A Level Computer Science 9618 — all 20 sections, free.

Three routes. AS Level only = Papers 1 and 2. A Level staged over two years = Papers 1 and 2 in year 1, Papers 3 and 4 in year 2. A Level in one series = all four papers.

1.1Data representation

Know denary, binary, hexadecimal and binary coded decimal (BCD), and convert freely between them. Understand why hexadecimal is used — one hex digit is exactly four bits, so it is a compact, human-readable shorthand for binary in MAC addresses, colour codes, memory dumps and error codes.

Two's complement represents signed integers: to negate, invert every bit and add one. An n-bit two's complement number covers −2ⁿ⁻¹ to +2ⁿ⁻¹ − 1, so 8 bits covers −128 to +127.

Character sets: ASCII uses 7 bits (128 characters), extended ASCII 8 bits; Unicode uses more bits per character so it can represent every writing system, at the cost of larger files.

1.2Multimedia — graphics and sound

Bitmap images store a colour value per pixel: resolution is the number of pixels, colour depth is the bits per pixel. Enlarging a bitmap causes pixellation. Vector images store a list of drawing objects with properties, so they scale without loss of quality and are usually much smaller — but they are unsuitable for photographs.

Sound is captured by sampling: the sampling rate is samples per second, the sampling resolution is bits per sample. Higher values give a more accurate reproduction and a larger file.

1.3Compression

Lossless compression allows the original file to be reconstructed exactly — run-length encoding, and dictionary methods used in ZIP and PNG. Essential for text, program code and spreadsheets. Lossy compression discards data permanently — JPEG, MP3 — giving far smaller files, acceptable for photographs and music where small losses are imperceptible.

2.1Networks including the internet

LAN covers a small geographic area under one organisation's control; WAN spans large distances using third-party infrastructure. Topologies: bus, star, mesh, hybrid — know an advantage and a disadvantage of each.

Hardware: a switch forwards frames only to the destination port using MAC addresses; a router connects different networks and forwards packets using IP addresses; a NIC gives a device its MAC address; a server provides services to clients.

Client–server centralises data and control, which makes backup and security easier but creates a single point of failure. Peer-to-peer shares resources directly between equal machines, with no central server.

The internet: IP addressing (IPv4 is 32-bit and running out; IPv6 is 128-bit), public versus private IP addresses, static versus dynamic addressing, URL structure, and the DNS which translates domain names into IP addresses. Understand the difference between the internet (the global network infrastructure) and the World Wide Web (a service running on it).

3.1Computers and their components

The von Neumann model: a processor, a single memory holding both data and instructions, and buses. Primary storage — RAM (volatile, working memory) and ROM (non-volatile, holds the bootstrap). Secondary storage — HDD (magnetic), SSD (flash, no moving parts, faster, more shock-resistant), optical discs. Off-line storage for backup and archive.

Match input and output devices to applications and justify the match: a barcode scanner at a supermarket checkout, a touch screen on a self-service kiosk, a 3D printer for prototyping.

3.2Logic gates and logic circuits

Build and interpret logic circuits, produce truth tables from circuits and circuits from problem statements or truth tables.

4.1CPU architecture

Components: ALU, control unit, system clock, immediate access store, and the registers — PC (program counter, address of the next instruction), MAR (memory address register), MDR (memory data register), CIR (current instruction register), ACC (accumulator), IX (index register), SR (status register).

Buses: the address bus is unidirectional and its width sets the maximum addressable memory; the data bus is bidirectional and its width affects the amount of data moved per cycle; the control bus carries signals such as read, write and interrupt.

Factors affecting performance: clock speed, number of cores, cache size and level, bus width. Explain each in terms of why it helps — more cache means fewer slow trips to main memory, more cores means genuine parallel execution.

Interrupts: a signal that causes the processor to suspend the current program, save its state on the stack, run the appropriate interrupt service routine, then restore the state and resume. The check for interrupts happens at the end of each fetch–execute cycle.

4.2Assembly language

Assembly is a low-level language with a one-to-one relationship to machine code, translated by an assembler. Learn the addressing modes: immediate (the operand is the value), direct (the operand is an address), indirect (the operand is the address of an address), indexed (the address plus the contents of IX), and relative.

You are given an instruction set in the exam. The skill is tracing a short program with a trace table, showing the contents of the accumulator, the index register and any memory locations after each instruction.

4.3Bit manipulation

Logical shifts move bits left or right, filling with zeros — a left shift multiplies by 2, a right shift divides by 2 (integer division). An arithmetic right shift preserves the sign bit. A cyclic shift moves the bit that falls off one end round to the other.

Bit masking uses AND to test or clear bits, OR to set bits, and XOR to toggle them — for example, AND with 0000 1111 keeps only the lower nibble.

5.1Operating systems

The OS manages memory, files, security, input/output, processes and the user interface, and provides a virtual machine — hiding hardware complexity so that applications and users need not deal with it directly.

Utility software: disk defragmenter, backup, file compression, anti-virus, formatting. The bootstrap in ROM loads the OS from secondary storage into RAM at start-up.

5.2Language translators

Integrated development environments support writing, editing and debugging: syntax highlighting, auto-completion, prettyprinting, breakpoints, single-stepping, variable watch windows and a report window.

6.1Data security

Threats: malware (virus, worm, trojan, spyware, ransomware), hacking, phishing, pharming, and accidental damage. Protection: firewalls, authentication (passwords, biometrics, two-factor), access rights, anti-malware, encryption, physical security, backups and a disaster recovery plan.

6.2Data integrity

Validation checks that data is reasonable and is done by the computer at input: range check, format check, length check, presence check, type check, existence check, limit check and check digit. Verification checks that data has been copied or entered accurately: double entry, visual check, and — for transfer — parity checks, checksums and Automatic Repeat reQuest (ARQ).

7.1Ethics and ownership

Professional ethics: the responsibilities of a computing professional — competence, honesty about capability, respect for privacy and confidentiality, avoiding harm, and acting in the public interest. Know the roles of professional bodies such as the BCS, IEEE and ACM.

Copyright and licensing: software is protected by copyright. Free software may be studied, changed and redistributed freely; open source makes the source code available under specified conditions; shareware is distributed free for a trial period; commercial software requires a paid licence. Compare their implications for the user and the developer.

8.1Database concepts

Limitations of a file-based approach: data duplication, data inconsistency, data dependence on the programs, and difficulty of controlling access. A relational database solves these through a single controlled store.

Key terms: entity, table, record, field, primary key (uniquely identifies a record), candidate key, composite key (two or more fields together), secondary key (an index for fast searching), foreign key (a field that links to a primary key in another table), and referential integrity (a foreign key must match an existing primary key or be null).

Relationships are one-to-one, one-to-many or many-to-many; a many-to-many relationship is resolved with a link table. Normalisation to 3NF: 1NF removes repeating groups, 2NF removes partial dependencies on part of a composite key, 3NF removes transitive (non-key) dependencies.

8.2Database management systems

A DBMS provides a data dictionary (metadata about the database), developer interface, query processor, and tools for creating forms, reports and queries. It enforces access rights and maintains integrity, so different applications share one consistent set of data.

8.3DDL and DML

DDL defines the structure: CREATE DATABASE, CREATE TABLE, ALTER TABLE, and the constraints PRIMARY KEY, FOREIGN KEY. DML manipulates the data: SELECT, FROM, WHERE, ORDER BY, GROUP BY, INNER JOIN, SUM, COUNT, AVG, INSERT INTO, UPDATE, DELETE.

9.1Computational thinking skills

Abstraction — keeping only the details relevant to the problem. Decomposition — breaking a problem into smaller sub-problems. Pattern recognition — spotting similarities that let one solution serve several cases. These are examined by asking you to apply them to an unfamiliar scenario, not to define them.

9.2Algorithms

Write and interpret algorithms as structured English, flowcharts and pseudocode, using input–process–output and the three basic constructs: sequence, selection and iteration.

Standard algorithms to know cold: linear search, binary search, bubble sort, insertion sort, and the routines for finding a total, a maximum, a minimum, a mean and a count.

10.1Data types and records

The pseudocode data types are INTEGER, REAL, CHAR, STRING, BOOLEAN and DATE. A record holds several related items of different types under one name.

10.2Arrays

An array holds many items of the same type under one identifier, accessed by index. Cambridge pseudocode declares the bounds explicitly.

