Again mystry outputs

[adarshakb]

:eek: I cant understand this can someone enligten me how floating point numbers are stored?

void main()
{
float a=0.7;
if(a<0.7)
printf("C");
else
printf("C++");
}

Output:
C

BUTTTT

void main()
{
if(0.7<0.7)
printf("C");
else
printf("C++");
}

Output:
C++


:confused::confused::confused::confused:

I am using Visual studio 2008 if that helps

[BrickInTheWall]

Look at this article:

http://steve.hollasch.net/cgindex/coding/ieeefloat.html

Thing is, even if you set a to be 0.7, it's not EXACTLY 0.7 since the accuracy of floating point numbers is limited. You should look for some tutorials on floating point arithmetic, as it is very important.

[adarshakb]

Look at this article:

http://steve.hollasch.net/cgindex/coding/ieeefloat.html

Thing is, even if you set a to be 0.7, it's not EXACTLY 0.7 since the accuracy of floating point numbers is limited. You should look for some tutorials on floating point arithmetic, as it is very important.

Ya i knew all those! and i worked it out.... here it is

sign=0

0.7 in binary = 0.101100110011001100....(1100 is repeating here)

this mantisa part= 0.10110011001100=in normalized form 1.0110011001100*2^-1

Exponent part=-1+127=126= 01111110

Mantissa part =

01100110011001100110011 ----->(23bits)

therefore out 0.7 is

00111111001100110011001100110011 -----in float


IN DOUBLE now

sign=0---- 1 bit
Exponent= 00001111110 ---- 11bits
Mantissa=011001100110011001100110011001100110011001100110011001100110

Hence 0.7 in Double is

000001111110011001100110011001100110011001100110011001100110011001100110



Here is the Problem now:p:D

a<0.7

What Happens is THIS i believe

0.7 is treated as DOUBLE

a is float so its converted to double...
Extra bits in sign as covered with 0s so that value remains same. This is done by prefixing Exponent with 0s and postfixing Mantissa with 0s


Sign -MATCHS 0=0

Exponent - MATCHES 01111110=00001111110

Mantissa -- is as follows

a's : Converted to
0011111100110011001100110011001100000000000......(fill it up with 0)

but 0.7's is 011001100110011001100110011001100110011001100110011001100110


(I might have missed a few 0s lol but no patience to check)

so this is what its adjudec "true" when a<0.7 is calculated


:thumbsup::thumbsup::thumbsup: