Alford C.digital Design VHDL Laboratory Notes.1996

DIGITAL DESIGN VHDL LABORATORY NOTES
CERL/EE
OCTOBER 16, 1996
VERSION 1.01
COPYRIGHT 1996
DR. CECIL ALFORD
TSAI CHI HUANG
CERL / EE
Digital Design VHDL Laboratory

pepe, 5/23/96
TABLE OF CONTENTS
1. VHDL - HARDWARE DESIGN SOFTWARE APPROACH
1.1 INTRODUCTION
1.1.1 SEQUENTIAL PROGRAMMING
1.1.2 PARALLEL PROGRAMMING
1.1.3 OBJECT BASED PROGRAMMING
1.1.4 PROGRAMMABLE CIRCUITS
1.2 PROBLEMS
6
6
6
8
9
10
2. VHDL BEHAVIOR AND STRUCTURE MODELS
13
2.1 INTRODUCTION
2.2 PROBLEMS
13
18
3. STATE MACHINES AND PROGRAMMABLE LOGIC DEVICES

3.1 INTRODUCTION
3.2 MEALY AND MOORE STATE MACHINES
3.3 VHDL PROGRAMS
3.3.1 MEALY STATE MACHINE
3.3.2 MOORE STATE MACHINE
3.4 PROBLEMS
19
19
19
22
22
23
24
4. DIGITAL DEVICE MODELING
26
4.1 INTRODUCTION
4.2 SRAM MEMORY
4.3 VHDL SRAM MEMORY DESIGN
4.4 SRAM VHDL LISTING
4.5 PROBLEMS
26
26
27
28
30
5. DIGITAL DESIGN USING DIVIDE AND CONQUER
32
5.1 INTRODUCTION
5.2 VHDL FEATURES
5.3 ALU FUNCTIONS
5.4 ALU COMPONENT IMPLEMENTATION EXAMPLES
5.4.1 RIPPLE CARRY ADDER IMPLEMENTATION
5.4.2 PACKAGE IARITH_PARTS_PKG LISTING
32
32
32
33
33
34
2
Copyright 1996, CERL / EE, V1.00

pepe, 5/23/96
5.4.3 TESTADD.VHD LISTING

5.4.4 CIRCULAR SHIFTER IMPLEMENTATION
5.4.5 CIRCULAR SHIFT TRACE DATA
5.4.6 PACKAGE LOGIC_PARTS_PKG LISTING
5.5 ALU IMPLEMENTATION EXAMPLE
5.6 PROBLEMS
6. ALU DATAPTH IMPLEMENTATION
36
37
37
37
40
42
43
6.1 INTRODUCTION
6.2 INSTRUCTION SET ARCHITECTURE (ISA)
6.3 A COMPUTER ARCHITECTURE IMPLEMENTATION
6.4 IMPLEMENTATION EXAMPLE
6.4.1 4X4 REGISTER FILE AND 8X4 MEMORY BLOCK
6.4.2 CONTROL MODULE
6.4.3 EXECUTION EXAMPLE
6.4.4 VHDL LISTING
6.4.4.1 regf4x4.vhd
6.4.4.2 mem4x2.vhd
6.4.4.3 ctrlm.vhd
6.4.4.4 BIA.vhd
6.5 PROBLEMS
7. APPENDICES
43
43
43
44
46
46
47
50
50
52
54
59
59
60
7.1 INSTALL WARP ON PC

7.2 INTRODUCTION TO WARP4 VHDL DEVELOPMENT SYSTEM
60
60
3

pepe, 5/23/96
LIST OF FIGURES
Figure 1-1.
Figure 1-2.
Figure 1-3.
Figure 1-4.
Figure 1-5.
Figure 2-1.
Figure 2-2.
Figure 2-3.
Figure 2-4.
Figure 2-5.
Figure 3-1.
Figure 3-2.
Figure 3-3.
Figure 3-4.
Figure 3-5.
Figure 3-6.
Figure 3-7.
Figure 3-8.
Figure 3-9.
Figure 4-1.
Figure 4-2.
Figure 4-3.
Figure 5-1.
Figure 5-2.
Figure 5-3.
Figure 6-1.
Figure 6-2.
Figure 6-3.
Figure 6-4.
Figure 6-5.
Figure 7-1.
Figure 7-2.
AND-OR Gates _____________________________________________________________ 7

AND-OR Gate Propagation Output _____________________________________________ 8
PLD Circuit Structure ______________________________________________________ 10
LED Control Circuit________________________________________________________ 10
LED Segment Control Label _________________________________________________ 11
Example of Discrete Components _____________________________________________ 13
RS Flip-flop ______________________________________________________________ 14
D Latch __________________________________________________________________ 14
D Latch with Reset Control __________________________________________________ 14
D Flip-flop _______________________________________________________________ 15
Mealy and Moore Machine Block Diagram _____________________________________ 19
Mealy Machine State Diagram _______________________________________________ 20
Mealy Machine Timing Diagram ______________________________________________ 20
Moore Machine State Diagram _______________________________________________ 21
Moore Machine Timing Diagram______________________________________________ 22
Moore Machine to Mealy Machine Conversion __________________________________ 22
Question #1_______________________________________________________________ 24
Question #2_______________________________________________________________ 24
State Flow Diagram ________________________________________________________ 25
SRAM Timing Diagram _____________________________________________________ 27
SRAM Simulation Waveform _________________________________________________ 27
SRAMvc Timing Diagram ___________________________________________________ 31
ALU Block Diagram________________________________________________________ 32
An ALU Design____________________________________________________________ 33
Ripple Carry Adder Waveform________________________________________________ 34
BIA Block Diagram ________________________________________________________ 45
lw Instruction Timing Diagram _______________________________________________ 45
4x4 Register File Block Diagram______________________________________________ 46
BIA State Diagram _________________________________________________________ 48
BIA Simulation Timing ______________________________________________________ 50
Project Window ___________________________________________________________ 60
Device Window ____________________________________________________________ 61
4

pepe, 5/23/96
LIST OF VHDL CODE

Listing 1-1.
Listing 1-2.
Listing 1-3.
Listing 2-1.
Listing 2-2.
Listing 2-3.
Listing 2-4.
Listing 3-1.
Listing 3-2.
Listing 4-1.
Listing 5-1.
Listing 5-2.
Listing 5-3.
Listing 5-4.
Listing 5-5.
Listing 6-1.
Listing 6-2.
Listing 6-3.
Simple VHDL Code for Dataflow Modeling ______________________________________ 7

Digital Design Goodness Criteria ______________________________________________ 9
Example VHDL Code_______________________________________________________ 11
Dataflow Modeling of AND2, OR2 and INV1 Gates _______________________________ 16
D Latch Structure Modeling _________________________________________________ 16
D Latch Behavior Modeling__________________________________________________ 17
D Flip Flop Behavior Modeling ______________________________________________ 18
Mealy Machine VHDL Code _________________________________________________ 23
Moore Machine VHDL Code _________________________________________________ 24
SRAM VHDL Code_________________________________________________________ 30
Iarith_parts_pkg VHDL Code Listing __________________________________________ 36
Testadd.vhd VHDL Code Listing ______________________________________________ 36
Circular Shift Trace Listing __________________________________________________ 37
Logic_parts_pkg VHDL Code Listing __________________________________________ 39
Example ALU VHDL Code Listing ____________________________________________ 42
Regf4x4.vhd VHDL Code____________________________________________________ 52
Mem4x2.vhd VHDL Code ___________________________________________________ 54
Ctrlm.h VHDL Code________________________________________________________ 59
5

1.VHDL - Hardware Design Software Approach
pepe, 5/23/96
1. VHDL - Hardware Design Software Approach

1.1 Introduction
Software programming, no doubt, is a common skill required by all college engineering students. Such
knowledge is indispensable for utilizing a computer as a tool for achieving greater engineering goals in
both the classroom and the real world. Programming tasks for applications range vary in level of detail
and complexity. The readability of programming language constructs span from a very low level, which
is almost at the hardware level such as PAL programming, to a high level such as application level
programming. Good examples of application programming are database, UNIX shell scripts, DOS batch
file, and SPICE circuit simulation. Almost all programming language lie between these two ends. The
major difference between programming languages is their programming complexity which affects such
things as development time. Sequential programming is most common among programming languages.
Examples of sequential languages are BASIC, FORTRAN, Pascal and C. Besides sequential
programming, there are other programming paradigms: parallel programming and object based
programming. These paradigms are new for most people, and they will be introduced during the process
of learning VHDL. VHDL is not much different from conventional programming languages, except it
uses sequential, parallel, and object based software paradigms to achieve its goal: namely, digital
hardware simulation and synthesis. In other words, a VHDL program is used to specify, to model, and to
create digital hardware.
VHDL stands for VHSIC (Very High Speed Integrated Circuit) Descriptive Language. VHDL is
promoted by the Department of Defense (DOD) mainly to facilitate the standardization of all digital
hardware design. VHDL serves as a documentation tool at the beginning to describe a system in a unified
manner across the cooperative group of a project. VHDL was started in early 1980 inside of DOD and
eventually got standardized by IEEE with the name standard 1076 (IEEE-STD-1076) in 1987. Although
further enhancements have been added to the later version of VHDL, for example standard 1164 (IEEESTD-1164), standard 1076 will be sufficient to learn and explore most of the language feature.
1.1.1 Sequential Programming

This type of programming is characterized by the sequential nature of program flow. Popular
programming languages such as Pascal, C, BASIC and FORTRAN are the examples in this category. It is
also called imperative programming because programming statements to be executed are imperative.
Programs are executed in sequential order and a statement is executed following the statement that
precedes it in the proogramm. There is always a clear decision as far as which the next program
statement is to be executed next. In summary, in sequential programming, there is no ambiguity as far as
the statements execution order is concerned.
In VHDL syntax, the statements after the reserved keyword process or within the process block will be
executed in sequential order. Typical VHDL sequential statements are condition statements, loop
statements, and assignment statements. The syntax of each type of statement can be found in any
reasonably thorough VHDL tutorial text book.
1.1.2 Parallel Programming

The parallel programming concept is a more difficult programming paradigm to grasp for people used to
sequential programming. The parallel programming concept is straightforward and is expressed through
6

pepe, 5/23/96
its name; statements are executed in parallel. Moreover, the process of executing statements in parallel
mimics the real world phenomena and thus coincide better with the behavior of digital circuits. In
parallel processing, the application of parallel programmingg, programs executed in parallel have the
advantage of boosting program execution speed versus the equivalent sequential program. In the area of
VHDL programming, the goal is to mimic the physical digital system. Due to the fact that hardware
operates executed in parallel, to capture the functionality of the hardware, it is important for VHDL to
have the features to facilitate the modeling of such behavior.
Parallel program statements, as the name suggests, are the statements that can be executed in parallel.
Parallel statements are less intuitive to use due to their nondeterministic nature. Ideally, all parallel
statements should be executed at the same time. However, in the popular Von Neumann machine or
sequential computer world (eg., PCs and worstations), only sequential statements can be directly executed.
Therefore, parallel programming must be simulated on a sequential machine. When simulating parallel
statements, due to the fact that statements are executed in parallel and random order, it is crucial that the
execution order be irrelevant and not affect the result.
An example of a parallel system is shown in Figure 1-1. It describes a digital hardware circuit where four
inputs and one output are involved. Inputs A, B, C and D are fed into two AND gates, and the outputs of
these AND gates, E and F, are then fed into an OR gate whose output is shown below(see Figure 1-1).
A
B
E
G
C
D
F
Figure 1-1. AND-OR Gates
A piece of simple VHDL code shown in Listing 1-1 describes the circuit above. Notice the section after
dataflow_machine architecture in the code. Three statements, a, b, and c describe AND and OR logical
assignments. These statements can be rearranged in any order, and the simulated output G should still be
the same.
-- Sum of product example
entity sop is
port (
a, b, c, d: in bit;
e, f: inout bit;
g: out bit);
end sop;
architecture dataflow_machine of sop is
begin
-- Begin parallel statements
g <= e or f;
e <= a and b;
f <= c and d;
-- Statement a
-- Statement b
-- Statement c
end dataflow_machine;
Listing 1-1. Simple VHDL Code for Dataflow Modeling
7

pepe, 5/23/96
The result G will be the same if the statements assigning E, F and G outputs were placed in any order.
Any data arriving at the inputs A, B, C, and D will pass through the AND and OR gates in continuous
time fashion. Whatever shown up at A, B, C or D inputs will be reflected at the intermediate outputs E
and F and at the final output G. Thus, output G should change instantly according to the essence of
parallel programming. In the world of hardware execution, there is always a delay when data pass
through gates. The VHDL code of hardware synthesis is mapped into a Cypress CPLD 375 device and
fitted with optimization disabled. The simulation of the mapped Cypress 375 PCLD shows that there is a
delay of about 5 ticks between each logical gate. Notice that ticks from VHDL simulation below do not
represent any real-time. Instead, a simulation tick is simply a unit of delay of simulation time. In the real
world, gates delay are on the order of nanoseconds (10-9 sec).
Figure 1-2. AND-OR Gate Propagation Output

