CS 2734 Computer Organization II

Read pages 565-573 (Appendix B) of Tanenbaum and pages 261-286 (Chapter 11) of Paul

• Scientific notation: $f\; x\; 10e$
  • f is the mantissa
  • e is the exponent
  • 10 is the base

• Floating point numbers as a model of the real numbers. (See Figure B-1 for an example.)

• Overflow and underflow

• Normalized numbers - adjust the exponent so that the fraction always starts with a one (e.g. .10111)

• IEEE floating point standards:
  • single (32 bits), double (64 bits) and extended (80 bits).
  • Instead of a normalized fraction, IEEE uses normalized significand in the range [1, 2).
  • Exponent uses excess notation.
  Single precision IEEE: $N\; =\; (-1)sx\; 2e\; -\; 127x\; 1.f$

• Excess representation for signed integers.

Revision Date: 4/12/97