Floats, Cont'd

• Currently, most manufacturers adopt IEEE standard:

  • s = 1st bit
  • Next 8 bits are e
  • Last 23 bits are M, expressed as a binary fraction, either 1.F, or, if e=0, 0.F, where F is in base 2.

• Largest single-precision float fmax = 2127  1038.

• Smallest (and least precise!) fmin = 2-149  10-45.

Notes: