Floating-Point Arithmetic Instructions - IBM 7090 Instruction-Reference

count of 438 will give the same result as DVH. A count greater than 438 causes part of
the quotient to be shifted into the AC, where it can be altered. A count of 60S or greater
will cause an I/A cycle, and the count field (12-17) will OR with 12-17 of the IA word.
Variable-Length Divide or Proceed
VDP +0225 (Min I, E) Figure 5.3-20
Max I, E, 12L)
The execution of this instruction is the same as VDH, except that the computer will
not stop for a divide check, but will proceed to the next instruction.
Round
RND +0760 ... 0010
(I, L)
Figure 5.3-21
This instruction examines the contents of MQ(l) and if it contains a one, the magni-
tude of the AC is increased by one. If MQ(l) contains a zero, the AC is unchanged. The
MQ is not changed in either case. AC overflow is possible. This is a primary opera-
tion 76 instruction. The contents of the AC are sent to the AD and, if MQ(l) contains
a one, a one is sent to AD(35) and the output of the AD replaces the contents of the AC.
Clear Magnitude CLM +0760 ... 0000
(I, L)
Figure 5.3-22
This instruction puts zeros in AC(Q-35). CLM is a primary operation 76 instruction.
The operation is accomplished by gating the AD to the AC, with nothing in the adders.
The AC(S) remains unchanged.
Complement Magnitude
COM +0760 ... 0006
(I, L)
Figure 5.3-22
The contents of the AC(Q-35) are complemented. Positions containing ones are
changed to zeros and positions containing zeros are changed to ones. This instruction
is executed by complementing the AC to the AD and replacing the contents of the AC
with this complement.
5.3.04 Floating-Point Arithmetic Instructions
The range of numbers anticipated during a calculation may be extremely large,
extremely small or, in some cases, unpredictable. Such situations make fixed-point
arithmetic difficult to work with for two reasons;
1.
The size of the number is limited by the size of the register
(35 binary bits or 10 decimal digits).
2. The programmer must keep track of the point in all numbers throughout
the calculation.
To meet the needs of large numbers and to automatically keep track of the point, an
alternative set of arithmetic instructions, called floating-point arithmetic instructions,
are available.
Floating-point arithmetic is merely arithmetic dealing with numbers in exponential
form. The numbers 5.6 x 10 3 or 56000 x 10- 4 have a familiar form.
The numbers
are made of three parts; a fraction (5.6 or 56000), an exponent (3 or -4), and a base (10).
Floating-point numbers in binary are similar to decimal floating-point numbers.
The major difference is the base. Numbers in the 7090 use 2 as a base, because it is
a binary computer. The other difference is one of terms. Instead of a decimal point,
we will call it a binary point.
The following chart gives a comparison of fixed-point binary numbers and floating-
point binary.
68