On Wed, Oct 29, 2008 at 3:16 AM, Alan McKinnon <[EMAIL PROTECTED]> wrote:
> On Wednesday 29 October 2008 00:55:42 Jorge Peixoto de Morais Neto wrote:
>> >> I mean to really know C,
>> >> that is, read a rigorous book such as "C: A Reference Manual" and be
>> >> able to write portable programs with well-defined behavior. Speaking
>> >> of well-defined behavior, do you know what happens when you cast a
>> >> float to an int, and the float is too big to fit into the int?
>> >
>> > Did oyu try it yourself and see?
>>
>> The point is that the behavior in this situation is "undefined". It
>> might do anything. Programming in C is different than programming in
>> Python.
>
> Most likely the compiler will try to treat the float as an int and use the
> first 4 bytes of the float, ignoring the rest.
Float is 4 bytes; the questions should be reworded s/float/double/g
But somehow, SSE version or fisttp (or whatever) doesn't set CF/OF on
overflow, the returned int is simply (1 << 31) ^ (1 << 31 - 1)
> This is insane though. I cannot think of any reason why one would ever want to
> treat the first 32 bits of a float as an int. It's not like you are casting a
> long to an int which can make sense - just discard the high bits.
> I reckon the standard would say this is undefined. Most compiler would bomb
> out with a compile error but give you an obscure flag to proceed anyway. If
> you want to commit suicide, C is quite happy to pass you the pills as long s
> you ask nicely
gcc does this without errors
Code [x86/64 only]:
#include <stdio.h>
#include <math.h>
#include <stdint.h>
int main() {
double f = 1.5e99;
int i = (int)f;
uint16_t z;
asm("pushf; popw %0;" : "=r"(z));
printf("%d %d %d\n", i, (z & (1 << 11)), (z & 1) );
return 0;
}
In short, your mileage shall vary, so sayeth the standard.
> --
--
Andrey Vul
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
