Problem with compiler C to use int16_t type for translation in Q15 format

Hello,

I have a problem with compiler C. I have this error message

Error using legacycode.LCT.legacyCodeImpl
The data type "int16_t" is neither a built-in data type, nor defined by
a Simulink.AliasType, Simulink.NumericType, Simulink.Bus object, or an
enumerated data type:
--> int16_t y1 = ControlloDig(int16_t u1)

Error in legacy_code (line 101)
[varargout{1:nargout}] = legacycode.LCT.legacyCodeImpl(action,
varargin{1:end});

Error in Discretizza_controllo_generazione_sfun (line 77)
legacy_code('sfcn_cmex_generate', def);

I did a function that works in Q15 format but, when I use the script to compile it, Matlab shows an error on the data type "int16_t". If you need the other files i can upload them.

Answer 1

[Sebastian K]

Hi there,

I am afraid it will not be possible to provide a very helpful answer based on the limited information you have provided. The error message itself is not quite useful. Ideally you should provide a detailed explanation of:

• What you are trying to do
• How you are trying to do it
• At what stage of the process you encounter the error

Just by looking at the error message, I am guessing that you are probably trying to use the Legacy Code Tool to build a C MEX S-function from some hand-written C code.

The main issue looks like the "int16_t" data type is not defined. One potential reason for this could be that you are missing a header file in your C code that contains the data type definition. Try adding the following "include" directive to your C code.

#include <tmwtypes.h>

I hope this helps.

Sebastian

Answer 2

[Bisca]

Hi Sebastian,

Thanks for the help, you guessed the problem. Sorry if I gave few details about the problem. I'm going to write an s-function for a dc-motor control. In particular, the s-function could be work in Q15 format and when i do this Matlab shows me that error. I try to do your suggest but the problem persist. I notice that the error disappear when I use only int16. For this reason, I think that the error is on the int16_t because Matlab doesn't reconize the datatype.

Now I write the code

clear all
close all
clc

%%Gain PI
Ki = 4;
Kp = 1;

% ---------------------------------
Ts=1/(2.5e3);   
method='tustin';
%  method = 'zoh';

contol's transfer function

PI=tf([Kp Ki],[1 0])

%%discretize
PI_discr=c2d(PI, Ts, method)

%%simulink
[numD, denD, Ts]=tfdata(PI_discr);

% TODO

A=numD{1,1};   
B=denD{1,1}(2:end); 

%%generazione codice versione con double
fID=fopen ('ControlloDig.h', 'wt');

fprintf(fID, '#include <stdint.h> \n\n');
fprintf(fID, '#include <tmwtypes.h> \n\n');
fprintf(fID, '#include <stdlib.h> \n\n');
fprintf(fID, '#ifndef _CONTROLLO_DIG_H_ \n');
fprintf(fID, '#define _CONTROLLO_DIG_H_ \n\n');

fprintf(fID, '#define Q15 \n\n');
fprintf(fID, '#define Q15(X) ((X < 0.0) ? (int16_t)(32768*(X) - 0.5) : (int16_t)(32767*(X) + 0.5))  \n\n');
fprintf(fID, '#define iQ15(X) ((X < 0.0) ? (double)(X+1)/32768.0 : (double)(X-1)/32767.0) \n\n');
fprintf(fID, '#define K   (1 << (Q - 1)) \n\n');

fprintf(fID, 'static int16_t A[2]; \n');
fprintf(fID, 'static int16_t B; \n');
fprintf(fID, 'static int16_t ek=0; \n');
fprintf(fID, 'static int16_t uk=0; \n\n');
fprintf(fID, 'static int16_t ek_1=0; \n');
fprintf(fID, 'static int16_t uk_1=0; \n\n');
fprintf(fID, 'static int16_t y1=0; \n\n');

fprintf(fID, 'int16_t ControlloDig(int16_t ek); \n\n');

fprintf(fID, '#endif \n');
fclose (fID);

%%compilazione
% per una guida più dettagliata, guarda link seguente: 
% http://www.mathworks.it/it/help/simulink/slref/legacy_code.html

def = legacy_code('initialize');

def.SourceFiles = {'ControlloDig.c'};
% def.HeaderFiles = {'Copy_of_ControlloDig.h'};
def.HeaderFiles = {'ControlloDig.h'};
def.SFunctionName = 'ControlloDigFun16';
def.OutputFcnSpec = 'int16_t y1 = ControlloDig(int16_t u1)';
def.SampleTime='parameterized';   % permette il settaggio dei parametri
legacy_code('sfcn_cmex_generate', def);

legacy_code('compile', def);

legacy_code('slblock_generate', def);

[Walter Roberson]

int16_t and so on are in stdint.h which was introduced in C99. Possibly you do not have the flags set to use C99 ?

[Bisca]

I don't know if I flagged or not. How can I check this?

[Walter Roberson]

You would generally be able to tell from the complete command output of the compiling process. On unix derived compilers look for std=c99 or std=c11 with std=c89 being before that. I do not recall the flag used for the visual studio compilers

[Walter Roberson]

Unfortunately it appears that Microsoft decided not to support C99, except for the parts that are also C++; see https://herbsutter.com/2012/05/03/reader-qa-what-about-vc-and-c99/. So do not expect a Visual Studio flag to indicate C99 conformance. However, the articles does note that stdint.h is supported.

I see that VS2015 started adding conformance flags, starting with /std:c++14 (no earlier conformance flag appears to be available.)

I am not clear as to which compiler you are using ?

[Bisca]

I'm using 'MinGW64 Compiler (C++)'

[Jan]

int16_T (uppercase T) is defined in tmwtypes.h .

[Bisca]

Unfortunately, it isn't work