How to use coder.ceval to call a void type C++ function

Hi,

I am trying to call an external C++ function, which is a void type function. Here is the code I wrote as an example

function [Y] = example(a,b)
    coder.updateBuildInfo('addSourceFiles','cpp_fun.cpp');
    coder.cinclude('cpp_fun.h');
    coder.ceval('cpp_fun', a, b, Y); 
end

The C++ function is defined as:

void cpp_fun(int a, float *b, float *Y);

Here int a, float *b are input of the function and *Y is suppose to be the output of the C++ function.

But when I use the Matlab Coder to convert the matlab function above, I get the following error message: "Variable 'Y' is not fully defined on some execution paths."

Shoud I use something like

Y = coder.opaque(type)

in the matlab function to define Y? If yes, what type should I use for Y?

Any help is greatly appreciated!

Answer 1

[Darshan Ramakant Bhat]

Try defining the variable 'y' in MATLAB function then try to use coder.wref()

Y = 0;
coder.ceval('cpp_fun', a, b, coder.wref(Y)); 

https://www.mathworks.com/help/simulink/slref/coder.wref.html

Similarly you may need to use coder.ref() for the second input 'b' since it is a pointer.

https://www.mathworks.com/help/simulink/slref/coder.ref.html

Hope this will help you.

[Erjian]

Hi Darshan,

Thanks fo your help! While I was trying your suggestions and I ran into other problems. I simplifed my C++ function to better understand how this should work. Here is the modified matlab function:

function [Y] = example(a,b)
if coder.target('MATLAB')
	% Executing in MATLAB, call function example
    Y = a+b;
else
    coder.updateBuildInfo('addSourceFiles','cpp_fun.cpp');
    coder.cinclude('cpp_fun.h');
    Y = 0;
    coder.ceval('cpp_fun', a, b, coder.wref(Y)); 
end
end

And here is the simplified C++ code:

void cpp_fun(int a, int b, float y)
{
    y = (float)a + (float)b;
}

And here is the screeshot of the error I got in the Matlab Coder Report Viewer:

How should I handle this data type inconsistency problem?

[Darshan Ramakant Bhat]

You need to use proper types at the interface. From the error I can see that you are passing "y" as pointer but in C++ code you are taking by value. This is giving the error. Also type of "y" is real_T, you need to make it as float.

I have modified your example and made it work. Please check the attached file and run doit.m file.

Hope this will help you.