Computer Architecture Basics

Computer's Main Components

• A computer is composed of central memory, a central processing unit (CPU), and input/output units.
• Central memory stores programs and data.
• The CPU executes programs loaded into central memory.
• Input/output units allow information exchange with peripheral devices.
• The central unit combines the CPU and central memory.

The Overall Diagram of an Architecture

• The computer's architecture consists of a central processing unit, central memory, and input/output units.
• The CPU is responsible for executing programs.
• Central memory stores programs and data.
• Input/output units facilitate data exchange between the computer and peripheral devices.
• Instructions in programs are sequentially processed by the control unit of the CPU.

The General Diagram of a Computer

• The CPU (Central Processing Unit) includes the control unit and arithmetic logic unit (ALU).
• The Control Unit decodes instructions and coordinates their execution.
• The ALU performs arithmetic and logical operations.
• Cache memory is a fast intermediary between main memory and the CPU.
• Input/output units manage communication with peripheral devices.
• Peripheral units include devices like printers, keyboards, mice, etc., along with storage devices.

The Execution of a Program

• Programs and data are loaded into main memory for execution.
• Instructions are sequentially fetched by the control unit.
• The control unit analyzes instructions and directs the ALU.
• Input/output operations may be part of the program's execution process.
• Cache memory speeds up data access.

Central Memory

• Central memory holds both instructions and data from programs and the operating system.
• Instructions are stored as machine code (e.g., an addition instruction in an Intel processor as 10000001).
• Data is stored using coding schemes like ASCII (e.g., 'F' encoded as 1011001).
• Characters are increasingly stored in 8-bit format for increased character capabilities.
• Memory capacity is typically measured in bytes (e.g., 2 GB = 2 × 2³⁰ bytes ≈ 2 billion bytes).

Memory Cells

• Each memory cell has a unique address and a content (instruction or data).
• Memory capacity is the number of cells.
• Memory is measured in bytes.
• Processors often handle multiple bytes at a time (memory word) for efficiency.
• Read and write operations usually take place on entire memory words.

Data and Memory Registers

• Modern processors manipulate data beyond a single byte (32/64 bits).
• A memory register is a memory cell with a specific function.
• Address registers hold memory addresses.
• Word registers hold memory contents (multiple successive bytes).
• Word size matches a memory word's size.
• Address registers have capacity to address all memory.

Address Registers and Memory Capacity

• Address registers must have sufficient bits (e.g. 10 bits for 1024 cells).
• 32-bit address registers can address 2³² different bytes (4 GB).
• Before 2008, most computers had 32-bit memory limitations (4 GB).
• 64-bit architectures theoretically support significantly larger memory capacities (2⁶⁴ bytes).

Operations in Central Memory

• Reading from memory copies memory cell contents to registers.
• Writing to memory overwrites the content of a specific memory cell or cells.

Access Time and RAM

• Access time is the time taken for reading or writing a memory word.
• It varies, but typical values are a few nanoseconds to a few hundred nanoseconds.
• Random Access Memory (RAM) has equal access time to any cell.

Central Processing Unit (CPU)

• The CPU comprises the control unit and the arithmetic logic unit (ALU).
• The control unit manages instruction execution.
• The ALU performs mathematical and logical operations.

Control Unit

• The control unit manages instruction execution.
• The instruction register (IR) holds current instructions.
• Fields exist with operation codes and operands.
• The Program Counter (PC) holds the address of the next instruction.

Program Counter and Instruction Execution

• Instructions are usually executed sequentially.
• The program counter (PC) is incremented after each instruction's execution.
• Instructions longer than one byte need appropriate PC increment.
• Manual PC modifications might be required for branching or similar operations.

Control Unit Components

• The control unit incorporates components like an instruction decoder and command sequencer.
• The instruction decoder interprets instructions, selecting the relevant circuits (instructions decoded and sent either to the decoder or to microprogram memory to execute the specific operation).
• The command sequencer activates the relevant circuits according to the current instructions' operation (activated by control signals based on the instruction to be executed, often using a clock signal from an external source).

