IEEE 754 Standard

Question

In IEEE $754$ single floating point format, how many numbers can be represented in the interval [10, 16)?

      A. $2^{21}$

      B. $3 * 2^{21}$

      C. $5 * 2^{21}$

      D. $2^{22}$

Comments

10 is written as

0 10000010 01<21 0's>

These rightmost 21 0's are actually redundant, and we can set them to anything.

 

16 is written as

0 10000011 <23 0's>

If we keep the exponent to 3 (to make add the manipulatable bits shift to one end)

0 10000010 1<22 0's>

These rightmost 22 0's are manipulatable.

 

So, $2^{21}+2^{22}=2^{21}(1+2)=3*2^{21}$

Answer 1

First write 10 and 16 in single precision floating point representation

10=0  10000010  01000000000000000000000

16=0  10000011  00000000000000000000000

Now lets traverse through all the numbers in between. Matissa of 10:  01000000000000000000000 to 01111111111111111111111 has  $2^{21}$ numbers. Then

100000000000000000000000 to 11111111111111111111111 has $2^{22}$ numbers. The next number after this will be nothing but 16. So total $2^{21}$+$2^{22}$=3*$2^{21}$ numbers are there.

Answer 2

In the IEEE 754 notation , the number is represented in 32 bits . The first bit is for the sign of the number , next 23 bits is normalise mantissa and next 8 bits is biased exponent.

$10$ is represented as = $1.01*2^{3} = 1.01*2^{130-127}$ , which in 32 bit format would be represented as ,

$001(19 \ 0's)10000010$.

And 16 is represented as $1.0*2^4 = 1.0*2^{131-127}=0(23 \ 0's)10000011$.

Thus we need to find the number of possible patterns of $0's$ and $1's$ in the following format.

Keeping the exponent same , the number of patterns of mantissa of the form $01(21 \ x's)$ and $1(22 \ x's)$.Where $x's$ are the positions to fill with 0's and 1's .

Thus total number of possible patterns = $2^{21}(of \ the \ 1^{st} \ kind)+2^{22}(of \ the \ 2^{nd} \ kind)$ = $3*2^{21}(Ans)$.

Comments

Why you are taking 1(22 x's) ? It should not be 2^23 as we have complete 23 zeros to select from?

---

1.00000...000 x 2^3 is  how much?

Answer 3

NOTE: the numbers that have to be represented here have to be in the de-normalized form i.e. there is just 1 significant bit  and its 0. The representation should be of the form “0. something something…..” Thus option C is correct