Skip to main content

PLCOPEN-CP8: Floating point comparison shall not be equality or inequality

Use a tolerance instead of = or <> when comparing REAL values.

Why is it bad?​

Floating-point arithmetic is approximate: the result of 0.1 + 0.2 is not exactly 0.3. Comparing two reals for equality almost never does what the author intended — small rounding errors that accumulate through arithmetic make a = b return FALSE when the values are mathematically equal.

The fix is to compare the absolute difference against a tolerance value appropriate for the application.

Example​

PROGRAM demo
  VAR x : REAL; END_VAR
  IF x = 0.0 THEN     (* PLCOPEN-CP8 *)
    x := x + 1;
  END_IF;
  IF x <> 0.0 THEN    (* PLCOPEN-CP8 *)
    x := x + 1;
  END_IF;
END_PROGRAM

Use instead:

PROGRAM demo
  VAR
    x : REAL;
    epsilon : REAL := 1.0E-6;
  END_VAR
  IF ABS(x) < epsilon THEN
    x := x + 1;
  END_IF;
END_PROGRAM

Resources​

  1. PLCOpen Software Construction Guidelines
  2. The Floating-Point Guide — Comparison