Process arrays with FOR loops; use nested FOR loops for two dimensions. You must be able to write a bubble sort and both search algorithms over an array.

10.3Files

Handle text files in pseudocode: open for READ, WRITE or APPEND, read or write a line, test for end of file, then close.

10.4Introduction to abstract data types

Understand what each is used for — a stack for storing return addresses and for undo functions, a queue for print spooling and buffers, a linked list where frequent insertion and deletion are needed without shifting data. At AS you use them and describe them with diagrams; writing the pseudocode implementation is A2 material.

11.1Programming basics

Declaration and use of variables and constants, assignment, arithmetic operators (+ - * / DIV MOD ^), relational operators, logical operators (AND OR NOT), and the built-in functions for strings (LENGTH, LEFT, RIGHT, MID, LCASE, UCASE), numbers (INT, RAND) and type conversion (NUM_TO_STR, STR_TO_NUM).

11.2Constructs

Selection: IF … THEN … ELSE … ENDIF and CASE OF … OTHERWISE … ENDCASE. Iteration: FOR … NEXT (count-controlled), WHILE … ENDWHILE (pre-condition, may run zero times), REPEAT … UNTIL (post-condition, always runs at least once).

11.3Structured programming

Procedures perform a task; functions return a value and can be used inside an expression. Parameters may be passed by value (a copy — changes do not affect the caller) or by reference (the actual variable — changes do affect the caller).

Scope: a local variable exists only inside its subroutine; a global variable is accessible everywhere. Prefer local variables and parameters — they prevent accidental side effects and make modules independently testable.

Benefits of a modular approach: the problem is decomposed into manageable parts, modules can be written and tested independently and by different people, code can be reused, and maintenance is easier.

12.1Program development life cycle

Analysis → design → coding → testing → maintenance. Compare the waterfall model (sequential, well documented, hard to revisit an earlier stage), iterative development (repeated cycles of refinement) and rapid application development (prototypes and user feedback, fast but with lighter documentation).

12.2Program design

Design tools: structure charts (showing the hierarchy of modules and the parameters passed between them), state-transition diagrams, flowcharts and pseudocode. Be able to read one and produce another from it.

12.3Program testing and maintenance

Types of error: syntax (breaks the rules of the language — caught by the translator), logic (runs but gives the wrong result), run-time (crashes during execution, such as division by zero).

Test data: normal (typical, should be accepted), abnormal/erroneous (should be rejected), extreme (at the limits of the valid range) and boundary (the pair of values either side of a limit — one accepted and one rejected).

Testing strategies: white-box (testing every path through the code), black-box (comparing outputs with expected outputs for given inputs), stub testing, integration testing, alpha, beta and acceptance testing. Maintenance is corrective (fixing faults), adaptive (responding to a changed environment) and perfective (improving performance or usability).

13.1User-defined data types

Non-composite user-defined types: enumerated (a named list of possible values) and pointer (holding a memory address). Composite: record, set and class. Justify the use of a user-defined type — it makes the program self-documenting, restricts values to those that are valid, and groups related data.

13.2File organisation and access

A hashing algorithm converts a key into a storage address. A collision occurs when two keys hash to the same address; handle it with overflow areas or by linear probing to the next free slot.

13.3Floating-point numbers

A normalised positive mantissa begins 0.1; a normalised negative mantissa begins 1.0. Normalisation gives the maximum precision for the available bits and makes each value's representation unique.

The trade-off: for a fixed total number of bits, more mantissa bits give greater precision but fewer exponent bits give a smaller range. Be ready to argue that trade-off both ways.

14.1Protocols

A protocol is an agreed set of rules for communication. Protocols are arranged in a stack of layers, each providing a service to the layer above, so that a change in one layer does not force changes in the others.

BitTorrent is the standard peer-to-peer example: a tracker coordinates peers, a swarm shares pieces of the file, and seeds hold complete copies. It spreads bandwidth cost across the peers instead of loading one server.

14.2Circuit switching and packet switching

A packet has a header (source and destination addresses, packet number, protocol), a payload, and a trailer (error-check bits and an end marker).

