Thread: Confusion regarding the storage of floating numbers on various architectures

Casting the variables to float and putting a .0 after a literal number will force all values to be float so it should give the same results.

Code:

int main(){
    printf("%f\n",7.0/2.0);
    return 0;
}

Code:

int main(){
    int a = 7;
    int b = 2;
    printf("%f\n",(float)a/(float)b);
    return 0;
}

Code:

int main(){
    int a = 7; //What's the point of this,
    printf("%f\n",7.0/2.0); //if you don't use it?
    return 0;
}

And exactly what output do you get from that - I'd expect that you'd get some pretty rubbish results, as you are printing an integer value as a floating point result. Since you are passing only an integer, half the result of your printout will depend on the content of where the other half of the double that the printf function expects. This is not at all clearly defined, and may be one of your integer values or some completely different random data - so yes, I expect that different bits of code would produce different results, but not in any way predictably - change the code around to printf("Hello, World\n") before the printf of the float and you'd get a different results. Add some calculations, change the compiler options, compile with a different compiler, etc, etc, and each time you change anything, it will change the results in some way [probably - there may be combinations that do not change the results - it is not easy to predict].

--
Mats

Actually, "7.0" is a double constant, "7.0f" is the float constant equivalent. Important, but not really at the same time

so, 7.0/2.0 results in a double.

Also, as far as I know there's nothing stopping certain architectures from rounding instead of truncating integer division? But that's not the problem here.

Originally Posted by zacs7

Actually, "7.0" is a double constant, "7.0f" is the float constant equivalent. Important, but not really at the same time

so, 7.0/2.0 results in a double.

And any double passed a variable arguments function will be converted to double by standard, so using double or float in the above code will give nearly identical results.

Also, as far as I know there's nothing stopping certain architectures from rounding instead of truncating integer division? But that's not the problem here.

Certainly not part of the problem here, but I'm pretty darn sure that the specification for integer division in C is that it truncates - a lot of code using division to split numbers into portions would fail if that was the case, and whilst most of us will not run on obscure machines, even the code in glibc that produces numbers for printf and such does use a iteration of divide by base to get the individual digits out of a number.

--
Mats

Originally Posted by matsp

I'm pretty darn sure that the specification for integer division in C is that it truncates

Yes, it does: "When integers are divided, the result of the / operator is the algebraic quotient with any fractional part discarded." A note at the end of that sentence says that: 'This is often called "truncation toward zero"'.

wow thanks laserlight and matsp, I've always wondered if it was in the standard. If not that'd certainly be a real hassle if you ever had to work with a machine like that.

Sorry for hijacking the thread