C: while loop using floating point never terminates

Question

By executing this code:

float f = 1.0;
while (f != 0.0) {
    f = f - 0.1;
    printf("%.1f\n", f);
}

It is expected that it would run 10 times and stop, but what turns out is that it inevitably goes into stack overflow. The same happens even if I change the while loop such that f goes to any other value below 1.0;

Anyone care to explain?

Are you sure that you have stack overflow rather than infinite loop? — Mikhail Vladimirov, Mar 22 '19 at 16:43

score 3 · Answer 1 · answered Mar 22 '19 at 16:43

3

It's dangerous to compare floats for equality/non-equality. float numbers are imprecise by nature because there are not enough bits to represent continuous nature of the float number.

Any time you compare floats, you should be comparing that the floats are within certain range of each other. For example:

while ((f - 0.0) > 0.00001)

answered Mar 22 '19 at 16:43

Kon

4,023
4
24
38

2

I read a similar post in VBA where the negative problem was solved with squaring where they do: `if ((f*f) > 0.00001) /*not equal*/` – Paul Ogilvie Mar 22 '19 at 16:46
...without which you should write `while (((f - 0.0) > 0.00001 || ((f - 0.0) < -0.00001)` – Paul Ogilvie Mar 22 '19 at 16:48
1

Comparing a _floating point_ to a constant takes the _floating_ out of floating point. A _relative_ compare more useful. – chux - Reinstate Monica Mar 22 '19 at 16:49

score 2 · Answer 2 · answered Mar 22 '19 at 16:55

Why did code loop forever?

Typical float can represent about 2³² numbers exactly, 0.1 is not one of them. So code is more like.

float f = 1.0;
while (f != 0.0) {
  f = f - a_number_close_to_one_tenth;
  printf("%.1f\n", f);
}

The repeated subtractions "miss" 0.0.

Instead use >= and code will iterate the expected number of times (or maybe once more). Use a higher precision print to see why.

float f = 1.0;
while (f >= 0.0) {
  f = f - a_number_close_to_one_tenth;
  printf("%.17e\n", f);
}

score 1 · Answer 3 · answered Mar 22 '19 at 16:51

1

It is a floating point issue. Try > instead of !=

float f = 1.0;
while (f > 0.0) {
    f = f - 0.1;
    printf("%.1f\n", f);
}

answered Mar 22 '19 at 16:51

Kevin Vandy

401
3
10

score 0 · Answer 4 · answered Mar 22 '19 at 17:02

0

Loops should better never be done on floating points.
It's even required by some coding standards like rule FLP30-C at https://wiki.sei.cmu.edu/confluence/display/c/FLP30-C.+Do+not+use+floating-point+variables+as+loop+counters

answered Mar 22 '19 at 17:02

B. Go

1,436
4
15
22

C: while loop using floating point never terminates

4 Answers4