chore: add minimum zig version

still not ideal imo

.github/workflows/main.yml

@@ -0,0 +1,59 @@
+name: Nightly
+
+on:
+  push:
+    paths:
+      - "**.zig"
+      - "dl_sdl2.ps1"
+    branches:
+      - main
+  schedule:
+    - cron: '0 0 * * *'
+  workflow_dispatch:
+
+jobs:
+  build:
+    strategy:
+      matrix:
+        os: [ubuntu-latest, windows-latest] # TODO: Figure out Apple Silicon macOS
+    runs-on: ${{matrix.os}}
+    steps:
+      - uses: goto-bus-stop/setup-zig@v2
+        with:
+          version: 0.13.0
+      - run: |
+            git config --global core.autocrlf false
+      - uses: actions/checkout@v3
+        with:
+          submodules: recursive
+      - name: prepare-linux
+        if: runner.os == 'Linux'
+        run: |
+            sudo apt update
+            sudo apt install libgtk-3-dev libsdl2-dev
+      - name: prepare-windows
+        if: runner.os == 'Windows'
+        run: |
+            .\dl_sdl2.ps1
+      - name: prepare-macos
+        if: runner.os == 'macOS'
+        run: |
+            brew install sdl2
+      - name: build 
+        run: zig build -Doptimize=ReleaseSafe -Dcpu=baseline
+      - name: upload
+        uses: actions/upload-artifact@v3
+        with:
+          name: zba-${{matrix.os}}
+          path: zig-out
+
+  lint:
+    runs-on: ubuntu-latest
+    steps: 
+      - uses: actions/checkout@v3
+        with:
+          submodules: recursive
+      - uses: goto-bus-stop/setup-zig@v2
+        with:
+          version: 0.13.0
+      - run: zig fmt --check {src,lib}/**/*.zig build.zig build.zig.zon

.gitignore

@@ -1,7 +1,8 @@
 /.vscode
 /bin
-/zig-cache
-/zig-out
+**/zig-cache
+**/.zig-cache
+**/zig-out
 /docs
 **/*.log
 **/*.bin
@@ -11,4 +12,8 @@
 /lib/SDL2
 
 # Any Custom Scripts for Debugging purposes
-*.sh
+*.sh
+
+
+# Dear ImGui
+**/imgui.ini

.gitmodules

@@ -1,15 +1,3 @@
 [submodule "lib/SDL.zig"]
 	path = lib/SDL.zig
-	url = https://github.com/MasterQ32/SDL.zig
-[submodule "lib/zig-clap"]
-	path = lib/zig-clap
-	url = https://github.com/Hejsil/zig-clap
-[submodule "lib/known-folders"]
-	path = lib/known-folders
-	url = https://github.com/ziglibs/known-folders
-[submodule "lib/zig-datetime"]
-	path = lib/zig-datetime
-	url = https://github.com/frmdstryr/zig-datetime
-[submodule "lib/zig-toml"]
-	path = lib/zig-toml
-	url = https://github.com/aeronavery/zig-toml
+	url = https://github.com/paoda/SDL.zig

.vscode/extensions.json

@@ -1,8 +0,0 @@
-{
-    "recommendations": [
-        "augusterame.zls-vscode",
-        "usernamehw.errorlens",
-        "vadimcn.vscode-lldb",
-        "dan-c-underwood.arm"
-    ]
-}

README.md

@@ -1,69 +1,96 @@
 # ZBA (working title)
-An in-progress Game Boy Advance Emulator written in Zig ⚡!
 
-## Tests 
-- [ ] [jsmolka's GBA Test Collection](https://github.com/jsmolka/gba-tests)
-    - [x] `arm.gba` and `thumb.gba`
-    - [x] `flash64.gba`, `flash128.gba`, `none.gba`, and `sram.gba`
-    - [x] `hello.gba`, `shades.gba`, and `stripes.gba`
-    - [x] `memory.gba`
-    - [x] `bios.gba`
-    - [ ] `nes.gba`
-- [ ] [DenSinH's GBA ROMs](https://github.com/DenSinH/GBARoms)
-    - [x] `eeprom-test` and `flash-test`
-    - [x] `midikey2freq`
-    - [ ] `swi-tests-random`
-- [ ] [destoer's GBA Tests](https://github.com/destoer/gba_tests)
-    - [x] `cond_invalid.gba`
-    - [x] `dma_priority.gba`
-    - [x] `hello_world.gba`
-    - [x] `if_ack.gba`
-    - [ ] `line_timing.gba`
-    - [ ] `lyc_midline.gba`
-    - [ ] `window_midframe.gba`
-- [x] [ladystarbreeze's GBA Test Collection](https://github.com/ladystarbreeze/GBA-Test-Collection)
-    - [x] `retAddr.gba`
-    - [x] `helloWorld.gba`
-    - [x] `helloAudio.gba`
-- [x] [`armwrestler-gba-fixed.gba`](https://github.com/destoer/armwrestler-gba-fixed)
-- [x] [FuzzARM](https://github.com/DenSinH/FuzzARM)
+A Game Boy Advance Emulator written in Zig ⚡!
 
-## Resources
-* [GBATEK](https://problemkaputt.de/gbatek.htm)
-* [TONC](https://coranac.com/tonc/text/toc.htm)
-* [ARM Architecture Reference Manual](https://www.intel.com/content/dam/www/programmable/us/en/pdfs/literature/third-party/ddi0100e_arm_arm.pdf)
-* [ARM7TDMI Data Sheet](https://www.dca.fee.unicamp.br/cursos/EA871/references/ARM/ARM7TDMIDataSheet.pdf)
+![ZBA running リズム天国](assets/screenshot.png)
+
+## Scope
+
+I'm hardly the first to write a Game Boy Advance Emulator nor will I be the last. This project isn't going to compete with the GOATs like [mGBA](https://github.com/mgba-emu) or [NanoBoyAdvance](https://github.com/nba-emu/NanoBoyAdvance). There aren't any interesting ideas either like in [DSHBA](https://github.com/DenSinH/DSHBA).
+
+This is a simple (read: incomplete) for-fun long-term project. I hope to get "mostly there", which to me means that I'm not missing any major hardware features and the set of possible improvements would be in memory timing or in UI/UX. With respect to that goal, here's what's outstanding:
+
+### TODO
+
+- [x] Affine Sprites
+- [ ] Windowing (see [this branch](https://git.musuka.dev/paoda/zba/src/branch/window))
+- [ ] Audio Resampler (Having issues with SDL2's)
+- [ ] Refactoring for easy-ish perf boosts
+
+## Usage
+
+ZBA supports both a CLI and a GUI. If running from the terminal, try using `zba --help` to see what you can do. If you want to use the GUI, feel free to just run `zba` without any arguments.
+
+ZBA does not feature any BIOS HLE, so providing one will be necessary if a ROM makes use of it. Need one? Why not try using the open-source [Cult-Of-GBA BIOS](https://github.com/Cult-of-GBA/BIOS) written by [fleroviux](https://github.com/fleroviux) and [DenSinH](https://github.com/DenSinH)?
+
+Finally it's worth noting that ZBA uses a TOML config file it'll store in your OS's data directory. See `example.toml` to learn about the defaults and what exactly you can mess around with.
 
 ## Compiling
-Most recently built on Zig [0.10.0-dev.3900+ab4b26d8a](https://github.com/ziglang/zig/tree/ab4b26d8a)
+
+Most recently built on Zig [v0.11.0](https://github.com/ziglang/zig/tree/0.11.0)
 
 ### Dependencies
-* [SDL.zig](https://github.com/MasterQ32/SDL.zig)
-    * [SDL2](https://www.libsdl.org/download-2.0.php)
-* [zig-clap](https://github.com/Hejsil/zig-clap)
-* [known-folders](https://github.com/ziglibs/known-folders)
-* [`bitfields.zig`](https://github.com/FlorenceOS/Florence/blob/aaa5a9e568197ad24780ec9adb421217530d4466/lib/util/bitfields.zig)
 
-`bitfields.zig` from [FlorenceOS](https://github.com/FlorenceOS) is included under `lib/util/bitfield.zig`.
+Dependency | Source
+--- | ---
+known-folders | <https://github.com/ziglibs/known-folders>
+nfd-zig | <https://github.com/fabioarnold/nfd-zig>
+SDL.zig | <https://github.com/MasterQ32/SDL.zig>
+tomlz | <https://github.com/mattyhall/tomlz>
+zba-gdbstub | <https://github.com/paoda/zba-gdbstub>
+zba-util | <https://git.musuka.dev/paoda/zba-util>
+zgui | <https://github.com/michal-z/zig-gamedev/tree/main/libs/zgui>
+zig-clap | <https://github.com/Hejsil/zig-clap>
+zig-datetime | <https://github.com/frmdstryr/zig-datetime>
+`bitfield.zig` | [https://github.com/FlorenceOS/Florence](https://github.com/FlorenceOS/Florence/blob/aaa5a9e568/lib/util/bitfields.zig)
+`gl.zig` | <https://github.com/MasterQ32/zig-opengl>
 
-Use `git submodule update --init` from the project root to pull the git submodules `SDL.zig`, `zig-clap`, and `known-folders`
+Use `git submodule update --init` from the project root to pull the git relevant git submodules
 
-Be sure to provide SDL2 using: 
-* Linux: Your distro's package manager
-* MacOS: ¯\\\_(ツ)_/¯
-* Windows: [`vcpkg`](https://github.com/Microsoft/vcpkg) (install `sdl2:x64-windows`)
+Be sure to provide SDL2 using:
 
-`SDL.zig` will provide a helpful compile error if the zig compiler is unable to find SDL2. 
+- Linux: Your distro's package manager
+- macOS: ¯\\\_(ツ)_/¯ (try [this formula](https://formulae.brew.sh/formula/sdl2)?)
+- Windows: [`vcpkg`](https://github.com/Microsoft/vcpkg) (install `sdl2:x64-windows`)
 
-Once you've got all the dependencies, execute `zig build -Drelease-fast`. The executable is located at `zig-out/bin/`. 
+`SDL.zig` will provide a helpful compile error if the zig compiler is unable to find SDL2.
+
+Once you've got all the dependencies, execute `zig build -Doptimize=ReleaseSafe`. The executable will be under `zig-out/bin` and the shared libraries (if enabled) under `zig-out/lib`. If working with shared libraries on windows, be sure to add all artifacts to the same directory. On Unix, you'll want to make use of `LD_PRELOAD`.
 
 ## Controls
-Key | Button
---- | ---
-<kbd>X</kbd> | A
-<kbd>Z</kbd> | B
-<kbd>A</kbd> | L
-<kbd>S</kbd> | R
-<kbd>Return</kbd> | Start
-<kbd>RShift</kbd> | Select
+
+Key | Button | | Key | Button
+--- | --- | --- | --- | ---
+<kbd>A</kbd> | L | | <kbd>S</kbd> | R
+<kbd>X</kbd> | A | | <kbd>Z</kbd> | B
+<kbd>Return</kbd> | Start | | <kbd>RShift</kbd> | Select
 Arrow Keys | D-Pad
+
+## Tests
+
+GBA Tests | [jsmolka](https://github.com/jsmolka/) | gba_tests | [destoer](https://github.com/destoer/)
+--- | --- | --- | ---
+`arm.gba`,  `thumb.gba` | PASS | `cond_invalid.gba` | PASS
+`memory.gba`, `bios.gba` | PASS | `dma_priority.gba` | PASS
+`flash64.gba`, `flash128.gba` | PASS | `hello_world.gba` | PASS
+`sram.gba` | PASS | `if_ack.gba` | PASS
+`none.gba` | PASS | `line_timing.gba` | FAIL
+`hello.gba`, `shades.gba`, `stripes.gba` | PASS | `lyc_midline.gba` | FAIL
+`nes.gba` | PASS | `window_midframe.gba` | FAIL
+
+GBARoms | [DenSinH](https://github.com/DenSinH/) | GBA Test Collection | [ladystarbreeze](https://github.com/ladystarbreeze)
+--- | --- | --- | ---
+`eeprom-test`, `flash-test` | PASS | `retAddr.gba` | PASS
+`midikey2freq` | PASS | `helloWorld.gba` | PASS
+`swi-tests-random` | FAIL | `helloAudio.gba` | PASS
+
+FuzzARM | [DenSinH](https://github.com/DenSinH/) |  arm7wrestler GBA Fixed | [destoer](https://github.com/destoer)
+--- | --- | --- | ---
+`main.gba` | PASS | `armwrestler-gba-fixed.gba` | PASS
+
+## Resources
+
+- [GBATEK](https://problemkaputt.de/gbatek.htm)
+- [TONC](https://coranac.com/tonc/text/toc.htm)
+- [ARM Architecture Reference Manual](https://www.intel.com/content/dam/www/programmable/us/en/pdfs/literature/third-party/ddi0100e_arm_arm.pdf)
+- [ARM7TDMI Data Sheet](https://www.dca.fee.unicamp.br/cursos/EA871/references/ARM/ARM7TDMIDataSheet.pdf)

build.zig

@@ -1,56 +1,65 @@
 const std = @import("std");
-const Sdk = @import("lib/SDL.zig/Sdk.zig");
+const builtin = @import("builtin");
 
-pub fn build(b: *std.build.Builder) void {
-    // Standard target options allows the person running `zig build` to choose
-    // what target to build for. Here we do not override the defaults, which
-    // means any target is allowed, and the default is native. Other options
-    // for restricting supported target set are available.
-    const target = b.standardTargetOptions(.{});
+const sdl = @import("lib/SDL.zig/build.zig");
 
-    // Standard release options allow the person running `zig build` to select
-    // between Debug, ReleaseSafe, ReleaseFast, and ReleaseSmall.
-    const mode = b.standardReleaseOptions();
+const SemVer = std.SemanticVersion;
 
-    const exe = b.addExecutable("zba", "src/main.zig");
-    exe.setTarget(target);
+const target_version = "0.13.0";
 
-    // Known Folders (%APPDATA%, XDG, etc.)
-    exe.addPackagePath("known_folders", "lib/known-folders/known-folders.zig");
-
-    // DateTime Library
-    exe.addPackagePath("datetime", "lib/zig-datetime/src/main.zig");
-
-    // Bitfield type from FlorenceOS: https://github.com/FlorenceOS/
-    // exe.addPackage(.{ .name = "bitfield", .path = .{ .path = "lib/util/bitfield.zig" } });
-    exe.addPackagePath("bitfield", "lib/util/bitfield.zig");
-
-    // Argument Parsing Library
-    exe.addPackagePath("clap", "lib/zig-clap/clap.zig");
-
-    // TOML Library
-    exe.addPackagePath("toml", "lib/zig-toml/src/toml.zig");
-
-    // Zig SDL Bindings: https://github.com/MasterQ32/SDL.zig
-    const sdk = Sdk.init(b);
-    sdk.link(exe, .dynamic);
-    exe.addPackage(sdk.getNativePackage("sdl2"));
-
-    exe.setBuildMode(mode);
-    exe.install();
-
-    const run_cmd = exe.run();
-    run_cmd.step.dependOn(b.getInstallStep());
-    if (b.args) |args| {
-        run_cmd.addArgs(args);
+pub fn build(b: *std.Build) void {
+    const actual_version = builtin.zig_version;
+    if (comptime actual_version.order(SemVer.parse(target_version) catch unreachable) != .eq) {
+        @compileError("ZBA must be built with Zig v" ++ target_version ++ ".");
     }
 
+    const target = b.standardTargetOptions(.{});
+    const optimize = b.standardOptimizeOption(.{});
+
+    const exe = b.addExecutable(.{
+        .name = "zba",
+        .root_source_file = b.path("src/main.zig"),
+        .target = target,
+        .optimize = optimize,
+    });
+
+    const sdk = sdl.init(b, null, null);
+    const zgui = b.dependency("zgui", .{ .shared = false, .with_implot = true, .backend = .sdl2_opengl3 });
+    const imgui = zgui.artifact("imgui");
+
+    exe.root_module.addImport("known_folders", b.dependency("known-folders", .{}).module("known-folders")); // https://github.com/ziglibs/known-folders
+    exe.root_module.addImport("datetime", b.dependency("zig-datetime", .{}).module("zig-datetime")); // https://github.com/frmdstryr/zig-datetime
+    exe.root_module.addImport("clap", b.dependency("zig-clap", .{}).module("clap")); // https://github.com/Hejsil/zig-clap
+    exe.root_module.addImport("zba-util", b.dependency("zba-util", .{}).module("zba-util")); // https://git.musuka.dev/paoda/zba-util
+    exe.root_module.addImport("tomlz", b.dependency("tomlz", .{}).module("tomlz")); // https://github.com/mattyhall/tomlz
+    exe.root_module.addImport("arm32", b.dependency("arm32", .{}).module("arm32")); // https://git.musuka.dev/paoda/arm32
+    exe.root_module.addImport("gdbstub", b.dependency("zba-gdbstub", .{}).module("zba-gdbstub")); // https://git.musuka.dev/paoda/gdbstub
+    exe.root_module.addImport("nfd", b.dependency("nfd", .{}).module("nfd")); // https://github.com/fabioarnold/nfd-zig
+    exe.root_module.addImport("zgui", zgui.module("root")); // https://git.musuka.dev/paoda/zgui
+    exe.root_module.addImport("sdl2", sdk.getNativeModule()); // https://github.com/MasterQ32/SDL.zig
+
+    exe.root_module.addAnonymousImport("bitfield", .{ .root_source_file = b.path("lib/bitfield.zig") }); // https://github.com/FlorenceOS/
+    exe.root_module.addAnonymousImport("gl", .{ .root_source_file = b.path("lib/gl.zig") }); // https://github.com/MasterQ32/zig-opengl
+    exe.root_module.addAnonymousImport("example.toml", .{ .root_source_file = b.path("example.toml") });
+
+    sdk.link(exe, .dynamic, .SDL2);
+    sdk.link(imgui, .dynamic, .SDL2);
+    exe.linkLibrary(imgui);
+
+    b.installArtifact(exe);
+
+    const run_cmd = b.addRunArtifact(exe);
+    run_cmd.step.dependOn(b.getInstallStep());
+    if (b.args) |args| run_cmd.addArgs(args);
+
     const run_step = b.step("run", "Run the app");
     run_step.dependOn(&run_cmd.step);
 
-    const exe_tests = b.addTest("src/main.zig");
-    exe_tests.setTarget(target);
-    exe_tests.setBuildMode(mode);
+    const exe_tests = b.addTest(.{
+        .root_source_file = b.path("src/main.zig"),
+        .target = target,
+        .optimize = optimize,
+    });
 
     const test_step = b.step("test", "Run unit tests");
     test_step.dependOn(&exe_tests.step);

build.zig.zon

@@ -0,0 +1,50 @@
+.{
+    .name = "zba",
+    .version = "0.1.0",
+    .paths = .{
+        "build.zig",
+        "build.zig.zon",
+        "lib/bitfield.zig",
+        "lib/gl.zig",
+        "src",
+    },
+    .minimum_zig_version = "0.13.0",
+    .dependencies = .{
+        .nfd = .{
+            .url = "git+https://github.com/paoda/nfd-zig#ad81729d33da30d5f4fd23718debec48245121ca",
+            .hash = "1220a679380847513262c8c5c474d4a415f9ecc4921c8c6aefbdbdce66cf2aa19ceb",
+        },
+        .@"known-folders" = .{
+            .url = "git+https://github.com/ziglibs/known-folders#1cceeb70e77dec941a4178160ff6c8d05a74de6f",
+            .hash = "12205f5e7505c96573f6fc5144592ec38942fb0a326d692f9cddc0c7dd38f9028f29",
+        },
+        .@"zig-datetime" = .{
+            .url = "git+https://github.com/frmdstryr/zig-datetime#70aebf28fb3e137cd84123a9349d157a74708721",
+            .hash = "122077215ce36e125a490e59ec1748ffd4f6ba00d4d14f7308978e5360711d72d77f",
+        },
+        .@"zig-clap" = .{
+            .url = "git+https://github.com/Hejsil/zig-clap#c0193e9247335a6c1688b946325060289405de2a",
+            .hash = "12207ee987ce045596cb992cfb15b0d6d9456e50d4721c3061c69dabc2962053644d",
+        },
+        .@"zba-util" = .{
+            .url = "git+https://git.musuka.dev/paoda/zba-util#bf0e744047ce1ec90172dbcc0c72bfcc29a063e3",
+            .hash = "1220d044ecfbeacc3b3cebeff131d587e24167d61435a3cb96dffd4d4521bb06aed0",
+        },
+        .@"zba-gdbstub" = .{
+            .url = "git+https://git.musuka.dev/paoda/zba-gdbstub#9a50607d5f48293f950a4e823344f2bc24582a5a",
+            .hash = "1220ac267744ed2a735f03c4620d7c6210fbd36d7bfb2b376ddc3436faebadee0f61",
+        },
+        .tomlz = .{
+            .url = "git+https://github.com/paoda/tomlz#9a16dd53927ef2012478b6494bafb4475e44f4c9",
+            .hash = "12204f922cab84980e36b5c058d354ec0ee169bda401c8e0e80a463580349b476569",
+        },
+        .arm32 = .{
+            .url = "git+https://git.musuka.dev/paoda/arm32#814d081ea0983bc48841a6baad7158c157b17ad6",
+            .hash = "12203c3dacf3a7aa7aee5fc5763dd7b40399bd1c34d1483330b6bd5a76bffef22d82",
+        },
+        .zgui = .{
+            .url = "git+https://git.musuka.dev/paoda/zgui#7f8d05101e96c64314d7926c80ee157dcb89da4e",
+            .hash = "1220bd81a1c7734892b1d4233ed047710487787873c85dd5fc76d1764a331ed2ff43",
+        },
+    },
+}

dl_sdl2.ps1

@@ -0,0 +1,36 @@
+$SDL2Version = "2.30.0"
+$ArchiveFile = ".\SDL2-devel-mingw.zip"
+$Json = @"
+{
+	"x86_64-windows-gnu": {
+		"include": ".build_config\\SDL2\\include",
+		"libs": ".build_config\\SDL2\\lib",
+		"bin": ".build_config\\SDL2\\bin"
+	}
+}	
+"@
+
+New-Item -Force -ItemType Directory -Path .\.build_config
+Set-Location -Path .build_config -PassThru
+
+if (!(Test-Path -PathType Leaf $ArchiveFile)) {
+	Invoke-WebRequest "https://github.com/libsdl-org/SDL/releases/download/release-$SDL2Version/SDL2-devel-$SDL2Version-mingw.zip" -OutFile $ArchiveFile
+}
+
+Expand-Archive $ArchiveFile
+
+if (Test-Path -PathType Container .\SDL2) {
+	Remove-Item -Recurse .\SDL2
+}
+
+New-Item -Force -ItemType Directory -Path .\SDL2 
+Get-ChildItem -Path ".\SDL2-devel-mingw\SDL2-$SDL2Version\x86_64-w64-mingw32" | Move-Item -Destination .\SDL2
+
+# #include <SDL.h>
+Move-Item -Force -Path .\SDL2\include\SDL2\* -Destination .\SDL2\include
+Remove-Item -Force .\SDL2\include\SDL2
+
+New-Item -Force .\sdl.json -Value $Json
+
+Remove-Item -Recurse .\SDL2-devel-mingw
+Set-Location -Path .. -PassThru

example.toml

@@ -0,0 +1,25 @@
+[host]
+# Using nearest-neighbour scaling, how many times the native resolution 
+# of the game bow should the screen be?
+win_scale = 3
+# Enable VSYNC on the UI thread
+vsync = true
+# Mute ZBA
+mute = false
+
+[guest]
+# Sync Emulation to Audio 
+audio_sync = true
+# Sync Emulation to Video
+video_sync = true
+# Force RTC support
+force_rtc = false
+# Skip BIOS
+skip_bios = false
+
+[debug]
+# Enable detailed CPU logs
+cpu_trace = false
+# When false and builtin.mode == .Debug, ZBA will panic
+# on unknown I/O reads
+unhandled_io = true

lib/bitfield.zig

@@ -26,13 +26,13 @@ fn BitType(comptime FieldType: type, comptime ValueType: type, comptime shamt: u
         }
 
         pub fn read(self: anytype) ValueType {
-            return @bitCast(ValueType, @truncate(u1, self.bits.field().* >> shamt));
+            return @bitCast(@as(u1, @truncate(self.bits.field().* >> shamt)));
         }
 
         // Since these are mostly used with MMIO, I want to avoid
         // reading the memory just to write it again, also races
         pub fn write(self: anytype, val: ValueType) void {
-            if (@bitCast(bool, val)) {
+            if (@as(bool, @bitCast(val))) {
                 self.set();
             } else {
                 self.unset();
@@ -67,17 +67,17 @@ pub fn Bitfield(comptime FieldType: type, comptime shamt: usize, comptime num_bi
         dummy: FieldType,
 
         fn field(self: anytype) PtrCastPreserveCV(@This(), @TypeOf(self), FieldType) {
-            return @ptrCast(PtrCastPreserveCV(@This(), @TypeOf(self), FieldType), self);
+            return @ptrCast(self);
         }
 
         pub fn write(self: anytype, val: ValueType) void {
             self.field().* &= ~self_mask;
-            self.field().* |= @intCast(FieldType, val) << shamt;
+            self.field().* |= @as(FieldType, @intCast(val)) << shamt;
         }
 
         pub fn read(self: anytype) ValueType {
             const val: FieldType = self.field().*;
-            return @intCast(ValueType, (val & self_mask) >> shamt);
+            return @intCast((val & self_mask) >> shamt);
         }
     };
 }

src/config.zig

@@ -1,11 +1,13 @@
 const std = @import("std");
-const toml = @import("toml");
+const tomlz = @import("tomlz");
 
 const Allocator = std.mem.Allocator;
 
+const log = std.log.scoped(.Config);
 var state: Config = .{};
 
 const Config = struct {
+    // FIXME: tomlz expects these to be case sensitive
     host: Host = .{},
     guest: Guest = .{},
     debug: Debug = .{},
@@ -18,6 +20,8 @@ const Config = struct {
         ///
         /// Note: This does not affect whether Emulation is synced to 59Hz
         vsync: bool = true,
+        /// Mute ZBA
+        mute: bool = false,
     };
 
     // Settings realted to the emulation itself
@@ -28,13 +32,15 @@ const Config = struct {
         video_sync: bool = true,
         /// Whether RTC I/O should always be enabled
         force_rtc: bool = false,
+        /// Skip BIOS
+        skip_bios: bool = false,
     };
 
     /// Settings related to debugging ZBA
     const Debug = struct {
         /// Enable CPU Trace logs
         cpu_trace: bool = false,
-        /// If false and ZBA is build in debug mode, ZBA will panic on unhandled I/O
+        /// If false and ZBA is built in debug mode, ZBA will panic on unhandled I/O
         unhandled_io: bool = true,
     };
 };
@@ -44,31 +50,14 @@ pub fn config() *const Config {
 }
 
 /// Reads a config file and then loads it into the global state
-pub fn load(allocator: Allocator, config_path: []const u8) !void {
-    var config_file = try std.fs.cwd().openFile(config_path, .{});
+pub fn load(allocator: Allocator, file_path: []const u8) !void {
+    var config_file = try std.fs.cwd().openFile(file_path, .{});
     defer config_file.close();
 
+    log.info("loaded from {s}", .{file_path});
+
     const contents = try config_file.readToEndAlloc(allocator, try config_file.getEndPos());
     defer allocator.free(contents);
 
-    const table = try toml.parseContents(allocator, contents, null);
-    defer table.deinit();
-
-    // TODO: Report unknown config options
-
-    if (table.keys.get("Host")) |host| {
-        if (host.Table.keys.get("win_scale")) |scale| state.host.win_scale = scale.Integer;
-        if (host.Table.keys.get("vsync")) |vsync| state.host.vsync = vsync.Boolean;
-    }
-
-    if (table.keys.get("Guest")) |guest| {
-        if (guest.Table.keys.get("audio_sync")) |sync| state.guest.audio_sync = sync.Boolean;
-        if (guest.Table.keys.get("video_sync")) |sync| state.guest.video_sync = sync.Boolean;
-        if (guest.Table.keys.get("force_rtc")) |forced| state.guest.force_rtc = forced.Boolean;
-    }
-
-    if (table.keys.get("Debug")) |debug| {
-        if (debug.Table.keys.get("cpu_trace")) |trace| state.debug.cpu_trace = trace.Boolean;
-        if (debug.Table.keys.get("unhandled_io")) |unhandled| state.debug.unhandled_io = unhandled.Boolean;
-    }
+    state = try tomlz.parser.decode(Config, allocator, contents);
 }

src/core/Bus.zig

@@ -1,7 +1,6 @@
 const std = @import("std");
 
-const AudioDeviceId = @import("sdl2").SDL_AudioDeviceID;
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
 const Bios = @import("bus/Bios.zig");
 const Ewram = @import("bus/Ewram.zig");
 const GamePak = @import("bus/GamePak.zig");
@@ -20,7 +19,7 @@ const log = std.log.scoped(.Bus);
 
 const createDmaTuple = @import("bus/dma.zig").create;
 const createTimerTuple = @import("bus/timer.zig").create;
-const rotr = @import("../util.zig").rotr;
+const rotr = @import("zba-util").rotr;
 
 const timings: [2][0x10]u8 = [_][0x10]u8{
     // BIOS, Unused, EWRAM, IWRAM, I/0, PALRAM, VRAM, OAM, ROM0, ROM0, ROM1, ROM1, ROM2, ROM2, SRAM, Unused
@@ -34,6 +33,11 @@ pub const fetch_timings: [2][0x10]u8 = [_][0x10]u8{
     [_]u8{ 1, 1, 6, 1, 1, 2, 2, 1, 4, 4, 4, 4, 4, 4, 8, 8 }, // 32-bit
 };
 
+// Fastmem Related
+const page_size = 1 * 0x400; // 1KiB
+const address_space_size = 0x1000_0000;
+const table_len = address_space_size / page_size;
+
 const Self = @This();
 
 pak: GamePak,
@@ -46,10 +50,19 @@ iwram: Iwram,
 ewram: Ewram,
 io: Io,
 
-cpu: ?*Arm7tdmi,
+cpu: *Arm7tdmi,
 sched: *Scheduler,
 
+read_table: *const [table_len]?*const anyopaque,
+write_tables: [2]*const [table_len]?*anyopaque,
+allocator: Allocator,
+
 pub fn init(self: *Self, allocator: Allocator, sched: *Scheduler, cpu: *Arm7tdmi, paths: FilePaths) !void {
+    const tables = try allocator.alloc(?*anyopaque, 3 * table_len); // Allocate all tables
+
+    const read_table = tables[0..table_len];
+    const write_tables = .{ tables[table_len .. 2 * table_len], tables[2 * table_len .. 3 * table_len] };
+
     self.* = .{
         .pak = try GamePak.init(allocator, cpu, paths.rom, paths.save),
         .bios = try Bios.init(allocator, paths.bios),
@@ -62,7 +75,17 @@ pub fn init(self: *Self, allocator: Allocator, sched: *Scheduler, cpu: *Arm7tdmi
         .io = Io.init(),
         .cpu = cpu,
         .sched = sched,
+
+        .read_table = read_table,
+        .write_tables = write_tables,
+        .allocator = allocator,
     };
+
+    self.fillReadTable(read_table);
+
+    // Internal Display Memory behaves differently on 8-bit reads
+    self.fillWriteTable(u32, write_tables[0]);
+    self.fillWriteTable(u8, write_tables[1]);
 }
 
 pub fn deinit(self: *Self) void {
@@ -71,171 +94,443 @@ pub fn deinit(self: *Self) void {
     self.pak.deinit();
     self.bios.deinit();
     self.ppu.deinit();
+
+    // This is so I can deallocate the original `allocator.alloc`. I have to re-make the type
+    // since I'm not keeping it around, This is very jank and bad though
+    // FIXME: please figure out another way
+    self.allocator.free(@as([*]const ?*anyopaque, @ptrCast(self.read_table[0..]))[0 .. 3 * table_len]);
     self.* = undefined;
 }
 
-pub fn dbgRead(self: *const Self, comptime T: type, address: u32) T {
-    const page = @truncate(u8, address >> 24);
-    const aligned_addr = forceAlign(T, address);
+pub fn reset(self: *Self) void {
+    self.bios.reset();
+    self.ppu.reset();
+    self.apu.reset();
+    self.iwram.reset();
+    self.ewram.reset();
 
-    return switch (page) {
-        // General Internal Memory
-        0x00 => blk: {
-            if (address < Bios.size)
-                break :blk self.bios.dbgRead(T, self.cpu.?.r[15], aligned_addr);
+    // https://github.com/ziglang/zig/issues/14705
+    {
+        comptime var i: usize = 0;
+        inline while (i < self.dma.len) : (i += 1) {
+            self.dma[0].reset();
+        }
+    }
 
-            break :blk self.readOpenBus(T, address);
-        },
-        0x02 => self.ewram.read(T, aligned_addr),
-        0x03 => self.iwram.read(T, aligned_addr),
-        0x04 => self.readIo(T, address),
+    // https://github.com/ziglang/zig/issues/14705
+    {
+        comptime var i: usize = 0;
+        inline while (i < self.tim.len) : (i += 1) {
+            self.tim[0].reset();
+        }
+    }
 
-        // Internal Display Memory
-        0x05 => self.ppu.palette.read(T, aligned_addr),
-        0x06 => self.ppu.vram.read(T, aligned_addr),
-        0x07 => self.ppu.oam.read(T, aligned_addr),
-
-        // External Memory (Game Pak)
-        0x08...0x0D => self.pak.dbgRead(T, aligned_addr),
-        0x0E...0x0F => blk: {
-            const value = self.pak.backup.read(address);
-
-            const multiplier = switch (T) {
-                u32 => 0x01010101,
-                u16 => 0x0101,
-                u8 => 1,
-                else => @compileError("Backup: Unsupported read width"),
-            };
-
-            break :blk @as(T, value) * multiplier;
-        },
-        else => self.readOpenBus(T, address),
-    };
+    self.io.reset();
 }
 
-fn readIo(self: *const Self, comptime T: type, unaligned_address: u32) T {
-    const maybe_value = io.read(self, T, forceAlign(T, unaligned_address));
-    return if (maybe_value) |value| value else self.readOpenBus(T, unaligned_address);
+pub fn replaceGamepak(self: *Self, file_path: []const u8) !void {
+    // Note: `save_path` isn't owned by `Backup`
+    const save_path = self.pak.backup.save_path;
+    self.pak.deinit();
+
+    self.pak = try GamePak.init(self.allocator, self.cpu, file_path, save_path);
+
+    const read_ptr: *[table_len]?*const anyopaque = @constCast(self.read_table);
+    const write_ptrs: [2]*[table_len]?*anyopaque = .{ @constCast(self.write_tables[0]), @constCast(self.write_tables[1]) };
+
+    self.fillReadTable(read_ptr);
+    self.fillWriteTable(u32, write_ptrs[0]);
+    self.fillWriteTable(u8, write_ptrs[1]);
 }
 
-fn readOpenBus(self: *const Self, comptime T: type, address: u32) T {
-    const r15 = self.cpu.?.r[15];
+fn fillReadTable(self: *Self, table: *[table_len]?*const anyopaque) void {
+    const vramMirror = @import("ppu/Vram.zig").mirror;
+
+    for (table, 0..) |*ptr, i| {
+        const addr: u32 = @intCast(page_size * i);
+
+        ptr.* = switch (addr) {
+            // General Internal Memory
+            0x0000_0000...0x0000_3FFF => null, // BIOS has it's own checks
+            0x0200_0000...0x02FF_FFFF => &self.ewram.buf[addr & 0x3FFFF],
+            0x0300_0000...0x03FF_FFFF => &self.iwram.buf[addr & 0x7FFF],
+            0x0400_0000...0x0400_03FF => null, // I/O
+
+            // Internal Display Memory
+            0x0500_0000...0x05FF_FFFF => &self.ppu.palette.buf[addr & 0x3FF],
+            0x0600_0000...0x06FF_FFFF => &self.ppu.vram.buf[vramMirror(addr)],
+            0x0700_0000...0x07FF_FFFF => &self.ppu.oam.buf[addr & 0x3FF],
+
+            // External Memory (Game Pak)
+            0x0800_0000...0x0DFF_FFFF => self.fillReadTableExternal(addr),
+            0x0E00_0000...0x0FFF_FFFF => null, // SRAM
+            else => null,
+        };
+    }
+}
+
+fn fillWriteTable(self: *Self, comptime T: type, table: *[table_len]?*const anyopaque) void {
+    comptime std.debug.assert(T == u32 or T == u16 or T == u8);
+    const vramMirror = @import("ppu/Vram.zig").mirror;
+
+    for (table, 0..) |*ptr, i| {
+        const addr: u32 = @intCast(page_size * i);
+
+        ptr.* = switch (addr) {
+            // General Internal Memory
+            0x0000_0000...0x0000_3FFF => null, // BIOS has it's own checks
+            0x0200_0000...0x02FF_FFFF => &self.ewram.buf[addr & 0x3FFFF],
+            0x0300_0000...0x03FF_FFFF => &self.iwram.buf[addr & 0x7FFF],
+            0x0400_0000...0x0400_03FF => null, // I/O
+
+            // Internal Display Memory
+            0x0500_0000...0x05FF_FFFF => if (T != u8) &self.ppu.palette.buf[addr & 0x3FF] else null,
+            0x0600_0000...0x06FF_FFFF => if (T != u8) &self.ppu.vram.buf[vramMirror(addr)] else null,
+            0x0700_0000...0x07FF_FFFF => if (T != u8) &self.ppu.oam.buf[addr & 0x3FF] else null,
+
+            // External Memory (Game Pak)
+            0x0800_0000...0x0DFF_FFFF => null, // ROM
+            0x0E00_0000...0x0FFF_FFFF => null, // SRAM
+            else => null,
+        };
+    }
+}
+
+fn fillReadTableExternal(self: *Self, addr: u32) ?*anyopaque {
+    // see `GamePak.zig` for more information about what conditions need to be true
+    // so that a simple pointer dereference isn't possible
+
+    std.debug.assert(addr & @as(u32, page_size - 1) == 0); // addr is guaranteed to be page-aligned
+
+    const start_addr = addr;
+    const end_addr = start_addr + page_size;
+
+    {
+        const data = start_addr <= 0x0800_00C4 and 0x0800_00C4 < end_addr; // GPIO Data
+        const direction = start_addr <= 0x0800_00C6 and 0x0800_00C6 < end_addr; // GPIO Direction
+        const control = start_addr <= 0x0800_00C8 and 0x0800_00C8 < end_addr; // GPIO Control
+
+        const has_gpio = data or direction or control;
+        const gpio_kind = self.pak.gpio.device.kind;
+
+        // There is a GPIO Device, and the current page contains at least one memory-mapped GPIO register
+        if (gpio_kind != .None and has_gpio) return null;
+    }
+
+    if (self.pak.backup.kind == .Eeprom) {
+        if (self.pak.buf.len > 0x100_000) {
+            // We are using a "large" EEPROM which means that if the below check is true
+            // this page has an address that's reserved for the EEPROM and therefore must
+            // be handled in slowmem
+            if (addr & 0x1FF_FFFF > 0x1FF_FEFF) return null;
+        } else {
+            // We are using a "small" EEPROM which means that if the below check is true
+            // (that is, we're in the 0xD address page) then we must handle at least one
+            // address in this page in slowmem
+            if (@as(u4, @truncate(addr >> 24)) == 0xD) return null;
+        }
+    }
+
+    // Finally, the GamePak has some unique behaviour for reads past the end of the ROM,
+    // so those will be handled by slowmem as well
+    const masked_addr = addr & 0x1FF_FFFF;
+    if (masked_addr >= self.pak.buf.len) return null;
+
+    return &self.pak.buf[masked_addr];
+}
+
+fn readIo(self: *const Self, comptime T: type, address: u32) T {
+    return io.read(self, T, address) orelse self.openBus(T, address);
+}
+
+fn openBus(self: *const Self, comptime T: type, address: u32) T {
+    @setCold(true);
+    const r15 = self.cpu.r[15];
 
     const word = blk: {
-        // If u32 Open Bus, read recently fetched opcode (PC + 8)
-        if (!self.cpu.?.cpsr.t.read()) break :blk self.dbgRead(u32, r15 + 4);
-        const page = @truncate(u8, r15 >> 24);
+        // If Arm, get the most recently fetched instruction (PC + 8)
+        //
+        // FIXME: This is most likely a faulty assumption.
+        // I think what *actually* happens is that the Bus has a latch for the most
+        // recently fetched piece of data, which is then returned during Open Bus (also DMA open bus?)
+        // I can "get away" with this because it's very statistically likely that the most recently latched value is
+        // the most recently fetched instruction by the pipeline
+        if (!self.cpu.cpsr.t.read()) break :blk self.cpu.pipe.stage[1].?;
+
+        const page: u8 = @truncate(r15 >> 24);
+
+        // PC + 2 = stage[0]
+        // PC + 4 = stage[1]
+        // PC + 6 = Need a Debug Read for this?
 
         switch (page) {
             // EWRAM, PALRAM, VRAM, and Game ROM (16-bit)
             0x02, 0x05, 0x06, 0x08...0x0D => {
-                // (PC + 4)
-                const halfword = self.dbgRead(u16, r15 + 2);
-
-                break :blk @as(u32, halfword) << 16 | halfword;
+                const halfword: u32 = @as(u16, @truncate(self.cpu.pipe.stage[1].?));
+                break :blk halfword << 16 | halfword;
             },
+
             // BIOS or OAM (32-bit)
             0x00, 0x07 => {
                 // Aligned: (PC + 6) | (PC + 4)
                 // Unaligned: (PC + 4) | (PC + 2)
-                const offset: u32 = if (address & 3 == 0b00) 2 else 0;
+                const aligned = address & 3 == 0b00;
 
