SEC204
1
Computer Architecture and Low Level Programming
Dr. Vasilios Kelefouras
Email: v.kelefouras@plymouth.ac.uk Website: https://www.plymouth.ac.uk/staff/vasilios -kelefouras
School of Computing (University of Plymouth)
Date
28/10/2019
2
Von Neumann architecture What is the CPU?
How CPU works?
• Arithmetic Logic Unit (ALU)
• Control Unit (CU)
• Bus
• Memory
• Registers
• Clock
Memory Hierarchy Secondary Memory Instruction Pipeline
Outline
3
The Von Neumann Architecture
All computers more or less based on the same basic design, the Von Neumann architecture
It is a Model for designing and building computers, based on the following three characteristics:
1.
The computer consists of four main sub-systems: Memory
ALU (Arithmetic/Logic Unit)
Control Unit
Input/Output System (I/O)
Program is stored in memory during execution Program instructions are executed sequentially
2.
3.
The architecture is named after the mathematician, John Von Neumann
A variation of this architecture is the Harvard architecture which separates data and instructions into two pathways
4
Processor (CPU)
Control Unit Registers
ALU
Input-Output I/O
Memory
The Von Neumann Architecture
The parts are connected to one another by a collection of
wires called a bus
Bus
Loads/Stores data and program
Executes program – Fetch, decode, execute
Handles communication with the
“outside world”, e.g.
• Screen
• Keyboard
• Storage devices •
Small memory – specific purpose only
Performs arithmetic/logic operations requested by program
5
Central Processing Unit (CPU)
Carries out the program’s instructions!
Operates on data it finds in the computer’s memory
Includes all binary circuits that carry out arithmetic & logic operations- reduced to a single Integrated Circuit
CPU has four key parts that we will examine: Control Unit, Arithmetic & Logic Unit, Registers, Clock
CPUs support a set of very simple instructions that typically fall into the following categories:
Data movement (load, store, copy…)
Arithmetic/logical (add, subtract, compare..) Programcontrol (branch,jump…)
Very primitive commands (operations) executed by the CPU
These commands are implemented as electronic binary circuits which can transform the 0s and 1s.
6
Processor (CPU)
Control Unit Registers
ALU
Central Processing Unit (CPU)
CPU is responsible for fetching program instructions, decoding each instruction that is fetched, and executing the indicated sequence of operations on the correct data
The key parts of the CPU are
1. Arithmetic Logic Unit (ALU)
2. Control Unit (CU)
3. Registers 4. Clock
7
CPU Clock (1)
Every computer contains an internal clock that regulates the rate at which instructions are executed and synchronizes all the various computer components
All CPU and bus operations are synchronized to the clock
Clock speeds are expressed in megahertz (MHz) or gigahertz ((GHz)
the beginning of each cycle is when the clock signal goes from “0” to “1”
e.g. CPU frequency 2 GHz -> clock cycle 0.5 ns
In the computer, all timings are measured in terms of clock cycles, e.g., an addition needs 2 cycles
8
CPU Clock (2)
The faster the clock, the more instructions the CPU can execute per second
But, to think that clock and performance is the same thing is the most common
misconception about processors
CPU frequency is not necessarily indicative of the execution speed; e.g. complex operations, cache misses etc
FLOPS: floating operations per second, a useful measure for (super)computers dedicated to extensive computations
A typical modern PC now has either four or five different clocks, running at different (but related) speeds, e.g., system (memory) bus, L2 cache bus, PCI bus
The entire system is tied to the speed of the system clock. This is why increasing the system clock speed is usually more important than increasing the CPU clock
Normally, the processor spends a significant amount of time waiting on data and signals from much slower devices, e.g., memory
9
Registers
In a computer, a register is the fastest memory
Registers are fast stand-alone storage locations that hold data temporarily
Multiple registers are needed to facilitate the operation of the CPU The main registers are:
The Instruction Register (IR)
The Program Counter (PC) or Instruction Pointer (IP) Data registers
10
Processor (CPU)
ALU
Central Processing Unit (CPU) (1)
The arithmetic logic unit (ALU) carries out the logic operations (such as comparisons) and arithmetic operations (add, shift, multiply) required during the program execution
The ALU knows
1. which operation to perform
2. Where are the input data
3. Where to store the output data
All the above are provided by the CU
The ALU performs
1.
•
2.
