What are the parameters I should take care while designing floating point adder?

[alam.tauqueer]

Can anyone have some idea what are the parameter I should take care while designing floating point adder.
Also explain me
what is guard bit, sticky bit and how we calculate NaN condition.

Tauqueer

 

[lordsathish]

Re: Floating point adder

this page might help you about guard round and sticky bits....
**broken link removed**

 

[FrankCh]

Floating point adder

Basically you need to first align both number to the same exponent, then add the mantissa. After that you need to adjust the exponent again and normalize mantissa to 1~10.0..

Guard bit is used to protect again overflow. I doubt whether you get a NaN in addition!

 

[alam.tauqueer]

Floating point adder

Thanks alot.Please explain me little on guard bit and sticky bit.what


Tauqueer

 

[angu]

Re: Floating point adder

During the prenormalization shifting phase (also known as the alignment step) three extra bits - Guard, Round, and Sticky
are generated. The Sticky bit being the value of logical OR of all bits that would be dropped by right shift. These three additional bits, attached at the rightmost of the significand, are required in order to avoid loss of precision.
Guard and Round bits are required when performing rounding-to- nearest even, the default rounding mode of IEEE-754. Sticky bit participates in subtraction and generates the necessary borrow.