GBATEK
Gameboy Advance Technical Info - Extracted from no$gba version 2.2b

 GBA Reference

Overview
Technical Data
Memory Map
I/O Map

Hardware Programming
LCD Video Controller
Sound Controller
Timers
DMA Transfers
Communication Ports
Keypad Input
Interrupt Control
System Control

Other
Cartridges
BIOS Functions
Unpredictable Things
External Connectors
 NDS Reference

DS I/O Maps
DS Memory Maps
DS Memory Control
DS Video
DS Sound
DS Various
DS DMA Transfers
DS Timers
DS Interrupts
DS Maths
DS Keypad
DS Inter Process Communication
DS Real-Time Clock (RTC)
DS Serial Peripheral Interface Bus
DS Touch Screen Controller (TSC)
DS Power Management
DS Cartridges, Encryption, Firmware
DS Xboo
DS Backwards-compatible GBA-Mode
 CPU Reference

General Information
CPU Overview
CPU Register Set
CPU Flags
CPU Exceptions

The Instruction Sets
THUMB Instruction Set
ARM Instruction Set
Pseudo Opcodes & Directives

Further Information
ARM System Control CP15
CPU Clock Cycles
CPU Versions
CPU Data Sheet

About GBATEK
About this Document


 Technical Data

CPU Modes
  ARM Mode     ARM7TDMI 32bit RISC CPU, 16.78MHz, 32bit opcodes (GBA)
  THUMB Mode   ARM7TDMI 32bit RISC CPU, 16.78MHz, 16bit opcodes (GBA)
  CGB Mode     Z80/8080-style 8bit CPU, 4.2MHz or 8.4MHz  (CGB compatibility)
  DMG Mode     Z80/8080-style 8bit CPU, 4.2MHz (monochrome gameboy compatib.)
Internal Memory
  BIOS ROM     16 KBytes
  Work RAM     288 KBytes (32K in-chip + 256K on-board)
  VRAM         96 KBytes
  OAM          1 KByte (128 OBJs 3x16bit, 32 OBJ-Rotation/Scalings 4x16bit)
  Palette RAM  1 KByte (256 BG colors, 256 OBJ colors)
Video
  Display      240x160 pixels (2.9 inch TFT color LCD display)
  BG layers    4 background layers
  BG types     Tile/map based, or Bitmap based
  BG colors    256 colors, or 16 colors/16 palettes, or 32768 colors
  OBJ colors   256 colors, or 16 colors/16 palettes
  OBJ size     12 types (in range 8x8 up to 64x64 dots)
  OBJs/Screen  max. 128 OBJs of any size (up to 64x64 dots each)
  OBJs/Line    max. 128 OBJs of 8x8 dots size (under best circumstances)
  Priorities   OBJ/OBJ: 0-127, OBJ/BG: 0-3, BG/BG: 0-3
  Effects      Rotation/Scaling, alpha blending, fade-in/out, mosaic, window
  Backlight    GBA SP only (optionally by light on/off toggle button)
Sound
  Analogue     4 channel CGB compatible (3x square wave, 1x noise)
  Digital      2 DMA sound channels
  Output       Built-in speaker (mono), or headphones socket (stereo)
Controls
  Gamepad      4 Direction Keys, 6 Buttons
Communication Ports
  Serial Port  Various transfer modes, 4-Player Link, Single Game Pak play
External Memory
  GBA Game Pak max. 32MB ROM or flash ROM + max 64K SRAM
  CGB Game Pak max. 32KB ROM + 8KB SRAM (more memory requires banking)
Case Dimensions
  Size (mm)    GBA: 145x81x25 - GBA SP: 82x82x24 (closed), 155x82x24 (stretch)
Power Supply
  Battery GBA  GBA: 2x1.5V DC (AA), Life-time approx. 15 hours
  Battery SP   GBA SP: Built-in rechargeable Lithium ion battery, 3.7V 600mAh
  External     GBA: 3.3V DC 350mA - GBA SP: 5.2V DC 320mA

 ----------------------------------------------------------------------------
        ____._____________...___.____               _______________________
   ____/    :  CARTRIDGE  SIO   :    \____         | _____________________ |
  | L       _____________________  LED  R |        ||                     ||
  |        |                     |        |        ||   2.9" TFT SCREEN   ||
  |   ||   |   2.9" TFT SCREEN   |    (A) |        || 240x160pix  61x40mm ||
  | |====| | 240x160pix  61x40mm | (B)    |        ||   WITH BACKLIGHT    ||
  |   ||   |    NO BACKLIGHT     |  ::::  |        ||                     ||
  |        |                     | SPEAKR |        ||_____________________||
  | STRT() |_____________________|  ::::  |        |  GAME BOY ADVANCE SP  |
  | SLCT()     GAME BOY ADVANCE    VOLUME |        |_______________________|
  |____  OFF-ON  BATTERY 2xAA PHONES  _==_|        |_|________|________|_|_|
       \__.##.__________________,,___/             |L    EXT1     EXT2    R|
                                   .::'            |          (*)      LEDSo
                                .::'     (OPENED)  (VOL_||_           (A)  o
   GBA SP SIDE VIEW          .::'                  |  |_  _| ,,,,,(B)      |
   (CLOSED)               .::'        (STRETCHED)  |    ||   ;SPK;         |
  ...................... _ ......................  |         '''''      ON #
  :_____________________(_).....................:  |       SLCT STRT    OFF#
  |. . . . . . . .'.'.   _|                        | CART.  ()   ()        |
  |_CARTRIDGE_:_BATT._:_|_| <-- EXT1/EXT2          |_:___________________:_|

 ----------------------------------------------------------------------------
     _____________________________________
    |        _____________________        |
    |       |                     |       |
    |       |   2.9" TFT SCREEN   |       |
    |       | 240x160pix  61x40mm |       |
    |       |      BACKLIGHT      |       |
    | ::::: |       3D GFX        | ::::: |
    | ::::: |_____________________| ::::: |
   _|        _          ______   _        |_
  |L|_______| |________|      |_| |_______|R|
  |_______   _____________________   _______|
  |  PWR  | |                     | |SEL STA|
  |   _   | |   2.9" TFT SCREEN   | |       |
  | _| |_ | | 240x160pix  61x40mm | |   X   |
  ||_   _|| |      BACKLIGHT      | | Y   A |
  |  |_|  | |    TOUCH SCREEN     | |   B   |
  |       | |_____________________| |       |
  |_______|             NintendoDS  |_______|
  |         MIC                LEDS         |
  |_________________________________________|
       VOL        SLOT2(GBA)     MIC/PHONES

The separate CPU modes cannot be operated simultaneously. Switching is allowed between ARM and THUMB modes only (that are the two GBA modes).
This manual does not describe CGB and DMG modes, both are completely different than GBA modes, and both cannot be accessed from inside of GBA modes anyways.

GBA SP Notes
Deluxe version of the original GBA. With backlight, new folded laptop-style case, and built-in rechargeable battery. Appears to be 100% compatible with GBA, there seems to be no way to detect SPs by software.

Nintendo DS (Dual Screen) Notes
New handheld with two screens, backwards compatible with GBA games, it is NOT backwards compatible with older 8bit games (mono/color gameboys) though..
Also, the DS has no link port, so that GBA games will thus work only in single player mode, link-port accessoires like printers cannot be used, and most unfortunately multiboot won't work (trying to press Select+Start at powerup will just lock up the DS).


 Memory Map

General Internal Memory
  00000000-00003FFF   BIOS - System ROM         (16 KBytes)
  00004000-01FFFFFF   Not used
  02000000-0203FFFF   WRAM - On-board Work RAM  (256 KBytes) 2 Wait
  02040000-02FFFFFF   Not used
  03000000-03007FFF   WRAM - In-chip Work RAM   (32 KBytes)
  03008000-03FFFFFF   Not used
  04000000-040003FE   I/O Registers
  04000400-04FFFFFF   Not used
Internal Display Memory
  05000000-050003FF   BG/OBJ Palette RAM        (1 Kbyte)
  05000400-05FFFFFF   Not used
  06000000-06017FFF   VRAM - Video RAM          (96 KBytes)
  06018000-06FFFFFF   Not used
  07000000-070003FF   OAM - OBJ Attributes      (1 Kbyte)
  07000400-07FFFFFF   Not used
External Memory (Game Pak)
  08000000-09FFFFFF   Game Pak ROM/FlashROM (max 32MB) - Wait State 0
  0A000000-0BFFFFFF   Game Pak ROM/FlashROM (max 32MB) - Wait State 1
  0C000000-0DFFFFFF   Game Pak ROM/FlashROM (max 32MB) - Wait State 2
  0E000000-0E00FFFF   Game Pak SRAM    (max 64 KBytes) - 8bit Bus width
  0E010000-0FFFFFFF   Not used
Unused Memory Area
  10000000-FFFFFFFF   Not used (upper 4bits of address bus unused)

Default WRAM Usage
By default, the 256 bytes at 03007F00h-03007FFFh in Work RAM are reserved for Interrupt vector, Interrupt Stack, and BIOS Call Stack. The remaining WRAM is free for whatever use (including User Stack, which is initially located at 03007F00h).

Address Bus Width and CPU Read/Write Access Widths
Shows the Bus-Width, supported read and write widths, and the clock cycles for 8/16/32bit accesses.
  Region        Bus   Read      Write     Cycles
  BIOS ROM      32    8/16/32   -         1/1/1
  Work RAM 32K  32    8/16/32   8/16/32   1/1/1
  I/O           32    8/16/32   8/16/32   1/1/1
  OAM           32    8/16/32   16/32     1/1/1 *
  Work RAM 256K 16    8/16/32   8/16/32   3/3/6 **
  Palette RAM   16    8/16/32   16/32     1/1/2 *
  VRAM          16    8/16/32   16/32     1/1/2 *
  GamePak ROM   16    8/16/32   -         5/5/8 **/***
  GamePak Flash 16    8/16/32   16/32     5/5/8 **/***
  GamePak SRAM  8     8         8         5     **
Timing Notes:
  *   Plus 1 cycle if GBA accesses video memory at the same time.
  **  Default waitstate settings, see System Control chapter.
  *** Separate timings for sequential, and non-sequential accesses.
  One cycle equals approx. 59.59ns (ie. 16.78MHz clock).
All memory (except GamePak SRAM) can be accessed by 16bit and 32bit DMA.

GamePak Memory
Only DMA3 (and the CPU of course) may access GamePak ROM. GamePak SRAM can be accessed by the CPU only - restricted to bytewise 8bit transfers. SRAM is supposed as external WRAM expansion - not for battery-buffered data storage - for that purpose it'd be more recommended to use a Flash ROM chip somewhere located in the ROM area.
For details about configuration of GamePak Waitstates, read respective chapter.
SRAM should be accessed only through library ???

VRAM, OAM, and Palette RAM Access
These memory regions can be accessed during H-Blank or V-Blank only (unless display is disabled by Forced Blank bit in DISPCNT register).
There is an additional restriction for OAM memory: Accesses during H-Blank are allowed only if 'H-Blank Interval Free' in DISPCNT is set (which'd reduce number of display-able OBJs though).
The CPU appears to be able to access VRAM/OAM/Palette at any time, a waitstate (one clock cycle) being inserted automatically in case that the display controller was accessing memory simultaneously. (Ie. unlike as in old 8bit gameboy, the data will not get lost.)

CPU Mode Performance
Note that the GamePak ROM bus is limited to 16bits, thus executing ARM instructions (32bit opcodes) from inside of GamePak ROM would result in a not so good performance. So, it'd be more recommended to use THUMB instruction (16bit opcodes) which'd allow each opcode to be read at once.
(ARM instructions can be used at best performance by copying code from GamePak ROM into internal Work RAM)

Data Format
Even though the ARM CPU itself would allow to select between Little-Endian and Big-Endian format by using an external circuit, in the GBA no such circuit exists, and the data format is always Little-Endian. That is, when accessing 16bit or 32bit data in memory, the least significant bits are stored in the first byte (smallest address), and the most significant bits in the last byte. (Ie. same as for 80x86 and Z80 CPUs.)


 I/O Map

LCD I/O Registers
  4000000h  2    R/W  DISPCNT   LCD Control
  4000002h  2    R/W  -         Undocumented - Green Swap
  4000004h  2    R/W  DISPSTAT  General LCD Status (STAT,LYC)
  4000006h  2    R    VCOUNT    Vertical Counter (LY)
  4000008h  2    R/W  BG0CNT    BG0 Control
  400000Ah  2    R/W  BG1CNT    BG1 Control
  400000Ch  2    R/W  BG2CNT    BG2 Control
  400000Eh  2    R/W  BG3CNT    BG3 Control
  4000010h  2    W    BG0HOFS   BG0 X-Offset
  4000012h  2    W    BG0VOFS   BG0 Y-Offset
  4000014h  2    W    BG1HOFS   BG1 X-Offset
  4000016h  2    W    BG1VOFS   BG1 Y-Offset
  4000018h  2    W    BG2HOFS   BG2 X-Offset
  400001Ah  2    W    BG2VOFS   BG2 Y-Offset
  400001Ch  2    W    BG3HOFS   BG3 X-Offset
  400001Eh  2    W    BG3VOFS   BG3 Y-Offset
  4000020h  2    W    BG2PA     BG2 Rotation/Scaling Parameter A (dx)
  4000022h  2    W    BG2PB     BG2 Rotation/Scaling Parameter B (dmx)
  4000024h  2    W    BG2PC     BG2 Rotation/Scaling Parameter C (dy)
  4000026h  2    W    BG2PD     BG2 Rotation/Scaling Parameter D (dmy)
  4000028h  4    W    BG2X      BG2 Reference Point X-Coordinate
  400002Ch  4    W    BG2Y      BG2 Reference Point Y-Coordinate
  4000030h  2    W    BG3PA     BG3 Rotation/Scaling Parameter A (dx)
  4000032h  2    W    BG3PB     BG3 Rotation/Scaling Parameter B (dmx)
  4000034h  2    W    BG3PC     BG3 Rotation/Scaling Parameter C (dy)
  4000036h  2    W    BG3PD     BG3 Rotation/Scaling Parameter D (dmy)
  4000038h  4    W    BG3X      BG3 Reference Point X-Coordinate
  400003Ch  4    W    BG3Y      BG3 Reference Point Y-Coordinate
  4000040h  2    W    WIN0H     Window 0 Horizontal Dimensions
  4000042h  2    W    WIN1H     Window 1 Horizontal Dimensions
  4000044h  2    W    WIN0V     Window 0 Vertical Dimensions
  4000046h  2    W    WIN1V     Window 1 Vertical Dimensions
  4000048h  2    R/W  WININ     Inside of Window 0 and 1
  400004Ah  2    R/W  WINOUT    Inside of OBJ Window & Outside of Windows
  400004Ch  2    W    MOSAIC    Mosaic Size
  400004Eh       -    -         Not used
  4000050h  2    R/W  BLDCNT    Color Special Effects Selection
  4000052h  2    W    BLDALPHA  Alpha Blending Coefficients
  4000054h  2    W    BLDY      Brightness (Fade-In/Out) Coefficient
  4000056h       -    -         Not used
Sound Registers
  4000060h  2  R/W  SOUND1CNT_L Channel 1 Sweep register       (NR10)
  4000062h  2  R/W  SOUND1CNT_H Channel 1 Duty/Length/Envelope (NR11, NR12)
  4000064h  2  R/W  SOUND1CNT_X Channel 1 Frequency/Control    (NR13, NR14)
  4000066h     -    -           Not used
  4000068h  2  R/W  SOUND2CNT_L Channel 2 Duty/Length/Envelope (NR21, NR22)
  400006Ah     -    -           Not used
  400006Ch  2  R/W  SOUND2CNT_H Channel 2 Frequency/Control    (NR23, NR24)
  400006Eh     -    -           Not used
  4000070h  2  R/W  SOUND3CNT_L Channel 3 Stop/Wave RAM select (NR30)
  4000072h  2  R/W  SOUND3CNT_H Channel 3 Length/Volume        (NR31, NR32)
  4000074h  2  R/W  SOUND3CNT_X Channel 3 Frequency/Control    (NR33, NR34)
  4000076h     -    -           Not used
  4000078h  2  R/W  SOUND4CNT_L Channel 4 Length/Envelope      (NR41, NR42)
  400007Ah     -    -           Not used
  400007Ch  2  R/W  SOUND4CNT_H Channel 4 Frequency/Control    (NR43, NR44)
  400007Eh     -    -           Not used
  4000080h  2  R/W  SOUNDCNT_L  Control Stereo/Volume/Enable   (NR50, NR51)
  4000082h  2  R/W  SOUNDCNT_H  Control Mixing/DMA Control
  4000084h  2  R/W  SOUNDCNT_X  Control Sound on/off           (NR52)
  4000086h     -    -           Not used
  4000088h  2  BIOS SOUNDBIAS   Sound PWM Control
  400008Ah  ..   -    -         Not used
  4000090h 2x10h R/W  WAVE_RAM  Channel 3 Wave Pattern RAM (2 banks!!)
  40000A0h  4    W    FIFO_A    Channel A FIFO, Data 0-3
  40000A4h  4    W    FIFO_B    Channel B FIFO, Data 0-3
  40000A8h       -    -         Not used
DMA Transfer Channels
  40000B0h  4    W    DMA0SAD   DMA 0 Source Address
  40000B4h  4    W    DMA0DAD   DMA 0 Destination Address
  40000B8h  2    W    DMA0CNT_L DMA 0 Word Count
  40000BAh  2    R/W  DMA0CNT_H DMA 0 Control
  40000BCh  4    W    DMA1SAD   DMA 1 Source Address
  40000C0h  4    W    DMA1DAD   DMA 1 Destination Address
  40000C4h  2    W    DMA1CNT_L DMA 1 Word Count
  40000C6h  2    R/W  DMA1CNT_H DMA 1 Control
  40000C8h  4    W    DMA2SAD   DMA 2 Source Address
  40000CCh  4    W    DMA2DAD   DMA 2 Destination Address
  40000D0h  2    W    DMA2CNT_L DMA 2 Word Count
  40000D2h  2    R/W  DMA2CNT_H DMA 2 Control
  40000D4h  4    W    DMA3SAD   DMA 3 Source Address
  40000D8h  4    W    DMA3DAD   DMA 3 Destination Address
  40000DCh  2    W    DMA3CNT_L DMA 3 Word Count
  40000DEh  2    R/W  DMA3CNT_H DMA 3 Control
  40000E0h       -    -         Not used
Timer Registers
  4000100h  2    R/W  TM0CNT_L  Timer 0 Counter/Reload
  4000102h  2    R/W  TM0CNT_H  Timer 0 Control
  4000104h  2    R/W  TM1CNT_L  Timer 1 Counter/Reload
  4000106h  2    R/W  TM1CNT_H  Timer 1 Control
  4000108h  2    R/W  TM2CNT_L  Timer 2 Counter/Reload
  400010Ah  2    R/W  TM2CNT_H  Timer 2 Control
  400010Ch  2    R/W  TM3CNT_L  Timer 3 Counter/Reload
  400010Eh  2    R/W  TM3CNT_H  Timer 3 Control
  4000110h       -    -         Not used
Serial Communication (1)
  4000120h  4    R/W  SIODATA32 SIO Data (Normal-32bit Mode) (shared with below!)
  4000120h  2    R/W  SIOMULTI0 SIO Data 0 (Parent)    (Multi-Player Mode)
  4000122h  2    R/W  SIOMULTI1 SIO Data 1 (1st Child) (Multi-Player Mode)
  4000124h  2    R/W  SIOMULTI2 SIO Data 2 (2nd Child) (Multi-Player Mode)
  4000126h  2    R/W  SIOMULTI3 SIO Data 3 (3rd Child) (Multi-Player Mode)
  4000128h  2    R/W  SIOCNT    SIO Control Register
  400012Ah  2    R/W  SIOMLT_SEND SIO Data (Local of Multi-Player) (shared below)
  400012Ah  2    R/W  SIODATA8  SIO Data (Normal-8bit and UART Mode)
  400012Ch       -    -         Not used
Keypad Input
  4000130h  2    R    KEYINPUT  Key Status
  4000132h  2    R/W  KEYCNT    Key Interrupt Control
Serial Communication (2)
  4000134h  2    R/W  RCNT      SIO Mode Select/General Purpose Data
  4000136h  -    -    IR        Ancient - Infrared Register (Prototypes only)
  4000138h       -    -         Not used
  4000140h  2    R/W  JOYCNT    SIO JOY Bus Control
  4000142h       -    -         Not used
  4000150h  4    R/W  JOY_RECV  SIO JOY Bus Receive Data
  4000154h  4    R/W  JOY_TRANS SIO JOY Bus Transmit Data
  4000158h  2    R/?  JOYSTAT   SIO JOY Bus Receive Status
  400015Ah       -    -         Not used
Interrupt, Waitstate, and Power-Down Control
  4000200h  2    R/W  IE        Interrupt Enable Register
  4000202h  2    R/W  IF        Interrupt Request Flags / IRQ Acknowledge
  4000204h  2    R/W  WAITCNT   Game Pak Waitstate Control
  4000206h       -    -         Not used
  4000208h  2    R/W  IME       Interrupt Master Enable Register
  400020Ah       -    -         Not used
  4000300h  1    R/W  POSTFLG   Undocumented - Post Boot Flag
  4000301h  1    W    HALTCNT   Undocumented - Power Down Control
  4000302h       -    -         Not used
  4000410h  ?    ?    ?         Undocumented - Purpose Unknown / Bug ??? 0FFh
  4000411h       -    -         Not used
  4000800h  4    R/W  ?         Undocumented - Internal Memory Control (R/W)
  4000804h       -    -         Not used
  4xx0800h  4    R/W  ?         Mirrors of 4000800h (repeated each 64K)

All further addresses at 4XXXXXXh are unused and do not contain mirrors of the I/O area, with the only exception that 4000800h is repeated each 64K (ie. mirrored at 4010800h, 4020800h, etc.)


 LCD Video Controller

Registers
LCD I/O Display Control
LCD I/O Interrupts and Status
LCD I/O BG Control
LCD I/O BG Scrolling
LCD I/O BG Rotation/Scaling
LCD I/O Window Feature
LCD I/O Mosaic Function
LCD I/O Color Special Effects

VRAM
LCD VRAM Overview
LCD VRAM Character Data
LCD VRAM BG Screen Data Format (BG Map)
LCD VRAM Bitmap BG Modes

Sprites
LCD OBJ - Overview
LCD OBJ - OAM Attributes
LCD OBJ - OAM Rotation/Scaling Parameters
LCD OBJ - VRAM Character (Tile) Mapping

Other
LCD Color Palettes
LCD Dimensions and Timings


 LCD I/O Display Control

4000000h - DISPCNT - LCD Control (Read/Write)
  Bit   Expl.
  0-2   BG Mode                (0-5=Video Mode 0-5, 6-7=Prohibited)
  3     Reserved for BIOS      (CGB Mode - cannot be changed after startup)
  4     Display Frame Select   (0-1=Frame 0-1) (for BG Modes 4,5 only)
  5     H-Blank Interval Free  (1=Allow access to OAM during H-Blank)
  6     OBJ Character VRAM Mapping (0=Two dimensional, 1=One dimensional)
  7     Forced Blank           (1=Allow access to VRAM,Palette,OAM)
  8     Screen Display BG0  (0=Off, 1=On)
  9     Screen Display BG1  (0=Off, 1=On)
  10    Screen Display BG2  (0=Off, 1=On)
  11    Screen Display BG3  (0=Off, 1=On)
  12    Screen Display OBJ  (0=Off, 1=On)
  13    Window 0 Display Flag   (0=Off, 1=On)
  14    Window 1 Display Flag   (0=Off, 1=On)
  15    OBJ Window Display Flag (0=Off, 1=On)

The table summarizes the facilities of the separate BG modes (video modes).
  Mode  Rot/Scal Layers Size               Tiles Colors       Features
  0     No       0123   256x256..512x515   1024  16/16..256/1 SFMABP
  1     Mixed    012-   (BG0,BG1 as above Mode 0, BG2 as below Mode 2)
  2     Yes      --23   128x128..1024x1024 256   256/1        S-MABP
  3     Yes      --?-   240x160            1     32768        --MABP
  4     Yes      --??   240x160            2     256/1        --MABP
  5     Yes      --??   160x128            2     32768        --MABP
Features: S)crolling, F)lip, M)osaic, A)lphaBlending, B)rightness, P)riority.

BG Modes 0-2 are Tile/Map-based. BG Modes 3-5 are Bitmap-based, in these modes 1 or 2 Frames (ie. bitmaps, or 'full screen tiles') exists, if two frames exist, either one can be displayed, and the other one can be redrawn in background.

Blanking Bits
Setting Forced Blank (Bit 7) causes the video controller to display white lines, and all VRAM, Palette RAM, and OAM may be accessed.
"When the internal HV synchronous counter cancels a forced blank during a display period, the display begins from the beginning, following the display of two vertical lines." What ???
Setting H-Blank Interval Free (Bit 5) allows to access OAM during H-Blank time - using this feature reduces the number of sprites that can be displayed per line.

Display Enable Bits
By default, BG0-3 and OBJ Display Flags (Bit 8-12) are used to enable/disable BGs and OBJ. When enabling Window 0 and/or 1 (Bit 13-14), color special effects may be used, and BG0-3 and OBJ are controlled by the window(s).

Frame Selection
In BG Modes 4 and 5 (Bitmap modes), either one of the two bitmaps/frames may be displayed (Bit 4), allowing the user to update the other (invisible) frame in background. In BG Mode 3, only one frame exists.
In BG Modes 0-2 (Tile/Map based modes), a similar effect may be gained by altering the base address(es) of BG Map and/or BG Character data.

4000002h - Undocumented - Green Swap (R/W)
Normally, red green blue intensities for a group of two pixels is output as BGRbgr (uppercase for left pixel at even xloc, lowercase for right pixel at odd xloc). When the Green Swap bit is set, each pixel group is output as BgRbGr (ie. green intensity of each two pixels exchanged).
  Bit   Expl.
  0     Green Swap  (0=Normal, 1=Swap)
  1-15  Not used
This feature appears to be applied to the final picture (ie. after mixing the separate BG and OBJ layers). Eventually intended for other display types (with other pin-outs). With normal GBA hardware it is just producing an interesting dirt effect.


 LCD I/O Interrupts and Status

4000004h - DISPSTAT - General LCD Status (Read/Write)
Display status and Interrupt control. The H-Blank conditions are generated once per scanline, including for the 'hidden' scanlines during V-Blank.
  Bit   Expl.
  0     V-Blank flag   (Read only) (1=VBlank)
  1     H-Blank flag   (Read only) (1=HBlank)
  2     V-Counter flag (Read only) (1=Match)
  3     V-Blank IRQ Enable         (1=Enable)
  4     H-Blank IRQ Enable         (1=Enable)
  5     V-Counter IRQ Enable       (1=Enable)
  6-7   Not used
  8-15  V-Count Setting            (0-227)
The V-Count-Setting value is much the same as LYC of older gameboys, when its value is identical to the content of the VCOUNT register then the V-Counter flag is set (Bit 2), and (if enabled in Bit 5) an interrupt is requested.

4000006h - VCOUNT - Vertical Counter (Read only)
Indicates the currently drawn scanline, values in range from 160-227 indicate 'hidden' scanlines within VBlank area.
  Bit   Expl.
  0-7   Current scanline (0-227)
  8-15  Not Used
Note: This is much the same than the 'LY' register of older gameboys.


 LCD I/O BG Control

4000008h - BG0CNT - BG0 Control (R/W) (BG Modes 0,1 only)
400000Ah - BG1CNT - BG1 Control (R/W) (BG Modes 0,1 only)
400000Ch - BG2CNT - BG2 Control (R/W) (BG Modes 0,1,2 only)
400000Eh - BG3CNT - BG3 Control (R/W) (BG Modes 0,2 only)
  Bit   Expl.
  0-1   BG Priority           (0-3, 0=Highest)
  2-3   Character Base Block  (0-3, in units of 16 KBytes) (=BG Tile Data)
  4-5   Not used (must be zero)
  6     Mosaic                (0=Disable, 1=Enable)
  7     Colors/Palettes       (0=16/16, 1=256/1)
  8-12  Screen Base Block     (0-31, in units of 2 KBytes) (=BG Map Data)
  13    Display Area Overflow (0=Transparent, 1=Wraparound; BG2CNT/BG3CNT only)
  14-15 Screen Size (0-3)
Internal Screen Size (dots) and size of BG Map (bytes):
  Value  Text Mode      Rotation/Scaling Mode
  0      256x256 (2K)   128x128   (256 bytes)
  1      512x256 (4K)   256x256   (1K)
  2      256x512 (4K)   512x512   (4K)
  3      512x512 (8K)   1024x1024 (16K)
In case that some or all BGs are set to same priority then BG0 is having the highest, and BG3 the lowest priority.

In 'Text Modes', the screen size is organized as follows: The screen consists of one or more 256x256 pixel (32x32 tiles) areas. When Size=0: only 1 area (SC0), when Size=1 or Size=2: two areas (SC0,SC1 either horizontally or vertically arranged next to each other), when Size=3: four areas (SC0,SC1 in upper row, SC2,SC3 in lower row). Whereas SC0 is defined by the normal BG Map base address (Bit 8-12 of BG#CNT), SC1 uses same address +2K, SC2 address +4K, SC3 address +6K. When the screen is scrolled it'll always wraparound.

In 'Rotation/Scaling Modes', the screen size is organized as follows, only one area (SC0) of variable size 128x128..1024x1024 pixels (16x16..128x128 tiles) exists (SC0). When the screen is rotated/scaled (or scrolled?) so that the LCD viewport reaches outside of the background/screen area, then BG may be either displayed as transparent or wraparound (Bit 13 of BG#CNT).


 LCD I/O BG Scrolling

4000010h - BG0HOFS - BG0 X-Offset (W)
4000012h - BG0VOFS - BG0 Y-Offset (W)
  Bit   Expl.
  0-8   Offset (0-511)
  9-15  Not used
Specifies the coordinate of the upperleft first visible dot of BG0 background layer, ie. used to scroll the BG0 area.

4000014h - BG1HOFS - BG1 X-Offset (W)
4000016h - BG1VOFS - BG1 Y-Offset (W)
Same as above BG0HOFS and BG0VOFS for BG1 respectively.

4000018h - BG2HOFS - BG2 X-Offset (W)
400001Ah - BG2VOFS - BG2 Y-Offset (W)
Same as above BG0HOFS and BG0VOFS for BG2 respectively.

400001Ch - BG3HOFS - BG3 X-Offset (W)
400001Eh - BG3VOFS - BG3 Y-Offset (W)
Same as above BG0HOFS and BG0VOFS for BG3 respectively.

The above BG scrolling registers are exclusively used in Text modes, ie. for all layers in BG Mode 0, and for the first two layers in BG mode 1.
In other BG modes (Rotation/Scaling and Bitmap modes) above registers are ignored. Instead, the screen may be scrolled by modifying the BG Rotation/Scaling Reference Point registers.


 LCD I/O BG Rotation/Scaling

4000028h - BG2X_L - BG2 Reference Point X-Coordinate, lower 16 bit (W)
400002Ah - BG2X_H - BG2 Reference Point X-Coordinate, upper 12 bit (W)
400002Ch - BG2Y_L - BG2 Reference Point Y-Coordinate, lower 16 bit (W)
400002Eh - BG2Y_H - BG2 Reference Point Y-Coordinate, upper 12 bit (W)
These registers are replacing the BG scrolling registers which are used for Text mode, ie. the X/Y coordinates specify the source position from inside of the BG Map/Bitmap of the pixel to be displayed at upper left of the GBA display. The normal BG scrolling registers are ignored in Rotation/Scaling and Bitmap modes.
  Bit   Expl.
  0-7   Fractional portion (8 bits)
  8-26  Integer portion    (19 bits)
  27    Sign               (1 bit)
  28-31 Not used
Because values are shifted left by eight, fractional portions may be specified in steps of 1/256 pixels (this would be relevant only if the screen is actually rotated or scaled). Normal signed 32bit values may be written to above registers (the most significant bits will be ignored and the value will be cut-down to 28bits, but this is no actual problem because signed values have set all MSBs to the same value).

Internal Reference Point Registers
The above reference points are automatically copied to internal registers during each vblank, specifying the origin for the first scanline. The internal registers are then incremented by dmx and dmy after each scanline.
Caution: Writing to a reference point register by software outside of the Vblank period does immediately copy the new value to the corresponding internal register, that means: in the current frame, the new value specifies the origin of the <current> scanline (instead of the topmost scanline).

4000020h - BG2PA - BG2 Rotation/Scaling Parameter A (alias dx) (W)
4000022h - BG2PB - BG2 Rotation/Scaling Parameter B (alias dmx) (W)
4000024h - BG2PC - BG2 Rotation/Scaling Parameter C (alias dy) (W)
4000026h - BG2PD - BG2 Rotation/Scaling Parameter D (alias dmy) (W)
  Bit   Expl.
  0-7   Fractional portion (8 bits)
  8-14  Integer portion    (7 bits)
  15    Sign               (1 bit)
See below for details.

400003Xh - BG3X_L/H, BG3Y_L/H, BG3PA-D - BG3 Rotation/Scaling Parameters
Same as above BG2 Reference Point, and Rotation/Scaling Parameters, for BG3 respectively.

dx (PA) and dy (PC)
When transforming a horizontal line, dx and dy specify the resulting gradient and magnification for that line. For example:
Horizontal line, length=100, dx=1, and dy=1. The resulting line would be drawn at 45 degrees, f(y)=1/1*x. Note that this would involve that line is magnified, the new length is SQR(100^2+100^2)=141.42. Yup, exactly - that's the old a^2 + b^2 = c^2 formula.

dmx (PB) and dmy (PD)
These values define the resulting gradient and magnification for transformation of vertical lines. However, when rotating a square area (which is surrounded by horizontal and vertical lines), then the desired result should be usually a rotated <square> area (ie. not a parallelogram, for example).
Thus, dmx and dmy must be defined in direct relationship to dx and dy, taking the example above, we'd have to set dmx=-1, and dmy=1, f(x)=-1/1*y.

Area Overflow
In result of rotation/scaling it may often happen that areas outside of the actual BG area become moved into the LCD viewport. Depending of the Area Overflow bit (BG2CNT and BG3CNT, Bit 13) these areas may be either displayed (by wrapping the BG area), or may be displayed transparent.
This works only in BG modes 1 and 2. The area overflow is ignored in Bitmap modes (BG modes 3-5), the outside of the Bitmaps is always transparent.

--- more details and confusing or helpful formulas ---

The following parameters are required for Rotation/Scaling
  Rotation Center X and Y Coordinates (x0,y0)
  Rotation Angle                      (alpha)
  Magnification X and Y Values        (xMag,yMag)
The display is rotated by 'alpha' degrees around the center.
The displayed picture is magnified by 'xMag' along x-Axis (Y=y0) and 'yMag' along y-Axis (X=x0).

Calculating Rotation/Scaling Parameters A-D
  A = Cos (alpha) / xMag    ;distance moved in direction x, same line
  B = Sin (alpha) / xMag    ;distance moved in direction x, next line
  C = Sin (alpha) / yMag    ;distance moved in direction y, same line
  D = Cos (alpha) / yMag    ;distance moved in direction y, next line

Calculating the position of a rotated/scaled dot
Using the following expressions,
  x0,y0    Rotation Center
  x1,y1    Old Position of a pixel (before rotation/scaling)
  x2,y2    New position of above pixel (after rotation scaling)
  A,B,C,D  BG2PA-BG2PD Parameters (as calculated above)
the following formula can be used to calculate x2,y2:
  x2 = A(x1-x0) + B(y1-y0) + x0
  y2 = C(x1-x0) + D(y1-y0) + y0


 LCD I/O Window Feature

The Window Feature may be used to split the screen into four regions. The BG0-3,OBJ layers and Color Special Effects can be separately enabled or disabled in each of these regions.

The DISPCNT Register
DISPCNT Bits 13-15 are used to enable Window 0, Window 1, and/or OBJ Window regions, if any of these regions is enabled then the "Outside of Windows" region is automatically enabled, too.
DISPCNT Bits 8-12 are kept used as master enable bits for the BG0-3,OBJ layers, a layer is displayed only if both DISPCNT and WININ/OUT enable bits are set.

4000040h - WIN0H - Window 0 Horizontal Dimensions (W)
4000042h - WIN1H - Window 1 Horizontal Dimensions (W)
  Bit   Expl.
  0-7   X2, Rightmost coordinate of window, plus 1
  8-15  X1, Leftmost coordinate of window
Garbage values of X2>240 or X1>X2 are interpreted as X2=240.

4000044h - WIN0V - Window 0 Vertical Dimensions (W)
4000046h - WIN1V - Window 1 Vertical Dimensions (W)
  Bit   Expl.
  0-7   Y2, Bottom-most coordinate of window, plus 1
  8-15  Y1, Top-most coordinate of window
Garbage values of Y2>160 or Y1>Y2 are interpreted as Y2=160.

4000048h - WININ - Control of Inside of Window(s) (R/W)
  Bit   Expl.
  0-3   Window 0 BG0-BG3 Enable Bits     (0=No Display, 1=Display)
  4     Window 0 OBJ Enable Bit          (0=No Display, 1=Display)
  5     Window 0 Color Special Effect    (0=Disable, 1=Enable)
  6-7   Not used
  8-11  Window 1 BG0-BG3 Enable Bits     (0=No Display, 1=Display)
  12    Window 1 OBJ Enable Bit          (0=No Display, 1=Display)
  13    Window 1 Color Special Effect    (0=Disable, 1=Enable)
  14-15 Not used

400004Ah - WINOUT - Control of Outside of Windows & Inside of OBJ Window (R/W)
  Bit   Expl.
  0-3   Outside BG0-BG3 Enable Bits      (0=No Display, 1=Display)
  4     Outside OBJ Enable Bit           (0=No Display, 1=Display)
  5     Outside Color Special Effect     (0=Disable, 1=Enable)
  6-7   Not used
  8-11  OBJ Window BG0-BG3 Enable Bits   (0=No Display, 1=Display)
  12    OBJ Window OBJ Enable Bit        (0=No Display, 1=Display)
  13    OBJ Window Color Special Effect  (0=Disable, 1=Enable)
  14-15 Not used

The OBJ Window
The dimension of the OBJ Window is specified by OBJs which are having the "OBJ Mode" attribute being set to "OBJ Window". Any non-transparent dots of any such OBJs are marked as OBJ Window area. The OBJ itself is not displayed.
The color, palette, and display priority of these OBJs are ignored. Both DISPCNT Bits 12 and 15 must be set when defining OBJ Window region(s).

Window Priority
In case that more than one window is enabled, and that these windows do overlap, Window 0 is having highest priority, Window 1 medium, and Obj Window lowest priority. Outside of Window is having zero priority, it is used for all dots which are not inside of any window region.


 LCD I/O Mosaic Function

400004Ch - MOSAIC - Mosaic Size (W)
The Mosaic function can be separately enabled/disabled for BG0-BG3 by BG0CNT-BG3CNT Registers, as well as for each OBJ0-127 by OBJ attributes in OAM memory. Also, setting all of the bits below to zero effectively disables the mosaic function.
  Bit   Expl.
  0-3   BG Mosaic H-Size  (minus 1)
  4-7   BG Mosaic V-Size  (minus 1)
  8-11  OBJ Mosaic H-Size (minus 1)
  12-15 OBJ Mosaic V-Size (minus 1)
Example: When setting H-Size to 5, then pixels 0-5 of each display row are colorized as pixel 0, pixels 6-11 as pixel 6, pixels 12-17 as pixel 12, and so on.

Normally, a 'mosaic-pixel' is colorized by the color of the upperleft covered pixel. In many cases it might be more desireful to use the color of the pixel in the center of the covered area - this effect may be gained by scrolling the background (or by adjusting the OBJ position, as far as upper/left rows/columns of OBJ are transparent).


 LCD I/O Color Special Effects

Two types of Special Effects are supported: Alpha Blending (Semi-Transparency) allows to combine colors of two selected surfaces. Brightness Increase/Decrease adjust the brightness of the selected surface.

4000050h - BLDCNT - Color Special Effects Selection (R/W)
  Bit   Expl.
  0     BG0 1st Target Pixel (Background 0)
  1     BG1 1st Target Pixel (Background 1)
  2     BG2 1st Target Pixel (Background 2)
  3     BG3 1st Target Pixel (Background 3)
  4     OBJ 1st Target Pixel (Top-most OBJ pixel)
  5     BD  1st Target Pixel (Backdrop)
  6-7   Color Special Effect (0-3, see below)
         0 = None                (Special effects disabled)
         1 = Alpha Blending      (1st+2nd Target mixed)
         2 = Brightness Increase (1st Target becomes whiter)
         3 = Brightness Decrease (1st Target becomes blacker)
  8     BG0 2nd Target Pixel (Background 0)
  9     BG1 2nd Target Pixel (Background 1)
  10    BG2 2nd Target Pixel (Background 2)
  11    BG3 2nd Target Pixel (Background 3)
  12    OBJ 2nd Target Pixel (Top-most OBJ pixel)
  13    BD  2nd Target Pixel (Backdrop)
  14-15 Not used
Selects the 1st Target layer(s) for special effects. For Alpha Blending/Semi-Transparency, it does also select the 2nd Target layer(s), which should have next lower display priority as the 1st Target.
However, any combinations are possible, including that all layers may be selected as both 1st+2nd target, in that case the top-most pixel will be used as 1st target, and the next lower pixel as 2nd target.

4000052h - BLDALPHA - Alpha Blending Coefficients (W)
Used for Color Special Effects Mode 1, and for Semi-Transparent OBJs.
  Bit   Expl.
  0-4   EVA Coefficient (1st Target) (0..16 = 0/16..16/16, 17..31=16/16)
  5-7   Not used
  8-12  EVB Coefficient (2nd Target) (0..16 = 0/16..16/16, 17..31=16/16)
  13-15 Not used
For this effect, the top-most non-transparent pixel must be selected as 1st Target, and the next-lower non-transparent pixel must be selected as 2nd Target, if so - and only if so, then color intensities of 1st and 2nd Target are mixed together by using the parameters in BLDALPHA register, for each pixel each R, G, B intensities are calculated separately:
  I = MIN ( 31, I1st*EVA + I2nd*EVB )
Otherwise - for example, if only one target exists, or if a non-transparent non-2nd-target pixel is moved between the two targets, or if 2nd target has higher display priority than 1st target - then only the top-most pixel is displayed (at normal intensity, regardless of BLDALPHA).

4000054h - BLDY - Brightness (Fade-In/Out) Coefficient (W)
Used for Color Special Effects Modes 2 and 3.
  Bit   Expl.
  0-4   EVY Coefficient (Brightness) (0..16 = 0/16..16/16, 17..31=16/16)
  5-15  Not used
For each pixel each R, G, B intensities are calculated separately:
  I = I1st + (31-I1st)*EVY   ;For Brightness Increase
  I = I1st - (I1st)*EVY      ;For Brightness Decrease
The color intensities of any selected 1st target surface(s) are increased or decreased by using the parameter in BLDY register.

Semi-Transparent OBJs
OBJs that are defined as 'Semi-Transparent' in OAM memory are always selected as 1st Target (regardless of BLDCNT Bit 4), and are always using Alpha Blending mode (regardless of BLDCNT Bit 6-7).
The BLDCNT register may be used to perform Brightness effects on the OBJ (and/or other BG/BD layers). However, if a semi-transparent OBJ pixel does overlap a 2nd target pixel, then semi-transparency becomes priority, and the brightness effect will not take place (neither on 1st, nor 2nd target).

The OBJ Layer
Before special effects are applied, the display controller computes the OBJ priority ordering, and isolates the top-most OBJ pixel. In result, only the top-most OBJ pixel is recursed at the time when processing special effects. Ie. alpha blending and semi-transparency can be used for OBJ-to-BG or BG-to-OBJ , but not for OBJ-to-OBJ.


 LCD VRAM Overview

The GBA contains 96 Kbytes VRAM built-in, located at address 06000000-06017FFF, depending on the BG Mode used as follows:

BG Mode 0,1,2 (Tile/Map based Modes)
  06000000-0600FFFF  64 KBytes shared for BG Map and Tiles
  06010000-06017FFF  32 KBytes OBJ Tiles
The shared 64K area can be split into BG Map area(s), and BG Tiles area(s), the respective addresses for Map and Tile areas are set up by BG0CNT-BG3CNT registers. The Map address may be specified in units of 2K (steps of 800h), the Tile address in units of 16K (steps of 4000h).

BG Mode 0,1 (Tile/Map based Text mode)
The tiles may have 4bit or 8bit color depth, minimum map size is 32x32 tiles, maximum is 64x64 tiles, up to 1024 tiles can be used per map.
  Item        Depth     Required Memory
  One Tile    4bit      20h bytes
  One Tile    8bit      40h bytes
  1024 Tiles  4bit      8000h (32K)
  1024 Tiles  8bit      10000h (64K) - excluding some bytes for BG map
  BG Map      32x32     800h (2K)
  BG Map      64x64     2000h (8K)

BG Mode 1,2 (Tile/Map based Rotation/Scaling mode)
The tiles may have 8bit color depth only, minimum map size is 16x16 tiles, maximum is 128x128 tiles, up to 256 tiles can be used per map.
  Item        Depth     Required Memory
  One Tile    8bit      40h bytes
  256  Tiles  8bit      4000h (16K)
  BG Map      16x16     100h bytes
  BG Map      128x128   4000h (16K)

BG Mode 3 (Bitmap based Mode for still images)
  06000000-06013FFF  80 KBytes Frame 0 buffer (only 75K actually used)
  06014000-06017FFF  16 KBytes OBJ Tiles

BG Mode 4,5 (Bitmap based Modes)
  06000000-06009FFF  40 KBytes Frame 0 buffer (only 37.5K used in Mode 4)
  0600A000-06013FFF  40 KBytes Frame 1 buffer (only 37.5K used in Mode 4)
  06014000-06017FFF  16 KBytes OBJ Tiles

Note
Additionally to the above VRAM, the GBA also contains 1 KByte Palette RAM (at 05000000h) and 1 KByte OAM (at 07000000h) which are both used by the display controller as well.


 LCD VRAM Character Data

Each character (tile) consists of 8x8 dots (64 dots in total). The color depth may be either 4bit or 8bit (see BG0CNT-BG3CNT).

4bit depth (16 colors, 16 palettes)
Each tile occupies 32 bytes of memory, the first 4 bytes for the topmost row of the tile, and so on. Each byte representing two dots, the lower 4 bits define the color for the left (!) dot, the upper 4 bits the color for the right dot.

8bit depth (256 colors, 1 palette)
Each tile occupies 64 bytes of memory, the first 8 bytes for the topmost row of the tile, and so on. Each byte selects the palette entry for each dot.


 LCD VRAM BG Screen Data Format (BG Map)

The display background consists of 8x8 dot tiles, the arrangement of these tiles is specified by the BG Screen Data (BG Map). The separate entries in this map are as follows:

Text BG Screen (2 bytes per entry)
Specifies the tile number and attributes. Note that BG tile numbers are always specified in steps of 1 (unlike OBJ tile numbers which are using steps of two in 256 color/1 palette mode).
  Bit   Expl.
  0-9   Tile Number     (0-1023) (a bit less in 256 color mode, because
                           there'd be otherwise no room for the bg map)
  10    Horizontal Flip (0=Normal, 1=Mirrored)
  11    Vertical Flip   (0=Normal, 1=Mirrored)
  12-15 Palette Number  (0-15)    (Not used in 256 color/1 palette mode)
A Text BG Map always consists of 32x32 entries (256x256 pixels), 400h entries = 800h bytes. However, depending on the BG Size, one, two, or four of these Maps may be used together, allowing to create backgrounds of 256x256, 512x256, 256x512, or 512x512 pixels, if so, the first map (SC0) is located at base+0, the next map (SC1) at base+800h, and so on.

Rotation/Scaling BG Screen (1 byte per entry)
In this mode, only 256 tiles can be used. There are no x/y-flip attributes, the color depth is always 256 colors/1 palette.
  Bit   Expl.
  0-7   Tile Number     (0-255)
The dimensions of Rotation/Scaling BG Maps depend on the BG size. For size 0-3 that are: 16x16 tiles (128x128 pixels), 32x32 tiles (256x256 pixels), 64x64 tiles (512x512 pixels), or 128x128 tiles (1024x1024 pixels).

The size and VRAM base address of the separate BG maps for BG0-3 are set up by BG0CNT-BG3CNT registers.


 LCD VRAM Bitmap BG Modes

In BG Modes 3-5 the background is defined in form of a bitmap (unlike as for Tile/Map based BG modes). Bitmaps are implemented as BG2, with Rotation/Scaling support. As bitmap modes are occupying 80KBytes of BG memory, only 16KBytes of VRAM can be used for OBJ tiles.

BG Mode 3 - 240x160 pixels, 32768 colors
Two bytes are associated to each pixel, directly defining one of the 32768 colors (without using palette data, and thus not supporting a 'transparent' BG color).
  Bit   Expl.
  0-4   Red Intensity   (0-31)
  5-9   Green Intensity (0-31)
  10-14 Blue Intensity  (0-31)
  15    Not used
The first 480 bytes define the topmost line, the next 480 the next line, and so on. The background occupies 75 KBytes (06000000-06012BFF), most of the 80 Kbytes BG area, not allowing to redraw an invisible second frame in background, so this mode is mostly recommended for still images only.

BG Mode 4 - 240x160 pixels, 256 colors (out of 32768 colors)
One byte is associated to each pixel, selecting one of the 256 palette entries. Color 0 (backdrop) is transparent, and OBJs may be displayed behind the bitmap.
The first 240 bytes define the topmost line, the next 240 the next line, and so on. The background occupies 37.5 KBytes, allowing two frames to be used (06000000-060095FF for Frame 0, and 0600A000-060135FF for Frame 1).

BG Mode 5 - 160x128 pixels, 32768 colors
Colors are defined as for Mode 3 (see above), but horizontal and vertical size are cut down to 160x128 pixels only - smaller than the physical dimensions of the LCD screen.
The background occupies exactly 40 KBytes, so that BG VRAM may be split into two frames (06000000-06009FFF for Frame 0, and 0600A000-06013FFF for Frame 1).

In BG modes 4,5, one Frame may be displayed (selected by DISPCNT Bit 4), the other Frame is invisible and may be redrawn in background.


 LCD OBJ - Overview

General
Objects (OBJs) are moveable sprites. Up to 128 OBJs (of any size, up to 64x64 dots each) can be displayed per screen, and under best circumstances up to 128 OBJs (of small 8x8 dots size) can be displayed per horizontal display line.

Maximum Number of Sprites per Line
The total available OBJ rendering cycles per line are
  1210  (=304*4-6)   If "H-Blank Interval Free" bit in DISPCNT register is 0
  954   (=240*4-6)   If "H-Blank Interval Free" bit in DISPCNT register is 1
The required rendering cycles are (depending on horizontal OBJ size)
  Cycles per <n> Pixels    OBJ Type              OBJ Type Screen Pixel Range
  n*1 cycles               Normal OBJs           8..64 pixels
  10+n*2 cycles            Rotation/Scaling OBJs 8..64 pixels   (area clipped)
  10+n*2 cycles            Rotation/Scaling OBJs 16..128 pixels (double size)
Caution:
The maximum number of OBJs per line is also affected by undisplayed (offscreen) OBJs which are having higher priority than displayed OBJs.
To avoid this, move displayed OBJs to the begin of OAM memory (ie. OBJ0 has highest priority, OBJ127 lowest).
Otherwise (in case that the program logic expects OBJs at fixed positions in OAM) at least take care to set the OBJ size of undisplayed OBJs to 8x8 with Rotation/Scaling disabled (this reduces the overload).
Does the above also apply for VERTICALLY OFFSCREEN (or VERTICALLY not on CURRENT LINE) sprites ???

VRAM - Character Data
OBJs are always combined of one or more 8x8 pixel Tiles (much like BG Tiles in BG Modes 0-2). However, OBJ Tiles are stored in a separate area in VRAM: 06010000-06017FFF (32 KBytes) in BG Mode 0-2, or 06014000-06017FFF (16 KBytes) in BG Mode 3-5.
Depending on the size of the above area (16K or 32K), and on the OBJ color depth (4bit or 8bit), 256-1024 8x8 dots OBJ Tiles can be defined.

OAM - Object Attribute Memory
This memory area contains Attributes which specify position, size, color depth, etc. appearance for each of the 128 OBJs. Additionally, it contains 32 OBJ Rotation/Scaling Parameter groups. OAM is located at 07000000-070003FF (sized 1 KByte).


 LCD OBJ - OAM Attributes

OBJ Attributes
There are 128 entries in OAM for each OBJ0-OBJ127. Each entry consists of 6 bytes (three 16bit Attributes). Attributes for OBJ0 are located at 07000000, for OBJ1 at 07000008, OBJ2 at 07000010, and so on.

As you can see, there are blank spaces at 07000006, 0700000E, 07000016, etc. - these 16bit values are used for OBJ Rotation/Scaling (as described in the next chapter) - they are not directly related to the separate OBJs.

OBJ Attribute 0 (R/W)
  Bit   Expl.
  0-7   Y-Coordinate           (0-255)
  8     Rotation/Scaling Flag  (0=Off, 1=On)
  When Rotation/Scaling used (Attribute 0, bit 8 set):
    9     Double-Size Flag     (0=Normal, 1=Double)
  When Rotation/Scaling not used (Attribute 0, bit 8 cleared):
    9     OBJ Disable          (0=Normal, 1=Not displayed)
  10-11 OBJ Mode  (0=Normal, 1=Semi-Transparent, 2=OBJ Window, 3=Prohibited)
  12    OBJ Mosaic             (0=Off, 1=On)
  13    Colors/Palettes        (0=16/16, 1=256/1)
  14-15 OBJ Shape              (0=Square,1=Horizontal,2=Vertical,3=Prohibited)
Caution: A very large OBJ (of 128 pixels vertically, ie. a 64 pixels OBJ in a Double Size area) located at Y>128 will be treated as at Y>-128, the OBJ is then displayed parts offscreen at the TOP of the display, it is then NOT displayed at the bottom.

OBJ Attribute 1 (R/W)
  Bit   Expl.
  0-8   X-Coordinate           (0-511)
  When Rotation/Scaling used (Attribute 0, bit 8 set):
    9-13  Rotation/Scaling Parameter Selection (0-31)
          (Selects one of the 32 Rotation/Scaling Parameters that
          can be defined in OAM, for details read next chapter.)
  When Rotation/Scaling not used (Attribute 0, bit 8 cleared):
    9-11  Not used
    12    Horizontal Flip      (0=Normal, 1=Mirrored)
    13    Vertical Flip        (0=Normal, 1=Mirrored)
  14-15 OBJ Size               (0..3, depends on OBJ Shape, see Attr 0)
          Size  Square   Horizontal  Vertical
          0     8x8      16x8        8x16
          1     16x16    32x8        8x32
          2     32x32    32x16       16x32
          3     64x64    64x32       32x64

OBJ Attribute 2 (R/W)
  Bit   Expl.
  0-9   Character Name          (0-1023=Tile Number)
  10-11 Priority relative to BG (0-3; 0=Highest)
  12-15 Palette Number   (0-15) (Not used in 256 color/1 palette mode)

Notes:

OBJ Mode
The OBJ Mode may be Normal, Semi-Transparent, or OBJ Window.
Semi-Transparent means that the OBJ is used as 'Alpha Blending 1st Target' (regardless of BLDCNT register, for details see chapter about Color Special Effects).
OBJ Window means that the OBJ is not displayed, instead, dots with non-zero color are used as mask for the OBJ Window, see DISPCNT and WINOUT for details.

OBJ Tile Number
There are two situations which may divide the amount of available tiles by two (by four if both situations apply):

1. When using the 256 Colors/1 Palette mode, only each second tile may be used, the lower bit of the tile number should be zero (in 2-dimensional mapping mode, the bit is completely ignored).

2. When using BG Mode 3-5 (Bitmap Modes), only tile numbers 512-1023 may be used. That is because lower 16K of OBJ memory are used for BG. Attempts to use tiles 0-511 are ignored (not displayed).

Priority
In case that the 'Priority relative to BG' is the same than the priority of one of the background layers, then the OBJ becomes higher priority and is displayed on top of that BG layer.
Caution: Take care not to mess up BG Priority and OBJ priority. For example, the following would cause garbage to be displayed:
  OBJ No. 0 with Priority relative to BG=1   ;hi OBJ prio, lo BG prio
  OBJ No. 1 with Priority relative to BG=0   ;lo OBJ prio, hi BG prio
That is, OBJ0 is always having priority above OBJ1-127, so assigning a lower BG Priority to OBJ0 than for OBJ1-127 would be a bad idea.


 LCD OBJ - OAM Rotation/Scaling Parameters

As described in the previous chapter, there are blank spaces between each of the 128 OBJ Attribute Fields in OAM memory. These 128 16bit gaps are used to store OBJ Rotation/Scaling Parameters.

Location of Rotation/Scaling Parameters in OAM
Four 16bit parameters (PA,PB,PC,PD) are required to define a complete group of Rotation/Scaling data. These are spread across OAM as such:
  1st Group - PA=07000006, PB=0700000E, PC=07000016, PD=0700001E
  2nd Group - PA=07000026, PB=0700002E, PC=07000036, PD=0700003E
  etc.
By using all blank space (128 x 16bit), up to 32 of these groups (4 x 16bit each) can be defined in OAM.

OBJ Rotation/Scaling PA,PB,PC,PD Parameters (R/W)
Each OBJ that uses Rotation/Scaling may select between any of the above 32 parameter groups. For details, refer to the previous chapter about OBJ Attributes.
The meaning of the separate PA,PB,PC,PD values is identical as for BG, for details read the chapter about BG Rotation/Scaling.

OBJ Reference Point & Rotation Center
The OBJ Reference Point is the upper left of the OBJ, ie. OBJ X/Y coordinates: X+0, Y+0.
The OBJ Rotation Center is always (or should be usually?) in the middle of the object, ie. for a 8x32 pixel OBJ, this would be at the OBJ X/Y coordinates: X+4, and Y+16.

OBJ Double-Size Bit (for OBJs that use Rotation/Scaling)
When Double-Size is zero: The sprite is rotated, and then display inside of the normal-sized (not rotated) rectangular area - the edges of the rotated sprite will become invisible if they reach outside of that area.
When Double-Size is set: The sprite is rotated, and then display inside of the double-sized (not rotated) rectangular area - this ensures that the edges of the rotated sprite remain visible even if they would reach outside of the normal-sized area. (Except that, for example, rotating a 8x32 pixel sprite by 90 degrees would still cut off parts of the sprite as the double-size area isn't large enough.)


 LCD OBJ - VRAM Character (Tile) Mapping

Each OBJ tile consists of 8x8 dots, however, bigger OBJs can be displayed by combining several 8x8 tiles. The horizontal and vertical size for each OBJ may be separately defined in OAM, possible H/V sizes are 8,16,32,64 dots - allowing 'square' OBJs to be used (such like 8x8, 16x16, etc) as well as 'rectangular' OBJs (such like 8x32, 64x16, etc.)

When displaying an OBJ that contains of more than one 8x8 tile, one of the following two mapping modes can be used. In either case, the tile number of the upperleft tile must be specified in OAM memory.

Two Dimensional Character Mapping (DISPCNT Bit 6 cleared)
This mapping mode assumes that the 1024 OBJ tiles are arranged as a matrix of 32x32 tiles / 256x256 pixels (In 256 color mode: 16x32 tiles / 128x256 pixels). Ie. the upper row of this matrix contains tiles 00h-1Fh, the next row tiles 20h-3Fh, and so on.
For example, when displaying a 16x16 pixel OBJ, with tile number set to 04h; The upper row of the OBJ will consist of tile 04h and 05h, the next row of 24h and 25h. (In 256 color mode: 04h and 06h, 24h and 26h.)

One Dimensional Character Mapping (DISPCNT Bit 6 set)
In this mode, tiles are mapped each after each other from 00h-3FFh.
Using the same example as above, the upper row of the OBJ will consist of tile 04h and 05h, the next row of tile 06h and 07h. (In 256 color mode: 04h and 06h, 08h and 0Ah.)


 LCD Color Palettes

Color Palette RAM
BG and OBJ palettes are using separate memory regions:
  05000000-050001FF - BG Palette RAM (512 bytes, 256 colors)
  05000200-050003FF - OBJ Palette RAM (512 bytes, 256 colors)
Each BG and OBJ palette RAM may be either split into 16 palettes with 16 colors each, or may be used as a single palette with 256 colors.
Note that some OBJs may access palette RAM in 16 color mode, while other OBJs may use 256 color mode at the same time. Same for BG0-BG3 layers.

Transparent Colors
Color 0 of all BG and OBJ palettes is transparent. Even though palettes are described as 16 (256) color palettes, only 15 (255) colors are actually visible.

Backdrop Color
Color 0 of BG Palette 0 is used as backdrop color. This color is displayed if an area of the screen is not covered by any non-transparent BG or OBJ dots.

Color Definitions
Each color occupies two bytes (same as for 32768 color BG modes):
  Bit   Expl.
  0-4   Red Intensity   (0-31)
  5-9   Green Intensity (0-31)
  10-14 Blue Intensity  (0-31)
  15    Not used

Intensities
Under normal circumstances (light source/viewing angle), the intensities 0-14 are practically all black, and only intensities 15-31 are resulting in visible medium..bright colors.

Note: The intensity problem appears in the 8bit CGB "compatibility" mode either. The original CGB display produced the opposite effect: Intensities 0-14 resulted in dark..medium colors, and intensities 15-31 resulted in bright colors. Any "medium" colors of CGB games will appear invisible/black on GBA hardware, and only very bright colors will be visible.


 LCD Dimensions and Timings

Horizontal Dimensions
The drawing time for each dot is 4 CPU cycles.
  Visible     240 dots,  57.221 us,    960 cycles - 78% of h-time
  H-Blanking   68 dots,  16.212 us,    272 cycles - 22% of h-time
  Total       308 dots,  73.433 us,   1232 cycles - ca. 13.620 kHz
VRAM and Palette RAM may be accessed during H-Blanking. OAM can accessed only if "H-Blank Interval Free" bit in DISPCNT register is set.

Vertical Dimensions
  Visible (*) 160 lines, 11.749 ms, 197120 cycles - 70% of v-time
  V-Blanking   68 lines,  4.994 ms,  83776 cycles - 30% of v-time
  Total       228 lines, 16.743 ms, 280896 cycles - ca. 59.737 Hz
All VRAM, OAM, and Palette RAM may be accessed during V-Blanking.
Note that no H-Blank interrupts are generated within V-Blank period.

System Clock
The system clock is 16.78MHz (16*1024*1024 Hz), one cycle is thus approx. 59.59ns.

(*) Even though vertical screen size is 160 lines, the upper 8 lines are not <really> visible, these lines are covered by a shadow when holding the GBA orientated towards a light source, the lines are effectively black - and should not be used to display important information.


 Sound Controller

The GBA supplies four 'analogue' sound channels for Tone and Noise (mostly compatible to CGB sound), as well as two 'digital' sound channels (which can be used to replay 8bit DMA sample data).

Sound Channel 1 - Tone & Sweep
Sound Channel 2 - Tone
Sound Channel 3 - Wave Output
Sound Channel 4 - Noise
Sound Channel A and B - DMA Sound

Sound Control Registers
Comparison of CGB and GBA Sound

The GBA includes only a single (mono) speaker built-in, each channel may be output to either left and/or right channels by using the external line-out connector (for stereo headphones, etc).


 Sound Channel 1 - Tone & Sweep

4000060h - SOUND1CNT_L (NR10) - Channel 1 Sweep register (R/W)
  Bit        Expl.
  0-2   R/W  Number of sweep shift      (n=0-7)
  3     R/W  Sweep Frequency Direction  (0=Increase, 1=Decrease)
  4-6   R/W  Sweep Time; units of 7.8ms (0-7, min=7.8ms, max=54.7ms)
  7-15  -    Not used
Sweep is disabled by setting Sweep Time to zero, if so, the direction bit should be set.
The change of frequency (NR13,NR14) at each shift is calculated by the following formula where X(0) is initial freq & X(t-1) is last freq:
  X(t) = X(t-1) +/- X(t-1)/2^n

4000062h - SOUND1CNT_H (NR11, NR12) - Channel 1 Duty/Len/Envelope (R/W)
  Bit        Expl.
  0-5   W    Sound length; units of (64-n)/256s  (0-63)
  6-7   R/W  Wave Pattern Duty                   (0-3, see below)
  8-10  R/W  Envelope Step-Time; units of n/64s  (1-7, 0=No Envelope)
  11    R/W  Envelope Direction                  (0=Decrease, 1=Increase)
  12-15 R/W  Initial Volume of envelope          (1-15, 0=No Sound)
Wave Duty:
  0: 12.5% ( -_______-_______-_______ )
  1: 25%   ( --______--______--______ )
  2: 50%   ( ----____----____----____ ) (normal)
  3: 75%   ( ------__------__------__ )
The Length value is used only if Bit 6 in NR14 is set.

4000064h - SOUND1CNT_X (NR13, NR14) - Channel 1 Frequency/Control (R/W)
  Bit        Expl.
  0-10  W    Frequency; 131072/(2048-n)Hz  (0-2047)
  11-13 -    Not used
  14    R/W  Length Flag  (1=Stop output when length in NR11 expires)
  15    W    Initial      (1=Restart Sound)

4000066h - Not used


 Sound Channel 2 - Tone

This sound channel works exactly as channel 1, except that it doesn't have a Tone Envelope/Sweep Register.

4000068h - SOUND2CNT_L (NR21, NR22) - Channel 2 Duty/Length/Envelope (R/W)
400006Ah - Not used
400006Ch - SOUND2CNT_H (NR23, NR24) - Channel 2 Frequency/Control (R/W)
400006Eh - Not used
For details, refer to channel 1 description.


 Sound Channel 3 - Wave Output

This channel can be used to output digital sound, the length of the sample buffer (Wave RAM) can be either 32 or 64 digits (4bit samples). This sound channel can be also used to output normal tones when initializing the Wave RAM by a square wave. This channel doesn't have a volume envelope register.

4000070h - SOUND3CNT_L (NR30) - Channel 3 Stop/Wave RAM select (R/W)
  Bit        Expl.
  0-4   -    Not used
  5     R/W  Wave RAM Dimension   (0=One bank/32 digits, 1=Two banks/64 digits)
  6     R/W  Wave RAM Bank Number (0-1, see below)
  7     R/W  Sound Channel 3 Off  (0=Stop, 1=Playback)
  8-15  -    Not used
The currently selected Bank Number (Bit 6) will be played back, while reading/writing to/from wave RAM will address the other (not selected) bank. When dimension is set to two banks, output will start by replaying the currently selected bank.

4000072h - SOUND3CNT_H (NR31, NR32) - Channel 3 Length/Volume (R/W)
  Bit        Expl.
  0-7   W    Sound length; units of (256-n)/256s  (0-255)
  8-12  -    Not used.
  13-14 R/W  Sound Volume  (0=Mute/Zero, 1=100%, 2=50%, 3=25%)
  15    R/W  Force Volume  (0=Use above, 1=Force 75% regardless of above)
The Length value is used only if Bit 6 in NR34 is set.

4000074h - SOUND3CNT_X (NR33, NR34) - Channel 3 Frequency/Control (R/W)
  Bit        Expl.
  0-10  W    Sample Rate; 2097152/(2048-n) Hz   (0-2047)
  11-13 -    Not used
  14    R/W  Length Flag  (1=Stop output when length in NR31 expires)
  15    W    Initial      (1=Restart Sound)
The above sample rate specifies the number of wave RAM digits per second, the actual tone frequency depends on the wave RAM content, for example:
  Wave RAM, single bank 32 digits   Tone Frequency
  FFFFFFFFFFFFFFFF0000000000000000  65536/(2048-n) Hz
  FFFFFFFF00000000FFFFFFFF00000000  131072/(2048-n) Hz
  FFFF0000FFFF0000FFFF0000FFFF0000  262144/(2048-n) Hz
  FF00FF00FF00FF00FF00FF00FF00FF00  524288/(2048-n) Hz
  F0F0F0F0F0F0F0F0F0F0F0F0F0F0F0F0  1048576/(2048-n) Hz

4000076h - Not used

4000090h - WAVE_RAM0_L - Channel 3 Wave Pattern RAM (W/R)
4000092h - WAVE_RAM0_H - Channel 3 Wave Pattern RAM (W/R)
4000094h - WAVE_RAM1_L - Channel 3 Wave Pattern RAM (W/R)
4000096h - WAVE_RAM1_H - Channel 3 Wave Pattern RAM (W/R)
4000098h - WAVE_RAM2_L - Channel 3 Wave Pattern RAM (W/R)
400009Ah - WAVE_RAM2_H - Channel 3 Wave Pattern RAM (W/R)
400009Ch - WAVE_RAM3_L - Channel 3 Wave Pattern RAM (W/R)
400009Eh - WAVE_RAM3_H - Channel 3 Wave Pattern RAM (W/R)
This area contains 16 bytes (32 x 4bits) Wave Pattern data which is output by channel 3. Data is played back ordered as follows: MSBs of 1st byte, followed by LSBs of 1st byte, followed by MSBs of 2nd byte, and so on - this results in a confusing ordering when filling Wave RAM in units of 16bit data - ie. samples would be then located in Bits 4-7, 0-3, 12-15, 8-11.

In the GBA, two Wave Patterns exists (each 32 x 4bits), either one may be played (as selected in NR30 register), the other bank may be accessed by the users. After all 32 samples have been played, output of the same bank (or other bank, as specified in NR30) will be automatically restarted.

Internally, Wave RAM is a giant shift-register, there is no pointer which is addressing the currently played digit. Instead, the entire 128 bits are shifted, and the 4 least significant bits are output.
Thus, when reading from Wave RAM, data might have changed its position. And, when writing to Wave RAM all data should be updated (it'd be no good idea to assume that old data is still located at the same position where it has been written to previously).


 Sound Channel 4 - Noise

This channel is used to output white noise. This is done by randomly switching the amplitude between high and low at a given frequency. Depending on the frequency the noise will appear 'harder' or 'softer'.

It is also possible to influence the function of the random generator, so the that the output becomes more regular, resulting in a limited ability to output Tone instead of Noise.

4000078h - SOUND4CNT_L (NR41, NR42) - Channel 4 Length/Envelope (R/W)
  Bit        Expl.
  0-5   W    Sound length; units of (64-n)/256s  (0-63)
  6-7   -    Not used
  8-10  R/W  Envelope Step-Time; units of n/64s  (1-7, 0=No Envelope)
  11    R/W  Envelope Direction                  (0=Decrease, 1=Increase)
  12-15 R/W  Initial Volume of envelope          (1-15, 0=No Sound)
The Length value is used only if Bit 6 in NR44 is set.

400007Ah - Not used

400007Ch - SOUND4CNT_H (NR43, NR44) - Channel 4 Frequency/Control (R/W)
The amplitude is randomly switched between high and low at the given frequency. A higher frequency will make the noise to appear 'softer'.
When Bit 3 is set, the output will become more regular, and some frequencies will sound more like Tone than Noise.
  Bit        Expl.
  0-2   R/W  Dividing Ratio of Frequencies (r)
  3     R/W  Counter Step/Width (0=15 bits, 1=7 bits)
  4-7   R/W  Shift Clock Frequency (s)
  8-13  -    Not used
  14    R/W  Length Flag  (1=Stop output when length in NR41 expires)
  15    W    Initial      (1=Restart Sound)
Frequency = 524288 Hz / r / 2^(s+1) ;For r=0 assume r=0.5 instead

400007Eh - Not used


 Sound Channel A and B - DMA Sound

The GBA contains two DMA sound channels (A and B), each allowing to replay digital sound (signed 8bit data, ie. -128..+127). Data can be transferred from INTERNAL memory (not sure if EXTERNAL memory works also ???) to FIFO by using DMA channel 1 or 2, the sample rate is generated by using one of the Timers.

40000A0h - FIFO_A_L - Sound A FIFO, Data 0 and Data 1 (W)
40000A2h - FIFO_A_H - Sound A FIFO, Data 2 and Data 3 (W)
These two registers may receive 32bit (4 bytes) of audio data (Data 0-3, Data 0 being located in least significant byte which is replayed first).
Internally, the capacity of the FIFO is 8 x 32bit (32 bytes), allowing to buffer a small amount of samples. As the name says (First In First Out), oldest data is replayed first.

40000A4h - FIFO_B_L - Sound B FIFO, Data 0 and Data 1 (W)
40000A6h - FIFO_B_H - Sound B FIFO, Data 2 and Data 3 (W)
Same as above, for Sound B.

Initializing DMA-Sound Playback
- Select Timer 0 or 1 in SOUNDCNT_H control register.
- Clear the FIFO.
- Manually write a sample byte to the FIFO.
- Initialize transfer mode for DMA 1 or 2.
- Initialize DMA Sound settings in sound control register.
- Start the timer.

DMA-Sound Playback Procedure
The pseudo-procedure below is automatically repeated.
  If Timer overflows then
    Move 8bit data from FIFO to sound circuit.
    If FIFO contains only 4 x 32bits (16 bytes) then
      Request more data per DMA
      Receive 4 x 32bit (16 bytes) per DMA
    Endif
  Endif
This playback mechanism will be repeated forever, regardless of the actual length of the sample buffer.

Synchronizing Sample Buffers
The buffer-end may be determined by counting sound Timer IRQs (each sample byte), or sound DMA IRQs (each 16th sample byte). Both methods would require a lot of CPU time (IRQ processing), and both would fail if interrupts are disabled for a longer period.
Better solutions would be to synchronize the sample rate/buffer length with V-blanks, or to use a second timer (in count up/slave mode) which produces an IRQ after the desired number of samples.

The Sample Rate
The GBA hardware does internally re-sample all sound output to 32.768kHz (default SOUNDBIAS setting). It'd thus do not make much sense to use higher DMA/Timer rates. Best re-sampling accuracy can be gained by using DMA/Timer rates of 32.768kHz, 16.384kHz, or 8.192kHz (ie. fragments of the physical output rate).


 Sound Control Registers

4000080h - SOUNDCNT_L (NR50, NR51) - Channel L/R Volume/Enable (R/W)
  Bit   Expl.
  0-2   Sound 1-4 Master volume RIGHT (0-7)
  3     Not used
  4-6   Sound 1-4 Master Volume LEFT (0-7)
  7     Not used
  8-11  Sound 1-4 Enable Flags RIGHT (each Bit 8-11, 0=Disable, 1=Enable)
  12-15 Sound 1-4 Enable Flags LEFT (each Bit 12-15, 0=Disable, 1=Enable)

4000082h - SOUNDCNT_H (GBA only) - DMA Sound Control/Mixing (R/W)
  Bit   Expl.
  0-1   Sound # 1-4 Volume   (0=25%, 1=50%, 2=100%, 3=Prohibited)
  2     DMA Sound A Volume   (0=50%, 1=100%)
  3     DMA Sound B Volume   (0=50%, 1=100%)
  4-7   Not used
  8     DMA Sound A Enable RIGHT (0=Disable, 1=Enable)
  9     DMA Sound A Enable LEFT  (0=Disable, 1=Enable)
  10    DMA Sound A Timer Select (0=Timer 0, 1=Timer 1)
  11    DMA Sound A Reset FIFO   (1=Reset)
  12    DMA Sound B Enable RIGHT (0=Disable, 1=Enable)
  13    DMA Sound B Enable LEFT  (0=Disable, 1=Enable)
  14    DMA Sound B Timer Select (0=Timer 0, 1=Timer 1)
  15    DMA Sound B Reset FIFO   (1=Reset)

4000084h - SOUNDCNT_X (NR52) - Sound on/off (R/W)
When not using sound output, write 00h to this register to save power consumption. While Bit 7 is cleared, all other sound registers cannot be accessed, and their content must be re-initialized when re-enabling sound.
  Bit   Expl.
  0     Sound 1 ON flag (Read Only)
  1     Sound 2 ON flag (Read Only)
  2     Sound 3 ON flag (Read Only)
  3     Sound 4 ON flag (Read Only)
  4-6   Not used
  7     All sound on/off  (0: stop all sound circuits) (Read/Write)
  8-15  Not used
Bits 0-3 are automatically set when starting sound output, and are automatically cleared when a sound ends. (Ie. when the length expires, as far as length is enabled. The bits are NOT reset when an volume envelope ends.)

4000086h - Not used

4000088h - SOUNDBIAS - Sound PWM Control (R/W, see below)
This register controls the final sound output. The default setting is 0200h, it is normally not required to change this value.
  Bit    Expl.
  0-9    Bias Level     (Default=200h, converting signed samples into unsigned)
  10-13  Not used
  14-15  Amplitude Resolution/Sampling Cycle (Default=0, see below)
Amplitude Resolution/Sampling Cycle (0-3):
  0  9bit / 32.768kHz   (Default, best for DMA channels A,B)
  1  8bit / 65.536kHz
  2  7bit / 131.072kHz
  3  6bit / 262.144kHz  (Best for FM channels 1-4)
For more information on this register, read the descriptions below.

400008Ah - Not used
400008Ch - Not used
400008Eh - Not used

Mixing of the separate channels into 10bit
The current output levels of all six channels are added together by hardware, resulting in a signed value, typically in range -512..+511. The bias level (typically 200h = 512 decimal) is added to the result to convert it into an unsigned 10bit value, range 0..+1023. Values smaller than 0 or greater than 1023 appear to be clipped.

Resampling to 32.768kHz / 9bit (default)
The FM channels 1-4 are internally generated at 262.144kHz, and DMA sound A-B could be theoretically generated at timer rates up to 16.78MHz. However, the final sound output is resampled to a rate of 32.768kHz, at 9bit depth (the above 10bit value, divided by two). If necessary, rates higher than 32.768kHz can be selected in the SOUNDBIAS register, that would result in a depth smaller than 9bit though.

PWM (Pulse Width Modulation) Output 16.78MHz / 1bit
Okay, now comes the actual output. The GBA can output only two voltages (low and high), these 'bits' are output at system clock speed (16.78MHz). If using the default 32.768kHz sampling rate, then 512 bits are output per sample (512*32K=16M). Each sample value (9bit range, N=0..511), would be then output as N low bits, followed by 512-N high bits. The resulting 'noise' is smoothed down by capacitors, by the speaker, and by human hearing, so that it will effectively sound like clean D/A converted 9bit voltages at 32kHz sampling rate.

Changing the BIAS Level
Normally use 200h for clean sound output. A value of 000h might make sense during periods when no sound is output (causing the PWM circuit to output low-bits only, which is eventually reducing the power consumption, and/or preventing 32KHz noise). Note: Using the SoundBias function (SWI 19h) allows to change the level by slowly incrementing or decrementing it (without hard scratch noise).


 Comparison of CGB and GBA Sound

The GBA sound controller is mostly the same than that of older monochrome gameboy and CGB. The following changes have been done:

New Sound Channels
Two new sound channels have been added that may be used to replay 8bit digital sound. Sample rate and sample data must be supplied by using a Timer and a DMA channel.

New Control Registers
The SOUNDCNT_H register controls the new DMA channels - as well as mixing with the four old channels. The SOUNDBIAS register controls the final sound output.

Sound Channel 3 Changes
The length of the Wave RAM is doubled by dividing it into two banks of 32 digits each, either one or both banks may be replayed (one after each other), for details check NR30 Bit 5-6. Optionally, the sound may be output at 75% volume, for details check NR32 Bit 7.

Changed Control Registers
NR50 is not supporting Vin signals (that's been an external sound input from cartridge).

Changed I/O Addresses
The GBAs sound register are located at 04000060-040000AE instead of at FF10-FF3F as in CGB and monochrome gameboy. However, note that there have been new blank spaces inserted between some of the separate registers - therefore it is NOT possible to port CGB software to GBA just by changing the sound base address.

Accessing I/O Registers
In some cases two of the old 8bit registers are packed into a 16bit register and may be accessed as such.


 Timers

The GBA includes four incrementing 16bit timers.
Timer 0 and 1 can be used to supply the sample rate for DMA sound channel A and/or B.

4000100h - TM0CNT_L - Timer 0 Counter/Reload (R/W)
4000104h - TM1CNT_L - Timer 1 Counter/Reload (R/W)
4000108h - TM2CNT_L - Timer 2 Counter/Reload (R/W)
400010Ch - TM3CNT_L - Timer 3 Counter/Reload (R/W)
Writing to these registers initializes the <reload> value (but does not directly affect the current counter value). Reading returns the current <counter> value (or the recent/frozen counter value if the timer has been stopped).
The reload value is copied into the counter only upon following two situations: Automatically upon timer overflows, or when the timer start bit becomes changed from 0 to 1.
Note: When simultaneously changing the start bit from 0 to 1, and setting the reload value at the same time (by a single 32bit I/O operation), then the newly written reload value is recognized as new counter value.

4000102h - TM0CNT_H - Timer 0 Control (R/W)
4000106h - TM1CNT_H - Timer 1 Control (R/W)
400010Ah - TM2CNT_H - Timer 2 Control (R/W)
400010Eh - TM3CNT_H - Timer 3 Control (R/W)
  Bit   Expl.
  0-1   Prescaler Selection (0=F/1, 1=F/64, 2=F/256, 3=F/1024)
  2     Count-up Timing   (0=Normal, 1=See below)
  3-5   Not used
  6     Timer IRQ Enable  (0=Disable, 1=IRQ on Timer overflow)
  7     Timer Start/Stop  (0=Stop, 1=Operate)
  8-15  Not used
When Count-up Timing is enabled, the prescaler value is ignored, instead the time is incremented each time when the previous counter overflows. This function cannot be used for Timer 0 (as it is the first timer).
F = System Clock (16.78MHz).


 DMA Transfers

Overview
The GBA includes four DMA channels, the highest priority is assigned to DMA0, followed by DMA1, DMA2, and DMA3. DMA Channels with lower priority are paused until channels with higher priority have completed.
The CPU is paused when DMA transfers are active, however, the CPU is operating during the periods when Sound/Blanking DMA transfers are paused.

Special features of the separate DMA channels
DMA0 - highest priority, best for timing critical transfers (eg. HBlank DMA).
DMA1 and DMA2 - can be used to feed digital sample data to the Sound FIFOs.
DMA3 - can be used to write to Game Pak ROM/FlashROM (but not GamePak SRAM).
Beside for that, each DMA 0-3 may be used for whatever general purposes.

40000B0h,0B2h - DMA0SAD - DMA 0 Source Address (W) (internal memory)
40000BCh,0BEh - DMA1SAD - DMA 1 Source Address (W) (any memory)
40000C8h,0CAh - DMA2SAD - DMA 2 Source Address (W) (any memory)
40000D4h,0D6h - DMA3SAD - DMA 3 Source Address (W) (any memory)
The most significant address bits are ignored, only the least significant 27 or 28 bits are used (max 07FFFFFFh internal memory, or max 0FFFFFFFh any memory - except SRAM ???!).

40000B4h,0B6h - DMA0DAD - DMA 0 Destination Address (W) (internal memory)
40000C0h,0C2h - DMA1DAD - DMA 1 Destination Address (W) (internal memory)
40000CCh,0CEh - DMA2DAD - DMA 2 Destination Address (W) (internal memory)
40000D8h,0DAh - DMA3DAD - DMA 3 Destination Address (W) (any memory)
The most significant address bits are ignored, only the least significant 27 or 28 bits are used (max. 07FFFFFFh internal memory or 0FFFFFFFh any memory - except SRAM ???!).

40000B8h - DMA0CNT_L - DMA 0 Word Count (W) (14 bit, 1..4000h)
40000C4h - DMA1CNT_L - DMA 1 Word Count (W) (14 bit, 1..4000h)
40000D0h - DMA2CNT_L - DMA 2 Word Count (W) (14 bit, 1..4000h)
40000DCh - DMA3CNT_L - DMA 3 Word Count (W) (16 bit, 1..10000h)
Specifies the number of data units to be transferred, each unit is 16bit or 32bit depending on the transfer type, a value of zero is treated as max length (ie. 4000h, or 10000h for DMA3).

40000BAh - DMA0CNT_H - DMA 0 Control (R/W)
40000C6h - DMA1CNT_H - DMA 1 Control (R/W)
40000D2h - DMA2CNT_H - DMA 2 Control (R/W)
40000DEh - DMA3CNT_H - DMA 3 Control (R/W)
  Bit   Expl.
  0-4   Not used
  5-6   Dest Addr Control  (0=Increment,1=Decrement,2=Fixed,3=Increment/Reload)
  7-8   Source Adr Control (0=Increment,1=Decrement,2=Fixed,3=Prohibited)
  9     DMA Repeat                   (0=Off, 1=On) (Must be zero if Bit 11 set)
  10    DMA Transfer Type            (0=16bit, 1=32bit)
  11    Game Pak DRQ  - DMA3 only -  (0=Normal, 1=DRQ <from> Game Pak, DMA3)
  12-13 DMA Start Timing  (0=Immediately, 1=VBlank, 2=HBlank, 3=Special)
          The 'Special' setting (Start Timing=3) depends on the DMA channel:
          DMA0=Prohibited, DMA1/DMA2=Sound FIFO, DMA3=Video Capture
  14    IRQ upon end of Word Count   (0=Disable, 1=Enable)
  15    DMA Enable                   (0=Off, 1=On)
After changing the Enable bit from 0 to 1, wait 2 clock cycles before accessing any DMA related registers.

When accessing OAM (7000000h) or OBJ VRAM (6010000h) by HBlank Timing, then the "H-Blank Interval Free" bit in DISPCNT register must be set.

Source and Destination Address and Word Count Registers
The SAD, DAD, and CNT_L registers are holding the initial start addresses, and initial length. The hardware does NOT change the content of these registers during or after the transfer.
The actual transfer takes place by using internal pointer/counter registers. The initial values are copied into internal regs under the following circumstances:
Upon DMA Enable (Bit 15) changing from 0 to 1: Reloads SAD, DAD, CNT_L.
Upon Repeat: Reloads CNT_L, and optionally DAD (Increment+Reload).

DMA Repeat bit
If the Repeat bit is cleared: The Enable bit is automatically cleared after the specified number of data units has been transferred.
If the Repeat bit is set: The Enable bit remains set after the transfer, and the transfer will be restarted each time when the Start condition (eg. HBlank, Fifo) becomes true. The specified number of data units is transferred <each> time when the transfer is (re-)started. The transfer will be repeated forever, until it gets stopped by software.

Sound DMA (FIFO Timing Mode) (DMA1 and DMA2 only)
In this mode, the DMA Repeat bit must be set, and the destination address must be FIFO_A (040000A0h) or FIFO_B (040000A4h).
Upon DMA request from sound controller, 4 units of 32bits (16 bytes) are transferred (both Word Count register and DMA Transfer Type bit are ignored). The destination address will not be incremented in FIFO mode.
Keep in mind that DMA channels of higher priority may offhold sound DMA. For example, when using a 64 kHz sample rate, 16 bytes of sound DMA data are requested each 0.25ms (4 kHz), at this time another 16 bytes are still in the FIFO so that there's still 0.25ms time to satisfy the DMA request. Thus DMAs with higher priority should not be operated for longer than 0.25ms. (This problem does not arise for HBlank transfers as HBlank time is limited to 16.212us.)

Game Pak DMA
Only DMA 4 may be used to transfer data to/from Game Pak ROM or Flash ROM - it cannot access Game Pak SRAM though (as SRAM data bus is limited to 8bit units). In normal mode, DMA is requested as long until Word Count becomes zero. When setting the 'Game Pack DRQ' bit, then the cartridge must contain an external circuit which outputs a /DREQ signal. Note that there is only one pin for /DREQ and /IREQ, thus the cartridge may not supply /IREQs while using DRQ mode.

Video Capture Mode (DMA3 only)
Intended to copy a bitmap from memory (or from external hardware/camera) to VRAM. When using this transfer mode, set the repeat bit, and write the number of data units (per scanline) to the word count register. Capture works similar like HBlank DMA, however, the transfer is started when VCOUNT=2, it is then repeated each scanline, and it gets stopped when VCOUNT=162.

Transfer End
The DMA Enable flag (Bit 15) is automatically cleared upon completion of the transfer. The user may also clear this bit manually in order to stop the transfer (obviously this is possible for Sound/Blanking DMAs only, in all other cases the CPU is stopped until the transfer completes by itself).

Transfer Rate/Timing
Except for the first data unit, all units are transferred by sequential reads and writes. For n data units, the DMA transfer time is:
  2N+2(n-1)S+xI
Of which, 1N+(n-1)S are read cycles, and the other 1N+(n-1)S are write cycles, actual number of cycles depends on the waitstates and bus-width of the source and destination areas (as described in CPU Instruction Cycle Times chapter). Internal time for DMA processing is 2I (normally), or 4I (if both source and destination are in gamepak memory area).

DMA lockup when stopping while starting ???
Capture delayed, Capture Enable=AutoCleared ???


 Communication Ports

The GBAs Serial Port may be used in various different communication modes. Normal mode may exchange data between two GBAs (or to transfer data from master GBA to several slave GBAs in one-way direction).
Multi-player mode may exchange data between up to four GBAs. UART mode works much like a RS232 interface. JOY Bus mode uses a standardized Nintendo protocol. And General Purpose mode allows to mis-use the 'serial' port as bi-directional 4bit parallel port.
Note: The Nintendo DS does not include a Serial Port.

SIO Normal Mode
SIO Multi-Player Mode
SIO UART Mode
SIO JOY BUS Mode
SIO General-Purpose Mode
SIO Control Registers Summary

Infrared Communication Adapters
Even though early GBA prototypes have been intended to support IR communication, this feature has been removed.
However, Nintendo is apparently considering to provide an external IR adapter (to be connected to the SIO connector, being accessed in General Purpose mode).
Also, it'd be theoretically possible to include IR ports built-in in game cartridges (as done for some older 8bit/monochrome Hudson games).


 SIO Normal Mode

This mode is used to communicate between two units.
Transfer rates of 256KBit/s or 2MBit/s can be selected, however, the fast 2MBit/s is intended ONLY for special hardware expansions that are DIRECTLY connected to the GBA link port (ie. without a cable being located between the GBA and expansion hardware). In normal cases, always use 256KBit/s transfer rate which provides stable results.
Transfer lengths of 8bit or 32bit may be used, the 8bit mode is the same as for older DMG/CGB gameboys, however, the voltages for "GBA cartridges in GBAs" are different as for "DMG/CGB cartridges in DMG/CGB/GBAs", ie. it is not possible to communicate between DMG/CGB games and GBA games.

4000134h - RCNT (R) - Mode Selection, in Normal/Multiplayer/UART modes (R/W)
  Bit   Expl.
  0-3   Undocumented (current SC,SD,SI,SO state, as for General Purpose mode)
  4-8   Not used     (Should be 0, bits are read/write-able though)
  9-13  Not used     (Always 0, read only)
  14    Not used     (Should be 0, bit is read/write-able though)
  15    Must be zero (0) for Normal/Multiplayer/UART modes

4000128h - SIOCNT - SIO Control, usage in NORMAL Mode (R/W)
  Bit   Expl.
  0     Shift Clock (SC)        (0=External, 1=Internal)
  1     Internal Shift Clock    (0=256KHz, 1=2MHz)
  2     SI State (opponents SO) (0=Low, 1=High/None) --- (Read Only)
  3     SO during inactivity    (0=Low, 1=High) (applied ONLY when Bit7=0)
  4-6   Not used                (Read only, always 0 ???)
  7     Start Bit               (0=Inactive/Ready, 1=Start/Active)
  8-11  Not used                (R/W, should be 0)
  12    Transfer Length         (0=8bit, 1=32bit)
  13    Must be "0" for Normal Mode
  14    IRQ Enable              (0=Disable, 1=Want IRQ upon completion)
  15    Not used                (Read only, always 0)
The Start bit is automatically reset when the transfer completes, ie. when all 8 or 32 bits are transferred, at that time an IRQ may be generated.

400012Ah - SIODATA8 - SIO Normal Communication 8bit Data (R/W)
For 8bit normal mode. Contains 8bit data (only lower 8bit are used). Outgoing data should be written to this register before starting the transfer. During transfer, transmitted bits are shifted-out (MSB first), and received bits are shifted-in simultaneously. Upon transfer completion, the register contains the received 8bit value.

4000120h - SIODATA32_L - SIO Normal Communication lower 16bit data (R/W)
4000122h - SIODATA32_H - SIO Normal Communication upper 16bit data (R/W)
Same as above SIODATA8, for 32bit normal transfer mode respectively.
SIOCNT/RCNT must be set to 32bit normal mode <before> writing to SIODATA32.

Initialization
First, initialize RCNT register. Second, set mode/clock bits in SIOCNT with startbit cleared. For master: select internal clock, and (in most cases) specify 256KHz as transfer rate. For slave: select external clock, the local transfer rate selection is then ignored, as the transfer rate is supplied by the remote GBA (or other computer, which might supply custom transfer rates).
Third, set the startbit in SIOCNT with mode/clock bits unchanged.

Recommended Communication Procedure for SLAVE unit (external clock)
- Initialize data which is to be sent to master.
- Set Start flag.
- Set SO to LOW to indicate that master may start now.
- Wait for IRQ (or for Start bit to become zero). (Check timeout here!)
- Set SO to HIGH to indicate that we are not ready.
- Process received data.
- Repeat procedure if more data is to be transferred.
(or is so=high done automatically ??? would be fine - more stable - otherwise master may still need delay)

Recommended Communication Procedure for SLAVE unit (external clock)
- Initialize data which is to be sent to master.
- Set Start=0 and SO=0 (SO=LOW indicates that slave is (almost) ready).
- Set Start=1 and SO=1 (SO=HIGH indicates not ready, applied after transfer).
  (Expl. Old SO=LOW kept output until 1st clock bit received).
  (Expl. New SO=HIGH is automatically output at transfer completion).
- Set SO to LOW to indicate that master may start now.
- Wait for IRQ (or for Start bit to become zero). (Check timeout here!)
- Process received data.
- Repeat procedure if more data is to be transferred.

Recommended Communication Procedure for MASTER unit (internal clock)
- Initialize data which is to be sent to slave.
- Wait for SI to become LOW (slave ready). (Check timeout here!)
- Set Start flag.
- Wait for IRQ (or for Start bit to become zero).
- Process received data.
- Repeat procedure if more data is to be transferred.

Cable Protocol
During inactive transfer, the shift clock (SC) is high. The transmit (SO) and receive (SI) data lines may be manually controlled as described above.
When master sends SC=LOW, each master and slave must output the next outgoing data bit to SO. When master sends SC=HIGH, each master and slave must read out the opponents data bit from SI. This is repeated for each of the 8 or 32 bits, and when completed SC will be kept high again.

Transfer Rates
Either 256KHz or 2MHz rates can be selected for SC, so max 32KBytes (256KBit) or 128KBytes (2MBit) can be transferred per second. However, the software must process each 8bit or 32bit of transmitted data separately, so the actual transfer rate will be reduced by the time spent on handling each data unit.
Only 256KHz provides stable results in most cases (such like when linking between two GBAs). The 2MHz rate is intended for special expansion hardware only.

Using Normal mode for One-Way Multiplayer communication
When using normal mode with multiplay-cables, data isn't exchanged between first and second GBA as usually. Instead, data is shifted from first to last GBA (the first GBA receives zero, because master SI is shortcut to GND).
This behaviour may be used for fast ONE-WAY data transfer from master to all other GBAs. For example (3 GBAs linked):
  Step         Sender      1st Recipient   2nd Recipient
  Transfer 1:  DATA #0 --> UNDEF      -->  UNDEF     -->
  Transfer 2:  DATA #1 --> DATA #0    -->  UNDEF     -->
  Transfer 3:  DATA #2 --> DATA #1    -->  DATA #0   -->
  Transfer 4:  DATA #3 --> DATA #2    -->  DATA #1   -->
The recipients should not output any own data, instead they should forward the previously received data to the next recipient during next transfer (just keep the incoming data unmodified in the data register).
Due to the delayed forwarding, 2nd recipient should ignore the first incoming data. After the last transfer, the sender must send one (or more) dummy data unit(s), so that the last data is forwarded to the 2nd (or further) recipient(s).


 SIO Multi-Player Mode

Multi-Player mode can be used to communicate between up to 4 units.

4000134h - RCNT (R) - Mode Selection, in Normal/Multiplayer/UART modes (R/W)
  Bit   Expl.
  0-3   Undocumented (current SC,SD,SI,SO state, as for General Purpose mode)
  4-8   Not used     (Should be 0, bits are read/write-able though)
  9-13  Not used     (Always 0, read only)
  14    Not used     (Should be 0, bit is read/write-able though)
  15    Must be zero (0) for Normal/Multiplayer/UART modes
Note: Even though undocumented, many Nintendo games are using Bit 0 to test current SC state in multiplay mode.

4000128h - SIOCNT - SIO Control, usage in MULTI-PLAYER Mode (R/W)
  Bit   Expl.
  0-1   Baud Rate     (0-3: 9600,38400,57600,115200 bps)
  2     SI-Terminal   (0=Parent, 1=Child)                  (Read Only)
  3     SD-Terminal   (0=Bad connection, 1=All GBAs Ready) (Read Only)
  4-5   Multi-Player ID     (0=Parent, 1-3=1st-3rd child)  (Read Only)
  6     Multi-Player Error  (0=Normal, 1=Error)            (Read Only)
  7     Start/Busy Bit      (0=Inactive, 1=Start/Busy) (Read Only for Slaves)
  8-11  Not used            (R/W, should be 0)
  12    Must be "0" for Multi-Player mode
  13    Must be "1" for Multi-Player mode
  14    IRQ Enable          (0=Disable, 1=Want IRQ upon completion)
  15    Not used            (Read only, always 0)
The ID Bits are undefined until the first transfer has completed.

400012Ah - SIOMLT_SEND - Data Send Register (R/W)
Outgoing data (16 bit) which is to be sent to the other GBAs.

4000120h - SIOMULTI0 - SIO Multi-Player Data 0 (Parent) (R/W)
4000122h - SIOMULTI1 - SIO Multi-Player Data 1 (1st child) (R/W)
4000124h - SIOMULTI2 - SIO Multi-Player Data 2 (2nd child) (R/W)
4000126h - SIOMULTI3 - SIO Multi-Player Data 3 (3rd child) (R/W)
These registers are automatically reset to FFFFh upon transfer start.
After transfer, these registers contain incoming data (16bit each) from all remote GBAs (if any / otherwise still FFFFh), as well as the local outgoing SIOMLT_SEND data.
Ie. after the transfer, all connected GBAs will contain the same values in their SIOMULTI0-3 registers.

Initialization
- Initialize RCNT Bit 14-15 and SIOCNT Bit 12-13 to select Multi-Player mode.
- Read SIOCNT Bit 3 to verify that all GBAs are in Multi-Player mode.
- Read SIOCNT Bit 2 to detect whether this is the Parent/Master unit.

Recommended Transmission Procedure
- Write outgoing data to SIODATA_SEND.
- Master must set Start bit.
- All units must process received data in SIOMULTI0-3 when transfer completed.
- After the first successful transfer, ID Bits in SIOCNT are valid.
- If more data is to be transferred, repeat procedure.
The parent unit blindly sends data regardless of whether childs have already processed old data/supplied new data. So, parent unit might be required to insert delays between each transfer, and/or perform error checking.
Also, slave units may signalize that they are not ready by temporarily switching into another communication mode (which does not output SD High, as Multi-Player mode does during inactivity).

Transfer Protocol
Beginning
- The masters SI pin is always LOW.
- When all GBAs are in Multiplayer mode (ready) SD is HIGH.
- When master starts the transfer, it sets SC=LOW, slaves receive Busy bit.
Step A
- ID Bits in master unit are set to 0.
- Master outputs Startbit (LOW), 16bit Data, Stopbit (HIGH) through SD.
- This data is written to SIOMULTI0 of all GBAs (including master).
- Master forwards LOW from its SO to 1st childs SI.
- Transfer ends if next child does not output data after certain time.
Step B
- ID Bits in 1st child unit are set to 1.
- 1st Child outputs Startbit (LOW), 16bit Data, Stopbit (HIGH) through SD.
- This data is written to SIOMULTI1 of all GBAs (including 1st child).
- 1st child forwards LOW from its SO to 2nd childs SI.
- Transfer ends if next child does not output data after certain time.
Step C
- ID Bits in 2nd child unit are set to 2.
- 2nd Child outputs Startbit (LOW), 16bit Data, Stopbit (HIGH) through SD.
- This data is written to SIOMULTI2 of all GBAs (including 2nd child).
- 2nd child forwards LOW from its SO to 3rd childs SI.
- Transfer ends if next child does not output data after certain time.
Step D
- ID Bits in 3rd child unit are set to 3.
- 3rd Child outputs Startbit (LOW), 16bit Data, Stopbit (HIGH) through SD.
- This data is written to SIOMULTI3 of all GBAs (including 3rd child).
- Transfer ends (this was the last child).
Transfer end
- Master sets SC=HIGH, all GBAs set SO=HIGH.
- The Start/Busy bits of all GBAs are automatically cleared.
- Interrupts are requested in all GBAs (as far as enabled).

Error Bit
This bit is set when a slave did not receive SI=LOW even though SC=LOW signalized a transfer (this might happen when connecting more than 4 GBAs, or when the previous child is not connected). Also, the bit is set when a Stopbit wasn't HIGH.
The error bit may be undefined during active transfer - read only after transfer completion (the transfer continues and completes as normal even if errors have occurred for some or all GBAs).
Don't know: The bit is automatically reset/initialized with each transfer, or must be manually reset ???

Transmission Time
The transmission time depends on the selected Baud rate. And on the amount of Bits (16 data bits plus start/stop bits for each GBA), delays between each GBA, plus final timeout (if less than 4 GBAs). That is, depending on the number of connected GBAs:
  GBAs    Bits    Delays   Timeout
  1       18      None     Yes
  2       36      1        Yes
  3       54      2        Yes
  4       72      3        None
(The average Delay and Timeout periods are unknown ???)
Above is not counting the additional CPU time that must be spent on initiating and processing each transfer.

Fast One-Way Transmission
Beside for the actual SIO Multiplayer mode, you could also use SIO Normal mode for fast one-way data transfer from Master unit to all Child unit(s). See chapter about SIO Normal mode for details.


 SIO UART Mode

This mode works much like a RS232 port, however, the voltages are unknown, probably 0/3V rather than +/-12V ???. SI and SO are data lines (with crossed wires), SC and SD signalize Clear to Send (with crossed wires also, which requires special cable when linking between two GBAs ???)

4000134h - RCNT (R) - Mode Selection, in Normal/Multiplayer/UART modes (R/W)
  Bit   Expl.
  0-3   Undocumented (current SC,SD,SI,SO state, as for General Purpose mode)
  4-8   Not used     (Should be 0, bits are read/write-able though)
  9-13  Not used     (Always 0, read only)
  14    Not used     (Should be 0, bit is read/write-able though)
  15    Must be zero (0) for Normal/Multiplayer/UART modes

4000128h - SCCNT_L - SIO Control, usage in UART Mode (R/W)
  Bit   Expl.
  0-1   Baud Rate  (0-3: 9600,38400,57600,115200 bps)
  2     CTS Flag   (0=Send always/blindly, 1=Send only when SC=LOW)
  3     Parity Control (0=Even, 1=Odd)
  4     Send Data Flag      (0=Not Full,  1=Full)    (Read Only)
  5     Receive Data Flag   (0=Not Empty, 1=Empty)   (Read Only)
  6     Error Flag          (0=No Error,  1=Error)   (Read Only)
  7     Data Length         (0=7bits,   1=8bits)
  8     FIFO Enable Flag    (0=Disable, 1=Enable)
  9     Parity Enable Flag  (0=Disable, 1=Enable)
  10    Send Enable Flag    (0=Disable, 1=Enable)
  11    Receive Enable Flag (0=Disable, 1=Enable)
  12    Must be "1" for UART mode
  13    Must be "1" for UART mode
  14    IRQ Enable          (0=Disable, 1=IRQ when any Bit 4/5/6 become set)
  15    Not used            (Read only, always 0)

400012Ah - SIODATA8 - usage in UART Mode (R/W)
Addresses the send/receive shift register, or (when FIFO is used) the send/receive FIFO. In either case only the lower 8bit of SIODATA8 are used, the upper 8bit are not used.
The send/receive FIFO may store up to four 8bit data units each. For example, while 1 unit is still transferred from the send shift register, it is possible to deposit another 4 units in the send FIFO, which are then automatically moved to the send shift register one after each other.

Send/Receive Enable, CTS Feedback
The receiver outputs SD=LOW (which is input as SC=LOW at the remote side) when it is ready to receive data (that is, when Receive Enable is set, and the Receive shift register (or receive FIFO) isn't full.
When CTS flag is set to always/blindly, then the sender transmits data immediately when Send Enable is set, otherwise data is transmitted only when Send Enable is set and SC is LOW.

Error Flag
The error flag is set when a bad stop bit has been received (stop bit must be 0), when a parity error has occurred (if enabled), or when new data has been completely received while the receive data register (or receive FIFO) is already full.
The error flag is automatically reset when reading from SIOCNT register.

Init & Initback
The content of the FIFO is reset when FIFO is disabled in UART mode, thus, when entering UART mode initially set FIFO=disabled.
The Send/Receive enable bits must be reset before switching from UART mode into another SIO mode!


 SIO JOY BUS Mode

This communication mode uses Nintendo's standardized JOY Bus protocol. When using this communication mode, the GBA is always operated as SLAVE!

In this mode, SI and SO pins are data lines (apparently synchronized by Start/Stop bits ???), SC and SD are set to low (including during active transfer ???), the transfer rate is unknown ???

4000134h - RCNT (R) - Mode Selection, in JOY BUS mode (R/W)
  Bit   Expl.
  0-3   Undocumented (current SC,SD,SI,SO state, as for General Purpose mode)
  4-8   Not used     (Should be 0, bits are read/write-able though)
  9-13  Not used     (Always 0, read only)
  14    Must be "1" for JOY BUS Mode
  15    Must be "1" for JOY BUS Mode

4000128h - SIOCNT - SIO Control, not used in JOY BUS Mode
This register is not used in JOY BUS mode.

4000140h - JOYCNT - JOY BUS Control Register (R/W)
  Bit   Expl.
  0     Device Reset Flag     (Command FFh)          (Read/Acknowledge)
  1     Receive Complete Flag (Command 14h or 15h?)  (Read/Acknowledge)
  2     Send Complete Flag    (Command 15h or 14h?)  (Read/Acknowledge)
  3-5   Not used
  6     IRQ when receiving a Device Reset Command  (0=Disable, 1=Enable)
  7-15  Not used
Bit 0-2 are working much like the bits in the IF register: Write a "1" bit to reset (acknowledge) the respective bit.
UNCLEAR: Interrupts can be requested for Send/Receive commands also ???

4000150h - JOY_RECV_L - Receive Data Register low (R/W)
4000152h - JOY_RECV_H - Receive Data Register high (R/W)
4000154h - JOY_TRANS_L - Send Data Register low (R/W)
4000156h - JOY_TRANS_H - Send Data Register high (R/W)
Send/receive data registers.

4000158h - JOYSTAT - Receive Status Register (R/W)
  Bit   Expl.
  0     Not used
  1     Receive Status Flag   (0=Remote GBA is/was receiving) (Read Only?)
  2     Not used
  3     Send Status Flag      (1=Remote GBA is/was sending)   (Read Only?)
  4-5   General Purpose Flag  (Not assigned, may be used for whatever purpose)
  6-15  Not used
Bit 1 is automatically set when writing to local JOY_TRANS.
Bit 3 is automatically reset when reading from local JOY_RECV.

Below are the four possible commands which can be received by the GBA. Note that the GBA (slave) cannot send any commands itself, all it can do is to read incoming data, and to provide 'reply' data which may (or may not) be read out by the master unit.

Command FFh - Device Reset
  Receive FFh (Command)
  Send    00h (GBA Type number LSB (or MSB?))
  Send    04h (GBA Type number MSB (or LSB?))
  Send    XXh (lower 8bits of SIOSTAT register)

Command 00h - Type/Status Data Request
  Receive 00h (Command)
  Send    00h (GBA Type number LSB (or MSB?))
  Send    04h (GBA Type number MSB (or LSB?))
  Send    XXh (lower 8bits of SIOSTAT register)

Command 15h - GBA Data Write (to GBA)
  Receive 15h (Command)
  Receive XXh (Lower 8bits of JOY_RECV_L)
  Receive XXh (Upper 8bits of JOY_RECV_L)
  Receive XXh (Lower 8bits of JOY_RECV_H)
  Receive XXh (Upper 8bits of JOY_RECV_H)
  Send    XXh (lower 8bits of SIOSTAT register)

Command 14h - GBA Data Read (from GBA)
  Receive 14h (Command)
  Send    XXh (Lower 8bits of JOY_TRANS_L)
  Send    XXh (Upper 8bits of JOY_TRANS_L)
  Send    XXh (Lower 8bits of JOY_TRANS_H)
  Send    XXh (Upper 8bits of JOY_TRANS_H)
  Send    XXh (lower 8bits of SIOSTAT register)


 SIO General-Purpose Mode

In this mode, the SIO is 'misused' as a 4bit bi-directional parallel port, each of the SI,SO,SC,SD pins may be directly controlled, each can be separately declared as input (with internal pull-up) or as output signal.

4000134h - RCNT (R) - SIO Mode, usage in GENERAL-PURPOSE Mode (R/W)
Interrupts can be requested when SI changes from HIGH to LOW, as General Purpose mode does not require a serial shift clock, this interrupt may be produced even when the GBA is in Stop (low power standby) state.
  Bit   Expl.
  0     SC Data Bit         (0=Low, 1=High)
  1     SD Data Bit         (0=Low, 1=High)
  2     SI Data Bit         (0=Low, 1=High)
  3     SO Data Bit         (0=Low, 1=High)
  4     SC Direction        (0=Input, 1=Output)
  5     SD Direction        (0=Input, 1=Output)
  6     SI Direction        (0=Input, 1=Output, but see below)
  7     SO Direction        (0=Input, 1=Output)
  8     SI Interrupt Enable (0=Disable, 1=Enable)
  9-13  Not used
  14    Must be "0" for General-Purpose Mode
  15    Must be "1" for General-Purpose or JOYBUS Mode
SI should be always used as Input to avoid problems with other hardware which does not expect data to be output there.

4000128h - SIOCNT - SIO Control, not used in GENERAL-PURPOSE Mode
This register is not used in general purpose mode. That is, the separate bits of SIOCNT still exist and are read- and/or write-able in the same manner as for Normal, Multiplay, or UART mode (depending on SIOCNT Bit 12,13), but are having no effect on data being output to the link port.


 SIO Control Registers Summary

Mode Selection (by RCNT.15-14 and SIOCNT.13-12)
  R.15 R.14 S.13 S.12 Mode
  0    x    0    0    Normal 8bit
  0    x    0    1    Normal 32bit
  0    x    1    0    Multiplay 16bit
  0    x    1    1    UART (RS232)
  1    0    x    x    General Purpose
  1    1    x    x    JOY BUS

SIOCNT
  Bit Normal Multi
         0      1    2     3      4 5 6   7     8    9      10   11
  Normal Master Rate SI/In SO/Out - - -   Start -    -      -    -
  Multi  Baud   Baud SI/In SD/In  ID# Err Start -    -      -    -
  UART   Baud   Baud CTS   Parity S R Err Bits  FIFO Parity Send Recv


 Infrared Communication

Early GBA prototypes have been intended to include a built-in IR port for sending and receiving IR signals. Among others, this port could have been used to communicate with other GBAs, or older CGB models, or TV Remote Controls, etc.

[ THE INFRARED COMMUNICATION FEATURE IS -NOT- SUPPORTED ANYMORE ]
Anyways, the prototype specifications have been as shown below...

Keep in mind that the IR signal may be interrupted by whatever objects moved between sender and receiver - the IR port isn't recommended for programs that require realtime data exchange (such like action games).

4000136h - IR - Infrared Register (R/W)
  Bit   Expl.
  0     Transmission Data  (0=LED Off, 1=LED On)
  1     READ Enable        (0=Disable, 1=Enable)
  2     Reception Data     (0=None, 1=Signal received) (Read only)
  3     AMP Operation      (0=Off, 1=On)
  4     IRQ Enable Flag    (0=Disable, 1=Enable)
  5-15  Not used
When IRQ is enabled, an interrupt is requested if the incoming signal was 0.119us Off (2 cycles), followed by 0.536us On (9 cycles) - minimum timing periods each.

Transmission Notes
When transmitting an IR signal, note that it'd be not a good idea to keep the LED turned On for a very long period (such like sending a 1 second synchronization pulse). The recipient's circuit would treat such a long signal as "normal IR pollution which is in the air" after a while, and thus ignore the signal.

Reception Notes
Received data is internally latched. Latched data may be read out by setting both READ and AMP bits.
Note: Provided that you don't want to receive your own IR signal, be sure to set Bit 0 to zero before attempting to receive data.

Power-consumption
After using the IR port, be sure to reset the register to zero in order to reduce battery power consumption.


 Keypad Input

The built-in GBA gamepad has 4 direction keys, and 6 buttons.

4000130h - KEYINPUT - Key Status (R)
  Bit   Expl.
  0     Button A        (0=Pressed, 1=Released)
  1     Button B        (etc.)
  2     Select          (etc.)
  3     Start           (etc.)
  4     Right           (etc.)
  5     Left            (etc.)
  6     Up              (etc.)
  7     Down            (etc.)
  8     Button R        (etc.)
  9     Button L        (etc.)
  10-15 Not used
It'd be usually recommended to read-out this register only once per frame, and to store the current state in memory. As a side effect, this method avoids problems caused by switch bounce when a key is newly released or pressed.

4000132h - KEYCNT - Key Interrupt Control (R/W)
The keypad IRQ function is intended to terminate the very-low-power Stop mode, it is not suitable for processing normal user input, to do this, most programs are invoking their keypad handlers from within VBlank IRQ.
  Bit   Expl.
  0     Button A        (0=Ignore, 1=Select)
  1     Button B        (etc.)
  2     Select          (etc.)
  3     Start           (etc.)
  4     Right           (etc.)
  5     Left            (etc.)
  6     Up              (etc.)
  7     Down            (etc.)
  8     Button R        (etc.)
  9     Button L        (etc.)
  10-13 Not used
  14    IRQ Enable Flag (0=Disable, 1=Enable)
  15    IRQ Condition   (0=Logical OR, 1=Logical AND)
In logical OR mode, an interrupt is requested when at least one of the selected buttons is pressed.
In logical AND mode, an interrupt is requested when ALL of the selected buttons are pressed.

Notes
In 8bit gameboy compatibility mode, L and R Buttons are used to toggle the screen size between normal 160x144 pixels and stretched 240x144 pixels.
The GBA SP is additionally having a * Button used to toggle the backlight on and off, as far as I know there's no way to detect the current button or backlight state by software.


 Interrupt Control

4000208h - IME - Interrupt Master Enable Register (R/W)
  Bit   Expl.
  0     Disable all interrupts         (0=Disable All, 1=See IE register)
  1-15  Not used

4000200h - IE - Interrupt Enable Register (R/W)
  Bit   Expl.
  0     LCD V-Blank                    (0=Disable)
  1     LCD H-Blank                    (etc.)
  2     LCD V-Counter Match            (etc.)
  3     Timer 0 Overflow               (etc.)
  4     Timer 1 Overflow               (etc.)
  5     Timer 2 Overflow               (etc.)
  6     Timer 3 Overflow               (etc.)
  7     Serial Communication           (etc.)
  8     DMA 0                          (etc.)
  9     DMA 1                          (etc.)
  10    DMA 2                          (etc.)
  11    DMA 3                          (etc.)
  12    Keypad                         (etc.)
  13    Game Pak (external IRQ source) (etc.)
  14-15 Not used
Note that there is another 'master enable flag' directly in the CPUs Status Register (CPSR) accessible in privileged modes, see CPU reference for details.

4000202h - IF - Interrupt Request Flags / IRQ Acknowledge (R/W, see below)
  Bit   Expl.
  0     LCD V-Blank                    (1=Request Interrupt)
  1     LCD H-Blank                    (etc.)
  2     LCD V-Counter Match            (etc.)
  3     Timer 0 Overflow               (etc.)
  4     Timer 1 Overflow               (etc.)
  5     Timer 2 Overflow               (etc.)
  6     Timer 3 Overflow               (etc.)
  7     Serial Communication           (etc.)
  8     DMA 0                          (etc.)
  9     DMA 1                          (etc.)
  10    DMA 2                          (etc.)
  11    DMA 3                          (etc.)
  12    Keypad                         (etc.)
  13    Game Pak (external IRQ source) (etc.)
  14-15 Not used
Interrupts must be manually acknowledged by writing a "1" to one of the IRQ bits, the IRQ bit will then be cleared.

"[Cautions regarding clearing IME and IE]
A corresponding interrupt could occur even while a command to clear IME or each flag of the IE register is being executed. When clearing a flag of IE, you need to clear IME in advance so that mismatching of interrupt checks will not occur." ???

"[When multiple interrupts are used]
When the timing of clearing of IME and the timing of an interrupt agree, multiple interrupts will not occur during that interrupt. Therefore, set (enable) IME after saving IME during the interrupt routine." ???

BIOS Interrupt handling
Upon interrupt execution, the CPU is switched into IRQ mode, and the physical interrupt vector is called - as this address is located in BIOS ROM, the BIOS will always always execute the following code before it forwards control to the user handler:
  00000018  b      128h                ;IRQ vector: jump to actual BIOS handler
  00000128  stmfd  r13!,r0-r3,r12,r14  ;save registers to SP_irq
  0000012C  mov    r0,4000000h         ;ptr+4 to 03FFFFFC (mirror of 03007FFC)
  00000130  add    r14,r15,0h          ;retadr for USER handler $+8=138h
  00000134  ldr    r15,[r0,-4h]        ;jump to [03FFFFFC] USER handler
  00000138  ldmfd  r13!,r0-r3,r12,r14  ;restore registers from SP_irq
  0000013C  subs   r15,r14,4h          ;return from IRQ (PC=LR-4, CPSR=SPSR)
As shown above, a pointer to the 32bit/ARM-code user handler must be setup in [03007FFCh]. By default, 160 bytes of memory are reserved for interrupt stack at 03007F00h-03007F9Fh.

Recommended User Interrupt handling
- If necessary switch to THUMB state manually (handler is called in ARM state)
- Determine reason(s) of interrupt by examining IF register
- User program may freely assign priority to each reason by own logic
- Process the most important reason of your choice
- User MUST manually acknowledge by writing to IF register
- If user wants to allow nested interrupts, save SPSR_irq, then enable IRQs.
- If using other registers than BIOS-pushed R0-R3, manually save R4-R11 also.
- Note that Interrupt Stack is used (which may have limited size)
- So, for memory consuming stack operations use system mode (=user stack).
- When calling subroutines in system mode, save LSR_usr also.
- Restore SPSR_irq and/or R4-R11 if you've saved them above.
- Finally, return to BIOS handler by BX LR (R14_irq) instruction.

Default memory usage at 03007FXX (and mirrored to 03FFFFXX)
  Addr. Size Expl.
  7FFCh 4    Pointer to user IRQ handler (32bit ARM code)
  7FF8h 4    Interrupt Check Flag (for IntrWait/VBlankIntrWait functions)
  7FF4h 4    Allocated Area
  7FF0h 4    Pointer to Sound Buffer
  7FE0h 16   Allocated Area
  7FA0h 64   Default area for SP_svc Supervisor Stack (4 words/time)
  7F00h 160  Default area for SP_irq Interrupt Stack (6 words/time)
Memory below 7F00h is free for User Stack and user data. The three stack pointers are initially initialized at the TOP of the respective areas:
  SP_svc=03007FE0h
  SP_irq=03007FA0h
  SP_usr=03007F00h
The user may redefine these addresses and move stacks into other locations, however, the addresses for system data at 7FE0h-7FFFh are fixed.

Not sure, is following free for user ???
Registers R8-R12_fiq, R13_fiq, R14_fiq, SPSR_fiq
Registers R13-R14_abt, SPSR_abt
Registers R13-R14_und, SPSR_und

Fast Interrupt (FIQ)
The ARM CPU provides two interrupt sources, IRQ and FIQ. In the GBA only IRQ is used. In normal GBAs, the FIQ signal is shortcut to VDD35, ie. the signal is always high, and there is no way to generate a FIQ by hardware. The registers R8..12_fiq could be used by software (when switching into FIQ mode by writing to CPSR) - however, this might make the game incompatible with hardware debuggers (which are reportedly using FIQs for debugging purposes).


 System Control

4000204h - WAITCNT - Waitstate Control (R/W)
This register is used to configure game pak access timings. The game pak ROM is mirrored to three address regions at 08000000h, 0A000000h, and 0C000000h, these areas are called Wait State 0-2. Different access timings may be assigned to each area (this might be useful in case that a game pak contains several ROM chips with different access times each).
  Bit   Expl.
  0-1   SRAM Wait Control          (0..3 = 4,3,2,8 cycles)
  2-3   Wait State 0 First Access  (0..3 = 4,3,2,8 cycles)
  4     Wait State 0 Second Access (0..1 = 2,1 cycles)
  5-6   Wait State 1 First Access  (0..3 = 4,3,2,8 cycles)
  7     Wait State 1 Second Access (0..1 = 4,1 cycles; unlike above WS0)
  8-9   Wait State 2 First Access  (0..3 = 4,3,2,8 cycles)
  10    Wait State 2 Second Access (0..1 = 8,1 cycles; unlike above WS0,WS1)
  11-12 PHI Terminal Output        (0..3 = Disable, 4.19MHz, 8.38MHz, 16.78MHz)
  13    Not used
  14    Game Pak Prefetch Buffer (Pipe) (0=Disable, 1=Enable)
  15    Game Pak Type Flag  (Read Only) (0=GBA, 1=CGB)
At startup, the default setting is 0000h. Currently manufactured cartridges are using the following settings: WS0/ROM=3,1 clks; SRAM=8 clks; WS2/EEPROM: 8,8 clks; prefetch enabled; that is, WAITCNT=4317h, for more info see "Cartridges" chapter.

First Access (Non-sequential) and Second Access (Sequential) define the waitstates for N and S cycles, the actual access time is 1 clock cycle PLUS the number of waitstates.
GamePak uses 16bit data bus, so that a 32bit access is split into TWO 16bit accesses (of which, the second fragment is always sequential, even if the first fragment was non-sequential).

GamePak Prefetch

NOTES:
The GBA forcefully uses non-sequential timing at the beginning of each 128K-block of gamepak ROM, eg. "LDMIA [801fff8h],r0-r7" will have non-sequential timing at 8020000h.
The PHI Terminal output (PHI Pin of Gamepak Bus) should be disabled.

4000300h - POSTFLG - BYTE - Undocumented - Post Boot / Debug Control (R/W)
After initial reset, the GBA BIOS initializes the register to 01h, and any further execution of the Reset vector (00000000h) will pass control to the Debug vector (0000001Ch) when sensing the register to be still set to 01h.
  Bit   Expl.
  0     Undocumented. First Boot Flag  (0=First, 1=Further)
  1-7   Undocumented. Not used.
Normally the debug handler rejects control unless it detects Debug flags in cartridge header, in that case it may redirect to a cut-down boot procedure (bypassing Nintendo logo and boot delays, much like nocash burst boot for multiboot software). I am not sure if it is possible to reset the GBA externally without automatically resetting register 300h though.

4000301h - HALTCNT - BYTE - Undocumented - Low Power Mode Control (W)
Writing to this register switches the GBA into battery saving mode.
In Halt mode, the CPU is paused as long as (IE AND IF)=0, this should be used to reduce power-consumption during periods when the CPU is waiting for interrupt events.
In Stop mode, most of the hardware including sound and video are paused, this very-low-power mode could be used much like a screensaver.
  Bit   Expl.
  0-6   Undocumented. Not used.
  7     Undocumented. Power Down Mode  (0=Halt, 1=Stop)
The current GBA BIOS addresses only the upper eight bits of this register (by writing 00h or 80h to address 04000301h), however, as the register isn't officially documented, some or all of the bits might have different meanings in future GBA models.
For best forwards compatibility, it'd generally be more recommended to use the BIOS Functions SWI 2 (Halt) or SWI 3 (Stop) rather than writing to this register directly.

4000410h - Undocumented - Purpose Unknown ??? 8bit (W)
The BIOS writes the 8bit value 0FFh to this address. Purpose Unknown.
Probably just another bug in the BIOS.

4000800h - 32bit - Undocumented - Internal Memory Control (R/W)
Supported by GBA and GBA SP only - NOT supported by DS (even in GBA mode).
Initialized to 0D000020h (by hardware). Unlike all other I/O registers, this register is mirrored across the whole 4XXXXXXh I/O area (in increments of 64K, ie. at 800h, 10800h, 20800h, etc.)
  Bit   Expl.
  0     Purpose Unknown  (Seems to lock up the GBA when set to 1)
  1-3   Purpose Unknown  (Read/Write able)
  4     Purpose Unknown  (Always zero - not used or write only)
  5     Purpose Unknown  (Seems to lock up the GBA when set to 0)
  6-23  Purpose Unknown  (Always zero - not used or write only)
  24-27 Wait Control WRAM 256K (0-14 = 15..1 Waitstates, 15=Lockup)
  28-31 Purpose Unknown  (Read/Write able)
The value 0Dh in Bits 24-27 selects 2 waitstates for 256K WRAM (ie. 3/3/6 cycles 8/16/32bit accesses). The fastest possible setting would be 0Eh (1 waitstate, 2/2/4 cycles for 8/16/32bit).

Note: One cycle equals approx. 59.59ns (ie. 16.78MHz clock).


 GamePak Prefetch

GamePak Prefetch can be enabled in WAITCNT register. When prefetch buffer is enabled, the GBA attempts to read opcodes from Game Pak ROM during periods when the CPU is not using the bus (if any). Memory access is then performed with 0 Waits if the CPU requests data which is already stored in the buffer. The prefetch buffer stores up to eight 16bit values.

GamePak ROM Opcodes
The prefetch feature works only with <opcodes> fetched from GamePak ROM. Opcodes executed in RAM or BIOS are not affected by the prefetch feature (even if that opcodes read <data> from GamePak ROM).

Prefetch Enable
For GamePak ROM opcodes, prefetch may occur in two situations:
  1) opcodes with internal cycles (I) which do not change R15, shift/rotate
     register-by-register, load opcodes (ldr,ldm,pop,swp), multiply opcodes
  2) opcodes that load/store memory (ldr,str,ldm,stm,etc.)

Prefetch Disable Bug
When Prefetch is disabled, the Prefetch Disable Bug will occur for
  <GamePak ROM opcodes with internal cycles which do not change R15>.
That are, Shift/rotate register-by-register opcodes, multiply opcodes, and load opcodes (ldr,ldm,pop,swp). The bug adds 2I cycles to any such opcodes, ie. a multiply opcode with 4I cycles will be delayed to 6I cycles.


 Cartridges

ROM
Cartridge Header
Cartridge ROM

Backup Media
Aside from ROM, cartridges may also include one of the following backup medias, used to store game positions, highscore tables, options, or other data.
Backup SRAM
Backup EEPROM
Backup Flash ROM
Backup DACS

Other Accessoires
Flashcards
Cheat Devices


 Cartridge Header

The first 192 bytes at 8000000h-80000BFh in ROM are used as cartridge header. The same header is also used for Multiboot images at 2000000h-20000BFh (plus some additional multiboot entries at 20000C0h and up).

Header Overview
  Address Bytes Expl.
  000h    4     ROM Entry Point  (32bit ARM branch opcode, eg. "B rom_start")
  004h    156   Nintendo Logo    (compressed bitmap, required!)
  0A0h    12    Game Title       (uppercase ascii, max 12 characters)
  0ACh    4     Game Code        (uppercase ascii, 4 characters)
  0B0h    2     Maker Code       (uppercase ascii, 2 characters)
  0B2h    1     Fixed value      (must be 96h, required!)
  0B3h    1     Main unit code   (00h for current GBA models)
  0B4h    1     Device type      (huh ???)
  0B5h    7     Reserved Area    (should be zero filled)
  0BCh    1     Software version (usually 00h)
  0BDh    1     Complement check (header checksum, required!)
  0BEh    2     Reserved Area    (should be zero filled)
  --- Additional Multiboot Header Entries ---
  0C0h    4     RAM Entry Point  (32bit ARM branch opcode, eg. "B ram_start")
  0C4h    1     Boot mode        (init as 00h - BIOS overwrites this value!)
  0C5h    1     Slave ID Number  (init as 00h - BIOS overwrites this value!)
  0C6h    26    Not used         (seems to be unused)
  0E4h    4     JOYBUS Entry Pt. (32bit ARM branch opcode, eg. "B joy_start")
Note: With all entry points, the CPU is initially set into system mode.

000h - Entry Point, 4 Bytes
Space for a single 32bit ARM opcode that redirects to the actual startaddress of the cartridge, this should be usually a "B <start>" instruction.
Note: This entry is ignored by Multiboot slave GBAs (in fact, the entry is then overwritten and redirected to a separate Multiboot Entry Point, as described below).

