.
Real-Time Workshop supports several approaches to the task of interfacing
block parameters and signals to your hand-written code.
The Model Parameter Configuration dialog enables you to declare how the
generated code allocates memory for variables used in your model. This allows
your supervisory software to read or write block parameter variables as your
model executes. Similarly, the Signal Properties dialog gives your code access
to selected signals within your model. Operation of these dialogs is described
in “Parameters: Storage, Interfacing, and Tuning” on page 5-2 and “Signals:
Storage, Optimization, and Interfacing” on page 5-17.
In addition, the MathWorks provides C and Target Language Compiler APIs
that give your code additional access to block outputs, and parameters that are
stored in global data structures and global variables created by Real-Time
Workshop. This section is an overview of these APIs. This section also includes
pointers to additional detailed API documents shipped with Real-Time
Workshop.
Signal Monitoring via Block Outputs
All block output data is written to the block outputs structure or specified
global variables with each time step in the model code. To access the output of
a given block in the generated code, your supervisory software must have the
following information, per port:
• The address of the field of the rtB structure, or the global variable where the
data is stored
• The number of output ports of the block
• The width of the signal
• The data type of the signal
14-70
Interfacing Parameters and Signals
This information is contained in the BlockIOSignals data structure. The TLC
code generation variable, BlockIOSignals, determines whether
BlockIOSignals data is generated. If BlockIOSignals is set to 1, a file
containing an array of BlockIOSignals structures is written during code
generation. This file is named model_bio.c, and by default is not generated.
BlockIOSignals is disabled by default. To enable generation of model_bio.c,
use the following statement in the Configure RTW code generation settings
section of your system target file:
%assign BlockIOSignals = 1
Alternatively, you can append the following command to the System target
file field on the Target configuration section of the Real-Time Workshop
pane.
-aBlockIOSignals=1
Note that depending on the size of your model, the BlockIOSignals array can
consume a considerable amount of memory.
BlockIOSignals and the Local Block Outputs Option
When the Local block outputs code generation option is selected, block
outputs are declared locally in functions instead of being declared globally in
the rtB structure when possible. The BlockIOSignals array in model_bio.c
will not contain information about such locally declared signals. (Note that
even when all outputs in the system are declared locally, enabling
BlockIOSignals will generate model_bio.c. In such a case the
BlockIOSignals array will contain only a null entry.)
Signals that are designated as test points via the Signal Properties dialog are
declared globally in the rtB structure, even when the Local block outputs
option is selected. Information about test-pointed signals is therefore written
to the BlockIOSignals array in model_bio.c. Similarly, signals whose storage
class is set are declared as global variables and represented in the
BlockIOSignals array.
Therefore, you can interface your code to selected signals by test-pointing them
or using storage classes, without losing the benefits of the Local block outputs
optimization for the other signals in your model.
14-71
14
Targeting Real-Time Systems
model_bio.c and the BlockIO Data Structure
The BlockIOSignals data structure is declared as follows.
typedef struct BlockIOSignals_tag {
char_T *blockName; /* Block's full pathname
(mangled by the Real-Time Workshop) */
char_T *signalName; /* Signal label (unmangled) */
uint_T portNumber; /* Block output port number (start at 0) */
uint_T signalWidth; /* Signal's width */
void *signalAddr;
/* Signal's address in the rtB vector */
char_T *dtName;
/* The C language data type name */
uint_T dtSize;
/* The size (# of bytes) for the data type*/
} BlockIOSignals;
The structure definition is in matlabroot/rtw/c/src/bio_sig.h. The
model_bio.c file includes bio_sig.h. Any source file that references the array
should also include bio_sig.h.
model_bio.c defines an array of BlockIOSignals structures. Each array
element, except the last, describes one output port for a block. The final
element is a sentinel, with all fields set to null values.
The code fragment below is an example of an array of BlockIOSignals
structures from a model_bio.c file.
#include "bio_sig.h"
/* Block output signal information */
static const BlockIOSignals rtBIOSignals[] =
{
/* blockName,
signalName, portNumber, signalWidth, signalAddr,
dtName, dtSize */
{
"simple/Constant",
NULL, 0, 1, &rtB.Constant,
"double", sizeof(real_T)
},
{
"simple/Constant1",
NULL, 0, 1, &rtB.Constant1,
"double", sizeof(real_T)
},
14-72
Interfacing Parameters and Signals
{
"simple/Gain",
"accel", 0, 2, &rtB.accel[0],
"double", sizeof(real_T)
},
{
NULL, NULL, 0, 0, 0, NULL, 0
}
};
Thus, a given block will have as many entries as it has output ports. In the
example above, the entry corresponding to the signal at output port 0 (indexing
is 0-based) of the block with path simple/Gain is named accel and has width 2.
Using BlockIOSignals to Obtain Block Outputs
The model_bio.c array is accessed via the name rtBIOSignals. To avoid
overstepping array bounds, you can do either of the following:
• Use the rtModel access macro rtmGetNumBlockIO to determine the number
of elements in the array.
• Use the rtModel access macro rtmGetModelMappingInfo to return the
mapping info corresponding to the model, and then access the array through
the mapping info.
• Test for a null blockName to identify the last element in the array.
You must then write code that iterates over the rtBIOSignals array and
chooses the signals to be monitored based on the blockName and signalName or
portNumber. How the signals are monitored is up to you. For example, you
could collect the signals at every time step. Alternatively, you could sample
signals asynchronously in a separate, lower priority task.
The following code example is drawn from the main program (rt_main.c) of the
Tornado target. The code illustrates how the StethoScope Graphical
Monitoring/Data Analysis Tool uses BlockIOSignals to collect signal
information in Tornado targets. The following function,
rtInstallRemoveSignals, selectively installs signals from the
BlockIOSignals array into the StethoScope Tool by calling
ScopeInstallSignal. The main simulation task then collects signals by calling
ScopeCollectSignals.
14-73
14
Targeting Real-Time Systems
static int_T rtInstallRemoveSignals(RT_MODEL *rtM, char_T
*installStr,
int_T fullNames, int_T install)
{
uint_T
i, w;
char_T
*blockName;
char_T
name[1024];
ModelMappingInfo mapInfo = rtmGetModelMappingInfo(rtM);
BlockIOSignals *rtBIOSignals = mapInfo.Signals.blockIOSignals;
int_T
ret = -1;
if (installStr == NULL) {
return -1;
}
i = 0;
while(rtBIOSignals[i].blockName != NULL) {
BlockIOSignals *blockInfo = &rtBIOSignals[i++];
if (fullNames) {
blockName = blockInfo->blockName;
} else {
blockName = strrchr(blockInfo->blockName, '/');
if (blockName == NULL) {
blockName = blockInfo->blockName;
} else {
blockName++;
}
}
if ((*installStr) == '*') {
} else if (strcmp("[A-Z]*", installStr) == 0) {
if (!isupper(*blockName)) {
continue;
}
} else {
if (strncmp(blockName, installStr, strlen(installStr)) !=
0) {
continue;
}
14-74
Interfacing Parameters and Signals
}
/*install/remove the signals*/
for (w = 0; w < blockInfo->signalWidth; w++) {
sprintf(name, "%s_%d_%s_%d", blockName,
blockInfo->portNumber,
!strcmp(blockInfo->signalName,"NULL")?"":blockInfo->signalName,
w);
if (install) { /*install*/
if (!ScopeInstallSignal(name, "units",
(void *)((int)blockInfo->signalAddr +
w*blockInfo->dtSize),
blockInfo->dtName, 0)) {
fprintf(stderr,"rtInstallRemoveSignals:
ScopeInstallSignal "
"possible error: over 256 signals.\n");
return -1;
} else {
ret =0;
}
} else { /*remove*/
if (!ScopeRemoveSignal(name, 0)) {
fprintf(stderr,"rtInstallRemoveSignals:
ScopeRemoveSignal\n"
"%s not found.\n",name);
} else {
ret =0;
}
}
}
}
return ret;
}
Below is an excerpt from an example routine that collects signals taken from
the main simulation loop.
/*******************************************
* Step the model for the base sample time *
*******************************************/
OUTPUTS(rtM,FIRST_TID);
14-75
14
Targeting Real-Time Systems
rtExtModeUploadCheckTrigger();
rtExtModeUpload(FIRST_TID,rtmGetTaskTime(rtM, FIRST_TID));
#ifdef MAT_FILE
if (rt_UpdateTXYLogVars(rtmGetRTWLogInfo(rtM),
rtmGetTPtr(rtM)) != NULL) {
fprintf(stderr,"rt_UpdateTXYLogVars() failed\n");
return(1);
}
#endif
#ifdef STETHOSCOPE
ScopeCollectSignals(0);
#endif
UPDATED(rtM,FIRST_TID);
if (rtmGetSampleTime(rtM,0) == CONTINUOUS_SAMPLE_TIME) {
rt_ODEUpdateContinuousStates(rtmGetRTWSolverInfo(rtM));
} else {
rt_SimUpdateDiscreteTaskTime(rtmGetTPtr(rtM),
rtmGetTimingData(rtM),0);
}
#if FIRST_TID == 1
rt_SimUpdateDiscreteTaskTime(rtmGetTPtr(rtM),
rtmGetTimingData(rtM),1);
#endif
rtExtModeCheckEndTrigger();
} /* end while(1) */
return(1);
} /* end tBaseRate */
<code continues ...>
See Chapter 12, “Targeting Tornado for Real-Time Applications” for more
information on using StethoScope.
14-76
Interfacing Parameters and Signals
C API for Parameter Tuning
Before reading this section, you should be familiar with the parameter storage
and tuning concepts described in “Parameters: Storage, Interfacing, and
Tuning” on page 5-2.
Overview
Real-Time Workshop provides data structures and a C API that enable a
running program to access model parameters without use of external mode.
Using the C API, you can
• Modify all occurrences of a MATLAB variable within a Simulink model
• Modify Stateflow machine data
• Modify a specified block parameter
• Modify a specific element within a block parameter
To access model parameters via the C API, you generate a model-specific
parameter mapping file, model_pt.c. This file contains parameter mapping
arrays containing information required for parameter tuning:
• The rtBlockTuning array contains information on all the modifiable block
parameters in the model by block name and parameter name. Each element
of the array is a BlockTuning struct. Note that if the Inline parameters
option is selected, an empty rtBlockTuning array is generated.
• The rtVariableTuning array contains information about all workspace
variables that were referenced as block parameters by one or more blocks or
Stateflow charts in the model. Each element of the array is a
VariableTuning struct. Note that if the Inline parameters option is not
selected, the elements of this array correspond to Stateflow sata of machine
scope.
• The rtParametersMap array, or map vector, contains the absolute base
address of all block or model parameters. The entries of the map are
initialized by the function model_InitializeParametersMap, which is called
during model initialization.
• The rtDimensionsMap array, or dimensions map, is a structure that contains
the dimensions sizes for parameters having dimensions greater than 2.
Your code should not access the data structures of model_pt.c directly.
Pointers to these arrays are loaded into a ModelMappingInfo structure that is
14-77
14
Targeting Real-Time Systems
cached in the rtModel data structure. Your code must obtain a pointer to the
ModelMappingInfo structure, using an accessor macro provided for the
purpose. Your code can then use the rtBlockTuning and rtVariableTuning
structures to access model parameters.
Real-Time Workshop provides sample code demonstrating how to use the
parameter mapping information. You can use this sample code as a starting
point in developing your own parameter tuning code.
The following sections discuss:
• How to generate the model_pt.c file
• Details of the parameter mapping structures
• Mapping of inlined and non-inlined parameters.
• Using the sample code
• Restrictions on the use of the parameter tuning API
• Summary of relevant source files
Generating the model.pt File
To generate the model_pt.c file, you must set the global TLC variable
ParameterTuning to 1 (by default, ParameterTuning is disabled.) You can use
the following statement in your system target file for this purpose.
%assign ParameterTuning = 1
Alternatively, you can append the following command to the System target
file field on the Target configuration section of the Real-Time Workshop
pane.
-aParameterTuning=1
The the model_pt.c file is written to the build directory.
Parameter Map Data Types and Data Structures
The file matlabroot/rtw/c/src/pt_info.h defines enumerated data types
and data structures used in the parameter map. Please refer to pt_info.h
while reading this discussion.
Enumerated Types. Two enumerations, ParamClass and ParamSource, are
defined in pt_info.h.
14-78
Interfacing Parameters and Signals
The ParamClass enumeration specifies how a parameter is to be updated. The
values rt_SCALAR and rt_VECTOR represent scalars and column vectors,
respectively. The C declarations for these types are
real_T scalarParam;
real_T vectorParam[width];
/* correpsponds to rt_SCALAR */
/* correpsponds to rt_VECTOR */
The value rt_MATRIX_ROW_MAJOR indicates that the parameter is a matrix that
is stored in memory in row major ordering. Conceptually, the C declaration for
a parameter of this type is
real_T param[nRows][nCols];
The value rt_MATRIX_COL_MAJOR specifies that the parameter is a matrix that
is stored in memory in column major ordering. Conceptually, the C declaration
for a parameter of this type is
real_T param[nCols][nRows];
The value rt_MATRIX_COL_MAJOR_ND specifies that the parameter is an
N-dimensional matrix. Conceptually, the C declaration for a parameter of this
type is
real_T param[dim2Size][dim1Size][dim3Size][dim4Size][...]
Note that Real-Time Workshop actually declares matrices as vectors in column
major order in each case. For example, a 2x3 matrix is represented as follows.
• In MATLAB:
matrix = [1,2,3; 4,5,6]
• In Real-Time Workshop:
real_T matrix[6] = {1.0, 4.0, 2.0, 5.0, 3.0, 6.0}
The ParamSource enumeration specifies the source of the parameter, which
may be one of the following:
• rt_SL_PARAM indicates a parameter used by a Simulink block.
• rt_SF_PARAM indicates Stateflow machine data.
• rt_SHARED_PARAM indicates data shared by Simulink and Stateflow.
14-79
14
Targeting Real-Time Systems
Map Vector. The map vector (rtParametersMap) is an array containing the
absolute base addresses of all block parameters that are members of rtP, the
global parameter data structure. The code fragment below shows an example
map vector. This example was generated from the model shown in Figure 14-1.
static void * const rtParametersMap[] = {
&rtP.amp,
/* 0: amp */
&rtP.freq,
/* 1: freq */
};
ParameterTuning, BlockTuning, and VariableTuning Structures. The ParameterTuning
structure contains the core of information stored in the BlockTuning and
VariableTuning structures. ParameterTuning is defined as follows:
typedef struct ParameterTuning_tag {
ParamClass paramClass;
/* Class of parameter */
int_T
nRows;
/* Number of rows */
int_T
nCols;
/* Number of columns */
int_T
nDims;
/* Number of dimensions */
int_T
dimsOffset;
/* Offset into dimensions vector */
ParamSource source;
/* Source of parameter */
uint_T
dataType;
/* data type enumeration */
uint_T
numInstances; /* Num of parameter instances */
int_T
mapOffset;
/* Offset into map vector */
} ParameterTuning;
The paramClass and source fields take on one of the enumerated values
mentioned in “Enumerated Types” on page 14-78.
The dataType field is the Simulink data type of the parameter, indicated by an
enumerated value such as SS_DOUBLE.
The mapOffset field is the offset to the parameter’s entry in the map vector.
Using mapOffset, your code can obtain the actual address of the parameter.
The numInstances field is described in “Mapping Parameter Instances in
Simulink and Stateflow” on page 14-85.
The fields nDims, nRows and nCols indicate the number of dimensions, rows and
columns in the parameter, respectively.
If the number of dimensions of the parameter is greater than 2, the value
dimsOffset is used to index into the dimensions map. This array contains the
14-80
Interfacing Parameters and Signals
dimensions sizes for parameters having dimensions greater than 2. If there are
no parameters having more than 2 dimensions, the dimensions map is empty.
The following table summarizes the relationship of the number of dimensions
to the dimensions information in the ParameterTuning structure.
Table 14-4: Parameter Tuning Dimensions Information
Number of
Dimensions
Dimensions Information Fields
<= 2
nRows and nCols valid, nDims = 2, dimsOffset = -1
> 2
nRows=-1, nCols=-1, nDims and dimsOffset valid
The BlockTuning structure, in addition to the ParameterTuning information,
contains the names of the originating block and parameter.
