Chapter 14  Preparing Models for Use with Reactis

To use Reactis on a Simulink / Stateflow model, you must ensure that the model satisfies certain constraints. This chapter describes what these constraints are. Please note that, while every effort is made to maintain the accuracy of the following list of supported and unsupported features, due to the complexity and continual evolution of the Simulink / Stateflow notation this description may not be complete. For example, some blocks listed as supported do not support all parameters for the block. A block or feature not listed as either supported or unsupported should be assumed to be unsupported.

We encourage all potential customers to take advantage of a free evaluation license for Reactis to determine if the subset of Simulink / Stateflow supported by Reactis is adequate for your models. If not, please let us know what features or blocks are lacking. Reactive Systems uses such feedback to prioritize enhancements to Reactis.

14.1  MATLAB

Reactis can process Simulink / Stateflow models that contain MATLAB code in callbacks or mask initializations, provided that certain conventions are observed. On the one hand, files that define workspace data items referred to in a model must be “connected” to the model appropriately, and certain MATLAB functions should be avoided. The remainder of this section elaborates on these points.

14.1.1  Workspace Data Items

Reactis invokes MATLAB to evaluate many MATLAB expressions, but it does not directly interact with an executing MATLAB session in the same way that Simulink does. For this reason, any workspace data items that a model uses must be included in .m files that are attached to a model using Simulink’s pre-load callback. For the cruise-control example, the file cruise_constants.m defines two workspace variables that are used in cruise.mdl. One attaches cruise_constants.m to cruise.mdl as follows:

1. Load cruise.mdl into Simulink.
2. From the Simulink window, select the File -> Model Properties menu item.
3. In the resulting dialog, select the Callbacks tab.
4. In the Model pre-load function entry box enter cruise_constants.
5. Save the model.

In general, using the pre-load callback in this manner is good modeling practice, since once the .m files are attached to a model file, loading the model file into Simulink (and not just Reactis) will automatically load the .m files as well.

14.1.2  Unsupported MATLAB Features

While Reactis supports most of MATLAB , the following exceptions may not be used in Simulink / Stateflow models, e.g. in callbacks or mask initializations. Note, that a feature added in V2010.2 adds partial support for these commands. To enable the feature, check “Propagate set_param changes by saving the model to a temporary file” in the Reactis Settings Dialog. If this setting is enabled, Reactis will cause Simulink to apply all the changes to the model, then automatically save them to a temporary file. Reactis then imports the temporary model to see the applied changes.

14.2  Simulink

Reactis currently supports Simulink releases R13 through R2011b. Most features of Simulink are supported; but the following are not supported by Reactis:

• Continuous-time blocks.
• Simulink signals with more than 2 dimensions.
• The use of complex values (i.e. values with real and imaginary parts).
• Models containing corresponding DataStoreWrite/DataStoreRead blocks whose execution order is not explicitly defined by either the model logic or priorities.

For the subset of Simulink blocks supported by Reactis please refer to Section 14.2.3. For blocks that can be either continuous- or discrete-time, only the discrete-time version is supported. For some blocks identified as supported not all settings are supported.

14.2.1  S-Functions

Reactis supports both C-Coded and M-File S-functions, with some restrictions. For C-Coded S-functions, the following are not supported by Reactis:

• Port-based sample times.
• Multiple sample times.
• Complex number signals.
• Zero-crossing detection.
• Output of function-calls.
• Level 1 S-functions (For a guide on how to convert Level 1 S-functions to Level 2 S-functions — which are supported by Reactis - please consult the MathWorks documentation)
• Calling any function from MATLAB’s “mex” library (including mexCallMATLAB, mexEvalString and mexGetVariable) from an S-Function.

For M-File S-functions, the following are not supported by Reactis:

• Multiple sample times.
• Complex number signals.
• Level 2 S-functions.

In addition to the above restrictions, care must be taken about any internal data that is stored by S-functions. In order to work properly, Reactis must be able to retrieve and reset the values of all internal states of any S-function occurring within a model. The best way to make internal states visible to Reactis (and Simulink) is to declare the appropriate number of discrete states in the mdlInitializeSizes() function and then use the state vector that Reactis and Simulink will provide. Reactis will also save and restore memory that an S-function allocates as a result of calling the ssGetNumRWork() and ssGetNumIWork() during mdlInitializeSizes.

Reactis has no way of knowing about any other persistent data that an S-function maintains by other means, such as:

• global or static variables in C-code;
• memory allocated by malloc() or mxMalloc() functions in C-code;
• Use of workspace variables in M-File S-functions.

Reactis will also not save and restore memory requested by ssGetNumPWork(), since otherwise pointers stored in this vector by your S-function might get lost or mangled, resulting in memory leaks or crashes.

If an S-function stores internal states in any of the unsupported ways described above, Reactis will seem to work properly, but the test suites generated by Tester may include wrong outputs. One sign of this can be if you run a test suite in Simulator and get an error message saying “Model fails test”. Another problem of such invalid use of internal states may be invalid outputs after using the “back” buttons in Simulator.

In general, Reactis passes S-Function parameters as fixed values at the time the S-Function is first initialized (i.e. when Tester or Simulator is started). Therefore, if a configuration variable is used as a parameter to an S-Function, the S-Function will not see any changes to the configuration variable unless the S-Function is designed to process such updates. To have an S-Function be updated on changes to its parameters, define a mdlProcessParameters function (see Simulink documentation) within the S-Function code. If this function is present, then Reactis will propagate the parameter changes into the S-Function by calling the S-Function’s mdlProcessParameters function at each step with the new parameter values. In the mdlProcessParameters function the S-Function can then take appropriate actions if any parameters have changed.

