Armada - RISC OS 256 bytes

A 256 bytes real-time intro for RISC OS 3.5+ (RISC PC+ / RPI, 640x512, 32 bpp), released at NOVA 2026.

It shows running XOR patterns behind a tumultuous "fog" that come and go.

The "fog" effect is powered by drifting HAKMEM 149 patterns plus a separate "blur" pass, the HAKMEM pattern is then XOR shaded.

This started from an early 2025 p5js accidental sketch where i was trying to do nice noise with the HAKMEM 149 algorithm, i decided to port this effect on RISC PC after X-87 to support some RO stuff at NOVA.

The port took few hours, it was close to 256 bytes without much effort, only had to fix top / bottom boundary artifacts and optimizing the blur pass / looking at the RISC OS API to replace some costly calls. (e.g. OS_ReadVduVariables by OS_ReadDynamicArea)

The combination of the cheap post processing pass and the unstable patterns + XOR shading adds the illusion of a depth effect.

The API dive was interesting as i didn't know that some of the mode flags can setup a greyscale palette (only RO 5.21+ though) or do some transforms (rotations), many bytes could be saved by going for 256 colors (with greyscale flag) but i prefered older RO compatibility so...

Assembled on RISC PC (RPCEmu) with a custom ARMv2 Forth / assembler, kinda served as a test bed for the tool as well.

ARM code

\ ARMADA \ 256b RISC OS intro

\ ---------> by grz / The Orz

: OS_Byte $6 ; immediate

: OS_ReadMonotonicTime $42 ; immediate

: OS_ScreenMode $65 ; immediate

: OS_ReadDynamicArea $5c ; immediate

: OS_RemoveCursors $36 ; immediate

: OS_ReadEscapeState $2c ; immediate

: OS_Exit $11 ; immediate



variable mode_block

variable l1

variable l2



create ARMADA_ARMv2_CODE

  $0 imm r0 mov

  mode_block r1 adr

  OS_ScreenMode swi



  $2 imm r0 mov

  OS_ReadDynamicArea swi

  r0 r14 mov



  OS_RemoveCursors swi



  \ compute offset

  $a0000 imm r11 mov

  $500 imm r11 r11 add



  l1 ->

    $13 imm r0 mov

    OS_Byte swi \ vsync



    OS_ReadMonotonicTime swi



    $10000 imm r1 mov

    l2 ->

      \ x

      $16 asr r2 r8 mov

      $2 lsl r8 r8 r4 add

      \ y

      $17 asr r3 r5 mov

      $2 lsl r5 r5 r12 add



      \ i

      $9 lsl r12 r11 r7 add

      $1 asr r4 r7 r7 add



      \ osc

      $80 imm r1 r8 and

      r7 r8 r8 add

      [] r8 r14 r6 ldrb

      $15 lsl r6 r2 r2 sub

      $2 asr r3 r2 r2 sub

      $1 asr r2 r3 r3 add

      $2 asr r3 r2 r2 sub



      $14 lsl r0 r2 r2 sub \ animate



      \ color

      r3 r2 r8 eor

      $18 lsr r8 r8 mov

      $150 imm r8 r8 rsb



      \ clamp

      $ff imm r8 cmp

      $ff imm r8 movgt



      \ fix top + bottom artifacts due to "blur"
pass

      $fe imm r5 r9 adds

      $1fc imm r9 cmp



      $8 lsl r8 r8 r8 orr

      $10 lsl r8 r8 r8 orr

      [] r7 r14 r8 strls



      $1 imm r1 r1 subs

      l2 <- bne



    \ "blur" pass

    $a00 imm r14 r1 add

    $140000 imm r14 r10 add

    $a00 imm r10 r10 sub

    l2 ->

      [] $4 negate imm r1 r6 ldrb

      [] $4 imm r1 r7 ldrb

      [] $a00 imm r1 r8 ldrb

      [] $a00 negate imm r1 r9 ldrb



      r7 r6 r6 add

      r8 r6 r6 add

      r9 r6 r6 add

      $2 asr r6 r6 mov

      $8 lsl r6 r6 r6 orr

      $10 lsl r6 r6 r6 orr



      $4 imm [] r1 r6 str



      r10 r1 cmp

      l2 <- bmi



    OS_ReadEscapeState swi

    OS_Exit swics

    l1 <- b



\  screen_mode --> \ X640 Y480 C16M F60 (mode string is
shorter than mode block but not compatible with old RO)

\    $58 c, $36 c, $34 c, $30 c, $20 c,

\    $59 c, $34 c, $38 c, $30 c, $20 c,

\    $43 c, $31 c, $36 c, $4d c, $20 c,

\    $46 c, $36 c, $30 c, $0 c,

  mode_block -->

    $1 l,

    $280 l,      \ 640

    $200 l,      \ 512

    $5 l,        \ C16M

    $1 negate l, \ default refresh rate

    $1 negate l,



ARMADA_ARMv2_CODE here $ff8 dump

p5js code

function setup() {

  createCanvas(640, 512)

  background(0)

}



let x = 0

let y = 0



function draw() {

  loadPixels()

  for (let l = 0; l < 16384*4; l += 1) {

    let i = (0xa0500 + ((((y >> 23) * 5) <<
9) + (((x >>> 22) * 5) >>> 1))) & ~3



    x -= pixels[i + 0 + (l & 128)] << 21

    x -= y >> 2

    y += x >> 1

    x -= y >> 2

    x -= frameCount << 22



    let c = 336 - ((x ^ y) >>> 24) // can also
remove the sub for a darker / busier version

    pixels[i + 0] = pixels[i + 1] = pixels[i + 2] = c

  }



  for (let i = 0; i < width * height * 4; i += 4) {

    let c = pixels[i + 4] + pixels[i - 4] + pixels[i +
width * 4] + pixels[i - width * 4] >> 2

    pixels[i + 0] = pixels[i + 1] = pixels[i + 2] = c //
+ 4 on index is cool

  }

  updatePixels()

}