004h..09Fh - Nintendo Logo, 156 Bytes
Contains the Nintendo logo which is displayed during the boot procedure. Cartridge won't work if this data is missing or modified.
In detail: This area contains Huffman compression data (but excluding the compression header which is hardcoded in the BIOS, so that it'd be probably not possible to hack the GBA by producing de-compression buffer overflows).
A copy of the compression data is stored in the BIOS, the GBA will compare this data and lock-up itself if the BIOS data isn't exactly the same as in the cartridge (or multiboot header). The only exception are the two entries below which are allowed to have variable settings in some bits.

09Ch Bit 2,7 - Debugging Enable
This is part of the above Nintendo Logo area, and must be commonly set to 21h, however, Bit 2 and Bit 7 may be set to other values.
When both bits are set (ie. A5h), the FIQ/Undefined Instruction handler in the BIOS becomes unlocked, the handler then forwards these exceptions to the user handler in cartridge ROM (entry point defined in 80000B4h, see below).
Other bit combinations currently do not seem to have special functions.

09Eh Bit 0,1 - Cartridge Key Number MSBs
This is part of the above Nintendo Logo area, and must be commonly set to F8h, however, Bit 0-1 may be set to other values.
During startup, the BIOS performs some dummy-reads from a stream of pre-defined addresses, even though these reads seem to be meaningless, they might be intended to unlock a read-protection inside of commercial cartridge. There are 16 pre-defined address streams - selected by a 4bit key number - of which the upper two bits are gained from 800009Eh Bit 0-1, and the lower two bits from a checksum across header bytes 09Dh..0B7h (bytewise XORed, divided by 40h).

0A0h - Game Title, Uppercase Ascii, max 12 characters
Space for the game title. If less than 12 chars, SPACE or ZERO padded ???

0ACh - Game Code, Uppercase Ascii, 4 characters
This is the same code as the AGB-UTTD code which is printed on the package and sticker on (commercial) cartridges (excluding the leading "AGB-" part).
  U  Unique Code          ("A", "B", "C", "D", etc.)
  TT Short Title          (eg. "PM" for Pac Man)
  D  Destination/Language ("J"=Japan, "E"=USA/English, "P"=Europe/Elsewhere)
The first character (U) is usually "A", unless if the same Short Title was already used by other game(s).

0B0h - Maker code, Uppercase Ascii, 2 characters
Identifies the (commercial) developer. For example, "01"=Nintendo.

0B2h - Fixed value, 1 Byte
Must be 96h.

0B3h - Main unit code, 1 Byte
Identifies the required hardware. Should be 00h for current GBA models.

0B4h - Device type, 1 Byte
Normally, this entry should be zero. With Nintendos hardware debugger Bit 7 identifies the debugging handlers entry point and size of DACS (Debugging And Communication System) memory: Bit7=0: 9FFC000h/8MBIT DACS, Bit7=1: 9FE2000h/1MBIT DACS. The debugging handler can be enabled in 800009Ch (see above), normal cartridges do not have any memory (nor any mirrors) at these addresses though.

0B5h - Reserved Area, 7 Bytes
Reserved, zero filled.

0BCh - Software version number
Version number of the game. Usually zero.

0BDh - Complement check, 1 Byte
Header checksum, cartridge won't work if incorrect. Calculate as such:
chk=0:for i=0A0h to 0BCh:chk=chk-[i]:next:chk=(chk-19h) and 0FFh

0BEh - Reserved Area, 2 Bytes
Reserved, zero filled.

Below required for Multiboot/slave programs only. For Multiboot, the above 192 bytes are required to be transferred as header-block (loaded to 2000000h-20000BFh), and some additional header-information must be located at the beginning of the actual program/data-block (loaded to 20000C0h and up). This extended header consists of Multiboot Entry point(s) which must be set up correctly, and of two reserved bytes which are overwritten by the boot procedure:

0C0h - Normal/Multiplay mode Entry Point
This entry is used only if the GBA has been booted by using Normal or Multiplay transfer mode (but not by Joybus mode).
Typically deposit a ARM-32bit "B <start>" branch opcode at this location, which is pointing to your actual initialization procedure.

0C4h (BYTE) - Boot mode
The slave GBA download procedure overwrites this byte by a value which is indicating the used multiboot transfer mode.
  Value  Expl.
  01h    Joybus mode
  02h    Normal mode
  03h    Multiplay mode
Typically set this byte to zero by inserting DCB 00h in your source.
Be sure that your uploaded program does not contain important program code or data at this location, or at the ID-byte location below.

0C5h (BYTE) - Slave ID Number
If the GBA has been booted in Normal or Multiplay mode, this byte becomes overwritten by the slave ID number of the local GBA (that'd be always 01h for normal mode).
  Value  Expl.
  01h    Slave #1
  02h    Slave #2
  03h    Slave #3
Typically set this byte to zero by inserting DCB 00h in your source.
When booted in Joybus mode, the value is NOT changed and remains the same as uploaded from the master GBA.

0C6h..0DFh - Not used
Appears to be unused.

0E0h - Joybus mode Entry Point
If the GBA has been booted by using Joybus transfer mode, then the entry point is located at this address rather than at 20000C0h. Either put your initialization procedure directly at this address, or redirect to the actual boot procedure by depositing a "B <start>" opcode here (either one using 32bit ARM code). Or, if you are not intending to support joybus mode (which is probably rarely used), ignore this entry.


 Cartridge ROM

ROM Size
The games F-ZERO and Super Mario Advance use ROMs of 4 MBytes each. Zelda uses 8 MBytes. Not sure if other sizes are manufactured.

ROM Waitstates
The GBA starts the cartridge with 4,2 waitstates (N,S) and prefetch disabled. The program may change these settings by writing to WAITCNT, the games F-ZERO and Super Mario Advance use 3,1 waitstates (N,S) each, with prefetch enabled.
Third-party flashcards are reportedly running unstable with these settings. Also, prefetch and shorter waitstates are allowing to read more data and opcodes from ROM is less time, the downside is that it increases the power consumption.

ROM Chip
Because of how 24bit addresses are squeezed through the Gampak bus, the cartridge must include a circuit that latches the lower 16 address bits on non-sequential access, and that increments these bits on sequential access. Nintendo includes this circuit directly in the ROM chip.
Also, the ROM must have 16bit data bus, or otherwise another circuit is required which converts two 8bit data units into one 16bit unit - by not exceeding the waitstate timings.


 Backup SRAM

32 KBytes - 256Kbit Battery buffered SRAM - Lifetime: Depends on battery
64 KBytes - reportedly used by newer games ???

Addressing and Waitstates
SRAM is mapped to E000000h-E007FFFh, it should be accessed with 8 waitstates (write a value of 3 into Bit0-1 of WAITCNT).

Databus Width
The SRAM databus is restricted to 8 bits, it should be accessed by LDRB, LDRSB, and STRB opcodes only.

Reading and Writing
Reading from SRAM should be performed by code executed in WRAM only (but not by code executed in ROM). There is no such restriction for writing.

Preventing Data Loss
The GBA SRAM carts do not include a write-protect function (unlike older 8bit gameboy carts). This seems to be a problem and may cause data loss when a cartridge is removed or inserted while the GBA is still turned on. As far as I understand, this is not so much a hardware problem, but rather a software problem, ie. theoretically you could remove/insert the cartridge as many times as you want, but you should take care that your program does not crash (and write blindly into memory).

Recommended Workaround
Enable the Gamepak Interrupt (it'll most likely get triggered when removing the cartridge), and hang-up the GBA in an endless loop when your interrupt handler senses a Gamepak IRQ. For obvious reason, your interrupt handler should be located in WRAM, ie. not in the (removed) ROM cartridge. The handler should process Gamepak IRQs at highest priority. Periods during which interrupts are disabled should be kept as short as possible, if necessary allow nested interrupts.

When to use the above Workaround
A program that relies wholly on code and data in WRAM, and that does not crash even when ROM is removed, may keep operating without having to use the above mechanism.
Do NOT use the workaround for programs that run without a cartridge inserted (ie. single gamepak/multiboot slaves), or for programs that use Gamepak IRQ/DMA for other purposes.
All other programs should use it. It'd be eventually a good idea to include it even in programs that do not use SRAM themselves (eg. otherwise removing a SRAM-less cartridge may lock up the GBA, and may cause it to destroy backup data when inserting a SRAM cartridge).

Note
SRAM is used by the game F-ZERO, and possibly others.

In SRAM cartridges, the /REQ pin (Pin 31 of Gamepak bus) should be a little bit shorter as than the other pins; when removing the cartridge, this causes the gamepak IRQ signal to get triggered before the other pins are disconnected.


 Backup EEPROM

512 Bytes (0200h) - 4Kbit EEPROM - Lifetime: 100,000 writes per address
8 KBytes (2000h) - 64Kbit EEPROM - Lifetime: No info.

Addressing and Waitstates
The eeprom is connected to Bit0 of the data bus, and to the upper 1 bit (or upper 17 bits in case of large 32MB ROM) of the cartridge ROM address bus, communication with the chip takes place serially.
The eeprom must be used with 8 waitstates (set WAITCNT=X3XXh; 8,8 clks in WS2 area), the eeprom can be then addressed at DFFFF00h..DFFFFFFh.
Respectively, with eeprom, ROM is restricted to 8000000h-9FFFeFFh (max. 1FFFF00h bytes = 32MB minus 256 bytes). On carts with 16MB or smaller ROM, eeprom can be alternately accessed anywhere at D000000h-DFFFFFFh.

Data and Address Width
Data can be read from (or written to) the EEPROM in units of 64bits (8 bytes). Writing automatically erases the old 64bits of data. Addressing works in units of 64bits respectively, that is, for 512 Bytes EEPROMS: an address range of 0-3Fh, 6bit bus width; and for 8KByte EEPROMs: a range of 0-3FFh, 14bit bus width (only the lower 10 address bits are used, upper 4 bits should be zero).

Set Address (For Reading)
Prepare the following bitstream in memory:
  2 bits "11" (Read Request)
  n bits eeprom address (MSB first, 6 or 14 bits, depending on EEPROM)
  1 bit "0"
Then transfer the stream to eeprom by using DMA.

Read Data
Read a stream of 68 bits from EEPROM by using DMA,
then decipher the received data as follows:
  4 bits - ignore these
 64 bits - data (conventionally MSB first)

Write Data to Address
Prepare the following bitstream in memory, then transfer the stream to eeprom by using DMA, it'll take ca. 108368 clock cycles (ca. 6.5ms) until the old data is erased and new data is programmed.
  2 bits "10" (Write Request)
  n bits eeprom address (MSB first, 6 or 14 bits, depending on EEPROM)
 64 bits data (conventionally MSB first)
  1 bit "0"
After the DMA, keep reading from the chip, by normal LDRH [DFFFF00h], until Bit 0 of the returned data becomes "1" (Ready). To prevent your program from locking up in case of malfunction, generate a timeout if the chip does not reply after 10ms or longer.

Using DMA
Transferring a bitstream to/from the EEPROM by LDRH/STRH opcodes does not work, this might be because of timing problems, or because how the GBA squeezes non-sequential memory addresses through the external address/data bus.
For this reason, a buffer in memory must be used (that buffer would be typically allocated temporarily on stack, one halfword for each bit, bit1-15 of the halfwords are don't care, only bit0 is of interest).
The buffer must be transfered as a whole to/from EEPROM by using DMA3 (only DMA 3 is valid to read & write external memory), use 16bit transfer mode, both source and destination address incrementing (ie. DMA3CNT=80000000h+length).
DMA channels of higher priority should be disabled during the transfer (ie. H/V-Blank or Sound FIFO DMAs). And, of course any interrupts that might mess with DMA registers should be disabled.

Pin-Outs
The EEPROM chips are having only 8 pins, these are connected, Pin 1..8, to ROMCS, RD, WR, AD0, GND, GND, A23, VDD of the GamePak bus. Carts with 32MB ROM must have A7..A22 logically ANDed with A23.

Notes
There seems to be no autodection mechanism, so that a hardcoded bus width must be used. The game Super Mario Advance uses a 512 Byte EEPROM, no idea which games use 8KBytes (if any) ???


 Backup Flash ROM

64 KBytes - 512Kbits Flash ROM - Lifetime: 10,000 writes per sector
128 KBytes - 1Mbit Flash ROM - Lifetime: ??? writes per sector

Chip Identification (all device types)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=90h  (enter ID mode)
  dev=[E000001h], man=[E000000h]                  (get device & manufacturer)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=F0h  (terminate ID mode)
Used to detect the type (and presence) of FLASH chips. See Device Types below.

Reading Data Bytes (all device types)
  dat=[E00xxxxh]                                  (read byte from address xxxx)

Erase Entire Chip (all device types)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=80h  (erase command)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=10h  (erase entire chip)
  wait until [E000000h]=FFh (or timeout)
Erases all memory in chip, erased memory is FFh-filled.

Erase 4Kbyte Sector (all device types, except Atmel)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=80h  (erase command)
  [E005555h]=AAh, [E002AAAh]=55h, [E00n000h]=30h  (erase sector n)
  wait until [E00n000h]=FFh (or timeout)
Erases memory at E00n000h..E00nFFFh, erased memory is FFh-filled.

Erase-and-Write 128 Bytes Sector (only Atmel devices)
  old=IME, IME=0                                  (disable interrupts)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=A0h  (erase/write sector command)
  [E00xxxxh+00h..7Fh]=dat[00h..7Fh]               (write 128 bytes)
  IME=old                                         (restore old IME state)
  wait until [E00xxxxh+7Fh]=dat[7Fh] (or timeout)
Interrupts (and DMAs) should be disabled during command/write phase. Target address must be a multiple of 80h.

Write Single Data Byte (all device types, except Atmel)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=A0h  (write byte command)
  [E00xxxxh]=dat                                  (write byte to address xxxx)
  wait until [E00xxxxh]=dat (or timeout)
The target memory location must have been previously erased.

Terminate Command after Timeout (only Macronix devices, ID=1CC2h)
  [E005555h]=F0h                            (force end of write/erase command)
Use if timeout occurred during "wait until" periods, for Macronix devices only.

Bank Switching (devices bigger than 64K only)
  [E005555h]=AAh, [E002AAAh]=55h, [E005555h]=B0h  (select bank command)
  [E000000h]=bnk                                  (write bank number 0..1)
Specifies 64K bank number for read/write/erase operations.
Required because gamepak flash/sram addressbus is limited to 16bit width.

Device Types
Nintendo puts different FLASH chips in commercial game cartridges. Developers should thus detect & support all chip types. For Atmel chips it'd be recommended to simulate 4K sectors by software, though reportedly Nintendo doesn't use Atmel chips in newer games anymore. Also mind that different timings should not disturb compatibility and performance.
  ID     Name       Size  Sectors  AverageTimings  Timeouts/ms   Waits
  D4BFh  SST        64K   16x4K    20us?,?,?       10,  40, 200  3,2
  1CC2h  Macronix   64K   16x4K    ?,?,?           10,2000,2000  8,3
  1B32h  Panasonic  64K   16x4K    ?,?,?           10, 500, 500  4,2
  3D1Fh  Atmel      64K   512x128  ?,?,?           ...40..,  40  8,8
  ???    ?          128K  ?        ?,?,?           ?    ?    ?    ?
  09C2h  Macronix ? 128K  ?        ?,?,?           ?    ?    ?    ?
Identification Codes MSB=Device Type, LSB=Manufacturer.
Size in bytes, and numbers of sectors * sector size in bytes.
Average medium Write, Erase Sector, Erase Chips timings are unknown ???
Timeouts im milliseconds for Write, Erase Sector, Erase Chips.
Waitstates for Writes, and Reads in clock cycles.

Accessing FLASH Memory
FLASH memory is located in the "SRAM" area at E000000h..E00FFFFh, which is restricted to 16bit address and 8bit data buswidths. Respectively, the memory can be accessed <only> by 8bit read/write LDRB/STRB opcodes.
Also, reading anything (data or status/busy information) can be done <only> by opcodes executed in WRAM (not from opcodes in ROM) (there's no such restriction for writing).

FLASH Waitstates
Use 8 clk waitstates for initial detection (WAITCNT Bits 0,1 both set). After detection of certain device types smaller wait values may be used for write/erase, and even smaller wait values for raw reading, see Device Types table.
In practice, games seem to use smaller values only for write/erase (even though those operations are slow anyways), whilst raw reads are always done at 8 clk waits (even though reads could actually benefit slightly from smaller wait values).

Verify Write/Erase and Retry
Even though device signalizes the completion of write/erase operations, it'd be recommended to read/confirm the content of the changed memory area by software. In practice, Nintendos "erase-write-verify-retry" function typically repeats the operation up to three times in case of errors.
Also, for SST devices only, the "erase-write" and "erase-write-verify-retry" functions repeat the erase command up to 80 times, additionally followed by one further erase command if no retries were needed, otherwise followed by six further erase commands.

Note
FLASH (64Kbytes) is used by the game Sonic Advance, and possibly others.


 Backup DACS

128 KBytes - 1Mbit DACS - Lifetime: 100,000 writes.
1024 KBytes - 8Mbit DACS - Lifetime: 100,000 writes.

DACS (Debugging And Communication System) is used in Nintendos hardware debugger only, DACS is NOT used in normal game cartridges.

Parts of DACS memory is used to store the debugging exception handlers (entry point/size defined in cartridge header), the remaining memory could be used to store game positions or other data. The address space is the upper end of the 32MB ROM area, the memory can be read directly by the CPU, including for ability to execute program code in this area.


 Flashcards

Flashcards are re-writable cartridges using FLASH memory, allowing to test even multiboot-incompatible GBA software on real hardware, providing a good development environment when used in combination with a reasonable software debugger.

The carts can be written to from external tools, or directly from GBA programs.
Below are pseudo code flowcharts for detect, erase, and write operations.
All flash reads/writes are meant to be 16bit (ldrh/strh) memory accesses.

detect_flashcard:
 configure_flashcard(9E2468Ah,9413h)    ;unlock flash advance cards
 turbo=1, send_command(8000000h,90h)    ;enter ID mode (both chips, if any)
 maker=[8000000h], device=[8000000h+2]
 IF maker=device THEN device=[8000000h+4] ELSE turbo=0
 flashcard_read_mode                    ;exit ID mode
 search (maker+device*10000h) in device_list
 total/erase/write_block_size = list_entry SHL turbo

flashcard_erase(dest,len):
 FOR x=1 to len/erase_block_size
  send_command(dest,20h)        ;erase sector command
  send_command(dest,D0h)        ;confirm erase sector
  dest=dest+erase_block_size
 IF wait_busy=okay THEN NEXT x
 enter_read_mode                ;exit erase/status mode

flashcard_write(src,dest,len):
 siz=write_block_size
 FOR x=1 to len/siz
  IF siz=2 THEN send_command(dest,10h)  ;write halfword command
  IF siz>2 THEN send_command(dest,E8h)  ;write to buffer command
  IF siz>2 THEN send_command(dest,16-1) ;buffer size 16 halfwords (per chip)
  FOR y=1 TO siz/2
   [dest]=[src], dest=dest+2, src=src+2 ;write data to buffer
  NEXT y
  IF siz>2 THEN send_command(dest,D0h)  ;confirm write to buffer
 IF wait_busy=okay THEN NEXT x
 enter_read_mode                        ;exit write/status mode

send_command(adr,val):
 [adr]=val
 IF turbo THEN [adr+2]=val

enter_read_mode:
 send_command(8000000h,FFh)     ;exit status mode
 send_command(8000000h,FFh)     ;again maybe more stable (as in jeff's source)

flashcard_wait_busy:
 start=time
 REPEAT
  stat=[8000000h] XOR 80h
  IF turbo THEN stat=stat OR ([8000000h+2] XOR 80h)
  IF (stat AND 7Fh)>0 THEN error
  IF (stat AND 80h)=0 THEN ready
  IF time-start>5secs THEN timeout
 UNTIL ready OR error OR timeout
 IF error OR timeout THEN send_command(8000000h,50h)    ;clear status

configure_flashcard(adr,val): ;required for Flash Advance cards only
 [930ECA8h]=5354h
 [802468Ah]=1234h, repeated 500 times
 [800ECA8h]=5354h
 [802468Ah]=5354h
 [802468Ah]=5678h, repeated 500 times
 [930ECA8h]=5354h
 [802468Ah]=5354h
 [8ECA800h]=5678h
 [80268A0h]=1234h
 [802468Ah]=ABCDh, repeated 500 times
 [930ECA8h]=5354h
 [adr]=val

init_backup: ;no info how to use that exactly
 configure_flashcard(942468Ah,???)

device_list: (id code, total/erase/write sizes in bytes)
  ID Code    Total   Erase  Write  Name
  -??-00DCh      ?       ?      ?  Hudson Cart (???)
  00160089h     4M    128K     32  Intel i28F320J3A (Flash Advance)
  00170089h     8M    128K     32  Intel i28F640J3A (Flash Advance)
  00180089h    16M    128K     32  Intel i28F128J3A (Flash Advance)
  00E200B0h      ?     64K      2  Sharp LH28F320BJE ? (Nintendo)

Notes
All flashcards should work at 4,2 waitstates (power on default), most commercial games change waits to 3,1 which may work unstable with some/older FA flashcards. Intel FLASH specified to have a lifetime of 100,000 erases, and average block erase time 1 second (up to 5 second in worst cases).
Aside from the main FLASH memory, Flash Advance (FA) (aka Visoly) cards additionally contain battery buffered SRAM backup, and FLASH backup, and in some cases also EEPROM backup.
Turbo FA cards are containing two chips interlaced (at odd/even halfword addresses), allowing to write/erase both chips simultaneously, resulting in twice as fast programming time.
Standard Nintendo flash carts have to be modified before you can actually write to them. This is done by removing resistor R7 and putting it at empty location R8.
Mind that write/erase/detect modes output status information in ROM area, so that in that modes all GBA program code (and any interrupt handlers) must be executed in WRAM, not in ROM.

Thanks to Jeff Frohwein for his FAQ and CARTLIB sample in FLGBA at devrs.com


 Cheat Devices

Codebreaker (US) aka Xploder (EUR).
Gameshark (US) aka Action Replay (EUR).

Cheat Codes - General Info
Cheat Codes - Codebreaker/Xploder
Cheat Codes - Gameshark/Action Replay V1/V2
Cheat Codes - Pro Action Replay V3


 Cheat Codes - General Info

Cheat devices are external adapters, connected between the GBA and the game cartridge. The devices include a BIOS ROM which is, among others, used to prompt the user to enter cheat codes.
These codes are used to patch specified memory locations for a certain GBA game, allowing the user to gain goodies such like Infinite sex, 255 Cigarettes, etc.

ROM and RAM Patches
For ROM Patches, the device watches the address bus, if it matches a specified address then it outputs a patched value to the data bus, that mechanism is implemented by hardware, aside from the Hook Enable Code some devices also allow a limited number of cheats to use ROM patches.
Most cheat codes are RAM patches, each time when the hook procedure is executed it will process all codes and overwrite the specified addresses in RAM (or VRAM or I/O area) by the desired values.

Enable Codes (Must Be On)
Enable codes usually consist of the Game ID, Hook Address, and eventually a third code used to encrypt all following codes. The Game ID is used to confirm that the correct cartridge is inserted, just a verification, though the device may insist on the ID code.
The Hook Address specifies an address in cartridge ROM, and should point to an opcode which is executed several times per second (eg. once per frame, many codes place the hook in the joypad handler). At the hook address, the device redirects to its own BIOS, processes the RAM patches, and does then return control to the game cartridge.
Note: The hook address should not point to opcodes with relative addressing (eg. B, BL, LDR Rd,=Imm, ADD Rd,=Imm opcodes - which are all relative to PC program counter register).

Alignment
Addresses for 16bit or 32bit values should be properly aligned.


 Cheat Codes - Codebreaker/Xploder

Codebreaker Codes
  0000xxxx 000y  Enable Code 1 - Game ID
  1aaaaaaa 000z  Enable Code 2 - Hook Address
  2aaaaaaa yyyy  [aaaaaaa]=[aaaaaaa] OR yyyy
  3aaaaaaa 00yy  [aaaaaaa]=yy
  4aaaaaaa yyyy  [aaaaaaa...]=yyyy repeated by parameters in next code
  cccccccc ssss  repeat count and address step parameters for above code
  5aaaaaaa xxxx  Write block (SUPER CODE)
  ........ ....  parameters for above code
  6aaaaaaa yyyy  [aaaaaaa]=[aaaaaaa] AND yyyy
  7aaaaaaa yyyy  IF [aaaaaaa]=yyyy THEN (next code)
  8aaaaaaa yyyy  [aaaaaaa]=yyyy
  9xyyxxxx xxxx  Enable Code 0 - Encrypt all following codes (optional)
  Aaaaaaaa yyyy  IF [aaaaaaa]<>yyyy THEN (next code)
  Baaaaaaa yyyy  IF [aaaaaaa]>yyyy THEN (next code) (signed comparison)
  Caaaaaaa yyyy  IF [aaaaaaa]<yyyy THEN (next code) (signed comparison)
  D0000020 yyyy  IF [joypad] AND yyyy = 0 THEN (next code)
  Eaaaaaaa yyyy  [aaaaaaa]=[aaaaaaa]+yyyy
  Faaaaaaa yyyy  IF [aaaaaaa] AND yyyy THEN (next code)

Codebreaker Enable Codes
Hook Address 'aaaaaaa' is a 25bit offset in ROM-image (0-1FFFFFFh).
Flag byte 'y' (usually 0Ah), Bit1=Disable IRQs, Bit3=CRC Exists.
Code Handler Store Address 'z' (0-7, usually 7) (8000100h+z*400000h).
Checksum 'xxxx' for first 64Kbytes of cartridge (no$gba pads by FFh if ROM is smaller than 64K). Calculated, by using unsigned 16bit values, as such:
  CRC=0FFFFh
  FOR Y=0 to 0FFFFh
   X=BYTE[Y] xor (CRC/100h)
   X=X xor (X/10h)
   CRC=(CRC*100h) xor (X*1001h) xor (X*20h)
  NEXT Y


 Cheat Codes - Gameshark/Action Replay V1/V2

Gameshark RAW Codes (These codes must be encrypted before using them)
  0aaaaaaa 000000xx  [aaaaaaa]=xx
  1aaaaaaa 0000xxxx  [aaaaaaa]=xxxx
  2aaaaaaa xxxxxxxx  [aaaaaaa]=xxxxxxxx
  3000cccc xxxxxxxx  write xxxxxxxx to (cccc-1) addresses (list in next codes)
  aaaaaaaa aaaaaaaa  parameter for above code, containing two addresses each
  aaaaaaaa 00000000  last parameter for above, zero-padded if only one address
  60aaaaaa y000xxxx  [8000000h+aaaaaa*2]=xxxx (ROM Patch)
  8a1aaaaa 000000xx  IF GS_Button_Down THEN [a0aaaaa]=xx
  8a2aaaaa 0000xxxx  IF GS_Button_Down THEN [a0aaaaa]=xxxx
  80F00000 0000xxxx  IF GS_Button_Down THEN slowdown xxxx * ??? cycles per hook
  Daaaaaaa 0000xxxx  IF [aaaaaaa]=xxxx THEN (next code)
  E0zzxxxx 0aaaaaaa  IF [aaaaaaa]=xxxx THEN (next 'zz' codes)
  Faaaaaaa 00000x0y  Enable Code - Hook Routine
  xxxxxxxx 001DC0DE  Enable Code - Game Code ID (value at [0ACh] in cartridge)
  DEADFACE 0000xxxx  Change Encryption Seeds

Enable Code - Hook Routine
Hook Address 'aaaaaaa' is a 28bit ROM address (8FFFFFFh-9FFFFFFh).
Used to insert the GS code handler routine where it will be executed at
least 20 times per second. Without this code, GSA can not write to RAM.
 y=1 - Executes code handler without backing up the LR register.
 y=2 - Executes code handler and backs up the LR register.
 y=3 - Replaces a 32-bit pointer used for long-branches.
 x=0 - Must turn GSA off before loading game.
 x=1 - Must not do that.

ROM Patch
This type allows GSA to intercept ROM reads and returns the value xxxx.
 y=0 wait for the code handler to enable the patch
 y=1 patch is enabled before the game starts
 y=2 unknown ???
Note: V1/V2 hardware can only have up to 1 user-defined rom patch max. V3 can have up to 4. Some enable code types can shorten the amount of user-defined rom patches available.

Gameshark Encryption
A=Left half, and V=Right half of code.
  IF V1V2 THEN S0=09F4FBBDh S1=9681884Ah S2=352027E9h S3=F3DEE5A7h
  IF V3   THEN S0=7AA9648Fh S1=7FAE6994h S2=C0EFAAD5h S3=42712C57h
  FOR I=1 TO 32
    A=A + (V*16+S0) XOR (V+I*9E3779B9h) XOR (V/32+S1)
    V=V + (A*16+S2) XOR (A+I*9E3779B9h) XOR (A/32+S3)
  NEXT I
All calculations truncated to unsigned 32bit integer values.


 Cheat Codes - Pro Action Replay V3

Pro Action Replay V3 - RAW Codes
  C4aaaaaa 0000yyyy  Enable Code - Hook Routine at [8aaaaaa]
  xxxxxxxx 001DC0DE  Enable Code - ID Code [080000AC]
  DEADFACE 0000xxxx  Enable Code - Change Encryption Seeds
  00aaaaaa xxxxxxyy  [a0aaaaa..a0aaaaa+xxxxxx]=yy
  02aaaaaa xxxxyyyy  [a0aaaaa..a0aaaaa+xxxx*2]=yyyy
  04aaaaaa yyyyyyyy  [a0aaaaa]=yyyyyyyy
  40aaaaaa xxxxxxyy  [ [a0aaaaa] + xxxxxx ]=yy   (Indirect)
  42aaaaaa xxxxyyyy  [ [a0aaaaa] + xxxx*2 ]=yyyy (Indirect)
  44aaaaaa yyyyyyyy  [ [a0aaaaa] ]=yyyyyyyy      (Indirect)
  80aaaaaa 000000yy  [a0aaaaa]=[a0aaaaa]+yy
  82aaaaaa 0000yyyy  [a0aaaaa]=[a0aaaaa]+yyyy
  84aaaaaa yyyyyyyy  [a0aaaaa]=[a0aaaaa]+yyyyyyyy
  C6aaaaaa 0000yyyy  [4aaaaaa]=yyyy              (I/O Area)
  C7aaaaaa yyyyyyyy  [4aaaaaa]=yyyyyyyy          (I/O Area)
  iiaaaaaa yyyyyyyy  IF [a0aaaaa] <cond> <value> THEN <action>
  00000000 60000000  ELSE (?)
  00000000 40000000  ENDIF (?)
  00000000 0800xx00  AR Slowdown : loops the AR xx times
  00000000 00000000  End of the code list
  00000000 10aaaaaa 000000zz 00000000  IF AR_BUTTON THEN [a0aaaaa]=zz
  00000000 12aaaaaa 0000zzzz 00000000  IF AR_BUTTON THEN [a0aaaaa]=zzzz
  00000000 14aaaaaa zzzzzzzz 00000000  IF AR_BUTTON THEN [a0aaaaa]=zzzzzzzz
  00000000 18aaaaaa 0000zzzz 00000000  [8000000+aaaaaa*2]=zzzz  (ROM Patch 1)
  00000000 1Aaaaaaa 0000zzzz 00000000  [8000000+aaaaaa*2]=zzzz  (ROM Patch 2)
  00000000 1Caaaaaa 0000zzzz 00000000  [8000000+aaaaaa*2]=zzzz  (ROM Patch 3)
  00000000 1Eaaaaaa 0000zzzz 00000000  [8000000+aaaaaa*2]=zzzz  (ROM Patch 4)

  00000000 80aaaaaa 000000yy ssccssss  repeat cc times [a0aaaaa]=yy
   (with yy=yy+ss, a0aaaaa=a0aaaaa+ssss after each step)

  00000000 82aaaaaa 0000yyyy ssccssss  repeat cc times [a0aaaaa]=yyyy
   (with yyyy=yyyy+ss, a0aaaaa=a0aaaaa+ssss*2 after each step)

  00000000 84aaaaaa yyyyyyyy ssccssss  repeat cc times [a0aaaaa]=yyyyyyyy
   (with yyyy=yyyy+ss, a0aaaaa=a0aaaaa+ssss*4 after each step)

Warning: There is a bug on the real AR (v2 upgraded to v3, and maybe on real v3) with the 32bit Increment Slide code. You HAVE to add a code (best choice is 80000000 00000000 : add 0 to value at address 0) right after it, else the AR will erase the 2 last 8 digits lines of the 32 Bits Inc. Slide code when you enter it !!!

Final Notes
The 'turn off all codes' makes an infinite loop (that can't be broken, unless the condition becomes True). - How? By Interrupt? Huh?
ROM Patch1 works on real V3 and, on V1/V2 upgraded to V3.
ROM Patch2,3,4 work on real V3 hardware only.

Pro Action Replay V3 Conditional Codes - iiaaaaaa yyyyyyyy
The 'ii' is composed of <cond> + <value> + <action>.
  <cond>           <value>            <action>
  08 Equal =       00 8bit zz         00 execute next code
  10 Not equal <>  02 16bit zzzz      40 execute next two codes
  18 Signed <      04 32bit zzzzzzzz  80 execute all following
  20 Signed >      06 (always false)     codes until ELSE or ENDIF
  28 Unsigned <                       C0 normal ELSE turn off all codes
  30 Unsigned >
  38 Logical AND
For example, ii=18h+02h+40h=5Ah, produces IF [a0aaaaa]<zzzz THEN next 2 codes.

Always... Codes
  For the "Always..." codes:
  - XXXXXXXX can be any authorised address except 00000000 (eg. use 02000000).
  - ZZZZZZZZ can be anything.
  - The "y" in the code data must be in the [1-7] range (which means not 0).
  typ=y,sub=0,siz=3   Always skip next line.
  typ=y,sub=1,siz=3   Always skip next 2 lines.
  typ=y,sub=2,siz=3   Always Stops executing all the codes below.
  typ=y,sub=3,siz=3   Always turn off all codes.

Code Format (ttaaaaaa xxxxyyzz)
 adr mask = 003FFFFF
 n/a mask = 00C00000 ;not used
 xtr mask = 01000000 ;used only by I/O write, and MSB of Hook
 siz mask = 06000000
 typ mask = 38000000 ;0=normal, other=conditional
 sub mask = C0000000


 BIOS Functions

The GBA BIOS includes several System Call Functions which can be accessed by SWI instructions. Incoming parameters are usually passed through registers R0,R1,R2,R3. Outgoing registers R0,R1,R3 are typically containing either garbage, or return value(s). All other registers (R2,R4-R14) are kept unchanged.

Caution
When invoking SWIs from inside of ARM state specify SWI NN*10000h, instead of SWI NN as in THUMB state.

Overview
BIOS Function Summary
BIOS Differences between GBA and NDS functions

All Functions Described
BIOS Arithmetic Functions
BIOS Rotation/Scaling Functions
BIOS Decompression Functions
BIOS Memory Copy
BIOS Halt Functions
BIOS Reset Functions
BIOS Misc Functions
BIOS Multi Boot (Single Game Pak)
BIOS Sound Functions

How BIOS Processes SWIs
SWIs can be called from both within THUMB and ARM mode. In ARM mode, only the upper 8bit of the 24bit comment field are interpreted.
Each time when calling a BIOS function 4 words (SPSR, R11, R12, R14) are saved on Supervisor stack (_svc). Once it has saved that data, the SWI handler switches into System mode, so that all further stack operations are using user stack.
In some cases the BIOS may allow interrupts to be executed from inside of the SWI procedure. If so, and if the interrupt handler calls further SWIs, then care should be taken that the Supervisor Stack does not overflow.


 BIOS Function Summary

  GBA  NDS7 NDS9 Function
  00h  00h  00h  SoftReset
  01h  -    -    RegisterRamReset
  02h  06h  06h  Halt
  03h  07h  -    Stop/Sleep
  04h  04h  04h  IntrWait
  05h  05h  05h  VBlankIntrWait
  06h  09h  09h  Div
  07h  -    -    DivArm
  08h  0Dh  0Dh  Sqrt
  09h  -    -    ArcTan
  0Ah  -    -    ArcTan2
  0Bh  0Bh  0Bh  CpuSet
  0Ch  0Ch  0Ch  CpuFastSet
  0Dh  -    -    GetBiosChecksum
  0Eh  -    -    BgAffineSet
  0Fh  -    -    ObjAffineSet
  10h  10h  10h  BitUnPack
  11h  11h  11h  LZ77UnCompWram
  12h  12h  12h  LZ77UnCompVram
  13h  13h  13h  HuffUnComp
  14h  14h  14h  RLUnCompWram
  15h  15h  15h  RLUnCompVram
  16h  -    16h  Diff8bitUnFilterWram
  17h  -    -    Diff8bitUnFilterVram
  18h  -    18h  Diff16bitUnFilter
  19h  08h  -    SoundBias
  1Ah  -    -    SoundDriverInit
  1Bh  -    -    SoundDriverMode
  1Ch  -    -    SoundDriverMain
  1Dh  -    -    SoundDriverVSync
  1Eh  -    -    SoundChannelClear
  1Fh  -    -    MidiKey2Freq
  20h  -    -    SoundWhatever0
  21h  -    -    SoundWhatever1
  22h  -    -    SoundWhatever2
  23h  -    -    SoundWhatever3
  24h  -    -    SoundWhatever4
  25h  -    -    MultiBoot
  26h  -    -    HardReset
  27h  1Fh  -    CustomHalt
  28h  -    -    SoundDriverVSyncOff
  29h  -    -    SoundDriverVSyncOn
  2Ah  -    -    SoundGetJumpList
  -    03h  03h  WaitByLoop
  -    0Eh  0Eh  GetCRC16
  -    0Fh  0Fh  IsDebugger
  -    1Ah  -    GetSineTable
  -    1Bh  -    GetPitchTable
  -    1Ch  -    GetVolumeTable
  -    1Dh  -    GetBootProcs
  -    -    1Fh  CustomPost
The BIOS SWI handler does not perform any range checks, so calling GBA SWI 2Bh-FFh or NDS SWI 20h-FFh will blindly jump to garbage addresses.
Also, NDS7 SWI 01h, 02h, 0Ah, 16h-19h, 1Eh, and NDS9 SWI 01h, 02h, 07h, 08h, 0Ah, 17h, 19h-1Eh will jump to zero (ie. the NDS7 reset vector, or NDS9 unused memory, which will be both redirected to the debug handler, if any).


 BIOS Differences between GBA and NDS functions

Differences between GBA and NDS BIOS functions
- SoftReset uses different addresses
- SWI numbers for Halt, Stop/Sleep, Div, Sqrt have changed
- Halt destroys r0 on NDS9, IntrWait bugged on NDS9
- CpuFastSet allows 4-byte blocks (nice), but...
- CpuFastSet works very SLOW because of a programming bug (uncool)
- LZ77UnCompVram, HuffUnComp, RLUnCompVram use callbacks
- SoundBias uses new delay parameter
And, a number of GBA functions have been removed, and some new NDS functions have been added, see:
BIOS Function Summary


 BIOS Arithmetic Functions

Div
DivArm
Sqrt
ArcTan
ArcTan2

SWI 06h (GBA) or SWI 09h (NDS7/NDS9) - Div
Signed Division, r0/r1.
  r0  signed 32bit Number
  r1  signed 32bit Denom
Return:
  r0  Number DIV Denom ;signed
  r1  Number MOD Denom ;signed
  r3  ABS (Number DIV Denom) ;unsigned
For example, incoming -1234, 10 should return -123, -4, +123.
The function usually gets caught in an endless loop upon division by zero.
Note: The NDS9 additionally supports hardware division, by math coprocessor, accessed via I/O Ports, however, the SWI function is a raw software division.

SWI 07h (GBA) - DivArm
Same as above (SWI 06h Div), but incoming parameters are exchanged, r1/r0 (r0=Denom, r1=number). For compatibility with ARM's library. Slightly slower (3 clock cycles) than SWI 06h.

SWI 08h (GBA) or SWI 0Dh (NDS7/NDS9) - Sqrt
Calculate square root.
  r0   unsigned 32bit number
Return:
  r0   unsigned 16bit number
The result is an integer value, so Sqrt(2) would return 1, to avoid this inaccuracy, shift left incoming number by 2*N as much as possible (the result is then shifted left by 1*N). Ie. Sqrt(2 shl 30) would return 1.41421 shl 15.
Note: The NDS9 additionally supports hardware square root calculation, by math coprocessor, accessed via I/O Ports, however, the SWI function is a raw software calculation.

SWI 09h (GBA) - ArcTan
Calculates the arc tangent.
  r0   Tan, 16bit (1bit sign, 1bit integral part, 14bit decimal part)
Return:
  r0   "-PI/2<THETA/<PI/2" in a range of C000h-4000h.
Note: there is a problem in accuracy with "THETA<-PI/4, PI/4<THETA".

SWI 0Ah (GBA) - ArcTan2
Calculates the arc tangent after correction processing.
Use this in normal situations.
  r0   X, 16bit (1bit sign, 1bit integral part, 14bit decimal part)
  r1   Y, 16bit (1bit sign, 1bit integral part, 14bit decimal part)
Return:
  r0   0000h-FFFFh for 0<=THETA<2PI.


 BIOS Rotation/Scaling Functions

BgAffineSet
ObjAffineSet

SWI 0Eh (GBA) - BgAffineSet
Used to calculate BG Rotation/Scaling parameters.
  r0   Pointer to Source Data Field with entries as follows:
        s32  Original data's center X coordinate (8bit fractional portion)
        s32  Original data's center Y coordinate (8bit fractional portion)
        s16  Display's center X coordinate
        s16  Display's center Y coordinate
        s16  Scaling ratio in X direction (8bit fractional portion)
        s16  Scaling ratio in Y direction (8bit fractional portion)
        u16  Angle of rotation (8bit fractional portion) Effective Range 0-FFFF
  r1   Pointer to Destination Data Field with entries as follows:
        s16  Difference in X coordinate along same line
        s16  Difference in X coordinate along next line
        s16  Difference in Y coordinate along same line
        s16  Difference in Y coordinate along next line
        s32  Start X coordinate
        s32  Start Y coordinate
  r2   Number of Calculations
Return: No return value, Data written to destination address.

SWI 0Fh (GBA) - ObjAffineSet
Calculates and sets the OBJ's affine parameters from the scaling ratio and angle of rotation.
The affine parameters are calculated from the parameters set in Srcp.
The four affine parameters are set every Offset bytes, starting from the Destp address.
If the Offset value is 2, the parameters are stored contiguously. If the value is 8, they match the structure of OAM.
When Srcp is arrayed, the calculation can be performed continuously by specifying Num.
  r0   Source Address, pointing to data structure as such:
        s16  Scaling ratio in X direction (8bit fractional portion)
        s16  Scaling ratio in Y direction (8bit fractional portion)
        u16  Angle of rotation (8bit fractional portion) Effective Range 0-FFFF
  r1   Destination Address, pointing to data structure as such:
        s16  Difference in X coordinate along same line
        s16  Difference in X coordinate along next line
        s16  Difference in Y coordinate along same line
        s16  Difference in Y coordinate along next line
  r2   Number of calculations
  r3   Offset in bytes for parameter addresses (2=continuous, 8=OAM)
Return: No return value, Data written to destination address.

For both Bg- and ObjAffineSet, Rotation angles are specified as 0-FFFFh (covering a range of 360 degrees), however, the GBA BIOS recurses only the upper 8bit; the lower 8bit may contain a fractional portion, but it is ignored by the BIOS.


 BIOS Decompression Functions

BitUnPack
Diff8bitUnFilterWram
Diff8bitUnFilterVram
Diff16bitUnFilter
HuffUnComp
LZ77UnCompWram
LZ77UnCompVram
RLUnCompVram
RLUnCompWram

SWI 10h (GBA/NDS7/NDS9) - BitUnPack
Used to increase the color depth of bitmaps or tile data. For example, to convert a 1bit monochrome font into 4bit or 8bit GBA tiles. The Unpack Info is specified separately, allowing to convert the same source data into different formats.
  r0  Source Address      (no alignment required)
  r1  Destination Address (must be 32bit-word aligned)
  r2  Pointer to UnPack information:
       16bit  Length of Source Data in bytes     (0-FFFFh)
       8bit   Width of Source Units in bits      (only 1,2,4,8 supported)
       8bit   Width of Destination Units in bits (only 1,2,4,8,16,32 supported)
       32bit  Data Offset (Bit 0-30), and Zero Data Flag (Bit 31)
      The Data Offset is always added to all non-zero source units.
      If the Zero Data Flag was set, it is also added to zero units.
Data is written in 32bit units, Destination can be Wram or Vram. The size of unpacked data must be a multiple of 4 bytes. The width of source units (plus the offset) should not exceed the destination width.
Return: No return value, Data written to destination address.

SWI 16h (GBA/NDS9) - Diff8bitUnFilterWram
SWI 17h (GBA) - Diff8bitUnFilterVram
SWI 18h (GBA/NDS9) - Diff16bitUnFilter
These aren't actually real decompression functions, destination data will have exactly the same size as source data. However, assume a bitmap or wave form to contain a stream of increasing numbers such like 10..19, the filtered/unfiltered data would be:
  unfiltered:   10  11  12  13  14  15  16  17  18  19
  filtered:     10  +1  +1  +1  +1  +1  +1  +1  +1  +1
In this case using filtered data (combined with actual compression algorithms) will obviously produce better compression results.
Data units may be either 8bit or 16bit used with Diff8bit or Diff16bit functions respectively. The 8bitVram function allows to write to VRAM directly (which uses 16bit data bus) by writing two 8bit values at once, the downside is that it is eventually slower as the 8bitWram function.
  r0  Source address (must be aligned by 4) pointing to data as follows:
       Data Header (32bit)
         Bit 0-3   Data size (must be 1 for Diff8bit, 2 for Diff16bit)
         Bit 4-7   Type (must be 8 for DiffFiltered)
         Bit 8-31  24bit size after decompression
       Data Units (each 8bit or 16bit depending on used SWI function)
         Data0          ;original data
         Data1-Data0    ;difference data
         Data2-Data1    ;...
         Data3-Data2
         ...
  r1  Destination address
Return: No return value, Data written to destination address.

SWI 13h (GBA/NDS7/NDS9) - HuffUnComp (NDS: with Callback)
Expands Huffman-compressed data and writes in units of 32bits.
If the size of the compressed data is not a multiple of 4, please adjust it as much as possible by padding with 0.
Align the source address to a 4Byte boundary.
  r0  Source Address, aligned by 4, pointing to:
       Data Header (32bit)
         Bit 0-3   Data size in bit units (normally 4 or 8)
         Bit 4-7   Compressed type (must be 2 for Huffman)
         Bit 8-31  24bit size of decompressed data in bytes
       Tree Table
        u8      tree table size/2-1
        Each of the nodes below defined as:
         u8
          6bit  offset to next node -1 (2 byte units)
          1bit  right node end flag (if set, data is in next node)
          1bit  left node end flag
        1 node  Root node
        2 nodes Left, and Right node
        4 nodes LeftLeft, LeftRight, RightLeft, and RightRight node
        ...
       Compressed data
        ...
  r1  Destination Address
  r2  Callback parameter (NDS SWI 13h only, see Callback notes below)
  r3  Callback structure (NDS SWI 13h only, see Callback notes below)
Return: No return value, Data written to destination address.

SWI 11h (GBA/NDS7/NDS9) - LZ77UnCompWram
SWI 12h (GBA/NDS7/NDS9) - LZ77UnCompVram (NDS: with Callback)
Expands LZ77-compressed data. The Wram function is faster, and writes in units of 8bits. For the Vram function the destination must be halfword aligned, data is written in units of 16bits.
If the size of the compressed data is not a multiple of 4, please adjust it as much as possible by padding with 0. Align the source address to a 4-Byte boundary.
  r0   Source address, pointing to data as such:
        Data header (32bit)
          Bit 0-3   Reserved
          Bit 4-7   Compressed type (must be 1 for LZ77)
          Bit 8-31  Size of decompressed data
        Repeat below. Each Flag Byte followed by eight Blocks.
        Flag data (8bit)
          Bit 0-7   Type Flags for next 8 Blocks, MSB first
        Block Type 0 - Uncompressed - Copy 1 Byte from Source to Dest
          Bit 0-7   One data byte to be copied to dest
        Block Type 1 - Compressed - Copy N+3 Bytes from Dest-Disp-1 to Dest
          Bit 0-3   Disp MSBs
          Bit 4-7   Number of bytes to copy (minus 3)
          Bit 8-15  Disp LSBs
  r1   Destination address
  r2  Callback parameter (NDS SWI 12h only, see Callback notes below)
  r3  Callback structure (NDS SWI 12h only, see Callback notes below)
Return: No return value.

SWI 14h (GBA/NDS7/NDS9) - RLUnCompWram
SWI 15h (GBA/NDS7/NDS9) - RLUnCompVram (NDS: with Callback)
Expands run-length compressed data. The Wram function is faster, and writes in units of 8bits. For the Vram function the destination must be halfword aligned, data is written in units of 16bits.
If the size of the compressed data is not a multiple of 4, please adjust it as much as possible by padding with 0. Align the source address to a 4Byte boundary.
  r0  Source Address, pointing to data as such:
       Data header (32bit)
         Bit 0-3   Reserved
         Bit 4-7   Compressed type (must be 3 for run-length)
         Bit 8-31  Size of decompressed data
       Repeat below. Each Flag Byte followed by one or more Data Bytes.
       Flag data (8bit)
         Bit 0-6   Expanded Data Length (uncompressed N-1, compressed N-3)
         Bit 7     Flag (0=uncompressed, 1=compressed)
       Data Byte(s) - N uncompressed bytes, or 1 byte repeated N times
  r1  Destination Address
  r2  Callback parameter (NDS SWI 15h only, see Callback notes below)
  r3  Callback structure (NDS SWI 15h only, see Callback notes below)
Return: No return value, Data written to destination address.

NDS Decompression Callbacks
On NDS7/NDS9, the SWI 12h, 13h, 15h functions are reading source data from callback functions (rather than directly from memory). The callback functions may read normal data from memory, or from other devices, such like directly from the gamepak bus, without storing the source data in memory. The downside is that the callback mechanism makes the function very slow, furthermore, NDS7/NDS9 SWI 12h, 13h, 15h are using THUMB code, and variables on stack, alltogether that makes the whole shit very-very-very slow.
  r2 = user defined callback parameter (passed on to Open function)
  r3 = pointer to callback structure
Callback structure (five 32bit pointers to callback functions)
  Open_and_get_32bit (eg. LDR r0,[r0], get header)
  Close              (optional, 0=none)
  Get_8bit           (eg. LDRB r0,[r0])
  Get_16bit          (not used)
  Get_32bit          (used by Huffman only)
All functions may use ARM or THUMB code (indicated by address bit0). The current source address (r0) is passed to all callback functions. Additionally, the initial destination address (r1), and a user defined parameter (r2) are passed to the Open function. All functions have return values in r0. The Open function normally returns the first word (containing positive length and type), alternatively it may return a negative error code to abort/reject decompression. The Close function, if it is defined, should return zero (or any positive value), or a negative errorcode. The other functions return raw data, without errorcodes. The SWI returns the length of decompressed data, or the signed errorcode from the Open/Close functions.


 BIOS Memory Copy

CpuFastSet
CpuSet

SWI 0Ch (GBA/NDS7/NDS9) - CpuFastSet
Memory copy/fill in units of 32 bytes. Memcopy is implemented as repeated LDMIA/STMIA [Rb]!,r2-r9 instructions. Memfill as single LDR followed by repeated STMIA [Rb]!,r2-r9. After processing all 32-byte-blocks, the NDS additonally processes the remaining words as 4-byte blocks. BUG: The NDS uses the fast 32-byte-block processing only for the first N bytes (not for the first N words), so only the first quarter of the memory block is FAST, the remaining three quarters are SLOWLY copied word-by-word.
The length is specifed as wordcount, ie. the number of bytes divided by 4.
On the GBA, the length must be a multiple of 8 words (32 bytes). On NDS, the length may be any number of words (4 bytes).
  r0    Source address        (must be aligned by 4)
  r1    Destination address   (must be aligned by 4)
  r2    Length/Mode
          Bit 0-20  Wordcount (GBA: must be a multiple of 8 words)
          Bit 24    Fixed Source Address (0=Copy, 1=Fill by WORD[r0])
Return: No return value, Data written to destination address.

SWI 0Bh (GBA/NDS7/NDS9) - CpuSet
Memory copy/fill in units of 4 bytes or 2 bytes. Memcopy is implemented as repeated LDMIA/STMIA [Rb]!,r3 or LDRH/STRH r3,[r0,r5] instructions. Memfill as single LDMIA or LDRH followed by repeated STMIA [Rb]!,r3 or STRH r3,[r0,r5].
The length must be a multiple of 4 bytes (32bit mode) or 2 bytes (16bit mode). The (half)wordcount in r2 must be length/4 (32bit mode) or length/2 (16bit mode), ie. length in word/halfword units rather than byte units.
  r0    Source address        (must be aligned by 4 for 32bit, by 2 for 16bit)
  r1    Destination address   (must be aligned by 4 for 32bit, by 2 for 16bit)
  r2    Length/Mode
          Bit 0-20  Wordcount (for 32bit), or Halfwordcount (for 16bit)
          Bit 24    Fixed Source Address (0=Copy, 1=Fill by {HALF}WORD[r0])
          Bit 26    Datasize (0=16bit, 1=32bit)
Return: No return value, Data written to destination address.

Note: On GBA and NDS7, these two functions will silently reject to do anything if the source start or end addresses are reaching into the BIOS area. The NDS9 doesn't have such read-proctection.


 BIOS Halt Functions

Halt
IntrWait
VBlankIntrWait
Stop/Sleep
CustomHalt

SWI 02h (GBA) or SWI 06h (NDS7/NDS9) - Halt
Halts the CPU until an interrupt request occurs. The CPU is switched into low-power mode, all other circuits (video, sound, timers, serial, keypad, system clock) are kept operating.
Halt mode is terminated when any enabled interrupts are requested, that is when (IE AND IF) is not zero, the GBA locks up if that condition doesn't get true. However, the state of CPUs IRQ disable bit in CPSR register, and the IME register are don't care, Halt passes through even if either one has disabled interrupts.
On GBA and NDS7, Halt is implemented by writing to HALTCNT, Port 4000301h. On NDS9, Halt is implemted by writing to System Control Coprocessor (mov p15,0,c7,c0,4,r0 opcode), this opcode hangs if IME=0.
No parameters, no return value.
(GBA/NDS7: all registers unchanged, NDS9: R0 destroyed)

SWI 04h (GBA/NDS7/NDS9) - IntrWait
Continues to wait in Halt state until one (or more) of the specified interrupt(s) do occur. When using multiple interrupts at the same time, this function is having less overhead than repeatedly calling the Halt function.
  r0    0=Return immediately if an old flag was already set (NDS9: bugged!)
        1=Discard old flags, wait until a NEW flag becomes set
  r1    Interrupt flag(s) to wait for (same format as IE/IF registers)
Caution: When using IntrWait or VBlankIntrWait, the user interrupt handler MUST update the BIOS Interrupt Flags value in RAM; when acknowleding processed interrupt(s) by writing a value to the IF register, the same value should be also ORed to the BIOS Interrupt Flags value, at following memory location:
  Host     GBA (16bit)  NDS7 (32bit)  NDS9 (32bit)
  Address  [3007FF8h]   [380FFF8h]    [DTCM+3FF8h]
NDS9: BUG: No Discard (r0=0) doesn't work. The function always waits for at least one IRQ to occur (no matter which, including IRQs that are not selected in r1), even if the desired flag was already set. NB. the same bug is also found in the GBA/NDS7 functions, but it's compensated by a second bug, ie. the GBA/NDS7 functions are working okay because their "bug doesn't work".
Return: No return value, the selected flag(s) are automatically reset in BIOS Interrupt Flags value in RAM upon return.

SWI 05h (GBA/NDS7/NDS9) - VBlankIntrWait
Continues to wait in Halt status until a new V-Blank interrupt occurs.
The function sets r0=1 and r1=1 and then executes IntrWait (SWI 04h), see IntrWait for details.
No parameters, no return value.

SWI 03h (GBA) - Stop
Switches the GBA into very low power mode (to be used similar as a screen-saver). The CPU, System Clock, Sound, Video, SIO-Shift Clock, DMAs, and Timers are stopped.
Stop state can be terminated by the following interrupts only (as far as enabled in IE register): Joypad, Game Pak, or General-Purpose-SIO.
"The system clock is stopped so the IF flag is not set."
Preparation for Stop:
Disable Video before implementing Stop (otherwise Video just freezes, but still keeps consuming battery power). Possibly required to disable Sound also ??? Obviously, it'd be also recommended to disable any external hardware (such like Rumble or Infra-Red) as far as possible.
No parameters, no return value.

SWI 07h (NDS7) - Sleep
No info, probably similar as GBA SWI 03h (Stop). Sleep is implemented for NDS7 only, not for NDS9, not sure if the functions stops BOTH NDS7 and NDS9?

SWI 27h (GBA) or SWI 1Fh (NDS7) - CustomHalt (Undocumented)
Writes the 8bit parameter value to HALTCNT, below values are equivalent to Halt and Stop/Sleep functions, other values reserved, purpose unknown.
  r2  8bit parameter (GBA: 00h=Halt, 80h=Stop) (NDS7: 80h=Halt, C0h=Sleep)
No return value.


 BIOS Reset Functions

SoftReset
RegisterRamReset
HardReset

SWI 00h - SoftReset
Clears 200h bytes of RAM (containing stacks, and BIOS IRQ vector/flags), initializes system, supervisor, and irq stack pointers, sets R0-R12, LR_svc, SPSR_svc, LR_irq, and SPSR_irq to zero, and enters system mode.
Note that the NDS9 stack registers are hardcoded (the DTCM base should be set to the default setting of 0800000h). The NDS9 function additionally flushes caches and write buffer, and sets the CP15 control register to 12078h.
  Host  sp_svc    sp_irq    sp_sys    zerofilled area       return address
  GBA   3007FE0h  3007FA0h  3007F00h  [3007E00h..3007FFFh]  Flag[3007FFAh]
  NDS7  380FFDCh  380FFB0h  380FF00h  [380FE00h..380FFFFh]  Addr[27FFE34h]
  NDS9  0803FC0h  0803FA0h  0803EC0h  [DTCM+3E00h..3FFFh]   Addr[27FFE24h]
The NDS7/NDS9 return addresses at [27FFE34h/27FFE24h] are usually containing copies of Cartridge Header [034h/024h] entry points, which may select ARM/THUMB state via bit0. The GBA return address 8bit flag is interpreted as 00h=8000000h (ROM), or 01h-FFh=2000000h (RAM), entered in ARM state.
Note: The reset is applied only to the CPU that has executed the SWI (ie. on the NDS, the other CPU will remain unaffected).
Return: Does not return to calling procedure, instead, loads the above return address into R14, and then jumps to that address by a "BX R14" opcode.

SWI 01h (GBA) - RegisterRamReset
Resets the I/O registers and RAM specified in ResetFlags. However, it does not clear the CPU internal RAM area from 3007E00h-3007FFFh.
  r0  ResetFlags
       Bit   Expl.
       0     Clear 256K on-board WRAM  ;-don't use when returning to WRAM
       1     Clear 32K in-chip WRAM    ;-excluding last 200h bytes
       2     Clear Palette
       3     Clear VRAM
       4     Clear OAM              ;-zerofilled! does NOT disable OBJs!
       5     Reset SIO registers    ;-switches to general purpose mode!
       6     Reset Sound registers
       7     Reset all other registers (except SIO, Sound)
Return: No return value.
Bug: LSBs of SIODATA32 are always destroyed, even if Bit5 of R0 was cleared.
The function always switches the screen into forced blank by setting DISPCNT=0080h (regardless of incoming R0, screen becomes white).

SWI 26h (GBA) - HardReset (Undocumented)
This function reboots the GBA (including for getting through the time-consuming nintendo intro, which is making the function particularly useless and annoying).
Parameters: None. Return: Never/Reboot.
Execution Time: About 2 seconds (!)


 BIOS Misc Functions

GetBiosChecksum
WaitByLoop
GetCRC16
IsDebugger
GetSineTable
GetPitchTable
GetVolumeTable
CustomPost
GetBootProcs

SWI 0Dh (GBA) - GetBiosChecksum (Undocumented)
Calculates the checksum of the BIOS ROM (by reading in 32bit units, and adding up these values). IRQ and FIQ are disabled during execution.
The checksum is BAAE187Fh (GBA and GBA SP), or BAAE1880h (DS in GBA mode, whereas the only difference is that the byte at [3F0Ch] is changed from 00h to 01h, otherwise the BIOS is 1:1 same as GBA BIOS, it does even include multiboot code).
Parameters: None. Return: r0=Checksum.

SWI 03h (NDS7/NDS9) - WaitByLoop
Performs a "LOP: SUB R0,1 / BGT LOP" wait loop, the loop is executed in BIOS memory, which provides reliable timings (regardless of the memory waitstates & cache state of the calling procedure). Intended only for short delays (eg. flash memory programming cycles).
  r0  Delay value (should be in range 1..7FFFFFFFh)
Execution Time: ARM7: R0*4 cycles, plus some overload on SWI handling.
Execution Time: ARM9: Unknown, clock cycles for ARMv5 are unspecified?
Return: No return value.

SWI 0Eh (NDS7/NDS9) - GetCRC16
  r0  Initial CRC value (16bit, usually FFFFh)
  r1  Start Address   (must be aligned by 2)
  r2  Length in bytes (must be aligned by 2)
CRC16 checksums can be calculated as such:
  val[0..7] = C0C1h,C181h,C301h,C601h,CC01h,D801h,F001h,A001h
  for i=start to end
    crc=crc xor byte[i]
    for j=0 to 7
      crc=crc shr 1:if carry then crc=crc xor (val[j] shl (7-j))
    next j
  next i
Return:
  r0  Calculated 16bit CRC Value
Additionally, if the length is nonzero, r3 contains the last processed halfword at [addr+len-2]. Unlike most other NDS7 SWI functions which reject to read from BIOS memory, this allows to dump the NDS7 BIOS (except for the memory region that is locked via BIOSPROT Port 4000308h).

SWI 0Fh (NDS7/NDS9) - IsDebugger
Detects if 4MB (normal) or 8MB (debug version) Main RAM installed.
Return: r0 = result (0=normal console 4MB, 1=debug version 8MB)
Destroys halfword at [27FFFFAh] (NDS7) or [27FFFF8h] (NDS9)!
The SWI 0Fh function doesn't work stable if it gets interrupted by an interrupt which is calling SWI 0Fh, which would destroy the above halfword scratch value (unless the IRQ handler has saved/restored the halfword).

SWI 1Ah (NDS7) - GetSineTable
  r0  Index (0..3Fh) (must be in that range, otherwise returns garbage)
Return: r0 = Desired Entry (0000h..7FF5h) ;SIN(0 .. 88.6 degrees)*8000h

SWI 1Bh (NDS7) - GetPitchTable
  r0  Index (0..2FFh) (must be in that range, otherwise returns garbage)
Return: r0 = Desired Entry (0000h..FF8Ah) (unsigned)

SWI 1Ch (NDS7) - GetVolumeTable
  r0  Index (0..2D3h) (must be in that range, otherwise returns garbage)
Return: r0 = Desired Entry (00h..7Fh) (unsigned)

SWI 1Fh (NDS9) - CustomPost
Writes to the POSTFLG register, probably for use by Firmware boot procedure.
  r0  32bit value, to be written to POSTFLG, Port 4000300h
Return: No return value.

SWI 1Dh (NDS7) - GetBootProcs
Returns addresses of Gamecart boot procedure/interrupt handler, probably for use by Firmware boot procedure. The return values are somewhat XORed by each other. Most of the returned functions won't work if the POSTFLG register is set.


 BIOS Multi Boot (Single Game Pak)

MultiBoot

SWI 25h (GBA) - MultiBoot
This function uploads & starts program code to slave GBAs, allowing to launch programs on slave units even if no cartridge is inserted into the slaves (this works because all GBA BIOSes contain built-in download procedures in ROM).
However, the SWI 25h BIOS upload function covers only 45% of the required Transmission Protocol, the other 55% must be coded in the master cartridge (see Transmission Protocol below).
  r0  Pointer to MultiBootParam structure
  r1  Transfer Mode (undocumented)
       0=256KHz, 32bit, Normal mode    (fast and stable)
       1=115KHz, 16bit, MultiPlay mode (default, slow, up to three slaves)
       2=2MHz,   32bit, Normal mode    (fastest but maybe unstable)
  Note: HLL-programmers that are using the MultiBoot(param_ptr) macro cannot
  specify the transfer mode and will be forcefully using MultiPlay mode.
Return:
  r0  0=okay, 1=failed
See below for more details.

