Unit 6

Fixed point Computer Arithmetic

Arithmetic instructions manipulate data to produce solution for computational problems. The 4 basic
arithmetic operations are addition, subtraction, multiplication and division. From these 4, it is possible
to formulate other scientific problems by means of numerical analysis methods. Here, we’ll discuss
these 4 operations only on fixed-point binary data (there are other types too, viz. floating point binary
data, binary-coded decimal data) and hence the unit named.

Addition and Subtraction

There are 3 ways of representing negative fixe-point binary numbers: signed magnitude, signed 1’s
complement or signed 2’s complement. Singed 2’s complemented form used most but occasionally we
deal with signed magnitude representation.

Addition and Subtraction with signed-magnitude data

Everyday arithmetic calculations with paper and pencil for signed binary numbers are straight forward
and are helpful on deriving hardware algorithm. When two signed numbers A and B are added are
added are subtracted, we find 8 different conditions to consider as described in following table:

Note: Brackets () for subtraction

Addition (subtraction) algorithm:

when the signs of A and B are
identical (different), add
magnitudes and attach the sign of
A to result. When the signs of A
and b are different (identical),
compare the magnitudes and
Table: addition and subtraction of signed-magnitude numbers subtract the smaller form larger.

Hardware Implementation
To implement the two arithmetic operations with hardware, we have to store numbers into two register
A and B. let As and Bs be two flip-flops that holds corresponding signs. The result is transferred to A and
As. A and As together form a accumulator.

We need:
 Consists of two resisters A and B
and sign flip-flops As and Bs.
 A magnitude comparator: to
check if A>B, A<B or A=B.
 A parallel adder: to perform A+B
 Two parallel subtractors: for A-B
and B-A
 The sign relationships are
determined from an exclusive-
OR gate with As and Bs as inputs.
Fig: hardware for signed-magnitude addition and subtraction

Page 1
Block Diagram Description: hardware above consists of registers A and B and sign flip-flops As and Bs.
subtraction is done by adding A to the 2’s complement of B. Output carry is transferred to flip-flop E,
where it can be checked to determine the relative magnitude of two numbers. Add-overflow flip-flop
AVF holds overflow bit when A and B are added. Addition of A and B is done through the parallel adder.
The S output of adder is applied to A again. The complementer provides an output of B or B’ depending
on mode input M. Recalling unit 2, when M = 0, the output of B is transferred to the adder, the input
carry is 0 and thus output of adder is A+B. when M=1, 1’s complement of B is applied to the adder, input
carry is 1 and output is S = A+B’+1 (i.e. A-B).

Hardware Algorithm
The flowchart for the H/W algorithm is given below:

As and Bs are compared by an exclusive-OR

gate. If output = 0, signs are identical, if 1 signs are
different.

For add operation identical signs dictate

addition of magnitudes and for subtraction,
different magnitudes dictate magnitudes be
added. Magnitudes are added with a
microoperation EAA+B (EA is a resister that
combines A and E). if E = 1, overflow occurs and is
transferred to AVF.

Two magnitudes are subtracted if signs are

different for add operation and identical for
subtract operation. Magnitudes are subtracted
with a microoperation EAA+B’+1. No overflow
occurs if the numbers are subtracted so AVF is
cleared to 0. E=1 indicates A>=B and number (this
number is checked again for 0 to make positive 0
[As=0]) in A is correct result. E=0 indicates A<B, so
we take 2’s complement of A.

Fig: flowchart for add and subtract operations

Addition and Subtraction with signed 2’s complement data

Guys, refer unit 1 once, addition and subtraction with signed 2’s complement data are introduced there.
Anyway, in signed 2’s complement representation, the leftmost bit represents sign (0-positive and 1-
negative). If sign bit is 1, entire number is represented in 2’s complement form (+33=00100001 and -
33=2’s complement of 00100001 =11011111).
Addition: sign bits treated as other bits of the number. Carry out of the sign bit is discarded.
Subtraction: consists of first taking 2’s complement of the subtrahend and then adding it to minuend.
When two numbers of n-digits each are added and the sum occupies n+1 bits, overflow occurs which is
detected by applying last two carries out of the addition to XOR gate. The overflow occurs when output
of the gate is 1.

Page 2
 Register configuration is same as signed-
magnitude representation except sign bits are not
separated. The leftmost bits in AC and BR represent
sign bits.

Significant difference: sign bits are added are

subtracted together with the other bits in
complementer and parallel adder. The overflow flip-
flop V is set to 1 if there is an overflow. Output carry
in this case is discarded.

Fig: hardware for signed-2’s complement

addition and subtraction

Example: 33 + (-35)
AC = 33 = 00100001
BR = -35 = 2’s complement of 35 = 11011101
AC + BR = 11111110 = -2 which is the result

Comparing this algorithm with its signed-

magnitude counterpart, it is much easier to
add and subtract numbers. For this reason
most computers adopt this representation
over the more familiar signed-magnitude.

Fig: algorithm for addition & subtraction of

numbers in signed-2’s complement representation

Multiplication

Signed-magnitude representation
For this representation, multiplication is done by a process of successive shift and adds operations. As an
example:

Process consists of looking successive bits of the

multiplier, least significant bits first. If the multiplier bit
is 1, the multiplicand is copied down; otherwise, zeros
are copied down. Numbers copied down in successive
lines are shifted one position Shifted left one position.
Finally, numbers are added to form a product.

Page 3
Hardware implementation for signed-magnitude data
It needs same hardware as that of addition and subtraction of signed-magnitude. In addition it needs
two more registers Q and SC.

 Successively accumulate partial

products and shift it right.
 Q  multiplier and Qs  sign.
 SC  no. of bits in multiplier
(magnitude only).
 SC is decremented after forming
each partial product. When SC is 0,
process halts and final product is
formed.
 B  multiplicand, Bs  sign
 Sum of A and B forms a partial
product

Fig: Hardware for multiply operation

Hardware Algorithm
Flowchart below shows a hardware multiply algorithm.

Example: B = 10111 (Multiplicand)

Q = 10011 (Multiplier)
Operation E A Q SC
Initial conf. 0 00000 10011 101
Iteration 1 (Qn = 1) 00000
EA A+B + 10111
PP1--> 0 10111
shr EAQ, SCSC-1 0 01011 11001 100
Iteration 2 (Qn = 1) 01011
EA A+B + 10111
PP2--> 1 00010 11001
shr EAQ, SCSC-1 0 10001 01100 011
Iteration 3 (Qn = 0)
shr EAQ, SCSC-1 0 01000 10110 010
Iteration 4 (Qn = 0)
shr EAQ, SCSC-1 0 00100 01011 001
Iteration 5 (Qn = 1) 00100
EA A+B + 10111
PP3--> 0 11011 01011
shr EAQ, SCSC-1 0 01101 10101 000
Final Product in AQ 0110110101

Page 4
Signed 2’s complement representation

Booth multiplication Algorithm

Booth algorithm gives a procedure for multiplying binary integers in signed 2’s complement notation.

Inspiration: String of 1’s in the multiplier from bit weight 2k to weight 2m can be treated as 2k+1-2m. As an
example, binary number 001110 (+14) has string of 1’s from 23 to 21 (k=3, m=1). So, this number can be
represented as 2k+1 - 2m = 24 - 21 = 16 – 2 = 14 (case is similar for -14 (110010) = -24+22-21). Thus, M * 14 =
M * 24 – M * 21; product can be obtained by shifting multiplicand M four times left and subtracted M
shifted left once.
As in other multiplication schemes, Booth algorithm also requires examination of multiplier bits and
shifting of the partial product. Prior to shifting multiplicand may be:
Subtracted <-- upon the encountering first least significant 1 in the string of 1’s in the multiplier.
Added <-- upon encountering first 0 (left of it must be 1) in string of 0’s in the multiplier.
Unchanged <-- when multiplier bit (Qn) is identical to previous multiplier bit (Qn+1)

Hardware for Booth algorithm

 Here, sign bits are not separated.
 Registers A, B and Q are renamed to AC, BR and QR.
 Extra flip-flop Qn+1 appended to QR is needed to store
almost lost right shifted bit of the multiplier (which
along with current Qn gives information about bit
sequencing of multiplier, in fact no. of 1’s gathered
together).
 Pair QnQn+1 inspect double bits of the multiplier.

Hardware Booth algorithm

Numerical Example: Booth algorithm

BR = 10111 (Multiplicand)
QR = 10011 (Multiplier)

Page 5
Array Multiplier
Checking the bits of the multiplier one at a time and forming partial products is a sequential operation
requiring sequence of add and shift microoperations. The multiplication of two binary numbers can be
done with one microoperation by using combinational circuit that forms product bits all at once. This is a
fast way of multiplying two numbers since all it takes is the time to propagate through the gates that
form the multiplication array.
Consider multiplication of two 2-bit numbers: Multiplicand = b1b0, Multiplier = a1a0, Product = c3c2c1c0

• Since multiplication of
two bits is identical to
AND operation and hence
can be implemented with
AND gate.
• In the diagram, partial
products and formed and
added by means of HA
(half adders).

