This is kind of cool:
float Lifetime = FLT_MAX; // Use FLT_MAX for “infinite lifetime”
Lifetime -= 1.0f; // Or elapsed time or whatever, small value
=> the IEEE float machinery is such that “Lifetime” doesn’t change in this case. So an infinite lifetime is really infinite, automagically 
March 12th, 2008 at 4:40 am
Actually, you can do slightly better by actually using infinity, which exists in the IEEE representation.
The following code gives an example:
float f = numeric_limits::max();
float f1 = f - 1.0f;
float f2 = f * 0.5f;
float f3 = f - powf(2.0f, 103.0f);
cout << f << " " << (f == f1) << " " << (f == f2) << " " << (f == f3) << endl;
f = numeric_limits::infinity();
f1 = f - 1.0f;
f2 = f * 0.5f;
f3 = f - powf(2.0f, 103.0f);
cout << f << " " << (f == f1) << " " << (f == f2) << " " << (f == f3) << endl;
As you can notice, your FLT_MAX (described above using the numeric_limits scheme in C++) is a finite value which can actually be affected by a sufficiently large value (or by scaling).
On the other hand, if you use the actual representation of infinity, it will stay unaffected by those operations.
The hexadecimal representation of +inf is 0×7F800000 (fill the exponent with 1s and fill the mantissa with 0s), so if you don’t want to use numeric_limits, you can always typecast an int’s bits into a float (I believe C has no standard #define for infinity).
March 12th, 2008 at 6:11 am
Yes, you can also use +INF for this. I usually avoid it because I never remember if there’s an extra cost associated with it (like there is with NaNs). Just in case, I prefer avoiding those special float values. But it might just be paranoia