-                break :blk @as(u32, self.dbgRead(u16, r15 + 2 + offset)) << 16 | self.dbgRead(u16, r15 + offset);
+                // TODO: What to do on PC + 6?
+                const high: u32 = if (aligned) self.dbgRead(u16, r15 + 4) else @as(u16, @truncate(self.cpu.pipe.stage[1].?));
+                const low: u32 = @as(u16, @truncate(self.cpu.pipe.stage[@intFromBool(aligned)].?));
+
+                break :blk high << 16 | low;
             },
+
             // IWRAM (16-bit but special)
             0x03 => {
                 // Aligned: (PC + 2) | (PC + 4)
                 // Unaligned: (PC + 4) | (PC + 2)
-                const offset: u32 = if (address & 3 == 0b00) 2 else 0;
+                const aligned = address & 3 == 0b00;
 
-                break :blk @as(u32, self.dbgRead(u16, r15 + 2 - offset)) << 16 | self.dbgRead(u16, r15 + offset);
+                const high: u32 = @as(u16, @truncate(self.cpu.pipe.stage[1 - @intFromBool(aligned)].?));
+                const low: u32 = @as(u16, @truncate(self.cpu.pipe.stage[@intFromBool(aligned)].?));
+
+                break :blk high << 16 | low;
+            },
+            else => {
+                log.err("THUMB open bus read from 0x{X:0>2} page @0x{X:0>8}", .{ page, address });
+                @panic("invariant most-likely broken");
             },
-            else => unreachable,
         }
     };
 
-    return @truncate(T, rotr(u32, word, 8 * (address & 3)));
+    return @truncate(word);
 }
 
-pub fn read(self: *Self, comptime T: type, address: u32) T {
-    const page = @truncate(u8, address >> 24);
-    const aligned_addr = forceAlign(T, address);
+pub fn read(self: *Self, comptime T: type, unaligned_address: u32) T {
+    const bits = @typeInfo(std.math.IntFittingRange(0, page_size - 1)).Int.bits;
+    const page = unaligned_address >> bits;
+    const offset = unaligned_address & (page_size - 1);
 
-    self.sched.tick += timings[@boolToInt(T == u32)][@truncate(u4, page)];
+    // whether or not we do this in slowmem or fastmem, we should advance the scheduler
+    self.sched.tick += timings[@intFromBool(T == u32)][@as(u4, @truncate(unaligned_address >> 24))];
+
+    // We're doing some serious out-of-bounds open-bus reads
+    if (page >= table_len) return self.openBus(T, unaligned_address);
+
+    if (self.read_table[page]) |some_ptr| {
+        // We have a pointer to a page, cast the pointer to it's underlying type
+        const ptr: [*]const T = @ptrCast(@alignCast(some_ptr));
+
+        // Note: We don't check array length, since we force align the
+        // lower bits of the address as the GBA would
+        return ptr[forceAlign(T, offset) / @sizeOf(T)];
+    }
+
+    return self.slowRead(T, unaligned_address);
+}
+
+pub fn dbgRead(self: *const Self, comptime T: type, unaligned_address: u32) T {
+    const bits = @typeInfo(std.math.IntFittingRange(0, page_size - 1)).Int.bits;
+    const page = unaligned_address >> bits;
+    const offset = unaligned_address & (page_size - 1);
+
+    // We're doing some serious out-of-bounds open-bus reads
+    if (page >= table_len) return self.openBus(T, unaligned_address);
+
+    if (self.read_table[page]) |some_ptr| {
+        // We have a pointer to a page, cast the pointer to it's underlying type
+        const ptr: [*]const T = @ptrCast(@alignCast(some_ptr));
+
+        // Note: We don't check array length, since we force align the
+        // lower bits of the address as the GBA would
+        return ptr[forceAlign(T, offset) / @sizeOf(T)];
+    }
+
+    return self.dbgSlowRead(T, unaligned_address);
+}
+
+fn slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
+    @setCold(true);
+
+    const page: u8 = @truncate(unaligned_address >> 24);
+    const address = forceAlign(T, unaligned_address);
 
     return switch (page) {
         // General Internal Memory
         0x00 => blk: {
             if (address < Bios.size)
-                break :blk self.bios.read(T, self.cpu.?.r[15], aligned_addr);
+                break :blk self.bios.read(T, self.cpu.r[15], unaligned_address);
 
-            break :blk self.readOpenBus(T, address);
+            break :blk self.openBus(T, address);
         },
-        0x02 => self.ewram.read(T, aligned_addr),
-        0x03 => self.iwram.read(T, aligned_addr),
+        0x02 => unreachable, // completely handled by fastmeme
+        0x03 => unreachable, // completely handled by fastmeme
         0x04 => self.readIo(T, address),
 
         // Internal Display Memory
-        0x05 => self.ppu.palette.read(T, aligned_addr),
-        0x06 => self.ppu.vram.read(T, aligned_addr),
-        0x07 => self.ppu.oam.read(T, aligned_addr),
+        0x05 => unreachable, // completely handled by fastmeme
+        0x06 => unreachable, // completely handled by fastmeme
+        0x07 => unreachable, // completely handled by fastmeme
 
         // External Memory (Game Pak)
-        0x08...0x0D => self.pak.read(T, aligned_addr),
-        0x0E...0x0F => blk: {
-            const value = self.pak.backup.read(address);
-
-            const multiplier = switch (T) {
-                u32 => 0x01010101,
-                u16 => 0x0101,
-                u8 => 1,
-                else => @compileError("Backup: Unsupported read width"),
-            };
-
-            break :blk @as(T, value) * multiplier;
-        },
-        else => self.readOpenBus(T, address),
+        0x08...0x0D => self.pak.read(T, address),
+        0x0E...0x0F => self.readBackup(T, unaligned_address),
+        else => self.openBus(T, address),
     };
 }
 
-pub fn write(self: *Self, comptime T: type, address: u32, value: T) void {
-    const page = @truncate(u8, address >> 24);
-    const aligned_addr = forceAlign(T, address);
+fn dbgSlowRead(self: *const Self, comptime T: type, unaligned_address: u32) T {
+    const page: u8 = @truncate(unaligned_address >> 24);
+    const address = forceAlign(T, unaligned_address);
 
-    self.sched.tick += timings[@boolToInt(T == u32)][@truncate(u4, page)];
+    return switch (page) {
+        // General Internal Memory
+        0x00 => blk: {
+            if (address < Bios.size)
+                break :blk self.bios.dbgRead(T, self.cpu.r[15], unaligned_address);
+
+            break :blk self.openBus(T, address);
+        },
+        0x02 => unreachable, // handled by fastmem
+        0x03 => unreachable, // handled by fastmem
+        0x04 => self.readIo(T, address),
+
+        // Internal Display Memory
+        0x05 => unreachable, // handled by fastmem
+        0x06 => unreachable, // handled by fastmem
+        0x07 => unreachable, // handled by fastmem
+
+        // External Memory (Game Pak)
+        0x08...0x0D => self.pak.dbgRead(T, address),
+        0x0E...0x0F => self.readBackup(T, unaligned_address),
+        else => self.openBus(T, address),
+    };
+}
+
+fn readBackup(self: *const Self, comptime T: type, unaligned_address: u32) T {
+    const value = self.pak.backup.read(unaligned_address);
+
+    const multiplier = switch (T) {
+        u32 => 0x01010101,
+        u16 => 0x0101,
+        u8 => 1,
+        else => @compileError("Backup: Unsupported read width"),
+    };
+
+    return @as(T, value) * multiplier;
+}
+
+pub fn write(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    const bits = @typeInfo(std.math.IntFittingRange(0, page_size - 1)).Int.bits;
+    const page = unaligned_address >> bits;
+    const offset = unaligned_address & (page_size - 1);
+
+    // whether or not we do this in slowmem or fastmem, we should advance the scheduler
+    self.sched.tick += timings[@intFromBool(T == u32)][@as(u4, @truncate(unaligned_address >> 24))];
+
+    // We're doing some serious out-of-bounds open-bus writes, they do nothing though
+    if (page >= table_len) return;
+
+    if (self.write_tables[@intFromBool(T == u8)][page]) |some_ptr| {
+        // We have a pointer to a page, cast the pointer to it's underlying type
+        const ptr: [*]T = @ptrCast(@alignCast(some_ptr));
+
+        // Note: We don't check array length, since we force align the
+        // lower bits of the address as the GBA would
+        ptr[forceAlign(T, offset) / @sizeOf(T)] = value;
+    } else {
+        // we can return early if this is an 8-bit OAM write
+        if (T == u8 and @as(u8, @truncate(unaligned_address >> 24)) == 0x07) return;
+
+        self.slowWrite(T, unaligned_address, value);
+    }
+}
+
+/// Mostly Identical to `Bus.write`, slowmeme is handled by `Bus.dbgSlowWrite`
+pub fn dbgWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    const bits = @typeInfo(std.math.IntFittingRange(0, page_size - 1)).Int.bits;
+    const page = unaligned_address >> bits;
+    const offset = unaligned_address & (page_size - 1);
+
+    // We're doing some serious out-of-bounds open-bus writes, they do nothing though
+    if (page >= table_len) return;
+
+    if (self.write_tables[@intFromBool(T == u8)][page]) |some_ptr| {
+        // We have a pointer to a page, cast the pointer to it's underlying type
+        const ptr: [*]T = @ptrCast(@alignCast(some_ptr));
+
+        // Note: We don't check array length, since we force align the
+        // lower bits of the address as the GBA would
+        ptr[forceAlign(T, offset) / @sizeOf(T)] = value;
+    } else {
+        // we can return early if this is an 8-bit OAM write
+        if (T == u8 and @as(u8, @truncate(unaligned_address >> 24)) == 0x07) return;
+
+        self.dbgSlowWrite(T, unaligned_address, value);
+    }
+}
+
+fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    @setCold(true);
+
+    const page: u8 = @truncate(unaligned_address >> 24);
+    const address = forceAlign(T, unaligned_address);
 
     switch (page) {
         // General Internal Memory
-        0x00 => self.bios.write(T, aligned_addr, value),
-        0x02 => self.ewram.write(T, aligned_addr, value),
-        0x03 => self.iwram.write(T, aligned_addr, value),
-        0x04 => io.write(self, T, aligned_addr, value),
+        0x00 => self.bios.write(T, address, value),
+        0x02 => unreachable, // completely handled by fastmem
+        0x03 => unreachable, // completely handled by fastmem
+        0x04 => io.write(self, T, address, value),
 
         // Internal Display Memory
-        0x05 => self.ppu.palette.write(T, aligned_addr, value),
-        0x06 => self.ppu.vram.write(T, self.ppu.dispcnt, aligned_addr, value),
-        0x07 => self.ppu.oam.write(T, aligned_addr, value),
+        0x05 => self.ppu.palette.write(T, address, value),
+        0x06 => self.ppu.vram.write(T, self.ppu.dispcnt, address, value),
+        0x07 => unreachable, // completely handled by fastmem
 
         // External Memory (Game Pak)
-        0x08...0x0D => self.pak.write(T, self.dma[3].word_count, aligned_addr, value),
-        0x0E...0x0F => {
-            const rotate_by = switch (T) {
-                u32 => address & 3,
-                u16 => address & 1,
-                u8 => 0,
-                else => @compileError("Backup: Unsupported write width"),
-            };
-
-            self.pak.backup.write(address, @truncate(u8, rotr(T, value, 8 * rotate_by)));
-        },
+        0x08...0x0D => self.pak.write(T, self.dma[3].word_count, address, value),
+        0x0E...0x0F => self.pak.backup.write(unaligned_address, @truncate(rotr(T, value, 8 * rotateBy(T, unaligned_address)))),
         else => {},
     }
 }
 
-fn forceAlign(comptime T: type, address: u32) u32 {
+fn dbgSlowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    @setCold(true);
+
+    const page: u8 = @truncate(unaligned_address >> 24);
+    const address = forceAlign(T, unaligned_address);
+
+    switch (page) {
+        // General Internal Memory
+        0x00 => self.bios.write(T, address, value),
+        0x02 => unreachable, // completely handled by fastmem
+        0x03 => unreachable, // completely handled by fastmem
+        0x04 => return, // FIXME: Let debug writes mess with I/O
+
+        // Internal Display Memory
+        0x05 => self.ppu.palette.write(T, address, value),
+        0x06 => self.ppu.vram.write(T, self.ppu.dispcnt, address, value),
+        0x07 => unreachable, // completely handled by fastmem
+
+        // External Memory (Game Pak)
+        0x08...0x0D => return, // FIXME: Debug Write to Backup/GPIO w/out messing with state
+        0x0E...0x0F => return, // FIXME: Debug Write to Backup w/out messing with state
+        else => {},
+    }
+}
+
+inline fn rotateBy(comptime T: type, address: u32) u32 {
     return switch (T) {
-        u32 => address & 0xFFFF_FFFC,
-        u16 => address & 0xFFFF_FFFE,
+        u32 => address & 3,
+        u16 => address & 1,
+        u8 => 0,
+        else => @compileError("Unsupported write width"),
+    };
+}
+
+pub inline fn forceAlign(comptime T: type, address: u32) u32 {
+    return switch (T) {
+        u32 => address & ~@as(u32, 3),
+        u16 => address & ~@as(u32, 1),
         u8 => address,
         else => @compileError("Bus: Invalid read/write type"),
     };

src/core/apu.zig

@@ -3,9 +3,8 @@ const SDL = @import("sdl2");
 const io = @import("bus/io.zig");
 const util = @import("../util.zig");
 
-const AudioDeviceId = SDL.SDL_AudioDeviceID;
-
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("Bus.zig");
 const Scheduler = @import("scheduler.zig").Scheduler;
 const ToneSweep = @import("apu/ToneSweep.zig");
 const Tone = @import("apu/Tone.zig");
@@ -14,133 +13,216 @@ const Noise = @import("apu/Noise.zig");
 
 const SoundFifo = std.fifo.LinearFifo(u8, .{ .Static = 0x20 });
 
-const intToBytes = @import("../util.zig").intToBytes;
-const setHi = @import("../util.zig").setHi;
-const setLo = @import("../util.zig").setLo;
+const getHalf = util.getHalf;
+const setHalf = util.setHalf;
 
 const log = std.log.scoped(.APU);
 
-pub const host_sample_rate = 1 << 15;
+pub const host_rate = @import("../platform.zig").sample_rate;
+pub const host_format = @import("../platform.zig").sample_format;
 
 pub fn read(comptime T: type, apu: *const Apu, addr: u32) ?T {
-    const byte = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     return switch (T) {
-        u16 => switch (byte) {
-            0x60 => apu.ch1.sound1CntL(),
-            0x62 => apu.ch1.sound1CntH(),
+        u32 => switch (byte_addr) {
+            0x60 => @as(T, apu.ch1.sound1CntH()) << 16 | apu.ch1.sound1CntL(),
             0x64 => apu.ch1.sound1CntX(),
             0x68 => apu.ch2.sound2CntL(),
             0x6C => apu.ch2.sound2CntH(),
-
-            0x70 => apu.ch3.select.raw & 0xE0, // SOUND3CNT_L
-            0x72 => apu.ch3.sound3CntH(),
-            0x74 => apu.ch3.freq.raw & 0x4000, // SOUND3CNT_X
-
+            0x70 => @as(T, apu.ch3.sound3CntH()) << 16 | apu.ch3.sound3CntL(),
+            0x74 => apu.ch3.sound3CntX(),
             0x78 => apu.ch4.sound4CntL(),
             0x7C => apu.ch4.sound4CntH(),
-
-            0x80 => apu.psg_cnt.raw & 0xFF77, // SOUNDCNT_L
-            0x82 => apu.dma_cnt.raw & 0x770F, // SOUNDCNT_H
+            0x80 => @as(T, apu.dma_cnt.raw) << 16 | apu.psg_cnt.raw, // SOUNDCNT_H, SOUNDCNT_L
             0x84 => apu.soundCntX(),
+            0x88 => apu.bias.raw, // SOUNDBIAS, high is unused
+            0x8C => null,
+            0x90, 0x94, 0x98, 0x9C => apu.ch3.wave_dev.read(T, apu.ch3.select, addr),
+            0xA0 => null, // FIFO_A
+            0xA4 => null, // FIFO_B
+            else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
+        },
+        u16 => switch (byte_addr) {
+            0x60 => apu.ch1.sound1CntL(),
+            0x62 => apu.ch1.sound1CntH(),
+            0x64 => apu.ch1.sound1CntX(),
+            0x66 => 0x0000, // suite.gba expects 0x0000, not 0xDEAD
+            0x68 => apu.ch2.sound2CntL(),
+            0x6A => 0x0000,
+            0x6C => apu.ch2.sound2CntH(),
+            0x6E => 0x0000,
+            0x70 => apu.ch3.sound3CntL(),
+            0x72 => apu.ch3.sound3CntH(),
+            0x74 => apu.ch3.sound3CntX(),
+            0x76 => 0x0000,
+            0x78 => apu.ch4.sound4CntL(),
+            0x7A => 0x0000,
+            0x7C => apu.ch4.sound4CntH(),
+            0x7E => 0x0000,
+            0x80 => apu.soundCntL(),
+            0x82 => apu.soundCntH(),
+            0x84 => apu.soundCntX(),
+            0x86 => 0x0000,
             0x88 => apu.bias.raw, // SOUNDBIAS
+            0x8A => 0x0000,
+            0x8C, 0x8E => null,
+            0x90, 0x92, 0x94, 0x96, 0x98, 0x9A, 0x9C, 0x9E => apu.ch3.wave_dev.read(T, apu.ch3.select, addr),
+            0xA0, 0xA2 => null, // FIFO_A
+            0xA4, 0xA6 => null, // FIFO_B
+            else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
+        },
+        u8 => switch (byte_addr) {
+            0x60, 0x61 => @truncate(@as(u16, apu.ch1.sound1CntL()) >> getHalf(byte_addr)),
+            0x62, 0x63 => @truncate(apu.ch1.sound1CntH() >> getHalf(byte_addr)),
+            0x64, 0x65 => @truncate(apu.ch1.sound1CntX() >> getHalf(byte_addr)),
+            0x66, 0x67 => 0x00, // assuming behaviour is identical to that of 16-bit reads
+            0x68, 0x69 => @truncate(apu.ch2.sound2CntL() >> getHalf(byte_addr)),
+            0x6A, 0x6B => 0x00,
+            0x6C, 0x6D => @truncate(apu.ch2.sound2CntH() >> getHalf(byte_addr)),
+            0x6E, 0x6F => 0x00,
+            0x70, 0x71 => @truncate(@as(u16, apu.ch3.sound3CntL()) >> getHalf(byte_addr)), // SOUND3CNT_L
+            0x72, 0x73 => @truncate(apu.ch3.sound3CntH() >> getHalf(byte_addr)),
+            0x74, 0x75 => @truncate(apu.ch3.sound3CntX() >> getHalf(byte_addr)), // SOUND3CNT_L
+            0x76, 0x77 => 0x00,
+            0x78, 0x79 => @truncate(apu.ch4.sound4CntL() >> getHalf(byte_addr)),
+            0x7A, 0x7B => 0x00,
+            0x7C, 0x7D => @truncate(apu.ch4.sound4CntH() >> getHalf(byte_addr)),
+            0x7E, 0x7F => 0x00,
+            0x80, 0x81 => @truncate(apu.soundCntL() >> getHalf(byte_addr)), // SOUNDCNT_L
+            0x82, 0x83 => @truncate(apu.soundCntH() >> getHalf(byte_addr)), // SOUNDCNT_H
+            0x84, 0x85 => @truncate(@as(u16, apu.soundCntX()) >> getHalf(byte_addr)),
+            0x86, 0x87 => 0x00,
+            0x88, 0x89 => @truncate(apu.bias.raw >> getHalf(byte_addr)), // SOUNDBIAS
+            0x8A, 0x8B => 0x00,
+            0x8C...0x8F => null,
             0x90...0x9F => apu.ch3.wave_dev.read(T, apu.ch3.select, addr),
-            else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
+            0xA0, 0xA1, 0xA2, 0xA3 => null, // FIFO_A
+            0xA4, 0xA5, 0xA6, 0xA7 => null, // FIFO_B
+            else => util.io.read.err(T, log, "unexpected {} read from 0x{X:0>8}", .{ T, addr }),
         },
-        u8 => switch (byte) {
-            0x60 => apu.ch1.sound1CntL(), // NR10
-            0x62 => apu.ch1.duty.raw, // NR11
-            0x63 => apu.ch1.envelope.raw, // NR12
-            0x68 => apu.ch2.duty.raw, // NR21
-            0x69 => apu.ch2.envelope.raw, // NR22
-            0x73 => apu.ch3.vol.raw, // NR32
-            0x79 => apu.ch4.envelope.raw, // NR42
-            0x7C => apu.ch4.poly.raw, // NR43
-            0x81 => @truncate(u8, apu.psg_cnt.raw >> 8), // NR51
-            0x84 => apu.soundCntX(),
-            0x89 => @truncate(u8, apu.bias.raw >> 8), // SOUNDBIAS_H
-            else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
-        },
-        u32 => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
         else => @compileError("APU: Unsupported read width"),
     };
 }
 
 pub fn write(comptime T: type, apu: *Apu, addr: u32, value: T) void {
-    const byte = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
+
+    if (byte_addr <= 0x81 and !apu.cnt.apu_enable.read()) return;
 
     switch (T) {
-        u32 => switch (byte) {
-            0x60 => apu.ch1.setSound1Cnt(value),
-            0x64 => apu.ch1.setSound1CntX(&apu.fs, @truncate(u16, value)),
-            0x68 => apu.ch2.setSound2CntL(@truncate(u16, value)),
-            0x6C => apu.ch2.setSound2CntH(&apu.fs, @truncate(u16, value)),
-            0x70 => apu.ch3.setSound3Cnt(value),
-            0x74 => apu.ch3.setSound3CntX(&apu.fs, @truncate(u16, value)),
-            0x78 => apu.ch4.setSound4CntL(@truncate(u16, value)),
-            0x7C => apu.ch4.setSound4CntH(&apu.fs, @truncate(u16, value)),
+        u32 => {
+            // 0x80 and 0x81 handled in setSoundCnt
+            if (byte_addr < 0x80 and !apu.cnt.apu_enable.read()) return;
 
-            0x80 => apu.setSoundCnt(value),
-            // WAVE_RAM
-            0x90...0x9F => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
-            0xA0 => apu.chA.push(value), // FIFO_A
-            0xA4 => apu.chB.push(value), // FIFO_B
-            else => util.io.write.undef(log, "Tried to write 0x{X:0>8}{} to 0x{X:0>8}", .{ value, T, addr }),
+            switch (byte_addr) {
+                0x60 => apu.ch1.setSound1Cnt(value),
+                0x64 => apu.ch1.setSound1CntX(&apu.fs, @truncate(value)),
+
+                0x68 => apu.ch2.setSound2CntL(@truncate(value)),
+                0x6C => apu.ch2.setSound2CntH(&apu.fs, @truncate(value)),
+
+                0x70 => apu.ch3.setSound3Cnt(value),
+                0x74 => apu.ch3.setSound3CntX(&apu.fs, @truncate(value)),
+
+                0x78 => apu.ch4.setSound4CntL(@truncate(value)),
+                0x7C => apu.ch4.setSound4CntH(&apu.fs, @truncate(value)),
+
+                0x80 => apu.setSoundCnt(value),
+                0x84 => apu.setSoundCntX(value >> 7 & 1 == 1),
+                0x88 => apu.bias.raw = @truncate(value),
+                0x8C => {},
+
+                0x90, 0x94, 0x98, 0x9C => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
+                0xA0 => apu.chA.push(value), // FIFO_A
+                0xA4 => apu.chB.push(value), // FIFO_B
+                else => util.io.write.undef(log, "Tried to write 0x{X:0>8}{} to 0x{X:0>8}", .{ value, T, addr }),
+            }
         },
-        u16 => switch (byte) {
-            0x60 => apu.ch1.setSound1CntL(@truncate(u8, value)), // SOUND1CNT_L
-            0x62 => apu.ch1.setSound1CntH(value),
-            0x64 => apu.ch1.setSound1CntX(&apu.fs, value),
+        u16 => {
+            if (byte_addr <= 0x81 and !apu.cnt.apu_enable.read()) return;
 
-            0x68 => apu.ch2.setSound2CntL(value),
-            0x6C => apu.ch2.setSound2CntH(&apu.fs, value),
+            switch (byte_addr) {
+                0x60 => apu.ch1.setSound1CntL(@truncate(value)), // SOUND1CNT_L
+                0x62 => apu.ch1.setSound1CntH(value),
+                0x64 => apu.ch1.setSound1CntX(&apu.fs, value),
+                0x66 => {},
 
-            0x70 => apu.ch3.setSound3CntL(@truncate(u8, value)),
-            0x72 => apu.ch3.setSound3CntH(value),
-            0x74 => apu.ch3.setSound3CntX(&apu.fs, value),
+                0x68 => apu.ch2.setSound2CntL(value),
+                0x6A => {},
+                0x6C => apu.ch2.setSound2CntH(&apu.fs, value),
+                0x6E => {},
 
-            0x78 => apu.ch4.setSound4CntL(value),
-            0x7C => apu.ch4.setSound4CntH(&apu.fs, value),
+                0x70 => apu.ch3.setSound3CntL(@truncate(value)),
+                0x72 => apu.ch3.setSound3CntH(value),
+                0x74 => apu.ch3.setSound3CntX(&apu.fs, value),
+                0x76 => {},
 
-            0x80 => apu.psg_cnt.raw = value, // SOUNDCNT_L
-            0x82 => apu.setSoundCntH(value),
-            0x84 => apu.setSoundCntX(value >> 7 & 1 == 1),
-            0x88 => apu.bias.raw = value, // SOUNDBIAS
-            // WAVE_RAM
-            0x90...0x9F => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
-            else => util.io.write.undef(log, "Tried to write 0x{X:0>4}{} to 0x{X:0>8}", .{ value, T, addr }),
+                0x78 => apu.ch4.setSound4CntL(value),
+                0x7A => {},
+                0x7C => apu.ch4.setSound4CntH(&apu.fs, value),
+                0x7E => {},
+
+                0x80 => apu.setSoundCntL(value),
+                0x82 => apu.setSoundCntH(value),
+                0x84 => apu.setSoundCntX(value >> 7 & 1 == 1),
+                0x86 => {},
+                0x88 => apu.bias.raw = value, // SOUNDBIAS
+                0x8A, 0x8C, 0x8E => {},
+
+                0x90, 0x92, 0x94, 0x96, 0x98, 0x9A, 0x9C, 0x9E => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
+                0xA0, 0xA2 => log.err("Tried to write 0x{X:0>4}{} to FIFO_A", .{ value, T }),
+                0xA4, 0xA6 => log.err("Tried to write 0x{X:0>4}{} to FIFO_B", .{ value, T }),
+                else => util.io.write.undef(log, "Tried to write 0x{X:0>4}{} to 0x{X:0>8}", .{ value, T, addr }),
+            }
         },
-        u8 => switch (byte) {
-            0x60 => apu.ch1.setSound1CntL(value),
-            0x62 => apu.ch1.setNr11(value),
-            0x63 => apu.ch1.setNr12(value),
-            0x64 => apu.ch1.setNr13(value),
-            0x65 => apu.ch1.setNr14(&apu.fs, value),
+        u8 => {
+            if (byte_addr <= 0x81 and !apu.cnt.apu_enable.read()) return;
 
-            0x68 => apu.ch2.setNr21(value),
-            0x69 => apu.ch2.setNr22(value),
-            0x6C => apu.ch2.setNr23(value),
-            0x6D => apu.ch2.setNr24(&apu.fs, value),
+            switch (byte_addr) {
+                0x60 => apu.ch1.setSound1CntL(value),
+                0x61 => {},
+                0x62 => apu.ch1.setNr11(value),
+                0x63 => apu.ch1.setNr12(value),
+                0x64 => apu.ch1.setNr13(value),
+                0x65 => apu.ch1.setNr14(&apu.fs, value),
+                0x66, 0x67 => {},
 
-            0x70 => apu.ch3.setSound3CntL(value), // NR30
-            0x72 => apu.ch3.setNr31(value),
-            0x73 => apu.ch3.vol.raw = value, // NR32
-            0x74 => apu.ch3.setNr33(value),
-            0x75 => apu.ch3.setNr34(&apu.fs, value),
+                0x68 => apu.ch2.setNr21(value),
+                0x69 => apu.ch2.setNr22(value),
+                0x6A, 0x6B => {},
+                0x6C => apu.ch2.setNr23(value),
+                0x6D => apu.ch2.setNr24(&apu.fs, value),
+                0x6E, 0x6F => {},
 
-            0x78 => apu.ch4.setNr41(value),
-            0x79 => apu.ch4.setNr42(value),
-            0x7C => apu.ch4.poly.raw = value, // NR 43
-            0x7D => apu.ch4.setNr44(&apu.fs, value),
+                0x70 => apu.ch3.setSound3CntL(value), // NR30
+                0x71 => {},
+                0x72 => apu.ch3.setNr31(value),
+                0x73 => apu.ch3.vol.raw = value, // NR32
+                0x74 => apu.ch3.setNr33(value),
+                0x75 => apu.ch3.setNr34(&apu.fs, value),
+                0x76, 0x77 => {},
 
-            0x80 => apu.setNr50(value),
-            0x81 => apu.setNr51(value),
-            0x82 => apu.setSoundCntH(setLo(u16, apu.dma_cnt.raw, value)),
-            0x83 => apu.setSoundCntH(setHi(u16, apu.dma_cnt.raw, value)),
-            0x84 => apu.setSoundCntX(value >> 7 & 1 == 1), // NR52
-            0x89 => apu.setSoundBiasH(value),
-            0x90...0x9F => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
-            else => util.io.write.undef(log, "Tried to write 0x{X:0>2}{} to 0x{X:0>8}", .{ value, T, addr }),
+                0x78 => apu.ch4.setNr41(value),
+                0x79 => apu.ch4.setNr42(value),
+                0x7A, 0x7B => {},
+                0x7C => apu.ch4.poly.raw = value, // NR 43
+                0x7D => apu.ch4.setNr44(&apu.fs, value),
+                0x7E, 0x7F => {},
+
+                0x80, 0x81 => apu.setSoundCntL(setHalf(u16, apu.psg_cnt.raw, byte_addr, value)),
+                0x82, 0x83 => apu.setSoundCntH(setHalf(u16, apu.dma_cnt.raw, byte_addr, value)),
+                0x84 => apu.setSoundCntX(value >> 7 & 1 == 1),
+                0x85 => {},
+                0x86, 0x87 => {},
+                0x88, 0x89 => apu.bias.raw = setHalf(u16, apu.bias.raw, byte_addr, value), // SOUNDBIAS
+                0x8A...0x8F => {},
+
+                0x90...0x9F => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
+                0xA0...0xA3 => log.err("Tried to write 0x{X:0>2}{} to FIFO_A", .{ value, T }),
+                0xA4...0xA7 => log.err("Tried to write 0x{X:0>2}{} to FIFO_B", .{ value, T }),
+                else => util.io.write.undef(log, "Tried to write 0x{X:0>2}{} to 0x{X:0>8}", .{ value, T, addr }),
+            }
         },
         else => @compileError("APU: Unsupported write width"),
     }
@@ -189,7 +271,7 @@ pub const Apu = struct {
             .bias = .{ .raw = 0x0200 },
 
             .sampling_cycle = 0b00,
-            .stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, 1 << 15, SDL.AUDIO_U16, 2, host_sample_rate).?,
+            .stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, 1 << 15, host_format, 2, host_rate).?,
             .sched = sched,
 
             .capacitor = 0,
@@ -197,27 +279,69 @@ pub const Apu = struct {
             .is_buffer_full = false,
         };
 
-        sched.push(.SampleAudio, apu.interval());
-        sched.push(.{ .ApuChannel = 0 }, @import("apu/signal/Square.zig").interval);
-        sched.push(.{ .ApuChannel = 1 }, @import("apu/signal/Square.zig").interval);
-        sched.push(.{ .ApuChannel = 2 }, @import("apu/signal/Wave.zig").interval);
-        sched.push(.{ .ApuChannel = 3 }, @import("apu/signal/Lfsr.zig").interval);
-        sched.push(.FrameSequencer, FrameSequencer.interval);
+        Self.initEvents(apu.sched, apu.interval());
 
         return apu;
     }
 
-    fn reset(self: *Self) void {
+    fn initEvents(scheduler: *Scheduler, apu_interval: u64) void {
+        scheduler.push(.SampleAudio, apu_interval);
+        scheduler.push(.{ .ApuChannel = 0 }, @import("apu/signal/Square.zig").interval);
+        scheduler.push(.{ .ApuChannel = 1 }, @import("apu/signal/Square.zig").interval);
+        scheduler.push(.{ .ApuChannel = 2 }, @import("apu/signal/Wave.zig").interval);
+        scheduler.push(.{ .ApuChannel = 3 }, @import("apu/signal/Lfsr.zig").interval);
+        scheduler.push(.FrameSequencer, FrameSequencer.interval);
+    }
+
+    /// Used when resetting the emulator
+    pub fn reset(self: *Self) void {
+        // FIXME: These reset functions are meant to emulate obscure APU behaviour. Write proper emu reset fns
         self.ch1.reset();
         self.ch2.reset();
         self.ch3.reset();
         self.ch4.reset();
+
+        self.chA.reset();
+        self.chB.reset();
+
+        self.psg_cnt = .{ .raw = 0 };
+        self.dma_cnt = .{ .raw = 0 };
+        self.cnt = .{ .raw = 0 };
+        self.bias = .{ .raw = 0x200 };
+
+        self.sampling_cycle = 0;
+        self.fs.reset();
+
+        Self.initEvents(self.sched, self.interval());
+    }
+
+    /// Emulates the reset behaviour of the APU
+    fn _reset(self: *Self) void {
+        // All PSG Registers between 0x0400_0060..0x0400_0081 are zeroed
+        // 0x0400_0082 and 0x0400_0088 retain their values
+        self.ch1.reset();
+        self.ch2.reset();
+        self.ch3.reset();
+        self.ch4.reset();
+
+        // GBATEK says 4000060h..4000081h I take this to mean inclusive
+        self.psg_cnt.raw = 0x0000;
     }
 
     /// SOUNDCNT
     fn setSoundCnt(self: *Self, value: u32) void {
-        self.psg_cnt.raw = @truncate(u16, value);
-        self.setSoundCntH(@truncate(u16, value >> 16));
+        if (self.cnt.apu_enable.read()) self.setSoundCntL(@truncate(value));
+        self.setSoundCntH(@truncate(value >> 16));
+    }
+
+    /// SOUNDCNT_L
+    pub fn soundCntL(self: *const Self) u16 {
+        return self.psg_cnt.raw & 0xFF77;
+    }
+
+    /// SOUNDCNT_L
+    pub fn setSoundCntL(self: *Self, value: u16) void {
+        self.psg_cnt.raw = value;
     }
 
     /// SOUNDCNT_H
@@ -232,6 +356,11 @@ pub const Apu = struct {
         self.dma_cnt = new;
     }
 
+    /// SOUNDCNT_H
+    pub fn soundCntH(self: *const Self) u16 {
+        return self.dma_cnt.raw & 0x770F;
+    }
+
     /// NR52
     pub fn setSoundCntX(self: *Self, value: bool) void {
         self.cnt.apu_enable.write(value);
@@ -240,42 +369,31 @@ pub const Apu = struct {
             self.fs.step = 0; // Reset Frame Sequencer
 
             // Reset Square Wave Offsets
-            self.ch1.square.pos = 0;
-            self.ch2.square.pos = 0;
+            self.ch1.square.reset();
+            self.ch2.square.reset();
 
-            // Reset Wave Device Offsets
-            self.ch3.wave_dev.offset = 0;
+            // Reset Wave
+            self.ch3.wave_dev.reset();
+
+            // Rest Noise
+            self.ch4.lfsr.reset();
         } else {
-            self.reset();
+            self._reset();
         }
     }
 
     /// NR52
     pub fn soundCntX(self: *const Self) u8 {
-        const apu_enable: u8 = @boolToInt(self.cnt.apu_enable.read());
+        const apu_enable: u8 = @intFromBool(self.cnt.apu_enable.read());
 
-        const ch1_enable: u8 = @boolToInt(self.ch1.enabled);
-        const ch2_enable: u8 = @boolToInt(self.ch2.enabled);
-        const ch3_enable: u8 = @boolToInt(self.ch3.enabled);
-        const ch4_enable: u8 = @boolToInt(self.ch4.enabled);
+        const ch1_enable: u8 = @intFromBool(self.ch1.enabled);
+        const ch2_enable: u8 = @intFromBool(self.ch2.enabled);
+        const ch3_enable: u8 = @intFromBool(self.ch3.enabled);
+        const ch4_enable: u8 = @intFromBool(self.ch4.enabled);
 
         return apu_enable << 7 | ch4_enable << 3 | ch3_enable << 2 | ch2_enable << 1 | ch1_enable;
     }
 
-    /// NR50
-    pub fn setNr50(self: *Self, byte: u8) void {
-        self.psg_cnt.raw = (self.psg_cnt.raw & 0xFF00) | byte;
-    }
-
-    /// NR51
-    pub fn setNr51(self: *Self, byte: u8) void {
-        self.psg_cnt.raw = @as(u16, byte) << 8 | (self.psg_cnt.raw & 0xFF);
-    }
-
-    pub fn setSoundBiasH(self: *Self, byte: u8) void {
-        self.bias.raw = (@as(u16, byte) << 8) | (self.bias.raw & 0xFF);
-    }
-
     pub fn sampleAudio(self: *Self, late: u64) void {
         self.sched.push(.SampleAudio, self.interval() -| late);
 
@@ -327,19 +445,18 @@ pub const Apu = struct {
         right += if (self.dma_cnt.chB_right.read()) chB_sample else 0;
 
         // Add SOUNDBIAS
-        // FIXME: Is SOUNDBIAS 9-bit or 10-bit?
-        const bias = @as(i16, self.bias.level.read()) << 1;
+        // FIXME: SOUNDBIAS is 10-bit but The waveform is centered around 0 if I treat it as 11-bit
+        const bias = @as(i16, self.bias.level.read()) << 2;
         left += bias;
         right += bias;
 
-        const clamped_left = std.math.clamp(@bitCast(u16, left), std.math.minInt(u11), std.math.maxInt(u11));
-        const clamped_right = std.math.clamp(@bitCast(u16, right), std.math.minInt(u11), std.math.maxInt(u11));
+        const clamped_left = std.math.clamp(@as(u16, @bitCast(left)), 0, std.math.maxInt(u11));
+        const clamped_right = std.math.clamp(@as(u16, @bitCast(right)), 0, std.math.maxInt(u11));
 
         // Extend to 16-bit signed audio samples
         const ext_left = (clamped_left << 5) | (clamped_left >> 6);
         const ext_right = (clamped_right << 5) | (clamped_right >> 6);
 
-        // FIXME: This rarely happens
         if (self.sampling_cycle != self.bias.sampling_cycle.read()) self.replaceSDLResampler();
 
         _ = SDL.SDL_AudioStreamPut(self.stream, &[2]u16{ ext_left, ext_right }, 2 * @sizeOf(u16));
@@ -356,7 +473,7 @@ pub const Apu = struct {
         defer SDL.SDL_FreeAudioStream(old_stream);
 
         self.sampling_cycle = self.bias.sampling_cycle.read();
-        self.stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, @intCast(c_int, sample_rate), SDL.AUDIO_U16, 2, host_sample_rate).?;
+        self.stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, @intCast(sample_rate), host_format, 2, host_rate).?;
     }
 
     fn interval(self: *const Self) u64 {
@@ -404,14 +521,20 @@ pub const Apu = struct {
     pub fn onDmaAudioSampleRequest(self: *Self, cpu: *Arm7tdmi, tim_id: u3) void {
         if (!self.cnt.apu_enable.read()) return;
 
-        if (@boolToInt(self.dma_cnt.chA_timer.read()) == tim_id) {
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+        if (@intFromBool(self.dma_cnt.chA_timer.read()) == tim_id) {
+            if (!self.chA.enabled) return;
+
             self.chA.updateSample();
-            if (self.chA.len() <= 15) cpu.bus.dma[1].requestAudio(0x0400_00A0);
+            if (self.chA.len() <= 15) bus_ptr.dma[1].requestAudio(0x0400_00A0);
         }
 
-        if (@boolToInt(self.dma_cnt.chB_timer.read()) == tim_id) {
+        if (@intFromBool(self.dma_cnt.chB_timer.read()) == tim_id) {
+            if (!self.chB.enabled) return;
+
             self.chB.updateSample();
-            if (self.chB.len() <= 15) cpu.bus.dma[2].requestAudio(0x0400_00A4);
+            if (self.chB.len() <= 15) bus_ptr.dma[2].requestAudio(0x0400_00A4);
         }
     }
 };
@@ -423,17 +546,31 @@ pub fn DmaSound(comptime kind: DmaSoundKind) type {
         fifo: SoundFifo,
         kind: DmaSoundKind,
         sample: i8,
+        enabled: bool,
 
         fn init() Self {
             return .{
                 .fifo = SoundFifo.init(),
                 .kind = kind,
                 .sample = 0,
+                .enabled = false,
             };
         }
 