The VariableTuning structure, in addition to the ParameterTuning
information, contains the name of the workspace variable.
Inlining Parameters
The Inline parameters option affects the information generated in the
rtBlockTuning and rtVariableTuning arrays.
If Inline parameters is deselected:
• The rtBlockTuning array contains an entry for every modifiable parameter
of every block in the model.
• The rtVariableTuning array contains only Stateflow data of machine scope
(it contains only a null entry in the absence of such data).
If Inline parameters is selected:
• The rtBlockTuning array is empty (it contains only a null entry).
• The rtVariableTuning array contains an entry for all workspace variables
that are referenced as tunable Simulink block parameters or Stateflow data
of machine scope.
Example Parameter Maps. In this section, we will examine parameter mapping
information generated from a simple model. In the example model, the
14-81
14
Targeting Real-Time Systems
amplitude and frequency of the Sine Wave block are controlled by the
workspace variables amp and freq, as shown below.
Figure 14-10: Example Model Referencing Workspace Variables as
Parameters
The following code fragment shows the rtBlockTuning and rtVariableTuning
arrays generated from this model (in model_pt.c), as well as the parameter
map and the function initializing the map, with Inline parameters off.
/* Tunable block parameters */
static const BlockTuning rtBlockTuning[] = {
/* blockName, parameterName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
14-82
Interfacing Parameters and Signals
*/
/* Sin */
{"simple/Sine Wave", "Amplitude",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 0}
},
/* Sin */
{"simple/Sine Wave", "Bias",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 1}
},
/* Sin */
{"simple/Sine Wave", "Frequency",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 2}
},
/* Sin */
{"simple/Sine Wave", "Phase",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 3}
},
/* Gain */
{"simple/Gain", "Gain",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 4}
},
{NULL, NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
/* Tunable variable parameters */
static const VariableTuning rtVariableTuning[] = {
/* variableName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
*/
{NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
static void * rtParametersMap[5];
void simple_InitializeParametersMap(void) {
rtParametersMap[0] = &rtP.Sine_Wave_Amp; /* 0 */
rtParametersMap[1] = &rtP.Sine_Wave_Bias; /* 1 */
rtParametersMap[2] = &rtP.Sine_Wave_Freq; /* 2 */
rtParametersMap[3] = &rtP.Sine_Wave_Phase; /* 3 */
rtParametersMap[4] = &rtP.Gain_Gain; /* 4 */
}
14-83
14
Targeting Real-Time Systems
The following code fragment shows the rtBlockTuning and rtVariableTuning
arrays generated (in model_pt.c), as well as the parameter map and the
function initializing the map, with Inline parameters on. The workspace
variables amp and freq have been declared tunable with storage class
SimulinkGlobal(Auto).
/* Individual block tuning is not valid when inline parameters is selected. *
* An empty map is produced to provide a consistent interface independent
*
* of inlining parameters.
*/
static const BlockTuning rtBlockTuning[] = {
/* blockName, parameterName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
*/
{NULL, NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
/* Tunable variable parameters */
static const VariableTuning rtVariableTuning[] = {
/* variableName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
*/
{"amp",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 0}
},
{"freq",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 1}
},
{NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
static void * rtParametersMap[2];
void simple_inline_InitializeParametersMap(void) {
rtParametersMap[0] = &rtP.amp;
/* 0: amp */
rtParametersMap[1] = &rtP.freq;
/* 1: freq */
}
14-84
Interfacing Parameters and Signals
Mapping Parameter Instances in Simulink and Stateflow
Simulink and Stateflow can have a shared or nonshared mapping of a
parameter, depending on the parameter's Simulink storage class and Stateflow
scope. A shared mapping is one in which the address of the parameter is the
same in the code generated for Simulink blocks and Stateflow charts. This
table shows how Simulink storage class and Stateflow scope affect the sharing
of parameters in Simulink and Stateflow.
Simulink
SimulinkGlobal
storage class
Simulink
ExportedGlobal
Simulink
ImportedExtern
storage class
storage class
Stateflow
imported
scope
storage class
Nonshared
Shared
(see note b)
Shared
(see note b)
Error
Nonshared
Error
Shared
(recommended:
see note a)
Error
scope
Stateflow
exported
Simulink
ImportedExternPointer
Note a: Recommended; does not require any user defined data definition.
Note b: Requires user defined data definition.
Therefore, to best share data between Simulink and Stateflow, define
parameters as exported in Stateflow and as ImportedExtern in Simulink.
When the mapping is nonshared, separate instances of that parameter appear
in the code generated for Simulink and Stateflow.
As an example, consider the model shown in this picture.In this model, the
MATLAB variable Kp is specified in two Gain blocks and as data of machine
scope in a Stateflow chart.
14-85
14
Targeting Real-Time Systems
When Inline parameters is selected, both Gain blocks share a single instance
of Kp, and the Stateflow chart references a second instance. In such cases, the
numInstances and mapOffset fields of the ParameterTuning structure are
used in conjunction. The numInstances field specifies the number of parameter
instances, while mapOffset is the offset into the map vector
(rtParametersMap). The map vector determines the actual address of each
instance from its source.
The following code shows the rtVariableTuning and rtParametersMap arrays
for this case.
/* Tunable variable parameters */
static const VariableTuning rtVariableTuning[] = {
/* variableName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
*/
{"Kp",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 2, 0}
},
{NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
static void * rtParametersMap[2];
void complex_inline_InitializeParametersMap(void) {
rtParametersMap[0] = &rtP.Kp;
/* 0: Kp */
rtParametersMap[1] = &Kp;
/* 1: Kp */
}
static uint_T const rtDimensionsMap[] = {
0
/* Dummy */
14-86
Interfacing Parameters and Signals
};
When Inline parameters is not selected, Real-Time Workshop creates two
instances of Kp in the global parameters structure rtP for the two Simulink
Gain blocks referencing Kp, and a third instance as a global variable for the
Stateflow chart. All three instances must be updated to reflect any change in
Kp. In the code example below, the entries for the Gain blocks in rtBlockTuning
correspond to the two instances of Kp for those blocks. In addition, the entry for
Kp in rtVariableTuning corresponds to the third instance for the Stateflow
chart.
/* Tunable block parameters */
static const BlockTuning rtBlockTuning[] = {
/* blockName, parameterName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
*/
/* Sin */
{"complex_noninline/Sine Wave", "Amplitude",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 0}
},
/* Sin */
{"complex_noninline/Sine Wave", "Bias",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 1}
},
/* Sin */
{"complex_noninline/Sine Wave", "Frequency",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 2}
},
/* Sin */
{"complex_noninline/Sine Wave", "Phase",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 3}
},
/* Gain */
{"complex_noninline/Gain", "Gain",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 4}
},
/* Sin */
14-87
14
Targeting Real-Time Systems
{"complex_noninline/Sine Wave1", "Amplitude",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 5}
},
/* Sin */
{"complex_noninline/Sine Wave1", "Bias",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 6}
},
/* Sin */
{"complex_noninline/Sine Wave1", "Frequency",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 7}
},
/* Sin */
{"complex_noninline/Sine Wave1", "Phase",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 8}
},
/* Gain */
{"complex_noninline/Gain1", "Gain",
{rt_SCALAR, 1, 1, 2, -1, rt_SL_PARAM, SS_DOUBLE, 1, 9}
},
{NULL, NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
/* Tunable variable parameters */
static const VariableTuning rtVariableTuning[] = {
/* variableName,
* class, nRows, nCols, nDims, dimsOffset, source, dataType, numInstances,
* mapOffset
*/
{"Kp",
{rt_SCALAR, 1, 1, 2, -1, rt_SF_PARAM, SS_DOUBLE, 1, 10}
},
{NULL,
{(ParamClass)0, 0, 0, 0, 0, (ParamSource)0, 0, 0, 0}
}
};
static void * rtParametersMap[11];
void complex_noninline_InitializeParametersMap(void) {
rtParametersMap[0] = &rtP.Sine_Wave_Amp; /* 0 */
14-88
Interfacing Parameters and Signals
rtParametersMap[1] = &rtP.Sine_Wave_Bias; /* 1 */
rtParametersMap[2] = &rtP.Sine_Wave_Freq; /* 2 */
rtParametersMap[3] = &rtP.Sine_Wave_Phase; /* 3 */
rtParametersMap[4] = &rtP.Gain_Gain; /* 4 */
rtParametersMap[5] = &rtP.Sine_Wave1_Amp; /* 5 */
rtParametersMap[6] = &rtP.Sine_Wave1_Bias; /* 6 */
rtParametersMap[7] = &rtP.Sine_Wave1_Freq; /* 7 */
rtParametersMap[8] = &rtP.Sine_Wave1_Phase; /* 8 */
rtParametersMap[9] = &rtP.Gain1_Gain; /* 9 */
rtParametersMap[10] = &Kp;
/* 10: Kp */
}
static uint_T const rtDimensionsMap[] = {
0
/* Dummy */
};
Accessing the Parameter Mapping Structures.
The parameter mapping arrays in model_pt.c are declared static. Pointers to
the parameter mapping arrays are stored in a ModelMappingInfo structure,
defined as follows in matlabroot/rtw/c/src/mdl_info.h.
typedef struct ModelMappingInfo_tag {
/* block signal monitoring */
struct {
BlockIOSignals const *blockIOSignals;
uint_T
numBlockIOSignals;
} Signals;
/* parameter tuning */
struct {
BlockTuning const
VariableTuning const
void * const
uint_T const
uint_T
uint_T
} Parameters;
} ModelMappingInfo;
*blockTuning;
*variableTuning;
*parametersMap;
*dimensionsMap;
numBlockTuning;
numVariableTuning;
/* Block signals map
/* Num signals in map
*/
*/
/*
/*
/*
/*
/*
/*
*/
*/
*/
*/
*/
*/
Block parameters map
Variable parameters map
Parameter index map
Dimensions index map
Num block parameters in map
Num variable parameter in map
The ModelMappingInfo structure is cached in the rtModel data structure. Use
the rtmGetModelMappingInfo macro to obtain a pointer to the
ModelMappingInfo structure, as in the following example.
#include "mdl_info.h"
/* note: rTM is a pointer to the real-time Model Object */
.
.
.
14-89
14
Targeting Real-Time Systems
ModelMappingInfo *MMI = rtmGetModelMappingInfo(rtM);
In mdl_info.h, Real-Time Workshop provides additional macros that let you
access members of the ModelMappingInfo structurevia a ModelMappingInfo
pointer.
Using the Example Code
Real-Time Workshop provides example code that uses the parameter tuning
API in matlabroot/rtw/c/src/pt_print.c. This file contains three functions:
• rt_PrintParamInfo displays all the tunable block parameters and MATLAB
variables to the standard output.
• rt_PrintPTRec prints and then tests a parameter tuning record.
• rt_ModifyParameter updates all parameters associated with a specified
parameter tuning record.
This code is intended as a starting point for your parameter tuning code. For
more information see the function abstracts preceding each function.
To become familiar with the example code, we suggest building a model that
displays all the tunable block parameters and MATLAB variables to the
screen. You can use ptdemo, the parameter tuning demo model, for this
purpose. First, run the demo with Inline parameters on:
1 Open the ptdemo model.
ptdemo
2 From the Simulation menu, choose Simulation Parameters.
3 Select the Advanced tab. Make sure that the Inline parameters option is
selected. Click Apply if necessary.
4 Click on the Real-Time Workshop tab of the Simulation Parameters dialog
box. The Real-Time Workshop pane activates. Select Target configuration
from the Category menu. Note the System target file edit field contains:
grt.tlc -aParameterTuning=1 -aParameterTuningTestFile="ptinfotestfile.tlc" -p0
The second argument:
14-90
Interfacing Parameters and Signals
-aParameterTuningTestFile="ptinfotestfile.tlc"
includes the TLC file matlabroot/rtw/c/tlc/mw/ptinfotestfile.tlc.
This file generates code required to display the parameter tuning
information.
5 Click the Build button.
6 When the build completes, run the executable program:
!ptdemo
Parameter information will be displayed in the MATLAB command window.
You can inspect the parameter map in the build directory
(./ptdemo_grt_rtw/ptdemo_pt.c).
Next, run the demo with Inline parameters off:
1 Select the Advanced tab of the Simulation Parameters dialog. Make sure
that the Inline parameters option is deselected. Click Apply if necessary.
2 Repeat steps 4-6 above. Note the difference, in the displayed parameter
information and the ptdemo_pt.c file, with Inline parameters on versus off.
Restrictions
The parameter tuning C API does not support:
• Complex parameters (e.g., k=1+i)
• Parameters local to Stateflow (e.g., chart parented data)
• Parameters transformed by Simulink (e.g., parameters of zero-pole TF)
• Parameters transformed by mask initialization code
Note, however, that transformations that do not change the value of a
masked parameter (such as a=k) are supported.
• The S-function code format
• The Simulink Accelerator
• Fixed-point parameters. Fixed-point Blockset parameters are not supported
unless they have a nominal scaling.
Note that Simulink built-in data types are supported. This includes signed
and unsigned 8, 16, and 32-bit integer, double, single and boolean.
14-91
14
Targeting Real-Time Systems
The coefficients of the Transfer Fcn, State-Space, Discrete Filter, Discrete
Transfer Function and Discrete State-Space blocks are tunable, subject to
requirements described in “Tunability of Linear Block Parameters” on
page 5-14.
Summary of Parameter Tuning Source Files
• matlabroot/rtw/c/src/mdl_info.h: model mapping structure definition
• matlabroot/rtw/c/src/pt_info.h: parameter tuning structure definitions
• matlabroot/rtw/c/tlc/ptinfo.tlc: TLC file to produce model_pt.c
• matlabroot/rtw/c/tlc/mw/ptinfotestfile.tlc: TLC file to hook in
example print code
• matlabroot/rtw/c/src/pt_print.c: example code to print/modify
parameters
Target Language Compiler API for
Signals and Parameters
Real-Time Workshop provides a TLC function library that lets you create a
global data map record. The global data map record, when generated, is added
to the CompiledModel structure in the model.rtw file. The global data map
record is a database containing all information required for accessing memory
in the generated code, including:
• Signals (Block I/O)
• Parameters
• Data type work vectors (DWork)
• External inputs
• External outputs
Use of the global data map requires knowledge of the Target Language
Compiler and of the structure of the model.rtw file. See the Target Language
Compiler documentation for information on these topics.
The TLC functions that are required to generate and access the global data
map record are contained in matlabroot/rtw/c/tlc/mw/globalmaplib.tlc.
The comments in the source code fully document the global data map
structures and the library functions.
14-92
Interfacing Parameters and Signals
Note The global data map structures and functions maybe modified and/or
enhanced in future releases.
14-93
14
Targeting Real-Time Systems
Creating an External Mode Communication Channel
This section provides information you will need in order to support external
mode on your custom target, using your own low-level communications layer.
This information includes:
• An overview of the design and operation of external mode
• A description of external mode source files
• Guidelines for modifying the external mode source files and rebuilding the
ext_comm MEX-file
This section assumes that you are familiar with the execution of Real-Time
Workshop programs, and with the basic operation of external mode. These
topics are described in Chapter 7, “Program Architecture” and Chapter 6,
“External Mode.”
The Design of External Mode
External mode communication between Simulink and a target system is based
on a client/server architecture. The client (Simulink) transmits messages
requesting the server (target) to accept parameter changes or to upload signal
data. The server responds by executing the request.
A low-level transport layer handles physical transmission of messages. Both
Simulink and the model code are independent of this layer. Both the transport
layer and code directly interfacing to the transport layer are isolated in
separate modules that format, transmit and receive messages and data
packets.
This design makes it possible for different targets to use different transport
layers. For example, the GRT, GRT malloc, ERT, and Tornado targets support
host/target communication via TCP/IP, whereas the xPC Target supports both
RS232 (serial) and TCP/IP communication.
Real-Time Workshop provides full source code for both the client and
server-side external mode modules, as used by the GRT, GRT malloc, ERT,
rapid simulation, real-time Windows, xPC, and Tornado targets. The main
client-side module is ext_comm.c. The main server-side module is ext_svr.c.