Multiboot Parameter Structure
Size of parameter structure should be 4Ch bytes (the current GBA BIOS uses only first 44h bytes though). The following entries must be set before calling SWI 25h:
  Addr Size Name/Expl.
  14h  1    handshake_data (entry used for normal mode only)
  19h  3    client_data[1,2,3]
  1Ch  1    palette_data
  1Eh  1    client_bit (Bit 1-3 set if child 1-3 detected)
  20h  4    boot_srcp  (typically 8000000h+0C0h)
  24h  4    boot_endp  (typically 8000000h+0C0h+length)
The transfer length (excluding header data) should be a multiple of 10h, minimum length 100h, max 3FF40h (ca. 256KBytes). Set palette_data as "81h+color*10h+direction*8+speed*2", or as "0f1h+color*2" for fixed palette, whereas color=0..6, speed=0..3, direction=0..1. The other entries (handshake_data, client_data[1-3], and client_bit) must be same as specified in Transmission Protocol (see below hh,cc,y).

Multiboot Transfer Protocol
Below describes the complete transfer protocol, normally only the Initiation part must be programmed in the master cartridge, the main data transfer can be then performed by calling SWI 25h, the slave program is started after SWI 25h completion.
The ending handshake is normally not required, when using it, note that you will need custom code in BOTH master and slave programs.
  Times  Send   Receive  Expl.
  -----------------------Required Transfer Initiation in master program
  ...    6200   FFFF     Slave not in multiplay/normal mode yet
  1      6200   0000     Slave entered correct mode now
  15     6200   720x     Repeat 15 times, if failed: delay 1/16s and restart
  1      610y   720x     Recognition okay, exchange master/slave info
  60h    xxxx   NN0x     Transfer C0h bytes header data in units of 16bits
  1      6200   000x     Transfer of header data completed
  1      620y   720x     Exchange master/slave info again
  ...    63pp   720x     Wait until all slaves reply 73cc instead 720x
  1      63pp   73cc     Send palette_data and receive client_data[1-3]
  1      64hh   73uu     Send handshake_data for final transfer completion
  -----------------------Below is SWI 25h MultiBoot handler in BIOS
  DELAY  -      -        Wait 1/16 seconds at master side
  1      llll   73rr     Send length information and receive random data[1-3]
  LEN    yyyy   nnnn     Transfer main data block in units of 16 or 32 bits
  1      0065   nnnn     Transfer of main data block completed, request CRC
  ...    0065   0074     Wait until all slaves reply 0075 instead 0074
  1      0065   0075     All slaves ready for CRC transfer
  1      0066   0075     Signalize that transfer of CRC follows
  1      zzzz   zzzz     Exchange CRC must be same for master and slaves
  -----------------------Optional Handshake (NOT part of master/slave BIOS)
  ...    ....   ....     Exchange whatever custom data
Legend for above Protocol
  y     client_bit, bit(s) 1-3 set if slave(s) 1-3 detected
  x     bit 1,2,or 3 set if slave 1,2,or 3
  xxxx  header data, transferred in 16bit (!) units (even in 32bit normal mode)
  nn    response value for header transfer, decreasing 60h..01h
  pp    palette_data
  cc    random client_data[1..3] from slave 1-3, FFh if slave not exists
  hh    handshake_data, 11h+client_data[1]+client_data[2]+client_data[3]
  uu    random data, not used, ignore this value
Below automatically calculated by SWI 25h BIOS function (don't care about)
  llll  download length/4-34h
  rr    random data from each slave for encryption, FFh if slave not exists
  yyyy  encoded data in 16bit (multiplay) or 32bit (normal mode) units
  nnnn  response value, lower 16bit of destadr in GBA memory (00C0h and up)
  zzzz  16bit download CRC value, must be same for master and slaves
Pseudo Code for SWI 25h Transfer with Checksum and Encryption calculations
  if normal_mode    then c=C387h:x=C37Bh:k=43202F2Fh
  if multiplay_mode then c=FFF8h:x=A517h:k=6465646Fh
  m=dword(pp,cc,cc,cc):f=dword(hh,rr,rr,rr)
  for ptr=000000C0h to (file_size-4) step 4
    c=c xor data[ptr]:for i=1 to 32:c=c shr 1:if carry then c=c xor x:next
    m=(6F646573h*m)+1
    send_32_or_2x16 (data[ptr] xor (-2000000h-ptr) xor m xor k)
  next
  c=c xor f:for i=1 to 32:c=c shr 1:if carry then c=c xor x:next
  wait_all_units_ready_for_checksum:send_32_or_1x16 (c)
Whereas, explained: c=chksum,x=chkxor,f=chkfin,k=keyxor,m=keymul

Multiboot Communication
In Multiplay mode, master sends 16bit data, and receives 16bit data from each slave (or FFFFh if none). In Normal mode, master sends 32bit data (upper 16bit zero, lower 16bit as for multipay mode), and receives 32bit data (upper 16bit as for multiplay mode, and lower 16bit same as lower 16bit previously sent by master). Because SIODATA32 occupies same addresses as SIOMULTI0-1, the same transfer code can be used for both multiplay and normal mode (in normal mode SIOMULTI2-3 should be forced to FFFFh though). After each transfer, master should wait for Start bit cleared in SIOCNT register, followed by a 36us delay.
Note: The multiboot slave would also recognize data being sent in Joybus mode, however, master GBAs cannot use joybus mode (because GBA hardware cannot act as master in joybus mode).

Multiboot Slave Header
The transferred Header block is written to 2000000-20000BFh in slave RAM, the header must contain valid data (identically as for normal ROM-cartridge headers, including a copy of the Nintendo logo, correct header CRC, etc.), in most cases it'd be recommended just to transfer a copy of the master cartridges header from 8000000h-80000BFh.

Multiboot Slave Program/Data
The transferred main program/data block is written to 20000C0h and up (max 203FFFFh) in slave RAM, note that absolute addresses in the program must be then originated at 2000000h rather than 8000000h. In case that the master cartridge is 256K or less, it could just transfer a copy of the whole cartridge at 80000C0h and up, the master should then copy & execute its own ROM data into RAM as well.

Multiboot Slave Extended Header
For Multiboot slaves, separate Entry Point(s) must be defined at the beginning of the Program/Data block (the Entry Point in the normal header is ignored), also some reserved bytes in this section are overwritten by the Multiboot procedure. For more information see chapter about Cartridge Header.

Multiboot Slave with Cartridge
Beside for slaves without cartridge, multiboot can be also used for slaves which do have a cartridge inserted, if so, SELECT and START must be kept held down during power-on in order to switch the slave GBA into Multiboot mode (ie. to prevent it from starting the cartridge as normally).
The general idea is to enable newer programs to link to any existing older GBA programs, even if these older programs originally didn't have been intended to support linking.
The uploaded program may access the slaves SRAM, Flash ROM, or EEPROM (if any, allowing to read out or modify slave game positions), as well as cartridge ROM at 80000A0h-8000FFFh (the first 4KBytes, excluding the nintendo logo, allowing to read out the cartridge name from the header, for example).
The main part of the cartridge ROM is meant to be locked out in order to prevent software pirates from uploading "intruder" programs which would send back a copy of the whole cartridge to the master, however, for good or evil, at present time, current GBA models and GBA carts do not seem to contain any such protection.

Uploading Programs from PC
Beside for the ability to upload a program from one GBA to another, this feature can be also used to upload small programs from a PC to a GBA. For more information see chapter about External Connectors.

Nintendo DS
The GBA multiboot function requires a link port, and so, works on GBA and GBA SP only. The Nintendo DS in GBA mode does include the multiboot BIOS function, but it won't be of any use as the DS doesn't have a link port.


 BIOS Sound Functions

MidiKey2Freq
SoundBias
SoundChannelClear
SoundDriverInit
SoundDriverMain
SoundDriverMode
SoundDriverVSync
SoundDriverVSyncOff
SoundDriverVSyncOn
SoundWhatever0..4
SoundGetJumpList

SWI 1Fh (GBA) - MidiKey2Freq
Calculates the value of the assignment to ((SoundArea)sa).vchn[x].fr when playing the wave data, wa, with the interval (MIDI KEY) mk and the fine adjustment value (halftones=256) fp.
  r0  WaveData* wa
  r1  u8 mk
  r2  u8 fp
Return:
  r0  u32
This function is particularly popular because it allows to read from BIOS memory without copy protection range checks. The formula to read one byte (a) from address (i, 0..3FFF) is:
a = (MidiKey2Freq(i-(((i AND 3)+1)OR 3), 168, 0) * 2) SHR 24

SWI 19h (GBA) or SWI 08h (NDS7) - SoundBias
Increments or decrements the current level of the SOUNDBIAS register (with short delays) until reaching the desired new level. The upper bits of the register are kept unchanged.
  r0   BIAS level (0=Level 000h, any other value=Level 200h)
  r1   Delay Count (NDS only) (GBA uses a fixed delay count of 8)
Return: No return value.

SWI 1Eh (GBA) - SoundChannelClear
Clears all direct sound channels and stops the sound.
This function may not operate properly when the library which expands the sound driver feature is combined afterwards. In this case, do not use it.
No parameters, no return value.

SWI 1Ah (GBA) - SoundDriverInit
Initializes the sound driver. Call this only once when the game starts up.
It is essential that the work area already be secured at the time this function is called.
You cannot execute this driver multiple times, even if separate work areas have been prepared.
  r0  Pointer to work area for sound driver, SoundArea structure as follows:
       SoundArea (sa) Structure
        u32    ident      Flag the system checks to see whether the
                          work area has been initialized and whether it
                          is currently being accessed.
        vu8    DmaCount   User access prohibited
        u8     reverb     Variable for applying reverb effects to direct sound
        u16    d1         User access prohibited
        void   (*func)()  User access prohibited
        int    intp       User access prohibited
        void*  NoUse      User access prohibited
        SndCh  vchn[MAX]  The structure array for controlling the direct
                          sound channels (currently 8 channels are
                          available). The term "channel" here does
                          not refer to hardware channels, but rather to
                          virtual constructs inside the sound driver.
        s8     pcmbuf[PCM_BF*2]
       SoundChannel Structure
        u8         sf     The flag indicating the status of this channel.
                          When 0 sound is stopped.
                          To start sound, set other parameters and
                          then write 80h to here.
                          To stop sound, logical OR 40h for a
                          release-attached off (key-off), or write zero
                          for a pause. The use of other bits is
                          prohibited.
        u8         r1     User access prohibited
        u8         rv     Sound volume output to right side
        u8         lv     Sound volume output to left side
        u8         at     The attack value of the envelope. When the
                          sound starts, the volume begins at zero and
                          increases every 1/60 second. When it
                          reaches 255, the process moves on to the
                          next decay value.
        u8         de     The decay value of the envelope. It is
                          multiplied by "this value/256" every 1/60
                          sec. and when sustain value is reached, the
                          process moves to the sustain condition.
        u8         su     The sustain value of the envelope. The
                          sound is sustained by this amount.
                          (Actually, multiplied by rv/256, lv/256 and
                          output left and right.)
        u8         re     The release value of the envelope. Key-off
                          (logical OR 40h in sf) to enter this state.
                          The value is multiplied by "this value/256"
                          every 1/60 sec. and when it reaches zero,
                          this channel is completely stopped.
        u8         r2[4]  User access prohibited
        u32        fr     The frequency of the produced sound.
                          Write the value obtained with the
                          MidiKey2Freq function here.
        WaveData*  wp     Pointer to the sound's waveform data. The waveform
                          data can be generated automatically from the AIFF
                          file using the tool (aif2agb.exe), so users normally
                          do not need to create this themselves.
        u32        r3[6]  User access prohibited
        u8         r4[4]  User access prohibited
       WaveData Structure
        u16   type    Indicates the data type. This is currently not used.
        u16   stat    At the present time, non-looped (1 shot) waveform
                      is 0000h and forward loop is 4000h.
        u32   freq    This value is used to calculate the frequency.
                      It is obtained using the following formula:
                      sampling rate x 2^((180-original MIDI key)/12)
        u32   loop    Loop pointer (start of loop)
        u32   size    Number of samples (end position)
        s8    data[]  The actual waveform data. Takes (number of samples+1)
                      bytes of 8bit signed linear uncompressed data. The last
                      byte is zero for a non-looped waveform, and the same
                      value as the loop pointer data for a looped waveform.
Return: No return value.

SWI 1Ch (GBA) - SoundDriverMain
Main of the sound driver.
Call every 1/60 of a second. The flow of the process is to call SoundDriverVSync, which is explained later, immediately after the V-Blank interrupt.
After that, this routine is called after BG and OBJ processing is executed.
No parameters, no return value.

SWI 1Bh (GBA) - SoundDriverMode
Sets the sound driver operation mode.
  r0  Sound driver operation mode
       Bit    Expl.
       0-6    Direct Sound Reverb value (0-127, default=0) (ignored if Bit7=0)
       7      Direct Sound Reverb set (0=ignore, 1=apply reverb value)
       8-11   Direct Sound Simultaneously-produced (1-12 channels, default 8)
       12-15  Direct Sound Master volume (1-15, default 15)
       16-19  Direct Sound Playback Frequency (1-12 = 5734,7884,10512,13379,
              15768,18157,21024,26758,31536,36314,40137,42048, def 4=13379 Hz)
       20-23  Final number of D/A converter bits (8-11 = 9-6bits, def. 9=8bits)
       24-31  Not used.
Return: No return value.

SWI 1Dh (GBA) - SoundDriverVSync
An extremely short system call that resets the sound DMA. The timing is extremely critical, so call this function immediately after the V-Blank interrupt every 1/60 second.
No parameters, no return value.

SWI 28h (GBA) - SoundDriverVSyncOff
Due to problems with the main program if the V-Blank interrupts are stopped, and SoundDriverVSync cannot be called every 1/60 a second, this function must be used to stop sound DMA.
Otherwise, even if you exceed the limit of the buffer the DMA will not stop and noise will result.
No parameters, no return value.

SWI 29h (GBA) - SoundDriverVSyncOn
This function restarts the sound DMA stopped with the previously described SoundDriverVSyncOff.
After calling this function, have a V-Blank occur within 2/60 of a second and call SoundDriverVSync.
No parameters, no return value.

SWI 20h..24h (GBA) - SoundWhatever0..4 (Undocumented)
Whatever undocumented sound-related BIOS functions.

SWI 2Ah (GBA) - SoundGetJumpList (Undocumented)
Receives pointers to 36 additional sound-related BIOS functions.
  r0  Destination address (must be aligned by 4) (120h bytes buffer)


 Unpredictable Things

Forward
Most of the below is caused by 'traces' from previous operations which have used the databus. No promises that the results are stable on all current or future GBA models, and/or under all temperature and interference circumstances.
Also, below specifies 32bit data accesses only. When reading units less than 32bit, data is rotated depending on the alignment of the originally specified address, and 8bit or 16bit are then isolated from the 32bit value as usually.

Reading from BIOS Memory (00000000-00003FFF)
The BIOS memory is protected against reading, the GBA allows to read opcodes or data only if the program counter is located inside of the BIOS area. If the program counter is not in the BIOS area, reading will return the most recent successfully fetched BIOS opcode (eg. the opcode at [00DCh+8] after startup and SoftReset, the opcode at [0134h+8] during IRQ execution, and opcode at [013Ch+8] after IRQ execution, and opcode at [0188h+8] after SWI execution).

Reading from Unused Memory (00004000-1FFFFFF0,10000000-FFFFFFFF)
Accessing unused memory returns the recently pre-fetched opcode, ie. the 32bit opcode at $+8 in ARM state, or the 16bit-opcode at $+4 in THUMB state, in the later case the 16bit opcode is mirrored across both upper/lower 16bits of the returned 32bit data.
Note: This is caused by the prefetch pipeline in the CPU itself, not by the external gamepak prefetch, ie. it works for code in RAM as well.

Reading from Unused or Write-Only I/O Ports
Works like above unused memory when the entire 32bit memory fragment is Unused (eg. 0E0h) and/or Write-Only (eg. DMA0SAD). And otherwise, returns zero if the lower 16bit fragment is readable (eg. 04C=MOSAIC, 04E=NOTUSED/ZERO).

Reading from GamePak ROM when no Cartridge is inserted
Because Gamepak uses the same signal-lines for both 16bit data and for lower 16bit halfword address, the entire gamepak ROM area is effectively filled by incrementing 16bit values (Address/2 AND FFFFh).

Memory Mirrors
Most internal memory is mirrored across the whole 24bit/16MB address space in which it is located: On-board RAM at 2XXXXXX, In-Chip RAM at 3XXXXXXh, Palette RAM at 5XXXXXXh, VRAM at 6XXXXXXh, and OAM at 7XXXXXXh. Even though VRAM is sized 96K (64K+32K), it is repeated in steps of 128K (64K+32K+32K, the two 32K blocks itself being mirrors of each other).
BIOS ROM, Normal ROM Cartridges, and I/O area are NOT mirrored, the only exception is the undocumented I/O port at 4000800h (repeated each 64K).
The 64K SRAM area is mirrored across the whole 32MB area at E000000h-FFFFFFFh, also, inside of the 64K SRAM field, 32K SRAM chips are repeated twice.

Writing 8bit Data to Video Memory
Video Memory (BG, OBJ, OAM, Palette) can be written to in 16bit and 32bit units only. Attempts to write 8bit data (by STRB opcode) won't work:
Writes to OBJ (6010000h-6017FFFh) and OAM (7000000h-70003FFh) are ignored, the memory content remains unchanged. Writes to BG (6000000h-600FFFFh) and Palette (5000000h-50003FFh) are writing the new 8bit value to BOTH upper and lower 8bits of the addressed halfword, ie. "[addr AND NOT 1]=data*101h".

Accessing SRAM Area by 16bit/32bit
Reading retrieves 8bit value from specified address, multiplied by 0101h (LDRH) or by 01010101h (LDR). Writing changes the 8bit value at the specified address only, being set to LSB of (source_data ROR (address*8)).


 External Connectors

External Connectors
AUX Game Pak Bus
AUX DS Game Card Slot
AUX Link Port
AUX Sound/Headphone Socket and Battery/Power Supply

Getting access to Internal Pins
AUX Opening the GBA
AUX Mainboard

Xboo Multiboot Cable
AUX Xboo PC-to-GBA Multiboot Cable
AUX Xboo Flashcard Upload
AUX Xboo Burst Boot Backdoor
DS Xboo


 AUX Game Pak Bus

Game Pak Bus - 32pin cartridge slot
The cartridge bus may be used for both CGB and GBA game paks. In GBA mode, it is used as follows:
 Pin    Name    Dir  Expl.
 1      VDD     O    Power Supply 3.3V DC
 2      PHI     O    System Clock (selectable none, 4.19MHz, 8.38MHz, 16.78MHz)
 3      /WR     O    Write Select
 4      /RD     O    Read Select
 5      /CS     O    ROM Chip Select
 6-21   AD0-15  I/O  lower 16bit Address    and/or  16bit ROM-data (see below)
 22-29  A16-23  I/O  upper 8bit ROM-Address   or    8bit SRAM-data (see below)
 30     /CS2    O    SRAM Chip Select
 31     /REQ    I    Interrupt request (/IREQ) or DMA request (/DREQ)
 32     GND     O    Ground 0V
When accessing game pak SRAM, a 16bit address is output through AD0-AD15, then 8bit of data are transferred through A16-A23.
When accessing game pak ROM, a 24bit address is output through AD0-AD15 and A16-A23, then 16bit of data are transferred through AD0-AD15. The 24bit address is formed from the actual 25bit memory address (byte-steps), divided by two (halfword-steps).

8bit-Gamepak-Switch (GBA, GBA SP only) (not DS)
A small switch is located inside of the cartridge slot, the switch is pushed down when an 8bit cartridge is inserted, it is released when a GBA cartridge is inserted (or if no cartridge is inserted).
The switch mechanically controls whether VDD3 or VDD5 are output at VDD35; ie. in GBA mode 3V power supply/signals are used for the cartridge slot and link port, while in 8bit mode 5V are used.
The current state of the switch can be determined GBA mode via Port 204h (WAITCNT), if (and only if) the switch is pushed, then CGB mode can be activated via Port 000h (DISPCNT.3). The GBA boot procedure in BIOS uses this to detect 8bit carts and to set the GBA into 8bit mode.
In 8bit mode, the cartridge bus works much like for GBA SRAM, however, the 8bit /CS signal is expected at Pin 5, while GBA SRAM /CS2 at Pin 30 is interpreted as /RESET signal by the 8bit MBC chip (if any). In practice, this appears to result in 00h being received as data when attempting to read-out 8bit cartridges from inside of GBA mode.


 AUX DS Game Card Slot

  Pin    Dir  Name  Connection in cartridge
  1   >  -    GND   (ROM all unused Pins, EPROM Pin 4)
  2      Out  CLK   (4MB/s, ROM Pin 5, EPROM Pin 6)
  3   N  ?    ?     (ROM Pin 17) (Seems to be not connected in console)
  4   i  Out  /CS1  (ROM Pin 44) ROM Chipselect
  5   n  Out  /RES  (ROM Pin 42) Reset, switches ROM to unencrypted mode
  6   t  Out  /CS2  (EPROM Pin 1) EEPROM Chipselect
  7   e  In   IRQ   (GND)
  8   n  -    3.3V  (ROM Pins 2, 23, EPROM Pins 3, 7, 8)
  9   d  I/O  D0    (ROM Pin 18)
  10  o  I/O  D1    (ROM Pin 19)
  11     I/O  D2    (ROM Pin 20)
  12  C  I/O  D3    (ROM Pin 21)
  13  0  I/O  D4    (ROM Pin 24)
  14  1  I/O  D5    (ROM Pin 25)
  15  -  I/O  D6    (ROM Pin 26, EPROM Pin 2)
  16  0  I/O  D7    (ROM Pin 27, EPROM Pin 5)
  17  1  -    GND   (ROM all unused Pins, EPROM Pin 4)

Chipselect High-to-Low transitions are invoking commands, which are transmitted through data lines during next following eight CLK pulses, after the command transmission, further CLK pulses are used to transfer data, the data transfer ends at chipselect Low-to-High transition.
Data should be stable during CLK=LOW period throughout CLK rising edge.
Note: Supply Pins (1,8,17) are slightly longer than other pins.

The DS does also have a 32pin cartridge slot, that slot is used to run GBA carts in GBA mode, it can be also used as expansion port in DS mode.


 AUX Link Port

Serial Link Port Pin-Out (GBA:"EXT" - GBA SP:"EXT.1")
  Pin  Name  Cable
  1    VDD35 N/A       GBA Socket     GBA Plug   Old "8bit" Plug
  2    SO    Red       ___________    _________    ___________
  3    SI    Orange   |  2  4  6  |  / 2  4  6 \  |  2  4  6  |
  4    SD    Brown     \_1_ 3 _5_/   \_1_ 3 _5_/   \_1__3__5_/
  5    SC    Green         '-'           '-'
  6    GND   Blue      Socket Outside View / Plug Inside View
  Shield     Shield
Note: The pin numbers and names are printed on the GBA mainboard, colors as used in Nintendos AGB-005 and older 8bit cables.

Cable Diagrams (Left: GBA Cable, Right: 8bit Gameboy Cable)
  Big Plug  Middle Socket  Small Plug    Plug 1         Plug 2
   SI _________________     ____ SI       SI ______  ______SI
   SO ____________SO   |__ | ___ SO       SO ______><______SO
   GND____________GND______|____GND       GND_____________GND
   SD ____________SD____________ SD       SD               SD
   SC ____________SC____________ SC       SC _____________ SC
   Shield_______Shield_______Shield       Shield_______Shield

Normal Connection
Just connect the plugs to the two GBAs and leave the Middle Socket disconnected, in this mode both GBAs may behave as master or slave, regardless of whether using big or small plugs.
The GBA is (NOT ???) able to communicate in Normal mode with MultiPlay cables which do not have crossed SI/SO lines.

Multi-Play Connection
Connect two GBAs as normal, for each further GBAs connect an additional cable to the Middle socket of the first (or further) cable(s), up to four GBAs may be connected by using up to three cables.
The GBA which is connected to a Small Plug is master, the slaves are all connected to Large Plugs. (Only small plugs fit into the Middle Socket, so it's not possible to mess up something here).

Multi-Boot Connection
MultiBoot (SingleGamepak) is typically using Multi-Play communication, in this case it is important that the Small plug is connected to the master/sender (ie. to the GBA that contains the cartridge).

Non-GBA Mode Connection
First of all, it is not possible to link between 32bit GBA games and 8bit games, parts because of different cable protocol, and parts because of different signal voltages.
However, when a 8bit cartridge is inserted (the GBA is switched into 8bit compatibility mode) it may be connected to other 8bit games (monochrome gameboys, CGBs, or to other GBAs which are in 8bit mode also, but not to GBAs in 32bit mode).
When using 8bit link mode, an 8bit link cable must be used. The GBA link cables won't work, see below modification though.

Using a GBA 32bit cable for 8bit communication
Open the middle socket, and disconnect Small Plugs SI from GND, and connect SI to Large Plugs SO instead. You may also want to install a switch that allows to switch between SO and GND, the GND signal should be required for MultiPlay communication only though.
Also, cut off the plastic ledge from the plugs so that they fit into 8bit gameboy sockets.

Using a GBA 8bit cable for 32bit communication
The cable should theoretically work as is, as the grounded SI would be required for MultiPlay communication only. However, software that uses SD for Slave-Ready detection won't work unless when adding a SD-to-SD connection (the 8bit plugs probably do not even contain SD pins though).


 AUX Sound/Headphone Socket and Battery/Power Supply

GBA and DS: Stereo Sound Connector (3.5mm, female)
  Pin     Expl.
  Tip     Sound Left
  Middle  Sound Right
  Base    Ground

GBA: Power Supply Input
The GBA does not have a separate power supply input, however external power supplies can be connected to the battery socket. The recommended external voltage is 3.3V DC.

GBA SP and DS: EXT.2 Socket (Power Supply and Headphones)
  Pin     Expl.                   ___________
  A       PWR(-) GND             | X  ___  Y |
  D       PWR(+) 5.2V DC         | ---   --- |
  Y       Sound Left             |_A_B   C_D_|
  C       Sound Right                 \_/
  X,B     Unknown ???

External power input is used to charge the built-in battery, it cannot be used to run the SP without that battery. The switch in the middle between X and Y pins is probably used to disable the internal speaker when headphones are connected... or to sense the external power supply ???

Using PC +5V DC as Power Supply
Developers whom are using a PC for GBA programming will probably want to use the PC power supply (gained from disk drive power supply cable) for the GBA as well rather than dealing with batteries or external power supplies.
GBA: To lower the voltage to approximately 3 Volts use two diodes, type 1N 4004 or similar, the ring printed onto the diodes points towards the GBA side, connected as such:
  PC +5V (red)   --------|>|---|>|--------  GBA BT+
  PC GND (black) -------------------------  GBA BT-
GBA SP and DS: Works directly at +5V connected to EXT.2 socket (not to the internal battery pins), without any diodes.

Using Parallel Port Data Lines as Power Supply
When using a Multiboot cable connected to PC parallel port it'd be comfortable to use the existing cable for power supply as well. This method definitely exceeds all official ratings, in fact almost, it is ethically wrong, absolutely no warranty that it will work and that it won't destroy hardware and/or burn down the house, USE AT OWN RISK, among others the following may cause problems:
The parallel port data signals aren't intended to be mis-used as power supply, many BIOSes and operating systems will initialize some data lines as LOW and some as HIGH, when directly shortcutting data lines any low signals will most likely forcefully pull-down high signals, different parallel ports output between 3.66V and 5.00V data HIGH, and not all parallel ports will provide enough Watts even when all data lines are high.
The most violent method is to shortcut all Data Lines (Pin 2-9) and connect these to the GBA BT+ input, also connect Ground (Pin 19 or else) to GBA BT- input (unless already connected to GBA link port GND). If necessary insert 1 or 2 diodes (depending on the parallel port type) between Data Lines and BT+ (as describe for 5V input above). The above Data/Ground pin numbers are the same for 25-pin PC sockets and 36-pin printer plugs.
I've tried above with 4 different parallel ports with different results: One didn't worked at all (not enough power), one worked even though power LED signalized low power, another worked just fine, and the fourth worked only after inserting two diodes. Note that the GBA would consume even more power when inserting cartridges into it.
Finally, to turn the power on, output FFh to parallel port base address, and if necessary output 0Bh to base address+2. (Or launch the no$gba multiboot function which automatically initializes these ports).


 AUX Opening the GBA

Since Nintendo uses special screws with Y-shaped heads to seal the GBA (as well as older 8bit gameboys), it's always a bit difficult to loosen these screws.

Using Screwdrivers
One possible method is to use a small flat screwdriver, which might work, even though it'll most likely damage the screwdriver.
Reportedly, special Y-shaped screwdrivers for gameboys are available for sale somewhere (probably not at your local dealer, but you might find some in the internet or elsewhere).

Destroying the Screws
A more violent method is to take an electric drill, and drill-off the screw heads, this might also slightly damage the GBA plastic chase, also take care that the metal spoons from the destroyed screws don't produce shortcuts on the GBA mainboard.

Using a selfmade Screwdriver
A possible method is to take a larger screw (with a normal I-shaped, or X-shaped head), and to cut the screw-tip into Y-shape, you'll then end up with an "adapter" which can be placed in the middle between a normal screwdriver and gameboy screws.
Preferably, first cut the screw-tip into a shape like a "sharp three sided pyramid", next cut notches into each side. Access to a grinding-machine will be a great benefit, but you might get it working by using a normal metal-file as well.


 AUX Mainboard

Other possibly useful signals on the mainboard...

FIQ Signal
The FIQ (Fast Interrupt) signal (labeled FIQ on the mainboard) could be used as external interrupt (or debugging break) signal.
Caution: By default, the FIQ input is directly shortcut to VDD35 (+3V or +5V power supply voltage), this can be healed by scratching off the CL1 connection located close to the FIQ pin (FIQ still appears to have an internal pull-up, so that an external resistor is not required).
The GBA BIOS rejects FIQs if using normal ROM cartridge headers (or when no cartridge is inserted). When using a FIQ-compatible ROM header, Fast Interrupts can be then requested by pulling FIQ to ground, either by a push button, or by remote controlled signals.

RESET Signal
The RESET signal (found on the mainboard) could be used to reset the GBA by pulling the signal to ground for a few microseconds (or longer). The signal can be directly used (it is not shortcut to VDD35, unlike FIQ).
Note: A reset always launches nintendos time-consuming and annoying boot/logo procedure, so that it'd be recommend to avoid this "feature" when possible.

Joypad Signals
The 10 direction/button signals are each directly shortcut to ground when pressed, and pulled up high otherwise (unlike 8bit gameboys which used a 2x4 keyboard matrix), it'd be thus easy to connect a remote keyboard, keypad, joypad, or read-only 12bit parallel port.


 AUX Xboo PC-to-GBA Multiboot Cable

Below describes how to connect a PC parallel port to the GBA link port, allowing to upload small programs (max 256 KBytes) from no$gba's Utility menu into real GBAs.

This is possible because the GBA BIOS includes a built-in function for downloading & executing program code even when no cartridge is inserted. The program is loaded to 2000000h and up in GBA memory, and must contain cartridge header information just as for normal ROM cartridges (nintendo logo, checksum, etc., plus some additional multiboot info).

Basic Cable Connection
The general connection is very simple (only needs four wires), the only problem is that you need a special GBA plug or otherwise need to solder wires directly to the GBA mainboard (see Examples below).
  GBA  Name  Color                 SUBD CNTR Name
  2    SO    Red     ------------- 10   10   /ACK
  3    SI    Orange  ------------- 14   14   /AUTOLF
  5    SC    Green   ------------- 1    1    /STROBE
  6    GND   Blue    ------------- 19   19   GND
Optionally, also connect the following signals (see notes below):
  4    SD    Brown   ------------- 17   36   /SELECT  (double speed burst)
  3    SI    Orange  ----[===]---- 2..9 2..9 D0..7    (pull-up, 560 Ohm)
  5    SC    Green   ----[===]---- 2..9 2..9 D0..7    (pull-up, 560 Ohm)
  4    SD    Brown   ----[===]---- 2..9 2..9 D0..7    (pull-up, 560 Ohm)
  START  (mainboard) -----|>|----- 16   31   /INIT    (auto-reset, 1N4148)
  SELECT (mainboard) -----|>|----- 16   31   /INIT    (auto-reset, 1N4148)
  RESET  (mainboard) -----||------ 16   31   /INIT    (auto-reset, 300nF)
Notes: The GBA Pins are arranged from left to right as 2,4,6 in upper row, and 1,3,5 in lower row; outside view of GBA socket; flat side of socket upside. The above "Colors" are as used in most or all standard Nintendo link cables, note that Red/Orange will be exchanged at one end in cables with crossed SO/SI lines. At the PC side, use the SUBD pin numbers when connecting to a 25-pin SUBD plug, or CNTR pin numbers for 36-pin Centronics plug.

Optional SD Connection (Double Speed Burst)
The SD line is used for Double Speed Burst transfers only, in case that you are using a gameboy link plug for the connection, and if that plug does not have a SD-pin (such like from older 8bit gameboy cables), then you may leave out this connection. Burst Boot will then only work half as fast though.

Optional Pull-Ups (Improves Low-to-High Transition Speed)
If your parallel port works only with medium or slow delay settings, try to connect 560 Ohm resistors to SI/SC/SD inputs each, and the other resistor pin to any or all of the parallel port data lines (no$gba outputs high to pins 2..9).

Optional Reset Connection (CAUTION: Connection changed September 2004)
The Reset connection allows to automatically reset & upload data even if a program in the GBA has locked up (or if you've loaded a program that does not support nocash burst boot), without having to reset the GBA manually by switching it off and on (and without having to press Start+Select if a cartridge is inserted).
The two diodes should be 1N4148 or similar, the capacitor should be 300nF (eg. three 100nF capacitors in parallel). The signals are labeled on the mainboard, and can be found at following names / CPU pin numbers: RESET/CPU.125, SELECT/TP2/CPU.126, START/TP3/CPU.127.

Optional Power Supply Connection
Also, you may want to connect the power supply to parallel port data lines, see chapter Power Supply for details.

Transmission Speed
The first transfer will be very slow, and the GBA BIOS will display the boot logo for at least 4 seconds, even if the transfer has completed in less time. Once when you have uploaded a program with burst boot backdoor, further transfers will be ways faster. The table below shows transfer times for 0KByte - 256KByte files:
  Boot Mode_____Delay 0_______Delay 1_______Delay 2_____
  Double Burst  0.1s - 1.8s   0.1s - 3.7s   0.1s - 5.3s
  Single Burst  0.1s - 3.6s   0.1s - 7.1s   0.1s - 10.6s
  Normal Bios   4.0s - 9.0s   4.0s - 12.7s  4.0s - 16.3s
All timings measured on a 66MHz computer, best possible transmission speed should be 150KBytes/second. Timings might slightly vary depending on the CPU speed and/or operating system. Synchronization is done by I/O waitstates, that should work even on faster computers. Non-zero delays are eventually required for cables without pull-ups.

Requirements
Beside for the cable and plugs, no special requirements.
The cable should work with all parallel ports, including old-fashioned one-directional printer ports, as well as modern bi-directional EPP ports. Transfer timings should work stable regardless of the PCs CPU speed (see above though), and regardless of multitasking interruptions.
Both no$gba and the actual transmission procedure are using some 32bit code, so that either one currently requires 80386SX CPUs or above.

Connection Examples
As far as I can imagine, there are four possible methods how to connect the cable to the GBA. The first two methods don't require to open the GBA, and the other methods also allow to connect optional power supply and reset signal.
  1) Connect it to the GBA link port. Advantage: No need to
     open/modify the GBA. Disadvantage: You need a special plug,
     (typically gained by removing it from a gameboy link cable).
  2) Solder the cable directly to the GBA link port pins. Advantages:
     No plug required & no need to open the GBA. Disadvantages:
     You can't remove the cable, and the link port becomes unusable.
  3) Solder the cable directly to the GBA mainboard. Advantage: No
     plug required at the GBA side. Disadvantage: You'll always
     have a cable leaping out of the GBA even when not using it,
     unless you put a small standard plug between GBA and cable.
  4) Install a Centronics socket in the GBA (between power switch
     and headphone socket). Advantage: You can use a standard
     printer cable. Disadvantages: You need to cut a big hole into
     the GBAs battery box (which cannot be used anymore), the big
     cable might be a bit uncomfortable when holding the GBA.
Personally, I've decided to use the lastmost method as I don't like ending up with hundreds of special cables for different purposes, and asides, it's been fun to damage the GAB as much as possible.

Note
The above used PC parallel port signals are typically using 5V=HIGH while GBA link ports deal with 3V=HIGH. From my experiences, the different voltages do not cause communication problems (and do not damage the GBA and/or PC hardware), and after all real men don't care about a handful of volts, however, use at own risk.


 AUX Xboo Flashcard Upload

Flashcard Upload
Allows to write data to flashcards which are plugged into GBA cartridge slot, cartridge is automatically started after writing. On initial power-up, hold down START+SELECT to prevent the GBA from booting the old program in the flashcard.
The Upload function in Utility menu uses flashcard mode for files bigger than 256KB (otherwise uses multiboot mode automatically). Also, there's a separate Upload to Flashcard function in Remote Access submenu, allowing to write files of 256KB or less to flashcard if that should be desired.

Supported Flashcards
Function currently tested with Visoly Flash Advance (FA) 256Mbit (32MB) Turbo cartridge. Should also work with older FA versions. Please let me know if you are using other flashcards which aren't yet supported.

Flashcard Performance
Writing to flashcards may become potentially slow because of chip erase/write times, cable transmission time, and the sheer size of larger ROM-images. However, developers whom are testing different builts of their project usually won't need to rewrite the complete flashcard, Xboo uses a highspeed checksum mechanism (16MB/sec) to determine which flashcard sector(s) have changed, and does then re-write only these sector(s).
To eliminate transmission time, data transfer takes place in the erase phases. Erase/write time depends on the flashcard type, should be circa 1-2 seconds per 256KB sector. Because the cartridge is programmed directly in the GBA there's no need to remove it from the GBA when writing to it.

Developers Advice
Locate your program fragments at fixed addresses, for example, code and data blocks each aligned to 64K memory boundaries, so that data remains at the same location even when the size of code changes. Fill any blank spaces by value FFh for faster write time. Reduce the size of your ROM-image by efficient memory use (except for above alignment trick). Include the burst boot backdoor in your program, allowing to re-write the flashcard directly without resetting the GBA.

Lamers Advice
Xboo Flashcard support does not mean to get lame & to drop normal multiboot support, if your program fits into 256KB then make it <both> flashcard <and> multiboot compatible - multiboot reduces upload time, increases your flashcard lifetime, and will also work for people whom don't own flashcards.


 AUX Xboo Burst Boot Backdoor

When writing Xboo compatible programs, always include a burst boot "backdoor", this will allow yourself (and other people) to upload programs much faster as when using the normal GBA BIOS multiboot function. Aside from the improved transmission speed, there's no need to reset the GBA each time (eventually manually if you do not have reset connect), without having to press Start+Select (if cartridge inserted), and, most important, the time-consuming nintendo-logo intro is bypassed.

The Burst Boot Protocol
In your programs IRQ handler, add some code that watches out for burst boot IRQ requests. When sensing a burst boot request, download the actual boot procedure, and pass control to that procedure.
  Send (PC)    Reply (GBA)
  "BRST"       "BOOT"        ;request burst, and reply <prepared> for boot
  <wait 1/16s> <process IRQ> ;long delay, allow slave to enter IRQ handler
  llllllll     "OKAY"        ;send length in bytes, reply <ready> to boot
  dddddddd     --------      ;send data in 32bit units, reply don't care
  cccccccc     cccccccc      ;exchange crc (all data units added together)
Use normal mode, 32bit, external clock for all transfers. The received highspeed loader (currently approx. 180h bytes) is to be loaded to and started at 3000000h, which will then handle the actual download operation.

Below is an example program which works with multiboot, burstboot, and as normal rom/flashcard. The source can be assembled with a22i (the no$gba built-in assembler, see no$gba utility menu). When using other/mainstream assemblers, you'll eventually have to change some directives, convert numbers from NNNh into 0xNNN format, and define the origin somewhere in linker/makefile instead of in source code.

 .arm            ;select 32bit ARM instruction set
 .gba            ;indicate that it's a gameboy advance program
 .fix            ;automatically fix the cartridge header checksum
 org 2000000h    ;origin in RAM for multiboot-cable/no$gba-cutdown programs
 ;------------------
 ;cartridge header/multiboot header
  b     rom_start                ;-rom entry point
  dcb   ...insert logo here...   ;-nintento logo (156 bytes)
  dcb   'XBOO SAMPLE '           ;-title (12 bytes)
  dcb   0,0,0,0,  0,0            ;-game code (4 bytes), maker code (2 bytes)
  dcb   96h,0,0                  ;-fixed value 96h, main unit code, device type
  dcb   0,0,0,0,0,0,0            ;-reserved (7 bytes)
  dcb   0                        ;-software version number
  dcb   0                        ;-header checksum (set by .fix)
  dcb   0,0                      ;-reserved (2 bytes)
  b     ram_start                ;-multiboot ram entry point
  dcb   0,0                      ;-multiboot reserved bytes (destroyed by BIOS)
  dcb   0,0                      ;-blank padded (32bit alignment)
 ;------------------
 irq_handler:  ;interrupt handler (note: r0-r3 are pushed by BIOS)
  mov    r1,4000000h             ;\get I/O base address,
  ldr    r0,[r1,200h] ;IE/IF     ; read IE and IF,
  and    r0,r0,r0,lsr 16         ; isolate occurred AND enabled irqs,
  add    r3,r1,200h   ;IF        ; and acknowledge these in IF
  strh   r0,[r3,2]               ;/
  ldrh   r3,[r1,-8]              ;\mix up with BIOS irq flags at 3007FF8h,
  orr    r3,r3,r0                ; aka mirrored at 3FFFFF8h, this is required
  strh   r3,[r1,-8]              ;/when using the (VBlank-)IntrWait functions
  and    r3,r0,80h ;IE/IF.7 SIO  ;\
  cmp    r3,80h                  ; check if it's a burst boot interrupt
  ldreq  r2,[r1,120h] ;SIODATA32 ; (if interrupt caused by serial transfer,
  ldreq  r3,[msg_brst]           ; and if received data is "BRST",
  cmpeq  r2,r3                   ; then jump to burst boot)
  beq    burst_boot              ;/
  ;... insert your own interrupt handler code here ...
  bx     lr                      ;-return to the BIOS interrupt handler
 ;------------------
 burst_boot:     ;requires incoming r1=4000000h
  ;... if your program uses DMA, disable any active DMA transfers here ...
  ldr   r4,[msg_okay]            ;\
  bl    sio_transfer             ; receive transfer length/bytes & reply "OKAY"
  mov   r2,r0 ;len               ;/
  mov   r3,3000000h   ;dst       ;\
  mov   r4,0  ;crc               ;
 @@lop:                          ;
  bl    sio_transfer             ; download burst loader to 3000000h and up
  stmia [r3]!,r0      ;dst       ;
  add   r4,r4,r0      ;crc       ;
  subs  r2,r2,4       ;len       ;
  bhi   @@lop                    ;/
  bl    sio_transfer             ;-send crc value to master
  b     3000000h  ;ARM state!    ;-launch actual transfer / start the loader
 ;------------------
 sio_transfer:  ;serial transfer subroutine, 32bit normal mode, external clock
  str   r4,[r1,120h]  ;siodata32 ;-set reply/send data
  ldr   r0,[r1,128h]  ;siocnt    ;\
  orr   r0,r0,80h                ; activate slave transfer
  str   r0,[r1,128h]  ;siocnt    ;/
 @@wait:                         ;\
  ldr   r0,[r1,128h]  ;siocnt    ; wait until transfer completed
  tst   r0,80h                   ;
  bne   @@wait                   ;/
  ldr   r0,[r1,120h]  ;siodata32 ;-get received data
  bx    lr
 ;---
 msg_boot dcb 'BOOT'     ;\
 msg_okay dcb "OKAY"     ; ID codes for the burstboot protocol
 msg_brst dcb "BRST"     ;/
 ;------------------
 download_rom_to_ram:
  mov  r0,8000000h  ;src/rom     ;\
  mov  r1,2000000h  ;dst/ram     ;
  mov  r2,40000h/16 ;length      ; transfer the ROM content
 @@lop:                          ; into RAM (done in units of 4 words/16 bytes)
  ldmia [r0]!,r4,r5,r6,r7        ; currently fills whole 256K of RAM,
  stmia [r1]!,r4,r5,r6,r7        ; even though the proggy is smaller
  subs  r2,r2,1                  ;
  bne   @@lop                    ;/
  sub   r15,lr,8000000h-2000000h ;-return (retadr rom/8000XXXh -> ram/2000XXXh)
 ;------------------
 init_interrupts:
  mov  r4,4000000h               ;-base address for below I/O registers
  ldr  r0,=irq_handler           ;\install IRQ handler address
  str  r0,[r4,-4]   ;IRQ HANDLER ;/at 3FFFFFC aka 3007FFC
  mov  r0,0008h                  ;\enable generating vblank irqs
  strh r0,[r4,4h]   ;DISPSTAT    ;/
  mrs  r0,cpsr                   ;\
  bic  r0,r0,80h                 ; cpu interrupt enable (clear i-flag)
  msr  cpsr,r0                   ;/
  mov  r0,0                      ;\
  str  r0,[r4,134h] ;RCNT        ; init SIO normal mode, external clock,
  ldr  r0,=5080h                 ; 32bit, IRQ enable, transfer started
  str  r0,[r4,128h] ;SIOCNT      ; output "BOOT" (indicate burst boot prepared)
  ldr  r0,[msg_boot]             ;
  str  r0,[r4,120h] ;SIODATA32   ;/
  mov  r0,1                      ;\interrupt master enable
  str  r0,[r4,208h] ;IME=1       ;/
  mov  r0,81h                    ;\enable execution of vblank IRQs,
  str  r0,[r4,200h] ;IE=81h      ;/and of SIO IRQs (burst boot)
  bx   lr
 ;------------------
 rom_start:   ;entry point when booted from flashcart/rom
  bl   download_rom_to_ram       ;-download ROM to RAM (returns to ram_start)
 ram_start:   ;entry point for multiboot/burstboot
  mov  r0,0feh                   ;\reset all registers, and clear all memory
  swi  10000h ;RegisterRamReset  ;/(except program code in wram at 2000000h)
  bl   init_interrupts           ;-install burst boot irq handler
  mov  r4,4000000h               ;\enable video,
  strh r4,[r4,000h] ;DISPCNT     ;/by clearing the forced blank bit
 @@mainloop:
  swi  50000h ;VBlankIntrWait    ;-wait one frame (cpu in low power mode)
  mov  r5,5000000h               ;\increment the backdrop palette color
  str  r8,[r5]                   ; (ie. display a blinking screen)
  add  r8,r8,1                   ;/
  b    @@mainloop
 ;------------------
 .pool
 end


 CPU Reference

General ARM7TDMI Information
CPU Overview
CPU Register Set
CPU Flags
CPU Exceptions

The ARM7TDMI Instruction Sets
THUMB Instruction Set
ARM Instruction Set
Pseudo Instructions and Directives

Further Information
ARM CP15 System Control Coprocessor
CPU Instruction Cycle Times
CPU Versions
CPU Data Sheet


 CPU Overview

The ARM7TDMI is a 32bit RISC (Reduced Instruction Set Computer) CPU, designed by ARM (Advanced RISC Machines), and designed for both high performance and low power consumption.

Fast Execution
Depending on the CPU state, all opcodes are sized 32bit or 16bit (that's counting both the opcode bits and its parameters bits) providing fast decoding and execution. Additionally, pipelining allows - (a) one instruction to be executed while (b) the next instruction is decoded and (c) the next instruction is fetched from memory - all at the same time.

Data Formats
The CPU manages to deal with 8bit, 16bit, and 32bit data, that are called:
   8bit - Byte
  16bit - Halfword
  32bit - Word

The two CPU states
As mentioned above, two CPU states exist:
- ARM state: Uses the full 32bit instruction set (32bit opcodes)
- THUMB state: Uses a cutdown 16bit instruction set (16bit opcodes)
Regardless of the opcode-width, both states are using 32bit registers, allowing 32bit memory addressing as well as 32bit arithmetic/logical operations.

When to use ARM state
Basically, there are two advantages in ARM state:
 - Each single opcode provides more functionality, resulting
   in faster execution when using a 32bit bus memory system
   (such like opcodes stored in GBA Work RAM).
 - All registers R0-R15 can be accessed directly.
The downsides are:
 - Not so fast when using 16bit memory system
   (but it still works though).
 - Program code occupies more memory space.

When to use THUMB state
There are two major advantages in THUMB state:
 - Faster execution up to approx 160% when using a 16bit bus
   memory system (such like opcodes stored in GBA GamePak ROM).
 - Reduces code size, decreases memory overload down to approx 65%.
The disadvantages are:
 - Not as multi-functional opcodes as in ARM state, so it will
   be sometimes required use more than one opcode to gain a
   similar result as for a single opcode in ARM state.
 - Most opcodes allow only registers R0-R7 to be used directly.

Combining ARM and THUMB state
Switching between ARM and THUMB state is done by a normal branch (BX) instruction which takes only a handful of cycles to execute (allowing to change states as often as desired - with almost no overload).

Also, as both ARM and THUMB are using the same register set, it is possible to pass data between ARM and THUMB mode very easily.

The best memory & execution performance can be gained by combining both states: THUMB for normal program code, and ARM code for timing critical subroutines (such like interrupt handlers, or complicated algorithms).

Note: ARM and THUMB code cannot be executed simultaneously.

Automatic state changes
Beside for the above manual state switching by using BX instructions, the following situations involve automatic state changes:
- CPU switches to ARM state when executing an exception
- User switches back to old state when leaving an exception


 CPU Register Set

Overview
The following table shows the ARM7TDMI register set which is available in each mode. There's a total of 37 registers (32bit each), 31 general registers (Rxx) and 6 status registers (xPSR).
Note that only some registers are 'banked', for example, each mode has it's own R14 register: called R14, R14_fiq, R14_svc, etc. for each mode respectively.
However, other registers are not banked, for example, each mode is using the same R0 register, so writing to R0 will always affect the content of R0 in other modes also.

  System/User FIQ       Supervisor Abort     IRQ       Undefined
  --------------------------------------------------------------
  R0          R0        R0         R0        R0        R0
  R1          R1        R1         R1        R1        R1
  R2          R2        R2         R2        R2        R2
  R3          R3        R3         R3        R3        R3
  R4          R4        R4         R4        R4        R4
  R5          R5        R5         R5        R5        R5
  R6          R6        R6         R6        R6        R6
  R7          R7        R7         R7        R7        R7
  --------------------------------------------------------------
  R8          R8_fiq    R8         R8        R8        R8
  R9          R9_fiq    R9         R9        R9        R9
  R10         R10_fiq   R10        R10       R10       R10
  R11         R11_fiq   R11        R11       R11       R11
  R12         R12_fiq   R12        R12       R12       R12
  R13 (SP)    R13_fiq   R13_svc    R13_abt   R13_irq   R13_und
  R14 (LR)    R14_fiq   R14_svc    R14_abt   R14_irq   R14_und
  R15 (PC)    R15       R15        R15       R15       R15
  --------------------------------------------------------------
  CPSR        CPSR      CPSR       CPSR      CPSR      CPSR
  --          SPSR_fiq  SPSR_svc   SPSR_abt  SPSR_irq  SPSR_und
  --------------------------------------------------------------

R0-R12 Registers (General Purpose Registers)
These thirteen registers may be used for whatever general purposes. Basically, each is having same functionality and performance, ie. there is no 'fast accumulator' for arithmetic operations, and no 'special pointer register' for memory addressing.
However, in THUMB mode only R0-R7 (Lo registers) may be accessed freely, while R8-R12 and up (Hi registers) can be accessed only by some instructions.

R13 Register (SP)
This register is used as Stack Pointer (SP) in THUMB state. While in ARM state the user may decided to use R13 and/or other register(s) as stack pointer(s), or as general purpose register.
As shown in the table above, there's a separate R13 register in each mode, and (when used as SP) each exception handler may (and MUST!) use its own stack.

R14 Register (LR)
This register is used as Link Register (LR). That is, when calling to a sub-routine by a Branch with Link (BL) instruction, then the return address (ie. old value of PC) is saved in this register.
Storing the return address in the LR register is obviously faster than pushing it into memory, however, as there's only one LR register for each mode, the user must manually push its content before issuing 'nested' subroutines.
Same happens when an exception is called, PC is saved in LR of new mode.
Note: In ARM mode, R14 may be used as general purpose register also, provided that above usage as LR register isn't required.

R15 Register (PC)
R15 is always used as program counter (PC). Note that when reading R15, this will usually return a value of PC+nn because of read-ahead (pipelining), whereas 'nn' depends on the instruction and on the CPU state (ARM or THUMB).

CPSR and SPSR (Program Status Registers) (ARMv3 and up)
The current condition codes (flags) and CPU control bits are stored in the CPSR register. When an exception arises, the old CPSR is saved in the SPSR of the respective exception-mode (much like PC is saved in LR).
For details refer to chapter about CPU Flags.


 CPU Flags

Current Program Status Register (CPSR)
  Bit   Expl.
  31    N - Sign Flag       (0=Not Signed, 1=Signed)
  30    Z - Zero Flag       (0=Not Zero, 1=Zero)
  29    C - Carry Flag      (0=No Carry, 1=Carry)
  28    V - Overflow Flag   (0=No Overflow, 1=Overflow)
  27    Q - Sticky Overflow (1=Sticky Overflow, ARMv5TE and up only)
  26-8  Reserved            (For future use) - Do not change manually!
  7     I - IRQ disable     (0=Enable, 1=Disable)
  6     F - FIQ disable     (0=Enable, 1=Disable)
  5     T - State Bit       (0=ARM, 1=THUMB) - Do not change manually!
  4-0   M4-M0 - Mode Bits   (See below)

Bit 31-28: Condition Code Flags (N,Z,C,V)
These bits reflect results of logical or arithmetic instructions. In ARM mode, it is often optionally whether an instruction should modify flags or not, for example, it is possible to execute a SUB instruction that does NOT modify the condition flags.
In ARM state, all instructions can be executed conditionally depending on the settings of the flags, such like MOVEQ (Move if Z=1). While In THUMB state, only Branch instructions (jumps) can be made conditionally.

Bit 27: Sticky Overflow Flag (Q) - ARMv5TE and ARMv5TExP and up only
Used by QADD, QSUB, QDADD, QDSUB, SMLAxy, and SMLAWy only. These opcodes set the Q-flag in case of overflows, but leave it unchanged otherwise. The Q-flag can be tested/reset by MSR/MRS opcodes only.

Bit 27-8: Reserved Bits (except Bit 27 on ARMv5TE and up, see above)
These bits are reserved for possible future implementations. For best forwards compatibility, the user should never change the state of these bits, and should not expect these bits to be set to a specific value.

Bit 7-0: Control Bits (I,F,T,M4-M0)
These bits may change when an exception occurs. In privileged modes (non-user modes) they may be also changed manually.
The interrupt bits I and F are used to disable IRQ and FIQ interrupts respectively (a setting of "1" means disabled).
The T Bit signalizes the current state of the CPU (0=ARM, 1=THUMB), this bit should never be changed manually - instead, changing between ARM and THUMB state must be done by BX instructions.
The Mode Bits M4-M0 contain the current operating mode.
  Binary Hex Dec  Expl.
  10000b 10h 16 - User (non-privileged)
  10001b 11h 17 - FIQ
  10010b 12h 18 - IRQ
  10011b 13h 19 - Supervisor (SWI)
  10111b 17h 23 - Abort
  11011b 1Bh 27 - Undefined
  11111b 1Fh 31 - System (privileged 'User' mode) (ARMv4 and up)
Writing any other values into the Mode bits is not allowed.

Saved Program Status Registers (SPSR_<mode>)
Additionally to above CPSR, five Saved Program Status Registers exist:
SPSR_fiq, SPSR_svc, SPSR_abt, SPSR_irq, SPSR_und
Whenever the CPU enters an exception, the current status register (CPSR) is copied to the respective SPSR_<mode> register. Note that there is only one SPSR for each mode, so nested exceptions inside of the same mode are allowed only if the exception handler saves the content of SPSR in memory.
For example, for an IRQ exception: IRQ-mode is entered, and CPSR is copied to SPSR_irq. If the interrupt handler wants to enable nested IRQs, then it must first push SPSR_irq before doing so.


 CPU Exceptions

Exceptions are caused by interrupts or errors. In the ARM7TDMI the following exceptions may arise, sorted by priority, starting with highest priority:
- Reset
- Data Abort
- FIQ
- IRQ
- Prefetch Abort
- Software Interrupt
- Undefined Instruction

Exception Vectors
The following are the exception vectors in memory. That is, when an exception arises, CPU is switched into ARM state, and the program counter (PC) is loaded by the respective address.
  Address    Exception                  Mode on Entry      Interrupt Flags
  BASE+00h   Reset                      Supervisor (_svc)  I=1, F=1
  BASE+04h   Undefined Instruction      Undefined  (_und)  I=1, F=unchanged
  BASE+08h   Software Interrupt (SWI)   Supervisor (_svc)  I=1, F=unchanged
  BASE+0Ch   Prefetch Abort             Abort      (_abt)  I=1, F=unchanged
  BASE+10h   Data Abort                 Abort      (_abt)  I=1, F=unchanged
  BASE+14h   (Reserved)                 -          -       -
  BASE+18h   Normal Interrupt (IRQ)     IRQ        (_irq)  I=1, F=unchanged
  BASE+1Ch   Fast Interrupt (FIQ)       FIQ        (_fiq)  I=1, F=1
BASE is normally 00000000h, but may be optionally FFFF0000h in some ARM CPUs.
As there's only space for one ARM opcode at each of the above addresses, it'd be usually recommended to deposit a Branch opcode into each vector, which'd then redirect to the actual exception handlers address.

Actions performed by CPU when entering an exception
  - R14=PC+nn              ;save old PC, ie. return address
  - SPSR_<new mode>=CPSR   ;save old flags
  - CPSR new T,M bits      ;set to T=0 (ARM state), and M4-0=new mode
  - CPSR new I bit         ;IRQs disabled (I=1), done by ALL exceptions
  - CPSR new F bit         ;FIQs disabled (F=1), done by Reset and FIQ only
  - PC=exception_vector    ;see table above
Above "PC+nn" depends on the type of exception. Basically, in ARM state that nn-offset is caused by pipelining, and in THUMB state an identical ARM-style 'offset' is generated (even though the 'base address' may be only halfword-aligned).

Required user-handler actions when returning from an exception
Restore any general registers (R0-R14) which might have been modified by the exception handler. Use return-instruction as listed in the respective descriptions below, this will both restore PC and CPSR - that automatically involves that the old CPU state (THUMB or ARM) as well as old state of FIQ and IRQ disable flags are restored.
As mentioned above (see action on entering...), the return address is always saved in ARM-style format, so that exception handler may use the same return-instruction, regardless of whether the exception has been generated from inside of ARM or THUMB state.

FIQ (Fast Interrupt Request)
This interrupt is generated by a LOW level on the nFIQ input. It is supposed to process timing critical interrupts at a high priority, as fast as possible.
Additionally to the common banked registers (R13_fiq,R14_fiq), five extra banked registers (R8_fiq-R12_fiq) are available in FIQ mode. The exception handler may freely access these registers without modifying the main programs R8-R12 registers (and without having to save that registers on stack).
In privileged (non-user) modes, FIQs may be also manually disabled by setting the F Bit in CPSR.

IRQ (Normal Interrupt Request)
This interrupt is generated by a LOW level on the nIRQ input. Unlike FIQ, the IRQ mode is not having its own banked R8-R12 registers.
IRQ is having lower priority than FIQ, and IRQs are automatically disabled when a FIQ exception becomes executed. In privileged (non-user) modes, IRQs may be also manually disabled by setting the I Bit in CPSR.
To return from IRQ Mode (continuing at following opcode):
  SUBS PC,R14,4   ;both PC=R14_irq-4, and CPSR=SPSR_irq

Software Interrupt
Generated by a software interrupt instruction (SWI). Recommended to request a supervisor (operating system) function. The SWI instruction may also contain a parameter in the 'comment field' of the opcode:
In case that your main program issues SWIs from both inside of THUMB and ARM states, then your exception handler must separate between 24bit comment fields in ARM opcodes, and 8bit comment fields in THUMB opcodes (if necessary determine old state by examining T Bit in SPSR_svc); However, in Little Endian mode, you could use only the most significant 8bits of the 24bit ARM comment field (as done in the GBA, for example) - the exception handler could then process the BYTE at [R14-2], regardless of whether it's been called from ARM or THUMB state.
To return from Supervisor Mode (continuing at following opcode):
  MOVS PC,R14   ;both PC=R14_svc, and CPSR=SPSR_svc
Note: Like all other exceptions, SWIs are always executed in ARM state, no matter whether it's been caused by an ARM or THUMB state SWI instruction.

Undefined Instruction Exception (supported by ARMv3 and up)
This exception is generated when the CPU comes across an instruction which it cannot handle. Most likely signalizing that the program has locked up, and that an errormessage should be displayed.
However, it might be also used to emulate custom functions, ie. as an additional 'SWI' instruction (which'd use R14_und and SPSR_und though, and it'd thus allow to execute the Undefined Instruction handler from inside of Supervisor mode without having to save R14_svc and SPSR_svc).
To return from Undefined Mode (continuing at following opcode):
  MOVS PC,R14   ;both PC=R14_und, and CPSR=SPSR_und
Note that not all unused opcodes are necessarily producing an exception, for example, an ARM state Multiply instruction with Bit 6 set to "1" would be blindly accepted as 'legal' opcode.

Abort (supported by ARMv3 and up)
Aborts (page faults) are mostly supposed for virtual memory systems (ie. not used in GBA, as far as I know), otherwise they might be used just to display an error message. Two types of aborts exists:
- Prefetch Abort (occurs during an instruction prefetch)
- Prefetch Abort (also occurs on BKPT opcodes, ARMv5 and up)
- Data Abort (occurs during a data access)
A virtual memory systems abort handler would then most likely determine the fault address: For prefetch abort that's just "R14_abt-4". For Data abort, the THUMB or ARM instruction at "R14_abt-8" needs to be 'disassembled' in order to determine the addressed data in memory.
The handler would then fix the error by loading the respective memory page into physical memory, and then retry to execute the SAME instruction again, by returning as follows:
  prefetch abort: SUBS PC,R14,#4   ;PC=R14_abt-4, and CPSR=SPSR_abt
  data abort:     SUBS PC,R14,#8   ;PC=R14_abt-8, and CPSR=SPSR_abt
Separate exception vectors for prefetch/data abort exists, each should use the respective return instruction as shown above.

Reset
Forces PC=VVVV0000h, and forces control bits of CPSR to T=0 (ARM state), F=1 and I=1 (disable FIQ and IRQ), and M4-0=10011b (Supervisor mode).


 THUMB Instruction Set

When operating in THUMB state, cut-down 16bit opcodes are used.
THUMB supported on T-variants of ARMv4 and up, ie. ARMv4T, ARMv5T, etc.

Summary
THUMB Instruction Summary

Register Operations
THUMB.1: move shifted register
THUMB.2: add/subtract
THUMB.3: move/compare/add/subtract immediate
THUMB.4: ALU operations
THUMB.5: Hi register operations/branch exchange

Memory Addressing Operations
THUMB.6: load PC-relative
THUMB.7: load/store with register offset
THUMB.8: load/store sign-extended byte/halfword
THUMB.9: load/store with immediate offset
THUMB.10: load/store halfword
THUMB.11: load/store SP-relative
THUMB.12: get relative address
THUMB.13: add offset to stack pointer
THUMB.14: push/pop registers
THUMB.15: multiple load/store

Jumps and Calls
THUMB.16: conditional branch
THUMB.17: software interrupt and breakpoint
THUMB.18: unconditional branch
THUMB.19: long branch with link
(See also THUMB.5: BX Rs, and ADD/MOV PC,Rs.)

Note:
Switching between ARM and THUMB state can be done by using the Branch and Exchange (BX) instruction.


 THUMB Instruction Summary

The table below lists all THUMB mode instructions with clock cycles, affected CPSR flags, Format/chapter number, and description.
Only register R0..R7 can be used in thumb mode (unless R8-15,SP,PC are explicitly mentioned).

