signed integer with only signed bit set. (-0)

[caibbor]

-0 doesn't seem possible with integers (although I can get floats to be -0.0 in some cases). but what exactly is the value of an int with only the signed bit set? I tried this:

int i=1<<31;

and cout the value, and it prints -2147483648, which after popping this value into a calculator, is exactly the value I suggested, except that it's negative. it's like the computer is taking the numerical value as if it were unsigned, but yet also considering the highest-order bit a floating point indicator.

I thought this was neat. I presume this is undefined behavior, but if anyone has any insight into this, I'd be interested to hear.

I don't really have a question, I just thought this was neat.

[Athar]

This article should answer all questions:
http://en.wikipedia.org/wiki/Two%27s_complement

[helios]

-0 doesn't seem possible with integers (although I can get floats to be -0.0 in some cases).

The representation of integer values is implementation-defined. Some encodings allow negative zero.

but what exactly is the value of an int with only the signed bit set?

Integer encodings don't necessarily have a "sign bit". Two's complement, for example, is defined as such:
1. A string of n zeroes is the integer 0.
2. If x ≠ 2^(n-1) - 1, then adding 1 to the representation of x yields a representation of y such that y=x+1
3. If x = 2^(n-1), then adding 1 to the representation of x yields a representation of y such that y=-(x+1)
The sign of a value is implied by whether it's higher or lower than a certain value. There isn't a sign bit that you can flip to get the negative of a number, like with floats.

I thought this was neat. I presume this is undefined behavior

Yep. 1<<31 is a 1 followed by 31 zeroes, which happens to correspond with how your platform represents the value -2^31.
By the way, a 1 followed by 30 zeroes and then another 1 is -2^31+1 .

http://en.wikipedia.org/wiki/Two%27s_complement

[Cubbi]

Brandon Captain wrote:

I presume this is undefined behavior

That is correct. Left-shifting a signed integer so far that it overflows is undefined behavior. Doesn't matter what 2's complement arithmetic says, the compiler is allowed to drop code that depends, for example, on a signed integer changing sign after left shift, because it "can't happen".

[caibbor]

I didn't think it was an overflow, considering (unless I'm rusty with my bitwise operations) 1<<31 set's the leftmost bit in the 32-bit integer. which, as I had always thought, was the bit that denotes a negative (signed) integer.

and floats, you say they have a signed-bit, helios? I attempted to figure out how floats at the bit-level once. it was pretty complicated, really.

[Cubbi]

I didn't think it was an overflow, considering (unless I'm rusty with my bitwise operations) 1<<31 set's the leftmost bit in the 32-bit integer.

It's not an overflow if you do 1u<<31 (on a 32-bit integer platform)

[JLBorges]

> I attempted to figure out how floats at the bit-level once. it was pretty complicated, really.

If std::numeric_limits<double>::is_iec559 is true (it typically is):
http://en.wikipedia.org/wiki/IEEE_754-1985

[helios]

I'm pretty sure the signedness of the left operand is irrelevant. What causes the undefined behavior is the storing of the result into an int.

[cire]

What causes the undefined behavior is the storing of the result into an int.

1 and 31 are of type int. the result of 1<<31 is of type int. Seems counter-intuitive to me that assigning an int to an int would result in undefined behavior.

[JLBorges]

> 1<<31 is of type int

If std::numeric_limits<int>::digits == 31 and std::numeric_limits<int>::max() == 231 - 1
(32-bit integer), 1 << 31 is undefined.

The behavior is undefined if the right operand is negative, or greater than or equal to the length in bits of the promoted left operand.
...
The value of E1 << E2 is E1 left-shifted E2 bit positions...
if E1 has a signed type and non-negative value, and E1 × 2^E2 is representable in the result type, then that is the resulting value; otherwise, the behavior is undefined.

[caibbor]

wow. of all the books, articles, conversations, etc... this is the first time I've ever even heard of numeric literal suffixes other than f or F for float. in fact, I wasn't even sure what to call them when I googled it.

so shifting also considers the type of the lvalue supplied?

unsigned int x = (unsigned short int)1 << 24; // undefined behavior because of the cast?

[JLBorges]

> undefined behavior because of the cast?

No. The unsigned short is promoted before the shift operation is applied.

The operands shall be of integral or unscoped enumeration type and integral promotions are performed.