These two modules call the TCP/IP transport layer. ext_transport.c
implements the client-side transport functions. ext_svr_transport.c
14-94
Creating an External Mode Communication Channel
contains the corresponding server-side functions. You can modify these files to
support external mode via your own low-level communications layer.
You need only modify those parts of the code that implement low-level
communications. You need not be concerned with issues such as data
conversions between host and target, or with the formatting of messages. Code
provided by Real-Time Workshop handles these functions.
On the client (Simulink) side, communications are handled by ext_comm, a C
MEX-file. This component is implemented as a DLL on Windows, or as a
shared library on UNIX.
On the server (target) side, external mode modules are linked into the target
executable. This takes place automatically if the External mode code
generation option is selected at code generation time. These modules, called
from the main program and the model execution engine, are independent of the
generated model code.
To implement your own low-level protocol:
• On the client side, you must replace low-level TCP/IP calls in
ext_transport.c with your own communication calls, and rebuild ext_comm
using the mex command. You should then designate your custom ext_comm
component as the MEX-file for external interface in the Simulink External
Target Interface dialog.
• On the server side, you must replace low-level TCP/IP calls in
ext_svr_transport.c with your own communication calls. If you are writing
your own system target file and/or template makefile, make sure that the
EXT_MODE code generation option is defined. The generated makefile will then
link ext_svr_transport.c and other server code into your executable.
• Define symbols and functions common to both the client and server sides in
ext_transport_share.h.
External Mode Communications Overview
This section gives a high-level overview of how a Real-Time Workshop
generated program communicates with Simulink in external mode. This
description is based on the TCP/IP version of external mode that ships with
Real-Time Workshop.
14-95
14
Targeting Real-Time Systems
For communication to take place:
• The server (target) program must have been built with the conditional
EXT_MODE defined. EXT_MODE is defined in the model.mk file if the External
mode code generation option was selected at code generation time.
• Both the server program and Simulink must be executing. Note that this
does not mean that the model code in the server system must be executing.
The server may be waiting for Simulink to issue a command to start model
execution.
The client and server communicate via two sockets. Each socket supports a
distinct channel. The message channel is bidirectional; it carries commands,
responses, and parameter downloads. The unidirectional upload channel is for
uploading signal data to the client. The message channel is given higher
priority.
If the target program was invoked with the -w command line option, the
program enters a wait state until it receives a message from the host.
Otherwise, the program begins execution of the model. While the target
program is in a wait state, Simulink can download parameters to the target
and configure data uploading.
When the user chooses the Connect to target option from the Simulation
menu, the host initiates a handshake by sending an EXT_CONNECT message. The
server responds with information about itself. This information includes:
• Checksums. The host uses model checksums to determine that the target
code is an exact representation of the current Simulink model.
• Data format information. The host uses this information when formatting
data to be downloaded, or interpreting data that has been uploaded.
At this point, host and server are connected. The server is either executing the
model or in the wait state. (In the latter case, the user can begin model
execution by selecting Start real-time code from the Simulation menu.)
During model execution, the message server runs as a background task. This
task receives and processes messages such as parameter downloads.
Data uploading comprises both foreground execution and background servicing
of the upload channel. As the target computes model outputs, it also copies
signal values into data upload buffers. This occurs as part of the task
associated with each task identifier (tid). Therefore, data collection occurs in
14-96
Creating an External Mode Communication Channel
the foreground. Transmission of the collected data, however, occurs as a
background task. The background task sends the data in the collection buffers
to Simulink via the upload channel.
The host initiates most exchanges on the message channel. The target usually
sends a response confirming that it has received and processed the message.
Examples of messages and commands are:
• Connection message / connection response
• Start target simulation / start response
• Parameter download / parameter download response
• Arm trigger for data uploading
• Terminate target simulation / target shutdown response
Model execution terminates when the model reaches its final time, when the
host sends a terminate command, or when a Stop Simulation block terminates
execution. On termination, the server informs the host that model execution
has stopped, and shuts down both sockets. The host also shuts down its sockets,
and exits external mode.
External Mode Source Files
Host (ext_comm) Source Files
The source files for the ext_comm component are located in the directory
matlabroot/rtw/ext_mode:
• ext_comm.c
This file is the core of external mode communication. It acts as a relay station
between the target and Simulink. ext_comm.c communicates to Simulink via
a shared data structure, ExternalSim. It communicates to the target via calls
to the transport layer.
Tasks carried out by ext_comm include establishment of a connection with
the target, downloading of parameters, and termination of the connection
with the target.
• ext_transport.c
This file implements required transport layer functions. (Note that
ext_transport.c includes ext_transport_share.h, which contains
functions common to client and server sides.) The version of
14-97
14
Targeting Real-Time Systems
ext_transport.c shipped with Real-Time Workshop uses TCP/IP functions
including recv(), send(), and socket().
• ext_main.c
This file is a MEX-file wrapper for external mode. ext_main interfaces to
Simulink via the standard mexFunction call. (See “External Interfaces/API”
in the MATLAB online documentation for information on mexFunction.)
ext_main contains a function dispatcher, esGetAction, that sends requests
from Simulink to ext_comm.
• ext_convert.c
This file contains functions used for converting data from host to target
formats (and vice versa). Functions include byte-swapping (big to littleendian), conversion from non-IEEE floats to IEEE doubles, and other
conversions. These functions are called both by ext_comm.c and directly by
Simulink (via function pointers).
Note You do not need to customize ext_convert in order to implement a
custom transport layer. However, it may be necessary to customize
ext_convert for the intended target. For example, if the target represents the
float data type in Texas Instruments (TI) format, ext_convert must be
modified to perform a TI to IEEE conversion.
• extsim.h
This file defines the ExternalSim data structure and access macros. This
structure is used for communication between Simulink and ext_comm.c.
• extutil.h
This file contains only conditionals for compilation of the assert macro.
Target (Server) Source Files
These files are part of the run-time interface and are linked into the model.exe
executable. They are located in the directory matlabroot/rtw/c/src.
• ext_svr.c
ext_svr.c is analogous to ext_comm.c on the host, but generally is
responsible for more tasks. It acts as a relay station between the host and the
generated code. Like ext_comm.c, ext_svr.c carries out tasks such as
establishing and terminating connection with the host. ext_svr.c also
14-98
Creating an External Mode Communication Channel
contains the background task functions that either write downloaded
parameters to the target model, or extract data from the target data buffers
and send it back to the host.
The version of ext_svr.c shipped with Real-Time Workshop uses TCP/IP
functions including recv(), send(), and socket().
• ext_svr_transport.c
This file implements required transport layer functions. (Note that
ext_svr_transport.c includes ext_transport_share.h, which contains
functions common to client and server sides.) The version of
ext_svr_transport.c shipped with Real-Time Workshop uses TCP/IP
functions including recv(), send(), and socket().
• updown.c
updown.c handles the details of interacting with the target model. During
parameter downloads, updown.c does the work of installing the new
parameters into the model’s parameter vector. For data uploading, updown.c
contains the functions that extract data from the model’s blockio vector and
write the data to the upload buffers. updown.c provides services both to
ext_svr.c and to the model code (e.g., grt_main.c). It contains code that is
called via the background tasks of ext_svr.c as well as code that is called as
part of the higher priority model execution.
• dt_info.h and model.dt
These files contain data type transition information that allows access to
multi-data type structures across different computer architectures. This
information is used in data conversions between host and target formats.
• updown_util.h
This file contains only conditionals for compilation of the assert macro.
Other Files
• ext_share.h
Contains message code definitions and other definitions required by both the
host and target modules.
• ext_transport_share.h
Contains functions and data structures required by both the host and target
modules of the transport layer. The version of ext_transport_share.h
shipped with Real-Time Workshop is specific to TCP/IP communications.
14-99
14
Targeting Real-Time Systems
Guidelines for Implementing the Transport Layer
Requirements
• ext_svr.c and updown.c use malloc to allocate buffers in target memory for
messages, data collection, and other purposes. If your target uses some other
memory allocation scheme, you must modify these modules appropriately.
• The target is assumed to support both int32_T and uint32_T data types.
Modifying ext_transport
The function prototypes in ext_transport.h define the calling interface for the
host (client) side transport layer functions. The implementation is in
ext_transport.c.
To implement the host side of your transport layer:
• Replace the functions in the “Visible Functions” section of ext_transport.c
with functions that call your low-level communications primitives. The
visible functions are called from other external mode modules such as
ext_comm.c. You must implement all the functions defined in
ext_transport.h. Your implementations must conform to the prototypes
defined in ext_transport.h.
• Supply a definition for the UserData structure in ext_transport.c. This
structure is required. If UserData is not necessary for your external mode
implementation, define a UserData structure with one dummy field.
• Replace the functions in ext_transport_share.h with functions that call
your low-level communications primitives, or remove these functions.
Functions defined in ext_transport_share.h are common to the host and
target, and are not part of the public interface to the transport layer.
• Rebuild the ext_comm MEX-file, using the MATLAB mex command. This
requires a compiler supported by the MATLAB API. See “External
Interfaces/API” in the MATLAB online documentation for more information
14-100
Creating an External Mode Communication Channel
on the mex command. The following table lists the form of the commands to
build the standard ext_comm module on PC and UNIX platforms.
Table 14-5: Commands to Rebuild ext_comm MEX-Files
Platform
Commands
PC
cd matlabroot\toolbox\rtw
mex matlabroot\rtw\ext_mode\ext_comm.c
matlabroot\rtw\ext_mode\ext_convert.c
matlabroot\rtw\ext_mode\ext_transport.c
–Imatlab\rtw\c\src –DWIN32
compiler_library_path\wsock32.lib
UNIX
cd matlabroot/toolbox/rtw
mex matlabroot/rtw/ext_mode/ext_comm.c
matlabroot/rtw/ext_mode/ext_convert.c
matlabroot/rtw/ext_mode/ext_transport.c
–Imatlab/rtw/c/src
The ext_transport and ext_transport_share source code modules are fully
commented. See these files for further details.
Guidelines for Modifying ext_svr_transport
The function prototypes in ext_svr_transport.h define the calling interface
for the target (server) side transport layer functions. The implementation is in
ext_svr_transport.c.
To implement the target side of your transport layer:
• Replace the functions in ext_svr_transport.c with functions that call your
low-level communications primitives. These are the functions called from
other target modules such as the main program. You must implement all the
functions defined in ext_svr_transport.h. Your implementations must
conform to the prototypes defined in ext_svr_transport.h.
• Supply a definition for the ExtUserData structure in ext_svr_transport.c.
This structure is required. If ExtUserData is not necessary for your external
mode implementation, define an ExtUserData structure with one dummy
field.
• Define the EXT_BLOCKING conditional in ext_svr_transport.c as needed:
14-101
14
Targeting Real-Time Systems
- Define EXT_BLOCKING as 0 to poll for a connection to the host (appropriate
for single-threaded applications).
- Define EXT_BLOCKING as 1 in multi-threaded applications where tasks are
able to block for a connection to the host without blocking the entire
program.
See also the comments on EXT_BLOCKING in ext_svr_transport.c.
The ext_svr_transport source code modules are fully commented. See these
files for further details.
14-102
Combining Multiple Models
Combining Multiple Models
If you want to combine several models (or several instances of the same model)
into a single executable, Real-Time Workshop offers several options.
One solution is to use the S-function target to combine the models into a single
model, and then generate an executable using either the GRT or GRT malloc
targets. Simulink and Real-Time workshop implicitly handle connections
between models, sequencing of calls to the models, and multiple sample rates.
This is the simplest solution in many cases. See Chapter 10, “The S-Function
Target” for further information.
A second option, for embedded systems development, is to generate code from
your models using the Real-Time Workshop Embedded Coder target. You can
interface the model code to a common harness program by directly calling the
entry points to each model. The Real-Time Workshop Embedded Coder target
has certain restrictions that may not be appropriate for your application. For
more information, see the Real-Time Workshop Embedded Coder
documentation.
The GRT malloc target is a third solution. It is appropriate in situations where
you want do any or all of the following:
• Selectively control calls to more than one model.
• Use dynamic memory allocation.
• Include models that employ continuous states.
• Log data to multiple files.
• Run one of the models in external mode.
This section discusses how to use the GRT malloc target to combine models into
a single program. Before reading this section, you should become familiar with
model execution in Real-Time Workshop programs. (See Chapter 7, “Program
Architecture” and Chapter 8, “Models with Multiple Sample Rates.”) It will be
helpful to refer to grt_malloc_main.c while reading these chapters.
The procedure for building a multiple-model executable is fairly
straightforward. The first step is to generate and compile code from each of the
models that are to be combined. Next, the makefiles for each of the models are
combined into one makefile for creating the final multimodel executable. The
next step is create a combined simulation engine by modifying
grt_malloc_main.c to initialize and call the models correctly. Finally, the
14-103
14
Targeting Real-Time Systems
combination makefile links the object files from the models and the main
program into an executable. “Example Mutliple-Model Program Using the
GRT_malloc Target” on page 14-105 discusses an example implementation.
Sharing Data Across Models
We recommend using unidirectional signal connections between models. This
affects the order in which models are called. For example, if an output signal
from modelA is used as input to modelB, modelA’s output computation should
be called first.
Timing Issues
You must generate all the models you are combining with the same solver mode
(either all singletasking or all multitasking.) In addition, if the models employ
continuous states, the same solver should be used for all models.
Since each model has its own model-specific definition of the rtModel data
structure, an alternative mechanism must be used to control model execution.
The file rtw/c/src/rtmcmacros.h provides an rtModel API clue that can be
used to call the rt_OneStep procedure. The rtmcmacros.h header file defines
the rtModelCommon data structure which has the minimum common elements
in the rtModel structure required to step a model forward one time step. The
#define rtmcsetCommon populates an object of type rtModelCommon by copying
the respective similar elements in the model's rtModel object. Your main
routine must create one rtModelCommon structure for each model being called
by the main routine. The main routine will subsequently invoke rt_OneStep
with a pointer to the rtModelCommon structure instead of a pointer to the
rtModel structure.
If the base rates for the models are not the same, the main program (such as
grt_malloc_main) must set up the timer interrupt to occur at the greatest
common divisor rate of the models. The main program is responsible for calling
each of the models at the appropriate time interval.
Data Logging and External Mode Support
A multiple-model program can log data to separate MAT-files for each model
(as in the example program discussed below.)
Only one of the models in a multiple-model program can use external mode.
14-104
Combining Multiple Models
Example Mutliple-Model Program Using the GRT_malloc Target
An demonstration of combining multiple-models, distributed with Real-Time
Workshop, is located at matlabroot/toolbox/rtw/rtwdemos. This example
combines two models, multimallockP (a plant model) and multimallocK (a
controller model). Both models have the same base rate and the same number
of sample times. Each model logs outputs and simulation time to a separate
model.mat file. The plant model also logs states. You can run the demo by
typing
multimallocdemo
at the MATLAB prompt. The interface for multimallocdemo is shown below.
Double-click the models to see each system’s blocks and how they work
together. Double-click on the blue labels in order from top to bottom to generate
14-105
14
Targeting Real-Time Systems
code, and see how main programs and makefiles were modified to combine the
two models.
When reviewing the differences between grt_malloc_main.c and
combine_malloc_main.c, search for comments containing “customize”. The
string “customize” denotes regions in the main routine which you must change
in order to customize this file to work with your models.
14-106
DSP Processor Support
DSP Processor Support
Real-Time Workshop now supports target processors that have only one
register size (e.g., 32-bit). This makes data type emulation of 8 and 16 bits on
the TCI_C30/C40 DSP and similar processors possible.
To support these processors:
• Add the command
-DDSP32=1
to your template makefile.
• Add the statement
%assign DSP32=1
to your system target file.
For DSP Blockset Users
Note that previous releases of the DSP Blockset did not fully support Simulink
Accelerator, Generic real-time malloc (GRT malloc), and S-function targets.
The current DSP Blockset supports code generation for all packaged targets.
14-107
14
Targeting Real-Time Systems
14-108
A
Glossary
A
Glossary
Application modules — With respect to Real-Time Workshop program
architecture, these are collections of programs that implement functions
carried out by the system dependent, system independent, and application
components.
