IOSR Journal of VLSI and Signal Processing (IOSR-JVSP)

Volume 5, Issue 5, Ver. I (Sep - Oct. 2015), PP 54-59

e-ISSN: 2319 4200, p-ISSN No. : 2319 4197
www.iosrjournals.org

Design and Implementation of Floating Point ALU with Parity

Generator Using Verilog HDL
Mohammad Ziaullah1, Abdul Munaff2
1

(Asst. Professor, Electronics and Communication Engineering, S.I.E.T Vijayapur, Karnataka, India)
2
(Asst. Professor, Electrical and Electronics Engineering, S.I.E.T Vijayapur, Karnataka, India)

Abstract: A floating-point unit (FPU) colloquially is a math coprocessor, which is a part of a computer system
specially designed to carry out operations on floating point numbers. Typical operations that are handled by
FPU are addition, subtraction, multiplication and division. The functions performed are handling of Floating
Point data, converting data to IEEE754 format, perform any one of the following arithmetic operations like
addition, subtraction, multiplication, division . All the above algorithms have been evaluated under Modelsim
environment. All the functions are built by possible efficient algorithms with several changes incorporated at
our end as far as the scope permitted. Consequently all of the unit functions are unique in certain aspects and
given the right environment these functions will tend to show comparable efficiency and speed ,and if pipelined
then higher throughput.
Keywords: Floating, Algorithm, coprocessor, pipelined, throughput.

I. Introduction
In computing, an arithmetic logic unit (ALU) is a digital circuit that performs arithmetic and logical
operations. The ALU is a fundamental building block of the central processing unit (CPU) of a computer, and
even the simplest microprocessors contain one for purposes such as maintaining timers. The processors found
inside modern CPUs and graphics processing units (GPUs) accommodate very powerful and very complex
ALUs; a single component may contain a number of ALUs. ALU is a necessity for a computer because it is
guaranteed that a computer will have to compute basic mathematical operations, including addition, subtraction,
multiplication, and division.

Fig1:General ALU

II.

Floating Point ALU

2.1 Floating Point Unit

When a CPU executes a program that is calling for a floating-point (FP) operation, there are
three ways by which it can carry out the operation. Firstly, it may call a floating-point unit emulator, which is a
floating-point library, using a series of simple fixed-point arithmetic operations which can run on the integer
ALU. These emulators can save the added hardware cost of a FPU but are significantly slow. Secondly, it may
use an add-on FPUs that are entirely separate from the CPU, and are typically sold as an optional add-ons which
are purchased only when they are needed to speed up math-intensive operations. Else it may use integrated FPU
present in the system . The FPU designed by us is a single precision IEEE754 compliant integrated unit. It can
handle not only basic floating point operations like addition, subtraction, multiplication and division but can also
handle operations like shifting, logical operations.

DOI: 10.9790/4200-05515459

www.iosrjournals.org

54 | Page

Design and Implementation of Floating Point ALU with Parity Generator Using Verilog HDL
2.2 Floating point ALU Model

Fig2: Top model

Floating point ALUs are used for high precision computing. This ALU uses 32 bit numbers, which is
the common computer word length. The numbers are represented in IEEE 754 standard. This standard is widely
used in floating point arithmetic. The ALU can perform the arithmetic operation: addition, subtraction,
multiplication and division. It can also perform some logical operations such as logical NOT, logical NAND and
shift right. To support RISC design the arithmetic operations are done in pipelined fashion. There are two
pipelined units in this ALU: the add/subtract pipeline and mult/div pipeline. The addition and subtraction
operations are done in the add/subtract pipeline and the multiplication and division operations are done in the
mult/div pipeline

