|
|
|
| R2012a Documentation → Simulink Fixed Point | |
Learn more about Simulink Fixed Point |
|
| Contents | Index |
This table summarizes what's new in Version 6.3 (R2010a):
| New Features and Changes | Version Compatibility Considerations | Fixed Bugs and Known Problems |
|---|---|---|
| Yes Details below | No | Bug
Reports Includes fixes |
New features and changes introduced in this version are
Trigonometric Function Block Supports CORDIC Algorithm and Fixed-Point Signals
Improved Automatic Scaling Handles Data Type Constraints for Several Simulink Blocks
Direct Lookup Table (n-D) Block Supports Fixed-Point Signals
Stateflow Support for Chart-Level Data with Fixed-Point Word Lengths Up to 128 Bits
Root Inport Support for Fixed-Point Data Contained in a Structure
The Trigonometric Function block now accepts and outputs fixed-point signals when you select sin, cos, or sincos and the approximation method is CORDIC.
In R2010a, casting from single to fixed-point data types no longer inserts unnecessary intermediate double-precision variables. Removing these intermediate variables:
Results in more efficient code, particularly for embedded targets.
Enables single to fixed-point casts for targets without double-precision arithmetic support.
Autoscaling with the Fixed-Point Tool and Fixed-Point Advisor now handles data type constraints for ports on several Simulink blocks. For example, autoscaling now takes into account that:
The index port of the Selector and Assignment blocks support only double, single, and built-in integer data types.
The input port of the Data Type Conversion supports only built-in integer data types when the block is configured to output an enumerated type.
The index port of the Interpolation Using Prelookup supports only integer data types.
This improved autoscaling reduces data type mismatch errors and enables the Fixed-Point Tool and Fixed-Point Advisor to provide additional diagnostic information when you accept autoscaling proposals. For more information, see Constrained Data Type Summary in the Simulink Fixed Point User's Guide.
The Direct Lookup Table (n-D) block accepts fixed-point data types for the table input port.
In R2010a, the Real-Time Workshop® software no longer unconditionally replaces divisions by constant power of 2 with casts. The software now replaces division by constant power of 2 with a cast only if this replacement results in less generated code. This enhancement relies on the target compiler to optimize the division appropriate to the target processor.
The decision whether to replace the division is based on these guidelines:
If the replacement by a cast results in extra rounding code, Real-Time Workshop® does not replace the division.
If the division requires a helper function, Real-Time Workshop® replaces the division with a cast even if the cast requires extra rounding code.
The Fixed-Point Advisor now supports restore points. Restore points provide you with the ability to:
Save a snapshot of your model at any time during conversion from floating point to fixed point.
Revert any changes made in response to advice from the Fixed-Point Advisor.
Load and rerun from any restore point without the need to run through the entire conversion process.
For more information, see Restore Points in the Simulink Fixed Point User's Guide.
The Fixed-Point Advisor now provides:
A wired-subsystem replacement for the State-Space block.
A preview of the replacement options for an unsupported block, when available.
A new context menu option to replace an unsupported block.
For more information, see Address unsupported blocks in the Fixed-Point Advisor Reference.
Target Function Library enhancements include:
Ability to create custom Target Function Library entries
TFLs now support custom entries that allow you to specify near-arbitrary match criteria. You first create your own TFL entry class, derived from either RTW.TflCFunctionEntryML (for function replacement) or RTW.TflCOperationEntryML (for operation replacement). In your derived class, you implement a do_match method with a fixed preset signature and customize the match criteria. You also can modify the implementation signature to meet your application needs. For more information, see Refine CRL Matching and Replacement Using Custom CRL Table Entries in the Real-Time Workshop® Embedded Coder™ documentation.
Additional scalar operator replacements
New TFL support for replacing scalar complex operations, including addition, subtraction, multiplication, cast, and complex conjugate. Mixed types are supported.
Additionally, you can now replace fixed-point shift right for all fixed-point input types.
Stateflow chart-level data now support up to 128 bits of fixed-point precision for the following scopes:
Input
Output
Parameter
Data Store Memory
This increase in maximum precision from 32 to 128 bits provides these enhancements:
Supports generating efficient code for targets with non-standard word sizes
Allows charts to work with large fixed-point signals
You can explicitly pass chart-level data with these fixed-point word lengths as inputs and outputs of the following functions:
Embedded MATLAB® functions
Simulink functions
Truth table functions that use Embedded MATLAB action language
For more information, see Using Fixed-Point Data in Stateflow Charts in the Stateflow and Simulink Coder User's Guide.
You can now use a root (top-level) Inport block to supply fixed-point data that is contained in a structure. In releases before R2010a, you had to use a Simulink.Timeseries object instead of a structure.
The To File and From File blocks now support fixed-point data with a word length of up to 32 bits.
 | Version 6.4 (R2010b) Simulink Fixed Point Software | Version 6.2 (R2009b) Simulink Fixed Point Software |  |
Learn more about Simulink through this collection of videos, articles, technical literature and the Getting Started with Simulink Guide.
| © 1984-2012- The MathWorks, Inc. - Site Help - Patents - Trademarks - Privacy Policy - Preventing Piracy - RSS |