Atomic subsystem — A subsystem whose blocks are executed as a unit before
moving on. Conditionally executed subsystems are atomic, and atomic
subsystems are nonvirtual. Unconditionally executed subsystems are virtual
by default, but can be designated as atomic. Real-Time Workshop can generate
reusable code only for nonvirtual subsystems.
Block target file — A file that describes how a specific Simulink block is to be
transformed to a language such as C, based on the block’s description in the
Real-Time Workshop file (model.rtw). Typically, there is one block target file
for each Simulink block.
Code reuse — An optimization whereby code generated for identical
nonvirtual subsystems is collapsed into one function that is called for each
subsystem instance with appropriate parameters. Code reuse, along with
expression folding, can dramatically reduce the amount of generated code.
Embedded Real-Time (ERT) target − A target configuration that generates
model code for execution on an independent embedded real-time system.
Requires Real-Time Workshop Embedded Coder.
Expression folding — A code optimization technique that minimizes the
computation of intermediate results at block outputs and the storage of such
results in temporary buffers or variables. It can dramatically improve the
efficiency of generated code, achieving results that compare favorably to
hand-optimized code.
File extensions — The table below lists the file extensions associated with
Simulink, the Target Language Compiler, and Real-Time Workshop.
A-2
Extension
Created by
Description
.c
Target Language
Compiler
The generated C code
.h
Target Language
Compiler
A C include header file used by the .c
program
Extension
Created by
Description
.mdl
Simulink
Contains structures associated with
Simulink block diagrams
.mk
Real-Time Workshop
A makefile specific to your model that
is derived from the template makefile
.rtw
Real-Time Workshop
An intermediate compilation
(“model.rtw”) of a .mdl file used in
generating C code
.tlc
The MathWorks and
Real-Time Workshop
users
Target Language Compiler script
files that Real-Time Workshop uses
to generate code for targets and
blocks
.tmf
Supplied with
Real-Time Workshop
Template makefiles
.tmw
Real-Time Workshop
A project marker file inside a build
directory that identifies the date and
product version of generated code
Generic Real-Time (GRT) target — A target configuration that generates
model code for a real-time system, with the resulting code executed on your
workstation. (Note that execution is not tied to a real-time clock.) You can use
GRT as a starting point for targeting custom hardware.
Host system — The computer system on which you create and may compile
your real-time application.
Inline — Generally, this means to place something directly in the generated
source code. You can inline parameters and S-functions using Real-Time
Workshop.
Inlined parameters (Target Language Compiler Boolean global variable:
InlineParameters) — The numerical values of the block parameters are hard
coded into the generated code. Advantages include faster execution and less
memory use, but you lose the ability to change the block parameter values at
run-time.
A-3
A
Glossary
Inlined S-function — An S-function can be inlined into the generated code by
implementing it as a .tlc file. The code for this S-function is placed in the
generated model code itself. In contrast, noninlined S-functions require a
function call to S-function residing in an external MEX-file.
Interrupt Service Routine (ISR) — A piece of code that your processor
executes when an external event, such as a timer, occurs.
Loop rolling (Target Language Compiler global variable: RollThreshold) —
Depending on the block's operation and the width of the input/output ports, the
generated code uses a for statement (rolled code) instead of repeating identical
lines of code (flat code) over the signal width.
Make — A utility to maintain, update, and regenerate related programs and
files. The commands to be executed are placed in a makefile.
Makefiles — Files that contain a collection of commands that allow groups of
programs, object files, libraries, etc., to interact. Makefiles are executed by your
development system’s make utility.
Multitasking — A process by which your microprocessor schedules the
handling of multiple tasks. The number of tasks is equal to the number of
sample times in your model.
Noninlined S-function — In the context of Real-Time Workshop, this is any C
MEX S-function that is not implemented using a customized .tlc file. If you
create an C MEX S-function as part of a Simulink model, it is by default
noninlined unless you write your own .tlc file that inlines it.
Nonreal-time — A simulation environment of a block diagram provided for
high-speed simulation of your model. Execution is not tied to a real-time clock.
Nonvirtual block — Any block that performs some algorithm, such as a Gain
block. Real-Time Workshop generates code for all nonvirtual blocks, either
inline or as separate functions and files, as directed by users.
Pseudomultitasking — n processors that do not offer multitasking support,
you can perform pseudomultitasking by scheduling events on a fixed
time-sharing basis.
Real-time model data structure — Real-Time Workshop encapsulates
information about the root model in the real-time model data structure, often
abbreviated as rtM. rtM contains global information related to timing, solvers,
and logging, and model data such as inputs, outputs, states, and parameters.
A-4
Real-time system — A computer that processes real-world events as they
happen, under the constraint of a real-time clock, and which may implement
algorithms in dedicated hardware. Examples include mobile telephones, test
and measurement devices, and avionic and automotive control systems.
Run-time interface — A wrapper around the generated code that can be built
into a stand-alone executable. The run-time interface consists of routines to
move the time forward, save logged variables at the appropriate time steps, etc.
The run-time interface is responsible for managing the execution of the
real-time program created from your Simulink block diagram.
S-function — A customized Simulink block written in C, Fortran, or M-code.
Real-Time Workshop can target C-code S-functions “as is” or users can inline
C-code S-functions through preparing TLC scripts for them.
Simstruct — A Simulink data structure and associated application
programming interface (API) that enables S-functions to communicate with
other entities in models. Simstructs are included in code generated by
Real-Time Workshop for noninlined S-functions.
Singletasking — A mode in which a model runs in one task.
System target file — The entry point to the Target Language Compiler
program, used to transform the Real-Time Workshop file into target specific
code.
Target file — A file that is compiled and executed by the Target Language
Compiler. The block and system target TLC files used specify how to transform
the Real-Time Workshop file (model.rtw) into target-specific code.
Targeting — The process of creating software modules appropriate for
execution on your target system.
Target Language Compiler (TLC) — A program that compiles and executes
system and target files by translating a model.rtw file into a target language
by means of TLC scripts and template makefiles.
Target Language Compiler program — One or more TLC script files that
describe how to convert a model.rtw file into generated code. There is one TLC
file for the target, plus one for each built-in block. Users can provide their own
TLC files in order to inline S-functions or to “wrap” existing user code.
Target system — The specific or generic computer system on which your
real-time application executes.
A-5
A
Glossary
Template makefile — A line-for-line makefile used by a make utility. The
template makefile is converted to a makefile by copying the contents of the
template makefile (usually system.tmf) to a makefile (usually system.mk)
replacing tokens describing your model’s configuration.
Task identifier (tid) — In generated code, each sample rate in a multirate
model is assigned a task identifier (tid). The tid is passed to the model output
and update routines to control which portion of your model should execute at a
given time step. Single-rate systems ignore the tid.
Virtual block — A connection or graphical block, for example a Mux block, that
has no algorithmic functionality. Virtual blocks incur no real-time overhead as
no code is generated for them.
A-6
B
Blocks That Depend on
Absolute Time
Some Simulink blocks use the value of absolute time (i.e., the time from the
beginning of the program to the present time) to calculate their outputs. If you
are designing a program that is intended to run indefinitely, then you cannot
use blocks that have a dependency on absolute time.
The problem arises when the value of time reaches the largest value that can
be represented by a double precision number. At that point, time is no longer
incremented and the output of the block is no longer correct.
Note In addition to the blocks listed below, logging Time (in the Workspace
I/O page of the Simulation Parameters dialog box) also requires absolute
time.
The following Simulink blocks depend on absolute time:
• Continuous Blocks
- Derivative
- Transport Delay
- Variable Transport Delay
• Discrete Blocks
- Discrete-Time Integrator (when used in triggered subsystems)
• Nonlinear Blocks
- Rate Limiter
• Sinks
- Scope
- To File
- To Workspace (only if logging to StructureWithTime format)
• Sources
B
Blocks That Depend on Absolute Time
The following Simulink blocks depend on absolute time:
• Chirp Signal
• Clock
• Digital Clock
• From File
• From Workspace
• Pulse Generator
• Ramp
• Repeating Sequence
• Signal Generator
• SineWave
• Step
Note The Sine Wave block is dependent on absolute time only when the Sine
type parameter is set to Time-based. Set this parameter to Sample-based to
avoid absolute time computations.
In addition to the Simulink block above:
• Blocks in other Blocksets may reference absolute time. Please refer to the
documentation for the Blocksets that you use.
• Stateflow charts that use time are dependent on absolute time.
B-2
C
Targeting DOS for
Real-Time Applications
The following sections describe the DOS target, which developments in Microsoft Windows and other
technologies have rendered oboslete. The DOS target is currently unsupported, and the information
provided here is for demonstration purposes only, particularly as a guide for developing device drivers
for other Real-Time Workshop targets. We cover the following DOS target topics:
DOS Target Basics (p. C-2)
Gaining an overview of the DOS target
Implementation Overview (p. C-4)
Understanding code architecture and hardware/software
requirements
Device Driver Blocks (p. C-10)
Using the DOS Device Drivers library
Building the Program (p. C-18)
Building the executable and running it under DOS
C
Targeting DOS for Real-Time Applications
DOS Target Basics
The discussion that follows describes using Real-Time Workshop in a DOS
environment. Advances in computing technology have resulted in DOS-based
systems being replaced by a variety of alternative platforms. Particularly, the
xPC Target and the Real-Time Windows Target provide significantly greater
capabilities than does the DOS target. We recommend use of these targets for
real-time development on PC platforms. For detailed information, see the
Real-Time Windows Documentation and the xPC documentation.
Note The DOS target is provided only as an unsupported example. Also, note
that the DOS target requires the Watcom C compiler. See “A Note on the
Watcom Compiler” on page C-6.
This chapter includes a discussion of:
• DOS-based Real-Time Workshop applications
• Supported compilers and development tools
• Device driver blocks — adding them to your model and configuring them for
use with your hardware
• Building the program
The DOS target creates an executable for DOS, using Watcom for DOS. The
executable runs on a computer running the DOS operating system. It will not
run under the Microsoft Windows DOS command prompt. This executable
installs interrupt service routines and effectively takes over the computer,
which allows the generated code to run in real time. If you want to run the
generated code in real time under Microsoft Windows, you should use the
Real-Time Windows Target. See the Real-Time Windows Target User’s Guide
for more information about this product.
DOS Device Drivers Library
Real-Time Workshop provides DOS-compatible analog and digital I/O device
driver blocks in the DOS Device Drivers library. Select DOS Device Drivers
under the Real-Time Workshop library in the Simulink Library Browser to
open the DOS Device Drivers library.
C-2
DOS Target Basics
C-3
C
Targeting DOS for Real-Time Applications
Implementation Overview
Real-Time Workshop includes DOS run-time interface modules designed to
implement programs that execute in real time under DOS. These modules,
when linked with the code generated from a Simulink model, build a complete
program that is capable of executing the model in real time. The DOS run-time
interface files can be found in the matlabroot/rtw/c/dos/rti directory.
Real-Time Workshop DOS run-time interface modules and the generated code
for the f14 demonstration model are shown in Figure C-1.
Generated Code
f14.c
f14.h
Template
Makefile
Timer
drt_time.c
Keyboard
drt_key.c
Interrupt
Support
drt_cpu.c
drt_cpu.h
drt_fpu.asm
Main Program
drt_watc.tmf
drt_main.c
Model
Execution
rt_sim.c
Makefile
f14.mk
Executable File
f14[.exe]
Integration
Module
ode5.c
Data Logger
rtwlog.c
Simulink
Data Structure
simstruc.h
Figure C-1: Source Modules Used to Build the DOS Real-Time Program
C-4
Implementation Overview
This diagram illustrates the code modules that are used to build a DOS
real-time program.
To execute the code in real time, the program runs under the control of an
interrupt driven timing mechanism. The program installs its own interrupt
service routine (ISR) to execute the model code periodically at predefined
sample intervals. The PC-AT’s 8254 Programmable Interval Timer is used to
time these intervals.
In addition to the modules shown in Figure C-1, the DOS run-time interface
also consists of device driver modules to read from and write to I/O devices
installed on the DOS target.
Figure C-2 shows the recommended hardware setup for designing control
systems using Simulink, and then building them into DOS real-time
applications using Real-Time Workshop. The figure shows a robotic arm being
controlled by a program (i.e., the controller) executing on the target PC. The
controller senses the arm position and applies inputs to the motors accordingly,
via the I/O devices on the target PC. The controller code executes on the PC and
communicates with the apparatus it controls via I/O hardware.
Host Workstation PC
Target PC
Running Windows
with MATLAB, Simulink
and Real-Time Workshop
DOS
Executing
Real-time
Program
I/O Devices
A/D
D/A
Motor
Drive
Control
Position
Sensor
Output
Figure C-2: Typical Hardware Setup
System Configuration
You can use Real-Time Workshop with a variety of system configurations, as
long as these systems meet the following hardware and software requirements.
C-5
C
Targeting DOS for Real-Time Applications
Hardware Requirements
The hardware needed to develop and run a real-time program includes:
• A workstation running Windows and capable of running MATLAB/Simulink.
This workstation is the host where the real-time program is built.
• A PC-AT (386 or later) running DOS. This system is the target, where the
real-time program executes.
• I/O boards, which include analog to digital converter and digital to analog
converters (collectively referred to as I/O devices), on the target.
• Electrical connections from the I/O devices to the apparatus you want to
control (or to use as inputs and outputs to the program in the case of
hardware-in-the-loop simulations).
Once built, you can run the executable on the target hardware as a stand-alone
program that is independent of Simulink.
Software Requirements
The development host must have the following software:
• MATLAB and Simulink to develop the model, and Real-Time Workshop to
create the code for the model. You also need the run-time interface modules
included with Real-Time Workshop. These modules contain the code that
handles timing, interrupts, data logging, and background tasks.
• Watcom C/C++ compiler, Version 10.6 or 11.0. (see “A Note on the Watcom
Compiler” below.)
The target PC must have the following software:
• DOS4GW extender dos4gw.exe, included with your Watcom compiler
package) must be on the search path on the DOS-targeted PC.
You can compile the generated code (i.e., the files model.c, model.h, etc.) along
with user-written code using other compilers. However, the use of 16-bit
compilers is not recommended for any application.
A Note on the Watcom Compiler
As of this writing, the Watcom C compiler is no longer available from the
manufacturer. Real-Time Workshop continues to ship Watcom-related target
C-6
Implementation Overview
configurations at this time. However, this policy may be subject to change in
the future.
Device Drivers
If your application needs to access its I/O devices on the target, then the
real-time program must contain device driver code to handle communication
with the I/O boards. The Real-Time Workshop DOS run-time interface includes
source code of the device drivers for the Keithley Metrabyte DAS-1600/1400
Series I/O boards. See “Device Driver Blocks” on page C-10 for information on
how to use these blocks.
Simulink Host
The development host must have Windows to run Simulink. However, the
real-time target requires only DOS, since the executable built from the
generated code is not a Windows application. The real-time target will not run
in a “DOS box” (i.e., a DOS window on Windows 95/98/NT).
Although it is possible to reboot the host PC under DOS for real-time execution,
the computer would need to be rebooted under Windows for any subsequent
changes to the block diagram in Simulink. Since this process of repeated
rebooting the computer is inconvenient, we recommend a second PC running
only DOS as the real-time target.
Sample Rate Limits
Program timing is controlled by installing an interrupt service routine that
executes the model code. The target PC’s CPU is then interrupted at the
specified rate (this rate is determined from the step size).
The rate at which interrupts occur is controlled by application code supplied
with Real-Time Workshop. This code uses the PC-AT’s 8254 Counter/Timer to
determine when to generate interrupts.
The code that sets up the 8254 Timer is in drt_time.c, which is in the
matlabroot\rtw\c\dos\rti directory. It is automatically linked in when you
build the program using the DOS real-time template makefile.
The 8254 Timer is a 16-bit counter that operates at a frequency of 1.193 MHz.
However, the timing module drt_time.c in the DOS run-time interface can
extend the range by an additional 16 bits in software, effectively yielding a
C-7
C
Targeting DOS for Real-Time Applications
32-bit counter. This means that the slowest base sample rate your model can
have is
6
1
3600
1.193 ×10 ÷ ( 2 32 – 1 ) ≈ ------------- Hz
This corresponds to a maximum base step size of approximately one hour.