+        /// Used when resetting hte emulator (not emulation code)
+        fn reset(self: *Self) void {
+            self.* = Self.init();
+        }
+
         pub fn push(self: *Self, value: u32) void {
-            self.fifo.write(&intToBytes(u32, value)) catch |e| log.err("{} Error: {}", .{ kind, e });
+            if (!self.enabled) self.enable();
+
+            self.fifo.write(std.mem.asBytes(&value)) catch |e| log.err("{} Error: {}", .{ kind, e });
+        }
+
+        fn enable(self: *Self) void {
+            @setCold(true);
+            self.enabled = true;
         }
 
         pub fn len(self: *const Self) usize {
@@ -441,11 +578,11 @@ pub fn DmaSound(comptime kind: DmaSoundKind) type {
         }
 
         pub fn updateSample(self: *Self) void {
-            if (self.fifo.readItem()) |sample| self.sample = @bitCast(i8, sample);
+            if (self.fifo.readItem()) |sample| self.sample = @bitCast(sample);
         }
 
         pub fn amplitude(self: *const Self) i16 {
-            return @as(i16, self.sample);
+            return self.sample;
         }
     };
 }
@@ -456,13 +593,17 @@ const DmaSoundKind = enum {
 };
 
 pub const FrameSequencer = struct {
-    const interval = (1 << 24) / 512;
     const Self = @This();
+    pub const interval = (1 << 24) / 512;
 
-    step: u3,
+    step: u3 = 0,
 
     pub fn init() Self {
-        return .{ .step = 0 };
+        return .{};
+    }
+
+    pub fn reset(self: *Self) void {
+        self.* = .{};
     }
 
     pub fn tick(self: *Self) void {

src/core/apu/Noise.zig

@@ -49,10 +49,13 @@ pub fn init(sched: *Scheduler) Self {
 }
 
 pub fn reset(self: *Self) void {
-    self.len = 0;
-    self.envelope.raw = 0;
-    self.poly.raw = 0;
-    self.cnt.raw = 0;
+    self.len = 0; // NR41
+    self.envelope.raw = 0; // NR42
+    self.poly.raw = 0; // NR43
+    self.cnt.raw = 0; // NR44
+
+    self.len_dev.reset();
+    self.env_dev.reset();
 
     self.sample = 0;
     self.enabled = false;
@@ -73,14 +76,14 @@ pub fn sound4CntL(self: *const Self) u16 {
 
 /// NR41, NR42
 pub fn setSound4CntL(self: *Self, value: u16) void {
-    self.setNr41(@truncate(u8, value));
-    self.setNr42(@truncate(u8, value >> 8));
+    self.setNr41(@truncate(value));
+    self.setNr42(@truncate(value >> 8));
 }
 
 /// NR41
 pub fn setNr41(self: *Self, len: u8) void {
-    self.len = @truncate(u6, len);
-    self.len_dev.timer = @as(u7, 64) - @truncate(u6, len);
+    self.len = @truncate(len);
+    self.len_dev.timer = @as(u7, 64) - self.len;
 }
 
 /// NR42
@@ -96,8 +99,8 @@ pub fn sound4CntH(self: *const Self) u16 {
 
 /// NR43, NR44
 pub fn setSound4CntH(self: *Self, fs: *const FrameSequencer, value: u16) void {
-    self.poly.raw = @truncate(u8, value);
-    self.setNr44(fs, @truncate(u8, value >> 8));
+    self.poly.raw = @truncate(value);
+    self.setNr44(fs, @truncate(value >> 8));
 }
 
 /// NR44

src/core/apu/Tone.zig

@@ -43,9 +43,12 @@ pub fn init(sched: *Scheduler) Self {
 }
 
 pub fn reset(self: *Self) void {
-    self.duty.raw = 0;
-    self.envelope.raw = 0;
-    self.freq.raw = 0;
+    self.duty.raw = 0; // NR21
+    self.envelope.raw = 0; // NR22
+    self.freq.raw = 0; // NR32, NR24
+
+    self.len_dev.reset();
+    self.env_dev.reset();
 
     self.sample = 0;
     self.enabled = false;
@@ -74,14 +77,14 @@ pub fn sound2CntL(self: *const Self) u16 {
 
 /// NR21, NR22
 pub fn setSound2CntL(self: *Self, value: u16) void {
-    self.setNr21(@truncate(u8, value));
-    self.setNr22(@truncate(u8, value >> 8));
+    self.setNr21(@truncate(value));
+    self.setNr22(@truncate(value >> 8));
 }
 
 /// NR21
 pub fn setNr21(self: *Self, value: u8) void {
     self.duty.raw = value;
-    self.len_dev.timer = @as(u7, 64) - @truncate(u6, value);
+    self.len_dev.timer = @as(u7, 64) - @as(u6, @truncate(value));
 }
 
 /// NR22
@@ -97,8 +100,8 @@ pub fn sound2CntH(self: *const Self) u16 {
 
 /// NR23, NR24
 pub fn setSound2CntH(self: *Self, fs: *const FrameSequencer, value: u16) void {
-    self.setNr23(@truncate(u8, value));
-    self.setNr24(fs, @truncate(u8, value >> 8));
+    self.setNr23(@truncate(value));
+    self.setNr24(fs, @truncate(value >> 8));
 }
 
 /// NR23

src/core/apu/ToneSweep.zig

@@ -50,12 +50,14 @@ pub fn init(sched: *Scheduler) Self {
 }
 
 pub fn reset(self: *Self) void {
-    self.sweep.raw = 0;
-    self.sweep_dev.calc_performed = false;
+    self.sweep.raw = 0; // NR10
+    self.duty.raw = 0; // NR11
+    self.envelope.raw = 0; // NR12
+    self.freq.raw = 0; // NR13, NR14
 
-    self.duty.raw = 0;
-    self.envelope.raw = 0;
-    self.freq.raw = 0;
+    self.len_dev.reset();
+    self.sweep_dev.reset();
+    self.env_dev.reset();
 
     self.sample = 0;
     self.enabled = false;
@@ -79,8 +81,8 @@ pub fn onToneSweepEvent(self: *Self, late: u64) void {
 
 /// NR10, NR11, NR12
 pub fn setSound1Cnt(self: *Self, value: u32) void {
-    self.setSound1CntL(@truncate(u8, value));
-    self.setSound1CntH(@truncate(u16, value >> 16));
+    self.setSound1CntL(@truncate(value));
+    self.setSound1CntH(@truncate(value >> 16));
 }
 
 /// NR10
@@ -92,10 +94,9 @@ pub fn sound1CntL(self: *const Self) u8 {
 pub fn setSound1CntL(self: *Self, value: u8) void {
     const new = io.Sweep{ .raw = value };
 
-    if (self.sweep.direction.read() and !new.direction.read()) {
-        // Sweep Negate bit has been cleared
-        // If At least 1 Sweep Calculation has been made since
-        // the last trigger, the channel is immediately disabled
+    if (!new.direction.read()) {
+        // If at least one (1) sweep calculation has been made with
+        // the negate bit set (since last trigger), disable the channel
 
         if (self.sweep_dev.calc_performed) self.enabled = false;
     }
@@ -110,14 +111,14 @@ pub fn sound1CntH(self: *const Self) u16 {
 
 /// NR11, NR12
 pub fn setSound1CntH(self: *Self, value: u16) void {
-    self.setNr11(@truncate(u8, value));
-    self.setNr12(@truncate(u8, value >> 8));
+    self.setNr11(@truncate(value));
+    self.setNr12(@truncate(value >> 8));
 }
 
 /// NR11
 pub fn setNr11(self: *Self, value: u8) void {
     self.duty.raw = value;
-    self.len_dev.timer = @as(u7, 64) - @truncate(u6, value);
+    self.len_dev.timer = @as(u7, 64) - @as(u6, @truncate(value));
 }
 
 /// NR12
@@ -133,8 +134,8 @@ pub fn sound1CntX(self: *const Self) u16 {
 
 /// NR13, NR14
 pub fn setSound1CntX(self: *Self, fs: *const FrameSequencer, value: u16) void {
-    self.setNr13(@truncate(u8, value));
-    self.setNr14(fs, @truncate(u8, value >> 8));
+    self.setNr13(@truncate(value));
+    self.setNr14(fs, @truncate(value >> 8));
 }
 
 /// NR13

src/core/apu/Wave.zig

@@ -42,10 +42,13 @@ pub fn init(sched: *Scheduler) Self {
 }
 
 pub fn reset(self: *Self) void {
-    self.select.raw = 0;
-    self.length = 0;
-    self.vol.raw = 0;
-    self.freq.raw = 0;
+    self.select.raw = 0; // NR30
+    self.length = 0; // NR31
+    self.vol.raw = 0; // NR32
+    self.freq.raw = 0; // NR33, NR34
+
+    self.len_dev.reset();
+    self.wave_dev.reset();
 
     self.sample = 0;
     self.enabled = false;
@@ -61,8 +64,8 @@ pub fn tick(self: *Self, comptime kind: Tick) void {
 
 /// NR30, NR31, NR32
 pub fn setSound3Cnt(self: *Self, value: u32) void {
-    self.setSound3CntL(@truncate(u8, value));
-    self.setSound3CntH(@truncate(u16, value >> 16));
+    self.setSound3CntL(@truncate(value));
+    self.setSound3CntH(@truncate(value >> 16));
 }
 
 /// NR30
@@ -71,6 +74,11 @@ pub fn setSound3CntL(self: *Self, value: u8) void {
     if (!self.select.enabled.read()) self.enabled = false;
 }
 
+/// NR30
+pub fn sound3CntL(self: *const Self) u8 {
+    return self.select.raw & 0xE0;
+}
+
 /// NR31, NR32
 pub fn sound3CntH(self: *const Self) u16 {
     return @as(u16, self.length & 0xE0) << 8;
@@ -78,8 +86,8 @@ pub fn sound3CntH(self: *const Self) u16 {
 
 /// NR31, NR32
 pub fn setSound3CntH(self: *Self, value: u16) void {
-    self.setNr31(@truncate(u8, value));
-    self.vol.raw = (@truncate(u8, value >> 8));
+    self.setNr31(@truncate(value));
+    self.vol.raw = @truncate(value >> 8);
 }
 
 /// NR31
@@ -90,8 +98,13 @@ pub fn setNr31(self: *Self, len: u8) void {
 
 /// NR33, NR34
 pub fn setSound3CntX(self: *Self, fs: *const FrameSequencer, value: u16) void {
-    self.setNr33(@truncate(u8, value));
-    self.setNr34(fs, @truncate(u8, value >> 8));
+    self.setNr33(@truncate(value));
+    self.setNr34(fs, @truncate(value >> 8));
+}
+
+/// NR33, NR34
+pub fn sound3CntX(self: *const Self) u16 {
+    return self.freq.raw & 0x4000;
 }
 
 /// NR33

src/core/apu/device/Envelope.zig

@@ -3,12 +3,16 @@ const io = @import("../../bus/io.zig");
 const Self = @This();
 
 /// Period Timer
-timer: u3,
+timer: u3 = 0,
 /// Current Volume
-vol: u4,
+vol: u4 = 0,
 
 pub fn create() Self {
-    return .{ .timer = 0, .vol = 0 };
+    return .{};
+}
+
+pub fn reset(self: *Self) void {
+    self.* = .{};
 }
 
 pub fn tick(self: *Self, nrx2: io.Envelope) void {

src/core/apu/device/Length.zig

@@ -1,9 +1,13 @@
 const Self = @This();
 
-timer: u9,
+timer: u9 = 0,
 
 pub fn create() Self {
-    return .{ .timer = 0 };
+    return .{};
+}
+
+pub fn reset(self: *Self) void {
+    self.* = .{};
 }
 
 pub fn tick(self: *Self, enabled: bool, ch_enable: *bool) void {

src/core/apu/device/Sweep.zig

@@ -3,19 +3,18 @@ const ToneSweep = @import("../ToneSweep.zig");
 
 const Self = @This();
 
-timer: u8,
-enabled: bool,
-shadow: u11,
+timer: u8 = 0,
+enabled: bool = false,
+shadow: u11 = 0,
 
-calc_performed: bool,
+calc_performed: bool = false,
 
 pub fn create() Self {
-    return .{
-        .timer = 0,
-        .enabled = false,
-        .shadow = 0,
-        .calc_performed = false,
-    };
+    return .{};
+}
+
+pub fn reset(self: *Self) void {
+    self.* = .{};
 }
 
 pub fn tick(self: *Self, ch1: *ToneSweep) void {
@@ -24,14 +23,13 @@ pub fn tick(self: *Self, ch1: *ToneSweep) void {
     if (self.timer == 0) {
         const period = ch1.sweep.period.read();
         self.timer = if (period == 0) 8 else period;
-        if (!self.calc_performed) self.calc_performed = true;
 
         if (self.enabled and period != 0) {
             const new_freq = self.calculate(ch1.sweep, &ch1.enabled);
 
             if (new_freq <= 0x7FF and ch1.sweep.shift.read() != 0) {
-                ch1.freq.frequency.write(@truncate(u11, new_freq));
-                self.shadow = @truncate(u11, new_freq);
+                ch1.freq.frequency.write(@as(u11, @truncate(new_freq)));
+                self.shadow = @truncate(new_freq);
 
                 _ = self.calculate(ch1.sweep, &ch1.enabled);
             }
@@ -45,7 +43,10 @@ pub fn calculate(self: *Self, sweep: io.Sweep, ch_enable: *bool) u12 {
     const shadow_shifted = shadow >> sweep.shift.read();
     const decrease = sweep.direction.read();
 
-    const freq = if (decrease) shadow - shadow_shifted else shadow + shadow_shifted;
+    const freq = if (decrease) blk: {
+        self.calc_performed = true;
+        break :blk shadow - shadow_shifted;
+    } else shadow + shadow_shifted;
     if (freq > 0x7FF) ch_enable.* = false;
 
     return freq;

src/core/apu/signal/Lfsr.zig

@@ -1,9 +1,7 @@
+//! Linear Feedback Shift Register
 const io = @import("../../bus/io.zig");
 
-/// Linear Feedback Shift Register
 const Scheduler = @import("../../scheduler.zig").Scheduler;
-const FrameSequencer = @import("../../apu.zig").FrameSequencer;
-const Noise = @import("../Noise.zig");
 
 const Self = @This();
 pub const interval: u64 = (1 << 24) / (1 << 22);
@@ -21,6 +19,11 @@ pub fn create(sched: *Scheduler) Self {
     };
 }
 
+pub fn reset(self: *Self) void {
+    self.shift = 0;
+    self.timer = 0;
+}
+
 pub fn sample(self: *const Self) i8 {
     return if ((~self.shift & 1) == 1) 1 else -1;
 }
@@ -35,7 +38,7 @@ pub fn reload(self: *Self, poly: io.PolyCounter) void {
 }
 
 /// Scheduler Event Handler for LFSR Timer Expire
-/// FIXME: This gets called a lot, clogging up the Scheduler
+/// FIXME: This gets called a lot, slowing down the scheduler
 pub fn onLfsrTimerExpire(self: *Self, poly: io.PolyCounter, late: u64) void {
     // Obscure: "Using a noise channel clock shift of 14 or 15
     // results in the LFSR receiving no clocks."

src/core/apu/signal/Square.zig

@@ -2,7 +2,6 @@ const std = @import("std");
 const io = @import("../../bus/io.zig");
 
 const Scheduler = @import("../../scheduler.zig").Scheduler;
-const FrameSequencer = @import("../../apu.zig").FrameSequencer;
 const ToneSweep = @import("../ToneSweep.zig");
 const Tone = @import("../Tone.zig");
 
@@ -21,6 +20,11 @@ pub fn init(sched: *Scheduler) Self {
     };
 }
 
+pub fn reset(self: *Self) void {
+    self.timer = 0;
+    self.pos = 0;
+}
+
 /// Scheduler Event Handler for Square Synth Timer Expire
 pub fn onSquareTimerExpire(self: *Self, comptime T: type, nrx34: io.Frequency, late: u64) void {
     comptime std.debug.assert(T == ToneSweep or T == Tone);

src/core/apu/signal/Wave.zig

@@ -2,8 +2,6 @@ const std = @import("std");
 const io = @import("../../bus/io.zig");
 
 const Scheduler = @import("../../scheduler.zig").Scheduler;
-const FrameSequencer = @import("../../apu.zig").FrameSequencer;
-const Wave = @import("../Wave.zig");
 
 const buf_len = 0x20;
 pub const interval: u64 = (1 << 24) / (1 << 22);
@@ -20,7 +18,7 @@ pub fn read(self: *const Self, comptime T: type, nr30: io.WaveSelect, addr: u32)
     const base = if (!nr30.bank.read()) @as(u32, 0x10) else 0; // Read from the Opposite Bank in Use
 
     const i = base + addr - 0x0400_0090;
-    return std.mem.readIntSliceLittle(T, self.buf[i..][0..@sizeOf(T)]);
+    return std.mem.readInt(T, self.buf[i..][0..@sizeOf(T)], .little);
 }
 
 pub fn write(self: *Self, comptime T: type, nr30: io.WaveSelect, addr: u32, value: T) void {
@@ -28,7 +26,7 @@ pub fn write(self: *Self, comptime T: type, nr30: io.WaveSelect, addr: u32, valu
     const base = if (!nr30.bank.read()) @as(u32, 0x10) else 0; // Write to the Opposite Bank in Use
 
     const i = base + addr - 0x0400_0090;
-    std.mem.writeIntSliceLittle(T, self.buf[i..][0..@sizeOf(T)], value);
+    std.mem.writeInt(T, self.buf[i..][0..@sizeOf(T)], value, .little);
 }
 
 pub fn init(sched: *Scheduler) Self {
@@ -40,6 +38,13 @@ pub fn init(sched: *Scheduler) Self {
     };
 }
 
+pub fn reset(self: *Self) void {
+    self.timer = 0;
+    self.offset = 0;
+
+    // sample buffer isn't reset because it's outside of the range of what NR52{7}'s effects
+}
+
 /// Reload internal Wave Timer
 pub fn reload(self: *Self, value: u11) void {
     self.sched.removeScheduledEvent(.{ .ApuChannel = 2 });
@@ -65,7 +70,7 @@ pub fn sample(self: *const Self, nr30: io.WaveSelect) u4 {
     const base = if (nr30.bank.read()) @as(u32, 0x10) else 0;
 
     const value = self.buf[base + self.offset / 2];
-    return if (self.offset & 1 == 0) @truncate(u4, value >> 4) else @truncate(u4, value);
+    return if (self.offset & 1 == 0) @truncate(value >> 4) else @truncate(value);
 }
 
 /// TODO: Write comment

src/core/bus/Bios.zig

@@ -3,6 +3,9 @@ const std = @import("std");
 const Allocator = std.mem.Allocator;
 const log = std.log.scoped(.Bios);
 
+const rotr = @import("zba-util").rotr;
+const forceAlign = @import("../Bus.zig").forceAlign;
+
 /// Size of the BIOS in bytes
 pub const size = 0x4000;
 const Self = @This();
@@ -10,21 +13,37 @@ const Self = @This();
 buf: ?[]u8,
 allocator: Allocator,
 
-addr_latch: u32,
+addr_latch: u32 = 0,
 
-pub fn read(self: *Self, comptime T: type, r15: u32, addr: u32) T {
+// https://github.com/ITotalJustice/notorious_beeg/issues/106
+pub fn read(self: *Self, comptime T: type, r15: u32, address: u32) T {
     if (r15 < Self.size) {
+        const addr = forceAlign(T, address);
+
         self.addr_latch = addr;
         return self._read(T, addr);
     }
 
-    log.debug("Rejected read since r15=0x{X:0>8}", .{r15});
-    return @truncate(T, self._read(T, self.addr_latch + 8));
+    log.warn("Open Bus! Read from 0x{X:0>8}, but PC was 0x{X:0>8}", .{ address, r15 });
+    const value = self._read(u32, self.addr_latch);
+
+    return @truncate(rotr(u32, value, 8 * rotateBy(T, address)));
 }
 
-pub fn dbgRead(self: *const Self, comptime T: type, r15: u32, addr: u32) T {
-    if (r15 < Self.size) return self._read(T, addr);
-    return @truncate(T, self._read(T, self.addr_latch + 8));
+fn rotateBy(comptime T: type, address: u32) u32 {
+    return switch (T) {
+        u8 => address & 3,
+        u16 => address & 2,
+        u32 => 0,
+        else => @compileError("bios: unsupported read width"),
+    };
+}
+
+pub fn dbgRead(self: *const Self, comptime T: type, r15: u32, address: u32) T {
+    if (r15 < Self.size) return self._read(T, forceAlign(T, address));
+
+    const value = self._read(u32, self.addr_latch);
+    return @truncate(rotr(u32, value, 8 * rotateBy(T, address)));
 }
 