•
Integer arithmetic operations
Add, subtract, increment, decrement
Bitwise logical operations
AND, OR, XOR, NOT, Arithmetic shift, logical shift, rotate
11
Central Processing Unit (CPU) (2)
ALUs often handle the multiplication of two integers, since the result is also an integer
ALUs typically do not perform division operations, since the result may be a fraction, or a “floating point” number
The floating-point unit (FPU) performs operations on floating point numbers
Modern CPUs have separate unit to perform multiplication faster
Processors include a coprocessor hardware unit which is used to perform much more complex mathematical operations such as arcsine, cosine, floating-point division, etc
12
+
Central Processing Unit (CPU) (3)
The multiply-accumulate operation is a common step that computes the product of two numbers and adds that product to an accumulator
It speeds up many computations that involve the accumulation of products, e.g., Matrix-matrix Multiplication
The hardware unit that performs the operation is known as a Multiplier
Accumulator unit (MAC, or MAC unit) B the operation is called MAC or MAC operation
The MAC operation performs
C
A = A +B x C
X
A
13
Central Processing Unit (CPU) (4)
Arithmetic shift: when shifting to the right, the leftmost bit (the vacant MSB) is filled with the value of the previous MSB (sign)
Ideal for signed two’s complement binary numbers
Logical shift : The vacant bits are filled with zero
the logical and arithmetic left-shifts are exactly the same
Ideal for unsigned binary numbers
Circular shift or bit rotation : In this operation, the bits are “rotated” as if the left and right ends of the register were joined. The value that is shifted in on the right during a left-shift is whatever value was shifted out on the left, and vice versa
14
1
0
1
1
0
0
1
0
Think Pair Share
Apply logical and circular shift to the following register:
15
Control Unit (CU) (1)
A Control Unit is the unit that handles the central work of the computer There are two registers in the control unit
Theinstructionregister(IR)containstheinstructionthatisbeingexecuted Theprogramcounter(PC)containstheaddressoftheinstructionbeing
executed
The CU is responsible of executing the right instruction and organizes the other function units appropriately
In every clock cycle the CU:
Loads from memory the next instruction to be executed – PC register contains
that address
The instruction is stored into IR and is decoded
The CU sends the appropriate signals to the ALU, memory, I/O devices in order to execute the instruction
The CU increments PC to show next instrution
16
Control Unit (CU) (2)
Program is stored in memory
machine language instructions are in binary format
The task of the control unit is to execute programs by repeatedly:
• Fetch from memory the next instruction to be executed
• Decode it, that is, determine what is to be done
• Execute it by issuing the appropriate signals to the ALU, memory, and I/O subsystems
• Continues until the program terminates (HALT instruction)
17
Main Memory
0000 0001 0010 0011 0100 0101 0110 0111 1000
18
Main Memory
Computer memory consists of a linear array of addressable storage cells that are similar to registers
Both program and data are stored into memory
Load/store operations are performed on both instructions and data
Consists of many memory cells (storage units) of a fixed size Each cell has an address associated with it
All accesses to memory are to a specified address
The time it takes to fetch/store a word is the same for all words
19
Main Memory – Address Space
To access a word in memory requires an identifier. Although programmers use a name to identify a word (or a collection of words), at the hardware level each word is identified by an address
The total number of uniquely identifiable locations in memory is called the address space. For example, a memory with 64 kilobytes and a word size of 1 byte has an address space that ranges from 0 to 65,535
If memory contains N words, then log2N bits are needed to address all words in memory
If the memory address space needs N bits, then there are 2N words in memory
20
Think Pair Share
1. A computer has 32 MB (megabytes) of memory. How many bits are needed to address any single byte in memory?
×
2. If main memory is of 64Mbyte and every word is of 2 bytes how many bits do we need to address any single word in memory?
25
, or 25 bits, to address each byte
The memory address space is 32 MB, or 2
25 2
2
). This means that we need
5 (2
20
log 2
The memory address space is 64 MB, which means 2
1
(
This means that we need log
26
. However, each word is
two (2
25 ) bytes, which means that we have 2
words. Note that , or 25 bits, to address each word
Mem.size=number.words x word.size
25 2
)
2
21
How Main Memory and CPU are connected?
Control Bus – sends appropriate signal whether store or load Address Bus – sends the memory address
Data Bus – sends the data
22
Buses (1)
Bus: a group of wires that transfer data from one part to another (data, address, control)
Data bus:
• bi-directional (read/write)
• 8, 16, 32, 64-bit wide (same as ‘word size’)
Address bus:
• specifies memory location in RAM/ROM/interface device to be accessed;
monodirectional
• address space: 16-bit wide -> 216 words= 64×210 = 64KB
• 32-bit wide -> 232 = 4GB – this is why in 32-bit PCs we cannot use more than 4Gbyte of RAM
Control bus:
• carries commands from the CPU and returns status signals from the devices
23
Buses (2)
How many wires needed for the?
