how to check each bit of a float

[geliangwei]

The 32-bit float number is defined by IEEE 754.
Given a float variable, f, how to check whether each of the 32 bits is '1' or '0'?

One possible solution is access f as a 32-bit int.
That is define an integer i that accesses the same memory address where f is stored.
Then, I can right shift i by one bit for 32 times and check whether the least significant bit is '1' or '0'.
But how to let i and f share the same memory address?

Thank you!

[drhowarddrfine]

Just i=&f. You can have any number of pointers point to a variable.

[AdrianH]

Just i=&f. You can have any number of pointers point to a variable.

Well, you do need to cast it.

unsigned int i=reinterpret_cast<unsigned int&>(f);

or if using C:
unsigned int* pI = (unsigned int*)&f;

and then dereference pI using the unary * operator. I.e.

*pI>>shiftValue

Hope this helps.


Adrian

[geliangwei]

Thank you very much for your replies!

I found another solution from Internet just now by using 'union'

union four_byte
{
int i;
float f;
} share;

share.f = 2.34;

Then, I can access share.i as an integer. This seems easier.
Is that ok?

[AdrianH]

Thank you very much for your replies!

I found another solution from Internet just now by using 'union'

union four_byte
{
int i;
float f;
} share;

share.f = 2.34;

Then, I can access share.i as an integer. This seems easier.
Is that ok?

Yeah, that is fine since int and float are both usually 4 bytes in size.

Good going finding another solution. I had actually forgoten about that one.


Adrian

[DeMan]

Almost always more than one way to skin a cat.....

[drhowarddrfine]

**sigh**

In assembly it's:

mov esi,[address of float]

Life was so much easier then. :)

[AdrianH]

**sigh**

In assembly it's:

mov esi,[address of float]

Life was so much easier then. :)

That would be 8x86 assembly. :p