Topic title: What does IB consider to be binary "normalization"? Topic summary: Created on: 22/03/2006 07:30:49 AM
22/03/2006 07:30:49 AM
Gerry Donaldson
Posts: 67
Joined: 20/03/2003
Binary normalization looks like scientific notation such as 1.01001 x 2^3 for example.
IEEE
drops the "1." as redundant since, in binary, there will always be a
"1.", thus the above example would be stored as "01001" rather than
"1.01001"?
The quandry is this: IEEE does set the international
standard for such things, but I have not seen IB documentation using
the IEEE standard on this. In the May papers, which convention should
students assume?
-------------------------
Gerry Donaldson Computer Science Teacher Sir Winston Churchill High School Calgary, Alberta, Canada
The
Subject Guide does not specify IEEE or indeed any other system of
storing floating point binary. For me, what you are talking about here
is a specific "representation" and not "normalisation".
Normalisation
refers to setting the fractional part such that it is stored with the
most possible precision, thus the first bit after the (imaginary) point
is always a 1. In IEEE representation this bit is never stored 'cos it
will always be 1) as you say.
However, your 1.01001 x 2^3 seems
to me to be a negative number because the MSB is a 1, so I would have
said it was already normalised (since in 2's comp the 0 is really a 1).
I'm pretty sure IEEE uses 2's comp for the mantissa and excess-n
notation (not in syllabus) for the exponent - but I'm prepared to be
corrected.
If it is positive, for me, normalised it would look
like: 0.101001 x 2^4. IEEE would then represent it as 001001 plus a
bunch more zeroes and the exponent.
Light blue touch paper and retire to safe distance :-)
Richard
-------------------------
Richard Jones Between occupations :-) richard@ib-computing.com
In conventional method the normalissation is to shift the point so as
to make the number between 1 and 2 ,ie the mantissa should be in the
form 1.*********** and while the mantissa is written, only
the digits after the point is written ( as always the integer part will
be 1) eg 1111.01 = 1.11101 x 2^3
and mantissa is 11101 exponent is 3 (11)
But
it is mentioned in IBO that the normalissation should be done in
the method where the number should be made less than 1 that means 1111.01
= .111101 x 2^4
where the mantissa is 111101 and exponent is 4 (100).