A. Data bus?
B. Address bus?
C. Control Bus?
A. B. C.
As many bits as the memory word contains As many bits as the memory address contains 1 bit is enough
24
Think Pair Share
Q:
If main memory is of 64Kbyte and every word is of 8 bytes how many wires do we need for the address bus?
25
(c) Yngvi Bjornsson
CMPUT101 Introduction to Computing
26
Machine Language or Machine Code (1)
• A program consists of a sequence of instructions (in binary)
• EACH instruction specifies both:
– The operation to perform – The address of the data
• Instructions are stored and processed in machine language–also called microcode
• Like everything else (e.g. like ASCII characters) machine language consists solely of bit patterns
27
Machine Language or Machine Code (2)
• Machine languages consist entirely of binary numbers and are almost impossible for humans to read and write
• Assembly languages have the same structure and set of commands as machine languages, but they enable a programmer to use names instead of numbers
• Each type of CPU has its own machine language and assembly language
an assembly language program written for one type of CPU won’t run on another
• In the early days of programming, all programs were written in assembly language
• Now, most programs are written in a high-level language such as Java, Python, C/C++
• Programmers still use assembly language when speed is essential
28
00001001
0000000001100011
0000000001100100
Machine Language or Machine Code (2)
A machine language instruction consists of:
Operation code, specifying which operation to perform
Address field(s), specifying the memory addresses of the values on which the operation works
Opcode (8 bits) Address 1 (16 bits) Address 2 (16 bits)
The above could be (ADD 99, 100) , assuming that the opcode for ADD instruction is 9
Add content of memory locations 99 and 100, and store back in memory location 99)
Instructions are given to the processor in the form of a program … so it knows what circuits to use, in what order; and from where the data should be read or to where it should be stored
29
Assembly basic instructions
Depending on the target CPU, different assembly instructions exist
Instr:
Store 0xA2B , R1 ADD R1, R2
SUB R1, R2, R3 MUL R1, R2
DIV R1, R2 INC R4 HALT
CMP R1, 10 JMP EQ S1
Meaning:
Store R1 in A2B16 memory location
Add R1 and R2 and store the result in R1
R1 = R2- R3
R1= R1 * R2
R1=R1/R2
R4=R4+1
Stops program execution
If R1==10, then set the register EQ=1, else EQ=0
Load next instruction from memory location S1, if EQ==1
Assemblers translate instructions that are comprehensible to humans into the machine language that is comprehensible to computers
30
High level language VS Assembly language
Easily understandable
Portable – run on just any computer
Debugging of the code is easy
High level languages like java, C++, etc. have one to many relationship with assembly, i.e., one statement of java expands into many assembly language commands
Highlevellanguageisalwaysconverted into assembly language
The high level language programmer doesn’t need to know the HW details
Hard to understand
Runs only on the target CPU only
Debugging is very hard
One to one or one to a few relationship
The assembly language programmer must know about the hardware such as registers, etc.
Assembly language can control the machine code better
Assemblyismuchfaster
31
High level language VS Assembly language
High Level Language (HLL) a = x + y – z
Assembly Language (AL): load x into r1
load y into r2
load z into r0
r3 ← r1 + r2
r0 ← r3 – r0
store r0 into a
• In HLL we write a=5 and we assume that it is stored somewhere – we don’t care
• But the computer does not work like that – it has to store every variable in an exact memory location
• In assembly we must specify the Hardware registers as well as the memory locations
32
Program Execution
– 1. PC is set to the address where the first program instruction is stored in memory
– 2. Repeat until HALT instruction or fatal error
2a. Fetch instruction
• Fetches instruction (from memory) at address given by PC; copies it into storage register
2b. Decode instruction
• Copies op code into IR and operands into address registers
• Interprets instruction
• ALU is invoked
2c. Execute instruction
• Execution cycles vary, depending on the op code (instruction), e.g., Load copies data from memory to ALU register, ADD adds values inside the ALU
2d. Increment PC by one, now contains the memory address of the next instruction that will be fetched
33
A simplified example
Load R1, 0x2F22 Load R2, 0x2F24 ADD R3, R1, R2 Store 0x2F24, R3
A simplified example
,4
PC
Instr. reg.
Fetch
CONTROL UNIT 2b 110011
PROCESSOR
ALU
execute cycle
000001
Enter
000011
Address reg
Address reg
3
R1
0101
111100
R2
1
2a
Copy & Decode
0101 110011 111100
Storage register Instructions
#000001 #000011
Data
MEMORY
3
R3
0101
1100 11
11 1100
0000 0011
#110011…002 or 0x2F2216 #111100… 002 or 0x2F2416
0101
1101 10
11 1101
0000 00100
A simplified example
,4
PC
Instr. reg.