Intuitively, the parallel VHDL statements above should be executed in the order: c, a and b. However, the
hardware VHDL simulator handles the situation differently. The VHDL simulator executes the parallel
statements in clock random order fashion.
Notice that VHDL statements written within the process block have been executed sequentially.
Therefore, all VHDL statements outside of the process block will be executed in parallel. For example,
two process blocks can be executed in parallel. To resolve sequential statements inside of each process
block, VHDL uses the concept of delta time slots for each clock tick. In each clock tick, the simulator
allocates however many sub-ticks called delta for simulating the instantaneous statement execution.
Individual parallel statement are being executed at delta ticks. A simulated clock tick will only move
forward when the delta tick yields consistent outputs. An example above would be stable E, F, and G
outputs.
1.1.3 Object Based Programming

Objected based programming is different from the sequential and parallel paradigms. It should be treated
more like a methodology than a paradigm. Objected based methodology basically enforces structure into
VHDL. This facilitates the systematic construction of large and complicate digital system. Object based
programming also supports the concept of abstract data types enabling the VHDL engineer to construct
custom data types and operations and then to store them systematically for future VHDL projects. This is
the concept of VHDL code reuse. Since the feature of breaking VHDL code into small components and
libraries is already built into the language, it provides the VHDL designer with the structural divide and
conquer technique when building larger systems. Moreover, the concept of code reuse definitely saves
8

pepe, 5/23/96
system development time. In summary, object based features are extremely useful for VHDL code
development, management, and reusability.
1.1.4 Programmable Circuits

Digital design technology has evolved greatly in recent years. This can be seen in the field of ever
changing microprocessor design where complexity and performance double every year. With
improvements to digital hardware technology, the criteria used to determine a designs goodness remain
Correctness
Ability to exactly perform their tasks, as defined by the requirement and specification.
Robustness
Ability to function even in abnormal conditions.
Extendibility
Ability to ease change in specification change.
Reusability
Ability to be reused, in whole or in part, for new applications.
Compatibility
Ability to combine with others.
Listing 1-2. Digital Design Goodness Criteria

the same, and they are listed in Listing 1-2.
These five criteria are borrowed from the book, Object-oriented Software Construction, by Bertrand
Meyer. Although, they were used to describe software quality, the concepts are also appropriate to be used
on hardware. Notice that hardware design technical merits such as speed and cost are not mentioned
because they highly depend on the state of technology and market demand.
The word, flexibility, sums up all these criteria above. The practice of digital design nowadays must be
flexible. Thus, the trend is to replace ridged discrete logic component with programmable devices to
maximize hardware flexibility.
Traditional programmable circuits comes in three flavors of simple combinational logic circuit:
Programmable Read Only Memory(PROM), Programmable Logic Array(PLA), and Programmable Array
Logic(PAL). All these three devices contain similar hardware structure of AND and OR arrays. Input
goes from the AND array to the OR array. The difference is that PROM contains fixed AND and
programmable OR arrays, PAL contains programmable AND and fixed OR arrays, and PLA contains
programmable AND and OR arrays. (See the figure Figure 1-3.)
9

pepe, 5/23/96
Programmable Combinational Logic
PAL's Programming Points
PLA's
Programming
Points
PROM's Programming Points
Figure 1-3. PLD Circuit Structure

Most of the time, a digital hardware designer doesnt need to know the circuit structure of the
programmable logic device (PLD) that he is using except he is conscious about the designs speed and
capacity. For example, he might want to try to fit a large piece of circuit into a given PLD or to select a
type of fast PLD to meet a timing limit. Digital hardware can be programmed to fit a circuit into a PLD
using hardware programming languages such as CUPL, Abel, PALism, and VHDL.
For more complex PLDs (CPLDs), memory elements such as flip flop devices are involved and they will
be introduced later.
1.2 Problems
1) Binary Coded Decimal (BCD) is a useful decimal number representation using binary numbers.
Many times, it is the only representation that digital hardware uses to communicate with humans
since humans are custom to the decimal numbering system. One example is to use BCD to control
the decimal LED display panel of a calculator. Each LED decimal digit has eight segments and each
segment contains an LED source as shown in Figure 1-4.
+5 V
Input
Figure 1-4. LED Control Circuit

At high input 1, current flows through the LED and resistor, and the LED glows. The typical
current is about 20mA, and voltage drops 1.6 Volts(V) across the LED and 3 V across the resistor. If
input is low 0, no current will flow through LED and it will become dim.
The LED segment specification is illustrated in Figure 1-5.
10

pepe, 5/23/96
a
f
b
c
d
Figure 1-5. LED Segment Control Label
The input consists of 5 bits, and the output consists of 7 bits. Out of 5 input bits, 4 bits are used to
specify binary value, and one bit is used to blank all LED segments. For all output bits, one bit is
used for controlling each of the segment, and one bit is used to indicate overflow of the input binary
number, for example binary input greater than 9. The LED logic specification is shown in Table 1-1.
An LED segment control decoder is used to control a one-digit LED display panel. Construct such a
decoder using VHDL. This can be done using if statements nested inside of a process statement or
when statements alone. The form of such decoder could look like the VHDL code segment in Listing
1-3.
entity XXX is
port( .... );
end XXX;
architecture XXXX of XXX is
begin
-- option 1
option1: process ( input)
begin
if <condition> then
<output> <= <input>;
elsif <condition> then
.....
else
end if;
end process option1;
-- Or use option 2
output <= <input> when <condition> else
<input> when <condition> else
.....
<input> when <condition> else
<input>;
end XXXX;
Listing 1-3. Example VHDL Code

After completing the implementation, test the decoder for various inputs and simulate the resulting output
using the NOVA simulator. The simulation session can be saved into the trace file with the PSD
extension. This is done by using the mouse to select FILE and then Write Trace (*.PSD) on the top of the
NOVA menu bar. In the saved PSD file, annotate the clock tick that corresponds to the decoder action.
For example, the input that leads to the output display.
11

pepe, 5/23/96
INPUT
OUTPUT
Comment Blank in1 in2 in3 in4 Overflow a

Blank
0
1
2
3
4
5
6
7
8
9
a
b
c
d
e
f
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
x
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
x
0
0
0
0
1
1
1
1
0
0
0
0
1
1
1
1
x
0
0
1
1
0
0
1
1
0
0
1
1
0
0
1
1
x
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
0
1
0
0
1
0
0
0
0
0
0
1
0
1
0
0
1
0
0
0
0
0
1
1
0
0
0
0
1
1
0
0
1
1
0
0
1
0
0
0
0
0
0
0
0
0
1
0
1
1
1
0
1
0
0
1
0
0
1
0
1
0
0
0
0
0
1
1
0
1
0
1
1
1
0
1
0
1
0
0
0
0
0
0
1
0
1
1
1
0
0
0
1
0
0
1
0
0
1
0
0
1
1
1
0
0
0
0
0
1
0
0
0
0
1
0
0
0
Table 1-1. LED-Control Truth Table
12

2.VHDL Behavior and Structure models
pepe, 5/23/96
2. VHDL Behavior and Structure models

2.1 Introduction
The ability to remember is important in the field of digital systems. The digital memory element is a
necessary and fundamental block for building complex digital systems. Despite its importance, it takes no
more than basic digital logic for its construction. There are many ways to construct a digital memory
element using basic logic gates such as NOT, AND, and OR.
There are three ways to model a digital system using VHDL. A piece of VHDL hardware can be modeled
in structural, data flow and behavior methodologies. In this lesson, methods of building memory elements
such as latches and flip flops are studied using both single and mixed methodologies.
A natural way to describe discrete components is to use a data flow model. Discrete gates such as a one
input and one output NOT gate, INV1, two inputs and one output AND gate, AND2, and two inputs and
one output OR gate, OR2, can be modeled fairly easily, and they are illustrated in Figure 2-1. The logic
gates in Figure 2-1 have their inputs and outputs labeled. These components will be used to illustrate
more examples of elaborate logic devices, such as a digital memory element.
Example of Discrete Components

n
nx
INV1 Gate
x
y
z
OR2 Gate
x
z
y
AND2 Gate
Figure 2-1. Example of Discrete Components

There are many types of memory elements that one can create using the basic gates shown in Figure 2-1.
A simple memory block that one can make is an RS flip flop which uses two OR2 and INV1 gates. For
the sake of consistency with the previous example, NOR2 is not included for now. Even though OR2 is
used to create an RS flip flop, DeMorgans law would also permit AND2 to be used as well.
Notice the flip flop is named RS because the flip and flop can be set (output=1) or reset (output=0)
depending on the inputs (R and S). When both R and S inputs are set to zero, outputs (q and nq) will be
suspended to the previous values of R and S. Be aware that both R and S, at the same time, cant be set to
one for two reasons. The first reason is that once R and S become one, output q and nq will no longer be
complementary to each other. Both q and nq will become zero, and this violates the assumption of
complementing flip flop output. Second, once the output q and nq become zero, the next state after R and
S change, outputs q and nq are impossible to predict due to the feedback of q and nq values. See Figure 22 for the RS flip flop logic truth table.
Although an RS flip flop is easy to implement, it is not an ideal memory device due to its incomplete input
states. From the perspective of a digital designer, it is cumbersome to guarantee the partial input state. A
better way to use the same memory block is to modify the RS flip flop to take care of the partial input
problem. One simple solution is to add two AND2 gates at the front of RS flip flop to prevent R and S
from becoming one at the same time. The resulting circuitry is called a D latch. See Figure 2-3 for this
modified memory element..
13

pepe, 5/23/96
RS Flip Flop
R
0
0
0
0
1
1
1
1
o1
n2
nq
n3
o2
S
0
0
1
1
0
0
1
1
(next)
q -> q
0
0
1
1
0
1
1
1
0
0
1
0
0
x
1
x
Figure 2-2. RS Flip-flop

A D latch functions very similar to an RS flip flop because it remembers past states. However, it works
differently because output is triggered by the level of the input signal instead of the input signal edge. As
a matter of fact, a flip flop works differently from a latch because the flip flop output signal is triggered at
the clock edge and the latch output signal is triggered at the level of clock signal.
D Latch
o1
n2
n1
a1
o2
a2
n_q
n3
Figure 2-3. D Latch

Digital circuitry during the beginning of operation, especially memory elements, will go to erroneous
states unless provision is made for startup operation. At the initial stage of the memory operation, the
value stored inside of the memory element is unknown, and, if this unknown is to be applied to rest of the
system, the system will almost be guaranteed to produce error. The modification to D latch is shown in
Figure 2-4. The new latch get reset circuitry so the internal value stored can be set to zero (low) whenever
the reset signal goes high.
D Latch with Reset Control
reset
o3
o1
n2
n1
a1
n3
a3
a2
o2
n3
n_q
Figure 2-4. D Latch with Reset Control

although the difference between flip flops and latches have been discussed, it is important to note the
application side of these memory elements. In the world of digital design, the majority of digital system is
14

pepe, 5/23/96
running synchronously with the system clock. Since flip flops activate at an instance in time to acquire
data, relatively speaking, flip flops are more useful than latches because the output of flip flops is held
constant for a full clock cycle instead of half a clock cycle.
Cascading two D latches can produce a D flip flop. The clock signal that runs through these two latches
is inverted so data can pass through the latches in pipeline fashion. Due to the holding of input data at the
lower phase clock signal, input signal of the first latch is preserved for the instance when the clock signal
goes down. Simultaneously, the signal output from the first latch is acquired by the second latch, and the
signal is held constant until the second latchs clock goes down. A D flip flop drawing made by cascading
two D latches is shown below in Figure 2-5.
D Flip Flip
reset
d
reset
reset
nq
D Latch
nq
D Latch
c
Figure 2-5. D Flip-flop
Now the memory basics have been explained, and it is time to model some of the memory block using
VHDL. The difference between structure, dataflow, and behavior models will be shown in later examples.
For example, in Listing 2-1, each discrete logic component, AND2, OR2 and INV1, is defined at the
beginning using the dataflow model as state before. Notice that VHDL programming structure is declared
with entity statement first and followed by an architecture statement. Signals defined in the entity block
under the keyword port are visible inside of the architecture block.
-- INV1 Gate
-entity inv1 is
port(
x: in bit;
nx: out bit);
end inv1;
architecture behav of inv1 is

begin
nx <= not x;
end behav;
--AND2 Gate
-entity and2 is
port(
x, y: in bit;
z: out bit);
end and2;
architecture behav of and2 is

begin
z <= x and y;
end behav;
-- OR2 Gate
-entity or2 is
port(
x, y: in bit;
z: out bit);
end or2;
15

pepe, 5/23/96
architecture behav of or2 is

begin
z <= x or y;
end behav;
Listing 2-1. Dataflow Modeling of AND2, OR2 and INV1 Gates

