float value condition

Question

I worked out the following code and got a bizarre output. Can anyone explain how it works?

main()
{
    float a=0.8;
    float b=0.25;
    if(a==0.8)
        printf("HELLO")
    if(b==0.25)
        printf("WORLD")
}

And the output that I got is surprisingly

WORLD

thanks in advance

score 5 · Accepted Answer · answered Mar 26 '13 at 13:38

5

This is because 0.25 is a power of two (i.e. 2^-2), while 0.8 is not. Only exact sums of powers of two can be represented exactly; all other numbers, including 0.8, are represented as an approximation, which has a different precision between float and double. The 0.8 in a==0.8 is a double, while a is a float. Their representations are different, and so are their values.

answered Mar 26 '13 at 13:38

Sergey Kalinichenko

714,442
84
1,110
1,523

you are the fastest gun in SO , dude. I was clicking "post your answer", and I AM really quick. – Ander Biguri Mar 26 '13 at 13:39
just add a printf("%1.12f", a); to your code and you will see the rounding error – Sibster Mar 26 '13 at 13:47

score 3 · Answer 2 · answered Mar 26 '13 at 13:39

You must never compare float-values against absolute values as you did. There are usually slight rounding errors, as a float is represented according IEEE 754 and the machine is not capable of providing exact float-values.

Have a look here for your explanation and especially the rounding-rules.

score 2 · Answer 3 · answered Mar 26 '13 at 14:23

2

you are comparing float against double. Try putting an f after the number

if(a==0.8f)
    printf("HELLO")
if(b==0.25f)
    printf("WORLD")

answered Mar 26 '13 at 14:23

Lefteris E

2,806
1
24
23

score 0 · Answer 4 · answered Mar 26 '13 at 13:44

0

the given answers are right... on Dr Dobb's Andrew Koenig is writing about order relationship with floats, have a view:

its-hard-to-compare-floating-point-numbers

comparing-an-integer-with-a-floating-point1

answered Mar 26 '13 at 13:44

Exceptyon

1,584
16
23

I don't see the answer to the question at hand in these two articles. The answer is that `a` is `0.8f` and that `0.8f` is different from the double-precision constant `0.8`. – Pascal Cuoq Mar 26 '13 at 17:12

score 0 · Answer 5 · answered Mar 26 '13 at 14:33

0

Everything is stored eventually in bits.... So floating points get rounded off in case their binary equivalents are recurring...

answered Mar 26 '13 at 14:33

Darshan Shah

57
1
7

float value condition

5 Answers5