Bug 141848

Summary: Weird rounding of floats
Product: [Fedora] Fedora Reporter: Bastien Nocera <bnocera>
Component: gccAssignee: Jakub Jelinek <jakub>
Status: CLOSED NOTABUG QA Contact:
Severity: medium Docs Contact:
Priority: medium    
Version: 3   
Target Milestone: ---   
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2004-12-06 07:04:05 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Bastien Nocera 2004-12-04 00:19:48 UTC 
gcc-3.4.2-6.fc3
Also happens with gcc 3.2 and 4.0-pre (as shipped with FC3)

$ cat test.c

void
test (float zoom)
{
        printf ("%d %f\n", (int) (zoom * 100.0), (zoom * 100));
}

int main (int argc, char **argv)
{
        test (1.01);
        return 0;
}

$ gcc -o test test.c
$ ./test
100 100.999999

It's supposed to print:
100 101.000000
Comment 1 Jakub Jelinek 2004-12-06 07:04:05 UTC 
There is nothing wrong on this, that's how floating point works.
printf does no rounding.  1.01f is rounded when stored in float and the result
is inexact too.
E.g. 1.01 rounded to fit into float and double is:
float 0x1.028f5cp+0
double 0x1.028f5c28f5c29p+0
and the exact number is actually 0x1.028f5c with period 0x28f5c.
As you can see, when rounded to float it is slightly smaller and double is
slightly bigger than the exact value.  Therefore it is no surpise if this
0x0.0000028f5c28f5c28f5c28f5c28f5c28f5c28f5c28f5c... difference multiplied
by 100 already shows up among the important bits.