Add pre-processor error when trying to synthesize mgc_floor

[gabeusse]

During synthesis, the calls to mgc_floor are stubbed to return 0.0, which can also be observed in the generated HDL. This change guards against unintended (indirect) uses of the function in code that is supposed to be translated into hardware. A basic test case to validate the change is:

#include <ac_int.h>
#include <iostream>

double t(double A[5], double inc) {
  for(int h = 0; h < 5; h++) {
    A[h] =  ac_private::mgc_floor(A[h] + inc);
  }
}

int main()
{   
  double A[5] = {1.2, 2.3, 3.4, 4.5, 5.6};
  t(A, 2.1);
  for(int h = 0; h < 5; h++)
    std::cout << "A[" << h << "] = " << A[h] << std::endl;
  return 0;
}

Where the output for simulation can be obtained with:

$ g++ -I/usr/local/bin/Mentor_Graphics/Catapult_Synthesis_10.6-912629/Mgc_home/shared/include t.cpp -o test_mgc
$ ./test_mgc 

A[0] = 3
A[1] = 4
A[2] = 5
A[3] = 6
A[4] = 7

And the behavior for synthesis (when SYNTHESIS is defined) is:

$ g++ -D__SYNTHESIS__ -I/usr/local/bin/Mentor_Graphics/Catapult_Synthesis_10.6-912629/Mgc_home/shared/include t.cpp -o test_mgc
In file included from t.cpp:331:0:
/usr/local/bin/Mentor_Graphics/Catapult_Synthesis_10.6-912629/Mgc_home/shared/include/ac_int.h:163:6: error: #error "mgc_floor on floating point types is not allowed for synthesis"
     #error "mgc_floor on floating point types is not allowed for synthesis"
      ^~~~~

Compilation was performed with g++ 7.5.0

[nkkav]

Can you include ac_int.h from the forked repository instead? I know that there is no test folder in ac_types so no good place to place your testcase.

I think you can also drop the f suffix from the float literals as you are using a double storage and the tested function uses double arguments (otherwise promotion will happen of course).