15.1Processors, parallel processing and virtual machines

RISC versus CISC: RISC has few, simple, fixed-length instructions executing in about one cycle, more registers, and relies on the compiler; it pipelines well and uses less power. CISC has many complex variable-length instructions, needing fewer instructions per program but more cycles each.

Pipelining overlaps the stages of the fetch–execute cycle for successive instructions, so throughput rises without the clock speed changing.

Flynn's taxonomy: SISD, SIMD, MISD, MIMD. SIMD suits applying the same operation to a large data set (graphics, arrays); MIMD is the multi-core model. Massively parallel computers link many processors for problems such as climate modelling.

A virtual machine is software emulating a computer. Uses: running an operating system inside another, running legacy software, testing safely in isolation, and consolidating servers. Limitations: slower than native execution, and it needs substantial host resources.

15.2Boolean algebra and logic circuits

Simplify Boolean expressions algebraically and with Karnaugh maps: group adjacent 1s in blocks of 1, 2, 4 or 8, in Gray-code order, wrapping around the edges, using the largest groups possible and as few as possible.

Flip-flops: the SR flip-flop stores one bit and has an invalid state when both inputs are 1; the JK flip-flop fixes this by toggling instead, and is clocked. Flip-flops are the basis of registers and of static RAM.

16.1Purposes of an operating system

Memory management: paging (fixed-size pages), segmentation (variable-size logical segments), and virtual memory — using disk as an extension of RAM, swapping pages in and out. Disk thrashing occurs when the system spends more time swapping pages than executing, and it is a favourite exam question.

Process management: a process moves between ready, running and blocked states. The scheduler chooses which ready process runs next, using round robin, first come first served, shortest job first or shortest remaining time. Interrupts and interrupt priorities let the OS respond to events; the low-level scheduler performs the context switch.

16.2Translation software

Stages of compilation: lexical analysis (removing comments and whitespace, producing tokens and building the symbol table), syntax analysis (parsing against the grammar, reporting syntax errors), code generation (producing object code) and optimisation (making it smaller or faster).

Backus-Naur Form (BNF) and syntax diagrams define a language's grammar formally. You should be able to read a BNF definition and decide whether a given string is valid, and to write simple BNF rules.

17.1Encryption, encryption protocols and digital certificates

Symmetric encryption uses one shared key for both encryption and decryption: fast, but the key must somehow be exchanged securely. Asymmetric encryption uses a public key to encrypt and a mathematically related private key to decrypt: it solves key distribution but is much slower. In practice, asymmetric encryption is used to exchange a symmetric session key, and the bulk of the data is then encrypted symmetrically.

Digital signatures: the sender hashes the message to produce a digest and encrypts the digest with their private key. The recipient decrypts it with the sender's public key and compares it with their own hash of the message. Matching digests prove both authenticity (only the sender's private key could have made it) and integrity (the message has not been altered).

Digital certificates are issued by a Certificate Authority and bind a public key to an identity, preventing an attacker from substituting their own public key. SSL/TLS uses certificates and an asymmetric handshake to set up an encrypted session — this is what HTTPS is.

18.1Artificial intelligence

Graph representation: AI search problems are modelled as graphs of nodes and weighted edges, stored as an adjacency matrix (fast lookup, wasteful for sparse graphs) or an adjacency list (compact for sparse graphs).

A* algorithm finds the lowest-cost path using f(n) = g(n) + h(n), where g is the cost so far and h is a heuristic estimate of the cost remaining. Dijkstra's algorithm is the special case where h = 0 — it explores in all directions, so A* is usually faster when a good heuristic exists.

Machine learning: supervised learning trains on labelled data (classification and regression); unsupervised learning finds structure in unlabelled data (clustering); reinforcement learning learns from rewards and penalties. Deep learning uses artificial neural networks with many hidden layers.

Artificial neural networks: input, hidden and output layers of nodes with weighted connections; training adjusts the weights, typically by back propagation of the error. Applications include image recognition, speech recognition, medical diagnosis and autonomous vehicles.

19.1Algorithms (A2)

Now you must write the pseudocode implementations of the abstract data types, not just describe them: push and pop on a stack, enqueue and dequeue on a queue, and insert, delete and traverse on a linked list — using an array of nodes with a free list.

Binary trees: insert a value, and traverse in pre-order (node, left, right), in-order (left, node, right — which outputs a binary search tree in sorted order) and post-order (left, right, node).