Fetch
CONTROL UNIT 2b 110110
PROCESSOR
ALU
execute cycle
000011
Enter
000101
Address reg
Address reg
3
R1
0101
111101
1
2a
Copy & Decode
0101 110110 111101
Storage register Instructions
#000001 #000011
Data #110011 ….
MEMORY
3
0101
1100 11
11 1100
0101
1101 10
11 1101
0000 00100
#111100 ….
5
R2
R3
0000 0011
A simplified example
,4
PC
Instr. reg.
Fetch
CONTROL UNIT 2b 10010
PROCESSOR
Enter ALU
execute cycle
000101
3 3+5=8
000111
Address reg
Address reg
3
R1
0001
1
2a
Copy & Decode
0001 100110 111100
111100
5
R2
Storage register Instructions
#000101 #000111
MEMORY
8
R3
Data #110011 ….
#111100 ….
0001
1000 10
11 1100
0000 0011
0111
1101 10
11 1101
0000 00100
A simplified example
,4
PC
Instr. reg.
Fetch
CONTROL UNIT 2b 10010
PROCESSOR
ALU
execute cycle
000111
Enter
001001
Address reg
Address reg
3
R1
0111
111101
1
2a
Copy & Decode
Storage register Instructions
#000101 #000111
Data
MEMORY
3
0001
1000 10
11 1100
#1111011 ….
0000 1000
5
R2
0111 100110 111101
8
R3
0111
1101 10
11 1101
38
Think Pair Share
Given the following high level code, first translate it into assembly pseudo code and second explain the CPU steps as before. Consider that a=0, x=2, y=3, z=4.
a = x + y – z;
Think Pair Share
PROCESSOR
ALU
R1
R2
R3
PC
Instr. reg.
Storage register
CONTROL UNIT
Address reg
Address reg
MEMORY
Instructions
#000001 #000011
Data #110011 ….
#111100 ….
40
Memory Types (1)
Random Access Memory (RAM) – Alternatively referred to as main memory
Static RAM (SRAM)
Dynamic RAM (DRAM) Read Only Memory (ROM)
Programmable read-only memory (PROM)
Erasable programmable read-only memory (EPROM)
Electrically erasable programmable read-only memory (EEPROM)
RAM loses any information it is holding when the power is turned off
ROM is meant for permanent storage, while RAM is for temporary storage
41
Memory Types (2)
ROM is used primarily in the start up process of a computer, whereas a RAM chip is used in the normal operations of a computer once the operating system has been loaded
A good example of ROM is the computer BIOS, a PROM chip that stores the programming needed to begin the initial computer start up process
Writing data to a ROM chip is a much slower process than writing it to a RAM chip
A RAM chip can store multiple gigabytes (GB) of data, ranging from 1 GB to 256 GB per chip. A ROM chip stores several megabytes (MB) of data, typically 4 MB or 8 MB per chip
42
Cache Memory
Cache is a high-speed static random access memory (SRAM) that a CPU can access more quickly than it can access regular random access memory (RAM)
This memory is typically integrated directly into the CPU chip
Cache memory is faster than main memory, but slower than the CPU and its
registers
Cache memory, which is normally small in size, is placed between the CPU and main memory
The purpose of cache memory is to store program instructions and data that are used repeatedly – The computer processor can access this information quickly from the cache rather than having to get it from computer’s main memory
Fast access to these instructions increases the overall speed of the program
43
for (i=0; i< 1000; i=i+1){ S1
S2 ... S10
}
Main memory
L2 unified cache
L1 data cache
L1 instruction cache
RF
Memory Hierarchy
users want unlimited fast memory
fast memory expensive, slow memory cheap cache: small, fast memory near CPU
large, slow memory (main memory, disk, ...) connected to faster memory (one level up)
CPU
44
Secondary memory (1)
Secondary memory is where programs and data are kept on a long- term basis
Common secondary storage devices are the hard disk and optical disks The hard disk has enormous storage capacity compared to main memory The hard disk is used for long-term storage of programs and data
Data and programs on the hard disk are organized into files A file is a collection of data on the disk that has a name
45
Secondary memory (1)
Running programs are always located in main memory
When creating a new file and type something it is stored into main memory; When you "save" your document, the characters are copied to a file on the hard disk
A permanent copy will also be in secondary memory on the hard disk
Main Memory
Fast Expensive Low Capacity
Hard Disc
Slow
Cheap
High Capacity
Question: Do you think that data transfer from the network is slower or faster than from main memory?
Answer: Data transfers from the network are much slower than from main memory
46
Questions?