3. Basic Computer Architecture
Recap Ch 1 Slide 6: Computers
• A computer is a machine that executes sequences of instructions (programs) which direct it to operate on (process) data.
Copyright By PowCoder代写 加微信 powcoder
• Strength: generality.
• Program instructs hardware to process incoming data in any chosen way.
• Compare with more detailed picture on next slide.
Systems and Networks Recap 2
A computer has several key subsystems:
– processor or central processing unit (CPU) which fetches, decodes and executes sequences of binary coded instructions, known as machine code (m/c).
– instruction memory where a program is stored as a list of m/c instructions.
– data memory where binary coded data is held.
– input/output subsystem which allows data to enter the computer via I/O devices like keyboards, screens, disks etc.
– system interconnect which allows the various components to communicate. This is usually implemented as one or more buses.
Computer Subsystems
Keyboards, scanners etc
Instructions fetched by CPU
Screens Printers etc
Systems and Networks Basic Computer Architecture 3
Electronic memory can store only binary coded information.
Each CPU has a set of m/c instructions it can interpret
This instruction set is different for different CPU designs
Instruction Memory
Processor (CPU)
intermediate results
Holds raw data and
Data Memory
• Instruction and data memory have same structure.
– Each is organised as an indexed array of registers called
locations each capable of holding a word.
– Each location labelled by unsigned binary index or address.
• CPU generates addresses when it wants to read from or write to a location.
• The length (in bits) of an address and of a memory word is set by each CPU design.
Location 0 Location 1 Location 2 Location 3 Location 4 Location 5 Location 6 Location 7 Location 8
Location 2n-1
– set of all addresses CPU can generate is its address space. • Memory hardware needs to be physically mapped to any
location CPU is to use.
– If no memory cells at an address, address is unpopulated.
– Unpopulated addresses cannot store anything!
– large address spaces are often only partly populated
Buses data/ instructions
Systems and Networks Basic Computer Architecture
Harvard and Princeton
• Randomly addressable memory such as that used for instructions and data is called primary memory.
• Instructions and data are stored in primary memory locations as binary coded words.
– If an instruction or data item has a code word too big to fit in one location it must be split across several.
• In most current computers the instructions and data are stored in a single address space (often on the same chips) accessed using the same buses.
– A computer where instructions and data are stored in the same address space is called a von Neumann or Princeton architecture;
– A computer with separate address spaces is called a Harvard architecture.
Systems and Networks Basic Computer Architecture 5
More Memory
• Physical memory is called volatile if its content is lost when power is removed; otherwise it is non-volatile.
• Some physical memory in the computer cannot be written to in normal use. Such memory is called read only memory (ROM).
• ROM is non-volatile and is used, for example, to hold programs that are essential to start the computer up. Writing to conventional ROM requires special hardware.
• Other physical memory is read/write and is usually called random access memory (RAM). Conventional RAM is volatile.
• Another type of memory is flash memory which behave like RAM but is non-volatile.
• Computers also have secondary memory in the form of disks and other mass storage devices. Such devices are not directly addressable by the CPU and are accessed via I/O interfaces.
Systems and Networks Basic Computer Architecture 6
• A CPU is a compound subsystem comprising:
– A control unit (CU) which controls the rest of the CPU and, through it, the computer.
– An execution unit (EU) which contains hardware to perform operations on binary words (e.g. includes an adder).
– A register file consisting of a set of general registers for holding data or addresses.
– A bus interface through which the CPU communicates with primary memory and I/O interfaces
• These devices are connected via buses internal to the CPU.
• A CPU on a single chip is a microprocessor or MPU.
• A complete computer on a single chip is an MCU
Systems and Networks Basic Computer Architecture 7
Control Unit
• The CU includes some special-purpose registers of its own.
• One of these is a register called the program counter (PC) that holds a memory address.
– This always points at the next instruction to be fetched.
• When an instruction has been fetched, the CU decodes it.
• It then controls execution of the instruction and using the other subsystems as necessary.
• Unless otherwise informed, the CU will assume that the next instruction is stored in the next memory address after the last one. Some instructions (e.g. jumps or branches) can change this normal flow.
• The sequence of fetch-decode-execute is called the instruction cycle and is repeated for each instruction in the program.
Systems and Networks Basic Computer Architecture 8
Execution Unit
• The EU is able to perform a variety of operations on one or two binary words supplied to it by the CU.
• The core of the EU is a device called an arithmetic logic unit (ALU). This device can typically:
– Perform Boolean logic operations such as (bitwise) NOT, AND, OR etc. These ops can be used on any binary words regardless of code being used.
– Add, subtract or compare two binary codewords (both these operations work on unsigned and two’s complement codes). The ALU will generate status bits (or flags) such as C and V to tell if overflow has occurred.
– Other status bits may also be generated to inform future instructions: e.g. a Z bit if the result is 0.
• An EU will also have a shifter: hardware to shift words left or right.
