| Title: | DIGITAL UNIX (FORMERLY KNOWN AS DEC OSF/1) |
| Notice: | Welcome to the Digital UNIX Conference |
| Moderator: | SMURF::DENHAM |
| Created: | Thu Mar 16 1995 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 10068 |
| Total number of notes: | 35879 |
Hi,
a customer of mine has a C program which doesn't round numbers
correct if they are printed with the printf command.
If the program:
#include <stdio.h>
main()
{
float f;
f=0.624;
printf("0.624=%2.2f\n",f);
f=0.625;
printf("0.625=%2.2f\n",f);
f=0.626;
printf("0.626=%2.2f\n",f);
printf("\n");
f=0.724;
printf("0.724=%2.2f\n",f);
f=0.725;
printf("0.725=%2.2f\n",f);
f=0.726;
printf("0.726=%2.2f\n",f);
}
is compilied with cc -o test test.c it produces the following output:
0.624=0.62
0.625=0.62
0.626=0.63
0.724=0.72
0.725=0.73
0.726=0.73
The rounding of the numbers ending with 5 is inconsistent. Rounding
"0.625" results in the correct value "0.62" while "0.725" leads to
the wrong value of "0.73".
Are any bugs in the C compiler known which cause that problem? Can
any compiler flags be used to avoid this unexpeted results?
Thank you for your help.
- David
[Posted by WWW Notes gateway]
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 9377.1 | Floating point is inexact. | QUARRY::reeves | Jon Reeves, UNIX compiler group | Thu Apr 03 1997 15:02 | 7 |
.625, expressed as a binary fraction, is exact. .725, expressed in binary then converted back to decimal, is 0.72499999999999998 which rounds to .72. This is inherent in floating point; if you don't like it, do your floating point arithmetic in decimal (I think COBOL lets you do this). | |||||