Logical Operations
  Instruction        Cycles Flags Format Expl.
  MOV Rd,Imm8bit      1S     NZ--  3   Rd=nn
  MOV Rd,Rs           1S     NZ00  2   Rd=Rs+0
  MOV R0..14,R8..15   1S     ----  5   Rd=Rs
  MOV R8..14,R0..15   1S     ----  5   Rd=Rs
  MOV R15,R0..15      2S+1N  ----  5   PC=Rs
  MVN Rd,Rs           1S     NZ--  4   Rd=NOT Rs
  AND Rd,Rs           1S     NZ--  4   Rd=Rd AND Rs
  TST Rd,Rs           1S     NZ--  4 Void=Rd AND Rs
  BIC Rd,Rs           1S     NZ--  4   Rd=Rd AND NOT Rs
  ORR Rd,Rs           1S     NZ--  4   Rd=Rd OR Rs
  EOR Rd,Rs           1S     NZ--  4   Rd=Rd XOR Rs
  LSL Rd,Rs,Imm5bit   1S     NZc-  1   Rd=Rs SHL nn
  LSL Rd,Rs           1S+1I  NZc-  4   Rd=Rd SHL (Rs AND 0FFh)
  LSR Rd,Rs,Imm5bit   1S     NZc-  1   Rd=Rs SHR nn
  LSR Rd,Rs           1S+1I  NZc-  4   Rd=Rd SHR (Rs AND 0FFh)
  ASR Rd,Rs,Imm5bit   1S     NZc-  1   Rd=Rs SRA nn
  ASR Rd,Rs           1S+1I  NZc-  4   Rd=Rd SRA (Rs AND 0FFh)
  ROR Rd,Rs           1S+1I  NZc-  4   Rd=Rd ROR (Rs AND 0FFh)
  NOP                 1S     ----  5   R8=R8
Carry flag affected only if shift amount is non-zero.

Arithmetic Operations and Multiply
  Instruction        Cycles Flags Format Expl.
  ADD Rd,Rs,Imm3bit   1S     NZCV  2   Rd=Rs+nn
  ADD Rd,Imm8bit      1S     NZCV  3   Rd=Rd+nn
  ADD Rd,Rs,Rn        1S     NZCV  2   Rd=Rs+Rn
  ADD R0..14,R8..15   1S     ----  5   Rd=Rd+Rs
  ADD R8..14,R0..15   1S     ----  5   Rd=Rd+Rs
  ADD R15,R0..15      2S+1N  ----  5   PC=Rd+Rs
  ADD Rd,PC,Imm8bit*4 1S     ---- 12   Rd=(($+4) AND NOT 2)+nn
  ADD Rd,SP,Imm8bit*4 1S     ---- 12   Rd=SP+nn
  ADD SP,Imm7bit*4    1S     ---- 13   SP=SP+nn
  ADD SP,-Imm7bit*4   1S     ---- 13   SP=SP-nn
  ADC Rd,Rs           1S     NZCV  4   Rd=Rd+Rs+Cy
  SUB Rd,Rs,Imm3Bit   1S     NZCV  2   Rd=Rs-nn
  SUB Rd,Imm8bit      1S     NZCV  3   Rd=Rd-nn
  SUB Rd,Rs,Rn        1S     NZCV  2   Rd=Rs-Rn
  SBC Rd,Rs           1S     NZCV  4   Rd=Rd-Rs-NOT Cy
  NEG Rd,Rs           1S     NZCV  4   Rd=0-Rs
  CMP Rd,Imm8bit      1S     NZCV  3 Void=Rd-nn
  CMP Rd,Rs           1S     NZCV  4 Void=Rd-Rs
  CMP R0-15,R8-15     1S     NZCV  5 Void=Rd-Rs
  CMP R8-15,R0-15     1S     NZCV  5 Void=Rd-Rs
  CMN Rd,Rs           1S     NZCV  4 Void=Rd+Rs
  MUL Rd,Rs           1S+mI  NZx-  4   Rd=Rd*Rs

Jumps and Calls
  Instruction        Cycles    Flags Format Expl.
  B disp              2S+1N     ---- 18  PC=$+/-2048
  BL disp             3S+1N     ---- 19  PC=$+/-4M, LR=$+5
  B{cond=true} disp   2S+1N     ---- 16  PC=$+/-0..256
  B{cond=false} disp  1S        ---- 16  N/A
  BX R0..15           2S+1N     ----  5  PC=Rs, ARM/THUMB (Rs bit0)
  SWI Imm8bit         2S+1N     ---- 17  PC=8, ARM SVC mode, LR=$+2
  BKPT Imm8bit        ???       ---- 17  ??? ARM9 Prefetch Abort
  BLX disp            ???       ---- ??? ??? ARM9
  BLX R0..R14         ???       ---- ??? ??? ARM9
  POP {Rlist,}PC   (n+1)S+2N+1I ---- 14
  MOV R15,R0..15      2S+1N     ----  5  PC=Rs
  ADD R15,R0..15      2S+1N     ----  5  PC=Rd+Rs
The thumb BL instruction occupies two 16bit opcodes, 32bit in total.

Memory Load/Store
  Instruction        Cycles    Flags Format Expl.
  LDR  Rd,[Rb,5bit*4] 1S+1N+1I  ----  9  Rd = WORD[Rb+nn]
  LDR  Rd,[PC,8bit*4] 1S+1N+1I  ----  6  Rd = WORD[PC+nn]
  LDR  Rd,[SP,8bit*4] 1S+1N+1I  ---- 11  Rd = WORD[SP+nn]
  LDR  Rd,[Rb,Ro]     1S+1N+1I  ----  7  Rd = WORD[Rb+Ro]
  LDRB Rd,[Rb,5bit*1] 1S+1N+1I  ----  9  Rd = BYTE[Rb+nn]
  LDRB Rd,[Rb,Ro]     1S+1N+1I  ----  7  Rd = BYTE[Rb+Ro]
  LDRH Rd,[Rb,5bit*2] 1S+1N+1I  ---- 10  Rd = HALFWORD[Rb+nn]
  LDRH Rd,[Rb,Ro]     1S+1N+1I  ----  8  Rd = HALFWORD[Rb+Ro]
  LDSB Rd,[Rb,Ro]     1S+1N+1I  ----  8  Rd = SIGNED_BYTE[Rb+Ro]
  LDSH Rd,[Rb,Ro]     1S+1N+1I  ----  8  Rd = SIGNED_HALFWORD[Rb+Ro]
  STR  Rd,[Rb,5bit*4] 2N        ----  9  WORD[Rb+nn] = Rd
  STR  Rd,[SP,8bit*4] 2N        ---- 11  WORD[SP+nn] = Rd
  STR  Rd,[Rb,Ro]     2N        ----  7  WORD[Rb+Ro] = Rd
  STRB Rd,[Rb,5bit*1] 2N        ----  9  BYTE[Rb+nn] = Rd
  STRB Rd,[Rb,Ro]     2N        ----  7  BYTE[Rb+Ro] = Rd
  STRH Rd,[Rb,5bit*2] 2N        ---- 10  HALFWORD[Rb+nn] = Rd
  STRH Rd,[Rb,Ro]     2N        ----  8  HALFWORD[Rb+Ro]=Rd
  PUSH {Rlist}{LR}    (n-1)S+2N ---- 14
  POP  {Rlist}{PC}              ---- 14  (ARM9: with mode switch)
  STMIA Rb!,{Rlist}   (n-1)S+2N ---- 15
  LDMIA Rb!,{Rlist}   nS+1N+1I  ---- 15

THUMB Binary Opcode Format
This table summarizes the position of opcode/parameter bits for THUMB mode instructions, Format 1-19.

 Form|_15|_14|_13|_12|_11|_10|_9_|_8_|_7_|_6_|_5_|_4_|_3_|_2_|_1_|_0_|
 __1_|_0___0___0_|__Op___|_______Offset______|____Rs_____|____Rd_____|Shifted
 __2_|_0___0___0___1___1_|_I,_Op_|___Rn/nn___|____Rs_____|____Rd_____|ADD/SUB
 __3_|_0___0___1_|__Op___|____Rd_____|_____________Offset____________|Immedi.
 __4_|_0___1___0___0___0___0_|______Op_______|____Rs_____|____Rd_____|AluOp
 __5_|_0___1___0___0___0___1_|__Op___|Hd_|Hs_|____Rs_____|____Rd_____|HiReg/BX
 __6_|_0___1___0___0___1_|____Rd_____|_____________Word______________|LDR PC
 __7_|_0___1___0___1_|__Op___|_0_|___Ro______|____Rb_____|____Rd_____|LDR/STR
 __8_|_0___1___0___1_|__Op___|_1_|___Ro______|____Rb_____|____Rd_____|""H/SB/SH
 __9_|_0___1___1_|__Op___|_______Offset______|____Rb_____|____Rd_____|""{B}
 _10_|_1___0___0___0_|Op_|_______Offset______|____Rb_____|____Rd_____|""H
 _11_|_1___0___0___1_|Op_|____Rd_____|_____________Word______________|"" SP
 _12_|_1___0___1___0_|Op_|____Rd_____|_____________Word______________|ADD PC/SP
 _13_|_1___0___1___1___0___0___0___0_|_S_|___________Word____________|ADD SP,nn
 _14_|_1___0___1___1_|Op_|_1___0_|_R_|____________Rlist______________|PUSH/POP
 _17_|_1___0___1___1___1___1___1___0_|___________User_Data___________|BKPT ARM9
 _15_|_1___1___0___0_|Op_|____Rb_____|____________Rlist______________|STM/LDM
 _16_|_1___1___0___1_|_____Cond______|_________Signed_Offset_________|B{cond}
 _U__|_1___1___0___1___1___1___1___0_|_____________var_______________|UNDEF ARM9
 _17_|_1___1___0___1___1___1___1___1_|___________User_Data___________|SWI
 _18_|_1___1___1___0___0_|________________Offset_____________________|B
 _19_|_1___1___1___0___1_|_________________________var___________|_0_|BLXsuf ARM9
 _U__|_1___1___1___0___1_|_________________________var___________|_1_|UNDEF ARM9
 _19_|_1___1___1___1_|_H_|______________Offset_Low/High______________|BL (BLX ARM9)


Further UNDEFS ??? ARM9?
 1011 0001 xxxxxxxx (reserved)
 1011 0x1x xxxxxxxx (reserved)
 1011 10xx xxxxxxxx (reserved)
 1011 1111 xxxxxxxx (reserved)
 1101 1110 xxxxxxxx (free for user)



 THUMB.1: move shifted register

Opcode Format
  Bit    Expl.
  15-13  Must be 000b for 'move shifted register' instructions
  12-11  Opcode
           00b: LSL Rd,Rs,#Offset   (logical/arithmetic shift left)
           01b: LSR Rd,Rs,#Offset   (logical    shift right)
           10b: ASR Rd,Rs,#Offset   (arithmetic shift right)
           11b: Reserved (used for add/subtract instructions)
  10-6   Offset                     (0-31)
  5-3    Rs - Source register       (R0..R7)
  2-0    Rd - Destination register  (R0..R7)
Example: LSL Rd,Rs,#nn ; Rd = Rs << nn ; ARM equivalent: MOVS Rd,Rs,LSL #nn
Zero shift amount is having special meaning (same as for ARM shifts), LSL#0 performs no shift (the the carry flag remains unchanged), LSR/ASR#0 are interpreted as LSR/ASR#32. Attempts to specify LSR/ASR#0 in source code are automatically redirected as LSL#0, and source LSR/ASR#32 is redirected as opcode LSR/ASR#0.
Execution Time: 1S
Flags: Z=zeroflag, N=sign, C=carry (except LSL#0: C=unchanged), V=unchanged.


 THUMB.2: add/subtract

Opcode Format
  Bit    Expl.
  15-11  Must be 00011b for 'add/subtract' instructions
  10-9   Opcode (0-3)
           0: ADD Rd,Rs,Rn   ;add register        Rd=Rs+Rn
           1: SUB Rd,Rs,Rn   ;subtract register   Rd=Rs-Rn
           2: ADD Rd,Rs,#nn  ;add immediate       Rd=Rs+nn
           3: SUB Rd,Rs,#nn  ;subtract immediate  Rd=Rs-nn
         Pseudo/alias opcode with Imm=0:
           2: MOV Rd,Rs      ;move (affects cpsr) Rd=Rs+0
  8-6    For Register Operand:
           Rn - Register Operand (R0..R7)
         For Immediate Operand:
           nn - Immediate Value  (0-7)
  5-3    Rs - Source register       (R0..R7)
  2-0    Rd - Destination register  (R0..R7)
Return: Rd contains result, N,Z,C,V affected (including MOV).
Execution Time: 1S


 THUMB.3: move/compare/add/subtract immediate

Opcode Format
  Bit    Expl.
  15-13  Must be 001b for this type of instructions
  12-11  Opcode
           00b: MOV Rd,#nn      ;move     Rd   = #nn
           01b: CMP Rd,#nn      ;compare  Void = Rd - #nn
           10b: ADD Rd,#nn      ;add      Rd   = Rd + #nn
           11b: SUB Rd,#nn      ;subtract Rd   = Rd - #nn
  10-8   Rd - Destination Register  (R0..R7)
  7-0    nn - Unsigned Immediate    (0-255)
ARM equivalents for MOV/CMP/ADD/SUB are MOVS/CMP/ADDS/SUBS same format.
Execution Time: 1S
Return: Rd contains result (except CMP), N,Z,C,V affected (for MOV only N,Z).


 THUMB.4: ALU operations

Opcode Format
  Bit    Expl.
  15-10  Must be 010000b for this type of instructions
  9-6    Opcode (0-Fh)
           0: AND Rd,Rs     ;AND logical       Rd = Rd AND Rs
           1: EOR Rd,Rs     ;XOR logical       Rd = Rd XOR Rs
           2: LSL Rd,Rs     ;log. shift left   Rd = Rd << (Rs AND 0FFh)
           3: LSR Rd,Rs     ;log. shift right  Rd = Rd >> (Rs AND 0FFh)
           4: ASR Rd,Rs     ;arit shift right  Rd = Rd SRA Rs
           5: ADC Rd,Rs     ;add with carry    Rd = Rd + Rs + Cy
           6: SBC Rd,Rs     ;sub with carry    Rd = Rd - Rs - NOT Cy
           7: ROR Rd,Rs     ;rotate right      Rd = Rd ROR (Rs AND 0FFh)
           8: TST Rd,Rs     ;test            Void = Rd AND (Rs AND 0FFh)
           9: NEG Rd,Rs     ;negate            Rd = 0 - Rs
           A: CMP Rd,Rs     ;compare         Void = Rd - Rs
           B: CMN Rd,Rs     ;neg.compare     Void = Rd + Rs
           C: ORR Rd,Rs     ;OR logical        Rd = Rd OR Rs
           D: MUL Rd,Rs     ;multiply          Rd = Rd * Rs
           E: BIC Rd,Rs     ;bit clear         Rd = Rd AND NOT Rs
           F: MVN Rd,Rs     ;not               Rd = NOT Rs
  5-3    Rs - Source Register       (R0..R7)
  2-0    Rd - Destination Register  (R0..R7)
ARM equivalent for NEG would be RSBS.
Return: Rd contains result (except TST,CMP,CMN),
Affected Flags:
  N,Z,C,V for  ADC,SBC,NEG,CMP,CMN
  N,Z,C   for  LSL,LSR,ASR,ROR (carry flag unchanged if zero shift amount)
  N,Z,C   for  MUL on ARMv4 and below: carry flag destroyed
  N,Z     for  MUL on ARMv5 and above: carry flag unchanged
  N,Z     for  AND,EOR,TST,ORR,BIC,MVN
Execution Time:
  1S      for  AND,EOR,ADC,SBC,TST,NEG,CMP,CMN,ORR,BIC,MVN
  1S+1I   for  LSL,LSR,ASR,ROR
  1S+mI   for  MUL (m=1..4 depending on MSBs of incoming Rd value)


 THUMB.5: Hi register operations/branch exchange

Opcode Format
  Bit    Expl.
  15-10  Must be 010001b for this type of instructions
  9-8    Opcode (0-3)
           0: ADD Rd,Rs   ;add        Rd = Rd+Rs
           1: CMP Rd,Rs   ;compare  Void = Rd-Rs  ;CPSR affected
           2: MOV Rd,Rs   ;move       Rd = Rs
           2: NOP         ;nop        R8 = R8
           3: BX  Rs      ;jump       PC = Rs     ;may switch THUMB/ARM
           3: BLX Rs      ;call       PC = Rs     ;may switch THUMB/ARM (ARM9)
  7      MSBd - Destination Register most significant bit (or BL/BLX flag)
  6      MSBs - Source Register most significant bit
  5-3    Rs - Source Register        (together with MSBs: R0..R15)
  2-0    Rd - Destination Register   (together with MSBd: R0..R15)
Restrictions: For ADD/CMP/MOV, MSBs and/or MSBd must be set, ie. it is not allowed that both are cleared.
When using R15 (PC) as operand, the value will be the address of the instruction plus 4.
For BX, MSBs may be 0 or 1, MSBd must be zero, Rd is not used/zero.
For BLX, MSBs may be 0 or 1, MSBd must be set, Rd is not used/zero.
For BX/BLX, when Bit 0 of the value in Rs is zero:
  Processor will be switched into ARM mode!
  If so, Bit 1 of Rs must be cleared (32bit word aligned).
  Thus, BX PC (switch to ARM) may be issued from word-aligned address
  only, the destination is PC+4 (ie. the following halfword is skipped).
BLX may not use R15. BLX saves the return address as LR=PC+3 (with thumb bit).
Assemblers/Disassemblers should use MOV R8,R8 as NOP (in THUMB mode).
Return: Only CMP affects CPSR condition flags!
Execution Time:
 1S     for ADD/MOV/CMP
 2S+1N  for ADD/MOV with Rd=R15, and for BX


 THUMB.6: load PC-relative

Opcode Format
  Bit    Expl.
  15-11  Must be 01001b for this type of instructions
  N/A    Opcode (fixed)
           LDR Rd,[PC,#nn]      ;load 32bit    Rd = WORD[PC+nn]
  10-8   Rd - Destination Register   (R0..R7)
  7-0    nn - Unsigned offset        (0-1020 in steps of 4)
The value of PC will be interpreted as (($+4) AND NOT 2).
Return: No flags affected, data loaded into Rd.
Execution Time: 1S+1N+1I


 THUMB.7: load/store with register offset

Opcode Format
  Bit    Expl.
  15-12  Must be 0101b for this type of instructions
  11-10  Opcode (0-3)
          0: STR  Rd,[Rb,Ro]   ;store 32bit data  WORD[Rb+Ro] = Rd
          1: STRB Rd,[Rb,Ro]   ;store  8bit data  BYTE[Rb+Ro] = Rd
          2: LDR  Rd,[Rb,Ro]   ;load  32bit data  Rd = WORD[Rb+Ro]
          3: LDRB Rd,[Rb,Ro]   ;load   8bit data  Rd = BYTE[Rb+Ro]
  9      Must be zero (0) for this type of instructions
  8-6    Ro - Offset Register              (R0..R7)
  5-3    Rb - Base Register                (R0..R7)
  2-0    Rd - Source/Destination Register  (R0..R7)
Return: No flags affected, data loaded either into Rd or into memory.
Execution Time: 1S+1N+1I for LDR, or 2N for STR


 THUMB.8: load/store sign-extended byte/halfword

Opcode Format
  Bit    Expl.
  15-12  Must be 0101b for this type of instructions
  11-10  Opcode (0-3)
          0: STRH Rd,[Rb,Ro]  ;store 16bit data          HALFWORD[Rb+Ro] = Rd
          1: LDSB Rd,[Rb,Ro]  ;load sign-extended 8bit   Rd = BYTE[Rb+Ro]
          2: LDRH Rd,[Rb,Ro]  ;load zero-extended 16bit  Rd = HALFWORD[Rb+Ro]
          3: LDSH Rd,[Rb,Ro]  ;load sign-extended 16bit  Rd = HALFWORD[Rb+Ro]
  9      Must be set (1) for this type of instructions
  8-6    Ro - Offset Register              (R0..R7)
  5-3    Rb - Base Register                (R0..R7)
  2-0    Rd - Source/Destination Register  (R0..R7)
Return: No flags affected, data loaded either into Rd or into memory.
Execution Time: 1S+1N+1I for LDR, or 2N for STR


 THUMB.9: load/store with immediate offset

Opcode Format
  Bit    Expl.
  15-13  Must be 011b for this type of instructions
  12-11  Opcode (0-3)
          0: STR  Rd,[Rb,#nn]  ;store 32bit data   WORD[Rb+nn] = Rd
          1: LDR  Rd,[Rb,#nn]  ;load  32bit data   Rd = WORD[Rb+nn]
          2: STRB Rd,[Rb,#nn]  ;store  8bit data   BYTE[Rb+nn] = Rd
          3: LDRB Rd,[Rb,#nn]  ;load   8bit data   Rd = BYTE[Rb+nn]
  10-6   nn - Unsigned Offset              (0-31 for BYTE, 0-124 for WORD)
  5-3    Rb - Base Register                (R0..R7)
  2-0    Rd - Source/Destination Register  (R0..R7)
Return: No flags affected, data loaded either into Rd or into memory.
Execution Time: 1S+1N+1I for LDR, or 2N for STR


 THUMB.10: load/store halfword

Opcode Format
  Bit    Expl.
  15-12  Must be 1000b for this type of instructions
  11     Opcode (0-1)
          0: STRH Rd,[Rb,#nn]  ;store 16bit data   HALFWORD[Rb+nn] = Rd
          1: LDRH Rd,[Rb,#nn]  ;load  16bit data   Rd = HALFWORD[Rb+nn]
  10-6   nn - Unsigned Offset              (0-62, step 2)
  5-3    Rb - Base Register                (R0..R7)
  2-0    Rd - Source/Destination Register  (R0..R7)
Return: No flags affected, data loaded either into Rd or into memory.
Execution Time: 1S+1N+1I for LDR, or 2N for STR


 THUMB.11: load/store SP-relative

Opcode Format
  Bit    Expl.
  15-12  Must be 1001b for this type of instructions
  11     Opcode (0-1)
          0: STR  Rd,[SP,#nn]  ;store 32bit data   WORD[SP+nn] = Rd
          1: LDR  Rd,[SP,#nn]  ;load  32bit data   Rd = WORD[SP+nn]
  10-8   Rd - Source/Destination Register  (R0..R7)
  7-0    nn - Unsigned Offset              (0-1020, step 4)
Return: No flags affected, data loaded either into Rd or into memory.
Execution Time: 1S+1N+1I for LDR, or 2N for STR


 THUMB.12: get relative address

Opcode Format
  Bit    Expl.
  15-12  Must be 1010b for this type of instructions
  11     Opcode/Source Register (0-1)
          0: ADD  Rd,PC,#nn    ;Rd = (($+4) AND NOT 2) + nn
          1: ADD  Rd,SP,#nn    ;Rd = SP + nn
  10-8   Rd - Destination Register         (R0..R7)
  7-0    nn - Unsigned Offset              (0-1020, step 4)
Return: No flags affected, result in Rd.
Execution Time: 1S


 THUMB.13: add offset to stack pointer

Opcode Format
  Bit    Expl.
  15-8   Must be 10110000b for this type of instructions
  7      Opcode/Sign
          0: ADD  SP,#nn       ;SP = SP + nn
          1: ADD  SP,#-nn      ;SP = SP - nn
  6-0    nn - Unsigned Offset    (0-508, step 4)
Return: No flags affected, SP adjusted.
Execution Time: 1S


 THUMB.14: push/pop registers

Opcode Format
  Bit    Expl.
  15-12  Must be 1011b for this type of instructions
  11     Opcode (0-1)
          0: PUSH {Rlist}{LR}   ;store in memory, decrements SP (R13)
          1: POP  {Rlist}{PC}   ;load from memory, increments SP (R13)
  10-9   Must be 10b for this type of instructions
  8      PC/LR Bit (0-1)
          0: No
          1: PUSH LR (R14), or POP PC (R15)
  7-0    Rlist - List of Registers (R7..R0)
In THUMB mode stack is always meant to be 'full descending', ie. PUSH is equivalent to 'STMFD/STMDB' and POP to 'LDMFD/LDMIA' in ARM mode.

Examples:
 PUSH {R0-R3}     ;push R0,R1,R2,R3
 PUSH {R0,R2,LR}  ;push R0,R2,LR
 POP  {R4,R7}     ;pop R4,R7
 POP  {R2-R4,PC}  ;pop R2,R3,R4,PC
Note: When calling to a sub-routine, the return address is stored in LR register, when calling further sub-routines, PUSH {LR} must be used to save higher return address on stack. If so, POP {PC} can be later used to return from the sub-routine.
POP {PC} ignores the least significant bit of the return address (processor remains in thumb state even if bit0 was cleared), when intending to return with optional mode switch, use a POP/BX combination (eg. POP {R3} / BX R3).
ARM9: POP {PC} copies the LSB to thumb bit (switches to ARM if bit0=0).
Return: No flags affected, SP adjusted, registers loaded/stored.
Execution Time: nS+1N+1I (POP), (n+1)S+2N+1I (POP PC), or (n-1)S+2N (PUSH).


 THUMB.15: multiple load/store

Opcode Format
  Bit    Expl.
  15-12  Must be 1100b for this type of instructions
  11     Opcode (0-1)
          0: STMIA Rb!,{Rlist}   ;store in memory, increments Rb
          1: LDMIA Rb!,{Rlist}   ;load from memory, increments Rb
  10-8   Rb - Base register (modified) (R0-R7)
  7-0    Rlist - List of Registers     (R7..R0)
Both STM and LDM are incrementing the Base Register.
The lowest register in the list (ie. R0, if it's in the list) is stored/loaded at the lowest memory address.
Examples:
 STMIA R7!,{R0-R2}  ;store R0,R1,R2
 LDMIA R0!,{R1,R5}  ;store R1,R5
Return: No flags affected, Rb adjusted, registers loaded/stored.
Execution Time: nS+1N+1I for LDM, or (n-1)S+2N for STM.


 THUMB.16: conditional branch

Opcode Format
  Bit    Expl.
  15-12  Must be 1101b for this type of instructions
  11-8   Opcode/Condition (0-Fh)
          0: BEQ label   ;Z=1         ;equal (zero)
          1: BNE label   ;Z=0         ;not equal (nonzero)
          2: BCS label   ;C=1         ;unsigned higher or same (carry set)
          3: BCC label   ;C=0         ;unsigned lower (carry cleared)
          4: BMI label   ;N=1         ;negative (minus)
          5: BPL label   ;N=0         ;positive or zero (plus)
          6: BVS label   ;V=1         ;overflow (V set)
          7: BVC label   ;V=0         ;no overflowplus (V cleared)
          8: BHI label   ;C=1 and Z=0 ;unsigned higher
          9: BLS label   ;C=0 or Z=1  ;unsigned lower or same
          A: BGE label   ;N=V         ;greater or equal
          B: BLT label   ;N<>V        ;less than
          C: BGT label   ;Z=0 and N=V ;greater than
          D: BLE label   ;Z=1 or N<>V ;less or equal
          E: Undefined, should not be used
          F: Reserved for SWI instruction (see SWI opcode)
  7-0    Signed Offset, step 2 ($+4-256..$+4+254)
Destination address must by halfword aligned (ie. bit 0 cleared)
Return: No flags affected, PC adjusted if condition true
Execution Time:
  2S+1N   if condition true (jump executed)
  1S      if condition false


 THUMB.17: software interrupt and breakpoint

Opcode Format
  Bit    Expl.
  15-8   Opcode
          11011111b: SWI nn   ;software interrupt
          10111110b: BKPT nn  ;software breakpoint (ARMv5 and up)
  7-0    nn - Comment Immediate    (0-255)
SWI supposed for calls to the operating system - Enter Supervisor mode (SVC) in ARM state. BKPT intended for debugging - enters Abort mode in ARM state via Prefetch Abort vector.

Execution SWI/BKPT:
  R14_svc=PC+2     R14_abt=PC+4   ;save return address
  SPSR_svc=CPSR    SPSR_abt=CPSR  ;save CPSR flags
  CPSR=<changed>   CPSR=<changed> ;Enter svc/abt, ARM state, IRQs disabled
  PC=VVVV0008h     PC=VVVV000Ch   ;jump to SWI/PrefetchAbort vector address
Execution Time: 2S+1N

Interpreting the Comment Field:
The immediate parameter is ignored by the processor, the user interrupt handler may read-out this number by examining the lower 8bit of the 16bit opcode opcode at [R14_svc-2]. In case that your program executes SWI's from inside of ARM mode also: Your SWI handler must then examine the T Bit SPSR_svc in order to determine whether it's been a ARM SWI - if so, examining the lower 24bit of the 32bit opcode opcode at [R14_svc-4].

For Returning from SWI use this instruction:
  MOVS PC,R14
That instructions does both restoring PC and CPSR, ie. PC=R14_svc, and CPSR=SPRS_svc. In this case (as called from THUMB mode), this does also include restoring THUMB mode.

Nesting SWIs:
SPSR_svc and R14_svc should be saved on stack before either invoking nested SWIs, or (if the IRQ handler uses SWIs) before enabling IRQs.


 THUMB.18: unconditional branch

Opcode Format
  Bit    Expl.
  15-11  Must be 11100b for this type of instructions
  N/A    Opcode (fixed)
          B label   ;branch (jump)
  10-0   Signed Offset, step 2 ($+4-2048..$+4+2046)
Return: No flags affected, PC adjusted.
Execution Time: 2S+1N


 THUMB.19: long branch with link

Opcode Format
This may be used to call (or jump) to a subroutine, return address is saved in LR (R14).
Unlike all other THUMB mode instructions, this instruction occupies 32bit of memory which are split into two 16bit THUMB opcodes.

First Instruction - LR = PC+4+(nn SHL 12)
  Bit    Expl.
  15-11  Must be 11110b for BL/BLX type of instructions
  10-0   nn - Upper 11 bits of Target Address
Second Instruction - PC = LR + (nn SHL 1), and LR = PC+2 OR 1 (and BLX: T=0)
  Bit    Expl.
  15-11  Opcode
          11111b: BL label   ;branch long with link
          11101b: BLX label  ;branch long with link switch to ARM mode (ARM9)
  10-0   nn - Lower 11 bits of Target Address (BLX: Bit0 Must be zero)

The destination address range is (PC+4)-400000h..+3FFFFEh, ie. PC+/-4M.
Target must be halfword-aligned. As Bit 0 in LR is set, it may be used to return by a BX LR instruction (keeping CPU in THUMB mode).
Return: No flags affected, PC adjusted, return address in LR.
Execution Time: 3S+1N (first opcode 1S, second opcode 2S+1N).

Note
Exceptions may or may not occur between first and second opcode, this is "implementation defined" ???


 ARM Instruction Set

When operating in ARM state, full 32bit opcodes are used.

Summaries
ARM Instruction Summary
ARM Condition Field

Jumps and Calls
ARM.3: Branch and Exchange (BX, BLX)
ARM.4: Branch and Branch with Link (B, BL, BLX)
(Also, most various ALU, LDR, LDM opcodes can change PC.)

Register Operations
ARM.5: Data Processing
ARM.6: PSR Transfer (MRS, MSR)
ARM.7: Multiply and Multiply-Accumulate (MUL,MLA)

Memory Addressing Operations
ARM.9: Single Data Transfer (LDR, STR, PLD)
ARM.10: Halfword, Doubleword, and Signed Data Transfer
ARM.11: Block Data Transfer (LDM,STM)
ARM.12: Single Data Swap (SWP)

Exception Calls and Coprocessor
ARM.13: Software Interrupt (SWI,BKPT)
ARM.14: Coprocessor Data Operations (CDP)
ARM.15: Coprocessor Data Transfers (LDC,STC)
ARM.16: Coprocessor Register Transfers (MRC, MCR)
ARM.X: Coprocessor Double-Register Transfer (MCRR,MRRC)
ARM.17: Undefined Instruction

ARM.X: Count Leading Zeros
ARM.X: QADD/QSUB

ARM 26bit Memory Interface

Note:
Switching between ARM and THUMB state can be done by using the Branch and Exchange (BX) instruction.


 ARM Instruction Summary

Modification of CPSR flags is optional for all {S} instructions.

Logical Operations
  Instruction             Cycles   Flags Format Expl.
  MOV{cond}{S} Rd,Op2      1S+x+y   NZc- 5   Rd = Op2
  MVN{cond}{S} Rd,Op2      1S+x+y   NZc- 5   Rd = NOT Op2
  AND{cond}{S} Rd,Rn,Op2   1S+x+y   NZc- 5   Rd = Rn AND Op2
  TST{cond}{P}    Rn,Op2   1S+x     NZc- 5 Void = Rn AND Op2
  EOR{cond}{S} Rd,Rn,Op2   1S+x+y   NZc- 5   Rd = Rn XOR Op2
  TEQ{cond}{P}    Rn,Op2   1S+x     NZc- 5 Void = Rn XOR Op2
  ORR{cond}{S} Rd,Rn,Op2   1S+x+y   NZc- 5   Rd = Rn OR Op2
  BIC{cond}{S} Rd,Rn,Op2   1S+x+y   NZc- 5   Rd = Rn AND NOT Op2
Add x=1I cycles if Op2 shifted-by-register. Add y=1S+1N cycles if Rd=R15.
Carry flag affected only if Op2 contains a non-zero shift amount.

Arithmetic Operations
  Instruction             Cycles  Flags Format Expl.
  ADD{cond}{S} Rd,Rn,Op2   1S+x+y   NZCV 5   Rd = Rn+Op2
  ADC{cond}{S} Rd,Rn,Op2   1S+x+y   NZCV 5   Rd = Rn+Op2+Cy
  SUB{cond}{S} Rd,Rn,Op2   1S+x+y   NZCV 5   Rd = Rn-Op2
  SBC{cond}{S} Rd,Rn,Op2   1S+x+y   NZCV 5   Rd = Rn-Op2+Cy-1
  RSB{cond}{S} Rd,Rn,Op2   1S+x+y   NZCV 5   Rd = Op2-Rn
  RSC{cond}{S} Rd,Rn,Op2   1S+x+y   NZCV 5   Rd = Op2-Rn+Cy-1
  CMP{cond}{P}    Rn,Op2   1S+x     NZCV 5 Void = Rn-Op2
  CMN{cond}{P}    Rn,Op2   1S+x     NZCV 5 Void = Rn+Op2
Add x=1I cycles if Op2 shifted-by-register. Add y=1S+1N cycles if Rd=R15.

Multiply
  Instruction                     Cycles  Flags Format Expl.
  MUL{cond}{S} Rd,Rm,Rs            1S+mI     NZx- 7  Rd = Rm*Rs
  MLA{cond}{S} Rd,Rm,Rs,Rn         1S+mI+1I  NZx- 7  Rd = Rm*Rs+Rn
  UMULL{cond}{S} RdLo,RdHi,Rm,Rs   1S+mI+1I  NZx- 7  RdHiLo = Rm*Rs
  UMLAL{cond}{S} RdLo,RdHi,Rm,Rs   1S+mI+2I  NZx- 7  RdHiLo = Rm*Rs+RdHiLo
  SMULL{cond}{S} RdLo,RdHi,Rm,Rs   1S+mI+1I  NZx- 7  RdHiLo = Rm*Rs
  SMLAL{cond}{S} RdLo,RdHi,Rm,Rs   1S+mI+2I  NZx- 7  RdHiLo = Rm*Rs+RdHiLo
  SMLAxy{cond} Rd,Rm,Rs,Rn       ---q 7  Rd=HalfRm*HalfRs+Rn ARMv5TE(xP)
  SMLAWy{cond} Rd,Rm,Rs,Rn       ---q 7  Rd=(Rm*HalfRs)/10000h+Rn ARMv5TE(xP)
  SMULWy{cond} Rd,Rm,Rs          ---- 7  Rd=(Rm*HalfRs)/10000h    ARMv5TE(xP)
  SMLALxy{cond} RdLo,RdHi,Rm,Rs  ---- 7  RdHiLo=RdHiLo+HalfRm*HalfRs ARMv5TE(xP)
  SMULxy{cond} Rd,Rm,Rs          ---- 7  Rd=HalfRm*HalfRs ARMv5TE(xP)

Memory Load/Store
  Instruction                     Cycles       Flags Format Expl.
  LDR{cond}{B}{T} Rd,<Address>     1S+1N+1I +y   ---- 9  Rd=[Rn+/-<offset>]
  LDR{cond}H      Rd,<Address>     1S+1N+1I +y   ---- 10 Load Unsigned halfword
  LDR{cond}D      Rd,<Address>                   ---- 10 Load Dword ARMv5TE
  LDR{cond}SB     Rd,<Address>     1S+1N+1I +y   ---- 10 Load Signed byte
  LDR{cond}SH     Rd,<Address>     1S+1N+1I +y   ---- 10 Load Signed halfword
  LDM{cond}{amod} Rn{!},<Rlist>{^} nS+1N+1I +y   ---- 11 Load Multiple
  STR{cond}{B}{T} Rd,<Address>     2N            ---- 9  [Rn+/-<offset>]=Rd
  STR{cond}H      Rd,<Address>     2N            ---- 10 Store halfword
  STR{cond}D      Rd,<Address>                   ---- 10 Store Dword ARMv5TE
  STM{cond}{amod} Rn{!},<Rlist>{^} (n-1)S+2N     ---- 11 Store Multiple
  SWP{cond}{B}    Rd,Rm,[Rn]       1S+2N+1I      ---- 12 Rd=[Rn], [Rn]=Rm
  PLD             <Address>        1S            ---- 9  Prepare Cache ARMv5TE
For LDR/LDM, add y=1S+1N if Rd=R15, or if R15 in Rlist.

Jumps, Calls, CPSR Mode, and others
  Instruction              Cycles  Flags Format Expl.
  B{cond}   label           2S+1N    ---- 4   PC=$+8+/-32M
  BL{cond}  label           2S+1N    ---- 4   PC=$+8+/-32M, LR=$+4
  BX{cond}  Rn              2S+1N    ---- 3   PC=Rn, T=Rn.0 (THUMB/ARM)
  BLX{cond} Rn              2S+1N    ---- 3   PC=Rn, T=Rn.0, LR=PC+4, ARM9
  BLX       label           2S+1N    ---- 3   PC=PC+$+/-32M, LR=$+4, T=1, ARM9
  MRS{cond} Rd,Psr          1S       ---- 6   Rd=Psr
  MSR{cond} Psr{_field},Op  1S      (psr) 6   Psr[field]=Op
  SWI{cond} Imm24bit        2S+1N    ---- 13  PC=8, ARM Svc mode, LR=$+4
  BKPT      Imm16bit        ???      ---- ??? PC=C, ARM Abt mode, LR=$+4 ARM9
  The Undefined Instruction 2S+1I+1N ---- 17  PC=4, ARM Und mode, LR=$+4
  cond=false                1S       ---- ..  Any opcode with condition=false
  NOP                       1S       ---- 5   R0=R0

  CLZ{cond} Rd,Rm           ???      ---- ??? Count Leading Zeros ARMv5
  QADD{cond} Rd,Rm,Rn            ---q   Rd=Rm+Rn       ARMv5TE(xP)
  QSUB{cond} Rd,Rm,Rn            ---q   Rd=Rm-Rn       ARMv5TE(xP)
  QDADD{cond} Rd,Rm,Rn           ---q   Rd=Rm+Rn*2     ARMv5TE(xP)
  QDSUB{cond} Rd,Rm,Rn           ---q   Rd=Rm-Rn*2     ARMv5TE(xP)

Coprocessor Functions (if any)
  Instruction                         Cycles  Flags Format Expl.
  CDP{cond} Pn,<cpopc>,Cd,Cn,Cm{,<cp>} 1S+bI   ---- 14 Coprocessor specific
  STC{cond}{L} Pn,Cd,<Address>         (n-1)S+2N+bI 15 [address] = CRd
  LDC{cond}{L} Pn,Cd,<Address>         (n-1)S+2N+bI 15 CRd = [address]
  MCR{cond} Pn,<cpopc>,Rd,Cn,Cm{,<cp>} 1S+bI+1C     16 CRn = Rn {<op> CRm}
  MRC{cond} Pn,<cpopc>,Rd,Cn,Cm{,<cp>} 1S+(b+1)I+1C 16 Rn = CRn {<op> CRm}
  CDP2,STC2,LDC2,MCR2,MRC2 - ARMv5 Extensions similar above, without {cond}
  MCRR{cond} Pn,<cpopc>,Rd,Rn,Cm  ;write Rd,Rn to coproc ARMv5TE
  MRRC{cond} Pn,<cpopc>,Rd,Rn,Cm  ;read Rd,Rn from coproc ARMv5TE

Note that no sections 1-2 exist, that is because the sections numbers comply with chapter numbers of the official ARM docs, which described ARM opcodes in chapter 3-17.

ARM Binary Opcode Format

  |..3 ..................2 ..................1 ..................0|
  |1_0_9_8_7_6_5_4_3_2_1_0_9_8_7_6_5_4_3_2_1_0_9_8_7_6_5_4_3_2_1_0|
  |_Cond__|0_0_0|___Op__|S|__Rn___|__Rd___|__Shift__|Typ|0|__Rm___| DataProc
  |_Cond__|0_0_0|___Op__|S|__Rn___|__Rd___|__Rs___|0|Typ|1|__Rm___| DataProc
  |_Cond__|0_0_1|___Op__|S|__Rn___|__Rd___|_Shift_|___Immediate___| DataProc
  |_Cond__|0_0_1_1_0|P|1|0|_Field_|__Rd___|_Shift_|___Immediate___| PSR Imm
  |_Cond__|0_0_0_1_0|P|L|0|_Field_|__Rd___|0_0_0_0|0_0_0_0|__Rm___| PSR Reg
  |_Cond__|0_0_0_1_0_0_1_0_1_1_1_1_1_1_1_1_1_1_1_1|0_0|L|1|__Rn___| BX,BLX
  |1_1_1_0|0_0_0_1_0_0_1_0|_____immediate_________|0_1_1_1|_immed_| BKPT ARM9
  |_Cond__|0_0_0_1_0_1_1_0_1_1_1_1|__Rd___|1_1_1_1|0_0_0_1|__Rm___| CLZ ARM9
  |_Cond__|0_0_0_1_0|Op_|0|__Rn___|__Rd___|0_0_0_0|0_1_0_1|__Rm___| QALU ARM9
  |_Cond__|0_0_0_0_0_0|A|S|__Rd___|__Rn___|__Rs___|1_0_0_1|__Rm___| Multiply
  |_Cond__|0_0_0_0_1|U|A|S|_RdHi__|_RdLo__|__Rs___|1_0_0_1|__Rm___| MulLong
  |_Cond__|0_0_0_1_0|Op_|0|Rd/RdHi|Rn/RdLo|__Rs___|1|y|x|0|__Rm___| MulHalf
  |_Cond__|0_0_0_1_0|B|0_0|__Rn___|__Rd___|0_0_0_0|1_0_0_1|__Rm___| TransSwp12
  |_Cond__|0_0_0|P|U|0|W|L|__Rn___|__Rd___|0_0_0_0|1|S|H|1|__Rm___| TransReg10
  |_Cond__|0_0_0|P|U|1|W|L|__Rn___|__Rd___|OffsetH|1|S|H|1|OffsetL| TransImm10
  |_Cond__|0_1_0|P|U|B|W|L|__Rn___|__Rd___|_________Offset________| TransImm9
  |_Cond__|0_1_1|P|U|B|W|L|__Rn___|__Rd___|__Shift__|Typ|0|__Rm___| TransReg9
  |_Cond__|0_1_1|________________xxx____________________|1|__xxx__| Undefined
  |_Cond__|1_0_0|P|U|S|W|L|__Rn___|__________Register_List________| BlockTrans
  |_Cond__|1_0_1|L|___________________Offset______________________| B,BL,BLX
  |_Cond__|1_1_0|P|U|N|W|L|__Rn___|__CRd__|__CP#__|____Offset_____| CoDataTrans
  |_Cond__|1_1_0_0_0_1_0|L|__Rn___|__Rd___|__CP#__|_CPopc_|__CRm__| CoRR ARM9
  |_Cond__|1_1_1_0|_CPopc_|__CRn__|__CRd__|__CP#__|_CP__|0|__CRm__| CoDataOp
  |_Cond__|1_1_1_0|CPopc|L|__CRn__|__Rd___|__CP#__|_CP__|1|__CRm__| CoRegTrans
  |_Cond__|1_1_1_1|_____________Ignored_by_Processor______________| SWI


 ARM Condition Field

In ARM mode, all instructions can be conditionally executed depending on the state of the CPSR flags (C,N,Z,V). The respective suffixes {cond} must be appended to the mnemonics. For example: BEQ = Branch if Equal, MOVMI = Move if Signed.

  Code Suffix Flags         Meaning
  0:   EQ     Z=1           equal (zero)
  1:   NE     Z=0           not equal (nonzero)
  2:   CS     C=1           unsigned higher or same (carry set)
  3:   CC     C=0           unsigned lower (carry cleared)
  4:   MI     N=1           negative (minus)
  5:   PL     N=0           positive or zero (plus)
  6:   VS     V=1           overflow (V set)
  7:   VC     V=0           no overflowplus (V cleared)
  8:   HI     C=1 and Z=0   unsigned higher
  9:   LS     C=0 or Z=1    unsigned lower or same
  A:   GE     N=V           greater or equal
  B:   LT     N<>V          less than
  C:   GT     Z=0 and N=V   greater than
  D:   LE     Z=1 or N<>V   less or equal
  E:   AL     -             always
  F:   NV     -             never (ARMv1,v2 only) (Reserved ARMv3 and up)

To define a non-conditional instruction which is always to be executed (regardless of any flags), the AL suffix may be used - that is the same as if no suffix is specified. For example, MOVAL would be usually abbreviated to MOV.

ARMv5 and up includes a few additional opcodes without condition field and which cannot be made conditional, these opcodes are: BKPT, PLD, CDP2, LDC2, MCR2, MRC2, STC2, and BLX_imm (however BLX_reg can be conditional).

Execution Time: If condition=false: 1S cycle.
Otherwise as specified for the respective opcode.


 ARM.3: Branch and Exchange (BX, BLX)

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-8   Must be "0001.0010.1111.1111.1111" for this instruction
  7-4    Opcode
          0001b: BX{cond}  Rn    ;PC=Rn, T=Rn.0  (ARMv4T and ARMv5 and up)
          0011b: BLX{cond} Rn    ;PC=Rn, T=Rn.0, LR=PC+4    (ARMv5 and up)
  3-0    Rn - Operand Register  (R0-R14)
Switching to THUMB Mode: Set Bit 0 of the value in Rn to 1, program continues then at Rn-1 in THUMB mode.
Results in undefined behaviour if using R15 (PC+8 itself) as operand.
Execution Time: 2S + 1N
Return: No flags affected.


 ARM.4: Branch and Branch with Link (B, BL, BLX)

Opcode Format
Branch (B) is supposed to jump to a subroutine. Branch with Link is meant to be used to call to a subroutine, return address is then saved in R14.
  Bit    Expl.
  31-28  Condition (must be 1111b for BLX)
  27-25  Must be "101" for this instruction
  24     Opcode (0-1) (or Halfword Offset for BLX)
          0: B{cond} label    ;branch            PC=PC+8+nn*4
          1: BL{cond} label   ;branch/link       PC=PC+8+nn*4, LR=PC+4
          H: BLX label ;ARM9  ;branch/link/thumb PC=PC+8+nn*4+H*2, LR=PC+4, T=1
  23-0   nn - Signed Offset, step 4      (-32M..+32M in steps of 4)
Branch with Link can be used to 'call' to a sub-routine, which may then 'return' by MOV PC,R14 for example.
Execution Time: 2S + 1N
Return: No flags affected.


 ARM.5: Data Processing

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-26  Must be 00b for this instruction
  25     I - Immediate 2nd Operand Flag (0=Register, 1=Immediate)
  24-21  Opcode (0-Fh)               ;*=Arithmetic, otherwise Logical
           0: AND{cond}{S} Rd,Rn,Op2    ;AND logical       Rd = Rn AND Op2
           1: EOR{cond}{S} Rd,Rn,Op2    ;XOR logical       Rd = Rn XOR Op2
           2: SUB{cond}{S} Rd,Rn,Op2 ;* ;subtract          Rd = Rn-Op2
           3: RSB{cond}{S} Rd,Rn,Op2 ;* ;subtract reversed Rd = Op2-Rn
           4: ADD{cond}{S} Rd,Rn,Op2 ;* ;add               Rd = Rn+Op2
           5: ADC{cond}{S} Rd,Rn,Op2 ;* ;add with carry    Rd = Rn+Op2+Cy
           6: SBC{cond}{S} Rd,Rn,Op2 ;* ;sub with carry    Rd = Rn-Op2+Cy-1
           7: RSC{cond}{S} Rd,Rn,Op2 ;* ;sub cy. reversed  Rd = Op2-Rn+Cy-1
           8: TST{cond}{P}    Rn,Op2    ;test            Void = Rn AND Op2
           9: TEQ{cond}{P}    Rn,Op2    ;test exclusive  Void = Rn XOR Op2
           A: CMP{cond}{P}    Rn,Op2 ;* ;compare         Void = Rn-Op2
           B: CMN{cond}{P}    Rn,Op2 ;* ;compare neg.    Void = Rn+Op2
           C: ORR{cond}{S} Rd,Rn,Op2    ;OR logical        Rd = Rn OR Op2
           D: MOV{cond}{S} Rd,Op2       ;move              Rd = Op2
           E: BIC{cond}{S} Rd,Rn,Op2    ;bit clear         Rd = Rn AND NOT Op2
           F: MVN{cond}{S} Rd,Op2       ;not               Rd = NOT Op2
  20     S - Set Condition Codes (0=No, 1=Yes) (Must be 1 for opcode 8-B)
  19-16  Rn - 1st Operand Register (R0..R15) (including PC=R15)
              Must be 0000b for MOV/MVN.
  15-12  Rd - Destination Register (R0..R15) (including PC=R15)
              Must be 0000b {or 1111b) for CMP/CMN/TST/TEQ{P}.
  When above Bit 25 I=0 (Register as 2nd Operand)
    When below Bit 4 R=0 - Shift by Immediate
      11-7   Is - Shift amount   (1-31, 0=Special/See below)
    When below Bit 4 R=1 - Shift by Register
      11-8   Rs - Shift register (R0-R14) - only lower 8bit 0-255 used
      7      Reserved, must be zero  (otherwise multiply or undefined opcode)
    6-5    Shift Type (0=LSL, 1=LSR, 2=ASR, 3=ROR)
    4      R - Shift by Register Flag (0=Immediate, 1=Register)
    3-0    Rm - 2nd Operand Register (R0..R15) (including PC=R15)
  When above Bit 25 I=1 (Immediate as 2nd Operand)
    11-8   Is - ROR-Shift applied to nn (0-30, in steps of 2)
    7-0    nn - 2nd Operand Unsigned 8bit Immediate

Second Operand (Op2)
This may be a shifted register, or a shifted immediate. See Bit 25 and 11-0.
Unshifted Register: Specify Op2 as "Rm", assembler converts to "Rm,LSL#0".
Shifted Register: Specify as "Rm,SSS#Is" or "Rm,SSS Rs" (SSS=LSL/LSR/ASR/ROR).
Immediate: Specify as 32bit value, for example: "#000NN000h", assembler should automatically convert into "#0NNh,ROR#0ssh" as far as possible (ie. as far as a section of not more than 8bits of the immediate is non-zero).

Zero Shift Amount (Shift Register by Immediate, with Immediate=0)
LSL#0: No shift performed, ie. directly Op2=Rm, the C flag is NOT affected.
LSR#0: Interpreted as LSR#32, ie. Op2 becomes zero, C becomes Bit 31 of Rm.
ASR#0: Interpreted as ASR#32, ie. Op2 and C are filled by Bit 31 of Rm.
ROR#0: Interpreted as RRX#1 (RCR), like ROR#1, but Op2 Bit 31 set to old C.
In source code, LSR#32, ASR#32, and RRX#1 should be specified as such - attempts to specify LSR#0, ASR#0, or ROR#0 will be internally converted to LSL#0 by the assembler.

Using R15 (PC)
When using R15 as Destination (Rd), note below CPSR description and Execution time description.
When using R15 as operand (Rm or Rn), the returned value depends on the instruction: PC+12 if I=0,R=1 (shift by register), otherwise PC+8 (shift by immediate).

Returned CPSR Flags
If S=1, Rd<>R15, logical operations (AND,EOR,TST,TEQ,ORR,MOV,BIC,MVN):
  V=not affected
  C=carryflag of shift operation (not affected if LSL#0 or Rs=00h)
  Z=zeroflag of result
  N=signflag of result (result bit 31)
If S=1, Rd<>R15, arithmetic operations (SUB,RSB,ADD,ADC,SBC,RSC,CMP,CMN):
  V=overflowflag of result
  C=carryflag of result
  Z=zeroflag of result
  N=signflag of result (result bit 31)
IF S=1, with unused Rd bits=1111b, {P} opcodes (CMPP/CMNP/TSTP/TEQP):
  R15=result  ;modify PSR bits in R15, ARMv2 and below only.
  In user mode only N,Z,C,V bits of R15 can be changed.
  In other modes additionally I,F,M1,M0 can be changed.
  The PC bits in R15 are left unchanged in all modes.
If S=1, Rd=R15; should not be used in user mode:
  CPSR = SPSR_<current mode>
  PC = result
  For example: MOVS PC,R14  ;return from SWI (PC=R14_svc, CPSR=SPSR_svc).
If S=0: Flags are not affected (not allowed for CMP,CMN,TEQ,TST).

The instruction "MOV R0,R0" is used as "NOP" opcode in 32bit ARM state.
Execution Time: (1+p)S+rI+pN. Whereas r=1 if I=0 and R=1 (ie. shift by register); otherwise r=0. And p=1 if Rd=R15; otherwise p=0.


 ARM.6: PSR Transfer (MRS, MSR)

Opcode Format
These instructions occupy an unused area (TEQ,TST,CMP,CMN with S=0) of Data Processing opcodes (ARM.5).
  Bit    Expl.
  31-28  Condition
  27-26  Must be 00b for this instruction
  25     I - Immediate Operand Flag  (0=Register, 1=Immediate) (Zero for MRS)
  24-23  Must be 10b for this instruction
  22     Psr - Source/Destination PSR  (0=CPSR, 1=SPSR_<current mode>)
  21     Opcode
           0: MRS{cond} Rd,Psr          ;Rd = Psr
           1: MSR{cond} Psr{_field},Op  ;Psr[field] = Op
  20     Must be 0b for this instruction (otherwise TST,TEQ,CMP,CMN)
  For MRS:
    19-16   Must be 1111b for this instruction (otherwise SWP)
    15-12   Rd - Destination Register  (R0-R14)
    11-0    Not used, must be zero.
  For MSR:
    19      f  write to flags field     Bit 31-24 (aka _flg)
    18      s  write to status field    Bit 23-16 (reserved, don't change)
    17      x  write to extension field Bit 15-8  (reserved, don't change)
    16      c  write to control field   Bit 7-0   (aka _ctl)
    15-12   Not used, must be 1111b.
  For MSR Psr,Rm (I=0)
    11-4    Not used, must be zero. (otherwise BX)
    3-0     Rm - Source Register <op>  (R0-R14)
  For MSR Psr,Imm (I=1)
    11-8    Shift applied to Imm   (ROR in steps of two 0-30)
    7-0     Imm - Unsigned 8bit Immediate
    In source code, a 32bit immediate should be specified as operand.
    The assembler should then convert that into a shifted 8bit value.
MSR/MRS and CPSR/SPSR supported by ARMv3 and up.
ARMv2 and below contained PSR flags in R15, accessed by CMP/CMN/TST/TEQ{P}.
The field mask bits specify which bits of the destination Psr are write-able (or write-protected), one or more of these bits should be set, for example, CPSR_fsxc (aka CPSR aka CPSR_all) unlocks all bits (see below user mode restriction though).
Restrictions:
In non-privileged mode (user mode): only condition code bits of CPSR can be changed, control bits can't.
Only the SPSR of the current mode can be accessed; In User and System modes no SPSR exists.
The T-bit may not be changed; for THUMB/ARM switching use BX instruction.
Unused Bits in CPSR are reserved for future use and should never be changed (except for unused bits in the flags field).
Execution Time: 1S.

Note: The A22i assembler recognizes MOV as alias for both MSR and MRS because it is practically not possible to remember whether MSR or MRS was the load or store opcode, and/or whether it does load to or from the Psr register.


 ARM.7: Multiply and Multiply-Accumulate (MUL,MLA)

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-25  Must be 000b for this instruction
  24-21  Opcode
          0000b: MUL{cond}{S}   Rd,Rm,Rs        ;multiply   Rd = Rm*Rs
          0001b: MLA{cond}{S}   Rd,Rm,Rs,Rn     ;mul.& accumulate Rd = Rm*Rs+Rn
          0100b: UMULL{cond}{S} RdLo,RdHi,Rm,Rs ;multiply   RdHiLo=Rm*Rs
          0101b: UMLAL{cond}{S} RdLo,RdHi,Rm,Rs ;mul.& acc. RdHiLo=Rm*Rs+RdHiLo
          0110b: SMULL{cond}{S} RdLo,RdHi,Rm,Rs ;sign.mul.  RdHiLo=Rm*Rs
          0111b: SMLAL{cond}{S} RdLo,RdHi,Rm,Rs ;sign.m&a.  RdHiLo=Rm*Rs+RdHiLo
          1000b: SMLAxy{cond}   Rd,Rm,Rs,Rn     ;Rd=HalfRm*HalfRs+Rn
          1001b: SMLAWy{cond}   Rd,Rm,Rs,Rn     ;Rd=(Rm*HalfRs)/10000h+Rn
          1001b: SMULWy{cond}   Rd,Rm,Rs        ;Rd=(Rm*HalfRs)/10000h
          1010b: SMLALxy{cond}  RdLo,RdHi,Rm,Rs ;RdHiLo=RdHiLo+HalfRm*HalfRs
          1011b: SMULxy{cond}   Rd,Rm,Rs        ;Rd=HalfRm*HalfRs
  20     S - Set Condition Codes (0=No, 1=Yes) (Must be 0 for Halfword mul)
  19-16  Rd (or RdHi) - Destination Register (R0-R14)
  15-12  Rn (or RdLo) - Accumulate Register  (R0-R14) (Set to 0000b if unused)
  11-8   Rs - Operand Register               (R0-R14)
  For Non-Halfword Multiplies
    7-4  Must be 1001b for these instructions
  For Halfword Multiplies
    7    Must be 1 for these instructions
    6    y - Rs Top/Bottom flag (0=B=Lower 16bit, 1=T=Upper 16bit)
    5    x - Rm Top/Bottom flag (as above), or 0 for SMLAW, or 1 for SMULW
    4    Must be 0 for these instructions
  3-0    Rm - Operand Register               (R0-R14)

Multiply and Multiply-Accumulate (MUL,MLA)
Restrictions: Rd may not be same as Rm. Rd,Rn,Rs,Rm may not be R15.
Note: Only the lower 32bit of the internal 64bit result are stored in Rd, thus no sign/zero extension is required and MUL and MLA can be used for both signed and unsigned calculations!
Execution Time: 1S+mI for MUL, and 1S+(m+1)I for MLA. Whereas 'm' depends on whether/how many most significant bits of Rs are all zero or all one. That is m=1 for Bit 31-8, m=2 for Bit 31-16, m=3 for Bit 31-24, and m=4 otherwise.
Flags (if S=1): Z=zeroflag, N=signflag, C=destroyed (ARMv4 and below) or C=not affected (ARMv5 and up), V=not affected. MUL/MLA supported by ARMv2 and up.

Multiply Long and Multiply-Accumulate Long (MULL, MLAL)
Optionally supported, INCLUDED in ARMv3M, EXCLUDED in ARMv4xM/ARMv5xM.
Restrictions: RdHi,RdLo,Rm must be different registers. R15 may not be used.
Execution Time: 1S+(m+1)I for MULL, and 1S+(m+2)I for MLAL. Whereas 'm' depends on whether/how many most significant bits of Rs are "all zero" (UMULL/UMLAL) or "all zero or all one" (SMULL,SMLAL). That is m=1 for Bit 31-8, m=2 for Bit 31-16, m=3 for Bit 31-24, and m=4 otherwise.
Flags (if S=1): Z=zeroflag, N=signflag, C=destroyed (ARMv4 and below) or C=not affected (ARMv5 and up), V=destroyed??? (ARMv4 and below???) or V=not affected (ARMv5 and up).

Signed Halfword Multiply (SMLAxy,SMLAWy,SMLALxy,SMULxy,SMULWy)
Supported by E variants of ARMv5 and up, ie. ARMv5TE(xP).
Q-flag gets set on 32bit SMLAxy/SMLAWy addition overflows, however, the result is NOT truncated (as it'd be done with QADD opcodes).
Q-flag is NOT affected on (rare) 64bit SMLALxy addition overflows.
SMULxy/SMULWy cannot overflow, and thus leave Q-flag unchanged as well.
NZCV-flags are not affected by Halfword multiplies.


 ARM.9: Single Data Transfer (LDR, STR, PLD)

Opcode Format
  Bit    Expl.
  31-28  Condition (Must be 1111b for PLD)
  27-26  Must be 01b for this instruction
  25     I - Immediate Offset Flag (0=Immediate, 1=Shifted Register)
  24     P - Pre/Post (0=post; add offset after transfer, 1=pre; before trans.)
  23     U - Up/Down Bit (0=down; subtract offset from base, 1=up; add to base)
  22     B - Byte/Word bit (0=transfer word quantity, 1=transfer byte quantity)
  When above Bit 24 P=0 (Post-indexing, write-back is ALWAYS enabled):
    21     T - Memory Management (0=Normal, 1=Force non-privileged access)
  When above Bit 24 P=1 (Pre-indexing, write-back is optional):
    21     W - Write-back bit (0=no write-back, 1=write address into base)
  20     L - Load/Store bit (0=Store to memory, 1=Load from memory)
          0: STR{cond}{B}{T} Rd,<Address>   ;[Rn+/-<offset>]=Rd
          1: LDR{cond}{B}{T} Rd,<Address>   ;Rd=[Rn+/-<offset>]
         (1: PLD <Address> ;Prepare Cache for Load, see notes below)
          Whereas, B=Byte, T=Force User Mode (only for POST-Indexing)
  19-16  Rn - Base register               (R0..R15) (including R15=PC+8)
  15-12  Rd - Source/Destination Register (R0..R15) (including R15=PC+12)
  When above I=0 (Immediate as Offset)
    11-0   Unsigned 12bit Immediate Offset (0-4095, steps of 1)
  When above I=1 (Register shifted by Immediate as Offset)
    11-7   Is - Shift amount      (1-31, 0=Special/See below)
    6-5    Shift Type             (0=LSL, 1=LSR, 2=ASR, 3=ROR)
    4      Must be 0 (Reserved, see ARM.17, The Undefined Instruction)
    3-0    Rm - Offset Register   (R0..R14) (not including PC=R15)

Instruction Formats for <Address>
An expression which generates an address:
  <expression>                  ;an immediate used as address
  ;*** restriction: must be located in range PC+/-4095+8, if so,
  ;*** assembler will calculate offset and use PC (R15) as base.
Pre-indexed addressing specification:
  [Rn]                          ;offset = zero
  [Rn, <#{+/-}expression>]{!}   ;offset = immediate
  [Rn, {+/-}Rm{,<shift>} ]{!}   ;offset = register shifted by immediate
Post-indexed addressing specification:
  [Rn], <#{+/-}expression>      ;offset = immediate
  [Rn], {+/-}Rm{,<shift>}       ;offset = register shifted by immediate
Whereas...
  <shift>  immediate shift such like LSL#4, ROR#2, etc. (see ARM.5).
  {!}      exclamation mark ("!") indicates write-back (Rn will be updated).

Notes
Shift amount 0 has special meaning, as described in ARM.5 Data Processing.
When writing a word (32bit) to memory, the address should be word-aligned.
When reading a byte from memory, upper 24 bits of Rd are zero-extended.
LDR PC,<op> on ARMv4 leaves CPSR.T unchanged.
LDR PC,<op> on ARMv5 sets CPSR.T to <op> Bit0, (1=Switch to Thumb).

When reading a word from a halfword-aligned address (which is located in the middle between two word-aligned addresses), the lower 16bit of Rd will contain [address] ie. the addressed halfword, and the upper 16bit of Rd will contain [Rd-2] ie. more or less unwanted garbage. However, by isolating lower bits this may be used to read a halfword from memory. (Above applies to little endian mode, as used in GBA.)

In a virtual memory based environment (ie. not in the GBA), aborts (ie. page faults) may take place during execution, if so, Rm and Rn should not specify the same register when post-indexing is used, as the abort-handler might have problems to reconstruct the original value of the register.

Return: CPSR flags are not affected.
Execution Time: For normal LDR: 1S+1N+1I. For LDR PC: 2S+2N+1I. For STR: 2N.

