Computer Architecture: Key Concepts for A-Level | 计算机体系结构：A-Level考点精讲

📚 Computer Architecture: Key Concepts for A-Level | 计算机体系结构：A-Level考点精讲

Computer architecture forms the foundation of how a computer system is organised and operates at the hardware level. Understanding the components, buses, CPU cycles, and performance factors is essential for A-Level success. This guide breaks down each key topic with clear explanations in both English and Chinese.

计算机体系结构是计算机系统在硬件层面如何组织和运行的基础。理解组件、总线、CPU周期和性能因素对于A-Level考试至关重要。本文将以清晰的中英双语解释逐一剖析各个重要考点。

1. The von Neumann Architecture | 冯·诺依曼体系结构

The von Neumann architecture describes a system where instructions and data share the same memory space and are accessed over a single bus. This stored-program concept means programs can be loaded into memory, executed, and modified like data.

冯·诺依曼体系结构描述了一种指令和数据共享同一存储空间并通过单一总线访问的系统。这种存储程序概念意味着程序可以像数据一样被加载到内存中、执行和修改。

The key limitation is the ‘von Neumann bottleneck’: because instructions and data compete for the same bus, the CPU can be forced to wait, limiting overall speed.

其关键局限在于“冯·诺依曼瓶颈”：由于指令和数据争用同一条总线，CPU可能被迫等待，从而限制了整体速度。

A contrasting architecture is Harvard, which uses separate memory and buses for instructions and data, allowing simultaneous access and speeding up execution. Most microcontrollers use a modified Harvard architecture.

与之对比的是哈佛体系结构，它为指令和数据使用独立的存储器和总线，允许同时访问，从而加速执行。大多数微控制器采用改进型的哈佛架构。

2. Main Components: CPU, Memory, I/O | 主要组件：CPU、存储器、输入/输出

A computer system can be divided into three fundamental subsystems: the Central Processing Unit (CPU) for processing, main memory (RAM) for temporary storage, and Input/Output (I/O) devices for communication with the outside world.

计算机系统可以分为三个基本子系统：用于处理的中央处理器 (CPU)、用于临时存储的主存储器 (RAM) 以及用于与外界通信的输入/输出 (I/O) 设备。

The CPU contains the arithmetic logic unit (ALU), control unit (CU), and registers. Memory stores both data and instructions currently in use. I/O devices include keyboards, mice, monitors, disks, and network interfaces. These components communicate via system buses.

CPU包含算术逻辑单元(ALU)、控制单元(CU)和寄存器。内存存储当前正在使用的数据和指令。I/O设备包括键盘、鼠标、显示器、磁盘和网络接口。这些组件通过系统总线进行通信。

CPU executes instructions and controls data flow.
Memory holds the program and data while the program is running.
I/O enables interaction with users and persistent storage.

CPU 执行指令并控制数据流。
内存在程序运行时保存程序和数据。
I/O 实现与用户交互和持久存储。

3. System Buses: Data, Address, Control | 系统总线：数据总线、地址总线、控制总线

A bus is a set of parallel wires that connect the CPU, memory, and I/O controllers. Three main buses work together: the address bus carries memory locations, the data bus carries the actual data or instructions, and the control bus transmits timing and command signals.

总线是一组连接 CPU、存储器和 I/O 控制器的并行线路。三条主要总线协同工作：地址总线传送内存位置，数据总线传送实际数据或指令，控制总线传送时序和命令信号。

The width of the address bus determines the maximum addressable memory (e.g., 32 wires can address 2³² locations). The data bus width defines how many bits can be transferred simultaneously, affecting system performance. The control bus includes signals such as read, write, interrupt requests, and clock timing.

地址总线的宽度决定了最大可寻址内存（例如，32条线可寻址2³²个位置）。数据总线的宽度定义了每次可以同时传输多少位，影响系统性能。控制总线包括读、写、中断请求和时钟时序等信号。

Bus	Function	Direction
Address	Carries memory addresses	Unidirectional (CPU to memory/I/O)
Data	Transfers instructions and data	Bidirectional
Control	Sends operation signals and timing	Bidirectional (each signal has a fixed direction)

4. The Fetch-Decode-Execute Cycle | 取指-解码-执行循环

The CPU repeatedly follows the fetch-decode-execute cycle for every instruction. During fetch, the next instruction’s address is taken from the Program Counter (PC), placed on the address bus, and the instruction is loaded into the Instruction Register (IR). The PC is then incremented.

CPU 对每条指令不断重复取指-解码-执行周期。在取指阶段，下一条指令的地址从程序计数器 (PC) 取出，放到地址总线上，指令被加载到指令寄存器 (IR)。然后 PC 递增。

Decode interprets the operation code (opcode) and determines the operands needed. The control unit activates the relevant circuits. Execute carries out the operation – this may involve the ALU, memory access, or a branch operation that updates the PC. After execution, the cycle restarts unless an interrupt occurs.

解码解释操作码并确定所需的操作数。控制单元激活相关电路。执行阶段执行操作——这可能涉及 ALU、内存访问或更新 PC 的分支操作。执行后，除非发生中断，否则周期重新开始。

A typical cycle can be summarised as: PC → MAR → Read memory → MDR → IR → decode → execute → (optional store result).

一个典型周期可总结为：PC → MAR → 读内存 → MDR → IR → 解码 → 执行 → （可选存储结果）。

5. CPU Components: ALU, CU, Registers | CPU组件：算术逻辑单元、控制单元、寄存器

The Arithmetic Logic Unit (ALU) performs arithmetic operations (addition, subtraction, etc.) and logical operations (AND, OR, NOT, shifts). It receives operands from registers and outputs results back to a register.

算术逻辑单元 (ALU) 执行算术运算（加、减等）和逻辑运算（与、或、非、移位）。它从寄存器接收操作数并将结果输出到寄存器。

The Control Unit (CU) orchestrates the fetch-decode-execute cycle by issuing timing and control signals. It decodes instructions and directs data movement between components. Registers are small, ultra-fast storage locations inside the CPU. Key registers include the PC, IR, MAR (Memory Address Register), MDR (Memory Data Register), ACC (Accumulator), and general-purpose registers.

控制单元 (CU) 通过发出时序和控制信号来协调取指-解码-执行周期。它解码指令并指导组件之间的数据移动。寄存器是 CPU 内部极小且超快的存储位置。关键寄存器包括 PC、IR、MAR（内存地址寄存器）、MDR（内存数据寄存器）、ACC（累加器）和通用寄存器。

PC: holds address of next instruction
MAR: holds address to be read from/written to memory
MDR: holds data read from or to be written to memory
CIR/IR: holds current instruction

PC：保存下一条指令的地址
MAR：保存要从内存读写的内存地址
MDR：保存从内存读出的或要写入内存的数据
CIR/IR：保存当前指令

6. Performance Factors: Clock Speed, Cores, Cache | 性能因素：时钟速度、核心数、高速缓存

CPU performance is influenced by multiple factors that students must be able to explain. Clock speed, measured in Hertz (Hz), determines how many fetch-decode-execute cycles can be performed per second. A 3.5 GHz processor runs 3.5 × 10⁹ cycles per second.

CPU 性能受多个因素影响，学生必须能够解释这些因素。时钟速度以赫兹 (Hz) 为单位，决定了每秒可以执行多少个取指-解码-执行周期。一个 3.5 GHz 的处理器每秒运行 3.5 × 10⁹ 个周期。

Number of cores allows multiple instructions to be processed in parallel. A quad-core processor can handle four threads simultaneously, but the performance gain depends on software optimisation. Cache memory is high-speed memory located close to the CPU that stores frequently accessed data and instructions, reducing the need to access slower main memory. Levels L1 (fastest, smallest), L2, and sometimes L3 provide a hierarchy of speed vs size.

核心数允许多条指令并行处理。四核处理器可以同时处理四个线程，但性能提升取决于软件优化。高速缓存是位于 CPU 附近的高速存储器，存储频繁访问的数据和指令，从而减少访问较慢的主存储器的需要。L1（最快、最小）、L2、有时还有 L3 提供了速度与容量的层次结构。

Execution time for a program can be approximated by: Execution time = (Number of instructions × CPI) / Clock rate, where CPI is cycles per instruction. A lower CPI and higher clock rate reduce execution time.

程序的执行时间可以近似为：执行时间 = （指令数 × CPI）/ 时钟频率，其中 CPI 为每条指令所需周期数。较低的 CPI 和较高的时钟频率会减少执行时间。

7. Pipelining and Parallelism | 流水线与并行处理

Pipelining improves instruction throughput by overlapping the execution of multiple instructions. While one instruction is being decoded, the next can be fetched. A typical 5-stage pipeline includes Fetch, Decode, Execute, Memory access, and Write-back (FDEMW).

流水线通过重叠多条指令的执行来提高指令吞吐量。当一条指令正在解码时，下一条指令可以被取出。典型的 5 级流水线包括取指、解码、执行、存储器访问和写回 (FDEMW)。

Hazards can stall the pipeline: data hazards occur when an instruction depends on a previous result not yet ready; control hazards arise from branches changing the flow; structural hazards happen when hardware resources are insufficient. Techniques like branch prediction and data forwarding help mitigate these stalls.

冒险可能导致流水线停顿：数据冒险发生在指令依赖于尚未准备好的前一条结果时；控制冒险由改变流程的分支引起；结构冒险在硬件资源不足时发生。分支预测和数据转发等技术有助于减轻这些停顿。

Parallel architectures like SIMD (Single Instruction, Multiple Data) and MIMD (Multiple Instruction, Multiple Data) extend this idea to multiple processing elements, enabling vector processing or multi-core parallelism used in modern CPUs and GPUs.

SIMD（单指令多数据）和 MIMD（多指令多数据）等并行架构将这一思想扩展到多个处理单元，实现了现代 CPU 和 GPU 中使用的向量处理或多核并行。

8. Instruction Set Architecture (RISC vs CISC) | 指令集体系结构（RISC 与 CISC）

The instruction set is the collection of commands a CPU can execute. Two dominant design philosophies exist: RISC (Reduced Instruction Set Computer) uses a small set of simple, fixed-length instructions that execute in one clock cycle. CISC (Complex Instruction Set Computer) has a larger set of more powerful, variable-length instructions that may take multiple cycles.

指令集是 CPU 可以执行的命令集合。存在两种主流设计理念：RISC（精简指令集计算机）使用少量、简单的定长指令，在一个时钟周期内执行。CISC（复杂指令集计算机）拥有更大、功能更强、变长的指令集，可能需要多个周期。

RISC architectures (e.g., ARM) rely on software for complex operations, use many general-purpose registers, and enable efficient pipelining. CISC (e.g., x86) allows complex operations like string manipulation to be done with a single instruction, reducing the number of instructions per program at the cost of more complex hardware and microcoding.

RISC 体系结构（如 ARM）依赖软件完成复杂操作，使用多个通用寄存器，并实现高效流水线。CISC（如 x86）允许使用单一指令完成字符串操作等复杂操作，减少了每个程序的指令数，但代价是更复杂的硬件和微码。

Feature	RISC	CISC
Instruction size	Fixed (e.g., 32-bit)	Variable
Cycles per instruction	Typically 1	Multiple
Addressing modes	Fewer	Many
Pipeline ease	Easier	More difficult
Example	ARM, MIPS	x86, 68k

9. Memory Hierarchy and Cache Types | 存储层次与缓存类型

Memory hierarchy organises storage from fast but small and expensive (registers, cache) to slow but large and cheap (main RAM, disk). Registers are at the top, followed by cache (L1, L2, L3), main memory (RAM), and secondary storage (SSD/HDD). The goal is to minimise access time and maximise capacity at an acceptable cost.

存储层次将从快速但容量小、价格贵的存储（寄存器、高速缓存）到慢速但容量大、价格便宜的存储（主存、磁盘）进行组织。寄存器位于顶端，接着是高速缓存（L1、L2、L3）、主存 (RAM) 和辅助存储 (SSD/HDD)。其目标是在可接受的成本下最小化访问时间并最大化容量。

Caches exploit the principle of locality: temporal locality (recently accessed data is likely to be accessed again soon) and spatial locality (nearby addresses are likely to be accessed soon). Cache is organised into blocks or lines, and a cache hit occurs when the requested data is found in cache; otherwise, a cache miss requires fetching from slower memory.

高速缓存利用了局部性原理：时间局部性（最近访问的数据很可能再次被访问）和空间局部性（附近的地址很可能即将被访问）。缓存以块或行的形式组织，当请求的数据在缓存中找到时发生缓存命中；否则缓存缺失需要从较慢的存储器中取数据。

Direct-mapped cache: each memory block maps to exactly one cache line.
Fully associative cache: a block can be placed anywhere.
Set-associative cache: a compromise where each block maps to a small set of lines.

直接映射缓存：每个内存块恰好映射到一个缓存行。
全相联缓存：块可以放置在任意位置。
组相联缓存：折衷方案，每个块映射到一个小组行。

10. Interrupts and I/O Handling | 中断与输入/输出处理

An interrupt is a signal sent by hardware or software to the CPU that causes it to pause the current task and execute an interrupt service routine (ISR). After handling the interrupt, the CPU resumes the original task by restoring its state from the stack.

中断是由硬件或软件发送给 CPU 的信号，使其暂停当前任务并执行中断服务程序 (ISR)。处理完中断后，CPU 通过从堆栈恢复状态来继续原来的任务。

Interrupts allow efficient I/O handling. Without interrupts, the CPU would need to poll devices repeatedly, wasting cycles. With interrupts, a device can asynchronously request service. Vectored interrupts automatically provide the ISR address, reducing latency. Prioritised interrupts ensure critical tasks are serviced first.

中断可以实现高效的 I/O 处理。没有中断，CPU 就需要反复轮询设备，浪费周期。有了中断，设备可以异步请求服务。向量化中断自动提供 ISR 地址，减少延迟。优先级中断确保关键任务首先得到服务。

Direct Memory Access (DMA) is another I/O technique where a dedicated controller transfers blocks of data between memory and I/O devices without CPU intervention, freeing the CPU for other tasks during the transfer.

直接存储器访问 (DMA) 是另一种 I/O 技术，专用控制器在存储器和 I/O 设备之间传输数据块而无需 CPU 干预，在传输过程中释放 CPU 执行其他任务。

11. Embedded Systems and Their Architectures | 嵌入式系统及其体系结构

An embedded system is a dedicated computer system designed to perform one or a few specific functions, often with real-time constraints. Examples include microcontrollers in washing machines, automotive control units, and smart sensors. They are typically resource-constrained in terms of memory, power, and processing capability.

嵌入式系统是一种专用于执行一个或几个特定功能的计算机系统，通常具有实时约束。例子包括洗衣机中的微控制器、汽车控制单元和智能传感器。它们通常在内存、功耗和处理能力方面受到限制。

Embedded systems often use Harvard or modified Harvard architectures because separating instruction and data memories increases performance and security. They commonly employ SoC (System-on-Chip) designs, integrating the CPU, memory, timers, and I/O peripherals on a single chip. Low power consumption is critical, so many use RISC processors like ARM Cortex-M variants.

嵌入式系统通常采用哈佛或改良哈佛体系结构，因为分离指令和数据存储器可提高性能和安全性。它们通常采用 SoC（片上系统）设计，将 CPU、存储器、定时器和 I/O 外设集成在单个芯片上。低功耗至关重要，因此许多使用 ARM Cortex-M 变体等 RISC 处理器。

In an A-Level context, you may need to compare embedded systems with general-purpose computers, highlighting the differences in operating systems (often none or a real-time OS), user interface, and upgradeability.

在 A-Level 背景下，你可能需要比较嵌入式系统和通用计算机，重点突出操作系统（通常没有或使用实时操作系统）、用户界面和可升级性方面的差异。

12. Summary of Key Exam Points | 重点考点总结

To excel in architecture questions, ensure you can: explain the fetch-decode-execute cycle with register transfers; compare von Neumann vs Harvard and RISC vs CISC; describe the roles of buses and how their width affects performance; discuss how clock speed, cores, and cache interact to determine CPU performance; and explain pipelining and interrupts.

要在体系结构考题中取得好成绩，请确保你能够：用寄存器传输解释取指-解码-执行周期；比较冯·诺依曼与哈佛、RISC 与 CISC；描述总线的作用及其宽度如何影响性能；讨论时钟速度、核心数和缓存如何相互作用决定 CPU 性能；并解释流水线和中断。

Common mistakes include confusing MAR/MDR roles, forgetting that the von Neumann bottleneck arises from a shared bus, and misapplying performance equations. Remember that increasing clock speed alone does not guarantee linear performance improvement if the memory or bus cannot keep up.

常见错误包括混淆 MAR/MDR 的作用，忘记冯·诺依曼瓶颈源于共享总线，以及错误运用性能方程。请记住，如果内存或总线跟不上，仅提高时钟速度并不能保证线性性能提升。

Use precise terminology: ‘opcode’, ‘operand’, ‘pipeline hazard’, ‘cache hit/miss’, ‘ISR’, ‘DMA’. Diagrams can help in longer answer questions, so practising clear labelled diagrams of the CPU internals and bus connections is highly recommended.

使用精确的术语：’操作码’、’操作数’、’流水线冒险’、’缓存命中/缺失’、’ISR’、’DMA’。图解在较长的回答题中很有帮助，因此强烈建议练习清晰标注的 CPU 内部结构和总线连接图。

Published by TutorHao | Computer Science Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Computer Architecture: Key Concepts for A-Level | 计算机体系结构：A-Level考点精讲