How to prevent replacing variables with their values in matlab coder?

Let's say I have the following function in Matlab:

function y = dum_func(x)

my_constant = 5;

y=my_constant+x;

end

The corresponding C++ function generated by the coder is the following:

double dum_func(double x)

{

// 'dum_func:3' my_constant = 5;

// 'dum_func:4' y=my_constant+x;

return x + 5.0;

}

Question) Is there a way to keep my_constant as variable in the c++ code?

Thanks.

Answer 1

[Mukund Sankaran]

Hello,

What you are observing is due to an optimization called constant folding. This is explained here: https://www.mathworks.com/help/coder/ug/matlab-coder-optimizations-in-generated-cc-code.html

For the particular example you shared, unfortunately, there is no straightforward way to get the generated code looking the way you expect today. However, FWIW, if your MATLAB code is rewritten to look like this:

function y = dum_func(x)
   my_constant = 5;
   coder.ceval('(void)', coder.ref(my_constant));
   y = my_constant + x;
end

Then the generated code will look like this:

double dum_func(double x)
{
  double my_constant;
  my_constant = 5.0;
  (void)(&my_constant);
  return my_constant + x;
}

The use of ceval might stop all optimizations on the variable my_constant and yield the result above.

Hope this helps!