III. IEEE 754 Standards

IEEE754 standard is a technical standard established by IEEE and the most widely used standard for
floating-point computation, followed by many hardware (CPU and FPU) and software implementations [3].
Single-precision floating-point format is a computer number format that occupies 32 bits in a computer memory
and represents a widedynamic range of values by using a floating point. In IEEE 754-2008, the 32-bit with base
2 format is officially referred to as single precision or binary32. It was called single in IEEE 754-1985. The
IEEE 754 standard specifies a single precision number as having sign bit which is of 1 bit length, an exponent of
width 8 bits and a significant precision of 24 bits out of which 23 bits are explicitly stored and 1 bit is implicit 1.
Sign bit determines the sign of the number where 0 denotes a positive number and 1 denotes a negative
number. It is the sign of the mantissa as well. Exponent is an 8 bit signed integer from -128 to 127 (2's
Complement) or can be an 8 bit unsigned integer from 0 to 255 which is the accepted biased form in IEEE 754
single precision definition. In this case an exponent with value 127 represents actual zero. The true mantissa
includes 23 fraction bits to the right of the binary point and an implicit leading bit (to the left of the binary point)
with value 1 unless the exponent is stored with all zeros. Thus only 23 fraction bits of the mantissa appear in the
memory format but the total precision is 24 bits. IEEE754 also defines certain formats which are a set of
representation of numerical values and symbols. It may also include how the sets are encoded.
3.1 Conversion Of The Binary Integer To Its Ieee754 Format
As our FPU is IEEE754 compliant, the next step is to convert the input (here the effective operand into
the IEEE specified format. IEEE754 single precision can be encoded into 32 bits using 1 bit for the sign bit (the
most significant i.e. 31st bit), next eight bits are used for the exponent part and finally rest 23 bits are used for
the mantissa part.
For example:
S EEEEEEEE FFFFFFFFFFFFFFFFFFFFFFF
31 30
23 22
0
However, it uses an implicit bit, so the significant part becomes 24 bits, even though it usually is
encoded using 23 bits.
This conversion can be done using the below algorithm of above example
Step1: Sign bit of the binary number becomes the sign bit (31st bit) of the IEEE
equivalent.
Step 2: 30th bit to 8th bit of the binary number becomes the mantissa part of the IEEE equivalent.
Step 3: The exponent part is calculated by subtracting the position of the 1st one obtained in the algorithm
described in section 2.2.1.
Step 4: A bias of 127 is added to the above exponent value.
DOI: 10.9790/4200-05515459

www.iosrjournals.org

55 | Page

Design and Implementation of Floating Point ALU with Parity Generator Using Verilog HDL

3.2 Pre-Normalization Of The Operands

Pre-normalization is the process of equalizing the exponents of the operands and accordingly adjusting
the entire IEEE754 expression of the inputs to produce correct results maintaining the IEEE754 standard
throughout all calculation steps inclusive of the intermediate calculations and their outputs.
This conversion can be done using the below algorithm
Step 1: Insert the implicit 1 in the mantissa part of each of the operands.
Step 2: Find positive difference between the exponents of the operands
Step 3: Set the lower operand.s exponent same as that of the operand with higher exponent.
Step 4: Right shift mantissa of the lower operand by steps equal to difference calculated.
3.3 Performing The Selected Operation
After completion of the preliminary steps the next step is to perform the actual operation. The choice of
operation is taken as input via a 4 bit wire oper. Following is the table 2.1 that describes the functions and their
corresponding operation code.
fpu_op
000
001
010
011
100