The fastest sample rate you can define is determined by the minimum value
from which the counter can count down. This value is 3, hence the fastest
sample rate that the 8254 is capable of achieving is
1.193 × 10 6 ÷ 3 ≈ 4 × 10 5 Hz
This corresponds to a minimum base step size of
1 ÷ 4 × 10 5 ≈ 2.5 × 10 – 6 sec onds
However, note that the above number corresponds to the fastest rate at which
the timer can generate interrupts. It does not account for execution time for the
model code, which would substantially reduce the fastest sample rate possible
for the model to execute in real time. Execution speed is machine dependent
and varies with the type of processor and the clock rate of the processor on the
target PC.
The slowest and fastest rates computed above refer to the base sample times in
the model. In a model with more than one sample time, you can define blocks
that execute at slower rates as long as the sample times are an integer multiple
of the base sample time.
Modifying Program Timing
If you have access to an alternate timer (e.g., some I/O boards include their own
clock devices), you can replace the file drt_time.c with an equivalent file that
makes use of the separate clock source. See the comments in drt_time.c to
understand how the code works.
You can use your version of the timer module by redefining the TIMER_OBJS
macros with the build command. For example, in the Real-Time Workshop
pane of the Simulation parameters dialog box, changing the build command
to
make_rtw TIMER_OBJS=my_timer.obj
C-8
Implementation Overview
replaces the file drt_time.c with my_timer.c in the list of source files used to
build the program.
C-9
C
Targeting DOS for Real-Time Applications
Device Driver Blocks
The real-time program communicates with external hardware via a set of
device drivers. These device drivers contain the necessary code for interfacing
to specific I/O devices.
Real-Time Workshop includes device drivers for commercially available
Keithley Metrabyte DAS-1600/1400 Series I/O boards. These device drivers are
implemented as C MEX S-functions to interface with Simulink. This means
you can add them to your model like any other block.
In addition, each of these S-function device drivers has a corresponding target
file to inline the device driver in the model code. See “Creating Device Drivers”
on page 14-39 for information on implementing your own device drivers.
Since the device drivers are provided as source code, you can use these device
drivers as a template to serve a a starting point for creating custom device
drivers for other I/O boards.
Device Driver Block Library
The device driver blocks for the Keithley Metrabyte DAS-1600/1400 Series I/O
boards designed for use with DOS applications are contained in a block library
called doslib (matlabroot\toolbox\rtw\doslib.mdl). To display this library,
type
doslib
at the MATLAB prompt. This window will appear.
C-10
Device Driver Blocks
To access the device driver blocks, double-click on the sublibrary icon.
The blocks in the library contain device drivers that can be used for the
DAS-1600/1400 Series I/O boards. The DAS-1601/1602 boards have 16 analog
input (ADC) channels, two 12-bit analog output (DAC) channels and 4-bits of
digital I/O. The DAS-1401/1402 boards do not have DAC channels. The
DAS-1601/1401 boards have high programmable gains (1, 10, 100 and 500),
while the DAS-1602/1402 boards offer low programmable gains (1, 2, 4 and 8).
For more information, contact the manufacturer via the Web site:
http://www.keithley.com. The documentation for the DAS-1600/1400 Series
I/O boards is the DAS-1600/1400 Series User’s Guide, Revision B (Part
Number: 80940).
Configuring Device Driver Blocks
Each device driver block has a dialog box that you use to set configuration
parameters. As with all Simulink blocks, double-clicking on the block displays
the dialog box. Some of the device driver block parameters (such as Base I/O
Address) are hardware specific and are set either at the factory or configured
via DIP switches at the time of installation.
C-11
C
Targeting DOS for Real-Time Applications
Analog Input (ADC) Block Parameters
• Base I/O Address: The beginning of the I/O address space assigned to the
board. The value specified here must match the board’s configuration. Note
that this parameter is a hexadecimal number and must be entered in the
dialog as a MATLAB string (e.g., '0x300').
• Analog Input Range: This two-element vector specifies the range of values
supported by the ADC. The specified range must match the
I/O board’s settings. Specifically, the DAS-1600/1400 Series boards switches
can be configured to either [0 10] for unipolar or [-10 10] for bipolar input
signals.
• Hardware Gain: This parameter specifies the programmable gain that is
applied to the input signal before presenting it to the ADC. Specifically, the
DAS-1601/1401 boards have programmable gains of 1, 10, 100, and 500. The
DAS-1602/1402 boards have programmable gains of 1, 2, 4, and 8. Configure
the Analog Input Range and the Hardware Gain parameters depending on
the type and range of the input signal being measured. For example, a
DAS-1601 board in bipolar configuration with a programmable gain of 100 is
C-12
Device Driver Blocks
best suited to measure input signals in the range between [±10v] ÷ 100 =
±0.1v.
Voltage levels beyond this range will saturate the block output form the ADC
block. Please adhere to manufacturers’ electrical specifications to avoid
damage to the board.
• Number of Channels: The number of analog input channels enabled on the
I/O board. The DAS-1600/1400 Series boards offer up to 16 ADC channels
when configured in unipolar mode (8 ADC channels if you select differential
mode). The output port width of the ADC block is equal to the number of
channels enabled.
• Sample Time (sec): Device drivers are discrete blocks that require you to
specify a sample time. In the generated code, these blocks are executed at the
specified rate. Specifically, when the ADC block is executed, it causes the
ADC to perform a single conversion on the enabled channels, and the
converted values are written to the block output vector.
C-13
C
Targeting DOS for Real-Time Applications
Analog Output (DAC) Block Parameters
• Base I/O Address: The beginning of the I/O address space assigned to the
board. The value specified here must match the board’s configuration. Note
that this parameter is a hexadecimal number and must be entered in the
dialog as a MATLAB string (e.g., '0x300').
• Analog Output Range: This parameter specifies the output range settings
of the DAC section of the I/O board. Typically, unipolar ranges are between
[0 10] volts and bipolar ranges are between [-10 10] volts. Refer to the
DAS-1600 documentation for other supported output ranges.
• Initial Output(s): This parameter can be specified either as a scalar or as
an N element vector, where N is the number of channels. If a single scalar
value is entered, the same scalar is applied to output. The specified initial
output(s) is written to the DAC channels in the mdlInitializeConditions
function.
• Final Output(s): This parameter is specified in a manner similar to the
Initial Output(s) parameter except that the specified final output values are
written out to the DAC channels in the mdlTerminate function. Once the
C-14
Device Driver Blocks
generated code completes execution, the code sets the final output values
prior to terminating execution.
• Number of Channels: Number of DAC channels enabled. The DAS-1600
Series I/O boards have two 12-bit DAC channels. The DAS-1400 Series I/O
boards do not have any DAC channels. The input port width of this block is
equal to the number of channels enabled.
• Sample Time (sec): DAC device drivers are discrete blocks that require you
to specify a sample time. In the generated code, these blocks are executed at
the specified rate. Specifically, when the DAC block is executed, it causes the
DAC to convert a single value on each of the enabled DAC channels, which
produces a corresponding voltage on the DAC output pin(s).
Digital Input Block Parameters
• Base I/O Address: The beginning of the I/O address space assigned to the
board. The value specified here must match the board’s configuration. Note
that this parameter is a hexadecimal number and must be entered in the
dialog as a MATLAB string (e.g., '0x300').
• Number of Channels: This parameter specifies the number of 1-bit digital
input channels being enabled. This parameter also determines the output
port width of the block in Simulink. Specifically, the DAS-1600/1400 Series
boards provide four bits (i.e., channels) for digital I/O.
C-15
C
Targeting DOS for Real-Time Applications
• Sample Time (sec): Digital input device drivers are discrete blocks that
require you to specify a sample time. In the generated code, these blocks are
executed at the specified rate. Specifically, when the digital input block is
executed, it reads a boolean value from the enabled digital input channels.
The corresponding input values are written to the block output vector.
Digital Output Block Parameters
• Base I/O Address: The beginning of the I/O address space assigned to the
board. The value specified here must match the board’s configuration. Note
that this parameter is a hexadecimal number and must be entered in the
dialog as a MATLAB string (e.g., '0x300').
• Low/High Threshold Values: This parameter specifies the threshold
levels, [lo hi], for converting the block inputs into 0/1 digital values. The
signal in the block diagram connected to the block input should rise above
the high threshold level for a 0 to 1 transition in the corresponding digital
output channel on the I/O board. Similarly, the input should fall below the
low threshold level for a 1 to 0 transition.
C-16
Device Driver Blocks
• Initial Output(s): Same as the Analog Output block, except the specified
values are converted to 0 or 1 based on the lower threshold value before they
are written to the corresponding digital output channel.
• Final Output(s): Same as the Analog Output block, except the specified
values are converted to 0 or 1 based on the lower threshold value before they
are written to the corresponding digital output channel on the I/O board.
• Number of Channels: This parameter specifies the number of 1-bit digital
I/O channels being enabled. This parameter also determines the output port
width of the block. Specifically, the DAS-1600/1400 Series boards provide
four bits (i.e., channels) for digital I/O.
• Sample Time (sec): Digital output device drivers are discrete blocks that
require you to specify a sample time. In the generated code, these blocks are
executed at the specified rate. Specifically, when the digital output block is
executed, it causes corresponding boolean values to be output from the
board’s digital I/O channels.
Adding Device Driver Blocks to the Model
Add device driver blocks to the Simulink block diagram as you would any other
block — simply drag the block from the block library and insert it into the
model. Connect the ADC or Digital Input block to the model’s inputs and
connect the DAC or Digital Output block to the model’s outputs.
Including Device Driver Code
Device driver blocks are implemented as S-functions written in C. The C code
for a device driver block is compiled as a MEX-file so that it can be called by
Simulink. See “External Interfaces/API” in the MATLAB online
documentation for information on MEX-files.
The same C code can also be compiled and linked to the generated code just like
any other C-coded, S-function. However, by using the target (.tlc) file that
corresponds to each of the C file S-functions, the device driver code is inlined in
the generated code.
The matlabroot\rtw\c\dos\devices directory contains the MEX-files, C files,
and target files (.tlc) for the device driver blocks included in doslib. This
directory is automatically added to your MATLAB path when you include any
of the blocks from doslib in your model.
C-17
C
Targeting DOS for Real-Time Applications
Building the Program
Once you have created your Simulink model and added the appropriate device
driver blocks, you are ready to build a DOS target application. To do this, select
the Real-Time Workshop pane of the Simulink parameters dialog box, and
select Target configuration from the Category menu.
Click Browse to open the System Target File Browser. Select drt.tlc; this
automatically fills in the correct files as shown above:
• drt.tlc as the System target file
• drt_watc.tmf as the Template makefile. This is used with the Watcom
compiler, assembler, linker, and WMAKE utility.
• make_rtw as the Make command
You can specify Target Language Compiler options in the System target file
field following drt.tlc. You can also specify and make options in the Make
command field. See Chapter 2, “Code Generation and the Build Process” for
descriptions of the available Target Language Compiler and make options.
The DOS system target file, drt.tlc, and the template makefile,
drt_watc.tmf, are located in the matlab\rtw\c\dos directory.
The template makefile assumes that the Watcom C/386 compiler, assembler,
and linker have been correctly installed on the host workstation. You can verify
C-18
Building the Program
this by checking the environment variable, WATCOM, which correctly points to
the directory where the Watcom files are installed.
The program builder invokes the Watcom wmake utility on the generated
makefile, so the directory where wmake is installed must be on your path.
Running the Program
The result of the build process is a DOS 32-bit protected-mode executable. The
default name of the executable is model.exe, where model is the name of your
Simulink model. You must run this executable in DOS; you cannot run the
executable in Windows.
C-19
C
Targeting DOS for Real-Time Applications
C-20
D
The Real-Time Workshop
Development Process
The following sections summarize the capabilities of Real-Time Workshop from a software
development perspective, discussing, among other topics, its code generation architecture, key
features and benefits, target environments supported, and code optimization features.
Introduction (p. D-2)
What Real-Time Workshop provides
A Next-Generation Development Tool
(p. D-3)
How Real-Time Workshop streamlines software design
and development
How MathWorks Tools Streamline
Development (p. D-12)
Maximizing the efficiency of software design,
development and deployment
Code Formats (p. D-18)
How Real-Time Workshop saves time and effort by
packaging code in target-specific ways
An Open and Extensible Environment
(p. D-33)
Real-Time Workshop features you can customize and
elaborate
D
The Real-Time Workshop Development Process
Introduction
The primary features of Real-Time Workshop are
• Simulink Code Generator: Automatically generates C code from your
Simulink model.
• Make Process: Real-Time Workshop’s user-extensible make process lets you
create your own production or rapid prototyping target.
• Simulink External Mode: External mode enables communication between
Simulink and a model executing on a real-time test environment, or in
another process on the same machine. External mode lets you perform
real-time parameter tuning and data viewing using Simulink as a front end.
• Targeting Support: Using the Real-Time Workshop bundled targets, you
can build systems for a number of environments, including Tornado and
DOS. The generic real-time and embedded real-time targets provide a
framework for developing customized rapid prototyping or production target
environments. In addition to the bundled targets, the Real-Time Windows
Target and/or the xPC Target let you turn a PC of any form factor into a rapid
prototyping target, or a small to medium volume production target.
• Rapid Simulations: Using Simulink Accelerator (part of the Simulink
Performance Tools product), S-Function Target, or Rapid Simulation Target,
you can accelerate your simulations by 5 to 20 times on average. Executables
built with these targets bypass Simulink normal interpretive simulation
modes, which must handle all configurations of each basic modeling
primitive. The code generated by Simulink Accelerator, S-Function Target,
or Rapid Simulation Target is optimized to execute only the algorithms used
in your specific model. In addition, these targets apply many optimizations,
such as eliminating ones and zeros in computations for filter blocks.
D-2
A Next-Generation Development Tool
A Next-Generation Development Tool
The MathWorks tools, including Simulink and Real-Time Workshop, are
revolutionizing the way embedded systems are designed. Simulink is a very
high-level language (VHLL) — a next-generation programing language. A brief
look at the history of dynamic and embedded system design methodologies
reveals a steady progression toward higher-level design tools and processes:
• Design -> analog components: Before the introduction of microcontrollers,
design was done on paper and realized using analog components.
• Design -> hand written assembly -> early microcontrollers: In the early
microprocessor era, design was done on paper and realized by writing
assembly code and placing it on microcontrollers. Today, very low-end
applications still use assembly language, but advancements in Real-Time
Workshop and C compiler technology are obsolescing such techniques.
• Design -> high-level language (HLL) -> object code -> microcontroller:
The advent of efficient HLL compilers led to the realization of paper designs
in languages such as C. HLL code, transformed to assembly language by a
compiler, was then placed on a microcontroller. In the early days of
high-level languages, programmers often inspected the machine-generated
assembly code produced by compilers for correctness. Today, it is taken for
granted that the assembly code is correct.
• Design -> modeling tool -> manual HLL coding -> object code ->
microcontroller: When design tools such as Simulink appeared, designers
were able to express system designs graphically and simulate them for
correctness. While this process saved considerable time and improved
performance, designs were still translated to C code manually before being
placed on a microcontroller. This translation process was both time
consuming and error prone.
• Design -> Simulink -> Real-Time Workshop (automatic code generation)
-> object code -> microcontroller: With the addition of Real-Time
Workshop, Simulink itself becomes a VHLL. Modeling constructs in
Simulink are the basic elements of the language. Real-Time Workshop then
compiles models to produce C code. This machine-generated code is produced
quickly and correctly. The manual process of transforming designs to code
D-3
D
The Real-Time Workshop Development Process
has now been eliminated, yielding significant improvements in system
design.
The Simulink code generator included within Real-Time Workshop is a
next-generation graphical block diagram compiler. Real-Time Workshop has
capabilities beyond those of a typical HLL compiler. Generated code is highly
readable and customizable. It is normally unnecessary to read the object code
produced by the HLL compiler.. You can use Real-Time Workshop in a wide
variety of applications, improving your design process.