Comparing algorithms: understand time and space efficiency in broad terms — linear search is O(n), binary search O(log n), bubble and insertion sort O(n²) — and be able to justify choosing one over another for a given data set.

19.2Recursion

A recursive routine calls itself. Every recursive definition needs a base case that stops the recursion and a general case that moves towards it. Recursion is implemented using a stack: each call pushes its parameters, local variables and return address, and they are popped as the calls unwind. Too deep a recursion causes stack overflow.

20.1Programming paradigms

Object-oriented programming. A class is a template defining properties (attributes) and methods; an object is an instance of a class, created by a constructor. The three principles:

1. Encapsulation — properties are PRIVATE and accessed only through PUBLIC get and set methods, so an object controls its own data and cannot be put into an invalid state.
2. Inheritance — a subclass inherits the properties and methods of its superclass and can add its own, so common code is written once.
3. Polymorphism — a subclass overrides an inherited method, so the same call produces the behaviour appropriate to the actual object.

Declarative programming in Prolog: facts such as parent(ali, sara)., rules such as grandparent(X,Z) :- parent(X,Y), parent(Y,Z)., and queries such as ?- grandparent(ali, Who). You state what is true, not how to compute it. Declarative programming is not examined in Paper 4 — only in Paper 3.

20.2File processing and exception handling

Random file access in pseudocode: OPENFILE … FOR RANDOM, SEEK to a record position, then GETRECORD or PUTRECORD. Serial and sequential files use READFILE and WRITEFILE as at AS.

Exception handling deals with run-time errors without crashing the program — a missing file, a division by zero, a non-numeric input. The structure is try / catch (or except) / finally.

PSConventions, declarations and constructs

Conventions: keywords in CAPITALS, identifiers in MixedCase, assignment with ←, comments after //, and consistent indentation for every block.

NBThe conversions you must be able to do on paper

Place values to memorise: 128 · 64 · 32 · 16 · 8 · 4 · 2 · 1 for a byte, and 4096 · 256 · 16 · 1 for four hexadecimal digits.

P4What to expect and how to prepare

2 hours 30 minutes, 75 marks, sat on a computer with no internet or email. It assesses sections 19 and 20 — except low-level and declarative programming, so no assembly and no Prolog. You submit complete program code and evidence of testing, and you work in Java, VB.NET or Python, all in console mode.

What you will be asked to implement, in your chosen language:

• Classes with private attributes, constructors, get and set methods, inheritance and overridden methods.
• Stacks, queues, linked lists and binary trees, with their operations.
• Recursive routines.
• Reading from and writing to text files, and random access to record files.
• Exception handling around file access, division and type conversion.
• Validation of input, with sensible messages.

Evidence of testing means screenshots or captured console output showing the program running with the data specified in the question. Marks are lost every session for code that works but has no evidence attached.

LearnThe definitions examiners want verbatim

Loading…

Clear my ticks

Official (free)

• Cambridge International — 9618 subject page: syllabus, specimen papers, past papers, mark schemes and examiner reports.
• Examiner reports name the exact questions candidates got wrong each series — read them for every paper you attempt.

Free archives

• GCE Guide · PastPapers.co — full CAIE past-paper archives.
• Physics & Maths Tutor — topic-sorted questions.

How to revise this subject

1. Write pseudocode by hand, every week. Paper 2 is answered entirely in pseudocode, and the only way to get fast at it is to write it — not read it.
2. Learn the official pseudocode conventions. Cambridge publishes a Pseudocode Guide; the reference section below covers the constructs you need. Arrays are declared with explicit bounds, assignment uses ←, and keywords are in capitals.
3. Trace tables win marks. When asked to dry-run an algorithm, draw the table, one column per variable, one row per iteration. Show every change.
4. Do the number conversions on paper. No calculator means you need the doubling method, the 16s-and-1s method for hex, and two's complement by "invert and add one".
5. Pick your Paper 4 language early and stick to it. Java, VB.NET or Python. Learn its file handling, exception handling and OOP syntax cold — not three languages badly.
6. Answer theory questions with the technical term. "Interrupt", "cache", "pipelining", "normalisation", "referential integrity", "encapsulation" — the term is usually the mark.