2/21/19

Introduction to Real-Time Digital Signal Processing 19

N N
infrastructures and wireless base stations, the performance, ease of development, and

Ê Ê
multiprocessor configurations are paramount.

K I Example 1.6

K I
. .
A number of DSP applications along with the relative importance for performance, price,

T T
and power consumption are listed in Table 1.2. This table shows, for handheld devices, that
XỬ LÝ TÍN HIỆU) SỐ Chương 1 – Giới thiệu chung về xử lý
the primary concern is power efficiency; however, the main criterion for the communica-

n)
G n
G .e d u
tion infrastructures is performance.
THỜI GIAN v tín hiệu số thời gian thực
v
u. THỰC .
N N
When processing speed is at a premium, the only valid comparison between processors is on an

e d algorithm implementation basis. Optimum code must be written for all candidates and then the

it .
Ơ p Ơ p t it
Bài giảng 03 – Giới thiệu chung về Lịchtime
execution sửmust phát triển của
be compared. DSP factors are memory usage and on-chip
Other important
t
Ư RƯien
peripheral devices, such as on-chip converters and I/O interfaces. In addition, a full set of
Xử lý tín hiệu số thời gian thực (tiếp theo) Các đặc điểm chính
tt@ @
development tools and support listed as follows are important for digital signal processor selection:
t
T R ie n
(k
Giảng viên: Trương Trung Kiên
Đơn vị: Bộ môn Xử lý tín hiệu và Truyền thông
Quy trình thiết kế, tnghiên

T
cứu và phát triển
1. Software development tools such as C compilers, assemblers, linkers, debuggers, and
simulators.
(k
© ©
Khoa Kỹ thuật Điện tử 1 2. Commercially available DSP boards for software development and testing before the target
DSP hardware is available.
Email: kientt@ptit.edu.vn, votuyen@gmail.com 3. Hardware testing tools such as in-circuit emulators and logic analyzers.
4. Development assistance such as application notes, DSP function libraries, application
libraries, data books, low-cost prototyping, and so on.
Hà Nội 02/2019 (c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 2

1.4.3 Software Development

There are four common measures of good DSP software: reliability, maintainability,
extensibility, and efficiency. A reliable program is one that seldom (or never) fails. Since
most programs will occasionally fail, a maintainable program is one that is easy to correct. A
truly maintainable program is one that can be fixed by someone other than the original
Introduction to Real-Time Digital Signal Processing 17 programmers. An extensible program is one that can be easily modified when the require-
Quy trình thiết kế hệ thống XLTHS TTT Lựa chọn xây dựng bộ tiêu chí kỹ thuật
ments change. A good DSP program often contains many small functions with only one

N N
Application purpose, which can be easily reused by other programs for different purposes.

Ê Ê
Phát triển thuật

I I
System specifications toán
Table 1.2 Some DSP applications with the relative importance rating (adapted from [19])

.
Algorithm development and simulation

K Application Performance

. K Price Power
consumption
Select DSP processor

T .v n ) Audio receiver
DSP hearing aid
1
2
2
3
T .v n ) 3
1

G .e d u G .e d u
S H
O
Software Hardware
A
Phát triển hệ MP3 player 3 1 2
F R

N N
T architecture schematic D thống: phần cứng Portable video recorder 2 1 3
W W và phần mềm Desktop computer 1 2 3

Ơ p t it Ơ p t it
A A
R Coding and Hardware R
Notebook computer 3 2 1
E debugging prototype E Cell phone handset 3 1 2

Ư Ư
Cellular base station 1 2 3

tt@ tt@
Rating: 1 to 3 with 1 being the most important

R ie R ie
System integration and debug
Tích hợp, thử

n n
T T
nghiệm và đo kiểm
System testing and release
Cần đánh giá tầm quan trọng của mỗi tiêu chí kỹ thuật
Source: Kuo et al, “Real-time digital
(k (k
© ©
signal processing,” Wiley, 2013 Figure 1.9 Simplified DSP system design flow
Câu hỏi: Hãy giải thích sự so sánh tầm quan trọng tương đối của
Quy trình thiết kế, nghiên cứu và phát triển tổng quát
At the algorithm development stage, it is easier to use high-level language tools (such as ba nhóm hiệu năng trên đối với từng ứng dụng?
ª ĐãMATLAB
được 1đơn giản hoá
or C/Cþþ) để algorithmic-level
for the minh hoạ ý tưởng và các
simulations. bước
A DSP thựccan
algorithm hiện
be chính
simulated using a general-purpose computer to test and analyze its performance. A block
ª Trong
diagramthực tế, mộtdevelopment
of software số bước phải using lặp đi lặp lại nhiều
a general-purpose lần is illustrated in
computer Source: Kuo et al, “Real-time digital signal processing,” Wiley, 2013