As its name implies, structures modeling uses the gates defined above to connect to others to make into a
system. The process of describing the system structure in gate connection is shown in Listing 2-2. The
VHDL code after the gate definitions shown in Listing 2-1 shows how to connect gates, and this is done
through the use of VHDL statement map.
--- structure modeling of D latch
--
entity lat_ds1 is
port(
c, d: in bit; -- input data
q, nq: inout bit);
end lat_ds1;
architecture struct of lat_ds1 is
signal i1_nx: bit;
signal i2_nx: bit;
signal i3_nx: bit;
signal a1_z: bit;
signal a2_z: bit;
signal o1_z: bit;
signal o2_z: bit;
component and2
port(x, y: in bit; z: out bit);
end component;
component or2
port(x, y: in bit; z: out bit);
end component;
component inv1
port(x: in bit; nx: out bit);
end component;
begin
-- structure connection
a1: and2 port map(c, i1_nx, a1_z);
a2: and2 port map(c, d, a2_z);
o1: or2 port map(a1_z, nq, o1_z);
o2: or2 port map(a2_z, q, o2_z);
i1: inv1 port map(d, i1_nx);
i2: inv1 port map(o1_z, q);
i3: inv1 port map(o2_z, nq);
end struct;
Listing 2-2. D Latch Structure Modeling

The variables inside of the mapping statements are declared after the D latch architecture declaration,
architecture struct of lat_ds1 is, and before the begin statement. These signals can only be used here, and
they are not accessible by other architecture code blocks.
16

pepe, 5/23/96
Gate components used here are also required to be declared. The declaration of a component is similar to
C, FORTRAN or Pascal programming where all procedures or subroutines are required to be declared
before use. However, it is not exactly the same as doing programming because each component declared
could be physical part rather than a conceptual abstract. For example, the ports declared by component do
have corresponding physical meaning, namely wires to carry signals. This concept should be understood
from the digital hardware engineer perspective.
Now, consider taking the same D latch circuit and describing or modeling it behaviorally in VHDL. This
code is shown in Listing 2-3. As we have noticed, the VHDL code here is shorter than its corresponding
structure model because the circuit behavior is directly described. It is a higher level modeling, and it can
be done easily if the designer knows exactly how his circuit behaves. Many times, the digital circuit
designer knows how the circuit should behave but is not sure if the actual circuit behaves as he predicts.
Therefore, it is useful to construct the circuit as a structure model and verifies its working order. Behavior
modeling saves time and it is based on the assumption that the designer knows his circuits behavior well.
If the assumption about the circuit is wrong, then behavior modeling can lead to a disastrous result.
Remember, behavior modeling should be used as much as possible when modeling digital circuits because
it will save much development time.
In the field of VLSI design, the translation of a behavior model to a structure model is called hardware
synthesis, and it relies much on the software intelligence. During the process of mapping a VHDL circuit
to the actual hardware, it is best to specify as much structure as possible if the speed and the size
requirement of the circuit are a concern. As an analogy, think of a programmer who wants his code to be
as fast and efficient as possible. The programmr does better write the code in assembly (or even better machine language) instead of relying on a the high level language compiler. For the exercises here,
problems mainly focus on translating VHDL code to hardware. In other words, VHDL will be used as a
programming language to map into hardware. Again, remember the concept of hardware synthesis. In
hardware synthesis, the digital designer needs to be conscious of the final form of the mapped circuit.
--- behavior modeling of D latch
--
entity lat_db1 is
port(
c, d: in bit;
-- input data
q, nq: inout bit
);
end lat_db1;
architecture behavior of lat_db1 is
begin
-- behavior model
behavior: process (c)
begin
if c='1' then
q
<= d;
nq <= not d;
end if;
end process behavior;
end behavior;
Listing 2-3. D Latch Behavior Modeling
17

pepe, 5/23/96
Another example of behavior modeling is again the D flip flop construction using two D latches which
was illustrated earlier in Figure 2-5. Listing 2-4, however, shows the construction of a D flip flop without
using any D latches. By using the behavior modeling technique, the VHDL code length was reduced
drastically. A circuit similar to a D flip flop in size would be tedious to describe totally described in a
structural model.
--- behavior modeling of D latch
-entity rdffb is
port(
reset: in bit;
-- reset signal
c, d: in bit;
-- input data, c=clock, d=data
q, nq: out bit ); -- output q and not q
end rdffb;
-- behavior model
architecture behavior of rdffb is
begin
behavior: process (c, reset)
begin
if reset='1' then
q
<= '0';
nq <= '1';
elsif (c='1' and c'event) then
q
<= d;
nq <= not d;
end if;
end process behavior;
end behavior;
Listing 2-4. D Flip Flop Behavior Modeling
2.2 Problems
1) Implement a D latch with reset using VHDL structure modeling. Demonstrate its functionality in the
Cypress NOVA simulator by resetting the circuit, latching in a high value, and resetting again at the
next clock cycle when the input is kept high. Save the simulation session to a stimulus file (*.SIM)
and trace file (*.PSD) using the NOVA FILE menu items.
2) A T flip flop can be defined as a flip flop that toggles the output at the edge of the clock cycle, for
example, rising or falling edge. Implement T flip flop behaviorally. Again, save the simulation
session in both stimulus file and trace file.
3) Implement an RS flip flop using an AND gate, and model it structurally. Demonstrate and show its
functionality.
4) In VHDL, implement the circuit described below using two D flip flops with /RESET input. The
circuit should first be implemented in a behavior model and, then, in a structure model. Show the
functionality using the VHDL simulator. (Hint: Utilize /RESET input).
INPUT:
Two input lines, A and B.
OUTPUT: One output C.
CONDITIONS:
a) The output goes to 1 on every positive transition of the A line.
b) The output goes to 0 on every positive transition of the B line.
c) The output must be capable of set to 1 again after it being to 0.
d) Input B takes precedence over Input A to affect the output.
18

3.State Machines and Programmable Logic Devices
Pepe, 4/5/96. 5/23/96
3. State Machines and Programmable Logic Devices

3.1 Introduction
Simple programmable devices that we saw earlier can be extended into state machine by a combining
memory elements (such as latches and flip flops) with combinational logic elements such as an AND-OR
array. A classical example of constructing state machine uses a PROM and latches. Although the
construction is simple, the design lacks flexibility for testing. In addition, the use of discrete PROM and
latches for creating state machine suffers performance loss due to the hardware interconnection overhead.
Modern state machine construction uses programmable devices such as PLDs that integrate memory and
discrete elements into one block of hardware real-estate.
3.2 Mealy and Moore State Machines

There are two classical state machine constructions, and they are described in most of the digital design
textbooks. They are Mealy and Moore state machines. It can be seen from Figure 3-1 that both state
machines are very similar. The Moore machine uses only the output of a delay element to generate the
system output whereas the Mealy machine uses both the system input and a delay element output for
system output. In theory, both machines are equivalent, so one can convert a synchronized digital
hardware design from a Mealy to a Moore state machines or vice versa. Such a statement is not true in
the real world of digital design. The Moore state machine generates output based only on its current state.
Therefore, the Moore machine output is usually synchronized with the system clock or the clock edge of
the output D flip flop (the delay element). In the Mealy machine, because output depends on both the
input and current state, output will change whenever the input changes. In other words, in the Mealy
system, when output is released at a clock edge, the input is usually latched and considered prior to the
release of latched output.
Mealy and Moore State Machine Models
Delay Element
Moore State Machine Output

Mealy State Machine Output
Combinational Logic
Input
Figure 3-1. Mealy and Moore Machine Block

Diagram
The Mealy machine is more difficult to understand than the Moore machine. Because the Mealy machine
outputs based on the input and the current state, the correct initiating output process in a Mealy machine
is first to change the input value. Due to the dependency of input and the current state, Mealy machine
output changes next. However, the change at the input will also trigger the change at the machine state.
At last, the output will change again due the new state. Notice that Mealy machine output only at the
edge of the state transition, see Figure 3-3. whereas the Moore machine output coincide with the state
change, see Figure 3-5.
An VHDL example of a rising edge detector will be constructed using both machines. First, the state
diagram of such device as a Mealy machine is shown in Figure 3-2.
19

Pepe, 4/5/96. 5/23/96
0/0
1/0
0/0
HIGH
LOW
1/1
Rising Edge Detector

(Mealy Machine)
Figure 3-2. Mealy Machine State Diagram

There are two states in the state machine: HIGH and LOW. Notation x/y next to an arc between states
stands for input/output values during the state transition. The machine advances to the next state when it
sees an input x. In mean time, the machine also outputs y. The VHDL code for the state machine shown
in Figure 3-2 is listed in Listing 3-1.
Listing 3-1 contains the usual VHDL components such as entity, architecture declaration and the VHDL
body. A type statement after the architecture declaration declares the enumerated type StateType. It is
advantageous to use VHDL enumerated type so each state of the state machine can have meaningful name
(for example, HIGH and LOW). Two signal identifiers present_state and next_state are declared and
used like any identifiers, except they can only be assigned an enuerated state value, HIGH or LOW. The
use of the identifiers present_state and next_state can be found in the process blocks: state_machine,
state_clocking, and mealy_output process statements. These blocks of the VHDL program execute their
statements in parallel. In particular, depending on the input, state_machine controls the state to which to
advance. Mealy_output generates output depending on the current state and input. State_clocking
dictates the synchronization between the state machine and clock rising edge. The timing diagram is
shown in Figure 3-3. This figure shows that input triggers output instantly and that present_state does not
change its value until the beginning of the next rising clock edge. Thus the timing diagram, Mealy
machine timing diagram, agrees with which statement stated above where outputs at the state transition of
clock rising edges. In addition, it also shows that Mealy machine output is not registered and the machine
will reflect whatever input value it sees on its output.
Figure 3-3. Mealy Machine Timing Diagram

Given a Mealy machine, it is possible to convert it to an equivalent Moore machine using a deterministic
algorithm. With the differences of Mealy and Moore machines in mind, the algorithm consists of the
folllowing steps:
20

Pepe, 4/5/96. 5/23/96
1) Put the state machine diagram of edge detector in Figure 3-2 into table form, as shown in Table 3-1.
Realize that the table entry is constructed based on state transitions. Because there are four
transitions, four table entries (four columns) are shown.
2) Combine the last two rows of table entries, output state and output value, to form new Moore machine
states. According to the table, three unique states result: HIGH0, HIGH1, and LOW0.
Table 3-1.
Entry #
Input State
HIGH
LOW
LOW
HIGH
Input Value
Output State
HIGH
HIGH
LOW
LOW
Output Value
3) Create the new Moore machine state table from Table 3-1 using the new states. For example, the
Mealy state machine transition table entry column 1 contains four values: input state, input value,
output state, and output value. As shown before, output state and output value are used to generate
the corresponding Moore machine state, HIGH0. Input state and input value are to be used to
generate the corresponding Moore machine state transitions. Here, Mealy input state HIGH is
replaced by the Moore state HIGH0 and HIGH1 (HIGH plus all possible outputs). Now, two
corresponding Moore state transitions are derived; HIGH0 and input 1 goes to HIGH0, HIGH1 and
input 1 goes to HIGH0. The same process applies to the other entry columns (2, 3, and 4), see Table
3-2.
Table 3-2.
Moore States
Input 0
Input 1
Output Value
LOW0
LOW0
HIGH1
HIGH0
LOW0
HIGH0
HIGH1
LOW0
HIGH0
4) Make a Moore machine state diagram based on this state transition table, as shown in Figure 3-4.
1
HIGH0
0
1
0
0
LOW0
HIGH1
Rising Edge Detector

(Moore Machine)
Figure 3-4. Moore Machine State

Diagram
5) Describe the Moore machine in VHDL constructs, as shown in Listing 3-2.
21

Pepe, 4/5/96. 5/23/96
6) Express the Moore machine operation in the form of a timing diagram. The timing Diagram of the
derived Moore machine is shown in Figure 3-5. In the figure, the output signal coincides with the
state transition, as is expected from a Moore state machine.
Figure 3-5. Moore Machine Timing Diagram

The inverse process of converting a Moore state machine to a Mealy state machine is simpler. The
steps are rather trivial and consist of changing every state output in the Moore machine to the out
going transition output of the corresponding Mealy machine. This process effectively converts a
machine from a state output state machine to a transition output state machine. See Figure 3-6.
input
State0
input/output
State1
State0
State1
output
Moore Machine
Mealy Machine
Figure 3-6. Moore Machine to Mealy Machine Conversion

When selecting Mealy versus Moore state machine types for a design, one needs to keep in mind the
advantages and disadvantages of each construction. For a Mealy machine, system equations are intuitive
to derive because input is readily available. Due to its simplicity, a Mealy machine contains fewer
memory elements than a Moore machine. However, because of input dependency, Mealy output also
depend highly on the input. For slow input, this phenomenon can translate into a performance bottleneck.
Moore machine, on the other hand, does not suffer such input dependency since it buffers the system
input. However, the tradeoff will be that Moore machine needs more hardware to construct.
3.3 VHDL Programs