PLD <Address> ;Prepare Cache for Load
PLD must use following settings cond=1111b, P=1, B=1, W=0, L=1, Rd=1111b, the address may not use post-indexing, and may not use writeback, the opcode is encoded identical as LDRNVB R15,<Address>.
PLD signalizes to the memory system that a specific memory address will be soon accessed, the memory system may use this hint to prepare caching/pipelining, aside from that, PLD does not have any affect to the program logic, and behaves identical as NOP.
PLD supported by ARMv5TE only, not ARMv5, not ARMv5TExP.


 ARM.10: Halfword, Doubleword, and Signed Data Transfer

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-25  Must be 000b for this instruction
  24     P - Pre/Post (0=post; add offset after transfer, 1=pre; before trans.)
  23     U - Up/Down Bit (0=down; subtract offset from base, 1=up; add to base)
  22     I - Immediate Offset Flag (0=Register Offset, 1=Immediate Offset)
  When above Bit 24 P=0 (Post-indexing, write-back is ALWAYS enabled):
    21     Not used, must be zero (0)
  When above Bit 24 P=1 (Pre-indexing, write-back is optional):
    21     W - Write-back bit (0=no write-back, 1=write address into base)
  20     L - Load/Store bit (0=Store to memory, 1=Load from memory)
  19-16  Rn - Base register                (R0-R15) (Including R15=PC+8)
  15-12  Rd - Source/Destination Register  (R0-R15) (Including R15=PC+12)
  11-8   When above Bit 22 I=0 (Register as Offset):
           Not used. Must be 0000b
         When above Bit 22 I=1 (immediate as Offset):
           Immediate Offset (upper 4bits)
  7      Reserved, must be set (1)
  6-5    Opcode (0-3)
         When Bit 20 L=0 (Store) (and Doubleword Load/Store):
          0: Reserved for SWP instruction (see ARM.12 Single Data Swap)
          1: STR{cond}H  Rd,<Address>  ;Store halfword   [a]=Rd
          2: LDR{cond}D  Rd,<Address>  ;Load Doubleword  R(d)=[a], R(d+1)=[a+4]
          3: STR{cond}D  Rd,<Address>  ;Store Doubleword [a]=R(d), [a+4]=R(d+1)
         When Bit 20 L=1 (Load):
          0: Reserved.
          1: LDR{cond}H  Rd,<Address>  ;Load Unsigned halfword (zero-extended)
          2: LDR{cond}SB Rd,<Address>  ;Load Signed byte (sign extended)
          3: LDR{cond}SH Rd,<Address>  ;Load Signed halfword (sign extended)
  4      Reserved, must be set (1)
  3-0    When above Bit 22 I=0:
           Rm - Offset Register            (R0-R14) (not including R15)
         When above Bit 22 I=1:
           Immediate Offset (lower 4bits)  (0-255, together with upper bits)
STRH,LDRH,LDRSB,LDRSH supported on ARMv4 and up.
STRD/LDRD supported on ARMv5TE only, not ARMv5, not ARMv5TExP.
STRD/LDRD: base writeback: Rn should not be same as R(d) or R(d+1).
STRD: index register: Rm should not be same as R(d) or R(d+1).
STRD/LDRD: Rd must be an even numbered register (R0,R2,R4,R6,R8,R10,R12).
STRD/LDRD: Address must be double-word aligned (multiple of eight).

Instruction Formats for <Address>
An expression which generates an address:
  <expression>                  ;an immediate used as address
  ;*** restriction: must be located in range PC+/-255+8, if so,
  ;*** assembler will calculate offset and use PC (R15) as base.
Pre-indexed addressing specification:
  [Rn]                          ;offset = zero
  [Rn, <#{+/-}expression>]{!}   ;offset = immediate
  [Rn, {+/-}Rm]{!}              ;offset = register
Post-indexed addressing specification:
  [Rn], <#{+/-}expression>      ;offset = immediate
  [Rn], {+/-}Rm                 ;offset = register
Whereas...
  {!}      exclamation mark ("!") indicates write-back (Rn will be updated).

Return: No Flags affected.
Execution Time: For Normal LDR, 1S+1N+1I. For LDR PC, 2S+2N+1I. For STRH 2N.


 ARM.11: Block Data Transfer (LDM,STM)

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-25  Must be 100b for this instruction
  24     P - Pre/Post (0=post; add offset after transfer, 1=pre; before trans.)
  23     U - Up/Down Bit (0=down; subtract offset from base, 1=up; add to base)
  22     S - PSR & force user bit (0=No, 1=load PSR or force user mode)
  21     W - Write-back bit (0=no write-back, 1=write address into base)
  20     L - Load/Store bit (0=Store to memory, 1=Load from memory)
          0: STM{cond}{amod} Rn{!},<Rlist>{^}  ;Store (Push)
          1: LDM{cond}{amod} Rn{!},<Rlist>{^}  ;Load  (Pop)
          Whereas, {!}=Write-Back (W), and {^}=PSR/User Mode (S)
  19-16  Rn - Base register                (R0-R14) (not including R15)
  15-0   Rlist - Register List
  (Above 'offset' is meant to be the number of words specified in Rlist.)

Addressing Modes {amod}
The IB,IA,DB,DA suffixes directly specify the desired U and P bits:
  IB  increment before          ;P=1, U=1
  IA  increment after           ;P=0, U=1
  DB  decrement before          ;P=1, U=0
  DA  decrement after           ;P=0, U=0
Alternately, FD,ED,FA,EA could be used, mostly to simplify mnemonics for stack transfers.
  ED  empty stack, descending   ;LDM: P=1, U=1  ;STM: P=0, U=0
  FD  full stack,  descending   ;     P=0, U=1  ;     P=1, U=0
  EA  empty stack, ascending    ;     P=1, U=0  ;     P=0, U=1
  FA  full stack,  ascending    ;     P=0, U=0  ;     P=1, U=1
Ie. the following expressions are aliases for each other:
  STMFD=STMDB=PUSH   STMED=STMDA   STMFA=STMIB   STMEA=STMIA
  LDMFD=LDMIA=POP    LDMED=LDMIB   LDMFA=LDMDA   LDMEA=LDMDB
Note: The equivalent THUMB functions use fixed organization:
  PUSH/POP: full descending     ;base register SP (R13)
  LDM/STM:  increment after     ;base register R0..R7
Descending is common stack organization as used in 80x86 and Z80 CPUs, SP is decremented when pushing/storing data, and incremented when popping/loading data.

When S Bit is set (S=1)
If instruction is LDM and R15 is in the list: (Mode Changes)
  While R15 loaded, additionally: CPSR=SPSR_<current mode>
Otherwise: (User bank transfer)
  Rlist is referring to User Bank Registers, R0-R15 (rather than
  register related to the current mode, such like R14_svc etc.)
  Base write-back should not be used for User bank transfer.
  !  When instruction is LDM:                                   !
  !  If the following instruction reads from a banked register, !
  !  like R14_svc, then CPU might still read R14 instead. If    !
  !  necessary insert a dummy instruction such like MOV R0,R0.  !

Notes
The lowest Register in Rlist (R0 if its in the list) will be loaded/stored to/from the lowest memory address.
The base address should be usually word-aligned.
LDM Rn,...,PC on ARMv4 leaves CPSR.T unchanged.
LDR Rn,...,PC on ARMv5 sets CPSR.T to <op> Bit0, (1=Switch to Thumb).

Inclusion of the base in the register list
When write-back is specified, the base is written back at the end of the second cycle of the instruction. During a STM, the first register is written out at the start of the second cycle. A STM which includes storing the base as the first register to be stored, will therefore store the unchanged value, whereas with the base second or later in the transfer order, will store the modified value.
A LDM will always overwrite the updated base if the base is in the list.

Return: No Flags affected.
Execution Time: For normal LDM, nS+1N+1I. For LDM PC, (n+1)S+2N+1I. For STM (n-1)S+2N. Where n is the number of words transferred.


 ARM.12: Single Data Swap (SWP)

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-23  Must be 00010b for this instruction
         Opcode (fixed)
           SWP{cond}{B} Rd,Rm,[Rn]      ;Rd=[Rn], [Rn]=Rm
  22     B - Byte/Word bit (0=swap word quantity, 1=swap byte quantity)
  21-20  Must be 00b for this instruction
  19-16  Rn - Base register                     (R0-R14)
  15-12  Rd - Destination Register              (R0-R14)
  11-4   Must be 00001001b for this instruction
  3-0    Rm - Source Register                   (R0-R14)
SWP/SWPB supported by ARMv2a and up.
Swap works properly including if Rm and Rn specify the same register.
R15 may not be used for either Rn,Rd,Rm. (Rn=R15 would be MRS opcode).
Upper bits of Rd are zero-expanded when using Byte quantity. For info about byte and word data memory addressing, read LDR and STR opcode description.
Execution Time: 1S+2N+1I. That is, 2N data cycles, 1S code cycle, plus 1I.


 ARM.13: Software Interrupt (SWI,BKPT)

Opcode Format
  Bit    Expl.
  31-28  Condition (must be 1110b for BKPT, ie. Condition=always)
  27-24  Opcode
          1111b: SWI{cond} nn   ;software interrupt
          0001b: BKPT      nn   ;breakpoint (ARMv5 and up)
  For SWI:
   23-0   nn - Comment Field, ignored by processor (24bit value)
  For BKPT:
   23-20  Must be 0010b for BKPT
   19-8   nn - upper 12bits of comment field, ignored by processor
   7-4    Must be 0111b for BKPT
   3-0    nn - lower 4bits of comment field, ignored by processor
SWI supposed for calls to the operating system - Enter Supervisor mode (SVC) in ARM state. BKPT intended for debugging - enters Abort mode in ARM state via Prefetch Abort vector.

Execution SWI/BKPT:
  R14_svc=PC+4     R14_abt=PC+4   ;save return address
  SPSR_svc=CPSR    SPSR_abt=CPSR  ;save CPSR flags
  CPSR=<changed>   CPSR=<changed> ;Enter svc/abt, ARM state, IRQs disabled
  PC=VVVV0008h     PC=VVVV000Ch   ;jump to SWI/PrefetchAbort vector address
Execution Time: 2S+1N

Interpreting the Comment Field:
The immediate parameter is ignored by the processor, the user interrupt handler may read-out this number by examining the lower 24bit of the 32bit opcode opcode at [R14_svc-4]. In case that your program executes SWI's from inside of THUMB mode also: Your SWI handler must then examine the T Bit SPSR_svc in order to determine whether it's been a THUMB SWI - if so, examining the lower 8bit of the 16bit opcode opcode at [R14_svc-2].

For Returning from SWI use this instruction:
  MOVS PC,R14
That instructions does both restoring PC and CPSR, ie. PC=R14_svc, and CPSR=SPRS_svc.

Nesting SWIs:
SPSR_svc and R14_svc should be saved on stack before either invoking nested SWIs, or (if the IRQ handler uses SWIs) before enabling IRQs.


 ARM.14: Coprocessor Data Operations (CDP)

Opcode Format
  Bit    Expl.
  31-28  Condition (or 1111b for CDP2 opcode on ARMv5 and up)
  27-24  Must be 1110b for this instruction
         ARM-Opcode (fixed)
           CDP{cond} Pn,<cpopc>,Cd,Cn,Cm{,<cp>}
           CDP2      Pn,<cpopc>,Cd,Cn,Cm{,<cp>}
  23-20  CP Opc - Coprocessor operation code       (0-15)
  19-16  Cn     - Coprocessor operand Register     (C0-C15)
  15-12  Cd     - Coprocessor destination Register (C0-C15)
  11-8   Pn     - Coprocessor number               (P0-P15)
  7-5    CP     - Coprocessor information          (0-7)
  4      Reserved, must be zero (otherwise MCR/MRC opcode)
  3-0    Cm     - Coprocessor operand Register     (C0-C15)
CDP supported by ARMv2 and up, CDP2 by ARMv5 and up.
Execution time: 1S+bI, b=number of cycles in coprocessor busy-wait loop.
Return: No flags affected, no ARM-registers used/modified.
For details refer to original ARM docs, irrelevant in GBA because no coprocessor exists.


 ARM.15: Coprocessor Data Transfers (LDC,STC)

Opcode Format
  Bit    Expl.
  31-28  Condition (or 1111b for LDC2/STC2 opcodes on ARMv5 and up)
  27-25  Must be 110b for this instruction
  24     P - Pre/Post (0=post; add offset after transfer, 1=pre; before trans.)
  23     U - Up/Down Bit (0=down; subtract offset from base, 1=up; add to base)
  22     N - Transfer length (0-1, interpretation depends on co-processor)
  21     W - Write-back bit (0=no write-back, 1=write address into base)
  20     Opcode (0-1)
          0: STC{cond}{L} Pn,Cd,<Address>  ;Store to memory (from coprocessor)
          0: STC2{L}      Pn,Cd,<Address>  ;Store to memory (from coprocessor)
          1: LDC{cond}{L} Pn,Cd,<Address>  ;Read from memory (to coprocessor)
          1: LDC2{L}      Pn,Cd,<Address>  ;Read from memory (to coprocessor)
          whereas {L} indicates long transfer (Bit 22: N=1)
  19-16  Rn     - ARM Base Register              (R0-R15)     (R15=PC+8)
  15-12  Cd     - Coprocessor src/dest Register  (C0-C15)
  11-8   Pn     - Coprocessor number             (P0-P15)
  7-0    Offset - Unsigned Immediate, step 4     (0-1020, in steps of 4)
LDC/STC supported by ARMv2 and up, LDC2/STC2 by ARMv5 and up.
Execution time: (n-1)S+2N+bI, n=number of words transferred.
For details refer to original ARM docs, irrelevant in GBA because no coprocessor exists.


 ARM.16: Coprocessor Register Transfers (MRC, MCR)

Opcode Format
  Bit    Expl.
  31-28  Condition (or 1111b for MRC2/MCR2 opcodes on ARMv5 and up)
  27-24  Must be 1110b for this instruction
  23-21  CP Opc - Coprocessor operation code         (0-7)
  20     ARM-Opcode (0-1)
          0: MCR{cond} Pn,<cpopc>,Rd,Cn,Cm{,<cp>}   ;move from ARM to CoPro
          0: MCR2      Pn,<cpopc>,Rd,Cn,Cm{,<cp>}   ;move from ARM to CoPro
          1: MRC{cond} Pn,<cpopc>,Rd,Cn,Cm{,<cp>}   ;move from CoPro to ARM
          1: MRC2      Pn,<cpopc>,Rd,Cn,Cm{,<cp>}   ;move from CoPro to ARM
  19-16  Cn     - Coprocessor source/dest. Register  (C0-C15)
  15-12  Rd     - ARM source/destination Register    (R0-R15)
  11-8   Pn     - Coprocessor number                 (P0-P15)
  7-5    CP     - Coprocessor information            (0-7)
  4      Reserved, must be one (1) (otherwise CDP opcode)
  3-0    Cm     - Coprocessor operand Register       (C0-C15)
MCR/MRC supported by ARMv2 and up, MCR2/MRC2 by ARMv5 and up.
A22i syntax allows to use MOV with Rd specified as first (dest), or last (source) operand. Native MCR/MRC syntax uses Rd as middle operand, <cp> can be ommited if <cp> is zero.
When using MCR with R15: Coprocessor will receive a data value of PC+12.
When using MRC with R15: Bit 31-28 of data are copied to Bit 31-28 of CPSR (ie. N,Z,C,V flags), other data bits are ignored, CPSR Bit 27-0 are not affected, R15 (PC) is not affected.
Execution time: 1S+bI+1C for MCR, 1S+(b+1)I+1C for MRC.
Return: For MRC only: Either R0-R14 modified, or flags affected (see above).
For details refer to original ARM docs, irrelevant in GBA because no coprocessor exists.


 ARM.X: Coprocessor Double-Register Transfer (MCRR,MRRC)

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-21  Must be 1100010b for this instruction
  20     L - Opcode (Load/Store)
          0: MCRR{cond} Pn,opcode,Rd,Rn,Cm  ;write Rd,Rn to coproc
          1: MRRC{cond} Pn,opcode,Rd,Rn,Cm  ;read Rd,Rn from coproc
  19-16  Rn - Second source/dest register      (R0-R14)
  15-12  Rd - First source/dest register       (R0-R14)
  11-8   Pn     - Coprocessor number           (P0-P15)
  7-4    CP Opc - Coprocessor operation code   (0-15)
  3-0    Cm     - Coprocessor operand Register (C0-C15)
Supported by ARMv5TE only, not ARMv5, not ARMv5TExP.


 ARM.17: Undefined Instruction

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-25  Must be 011b for this instruction
  24-5   Reserved for future use
  4      Must be 1b for this instruction
  3-0    Reserved for future use
No assembler mnemonic exists, following bitstreams are (not) reserved.
  cond011xxxxxxxxxxxxxxxxxxxx1xxxx - reserved for future use (except below).
  cond01111111xxxxxxxxxxxx1111xxxx - free for user.
Execution time: 2S+1I+1N.


 ARM.X: Count Leading Zeros

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-16  Must be 0001.0110.1111b for this instruction
         Opcode (fixed)
           CLZ{cond} Rd,Rm  ;Rd=Number of leading zeros in Rm
  15-12  Rd - Destination Register              (R0-R14)
  11-4   Must be 1111.0001b for this instruction
  3-0    Rm - Source Register                   (R0-R14)
CLZ supported by ARMv5 and up.
Return: No Flags affected. Rd=0..32.


 ARM.X: QADD/QSUB

Opcode Format
  Bit    Expl.
  31-28  Condition
  27-24  Must be 0001b for this instruction
  23-20  Opcode
          0000b: QADD{cond}  Rd,Rm,Rn    ;Rd=Rm+Rn
          0010b: QSUB{cond}  Rd,Rm,Rn    ;Rd=Rm-Rn
          0100b: QDADD{cond} Rd,Rm,Rn    ;Rd=Rm+Rn*2 (doubled)
          0110b: QDSUB{cond} Rd,Rm,Rn    ;Rd=Rm-Rn*2 (doubled)
  19-16  Rn - Second Source Register            (R0-R14)
  15-12  Rd - Destination Register              (R0-R14)
  11-4   Must be 00000101b for this instruction
  3-0    Rm - First Source Register             (R0-R14)
Supported by E variants of ARMv5 and up, ie. ARMv5TE(xP).
Results truncated to signed 32bit range in case of overflows, with the Q-flag being set (and being left unchanged otherwise). NZCV flags are not affected.
Note: Rn*2 is internally processed first, and may get truncated - even if the final result would fit into range.


 ARM 26bit Memory Interface

The 26bit Memory Interface was used by ARMv1 and ARMv2. The 32bit interface is used by ARMv3 and newer, however, 26bit backward compatibility was included in all ARMv3 (except ARMv3G), and optionally in some non-T variants of ARMv4.

Format of R15 in 26bit Mode (Program Counter Register)
  Bit   Name     Expl.
  31-28 N,Z,C,V  Flags (Sign, Zero, Carry, Overflow)
  27-26 I,F      Interrupt Disable bits (IRQ, FIQ) (1=Disable)
  25-2  PC       Program Counter, 24bit, Step 4 (64M range)
  1-0   M1,M0    Mode (0=User, 1=FIQ, 2=IRQ, 3=Supervisor)
Branches with +/-32M range wrap the PC register, and can reach all 64M memory.

Reading from R15
If R15 is specified in bit16-19 of an opcode, then NZCVIF and M0,1 are masked (zero), otherwise the full 32bits are used.

Writing to R15
Data Processing opcodes with S=1, and LDM opcodes with PSR=1 can write to all 32bits in R15 (in 26bit mode, that is allowed even in user mode, though it does then affect only NZCF, not the write protected IFMM bits ???), other opcodes which write to R15 will modify only the program counter bits. Also, special CMP/CMN/TST/TEQ{P} opcodes can be used to write to the PSR bits in R15 without modifying the PC bits.

Exceptions
SWIs, Reset, Data/Prefetch Aborts and Undefined instructions enter Supervisor mode. Interrupts enter IRQ and FIQ mode. Additionally, a special 26bit Address Exception exists, which enters Supervisor mode on accesses to memory addresses>=64M as follows:
  R14_svc = PC ($+8, including old PSR bits)
  M1,M0 = 11b = supervisor mode, F=same, I=1, PC=14h
to continue at the fault location, return by SUBS PC,LR,8.

26bit Backwards Compatibility on 32bit ARMv3 and up
CPSR M4=0 = 26bit mode (with USR,FIQ,IRQ,SVC modes in M1,M0)
32bit CPUs with 26bit compatibility mode can be configured to switch into 32bit mode when encountering exceptions.


 Pseudo Instructions and Directives

ARM Pseudo Instructions
  nop              mov r0,r0
  ldr Rd,=Imm      ldr Rd,[r15,disp] ;use .pool as parameter field)
  add Rd,=addr     add/sub Rd,r15,disp
  adr Rd,addr      add/sub Rd,r15,disp
  adrl Rd,addr     two add/sub opcodes with disp=xx00h+00yyh
  mov Rd,Imm       mvn Rd,NOT Imm    ;or vice-versa
  and Rd,Rn,Imm    bic Rd,Rn,NOT Imm ;or vice-versa
  cmp Rd,Rn,Imm    cmn Rd,Rn,-Imm    ;or vice-versa
  add Rd,Rn,Imm    sub Rd,Rn,-Imm    ;or vice-versa
All above opcodes may be made conditional by specifying a {cond} field.

THUMB Pseudo Instructions
  nop              mov r8,r8
  ldr Rd,=Imm      ldr Rd,[r15,disp] ;use .pool as parameter field
  add Rd,=addr     add Rd,r15,disp
  adr Rd,addr      add Rd,r15,disp
  mov Rd,Rs        add Rd,Rs,0       ;with Rd,Rs in range r0-r7 each

A22i Directives
  org  adr     assume following code from this address on
  .gba         indicate GBA program
  .fix         fix GBA header checksum
  .norewrite   do not delete existing output file (keep following data in file)
  .data?       following defines RAM data structure (assembled to nowhere)
  .code        following is normal ROM code/data (assembled to ROM image)
  .include     includes specified source code file (no nesting/error handling)
  .import      imports specified binary file (optional parameters: ,begin,len)
  .radix nn    changes default numeric format (nn=2,8,10,16 = bin/oct/dec/hex)
  .errif expr  generates an error message if expression is nonzero
  .if expr     assembles following code only if expression is nonzero
  .else        invert previous .if condition
  .endif       terminate .if/.ifdef/.ifndef
  .ifdef sym   assemble following only if symbol is defined
  .ifndef sym  assemble following only if symbol is not defined
  .align nn    aligns to an address divisible-by-nn, inserts 00's
  .msg         defines a no$gba debugmessage string, such like .msg 'Init Okay'
  .brk         defines a no$gba source code break opcode
  l equ n      l=n
  l:   [cmd]   l=$   (global label)
  @@l: [cmd]   @@l=$ (local label, all locals are reset at next global label)
  end          end of source code
  db ...       define 8bit data (bytes)
  dw ...       define 16bit data (halfwords)
  dd ...       define 32bit data (words)
  defs nn      define nn bytes space (zero-filled)
  ;...         defines a comment (ignored by the assembler)
  //           alias for CRLF, eg. allows <db 'Text',0 // dw addr> in one line

A22i Alias Directives (for compatibility with other assemblers)
  align        .align 4          code16    .thumb
  align nn     .align nn         .code 16  .thumb
  % nn         defs nn           code32    .arm
  .space nn    defs nn           .code 32  .arm
  ..ds nn      defs nn           ltorg     .pool
  x=n          x equ n           .ltorg    .pool
  .equ x,n     x equ n           ..ltorg   .pool
  .define x n  x equ n           dcb       db (8bit data)
  incbin       .import           defb      db (8bit data)
  @@@...       ;comment          .byte     db (8bit data)
  @ ...        ;comment          .ascii    db (8bit string)
  @*...        ;comment          dcw       dw (16bit data)
  @...         ;comment          defw      dw (16bit data)
  .text        .code             .hword    dw (16bit data)
  .bss         .data?            dcd       dd (32bit data)
  .global      (ignored)         defd      dd (32bit data)
  .extern      (ignored)         .long     dd (32bit data)
  .thumb_func  (ignored)         .word     dw/dd, don't use
  #directive   .directive        .end      end
  .fill nn,1,0 defs nn

Alias Conditions, Opcodes, Operands
  hs   cs   ;condition higher or same = carry set
  asl  lsl  ;arithmetic shift left = logical shift left

A22i Numeric Formats & Dialects
  Type          Normal       Alias
  Decimal       85           #85  &d85
  Hexadecimal   55h          #55h  0x55  #0x55  $55  &h55
  Octal         125o         0o125  &o125
  Ascii         'U'          "U"
  Binary        01010101b    %01010101  0b01010101  &b01010101
  Roman         &rLXXXV      (very useful for arrays of kings and chapters)
Note: The default numeric format can be changed by the .radix directive (usually 10=decimal). For example, with radix 16, values like "85" and "0101b" are treated as hexadecimal numbers (in that case, decimal and binary numbers can be still defined with prefixes &d and &b).

A22i Numeric Operators Priority
  Prio  Operator           Aliases
  8     (,) brackets
  7     +,- sign
  6     *,/,MOD,SHL,SHR    MUL,DIV,<<,>>
  5     +,- operation
  4     EQ,GE,GT,LE,LT,NE  =,>=,>,<=,<,<>,==,!=
  3     NOT
  2     AND
  1     OR,XOR             EOR
Operators of same priority are processed from left to right.
Boolean operators (priority 4) return 1=TRUE, 0=FALSE.

A22i Nocash Syntax
Even though A22i does recognize the official ARM syntax, it's also allowing to use friendly code:
  mov   r0,0ffh         ;no C64-style "#", and no C-style "0x" required
  stmia [r7]!,r0,r4-r5  ;square [base] brackets, no fancy {rlist} brackets
  mov   r0,cpsr         ;no confusing MSR and MRS (whatever which is which)
  mov   r0,p0,0,c0,c0,0 ;no confusing MCR and MRC (whatever which is which)
  ldr   r0,[score]      ;allows to use clean brackets for relative addresses
  push  rlist           ;alias for stmfd [r13]!,rlist (and same for pop/ldmfd)
  label:                ;label definitions recommended to use ":" colons

[A22i is the no$gba debug version's built-in source code assembler.]


 ARM CP15 System Control Coprocessor

ARM CP15 Overview
ARM CP15 ID Codes
ARM CP15 Control Register
ARM CP15 Memory Managment Unit (MMU)
ARM CP15 Protection Unit (PU)
ARM CP15 Cache Control
ARM CP15 Fast Context Switch Extension (FCSE)
ARM CP15 Tightly Coupled Memory (TCM)


 ARM CP15 Overview

CP15
In many ARM CPUs, particulary such with memory control facilities, coprocessor number 15 (CP15) is used as built-in System Control Coprocessor.
CPUs without memory control functions typically do include a CP15 at all, in that case even an attempt to read the Main ID register will cause an Undefined Instruction exception.

CP15 Opcodes
CP15 can be accessed via MCR and MRC opcodes, with Pn=P15, and <cpopc>=0.
  MCR{cond} P15,0,Rd,Cn,Cm,<cp>   ;move from ARM to CP15
  MRC{cond} P15,0,Rd,Cn,Cm,<cp>   ;move from CP15 to ARM
Rd can be any ARM register in range R0-R14, R15 should not be used with P15.
Cn,Cm,<cp> are used to select a CP15 register, eg. C0,C0,0 = Main ID Register.
Other coprocessor opcodes (CDP, LDC, STC) cannot be used with P15.

CP15 Register List
  Register     Expl.
  C0,C0,0      Main ID Register (R)
  C0,C0,1      Cache Type and Size (R)
  C0,C0,2      TCM Physical Size (R)
  C1,C0,0      Control Register (R/W, or R=Fixed)
  C2,C0,0      PU Cachability Bits for Data/Unified Protection Region
  C2,C0,1      PU Cachability Bits for Instruction Protection Region
  C3,C0,0      PU Write-Bufferability Bits for Data Protection Regions
  C5,C0,0      PU Access Permission Data/Unified Protection Region
  C5,C0,1      PU Access Permission Instruction Protection Region
  C5,C0,2      PU Extended Access Permission Data/Unified Protection Region
  C5,C0,3      PU Extended Access Permission Instruction Protection Region
  C6,C0..C7,0  PU Protection Unit Data/Unified Region 0..7
  C6,C0..C7,1  PU Protection Unit Instruction Region 0..7
  C7,Cm,Op2    Cache Commands and Halt Function (W)
  C9,C0,0      Cache Data Lockdown
  C9,C0,1      Cache Instruction Lockdown
  C9,C1,0      TCM Data TCM Base and Virtual Size
  C9,C1,1      TCM Instruction TCM Base and Virtual Size
  C13,C0,0     Process ID for Fast Context Switch Extension (FCSE)
  C15,Cm,Op2   Implementation Defined

Data/Unified Registers
Some Cache/PU/TCM registers are declared as "Data/Unified".
That registers are used for Data accesses in case that the CPU contains separate Data and Instruction registers, otherwise the registers are used for both (unified) Data and Instruction accesses.


 ARM CP15 ID Codes

C0,C0,0 - Main ID Register (R)
  12-15 ARM Era (0=Pre-ARM7, 7=ARM7, other=Post-ARM7)
Post-ARM7 Processors
  0-3   Revision Number
  4-15  Primary Part Number (Bit12-15 must be other than 0 or 7)
        (eg. 946h for ARM946)
  16-19 Architecture        (1=v4, 2=v4T, 3=v5, 4=v5T, 5=v5TE)
  20-23 Variant Number
  24-31 Implementor         (41h=ARM, 44h=Digital Equipment Corp, 69h=Intel)
ARM7 Processors
  0-3   Revision Number
  4-15  Primary Part Number (Bit12-15 must be 7)
  16-22 Variant Number
  23    Architecture        (0=v3, 1=v4T)
  24-31 Implementor         (41h=ARM, 44h=Digital Equipment Corp, 69h=Intel)
Pre-ARM7 Processors
  0-3   Revision Number
  4-11  Processor ID LSBs (30h=ARM3/v2, 60h,61h,62=ARM600,610,620/v3)
  12-31 Processor ID MSBs (fixed, 41560h)

C0,C0,1 - Cache Type Register (R)
  0-11  Instruction Cache (bits 0-1=len, 2=m, 3-5=assoc, 6-8=size, 9-11=zero)
  12-23 Data Cache        (bits 0-1=len, 2=m, 3-5=assoc, 6-8=size, 9-11=zero)
  24    Separate Cache Flag (0=Unified, 1=Separate Data/Instruction Caches)
  25-28 Cache Type (0,1,2,6,7=see below, other=reserved)
         Type Method         Cache cleaning         Cache lock-down
         0    Write-through  Not needed             Not supported
         1    Write-back     Read data block        Not supported
         2    Write-back     Register 7 operations  Not supported
         6    Write-back     Register 7 operations  Format A
         7    Write-back     Register 7 operations  Format B
  29-31 Reserved (zero)
The 12bit Instruction/Data values are decoded as shown below,
  Cache Absent  = (ASSOC=0 and M=1)       ;in that case overriding below
  Cache Size    = 200h+(100h*M) shl SIZE  ;min 0.5Kbytes, max 96Kbytes
  Associativity = (1+(0.5*M)) shl ASSOC   ;min 1-way,     max 192-way
  Line Length   = 8 shl LEN               ;min 8 bytes,   max 64 bytes
For Unified cache (Bit 24=0), Instruction and Data values are identical.

C0,C0,2 - Tightly Coupled Memory (TCM) Size Register (R)
  0-1   Reserved    (0)
  2     ITCM Absent (0=Present, 1=Absent)
  3-5   Reserved    (0)
  6-9   ITCM Size   (Size = 512 SHL N) (or 0=None)
  10-13 Reserved    (0)
  14    DTCM Absent (0=Present, 1=Absent)
  15-17 Reserved    (0)
  18-21 DTCM Size   (Size = 512 SHL N) (or 0=None)
  22-31 Reserved    (0)

C0,C0,3..7 - Reserved (R)
Unused/Reserved registers containing the same value as C0,C0,0.


 ARM CP15 Control Register

C1,C0,0 - Control Register (R/W, or R=Fixed)
  0  MMU/PU Enable         (0=Disable, 1=Enable) (Fixed 0 if none)
  1  Alignment Fault Check (0=Disable, 1=Enable) (Fixed 0/1 if none/always on)
  2  Data/Unified Cache    (0=Disable, 1=Enable) (Fixed 0/1 if none/always on)
  3  Write Buffer          (0=Disable, 1=Enable) (Fixed 0/1 if none/always on)
  4  Exception Handling    (0=26bit, 1=32bit)    (Fixed 1 if always 32bit)
  5  26bit-address faults  (0=Enable, 1=Disable) (Fixed 1 if always 32bit)
  6  Abort Model (pre v4)  (0=Early, 1=Late Abort) (Fixed 1 if ARMv4 and up)
  7  Endian                (0=Little, 1=Big)     (Fixed 0/1 if fixed)
  8  System Protection bit (MMU-only)
  9  ROM Protection bit    (MMU-only)
  10 Implementation defined
  11 Branch Prediction     (0=Disable, 1=Enable)
  12 Instruction Cache     (0=Disable, 1=Enable) (ignored if Unified cache)
  13 Exception Vectors     (0=00000000h, 1=FFFF0000h)
  14 Cache Replacement     (0=Normal, 1=Predictable)
  15 Pre-ARMv5 Mode        (0=Normal, 1=Pre ARMv5; LDM/LDR/POP)
  16 DTCM Enable           (0=Disable, 1=Enable)
  17 DTCM Load Mode        (0=R/W, 1=DTCM Write-only)
  18 ITCM Enable           (0=Disable, 1=Enable)
  19 ITCM Load Mode        (0=R/W, 1=ITCM Write-only)
  20-31 Reserved           (keep these bits unchanged) (usually zero)
Various bits in this register may be read-only (fixed 0 if unsupported, or fixed 1 if always activated).


 ARM CP15 Memory Managment Unit (MMU)

Function of some registers depends on wheter the CPU contains a MMU or PU.
MMU handles virtual addressing tables.
  C2,Cm,Op2  MMU Translation Table Base
  C3,Cm,Op2  MMU Domain Access Control
  C5,Cm,Op2  MMU Fault Status
  C6,Cm,Op2  MMU Fault Address
  C8,Cm,Op2  MMU TLB Control
  C10,Cm,Op2 MMU TLB Lockdown
The GBA, and Nintendo DS do not have a MMU.


 ARM CP15 Protection Unit (PU)

Protection Unit can be enabled in Bit0 of C1,C0,0 (Control Register).

C2,C0,0 - Cachability Bits for Data/Unified Protection Region (R/W)
C2,C0,1 - Cachability Bits for Instruction Protection Region (if any) (R/W)
  0-7  Cachable (C) bits for region 0-7
  8-31 Reserved/zero

C3,C0,0 - Write-Bufferability Bits for Data Protection Regions (R/W)
  0-7  Bufferable (B) bits for region 0-7
  8-31 Reserved/zero
Instruction fetches are, obviously, always read-operations. So, there are no write-bufferability bits for Instruction Protection Regions.

C5,C0,0 - Access Permission Data/Unified Protection Region (R/W)
C5,C0,1 - Access Permission Instruction Protection Region (if any) (R/W)
C5,C0,2 - Extended Access Permission Data/Unified Protection Region (R/W)
C5,C0,3 - Extended Access Permission Instruction Protection Region (if any) (R/W/W)
For C5,C0,0 and C5,C0,1:
  0-15  Access Permission (AP) bits for region 0-7 (Bits 0-1=AP0, 2-3=AP1, etc)
  16-31 Reserved/zero
For C5,C0,2 and C5,C0,3 (Extended):
  0-31  Access Permission (AP) bits for region 0-7 (Bits 0-3=AP0, 4-7=AP1, etc)
The possible AP settings (0-3 for C5,C0,0..1, or 0-15 for C5,C0,2..3) are:
  AP  Privileged User
  0   -          -
  1   R/W        -
  2   R/W        R
  3   R/W        R/W
  5   R          -
  6   R          R
Settings 5,6 only for Extended Registers, settings 4,7..15 are Reserved.

C6,C0..C7,0 - Protection Unit Data/Unified Region 0..7 (R/W)
C6,C0..C7,1 - Protection Unit Instruction Region 0..7 (R/W) if any
  0     Protection Region Enable (0=Disable, 1=Enable)
  1-5   Protection Region Size   (2 SHL X) ;min=(X=11)=4KB, max=(X=31)=4GB
  6-11  Reserved/zero
  12-31 Protection Region Base address (Addr = Y*4K; must be SIZE-aligned)
Overlapping Regions are allowed, Region 7 is having highest priority, region 0 lowest priority.

Background Region
Additionally, any memory areas outside of the eight Protection Regions are handled as Background Region, this region has neither Read nor Write access.

Unified Region Note
On some systems (eg. NDS), the Region registers may be unified (same settings used for code and data), whilst the Cachabilty and Permission registers may be non-unified (separate registers for code and data settings).


 ARM CP15 Cache Control

Cache enabled/controlled by Bit 2,3,12,14 in Control Register.
Cache type detected in Cache Type Register.

C7,C0..C15,0..7 - Cache Commands (W)
Write-only Cache Command Register. Cm,Op2 operands used to select a specific command, with parameter value in Rd.
  Cn,Cm,Op2 Rd   Command
  C7,C0,4   0    Wait For Interrupt (Halt)
  C7,C5,0   0    Invalidate Entire Instruction Cache
  C7,C5,1   VA   Invalidate Instruction Cache Line
  C7,C5,2   S/I  Invalidate Instruction Cache Line
  C7,C5,4   0    Flush Prefetch Buffer
  C7,C5,6   0    Flush Entire Branch Target Cache
  C7,C5,7   IMP  Flush Branch Target Cache Entry
  C7,C6,0   0    Invalidate Entire Data Cache
  C7,C6,1   VA   Invalidate Data Cache Line
  C7,C6,2   S/I  Invalidate Data Cache Line
  C7,C7,0   0    Invalidate Entire Unified Cache
  C7,C7,1   VA   Invalidate Unified Cache Line
  C7,C7,2   S/I  Invalidate Unified Cache Line
  C7,C8,2   0    Wait For Interrupt (Halt), alternately to C7,C0,4
  C7,C10,1  VA   Clean Data Cache Line
  C7,C10,2  S/I  Clean Data Cache Line
  C7,C10,4  0    Drain Write Buffer
  C7,C11,1  VA   Clean Unified Cache Line
  C7,C11,2  S/I  Clean Unified Cache Line
  C7,C13,1  VA   Prefetch Instruction Cache Line
  C7,C14,1  VA   Clean and Invalidate Data Cache Line
  C7,C14,2  S/I  Clean and Invalidate Data Cache Line
  C7,C15,1  VA   Clean and Invalidate Unified Cache Line
  C7,C15,2  S/I  Clean and Invalidate Unified Cache Line
Parameter values (Rd) formats:
  0    Not used, should be zero
  VA   Virtual Address
  S/I  Set/index; Bit 31..(32-A) = Index, Bit (L+S-1)..L = Set ?
  IMP  ?

C9,C0,0 - Data Cache Lockdown
C9,C0,1 - Instruction Cache Lockdown
(Width (W) of index field depends on cache ASSOCIATIVETY.)
Format A:
  0..(31-W)  Reserved/zero
  (32-W)..31 Lockdown Block Index
Format B:
  0..(W-1)   Lockdown Block Index
  W..30      Reserved/zero
  31         L

Cache/Write-buffer should not be enabled for the whole 4GB memory area, high-speed TCM memory doesn't require caching, and caching would have fatal results on I/O ports. So, cache can be used only in combination with the Protection Unit, which allows to enable/disable caching in specified regions.

Note
ARMv5 instruction set supports a Cache Prepare for Load opcode (PLD), see
ARM.9: Single Data Transfer (LDR, STR, PLD)


 ARM CP15 Fast Context Switch Extension (FCSE)

C13,C0,0 - Process ID for Fast Context Switch Extension (FCSE) (R/W)
  0-24  Reserved/zero
  25-31 Process ID (PID) (0-127) (0=Disable)
The FCSE allows different processes (each assembled with ORG 0) to be located at virtual addresses in the 1st 32MB area. The FCSE splits the total 4GB address space into blocks of 32MB, accesses to Block(0) are redirected to Block(PID):
  IF addr<32M then addr=addr+PID*32M
  Respectively, with PID=0, the address remains unchanged (FCSE disabled).
The CPU-to-Memory address handling is shown below:
  1. CPU outputs a virtual address (VA)
  2. FCSE adjusts the VA to a modified virtual address (MVA)
  3. Cache hits determined by examining the MVA, continue below if no hit
  4. MMU translates MVA to physical address (PA) (if no MMU present: PA=MVA)
  5. Memory access occurs at PA
The FCSE allows limited virtual addressing even if no MMU is present.
If the MMU is present, then either the FCSE and/or the MMU can be used for virtual addressing; the advantage of using the FCSE (a single write to C13,C0,0) is less overload; using the MMU for the same purpose would require to change virtual address translation table in memory, and to flush the cache.


 ARM CP15 Tightly Coupled Memory (TCM)

TCM is high-speed memory, directly contained in the ARM CPU core.

TCM and DMA
TCM doesn't use the ARM bus. A minor disadvantage is that TCM cannot be accessed by DMA. However, the main advantage is that, when using TCM, the CPU can be kept running without any waitstates even while the bus is used for DMA transfers. Operation during DMA works only if all code/data is located in TCM, waitstates are generated if any code/data outside TCM is accessed; in worst case (if there are no gaps in the DMA) then the CPU is halted until the DMA finishes.

TCM and DMA and IRQ
No idea if/how IRQs are handled during DMA? Eventually (unlikely) code in TCM is kept executed until DMA finishes (ie. until the IRQ vector can be accessed. Eventually the IRQ vector is instantly accessed (causing to halt the CPU until DMA finishes). In both cases: Assuming that IRQs are enabled, and that the IRQ vector and/or IRQ handler are located outside TCM.

Separate Instruction (ITCM) and Data (DTCM) Memory
DTCM can be used only for Data accesses, typically used for stacks and other frequently accessed data.
ITCM is primarily intended for instruction accesses, but it can be also used for Data accesses (among others allowing to copy code to ITCM), however, performance isn't optimal when simultaneously accessing ITCM for code and data (such like opcodes in ITCM that use literal pool values in ITCM).

TCM Enable, TCM Load Mode
CP15 Control Register allows to enable ITCM and DTCM, and to switch ITCM/DTCM into Load Mode. In Load Mode (when TCM is enabled), TCM becomes write-only; this allows to read data from source addresses in main memory, and to write data to destination addresses in TCM by using the same addresses; useful for initializing TCM with overlapping source/dest addresses; Load mode works with all Load/Store opcodes, it does NOT work with SWP/SWPB opcodes.

TCM Physical Size can be detected in 3rd ID Code Register. (C0,C0,2)

C9,C1,0 - Data TCM Size/Base (R/W)
C9,C1,1 - Instruction TCM Size/Base (R/W)
  0     Reserved     (0)
  1-5   Virtual Size (Size = 512 SHL N) ;min=(N=3)=4KB, max=(N=23)=4GB
  6-11  Reserved     (0)
  12-31 Region Base  (Base = X SHL 12)  ;Base must be Size-aligned
The ITCM region base may be fixed (read-only). The Virtual size settings should be normally same as the Physical sizes (see C0,C0,2). However, smaller sizes are allowed (using only the 1st some KB), as well as bigger sizes (TCM area is then filled with mirrors of physical TCM).

TCM and PU
TCM can be used without Protection Unit.
When the protection unit is enabled, TCM is controlled by the PU just like normal memory, the PU should provide R/W Access Permission for TCM regions; cache and write-buffer are not required for high-speed TCM (so both should be disabled for TCM regions).


 CPU Instruction Cycle Times

Instruction Cycle Summary

  Instruction      Cycles      Additional
  ---------------------------------------------------------------------
  Data Processing  1S          +1S+1N if R15 loaded, +1I if SHIFT(Rs)
  MSR,MRS          1S
  LDR              1S+1N+1I    +1S+1N if R15 loaded
  STR              2N
  LDM              nS+1N+1I    +1S+1N if R15 loaded
  STM              (n-1)S+2N
  SWP              1S+2N+1I
  BL (THUMB)       3S+1N
  B,BL             2S+1N
  SWI,trap         2S+1N
  MUL              1S+ml
  MLA              1S+(m+1)I
  MULL             1S+(m+1)I
  MLAL             1S+(m+2)I
  CDP              1S+bI
  LDC,STC          (n-1)S+2N+bI
  MCR              1N+bI+1C
  MRC              1S+(b+1)I+1C
  {cond} false     1S
Whereas,
  n = number of words transferred
  b = number of cycles spent in coprocessor busy-wait loop
  m = depends on most significant byte(s) of multiplier operand
Above 'trap' is meant to be the execution time for exceptions. And '{cond} false' is meant to be the execution time for conditional instructions which haven't been actually executed because the condition has been false.

The separate meaning of the N,S,I,C cycles is:

N - Non-sequential cycle
Requests a transfer to/from an address which is NOT related to the address used in the previous cycle. (Called 1st Access in GBA language).
The execution time for 1N is 1 clock cycle (plus non-sequential access waitstates).

S - Sequential cycle
Requests a transfer to/from an address which is located directly after the address used in the previous cycle. Ie. for 16bit or 32bit accesses at incrementing addresses, the first access is Non-sequential, the following accesses are sequential. (Called 2nd Access in GBA language).
The execution time for 1S is 1 clock cycle (plus sequential access waitstates).

I - Internal Cycle
CPU is just too busy, not even requesting a memory transfer for now.
The execution time for 1I is 1 clock cycle (without any waitstates).

C - Coprocessor Cycle
The CPU uses the data bus to communicate with the coprocessor (if any), but no memory transfers are requested.

Memory Waitstates
Ideally, memory may be accessed free of waitstates (1N and 1S are then equal to 1 clock cycle each). However, a memory system may generate waitstates for several reasons: The memory may be just too slow. Memory is currently accessed by DMA, eg. sound, video, memory transfers, etc. Or when data is squeezed through a 16bit data bus (in that special case, 32bit access may have more waitstates than 8bit and 16bit accesses). Also, the memory system may separate between S and N cycles (if so, S cycles would be typically faster than N cycles).

Memory Waitstates for Different Memory Areas
Different memory areas (eg. ROM and RAM) may have different waitstates. When executing code in one area which accesses data in another area, then the S+N cycles must be split into code and data accesses: 1N is used for data access, plus (n-1)S for LDM/STM, the remaining S+N are code access. If an instruction jumps to a different memory area, then all code cycles for that opcode are having waitstate characteristics of the NEW memory area (except Thumb BL which still executes 1S in OLD area).


 CPU Versions

Version Numbers
ARM CPUs are distributed by name ARM#, and are described as ARMv# in specifications, whereas "#" is NOT the same than "v#", for example, ARM7TDMI is ARMv4TM. That is so confusing, that ARM didn't even attempt to clarify the relationship between the various "#" and "v#" values.

Version Variants
Suffixes like "M" (long multiply), "T" (Thumb support), "E" (Enhanced DSP) indicate presence of special features, additionally to the standard instruction set of a given version, or, when preceded by an "x", indicate the absence of that features.

ARMv1 aka ARM1
Some sort of a beta version, according to ARM never been used
in any commercial products.

ARMv2 and up
MUL,MLA
CDP,LDC,MCR,MRC,STC
SWP/SWPB (ARMv2a and up only)
Two new FIQ registers

ARMv3 and up
MRS,MSR opcodes (instead CMP/CMN/TST/TEQ{P} opcodes)
CPSR,SPSR registers (instead PSR bits in R15)
Removed never condition, cond=NV no longer valid
32bit addressing (instead 26bit addressing in older versions)
26bit addressing backwards comptibility mode (except v3G)
Abt and Und modes (instead handling aborts/undefined in Svc mode)
SMLAL,SMULL,UMLAL,UMULL (optionally, INCLUDED in v3M, EXCLUDED in v4xM/v5xM)

ARMv4 aka ARM7 and up
LDRH,LDRSB,LDRSH,STRH
Sys mode (privileged user mode)
BX (only ARMv4T, and any ARMv5 or ARMv5T and up)
THUMB code (only T variants, ie. ARMv4T, ARMv5T)

ARMv5 and up
BKPT,BLX,CLZ (BKPT,BLX also in THUMB mode)
LDM/LDR/POP PC with mode switch (POP PC also in THUMB mode)
CDP2,LDC2,MCR2,MRC2,STC2 (new coprocessor opcodes)
C-flag unchanged by MUL (instead undefined flag value)
changed instruction cycle timings / interlock ??? or not ???
QADD,QDADD,QDSUB,QSUB opcodes, CPSR.Q flag (v5TE and V5TExP only)
SMLAxy,SMLALxy,SMLAWy,SMULxy,SMULWy (v5TE and V5TExP only)
LDRD,STRD,PLD,MCRR,MRRC (v5TE only, not v5, not v5TExP)

ARMv6
No public specifications available.

A Milestone in Computer History
Original ARMv2 has been used in the relative rare and expensive Archimedes deluxe home computer series in the late eighties, the Archimedes has caught a lot of attention, particularly for being the first home computer that used a BIOS being programmed in BASIC language - which has been a absolutely revolutionary decadency at that time.
Inspired, programmers all over the world have successfully developed even slower and much more inefficient programming languages, which are nowadays consequently used by nearly all ARM programmers, and by most non-ARM programmers as well.


 CPU Data Sheet

This present document is an attempt to supply a brief ARM7TDMI reference, hopefully including all information which is relevant for programmers.

Some details that I have treated as meaningless for GBA programming aren't included - such like Big Endian format, and Virtual Memory data aborts, and most of the chapters listed below.

Have a look at the complete data sheet (URL see below) for more detailed verbose information about ARM7TDMI instructions. That document also includes:

- Signal Description
  Pins of the original CPU, probably other for GBA.
- Memory Interface
  Optional virtual memory circuits, etc. not for GBA.
- Coprocessor Interface
  As far as I know, none such in GBA.
- Debug Interface
  For external hardware-based debugging.
- ICEBreaker Module
  For external hardware-based debugging also.
- Instruction Cycle Operations
  Detailed: What happens during each cycle of each instruction.
- DC Parameters (Power supply)
- AC Parameters (Signal timings)

The official ARM7TDMI data sheet can be downloaded from ARMs webpage,
  http://www.arm.com/Documentation/UserMans/PDF/ARM7TDMI.html
Be prepared for bloated PDF Format, approx 1.3 MB, about 200 pages.


 NDS Reference

DS I/O Maps
DS Memory Maps
DS Memory Control
DS Video
DS Sound
DS Various
DS DMA Transfers
DS Timers
DS Interrupts
DS Maths
DS Keypad
DS Inter Process Communication (IPC)
DS Real-Time Clock (RTC)
DS Serial Peripheral Interface Bus (SPI)
DS Touch Screen Controller (TSC)
DS Power Management
DS Cartridges, Encryption, Firmware
DS Xboo
DS Backwards-compatible GBA-Mode
BIOS Functions
CPU Reference


 DS I/O Maps

ARM9 I/O Map
Display Engine A
  4000000h  56h  2D Engine A (same registers as GBA, some changed bits)
  4000060h  2    DISP3DCNT - ?
  4000064h  4    DISPCAPCNT - Display Capture Control Register (R/W)
  4000068h  4    DISP_MMEM_FIFO - Main Memory Display FIFO (R?/W)
  400006Ch  2    MASTER_BRIGHT - Master Brightness Up/Down
DMA, Timers, and Keypad
  40000B0h  30h  DMA Channel 0..3
  40000E0h  10h  DMA FILL Registers for Channel 0..3
  4000100h  10h  Timers 0..3
  4000130h  2    KEYINPUT
  4000132h  2    KEYCNT
IPC/ROM
  4000180h  2  IPCSYNC - IPC Synchronize Register (R/W)
  4000184h  2  IPCFIFOCNT - IPC Fifo Control Register (R/W)
  4000188h  4  IPCFIFOSEND - IPC Send Fifo (W)
  40001A1h  1  rom... - undoc
  40001A4h  4  romctrl - undoc
  40001A8h  8  romcmd - undoc
  40001B0h     romcrypt - (not sure if encryption can be accessed by arm9...?)
Memory and IRQ Control
  4000204h  2  EXMEMCNT - External Memory Control (R/W)
  4000208h  2  IME - Interrupt Master Enable (R/W)
  4000210h  4  IE  - Interrupt Enable (R/W)
  4000214h  4  IF  - Interrupt Request Flags (R/W)
  4000240h  1  VRAMCNT_A - VRAM-A (128K) Bank Control (W)
  4000241h  1  VRAMCNT_B - VRAM-B (128K) Bank Control (W)
  4000242h  1  VRAMCNT_C - VRAM-C (128K) Bank Control (W)
  4000243h  1  VRAMCNT_D - VRAM-D (128K) Bank Control (W)
  4000244h  1  VRAMCNT_E - VRAM-E (64K) Bank Control (W)
  4000245h  1  VRAMCNT_F - VRAM-F (16K) Bank Control (W)
  4000246h  1  VRAMCNT_G - VRAM-G (16K) Bank Control (W)
  4000247h  1  WRAMCNT   - WRAM Bank Control (W)
  4000248h  1  VRAMCNT_H - VRAM-H (32K) Bank Control (W)
  4000249h  1  VRAMCNT_I - VRAM-I (16K) Bank Control (W)
Maths
  4000280h  2  DIVCNT - Division Control (R/W)
  4000290h  8  DIV_NUMER - Division Numerator (R/W)
  4000298h  8  DIV_DENOM - Division Denominator (R/W)
  40002A0h  8  DIV_RESULT - Division Quotient (=Numer/Denom) (R/W?)
  40002A8h  8  DIVREM_RESULT - Division Remainder (=Numer MOD Denom) (R/W?)
  40002B0h  2  SQRTCNT - Square Root Control (R/W)
  40002B4h  4  SQRT_RESULT - Square Root Result (R/W?)
  40002B8h  8  SQRT_PARAM - Square Root Parameter Input (R/W)
  4000300h  4  POSTFLG - Undoc
  4000304h  2  POWCNT1 - Graphics Power Control Register (R/W)
3D Display Engine
  4000320h..6A3h
Display Engine B
  4001000h  56h  2D Engine B (same registers as GBA, some changed bits)
                 (above Engine B probably excludes separate DISPSTAT/VCOUNT?)
  400106Ch  2    DB_MASTER_BRIGHT - 16bit - Master Brightness Up/Down

  4100000h  4    IPCFIFORECV - IPC Receive Fifo (R)
  4100010h  4    undoc
Main Memory Control
  27FFFFEh  2    Main Memory Control
Further Memory Control Registers
ARM CP15 System Control Coprocessor

ARM7 I/O Map
  4000004h  2   DISPSTAT
  4000006h  2   VCOUNT
  40000B0h  30h DMA Channels 0..3
  4000100h  10h Timers 0..3
  4000120h  4   debug siodata32
  4000128h  4   debug siocnt
  4000130h  2   keyinput
  4000132h  2   keycnt
  4000134h  2   debug rcnt
  4000136h  2   EXTKEYIN
  4000138h  1   RTC Realtime Clock Bus
  4000180h  2   IPCSYNC - IPC Synchronize Register (R/W)
  4000184h  2   IPCFIFOCNT - IPC Fifo Control Register (R/W)
  4000188h  4   IPCFIFOSEND - IPC Send Fifo (W)
  40001A1h  1   Gamecard bus whatever
  40001A4h  4   Gamecard bus timing/control
  40001A8h  8   Gamecard bus 8-byte command out
  40001B0h  4   Gamecard Encryption
  40001B4h  4   Gamecard Encryption
  40001B8h  2   Gamecard Encryption
  40001BAh  2   Gamecard Encryption
  40001C0h  2   Firmware SPI bus Control
  40001C2h  2   Firmware SPI bus Data
  4000204h  2   EXMEMSTAT - External Memory Status
  4000206h  2   Unknown (set to 0030h) maybe bug, or WLAN/POWCNT related?
  4000208h  4   IME
  4000210h  4   IE
  4000214h  4   IF
  4000240h  1   VRAMSTAT - VRAM-C,D Bank Status (R)
  4000241h  1   WRAMSTAT - WRAM Bank Status (R)
  4000300h  1   POSTFLG
  4000301h  1   HALTCNT (different bits than on GBA) (plus NOP delay)
  4000304h  2   POWCNT2  Sound/Wifi Power Control Register (R/W)
  4000308h  4   BIOSPROT - Bios-data-read-protection address
Sound Registers
  4000400h 100h Sound Channel 0..15 (10h bytes each)
  40004x0h  4  SOUNDxCNT - Sound Channel X Control Register (R/W)
  40004x4h  4  SOUNDxSAD - Sound Channel X Data Source Register (W)
  40004x8h  2  SOUNDxTMR - Sound Channel X Timer Register (W)
  40004xAh  2  SOUNDxPNT - Sound Channel X Loopstart Register (W)
  40004xCh  4  SOUNDxLEN - Sound Channel X Length Register (W)
  4000500h  2  SOUNDCNT - Sound Control Register (R/W)
  4000504h  2  SOUNDBIAS - Sound Bias Register (R/W)
  4000508h  1  SNDCAP0CNT - Sound Capture 0 Control Register (R/W)
  4000509h  1  SNDCAP1CNT - Sound Capture 1 Control Register (R/W)
  4000510h  4  SNDCAP0DAD - Sound Capture 0 Destination Address (W?)
  4000514h  2  SNDCAP0LEN - Sound Capture 0 Length (R/W)
  4000518h  4  SNDCAP1DAD - Sound Capture 1 Destination Address (W?)
  400051Ch  2  SNDCAP1LEN - Sound Capture 1 Length (R/W)
gamecart...
  4100000h  4   IPCFIFORECV - IPC Receive Fifo (R)
  4100010h  4   Gamecard bus 4-byte data in, for manual or dma read
WLAN Registers
  4808036h  2   W
  4808158h  2   W
  480815Ah  2   W
  480815Ch  2   R
  480815Eh  2   R
  4808160h  2   W
  4808168h  2   W
  480817Ch  2   W
  480817Eh  2   W
  4808180h  2   R
  4808184h  2   W


 DS Memory Maps

ARM9
  01000000h  Instruction TCM (32KB) (moveable)
  02000000h  Main Memory (4MB)
  027C0000h  Data TCM (16KB) (moveable)
  03000000h  Shared WRAM (0KB, 16KB, or 32KB can be allocated to ARM9)
  04000000h  ARM9-I/O Ports
  05000000h  Standard Palettes (2KB) (Engine A BG/OBJ, Engine B BG/OBJ)
  06000000h  VRAM - Engine A, BG VRAM (max 512KB)
  06200000h  VRAM - Engine B, BG VRAM (max 128KB)
  06400000h  VRAM - Engine A, OBJ VRAM (max 256KB)
  06600000h  VRAM - Engine B, OBJ VRAM (max 128KB)
  06800000h  VRAM - "LCDC"-allocated (max 656KB)
  07000000h  OAM (2KB) (Engine A, Engine B)
  08000000h  GBA Slot ROM (max. 32MB)
  0A000000h  GBA Slot RAM (max. 64KB)
  FFFF0000h  ARM9-BIOS (32KB) (only 3K used)
The ARM9 Exception Vectors are located at FFFF0000h. The IRQ handler redirects to [DTCM+3FFCh].

ARM7
  00000000h  ARM7-BIOS (16KB)
  02000000h  Main Memory (4MB)
  03000000h  Shared WRAM (0KB, 16KB, or 32KB can be allocated to ARM7)
  03800000h  ARM7-WRAM (64KB)
  04000000h  ARM7-I/O Ports
  04800000h  Wireless Communications Wait State 0
  04808000h  Wireless Communications Wait State 1
  06000000h  VRAM allocated as Work RAM to ARM7 (max. 256K)
  08000000h  GBA Slot ROM (max. 32MB)
  0A000000h  GBA Slot RAM (max. 64KB)
The ARM7 Exception Vectors are located at 00000000h. The IRQ handler redirects to [3FFFFFCh aka 380FFFCh].

Further Memory (not mapped to ARM9/ARM7 bus)
  3D Engine Polygon RAM (52KBx2)
  3D Engine Vertex RAM (72KBx2)
  Firmware (256KB) (built-in serial flash memory)
  GBA-BIOS (16KB) (not used in NDS mode)
  NDS Slot ROM (serial 8bit-bus, max. 4GB with default protocol)
  NDS Slot EEPROM (serial 1bit-bus)

Shared-RAM
Even though Shared WRAM begins at 3000000h, programs are commonly using mirrors at 37F8000h (both ARM9 and ARM7). At the ARM7-side, this allows to use 32K Shared WRAM and 64K ARM7-WRAM as a continous 96K RAM block.

Undefined I/O Ports
On the NDS (at the ARM9-side at least) undefined I/O ports are always zero.

Undefined Memory Regions
16MB blocks that do not contain any defined memory regions (or that contain only mapped TCM regions) are typically completely undefined.
16MB blocks that do contain valid memory regions are typically containing mirrors of that memory in the unused upper part of the 16MB area (only exceptions are TCM and BIOS which are not mirrored).


 DS Memory Control

DS Memory Control - Cache and TCM
DS Memory Control - Cartridges and Main RAM
DS Memory Control - WRAM
DS Memory Control - VRAM
DS Memory Control - BIOS


 DS Memory Control - Cache and TCM

TCM and Cache are controlled by the System Control Coprocessor,
ARM CP15 System Control Coprocessor

The specifications for the NDS9 are:

Tightly Coupled Memory (TCM)
  ITCM 32K (default address 1000000h)
  DTCM 16K (default address 27C0000h)

Cache
  Data Cache 4KB, Instruction Cache 8KB
  4-way set associative method
  Cache line 8 words (32 bytes)
  Read-allocate method (ie. writes are not allocating cache lines)
  Round-robin and Pseudo-random replacement algorithms selectable
  Cache Lockdown, Instruction Prefetch, Data Preload
  Data write-through and write-back modes selectable

Protection Unit (PU)
Recommended/default settings are:
  Region  Name            Address   Size   Cache WBuf Code Data
  -       Background      00000000h 4GB    -     -    -    -
  0       I/O and VRAM    04000000h 64MB   -     -    R/W  R/W
  1       Main Memory     02000000h 4MB    On    On   R/W  R/W
  2       ARM7-dedicated  027C0000h 256KB  -     -    -    -
  3       GBA Slot        08000000h 128MB  -     -    -    R/W
  4       DTCM            027C0000h 16KB   -     -    -    R/W
  5       ITCM            01000000h 32KB   -     -    R/W  R/W
  6       BIOS            FFFF0000h 32KB   On    -    R    R
  7       Shared Work     027FF000h 4KB    -     -    -    R/W
Notes: In Nintendos hardware-debugger, Main Memory is expanded to 8MB (for that reason, some addresses are at 27NN000h instead 23NN000h) (some of the extra memory is reserved for the debugger, some can be used for game development). Region 2 and 7 are not understood? GBA Slot should be max 32MB+64KB, rounded up to 64MB, no idea why it is 128MB? DTCM and ITCM do not use Cache and Write-Buffer because TCM is fast. Above settings do not allow to access Shared Memory at 37F8000h? Do not use cache/wbuf for I/O, doing so might suppress writes, and/or might read outdated values.
The main purpose of the Protection Unit is debugging, a major problem with GBA programs have been faulty accesses to memory address 00000000h and up (due to [base+offset] addressing with uninitialized (zero) base values). This problem has been fixed in the NDS, for the ARM9 processor at least, still there are various leaks: For example, the 64MB I/O and VRAM area contains only ca. 660KB valid addresses, and the ARM7 probably doesn't have a Protection Unit at all. Alltogether, the protection is better than in GBA, but it's still pretty crude compared with software debugging tools.
Region address/size are unified (same for code and data), however, cachabilty and access rights are non-unified (and may be separately defined for code and data).

Note: The NDS7 doesn't have any TCM, Cache, or CP15.