Arithmetic and Logic Unit (ALU)

• The ALU encompasses all electronic circuits for the desired operations (mathematical and logical operations).
• Operations include addition, subtraction, multiplication, division, bitwise negation (bit inversion), and Boolean logic.
• Operands for operations are stored in ALU registers.

The General Diagram of the Operation of the CPU

• A diagram illustrating the components and general operation of the CPU.

Arithmetic and Logic Unit (ALU)

• Modern ALUs handle a wide variety of operations.

Single Operand Operations

• Single register, single operand operations (e.g., resetting, complement, shifting, incrementing).

Two Operand Operations

• Operations with two operands (e.g., addition, subtraction, logical operations).

Accumulator Extension

• Some operations utilize accumulators to support double-precision or expanded accumulator capability.

Floating Point and Double Precision Operations

• Historically, floating-point operations were specific to scientific computers.
• Modern processors integrate floating-point calculations within their ALU.

Processing in the ALU

• Data processing entirely occurs inside the ALU.
• The ALU’s circuits perform essential operations which form the basis for algorithms.
• The control unit directs the entire ALU operation.

Registers of the ALU

• Arithmetic, general-purpose, base, or index registers used for specific operations.

ALU Registers

• Status register (PSW or program status word) holds information about the system's state (e.g. overflow, carry status).

Access Time and Memory Performance

• Register access time is significantly faster than main memory access time.

Input/Output Units

• Input/output units facilitate data exchange between the CPU and peripheral devices.
• The most common components are buses, Direct Memory Access, and channels.
• Transfer speeds decrease away from the CPU.
• I/O operations demand faster specialized processors; dedicated processors manage I/O more efficiently.

Peripheral Units

• Peripheral devices provide data exchange with the outside world.
• Auxiliary memory (storage) allows for substantial data storage (magnetic disks, tapes, cartridges).
• Peripheral devices require dedicated controllers for management and interaction.

Mass Storage and Controllers

• Auxiliary memories (mass storage) provide permanent (non-volatile) storage compared to RAM, which is volatile storage.
• Controllers are required for management and communication of each peripheral device category.

Buses

• Buses establish connections between devices.
• Address, data, and control signals travel via these lines; they are shared resources used for communication during exchange.
• I/O devices have buffer registers combined with interface circuits for computer communication.
• Buses contain components like data, address, and control bus.

Microcomputer Architecture

• The data bus facilitates two-way information transfer.
• Data bus line quantity corresponds to the microprocessor’s bit width.

CPU Registers

• CPU register numbers and types influence programming and are integral to the computer architecture.
• Basic functionalities are generally similar across different CPU manufacturers.
• Description of important registers, function, program modification, and access will be presented.

Instruction Register (IR)

• When fetching instructions, the IR holds the fetched instruction.
• Instruction size corresponds to the instruction register’s size.
• Programmers lack direct access to the instruction register.

Program Counter (PC)

• The PC always holds the next address for instruction execution.
• The PC automatically increments after each instruction's use.
• Program alterations affect sequential instructions via jump type instructions.
• Modifying the PC with new addresses is required for jumps and branching.

Program Counter (PC)

• The PC change happens during instruction decoding.
• The PC changes based on the size of the memory addresses.
• Access to the PC is restricted.

Accumulator (ACC)

• The Accumulator is prominent in single-address machines and the ALU.
• It handles operands in most operations.
• Useful as a buffer in I/O operations.
• Typically sized the same as the ALU, but it can have an extension.

Accumulator (ACC)

• It acts as a store for multiplication and division results and holds most/least significant bits in double-precision operations.
• It has easy access to programmers for data processing; some processors have multiple accumulators; operations specify which one to use.

General Purpose Registers

• General-purpose registers, sometimes called scratchpad registers, are helpful for storing frequently accessed data.
• Using scratchpad registers avoids constant memory accesses (improving execution speed).
• Programmers have access to these registers to manipulate them via instructions.