Fig: 2-bit by 2-bit array multiplier

A combinational circuit binary multiplier with more bits can be constructed in similar fashion. For j
multiplier bits and k multiplicand bits, we need j*k AND gates and (j-1) k-bit adders to produce a product
of j+k bits.

Fig: 4-bit by 3-bit array multiplier

Page 6
Division Algorithms
Division of fixed-point binary numbers in signed-magnitude representation is done with successive
compare, shift and subtract operations.

Example:

• Easier than decimal since

quotient digits are 0 or 1.
• B  divisor, A  dividend,
Q Quotient
• Process consists of
comparing a partial
remainder with a divisor.

Hardware Implementation for Signed-Magnitude Data

While implementing division in digital system, we adopt slightly different approach. Instead of shifting
divisor right, the partial remainder (or dividend) is shifted left. Hardware is similar to multiplication
algorithm (not booth). Register EAQ is now shifted left with 0 inserted into Qn (Obviously, previous value
of E is lost). (I am not redrawing the diagram guys, it’s all same as multiplication but EAQ is shifted left so
change the direction of arrows at bottom).

Divide Overflow
 Division operation may result in a quotient with an overflow when working with finite size registers.
 Storing divisor in n-bit resister and dividend in 2 n-bit registers, then if quotient occupies n+1 bits,
we say divide-overflow has occurred (since n+1 bit quotient can not be stored in standard n-bit Q-
register and/or memory word).
 Talking about special case: size (dividend) = 2 * size (divisor). Divide-overflow condition will occur if
high-order half bits of the dividend >= divisor. This condition is detected by DVF (Divide-overflow
Flip-flop).

Handling of overflow: its programmer’s responsibility to detect DVF and take corrective measure. The
best way is to use floating point data.

Hardware algorithm (Restoring algorithm)

Flowchart for hardware algorithm is shown below:

Page 7
  B: Divisor, AQ: Dividend
 If A>=B (oh yes, magnitudes are compared
subtracting one from another and testing E flip-
flop), DVF is set and operation is terminated
prematurely. If A<B, no overflow and dividend
is restored by adding B to A (since B was
subtracted previously to compare magnitudes).
 Division starts by left shifting AQ (dividend)
with high order bit shifted to E. Then E=1, EA>B
so B is subtracted from EA and Qn is set to 1. If
E=0, result of subtraction is stored in EA, again
E is tested. E=1 signifies A>=B, thus Qn is set to
1 and E=0 denotes A<B, so original number is
restored by adding B to A and we leave 0 in Qn.
 Process is repeated again with register A
holding partial remainder. After n-1 times Q
contains magnitude of Quotient and A contains
remainder. Quotient sign in Qs and remainder
sign in As.

Fig: flow chart for divide operation

Numerical Example: Binary division with digital hardware

This is the restoring step. Different variant of
division algorithm only have distinction at this
step.

HEY! You may face Nonrestoring or comparison

methods as long questions. Don’t blame me for
that since everything (hardware implementation
and hardware algorithm) is same. Only difference
is at this step.

!!!HEY: In each iteration, just after left -

shifting EAQ, we test it for 0 or 1 and proceed
accordingly which is not noted in example
(Example is taken such that E is always 0 just
after shifting).

Page 8
Other division algorithms
Method described above is restoring method in which partial remainder is restored by adding the
divisor to the negative result. Other methods:

Comparison method: A and B are compared prior to subtraction. Then if A>=B, B is subtracted form A. if
A<B nothing is done. The partial remainder is then shifted left and numbers are compared again.
Comparison inspects end-carry out of the parallel adder before transferring to E.

Nonrestoring method: In contrast to restoring method, when A-B is negative, B is not added to restore
A but instead, negative difference is shifted left and then B is added. How is it possible? Let’s argue:
 In flowchart for restoring method, when A<B, we restore A by operation A-B+B. Next tine in a loop,
this number is shifted left (multiplied by 2) and B subtracted again, which gives: 2(A - B + B) – B =
2A-B.
 In Nonrestoring method, we leave A-B as it is. Next time around the loop, the number is shifted left
and B is added: 2(A-B)+B = 2A-B (same as above).

Exercises: textbook ch 10  10.5, 10.9, 10.10, 10.15

10.5 solution

10.9 and 10.10 solution: do it yourself

10.15 solution:

Page 9
Page 10