Control Data 3100 Reference Manual page 177

As an example, compare the unbiased exponents of +528 and +0.028 (Example 1).
EXAMPLE 1
o
Coefficient
Sign
o
Coefficient
Sign
Number
=
+52
o
0 000 000 11 0
Exponent
Number
=
+0.02
111
111 011
Exponent
(36 bits)
Coefficient
(36 bits)
Coefficient
In this case +0.02 appears to be larger than +52 because of the larger exponent. If,
however, both exponents are biased (Example 2), changing the sign of both exponents
makes +52 greater than +0.02.
EXAMPLE 2
o
Coefficient
Sign
o
Coefficient
Sign
o
Number
=
+528
o
000 000 110
Exponent
Number
=
+0.028
111
111
011
Exponent
(36 bits)
Coefficient
(36 bits)
Coefficient
When bias is used with the exponent, floating point operation is more versatile since
floating point operands can be compared with each other in the normal fixed point mode.
All floating point operations involve the A, Q, and E registers, plus two consecutive
storage locations M and M
+
1. The A and Q registers are treated as one 4S-bit register.
Indirect addressing and address modification are applicable to this whole group of in-
structions.
Operand Formats
The AQ register and the storage address contents have identical formats.
In both cases the maximum possible shift is 64
(778)
bit positions. Since the coefficient
consists of only 36 bits at the start, any shift greater than 36 positions will, of course,
always result in an answer equal to the larger of the two original operands.
B-10