General Purpose Registers

• Loading data from memory or other registers, storing register contents in memory, transferring data to the accumulator, and incrementing/decrementing operations are typical.

Index Registers (XR)

• Index registers resemble general-purpose registers.
• They support data manipulation (arrays).
• Index registers support indexed addressing schemes (addresses modified using specific indices).

Base Registers

• Base registers mirror index registers for effective address calculations (especially with base addressing).
• Useful for dynamic relocation.
• Base registers extend address handling beyond the standard instruction field’s capacity.

Program Status Word (PSW)

• The PSW (Program Status Word), also called the state register, displays the CPU’s current state reflected in registers.
• Bits (flags) in the PSW show the specific conditions.

Program Status Word

• The zero flag (ZF) indicates if the result is zero.
• The carry flag (CF) indicates a carry in an arithmetic operation.
• Flags aid the CPU, signaling interrupt conditions or operational states (interrupt indicators or levels).
• Flags assist in controlling program execution by indicating conditions (e.g., zero result, carry).

Stack Pointer (SP)

• The SP (Stack Pointer) simulates a stack in main memory, reserving a segment for the stack.
• Functions like a memory address register dedicated to stack operations (specifically RAM's stack portion).
• Pushing/popping bytes updates the SP (stack pointer); operations reference last pushed byte (Last-In-First-Out approach); addresses are recorded, then read in reverse.

Stack Operation (LIFO)

• Data storage follows LIFO (Last-In-First-Out) in the stack.
• Reading data follows the reverse order of storage in the stack.

Specialized registers

• Simulating multiple LIFO stacks requires multiple Stack Pointers (SPs).
• Special-purpose registers exist for specific operations - e.g., shift registers for bit manipulation, registers for decimal arithmetic.

Memory

• A memory device records, stores, and retrieves information.
• Memory is organized in hierarchical levels (e.g. CPU registers, cache, primary mass, auxiliary memory).
• These hierarchical levels vary dramatically based on access time and size.

Human Memory vs Computer Memory

• Human memory exceeds computer capabilities in activation and retrieval mechanisms, though computers can surpass human capacity in storage.
• Human brains utilize processing and storage in parallel, which gives the brain a distinct advantage in certain types of tasks.

Parallel Processing vs Sequential Processing

• The brain processes data simultaneously, while computers primarily process sequentially.
• This explains differences in performance between brain and computer processing capabilities.
• In computers, memory is passive, while in the brain, it is actively involved in processing.

Memory Hierarchy

• A diagram illustrating the levels of memory hierarchy based on speed and capacity (e.g. CPU registers -> cache memory -> main memory-> primary storage -> secondary storage -> mass storage).

Memory Hierarchy in a Computer

• Memory hierarchy is based on three criteria: access time, capacity, and cost per bit.
• Access time decreases with proximity to the CPU, but cost per bit increases, and storage capacity increases.

CPU Registers

• High-speed registers are within the CPU with ultra-fast access times, used for calculations/intermediate results.

Cache Memory

• Cache memory acts as a buffer between the CPU and main memory.
• It reduces memory access time.

Main Memory (Central Memory)

• Main memory serves as the primary workspace for the CPU.
• Programs must be loaded here to execute properly.

Mass Storage

• Primary mass storage includes hard drives and solid-state drives (SSDs); it provides long-term data storage.
• Secondary mass storage (auxiliary memory) uses slower but substantial capacity devices (e.g., magnetic disks, tapes, magneto-optical discs).

Permanent Storage Systems

• Various permanent storage options exist; online storage (SSD/Hard disks) provides fast access.
• Network storage (SAN/NAS) enables remote or local access via a network (e.g., accessing data from your home computer to a server).
• Cloud storage employs remote servers accessible through the internet to store data.

Near-Line and Offline Storage

• Near-line storage (based on devices like magnetic tape or optical libraries) offers slow but large-scale storage. - Offline storage stores data on shelves without automated access.

Characteristics of Memory

• Address: Unique identifier for memory location.
• Capacity: Total amount of data stored.
• Access time: Time to retrieve data.
• Memory cycle: Time between successive access operations.