Operation
Add
Subtract
Multiply
Divide
Shifting

3.4 Module Add

Addition is a mathematical operation which represents combining a collection of objects together to
form larger collection. The process of developing an efficient addition module in our FPU was an iterative
process and with gradual improvement at each attempt.
3.4.1 Add Using The + Operator
The initial attempt was to add using the simple in-built + operator available in Verilog library. It
used a 23 bit register sum and a 1 bit register Co (for carry). The algorithm for the addition can be described
below
Step 1: Check if oper = 4.b000
Step 2: {Co,Sum} = Temp_op1_ieee[22:0] + Temp_op2_ieee[22:0]
Step 3: If carry is 1, then. Resultant_exponent = Larger_exponent + 1;
Else if carry is 0, then do
Resultant_exponent = Larger_exponent . (21-difference) (difference as in sec.2.2.3)
Step 4: Check for overflow and underflow.If for any of the operands (sign(operand with greater exponent)==0 & (exp_greater + 1 > 255)) then, Set the
overflow flag to 1.
Else if (sign(operand with lesser exponent==0) & (exp_lesser<0)), then set the underflow flag to 1
Step 5: Aggregate the result as concatenation of {Sign_bit,Resultant_exponent,Sum}
3.5 Subtract Module
Subtraction is an operation which is treated as inverse of addition operation. The process of developing
an efficient SUB module followed the iterative development of the ADD module.
3.5.1 Sub Using The - Operator
The initial attempt was to subtract using the simple in-built - operator available in Verilog library. It
used a 23 bit register diff and a 1 bit register borrow (for borrow). The algorithm for the subtraction module can
be described.
Step 1: Check if oper = 4.b0001
Step 2: {borrow.diff} = Temp_op1_ieee[22:0] - Temp_op2_ieee[22:0]
Step 3: Resultant_exponent = Larger_exponent + (21-difference)
Step 4: Check for overflow and underflowIf for any operand (sign(operand with greater exponent)==1 AND (exp_greater + 1 < 0))
DOI: 10.9790/4200-05515459

www.iosrjournals.org

56 | Page

Design and Implementation of Floating Point ALU with Parity Generator Using Verilog HDL
Set the overflow flag to 1
If for any operand (sign(operand with exponent)==1'b1 AND (exp_lesser>8'd255))
Set the underflow flag to 1
Step 5: Aggregate the result as concatenation of {Sign_bit,Resultant_exponent,diff}
3.6 Multiplication Module
The process of developing an efficient multiplication module was iterative and with gradual
improvement at each attempt. The product of two n-digit operands can be accommodated in 2n-digit operand.
3.6.1multiplication Using * Operator
It used a 47 bit register to store the product. The algorithm is explained
Step 1: Check if oper = 4.b0010
Step 2: product = Temp_op1_ieee[22:0] * Temp_op2_ieee[22:0]
Step 3: Resultant_exponent = op1_ieee[30:23] + op2_ieee[30:23] - 127
Step 4: If for product ( Resultant_exponent >255 ), then do,
. Set the overflow flag to 1
Step 5: Sign_bit = op1_ieee[31] ^op2_ieee[31]
Step 6: Aggregate the result as concatenation of { Sign_bit, Resultant_exponent, product }
3.7 Module Division
Division is regarded as the most complex and time-consuming of the four basic arithmetic operations.
Given two inputs, a dividend and a divisor, division operation has two components as its result, quotient and a
remainder.
3.7.1 Division Using ./. Operator
The initial attempt was to divide two numbers using the simple in-built / operator available in
Verilog library. It used a 32 bit result_div_ieee register to store the quotient and register remainder to store the
remainder of the division operation.
Step 1: Check if the oper = 4 bit 0100
Step 2: result_div_ieee = temp_op1_ieee[22:0] / temp_op2_ieee[22:0]
Step 3: If op2_ieee[30:0] is all 0
. Set div_bby_zero flag to 1
Step 4: Aggregate the result as concatenation of {Sign_bit, Resultant_exponent, result_div_ieee}

IV. Block Diagram

Fig.3 Block Diagram of ALU

In the ALU top module there are in total 70 I/O pins. It include 32 pins for input operand , 32 pins for
output result, 3 pins for selop, 1 pin for rst,1 pin for clk and 1 output pin for error. For actual arithmetic
DOI: 10.9790/4200-05515459

www.iosrjournals.org

57 | Page

Design and Implementation of Floating Point ALU with Parity Generator Using Verilog HDL
calculations we need two operands but in a microprocessor there is only one databus, and we have to multiplex
this databus. If selop is 000, it is considered as first operand and if selop have any other values, then the operand
that present at that time is considered as the second operand.
Selop
000
001
010
011
100
101
110
111

