IBM 7090 Instruction-Reference page 79

When signs are unlike, a complement addition occurs. A column 9 carry resulting
from this addition means the AC fraction was smaller than the SR fraction and the re-
sult placed in the AC is a true number. Assuming the MQ to be zero, this true number
is 1 less than the sum, because the 1 's complement of the AC fraction was used for addi-
tion. Therefore, a one is added to the AC fraction.
If
the MQ contains something other
than zero, the result of the addition is 1 minus MQ fraction less than it should be. There-
fore, one is not added to the AC fraction, but the MQ is subtracted from this one. This
is done by, in effect, complementing the MQ. Because no provisions are in the circuits
to complement the MQ, the MQ contents are moved to the AC through the SR. The re-
sult of the addition is moved to the SR when the MQ contents are brought into the AC.
Fourth Step L Time. Fourth step L time is used to:
1. Complement the AC fraction.
a. AC contains original MQ fraction if the MQ was not zero.
b. If original MQ fraction was zero, it was not moved to the AC; therefore,
this complementing is used to test the AC fraction for a zero result of the
addition.
2. If the MQ and AC are zero, clear the AC to give a zero characteristic.
3. End operation if normalizing is not needed.
a, If AC and MQ are zero.
b. If AC (9) contains a one.
After complementing the AC fraction with the FP trigger on (MQ was not zero), the
complement is sent to the MQ and the result of the addition is moved from the SR back
to the AC. Again available circuits must be used to get the complement to the MQ.
Therefore, the AC is gated to the SR and the SR is gated to the MQ.
Fifth Step L Time. Fifth step L time is used for normalizing the result and ending
operation. Normalizing is done by (1) shifting the AC and MQ fractions left until AC(9)
receives a one, (2) counting the number of shifts, and (3) subtracting the number of
shifts from the characteristic. The SC is used to count the number of shifts. Since the
SC is a count down counter which starts from zero, the 2's complement of the number
of shifts will be in the SC when shifting is stopped. Adding this complement to the
characteristic of the AC gives the correct adjusted characteristic.
I Time. I time of the next instruction sets the MQ(S) and characteristic. The AC(S)
is gated to the MQ(S) so both signs are the same.
For an MQ and AC fraction that is not zero, the characteristic of the MQ is set 2710
(338) less than the AC characteristic. This is done by adding the 2's complement of 338
to the AC characteristic and putting the result into the MQ characteristic.
Unnormalized Floating Add
UF A -0300 (Min I, E, 3L)
(Max I, E, 8L)
This instruction algebraically adds two floating-point numbers in the same manner
as FAD. The result, however, is not normalized. The sequence of operation for UF A
is the same as FAD with these exceptions:
1. For like SR and AC signs, UF A ends operation at the end of the third step
because a normalizing step is not needed.
2. UF A ends operation at the end of the third step if AC and SR signs are unlike
and there is no column 9 carry as a result of the addition. This is because
the correct unnormalized result is in the AC and MQ at the end of the 3rd step.
78