3.3.1 Mealy State Machine
entity mea_edge is port (
reset, clk, input: in bit;
output:
inout bit;
mea_output: out bit );
end mea_edge;
-------------------------------------------------------------------architecture mealy_machine of mea_edge is
type StateType is (LOW, HIGH);
signal present_state, next_state: StateType;
begin
---------------------------------------------------------------------
22

Pepe, 4/5/96. 5/23/96
state_machine: process(present_state) begin
case present_state is
when LOW => if (input='0') then
next_state
<= LOW;
else
next_state <= HIGH;
end if;
when HIGH => if (input='0') then
next_state
<= LOW;
else
next_state
<= HIGH;
end if;
end case;
end process state_machine;
---------------------------------------------------------------------state_clocking: process(clk) begin
if (clk'event and clk='1') then
present_state <= next_state;
mea_output <= output;
end if;
end process state_clocking;
----------------------------------------------------------------------mealy_output: process(input, present_state) begin
if (present_state=LOW and input='1') then
output <= '1';
else
output <= '0';
end if;
end process mealy_output;
end mealy_machine;
Listing 3-1. Mealy Machine VHDL Code
3.3.2 Moore State Machine

entity moo_edge is port (
reset, clk, input: in bit;
output:
out bit );
end moo_edge;
architecture moore_machine of moo_edge is
type StateType is (LOW, HIGH0, HIGH1);
signal present_state, next_state: StateType;
begin
state_machine: process(present_state) begin
case present_state is
when LOW => if (input='0') then
next_state
<= LOW;
else
next_state <= HIGH1;
end if;
when HIGH0=> if (input='0') then
next_state
<= LOW;
else
next_state
<= HIGH1;
end if;
when HIGH1=> if (input='0') then
next_state
<= LOW;
else
next_state
<= HIGH0;
end if;
end case;
end process state_machine;
23

Pepe, 4/5/96. 5/23/96
state_clocking: process(clk) begin
if (clk'event and clk='1') then
present_state <= next_state;
end if;
end process state_clocking;
moore_output: process(present_state) begin

if (present_state=HIGH1) then
output <= '1';
else
output <= '0';
end if;
end process moore_output;
end moore_machine;
Listing 3-2. Moore Machine VHDL Code
3.4 Problems
1) Convert the Moore machine in Figure 3-7 into a Mealy machine. Implement the Mealy machine in
VHDL and simulate it with Cypress NOVA simulator.
0/1
Start
q1
0/0
q0
1/0
0/0
1/1
1/0
q2
Figure 3-7. Question #1

2) Simulate the state machine shown in Figure 3-8 with VHDL and describe what it does (Hint: See
next question).
1
0
1
q0/0
q1/1
q2/2
Start
Figure 3-8. Question #2

3) Design a Mealy state machine that computes the remainder of value 5 instead 3. Draw the state
diagram and simulate it with VHDL. Test the machine using input 3, 4, 5, 6, 76, and 101.
4) Design the state machine of the state flow diagram in Figure 3-9 and the output table in Table 3-3.
24

Pepe, 4/5/96. 5/23/96
Table 3-3.
frame
IDLE
/frame
frame
DECODE
/hit
hit
/frame
/frame
XFER
BUSY
State / Output
OE
GO
ACT
IDLE
DECODE
BUSY
XFER
XFER2
XFER2
Figure 3-9. State Flow Diagram
25

pepe, 4/30/96
4. Digital Device Modeling

4.1 Introduction
Several VHDL concepts have been addressed in the previous lessons. VHDL topics covered in the
previous lessons are various digital circuit modeling methodologies: structural, behavioral, and dataflow
models. In addition to digital modeling, the text also covered the implementation of two types of state
machines: Moore and Mealy state machines using the VHDL process statement. The Process statement
is a unique feature in comparison to conventional imperative programming languages such as C, Pascal,
FORTRAN, Ada, C++, etc. Due to the concurrent nature of VHDL programming, Process blocks
execute/simulate in parallel. Statements inside of the process block, on the other hand, are executed
sequentially. Parallel simulation is important because real world digital systems operate and behave in
parallel. In the following lesson, this same concept of parallel executing process blocks will be addressed
again. The emphasis here will be more towards the solving of real world application problems such as
digital component modeling.
4.2 SRAM Memory

Memory plays an important role in the advancement of modern computers. When design a computing
system, different types of memory are hierarchically structured into layers according to their data handling
speed. The CPU is usually the fastest component and it is surrounded by various levels of memory. The
various types of memory work in conjunction with the CPU: for example, on-chip and off-chip memory,
memory chips surround the CPU on the motherboard, hard disk memory, and secondary memory device
such as tape backups. To make computer efficient so the CPU will not be bogged down by the slower
memory, different speeds of memory need to be used. Memory used by CPU are classified in general as
SRAM (static RAM) and DRAM (dynamic RAM). SRAM memory runs faster than DRAM, and it
interfaces directly with the CPU or the CPU on chip memory such as the L1 cache of the system. DRAM,
on the other hand, is slower than SRAM by an order and requires more control circuitry, but has the
advantages of smaller cell size and lower cost.
DRAM memory is denser than SRAM due to its circuit simplicity. Unlike SRAM which uses flip flops,
DRAM stores information in capacitors, and information will only be kept until the charges in the
capacitors leak out. As a consequence, more effort is needed to design DRAM refresh circuitry to
preserve the stored DRAM information. Refreshing DRAM means re-charging the capacitors.
SRAM memory is much simpler to use than DRAM because it does not require a refresh operation. A
generalized SRAM timing diagram is shown in Figure 4-1. SRAM is an asynchronous device. Read and
write operations depend on the chips selecting signal nce shown in Figure 4-1. Typical timing
parameters found in an SRAM data book include signal nce to various signal edges, which also shown in
Figure 4-1.
The SRAM signals shown in Figure 4-1 are described in Table 4-1. The signal type entry in the table
describes the signal direction attribute such as input, output or bidirectional. The line bidata is set to high
impedence when it is not used. By doing so, bidata will not cause data line conflict.
Timing parameters for an SRAM are also illustrated in Figure 4-1. They are the minimum times starting
from the signal nce to various signals and they are the same for both read and write operations. Signal
nwrite indicates read or write operation. If the SRAM is in write mode, bidata will be the input for
capturing data. Oppositely, if the SRAM is in read mode, bidata will be used to output data.
26

pepe, 4/30/96
Figure 4-1. SRAM Timing Diagram
Table 4-1. SRAM Signal Description

Signal Name
Signal Type
Signal Description
NCE
IN
Not chip select signal.
NWrite
IN
Not chip write or chip read signal.
Address
IN
Use for read and write address.
Data
IN/OUT
Use for data input and data output.
4.3 VHDL SRAM Memory Design

SRAM modeling can be implemented using the above timing information and previous VHDL lessons. A
VHDL program that models a 2 bit by 4 word SRAM device is listed in Section 4.4. This program was
compiled and simulated using the Cypress Warp tool. The simulation waveform is shown in Figure 4-2.
From the figure, it can be seen that the data value 2 was written into the memory location 3 (bidata(1)=1
and bidata(0)=0) and read back afterwards.
Figure 4-2. SRAM Simulation Waveform

Several VHDL concepts were used in this program. The first is the use of the generic statement. The
Generic statement is similar to constant declaration in programming languages except it is more useful
27

pepe, 4/30/96
due to the hardware design application. Notice that we have discussed a VHDL method for declaring
constants in earlier lessons. Generic statement is different because it specifies where those constant are
valid. In other words, the generic statement defines the scope of its constants. The Generic statement is
to be used inside of an entity statement (see previous lessons), and the architecture body associated with
this entity will be able to see and use these generic constants. It is a very powerful concept because it
allows generalized hardware modeling. In the SRAM example, the size of the SRAM does not need to be
fixed at 8 words deep and 2 bits per word. Different size SRAMs can be created by changing only the
declaration inside of the generic statement.
The second concept is the use of the sub-program procedure statement. Although the component
statement was introduced first, the procedure statement is the preferred construct due to its simplicity.
The component structure is very much the same as the architecture structure. It uses an entity statement
to specify input and output signals and an architecture body to define the relationship(s) of these signals.
Declaration of a component is also needed in the main architecture body. As clear as the component
concept may seem it, there are many hassles in creating and using component for simulation. Unless the
component is useful or has already been created and put into a library, a procedure is the prefer choice.
Another reason why a procedure construct is preferred over a component construct is for ease of
understanding VHDL code. In general, VHDL code for structure modeling is less likely to be
comprehended than the corresponding behavior modeling which is naturally associated with procedure
statement. The fact that behavior modeling is easier to grasp was shown in the previous lesson where a D
flip-flop was represented in different VHDL models.
The third concept is the use of library routines. The use statement can be found between the entity
statement and architecture statement. Since the Cypress VHDL compiler is mainly geared towards
hardware synthesis, the behavior modeling aspect of Cypress VHDL compiler is weak because it doesnt
have good support for modeling a resistor. For example, in the SRAM modeling below, to multiplex bidirectional line, bidata, component tristate buffer, bufoe, from the Cypress library rtlpkg would have to be
used.
4.4 SRAM VHDL Listing

---------------------------------------------------------------------- Entity statement
--- Specifies input, output and bidirectional signals
-- Specifies
--------------------------------------------------------------------entity sram0 is
generic(
address_width : integer := 2; -- column address width
address_size : integer := 4;
word_width : integer := 2; -- word width
zero_word : bit_vector := "00" -- To nullify a memory
);
port(
address
nwrite
nce
bidata
: in bit_vector (address_width-1 downto 0);

: in bit;
: in bit;
: inout x01z_vector (word_width-1 downto 0)
);
end sram0;
use work.rtlpkg.all;
-- bufoe(x, oe, y, yfb): Three state buffer with feedback.
use work.int_math.all; -- See Cypress VHDL synthesis reference page 4-86
-- i2bv(i,w): converts integer i to w width binary
28

pepe, 4/30/96
-signal.
-- bv2i(a): inverse of function i2bv
--------------------------------------------------------------------architecture sram0arch of sram0 is

type memory_elmt is array (address_size downto 0) of
bit_vector (address_width downto 0);
---------------------------------------------------------------------- Read from the memory

--------------------------------------------------------------------procedure ReadMemory
( nce, nwrite: in bit;
address: in bit_vector (address_width-1 downto 0);
memory_word: in memory_elmt;
dataout: out bit_vector (word_width-1 downto 0)) is
variable i, w: integer;
variable addressi: integer;
begin
if (nce='0') and nce'event then
if (nwrite='1') then
addressi := bv2i(address);
for i in 0 to address_size-1 loop
if(i=addressi) then
for w in 0 to word_width-1 loop
dataout(w) <= memory_word(i)(w);
end loop;
end if;
end loop;
end if;
end if;
end ReadMemory;
---------------------------------------------------------------------- Write to the memory
--------------------------------------------------------------------procedure WriteMemory
( nce, NWrite: in bit;
Address: in bit_vector (address_width-1 downto 0);
memory_word: out memory_elmt;
datain: in bit_vector (word_width-1 downto 0)) is
variable i, w: integer;
variable addressi: integer;
begin
if (nce='0') and nce'event then
if (nwrite='0') then
addressi := bv2i(address);
for i in 0 to address_size-1 loop
if(i=addressi) then
for w in 0 to word_width-1 loop
memory_word(i)(w) <= datain(w);
end loop;
end if;
end loop;
end if;
end if;
end WriteMemory;
---------------------------------------------------------------------- Main program

---------------------------------------------------------------------
29

pepe, 4/30/96
signal datain, dataout: bit_vector (word_width-1 downto 0);
signal memory_word: memory_elmt;
signal w: integer;
begin
-- Memory read cycle
process (nce)
begin
ReadMemory(nce, nwrite, address, memory_word, dataout);
end process;
-- Memory write cycle
process (nce)
begin
WriteMemory(nce, nwrite, address, memory_word, datain);
end process;
-- Tristate input and output
bif0: bufoe port map(dataout(0), nwrite, bidata(0), datain(0));
bif1: bufoe port map(dataout(1), nwrite, bidata(1), datain(1));
end sram0arch;
Listing 4-1. SRAM VHDL Code
4.5 Problems
1) The Cypress Warp tool uses the time unit tick. Figure 4-2 shows an example of tick value 68. From
Figure 4-1, use the Warp simulator to determine the minimum times for signals Tnce2datain, Tnce2nwr and
Tnce2addr that produce a valid Tnce2dataout. Also, what is the value of Tnce2dataout?
2) The timing diagram in Figure 4-3 below shows a modified SRAM that contains an extra field for
storing a valid bit. This memory is to be named SRAMvc (SRAM with valid control) memory. To
put data into a memory location, the signal nvalid has to be set to zero prior to signal nces going
during the SRAMvc reading cycle, if the signal nvalid at the requested address location is high,
bidata should output zero. During the SRAMvc writing cycle, data will write into the memory along
with the nvalid signal. If the signal nvalid is high, the maximum data value (all bits set to one)
should write into the corresponding memory location.
Before using SRAMvc, a signal nreset should be applied to invalidate signal nvalid all memory
locations (for example, nvalid<=0). In result, all nvalid bits will be set to zero, and all bidata value
will be set to max. See Figure 4-3 for the timing relationship of nreset and nce.
Construct SRAMvc using VHDL and prove its functional by simulation. The simulator should
produce timing waveforms similar to those shown in Figure 4-3.
From the simulation, find all the timing parameters in Figure 4-3 based on the simulation of your
SRAMvc VHDL model.
30

pepe, 4/30/96
Figure 4-3. SRAMvc Timing Diagram
31

5.Digital Design using Divide and Conquer
pepe, 4/30/96
5. Digital Design using Divide and Conquer