 /// Read without the GBA safety checks
@@ -32,7 +51,7 @@ fn _read(self: *const Self, comptime T: type, addr: u32) T {
     const buf = self.buf orelse std.debug.panic("[BIOS] ZBA tried to read {} from 0x{X:0>8} but not BIOS was present", .{ T, addr });
 
     return switch (T) {
-        u32, u16, u8 => std.mem.readIntSliceLittle(T, buf[addr..][0..@sizeOf(T)]),
+        u32, u16, u8 => std.mem.readInt(T, buf[addr..][0..@sizeOf(T)], .little),
         else => @compileError("BIOS: Unsupported read width"),
     };
 }
@@ -43,18 +62,28 @@ pub fn write(_: *Self, comptime T: type, addr: u32, value: T) void {
 }
 
 pub fn init(allocator: Allocator, maybe_path: ?[]const u8) !Self {
-    const buf: ?[]u8 = if (maybe_path) |path| blk: {
-        const file = try std.fs.cwd().openFile(path, .{});
-        defer file.close();
+    if (maybe_path == null) return .{ .buf = null, .allocator = allocator };
+    const file_path = maybe_path.?;
 
-        break :blk try file.readToEndAlloc(allocator, try file.getEndPos());
-    } else null;
+    const buf = try allocator.alloc(u8, Self.size);
+    errdefer allocator.free(buf);
 
-    return Self{
-        .buf = buf,
-        .allocator = allocator,
-        .addr_latch = 0,
-    };
+    var self: Self = .{ .buf = buf, .allocator = allocator };
+    try self.load(file_path);
+
+    return self;
+}
+
+pub fn load(self: *Self, file_path: []const u8) !void {
+    const file = try std.fs.cwd().openFile(file_path, .{});
+    defer file.close();
+
+    const len = try file.readAll(self.buf orelse return error.UnallocatedBuffer);
+    if (len != Self.size) log.err("Expected BIOS to be {}B, was {}B", .{ Self.size, len });
+}
+
+pub fn reset(self: *Self) void {
+    self.addr_latch = 0;
 }
 
 pub fn deinit(self: *Self) void {

src/core/bus/Ewram.zig

@@ -11,7 +11,7 @@ pub fn read(self: *const Self, comptime T: type, address: usize) T {
     const addr = address & 0x3FFFF;
 
     return switch (T) {
-        u32, u16, u8 => std.mem.readIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)]),
+        u32, u16, u8 => std.mem.readInt(T, self.buf[addr..][0..@sizeOf(T)], .little),
         else => @compileError("EWRAM: Unsupported read width"),
     };
 }
@@ -20,14 +20,14 @@ pub fn write(self: *const Self, comptime T: type, address: usize, value: T) void
     const addr = address & 0x3FFFF;
 
     return switch (T) {
-        u32, u16, u8 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
+        u32, u16, u8 => std.mem.writeInt(T, self.buf[addr..][0..@sizeOf(T)], value, .little),
         else => @compileError("EWRAM: Unsupported write width"),
     };
 }
 
 pub fn init(allocator: Allocator) !Self {
     const buf = try allocator.alloc(u8, ewram_size);
-    std.mem.set(u8, buf, 0);
+    @memset(buf, 0);
 
     return Self{
         .buf = buf,
@@ -35,6 +35,10 @@ pub fn init(allocator: Allocator) !Self {
     };
 }
 
+pub fn reset(self: *Self) void {
+    @memset(self.buf, 0);
+}
+
 pub fn deinit(self: *Self) void {
     self.allocator.free(self.buf);
     self.* = undefined;

src/core/bus/GamePak.zig

@@ -1,11 +1,7 @@
 const std = @import("std");
 const config = @import("../../config.zig");
 
-const Bit = @import("bitfield").Bit;
-const Bitfield = @import("bitfield").Bitfield;
-const DateTime = @import("datetime").datetime.Datetime;
-
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
 const Backup = @import("backup.zig").Backup;
 const Gpio = @import("gpio.zig").Gpio;
 const Allocator = std.mem.Allocator;
@@ -34,7 +30,7 @@ pub fn read(self: *Self, comptime T: type, address: u32) T {
             // Addresses 0x0D00_0000 to 0x0DFF_FFFF are reserved for EEPROM accesses if
             // * Backup type is EEPROM
             // * Small ROM (less than 16MB)
-            if (@truncate(u8, address >> 24) == 0x0D)
+            if (@as(u8, @truncate(address >> 24)) == 0x0D)
                 return self.backup.eeprom.read();
         }
     }
@@ -81,7 +77,7 @@ inline fn get(self: *const Self, i: u32) u8 {
     if (i < self.buf.len) return self.buf[i];
 
     const lhs = i >> 1 & 0xFFFF;
-    return @truncate(u8, lhs >> 8 * @truncate(u5, i & 1));
+    return @truncate(lhs >> 8 * @as(u5, @truncate(i & 1)));
 }
 
 pub fn dbgRead(self: *const Self, comptime T: type, address: u32) T {
@@ -98,7 +94,7 @@ pub fn dbgRead(self: *const Self, comptime T: type, address: u32) T {
             // Addresses 0x0D00_0000 to 0x0DFF_FFFF are reserved for EEPROM accesses if
             // * Backup type is EEPROM
             // * Small ROM (less than 16MB)
-            if (@truncate(u8, address >> 24) == 0x0D)
+            if (@as(u8, @truncate(address >> 24)) == 0x0D)
                 return self.backup.eeprom.dbgRead();
         }
     }
@@ -109,14 +105,13 @@ pub fn dbgRead(self: *const Self, comptime T: type, address: u32) T {
 
         switch (T) {
             u32 => switch (address) {
-                // TODO: Do I even need to implement these?
+                // FIXME: Do I even need to implement these?
                 0x0800_00C4 => std.debug.panic("Handle 32-bit GPIO Data/Direction Reads", .{}),
                 0x0800_00C6 => std.debug.panic("Handle 32-bit GPIO Direction/Control Reads", .{}),
                 0x0800_00C8 => std.debug.panic("Handle 32-bit GPIO Control Reads", .{}),
                 else => {},
             },
             u16 => switch (address) {
-                // FIXME: What do 16-bit GPIO Reads look like?
                 0x0800_00C4 => return self.gpio.read(.Data),
                 0x0800_00C6 => return self.gpio.read(.Direction),
                 0x0800_00C8 => return self.gpio.read(.Control),
@@ -144,7 +139,7 @@ pub fn write(self: *Self, comptime T: type, word_count: u16, address: u32, value
     const addr = address & 0x1FF_FFFF;
 
     if (self.backup.kind == .Eeprom) {
-        const bit = @truncate(u1, value);
+        const bit: u1 = @truncate(value);
 
         if (self.buf.len > 0x100_0000) { // Large
             // Addresses 0x1FF_FF00 to 0x1FF_FFFF are reserved from EEPROM accesses if
@@ -156,7 +151,7 @@ pub fn write(self: *Self, comptime T: type, word_count: u16, address: u32, value
             // Addresses 0x0D00_0000 to 0x0DFF_FFFF are reserved for EEPROM accesses if
             // * Backup type is EEPROM
             // * Small ROM (less than 16MB)
-            if (@truncate(u8, address >> 24) == 0x0D)
+            if (@as(u8, @truncate(address >> 24)) == 0x0D)
                 return self.backup.eeprom.write(word_count, &self.backup.buf, bit);
         }
     }
@@ -164,19 +159,19 @@ pub fn write(self: *Self, comptime T: type, word_count: u16, address: u32, value
     switch (T) {
         u32 => switch (address) {
             0x0800_00C4 => {
-                self.gpio.write(.Data, @truncate(u4, value));
-                self.gpio.write(.Direction, @truncate(u4, value >> 16));
+                self.gpio.write(.Data, @as(u4, @truncate(value)));
+                self.gpio.write(.Direction, @as(u4, @truncate(value >> 16)));
             },
             0x0800_00C6 => {
-                self.gpio.write(.Direction, @truncate(u4, value));
-                self.gpio.write(.Control, @truncate(u1, value >> 16));
+                self.gpio.write(.Direction, @as(u4, @truncate(value)));
+                self.gpio.write(.Control, @as(u1, @truncate(value >> 16)));
             },
             else => log.err("Wrote {} 0x{X:0>8} to 0x{X:0>8}, Unhandled", .{ T, value, address }),
         },
         u16 => switch (address) {
-            0x0800_00C4 => self.gpio.write(.Data, @truncate(u4, value)),
-            0x0800_00C6 => self.gpio.write(.Direction, @truncate(u4, value)),
-            0x0800_00C8 => self.gpio.write(.Control, @truncate(u1, value)),
+            0x0800_00C4 => self.gpio.write(.Data, @as(u4, @truncate(value))),
+            0x0800_00C6 => self.gpio.write(.Direction, @as(u4, @truncate(value))),
+            0x0800_00C8 => self.gpio.write(.Control, @as(u1, @truncate(value))),
             else => log.err("Wrote {} 0x{X:0>4} to 0x{X:0>8}, Unhandled", .{ T, value, address }),
         },
         u8 => log.debug("Wrote {} 0x{X:0>2} to 0x{X:0>8}, Ignored.", .{ T, value, address }),
@@ -184,23 +179,30 @@ pub fn write(self: *Self, comptime T: type, word_count: u16, address: u32, value
     }
 }
 
-pub fn init(allocator: Allocator, cpu: *Arm7tdmi, rom_path: []const u8, save_path: ?[]const u8) !Self {
-    const file = try std.fs.cwd().openFile(rom_path, .{});
-    defer file.close();
+pub fn init(allocator: Allocator, cpu: *Arm7tdmi, maybe_rom: ?[]const u8, maybe_save: ?[]const u8) !Self {
+    const Device = Gpio.Device;
 
-    const file_buf = try file.readToEndAlloc(allocator, try file.getEndPos());
-    const title = file_buf[0xA0..0xAC].*;
-    const kind = Backup.guess(file_buf);
-    const device = if (config.config().guest.force_rtc) .Rtc else guessDevice(file_buf);
+    const items: struct { []u8, [12]u8, Backup.Kind, Device.Kind } = if (maybe_rom) |file_path| blk: {
+        const file = try std.fs.cwd().openFile(file_path, .{});
+        defer file.close();
 
-    logHeader(file_buf, &title);
+        const buffer = try file.readToEndAlloc(allocator, try file.getEndPos());
+        const title = buffer[0xA0..0xAC];
+        logHeader(buffer, title);
+
+        const device_kind = if (config.config().guest.force_rtc) .Rtc else guessDevice(buffer);
+
+        break :blk .{ buffer, title.*, Backup.guess(buffer), device_kind };
+    } else .{ try allocator.alloc(u8, 0), [_]u8{0} ** 12, .None, .None };
+
+    const title = items[1];
 
     return .{
-        .buf = file_buf,
+        .buf = items[0],
         .allocator = allocator,
         .title = title,
-        .backup = try Backup.init(allocator, kind, title, save_path),
-        .gpio = try Gpio.init(allocator, cpu, device),
+        .backup = try Backup.init(allocator, items[2], title, maybe_save),
+        .gpio = try Gpio.init(allocator, cpu, items[3]),
     };
 }
 
@@ -218,25 +220,24 @@ fn guessDevice(buf: []const u8) Gpio.Device.Kind {
     // Try to Guess if ROM uses RTC
     const needle = "RTC_V"; // I was told SIIRTC_V, though Pokemen Firered (USA) is a false negative
 
+    // TODO: Use new for loop syntax?
     var i: usize = 0;
     while ((i + needle.len) < buf.len) : (i += 1) {
         if (std.mem.eql(u8, needle, buf[i..(i + needle.len)])) return .Rtc;
     }
 
     // TODO: Detect other GPIO devices
-
     return .None;
 }
 
 fn logHeader(buf: []const u8, title: *const [12]u8) void {
-    const code = buf[0xAC..0xB0];
-    const maker = buf[0xB0..0xB2];
     const version = buf[0xBC];
 
     log.info("Title: {s}", .{title});
     if (version != 0) log.info("Version: {}", .{version});
-    log.info("Game Code: {s}", .{code});
-    log.info("Maker Code: {s}", .{maker});
+
+    log.info("Game Code: {s}", .{buf[0xAC..0xB0]});
+    log.info("Maker Code: {s}", .{buf[0xB0..0xB2]});
 }
 
 test "OOB Access" {

src/core/bus/Iwram.zig

@@ -11,7 +11,7 @@ pub fn read(self: *const Self, comptime T: type, address: usize) T {
     const addr = address & 0x7FFF;
 
     return switch (T) {
-        u32, u16, u8 => std.mem.readIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)]),
+        u32, u16, u8 => std.mem.readInt(T, self.buf[addr..][0..@sizeOf(T)], .little),
         else => @compileError("IWRAM: Unsupported read width"),
     };
 }
@@ -20,14 +20,14 @@ pub fn write(self: *const Self, comptime T: type, address: usize, value: T) void
     const addr = address & 0x7FFF;
 
     return switch (T) {
-        u32, u16, u8 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
+        u32, u16, u8 => std.mem.writeInt(T, self.buf[addr..][0..@sizeOf(T)], value, .little),
         else => @compileError("IWRAM: Unsupported write width"),
     };
 }
 
 pub fn init(allocator: Allocator) !Self {
     const buf = try allocator.alloc(u8, iwram_size);
-    std.mem.set(u8, buf, 0);
+    @memset(buf, 0);
 
     return Self{
         .buf = buf,
@@ -35,6 +35,10 @@ pub fn init(allocator: Allocator) !Self {
     };
 }
 
+pub fn reset(self: *Self) void {
+    @memset(self.buf, 0);
+}
+
 pub fn deinit(self: *Self) void {
     self.allocator.free(self.buf);
     self.* = undefined;

src/core/bus/backup.zig

@@ -6,7 +6,6 @@ const Eeprom = @import("backup/eeprom.zig").Eeprom;
 const Flash = @import("backup/Flash.zig");
 
 const escape = @import("../../util.zig").escape;
-const span = @import("../../util.zig").span;
 
 const Needle = struct { str: []const u8, kind: Backup.Kind };
 const backup_kinds = [6]Needle{
@@ -33,7 +32,7 @@ pub const Backup = struct {
     flash: Flash,
     eeprom: Eeprom,
 
-    const Kind = enum {
+    pub const Kind = enum {
         Eeprom,
         Sram,
         Flash,
@@ -78,7 +77,7 @@ pub const Backup = struct {
 
                 switch (addr) {
                     0x0000 => if (self.kind == .Flash1M and self.flash.set_bank) {
-                        self.flash.bank = @truncate(u1, byte);
+                        self.flash.bank = @truncate(byte);
                     },
                     0x5555 => {
                         if (self.flash.state == .Command) {
@@ -111,7 +110,7 @@ pub const Backup = struct {
         };
 
         const buf = try allocator.alloc(u8, buf_size);
-        std.mem.set(u8, buf, 0xFF);
+        @memset(buf, 0xFF);
 
         var backup = Self{
             .buf = buf,
@@ -138,6 +137,7 @@ pub const Backup = struct {
         for (backup_kinds) |needle| {
             const needle_len = needle.str.len;
 
+            // TODO: Use new for loop syntax?
             var i: usize = 0;
             while ((i + needle_len) < rom.len) : (i += 1) {
                 if (std.mem.eql(u8, needle.str, rom[i..][0..needle_len])) return needle.kind;
@@ -151,8 +151,8 @@ pub const Backup = struct {
         const file_path = try self.savePath(allocator, path);
         defer allocator.free(file_path);
 
-        // FIXME: Don't rely on this lol
-        if (std.mem.eql(u8, file_path[file_path.len - 12 .. file_path.len], "untitled.sav")) {
+        const expected = "untitled.sav";
+        if (std.mem.eql(u8, file_path[file_path.len - expected.len .. file_path.len], expected)) {
             return log.err("ROM header lacks title, no save loaded", .{});
         }
 
@@ -163,7 +163,7 @@ pub const Backup = struct {
         switch (self.kind) {
             .Sram, .Flash, .Flash1M => {
                 if (self.buf.len == file_buf.len) {
-                    std.mem.copy(u8, self.buf, file_buf);
+                    @memcpy(self.buf, file_buf);
                     return log.info("Loaded Save from {s}", .{file_path});
                 }
 
@@ -174,7 +174,7 @@ pub const Backup = struct {
                     self.eeprom.kind = if (file_buf.len == 0x200) .Small else .Large;
 
                     self.buf = try allocator.alloc(u8, file_buf.len);
-                    std.mem.copy(u8, self.buf, file_buf);
+                    @memcpy(self.buf, file_buf);
                     return log.info("Loaded Save from {s}", .{file_path});
                 }
 
@@ -195,7 +195,7 @@ pub const Backup = struct {
     }
 
     fn saveName(self: *const Self, allocator: Allocator) ![]const u8 {
-        const title_str = span(&escape(self.title));
+        const title_str = std.mem.sliceTo(&escape(self.title), 0);
         const name = if (title_str.len != 0) title_str else "untitled";
 
         return try std.mem.concat(allocator, u8, &[_][]const u8{ name, ".sav" });

src/core/bus/backup/Flash.zig

@@ -44,7 +44,7 @@ pub fn handleCommand(self: *Self, buf: []u8, byte: u8) void {
         0xB0 => self.set_bank = true,
         0x80 => self.prep_erase = true,
         0x10 => {
-            std.mem.set(u8, buf, 0xFF);
+            @memset(buf, 0xFF);
             self.prep_erase = false;
         },
         0xA0 => self.prep_write = true,
@@ -61,7 +61,7 @@ pub fn shouldEraseSector(self: *const Self, addr: usize, byte: u8) bool {
 pub fn erase(self: *Self, buf: []u8, sector: usize) void {
     const start = self.address() + (sector & 0xF000);
 
-    std.mem.set(u8, buf[start..][0..0x1000], 0xFF);
+    @memset(buf[start..][0..0x1000], 0xFF);
     self.prep_erase = false;
     self.state = .Ready;
 }

src/core/bus/backup/eeprom.zig

@@ -58,12 +58,14 @@ pub const Eeprom = struct {
                     log.err("Failed to resize EEPROM buf to {} bytes", .{len});
                     std.debug.panic("EEPROM entered irrecoverable state {}", .{e});
                 };
-                std.mem.set(u8, buf.*, 0xFF);
+
+                // FIXME: ptr to a slice?
+                @memset(buf.*, 0xFF);
             }
         }
 
         if (self.state == .RequestEnd) {
-            if (bit != 0) log.debug("EEPROM Request did not end in 0u1. TODO: is this ok?", .{});
+            // if (bit != 0) log.debug("EEPROM Request did not end in 0u1. TODO: is this ok?", .{});
             self.state = .Ready;
             return;
         }
@@ -106,7 +108,7 @@ pub const Eeprom = struct {
         switch (self.state) {
             .Ready => {
                 if (self.writer.len() == 2) {
-                    const req = @intCast(u2, self.writer.finish());
+                    const req: u2 = @intCast(self.writer.finish());
                     switch (req) {
                         0b11 => self.state = .Read,
                         0b10 => self.state = .Write,
@@ -118,8 +120,8 @@ pub const Eeprom = struct {
                 switch (self.kind) {
                     .Large => {
                         if (self.writer.len() == 14) {
-                            const addr = @intCast(u10, self.writer.finish());
-                            const value = std.mem.readIntSliceLittle(u64, buf[@as(u13, addr) * 8 ..][0..8]);
+                            const addr: u10 = @intCast(self.writer.finish());
+                            const value = std.mem.readInt(u64, buf[@as(u13, addr) * 8 ..][0..8], .little);
 
                             self.reader.configure(value);
                             self.state = .RequestEnd;
@@ -128,8 +130,8 @@ pub const Eeprom = struct {
                     .Small => {
                         if (self.writer.len() == 6) {
                             // FIXME: Duplicated code from above
-                            const addr = @intCast(u6, self.writer.finish());
-                            const value = std.mem.readIntSliceLittle(u64, buf[@as(u13, addr) * 8 ..][0..8]);
+                            const addr: u6 = @intCast(self.writer.finish());
+                            const value = std.mem.readInt(u64, buf[@as(u13, addr) * 8 ..][0..8], .little);
 
                             self.reader.configure(value);
                             self.state = .RequestEnd;
@@ -142,13 +144,13 @@ pub const Eeprom = struct {
                 switch (self.kind) {
                     .Large => {
                         if (self.writer.len() == 14) {
-                            self.addr = @intCast(u10, self.writer.finish());
+                            self.addr = @as(u10, @intCast(self.writer.finish()));
                             self.state = .WriteTransfer;
                         }
                     },
                     .Small => {
                         if (self.writer.len() == 6) {
-                            self.addr = @intCast(u6, self.writer.finish());
+                            self.addr = @as(u6, @intCast(self.writer.finish()));
                             self.state = .WriteTransfer;
                         }
                     },
@@ -157,7 +159,7 @@ pub const Eeprom = struct {
             },
             .WriteTransfer => {
                 if (self.writer.len() == 64) {
-                    std.mem.writeIntSliceLittle(u64, buf[self.addr * 8 ..][0..8], self.writer.finish());
+                    std.mem.writeInt(u64, buf[self.addr * 8 ..][0..8], self.writer.finish(), .little);
                     self.state = .RequestEnd;
                 }
             },
@@ -184,11 +186,9 @@ const Reader = struct {
     fn read(self: *Self) u1 {
         if (!self.enabled) return 1;
 