Figure 1.10. The testing signals

(c) 2019-nay, may
Trương beKiên
Trung internally generated
(kientt@ptit.edu.vn, by signal generators, digitized
votuyen@gmail.com) 3 (c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 4
from the experimental setup or real environment based on the given application, or received
from other computers via the networks. The simulation program uses the signal samples
stored in data file(s) as input(s) to produce output signal(s) that will be saved as data file(s) for
further analysis.

1
DSP
algorithms

MATLAB® or C/C++
ADC DAC
 System testing and release

Figure 1.9 Simplified DSP system design flow

At the algorithm development stage, it is easier to use high-level language tools (such as
2/21/19
MATLAB1 or C/Cþþ) for the algorithmic-level simulations. A DSP algorithm can be
simulated using a general-purpose computer to test and analyze its performance. A block
diagram of software development using a general-purpose computer is illustrated in
Figure 1.10. The testing signals may be internally generated by signal generators, digitized
from the experimental setup or real environment based on the given application, or received
from other
Introduction computers
to Real-Time via the
Digital networks.
Signal The simulation program uses the signal samples 11
Processing
stored in data file(s) as input(s) to produce output signal(s) that will be saved as data file(s) for
Phát triển thuật toán XLTHS TTT
further analysis. Lựa chọn phát triển phần mềm
Introduction to Real-Time Digital Signal Processing 21
Source: Kuo et al, “Real-time digital

ÊN N
1.3.1 DSP Hardware Options signal processing,” Wiley, 2013
Libraries Data

Ê
Source: Kuo et al, “Real-time digital
As shown in Figure 1.1, the processing of DSPdigital signal x(n) is performed using the DSP

I I
signal processing,” Wiley, 2013 Machine Program
algorithms
hardware. Although it is possible to implement DSP algorithms on different digital hardware, C program code output
C compiler Linker/loader Execution

K K
the given application determines the optimum hardware platform. The following hardware
MATLAB® or C/C++
(Source) (Object)

options are widely used for DSP systems:

T. .vn) .
DAC Figure 1.11 Program compilation, linking, and execution flow
ADC

T .v n )
Data DSP Data
1. Special-purpose (custom)files
chips such assoftware
application-specific
files integrated circuit (ASICs). Các yêu cầu đối với phần mềm DSP
developing DSP systems. C is an efficient tool for performing signal processing and is
portable over different DSP platforms.
2. Field-programmable
Other gate arrays (FPGAs). Other
ª Độ tin cậy is an(reliability)

G .e d u G .e d u
1
computers computers MATLAB Viết chương trình chạy ra
interactive, technical computing environment for scientific and engineering
3. General-purpose microprocessors or microcontrollers (mP/mC). Analysis numerical analysis, computation, and visualization. Its strength lies in the fact that complex
Signal generators kết quả chính xác chỉ là
4. General-purpose digital signal processors. Ø Rấtproblems
ít khi (thậm
can bechí không baoin giời) bị lỗiof the time required by programming

N N
18 Real-Time Digital Signal Processing numerical solved easily a fraction một phần nhỏ công việc
5. Digital signal processors with application-specific hardware (HW) accelerators. languages such as C/Cþþ. By using its relatively simple programming capability, MATLAB1
ª canKhả năng bảo trì (maintainability) phát triển phần mềm thuật
Figure 1.10 DSP software developments using a general-purpose computer be easily extended to create new functions, and is further enhanced by numerous toolboxes.