Key Features
The general goal of the MathWorks tools, including Real-Time Workshop, is to
enable you to accelerate your design process while reducing cost, decreasing
time to market, and improving quality.
labor
Traditional
development:
Area under
curve indicates
the development cost.
design,
implementation,
test
start
In traditional
development practices
products often
ship before they
are completely tested,
resulting in a product
with defects.
labor
Development via
the MathWorks tools:
time
product
release
time
product
release
Traditional development practices tend to be very labor intensive. Poor tools
often lead to a proliferation of ad hoc software projects that fail to deliver
reusable code. With the MathWorks tools, you can focus energy on design and
achieve better results in less time with fewer people.
D-4
A Next-Generation Development Tool
Real-Time Workshop, along with other components of the MathWorks tools,
provides
• A rapid and direct path from system design to implementation
• Seamless integration with MATLAB and Simulink
• A simple graphical user interface
• An open and extensible architecture
The following features of Real-Time Workshop enable you to reach the above
goal:
• Code generator for Simulink models
- Generates optimized, customizable code. There are several styles of
generated code, which can be classified as either embedded (production
phase) or rapid prototyping.
- Supports all Simulink features, including 8, 16, and 32 bit integers and
floating-point data types.
- Fixed-Point Blockset and Real-Time Workshop allow for scaling of integer
words ranging from 2 to 128 bits. Code generation is limited by the
implementation of char, short, int, and long in embedded C compiler
environments (usually 8, 16, and 32 bits).
- Generated code is processor independent. The generated code represents
your model exactly. A separate run-time interface is used to execute this
code. We provide several example run-time interfaces as well as
production run-time interfaces.
- Supports any single or multitasking operating system. Also supports
“bare-board” (no operating system) environments.
- The Target Language Compiler allows extensive customization of the
generated code via TLC scripting.
- Enables custom code generation for S-functions (user-created blocks)
using TLC files, enabling you to embed very efficient custom code into the
model’s generated code.
• Extensive model-based debugging support
- External mode enables you to examine what the generated code is doing
by uploading data from your target to the graphical display elements in
your model. There is no need to use a conventional C debugger to look at
your generated code.
D-5
D
The Real-Time Workshop Development Process
- External mode also enables you to tune the generated code via your
Simulink model. When you change a parametric value of a block in your
model, the new value is passed down to the generated code, running on
your target, and the corresponding target memory location is updated.
Again, there is no need to use an embedded compiler debugger to perform
this type of operation. Your model is your debugger user interface.
• Integration with Simulink
- Code validation. You can generate code for your model and create a
standalone executable that exercises the generated code and produces a
MAT-file containing the execution results.
- Generated code contains system/block identification tags to help you
identify the block, in your source model, that generated a given line of
code. The MATLAB command hilite_system recognizes these tags and
highlights the corresponding blocks in your model.
- Support for Simulink data objects lets you define how your signals and
block parameters interface to the external world.
• Rapid simulations
- Real-Time Workshop supports several ways to speed up your simulations
by creating optimized, model-specific executables.
• Target support
- Turnkey solutions for rapid prototyping substantially reduce design
cycles, allowing for fast turnaround of design iterations.
- Bundled rapid prototyping example targets are provided.
- Add-on targets (Real-Time Windows Target and xPC Target) for PC-based
hardware are available from The MathWorks. These targets enable you to
turn a PC with fast, high-quality, low cost hardware into a rapid
prototyping system.
- Supports a variety of third-party hardware and tools, with extensible
device driver support.
• Extensible make process
- Allows for easy integration with any embedded compiler and linker.
- Provides for easy linkage with your hand-written supervisory or
supporting code.
D-6
A Next-Generation Development Tool
• Real-Time Workshop Embedded Coder
- Customizable, portable, and readable C code that is designed to be placed
in a production embedded environment.
- More efficient code is created, because inlined S-functions are required
and continuous time states are not allowed.
- Software-in-the-loop. With Real-Time Workshop Embedded Coder, you
can generate code for your embedded application and bring it back into
Simulink for verification via simulation.
- Web-viewable code generation report describes code modules, analyzes the
generated code, and helps to identify code generation optimizations
relevant to your program.
- Annotation of the generated code using the Description block property.
- Hooks for external parameter tuning and signal monitoring are provided
enabling easy interfacing of the generated code in your real-time system.
Benefits
You can benefit by using Real-Time Workshop in the following applications.
This is not an exhaustive list, but a general survey:
• Production Embedded Real-Time Applications
Real-Time Workshop lets you generate, cross-compile, link, and download
production C code for real-time systems (such as controllers or DSP
applications) onto your target processor directly from Simulink. You can
customize the generated code by inserting S-functions into your model and
specifying, via the Target Language Compiler, what the generated code
should look like. Using the optimized, automatically generated code, you can
focus your coding efforts on specific features of your product, such as device
drivers and general device interfacing.
• Rapid Prototyping
As a rapid prototyping tool, Real-Time Workshop enables you to implement
your embedded systems designs quickly, without lengthy hand-coding and
debugging. Rapid prototyping is typically used in the software/hardware
integration and testing phases of the design cycle enabling you to
- Conceptualize solutions graphically in a block diagram modeling
environment.
D-7
D
The Real-Time Workshop Development Process
- Evaluate system performance early on - prior to laying out hardware,
coding production software, or committing to a fixed design.
- Refine your design by rapid iteration between algorithm design and
prototyping.
- Tune parameters while your real-time model runs, using Simulink
operating in external mode as a graphical front end.
You can use Real-Time Workshop to generate downloadable, targeted C code
that runs on top of a real-time operating system (RTOS). Alternatively, you
can generate code to run on the bare hardware at interrupt level, using a
simple rate monotonic scheduling executive that you create from examples
provided with Real-Time Workshop. There are many rapid prototyping
targets provided; or you can create your own.
During rapid prototyping, the generated code is fully instrumented enabling
direct access via Simulink external mode for easy monitoring and debugging.
The generated code contains a data structure that encapsulates the details
of your model. This data structure is used in the bidirectional connection to
Simulink running in external mode. Using Simulink external mode, you can
monitor signal and tune parameters to further refine your model in rapid
iterations enabling you to achieve desired results quickly.
• Real-Time Simulation
You can create and execute code for an entire system or specified subsystems
for hardware-in-the-loop simulations. Typical applications include training
simulators, real-time model validation, and prototype testing.
• Turnkey Solutions
Bundled Real-Time Workshop targets and third-party turnkey solutions
support a variety of control and DSP applications. The target environments
include embedded PC, PCI, ISA, VME, and custom hardware, running
off-the-shelf real-time operating systems, DOS, or Microsoft Windows.
Target system processor architectures include Motorola MC680x0 and
PowerPC processors, Intel-80x86 and compatibles, Alpha, and Texas
Instruments DSPs. Third-party vendors are regularly adding other
architectures. For a current list of third-party turnkey solutions, see the
MATLAB Connections Web page:
http://www.mathworks.com/products/connections.
The open environment of Real-Time Workshop also lets you create your own
turnkey solution.
D-8
A Next-Generation Development Tool
• Intellectual Property Protection
The S-Function Target, in addition to speeding up your simulation, allows
you to protect your intellectual property: the designs and algorithms
embodied in your models. Using the S-Function Target, you can generate
and distribute binaries from your models or subsystems. End users have
access to the interface, but not to the body, of your algorithms.
• Rapid Simulations
The MathWorks tools can be used in the design of most dynamic systems.
Generally Simulink is either used to model a high-fidelity dynamic system
(e.g., an engine) or a real-time system (such as an engine controller or a
signal processing system).
When modeling high-fidelity systems, you can use Real-Time Workshop to
accelerate the design process by speeding up your simulations. This is
achieved by using one of the following Real-Time Workshop components:
- Simulink Accelerator: Creates a dynamically linked library (MEX-file)
from code optimized and generated for your specific model configuration.
This executable is used in place of the normal interpretive mode of
simulation. Typical speed improvements range from 2 to 8 times faster
than normal simulation time. Simulink Accelerator supports both fixed
and variable step solvers. Simulink Accelerator is part of the Simulink
Performance Tools product.
- Rapid Simulation Target: Creates a stand-alone executable from code
optimized and generated for your specific model configuration. This
stand-alone executable does not need to interact with a graphics
subsystem. Typical speed improvements range from 5 to 20 times faster
than normal simulation times. The Rapid Simulation Target is ideal for
repetitive (batch) simulations where you are adjusting model parameters
or coefficients. Rapid Simulation Target supports only fixed-step solvers.
- S-Function Target: This target, like Simulink Accelerator, creates a
dynamically linked library (MEX-file) from a model. You can incorporate
this component into another model using the Simulink S-function block.
Integration with Simulink
If the Real-Time Workshop target you are using supports Simulink external
mode, you can use Simulink as the monitoring/debugging interface for the
generated code. With external mode, you can
D-9
D
The Real-Time Workshop Development Process
• Change parameters via the block dialogs, gauges, and the set_param
MATLAB command. The set_param command lets you interact
programmatically with your target.
• View target signals in Scope blocks, Display blocks, general S-Function
blocks, and via gauges.
These concepts are illustrated by Figure D-1 and Figure D-2.
You can view the
state of target data
values using display
devices.
You can change
block parameter values
on the target while your
model is executing.
You can view
target data values
using display blocks.
You can view
target signals
using scope blocks.
TCP/IP, serial, shared memory or other
communication link
Target system
Figure D-1: Signal Viewing and Parameter Tuning in External Mode
D-10
A Next-Generation Development Tool
Simulink
model
TCP/IP, serial, shared memory or other
communication link
Target system
Figure D-2: Dials and Gauges Provide Front End to Target System
D-11
D
The Real-Time Workshop Development Process
How MathWorks Tools Streamline Development
Figure D-3 is a high-level view of a traditional development process without the
MathWorks tools.
Problem/task
formulation
Requirements
specification
Unrealizable or
incorrect requirements
System level
design
System specification,
component interface
specification,
and validation plan
Unfeasible system
specification
Detailed
design
Components specifications
and component validation
Unrealizable or
incorrect design
Software and
hardware
implementation
Prototype system
and test plans
Design error found
during testing
Design validation
via testing
Production system
ready for deployment
Field failure or
manufacturing
problem
Production and
manufacturing
testing
Figure D-3: Traditional Development Process Without MathWorks Tools
In Figure D-3, each block represents a work phase. Documents are used to
coordinate the different work phases. In this environment, it is easy to go back
one work phase, but hard to go back multiple work phases. In this
environment, design engineers (such as control system engineers or signal
processing engineers) are not usually involved in the prototyping phase until
D-12
How MathWorks Tools Streamline Development
many months after they have specified the design. This can result in poor time
to market and inferior quality.
In this environment, different tools are used in each phase. Designs are
communicated via paper. This enforces a serial, rather than an iterative,
development process. Developers must reenter the result of the previous phase
before they can begin work on a new phase. This leads to miscommunication
and errors, resulting in lost work hours. Errors found in later phases are very
expensive and time consuming to correct. Correction often involves going back
several phases. This is difficult because of the poor communication between the
phases.
The MathWorks does not suggest or impose a development process. The
MathWorks tools can be used to complement any development process. In the
above process, use of our tools in each phase can help eliminate paper work.
Our tools also lends itself well to the spiral design process shown in Figure D-4.
D-13
D
The Real-Time Workshop Development Process
tai
l ed
de
sig
n
System level design
De
ob
Pr
l em
f
sk
/t a
or
la
mu
ti o
n
Done
Start
Design validation
via testing
nd
re a
a
w
Soft dware tation
har lemen
imp
Pro
manductio
test ufact n and
ing urin
g
Figure D-4: Spiral Design Process
Using the MathWorks tools, your model represents your understanding of your
system. This understanding is passed from phase to phase in the model,
reducing the need to go back to a previous phase. In the event that rework is
necessary in a previous phase, it is easier to transition back one or more
phases, because the same model and tools are used in all phases.
A spiral design process iterates quickly between phases, enabling engineers to
work on innovative features. The only way to do this cost effectively is to use
tools that make it easy to move from one phase to another. For example, in a
matter of minutes a control system engineer or a signal processing engineer
can validate an algorithm on a real-world rapid prototyping system. The spiral
process lends itself naturally to parallelism in the overall development process.
You can provide early working models to validation and production groups,
D-14
How MathWorks Tools Streamline Development
involving them in your system development process from the start. This helps
compress the overall development cycle while increasing quality.
Another advantage of the MathWorks tools is that it enables people to work on
tasks that they are good at and enjoy doing. For example, control system
engineers specialize in design control rules, while embedded system engineers
enjoy assembling systems consisting of hardware and low-level software. It is
possible to have very talented people perform different roles, but it is not
efficient. Embedded system engineers, for example, are rewarded by specifying
and building the hardware and creating low-level software such as device
drivers, or real-time operating systems. They do not find data entry operations,
such as the manual conversion of a set of equations to efficient code, to be
rewarding. This is where the MathWorks tools shines. The equations are
represented as models and Real-Time Workshop converts them to highly
efficient code ready for deployment.
D-15
D
The Real-Time Workshop Development Process
Role of the MathWorks Tools in Your Development Process
The following figure outlines where the MathWorks tools, including Real-Time
Workshop, helps you in your development process.
Interactive design
Interactive modeling and simulation
MATLAB
and
Toolboxes
High speed simulation
Accelerator,
Simulink,
Stateflow, and Blocksets
S-Function Targets
Batc
Design
Cycle
Sy
Deployed system
Embedded Code
in Custom Target
System testing and tuning
Embedded Code
in Custom Target
Software integration
valid
R ap
ation
id S
imu
Targ lation
et
m
de
Ra
vel
p
op
i
Ta d
me
rg Pr
nt
ets ot
t
(re otyp estin
al- in
g
tim g
e)
g
in
st
te
e
it
od
un
C
e
ar
ed les
ftw edd du
So
b o
m M
E
Customer defined
monitoring and
parameter tuning
ste
h de
sign
Embedded Code
in Custom Target
Figure D-5: Roles of MathWorks Tools in Software Design
Early in the design phase, you will start with MATLAB and Simulink to help
you formulate your problems and create your initial design. Real-Time
Workshop helps with this process by enabling high-speed simulations via
Simulink Accelerator (also part of Simulink Performance Tools), and the
S-function Target for componentization and model speed-up.
D-16
How MathWorks Tools Streamline Development
After you have a functional model, you may need to tune your model’s
coefficients. This can be done quickly using Real-Time Workshop Rapid
Simulation Target for Monte-Carlo type simulations (varying coefficients over
many simulations).
After you’ve tuned your model, you can move into system development testing
by exercising your model on a rapid prototyping system such as the Real-Time
Windows Target or the xPC Target. With a rapid prototyping target, you
connect your model to your physical system. This lets you locate design flaws
or modeling errors quickly.
After your prototype system is created, you can use the Real-Time Workshop
Embedded Coder to create embeddable code for deployment on your custom
target. The signal monitoring and parameter tuning capabilities enable you to
easily integrate the embedded code into a production environment equipped
with debugging and upgrade capabilities.
D-17
D
The Real-Time Workshop Development Process
Code Formats
The Real-Time Workshop code generator transforms your model to HLL code.
Real-Time Workshop supports a variety of code formats designed for different
execution environments, or targets.
In the traditional embedded system development process, an engineer develops
an algorithm (or equations) to be implemented in an embedded system. These
algorithms are manually converted to a computer language such as C. This
translation process, usually done by an embedded system engineer, is much
like data entry.
Using Simulink to specify the algorithm (or equations), and Real-Time
Workshop to generate corresponding code, engineers can bypass this
redundant translation step. This enables embedded system engineers to focus
on the key issues involved in creating an embedded system: the hardware
configuration, device drivers, supervisory logic, and supporting logic for the
model equations. Simulink itself is the programming language that expresses
the algorithmic portion of the system.
D-18
Code Formats
The Simulink code generator provided with Real-Time Workshop is an open
“graphical compiler” supporting a variety of code formats. The relationship
between code formats and targets is shown below.