-        const bit = if (self.i < 4) blk: {
-            break :blk 0;
-        } else blk: {
-            const idx = @intCast(u6, 63 - (self.i - 4));
-            break :blk @truncate(u1, self.data >> idx);
+        const bit: u1 = if (self.i < 4) 0 else blk: {
+            const idx: u6 = @intCast(63 - (self.i - 4));
+            break :blk @truncate(self.data >> idx);
         };
 
         self.i = (self.i + 1) % (64 + 4);
@@ -200,11 +200,11 @@ const Reader = struct {
     fn dbgRead(self: *const Self) u1 {
         if (!self.enabled) return 1;
 
-        const bit = if (self.i < 4) blk: {
+        const bit: u1 = if (self.i < 4) blk: {
             break :blk 0;
         } else blk: {
-            const idx = @intCast(u6, 63 - (self.i - 4));
-            break :blk @truncate(u1, self.data >> idx);
+            const idx: u6 = @intCast(63 - (self.i - 4));
+            break :blk @truncate(self.data >> idx);
         };
 
         return bit;
@@ -228,7 +228,7 @@ const Writer = struct {
     }
 
     fn requestWrite(self: *Self, bit: u1) void {
-        const idx = @intCast(u1, 1 - self.i);
+        const idx: u1 = @intCast(1 - self.i);
         self.data = (self.data & ~(@as(u64, 1) << idx)) | (@as(u64, bit) << idx);
         self.i += 1;
     }
@@ -242,13 +242,13 @@ const Writer = struct {
             .Unknown => unreachable,
         };
 
-        const idx = @intCast(u4, size - self.i);
+        const idx: u4 = @intCast(size - self.i);
         self.data = (self.data & ~(@as(u64, 1) << idx)) | (@as(u64, bit) << idx);
         self.i += 1;
     }
 
     fn dataWrite(self: *Self, bit: u1) void {
-        const idx = @intCast(u6, 63 - self.i);
+        const idx: u6 = @intCast(63 - self.i);
         self.data = (self.data & ~(@as(u64, 1) << idx)) | (@as(u64, bit) << idx);
         self.i += 1;
     }

src/core/bus/dma.zig

@@ -3,91 +3,143 @@ const util = @import("../../util.zig");
 
 const DmaControl = @import("io.zig").DmaControl;
 const Bus = @import("../Bus.zig");
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
 
-pub const DmaTuple = std.meta.Tuple(&[_]type{ DmaController(0), DmaController(1), DmaController(2), DmaController(3) });
+pub const DmaTuple = struct { DmaController(0), DmaController(1), DmaController(2), DmaController(3) };
 const log = std.log.scoped(.DmaTransfer);
 
-const setHi = util.setHi;
-const setLo = util.setLo;
+const getHalf = util.getHalf;
+const setHalf = util.setHalf;
+const setQuart = util.setQuart;
+const handleInterrupt = @import("../cpu_util.zig").handleInterrupt;
+
+const rotr = @import("zba-util").rotr;
 
 pub fn create() DmaTuple {
     return .{ DmaController(0).init(), DmaController(1).init(), DmaController(2).init(), DmaController(3).init() };
 }
 
 pub fn read(comptime T: type, dma: *const DmaTuple, addr: u32) ?T {
-    const byte = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     return switch (T) {
-        u32 => switch (byte) {
-            0xB8 => @as(T, dma.*[0].cnt.raw) << 16,
-            0xC4 => @as(T, dma.*[1].cnt.raw) << 16,
-            0xD0 => @as(T, dma.*[2].cnt.raw) << 16,
-            0xDC => @as(T, dma.*[3].cnt.raw) << 16,
-            else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
+        u32 => switch (byte_addr) {
+            0xB0, 0xB4 => null, // DMA0SAD, DMA0DAD,
+            0xB8 => @as(T, dma.*[0].dmacntH()) << 16, // DMA0CNT_L is write-only
+            0xBC, 0xC0 => null, // DMA1SAD, DMA1DAD
+            0xC4 => @as(T, dma.*[1].dmacntH()) << 16, // DMA1CNT_L is write-only
+            0xC8, 0xCC => null, // DMA2SAD, DMA2DAD
+            0xD0 => @as(T, dma.*[2].dmacntH()) << 16, // DMA2CNT_L is write-only
+            0xD4, 0xD8 => null, // DMA3SAD, DMA3DAD
+            0xDC => @as(T, dma.*[3].dmacntH()) << 16, // DMA3CNT_L is write-only
+            else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
         },
-        u16 => switch (byte) {
-            0xBA => dma.*[0].cnt.raw,
-            0xC6 => dma.*[1].cnt.raw,
-            0xD2 => dma.*[2].cnt.raw,
-            0xDE => dma.*[3].cnt.raw,
-            else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
+        u16 => switch (byte_addr) {
+            0xB0, 0xB2, 0xB4, 0xB6 => null, // DMA0SAD, DMA0DAD
+            0xB8 => 0x0000, // DMA0CNT_L, suite.gba expects 0x0000 instead of 0xDEAD
+            0xBA => dma.*[0].dmacntH(),
+
+            0xBC, 0xBE, 0xC0, 0xC2 => null, // DMA1SAD, DMA1DAD
+            0xC4 => 0x0000, // DMA1CNT_L
+            0xC6 => dma.*[1].dmacntH(),
+
+            0xC8, 0xCA, 0xCC, 0xCE => null, // DMA2SAD, DMA2DAD
+            0xD0 => 0x0000, // DMA2CNT_L
+            0xD2 => dma.*[2].dmacntH(),
+
+            0xD4, 0xD6, 0xD8, 0xDA => null, // DMA3SAD, DMA3DAD
+            0xDC => 0x0000, // DMA3CNT_L
+            0xDE => dma.*[3].dmacntH(),
+            else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
+        },
+        u8 => switch (byte_addr) {
+            0xB0...0xB7 => null, // DMA0SAD, DMA0DAD
+            0xB8, 0xB9 => 0x00, // DMA0CNT_L
+            0xBA, 0xBB => @truncate(dma.*[0].dmacntH() >> getHalf(byte_addr)),
+
+            0xBC...0xC3 => null, // DMA1SAD, DMA1DAD
+            0xC4, 0xC5 => 0x00, // DMA1CNT_L
+            0xC6, 0xC7 => @truncate(dma.*[1].dmacntH() >> getHalf(byte_addr)),
+
+            0xC8...0xCF => null, // DMA2SAD, DMA2DAD
+            0xD0, 0xD1 => 0x00, // DMA2CNT_L
+            0xD2, 0xD3 => @truncate(dma.*[2].dmacntH() >> getHalf(byte_addr)),
+
+            0xD4...0xDB => null, // DMA3SAD, DMA3DAD
+            0xDC, 0xDD => 0x00, // DMA3CNT_L
+            0xDE, 0xDF => @truncate(dma.*[3].dmacntH() >> getHalf(byte_addr)),
+            else => util.io.read.err(T, log, "unexpected {} read from 0x{X:0>8}", .{ T, addr }),
         },
-        u8 => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
         else => @compileError("DMA: Unsupported read width"),
     };
 }
 
 pub fn write(comptime T: type, dma: *DmaTuple, addr: u32, value: T) void {
-    const byte = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     switch (T) {
-        u32 => switch (byte) {
+        u32 => switch (byte_addr) {
             0xB0 => dma.*[0].setDmasad(value),
             0xB4 => dma.*[0].setDmadad(value),
             0xB8 => dma.*[0].setDmacnt(value),
+
             0xBC => dma.*[1].setDmasad(value),
             0xC0 => dma.*[1].setDmadad(value),
             0xC4 => dma.*[1].setDmacnt(value),
+
             0xC8 => dma.*[2].setDmasad(value),
             0xCC => dma.*[2].setDmadad(value),
             0xD0 => dma.*[2].setDmacnt(value),
+
             0xD4 => dma.*[3].setDmasad(value),
             0xD8 => dma.*[3].setDmadad(value),
             0xDC => dma.*[3].setDmacnt(value),
             else => util.io.write.undef(log, "Tried to write 0x{X:0>8}{} to 0x{X:0>8}", .{ value, T, addr }),
         },
-        u16 => switch (byte) {
-            0xB0 => dma.*[0].setDmasad(setLo(u32, dma.*[0].sad, value)),
-            0xB2 => dma.*[0].setDmasad(setHi(u32, dma.*[0].sad, value)),
-            0xB4 => dma.*[0].setDmadad(setLo(u32, dma.*[0].dad, value)),
-            0xB6 => dma.*[0].setDmadad(setHi(u32, dma.*[0].dad, value)),
+        u16 => switch (byte_addr) {
+            0xB0, 0xB2 => dma.*[0].setDmasad(setHalf(u32, dma.*[0].sad, byte_addr, value)),
+            0xB4, 0xB6 => dma.*[0].setDmadad(setHalf(u32, dma.*[0].dad, byte_addr, value)),
             0xB8 => dma.*[0].setDmacntL(value),
             0xBA => dma.*[0].setDmacntH(value),
 
-            0xBC => dma.*[1].setDmasad(setLo(u32, dma.*[1].sad, value)),
-            0xBE => dma.*[1].setDmasad(setHi(u32, dma.*[1].sad, value)),
-            0xC0 => dma.*[1].setDmadad(setLo(u32, dma.*[1].dad, value)),
-            0xC2 => dma.*[1].setDmadad(setHi(u32, dma.*[1].dad, value)),
+            0xBC, 0xBE => dma.*[1].setDmasad(setHalf(u32, dma.*[1].sad, byte_addr, value)),
+            0xC0, 0xC2 => dma.*[1].setDmadad(setHalf(u32, dma.*[1].dad, byte_addr, value)),
             0xC4 => dma.*[1].setDmacntL(value),
             0xC6 => dma.*[1].setDmacntH(value),
 
-            0xC8 => dma.*[2].setDmasad(setLo(u32, dma.*[2].sad, value)),
-            0xCA => dma.*[2].setDmasad(setHi(u32, dma.*[2].sad, value)),
-            0xCC => dma.*[2].setDmadad(setLo(u32, dma.*[2].dad, value)),
-            0xCE => dma.*[2].setDmadad(setHi(u32, dma.*[2].dad, value)),
+            0xC8, 0xCA => dma.*[2].setDmasad(setHalf(u32, dma.*[2].sad, byte_addr, value)),
+            0xCC, 0xCE => dma.*[2].setDmadad(setHalf(u32, dma.*[2].dad, byte_addr, value)),
             0xD0 => dma.*[2].setDmacntL(value),
             0xD2 => dma.*[2].setDmacntH(value),
 
-            0xD4 => dma.*[3].setDmasad(setLo(u32, dma.*[3].sad, value)),
-            0xD6 => dma.*[3].setDmasad(setHi(u32, dma.*[3].sad, value)),
-            0xD8 => dma.*[3].setDmadad(setLo(u32, dma.*[3].dad, value)),
-            0xDA => dma.*[3].setDmadad(setHi(u32, dma.*[3].dad, value)),
+            0xD4, 0xD6 => dma.*[3].setDmasad(setHalf(u32, dma.*[3].sad, byte_addr, value)),
+            0xD8, 0xDA => dma.*[3].setDmadad(setHalf(u32, dma.*[3].dad, byte_addr, value)),
             0xDC => dma.*[3].setDmacntL(value),
             0xDE => dma.*[3].setDmacntH(value),
             else => util.io.write.undef(log, "Tried to write 0x{X:0>4}{} to 0x{X:0>8}", .{ value, T, addr }),
         },
-        u8 => util.io.write.undef(log, "Tried to write 0x{X:0>2}{} to 0x{X:0>8}", .{ value, T, addr }),
+        u8 => switch (byte_addr) {
+            0xB0, 0xB1, 0xB2, 0xB3 => dma.*[0].setDmasad(setQuart(dma.*[0].sad, byte_addr, value)),
+            0xB4, 0xB5, 0xB6, 0xB7 => dma.*[0].setDmadad(setQuart(dma.*[0].dad, byte_addr, value)),
+            0xB8, 0xB9 => dma.*[0].setDmacntL(setHalf(u16, dma.*[0].word_count, byte_addr, value)),
+            0xBA, 0xBB => dma.*[0].setDmacntH(setHalf(u16, dma.*[0].cnt.raw, byte_addr, value)),
+
+            0xBC, 0xBD, 0xBE, 0xBF => dma.*[1].setDmasad(setQuart(dma.*[1].sad, byte_addr, value)),
+            0xC0, 0xC1, 0xC2, 0xC3 => dma.*[1].setDmadad(setQuart(dma.*[1].dad, byte_addr, value)),
+            0xC4, 0xC5 => dma.*[1].setDmacntL(setHalf(u16, dma.*[1].word_count, byte_addr, value)),
+            0xC6, 0xC7 => dma.*[1].setDmacntH(setHalf(u16, dma.*[1].cnt.raw, byte_addr, value)),
+
+            0xC8, 0xC9, 0xCA, 0xCB => dma.*[2].setDmasad(setQuart(dma.*[2].sad, byte_addr, value)),
+            0xCC, 0xCD, 0xCE, 0xCF => dma.*[2].setDmadad(setQuart(dma.*[2].dad, byte_addr, value)),
+            0xD0, 0xD1 => dma.*[2].setDmacntL(setHalf(u16, dma.*[2].word_count, byte_addr, value)),
+            0xD2, 0xD3 => dma.*[2].setDmacntH(setHalf(u16, dma.*[2].cnt.raw, byte_addr, value)),
+
+            0xD4, 0xD5, 0xD6, 0xD7 => dma.*[3].setDmasad(setQuart(dma.*[3].sad, byte_addr, value)),
+            0xD8, 0xD9, 0xDA, 0xDB => dma.*[3].setDmadad(setQuart(dma.*[3].dad, byte_addr, value)),
+            0xDC, 0xDD => dma.*[3].setDmacntL(setHalf(u16, dma.*[3].word_count, byte_addr, value)),
+            0xDE, 0xDF => dma.*[3].setDmacntH(setHalf(u16, dma.*[3].cnt.raw, byte_addr, value)),
+            else => util.io.write.undef(log, "Tried to write 0x{X:0>2}{} to 0x{X:0>8}", .{ value, T, addr }),
+        },
         else => @compileError("DMA: Unsupported write width"),
     }
 }
@@ -99,6 +151,7 @@ fn DmaController(comptime id: u2) type {
 
         const sad_mask: u32 = if (id == 0) 0x07FF_FFFF else 0x0FFF_FFFF;
         const dad_mask: u32 = if (id != 3) 0x07FF_FFFF else 0x0FFF_FFFF;
+        const WordCount = if (id == 3) u16 else u14;
 
         /// Write-only. The first address in a DMA transfer. (DMASAD)
         /// Note: use writeSrc instead of manipulating src_addr directly
@@ -107,17 +160,19 @@ fn DmaController(comptime id: u2) type {
         /// Note: Use writeDst instead of manipulatig dst_addr directly
         dad: u32,
         /// Write-only. The Word Count for the DMA Transfer (DMACNT_L)
-        word_count: if (id == 3) u16 else u14,
+        word_count: WordCount,
         /// Read / Write. DMACNT_H
         /// Note: Use writeControl instead of manipulating cnt directly.
         cnt: DmaControl,
 
+        /// Internal. The last successfully read value
+        data_latch: u32,
         /// Internal. Currrent Source Address
         sad_latch: u32,
         /// Internal. Current Destination Address
         dad_latch: u32,
         /// Internal. Word Count
-        _word_count: if (id == 3) u16 else u14,
+        _word_count: WordCount,
 
         /// Some DMA Transfers are enabled during Hblank / VBlank and / or
         /// have delays. Thefore bit 15 of DMACNT isn't actually something
@@ -134,11 +189,17 @@ fn DmaController(comptime id: u2) type {
                 // Internals
                 .sad_latch = 0,
                 .dad_latch = 0,
+                .data_latch = 0,
+
                 ._word_count = 0,
                 .in_progress = false,
             };
         }
 
+        pub fn reset(self: *Self) void {
+            self.* = Self.init();
+        }
+
         pub fn setDmasad(self: *Self, addr: u32) void {
             self.sad = addr & sad_mask;
         }
@@ -148,7 +209,11 @@ fn DmaController(comptime id: u2) type {
         }
 
         pub fn setDmacntL(self: *Self, halfword: u16) void {
-            self.word_count = @truncate(@TypeOf(self.word_count), halfword);
+            self.word_count = @truncate(halfword);
+        }
+
+        pub fn dmacntH(self: *const Self) u16 {
+            return self.cnt.raw & if (id == 3) 0xFFE0 else 0xF7E0;
         }
 
         pub fn setDmacntH(self: *Self, halfword: u16) void {
@@ -158,7 +223,7 @@ fn DmaController(comptime id: u2) type {
                 // Reload Internals on Rising Edge.
                 self.sad_latch = self.sad;
                 self.dad_latch = self.dad;
-                self._word_count = if (self.word_count == 0) std.math.maxInt(@TypeOf(self._word_count)) else self.word_count;
+                self._word_count = if (self.word_count == 0) std.math.maxInt(WordCount) else self.word_count;
 
                 // Only a Start Timing of 00 has a DMA Transfer immediately begin
                 self.in_progress = new.start_timing.read() == 0b00;
@@ -168,32 +233,46 @@ fn DmaController(comptime id: u2) type {
         }
 
         pub fn setDmacnt(self: *Self, word: u32) void {
-            self.setDmacntL(@truncate(u16, word));
-            self.setDmacntH(@truncate(u16, word >> 16));
+            self.setDmacntL(@truncate(word));
+            self.setDmacntH(@truncate(word >> 16));
         }
 
         pub fn step(self: *Self, cpu: *Arm7tdmi) void {
+            const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
             const is_fifo = (id == 1 or id == 2) and self.cnt.start_timing.read() == 0b11;
-            const sad_adj = @intToEnum(Adjustment, self.cnt.sad_adj.read());
-            const dad_adj = if (is_fifo) .Fixed else @intToEnum(Adjustment, self.cnt.dad_adj.read());
+            const sad_adj: Adjustment = @enumFromInt(self.cnt.sad_adj.read());
+            const dad_adj: Adjustment = if (is_fifo) .Fixed else @enumFromInt(self.cnt.dad_adj.read());
 
             const transfer_type = is_fifo or self.cnt.transfer_type.read();
             const offset: u32 = if (transfer_type) @sizeOf(u32) else @sizeOf(u16);
 
             const mask = if (transfer_type) ~@as(u32, 3) else ~@as(u32, 1);
+            const sad_addr = self.sad_latch & mask;
+            const dad_addr = self.dad_latch & mask;
 
             if (transfer_type) {
-                cpu.bus.write(u32, self.dad_latch & mask, cpu.bus.read(u32, self.sad_latch & mask));
+                if (sad_addr >= 0x0200_0000) self.data_latch = cpu.bus.read(u32, sad_addr);
+                cpu.bus.write(u32, dad_addr, self.data_latch);
             } else {
-                cpu.bus.write(u16, self.dad_latch & mask, cpu.bus.read(u16, self.sad_latch & mask));
+                if (sad_addr >= 0x0200_0000) {
+                    const value: u32 = cpu.bus.read(u16, sad_addr);
+                    self.data_latch = value << 16 | value;
+                }
+
+                cpu.bus.write(u16, dad_addr, @as(u16, @truncate(rotr(u32, self.data_latch, 8 * (dad_addr & 3)))));
             }
 
-            switch (sad_adj) {
-                .Increment => self.sad_latch +%= offset,
-                .Decrement => self.sad_latch -%= offset,
-                // FIXME: Is just ignoring this ok?
-                .IncrementReload => log.err("{} is a prohibited adjustment on SAD", .{sad_adj}),
-                .Fixed => {},
+            switch (@as(u8, @truncate(sad_addr >> 24))) {
+                // according to fleroviux, DMAs with a source address in ROM misbehave
+                // the resultant behaviour is that the source address will increment despite what DMAXCNT says
+                0x08...0x0D => self.sad_latch +%= offset, // obscure behaviour
+                else => switch (sad_adj) {
+                    .Increment => self.sad_latch +%= offset,
+                    .Decrement => self.sad_latch -%= offset,
+                    .IncrementReload => log.err("{} is a prohibited adjustment on SAD", .{sad_adj}),
+                    .Fixed => {},
+                },
             }
 
             switch (dad_adj) {
@@ -207,13 +286,13 @@ fn DmaController(comptime id: u2) type {
             if (self._word_count == 0) {
                 if (self.cnt.irq.read()) {
                     switch (id) {
-                        0 => cpu.bus.io.irq.dma0.set(),
-                        1 => cpu.bus.io.irq.dma1.set(),
-                        2 => cpu.bus.io.irq.dma2.set(),
-                        3 => cpu.bus.io.irq.dma3.set(),
+                        0 => bus_ptr.io.irq.dma0.set(),
+                        1 => bus_ptr.io.irq.dma1.set(),
+                        2 => bus_ptr.io.irq.dma2.set(),
+                        3 => bus_ptr.io.irq.dma3.set(),
                     }
 