• An EU may also have additional hardware to perform fast multiply or divide. Without this hardware, operations like multiply have to be built up from addition and shifting.
Systems and Networks Basic Computer Architecture 9
CPU Registers
• A CPU has two types of register:
– Registers in the register file for general use by instructions and are numbered.
– Special purpose registers usually associated directly with the CU.
• In some CPUs some registers in the register file have special functions.
• The register file is often used to supply codewords to the EU during instruction execution and is often where results from the EU are written.
• There are two kinds of special purpose register:
– Registers which contain useful information at the end of an instruction.
• E.g. the PC; the status register (SR) present in most CPUs, which stores the status bits or flags (V,C,Z etc.) from the last instruction
– Registers which are only used inside an instruction and have no useful information after the instruction finishes execution.
• E.g. the instruction register (IR), holds the first code-word of the instruction currently being processed (often m/c instructions have only one word).
Systems and Networks Basic Computer Architecture 10
CPU Internal Structure
ADDRESS BUS
REGISTER 1
INTERNAL BUS
STATUS LINES
INSTRUCTION REGISTER STATUS REGISTER
MEMORY BUFFER REGISTER
REGISTER 0
PROGRAM COUNTER
REGISTER 2
REGISTER 3
CONTROL UNIT
EXECUTION UNIT
REGISTER FILE
CONTROL LINES
Systems and Networks
Basic Computer Architecture
MEMORY ADDRESS REGISTER
CONTROL SIGNALS: READ/WRITE ETC
Writing Programs
• A programmer can write a program directly as a list of m/c instructions.
• It is hard to write m/c directly. To help, each m/c instruction has a symbolic
form. A program written in this form is called assembly language.
• Assembly language can be translated into m/c automatically by a program
called an assembler.
• Because each type of CPU has its own m/c, programs written in assembly
language are machine-specific.
• Programs written in a high-level language (HLL) like Java or C must also be translated into m/c before being executed. The HLL is machine independent before it is translated by a compiler.
• A compiler must generate m/c for a specific CPU type and so the compiler itself is machine specific, even though the HLL is not
• A single HLL statement will compile into many m/c instructions.
• A single assembly statement will assemble to exactly one m/c instruction.
Systems and Networks Basic Computer Architecture 12
Instructions
• Instruction formats vary from one CPU type to another.
• In any given machine each instruction must tell the CU exactly what to do.
• To say ADD is not enough. The instruction must say where to find the data to be operated on (the operands) and where the result should be stored.
– The locations of operands and results may be in registers or in memory locations.
– Not all instructions allow total freedom to specify operand and result locations.
– For example, in some CPUs, arithmetic and logic instructions can only use CPU registers. Operands must be read from memory and results written to memory using other instructions (names like LOAD, STORE or MOVE but CPU specific).
• The set of CPU registers that can be used to carry information from one instruction to another is called the programmer’s model. The model typically contains all general purpose registers, the PC and the SR.
• Registers like the IR are not, however, included in the programmer’s model. A programmer may ignore these registers although they are relevant to a hardware designer.
Systems and Networks Basic Computer Architecture 13
The Instruction Cycle
1. CU sends PC address to Bus Interface Memory Address Register (MAR)
2. CPU reads first word of instruction from memory into Memory Buffer Register (MBR)
3. First word placed in Instruction Register (IR)
4. CU decodes instruction.
5. If instruction has more than one word, CU commands bus interface to get the rest.
6. CU organises execution. This depends on what type of instruction. Assume ADD.
7. CU advances PC to point to next instruction.
8. CU determines where operands are. If any operands are in memory, CU commands bus interface to get them (In our example CPU, ADD operands are only allowed in register file).
9. CU supplies EU with operands and commands EU to add.
10. CU determines where result goes and writes it. In some CPUs this can also mean a memory access (in our CPU, ADD result must also go into register file).
11. EU generates status bits aka condition codes (C, V, Z etc.) and stores them in a status register (SR) in some CPU designs called a condition code register.
12. Finished. Start fetch of first word of next instruction
Systems and Networks Basic Computer Architecture 14
Revision: von Neumann computer
Location 0 Location 1 Location 2 Location 3 Location 4 Location 5 Location 6 Location 7 Location 8
Location 2n-1
• CPU, primary memory, buses.
• Every memory location has a unique address and can store a fixed number of bits (content) which can be a machine code or data (or random) word.
• Instructions are binary-coded & stored in memory (machine code).
• List of instructions written to perform some task is a program.
• We can write programs in machine code (nobody does this anymore).
• Instead use symbolic language.
– Low-level (assembly) language – statements are
1-to-1 with m/c. This is CPU specific.
– High-level language (e.g. Java) – statements are
1-to-many with m/c. This is CPU independent.
• Data also stored as binary codewords in memory
(data codes).
Buses data/ instructions
Systems and Networks Basic Computer Architecture
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com