Ring Buffer

Overview

A ring buffer (circular buffer) is a fixed-size array that wraps around from end to beginning, treating the storage as a continuous loop. It enables efficient first-in-first-out (FIFO) operations without memory allocation, making it essential for audio processing, network packets, and game mechanics like snake body segments.

Fast Facts

Aspect	Detail
Also called	Circular buffer, cyclic buffer
Operations	O(1) read, O(1) write
Memory	Fixed, pre-allocated
Common uses	Audio, networking, input history

How It Works

Pointer	Purpose
Head (write)	Where the next write goes
Tail (read)	Where the next read comes from
Wrap	When a pointer reaches `size`, it wraps back to `0`

A write puts data at the head's current position and advances the head; a read takes data from the tail's position and advances the tail. Wrap-around is what makes the buffer circular — past the end, the next position is the start.

Initial state (3 items queued):
Buffer:   [A][B][C][.][.][.][.][.]
           ^tail        ^head
           (read here)  (write here)

After write D, write E:
Buffer:   [A][B][C][D][E][.][.][.]
           ^tail              ^head

After read A:
Buffer:   [.][B][C][D][E][.][.][.]
              ^tail           ^head

After 3 more writes (head wraps past end):
Buffer:   [H][B][C][D][E][F][G][.]
              ^tail        ^head
           (head wrote F, G, H — H landed at index 0)

Both pointers chase each other around the ring; the buffer is empty when they meet, full when the head is one slot behind the tail.

Implementation

Operation	Logic
Write	`buffer[head] = value; head = (head + 1) % size`
Read	`value = buffer[tail]; tail = (tail + 1) % size`
Full	`(head + 1) % size == tail`
Empty	`head == tail`

The "full = head one ahead of tail" convention costs one slot — usable capacity is size - 1 rather than size. The alternative is a separate count or full flag; faster on RISC, identical on most 8-bit chips. Pick the one that suits your code.

Gaming Applications

Snake Body

The snake's body is a ring buffer of segment positions. Confusingly, the visible tail of the snake is at the buffer's tail pointer (oldest segment, the one that disappears first), and the snake's head is at the buffer's head pointer (newest segment, the one just added).

Frame	Head pointer (snake's head)	Tail pointer (snake's tail)
Move	Write new position; advance head	Read (erase) old position; advance tail
Grow	Write new position; advance head	Leave the tail alone — the snake gains a segment
Shrink	Don't write	Advance the tail twice — segment falls off

For an 8-bit Snake with a 64-segment maximum body, that's two zero-page bytes of state (head + tail), 64 × 2 bytes for x/y positions, and four AND-masked pointer increments per frame.

Input Replay

Use	Benefit
Fighting games	Store last N inputs for combos
Replays	Record input history
Undo systems	Fixed-depth undo stack
Ghost racing	Store position history

Audio

Application	Why Ring Buffer
Streaming	Producer/consumer decoupling
Effects	Delay lines are ring buffers
Mixing	Lock-free audio threads
Amiga Paula	Each of Paula's four DMA channels reads samples from a hardware-managed ring; the CPU rewrites the sample-pointer / length each VBlank to chain new samples in
NES DMC	The DMC (sample) channel reads from a CPU-RAM ring and increments through it, re-firing CPU stalls each fetch

Vintage Implementation (6502)

8-bit CPUs have no divide instruction, so the modulo in (head + 1) % size is expensive — unless size is a power of two, in which case it collapses to a bitwise AND:

Buffer size	Wrap method	Cycles
256 bytes	Natural 8-bit overflow — no AND needed	0
128 bytes	`AND #$7F`	2
64 bytes	`AND #$3F`	2
Non-power-of-two	Compare and branch (`CMP size; BCC ok; SBC size`)	5-7

A 64-byte input ring on the C64:

; --- Ring state in zero page ---
ring_head:    .byte 0
ring_tail:    .byte 0
RING_MASK = $3F                ; size = 64

; --- Buffer in main RAM ---
ring_buffer:  .res 64

; Write A to the ring (caller checks not-full first)
ring_write:
    ldx ring_head
    sta ring_buffer,x
    inx
    txa
    and #RING_MASK
    sta ring_head
    rts

; Read from the ring into A (caller checks not-empty first)
ring_read:
    ldx ring_tail
    lda ring_buffer,x
    inx
    txa
    and #RING_MASK
    sta ring_tail
    rts

; Empty if head == tail
ring_empty:
    lda ring_head
    cmp ring_tail
    rts                        ; Z set if empty

For a 256-byte ring the AND disappears entirely — INX wraps naturally on overflow — and the routine drops to about 9 cycles per write.

Modern Applications

System	Usage
FMOD/Wwise	Audio streaming buffers
TCP/IP stacks	Packet receive buffers
GGPO (rollback)	Input history for netcode
Logging	Circular log files

Lock-Free Benefits

Ring buffers enable lock-free producer/consumer patterns:

Audio thread writes, main thread reads
Network thread writes, game thread reads
No mutexes needed with single producer/consumer

Trade-offs

Advantage	Disadvantage
No allocation	Fixed maximum size
O(1) operations	Can overflow if full
Cache-friendly	Must handle wrap-around
Lock-free possible	Power-of-two size optimal

Legacy

Ring buffers appear everywhere in systems programming. Understanding them connects game development to audio engineering, networking, and operating systems—fundamental computer science in action.

Circular data, fixed memory