ansaurus

Question

Answer 1

+3 A:

As you note in your question, the float only takes 7 digits, as opposed to the 10 digits in INT_MAX. I believe this would be the reason to give a C4244 warning at this conversion.

You 2010-08-04 00:26:38

@You: Yes I did see that, however, based on where the decimal is, shouldn't we be able to represent larger numbers, even thought there are less digits?

Tommy 2010-08-04 00:29:42

@Tommy: You may want to read http://en.wikipedia.org/wiki/Floating_point to learn more about how floating point values are represented, in particular the way the bits are divvied up between mantissa and exponent.

JaredReisinger 2010-08-04 00:32:33

@Tommy - float can handle the range, but will lose precision (ie., you'll lose some of the least significant digits). That's a loss of data, so the warning is appropriate.

Michael Burr 2010-08-04 00:33:00

e.g., try storing both 4_000_000_000 and 4_000_000_001 in your floats, and see what value they have.

sarnold 2010-08-04 00:46:01

Ok, I get it now, thank you.

Tommy 2010-08-04 01:27:56

Answer 2

+3 A:

While float supports a wider range of values than unsigned int, it does so with less accuracy. Floats have a 23-bit mantissa which, as you posted, is only about 7 decimal digits. unsigned ints support over 9 decimal digits.

Suggestion: use double, not float.

Edit: Actually, I retract that suggestion; floating-point and integer data types are fundamentally different and are not directly convertible. What integer value do you expect to get for Single.MaxValue? For Single.NegativeInfinity? Explaining why you want to convert from float to int would be helpful.

Dour High Arch 2010-08-04 00:28:33

@Dour... I tihnk he's converting the other way; from int to float. Still, knowing *why* would certainly help.

JaredReisinger 2010-08-04 00:58:56

Answer 3

+3 A:

'unisigned int' and 'float' both use 32 bits to store values. Since a float has a larger range, it necessarily sacrifices some precision. This means that there are some unsigned int values that cannot be accurately represented in a float. MSDN provides some more details.

JaredReisinger 2010-08-04 00:28:58

Answer 4

+2 A:

From wikipedia:

Single precision, called "float" in the C language family, and "real" or "real*4" in ? Fortran. This is a binary format that occupies 32 bits (4 bytes) and its significand has a precision of 24 bits (about 7 decimal digits).

With an int having 32 bits (generally) and a float having 32 bits (with part reserved for the exponent), there are a lot of int values that can't be exactly represented by the floating point type.

sje397 2010-08-04 00:31:57

Answer 5

+1 A:

try this to understand the situation

float f = 0.0f;
while (f < (INT_MAX-1))
   f++;

and you'll see that this is really an infinite loop on systems where int is 32 bits or less.

What will be interesting is to break into the infinite loop and see what the value is for f such that f == f + 1

franji1 2010-08-04 03:12:13

The limit is exp(2, 24), because float only has 24 significant bits. `printf("%f\n", float(1<<24)+1)` prints an even number which is obviously wrong.

FredOverflow 2010-08-04 11:21:53

ansaurus

tags:

views:

answers:

Conversion from unsigned int to float

related questions