5.1 Introduction
As the demand for digital system complexity increases, one needs to rely on traditional divide and conquer
methodologies for constructing complex digital systems. VHDL contains features that support the divide
and conquer technique for designing complex system design.
Some of the VHDL features that that permit the designer to build sub-systems have already been
introduced in previous lessons: for example, the use of VHDL component block and procedure and
functions statements. In addition, there is a benefit to breaking a large problem into many smaller
problems because the solution for each small problem can be reused and applied to other problems similar
in nature. For example, the design of an automobile fuel mixer controller might require the use of a two
stage 5 bit D flip flop. By coincidence, this same odd device (2-stages, 5-bit D flip flop) is also used
inside traffic light controller. One can imagine that the design for this device should not have to be
implemented twice, provided the design tool (I.e. VHDL) allows it. This is the concept of hardware reuse,
and for a long time it has been employed in the object oriented software world. In this lesson, the VHDL
package and library between which support hardware reuse will be introduced.
5.2 VHDL features

The use of VHDL component and procedure statements in previous lessons gave the VHDL designer a
systematic way of breaking a large and complicated problem into smaller, and simpler problems to solve.
The VHDL package clause feature lets the digital designer pack tested digital parts into a library
(storage). Then, he/she can use the VHDL keyword use or library to retrieve these parts for any design
afterwards. This enables designer to build more complex systems and also encourages the digital design
divide and conquer methodologies.
5.3 ALU Functions

The Arithmetic logic unit (ALU) is the most important part of the computation engine because it directly
affects the system performance. All computational arithmetic or logic operations are done in the ALU.
Therefore, it is clear that ALU output latency will affect the overall input to output delay. Figure 5-1
shows the general block diagram of an ALU.
Operand 2
Operand 1
Function Select
ALU
ALU Output
Figure 5-1. ALU Block Diagram

The types of functions built into an ALU depends on the nature of computation. For example, a number
crunching ALU such as a digital signal processor (DSP) will devote a good portion of ALU resources for
floating point number manipulation such as floating point add, multiply, etc. An intelligent embeded
32

pepe, 4/30/96
controller, on the other hand, would probably have an overall balanced arithmetic and logic functions,
with emphasis on integer manipulation.
There are many ways to construct an ALU. During the 1980s, when Large Scale Integration (LSI)
Integrated Chip (IC) and Medium Scale Integration IC were popular, bit-slice ALU design was the main
technology due to its modularity for LSI and MSI design. The bit-slice design concept was used because it
reduced computations to individual bits as to minimize each bits overlapping functionality. In the
current trend of integrated circuit technology (e.g. VLSI), ALU designers focus less on the ALUs bit
versatility and more towards the its overall speed performance. An ALU design example is shown in
Figure 5-2.
Operand 1
Operand 2
F0
Fn
F1
Function Select
ALU Output
Figure 5-2. An ALU Design

Figure 5-2 is a simplified ALU design, nevertheless, it shows a good overall ALU structure. In the ALU,
function boxes, F0, F1, to Fn are dedicated for specified ALU functions such as add, multiply, shift, etc.
The inputs to each function box are delivered by two buses, operand1 and operand2. Output of each
function box is tapped onto the output bus via tri-state buffers. In addition to the resulting data outputs,
there are also ALU flag outputs such as divide by zero, arithmetic overflow, underflow, etc. Each function
box is controlled by the control input bits. When functions are similar in nature, for example, left circular
and right circular shifts, control bits are used to distinguish these functions.
There are many arithmetic and logic functions that a computer architecture designer might like to put into
an ALU. However, due to design requirements and technology limitations the designer must compromise
all the factors for an optimal area and speed tradeoff.
5.4 ALU Component Implementation Examples

For simple laboratory work, one might like to design a 4-bit, 2s complement ALU which consists a 4-bit
adder for integer arithmetic, left circular shift, right circular shift, and bitwise AND and OR logic
functions.
VHDL supports modular hardware implementation by using the objected oriented software concept of
reusability. In VHDL, code or program pieces such as a procedure, function or component can be stored
into a library for later use. Whenever the parts are needed, they can be linked into the main VHDL
program and reused again. The VHDL feature that supports part reuse is called Package, and it is
described in the ALU implementation example.
5.4.1 Ripple Carry Adder Implementation

A ripple carry adder is the simplest adder and it is easiest to build. To implement a 4-bit adder, one starts
by paralleling a half adder with full adders. Ripple carry adder implementation can be found in many
digital design text books. this laboratory, half adder and full adder components were implemented and
33

pepe, 4/30/96
they have been placed inside of the library, Arith_parts_pkg package. The listing of the Arith_parts_pkg
package is shown in Section 5.4.2.
In Arith_parts_pkg, components half_adder and full_adder were combined to produce a 4-bit ripple carry
adder ripp_adder4. One example of hardware reuse is to test the functionality of ripp_adder4 without reimplementing the same module. This is done by creating a VHDL module, addtest.vhd, which is listed in
Section 5.4.3. In file addtest.vhd, after the entity declaration, the line beginning with the VHDL key word
use tells the VHDL compiler that all modules (procedure, function, and component) inside of the package
iarith_parts_pkg are visible to file addtest.vhd and therefore are available for use.
Only the module ripp_adder4 is used structurally inside of the addtest module. Remaining modules such
as full_adder and half adder are not shown inside of the addtest module; nevertheless, they were
important in building the ripple carry adder ripp_adder4. The test wave form which illustrates the
adders functionality is shown in Figure 5-3. In Figure 5-3, ain is set to 1111 and bin is set to 0011.
The output aout becomes 0010 with the overflow flag set to 1. Notice the glitches on the at aout lines
caused by the asynchronous ain and bin input signals.
Figure 5-3. Ripple Carry Adder Waveform
5.4.2 Package Iarith_parts_pkg Listing

package iarith_parts_pkg is
component half_adder
port(signal Ain, Bin: in bit;
signal Cout, Sout: out bit);
end component;
component full_adder
port(signal Cin, Ain, Bin: in bit;
34

pepe, 4/30/96
signal Cout, Sout:
end component;
out bit);
component ripp_adder4
port(signal Ain, Bin: in bit_vector (3 downto 0);
signal Sout:
out bit_vector (3 downto 0);
signal OverFlow: out bit);
end component;
end iarith_parts_pkg;
----------------------------------------------------------- Half adder component
entity half_adder is
port(signal Ain, Bin: in bit;
end half_adder;
architecture half_adder_arch of half_adder is
begin
process (Ain, Bin)
begin
if Ain='1' and Bin='1' then
Sout <= '0';
Cout <= '1';
else
Cout <= '0';
Sout <= Ain or Bin;
end if;
end process;
end half_adder_arch;
----------------------------------------------------------- Full adder component
entity full_adder is
port(signal Cin, Ain, Bin: in bit;
end full_adder;
architecture full_adder_arch of full_adder is
begin
process (Ain, Bin, Cin)
begin
if Ain='1' and Bin='1' and Cin='1' then
Sout <= '1';
Cout <= '1';
elsif Cin='0' then
if Ain='1' and Bin='1' then
Sout <= '0';
Cout <= '1';
else
Sout <= Ain or Bin;
Cout <= '0';
end if;
else -- Cin='1' and either Ain='1' or Bin='1'
Sout <= not (Ain or Bin);
Cout <= Ain or Bin;
end if;
end process;
end full_adder_arch;
----------------------------------------------------------
35

pepe, 4/30/96
-- 4 bit adder component
entity ripp_adder4 is
port(signal Ain, Bin: in bit_vector (3 downto 0);
signal Sout:
out bit_vector (3 downto 0);
signal OverFlow: out bit);
end ripp_adder4;
architecture ripp_adder4_arch of ripp_adder4 is
signal Cout: bit_vector (3 downto 1);
begin
ha0: half_adder
port map(Ain=>Ain(0), Bin=>Bin(0),
Cout=>Cout(1), Sout=>Sout(0));
fa1: full_adder
port map(Cin=>Cout(1), Ain=>Ain(1), Bin=>Bin(1),
fa2: full_adder
fa3: full_adder
Cout=>OverFLow, Sout=>Sout(3));
end ripp_adder4_arch;
Listing 5-1. Iarith_parts_pkg VHDL Code Listing
5.4.3 Testadd.vhd Listing

---------------------------------------------------- to test ripple adder routine using component
-- "ripp_adder4" of vhdl file iar.vhd.
--------------------------------------------------entity addtest is
generic( data_wd: integer := 3);
port (
Ain: in bit_vector (data_wd downto 0);

Bin: in bit_vector (data_wd downto 0);
Aout:out bit_vector(data_wd downto 0);
OVF: out bit);
end addtest;
use work.iarith_parts_pkg.all;
architecture addtest_arch of addtest is
begin
op4: ripp_adder4 port map(Ain, Bin, Aout, OVF);
end;
Listing 5-2. Testadd.vhd VHDL Code Listing
36

pepe, 4/30/96
5.4.4 Circular Shifter Implementation

The circular shifter unit is more difficult to implement than simple bitwise logical operations. To make
the implementation more challenging, the shifter was implemented in generalized fashion so it could
accommodate various bus widths. Again, by applying the concept of reuse, the implementation of the
circular shifter was put into a library for test. The library that contains the shifter is called
logic_parts_pkg, and the VHDL code is listed in Section 5.4.6. Again, the testing of circular shift logic
can be performed similar to the ripple carry adder test. The trace file for doing two right circular shifts on
ain bits 0010 is 1000 and is shown in the Section 5.4.5. Notice that the output 1000 took 3 ticks to
come out after applying inputs ain and bin. Ain is the bit to be shifted and bin is the shifted value. In the
cshift component of logic_parts_pkg, input bit RL was used to indicate shift direction: RL=0 for left
circular shift and 1 for right circular shift.
5.4.5 Circular Shift Trace Data

NOVA Simulation Printout
File: TEST.psd -
View: PINS and REGS
Printout produced: Thu Jun 13 01:46:57 1996
b b a a a a a a a a
i i i i i i o o o o
n n n n n n u u u u
_ _ _ _ _ _ t t t t
0 1 0 1 2 3 _ _ _ _
| | | | | | 0 1 2 3
| | | | | | | | | |
^^^^^^^^^^^^^^^^^^^^^^^^^^^
0: 0 1 0 0 1 0 L L L L
1: 0 1 0 0 1 0 L L L L
2: 0 1 0 0 1 0 L L L L
3: 0 1 0 0 1 0 H L L L
4: 0 1 0 0 1 0 H L L L
5: 0 1 0 0 1 0 H L L L
6: 0 1 0 0 1 0 H L L L
7: 0 1 0 0 1 0 H L L L
8: 0 1 0 0 1 0 H L L L
9: 0 1 0 0 1 0 H L L L
10: 0 1 0 0 1 0 H L L L
11: 0 1 0 0 1 0 H L L L
12: 0 1 0 0 1 0 H L L L
13: 0 1 0 0 1 0 H L L L
14: 0 1 0 0 1 0 H L L L
Listing 5-3. Circular Shift Trace Listing
5.4.6 Package Logic_parts_pkg Listing

package logic_parts_pkg is
component cshift
generic (data_wd: integer:= 3);
port (Ain: in bit_vector (data_wd downto 0);
Sin: in bit_vector (1 downto 0);
RL : in bit;
Sout: out bit_vector(data_wd downto 0));
end component;
component bAND
port (Ain: in bit_vector (3 downto 0);
Bin: in bit_vector (3 downto 0);
Cout: out bit_vector (3 downto 0));
37

pepe, 4/30/96
end component;
component bOR
port (Ain: in bit_vector (3 downto 0);
end component;
end logic_parts_pkg;
-- Use Cypress library

use work.bv_math.all;
use work.int_math.all;
------------------------------------------------------ Circular right or left shift