Ơ p t it Ơ p t it
The advantages of developing DSP algorithms using a general-purpose computer are: 1 toán XLTHS TGT
Dễ sửa
InØaddition, MATLAB chữa, chia sẻ many
provides giữa graphical
nhiều người phát triển
user interface (GUI) tools such as the Signal
The characteristics of these hardware options are summarized in Table 1.1. Processing Tool (SPTool) and Filter Design and Analysis Tool (FDATool).
ASIC devices arehigh-level
usually languages
designedsuch for as
specific tasks that require
or otherintensive computation
packages such ª Khả năngof mở rộng (extensibility)

Ư Ư
1
1. Using MATLAB , C/Cþþ, DSP software The purpose programming languages is to solve problems involving the manipulation of
on computers
as digital subscriber can significantly
loop (DSL) modems, save
oralgorithm
high-volume development
products time.that
In addition,
use maturethe proto-
algorithms

tt@ tt@
information.
Ø Dễ mở Therộng
purpose of DSP
và hiệu programs
chỉnh để đápis toứng
manipulate
các yêu signals to solve specific signal
cầu mới

R ie R ie
type C programs used for algorithm evaluation can be ported to different DSP hardware processing problems. High-level languages such as C/Cþþ are usually portable, so they can
such as fast Fourier transforms. These devices perform the required functions much faster
ª beHiệu quảand (efficiency)
n n
platforms. recompiled run on many different computer platforms. Although C/Cþþ is catego-

T T
because their 2. Itdedicated
is easier to architecture
debug and modify is optimized
high-levelfor the required
language programsoperations,
on computersbut they lack
using rized as a high-level language, it can also be used for low-level device drivers. In addition, C
flexibility to modify
integratedthe specific algorithmstools. and functions for new applications. They are Ø Cóare khả năng fortậnmost
dụng lại vàdigital
cho kết quả chính xác khi
C sử dụng tài isnguyên ít nhất

(k (k
software development compilers available modern signal processors. Thus, programming the
suitable for 3.
implementing
I/O operations basedwell-defined
on disk filesandarepopular DSP algorithms
easy to implement for high-volume
and the behaviors products,
of the system are most commonly used high-level language for DSP applications.
Ngôn
C hasngữ
becomelập trìnhof DSP
choice forchính:
many DSP Assembly vàengineers,
C not

© ©
or applicationseasy to analyze.extremely high speeds that can only be achieved by ASICs. Recently,
demanding the language software development
4. Floating-point data formats and arithmetic can be used for computer simulations, thus ease only because it has powerful commands and data structures, but also because it can easily be
the availabilityofof core modules for some common DSP functions can simplify ASIC design
development. Tốttonhất
ª ported nên
different kết
digital hợp
signal cả hai
processors andngôn
platforms.ngữ lập trình
C compilers are available for a wide
tasks, but the5.cost of prototyping
Bit-true simulations ofASIC devices,
the developed the longer
algorithms candesign cycle,using
be performed and MATLAB
the lack of 1 standard
or
ª range of computers and processors, thus making the C program the most portable software for
Chương trình chung được viết bằng C
development tool C/Cþþ support and reprogramming
for fixed-point DSP implementation. flexibility sometimes outweigh their benefits. DSP applications. The processes of compilation, linking/loading, and execution are outlined
FPGAs have been used in DSP systems for years as glue logics, bus bridges, and peripherals for ª inCác
Figurevòng
1.11. Thelặp và các khối
C programming chạy includes
environment mất nhiều the GUIthời gian
debugger, whichthìisviết
usefulbằng Assembly
5 in identifying errors in source programs. The debugger can display values stored in variables 6
1.4.2 costs
reducing system Selection of2019-nay,
(c) DSP Hardware
and affording higher
Trương Trunglevels of system votuyen@gmail.com)
Kiên (kientt@ptit.edu.vn, integration. Recently, FPGAs have (c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com)
at different points in the program, and step through the program line by line.
been gainingAsconsiderable
discussed earlier, attention in high-performance
digital signal processors are usedDSP applications,
in a wide and are emerging
range of applications from as
coprocessors [14] for standard
high-performance radardigital
systemssignal processors
to low-cost consumer thatelectronics.
need specific accelerators.
DSP system designersIn these 1.5 Experiments and Program Examples
require a full understanding of the application requirements
cases, FPGAs work in conjunction with digital signal processors for integrating pre- and post-in order to select the right DSP The Code Composer Studio (CCS) is an integrated development environment for DSP
hardware for the given application. The objective is to choose the processor that meets the applications. CCS has several built-in tools for software development including project build
processing project’s
functions. These devices are hardware reconfigurable, and thus allow system
requirements with the most cost-effective solution [18]. Some decisions can be made environment, source code editor, C/Cþþ compiler, debugger, profiler, simulator, and real-time
designers toatoptimize the hardware
an early stage architecture
based on arithmetic format,for implementing
performance, algorithms
price, power that require
consumption, ease of higher operating system. CCS allows users to create, edit, build, debug, and analyze software programs.
performancedevelopment
and lowerand production
integration,cost.
and soInon.
addition,
For real-timedesigners can implement
DSP applications, high-performance
the efficiency of data It also provides a project manager to handle multiple programming projects for building large
complex DSP flow into and out
functions of the
using processorofisthe
a fraction alsodevice,
critical.and use the rest of the device to implement
Lựa chọn phần cứng DSP (1) Lựa chọn phần cứng DSP (2)
applications. For software debugging, CCS supports breakpoints, watch windows for monitor-
system logic or interface functions, resulting in both lower costs and higher system integration. ing variables, memory, and registers, graphical display and analysis, program execution