 DS Memory Control - Cartridges and Main RAM

4000204h - NDS9 - EXMEMCNT - 16bit - External Memory Control (R/W)
4000204h - NDS7 - EXMEMSTAT - 16bit - External Memory Status (R/W..R)
  0-1   32-pin GBA Slot SRAM Access Waitstate    (0-3 = 10, 8, 6, 18 cycles)
  2-3   32-pin GBA Slot ROM 1st Access Waitstate (0-3 = 10, 8, 6, 18 cycles)
  4     32-pin GBA Slot ROM 2nd Access Waitstate (0-1 = 6, 4 cycles)
  5-6   32-pin GBA Slot PHI-pin out   (0-3 = Low, 4.19MHz, 8.38MHz, 16.76MHz)
  7     32-pin GBA Slot Access Rights     (0=ARM9, 1=ARM7)
  8-10  Not used (always zero)
  11    17-pin NDS Slot Access Rights     (0=ARM9, 1=ARM7)
  12    Not used (always zero)
  13    Not used (always set ?)
  14    Main Memory Interface Mode Switch (0=Async/GBA/Reserved, 1=Synchronous)
  15    Main Memory Access Priority       (0=ARM9 Priority, 1=ARM7 Priority)
Bit0-6 can be changed by both NDS9 and NDS7, changing these bits affects the local EXMEM register only, not that of the other CPU.
Bit7-15 can be changed by NDS9 only, changing these bits affects both EXMEM registers, ie. both NDS9 and NDS7 can read the current NDS9 setting.
Bit14=0 is intended for GBA mode, however, writes to this bit appear to be ignored?
DS Main Memory Control


 DS Memory Control - WRAM

4000247h - NDS9 - WRAMCNT - 8bit - WRAM Bank Control (R/W)
4000241h - NDS7 - WRAMSTAT - 8bit - WRAM Bank Status (R)
Should not be changed when using Nintendo's API.
  0-1   ARM9/ARM7 (0-3 = 32K/0K, 2nd 16K/1st 16K, 1st 16K/2nd 16K, 0K/32K)
  2-7   Not used
The ARM9 WRAM area is 3000000h-3FFFFFFh (16MB range).
The ARM7 WRAM area is 3000000h-37FFFFFh (8MB range).
The allocated 16K or 32K are mirrored everywhere in the above areas.
De-allocation (0K) is a special case: At the ARM9-side, the WRAM area is then empty (containing undefined data). At the ARM7-side, the WRAM area is then containing mirrors of the 64KB ARM7-WRAM (the memory at 3800000h and up).


 DS Memory Control - VRAM

4000240h - NDS7 - VRAMSTAT - 8bit - VRAM Bank Status (R)
  0     VRAM C enabled and allocated to NDS7  (0=No, 1=Yes)
  1     VRAM D enabled and allocated to NDS7  (0=No, 1=Yes)
  2-7   Not used (always zero)
The register indicates if VRAM C/D are allocated to NDS7 (as Work RAM), ie. if VRAMCNT_C/D are enabled (Bit7=1), with MST=2 (Bit0-2). However, it does not reflect the OFS value.

4000240h - VRAMCNT_A - 8bit - VRAM-A (128K) Bank Control (W)
4000241h - VRAMCNT_B - 8bit - VRAM-B (128K) Bank Control (W)
4000242h - VRAMCNT_C - 8bit - VRAM-C (128K) Bank Control (W)
4000243h - VRAMCNT_D - 8bit - VRAM-D (128K) Bank Control (W)
4000244h - VRAMCNT_E - 8bit - VRAM-E (64K) Bank Control (W)
4000245h - VRAMCNT_F - 8bit - VRAM-F (16K) Bank Control (W)
4000246h - VRAMCNT_G - 8bit - VRAM-G (16K) Bank Control (W)
4000248h - VRAMCNT_H - 8bit - VRAM-H (32K) Bank Control (W)
4000249h - VRAMCNT_I - 8bit - VRAM-I (16K) Bank Control (W)
  0-2   VRAM MST              ;Bit2 not used by VRAM-A,B,H,I
  3-4   VRAM Offset (0-3)     ;Offset not used by VRAM-E,H,I
  5-6   Not used
  7     VRAM Enable (0=Disable, 1=Enable)
There is a total of 656KB of VRAM in Blocks A-I.
Table below shows the possible configurations.
  VRAM    SIZE  MST  OFS   ARM9, Plain ARM9-CPU Access (so-called LCDC mode)
  A       128K  0    -     6800000h-681FFFFh
  B       128K  0    -     6820000h-683FFFFh
  C       128K  0    -     6840000h-685FFFFh
  D       128K  0    -     6860000h-687FFFFh
  E       64K   0    -     6880000h-688FFFFh
  F       16K   0    -     6890000h-6893FFFh
  G       16K   0    -     6894000h-6897FFFh
  H       32K   0    -     6898000h-689FFFFh
  I       16K   0    -     68A0000h-68A3FFFh
  VRAM    SIZE  MST  OFS   ARM9, 2D Graphics Engine A, BG-VRAM (max 512K)
  A,B,C,D 128K  1    0..3  6000000h+(20000h*OFS)
  E       64K   1    -     6000000h
  F,G     16K   1    0..3  6000000h+(4000h*OFS.0)+(10000h*OFS.1)
  VRAM    SIZE  MST  OFS   ARM9, 2D Graphics Engine A, OBJ-VRAM (max 256K)
  A,B     128K  2    0..1  6400000h+(20000h*OFS.0)  ;(OFS.1 must be zero)
  E       64K   2    -     6400000h
  F,G     16K   2    0..3  6400000h+(4000h*OFS.0)+(10000h*OFS.1)
  VRAM    SIZE  MST  OFS   2D Graphics Engine A, BG Extended Palette
  E       64K   4    -     Slot 0-3  ;only lower 32K used
  F,G     16K   4    0..1  Slot 0-1 (OFS=0), Slot 2-3 (OFS=1)
  VRAM    SIZE  MST  OFS   2D Graphics Engine A, OBJ Extended Palette
  F,G     16K   5    -     Slot 0  ;16K each (only lower 8K used)
  VRAM    SIZE  MST  OFS   Texture Image
  A,B,C,D 128K  3    0..3  Slot OFS(0-3) ... or Slot2-3=Clear image ?
  VRAM    SIZE  MST  OFS   Texture Palette
  E       64K   3    -     Slots 0-3                 ;OFS=don't care
  F,G     16K   3    0..3  Slot (OFS.0*1)+(OFS.1*4)  ;ie. Slot 0, 1, 4, or 5
  VRAM    SIZE  MST  OFS   ARM9, 2D Graphics Engine B, BG-VRAM (max 128K)
  C       128K  4    -     6200000h
  H       32K   1    -     6200000h
  I       16K   1    -     6208000h
  VRAM    SIZE  MST  OFS   ARM9, 2D Graphics Engine B, OBJ-VRAM (max 128K)
  D       128K  4    -     6600000h
  I       16K   2    -     6600000h
  VRAM    SIZE  MST  OFS   2D Graphics Engine B, BG Extended Palette
  H       32K   2    -     Slot 0-3
  VRAM    SIZE  MST  OFS   2D Graphics Engine B, OBJ Extended Palette
  I       16K   3    -     Slot 0  ;(only lower 8K used)
  VRAM    SIZE  MST  OFS   <ARM7>, Plain <ARM7>-CPU Access
  C,D     128K  2    0..1  6000000h+(20000h*OFS.0)  ;OFS.1 must be zero

Notes
In Plain-CPU modes, VRAM can be accessed only by the CPU (and by the Capture Unit, and by VRAM Display mode). In "Plain <ARM7>-CPU Access" mode, the VRAM blocks are allocated as Work RAM to the NDS7 CPU.
In BG/OBJ VRAM modes, VRAM can be accessed by the CPU at specified addresses, and by the display controller.
In Extended Palette and Texture Image/Palette modes, VRAM is not mapped to CPU address space, and can be accessed only by the display controller (so, to initialize or change the memory, it should be temporarily switched to Plain-CPU mode).

Other Video RAM
Aside from the map-able VRAM blocks, there are also some video-related memory regions at fixed addresses:
  5000000h Engine A Standard BG Palette (512 bytes)
  5000200h Engine A Standard OBJ Palette (512 bytes)
  5000400h Engine B Standard BG Palette (512 bytes)
  5000600h Engine B Standard OBJ Palette (512 bytes)
  7000000h Engine A OAM (1024 bytes)
  7000400h Engine B OAM (1024 bytes)


 DS Memory Control - BIOS

4000308h - NDS7 - BIOSPROT - Bios-data-read-protection address
Used to double-protect the first some KBytes of the NDS7 BIOS. The BIOS is split into two protection regions, one always active, one controlled by the BIOSPROT register. The overall idea is that only the BIOS can read from itself, any other attempts to read from that regions return FFh-bytes.
  Opcodes at...      Can read from      Expl.
  0..[BIOSPROT]-1    0..3FFFh           Double-protected (when BIOSPROT is set)
  [BIOSPROT]..3FFFh  [BIOSPROT]..3FFFh  Normal-protected (always active)
The initial BIOSPROT setting on power-up is zero (disabled). Before starting the cartridge, the BIOS boot code sets the register to 1204h (actually 1205h, but the mis-aligned low-bit is ignored). Once when initialized, further writes to the register are ignored.

The double-protected region contains the exception vectors, some bytes of code, and the cartridge KEY1 encryption seed (about 4KBytes). As far as I know, it is impossible to unlock the memory once when it is locked, however, with some trickery, it is possible execute code before it gets locked. Also, the two THUMB opcodes at 05ECh can be used to read all memory at 0..3FFFh,
  05ECh  ldrb r3,[r3,12h]      ;requires incoming r3=src-12h
  05EEh  pop  r2,r4,r6,r7,r15  ;requires dummy values & THUMB retadr on stack
Additionally most BIOS functions (eg. CpuSet), include a software-based protection which rejects source addresses in the BIOS area (the only exception is GetCRC16, though it still cannot bypass the BIOSPROT setting).

Note
The NDS9 BIOS doesn't include any software or hardware based read protection.


 DS Video

LCD Video Controller

DS Video Stuff
DS Video BG Modes / Control
DS Video OBJs
DS Video Extended Palettes
DS Video Capture and Main Memory Display Mode
DS Video Display System Block Diagram

For Display Power Control (and Display Swap), and VRAM Allocation, see
DS Power Management
DS Memory Control - VRAM


 DS Video Stuff

400006Ch - (DB_)MASTER_BRIGHT - 16bit - Master Brightness Up/Down
Supported for both Display Engine A (Port 400006Ch) and B (Port 400106Ch).
  0-4   Factor used for 6bit R,G,B Intensities (0-16, values >16 same as 16)
          Brightness up:   New = Old + (63-Old) * Factor/16
          Brightness down: New = Old - Old      * Factor/16
  5-13  Not used
  14-15 Mode (0=Disable, 1=Up, 2=Down, 3=Reserved)

DISPSTAT/VCOUNT
The LY and LYC values are in range 0..262, so LY/LYC values have been expanded to 9bit values: LY = VCOUNT Bit 0..8, and LYC=DISPSTAT Bit8..15,7.
VCOUNT register is write-able, allowing to synchronize linked DS consoles.
For proper synchronization:
  write new LY values only in range of 202..212
  write only while old LY values are in range of 202..212

VRAM Waitstates
The display controller performs VRAM-reads once every 6 clock cycles, a 1 cycle waitstate is generated if the CPU simultaneously accesses VRAM.
With capture enabled, additionally VRAM-writes take place once every 6 cycles, so the total VRAM-read/write access rate is then once every 3 cycles.

DS Window Glitches
The DS counts scanlines in range 0..262 (0..106h), of which only the lower 8bit are compared with the WIN0V/WIN1V register settings. Respectively, Y1 coordinates 00h..06h will be triggered in scanlines 100h-106h by mistake. That means, the window gets activated within VBlank period, and will be active in scanline 0 and up (that is no problem with Y1=0, but Y1=1..6 will appear as if if Y1 would be 0). Workaround would be to disable the Window during VBlank, or to change Y1 during VBlank (to a value that does not occur during VBlank period, ie. 7..191).
Also, there's a problem to fit the 256 pixel horizontal screen resolution into 8bit values: X1=00h is treated as 0 (left-most), X2=00h is treated as 100h (right-most). However, the window is not displayed if X1=X2=00h; the window width can be max 255 pixels.

2D Engines
Includes two 2D Engines, called A and B. Both engines are accessed by the ARM9 processor, each using different memory and register addresses:
  Region______Engine A______________Engine B___________
  I/O Ports   4000000h              4001000h
  Palette     5000000h (1K)         5000400h (1K)
  BG VRAM     6000000h (max 512K)   6200000h (max 128K)
  OBJ VRAM    6400000h (max 256K)   6600000h (max 128K)
  OAM         7000000h (1K)         7000400h (1K)
Engine A additionally supports 3D and large-screen 256-color Bitmaps, plus main-memory-display and vram-display modes, plus capture unit.

Viewing Angles
The LCD screens are best viewed at viewing angles of 90 degrees. Colors may appear distorted, and may even become invisible at other viewing angles.
When the console is handheld, both screens can be turned into preferred direction. When the console is settled on a table, only the upper screen can be turned, but the lower screen is stuck into horizontal position - which results rather bad visibility (unless the user moves his/her head directly above of it).


 DS Video BG Modes / Control

DS DISPCNT
  0-2   BG Mode
  3     A:BG0 2D/3D Selection (instead CGB Mode) (0=2D, 1=3D)
  3     B:Not used            (instead CGB Mode)
  4     Tile OBJ Mapping  (0=2D; max 32KB, 1=1D; max 32KB..256KB; see Bit20-21)
  5-6   Bitmap OBJ Mapping
  7-15  Same as GBA
  16-17 A: Display Mode (instead Green Swap, Green swap still supp. in GBA mode)
  18-19 A: VRAM block (0..3=VRAM A..D) (For Capture, and above Display Mode=2)
  16    B: Display Mode (instead Green Swap, Green swap still supp. in GBA mode)
  17-19 B: Not used
  20-21 A/B: Ext OBJ CH
  22    A: Ext OBJ BM
  22    B: Not used
  23    A/B: OBJ Processing during H-Blank (was located in Bit5 on GBA)
  24-26 A: Character Base (in 64K steps) (merged with 16K step in BGxCNT)
  27-29 A: Screen Base (in 64K steps) (merged with 2K step in BGxCNT)
  24-29 B: Not used
  30-21 A/B: Ext Palette

BG Mode
Engine A BG Mode (DISPCNT LSBs) (0-6, 7=Reserved)
  Mode  BG0      BG1      BG2      BG3
  0     Text/3D  Text     Text     Text
  1     Text/3D  Text     Text     Affine
  2     Text/3D  Text     Affine   Affine
  3     Text/3D  Text     Text     Extended
  4     Text/3D  Text     Affine   Extended
  5     Text/3D  Text     Extended Extended
  6     3D       -        Large    -
Engine B: Same as above, except that: Mode 6 is reserved (no Large screen bitmap), and BG0 is always Text (no 3D support).
Affine = formerly Rot/Scal mode (with 8bit BG Map entries)
Extended Affine = can be 3 different types (selected in BGnCNT register):
 1) rot/scal with 16bit BG Map entries (mixup of Text and Affine modes)
 2) rot/scal 256 color bitmap
 3) rot/scal direct color bitmap
Large Screen Bitmap = rot/scal 256 color bitmap (using all 512K of 2D VRAM)

Display Mode (DISPCNT.16-17):
  0  Display off (screen becomes white)
  1  Graphics Display (normal BG and OBJ layers)
  2  Engine A only: VRAM Display (Bitmap from block selected in DISPCNT.18-19)
  3  Engine A only: Main Memory Display (Bitmap DMA transfer from Main RAM)
Mode 2-3 display a raw direct color bitmap (15bit RGB values, the upper bit in each halfword is unused), without any further BG,OBJ,3D layers, these modes are completely bypassing the 2D/3D engines as well as any 2D effects, however the Master Brightness effect can be applied to these modes. Mode 2 is particulary useful to display captured 2D/3D images (in that case it can indirectly use the 2D/3D engine).

BGxCNT
character base extended from bit2-3 to bit2-5 (bit4-5 formerly unused)
  engine A screen base: BGxCNT.bits*2K + DISPCNT.bits*64K
  engine B screen base: BGxCNT.bits*2K + 0
  engine A char base: BGxCNT.bits*16K + DISPCNT.bits*64K
  engine B char base: BGxCNT.bits*16K + 0
char base is used only in tile/map modes (not bitmap modes)
screen base is used in tile/map modes,
screen base used in bitmap modes as BGxCNT.bits*16K, without DISPCNT.bits*64K
screen base however NOT used at all for Large screen bitmap mode
  bgcnt size  text     rotscal    bitmap   large bmp
  0           256x256  128x128    128x128  512x1024
  1           512x256  256x256    256x256  1024x512
  2           256x512  512x512    512x256  -
  3           512x512  1024x1024  512x512  -
bitmaps that require more than 128K VRAM are supported on engine A only.

Extended Affine modes (selected in DISPCNT) are sub-selected in BGxCNT,
(extended affine doesn't include 16-color modes, so color depth bit can be used for mode selection. Also, bitmap modes do not use charbase, so charbase.0 can be used for mode selection as well)
  ColorDepth  charbase.bit0
  0           x (charbase)      rot/scal tile/map mode with 16bit entries
  1           0 (mode)          256 color bitmap
  1           1 (mode)          direct color bitmap

for BG0, BG1 only: bit13 selects extended palette slot
                   (BG0: 0=Slot0, 1=Slot2, BG1: 0=Slot1, 1=Slot3)

Direct Color Bitmap BG, and Direct Color Bitmap OBJ
BG/OBJ Supports 32K colors (15bit RGB value) - so far same as GBAs BG.
However, the upper bit (Bit15) is used as Alpha flag.


 DS Video OBJs

DS OBJ Priority
The GBA has been assigning OBJ priority in respect to the 7bit OAM entry number, regardless of the OBJs 2bit BG-priority attribute (which allowed to specify invalid priority orders). That problem has been fixed in DS mode by combining the above two values into a 9bit priority value.

OBJ Tile Mapping (DISPCNT.4,20-21):
  Bit4  Bit20-21  Dimension Boundary Total
  0     x         2D        32       32K   ;Same as GBA 2D Mapping
  1     0         1D        32       32K   ;Same as GBA 1D Mapping
  1     1         1D        64       64K
  1     2         1D        128      128K
  1     3         1D        256      256K (Engine B: 128K)
TileVramAddress = TileNumber * BoundaryValue
Even if the boundary gets changed, OBJs are kept composed of 8x8 tiles.

Bitmap OBJ Mapping (DISPCNT.5-6,22):
  Bit5-6 Bit22  Dimension    Boundary Total
  0      x      2D/128 dots  32 bytes   32K  (Source Bitmap width 128 dots)
  1      x      2D/256 dots  32 bytes   32K  (Source Bitmap width 256 dots)
  2      0      1D           128 bytes  128K (Source Width = Target Width)
  2      1      1D           256 bytes  256K (Engine A only)
  3      x      Reserved
BitmapVramAddress = TileNumber * BoundaryValue

OBJ Attribute 0 and 2
Setting the OBJ Mode bits (Attr 0, Bit10-11) to a value of 3 has been prohibited in GBA, however, in NDS it selects the the new Bitmap OBJ mode; in that mode, the Color depth bit (Attr 0, Bit13) should be set to zero; also in that mode, the color bits (Attr 2, Bit 12-15) are used as Alpha-OAM value (instead of as palette setting).

OBJ Vertical Wrap
On the GBA, a large OBJ located near the bottom of the screen has been wrapped to the top of the screen (and was NOT displayed at the bottom. This problem has been "corrected" in the NDS (except in GBA mode), presumably by displaying the OBJ only at the screen bottom, or BOTH wrapped at top and bottom?


 DS Video Extended Palettes

Extended Palettes
When allocating extended palettes, the allocated memory is not mapped to the CPU bus, so the CPU can access extended palette only when temporarily de-allocating it.

BG Extended Palette enabled in DISPCNT Bit 31, when enabled,
 standard palette --> used for 16x16 color tiles, and 256 color bitmaps,
                      256 color tiles with 8bit bgmap entries (rot/scal mode)
 extended palette --> used for 256x16 color tiles (16bit bgmap entries)
color 0 of all standard/extended palettes is transparent, color 0 of BG standard palette 0 is used as backdrop. extended palette memory must be allocated to VRAM. Allocated VRAM is split into 4 slots of 8K each (32K used in total), normally BG0..3 are using Slot 0..3, however BG0 and BG1 can be optionally changed to BG0=Slot2, and BG1=Slot3 via BG0CNT and BG1CNT.

OBJ Extended Palette enabled in DISPCNT Bit 30, when enabled,
 16 colors x 16 palettes --> standard palette memory (=256 colors)
 256 colors x 16 palettes --> extended palette memory (=4096 colors)
color 0 of all standard/extended palettes is transparent, color 0 of BG standard palette 0 is used as backdrop. extended palette memory must be allocated to VRAM F, G, or I (which are 16K) of which only the first 8K are used for extended palettes (=1000h 16bit entries).


 DS Video Capture and Main Memory Display Mode

4000064h - DISPCAPCNT - 32bit - Display Capture Control Register (R/W)
Capture is supported for Display Engine A only.
  0-4   EVA               (0..16 = Blending Factor for Source A)
  5-7   Not used
  8-12  EVB               (0..16 = Blending Factor for Source B)
  13-15 Not used
  16-17 VRAM Write Block  (0..3 = VRAM A..D) (VRAM must be allocated to LCDC)
  18-19 VRAM Read Offset  (0=00000h, 0=08000h, 0=10000h, 0=18000h)
  20-21 Capture Size      (0=128x128, 1=256x64, 2=256x128, 3=256x192 dots)
  22-23 Not used
  24    Source A          (0=Graphics Screen BG+3D+OBJ, 1=3D Screen)
  25    Source B          (0=VRAM, 1=Main Memory Display FIFO)
  26-27 VRAM Write Offset (0=00000h, 0=08000h, 0=10000h, 0=18000h)
  28    Not used
  29-30 Capture Source    (0=Source A, 1=Source B, 2/3=Sources A+B blended)
  31    Capture Enable    (0=Disable/Ready, 1=Enable/Busy)
Notes:
VRAM Read Block (VRAM A..D) is selected in DISPCNT Bits 18-19.
VRAM Read Block can be (or must be ?) allocated to LCDC (MST=0).
VRAM Read Offset is ignored (zero) in VRAM Display Mode (DISPCNT.16-17).
VRAM Read/Write Offsets wrap to 00000h when exceeding 1FFFFh (max 128K).
After setting the Capture Enable bit, capture starts at next line 0.
The capture enable bit is automatically cleared at end of capture.

Capture data is 15bit color depth (even when capturing 18bit 3D-images).
Capture A: Dest_Intensity = SrcA_Intensitity ; Dest_Alpha=SrcA_Alpha.
Capture B: Dest_Intensity = SrcB_Intensitity ; Dest_Alpha=SrcB_Alpha.
Capture A+B (blending):
 Dest_Intensity = (  (SrcA_Intensitity * SrcA_Alpha * EVA)
                   + (SrcB_Intensitity * SrcB_Alpha * EVB) ) / 16
 Dest_Alpha = (SrcA_Alpha AND (EVA>0)) OR (SrcB_Alpha AND EVB>0))

Capture provides a couple of interesting effects.
For example, 3D Engine output can be captured via source A (to LCDC-allocated VRAM), in the next frame, either Graphics Engine A or B can display the captured 3D image in VRAM image as BG2, BG3, or OBJ (from BG/OBJ-allocated VRAM); this method requires to switch between LCDC- and BG/OBJ-allocation.
Another example would be to capture Engine A output, the captured image can be displayed (via VRAM Display mode) in the following frames, simultaneously the new Engine A output can be captured, blended with the old captured image; in that mode moved objects will leave traces on the screen; this method works with a single LCDC-allocated VRAM block.
DS Video Display System Block Diagram

4000068h - DISP_MMEM_FIFO - 32bit - Main Memory Display FIFO (R?/W)
Intended to send 256x192 pixel 32K color bitmaps by DMA directly
 - to Screen A             (set DISPCNT to Main Memory Display mode), or
 - to Display Capture unit (set DISPCAPCNT to Main Memory Source).
The FIFO can receive 4 words (8 pixels) at a time, each pixel is a 15bit RGB value (the upper bit, bit15, is unused).
Set DMA to Main Memory mode, 32bit transfer width, word count set to 4, destination address to DISP_MMEM_FIFO, source address must be in Main Memory.
Transfer starts at next frame.
Main Memory Display/Capture is supported for Display Engine A only.


 DS Video Display System Block Diagram
             _____________               __________
  VRAM A -->| 2D Graphics |--------OBJ->|          |
  VRAM B -->| Engine A    |--------BG3->| Layering |
  VRAM C -->|             |--------BG2->| and      |
  VRAM D -->|             |--------BG1->| Special  |
  VRAM E -->|             |   ___       | Effects  |
  VRAM F -->|             |->|SEL|      |          |          ______
  VRAM G -->| - - - - - - |  |BG0|-BG0->|          |----+--->|      |
            | 3D Graphics |->|___|      |__________|    |    |Select|
            | Engine      |                             |    |Video |
            |_____________|--------3D----------------+  |    |Input |
             _______      _______              ___   |  |    |      |
            |       |    |       |<-----------|SEL|<-+  |    |and   |-->
            |       |    |       |    _____   |A  |     |    |      |
  VRAM A <--|Select |    |Select |   |     |<-|___|<----+    |Master|
  VRAM B <--|Capture|<---|Capture|<--|Blend|   ___           |Bright|
  VRAM C <--|Dest.  |    |Source |   |_____|<-|SEL|<----+    |A     |
  VRAM D <--|       |    |       |            |B  |     |    |      |
            |_______|    |_______|<-----------|___|<-+  |    |      |
             _______                                 |  |    |      |
  VRAM A -->|Select |                                |  |    |      |
  VRAM B -->|Display|--------------------------------+------>|      |
  VRAM C -->|VRAM   |                                   |    |      |
  VRAM D -->|_______|   _____________                   |    |      |
                       |Main Memory  |                  |    |      |
  Main   ------DMA---->|Display FIFO |------------------+--->|______|
  Memory               |_____________|
             _____________               __________           ______
  VRAM C -->| 2D Graphics |--------OBJ->| Layering |         |      |
  VRAM D -->| Engine B    |--------BG3->| and      |         |Master|
  VRAM H -->|             |--------BG2->| Special  |-------->|Bright|-->
  VRAM I -->|             |--------BG1->| Effects  |         |B     |
            |_____________|--------BG0->|__________|         |______|


 DS Sound

The DS contains 16 hardware sound channels.
The console contains two speakers, arranged left and right of the upper screen, and so, provides stereo sound even without using the headphone socket.

DS Sound Channels 0..15
DS Sound Control Registers
DS Sound Capture
DS Sound Block Diagrams
DS Sound Notes

Power control
When restoring power supply to the sound circuit, do not output any sound during the first 15 milliseconds.


 DS Sound Channels 0..15

Each of the 16 sound channels occopies 16 bytes in the I/O region, starting with channel 0 at 4000400h..400040Fh, up to channel 15 at 40004F0h..40004FFh.

40004x0h - ARM7 - SOUNDxCNT - Sound Channel X Control Register (R/W)
  Bit0-6    Volume       (0..127=silent..loud)
  Bit7      Not used     (always zero)
  Bit8-9    Data Shift   (0=Normal, 1=Div2, 2=Div4, 3=Div16)
  Bit10-14  Not used     (always zero)
  Bit15     Hold         (0=Nothing, 1=Hold)               (?)
  Bit16-22  Panning      (0..127=left..right) (64=half volume on both speakers)
  Bit23     Not used     (always zero)
  Bit24-26  Wave Duty    (0..7) ;HIGH=(N+1)*12.5%, LOW=(7-N)*12.5% (PSG only)
  Bit27-28  Repeat Mode  (0=Manual, 1=Loop Infinite, 2=One-Shot, 3=Prohibited)
  Bit29-30  Format       (0=PCM8, 1=PCM16, 2=IMA-ADPCM, 3=PSG/Noise)
  Bit31     Start/Status (0=Stop, 1=Start/Busy)
All channels support ADPCM/PCM formats, PSG rectangular wave can be used only on channels 8..13, and white noise only on channels 14..15.

40004x4h - ARM7 - SOUNDxSAD - Sound Channel X Data Source Register (W)
  Bit0-26  Source Address
  Bit27-31 Not used

40004x8h - ARM7 - SOUNDxTMR - Sound Channel X Timer Register (W)
  Bit0-15  Timer Value, Sample frequency, timerval=-(16777216 / freq)
The PSG Duty Cycles are composed of eight "samples", and so, the frequency for Rectangular Wave is 1/8th of the selected sample frequency.
For PSG Noise, the noise frequency is equal to the sample frequency.

40004xAh - ARM7 - SOUNDxPNT - Sound Channel X Loopstart Register (W)
  Bit0-15  Loop Start, Sample loop start position

40004xCh - ARM7 - SOUNDxLEN - Sound Channel X Length Register (W)
The number of samples for N words is 4*N PCM8 samples, 2*N PCM16 samples, or 8*(N-1) ADPCM samples (the first word containing the ADPCM header). The Sound Length is not used in PSG mode.
  Bit0-21  Sound length (counted in words, ie. N*4 bytes)
  Bit22-31 Not used
Minimum length is 4 words (16 bytes), smaller values (0..3 words) are interpreted as zero length (no sound output).


 DS Sound Control Registers

4000500h - SOUNDCNT - ARM7 - Sound Control Register (R/W)
  Bit0-6   Master Volume          (0..127=silent..loud)
  Bit7     Not used               (always zero)
  Bit8-9   Left Out      (probably selects Mixer or "Bypassed" channels?)
  Bit10-11 Right Out     (probably selects Mixer or "Bypassed" channels?)
  Bit12    Output Sound Channel 1 (0=To Mixer, 1=Bypass Mixer)
  Bit13    Output Sound Channel 3 (0=To Mixer, 1=Bypass Mixer)
  Bit14    Not used               (always zero)
  Bit15    Master Enable          (0=Disable, 1=Enable)

4000504h - SOUNDBIAS - ARM7 - Sound Bias Register (R/W)
  Bit0-9   Sound Bias    (0..3FFh, usually 200h)
  Bit10-31 Not used      (always zero)
After applying the master volume, the signed left/right audio signals are in range -200h..+1FFh (with medium level zero), the Bias value is then added to convert the signed numbers into unsigned values (with medium level 200h).

The sampling frequency of the mixer is 1.04876 MHz with an amplitude resolution of 24 bits, but the sampling frequency after mixing with PWM modulation is 32.768 kHz with an amplitude resolution of 10 bits.


 DS Sound Capture

The DS contains 2 built-in sound capture devices that can capture output waveform data to memory.
Sound capture 0 can capture output from left-mixer or output from channel 0.
Sound capture 1 can capture output from right-mixer or output from channel 2.

4000508h - SNDCAP0CNT - ARM7 - Sound Capture 0 Control Register (R/W)
4000508h - SNDCAP1CNT - ARM7 - Sound Capture 1 Control Register (R/W)
  Bit0     Control of Associated Sound Channels
            SNDCAP0CNT: Output Sound Channel 1 (0=As such, 1=Add to Channel 0)
            SNDCAP1CNT: Output Sound Channel 3 (0=As such, 1=Add to Channel 2)
  Bit1     Capture Source Selection
            SNDCAP0CNT: Capture 0 Source (0=Left Mixer, 1=Channel 0)
            SNDCAP1CNT: Capture 1 Source (0=Right Mixer, 1=Channel 2)
  Bit2     Capture Repeat        (0=Loop, 1=One-shot)
  Bit3     Capture Format        (0=PCM16, 1=PCM8)
  Bit4-6   Not used              (always zero)
  Bit7     Capture Start/Status  (0=Stop, 1=Start/Busy)

4000510h - SNDCAP0DAD - ARM7 - Sound Capture 0 Destination Address (R/W)
4000518h - SNDCAP1DAD - ARM7 - Sound Capture 1 Destination Address (R/W)
  Bit0-26  Destination address
  Bit27-31 Not used       (always zero)
Capture start address (also used as re-start address for looped capture).

4000514h - SNDCAP0LEN - ARM7 - Sound Capture 0 Length (W)
400051Ch - SNDCAP1LEN - ARM7 - Sound Capture 1 Length (W)
  Bit0-15  Buffer length (1..FFFFh words) (ie. N*4 bytes)
  Bit16-31 Not used
Minimum length is 1 word (attempts to use 0 words are interpreted as 1 word).

SOUND1TMR - Sound Channel 1 Timer shared as Capture 0 Timer
SOUND3TMR - Sound Channel 3 Timer shared as Capture 1 Timer
There are no separate capture frequency registers, instead, the sample frequency of Channel 1/3 is shared for Capture 0/1. These channels are intended to output the captured data, so it makes sense that both capture and sound output use the same frequency.


 DS Sound Block Diagrams

Left Mixer with Capture 0
(Right Mixer with Capture 1, respectively)
                       _____
  Ch0.L ------------->|     |  +---------------------> to Capture 0
           ___        |     |  |      Ch0..Ch15  ___
  Ch1.L ->|Sel|------>|     |--+--------------->|   |
          |___|----+  |Left |                   |   |
  Ch2.L -----------|->|Mixer|                   |Sel|    ______
           ___     |  |     |                   |   |   |Master|
  Ch3.L ->|Sel|----|->|     |               Ch1 |   |-->|Volume|--> L
          |___|--+ |  |     |  +--------------->|   |   |______|
  Ch4.L ---------|-|->|     |  |                |   |
  ...   ---------|-|->|     |  |                |   |
  Ch15.L---------|-|->|_____|  |   ___          |   |
                 | +-----------+->|Add| Ch1+Ch3 |   |
                 +--------------->|___|-------->|___|

Channel 0 and 1, Capture 0 with input from Left Mixer
(Channel 2 and 3, Capture 1 with input from Right Mixer, respectively)
  ____     _________     ___     ___      ___
 |FIFO|-->|Channel 0|-->|Vol|-->|Add|-+->|Pan|--> Ch0.L
 |____|   |_________|   |___|   |___| |  |___|--> Ch0.R
  ____     _________     ___      ^   |
 |FIFO|<--|Capture 0|<--|Sel|<----|---+
 |____|   |_ _____ _|   |___|<----|-------------- Left Mixer
  ____     _:Timer:_     ___     _|_      ___
 |FIFO|-->|Channel 1|-->|Vol|-->|Sel|--->|Pan|--> Ch1.L
 |____|   |_________|   |___|   |___|    |___|--> Ch1.R

Channel 4 (Channel 5..15, respectively)
  ____     _________     ___              ___
 |FIFO|-->|Channel 4|-->|Vol|----------->|Pan|--> Ch4.L
 |____|   |_________|   |___|            |___|--> Ch4.R

The FIFO isn't used in PSG/Noise modes (supported on channel 8..15).


 DS Sound Notes

Sound delayed Start/Restart
A sound will be started/restarted when changing its start bit from 0 to 1, however, the sound won't start immediately: PSG/Noise starts after 1 sample, PCM starts after 3 samples, and ADPCM starts after 11 samples (3 dummy samples as for PCM, plus 8 dummy samples for the ADPCM header).

7bit Volume and Panning Values
  data.vol = data*N/128
  data.left = data*(128-N)/128
  data.right = data*N/128.
Register settings of 0..126,127 are interpreted as N=0..126,128.

PSG Sound
The output volume equals to PCM16 values +7FFFh (HIGH) and -7FFFh (LOW).
PSG sound is always Infinite (the SOUNDxLEN Register, and the SOUNDxCNT Repeat Mode bits have no effect). The PSG hardware doesn't support sound length, sweep, or volume envelopes, however, these effects can be produced by software with little overload (or, more typically, with enormous overload, depending on the programming language used).

PSG Wave Duty (channel 8..13 in PSG mode)
  0  12.5% "_______-_______-_______-"
  1  25.0% "______--______--______--"
  2  37.5% "_____---_____---_____---"
  3  50.0% "____----____----____----"
  4  62.5% "___-----___-----___-----"
  5  75.0% "__------__------__------"
  6  87.5% "_-------_-------_-------"
  7   0.0% "________________________"
Each duty cycle consists of eight HIGH or LOW samples, so the sound frequency is 1/8th of the selected sample rate. The duty cycle always starts at the begin of the LOW period when the sound gets (re-)started.

PSG Noise (channel 14..15 in PSG mode)
Noise randomly switches between HIGH and LOW samples, the output levels are calculated, at the selected sample rate, as such:
  X=X SHR 1, IF carry THEN Out=LOW, X=X XOR 6000h ELSE Out=HIGH
The initial value when (re-)starting the sound is X=7FFFh. The formula is more or less same as "15bit polynomial counter" used on 8bit Gameboy and GBA.

PCM8 and PCM16
Signed samples in range -80h..+7Fh (PCM8), or -8000h..+7FFFh (PCM16).
The output volume of PCM8=NNh is equal to PCM16=NN00h.

IMA-ADPCM Format
IMA-ADPCM is a Adaptive Differential Pulse Code Modulation (ADPCM) variant, designed by International Multimedia Association (IMA), the format is used, among others, in IMA-ADPCM compressed Windows .WAV files.
The NDS data consist of a 32bit header, followed by 4bit values (so each byte contains two values, the first value in the lower 4bits, the second in upper 4 bits). The 32bit header contains initial values:
  Bit0-15   Initial PCM16 Value (Pcm16bit = -8000h..+7FFF)
  Bit16-22  Table Index Initial Value (Index = 0..88)
  Bit23-31  Not used (zero)
The 4bit values are decoded into PCM16 values, as such:
  Diff = ((Data4bit AND 7)*2+1)*AdpcmTable[Index]/8
  IF (Data4bit AND 8) THEN Diff = -Diff
  Pcm16bit = MinMax (Pcm16bit+Diff,-8000h,+7FFFh)
  Index = MinMax (Index+IndexTable[Data4bit AND 7],0,88)
Whereas, IndexTable[0..7] = -1,-1,-1,-1,2,4,6,8. And AdpcmTable [0..88] =
  0007h,0008h,0009h,000Ah,000Bh,000Ch,000Dh,000Eh,0010h,0011h,0013h,0015h
  0017h,0019h,001Ch,001Fh,0022h,0025h,0029h,002Dh,0032h,0037h,003Ch,0042h
  0049h,0050h,0058h,0061h,006Bh,0076h,0082h,008Fh,009Dh,00ADh,00BEh,00D1h
  00E6h,00FDh,0117h,0133h,0151h,0173h,0198h,01C1h,01EEh,0220h,0256h,0292h
  02D4h,031Ch,036Ch,03C3h,0424h,048Eh,0502h,0583h,0610h,06ABh,0756h,0812h
  08E0h,09C3h,0ABDh,0BD0h,0CFFh,0E4Ch,0FBAh,114Ch,1307h,14EEh,1706h,1954h
  1BDCh,1EA5h,21B6h,2515h,28CAh,2CDFh,315Bh,364Bh,3BB9h,41B2h,4844h,4F7Eh
  5771h,602Fh,69CEh,7462h,7FFFh
The closest way to reproduce the AdpcmTable with 32bit integer maths appears:
  X=000776d2h, FOR I=0 TO 88, Table[I]=X SHR 16, X=X+(X/10), NEXT I
  Table[3]=000Ah, Table[4]=000Bh, Table[88]=7FFFh, Table[89..127]=0000h
When using ADPCM and loops, set the loopstart position to the data part, rather than the header. Do not change the ADPCM loop start position during playback.

Microphone Input
For Touchscreen (and Microphone) inputs, see
DS Touch Screen Controller (TSC)


 DS Various

System Clock
  Bus clock = somewhat 33MHz
  NDS7 clock = somewhat 33MHz (same as bus clock)
  NDS9 clock = somewhat 66MHz (twice bus clock)
Most timings in this document should be specified for 33MHz clock (not for the 66MHz clock). Respectively, NDS9 timings are counted in "half cycles".
All external memory access (and I/O) is delayed to bus clock (or slower in case of additional waitstates), so the full 66MHz can be used only internally in the NDS9 CPU core, ie. with cache and TCM.
The exact bus clock is specified as 33.513982 MHz (1FF61FEh Hertz).
On my own NDS, measured in relation to the RTC seconds IRQ, it appears more like 1FF6231h, that inaccuary of 1 cycle per 657138 cycles (about one second per week) on either oscillator, isn't too significant though.

DS Display Dimensions / Timings
Dot clock = 5.585664 MHz (=33.513982 MHz / 6)
H-Timing: 256 dots visible, 99 dots blanking, 355 dots total (15.7343KHz)
V-Timing: 192 lines visible, 71 lines blanking, 263 lines total (59.8261 Hz)
The V-Blank cycle for the 3D Engine consists of the 23 lines, 191..213.
Screen size 62.5mm x 47.0mm (each) (256x192 pixels)
Vertical space between screens 22mm (equivalent to 90 pixels)

RCNT
RCNT (134h) should be set to 80xxh (general purpose mode) before accessing EXTKEYIN (136h) or RTC (138h). No idea why (except for RTC/SI-interrupt).

DS Serial Port
The SI line is labeled "INT" on the NDS mainboard, it is connected to Pin 1 of the RTC chip (ie. the /INT interrupt pin).
I have no idea where to find SO, SC, and SD. I've written a test proggy that pulsed all four RCNT bits - but all I could find was the SI signal. However, the BIOS contains some code that uses SIO normal mode transfers (for the debug version), so at least SI, SO, SC should exist.

WLAN and 3D
Still missing in this document. Haven't done any research yet.


 DS DMA Transfers

The DS includes four DMA channels for each CPU (ie. eight channels in total), which are working more or less the same as on GBA:
DMA Transfers
All NDS9 and NDS7 DMA Registers are R/W. Word count is expanded to 21bits, max 1..1FFFFFh units, or 0=200000h units. All channels support the same regions and lengths. The gamepak bit (Bit 27) has been removed, instead, the mode bits have been expanded to a 3bit value (see below).

NDS9 DMACNT Bit27-29:
  0  Start Immediately
  1  Start at V-Blank
  2  Start at H-Blank (paused during V-Blank)
  3  Synchronize to start of display
  4  Main memory display
  5  DS Cartridge Slot
  6  GBA Cartridge Slot
  7  Geometry Command FIFO

NDS7 DMACNT Bit28-29: (Bit27 unused?)
  0  Start Immediately
  1  Start at V-Blank
  2  DS Cartridge Slot
  3  DMA0/DMA2: Wireless interrupt, DMA1/DMA3: GBA Cartridge Slot

40000E0h - NDS9 only - DMA0FILL - ARM9 - DMA 0 Filldata (R/W)
40000E4h - NDS9 only - DMA1FILL - ARM9 - DMA 1 Filldata (R/W)
40000E8h - NDS9 only - DMA2FILL - ARM9 - DMA 2 Filldata (R/W)
40000ECh - NDS9 only - DMA3FILL - ARM9 - DMA 3 Filldata (R/W)
  Bit0-31 Filldata
The DMA Filldata registers contain 16 bytes of general purpose WRAM, intended to be used as fixed source addresses for DMA memfill operations.
This is useful because DMA cannot read from TCM, and reading from Main RAM would require to recurse cache & write buffer.

NDS7 Sound DMA
The NDS additionally includes 16 Sound DMA channels, plus 2 Sound Capture DMA channels (see Sound chapter). The priority of these channels is unknown.

NDS9 Cache, Writebuffer, DTCM, and ITCM
Cache and tightly coupled memory are connected directly to the NDS9 CPU, without using the system bus. So that, DMA cannot access DTCM/ITCM, and access to cached memory regions must be handled with care: Drain the writebuffer before DMA-reads, and invalidate the cache after DMA-writes. See,
ARM CP15 System Control Coprocessor
The CPU can be kept running during DMA, provided that it is accessing only TCM (or cached memory), otherwise the CPU is halted until DMA finishes.
Respectively, interrupts executed during DMA will usually halt the CPU (unless the IRQ handler uses only TCM and cache; the IRQ vector at FFFF00xxh must be cached, or in TCM, and the IRQ handler may not access IE, IF, or other I/O ports).

NDS Sequential Main Memory DMA
Main RAM has different access time for sequential and non-sequential access. Normally DMA uses sequential access (except for the first word), however, if the source and destination addresses are both in Main RAM, then all accesses become non-sequential. In that case it would be faster to use two DMA transfers, one from Main RAM to a scratch buffer in WRAM, and one from WRAM to Main RAM.


 DS Timers

Same as GBA, except F = 33.514 MHz (for ARM9 at least).
Timers
Both ARM9 and ARM7 have four Timers each, eight Timers in total.


 DS Interrupts

4000208h - IME - 16bit - Interrupt Master Enable (R/W)
Same as GBA.

4000210h - IE - 32bit - Interrupt Enable (R/W)
4000214h - IF - 32bit - Interrupt Request Flags (R/W)
  Bit 0-6   Same as GBA
  Bit 7     NDS7 only: SIO/RCNT/RTC (Real Time Clock)
  Bit 8..   Same as GBA
  Bit 16    IPC Sync
  Bit 17    IPC Send FIFO Empty
  Bit 18    IPC Recv FIFO Not Empty
  Bit 19    Game Card Data Transfer Completion
  Bit 20    Game Card IREQ_MC
  Bit 21    NDS9 only: Geometry Command FIFO
  Bit 22    NDS7 only: Screens unfolding
  Bit 23    NDS7 only: SPI bus
  Bit 24    NDS7 only: Wifi
  Bit 25-31 Not used
Raw TCM-only IRQs can be processed even during DMA ?
For the "Same as GBA" Bits, see
Interrupt Control

DTCM+3FFCh - NDS9 IRQ Handler (hardcoded RAM address)
380FFFCh - NDS7 IRQ Handler (hardcoded RAM address)
  Bit 0-31  Pointer to IRQ Handler
NDS7 Handler must use ARM code, NDS9 Handler can be ARM/THUMB (Bit0=Thumb).

DTCM+3FF8h - NDS9 IRQ Check Bits (hardcoded RAM address)
380FFF8h - NDS7 IRQ Check Bits (hardcoded RAM address)
  Bit 0-31  IRQ Flags (same format as IE/IF registers)
When processing & acknowleding interrupts via IF register, the user interrupt handler should also set the corresponding bits of the IRQ Check value (required for BIOS IntrWait and VBlankIntrWait SWI functions).


 DS Maths

4000280h - NDS9 - DIVCNT - 16bit Division Control (R/W)
  0-1   Division Mode (0-2=See below, 3=Reserved)
  2-13  Not used
  14    Division by zero    (0=Okay, 1=Division by zero error)
  15    Busy                (0=Ready, 1=Busy) (Execution time see below)
Division Modes and Busy Execution Times
  Mode  Numer / Denum = Result, Remainder ; Cycles
  0     32bit / 32bit = 32bit , 32bit     ; 18 clks
  1     64bit / 32bit = 64bit , 32bit     ; 34 clks
  2     64bit / 64bit = 64bit , 64bit     ; 34 clks

4000290h - NDS9 - DIV_NUMER - 64bit Division Numerator (R/W)
4000298h - NDS9 - DIV_DENOM - 64bit Division Denominator (R/W)
40002A0h - NDS9 - DIV_RESULT - 64bit Division Quotient (=Numer/Denom) (R/W?)
40002A8h - NDS9 - DIVREM_RESULT - 64bit Remainder (=Numer MOD Denom) (R/W?)
All parameters are signed 64bit values. No info if values in 32bit-mode are signed and if sign is in bit31 or bit63? Division is started when writing to... what... presumably... Denom... Msb/Lsb?
The upper 32bit of DIV_DENOM must be zero in 32bit mode (the Division by Zero check always tests all 64bits, and fails if the unused/upper bits are non-zero).

40002B0h - NDS9 - SQRTCNT - 16bit - Square Root Control (R/W)
  0     Mode (0=32bit input, 1=64bit input)
  1-14  Not used
  15    Busy (0=Ready, 1=Busy) (Execution time is 13 clks, in either Mode)

40002B4h - NDS9 - SQRT_RESULT - 32bit - Square Root Result (R/W?)
40002B8h - NDS9 - SQRT_PARAM - 64bit - Square Root Parameter Input (R/W)
Unsigned 64bit parameter, and unsigned 32bit result.

Notes
Push all DIV/SQRT values (parameters, control, and result registers) when using DIV/SQRT registers on interrupt level.
The NDS9 and NDS7 BIOSes additionally contain software based division and square root functions, which are NOT using above hardware registers (even the NDS9 functions are raw software).
The Div/Sqrt timings are probably counted in 33.51MHz units?


 DS Inter Process Communication (IPC)

Allows to exchange status information between ARM7 and ARM9 CPUs.
The register can be accessed simultaneously by both CPUs (without violating access permissions, and without generating waitstates at either side).

4000180h - IPCSYNC - ARM9/ARM7 - IPC Synchronize Register (R/W)
  Bit  Dir  Expl.
  0-3  R    Data input from IPCSYNC Bit8-11 of remote CPU (00h..0Fh)
  4-7  -    Not used
  8-11 R/W  Data output to IPCSYNC Bit0-3 of remote CPU   (00h..0Fh)
  12   -    Not used
  13   W    Send IRQ to remote CPU      (0=None, 1=Send IRQ)
  14   R/W  Enable IRQ from remote CPU  (0=Disable, 1=Enable)
  15   -    Not used

4000184h - IPCFIFOCNT - ARM9/ARM7 - IPC Fifo Control Register (R/W)
  Bit  Dir  Expl.
  0    R    Send Fifo Empty Status      (0=Not Empty, 1=Empty)
  1    R    Send Fifo Full Status       (0=Not Full, 1=Full)
  2    R/W  Send Fifo Empty IRQ         (0=Disable, 1=Enable)
  3    W    Send Fifo Clear             (0=Nothing, 1=Flush Send Fifo)
  8    R    Receive Fifo Empty          (0=Not Empty, 1=Empty)
  9    R    Receive Fifo Full           (0=Not Full, 1=Full)
  10   R/W  Receive Fifo Not Empty IRQ  (0=Disable, 1=Enable)
  14   R/W  Error, Read Empty/Send Full (0=No Error, 1=Error/Acknowledge)
  15   R/W  Enable Send/Receive Fifo    (0=Disable, 1=Enable)

4000188h - IPCFIFOSEND - ARM9/ARM7 - IPC Send Fifo (W)
  Bit0-31  Send Fifo Data

4100000h - IPCFIFORECV - ARM9/ARM7 - IPC Receive Fifo (R)
  Bit0-31  Receive Fifo Data


 DS Keypad

For the GBA-buttons: Same as GBA, both ARM7 and ARM9 have keyboard input registers, and each its own keypad IRQ control register.
Keypad Input

For Touchscreen (and Microphone) inputs, see
DS Touch Screen Controller (TSC)

4000136h - ARM7 - EXTKEYIN - Key X/Y Input (R)
  0      Button X     (0=Pressed, 1=Released)
  1      Button Y     (0=Pressed, 1=Released)
  3      DEBUG button (0=Pressed, 1=Released/None such)
  6      Pen down     (0=Pressed, 1=Released/Disabled)
  7      Hinge/folded (0=Open, 1=Closed)
  2,4,5  Unknown / set
  8..15  Unknown / zero
The Hinge stuff is a magnetic sensor somewhere underneath of the Start/Select buttons, it will be triggered by the magnet field from the right speaker when the console is closed. The hinge generates an interrupt request (there seems to be no way to disable this, unlike as for all other IRQ sources), however, the interrupt execution can be disabled in IE register (as for other IRQ sources).
The Pen Down is the /PENIRQ signal from the Touch Screen Controller (TSC), if it is enabled in the TSC control register, then it will notify the program when the screen pressed, the program should then read data from the TSC (if there's no /PENIRQ then doing unneccassary TSC reads would just waste CPU power). However, the user may release the screen before the program performs the TSC read, so treat the screen as not pressed if you get invalid TSC values (even if /PENIRQ was LOW).
Not sure if the TSC /PENIRQ is actually triggering an IRQ in the NDS?
The Debug Button should be connected to R03 and GND (R03 is the large soldering point between the SL1 jumper and the VR1 potentiometer).
Interrupts are reportedly not supported for X,Y buttons.


 DS Real-Time Clock (RTC)

Seiko Instruments Inc. S-35180 (compatible with S-35190A)
Miniature 8pin RTC with 3-wire serial bus

4000138h - Real Time Clock Register
  Bit  Expl.
  0    Data I/O   (0=Low, 1=High)
  1    Clock Out  (0=Low, 1=High)
  2    Select Out (0=Low, 1=High/Select)
  4    Data  Direction  (0=Read, 1=Write)
  5    Clock Direction  (should be 1=Write)
  6    Select Direction (should be 1=Write)
  3,8-11   Unused I/O Lines
  7,12-15  Direction for Bit8-11 (usually 0)

Serial Transfer Flowchart
Chipselect and Command/Parameter Sequence:
  Init CS=LOW and /SCK=HIGH, and wait at least 1us
  Switch CS=HIGH, and wait at least 1us
  Send the Command byte (see bit-transfer below)
  Send/receive Parameter byte(s) associated with the command (see below)
  Switch CS to LOW
Bit transfer (repeat 8 times per cmd/param byte) (bits transferred LSB first):
  Output /SCK=LOW and SIO=databit (when writing), then wait at least 5us
  Output /SCK=HIGH, wait at least 5us, then read SIO=databit (when reading)
In this document <both> commands and parameters are specified in LSB-first transmission bit-order (unlike the original Seiko document which used mixed LSB-first and MSB-first conventions).

Command Register
  Command Register
    Fwd  Rev
    0-3  7-4 Fixed Code (must be 06h = 0110b) (same for Fwd and Rev)
    4-6  3-1 Command
             Fwd Rev Parameter bytes (read/write access)
             0   0   1 byte, status register 1
             4   1   1 byte, status register 2
             2   2   7 bytes, date & time (year,month,day,day_of_week,hour,minute, second)
             6   3   3 bytes, time (hour,minute,second)
             1*  4*  1 byte, int1, frequency duty setting
             1*  4*  3 bytes, int1, alarm time 1 (day_of_week, hour, minute)
             5   5   3 bytes, int2, alarm time 2 (day_of_week, hour, minute)
             3   6   1 byte, clock adjustment register
             7   7   1 byte, free register
    7    0   Parameter Read/Write Access (0=Write, 1=Read)
* INT1: Type and number of parameters depend on INT1 setting in stat reg2.
The "Fwd" bit numbers and command values for LSB-first command transfers (ie. both commands and parameters use the same bit-order).
The "Rev" numbers/values are for MSB-first command transfers (ie. commands using opposite bit-order than parameters, as being suggested by Seiko).

Control and Status Registers
  Status Register 1
    0   W   Reset                (0=Normal, 1=Reset)
    1   R/W 12/24 hour mode      (0=12 hour, 1=24 hour)
    2-3 R/W General purpose bits
    4   R   Interrupt 1 Flag (1=Yes)                      ;auto-cleared on read
    5   R   Interrupt 2 Flag (1=Yes)                      ;auto-cleared on read
    6   R   Power Low Flag (0=Normal, 1=Power is/was low) ;auto-cleared on read
    7   R   Power Off Flag (0=Normal, 1=Power was off)    ;auto-cleared on read
    Power off indicates that the battery was removed or fully discharged,
    all registers are reset to 00h (or 01h), and must be re-initialized.
  Status Register 2
    0-3 R/W INT1 Mode/Enable
            0000b Disable
            0x01b Selected Frequency steady interrupt
            0x10b Per-minute edge interrupt
            0011b Per-minute steady interrupt 1 (duty 30.0 secomds)
            0100b Alarm 1 interrupt
            0111b Per-minute steady interrupt 2 (duty 0.0079 secomds)
            1xxxb 32kHz output
    4-5 R/W General purpose bits
    6   R/W INT2 Enable
            0b    Disable
            1b    Alarm 2 interrupt
    7   R/W Test Mode (0=Normal, 1=Test, don't use) (cleared on Reset)
  Clock Adjustment Register (to compensate oscillator inaccuracy)
    0-7 R/W Adjustment (00h=Normal, no adjustment)
  Free Register
    0-7 R/W General purpose bits

Date Registers
  Year Register
    0-7 R/W Year     (BCD 00h..99h = 2000..2099)
  Month Register
    0-4 R/W Month    (BCD 01h..12h = January..December)
    5-7 -   Not used (always zero)
  Day Register
    0-5 R/W Day      (BCD 01h..28h,29h,30h,31h, range depending on month/year)
    6-7 -   Not used (always zero)
  Day of Week Register (septenary counter)
    0-2 R/W Day of Week (00h..06h, custom assignment, usually 0=Monday?)
    3-7 -   Not used (always zero)

Time Registers
  Hour Register
    0-5 R/W Hour     (BCD 00h..23h in 24h mode, or 00h..11h in 12h mode)
    6   *   AM/PM    (0=AM before noon, 1=PM after noon)
            * 24h mode: AM/PM flag is read only (PM=1 if hour = 12h..23h)
            * 12h mode: AM/PM flag is read/write-able
            * 12h mode: Observe that 12 o'clock is defined as 00h (not 12h)
    7   -   Not used (always zero)
  Minute Register
    0-6 R/W Minute   (BCD 00h..59h)
    7   -   Not used (always zero)
  Second Register
    0-6 R/W Minute   (BCD 00h..59h)
    7   -   Not used (always zero)

Alarm 1 and Alarm 2 Registers
  Alarm1 and Alarm2 Day of Week Registers (INT1 and INT2 each)
    0-2 R/W Day of Week (00h..06h)
    3-6 -   Not used (always zero)
    7   R/W Compare Enable (0=Alarm every day, 1=Alarm only at specified day)
  Alarm1 and Alarm2 Hour Registers (INT1 and INT2 each)
    0-5 R/W Hour     (BCD 00h..23h in 24h mode, or 00h..11h in 12h mode)
    6   R/W AM/PM    (0=AM, 1=PM) (must be correct even in 24h mode?)
    7   R/W Compare Enable (0=Alarm every hour, 1=Alarm only at specified hour)
  Alarm1 and Alarm2 Minute Registers (INT1 and INT2 each)
    0-6 R/W Minute   (BCD 00h..59h)
    7   R/W Compare Enable (0=Alarm every min, 1=Alarm only at specified min)
  Selected Frequency Steady Interrupt Register (INT1 only) (when Stat2/Bit2=0)
    0   R/W Enable 1Hz Frequency  (0=Disable, 1=Enable)
    1   R/W Enable 2Hz Frequency  (0=Disable, 1=Enable)
    2   R/W Enable 4Hz Frequency  (0=Disable, 1=Enable)
    3   R/W Enable 8Hz Frequency  (0=Disable, 1=Enable)
    4   R/W Enable 16Hz Frequency (0=Disable, 1=Enable)
            The signals are ANDed when two or more frequencies are enabled,
            ie. the /INT signal gets LOW when either of the signals is LOW.
    5-7 R/W General purpose bits
Note: There is only one register shared as "Selected Frequency Steady Interrupt" (accessed as single byte parameter when Stat2/Bit2=0) and as "Alarm1 Minute" (accessed as 3rd byte of 3-byte parameter when Stat2/Bit2=1), changing either value will also change the other value.

Interrupt
There's only one /INT signal, shared for both INT1 and INT2.
In the NDS, it is connected to the SI-input of the SIO unit (and so, also shared with SIO interrupts). To enable the interrupt, RCNT should be set to 8144h (Bit14-15=General Purpose mode, Bit8=SI Interrupt Enable, Bit6,2=SI Output/High).
Not sure why Output/High... maybe as pullup... however, it seems to be working also as Input... ie. RCNT=8100h.

Pin-Outs
  1 /INT      8 VDD
  2 XOUT      7 SIO
  3 XIN       6 /SCK
  4 GND       5 CS


 DS Serial Peripheral Interface Bus (SPI)

Serial Peripheral Interface Bus
SPI Bus is a 4-wire (Data In, Data Out, Clock, and Chipselect) serial bus.
The NDS supports the following SPI devices (each with its own chipselect).
DS Firmware Serial Flash Memory
DS Touch Screen Controller (TSC)
DS Power Management

40001C0h - SPICNT - NDS7 - SPI Bus Control/Status Register
  0-1 Baudrate   (0=4MHz/Firmware, 1=2MHz/Touchscr, 2=1MHz/Powerman., 3=512KHz)
  2-6 Not used            (Zero)
  7   Busy Flag           (0=Ready, 1=Busy) (presumably Read-only)
  8-9 Device Select       (0=Powerman., 1=Firmware, 2=Touchscr, 3=Reserved)
  10  Transfer Size       (0=8bit, 1=16bit)
  11  Chipselect Hold     (0=Deselect after transfer, 1=Keep selected)
  12  Unknown (usually 0) (set equal to Bit11 when BIOS accesses firmware)
  13  Unknown (usually 0) (set to 1 when BIOS accesses firmware)
  14  Interrupt Request   (0=Disable, 1=Enable)
  15  SPI Bus Enable      (0=Disable, 1=Enable)
The "Hold" flag should be cleared BEFORE transferring the LAST data unit, the chipselect will be then automatically cleared after the transfer, the program should issue a WaitByLoop(3) manually AFTER the LAST transfer.

40001C2h - SPIDATA - NDS7 - SPI Bus Data/Strobe Register (R/W)
The SPI transfer is started on writing to this register, so one must <write> a dummy value (should be zero) even when intending to <read> from SPI bus.
  0-7  Data
  8-15 Used only in 16bit mode, then containing first-or-second-8bits?
During transfer, the Busy flag in SPICNT is set, and the written SPIDATA value is transferred to the device (via output line), simultaneously data is received (via input line). Upon transfer completion, the Busy flag goes off (with optional IRQ), and the received value can be then read from SPIDATA, if desired.


 DS Touch Screen Controller (TSC)

Texas Instruments TSC2046
The Touch Screen Controller is accessed via SPI bus,
DS Serial Peripheral Interface Bus (SPI)

The lower LCD screen can be used as Touch Panel.

Control Byte (transferred MSB first)
  0-1  Power Down Mode Select
  2    Reference Select (0=Differential, 1=Single-Ended)
  3    Conversion Mode  (0=12bit, max CLK=2MHz, 1=8bit, max CLK=3MHz)
  4-6  Channel Select   (0-7, see below)
  7    Start Bit (Must be set to access Control Byte)

Channel
  0 Temperature 0 (requires calibration, step 2.1mV per 1'C accuracy)
  1 Touchscreen Y-Position  (somewhat 0B0h..F20h, or FFFh=released)
  2 Battery Voltage         (not used, connected to GND in NDS, always 000h)
  3 Touchscreen Z1-Position (diagonal position for pressure measurement)
  4 Touchscreen Z2-Position (diagonal position for pressure measurement)
  5 Touchscreen X-Position  (somewhat 100h..ED0h, or 000h=released)
  6 AUX Input               (connected to Microphone in the NDS)
  7 Temperature 1 (difference to Temp 0, without calibration, 2'C accuracy)
All channels can be accessed in Single-Ended mode.
In differential mode, only channel 1,3,4,5 (X,Z1,Z2,Y) can be accessed.

Power Down Mode
  Mode /PENIRQ   VREF  ADC   Recommended use
  0    Enabled   Auto  Auto  Differential Mode (Touchscreen, Penirq)
  1    Disabled  Off   On    Single-Ended Mode (Temperature, Microphone)
  2    Enabled   On    Off   Don't use
  3    Disabled  On    On    Don't use
Allows to enable/disable the /PENIRQ output, the internal reference voltage (VREF), and the Analogue-Digital Converter.

Reference Voltage (VREF)
VREF is used as reference voltage in single ended mode, at 12bit resolution one ADC step equals to VREF/4096. The TSC generates an internal VREF of 2.5V (+/-0.05V), however, the NDS uses as external VREF of 3.33V (sinks to 3.31V at low battery charge), the external VREF is always enabled, no matter if internal VREF is on or off. Power Down Mode 1 disables the internal VREF, which may reduce power consumption in single ended mode. After conversion, Power Down Mode 0 should be restored to re-enable the Penirq signal.

Sending the first Command after Chip-Select
Switch chipselect low, then output the command byte (MSB first).

Reply Data
The following reply data is received (via Input line) after the Command byte has been transferred: One dummy bit (zero), followed by the 8bit or 12bit conversion result (MSB first), followed by endless padding (zero).
Note: The returned ADC value may become unreliable if there are longer delays between sending the command, and receiving the reply byte(s).

Sending further Commands during/after receiving Reply Data
In general, the Output line should be LOW during the reply period, however, once when Data bit6 has been received (or anytime later), a new Command can be invoked (started by sending the HIGH-startbit, ie. Command bit7), simultanously, the remaining reply-data bits (bit5..0) can be received.
In other words, the new command can be output after receiving 3 bits in 8bit mode (the dummy bit, and data bits 7..6), or after receiving 7 bits in 12bit mode (the dummy bit, and data bits 11..6).
In practice, the NDS SPI register always transfers 8 bits at once, so that one would usually receive 8 bits (rather than above 3 or 7 bits), before outputting a new command.

Touchscreen Position
Read the X and Y positions in 12bit differential mode, then convert the touchscreen values (adc) to screen/pixel positions (scr), as such:
  scr.x = (adc.x-adc.x1) * (scr.x2-scr.x1) / (adc.x2-adc.x1) + (scr.x1-1)
  scr.y = (adc.y-adc.y1) * (scr.y2-scr.y1) / (adc.y2-adc.y1) + (scr.y1-1)
The X1,Y1 and X2,Y2 calibration points are found in Firmware User Settings,
DS Firmware User Settings
scr.x1,y1,x2,y2 are originated at 1,1 (converted to 0,0 by above formula).