Operation
Select operand a
Add operand a and b
Subtract operands a and b
Multiply operands a and b
Divide operands a and b
Logically NOT operand a
Logically NAND operand a and b
Shift right the operand a

4.2 Block diagram description

The floating point ALU is internally divided in to three logical modules: the adder/subtracter
module, the multiplier/divider module and the logical module. All the input parameters are given to the input
logic and also the output is taken from the output logic. The input parameters to the ALU include operand,
selop, clk and reset. The output from the ALU includes 32 bit result and error signal. Here selop is used to selet
an operation. In other words the selop is equivalent to the instruction set of a microprocessor. The selop is three
bit wide, and so there are totally seven different operations are possible. All these operations are given in the
table.
4.3 Input Logic
The input logic receives two 32 bit operands one by one and store it in registers a and b. According to
the selop signal, the input logic produces the semdl signal to activate any one of the logical module.
4.4 Adder/Sbtracter
The the adder/subtracter module performs the addition and subtraction operations according to the selop
signal. Internally it is a four level pipelined add/sub unit.
4.5 Multiplier/Divider
The the multiplier/divider module performs the multiplication and division operations according to the
selop signal. Internally it is a four level pipelined add/sub unit.
4.6 Logical Module
The logical module can perform logical NOT, logical NAND and logical right shift operations. These
operations are performed according to the selop signal.
4.7 Output Logic
The output logic accepts the three results res0, res1 and res2 respectively from adder/subtracter,
multiplier/divider and logical module. It directs any one of them to the final result according to the selop signal.

DOI: 10.9790/4200-05515459

www.iosrjournals.org

58 | Page

Design and Implementation of Floating Point ALU with Parity Generator Using Verilog HDL

V. Simulation Results

Fig 4:Simulation Result waveform

Synthesis Results
After the simulation of the code was successful, we proceeded for the synthesis analysis. The
simulation results gave a detailed description of the inputs used showing separately sign bit, exponent bits and
mantissa of inputs and outputs of the operation.

VI. Conclusion
Hence, Floating Point ALU is successfully designed and implemented by using verilog description
language and simulated using modelsim6.3g version. The addition of two floating point numbers are determined
by showing separately sign,exponent,mantissa.

References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]

IEEE 754-2008, IEEE Standard for Floating-Point Arithmetic, 2008.

L. Louca, T. A. Cook, and W. H. Johnson, Implementation of IEEE Single Precision Floating Point Addition and Multiplication on
FPGAs, Proceedings of 83 the IEEE Symposium on FPGAs for Custom Computing Machines (FCCM96), pp. 107116, 1996.
A. Jaenicke and W. Luk, "Parameterized Floating-Point Arithmetic on FPGAs", Proc. of IEEE ICASSP, 2001, vol. 2, pp. 897-900.
DesignChecker User Guide, HDL Designer Series 2010.2a, Mentor Graphics, 2010
Precision Synthesis Users Manual, Precision RTL plus 2010a update 2, Mentor Graphics, 2010.
Patterson, D. & Hennessy, J. (2005), Computer Organization and Design: The Hardware/software Interface , Morgan Kaufmann
] John G. Proakis and Dimitris G. Manolakis (1996), Digital Signal Processing: Principles,. Algorithms and Applications, Third
Edition.
R.V.K Pillai, D. Al-Khalili, and A.J. Al-Khalili. A Low Power Approach to Floating Point Adder Design. In Proceedings of ICCD
'97., Computer Design: VLSI in Computers and Processors, pages 178{185, Austin, TX, October 1997.ICCD.
Jinwoo Suh, Dong-In Kang, and Stephen P. Crago, Efficient Algorithms for Fixed-Point Arithmetic Operations In An Embedded
PIM, 2005, University of Southern California/Information Sciences Institute

DOI: 10.9790/4200-05515459

www.iosrjournals.org

59 | Page