Simulink,
Stateflow, and Blocksets
Modeling and simulation
Simulink Code
Generator
S-function/Accelerator
code format
Real-time
code format
(single-instance)
Stateflow Coder
Real-Time Malloc
code format
(multi-instance)
Embedded
code format
MEX-file
Interface
Generic
Run-Time
Interface
Turnkey
Run-Time
Interface
Rapid
Simulation
Run-Time
Interface
Embedded
Run-Time
Interface
.DLL or .so
for use with
Simulink
Any
computer or
processor
VME, PCI, ISA
and other
environments
Workstation
executable
Embedded
microcontroller /
processor
Integrated
Generic workstation
Turnkey
accelerated
target for model
prototyping
simulations
verification and for
targets
and intellectual use as a starting point
property
when creating a new
protection
rapid prototyping target
Rapid Simulation
Target
for fast
simulation and
batch processing
Embedded
targets
Figure D-6: Relationship Between Code Formats and Targets
D-19
D
The Real-Time Workshop Development Process
S-Function/Accelerator Code Format
This code format, used by the S-function Target and Simulink Accelerator,
generates code that conforms to Simulink C MEX S-function API.
Real-Time Code Format
The real-time code format is ideally suited for rapid prototyping. This code
format (C only) supports increased monitoring and tuning capabilities,
enabling easy connection with external mode. Real-time code format supports
continuous-time models, discrete-time singlerate or multirate models, and
hybrid continuous-time and discrete-time models. Real-time code format
supports both inlined and noninlined S-functions. Memory allocation is
declared statically at compile time.
Real-Time Malloc Code Format
The real-time malloc code format is similar to the real-time code format. The
primary difference is that the real-time malloc code format declares memory
dynamically. This supports multiple instances of the same model, with each
instance including a unique data set. Multiple models can be combined into one
executable without name clashing. Multiple instances of a given model can also
be created in one executable.
Embedded Code Format
The embedded code format is designed for embedded targets. The generated
code is optimized for speed, memory usage, and simplicity. Generally, this
format is used in deeply embedded or deployed applications. There are no
dynamic memory allocation calls; all persistent memory is statically allocated.
Real-Time Workshop can generate either C code in the embedded code format.
Generating embedded code format requires the Real-Time Workshop
Embedded Coder, a separate add-on product for use with Real-Time Workshop.
The embedded code format provides a simplified calling interface and reduced
memory usage. This format manages model and timing data in a compact
real-time model data structure. This contrasts with the other code formats,
which use a significantly larger data structure to manage the generated code.
The embedded code format improves readability of the generated code, reduces
code size, and speeds up execution. The embedded code format supports all
discrete-time singlerate or multirate models.
D-20
Code Formats
Because of its optimized and specialized data structures, the embedded code
format supports only inlined S-functions.
Target Environments
Real-Time Workshop supports many target environments. These include
ready-to-run configurations and third-party targets. You can also develop your
own custom target.
This section begins with a list of available target configurations. Following the
list, we summarize the characteristics of each target.
Available Target Configurations
Target Configurations Bundled with Real-Time Workshop. The MathWorks supplies the
following target configurations with Real-Time Workshop:
• DOS (4GW) Target (example only)
• Generic Real-Time (GRT) Target
• LE/O (Lynx Embedded OSEK) Real-Time Target (example only)
• Rapid Simulation Target
• Tornado (VxWorks) Real-Time Target
Target Configurations Bundled with Real-Time Workshop Embedded Coder. The
MathWorks supplies the following target configuration with Real-Time
Workshop Embedded Coder (a separate product from Real-Time Workshop):
• Real-Time Workshop Embedded Coder Target
Turnkey Rapid Prototyping Target Products. These self-contained solutions ( separate
products from Real-Time Workshop) include:
• Real-Time Windows Target
• xPC Target
DSP Target Products. See Developer's Kit for Texas Instruments DSP User’s Guide
for information on this target:
• Texas Instruments TMS320C6701 Evaluation Module Target
D-21
D
The Real-Time Workshop Development Process
Third-Party Targets. Numerous software vendors have developed customized
targets for Real-Time Workshop. For an up-to-date listing of third-party
targets, visit the MATLAB Connections Web page at
http://www.mathworks.com/products/connections
View Third-Party Solutions by Product Type, and then select Real-Time
Workshop Target from the drop-down list.
Custom Targets. Typically, to target custom hardware, you must write a harness
(main) program for your target system to execute the generated code, and I/O
device drivers to communicate with your hardware. You must also create a
system target file and a template makefile.
Real-Time Workshop supplies generic harness programs as starting points for
custom targeting. See Chapter 14, “Targeting Real-Time Systems” in the
Real-Time Workshop documentation for the information you will need to
develop a custom target.
Rapid Simulation Target
Rapid Simulation Target (RSIM) consists of a set of target files for
non-real-time execution on your host computer. RSIM enables you to use
Real-Time Workshop to generate fast, stand-alone simulations. RSIM allows
batch parameter tuning and downloading of new simulation data (signals)
from a standard MATLAB MAT-file without the need to recompile the model.
The speed of the generated code also makes RSIM ideal for Monte Carlo
simulations. The RSIM target enables the generated code to read and write
data from or to standard MATLAB MAT-files. RSIM reads new signals and
parameters from MAT-files at the start of simulation.
RSIM enables you to run stand-alone, fixed-step simulations on your host
computer or on additional computers. If you need to run 100 large simulations,
you can generate the RSIM model code, compile it, and run the executables on
10 identical computers. The RSIM target allows you to change the model
parameters and the signal data, achieving significant speed improvements by
using a compiled simulation.
S-Function and Accelerator Targets
S-Function Target provides the ability to transform a model into a Simulink
S-function component. Such a component can then be used in a larger model.
This allows you to speed up simulations and/or reuse code. You can include
D-22
Code Formats
multiple instances of the same S-function in the same model, with each
instance maintaining independent data structures. You can also share
S-function components without exposing the details of the a proprietary source
model.
The Accelerator Target is similar to the S-Function Target in that an
S-function is created for a model. The Accelerator Target differs from the
S-Function Target in that the generated S-function operates in the
background. It provides for faster simulations while preserving all existing
simulation capabilities (parameter change, signal visualization, full S-function
support, etc.).
Turnkey Rapid Prototyping Targets
The Real-Time Windows Target and the xPC Target are add-on products to
Real-Time Workshop. Both of these targets turn an Intel 80x86/Pentium or
compatible PC into a real-time system. Both support a large selection of
off-the-shelf I/O cards (both ISA and PCI).
With turnkey target systems, all you need to do is install the MathWorks
software and a compiler, and insert the I/O cards. You can then use a PC as a
real-time system connected to external devices via the I/O cards.
Real-Time Windows Target. The Real-Time Windows Target brings rapid
prototyping and hardware-in-the-loop simulation to your desktop. It is the
most portable solution available today for rapid prototyping and
hardware-in-the-loop simulation when used on a laptop outfitted with a
PCMCIA I/O card. The Real-Time Windows Target is ideal since a second PC
or other real-time hardware is often unnecessary, impractical or cumbersome.
D-23
D
The Real-Time Workshop Development Process
This picture shows the basic components of the Real-Time Windows Target.
MATLAB/Simulink
Real-Time Workshop
Real-time debugging
of your model using
Simulink external mode:
- Run-time parameter tuning
- Data uploading to scopes
- Data uploading to display blocks
- Data uploading to custom blocks
(S-functions)
- Full Dials and Gauges support
- Support for general simulation viewing
devices
Modeling and simulation
Code generation
Visual C/C++ or
Watcom C/C++ Compiler
Automated build and download process
Real-Time Task Running
Underneath Windows
Single-box solution
10 kHz + sample rates
I/O Boards in PC
Support for over 100 I/O boards
(more added on request)
Figure D-7: Real-Time Windows (rtwin) Target
As a prototyping environment, the Real-Time Windows Target is exceptionally
easy to use, due to tight integration with Simulink and external mode. It is
much like using Simulink itself, with the added benefit of gaining real-time
performance and connectivity to the real world through a wide selection of
supported I/O boards. You can control your real-time execution with buttons
located on the Simulink toolbar. Parameter tuning is done interactively, by
simply editing Simulink blocks and changing parameter values. For viewing
signals, the Real-Time Windows Target uses standard Simulink Scope blocks,
without any need to alter your Simulink block diagram. Signal data can also be
logged to a file or set of files for later analysis in MATLAB.
The Real-Time Windows Target is often called the “one-box rapid prototyping
system,” since both Simulink and the generated code run on the same PC. A
D-24
Code Formats
run-time interface enables you to run generated code on the same processor
that runs Windows The generated code executes in real time, allowing
Windows to execute when there are free CPU cycles. The Real-Time Windows
Target supports over 100 I/O boards, including ISA, PCI, CompactPCI, and
PCMCIA. Sample rates in excess of 10 to 20 kHz can be achieved on Pentium
PCs.
In universities, the Real-Time Windows Target provides a cost effective
solution since only a single computer is required. In commercial applications,
the Real-Time Windows Target is often used at an engineer’s desk prior to
taking a project to an expensive dedicated real-time testing environment. Its
portability is unrivaled, allowing you to use your laptop as a real-time test bed
for applications in the field.
Figure D-8 illustrates the use of the Real-Time Windows Target in a model
using magnetic levitation to suspend a metal ball in midair. The system is
controlled by the model shown in Figure D-9.
Electromagnet
Simulink &
Real-Time
Workshop
Metallic ball
PMCIA I/O
Proximity sensor
Sensor and actuator
connector cable
Figure D-8: Magnetic Levitation System
D-25
D
The Real-Time Workshop Development Process
Metallic ball height offset
oscillation frequency and amplitude
Strip chart shows height of metallic ball
Figure D-9: Model for Controlling Magnetic Levitation System
Rapid Prototyping Targets
There are two classes of rapid prototyping targets: those using the real-time
code format and those using the real-time malloc code format. These differ in
the way they allocate memory (statically versus dynamically). Most rapid
prototyping targets use the real-time code format.
We define two forms of rapid prototyping environments:
• Heterogeneous rapid prototyping environments use rapid prototyping
hardware (such as an Intel-80x86/Pentium or similar processor) that differs
D-26
Code Formats
from the final production hardware. For example, an Intel-80x86/Pentium or
similar processor might be used during rapid prototyping of a system that is
eventually deployed onto a fixed-point Motorola microcontroller.
• Homogeneous rapid prototyping environments are characterized by the
use of similar hardware for the rapid prototyping system and the final
production system. The main difference is that the rapid prototyping system
has extra memory and/or interfacing hardware to support increased
debugging capabilities, such as communication with external mode.
Homogeneous rapid prototyping environments eliminate uncertainty because
the rapid prototyping environment is closer to the final production system.
However, a turnkey system for your specific hardware may not exist. In this
case, you must weigh the advantages and disadvantages of using one of the
existing turnkey systems for heterogeneous rapid prototyping, versus creating
a homogeneous rapid prototyping environment.
Several rapid prototyping targets are bundled with Real-Time Workshop.
Generic Real-Time (GRT) Target. This target uses the real-time code format and
supports external mode communication. It is designed to be used as a starting
point when creating a custom rapid prototyping target, or for validating the
generated code on your workstation.
Generic Real-Time Malloc (GRTM) Target. This target is similar to the GRT target but
it uses the real-time malloc code format. This format uses the C malloc and
free routines to manage all data. With this code format, you can have multiple
instances of your model and/or multiple models in one executable.
Tornado Target. The Tornado target uses the real-time or real-time malloc code
format. A set of run-time interface files are provided to execute your models on
the Wind River System’s real-time operating system, VxWorks. The Tornado
target supports singletasking, multitasking, and hybrid continuous and
discrete-time models.
The Tornado run-time interface and device driver files can also be used as a
starting point when targeting other real-time operating system environments.
The run-time interface provides full support for external mode, enabling you to
take full advantage of the debugging capabilities for parameter tuning and
data monitoring via graphical devices.
D-27
D
The Real-Time Workshop Development Process
DOS Target. The DOS target (provided as an example only) uses the real-time
code format to turn a PC running the DOS operating system into a real-time
system. This target includes a set of run-time interface files for executing the
generated code. This run-time interface installs interrupt service routines to
execute the generated code and handle other interrupts. While the DOS target
is running, the user does not have access to the DOS operating system. Sample
device drivers are provided.
The MathWorks recommends that you use the Real-Time Windows Target or
the xPC Target as alternatives to the DOS Target. The DOS target is provided
only as an example and its support will be discontinued in the future.
OSEK Targets. The OSEK target (provided as an example only) lets you use the
automotive standard open real-time operating system. The run-time interface
and OSEK configuration files that are included with this target make it easy
to port applications to a wide range of OSEK environments.
Embedded Targets
The embedded real-time target is the main component of the Real-Time
Workshop Embedded Coder. It consists of a set of run-time interface files that
drive code, generated in the embedded code format, on your workstation. This
target is ideal for memory-constrained embedded applications. Real-Time
Workshop supports generation of embedded code in C.
In its default configuration, the embedded real-time target is designed for use
as a starting point for targeting custom embedded applications, and as a means
by which you can validate the generated code. To create a custom embedded
target, you start with the embedded real-time target run-time interface files
and edit them as needed for your application.
In the terminology of Real-Time Workshop, an embedded target is a deeply
embedded system. Note that it is possible to use a rapid prototyping target in
an embedded (production) environment. This may make more sense in your
application.
Code Generation Optimizations
The Simulink code generator included with Real-Time Workshop is packed
with optimizations to help create fast and minimal size code. The optimizations
are classified either as cross-block optimizations, or block specific
optimizations. Cross-block optimizations apply to groups of blocks or the
D-28
Code Formats
general structure of a model. Block specific optimizations are handled locally
by the object generating code for a given block. Listing each block specific
optimization here is not practical; suffice it to say that the Target Language
Compiler technology generates very tight and fast code for each block in your
model.
The following sections discuss some of the cross-block optimizations.
Multirate Support
One of the more powerful features of Simulink is its implicit support for
multirate systems. The ability to run different parts of a model at different
rates guarantees optimal use of the target processor. In addition, Simulink
enforces correctness by requiring that you create your model in a manner that
guarantees deterministic execution.
Inlining S-Function Blocks for Optimal Code
The ability to add blocks to Simulink via S-functions is enhanced by the Target
Language Compiler. You can create blocks that embed the minimal amount of
instructions into the generated code. For example, if you create a device driver
using an S-function, you can have the generated code produce one line for the
device read, as in the following code fragment:
mdlOutputs(void)
{
.
.
rtB.deviceout = READHW(); /* Macro to read hw device using
.
assembly code */
.
}
Note that the generic S-function API is suitable for any basic block-type
operation.
Loop Rolling Threshold
The code generated for blocks can contain for loops, or the loop iterations can
be “flattened out” into inline statements. For example, the general gain block
equation is
D-29
D
The Real-Time Workshop Development Process
for (i = 0; i < N; i++) {
y[i] = k[i] * u[i];
}
If N is less than a specified roll threshold, Real-Time Workshop expands out the
for loop, otherwise Real-Time Workshop retains the for loop.
Tightly Coupled Optimal Stateflow Interface
The generated code for models that combine Simulink blocks and Stateflow
charts is tightly integrated and very efficient.
Stateflow Optimizations
The Stateflow Coder contains a large number of optimizations that produce
highly readable and very efficient generated code.
Inlining of Systems
In Simulink, a system starting at a nonvirtual subsystem boundary (e.g. an
enabled, triggered, enabled and triggered, function-call, or atomic subsystem)
can be inlined by selecting the RTW inline subsystem option from the
subsystem block properties dialog. The default action is to inline the
subsystem, unless it is a function-call subsystem with multiple callers.
Block I/O Reuse
Consider a model with a D/A converter feeding a gain block (for scaling), then
feeding a transfer function block, then feeding a A/D block. If all signals refer
to the same memory location, then less memory will be used. This is referred
to as block I/O reuse. It is a powerful optimization technique for re-using
memory locations. It reduces the number of global variables improving the
executing speed (faster execution) and reducing the size of the generated code.
Declaration of Block I/O Variables in Local Scope
If input/output signal variables are not used across function scope, then they
can be placed in local scope. This optimization technique reduces code size and
improves the execution speed (faster execution).
