matlab interp C code generation, Waring: comparing floating point with == or != is unsafe

Question

da wu 2022-6-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1746170-matlab-interp-c-code-generation-waring-comparing-floating-point-with-or-is-unsafe

回答： Szabolcs Fodor 2023-10-25

I use interp1,interp2 in a .m file and generate C code, then the _sharedutils folder has a new xxxxx_interp1.c file to realize interp function, but the generated C code use "if floating point == floating point ", so C complier warn: comparing floating point with == or != is unsafe, how can I solve the problem in matlab and let it generate the rigtht C code.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

SlipperyGnome 2022-6-23

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1746170-matlab-interp-c-code-generation-waring-comparing-floating-point-with-or-is-unsafe#answer_991725

在 MATLAB Online 中打开

Hi da wu,

When comparing floating point numbers, because of very small rounding off errors, using '==' or '!=' will generate compiler warnings.

You can set an error tolerance upto the magnitude you need it to be same.

num1=0.8-0.5;
num2=0.3;
Error_tolerance= (1e-15)*max(num1,num2);
if (abs(num1-num2) < Error_tolerance)
    disp(0);
else
    disp("not 0")
end
     0

This tolerance value method would work accurately upto 1e-15. You can also go to this link for better understanding.

2 个评论
显示无隐藏无

da wu 2022-6-24

在 MATLAB Online 中打开

Hi SlipperyGnome,

Firstly, thank you very much for your answer.

I know, when I write a code by hand to realize " if single == single", I can use your method above. But the situation is that I use the matlab function " interp1", and generate C code automatically. Maybe a little different.

A test example:

function y = fcn(u)
y = interp1([0,20.5,50.2],[0,30.3,40],u);

Generated C code: (if single == single occurs)

void interp_test_step(void)
{
  static const real_T b[3] = { 0.0, 20.5, 50.2 };
  static const real_T c[3] = { 0.0, 30.3, 40.0 };
  int32_T low_i;
  real32_T r;
  /* Outport: '<Root>/y' incorporates:
   *  MATLAB Function: '<Root>/MATLAB Function'
   */
  interp_test_Y.y = (rtNaNF);
  /* MATLAB Function: '<Root>/MATLAB Function' incorporates:
   *  Inport: '<Root>/u'
   */
  if ((!rtIsNaNF(interp_test_U.u)) && ((!(interp_test_U.u > 50.2)) &&
       (!(interp_test_U.u < 0.0F)))) {
    low_i = 0;
    if (interp_test_U.u >= 20.5F) {
      low_i = 1;
    }
    r = (interp_test_U.u - (real32_T)b[low_i]) / (real32_T)(b[low_i + 1] -
      b[low_i]);
    if (r == 0.0F) {
      /* Outport: '<Root>/y' */
      interp_test_Y.y = (real32_T)c[low_i];
    } else if (r == 1.0F) {
      /* Outport: '<Root>/y' */
      interp_test_Y.y = (real32_T)c[low_i + 1];
    } else if (c[low_i + 1] == c[low_i]) {
      /* Outport: '<Root>/y' */
      interp_test_Y.y = 30.3F;
    } else {
      /* Outport: '<Root>/y' */
      interp_test_Y.y = (1.0F - r) * (real32_T)c[low_i] + (real32_T)c[low_i + 1]
        * r;
    }
  }
}