--- Input:
-- Ain - Input bits
-- Sin - Number of shift
-- RL - 0: left circular shift
-1: right circular shift
-- Output:
-- Sout - Output bits
----------------------------------------------------entity cshift is
generic(data_wd: integer := 3);
port( Ain: in bit_vector (data_wd downto 0);
Sin: in bit_vector (1 downto 0);
RL : in bit;
Sout: out bit_vector(data_wd downto 0));
end cshift;
----------------------------------------------------architecture cshift_arch of cshift is

signal temp: bit_vector (data_wd downto 0);
begin
process (Ain, Sin, RL)
variable i, j, k, l: integer := 0;
begin
-- i is the shifted value
for i in 0 to data_wd loop
if i=bv2i(Sin) then
for j in 0 to data_wd loop
-- Find right circular shift index k
if RL='1' then
k := j+(data_wd+1)-i;
-- Find left circular shift index k
else
k := j+i;
end if;
38

pepe, 4/30/96
-- Assign Output
if k > data_wd then
k := k-(data_wd+1);
end if;
for l in 0 to data_wd loop
if k=l then
temp(l) <= Ain(j);
end if;
end loop;
end loop;
end if;
end loop;
Sout <= temp;
end process;
end cshift_arch;
------------------------------------------------------ Bitwise AND
---------------------------------------------------entity bAND is
port( Ain: in bit_vector (3 downto 0);
end bAND;
---------------------------------------------------architecture bAND_arch of bAND is
begin
process (Ain, Bin)
begin
Cout(3 downto 0) <= Ain(3 downto 0) and Bin(3 downto 0);
end process;
end bAND_arch;
------------------------------------------------------ Bitwise OR
----------------------------------------------------entity bOR is
port(Ain: in bit_vector (3 downto 0);
end bOR;
---------------------------------------------------architecture bOR_arch of bOR is
begin
process (Ain, Bin)
begin
Cout(3 downto 0) <= Ain(3 downto 0) or Bin(3 downto 0);
end process;
end bOR_arch;
Listing 5-4. Logic_parts_pkg VHDL Code Listing
39

pepe, 4/30/96
5.5 ALU Implementation Example

Using components from Section 5.4 and the block diagram in Figure 5-2, one should be able effortlessly to
construct an ALU. See the ALU VHDL code in Listing 5-5.
use work.logic_parts_pkg.all;
------------------------------------------------------ alu_comp
--- Input:
--- Ain, Bin - Input bits
-- sel_bit - Calculation type selection bit
-0001 - Bitwise AND operation
-0010 - Bitwise OR operation
-0100 - Left circular shift
-0101 - Right circular shift
-1000 - Integer add.
--- Output:
--- flags - Bit 0 for Ain all zero flag, all time
-Bit 1 for Bin all zero flag, all time
-Bit 2 for latched adder overflow flag
----------------------------------------------------entity alu_comp is
generic ( data_wd: integer := 3);
port( Ain, Bin: in bit_vector (data_wd downto 0);
sel_bit: in bit_vector (data_wd downto 0);
clk:
in bit;
Aout: out bit_vector (data_wd downto 0);
flags:
out bit_vector (2 downto 0));
end alu_comp;
----------------------------------------------------architecture alu_body of alu_comp is
------------------------------------------------------ calc_ctrl
--- Calculation type decoder
--- logic_ctrl - * Bit 0 for Bitwise AND tri-buffer
-* Bit 1 for Bitwise OR tri-buffer
-* Bit 2,3 for circular shift controller
-Bit 4 for circular shift tri-buffer
-* Bit 5 for integer add tri-buffer
----------------------------------------------------procedure calc_ctrl
(sel_bit:
in bit_vector (3 downto 0);
ctrl: out bit_vector (5 downto 0)) is
begin
case sel_bit is
when "0001" => -- Bitwise AND operation
ctrl <= "000001";
when "0010" => -- Bitwise OR operation
ctrl <= "000010";
when "0100" => -- Left circular shift operation
40

pepe, 4/30/96
ctrl <= "010100";
when "0101" => -- Right circular shift operation
ctrl <= "011000";
when "1000" => -- Integer add operation
ctrl <= "100000";
when others => -- Invalid operation
ctrl <= "000000";
end case;
end calc_ctrl;
signal Ain_lat, Bin_lat: bit_vector (data_wd downto 0);

signal sel_bit_lat: bit_vector (3 downto 0);
signal ctrl: bit_vector (5 downto 0);
signal bAND_out: bit_vector (data_wd downto 0);
signal bOR_out: bit_vector (data_wd downto 0);
signal cshift_out:bit_vector (data_wd downto 0);
signal ripp_adder4_out: bit_vector (data_wd downto 0);
signal overflow: bit;
begin
-- Synchronize data
process(clk)
begin
if clk='1' and clk'event then
Ain_lat(3 downto 0) <= Ain(3 downto 0);
Bin_lat(3 downto 0) <= Bin(3 downto 0);
sel_bit_lat(3 downto 0) <= sel_bit(3 downto 0);
end if;
end process;
-- Input select line decode

process(sel_bit_lat)
begin
calc_ctrl(sel_bit_lat, ctrl);
end process;
-- Operations assignment
op1: bAND port map(Ain_lat, Bin_lat, bAND_out);
op2: bOR port map(Ain_lat, Bin_lat, bOR_out);
op3: cshift port map(Ain_lat, Bin_lat (1 downto 0), ctrl(3), cshift_out);
op4: ripp_adder4 port map(Ain_lat, Bin_lat, ripp_adder4_out, overflow);
-- Output stage
process(ctrl)
begin
if ctrl(0)='1' then
Aout <= bAND_out;
elsif ctrl(1)='1' then
Aout <= bOR_out;
Aout <= cshift_out;
Aout <= ripp_adder4_out;
else
Aout <= Ain;
end if;
end process;
-- Setting alu flags
41

pepe, 4/30/96
process(Ain)
begin
if
Ain="0000" then
flags(0) <= '1';
else
flags(0) <= '0';
end if;
end process;
process(Bin)
begin
if
Bin="0000" then
flags(1) <= '1';
else
flags(1) <= '0';
end if;
end process;
-- To latch overflow flag

process(ctrl(5))
begin
if ctrl(5)='1' then
flags(2) <= overflow;
end if;
end process;
end alu_body;
Listing 5-5. Example ALU VHDL Code Listing
5.6 Problems
1) Test component cshift by constructing a VHDL module cshtest.vhd. Test both shift directions, left
and right and make sure the test covers the circular wrap effect. The test vectors should be collected
in a trace file after running the cshtest.vhd simulation. In the trace file, write a comment at the side
of the test vectors that corresponding to each distinct operating period.
Specifically, how long is the input to output delay. Does the delay depend on the number of shift
and/or shift direction?
2) Implement a 4-bit arithmetic shift component ashift. Ashift takes a similar input to cshift except it
does an arithmetic shift instead of a circular shift. This component should also be included in the
package logic_parts_pkg.
3) Implement a 4-bit subtract unit that utilizes components already built, such as the adder. Notice that
one can subtract two numbers by complementing the subtractant then and add the numbers. Again,
this unit should be included in the package iarith_parts_pkg.
4) Implement the ALU shown in Figure 4.18 of Computer Organization and Design by Patterson and
Hennessy using the control logic described in Figure 4.19.
What is the worstcase operational propagation delay in this device that you just implemented?
42

6.ALU Datapth Implementation
pepe, 5/23/96
6. ALU Datapth Implementation

6.1 Introduction
After learning ALU modeling in VHDL, one can carry it further by designing an ALU datapath and
eventually building a complete CPU. Again, the divide and conquer approach will be used to illustrate the
example. A CPU datapath consists of several major components: register file, data memory, instruction
memory, and control module. To model the overall CPU datapath activity, it is necessary first to
implement and model each component separately. VHDL is the nature candidate for such a task because
it supports the building of small modules, for example with the. package concept.
6.2 Instruction Set Architecture (ISA)

Computer architecture design requires technology from both software and hardware areas. By
conventional practice, a computer system is built in a hierarchy of hardware to software layers. The
abstract layer where hardware and software interact is called the instruction set architecture (ISA). How
ISA is designed and implemented is important because it affects the machine in every aspect, such as
compatibility, performance, and cost. Usually, depending on the end user application, the machine
architect selects the best possible trade off for its design. By utilizing the same instruction set, one can
ensure compatibility problem among different machines. However, this choice might limit the machine in
performance because it must follow the same architecture path as other machines using the instruction set.
To illustrate ISA directly, think as the machines instruction set. A machines instruction set plays
important role in both software and hardware layers. On the hardware side, it sits on top and directly
controls how the machine manipulates data. In software side, it is used by all software running on the
machine because it is the lowest possible level. Thus, the ISA is obviously the software and hardware
interface level. Moreover, the instruction set is abstract, and it serves as a good starting point for one to
begin his/her computer architecture design.
6.3 A Computer Architecture Implementation

Perhaps MIPS is the most popular RISC machine used in academia for studying computer architecture.
This laboratory will take no exception by borrowing concepts from the the MIPS architecture and
implementing example datapath of a CPU in VHDL. This architecture is to be called BIA (Borrowed and
Incomplete Architecture).
Similar to the MIPS instruction format, the BIA instruction format is constant but shorter in length 16
bits. To steal more MIPS ideas, BIA instructions can be separated into six fields (see Table 6-1). Notice
that the field lengths sum to 16 bits.
Field
Description
Length
Bit Location
OP
Opcode
4 bits
[15:12]
rs
Operand 1
2 bits
[11:10]
rt
Operand 2
2 bits
[9:8]
rd
Distention
2 bits
[7:6]
sh
Shift amount
2 bits
[5:4]
43
Copyright 1996, EE / CERL, V1.00

pepe, 5/23/96
funct
Function
4 bits
[3:0]
Table 6-1. BIA Instruction Field

Shamelessly, BIA also contains three types of instructions: R-type, I-type and J-type, where R instructions
are for register type operations, I instructions are for data transfer operations and J instructions for
unconditional instruction jumps. Since the purpose of BIA is to demonstrate the ability to model a CPU
datapath in VHDL, the BIA architectural functionality need not to be complete. Only a few instructions
are sufficient for modeling purposes, and they are shown in Table 6-2.
Instructions
Description
Format
OP code
Function
lw
load word
14
sw
store word
15
add
add
sub
subtract
or
logical or
sll
logical left shift
srl
logical right shift
andi
and immediate
12
ori
or immediate
13
Table 6-2. BIA Instructions
6.4 Implementation Example

The BIA CPU block diagram shown in Figure 6-1 contains four major components: register file, ALU,
memory block, and control module. The BIA ALU implementation is simple and straight forward. Most
of the ALU functions, (for example, add, circular shift, logic operations, etc.), have already been presented
in the previous VHDL exercise so there is no need to re-implementing them. Since the main focus of this
laboratory is to learn the design of a modern microprocessor CPU, supporting logic that surrounds the
ALU, (such as datapaths and memory element entities), will be the main focus of this lesson. Again, these
items are shown in Figure 6-1.
44

pepe, 5/23/96
Figure 6-1. BIA Block Diagram

All BIA instructions shown in Table 6-2 take one clock cycle to execute, with the exception of the lw (load
word) instruction which needs two clock cycles. The clocked elements are the register file and memory in
the BIA datapath. The lw operation is triggered by setting the instruction signal aluop to 14. Next, the
load memory effective address is calculated. The calculation is done by adding the register content to the
address offset which is stored in the immediate field of lw instruction. Next, lw reads the memory by
supplying it with the effective address. Finally, the content of the memory is written back into a particular
register which is specified in the instruction word. The timing diagram of the lw instruction along with
other control signals is in drawn in Figure 6-2. In the figure, the BIA execution path coarsely resembles
to a typical CPU instruction cycle: Decode (TopcodeDelay), Fetch (TmemoutDelay), and Execute (TnregwriteHold).
Figure 6-2. lw Instruction Timing Diagram

Due to the complexity of the Cypress VHDL tool, only a fraction of an ALU block is implemented in BIA,
and it is shown in the VHDL listing. Important ALU functions such as add and circular shift are not
45

pepe, 5/23/96
included because the current CPLD fitter only supports up to 128 micro-cell devices. The BIA VHDL
implementation which is shown in a later section includes everything shown in Figure 6-1 with a basic
ALU block that mimic ALU functions.
6.4.1 4x4 Register File and 8x4 Memory Block

A Register file is not much different from a memory block as seen a few lessons earlier. A register file is
made by an array of registers, and each register bit is built from a D flip-flop. A Register file is rather
similar to a memory block in that they both are data storage devices. The difference between a register
file and a memory block is that a register file contains two read ports instead of one. Moreover, a register
file is a faster device because it is placed physically near the ALU. Memory, on the other hand, is located
farther from the ALU or even sometimes external to the CPU chip. No less important than the ALU block,
register file performance is also critical to how a carfully designed CPU performs.
The logic block diagram of the BIA register file is shown in Figure 6-3. Since BIA is a much reduced
MIPS CPU, only four 4-bit wide registers are implemented. The overall BIA register file is made up of
multiplexers, decoders, and D flip-flops. Two addresses, REG1a and REG2a, are used to select any two of
the registers for reading. Data are output on the corresponding REG1 and REG2 ports. When writing to
the register file, data to be written is input from REGin, and the writing register address from REGina is
decoded to enable the corresponding register. When reading and writing from the same register, the
reading port gets the value prior to the writing of the new value. In other words, the value returned to the
reading port will be the value written in an earlier clock cycle. The VHDL source code can be found in a
later sections.
Both the register file and memory block in BIA are clocked memory elements. To write or read from
these devices, control signals such as the read/write signal, address, and write data lines need to be
synchronized with a rising clock edge. This synchronization can be seen in the timing diagram Figure 62.
Figure 6-3. 4x4 Register File Block Diagram
6.4.2 Control Module

