Floating point number - verify please

[denian]

if the floatingpoint number is 1111111111111111111111111111111,
the answer should be NaN instead of infinity rite?

 

[HallsofIvy]

Since computers don't know about "infinity", yes.

(Although now, I'm going to waste time pondering what an "infinity rite" could be!)

 

[denian]

sorry. it should be right, not rite. pardon me.

from your statement, i conclude that the exact answer should be infinity.
am i right?

what if, i have this floating point number : 11111111100000000000000000000000
and i am to convert this floating point number to decimal, the answer should be infinity, isn't it?

then can i summarise that if the exponential is all 1's, it will be infinity when converted to decimal?

thank you.

 

[RiBBiT]

sorry. it should be right, not rite. pardon me.

from your statement, i conclude that the exact answer should be infinity.
am i right?

what if, i have this floating point number : 11111111100000000000000000000000
and i am to convert this floating point number to decimal, the answer should be infinity, isn't it?

then can i summarise that if the exponential is all 1's, it will be infinity when converted to decimal?

thank you.

Whether you convert it to decimal or not does not really matter. In the standard floating point systems (IEEE SP and IEEE DP) there are two special cases for representing Inf and NaN. I believe (though I'm not sure) that both of these have an exponent with only 1's. Which of the cases it is should then be decided by the value of the mantissa.

 

[denian]

tq. btw, i just found this in the internet.

# Infinity is encoded when the exponent field is all 1’s and the mantissa is all 0’s.
# If the exponent contains all 1’s but the mantissa does not contain all 0’s, then this is referred to as NaN or Not a Number.