Characteristics of Memory

• Data transfer rate: Number of data units handled per second.
• Volatility: Persistence of data in memory (volatile - data lost without power, non-volatile - retains data).

Different Types of Memory Access

• Sequential access: Data accessed by traversing sequentially (e.g., tapes).
• Direct access: Data access happens instantly using its address (e.g. memory cells in RAM).
• Semi-sequential access: Combination of both sequential and direct access approaches.

Main Memory

• Main memory contains programs and data (necessary components for operating system functions).

Technological Evolution of Main Memory

• Main memory evolution included various technologies (delay lines, vacuum tubes, magnetic drums).
• Semiconductor memories later replaced these earlier technologies.
• Technologies have improved efficiency and speed of operations.

Semiconductor Memories

• Semiconductor memories have become fundamental for modern main memory systems (especially since the 1970s).
• Technological advancements in integrated circuits led to better capacity and speed with reduced costs.
• RAM offers random/direct access to memory cells.

Read-Write and Read-Only Memory

• Read-write memory allows both reading and writing operations (volatile, current data).
• Read-only memory permits only read operations—data is written during manufacturing and remains unchanged.

Characteristics of Major Types of Semiconductor Memory

• Random-access memory (RAM) examples include static/dynamic RAM, and other variations; access times differ based on specific technology used.
• Data transfer rates and characteristics are given for each major RAM type.

Characteristics of Major Types of Semiconductor Memory

• VRAM: Video RAM, used in graphics cards due to two ports for concurrent writing/reading.
• MRAM: Magnetic RAM; using magnetization, data is non-volatile, hence it's retained even when power is off.

Characteristics of Major Types of Semiconductor Memory

• Access time is measured in nanoseconds; SDRAM access time is described by bus frequency (MHz).

ROM Memory

• ROM (read-only memory) stores data permanently; the data can only be read and not written to after production. (non-volatile memory).

ROM Memory

• ROM production is often costly; it's manufactured using MOS or bipolar technologies; ROM is often used for code conversion, character generation, and system programs.

PROM and EPROM

• Programmable ROMs (PROMs) can be programmed once by the user, beneficial especially during prototype construction.
• Erasable programmable ROMs (EPROMs) can be reprogrammed repetitively.
• EPROM utilizes ultraviolet light for erasure; this process often takes up to 30 minutes.

EEPROM

• EEPROM (electrically erasable programmable ROM) allows for electrical erasure; this offers more flexibility compared to EPROM.

Flash Memory

• Flash memory combines EEPROM advantages with permanent storage capabilities.
• Uses floating-gate transistors; this data storage ability is gigabytes in size.
• This design improves data transfer rates, reliability, and shock resistance.

Evolution of ROM Memory

• ROM's evolution transitioned from purely read-only functionalities to reprogrammable varieties with erasure capabilities.
• Flash memory combines ROM's advantages and supports more robust and flexible memory capabilities.

Cache Memory

• Cache memory addresses the speed disparity between the CPU and main memory by creating fast intermediate storage.
• This memory is not part of the main memory and includes essential information (instructions, data).
• Cache memory employs associative addressing; each entry has a key and its associated data stored in memory.

Searching in Cache Memory

• To find data in cache memory, a key search comparing with each cache entry is performed concurrently to determine if the sought instruction or data is present in the cache.
• This parallel comparison leads to faster access times to the memory for instructions.

Multi-Level Cache

• Modern computers often use multiple cache levels.
• Often, a few megabytes of cache memory are used, divided to specialize in handling instructions or data.
• Cache prioritizes frequently used program instructions and data to maximize execution speed.

Multi-Level Cache

• Two key ideas underpin cache efficiency, spatial locality—the next piece of information is near the previously processed one (important idea for programming efficiency)—and temporal locality—the tendency to reuse recently processed information (e.g., loops).

Mass Storage

• Diagram illustrates a comparison between the addressing of main memory and cache memory.
• Main memory utilizes address-based access.