14.2.2  Lookup Tables

This section describes the settings for each of the standard Simulink Lookup Table blocks that are supported by Reactis. Some settings might not be relevant depending on the version of MATLAB used. For example, many data type settings for the “Lookup Table (n-D)” block were only introduced in MATLAB R2009b and therefore are not relevant when using Reactis with MATLAB R2009a or earlier.

1-D Lookup Table

Note that this block was previously named Lookup Table block. All block settings and the following data type combinations are supported. “fixpoint” includes all fixed-point data types as well as built-in integer data types int8, uint8, int16, uint16, int32, and uint32.

2-D Lookup Table

Note this block was previously named Lookup Table (2-D). All block settings and the following data type combinations are supported. “fixpoint” includes all fixed-point data types as well as built-in integer data types int8, uint8, int16, uint16, int32 and uint32.

n-D Lookup Table

This block was previously named Lookup Table (n-D). Reactis supports a wide variety of n-D lookup tables, subject to restrictions which depend on the number of table dimensions. There are two levels of support, native and via S-function.

Native support means Reactis will natively execute the n-D lookup table block and track coverage for it. Only n-D lookup tables which satisfy the following requirements are natively supported in Reactis:

1. The number of table dimensions is at most two.
2. The interpolation and extrapolation methods are not cubic spline.
3. Each breakpoint data type matches the corresponding input data type.
4. The output data type matches the table data type.
5. If the table is 1-dimensional, then the fraction data type must satisfy at least one of the following requirements:
   1. The fraction, input, and output data types are floating-point types.
   2. The fraction and output data types are fixpoint types.
   3. The fraction and output data types are the same floating-point type and the input data type is a fixpoint type.
   4. The fraction and input data type are the same floating-point type and the output data type is a fixpoint type.
6. If the table is 2-dimensional, then both of the following requirements must be satisfied:
   1. Both input types must be the same.
   2. The intermediate data type matches the output data type.

Support via S-function means Reactis can execute the n-D lookup block but not track coverage. In order to execute the block, Reactis requires the sfun_lookupnd.mexw32 S-Function that comes with MATLAB R2007a or earlier. Starting with MATLAB R2007b, this S-Function is no longer included in the MATLAB distribution. However, if a sfun_lookupnd.mexw32 from R2007a or earlier is placed in Reactis’ search path, Reactis can use that S-Function even with R2007b and later.

n-D lookup tables which meet the following six requirements are supported via S-function in Reactis:

1. The input and output data types are the same floating-point type.
2. The fraction data type is set to Inherit via internal rule.
3. The table data type is set to Inherit: same as output.
4. The breakpoint data type is set to Inherit: same as corresponding input.
5. All inputs are scalars.
6. If the table has more than one dimension, then the intermediate data type is set to Inherit: same as output.

Direct Lookup Table (n-D)

Only floating-point (double and single) data types are supported.

This table is supported in Reactis via the sfun_nddirectlook.mexw32 S-Function that comes with MATLAB R2009a and earlier. Starting with MATLAB R2009b, this S-Function is no longer included in the MATLAB distribution. However, if a sfun_lookupnd.mexw32 from R2009a or earlier is placed in Reactis’ search path, Reactis can use that S-Function even with R2009b and later.

Lookup Table Dynamic

All parameters and the data type configurations listed for the “Lookup Table” block above are supported.

Prelookup

All parameters are supported. The following data type restrictions apply:

• Only floating-point (double and single) input data types are supported.
• Only “int32” and “uint32” are supported as the data type of the “index” output.
• Only “single” and “double” are supported as the data type of the “fraction” output.

Interpolation using Prelookup

All parameters are supported. The following data type restrictions apply:

• Only floating-point (double and single) data types are supported for “fraction” inputs.
• Data types of all “fraction” inputs must be the same as the output data type.
• Only “int32” and “uint32” are supported as the data type of the “index” input.
• Table data type must match output data type
• Intermediate results data type must match output data type.

14.2.3  Table of Supported Blocks

14.2.4  Simulink Extras

The following table lists the supported blocks from the “Simulink extras” library which is available in all MATLAB versions:

14.2.5  TargetLink 2.2.1 Library Blocks

The following table lists the supported blocks from the dSPACE TargetLink 2.2.1 library, if installed:

14.2.6  TargetLink 3.0 Library Blocks

The following table lists the supported blocks from the dSPACE TargetLink 3.0 library, if installed:

14.3  Stateflow

Reactis supports most of Stateflow. Some exceptions are the following unsupported features:

• Implicit “enter”, “exit” and “change” events
• Range limits for variables.
• “ml.” namespace operator and “ml()” function call.
• Embedded MATLAB code.
• When using expressions in the ’size’ field of a Stateflow variable declaration, all references to constants within the expression must refer to constants that are defined in the MATLAB workspace.
• Using Stateflow keywords as variable names. The Stateflow keywords are: at, after, before, change, du, during, enter, en, entry, every, ex, exit, in, on, ml, send, abs, acos, asin, atan, atan2, ceil, cos, cosh, exp, fabs, floor, fmod, labs, ldexp, log, log10, min, max, pow, rand, rem, sin, sinh, sqrt, tan, tanh, int8, int16, int32, uint8, uint16, uint32, double, boolean.
• The explicit type cast function (“cast”)
• The address operator (&) is supported only in calls to external C functions.
• The pointer operator (*) is supported only inside a literal C code section.
• Charts without trigger or sample time are supported only if the model has a fixed sample time.
• In some cases, the detection of inner transitions fails for odd-shaped transitions (for example, single transition segments which leave and reenter a state).
• Using “bind” actions.
• Graphical functions with multiple outputs.
• Absolute time in temporal logic operators (i.e. using the sec keyword).
• The temporalCount operator.
• Simulink function-call subsystems.
• Atomic Subcharts (introduced in R2010b).