Touchscreen Pressure
To calculate the pressure resistance, in respect to X/Y/Z positions and X/Y plate resistances, either of below formulas can be used,
  Rtouch = (Rx_plate*Xpos*(Z2pos/Z1pos-1))/4096
  Rtouch = (Rx_plate*Xpos*(4096/Z1pos-1)-Ry_plate*(1-Ypos))/4096
The second formula requires less CPU load (as it doesn't require to measure Z2), the downside is that one must know both X and Y plate resistance (or at least their ratio). The first formula doesn't require that ratio, and so Rx_plate can be set to any value, setting it to 4096 results in
  touchval = Xpos*(Z2pos/Z1pos-1)
Of course, in that case, touchval is just a number, not a resistance in Ohms.

Touchscreen Notes
It may be impossible to press locations close to the screen borders.
When pressing two or more locations the TSC values will be somewhere in the middle of these locations.
The TSC values may be garbage if the screen becomes newly pressed or released, to avoid invalid inputs: read TSC values at least two times, and ignore BOTH positions if ONE position was invalid.

Microphone / AUX Channel
Observe that the microphone amplifier is switched off after power up, see:
DS Power Management

Temperature Calculation
TP0 decreases by circa 2.1mV per degree Kelvin. The voltage difference between TP1 minus TP0 increases by circa 0.39mV (1/2573 V) per degree Kelvin. At VREF=3.33V, one 12bit ADC step equals to circa 0.8mV (VREF/4096).
Temperature can be calculated at best resolution when using the current TP0 value, and two calibration values (an ADC value, and the corresponding temperature in degrees kelvin):
  K = (CAL.TP0-ADC.TP0) * 0.4 + CAL.KELVIN
Alternately, temperature can be calculated at rather bad resolution, but without calibration, by using the difference between TP1 and TP0:
  K = (ADC.TP1-ADC.TP0) * 8568 / 4096
To convert Kelvin to other formats,
  Celsius:     C = (K-273.15)
  Fahrenheit:  F = (K-273.15)*9/5+32
  Reaumur:     R = (K-273.15)*4/5
  Rankine:     X = (K)*9/5
The Temperature Range for the TSC chip is -40'C .. +85'C. According to Nintendo, the DS should not be exposed to "extreme" heat or cold, the optimal battery charging temperature is specified as +10'C..+40'C.
The original firmware does not support temperature calibration, calibration is supported by nocash firmware (if present). See Extended Settings,
DS Firmware Extended Settings

Pin-Outs
         ________
  VCC  1|o       |16 DCLK
  X+   2|        |15 /CS
  Y+   3|  TSC   |14 DIN
  X-   4|  2046  |13 BUSY
  Y-   5|        |12 DOUT
  GND  6|        |11 /PENIRQ
  VBAT 7|        |10 IOVDD
  AUX  8|________|9  VREF


 DS Power Management

The DS contains several Power Managment functions, some accessed via I/O ports, some accessed via SPI bus (described later on below).

4000304h - POWCNT1 - NDS9 - Graphics Power Control Register (R/W)
  0     Enable Flag for both LCDs (0=Disable) (Prohibited, see notes)
  1     2D Graphics Engine A      (0=Disable) (Ports 008h-05Fh, Pal 5000000h)
  2     3D Rendering Engine       (0=Disable) (Ports 320h-3FFh)
  3     3D Geometry Engine        (0=Disable) (Ports 400h-6FFh)
  4-8   Not used
  9     2D Graphics Engine B      (0=Disable) (Ports 1008h-105Fh, Pal 5000400h)
  10-14 Not used
  15    Display Swap (0=Send Display A to Lower Screen, 1=To Upper Screen)
Use SwapBuffers command once after enabling Rendering/Geometry Engine.
Improper use of Bit0 may damage the hardware?
When disabled, corresponding Ports become Read-only, corresponding (palette-) memory becomes read-only-zero-filled.

4000304h - POWCNT2 - NDS7 - Sound/Wifi Power Control Register (R/W)
  Bit   Expl.
  0     Sound Speakers (0=Disable, 1=Enable)
  1     Wifi           (0=Disable, 1=Enable)
  2-15  Not used
Note: Bit0 disables the internal Speaker only, headphones are not disabled.

4000206h - NDS7 - Unknown
  Bit   Expl.
  0-15  Unknown (set to 0030h) somehow WLAN/POWCNT related?

4000301h - NDS7 - HALTCNT - Low Power Mode Control (R/W)
In Halt mode, the CPU is paused as long as (IE AND IF)=0.
In Sleep mode, most of the hardware including sound and video are paused, this very-low-power mode could be used much like a screensaver.
  Bit   Expl.
  0-5   Not used (zero)
  6-7   Power Down Mode  (0=No function, 1=Enter GBA Mode, 2=Halt, 3=Sleep)
The HALTCNT register should not be accessed directly. Instead, the BIOS Halt, Sleep, CustomHalt, IntrWait, or VBlankIntrWait SWI functions should be used.
BIOS Halt Functions
ARM CP15 System Control Coprocessor
The NDS9 does not have a HALTCNT register, instead, the Halt function uses the co-processor opcode "mcr p15,0,r0,c7,c0,4" - this opcode locks up if interrupts are disabled via IME=0 (unlike NDS7 HALTCNT method which doesn't check IME).

4000300h - NDS7/NDS9 - POSTFLG - BYTE - Post Boot Flag (R/W)
The NDS7 and NDS9 post boot flags are usually set upon BIOS/Firmware boot completion, once when set the reset vector is redirected to the debug handler of Nintendo's hardware debugger. That allows the NDS7 debugger to capture accidental jumps to address 0, that appears to be a common problem with HLL-programmers, asm-coders know that (and why) they should not jump to 0.
  Bit   Expl.
  0     Post Boot Flag (0=Boot in progress, 1=Boot completed)
  1     NDS7: Not used (always zero), NDS9: Bit1 is read-writeable
  2-7   Not used (always zero)
There are some write-restrictions: The NDS7 register can be written to only from code executed in BIOS. Bit0 of both NDS7 and NDS9 registers cannot be cleared (except by Reset) once when it is set.

Power Management Device
The Power Management Device is accessed via SPI bus,
DS Serial Peripheral Interface Bus (SPI)
To access the device, write the Index Register, then read or write the data register, and release the chipselect line when finished.

Index Register
  Bit0-1 Register Select          (0..3)
  Bit2-6 Not used
  Bit7   Register Direction       (0=Write, 1=Read)
Register 0 - Powermanagement Control (R/W)
  Bit0   Sound Amplifier          (0=Disable, 1=Enable)
         (When disabled, sound becomes very silent, but it is still audible)
  Bit1   Sound related?           (0=Disable, 1=Enable)
  Bit2   Lower Backlight          (0=Disable, 1=Enable)
  Bit3   Upper Backlight          (0=Disable, 1=Enable)
  Bit4   Power LED Blink Enable   (0=Always ON, 1=Blinking OFF/ON)
  Bit5   Power LED Blink Speed    (0=Slow, 1=Fast) (only if Blink enabled)
  Bit6   DS System Power          (0=Normal, 1=Shut Down)
  Bit7   Not used
Register 1 - Battery Status (R)
  Bit0   Battery Power LED Status (0=Power Good/Green, 1=Power Low/Red)
  Bit1-7 Not used
Register 2 - Microphone Amplifier Control (R/W)
  Bit0   Amplifier                (0=Disable, 1=Enable)
  Bit1-7 Not used
Register 3 - Microphone Amplifier Gain Control (R/W)
  Bit0-1 Gain                     (0..3=Gain 20, 40, 80, 160)
  Bit2-7 Not used

Backlight Dimming / Backlight caused Shut-Down(s)
The above bits are essentially used to switch Backlights on or off. However, there a number of strange effects. Backlight dimming is possible by pulse width modulation, ie. by using a timer interrupt to issue pulse widths of N% ON, and 100-N% OFF. Too long pulses are certainly resulting in flickering. Too short pulses are ignored, the backlights will remain OFF, even if the ON and OFF pulses are having the same length. Much too short pulses cause the power supply to shut-down; after changing the backlight state, further changes must not occur within the next (circa) 2500 clock cycles. The mainboard can be operated without screens & backlights connected, however, if so, the power supply will shut-down as soon as backlights are enabled.

Memory Power Down Functions
DS Main Memory Control
DS Firmware Serial Flash Memory


 DS Main Memory Control

Main Memory
The DS Main Memory is 2Mx16bit (4MByte), 1.8V Pseudo SRAM (PSRAM); all Dynamic RAM refresh is handled internally, the chip doesn't require any external refresh signals, and alltogether behaves like Static RAM. Non-sequential access time is 70ns, sequential (burst) access time is 12ns.

Main Memory Control
The memory chips contain built-in Control functions, which can be accessed via Port 27FFFFEh and/or by EXMEMCNT Bit 14. Nintendo is using at least two different types of memory chips in DS consoles, Fujitsu 82DBS02163C-70L, and ST M69AB048BL70ZA8, both appear to have different control mechanisms, other chips (with 8MB size) are used in the semi-professional DS hardware debuggers, and further chips may be used in future, so using the memory control functions may lead into compatibitly problems.

Power Consumption / Power Control
Power Consumption during operation (read/write access) is somewhat 30mA, in standby mode (no read/write access) consumption is reduced to 100uA.
Furthermore, a number of power-down modes are supported: In "Deep" Power Down mode the refresh is fully disabled, consumption is 10uA (and all data will be lost), in "Partial" Power Down modes only fragment of memory is refreshed, for smallest fragments, consumption goes to down to circa 50uA. The chip cannot be accessed while it is in Deep or Partial Power Down mode.

Fujitsu 82DBS02163C-70L
The Configuration Register (CR) can be written to by the following sequence:
  LDRH R0,[27FFFFEh]      ;read one value
  STRH R0,[27FFFFEh]      ;write should be same value as above
  STRH R0,[27FFFFEh]      ;write should be same value as above
  STRH R0,[27FFFFEh]      ;write any value
  STRH R0,[27FFFFEh]      ;write any value
  LDRH R0,[2400000h+CR*2] ;read, address-bits are defining new CR value
Do not access any other Main Memory addresses during above sequence (ie. disable interrupts, and do not execute the sequence by code located in Main Memory). The CR value is write-only. The CR bits are:
  Bit    Expl.
  0-6    Reserved         (Must be 7Fh)
  7      Write Control
           0=WE Single Clock Pulse Control without Write Suspend Function
           1=WE Level Control with Write Suspend Function)
          Burst Read/Single Write is not supported at WE Single Clock Mode.
  8      Reserved         (Must be 1)
  9      Valid Clock Edge (0=Falling Edge, 1=Rising Edge)
  10     Single Write     (0=Burst Read/Burst Write, 1=Burst Read/Single Write)
  11     Burst Sequence   (0=Reserved, 1=Sequential)
  12-14  Read Latency     (1=3 clocks, 2=4 clocks, 3=5 clocks, other=Reserved)
  15     Mode
           0=Synchronous:  Burst Read, Burst Write
           1=Asynchronous: Page Read, Normal Write
          In Mode 1 (Async), only the Partial Size bits are used,
          all other bits, CR bits 0..18, must be "1".
  16-18  Burst Length     (2=8 Words, 3=16Words, 7=Continous, other=Reserved)
  19-20  Partial Size     (0=1MB, 1=512KB, 2=Reserved, 3=Deep/0 bytes)
The Power Down mode is entered by setting CE2=LOW, this can be probably done by setting EXMEMCNT Bit14 to zero.

ST Microelectronics M69AB048BL70ZA8
The chip name decodes as PSRAM (M96), Asynchronous (A), 1.8V Burst (B), 2Mx16 (048), Two Chip Enables (B), Low Leakage (L), 70ns (70), Package (ZA), -30..+85'C (8).
There are three data sheets for different PSRAM chips available at www.st.com (unfortunately none for M69AB048BL70ZA8), each using different memory control mechanisms.

NDS9 BIOS
The NDS9 BIOS contains the following Main Memory initialization code, that method doesn't match up with any ST (nor Fujitsu) data sheets that I've seen. At its best, it looks like a strange (and presumably non-functional) mix-up of different ST control methods.
  STRH 2000h,[4000204h]
  LDRH R0,[27FFFFEh]
  STRH R0,[27FFFFEh]
  STRH R0,[27FFFFEh]
  STRH FFDFh,[27FFFFEh]
  STRH E732h,[27FFFFEh]
  LDRH R0,[27E57FEh]
  STRH 6000h,[4000204h]
In the above BIOS code, EXMEMCNT.14 appears to be used to unlock the control register. However, the NDS Firmware appears to use EXMEMCNT.14 to switch Main Memory into Power Down mode before entering GBA mode.


 DS Cartridges, Encryption, Firmware

Cartridges
DS Cartridge Header
DS Cartridge Secure Area
DS Cartridge Icon/Title
DS Cartridge Protocol
DS Cartridge I/O Ports
DS Cartridge NitroROM File System
DS Cartridge PassMe/PassThrough

Encryption
DS Encryption by Gamecode/Idcode (KEY1)
DS Encryption by Random Seed (KEY2)

Firmware
DS Firmware Serial Flash Memory
DS Firmware Header
DS Firmware User Settings
DS Firmware Extended Settings


 DS Cartridge Header

Header Overview (loaded from ROM Addr 0 to Main RAM 27FFE00h on Power-up)
  Address Bytes Expl.
  000h    12    Game Title  (Uppercase ASCII, padded with 00h)
  00Ch    4     Gamecode    (Uppercase ASCII, NTR-<code>)        (0=homebrew)
  010h    2     Makercode   (Uppercase ASCII, eg. "01"=Nintendo) (0=homebrew)
  012h    1     Unitcode    (00h=Nintendo DS)
  013h    1     Encryption Seed Select (00..07h, usually 00h)
  014h    1     Devicecapacity         (Chipsize = 128KB SHL nn) (eg. 7 = 16MB)
  015h    9     Reserved           (zero filled)
  01Eh    1     ROM Version        (usually 00h)
  01Fh    1     Autostart (Bit2: Skip "Press Button" after Health and Safety)
                (Also skips bootmenu, even in Manual mode & even Start pressed)
  020h    4     ARM9 rom_offset    (4000h and up, align 1000h)
  024h    4     ARM9 entry_address (2000000h..23BFE00h)
  028h    4     ARM9 ram_address   (2000000h..23BFE00h)
  02Ch    4     ARM9 size          (max 3BFE00h) (3839.5KB)
  030h    4     ARM7 rom_offset    (8000h and up)
  034h    4     ARM7 entry_address (2000000h..23BFE00h, or 37F8000h..3807E00h)
  038h    4     ARM7 ram_address   (2000000h..23BFE00h, or 37F8000h..3807E00h)
  03Ch    4     ARM7 size          (max 3BFE00h, or FE00h) (3839.5KB, 63.5KB)
  040h    4     File Name Table (FNT) offset
  044h    4     File Name Table (FNT) size
  048h    4     File Allocation Table (FAT) offset
  04Ch    4     File Allocation Table (FAT) size
  050h    4     File ARM9 overlay_offset
  054h    4     File ARM9 overlay_size
  058h    4     File ARM7 overlay_offset
  05Ch    4     File ARM7 overlay_size
  060h    4     Port 40001A4h setting for normal commands (usually 00586000h)
  064h    4     Port 40001A4h setting for KEY1 commands   (usually 001808F8h)
  068h    4     Icon_title_offset (0=None) (8000h and up)
  06Ch    2     Secure Area Checksum, CRC-16 of [ [20h]..7FFFh]
  06Eh    2     Secure Area Loading Timeout (usually 051Eh)
  070h    4     ARM9 Auto Load List RAM Address (?)
  074h    4     ARM7 Auto Load List RAM Address (?)
  078h    8     Secure Area Disable (by encrypted "NmMdOnly") (usually zero)
  080h    4     Total Used ROM size (remaining/unused bytes usually FFh-padded)
  084h    4     ROM Header Size (4000h)
  088h    38h   Reserved (zero filled)
  0C0h    9Ch   Nintendo Logo (compressed bitmap, same as in GBA Headers)
  15Ch    2     Nintendo Logo Checksum, CRC-16 of [0C0h-15Bh], fixed CF56h
  15Eh    2     Header Checksum, CRC-16 of [000h-15Dh]
  160h    4     Debug rom_offset   (0=none) (8000h and up)       ;only if debug
  164h    4     Debug size         (0=none) (max 3BFE00h)        ;version with
  168h    4     Debug ram_address  (0=none) (2400000h..27BFE00h) ;SIO and 8MB
  16Ch    4     Reserved (zero filled) (transferred, and stored, but not used)
  170h    90h   Reserved (zero filled) (transferred, but not stored in RAM)

For more info about CRC-16, see description of GetCRC16 BIOS function,
BIOS Misc Functions
For the Logo checksum, the BIOS verifies only [15Ch]=CF56h, it does NOT verify the actual data at [0C0h-15Bh] (nor it's checksum), however, the data is verified by the firmware.

Secure Area Loading Timeout (usually X=051Eh) used to initialize timer counter to (0-((X AND 3FFFh)+2)). (In case of Header checksum error it is ANDed with 1FFFh instead of 3FFFh, no idea why).


 DS Cartridge Secure Area

The Secure Area is located in ROM at 4000h..7FFFh, it can contain normal program code and data, however, it can be used only for ARM9 boot code, it cannot be used for ARM7 boot code, icon/title, filesystem, or other data.

Secure Area Size
The Secure Area exists if the ARM9 boot code ROM source address (src) is located within 4000h..7FFFh, if so, it will be loaded (by BIOS via KEY1 encrypted commands) in 4K portions, starting at src, aligned by 1000h, up to address 7FFFh. The secure area size if thus 8000h-src, regardless of the ARM9 boot code size entry in header.

Secure Area ID
The first 8 bytes of the secure area are containing the Secure Area ID, the ID is required (verified by BIOS boot code), the ID value changes during boot process:
  Value                Expl.
  "encryObj"           raw ID before encryption (raw ROM-image)
  (encrypted)          encrypted ID after encryption (encrypted ROM-image)
  "encryObj"           raw ID after decryption (verified by BIOS boot code)
  E7FFDEFFh,E7FFDEFFh  destroyed ID (overwritten by BIOS after verify)

Secure Area First 2K Encryption
The first 2K of the Secure Area (if it exists) are KEY1 encrypted. In Nintendo's cartridges, the first 2K of the secure area are containing only a few SWI calls, and the rest of it appears to be filled with random values (presumably for confusion purposes). Anyways, that confusion/random values aren't required, the first 2K may contain normal program code and data.


 DS Cartridge Icon/Title

The ROM offset of the Icon/Title is defined in Cartridge Header [68h].
If it is present (nonzero), then Icon/Title are displayed in the bootmenu.
  Addr  Siz  Expl.
  000h  2    Version  (0001h)
  002h  2    CRC16 across entries 020h..83Fh
  004h  1Ch  Reserved (zero-filled)
  020h  200h Icon Bitmap  (32x32 pix) (4x4 tiles, each 4x8 bytes, 4bit depth)
  220h  20h  Icon Palette (16 colors, 16bit, range 0000h-7FFFh)
             (Color 0 is transparent, so the 1st palette entry is ignored)
  240h  100h Title 0 Japanese (128 characters, 16bit Unicode)
  340h  100h Title 1 English  ("")
  440h  100h Title 2 French   ("")
  540h  100h Title 3 German   ("")
  640h  100h Title 4 Italian  ("")
  740h  100h Title 5 Spanish  ("")
  840h  -    End of Icon/Title structure (next 1C0h bytes usually FFh-filled)
Usually, for non-multilanguage games, the same (english) title is stored in all title entries. The title may consist of ASCII characters 0020h-007Fh, character 000Ah (linefeed), and should be terminated/padded by 0000h. The whole text should not exceed the dimensions of the DS cart field in the bootmenu. The title is usually split into a primary title, optional sub-title, and manufacturer, each separated by 000Ah character(s). For example, "America",000Ah,"The Axis of War",000Ah,"Cynicware",0000h


 DS Cartridge Protocol

Communication with Cartridge ROM relies on sending 8 byte commands to the cartridge, after the sending the command, a data stream can be received from the cartridge (the length of the data stream isn't fixed, below descriptions show the default length in brackets, but one may receive more, or less bytes, if desired).

Cartridge Memory Map
  0000h-0FFFh Header (unencrypted)
  1000h-3FFFh Not read-able (zero filled in ROM-images)
  4000h-7FFFh Secure Area, 16KBytes (first 2Kbytes with extra encryption)
  8000h-...   Main Data Area
Cartridge memory must be copied into Work RAM (the CPU cannot execute code in ROM).

Command Summary, Cmd/Reply-Encryption Type, Default Length
  Command/Params    Expl.                             Cmd  Reply Len
  -- Unencrypted Load --
  9F00000000000000h Dummy (read HIGH-Z bytes)         RAW  RAW   2000h
  0000000000000000h Get Cartridge Header              RAW  RAW   200h
  9000000000000000h 1st Get ROM Chip ID               RAW  RAW   4
  00aaaaaaaa000000h Unencrypted Data (debug ver only) RAW  RAW   200h
  3Ciiijjjxkkkkkxxh Activate KEY1 Encryption Mode     RAW  RAW   0
  -- Secure Area Load --
  4llllmmmnnnkkkkkh Activate KEY2 Encryption Mode     KEY1 FIX   910h+0
  1lllliiijjjkkkkkh 2nd Get ROM Chip ID               KEY1 KEY2  910h+4
  xxxxxxxxxxxxxxxxh Invalid - Get KEY2 Stream XOR 00h KEY1 KEY2  910h+...
  2bbbbiiijjjkkkkkh Get Secure Area Block (4Kbytes)   KEY1 KEY2  910h+11A8h
  6lllliiijjjkkkkkh Optional KEY2 Disable             KEY1 KEY2  910h+?
  Alllliiijjjkkkkkh Enter Main Data Mode              KEY1 KEY2  910h+0
  -- Main Data Load --
  B7aaaaaaaa000000h Encrypted Data Read               KEY2 KEY2  200h
  B800000000000000h 3rd Get ROM Chip ID               KEY2 KEY2  4
  xxxxxxxxxxxxxxxxh Invalid - Get KEY2 Stream XOR 00h KEY2 KEY2  ...
The parameter digits contained in above commands are:
  aaaaaaaa     32bit ROM address (command B7 can access only 8000h and up)
  bbbb         Secure Area Block number (0004h..0007h for addr 4000h..7000h)
  x,xx         Random, not used in further commands
  iii,jjj,llll Random, must be SAME value in further commands
  kkkkk        Random, must be INCREMENTED after FURTHER commands
  mmm,nnn      Random, used as KEY2-encryption seed

++++ Unencrypted Commands (First Part of Boot Procedure) ++++

Cartridge Reset
The /RES Pin switches the cartridge into unencrypted mode. After reset, the first two commands (9Fh and 00h) are transferred at 4MB/s CLK rate.

9F00000000000000h (2000h) - Dummy
Dummy command send after reset, returns endless stream of HIGH-Z bytes (ie. usually receiving FFh, immediately after sending the command, the first 1-2 received bytes may be equal to the last command byte).

0000000000000000h (200h) - Get Header
Returns RAW unencrypted cartridge header, repeated every 1000h bytes. The interesting area are the 1st 200h bytes, the rest is typically zero filled.
The Gamecode header entry is used later on to initialize the encryption. Also, the ROM Control entries define the length of the KEY1 dummy periods (typically 910h clocks), and the CLK transfer rate for further commands (typically faster than the initial 4MB/s after power up).

9000000000000000h (4) - 1st Get ROM Chip ID
Returns RAW unencrypted Chip ID (C2h,0Fh,00h,00h), repeated every 4 bytes.
The first byte identifies the manufacturer (C2h = Macronix).
The MSB of the Chip ID (bit7 of the fourth byte) selects the Secure Area Block transfer mode (8x200h or 1000h).

3Ciiijjjxkkkkkxxh (0) - Activate KEY1 Encryption Mode
The 3Ch command returns endless stream of HIGH-Z bytes, all following commands, and their return values, are encrypted. The random parameters iii,jjj,kkkkk must be re-used in further commands; the 20bit kkkkk value is to be incremented by one after each <further> command (it is <not> incremented after the 3Ch command).

++++ KEY1 Encrypted Commands (2nd Part of Boot procedure) ++++

4llllmmmnnnkkkkkh (910h) - Activate KEY2 Encryption Mode
KEY1 encrypted command, parameter mmmnnn is used to initialize the KEY2 encryption stream. Returns 910h dummy bytes (which are still subject to old KEY2 settings; at pre-initialization time, this is fixed: HIGH-Z, C5h, 3Ah, 81h, etc.). The new KEY2 seeds are then applied, and the first KEY2 byte is then precomputed. The 910h dummy stream is followed by that precomputed byte value endless repeated (this is the same value as that "underneath" of the first HIGH-Z dummy-byte of the next command).

1lllliiijjjkkkkkh (914h) - 2nd Get ROM Chip ID / Get KEY2 Stream
KEY1 encrypted command. Returns 910h dummy bytes, followed by KEY2 encrypted Chip ID repeated every 4 bytes, which must be identical as for the 1st Get ID command. The BIOS randomly executes this command once or twice. Changing the first command byte to any other value returns an endless KEY2 encrypted stream of 00h bytes, that is the easiest way to retrieve encryption values and to bypass the copyprotection.

2bbbbiiijjjkkkkkh (19B8h) - Get Secure Area Block
KEY1 encrypted command. Used to read a secure area block (bbbb in range 0004h..0007h for addr 4000h..7000h), each block is 4K, so it requires four Get Secure Area commands to receive the whole Secure Area (ROM locations 4000h-7FFFh), the BIOS is reading these blocks in random order.
Normally (if the upper bit of the Chip ID is set): Returns 910h dummy bytes, followed by 200h KEY2 encrypted Secure Area bytes, followed by 18h KEY2 encrypted 00h bytes, then the next 200h KEY2 encrypted Secure Area bytes, again followed by 18h KEY2 encrypted 00h bytes, and so on. That stream is repeated every 10C0h bytes (8x200h data bytes, plus 8x18h zero bytes).
Alternately (if the upper bit of the Chip ID is zero): Returns 910h dummy bytes, followed by 1000h KEY2 encrypted Secure Area bytes, presumably followed by 18h bytes, too.
Aside from above KEY2 encryption (which is done by hardware), the first 2K of the Secure Area is additionally KEY1 encrypted (which must be resolved after transfer by software).

6lllliiijjjkkkkkh (0) - Optional KEY2 Disable
KEY1 encrypted command. Returns 910h dummy bytes (which are still KEY2 affected), followed by endless stream of RAW 00h bytes. KEY2 encryption is disabled for all following commands.
This command is send only if firmware[18h] matches encrypted string "enPngOFF", and ONLY if firmware get_crypt_keys had completed BEFORE completion of secure area loading, this timing issue may cause unstable results.

Alllliiijjjkkkkkh (910h) - Enter Main Data Mode
KEY1 encrypted command. Returns 910h dummy bytes, followed by endless KEY2 encrypted stream of 00h bytes. All following commands are KEY2 encrypted.

++++ KEY2 Encrypted Commands (Main Data Transfer) ++++

B7aaaaaaaa000000h (200h) - Get Data
KEY2 encrypted command. The desired ROM address is specifed, MSB first, in parameter bytes (a). Can be used only for addresses 8000h and up, smaller addresses will be silently redirected to address "8000h+(addr AND 1FFh)". There is no alignment restriction for the address. However, the datastream wraps to the begin of the current 4K block when address+length crosses a 4K boundary (1000h bytes). Returned data is KEY2 encrypted.

B800000000000000h (4) - 3rd Get ROM Chip ID
KEY2 encrypted command. Returns KEY2 encrypted Chip ID repeated every 4 bytes.

xxxxxxxxxxxxxxxxh - Invalid Command
Any other command (anything else than above B7h and B8h) in KEY2 command mode causes communcation failures. The invalid command returns an endless KEY2 encrypted stream of 00h bytes. After the invalid command, the KEY2 stream is NOT advanced for further command bytes, further commands seems to return KEY2 encrypted 00h bytes, of which, the first returned byte appears to be HIGH-Z.
Ie. the cartridge seems to have switched back to a state similar to the KEY1-phase, although it doesn't seem to be possible to send KEY1 commands.

++++ Notes ++++

KEY1 Command Encryption / 910h Dummy Bytes
All KEY1 encrypted commands are followed by 910h dummy byte transfers, these 910h clock cycles are probably used to decrypt the command at the cartridge side; communication will fail when transferring less than 910h bytes.
The return values for the dummy transfer are: A single HIGH-Z byte, followed by 90Fh KEY2-encrypted 00h bytes. The KEY2 encryption stream is advanced for all 910h bytes, including for the HIGH-Z byte.
Note: Current cartridges are using 910h bytes, however, other carts might use other amounts of dummy bytes, the 910h value can be calculated based on ROM Control entries in cartridge header. For the KEY1 formulas, see:
DS Encryption by Gamecode/Idcode (KEY1)

KEY2 Command/Data Encryption
DS Encryption by Random Seed (KEY2)


 DS Cartridge I/O Ports

The Gamecard bus registers can be mapped to NDS7 or NDS9 via EXMEMCNT, see
DS Memory Control

40001A1h - Gamecard Bus Whatever
  0-5   Always zero
  6     Transfer Ready IRQ (0=Disable, 1=Enable) (see Port 40001A4h/Bit31)
  7     Always set (1=Enable?)
Maybe used to reset the cartridge? Maybe this (or another register) controls gamecard BACKUP memory?

40001A4h - Gamecard Bus ROMCTRL
  Bit   Expl.
  0-12  KEY1 length part1 (0-1FFFh) (forced min 08F8h by BIOS)
  13    Unknown?
  14    Unknown?
  15    Unknown? (read-only, or write-only?)
  16-21 KEY1 length part2 (0-3Fh)   (forced min 18h by BIOS)
  22    Unknown?
  23    Data-Word Status  (0=Busy, 1=Ready/DRQ) (Read-only)
  24-26 Data Block size   (0=None, 1..6=100h SHL (1..6) bytes, 7=4 bytes)
  27    Transfer CLK rate (0=6.7MHz=33.51MHz/5, 1=4.2MHz=33.51MHz/8)
  28    Secure Area Mode  (0=normal, 1=other)
  29    Unknown (always 1 ?)
  30    Unknown (always 0 ?)
  31    Block Start/Status (0=Ready, 1=Start/Busy) (IRQ See Port 40001A1h)
The cartridge header is booted at 4.2MHz CLK rate, and following transfers are then using ROMCTRL settings specified in cartridge header entries [060h] and [064h], which are usually using 6.7MHz CLK rate. When using other CLK rate, the Timeout in header [06Eh] must be probably adjusted accordingly.
Transfer length of null, four, and 200h..4000h bytes are supported by the console, however, regular cartridges support only max 1000h bytes.

40001A8h - Gamecard bus 8-byte Command Out
The separate commands are described in the Cartridge Protocol chapter, however, once when the BIOS boot procedure has completed, one would usually only need command "B7aaaaaaaa000000h", for reading data (usually 200h bytes) from address aaaaaaaah.
  0-7   1st Command Byte       (MSB, at 40001A8h)
  ...   2nd..7th Command Bytes (at 40001A9..1AEh)
  56-63 8th Command Byte       (LSB, at 40001AFh)
Observe that the command/parameter MSB is located at the smallest memory location (40001A8h), ie. compared with the CPU, the byte-order is reversed.

4100010h - Gamecard bus 4-byte Data In
  0-7   1st received Data Byte (at 4100010h)
  8-15  2nd received Data Byte (at 4100011h)
  16-23 3rd received Data Byte (at 4100012h)
  24-31 4th received Data Byte (at 4100013h)
After sending a command, data can be read from this register manually (when the DRQ bit is set), or by DMA (with DMASAD=4100010h, Fixed Source Address, Length=1, Size=32bit, Repeat=On, Mode=DS Gamecard).

40001B0h - 32bit - Encryption Seed 0 Lower 32bit
40001B4h - 32bit - Encryption Seed 1 Lower 32bit
40001B8h - 16bit - Encryption Seed 0 Upper 7bit (bit8-15 unused)
40001BAh - 16bit - Encryption Seed 1 Upper 7bit (bit8-15 unused)
These registers are used by the NDS7 BIOS to initialize KEY2 encryption,
after that, the registers should not (cannot?) be changed.
DS Encryption by Random Seed (KEY2)


 DS Cartridge NitroROM File System

NitroROM
The NitroROM Filesystem is used by many commercial games, at least those that have been developed with Nintendo's tools. The filesystem allows to load data from cartridge ROM by filenames and/or by Overlay IDs.
However, the DS hardware, BIOS, and Firmware do NOT contain any built-in filesystem functions. The ARM9/ARM7 boot code (together max 3903KB), and Icon/Title information are automatically loaded on power-up.
Programs that require to load additional data from cartridge ROM may do that either by implementing whatever functions to translate filenames to ROM addresses, or by reading from ROM directly.

File Allocation Table (FAT) (base/size defined in cart header)
Contains ROM addresses for up to 61440 files (File IDs 0000h and up).
  Addr Size Expl.
  00h  4    Start address of file in ROM (8000h and up)    (0=Unused Entry)
  04h  4    End address of file in ROM   (Start+Len...-1?) (0=Unused Entry)
Directories are fully defined in FNT area, and do not require FAT entries.

File Name Table (FNT) (base/size defined in cart header)
Consists of the FNT Directory Table, followed by one or more FNT Sub-Tables.
To interprete the directory tree: Start at the 1st Main-Table entry, which is referencing to a Sub-Table, any directories in the Sub-Table are referencing to Main-Table entries, which are referencing to further Sub-Tables, and so on.

FNT Directory Main-Table (base=FNT+0, size=[FNT+06h]*8)
Consists of a list of up to 4096 directories (Directory IDs F000h and up).
  Addr Size Expl.
  00h  4    Offset to Sub-table             (originated at FNT base)
  04h  2    ID of first file in Sub-table   (0000h..EFFFh)
For first entry (ID F000h, root directory):
  06h  2    Total Number of directories     (1..4096)
Further entries (ID F001h..FFFFh, sub-directories):
  06h  2    ID of parent directory (F000h..FFFEh)

FNT Sub-tables (base=FNT+offset, ends at Type/Length=00h)
Contains ASCII names for all files and sub-directories within a directory.
  Addr Size Expl.
  00h  1    Type/Length
              01h..7Fh File Entry          (Length=1..127, without ID field)
              81h..FFh Sub-Directory Entry (Length=1..127, plus ID field)
              00h      End of Sub-Table
              80h      Reserved
  01h  LEN  File or Sub-Directory Name, case-sensitive, without any ending
              zero, ASCII 20h..7Eh, except for characters \/?"<>*:;|
Below for Sub-Directory Entries only:
  LEN+1 2    Sub-Directory ID (F001h..FFFFh) ;see FNT+(ID AND FFFh)*8
File Entries do not have above ID field. Instead, File IDs are assigned in incrementing order (starting at the "First ID" value specified in the Directory Table).

ARM9 and ARM7 Overlay Tables (OVT) (base/size defined in cart header)
Somehow related to Nintendo's compiler, allows to assign compiler Overlay IDs to filesystem File IDs, and to define additional information such like load addresses.
  Addr Size Expl.
  00h  4    Overlay ID
  04h  4    RAM Address ;Point at which to load
  08h  4    RAM Size    ;Amount to load
  0Ch  4    BSS Size    ;Size of BSS data region
  10h  4    Static initialiser start address
  14h  4    Static initialiser end address
  18h  4    File ID  (0000h..EFFFh)
  1Ch  4    Reserved (zero)

Cartridge Header
The base/size of FAT, FNT, OVT areas is defined in cartridge header,
DS Cartridge Header


 DS Cartridge PassMe/PassThrough

PassMe is an adapter connected between the DS and an original NDS cartridge, used to boot unencrypted code from a flash cartridge in the GBA slot, it replaces the following entries in the original NDS cartridge header:
  Addr  Siz Patch
  004h  4   E59FF018h  ;opcode LDR PC,[027FFE24h] at 27FFE04h
  01Fh  1   04h        ;set autostart bit
  022h  1   01h        ;set ARM9 rom offset to nn01nnnnh (above secure area)
  024h  4   027FFE04h  ;patch ARM9 entry address to endless loop
  034h  4   080000C0h  ;patch ARM7 entry address in GBA slot
  15Eh  2   nnnnh      ;adjust header crc16
After having verified the encrypted chip IDs (from the original cartridge), the console thinks that it has successfully loaded a NDS cartridge, and then jumps to the (patched) entrypoints.

GBA Flashcard Format
Although the original PassMe requires only the entrypoint, PassMe programs should additionally contain one (or both) of the ID values below, allowing firmware patches to identify & start PassMe games without real PassMe hardware.
  0A0h  GBA-style Title    ("DSBooter")
  0ACh  GBA-style Gamecode ("PASS")
  0C0h  ARM7 Entrypoint    (32bit ARM code)
Of course, that applies only to early homebrew programs, newer games should use normal NDS cartridge headers.

ARM9 Entrypoint
The GBA-slot access rights in the EXMEMCNT register are initially assigned to the ARM7 CPU, so the ARM9 cannot boot from the flashcard, instead it is switched into an endless loop in Main RAM (which contains a copy of the cartridge header at 27FFE00h and up). The ARM7 must thus copy ARM9 code to Main RAM, and the set the ARM9 entry address by writing to [027FFE24h].


 DS Encryption by Gamecode/Idcode (KEY1)

KEY1 - Gamecode / Idcode Encryption
The KEY1 encryption relies only on the gamecode (or firmware idcode), it does not contain any random components. The fact that KEY1 encrypted commands appear random is just because the <unencrypted> commands contain random values, so the encryption result looks random.

KEY1 encryption is used for KEY1 encrypted gamecart commands (ie. for loading the secure area). It is also used for resolving the extra decryption of the first 2K of the secure area, and for firmware decryption, and to decode some encrypted values in gamecart/firmware header.

Below are KEY1 encryption formulas. The formulas can be used only with a copy of the key table (1048h bytes) from the NDS ARM7 BIOS (address 0030h..1077h).

crypt_64bit_up(ptr) / crypt_64bit_down(ptr)
  Y=[ptr+0]
  X=[ptr+4]
  FOR I=0 TO 0Fh (up), or FOR I=11h TO 02h (down)
    Z=[keybuf+I*4] XOR X
    X=[keybuf+048h+((Z SHR 24) AND FFh)*4]
    X=[keybuf+448h+((Z SHR 16) AND FFh)*4] + X
    X=[keybuf+848h+((Z SHR  8) AND FFh)*4] XOR X
    X=[keybuf+C48h+((Z SHR  0) AND FFh)*4] + X
    X=Y XOR X
    Y=Z
  NEXT I
  [ptr+0]=X XOR [keybuf+40h], or [ptr+0]=X XOR [keybuf+4h] (down)
  [ptr+4]=Y XOR [keybuf+44h], or [ptr+4]=Y XOR [keybuf+0h] (down)

apply_keycode(modulo)
  crypt_64bit_up(keycode+4)
  crypt_64bit_up(keycode+0)
  [scratch]=0000000000000000h   ;S=0 (64bit)
  FOR I=0 TO 44h STEP 4         ;xor with reversed byte-order (bswap)
    [keybuf+I]=[keybuf+I] XOR bswap_32bit([keycode+(I MOD modulo)])
  NEXT I
  FOR I=0 TO 1040h STEP 8
    crypt_64bit_up(scratch)     ;encrypt S (64bit) by keybuf
    [keybuf+I]=[scratch]        ;write S (64bit) to keybuf
  NEXT I

init_keycode(idcode,level,modulo)
  copy [arm7bios+0030h..1077h] to [keybuf+0..1047h]
  [keycode+0]=[idcode]
  [keycode+4]=[idcode]/2
  [keycode+8]=[idcode]*2
  IF level>=1 THEN apply_keycode(modulo) ;first apply (always)
  IF level>=2 THEN apply_keycode(modulo) ;second apply (optional)
  [keycode+4]=[keycode+4]*2
  [keycode+8]=[keycode+8]/2
  IF level>=3 THEN apply_keycode(modulo) ;third apply (optional)

firmware_decryption
  init_keycode(firmware_header+08h,1,0Ch) ;idcode (usually "MACP"), level 1
  crypt_64bit_down(firmware_header+18h)   ;rominfo
  init_keycode(firmware_header+08h,2,0Ch) ;idcode (usually "MACP"), level 2
  decrypt ARM9 and ARM7 bootcode by crypt_64bit_down (each 8 bytes)
  decompress ARM9 and ARM7 bootcode by LZ77 function (swi)
  calc CRC16 on decrypted/decompressed ARM9 bootcode followed by ARM7 bootcode
Note: The sizes of the compressed/encrypted bootcode areas are unknown (until they are fully decompressed), one way to solve that problem is to decrypt the next 8 bytes each time when the decompression function requires more data.

gamecart_decryption
  init_keycode(cart_header+0Ch,1,08h)   ;gamecode, level 1, modulo 8
  crypt_64bit_down(cart_header+78h)     ;rominfo (secure area disable)
  init_keycode(cart_header+0Ch,2,08h)   ;gamecode, level 2, modulo 8
  crypt_64bit_up all KEY1 commands (1st command byte in MSB of 64bit value)
  after loading the secure_area, calculate secure_area crc, then
  crypt_64bit_down(secure_area+0)       ;first 8 bytes of secure area
  init_keycode(cart_header+0Ch,3,08h)   ;gamecode, level 3, modulo 8
  crypt_64bit_down(secure_area+0..7F8h) ;each 8 bytes in first 2K of secure
After decryption, the ID field in the first 8 bytes should be "encryObj", if it matches then first 8 bytes are filled with E7FFDEFFh, otherwise the whole 2K are filled by that value.

Gamecart Command Register
Observe that the byte-order of the command register is reversed.
The CPU's STR [addr] opcode writes the MSB to [addr+0], and LSB to [addr+3], and, when using the "crypt_64bit_up" function for KEY1-encrypted commands, the MSB of the 64bit command should be at [addr+7], the LSB at [addr+0].
Respectively, the value in <memory> should be transferred MSB first, however, the DS transfers the <register> value LSB first. Thus, the byte order must be reversed when copying the value from memory to the command register.

Note
The KEY1 encryption is based on Bruce Schneier's "Blowfish Encryption Algorithm".


 DS Encryption by Random Seed (KEY2)

KEY2 39bit Seed Values
The pre-initialization settings at cartridge-side (after reset) are:
  Seed0 = 58C56DE0E8h
  Seed1 = 5C879B9B05h
The post-initialization settings (after sending command 4llllmmmnnnkkkkkh to the cartridge, and after writing the Seed values to Port 40001Bxh) are:
  Seed0 = (mmmnnn SHL 15)+6000h+Seedbyte
  Seed1 = 5C879B9B05h
The seedbyte is selected by Cartridge Header [013h].Bit0-2, this index value should be usually in range 0..5, however, possible values for index 0..7 are: E8h,4Dh,5Ah,B1h,17h,8Fh,99h,D5h.
The 24bit random value (mmmnnn) is derived from the real time clock setting, and also scattered by KEY1 encryption, anyways, it's just random and doesn't really matter where it comes from.

KEY2 Encryption
Relies on two 39bit registers (x and y), which are initialized as such:
  x = reversed_bit_order(seed0)  ;ie. LSB(bit0) exchanged with MSB(bit38), etc.
  y = reversed_bit_order(seed1)
During transfer, x, y, and transferred data are modified as such:
  x = (((x shr 5)xor(x shr 17)xor(x shr 18)xor(x shr 31)) and 0FFh)+(x shl 8)
  y = (((y shr 5)xor(y shr 23)xor(y shr 18)xor(y shr 31)) and 0FFh)+(y shl 8)
  data = (data xor x xor y) and 0FFh


 DS Firmware Serial Flash Memory

ST Microelectronics SPI Bus Compatible Serial Flash Memory
  ST M45PE20 - ID 20h, 40h, 12h - 256 KBytes (used in original DS)
  ST M25PE40 - ID 20h, 80h, 13h - 512 KBytes (used in some/all newer DS)
More than 100,000 Write Cycles, more than 20 Year Data Retention
The Firmware Flash Memory is accessed via SPI bus,
DS Serial Peripheral Interface Bus (SPI)

Instruction Codes
  06h  WREN Write Enable (No Parameters)
  04h  WRDI Write Disable (No Parameters)
  9Fh  RDID Read JEDEC Identification (Read 1..3 ID Bytes)
             (Manufacturer, Device Type, Capacity)
  05h  RDSR Read Status Register (Read Status Register, endless repeated)
             Bit7-2  Not used (zero)
             Bit1    WEL Write Enable Latch             (0=No, 1=Enable)
             Bit0    WIP Write/Program/Erase in Progess (0=No, 1=Busy)
  03h  READ Read Data Bytes (Write 3-Byte-Address, read endless data stream)
  0Bh  FAST Read Data Bytes at Higher Speed (Write 3-Byte-Address, write 1
             dummy-byte, read endless data stream) (max 25Mbit/s)
  0Ah  PW   Page Write (Write 3-Byte-Address, write 1..256 data bytes)
             (changing bits to 0 or 1) (reads unchanged data, erases the page,
             then writes new & unchanged data) (11ms typ, 25ms max)
  02h  PP   Page Program (Write 3-Byte-Address, write 1..256 data bytes)
             (changing bits from 1 to 0) (1.2ms typ, 5ms max)
  DBh  PE   Page Erase 100h bytes (Write 3-Byte-Address) (10ms typ, 20ms max)
  D8h  SE   Sector Erase 10000h bytes (Write 3-Byte-Address) (1s typ, 5s max)
  B9h  DP   Deep Power-down (No Parameters) (consumption 1uA typ, 10uA max)
             (3us) (ignores all further instructions, except RDP)
  ABh  RDP  Release from Deep Power-down (No Parameters) (30us)
Write/Program may not cross page-boundaries. Write/Program/Erase are rejected during first 1..10ms after power up. The WEL bit is automatically cleared on Power-Up, on /Reset, and on completion of WRDI/PW/PP/PE/SE instructions. WEL is set by WREN instruction (which must be issued before any write/program/erase instructions). Don't know how RDSR behaves when trying to write to the write-protected region?

Communication Protocol
  Set Chip Select LOW to invoke the command
  Transmit the instruction byte
  Transmit any parameter bytes
  Transmit/receive any data bytes
  Set Chip Select HIGH to finish the command
All bytes (and 3-byte addresses) transferred most significant bit/byte first.

Pin-Outs
  1   D    Serial Data In (latched at rising clock edge)          _________
  2   C    Serial Clock (max 25MHz)                             /|o        |
  3   /RES Reset                                            1 -| |         |- 8
  4   /S   Chip Select (instructions start at falling edge) 2 -| |         |- 7
  5   /W   Write Protect (makes first 256 pages read-only)  3 -| |_________|- 6
  6   VCC  Supply (2.7V..3.6V typ) (4V max)                 4 -|/          |- 5
  7   VSS  Ground                                              |___________|
  8   Q    Serial Data Out (changes at falling clock edge)


 DS Firmware Header

Firmware Memory Map
  00000h-001FFh  Firmware Header
  00200h-3FDFFh  Firmware Code/Data
  3FE00h-3FEFFh  User Settings Area 1
  3FF00h-3FFFFh  User Settings Area 2

Firmware Header (00000h-001FFh)
  Addr Size Expl.
Memory Addresses...
  000h 2    part3 romaddr/8 (arm9 code) (LZ/huffman compression)
  002h 2    part4 romaddr/8 (arm7 code) (LZ/huffman compression)
  004h 2    part3/4 CRC16
  006h 2    part1/2 CRC16 arm9/7 boot code
  008h 4    firmware identifier (usually nintendo "MACP") (or nocash "XBOO")
  00Ch 2    part1 arm9 boot code romaddr/2^(2+shift1) (LZSS compressed)
  00Eh 2    part1 arm9 boot code 2800000h-ramaddr/2^(2+shift2)
  010h 2    part2 arm7 boot code romaddr/2^(2+shift3) (LZSS compressed)
  012h 2    part2 arm7 boot code 3810000h-ramaddr/2^(2+shift4)
  014h 2    bit0-2=shift1, bit3-5=shift2, bit6-8=shift3, bit9-11=shift4
  016h 2    part5 data/gfx romaddr/8 (LZ/huffman compression)
  018h 8    Unknown (xx xx xx 1x 04 FF FF FF)
            (or encrypted "enPngOFF"=Cartridge KEY2 Disable)
  020h 2    User Settings Offset (div8) (usually 3FE00h/8)
  022h 2     C0 7E   Whatever (maybe div8) (7EC0h*8 = 3F600h ?)
             (maybe size of used memory, excluding header..)
  024h 2     40 7E       (=3F200h)
  026h 2    part5 data/gfx CRC16
WLAN Calibration Data...
  028h 2    Unknown/unused (FFh-filled)
  02Ah 2    CRC16 (with initial value 0) of [2Ch..2Ch+config_length-1]
  02Ch 2    config_length (usually 0138h)
  02Eh 8    Unknown/unused (00h-filled)
  036h 6    00 09 BF xx xx xx         48bit MAC address (WLAN)
  03Ch 2    list of enabled channels ANDed with 7FFE,
             each bit represents one channel
  03Eh 2    ?
  040h 1    ?
  041h 1    ?
  042h 1    ?
  043h 1    ?
  044h 20h  list of 16 2-byte MAC reg values for a list of 16 hardcoded MAC
            reg addresses (see list below) related to RX & TX
  064h 69h  list of 1-byte BBP reg values for BBP regs 0..x
  0CDh 1    Unknown
  0CEh ?    list of 1-byte RF values
  0D4h 6    Same as 1st entry at 00F2h
  0DAh 18h  Unknown
  0F2h 54h  list of 14 2x3-byte RF values for each channel (15 channels?),
            related to channel frequency, RF_Write( first 3-byte value),
            RF_Write( next 3-byte value)
  146h E    list of 14 1-byte BBP reg values, seem to go into BBP reg 1Eh
  154h E    list of 14 1-byte RF values, also channel frequency related,
            only lower 5 bits are used
  162h 1    Unknown
  163h 9Dh  Unused (FFh-filled)


 DS Firmware User Settings

Current Settings (RAM 27FFC80h-27FFCFFh)
User Settings 0 (Firmware 3FE00h-3FEFFh)
User Settings 1 (Firmware 3FF00h-3FFFFh)
  Addr Size Expl.
  000h  2   Version (5)
  002h  1   Favorite color (0..15) (0=Gray, 1=Brown, etc.)
  003h  1   Birthday month (1..12) (Binary, non-BCD)
  004h  1   Birthday day   (1..31) (Binary, non-BCD)
  005h  1
  006h  20  Nickname string in UTF-16 format
  01Ah  2   Nickname length in characters    (0..10)
  01Ch  52  Message string in UTF-16 format
  050h  2   Message length in characters     (0..26)
  052h  1   Alarm hour     (0..23) (Binary, non-BCD)
  053h  1   Alarm minute   (0..59) (Binary, non-BCD)
  054h
  056h  1   80h=enable alarm (huh?), bit 0..6=enable?
  057h      Zero (1 byte)
  058h  2x2 touch-screen calibration point (adc.x1,y1) 12bit ADC-position
  05Ch  2x1 touch-screen calibration point (scr.x1,y1) 8bit pixel-position
  05Eh  2x2 touch-screen calibration point (adc.x2,y2) 12bit ADC-position
  062h  2x1 touch-screen calibration point (scr.x2,y2) 8bit pixel-position
  064h  2   Language and Flags (see below)
  066h  2   Unknown
  068h  4   RTC Offset (difference in seconds when RTC time/date was changed)
  06Ch  4   FFh-filled
  070h  1/2 update counter (used to check latest)
  072h  2   CRC16 of entries 00h..6Fh
  074h  ..  FFh-filled

Language and Flags (Entry 64h)
  Bit
  0..2 Language (0=Japanese, 1=English, 2=French, 3=German,
       4=Italian, 5=Spanish) (this also implies time/data format)
  3    GBA mode screen selection. 0=upper, 1=lower
  6    Bootmenu Disable   (0=Manual/bootmenu, 1=Autostart game)
  9    User Settings Lost (0=Normal, 1=Prompt/Settings Lost)
  13
  14
 The Health and Safety message is skipped if Bit9=1, or if
 one or more of the following bits is zero: Bits 10,11,13,14,15.
 However, as soon as entering the bootmenu, the Prompt occurs.
Bit9: Prompt for User Info, and Language, and Calibration
Bit10: Prompt for User Info
Bit11: Same as Bit10
Bit12: No function
Bit13: Prompt for User Info, and Language
Bit14: Same as Bit10
Bit15: Same as Bit10


Note: There are two User Settings areas in the firmware chip, at offset 3FE00h and 3FF00h, if both areas have valid CRCs, then the current/newest area is that whose Update Counter is one bigger than in the other/older area.
 IF count1=((count0+1) AND 7Fh) THEN area1=newer ELSE area0=newer
When changing settings, the older area is overwritten with new data (and incremented Update Counter). The two areas allow to recover previous settings in case of a write-error (eg. on a battery failure during write).

Battery Removal
Even though the battery is required only for the RTC (not for the firmware flash memory), most of the firmware user settings are reset when removing the battery. This appears to be a strange bug-or-feature of the DS bios, at least, fortunately, it still keeps the rest of the firmware intact.


 DS Firmware Extended Settings

Extended Settings contain some additional information which is not supported by the original firmware (current century, date/time formats, temperature calibration, etc.), the settings are supported by Nocash Firmware, by the no$gba emulator, and may be eventually also supported by other emulators. If present, the values can be used by games, otherwise games should use either whatever default settings, or contain their own configuration menu.

Extended Settings - loaded to 23FEE00h (aka fragments of NDS9 boot code)
  Addr Siz Expl.
  00h  8  ID "XbooInfo"
  08h  2  CRC16 Value [0Ch..0Ch+Length-1]
  0Ah  2  CRC16 Length (from 0Ch and up)
  0Ch  1  Version (currently 01h)
  0Dh  1  Update Count (newer = (older+1) AND FFh)
  0Eh  1  Bootmenu Flags
            Bit6   Important Info  (0=Disable, 1=Enable)
            Bit7   Bootmenu Screen (0=Upper, 1=Lower)
  0Fh  1  GBA Border (0=Black, 1=Gray Line)
  10h  2  Temperature Calibration TP0 ADC value  (x16) (sum of 16 ADC values)
  12h  2  Temperature Calibration TP1 ADC value  (x16) (sum of 16 ADC values)
  14h  2  Temperature Calibration Degrees Kelvin (x100) (0=none)
  16h  1  Temperature Flags
            Bit0-1 Format (0=Celsius, 1=Fahrenheit, 2=Reaumur, 3=Kelvin)
  17h  1  Backlight Intensity (0=0ff .. FFh=Full)
  18h  4  Date Century Offset       (currently 20, for years 2000..2099)
  1Ch  1  Date Month Recovery Value (1..12)
  1Dh  1  Date Day Recovery Value   (1..31)
  1Eh  1  Date Year Recovery Value  (0..99)
  1Fh  1  Date/Time Flags
            Bit0-1 Date Format   (0=YYYY-MM-DD, 1=MM-DD-YYYY, 2=DD-MM-YYYY)
            Bit2   Friendly Date (0=Raw Numeric, 1=With Day/Month Names)
            Bit5   Time DST      (0=Hide DST, 1=Show DST=On/Off)
            Bit6   Time Seconds  (0=Hide Seconds, 1=Show Seconds)
            Bit7   Time Format   (0=24 hour, 1=12 hour)
  20h  1  Date Separator      (Ascii, usually Slash, or Dot)
  21h  1  Time Separator      (Ascii, usually Colon, or Dot)
  22h  1  Decimal Separator   (Ascii, usually Comma, or Dot)
  23h  1  Thousands Separator (Ascii, usually Comma, or Dot)
  24h  1  Daylight Saving Time (Nth)
             Bit 0-3 Activate on (0..4 = Last,1st,2nd,3rd,4th)
             Bit 4-7 Deactivate on (0..4 = Last,1st,2nd,3rd,4th)
  25h  1  Daylight Saving Time (Day)
             Bit 0-3 Activate on (0..7 = Mon,Tue,Wed,Thu,Fri,Sat,Sun,AnyDay)
             Bit 4-7 Deactivate on (0..7 = Mon,Tue,Wed,Thu,Fri,Sat,Sun,AnyDay)
  26h  1  Daylight Saving Time (of Month)
             Bit 0-3 Activate DST in Month   (1..12)
             Bit 4-7 Deactivate DST in Month (1..12)
  27h  1  Daylight Saving Time (Flags)
             Bit 0   Current DST State (0=Off, 1=On)
             Bit 1   Adjust DST Enable (0=Disable, 1=Enable)
Note: With the original firmware, the memory region at 23FEE00h and up contains un-initialized, non-zero-filled data (fragments of boot code).


 DS Backwards-compatible GBA-Mode

When booting a 32pin GBA cartridge, the NDS is automatically switched into GBA mode, in that mode all NDS related features are disabled, and the console behaves (almost) like a GBA.

GBA Features that are NOT supported on NDS in GBA Mode.
Unlike real GBA, the NDS does not support 8bit DMG/CGB cartridges.
The undocumented Internal Memory Control register (Port 800h) isn't supported, so the NDS doesn't allow to use 'overclocked' RAM.
The NDS doesn't have a link-port, so GBA games can be played only in single player mode, link-port accessories cannot be used, and the NDS cannot run GBA code via multiboot.

GBA Features that are slightly different on NDS in GBA Mode.
The CPU, Timers, and Sound Frequencies are probably clocked at 16.76MHz; 33.51Mhz/2; a bit slower than the original GBA's 16.78MHz clock?
In the BIOS, a single byte in a formerly 00h-filled area has been changed from 00h to 01h, resulting in SWI 0Dh returning a different BIOS checksum.
The GBA picture can be shown on upper or lower screen (selectable in boot-menu), the backlight for the selected screen is always on, resulting in different colors & much better visibility than original GBA. Unlike GBA-SP, the NDS doesn't have a backlight-button.

Screen Border in GBA mode
The GBA screen is centered in the middle of the NDS screen. The surrounding pixels are defined by 32K-color bitmap data in VRAM Block A and B. Each frame, the GBA picture is captured into one block, and is displayed in the next frame (while capturing new data to the other block).
To get a flicker-free border, both blocks should be initialized to contain the same image before entering GBA mode (usually both are zero-filled, resulting in a plain black border).
Note: When using two different borders, the flickering will be irregular - so there appears to be a frame inserted or skipped once every some seconds in GBA mode?!

Switching from NDS Mode to GBA Mode
  --- NDS9: ---
  ZEROFILL VRAM A,B     ;init black screen border (or other color/image)
  POWCNT=8003h          ;enable 2D engine A on upper screen (0003h=lower)
  EXMEMCNT=...          ;set Async Main Memory mode (clear bit14)
  IME=0                 ;disable interrupts
  SWI 06h               ;halt with interrupts disabled (lockdown)
  --- NDS7: ---
  POWERMAN.REG0=09h     ;enable sound amplifier & upper backlight (05h=lower)
  IME=0                 ;disable interrupts
  wait for VCOUNT=200   ;wait until VBlank
  SWI 1Fh with R2=40h   ;enter GBA mode, by CustomHalt(40h)
After that, the GBA BIOS will be booted, the GBA Intro will be displayed, and the GBA cartridge (if any) will be started.


 DS Xboo

DS XBOO Connection
  Console Pin/Names             Parallel Port Pin/Names
  RFU.9    FMW.1 D    ---|>|--- DSUB.14    CNTR.14    AutoLF
  RFU.6    FMW.2 C    ---|>|--- DSUB.1     CNTR.1     Strobe
  RFU.10   FMW.3 /RES ---|>|--- DSUB.16    CNTR.31    Init
  RFU.7    FMW.4 /S   ---|>|--- DSUB.17    CNTR.36    Select
  RFU.5    FMW.5 /W   --. SL1A  -          -          N.C.
  RFU.28   FMW.6 VCC  __| SL1B  -          -          N.C.
  RFU.2,12 FMW.7 VSS  --------- DSUB.18-25 CNTR.19-30 Ground
  RFU.8    FMW.8 Q    --------- DSUB.11    CNTR.11    Busy
  P00 Joypad-A        --------- DSUB.2     CNTR.2     D0
  P01 Joypad-B        --------- DSUB.3     CNTR.3     D1
  P02 Joypad-Select   --------- DSUB.4     CNTR.4     D2
  P03 Joypad-Start    --------- DSUB.5     CNTR.5     D3
  P04 Joypad-Right    --------- DSUB.6     CNTR.6     D4
  P05 Joypad-Left     --------- DSUB.7     CNTR.7     D5
  P06 Joypad-Up       --------- DSUB.8     CNTR.8     D6
  P07 Joypad-Down     --------- DSUB.9     CNTR.9     D7
  RTC.1 INT aka SI    --------- DSUB.10    CNTR.10    /Ack
Parts List: 15 wires, 4 "BAT 85" diodes, 1 parallel port socket.

DS XBOO Connection Notes
The Firmware chip (FMW.Pins) hides underneath of the RFU shielding plate, so it'd be easier to connect the wires to the RFU.Pins. The easiest way for the /W-to-VCC connection is to shortcut SL1 by putting some solder onto it.
The P00..P07 and INT signals are labeled on the switch-side of the mainboard, however, there should be more room for the cables when connecting them to via's at the bottom-side, image below may help to locate that pins,
  http://nocash.emubase.de/nds-pins.gif (3.5 KBytes)
At the parallel port side, DSUB.Pins or CNTR.Pins can be used for 25pin DSUB or 36pin Centronics sockets, the latter one allowing to use a standard printer cable.
The ring printed on the diodes is pointing towards parallel port side, the diodes are required to prevent the parallel port to pull-up LOW levels on the NDS side, be sure to use BAT85 diodes, cheaper ones like 1N4148 are loosing too much voltage and won't work stable.

DS XBOO Operation Notes
The main Upload function is found in no$gba Utility menu, together with further functions in Remote Access sub-menu.
Before uploading anything: download the original firmware, the file is saved as FIRMnnnn.BIN, whereas "nnnn" is equal to the last 16bit of the consoles 48bit MAC address, so Firmware-images from different consoles are having unique filenames. If you don't already have, also download the NDS BIOS, the BIOS contains encryption seed data required to encrypt/decrypt secure area; without having downloaded the BIOS, no$gba will be working only with unencrypted ROM-images. Next, select Patch Firmware to install the nocash firmware.

DS XBOO Troubleshooting
Be sure that the console is switched on, and that the XBOO cable is connected, and that you have selected the correct parallel port in no$gba setup (the "multiboot" options in Various Options screen), and, of course, try avoid to be fiddling with the joypad during uploads.
I've tested the cable on two computers, the overall upload/download stuff should work stable. The firmware access functions - which are required only for (un-)installation - worked only with one of the two computers; try using a different computer/parallel port in case of problems.

Nocash Firmware
The primary purpose is to receive uploaded NDS-images via parallel port connection, additionally it's containing bootmenu and setup screens similar to the original firmware. The user interface is having less cryptic symbols and should be alltogether faster and easier to use. Important Information about Whatever is supported (but it can be disabled). The setup contains a couple of additional options like automatic daylight saving time adjustment.
The bootmenu allows to boot normal NDS and GBA carts, it does additionally allow to boot NDS-images (or older PassMe-images) from flashcards in GBA slot. Furthermore, benefits of asm coding, the nocash firmware occupies less than 32KBytes, allowing to store (and boot) smaller NDS-images in the unused portion of the firmware memory (about 224KBytes), the zero-filled region between cart header and secure area, at 200h..3FFFh, is automatically excluded, so the image may be slightly bigger than the available free memory space.

Missing
Unlike the original firmware, the current version cannot boot via WLAN. Help or hardware would be welcome (I am having only one NDS, and my PC doesn't have WLAN, so I cannot test any WLAN transfers right now).


 About this Document

About
GBATEK written 2001-2006 by Martin Korth, programming specs for the GBA and NDS hardware, I've been trying to keep the specs both as short as possible, and as complete as possible. The document is part of the no$gba debuggers built-in help text.

Updates
The standalone docs in TXT and HTM format are updated when having added any major changes to the document. The no$gba built-in version will be updated more regularly, including for minor changes, along with all no$gba updates.

Homepage
http://nocash.emubase.de/gba.htm - no$gba homepage
http://nocash.emubase.de/gbatek.htm - gbatek html version
http://nocash.emubase.de/gbatek.txt - gbatek text version

Feedback
If you find any information in this document to be misleading, incomplete, or incorrect, please say something! My email address will be eventually found on the no$gba webpage - depending on the amount of incoming garbage; mail from programmers only, please.

Credits
Thanks for GBATEK fixes, and for info about GBA and NDS hardware,
- Christian Auby
- Damien Good
- DarkFader
- Dstek.xml, funny name, strange non-txt/htm format :-(
- Jeff Frohwein
- NDSTech Wiki, http://www.bottledlight.com/ds/
- Randy Linden
- Sebastian Rasmussen