Homework 3
Assigned 20 Oct, 1998
Due 5 Nov, 1998
Work out, but don't turn in
 Textbook problems 5.25.12 (as many problems as you think you need
to be certain you understand what's going on). See if you get the same
answers as your classmates.
 Textbook problems 6.36.9 (similarly).
To be turned in
 Textbook problems 5.13, 5.14
(handle only the case of shifting by one bit)
 Textbook problems 6.1, 6.2, 6.13
 Suppose you are given a large (hundreds or thousands) sorted list
of positive
floatingpoint numbers, and are asked to add them up on a computer.
Should you add them in increasing order, decreasing order,
some other order, or does it not matter?

The IBM 370 used a floatingpoint format somewhat like the modern
IEEE standard (which hadn't been invented yet), but not quite the same.
The biggest difference was that
the exponent field was treated as a power of 16, rather than as a power
of 2. Thus for example the number 80 = 5 * 16 would be normalized to
1.010000... * 2^{6} in IEEE, but to 0.101000... * 16^{2}
in IBM 370. Discuss the pros and cons of this approach
in contrast to the IEEE standard, with particular attention to precision
and range.
Last modified:
Tue Oct 20 14:41:05 EDT 1998
Stephen Bloch / sbloch@boethius.adelphi.edu