Inlining of Parameters
If you select the Inline parameters option, the numeric values of block
parameters that represent coefficients are embedded in the generated code. If
D-30
Code Formats
Inline parameters is off, block parameters that represent coefficients can be
changed while the model is executing.
Note that it is possible to specify which parameters to tune using the
Workspace parameter attributes dialog box.
Inlining of Invariant Signals
An invariant signal is a block output signal that does not change during
Simulink simulation. For example, the output of a sum block that is fed by two
constants cannot change. When Inline invariant signals is selected on the
General code generation options portion of the Real-Time Workshop pane,
a single numeric value is placed in the generated code to represent the output
value of the sum block. The Inline invariant signals option is available when
the Inline parameters option is on.
Parameter Pooling
The Parameter pooling option is available when Inline parameters is
selected. If Real-Time Workshop detects identical usage of parameters (e.g. two
lookup tables with same tables), it will pool these parameters together, thereby
reducing code size.
Block Reduction Optimizations
Real-Time Workshop can detect block patterns (e.g. an accumulator
represented by a constant, sum and a delay block) and reduce these patterns to
a single operation, resulting in very efficient generated code.
Creation of Contiguous Signals to Speed Block Computations
Some block algorithms (for example a matrix multiply) can be implemented
more efficiently if the signals entering the blocks are contiguous.
Noncontiguous signals occur because of the handling of virtual blocks. For
example, the output of a Mux block is noncontiguous. When this class of block
requires a contiguous signal, Simulink will insert (if needed) a copy block
operator to make the signal contiguous. This results in better code efficiency.
Support for Noncontiguous Signals by Blocks
Noncontiguous signals occur because of the block virtualization capabilities of
Simulink. For example, the output of a Mux block is generally a noncontiguous
signal (i.e., the output signal consists of signals from multiple sources). General
D-31
D
The Real-Time Workshop Development Process
blocks in Simulink support this behavior by generating very efficient code to
handle each different signal source in a noncontiguous signal.
Data Type Support
Simulink models support a wide range of data types. You can use double
precision values to represent real-world values and then when needed use
integers or Booleans for discrete valued signals. You can also use fixed-point
(integer scaling) capabilities to target models for fixed-point embedded
processors. The wide selection of data types in Simulink models enables you to
realize your models efficiently.
Frame Support
In signal processing, a frame of data represents time sampled sequences of an
input. Many devices have support in hardware for collecting frames of data.
With Simulink and the DSP Blockset, you can use frames and perform frame
based operations on the data. Frames are a very efficient way of handling high
frequency signal processing applications.
Matrix Support
Most blocks in Simulink support the use of matrices. This enables you to create
models that represent high levels of abstractions and produce very efficient
generated code.
Virtualization of Blocks
Nearly half of the blocks in a typical model are connection type blocks (e.g.
Virtual Subsystem, Inport, Outport, Goto, From, Selector, Bus Selector, Mux,
Demux, Ground, and Terminator). These blocks are provided to enable you to
create complex models with your desired levels of abstraction. Simulink treats
these blocks as virtual, meaning that they impose no overhead during
simulation or in the generated code.
D-32
An Open and Extensible Environment
An Open and Extensible Environment
The Simulink and Real-Time Workshop model-based software development
environment is extensible in several ways.
Custom Code Support
S-functions are dynamically linked objects (.dll or .so) that bind with
Simulink to extend the modeling environment. By developing S-functions, you
can add custom block algorithms to Simulink. Such S-functions provide
supporting logic for the model. S-functions are flexible, allowing you to
implement complex algorithmic equations or basic low-level device drivers.
Real-Time Workshop support for S-functions includes the ability to inline
S-function code directly into the generated code. Inlining, supported by the
Target Language Compiler, can significantly reduce memory usage and calling
overhead.
Support for Supervisory Code
The generated code implements an algorithm that corresponds exactly to the
algorithm defined in your model. With the embedded code format, you can call
the generated model code as a procedure. This enables you to incorporate the
generated code into larger systems that decide when to execute the generated
code. Conceptually, you can think of the generated code as set of equations,
wrapped in a function called by your supervisory code. This facilitates
integration of model code into large existing systems, or into environments that
consist of more than signal-flow processing (Simulink) and state machines
(Stateflow).
Monitoring and Parameter Tuning APIs
External mode provides a communication channel for interfacing the
generated code running on your target with Simulink. External mode lets you
use Simulink as a debugging front end for an executing model. Typically, the
external mode configuration works in conjunction with either the real-time
code format or the real-time malloc code format.
Real-Time Workshop provides other mechanisms for making model signals and
block parameters visible to your own monitoring and tuning interfaces. These
mechanisms, suitable for use on all code formats, include:
D-33
D
The Real-Time Workshop Development Process
• The Model Parameter Configuration dialog box, where you declare how to
allocate memory for variables that are used in your model. For example, if a
Gain block contains the variable k, you can declare k as an external variable,
a pointer to an external variable, a global variable, or let Real-Time
Workshop decide where and how to declare the variable.
The Model Parameter Configuration feature enables you to specify block
parameters as tunable or global. This gives your supervisory code complete
access to any block parameter variables that you may need to alter while
your model is executing. You can also use this feature to interface
parameters to specific constant read-only memory locations.
• You can mark signals in your model as test points. Declaring a test point
indicates that you may want to see the signal’s value while the model is
executing. After marking a signal as a test point, you specify how the
memory for the signal is to be allocated. This gives your supervisory code
complete read-only access to signals in your model, so that you can monitor
the internal workings of your model.
• C and Target Language Compiler APIs provide another form of access to the
signals and parameters in your model. The Target Language Compiler API
is a means to access the internal signals and parameters during code
generation. With this information, you can generate monitoring/tuning code
that is optimized specifically for your model or target.
Interrupt Support
Interrupt blocks enable you to create models that handle synchronous and
asynchronous events, including interrupt service routines (ISRs),
hardware-generated interrupts, and asynchronous read and write operations.
The blocks provided work with the Tornado target. You can use these blocks as
templates when creating new interrupt blocks for your target environment.
Interrupt blocks include
• Asynchronous Interrupt block
• Task Synchronization block
• Asynchronous Buffer block (read)
• Asynchronous Buffer block (write)
• Asynchronous Rate Transition block
D-34
Index
A
application modules
application-specific components 7-33
definition of 7-23
system-independent components 7-29
atomic subsystem 4-2
automatic S-function generation 10-11
See also S-function target
B
block reduction optimization 2-27
block states
Simulink data objects and 5-54
State Properties dialog and 5-51
storage and interfacing 5-49
storage classes for 5-50
symbolic names for 5-52
blocks
Asynchronous Buffer 13-16
Asynchronous Interrupt 13-5
depending on absolute time B-1
device driver 14-39
Rate Transition 8-12, 13-18
scope 2-24
S-Function 14-39
Task Synchronization 13-12
to file 2-24
to workspace 2-24
buffer reuse option 2-12
Build button 2-3
build command 2-3
C
code format
choosing 3-3
embedded C 3-11
real-time 3-6
real-time malloc 3-8
S-function 3-10
code generation 2-1
and simulation parameters 2-21
from nonvirtual subsystems 4-2
TLC variables for 14-18
code generation options
advanced 2-26
block reduction 2-27
Boolean logic signals 2-29
buffer reuse 2-12
see also signal storage reuse
Expression folding 9-3
external mode 2-17
force generation of parameter comments 2-12
general 2-7
generate HTML report 2-10
Ignore custom storage classes 2-17
inline invariant signals 2-11
inline parameters 2-26
local block outputs 2-12
see also signal storage reuse
loop rolling threshold 2-9
MAT-file variable name modifier 2-17
retain .rtw file 2-19
show eliminated statements 2-8
signal storage reuse 2-32
see also Llocal block outputs
Solver pane 2-21
target specific 2-16
TLC debugging 2-18
verbose builds 2-10
Workspace I/O pane 2-22
code reuse
I-1
Index
diagnostics for 4-13
enabling 4-10
presence of blocks preventing 4-11
subsystem characteristics preventing 4-11
code tracing
via <CF>hilite_system command 2-33
via HTML reports 2-33
combining models
in RTW Embedded Coder target 14-103
via grt_malloc target 14-103
via S-function target 14-103
communication
external mode 6-2
external mode API for 14-94
context-sensitive help 2-4
continuous states, integration of 7-28
custom target configuration
components of 14-3
tutorial 14-9
D
data logging 2-22
in single- and multitasking models 2-24
to MAT-files 2-22
via scope blocks 2-24
via to file blocks 2-24
via to workspace blocks 2-24
Data Store Memory blocks
Simulink data objects and 5-59
data structures in generated code
block I/O 7-20
block parameters 7-20
block states 7-20
external inputs 7-20
external outputs 7-20
device driver blocks
I-2
building 14-59
creating custom 14-39
DOS C-10
inlined 14-53
example 14-54
mdlRTW function in 14-57
when to inline 14-41
limitations of 14-42
noninlined 14-44
conditional compilation of 14-45
required defines and includes 14-46
required functions 14-47
parameterizing 14-43
requirements for 14-42
S-function wrappers for 14-59
VxWorks 12-13
device drivers
inlining with S-functions 14-53
directories
used in build process 2-51
DOS
Analog Input (ADC) block parameters C-12
Analog Output (DAC) block parameters C-14
building programs C-18
device driver blocks C-10
Digital Input block parameters C-15
Digital Output block parameters C-16
hardware requirements C-6
implementation overview C-4
interrupt service routine 8-3
Keithley Metrabyte board I/O driver C-10
modifying program timing C-8
program timing C-7
sample rate limits C-7
software requirements C-6
system configuration C-5
DOS Device Drivers library C-2
Index
doslib block library C-10
DSP processor support 14-107
E
Euler integration algorithm 7-28
Expression folding 9-3
configuring options for 9-6
in S-Functions 9-10
ext_comm MEX-file 6-28
external mode 6-2
API
host source files 14-97
implementing transport layer 14-100
target source files 14-98
architecture 6-24
blocks compatible with 6-19
code generation option 14-31
command line options for target program 6-31
communication channel creation 14-94
communications overview 14-95
configuring to use sockets 12-7
design of 14-94
download mechanism 6-23
error conditions 6-32
ext_comm MEX-file 6-28
optional arguments to 6-28
rebuilding 14-101
host and target systems in 6-2
limitations of 6-33
Signal Viewing Subsystems in 6-19
TCP implementation 6-26
using with VxWorks 12-6
Generate comments option 2-16
Generate scalar inlined parameters option
2-16
Include data type acronym in identifier 2-14
Include system hierarchy number in identifiers
2-14
Maximium identifier length 2-13
Prefix model name to global identifiers 2-15
generated code
operations performed by 7-29
generated files
contents of 2-47
dependencies among 2-48
model_pt.c 14-77
generated s-functions
tunable parameters in 10-9
H
harware-specific characteristics
communication of 14-7
hook files 14-7
host
in external mode 6-2
I
interrupt service routine
locking and unlocking 13-11
under DOS C-5
under VxWorks 8-3
Interrupt Templates library 13-5
L
G
general code appearance options
libraries
DOS Device Drivers C-2
I-3
Index
Interrupt Templates 13-5
VxWorks support 12-3
local block outputs option 2-12
M
make command 14-32
make_rtw 2-6
MAT-files
file naming convention 2-23
logging data to 2-22
variable names in 2-23
MATLAB D-5
mdlRTW function 14-57
model code
execution of 7-31
model execution
in real time 8-10
in Simulink 8-9
Simulink vs. real-time 8-9
Model Parameter Configuration dialog
tunable parameters and 5-2
using 5-8
model registration function 7-31
multiple models
combining 14-103
multitasking
building program for 8-8
enabling 8-8
example model 8-22
task identifiers in 8-5
task priorities 8-5
versus singletasking 8-3
N
nonvirtual subsystem code generation
I-4
Auto option 4-3
Function option 4-7
Inline option 4-5
Reusable function option 4-10
nonvirtual subsystems
atomic 4-2
categories of 4-2
conditionally executed 4-2
modularity of code generated from 4-13
O
operating system
tasking primitives 7-9
VxWorks 12-2
P
parameters
interfacing 5-2
storage declarations 5-2
tunable 5-2
tuning 5-2
C API for 14-77
priority
of sample rates 8-5
of VxWorks tasks 12-10
program architecture
embedded 7-34
initialization functions 7-25
main function 7-25
model execution 7-27
program execution 7-13
program termination 7-26
program timing 7-12
rapid prototyping 7-24
real-time 7-24
Index
termination functions 7-31
pseudomultitasking 8-5
R
rapid prototyping 1-5
for control systems 1-9
for digital signal processing 1-8
rapid simulation target
batch simulations (Monte Carlo) 11-15
command line options 11-7
limitations 11-16
output filename specification 11-12
parameter strructure access 11-8
signal data file specification 11-9
Simulink license checkout 11-3
Rate Transition block 8-12
real time
executing models in 8-10
integrating continuous states in 7-28
real-time malloc target 3-8
combining models with 14-103
Real-time model
description 7-30
Real-Time Workshop
open architecture of 1-11
user interface 2-2, 4-2, 5-2
Real-Time Workshop pane
Category menu 2-2
General code appearance options 2-13
opening 2-2
overview 2-2, 4-2
Target configuration options 2-5
Browse button 2-5
generate code only option 2-7
make command field 2-6
system target file field 2-5
template makefile field 2-6
target configuration options
Build button 2-3
rsim. See rapid simulation target
rtw_local_blk_outs 5-21
S
sample rate transitions 8-12
faster to slower
in real-time 8-16
in Simulink 8-16
slower to faster
in real-time 8-19
in Simulink 8-18
sample time overlaps 8-19
S-function target 3-10
applications of 10-2
automatic S-function generation 10-11
generating reusable components with 10-3
intellectual property protection in 10-2
restrictions 10-15
tunable parameters in 10-9
unsupported blocks 10-17
S-functions
API 7-32
generating automatically
models containing 7-31
noninlined 7-31
signal properties 5-30
setting via Signal Properties dialog 5-30
signal storage reuse option 2-32
Signal Viewing Subsystems 6-19
SimStruct data structure
and global registration function 7-31
definition of 7-29
Simulation Parameters dialog
I-5
Index
Real-Time Workshop pane 4-2, 5-2
Simulation parameters dialog
Advanced pane 2-26
Diagnostics pane 2-25
Solver options pane 2-21
Workspace I/O pane 2-22
simulation parameters dialog
Real-Time Workshop pane 2-2
Simulink D-5
interactions with Real-Time Workshop 2-35
block execution order 2-37
sample time propagation 2-36
simulation parameters
and code generation 2-21
Simulink data objects 5-32
parameter objects 5-34
signal objects 5-39
singletasking 8-8
building program for 8-8
enabling 8-9
example model 8-22
step size
of real-time continuous system 7-28
StethoScope
See VxWorks
subsystem
nonvirtual 4-2
System target file browser 2-40
T
target
available configurations
bundled D-21
custom D-22
third-party D-22
custom configuration 14-2
I-6
rapid simulation See rapid simulation target
real-time malloc See real-time malloc target
Target Language Compiler
code generation variables 14-18
debugging options 2-18
targets
API for setting hardware characteristics 14-7
available configurations 2-42
task identifier (tid) 8-5
template makefile
compiler-specific 2-54
default 2-51
structure of 14-28
tokens 14-29
template makefile options
Borland 2-57
LCC 2-58
UNIX 2-55
Visual C/C++ 2-55
Watcom 2-57
tokens 14-29
tooltips 2-4
tunable expressions 5-2
in masked subsystems 5-13
limitations on 5-12
tutorials
creating custom target configuration 14-9
V
VxWorks
and external mode 12-6
application overview 12-5
configuring
for external mode (sockets) 12-6
makefile template 12-13
connecting target to Ethernet 12-5
Index
downloading and running the executable
interactively 12-18
external mode options 12-7
GNU tools for 12-14
implementation overview 12-11
program build options 12-14
program execution 12-19
program monitoring 12-5
real-time operating system 12-2
runtime structure 12-8
StethoScope code generation option 12-17
support library 12-3
target
connecting to 12-18
downloading to 12-18
target CPU 12-5
tasks created by 12-9
template makefiles 12-13
I-7
Index
I-8
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement