📚 Computer Architecture Key Points for CCEA A-Level | CCEA A-Level 计算机：计算机体系结构考点精讲
Understanding computer architecture is fundamental to A-Level Computer Science, especially for the CCEA specification. This article breaks down the essential concepts you need to master, from the classic Von Neumann model to modern performance considerations, memory hierarchies, pipelining, and I/O techniques. By the end, you will have a clear, exam-ready understanding of how a computer's internal components are organised and how they interact to execute programs efficiently.
理解计算机体系结构是 A-Level 计算机科学的基础，尤其是 CCEA 考试大纲的重中之重。本文分解了你需要掌握的核心概念，从经典的冯·诺依曼模型到现代性能考量、存储层次、流水线技术以及输入输出方法。读完本文，你将清晰且具备应试能力地理解计算机内部组件如何组织，以及它们如何协同高效地执行程序。
1. Introduction to Computer Architecture | 计算机体系结构概论
Computer architecture defines the structure, behaviour, and design of a computer system. It encompasses the instruction set, memory addressing, and the interconnection of major hardware components.
计算机体系结构定义了计算机系统的结构、行为和设计，它包含指令集、内存寻址以及主要硬件组件之间的互连方式。
At this level, we distinguish between architecture (the programmer's view, e.g. x86, ARM) and organisation (the actual hardware implementation of that architecture). The CCEA syllabus focuses on both the logical layout and the physical elements that support execution.
在这个层面上，我们要区分体系结构（程序员的视角，例如 x86、ARM）和组织（该体系结构的实际硬件实现）。CCEA 教学大纲同时关注逻辑布局和支撑执行过程的物理元件。
2. The Von Neumann Architecture | 冯·诺依曼体系结构
The Von Neumann architecture is the foundational design for most modern computers. It stores both program instructions and data in a single read-write memory.
冯·诺依曼体系结构是大多数现代计算机的基础设计，它将程序指令和数据存储在同一个可读写内存中。
This unified memory approach creates the Von Neumann bottleneck, where the speed of the CPU is limited by the rate at which data and instructions can be transferred between memory and the processor. The key features include a control unit, an arithmetic logic unit (ALU), a memory unit, and input/output mechanisms all connected via buses.
这种统一内存的方案造成了冯·诺依曼瓶颈，即 CPU 的速度受限于内存与处理器之间数据和指令的传输速率。其关键特征包括控制单元、算术逻辑单元 (ALU)、存储单元以及通过总线连接的输入/输出机制。
In contrast, the Harvard architecture uses separate storage and pathways for instructions and data, allowing simultaneous access and overcoming the bottleneck to some extent. While pure Harvard designs appear in embedded systems, modern PCs often adopt a modified Harvard architecture where caches are split but main memory is shared.
相比之下，哈佛结构为指令和数据使用独立的存储和通路，允许同时访问并在一定程度上克服了瓶颈。纯哈佛设计常见于嵌入式系统，而现代个人计算机通常采用改进的哈佛结构，其中高速缓存是分开的，但主存是共享的。
3. CPU Components and Functions | CPU 组件与功能
The Central Processing Unit (CPU) is the brain of the computer. Its primary components are the Control Unit (CU), the Arithmetic Logic Unit (ALU), and a set of registers.
中央处理器 (CPU) 是计算机的大脑，其主要组件包括控制单元 (CU)、算术逻辑单元 (ALU) 和一组寄存器。
The Control Unit decodes instructions and directs the flow of data by generating control signals. The ALU performs all arithmetic operations (addition, subtraction, etc.) and logical operations (AND, OR, NOT, comparisons).
控制单元对指令进行译码，并通过生成控制信号来指挥数据流。ALU 执行所有算术运算（加法、减法等）和逻辑运算（与、或、非、比较）。
Registers are small, high-speed storage locations inside the CPU. Crucial registers for the CCEA syllabus include:
- Program Counter (PC): holds the address of the next instruction to fetch.
- Memory Address Register (MAR): holds the address of the memory location to be accessed.
- Memory Data Register (MDR): holds the data being transferred to or from memory.
- Current Instruction Register (CIR): stores the instruction currently being decoded and executed.
- Accumulator (ACC): temporarily stores results from the ALU.
寄存器是 CPU 内部的小型高速存储位置。CCEA 教学大纲中重要的寄存器包括：
- 程序计数器 (PC)：存放下一条要取指的指令地址。
- 内存地址寄存器 (MAR)：存放要访问的内存位置的地址。
- 内存数据寄存器 (MDR)：存放传入或传出内存的数据。
- 当前指令寄存器 (CIR)：存放当前正在译码和执行的指令。
- 累加器 (ACC)：临时存放来自 ALU 的结果。
Additionally, the Status Register (or Flags) stores bits indicating conditions such as zero, carry, or overflow. These flags are essential for conditional branching.
此外，状态寄存器（或标志位）存储指示零、进位或溢出等条件的位。这些标志对条件分支至关重要。
4. The Fetch-Decode-Execute Cycle | 取指-译码-执行周期
Every instruction is processed through a continuous cycle of fetch, decode, and execute. This sequential process is the heartbeat of the Von Neumann CPU.
每条指令都通过取指、译码、执行这一连续周期来处理。这个顺序过程是冯·诺依曼 CPU 的心跳。
During the Fetch stage:
- The address in the PC is copied to the MAR.
- The PC is incremented to point to the next instruction.
- The control unit issues a read signal, and the instruction from memory is placed into the MDR.
- The instruction is then copied to the CIR.
在取指阶段：
- PC 中的地址被复制到 MAR。
- PC 自增以指向下一条指令。
- 控制单元发出读信号，内存中的指令被放入 MDR。
- 随后指令被复制到 CIR。
During the Decode stage, the control unit interprets the opcode and addressing mode in the CIR, breaking down the instruction into the necessary micro-operations.
在译码阶段，控制单元解读 CIR 中的操作码和寻址模式，将指令分解为必要的微操作。
During the Execute stage, the ALU performs the required operation. This might involve reading operands from registers or memory, performing an arithmetic calculation, writing a result back to a register, or altering the PC for a branch.
在执行阶段，ALU 执行所需的操作。这可能包括从寄存器或内存读取操作数、执行算术计算、将结果写回寄存器，或为分支而更改 PC。
An interrupt may be checked at the end of each cycle. If an interrupt is pending, the CPU saves its state and services the interrupt before returning to the next fetch cycle.
每个周期结束时可能会检查中断。如果有一个中断待处理，CPU 会保存其状态并服务该中断，然后返回到下一个取指周期。
5. Factors Affecting CPU Performance | 影响 CPU 性能的因素
Several hardware characteristics determine how quickly a CPU can process instructions. The CCEA exam expects you to explain the following core factors.
多个硬件特性决定了 CPU 处理指令的速度。CCEA 考试要求你解释以下核心因素。
Clock Speed memory is fast enough. For instance, a 3.5 GHz processor runs 3.5 × 10⁹ cycles per second.
时钟速度：以赫兹为单位，指每秒的周期数。更高的时钟速度意味着每秒更多的取指-译码-执行周期，如果内存足够快，将直接提升吞吐量。例如，一个 3.5 GHz 的处理器每秒运行 3.5 × 10⁹ 个周期。
The period of one cycle T is given by
T = 1/f where f is the frequency in Hz.
一个周期的周期 T 由下式给出
T = 1/f，其中 f 是以 Hz 为单位的频率。
Number of Cores: a core is an independent processing unit within the CPU. Dual-core, quad-core, or many-core processors can execute multiple instructions simultaneously if the software is multi-threaded, improving multitasking and parallel performance. However, not all programs can be parallelised, so doubling cores does not double speed for all tasks.
核心数量：核心是 CPU 内部独立的处理单元。双核、四核或多核处理器可以在软件多线程时同时执行多条指令，从而提升多任务和并行性能。然而，并非所有程序都能并行化，因此核心数加倍并不会让所有任务的速度翻倍。
Cache Size and Levels: cache is a small amount of very fast memory located on the CPU chip. A larger L1, L2, or L3 cache reduces the need to access slower main memory. More cache means more data and instructions can be prefetched and kept close to the CPU, reducing the average memory access time.
缓存大小与层级：缓存是位于 CPU 芯片上的少量极快速内存。更大的 L1、L2 或 L3 缓存减少了对慢速主存的访问需求。更大的缓存意味着可以预取更多的数据和指令并将其保持在 CPU 附近，从而减少平均内存访问时间。
Other factors such as bus width, pipelining, and instruction set design (RISC vs CISC) also play roles but are often treated under separate topics.
总线宽度、流水线技术和指令集设计（RISC 与 CISC）等其他因素也起作用，但通常被放在单独的专题中讨论。
6. Memory Hierarchy | 存储层次结构
The memory hierarchy organises storage technologies by speed, capacity, and cost. The top is fastest and most expensive per byte, while the bottom is slowest and cheapest.
存储层次结构根据速度、容量和成本组织存储技术。顶层最快且每字节最贵，底层最慢且最便宜。
| Level | Technology | Typical Size | Access Time |
|---|---|---|---|
| Registers | Flip-flops inside CPU | Few hundred bytes | ~1 ns |
| L1/L2 Cache | SRAM | KB to a few MB | ~1-5 ns |
| Main Memory | DRAM | GB | ~50-100 ns |
| Secondary Storage | SSD / HDD | Hundreds of GB to TB | ~0.1-10 ms |
The principle of locality is exploited: temporal locality (recently accessed items will be accessed again soon) and spatial locality (items near a recently accessed item will likely be accessed). This allows the cache to be effective, bridging the speed gap between CPU and main memory.
利用了局部性原理：时间局部性（最近访问过的项不久后可能再次访问）和空间局部性（靠近最近访问项的附近项可能被访问）。这使得缓存能够有效地弥合 CPU 与主存之间的速度差距。
7. Cache Memory | 高速缓存
Cache memory is a small, high-speed buffer between the CPU and main memory. It stores copies of frequently used data and instructions to accelerate access.
高速缓存是 CPU 与主存之间的一个小型高速缓冲区，它存储常用数据和指令的副本以加速访问。
When the CPU needs data, it first checks the cache. If the data is found (a cache hit), it is quickly supplied. If not (a cache miss), a block containing the required data is fetched from main memory and placed in the cache, potentially evicting an existing block.
当 CPU 需要数据时，首先检查缓存。如果找到数据（缓存命中），则快速提供。如果没有找到（缓存未命中），则从主存中取出包含所需数据的块并将其放入缓存，可能驱逐现有块。
The hit rate h is the proportion of memory accesses found in the cache. The average access time can be modelled as:
ess Time = h × Cache Access Time + (1 - h) × (Main Memory Access Time + Cache Access Time)
命中率 h 是在缓存中找到的内存访问的比例。平均访问时间可建模为：
平均访问时间 = h × 缓存访问时间 + (1 - h) × (主存访问时间 + 缓存访问时间)
CCEA requires knowledge of cache mapping schemes at a conceptual level:
- Direct Mapped: each memory block maps to exactly one cache line. Simple but may cause conflicts.
- Fully Associative: a block can be placed in any cache line. Flexible but requires complex searching.
- Set Associative: a compromise where each block maps to a set of lines; usually n-way set associative.
CCEA 要求在概念层次上理解缓存映射方案：
- 直接映射：每个内存块映射到唯一的一个缓存行。简单，但可能引起冲突。
- 全相联：一个块可以放在任意缓存行。灵活，但需要复杂的搜索。
- 组相联：一种折衷方案，每个块映射到一组行；通常是 n 路组相联。
Replacement policies like Least Recently Used (LRU) and write policies like write-through and write-back are also relevant. In write-through, data is written to both cache and main memory simultaneously. In write-back, writes are done only to cache initially, and main memory is updated only when the block is replaced.
替换策略如最近最少使用 (LRU) 和写入策略如写直达和写回也相关。在写直达中，数据同时写入缓存和主存。在写回中，写入最初只对缓存进行，只有当块被替换时才更新主存。
8. CISC vs RISC Architectures | CISC 与 RISC 体系结构
Instruction Set Architecture (ISA) design philosophies are broadly classified into CISC (Complex Instruction Set Computer) and RISC (Reduced Instruction Set Computer).
指令集架构 (ISA) 设计理念大致分为 CISC（复杂指令集计算机）和 RISC（精简指令集计算机）。
| Feature | CISC | RISC |
|---|---|---|
| Instruction Set | Large, complex instructions | Small, simple instructions |
| Instruction Execution | Multi-cycle, microcode | Single-cycle (pipelined) |
| Memory Access | Instructions can access memory directly | Only load/store instructions access memory |
| Number of Registers | Fewer registers | Many general-purpose registers |
| Pipelining | Difficult due to variable instruction lengths | Easier due to fixed-length instructions |
| Examples | x86, VAX | ARM, MIPS, PowerPC |
CISC aims to make the compiler's job easier by providing powerful instructions that can perform multiple low-level operations. RISC, on the other hand, makes hardware simpler and enables aggressive pipelining, leading to higher clock speeds and lower power consumption per instruction.
CISC 旨在通过提供能执行多个底层操作的强大指令来减轻编译器的负担。而 RISC 则使硬件更简单，并允许积极的流水线设计，从而实现更高的时钟速度和每条指令更低的功耗。
In modern practice, many CISC processors (like Intel Core) internally translate CISC instructions into RISC-like micro-operations, blending philosophies. However, the theoretical distinction remains important for exams.
在现代实践中，许多 CISC 处理器（如 Intel Core）在内部将 CISC 指令翻译成类 RISC 的微操作，融合了两种理念。然而，理论上的区别对于考试仍然重要。
9. Pipelining | 流水线技术
Pipelining is a technique used to improve CPU throughput by overlapping the execution of multiple instructions. Think of it as an assembly line for instructions.
流水线是一种通过重叠执行多条指令来提高 CPU 吞吐量的技术。可以将其视为指令的流水生产线。
In a typical five-stage pipeline, the stages are: Instruction Fetch (IF), Instruction Decode (ID), Execute (EX), Memory Access (MEM), and Write Back (WB). While one instruction is being decoded, the next instruction can already be fetched.
在典型的五级流水线中，各个阶段为：取指 (IF)、译码 (ID)、执行 (EX)、访问内存 (MEM) 和写回 (WB)。当一条指令正在译码时，下一条指令已经可以开始取指。
The ideal throughput approaches one instruction per clock cycle, even though each instruction takes multiple cycles to complete. However, pipelining introduces hazards:
- Data Hazards: when an instruction depends on the result of a previous instruction not yet completed. Solved with forwarding (bypassing) or stalls (bubbles).
- Control Hazards: caused by branches; the next instruction cannot be determined until the branch resolves. Solved by branch prediction or flushing the pipeline.
- Structural Hazards: when two instructions require the same hardware resource simultaneously. Solved by duplicating resources or stalling.
理想吞吐量接近每个时钟周期完成一条指令，即使每条指令需要多个周期才能完成。然而，流水线引入了以下冒险：
- 数据冒险：当一条指令依赖于尚未完成的前一条指令的结果时发生。通过转发（旁路）或停顿（气泡）来解决。
- 控制冒险：由分支引起；在分支确定之前无法确定下一条指令。通过分支预测或刷新流水线来解决。
- 结构冒险：当两条指令同时需要相同的硬件资源时。通过复制资源或停顿来解决。
The CCEA specification expects you to understand the concept of pipeline stages and the impact of hazards, rather than designing complex pipeline logic.
CCEA 规范期望你理解流水线阶段的概念和冒险的影响，而不是设计复杂的流水线逻辑。
10. Input/Output Systems and Interrupts | 输入输出系统与中断
For a computer to interact with external devices, efficient I/O mechanisms are essential. Three main techniques are covered: Programmed I/O, Interrupt-driven I/O, and Direct Memory Access (DMA).
