how a number say (a.bc * 10 exp def ) is represented in a fpu ? say for example, 9.75 * 10 exp 263 .How this number is normalised?

Could you elaborate (c l a r i f y) a little, what you want ?

it's hard to normalize numbers manually. but I'm sure, there are some algos or formulas to do this. I know one method of normalization:
for example, we have decimal number 2.5, and we want to represent it as double-precision floting-point (DPFP) value...
we know the format of DPFP value: 11 bits for biased exponent, 52 bits for significand. biasing constant = 1023.
to normalize the number we should solve an equation: 1.x * y = 2.5; here will take y = 2^1. then x = 2.5/2 = 1.25
now we have x = 0.25 decimal or 01b.
now let's form the value: exponent = 1, then biased exponent = 1024 (or 10000000000b), and significand is 4000000000000h

2.5 decimal = 01000000000001000000000... 40 more zeros...0b

If we take small numbers it is fine.But i get confuse when considering large numbers.DPFP range is upto 1.79 * 10 exp 308.Lets declare a variable double d in some lang with value 12.83 * 10 exp 300.Now how the FPU going to normalise this.It must normalise as the number is within the range.