ÊN ÊN
Example 1.5 profiling, and display assembly and C instructions for single-step instruction traces.
This section uses experiments to introduce several key CCS features including basic
There are a number of ways to measure a processor’s execution speed, as follows:
Source: Kuo et al, “Real-time digital editing, memory configuration, and compiler and linker settings for building programs. We

I I
Table 1.1 Summary
1. MIPSof– DSP hardware
Millions implementations
of instructions per second.
signal processing,” Wiley, 2013 Implementation
will demonstrate DSPArea Powerand debugging
software development Flexibility
processesSw.
using Dev. Risk
CCS with the Remote
2. MOPS – Millions of operations per second. type
low-cost TMS320C5505 eZdsp USB stick. Finally, we will present real-time audio experi- Upgrade

K K
3.
ASIC FPGA mP/mC Digital signal
MFLOPS – Millions of floating-point operations per second.
Digital signal ments using eZdsp, which will be used as prototypes for building real-time experiments
processor processors with FPGA -- -- ++ - ++ ++

T. .vn) T. .vn)
4. MHz – clock rate in mega hertz.
5. MMACS – Millions of multiply–accumulate operations. HW accelerators Parallel - 0 0 -- 0 +
Flexibility In addition,None
there are Limited High
other metrics to be considered High (mW) for measuring
such as milliwatts Medium Homogeneous
Design time power consumption,
Long Medium
MIPS per mW, or MIPSShortper dollar. These
Short
numbers provideShort
a simple
Configurable 0 + - 0 -- 0

N G .e d u N G .e d u
Power consumption
indication Low
of performance, power, and price
Low–medium for the given application.
Medium–high Low–medium Low–medium
Performance High High Low–medium Medium–high High CPU
Development cost High Medium Low Low Low
As discussed earlier, hardware cost and product manufacture integration are important
CPU + + ++ -- + - -
Production cost Low Low–medium Medium–high Low–medium Medium accelerators

Ơ Ơ
factors for high-volume applications. For portable, battery-powered products, power con-

it
sumption is more critical. For low- to medium-volume applications, there will be trade-offs
it
pt pt
Một sốamong
lựadevelopment
chọn phần time, costcứng để triển
of development tools, andkhai
the cost of the hardware itself. The
Single speed ++ - + ++ + ++

Ư t@ Ư t@
likelihood of having higher performance processors with upwards-compatible software is also demon
ª Chipanthiết kế factor.
important riêngFor(ASIC – Application-specific
high-performance, low-volume applications integrated circuit)
such as communication

TR R
nt nt
Source: Francois Charlot, “Digital Signal Processor evolution over the last 30 years,” Mar. 2013 with info from Mirocroprocessor Report, "Mixed
ª FPGA (Field-programmable gate array) architectures dominate consumer sockets", 11-Aug-2008

ª i
Vi xử lý mục đích chung e(!P/!C - microprocessor/microcomputer)
(ksố mục đích chung (DSP – Digital signal processor) T (k i e
© ©
ª Các bộ xử lý tín hiệu So sánh tương đối với quy ước ký hiệu như sau
ª Các bộ xử lý tín hiệu số được tăng tốc bằng phần cứng (hardware (HW) ª "kém nhất --" => "trung bình 0" => "tốt nhất ++"
accelerators)
Lựa chọn phần cứng để triển khai DSP phụ thuộc nhiều yếu tố
(c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 7 (c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 8

