CODEWITHSUNDEEP
cvisual-studio-2012rounding
cxxl
cxxl

Reputation: 5379

Rounding behavior of VS 2012 printf()

The printf() function in Visual Studio 2012 (Compiler Version 17.00.61030 for x86) uses rounding away from zero when I supress decimal digits. I.e., printf("%.0f%, 0.5) prints 1. But when it comes to actually printing decimal digits it seems that it uses some IEEE-like rounding, but just to the nearest uneven. I.e., both printf("%.1f%, 0.05) and printf("%.1f%, 0.15) print 1. I find this strange and inconsistent.

Here is some code:

#include <stdio.h>
int main(int, char**) {
   printf("0.5 -> %.0f\n", 0.5);
   printf("1.5 -> %.0f\n", 1.5);
   printf("-0.5 -> %.0f\n", -0.5);
   printf("-1.5 -> %.0f\n", -1.5);
   printf("0.05 -> %.1f\n", 0.05);
   printf("0.15 -> %.1f\n", 0.15);
   printf("0.25 -> %.1f\n", 0.25);
   printf("-0.05 -> %.1f\n", -0.05);
   printf("-0.15 -> %.1f\n", -0.15);
   printf("-0.25 -> %.1f\n", -0.25);
   return 0;
}

And the matching output:

 0.5 -> 1
 1.5 -> 2
 -0.5 -> -1
 -1.5 -> -2
 0.05 -> 0.1  
 0.15 -> 0.1  
 0.25 -> 0.3  
 -0.05 -> -0.1
 -0.15 -> -0.1
 -0.25 -> -0.3

Am I misunderstanding something? I would assume that rounding to whatever many decimal places always works the same way.

Upvotes: 0

Views: 77

Answers (1)

Bathsheba
Bathsheba

Reputation: 234715

It's down to the fact that only very few numbers can be expressed precisely in floating point.

0.5 can be as it's a dyadic rational. Ditto 0.25.

0.15 can't be and the closest available double is lower than that. Hence the rounding down that you observe.

Upvotes: 3

Related Questions