When given instructions, BIA executes according to the state diagram shown in Figure 6-4 below. Notice
that BIA instructions are divided into four types: lw (load word), sw (store word), ori (or immediate), and
R (or register operations). Each instruction type has distinct execution flow and consequential control
signal values. The VHDL file ctrlm.vhd which is shown in 6.4.4.3 contains a VHDL entity, st-mach that
46

pepe, 5/23/96
implements functions dictating the BIA state flow. Entity ctrl_module uses the state value produced in
st_mach to generate control signals that go to every module of the BIA CPU.
6.4.3 Execution Example

An example session of running BIA is shown in Figure 6-5. To see how the BIA datapath handles
instruction and data traffic, one also needs to be familiar with the BIA block diagram of Figure 6-1. The
session was captured by executing five BIA instructions. A BIA instruction is first fed into the input port
inst. Next, the instructions is latched into registers ir. Notice that ir is associated with the state machine
that holds instruction (opcode+operand) for some number of clock cycles. For example, the R-type, sw,
and ori instructions take two cycles to complete and the lw instruction takes three. This instruction latch
state machine is implemented in entity instruction_latch and it can be found in the listing for bia.vhd.
After an instruction has been latched into the instruction register ir, each instruction field, such as opcode
and operands, is sent to the corresponding function block. For example, in the typical R-type instruction,
after the opcode is identified to be an R-type instruction, operands are sent to the register file (fast twoport read memory) for retrieving register values and getting them to the next processing stage, namely the
ALU. If the instruction in execution is simple R-type instruction, the result from the ALU stores directly
back into the register file, and it takes two cycles (cycle 07 and 08 in Figure 6-5). If the instruction is a
memory transfer instruction, depending on load word or store word, it can take either two or three clock
cycles. Remember that both the register file and memory are clocked elements, therefore, to store a
register word into memory, the first clock cycle is needed to release the data from the register and the
second clock cycle to store the data into the memory. Similarly in a load word operation, the first clock
cycle is to read data from the register (address), the second clock cycle to retrieve data, and the third clock
cycle to store back to the register. Again, all of these processes are executed when running the BIA code
in Table 6-3 below.
47

pepe, 5/23/96
Instruction
Comment
Opcode rs rt
rd sh function
[15:12]
[11:10] [9:8]
[7:6]
[5:4]
[3:0]
ori
r3, r0, #06
; $3 <- $0 OR 6
1101
00 00 11 00 0110
ori
r2, r0, #04
; $2 <- $0 OR 4
1101
00 00 10 00 0001
or
r1, r2, r3
; $1 <- $2 OR $3
0000
10 11 01 00 0010
sw
r1, #4(r0)
; Memory[$0+4] = $1
1111
00 00 01 00 0001
lw
r3, #4(r0)
; $3 = Memory[$0+4]
1110
00 00 11 00 0001
Figure 6-4. BIA State Diagram

nop
; no op
0001
00 00 00 00 0000
Table 6-3. BIA Code

The simulation run of the operations in Figure 6-5 contains two fields, inst and ir. Inst is the input
instruction corresponding to the code above. Ir is the instruction latched corresponding to the inst value.
For example, at clock cycle 2, instruction 0xd0c6 is input from inst and then latched by the ir register for
48

pepe, 5/23/96
two cycles. The reason that instructions are latched is because control signals at every stage or clock cycle
used by various modules are generated from the instruction field. To execute each BIA instruction
correctly, the instruction needs to be frozen during that instruction cycle. For example, instructions such
as ori, or, and lw need to be constant for two clock cycles, and the instruction lw needs three clock cycles.
Again the timing diagram in Figure 6-5 shows ir was assigned value 0xd0c6 for two cycles (cycle 3 and
4). Notice that this operation is different from the pipeline concept to be learned later.
The signals regf4x4_regd1, regf4x4_regd2, and regf4x4_regd3 are a register file that contains registers 1,
2 and 3. Notice that register 0 always contains a zero value, and therefore it was not shown in the figure.
Memory contents are stored in the signals mem2x4_regd0 and mem2x4_regd1. A memory word is 4 bits
wide, However, there are only two words implemented in BIA because of Cypress the VHDL compiler
limitation (BIA with more memory words cannot be fitted and simulated properly).
Although a more complete ALU can be used, the ALU implemenetd here contains only a few simple
operations such as the or, ori, and, and andi instructions. Due to the limited amount of logic that one can
fit into a Cypress CPLD family part, only these representative operations are used and shown for purposes
of this example. Similar is of the case in the memory word depth. The BIA module originally contain a 4
bit wide and 8 word deep (4x8) module. But, the fitter cant get that BIA logic into a 37x CPLD.
Therefore, the memory block size was reduced to 4x2 to make the logic fit.
Another comment about simulating these six BIA instructions in Table 6-3 is that the Cypress VHDL
simulator, NOVA, takes about three minutes to do the simulation on a Petium-90 processor with 16
Megabytes of RAM in the Windows 95 platform. NOVA is not a graphically incremental simulator. The
user does not see timing waveform change incrementally. The final wave shows only at the end of the
simulation. In a typical simulation session, one would first start the simulation and let it run. While the
machine is simulating, output timing waveforms are frozen for approximately three minutes, and then at
last, the waveforms are updated all at once when the simulation completes.
49

pepe, 5/23/96
Figure 6-5. BIA Simulation Timing
6.4.4 VHDL Listing

Sections below list VHDL code for major components and modules that make BIA architecture.
6.4.4.1 regf4x4.vhd
Below is the VHDL listing for the 4x4 register file block.
package regf_type is
constant bw: integer := 3;
constant adw: integer := 1;
component regf4x4
port(
clk :
nreset: in bit;
nrw: in bit;
in
50
bit;

pepe, 5/23/96
regin:
in
bit_vector(bw downto 0);
reg1, reg2:
out bit_vector(bw downto 0);
reg1a, reg2a, regina:
in
bit_vector(adw downto 0));
end component;
end regf_type;
use work.regf_type.all;
entity regf4x4 is
port(
clk : in bit;
nreset: in bit;
nrw: in bit;
regin: in bit_vector(bw downto 0);
reg1, reg2:
out bit_vector(bw downto 0);
reg1a, reg2a, regina:
in
bit_vector(adw downto 0));
end regf4x4;
architecture regf4x4_arch of regf4x4 is

signal regd0, regd1, regd2, regd3: bit_vector (bw downto 0);
begin
process (clk)
variable i: integer;
begin
if clk'event and clk='1' then
if nreset='0' then
regd0 <= "0000";
regd1 <= "0000";
regd2 <= "0000";
regd3 <= "0000";
else
-- writing to register
if nrw='0' then
if regina="00" then
regd0<=regin;
elsif regina="01" then
regd1<=regin;
elsif regina="10" then
regd2<=regin;
else
regd3<=regin;
end if;
end if;
-- reading from register
-- register 1 read
if reg1a="00" then
-reg1<=regd0;
reg1<="0000";
elsif reg1a="01" then
reg1<=regd1;
reg1<=regd2;
else
reg1<=regd3;
end if;
51

pepe, 5/23/96
-- register 2 read
if reg2a="00" then
-reg2<=regd0;
reg2<="0000";
reg2<=regd1;
reg2<=regd2;
else
reg2<=regd3;
end if;
end if;
end if;
end process;
end;
Listing 6-1. Regf4x4.vhd VHDL Code
6.4.4.2 mem4x2.vhd
Below is the construction of both 2x4 and 8x4 memory blocks in VHDL. Although the 2x4 memory block
is the one used in BIA.vhd, the 8x4 memory block illustrates a more generic method of constructing
memory of any word depth.
package memory_type is
constant bw: integer := 3;
constant adw: integer := 1;
constant total_word: integer := 3;
type one_mem_data is array (bw downto 0)of bit;
type mem_data is array (total_word downto 0) of one_mem_data;
component mem2x4
port(
clk: in bit;
mdatain: in bit_vector(3 downto 0);
mdataout: out bit_vector(3 downto 0);
address: in bit;
nw:
in bit);
end component;
component mem8x4
port(
clk: in bit;
mdatain: in one_mem_data;
mdataout: out one_mem_data;
address: in bit_vector (adw downto 0);
nw:
in bit);
end component;
end memory_type;
use work.int_math.all;
use work.memory_type.all;
52

pepe, 5/23/96
entity mem8x4 is
port(
clk: in bit;
mdatain: in one_mem_data;
mdataout: out one_mem_data;
address: in bit_vector (adw downto 0);
nw:
in bit);
end mem8x4;
architecture mem8x4_arch of mem8x4 is
signal mword: mem_data;
signal madd: bit_vector (adw downto 0);
begin
-- clocked memory!
process (clk)
variable j: integer;
begin
-- write process
if nw='0' then
for j in 0 to total_word loop
if address=i2bv(j,2) then
mword(j) <= mdatain;
end if;
end loop;
else
for j in 0 to total_word loop
if address=i2bv(j,2) then
mdataout <= mword(j);
end if;
end loop;
end if;
end if;
end process;
end mem8x4_arch;
-----------------------------------------------------------------------------use work.memory_type.all;
entity mem2x4 is
port(
clk: in bit;
mdatain: in bit_vector(3 downto 0);
mdataout: out bit_vector(3 downto 0);
address: in bit;
nw:
in bit);
end mem2x4;
architecture mem2x4_arch of mem2x4 is
signal word0, word1: bit_vector(3 downto 0);

begin
process (clk)
begin
if address='0' then
if nw='0' then
word0 <= mdatain;
else
mdataout <= word0;
53

pepe, 5/23/96
end if;
else
if nw='0' then
word1 <= mdatain;
else
mdataout <= word1;
end if;
end if;
end if;
end process;
end mem2x4_arch;
Listing 6-2. Mem4x2.vhd VHDL Code
6.4.4.3 ctrlm.vhd
The VHDL program ctrl_pkg below the VHDL code section main contains BIA state control logic. It
dictates BIA state given independent input such as instructions. At each BIA state, ctrlm.vhd also
generates control signals to control other logic blocks such as ALU, memory, and the register block.
These processes comprise in the component st_mach. Notice that st_mach execute according to the flow
diagram in Figure 6-4.
One more important piece of code is the alu_module block. alu_module is very similar to the one
described in the previous lesson except it contains operations custom to BIA architecture. Also,
operations such as addition and circular shift are not included (commented out) due the CPLD capacity.
The last remaining components are simple multiplexers of various bit widths.
package ctrl_pkg is
component alu_module
port (
ain, bin: in bit_vector (3 downto 0);
sel_bit: in bit_vector (3 downto 0);
shamt: in bit_vector (1 downto 0);
cout: out bit_vector (3 downto 0);
flags: out bit_vector (2 downto 0));
end component;
component mux1_2x1
port (
ain, bin: in bit;
cout: out bit;
sel: in bit);
end component;
component mux4_2x1
port (
sel: in bit);
end component;
component mux2_2x1
port (
sel: in bit);
end component;
component st_mach
54

pepe, 5/23/96
port( state: inout bit_vector (2 downto 0);
aluop: in bit_vector (3 downto 0);
clk: in bit;
nreset: in bit;
ir3_0: in bit_vector (3 downto 0);
aluselb: out bit;
alufunc: out bit_vector (3 downto 0);
mem2reg: out bit;
regdst: out bit;
nregwr: out bit;
nmemwr: out bit;
signmux: out bit);
end component;
end ctrl_pkg;
---------------------------------------------------------------use work.logic_parts_pkg.all;
entity alu_module is
port (
sel_bit: in bit_vector (3 downto 0);
shamt: in bit_vector (1 downto 0);
flags: out bit_vector (2 downto 0));
end alu_module;
architecture alu_module_arch of alu_module is
signal bAND_out, bOR_out: bit_vector (3 downto 0);
signal lcshift_out, rcshift_out: bit_vector (3 downto 0);
signal ripp_adder4_out: bit_vector (3 downto 0);
signal apass_out, bpass_out: bit_vector (3 downto 0);
signal overflow, b_zero, b_one: bit;
begin
op1: bAND port map(ain, bin, bAND_out);
op2: bOR port map(ain, bin, bOR_out);
b_zero <= '0';
b_one <= '1';
process (shamt)
begin
if shamt="00" then
lcshift_out <= "0000";
rcshift_out <= "0000";
end if;
end process;
-- use reg2 pass for memory transfer

ripp_adder4_out <= bin;
-- op3: cshift port map(ain, shamt, b_zero, lcshift_out);
-- op4: cshift port map(ain, shamt, b_one, rcshift_out);
-- op5: ripp_adder4 port map(ain, bin, ripp_adder4_out, overflow);
op6: apass_out <= ain;
op7: bpass_out <= bin;
process(sel_bit)
begin
if sel_bit="0001" then
cout <= bAND_out;
elsif sel_bit="0010" then
cout <= bOR_out;
55

pepe, 5/23/96
cout <= lcshift_out;
cout <= rcshift_out;
cout <= ripp_adder4_out;
cout <= apass_out;
cout <= bpass_out;
else
cout <= "1111";
end if;
end process;
-- flags(0) <= overflow;
flags(0) <= '0';
flags(1) <= '0';
flags(2) <= '0';
end alu_module_arch;
---------------------------------------------------------------entity mux1_2x1 is
port(
ain, bin: in bit;
cout: out bit;
sel: in bit);
end mux1_2x1;
architecture mux1_2x1_arch of mux1_2x1 is
begin
process (sel)
begin
if sel='0' then
cout <= ain;
else
cout <= bin;
end if;
end process;
end mux1_2x1_arch;
---------------------------------------------------------------entity mux4_2x1 is
port(
sel: in bit);
end mux4_2x1;
begin
process (sel)
begin
if sel='0' then
cout <= ain;
else
cout <= bin;
end if;
end process;
end mux4_2x1_arch;
56

pepe, 5/23/96
entity mux2_2x1 is
port(
sel: in bit);
end mux2_2x1;
begin
process (sel)
begin
if sel='0' then
cout <= ain;
else
cout <= bin;
end if;
end process;
end mux2_2x1_arch;
----------------------------------------------------------------- state machine

