Floating Point Addition - hardware/software

[coolul007]

Can someone explain to me how floating point addition is implemented on a x86 in hardware or software. I would like to find out what method is used to add varying number size. if I have a 1 X 10^-100 + 2 X 10^50. are the exponents average for a common ground or does the large one rule etc. or is it something else?

 

[rcgldr]

The fractional (mantissa) part of the number with the lower exponent is shifted to the right before addition is performed. This results in truncation error.

 

[coolul007]

The fractional (mantissa) part of the number with the lower exponent is shifted to the right before addition is performed. This results in truncation error.

Thanx, for the answer, this came up in a question in the math section "is it better to add an array of numbers from smaller to larger or larger to smaller". Everyone answered small to large, which got to me to thinking about what really goes on...Therefore the larger exponent rules.