2
 2/21/19
12 Real-Time Digital Signal Processing

Program address bus Data address bus

Program Processor Data

memory memory

12 Kiến trúc của các !P/!C Real-Time Digital Signal Processing

Program data bus Data data bus
Các bộ xử lý tín hiệu số DSP đầu tiên

N N
(a) Harvard architecture

Program address bus Data address bus Address bus

Program Processor Data Memory

I Ê Processor

I Ê
K K
memory memory

. .
Data bus
Program data bus Data data bus

T .v n ) T .v n )
(b) von Neumann architecture
(a) Harvard architecture
Figure 1.8 Different memory architectures
Address bus

!P/!C đang ngày

CácMemory Processor
càng được phổ
General-purpose

ª Xử lý nhanh các phép toánforđơn

Data bus
giản
engine,

N
brake,nhưG .e d u
mP/mCbiến
becomevà cóandhiệu
faster

cộng, trừ,
and suspension
năng
increasingly

vàanddịch
control
cao
capable of handling some DSP
applications. Many electronics products such as automotive controllers use microcontrollers
are often designed using these processors. If new

N G .e d u
Ơ p t it Ơ p t it
DSP functions are needed for an existing product based on mP/mC, it is preferable to
ª Xử lý(b)chậm các architecture
von Neumann phép toánimplement
phức tạp như nhân, chia
these functions in software than modify existing hardware.
Ø Do1.8yêu cầu thực hiệnarchitectures
một loạt nối tiếp các phép toán cộng,fall trừ,into
dịchtwo categories: Harvard architecture and von

Ư Ư
Figure Different memory General mP/mC architectures
Neumann architecture. As illustrated in Figure 1.8(a), Harvard architecture has separate

tt@ tt@
R ie R ie
memory spaces for the program and the data, thus both memories can be accessed

n n
General-purpose mP/mC become faster and increasingly capable simultaneously.
of handlingThe some vonDSPNeumann architecture assumes that the program and data are stored
Hai kiến trúc chính

T T
applications. Many electronics products such as automotive controllers in the sameuse memory as illustrated in Figure 1.8(b). Operations such as add, move, and subtract
microcontrollers
Bắt đầu xuất hiện DSP đơn chip (single-chip DSP) từ 5/1979
(k (k
for engine, brake, and suspension control and are often designed using are easy
thesetoprocessors.
perform on mP/mC. However, complex instructions such as multiplication and
If new
ª Kiến trúc Harvard: bộ nhớ mP/mC, riêng cho phần chương trình và cho phần dữ liệu
DSP functions are needed for an existing product based on division are
it isslow since they
preferable to need a series of conditional shift, addition, or subtraction ª Bell Labs's single-chip DSP-1 được dùng trong tổng đài chuyển mạch số