计算机要与外部设备交互，高效的 I/O 机制必不可少。涵盖了三种主要技术：程序控制 I/O、中断驱动 I/O 和直接存储器访问 (DMA)。
In Programmed I/O, the CPU continuously polls the status register of a device until it is ready. This is simple but wastes CPU cycles, making it inefficient for high-speed devices.
在程序控制 I/O 中，CPU 持续轮询设备的状态寄存器直到设备就绪。这种方法简单，但浪费 CPU 周期，对高速设备效率低下。
Interrupt-driven I/O allows the CPU to issue a command and then continue with other tasks. When the device is ready, it sends an interrupt signal to the CPU. The CPU finishes its current instruction, saves its state, and jumps to an Interrupt Service Routine (ISR). After servicing the device, it restores its state and resumes. This method improves efficiency but still involves the CPU for every data transfer.
中断驱动 I/O 允许 CPU 发出命令后继续执行其他任务。当设备就绪时，它向 CPU 发送中断信号。CPU 完成当前指令，保存其状态，并跳转到中断服务程序 (ISR)。服务完设备后，恢复状态并继续执行。这种方法提高了效率，但每次数据传输仍然涉及 CPU。
Direct Memory Access (DMA) further offloads the CPU. A DMA controller takes over the system buses to transfer a block of data directly between memory and an I/O device without CPU intervention, except for setting up and completing the transfer. DMA is vital for high-bandwidth devices like disk drives and graphic cards.
直接存储器访问 (DMA) 进一步减轻了 CPU 的负担。DMA 控制器接管系统总线，直接在内存和 I/O 设备之间传输数据块，除了设置和完成传输外无需 CPU 干预。DMA 对于磁盘驱动器和显卡等高带宽设备至关重要。
Interrupts themselves come in various types: maskable (can be ignored), non-maskable (cannot be ignored for critical events), and software interrupts generated by programs. The interrupt vector table holds the starting addresses of ISRs.
中断本身有多种类型：可屏蔽中断（可被忽略）、不可屏蔽中断（用于关键事件，不可忽略）以及由程序产生的软件中断。中断向量表保存了各个 ISR 的起始地址。
11. Buses and Data Transfer | 总线与数据传输
A bus is a communication pathway that connects the CPU, memory, and I/O devices. The system bus is typically divided into three separate buses: the address bus, the data bus, and the control bus.
总线是连接 CPU、内存和 I/O 设备的通信通路。系统总线通常分为三个独立的总线：地址总线、数据总线和控制总线。
The address bus carries memory addresses from the CPU to memory or an I/O device. It is unidirectional from the CPU's perspective. Its width determines the maximum addressable memory space — an n‑bit address bus can address 2ⁿ memory locations.
地址总线 将内存地址从 CPU 传送到内存或 I/O 设备。从 CPU 的角度看它是单向的。其宽度决定了最大可寻址内存空间——一条 n 位地址总线可以寻址 2ⁿ 个存储单元。
The data bus carries actual data between the CPU, memory, and I/O. It is bidirectional, allowing read and write operations. Its width (e.g. 32-bit, 64-bit) determines how much data can be transferred per cycle.
数据总线 在 CPU、内存和 I/O 之间传输实际数据。它是双向的，允许读写操作。其宽度（如 32 位、64 位）决定了每个周期能传输的数据量。
The control bus is a collection of lines that carry control and timing signals, such as memory read, memory write, I/O read, I/O write, interrupt request, and clock signals. These signals orchestrate the entire operation of the system.
控制总线 是一组承载控制和定时信号的线路，例如内存读、内存写、I/O 读、I/O 写、中断请求和时钟信号。这些信号协调整个系统的运作。
Modern systems may use multiple buses arranged in a hierarchy, with a faster front-side bus connecting the CPU to the chipset and slower buses (e.g. PCIe, SATA, USB) for peripherals. A bus arbitration mechanism decides which device gains control of the bus when multiple devices request access.
现代系统可能使用分层布置的多条总线，较快的北桥总线将 CPU 连接到芯片组，较慢的总线（如 PCIe、SATA、USB）用于外设。总线仲裁机制决定当多个设备请求访问时哪一个设备获得总线控制权。
Published by TutorHao | Computer Science Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图