3
Table 3-3. Extended Precision Binary Real Format
Memory Format:
Field Size (in Bits):
s = Sign
e = Biased Exponent
u = Zero, Reserved
j = Integer Part
f = Fraction
Total
Interpretation of Unused Bits:
Input
Output
Interpretation of Sign:
Positive Mantissa, s =
Negative Mantissa, s =
Normalized Numbers:
Bias of e
Range of e
j=
Range of f
j.f = Mantissa = Significand =
Relation to Representation of Real Numbers
Denormalized Numbers:
e = Format Minimum
Bias of e
j=
Range of f
j.f = Mantissa = Significand =
Relation to Representation of Real Numbers
Signed Zeros:
e = Format Minimum =
j.f = Mantissa = Significand =
Signed Infinities:
e = Format Maximum =
j=
j.f = Mantissa = Significand
NANs (Not-A-Numbers):
e =
j=
e = Format Maximum =
f=
Representation of f
xxx.., xxxx
f When Created by the FPCP
Ranges (Approximate):
Maximum Positive Normalized
Minimum Positive Normalized
Minimum Positive Denormalized
95 94 80 79 64 62 0
~ BIASEO l ZERO INTEGER PART ',
EXPONENT FRACTION i
1
15
15
1
63
95
Don't Care
All Zeros
0
1
+ 15383 ($3FFFI
0 < = e < 32757 ($7FFF)
t
Zero or Nonzero
l.f
(-1)s x 2e-16363 x j.f
0 I$O0001
+ 16363 ($3FFFI
0
Nonzero
O.f
{-1is x 2 -16383 x O.f
0 ($00001
0.0
32767 (S7FFF)
Don't Care
j.O00... 0000
Don't Care
Don't Care
32767 ($7FFF}
Nonzero
j.lxxx.., x~c~x, Non-Signaling
j.Ox~c(.., xxxx, Signaling
Non-Zero Bit Pattern
1.11111 ...111t
6 × 104831
8 x 10 -4933
9 x 104952
FREESCALE
3-12
MC68881/MC68882 USER'S MANUAL