---------------------------------------------------------------entity st_mach is
port( state: inout bit_vector (2 downto 0);
clk: in bit;
nreset: in bit;
aluselb: out bit;
alufunc: out bit_vector (3 downto 0);
mem2reg: out bit;
regdst: out bit;
nregwr: out bit;
nmemwr: out bit;
signmux: out bit);
end st_mach;
architecture st_mach_arch of st_mach is
begin
process(clk)
begin
if nreset='0' then
state <= "000";
alufunc <= "0101";
regdst <= '1';
nregwr <= '1';
nmemwr <= '1';
signmux <= '1';
aluselb <= '1';
mem2reg <= '1';
regdst <= '1';
else
-- lw instruction
if state="000" and aluop="1110" then
state <= "001";
elsif state="001" and aluop="1110" then
state <= "010";
mem2reg <='0';
regdst <='0';
nregwr <= '0';
57

pepe, 5/23/96
-- sw instruction
state <= "011";
nmemwr <= '0';
-- ORI instruct
state <= "110";
alufunc <= "0010";
nregwr <= '0';
signmux <= '0';
-- R-type instruction
state <= "100";
alufunc <= ir3_0 (3 downto 0);
aluselb <= '0';
nregwr <= '0';
-- back to starting state
else
state <= "000";
alufunc <= "0101";
regdst <= '1';
nregwr <= '1';
nmemwr <= '1';
signmux <= '1';
aluselb <= '1';
mem2reg <= '1';
regdst <= '1';
end if;
end if;
end if;
end process;
end st_mach_arch;
----------------------------------------------------------------- test program
---------------------------------------------------------------entity cmtest is
port (
whatop: inout bit_vector (3 downto 0);
state: inout bit_vector (2 downto 0);
alufunc: inout bit_vector (3 downto 0);
aluselb: out bit;
mem2reg: out bit;
regdst: out bit;
nregwr: out bit;
signmux: out bit;
nmemwr: out bit;
clk: in bit;
dummy: out bit;
nreset: in bit );
end cmtest;
architecture cmtest_arch of cmtest is
begin
state1: st_mach port map (state, aluop, clk, nreset, ir3_0, aluselb, alufunc, mem2reg,
regdst,nregwr, nmemwr,
signmux);
end cmtest_arch;
58

pepe, 5/23/96
Listing 6-3. Ctrlm.h VHDL Code
6.4.4.4 BIA.vhd
The VHDL code below describes the top-most BIA structure. BIA.vhd connects all major component
blocks with extra glue logic to make these blocks compatible. For example, an instruction is fed through
the instruction register ir. ir input is used by a few components such as alu_module and st_mach listed in
previous section. However, ir is one clock cycle behind the actual input inst shown in Figure 6-5 due to
latching. When latching inst, ir is required to hold its value for either two or three clock cycles depending
on the instruction. Therefore, the state machine ir_process inside of bia.vhd is used for this purpose.
Moreover, ir_process also delays the reset signal because of the delayed or latched instruction input into
ir.
See problems section for BIA.vhd VHDL code listing.
6.5 Problems
1) The program BIA.vhd is not given in section 6.4.4.4. However, the entity and the architecture header
statements are given. Please use all the information from this lesson to complete the program
BIA.vhd. The BIA.vhd JEDEC file is given to demonstrate the simulation.
use work.regf_type.all;
use work.ctrl_pkg.all;
use work.memory_type.all;
entity bia is
port (
nrw, nmw: inout bit;
aluselb, signmux: inout bit;
nreset_delay: inout bit;
clk, nreset: in bit;
inst: in bit_vector(15 downto 0);
regb2out, rega1out, aluselbout: inout bit_vector(3 downto 0);
address: inout bit_vector (3 downto 0);
mem2regout: inout bit_vector (3 downto 0);
state: inout bit_vector (2 downto 0));
end bia;
architecture bia_arch of bia is
signal pass: bit_vector (1 downto 0);

signal ir: bit_vector (15 downto 0);
signal regdstout: bit_vector (1 downto 0);
signal mem_out: bit_vector ( 3 downto 0);
signal flags: bit_vector (2 downto 0);
signal temp: bit_vector (3 downto 0);
signal regdst, mem2reg: bit;
signal alufunc: bit_vector (3 downto 0);
begin
<your vhdl code goes here>
end bia_arch;
Simulate your compiled program in the NOVA simulator and produce a timing waveform similar to
the one in Figure 6-5.
59
7. APPENDICES
The sections below describe miscellaneous procedures for beginners to use the Cypress VHDL WARP2
version 4.0 development system. The WARP2 program supports both Window 3.1X and Window 95.
7.1 Install Warp on PC

Cypress VHDL WARP2 version 4.0 requires the use of a CD ROM drive to install. Since the program
itself is big, the WARP2 program requires around 70 Mbytes of hard disk space on the PC. The following
steps are needed to setup the WARP program for a PC.
1) Go to the CD ROM drive and under the directory \pc, run setup.
2) Answer all questions the setup procedure asks, (for example, name, serial number, etc.).
3) Install everything by clicking the continue, next, and yes buttons as appropriate, (e.g., installing
WIN32S system [for Window 3.1X], WARP documentation, etc.).
4) After you are done installing WARP documents and programs, install the acrobat reader so you can
retrieve the on-line WARP user guide and reference manual. To install the acrobat reader, go to the
directory \pc\acroread\disk1 and run setup.
5) At last, you need to copy files uguide.pdf (WARP user guide) and refmanl.pdf (WARP reference
manual) from \doc directory in the CD to a directory on your computer so the Acrobat reader can use
them.
6) Now, the WARP2 system should be ready to go.
7.2 Introduction to WARP4 VHDL Development System

1) Cypress VHDL WARP2v4.0 comes with an integrated project manager, Galaxy, in which the VHDL
code developer can edit, compile, debug, simulate, and test the VHDL program under the same
environment. This section will give you a brief introduction on how to use the VHDL project
manager.
Figure 7-1. Project Window

2) Go into the Windows environment WARP program group, and select the Galaxy program. This
should enable the Galaxy program to pop up an integrated project development window. Click on the

pepe, 5/23/96
OK button if this were the first time WARP got start. After that, a project window will appear on the
screen. See Figure 7-1.
3) Click on the top Galaxy menu bar item Project, then select New command, and another window will
pop on the screen to prompt for the project name. Type prj_name for the project name. After that,
another interactive project development window named prj_name will pop on the screen.
4) Now click on the New button inside of the Edit box on right hand side of the prj_name project
window to start the editor.
5) Enter the VHDL code such as the one from lesson one, Listing 1-1.
6) After typing in the VHDL program, click on the top File button and select the Save As command to
save the file into fil_name.vhd. Quit the editor after the file has been saved. This should return the
program control back to the project manager window prj_name.
7) To include the VHDL program in the project manager window prj_name for compiling, click on the
Files command on the top menu bar and select the Add button to enable the VHDL program add
window. In the add window, first select the VHDL file by clicking on it (e.g. fil_name.vhd) in the left
box, and then click on the button to move the VHDL file to the right box. Select the OK button to
go back to the project window afterwards.
Figure 7-2. Device Window

8) To open a device window, click on the Device button inside of the Synthesis options box which is
located at the bottom right corner of the project window. In the device window, see Figure 7-2, select
C375 on the top left corner under the device category. Click on the OK button to go back to the
project window again.
9) In the project window, click on the Set top button to tell the project manager that the VHDL program
fil_name just entered is the main program for simulation.
10) Now, the program fil_name is ready to compile and simulate. To compile the VHDL program, first
select the file fil_name.vhd, then click the Smart button inside of the compile box in project window.
Another window should pop up and display the compilation status.
61

pepe, 5/23/96
11) If there were VHDL syntax errors, one can easily go back to the VHDL program by selecting the
Error buttons that got the magnifying glass icon on the top of the compilation window menu bar.
12) After the compilation process is completed, quit the compilation window by clicking on the Close
button on the top of compilation window menu bar.
13) To start the NOVA simulator, click on the top Tools button and select the NOVA command to get to
the NOVA window. In the NOVA window, click on the Files button and select Open to specify the
simulation object file with the file extension *.jed. In this example, fil_name.jed is used.
14) In the NOVA simulator, one can examine the circuits functionality by studying the input and output
relationship. To simulate the circuit, one needs to set the input waveform(s) first, then, click on the
top menu bar Simulate item and select the execute command to simulate. Hopefully, in a short
period, NOVA will display the simulated waveform in red.
62

Alford C.digital Design VHDL Laboratory Notes.1996

Uploaded by

Document Informationclick to expand document informationAlford C.digital Design VHDL Laboratory Notes.1996

Document Informationclick to expand document information

Copyright:

Available Formats

Alford C.digital Design VHDL Laboratory Notes.1996

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Alford C.digital Design VHDL Laboratory Notes.1996

Uploaded by

Copyright:

Available Formats

DIGITAL DESIGN VHDL LABORATORY NOTES

Digital Design VHDL Laboratory

2. VHDL BEHAVIOR AND STRUCTURE MODELS

3. STATE MACHINES AND PROGRAMMABLE LOGIC DEVICES

4. DIGITAL DEVICE MODELING

5. DIGITAL DESIGN USING DIVIDE AND CONQUER

Digital Design VHDL Laboratory

5.4.3 TESTADD.VHD LISTING

7.1 INSTALL WARP ON PC

Digital Design VHDL Laboratory

AND-OR Gates _____________________________________________________________ 7

Digital Design VHDL Laboratory

LIST OF VHDL CODE

Simple VHDL Code for Dataflow Modeling ______________________________________ 7

Digital Design VHDL Laboratory

1. VHDL - Hardware Design Software Approach

1.1.1 Sequential Programming

1.1.2 Parallel Programming

Digital Design VHDL Laboratory

Listing 1-1. Simple VHDL Code for Dataflow Modeling

Digital Design VHDL Laboratory

Figure 1-2. AND-OR Gate Propagation Output

1.1.3 Object Based Programming

Digital Design VHDL Laboratory

1.1.4 Programmable Circuits

Ability to function even in abnormal conditions.

Ability to ease change in specification change.

Ability to be reused, in whole or in part, for new applications.

Ability to combine with others.

Listing 1-2. Digital Design Goodness Criteria

Digital Design VHDL Laboratory

PROM's Programming Points

Figure 1-3. PLD Circuit Structure

Figure 1-4. LED Control Circuit

Digital Design VHDL Laboratory

Listing 1-3. Example VHDL Code

Digital Design VHDL Laboratory

Comment Blank in1 in2 in3 in4 Overflow a

Table 1-1. LED-Control Truth Table

Digital Design VHDL Laboratory

2. VHDL Behavior and Structure models

Example of Discrete Components

Figure 2-1. Example of Discrete Components

Copyright 1996, CERL / EE, V1.00

Digital Design VHDL Laboratory

Figure 2-2. RS Flip-flop

Figure 2-3. D Latch

Figure 2-4. D Latch with Reset Control

Copyright 1996, CERL / EE, V1.00

Digital Design VHDL Laboratory

architecture behav of inv1 is

architecture behav of and2 is

Copyright 1996, CERL / EE, V1.00

Digital Design VHDL Laboratory

architecture behav of or2 is

Listing 2-1. Dataflow Modeling of AND2, OR2 and INV1 Gates

Listing 2-2. D Latch Structure Modeling

Copyright 1996, CERL / EE, V1.00