IBM 1620 1 Manual page 28

1620 Automatic Floating-Point Operations
In 1620 Automatic Floating-Point Operations, a
floating-point number is a field consisting of a var-
iable length mantissa and a 2-digit exponent. The
exponent is in the two low-order positions of the field,
and the mantissa is in the remaining high-order posi-
tions, as shown:
M,
MEE
The mantissa must have a minimum of two digits
and
ca~
have a maximum of 100 digits. However,
when two fields are operands (quantities being added,
subtracted, multiplied, divided), they must have man-
tissas of the same length. The extremity of the field
is marked by a flag over the high-order digit.
The exponent is established on the premise that
the mantissa is less than 1.0 and equal to or greater
than 0.1. The exponent is always two digits and
has a range of -99 to +99. The length of the ex-
ponent field is defined by a flag over the high-order
( tens) digit.
The mantissa and the exponent each have an al-
gebraic sign represented by the presence (negative)
or absence (positive) of a flag over the units position.
A floating-point number with a negative mantissa
and a negative exponent is represented as follows:
M .... MEE
Sign control of the results of all computations is
maintained according to the standard rules of arith-
metic operations.
Eight floating-point instructions are provided: four
are for arithmetic computations - Floading Add,
Floating Subtract, Floating Multiply, and Floating
Divide; three are used to control field size and loca-
tion - Floating Shift Right, Floating Shift Left, and
Transmit Floating. The eighth instruction provides
for Branch and Transmit in floating-point operations.
All instructions are in the 1620 format of a 2-digit Op
code, 5-digit P address, and 5-digit
Q
address. Float-
ing-point instructions depend on the presence of flags
over the high-order digits of the mantissa and expo-
nent. Therefore they should be used only with data
in the floating-point format.
As an aid to the programmer or operator in check-
ing program logic and computation results, the oper-
ation of the computer in aligning decimal points,
normalizing results, etc., is described with each in-
struction. These operations are automatic and need
not be programmed. Of particular note is Floating
Divide, which requires only one instruction; the divi-
dend is positioned, division is accomplished, and
the quotient is transmitted to the P field without
further command.
24
In deSCriptions of instructions and operations, the
following symbols are used for clarity and brevity:
M p mantissa of the field at the P address (P )
Mq mantissa of the field at the Q address (Q)
Ep exponent of the field at the P address
Eq exponent of the field at the
Q
address
L number of digits in the mantissa
d Ep - Eq
In all floating-point numbers, the decimal point
is assumed to be at the left of the high-order digit,
which must not be zero. Such a number is referred
to as "normalized." When anum ber has one or more
high-order zeros, it is considered to be "unnor-
malized." An unnormalized number resulting from
a floating-point computation is normalized auto-
matically, but unnormalized terms are not recognized
as such when entered as data. They will be processed
but correct results cannot be assured. Therefore, it
is necessary that all data be entered in normalized
form. For example, the number 0082349405 should
be entered as 6823494004, assuming the fixed-point
number is 6823.494, and an 8-digit mantissa is re-
quired.
With the exception of Floating Shift Right and
Floating Shift Left, the P address and
Q
address of
floating point fields are the addresses of the low-order
positions of the exponents.
Floating Add CFADD-Ol)
Description. Mq is added to Mp and the result
replaces Mp. Mp and Eq are not altered in core
storage. Dependent on L and the value of d, the
appropriate field is shifted to align decimal points
before addition is performed. If d
=
0, no shift is
made (Figure 19).
If d is greater than zero and less than L, in effect,
Mq is shifted d positions to the right before being
added to Mp. The number of low-order digits of Mq
equal to d are truncated as the shift is made (Figure
20). If d is less than zero, and the absolute value
Core Storage Locations
Core Storage Locations
01590_01599
Instruction
01590--01599
Before
After
.vp
I
Ep
I
Mq
T
Eq OP
I
P
I
Q
Mp
I
Ep
I
Mq
I
Eq
123\"04\789\04
o
1\0 1 5 9 4\0 1 5 9 9
9
1 2\0 4\7 8 9\0 4
Figure 19. Addition Without Mp or Mq Shift