asked on

floating point comparison

Haya guys,
little problem here, i'm trying to compare two floats:

float x, y;

Then let's pretend they receive their values from a bizarre function...

x = bizarreFct();
y = bizarreFct();

if( x > y )
{
doACapper();
}

What happens is x and y are both let's says equal to 5.000
But the fct doACapper (purely fictive func by the way -BG-) is still called, the debugger shows that both variable are == 5.

Anyone having a trick to cut the presision of floating point numbers to 3 or 4 decimal...

because x and y are supposed to be == 5 , but our friend the floating point makes them equal to:

x = 5.0000517
y = 5.0000345

Thanks a lot folks!

David MacDonald

ASKER

Btw, i don't want to work with integers, so no need to tell me to cast them, I really have to work with float (damb 3D programming!)

akshayxx

do like this...
#define PRECISION 0.0001
int compare;
compare=(int)((double)(x-y)/(double)PRECISION);

now
negative compare means x<y
positive compare means x>y
0 means both are equal

ASKER CERTIFIED SOLUTION

akshayxx

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

ssnmoorthy

what akshayxx has written is correct.

Suppose you define

float x=5;
float y=5;

It will be mostly taken as
x=4.999999 in one case and
y=5.000001 in another case. So when you compare like

if(x==y)
{ //this will not be executed
}

That is (x==y) is evaluated to false or 0.

So it is always good to use

if((fabs(x-y))>0.00001)
{ //do this
}

or you can define one like

#define eps 0.00001

and use

if((fabs(x-y))>eps)
{ //do this
}

So you can replace this when ever you compare any float values.

shlezman

you can try this :

if ( (int)(1000*x) > (int)(1000*y) )
{
foo();
}

then you get comparison with the desired precision

akshayxx

had he agreed on using int .. there can be so many solutions..

Kryp

I would suggest you read this if you intend to make extensive use of floating point
http://www.nondot.org/sabre/os/files/Processors/WECSSKAFloatingPoint.pdf

David MacDonald

ASKER

Thanks a lot akshayxx and all of you, thanks Kryp for the docs on floating points.