© ©
implement these functionsnhưngin softwarehaithan
bộmodify
nhớ existing
có thểhardware.
được
operations.truyThesenhập đồng
devices do thời
not have the architecture or on-chip facilities required for
General mP/mC architectures fall into two categories: Harvard efficient DSP operations,
architecture and von and they are not cost effective or power efficient for many DSP ESS của AT&T
Neumann architecture. ª AsKiến trúc invon
illustrated Neumann:
Figure 1.8(a), Harvard bộ nhớ chung
applications.
architecture It is
has cho
important
separatecả chương
to note that sometrình
modernvàmicroprocessors,
dữ liệu specifically for mobile
memory spaces for the program and the data, thus both memories and portable candevices,
be accessedcan run at high speed, consume low power, provide single-cycle
ª Họ thiết bị 16-bit lập trình được TMS 320 của TI xuất hiện 1980 và được
simultaneously. The von Neumann architecture assumes that the multiplication
program and data andare arithmetic
stored operations, have good memory bandwidth, and have many dùng trong nhiều ứng dụng như điện thoại di động, đồ chơi, ...
in the same memory as illustrated in Figure 1.8(b). Operations such as add, move,
supporting toolsandandsubtract
software available for ease of development. Source: https://www.computerhistory.org/siliconengine/single-chip-digital-signal-processor-introduced/
are easy to perform on mP/mC. However, complex instructions
(c) 2019-nay, Trương such
Trung as multiplication
A digital
Kiên signal processor
(kientt@ptit.edu.vn, andis basically a microprocessor with architecture and
votuyen@gmail.com) 9 instruction set (c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 10
division are slow since they need a series of conditional shift, addition,
designed or subtraction
specifically for DSP applications [15–17]. The rapid growth and exploitation of
operations. These devices do not have the architecture or on-chip digitalfacilities requiredtechnology
signal processor for is not a surprise, considering the commercial advantages
efficient DSP operations, and they are not cost effective or power efficient
in terms for many
of the DSP
fast, flexible, low-power consumption, and potentially low-cost design
applications. It is important to note that some modern microprocessors, specifically
capabilities offered forbymobile
these devices. In comparison to ASIC and FPGA solutions, digital
and portable devices, can run at high speed, consume low power, provide single-cycle
signal processors have advantages in ease of development and being reprogrammable in the
multiplication and arithmetic operations, have good memory bandwidth, field to upgrade and product
have many features or fix bugs. They are often more cost effective than custom
supporting tools and software available for ease of development. hardware such as ASIC and FPGA, especially for low-volume applications. In comparison to
A digital signal processor is basically a microprocessor with architecture and instruction
general-purpose set
mP/mC, digital signal processors have better speed, better energy efficiency

Đặc điểm của các bộ vi xử lý DSP (1/2)

designed specifically for DSP applications [15–17]. The rapid or growth
powerand exploitationand
consumption,
digital signal processor technology is not a surprise, considering the commercial advantages
of lower cost for many DSP applications.
Đặc điểm của các bộ vi xử lý DSP (2/2)

ÊN N
in terms of the fast, flexible, low-power consumption, and potentially low-cost design
capabilities offered by these devices. In comparison to ASIC and FPGA solutions, digital
Các khốiin MAC nhanhand (fast

I I Ê
signal processors have advantages ease of development being MAC
reprogrammable units) in the Điều khiển chương trình đặc biệt (special program control)
field to upgrade product features or fix bugs. They are often more cost effective than custom

K K
hardware such as ASIC ª and DSPFPGA, tích hợp for
especially các bộ nhân
low-volume và các Inbộ
applications. cộng vào
comparison to trong cùng tuyến dữ liệu để ª Cung cấp khả năng lặp không cần tiêu đề (zero-overhead looping)
general-purpose mP/mC, digital signal processors have better speed, better energy efficiency
có thể hoàn thành phép toán MAC trong một vòng/chu lệnh

T. .vn) .
or power consumption, and lower cost for many DSP applications.

Tập lệnh được tối ưu (optimal instruction set)

T .v n )
G .e d u G .e d u
Truy cập đa địa chỉ (Multiple memory access) ª Có tập lệnh được tối ưu để hỗ trợ các hoạt động tính toán nhiều
ª Kiến trúc Harvard hiệu chỉnh để sử dụng đồng thời các lệnh và bộ nhớ

N
ª Nhiều bus dữ liệu trên chip (on-chip bus), dải bộ nhớ độc lập (independent

Ơ p t it
memory banks) và bộ nhớ dữ liệu truy nhập kép trên chip (on-chip dual N
Giao diện song song hiệu quả (effective parallel interface)

Ơ p t it
ª Giao diện vào/ra (I/O interface) nối tiếp và song song hiệu năng cao để kết

Ư Ư
access data memory)
nối đến các thiết bị khác như ADC và DAC

tt@ tt@
R ie R ie
ª Cơ chế quản lý vào/ra (I/O handling) theo luồng để cho việc truyền dữ liệu
n n
T T
gần như không ảnh hưởng đến các khối tính toán
Các chế độ địa chỉ đặc biệt (special address modes)
(k
ª Có các khối tạo địa chỉ dành riêng để tạo địa chỉ dữ liệu song song với việc (k
© thực hiện các lệnh
ª Thường hỗ trợ việc đánh địa chỉ vòng (circular addressing) và đánh địa chỉ
ngược bit (bit-reversed addressing)
©
(c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 11 (c) 2019-nay, Trương Trung Kiên (kientt@ptit.edu.vn, votuyen@gmail.com) 12