-                    cpu.handleInterrupt();
+                    handleInterrupt(cpu);
                 }
 
                 // If we're not repeating, Fire the IRQs and disable the DMA
@@ -242,7 +321,7 @@ fn DmaController(comptime id: u2) type {
             // Reload internal DAD latch if we are in IncrementRelaod
             if (self.in_progress) {
                 self._word_count = if (self.word_count == 0) std.math.maxInt(@TypeOf(self._word_count)) else self.word_count;
-                if (@intToEnum(Adjustment, self.cnt.dad_adj.read()) == .IncrementReload) self.dad_latch = self.dad;
+                if (@as(Adjustment, @enumFromInt(self.cnt.dad_adj.read())) == .IncrementReload) self.dad_latch = self.dad;
             }
         }
 
@@ -266,11 +345,8 @@ fn DmaController(comptime id: u2) type {
     };
 }
 
-pub fn pollDmaOnBlank(bus: *Bus, comptime kind: DmaKind) void {
-    bus.dma[0].poll(kind);
-    bus.dma[1].poll(kind);
-    bus.dma[2].poll(kind);
-    bus.dma[3].poll(kind);
+pub fn onBlanking(bus: *Bus, comptime kind: DmaKind) void {
+    inline for (0..4) |i| bus.dma[i].poll(kind);
 }
 
 const Adjustment = enum(u2) {

src/core/bus/gpio.zig

@@ -1,11 +1,14 @@
 const std = @import("std");
 const Bit = @import("bitfield").Bit;
-const Bitfield = @import("bitfield").Bitfield;
 const DateTime = @import("datetime").datetime.Datetime;
 
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("../Bus.zig");
+const Scheduler = @import("../scheduler.zig").Scheduler;
 const Allocator = std.mem.Allocator;
 
+const handleInterrupt = @import("../cpu_util.zig").handleInterrupt;
+
 /// GPIO Register Implementation
 pub const Gpio = struct {
     const Self = @This();
@@ -28,8 +31,8 @@ pub const Gpio = struct {
         fn step(self: *Device, value: u4) u4 {
             return switch (self.kind) {
                 .Rtc => blk: {
-                    const clock = @ptrCast(*Clock, @alignCast(@alignOf(*Clock), self.ptr.?));
-                    break :blk clock.step(Clock.Data{ .raw = value });
+                    const clock: *Clock = @ptrCast(@alignCast(self.ptr.?));
+                    break :blk clock.step(.{ .raw = value });
                 },
                 .None => value,
             };
@@ -68,9 +71,11 @@ pub const Gpio = struct {
         log.info("Device: {}", .{kind});
 
         const self = try allocator.create(Self);
+        errdefer allocator.destroy(self);
+
         self.* = .{
             .data = 0b0000,
-            .direction = 0b1111, // TODO: What is GPIO DIrection set to by default?
+            .direction = 0b1111, // TODO: What is GPIO Direction set to by default?
             .cnt = 0b0,
 
             .device = switch (kind) {
@@ -89,7 +94,7 @@ pub const Gpio = struct {
 
     pub fn deinit(self: *Self, allocator: Allocator) void {
         switch (self.device.kind) {
-            .Rtc => allocator.destroy(@ptrCast(*Clock, @alignCast(@alignOf(*Clock), self.device.ptr.?))),
+            .Rtc => allocator.destroy(@as(*Clock, @ptrCast(@alignCast(self.device.ptr.?)))),
             .None => {},
         }
 
@@ -141,16 +146,16 @@ pub const Clock = struct {
         /// 2. A `count`, which keeps track of which byte is currently being read
         /// 3. An index, which keeps track of which bit of the byte determined by `count` is being read
         fn read(self: *Reader, clock: *const Clock, register: Register) u1 {
-            const idx = @intCast(u3, self.i);
+            const idx: u3 = @intCast(self.i);
             defer self.i += 1;
 
             // FIXME: What do I do about the unused bits?
             return switch (register) {
-                .Control => @truncate(u1, switch (self.count) {
+                .Control => @truncate(switch (self.count) {
                     0 => clock.cnt.raw >> idx,
                     else => std.debug.panic("Tried to read from byte #{} of {} (hint: there's only 1 byte)", .{ self.count, register }),
                 }),
-                .DateTime => @truncate(u1, switch (self.count) {
+                .DateTime => @truncate(switch (self.count) {
                     // Date
                     0 => clock.year >> idx,
                     1 => @as(u8, clock.month) >> idx,
@@ -163,7 +168,7 @@ pub const Clock = struct {
                     6 => @as(u8, clock.second) >> idx,
                     else => std.debug.panic("Tried to read from byte #{} of {} (hint: there's only 7 bytes)", .{ self.count, register }),
                 }),
-                .Time => @truncate(u1, switch (self.count) {
+                .Time => @truncate(switch (self.count) {
                     0 => @as(u8, clock.hour) >> idx,
                     1 => @as(u8, clock.minute) >> idx,
                     2 => @as(u8, clock.second) >> idx,
@@ -202,7 +207,7 @@ pub const Clock = struct {
 
         /// Append a bit to the internal bit buffer (aka an integer)
         fn push(self: *Writer, value: u1) void {
-            const idx = @intCast(u3, self.i);
+            const idx: u3 = @intCast(self.i);
             self.buf = (self.buf & ~(@as(u8, 1) << idx)) | @as(u8, value) << idx;
             self.i += 1;
         }
@@ -285,20 +290,22 @@ pub const Clock = struct {
             .gpio = gpio, // Can't use Arm7tdmi ptr b/c not initialized yet
         };
 
-        cpu.sched.push(.RealTimeClock, 1 << 24); // Every Second
+        const sched_ptr: *Scheduler = @ptrCast(@alignCast(cpu.sched.ptr));
+        sched_ptr.push(.RealTimeClock, 1 << 24); // Every Second
     }
 
     pub fn onClockUpdate(self: *Self, late: u64) void {
-        self.cpu.sched.push(.RealTimeClock, (1 << 24) -| late); // Reschedule
+        const sched_ptr: *Scheduler = @ptrCast(@alignCast(self.cpu.sched.ptr));
+        sched_ptr.push(.RealTimeClock, (1 << 24) -| late); // Reschedule
 
         const now = DateTime.now();
-        self.year = bcd(u8, @intCast(u8, now.date.year - 2000));
-        self.month = bcd(u5, now.date.month);
-        self.day = bcd(u6, now.date.day);
-        self.weekday = bcd(u3, (now.date.weekday() + 1) % 7); // API is Monday = 0, Sunday = 6. We want Sunday = 0, Saturday = 6
-        self.hour = bcd(u6, now.time.hour);
-        self.minute = bcd(u7, now.time.minute);
-        self.second = bcd(u7, now.time.second);
+        self.year = bcd(@intCast(now.date.year - 2000));
+        self.month = @truncate(bcd(now.date.month));
+        self.day = @truncate(bcd(now.date.day));
+        self.weekday = @truncate(bcd((now.date.weekday() + 1) % 7)); // API is Monday = 0, Sunday = 6. We want Sunday = 0, Saturday = 6
+        self.hour = @truncate(bcd(now.time.hour));
+        self.minute = @truncate(bcd(now.time.minute));
+        self.second = @truncate(bcd(now.time.second));
     }
 
     fn step(self: *Self, value: Data) u4 {
@@ -314,14 +321,14 @@ pub const Clock = struct {
                     }
                 }
 
-                break :blk @truncate(u4, value.raw);
+                break :blk @truncate(value.raw);
             },
             .Command => blk: {
                 if (!value.cs.read()) log.err("Expected CS to be set during {}, however CS was cleared", .{self.state});
 
                 // If SCK rises, sample SIO
                 if (!cache.sck.read() and value.sck.read()) {
-                    self.writer.push(@boolToInt(value.sio.read()));
+                    self.writer.push(@intFromBool(value.sio.read()));
 
                     if (self.writer.finished()) {
                         self.state = self.processCommand(self.writer.buf);
@@ -331,14 +338,14 @@ pub const Clock = struct {
                     }
                 }
 
-                break :blk @truncate(u4, value.raw);
+                break :blk @truncate(value.raw);
             },
             .Write => |register| blk: {
                 if (!value.cs.read()) log.err("Expected CS to be set during {}, however CS was cleared", .{self.state});
 
                 // If SCK rises, sample SIO
                 if (!cache.sck.read() and value.sck.read()) {
-                    self.writer.push(@boolToInt(value.sio.read()));
+                    self.writer.push(@intFromBool(value.sio.read()));
 
                     const register_width: u32 = switch (register) {
                         .Control => 1,
@@ -357,7 +364,7 @@ pub const Clock = struct {
                     }
                 }
 
-                break :blk @truncate(u4, value.raw);
+                break :blk @truncate(value.raw);
             },
             .Read => |register| blk: {
                 if (!value.cs.read()) log.err("Expected CS to be set during {}, however CS was cleared", .{self.state});
@@ -383,7 +390,7 @@ pub const Clock = struct {
                     }
                 }
 
-                break :blk @truncate(u4, ret.raw);
+                break :blk @truncate(ret.raw);
             },
         };
     }
@@ -396,11 +403,13 @@ pub const Clock = struct {
     }
 
     fn irq(self: *Self) void {
+        const bus_ptr: *Bus = @ptrCast(@alignCast(self.cpu.bus.ptr));
+
         // TODO: Confirm that this is the right behaviour
         log.debug("Force GamePak IRQ", .{});
 
-        self.cpu.bus.io.irq.game_pak.set();
-        self.cpu.handleInterrupt();
+        bus_ptr.io.irq.game_pak.set();
+        handleInterrupt(self.cpu);
     }
 
     fn processCommand(self: *Self, raw_command: u8) State {
@@ -420,7 +429,7 @@ pub const Clock = struct {
         log.debug("Handling Command 0x{X:0>2} [0b{b:0>8}]", .{ command, command });
 
         const is_write = command & 1 == 0;
-        const rtc_register = @truncate(u3, command >> 1 & 0x7);
+        const rtc_register: u3 = @truncate(command >> 1 & 0x7);
 
         if (is_write) {
             return switch (rtc_register) {
@@ -448,16 +457,8 @@ pub const Clock = struct {
     }
 };
 
-fn bcd(comptime T: type, value: u8) T {
-    var input = value;
-    var ret: u8 = 0;
-    var shift: u3 = 0;
-
-    while (input > 0) {
-        ret |= (input % 10) << (shift << 2);
-        shift += 1;
-        input /= 10;
-    }
-
-    return @truncate(T, ret);
+/// Converts an 8-bit unsigned integer to its BCD representation.
+/// Note: Algorithm only works for values between 0 and 99 inclusive.
+fn bcd(value: u8) u8 {
+    return ((value / 10) << 4) + (value % 10);
 }

src/core/bus/io.zig

@@ -1,18 +1,16 @@
 const std = @import("std");
-const builtin = @import("builtin");
 const timer = @import("timer.zig");
 const dma = @import("dma.zig");
 const apu = @import("../apu.zig");
+const ppu = @import("../ppu.zig");
 const util = @import("../../util.zig");
 
 const Bit = @import("bitfield").Bit;
 const Bitfield = @import("bitfield").Bitfield;
 const Bus = @import("../Bus.zig");
-const DmaController = @import("dma.zig").DmaController;
-const Scheduler = @import("../scheduler.zig").Scheduler;
 
-const setHi = util.setLo;
-const setLo = util.setHi;
+const getHalf = util.getHalf;
+const setHalf = util.setHalf;
 
 const log = std.log.scoped(.@"I/O");
 
@@ -24,23 +22,29 @@ pub const Io = struct {
     ie: InterruptEnable,
     irq: InterruptRequest,
     postflg: PostFlag,
+    waitcnt: WaitControl,
     haltcnt: HaltControl,
-    keyinput: KeyInput,
+    keyinput: AtomicKeyInput,
 
     pub fn init() Self {
         return .{
             .ime = false,
             .ie = .{ .raw = 0x0000 },
             .irq = .{ .raw = 0x0000 },
-            .keyinput = .{ .raw = 0x03FF },
+            .keyinput = AtomicKeyInput.init(.{ .raw = 0x03FF }),
+            .waitcnt = .{ .raw = 0x0000_0000 }, // Bit 15 == 0 for GBA
             .postflg = .FirstBoot,
             .haltcnt = .Execute,
         };
     }
 
+    pub fn reset(self: *Self) void {
+        self.* = Self.init();
+    }
+
     fn setIrqs(self: *Io, word: u32) void {
-        self.ie.raw = @truncate(u16, word);
-        self.irq.raw &= ~@truncate(u16, word >> 16);
+        self.ie.raw = @truncate(word);
+        self.irq.raw &= ~@as(u16, @truncate(word >> 16));
     }
 };
 
@@ -48,9 +52,10 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
     return switch (T) {
         u32 => switch (address) {
             // Display
-            0x0400_0000 => bus.ppu.dispcnt.raw,
-            0x0400_0004 => @as(T, bus.ppu.vcount.raw) << 16 | bus.ppu.dispstat.raw,
-            0x0400_0006 => @as(T, bus.ppu.bg[0].cnt.raw) << 16 | bus.ppu.vcount.raw,
+            0x0400_0000...0x0400_0054 => ppu.read(T, &bus.ppu, address),
+
+            // Sound
+            0x0400_0060...0x0400_00A4 => apu.read(T, &bus.apu, address),
 
             // DMA Transfers
             0x0400_00B0...0x0400_00DC => dma.read(T, &bus.dma, address),
@@ -68,26 +73,18 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
             0x0400_0150 => util.io.read.todo(log, "Read {} from JOY_RECV", .{T}),
 
             // Interrupts
-            0x0400_0200 => @as(T, bus.io.irq.raw) << 16 | bus.io.ie.raw,
-            0x0400_0208 => @boolToInt(bus.io.ime),
+            0x0400_0200 => @as(u32, bus.io.irq.raw) << 16 | bus.io.ie.raw,
+            0x0400_0204 => bus.io.waitcnt.raw,
+            0x0400_0208 => @intFromBool(bus.io.ime),
+            0x0400_0300 => @intFromEnum(bus.io.postflg),
             else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, address }),
         },
         u16 => switch (address) {
             // Display
-            0x0400_0000 => bus.ppu.dispcnt.raw,
-            0x0400_0004 => bus.ppu.dispstat.raw,
-            0x0400_0006 => bus.ppu.vcount.raw,
-            0x0400_0008 => bus.ppu.bg[0].cnt.raw,
-            0x0400_000A => bus.ppu.bg[1].cnt.raw,
-            0x0400_000C => bus.ppu.bg[2].cnt.raw,
-            0x0400_000E => bus.ppu.bg[3].cnt.raw,
-            0x0400_004C => util.io.read.todo(log, "Read {} from MOSAIC", .{T}),
-            0x0400_0050 => bus.ppu.bldcnt.raw,
-            0x0400_0052 => bus.ppu.bldalpha.raw,
-            0x0400_0054 => bus.ppu.bldy.raw,
+            0x0400_0000...0x0400_0054 => ppu.read(T, &bus.ppu, address),
 
             // Sound
-            0x0400_0060...0x0400_009E => apu.read(T, &bus.apu, address),
+            0x0400_0060...0x0400_00A6 => apu.read(T, &bus.apu, address),
 
             // DMA Transfers
             0x0400_00B0...0x0400_00DE => dma.read(T, &bus.dma, address),
@@ -99,32 +96,38 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
             0x0400_0128 => util.io.read.todo(log, "Read {} from SIOCNT", .{T}),
 
             // Keypad Input
-            0x0400_0130 => bus.io.keyinput.raw,
+            0x0400_0130 => bus.io.keyinput.load(.monotonic),
 
             // Serial Communication 2
             0x0400_0134 => util.io.read.todo(log, "Read {} from RCNT", .{T}),
+            0x0400_0136 => 0x0000,
+            0x0400_0142 => 0x0000,
+            0x0400_015A => 0x0000,
 
             // Interrupts
             0x0400_0200 => bus.io.ie.raw,
             0x0400_0202 => bus.io.irq.raw,
-            0x0400_0204 => util.io.read.todo(log, "Read {} from WAITCNT", .{T}),
-            0x0400_0208 => @boolToInt(bus.io.ime),
+            0x0400_0204 => bus.io.waitcnt.raw,
+            0x0400_0206 => 0x0000,
+            0x0400_0208 => @intFromBool(bus.io.ime),
+            0x0400_020A => 0x0000,
+            0x0400_0300 => @intFromEnum(bus.io.postflg),
+            0x0400_0302 => 0x0000,
             else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, address }),
         },
         u8 => return switch (address) {
             // Display
-            0x0400_0000 => @truncate(T, bus.ppu.dispcnt.raw),
-            0x0400_0004 => @truncate(T, bus.ppu.dispstat.raw),
-            0x0400_0005 => @truncate(T, bus.ppu.dispcnt.raw >> 8),
-            0x0400_0006 => @truncate(T, bus.ppu.vcount.raw),
-            0x0400_0008 => @truncate(T, bus.ppu.bg[0].cnt.raw),
-            0x0400_0009 => @truncate(T, bus.ppu.bg[0].cnt.raw >> 8),
-            0x0400_000A => @truncate(T, bus.ppu.bg[1].cnt.raw),
-            0x0400_000B => @truncate(T, bus.ppu.bg[1].cnt.raw >> 8),
+            0x0400_0000...0x0400_0055 => ppu.read(T, &bus.ppu, address),
 
             // Sound
             0x0400_0060...0x0400_00A7 => apu.read(T, &bus.apu, address),
 
+            // DMA Transfers
+            0x0400_00B0...0x0400_00DF => dma.read(T, &bus.dma, address),
+
+            // Timers
+            0x0400_0100...0x0400_010F => timer.read(T, &bus.tim, address),
+
             // Serial Communication 1
             0x0400_0128 => util.io.read.todo(log, "Read {} from SIOCNT_L", .{T}),
 
@@ -133,10 +136,20 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
 
             // Serial Communication 2
             0x0400_0135 => util.io.read.todo(log, "Read {} from RCNT_H", .{T}),
+            0x0400_0136, 0x0400_0137 => 0x00,
+            0x0400_0142, 0x0400_0143 => 0x00,
+            0x0400_015A, 0x0400_015B => 0x00,
 
             // Interrupts
-            0x0400_0200 => @truncate(T, bus.io.ie.raw),
-            0x0400_0300 => @enumToInt(bus.io.postflg),
+            0x0400_0200, 0x0400_0201 => @truncate(bus.io.ie.raw >> getHalf(@truncate(address))),
+            0x0400_0202, 0x0400_0203 => @truncate(bus.io.irq.raw >> getHalf(@truncate(address))),
+            0x0400_0204, 0x0400_0205 => @truncate(bus.io.waitcnt.raw >> getHalf(@truncate(address))),
+            0x0400_0206, 0x0400_0207 => 0x00,
+            0x0400_0208, 0x0400_0209 => @truncate(@as(u16, @intFromBool(bus.io.ime)) >> getHalf(@truncate(address))),
+            0x0400_020A, 0x0400_020B => 0x00,
+            0x0400_0300 => @intFromEnum(bus.io.postflg),
+            0x0400_0301 => null,
+            0x0400_0302, 0x0400_0303 => 0x00,
             else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, address }),
         },
         else => @compileError("I/O: Unsupported read width"),
@@ -147,34 +160,7 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
     return switch (T) {
         u32 => switch (address) {
             // Display
-            0x0400_0000 => bus.ppu.dispcnt.raw = @truncate(u16, value),
-            0x0400_0004 => {
-                bus.ppu.dispstat.raw = @truncate(u16, value);
-                bus.ppu.vcount.raw = @truncate(u16, value >> 16);
-            },
-            0x0400_0008 => bus.ppu.setAdjCnts(0, value),
-            0x0400_000C => bus.ppu.setAdjCnts(2, value),
-            0x0400_0010 => bus.ppu.setBgOffsets(0, value),
-            0x0400_0014 => bus.ppu.setBgOffsets(1, value),
-            0x0400_0018 => bus.ppu.setBgOffsets(2, value),
-            0x0400_001C => bus.ppu.setBgOffsets(3, value),
-            0x0400_0020 => bus.ppu.aff_bg[0].writePaPb(value),
-            0x0400_0024 => bus.ppu.aff_bg[0].writePcPd(value),
-            0x0400_0028 => bus.ppu.aff_bg[0].setX(bus.ppu.dispstat.vblank.read(), value),
-            0x0400_002C => bus.ppu.aff_bg[0].setY(bus.ppu.dispstat.vblank.read(), value),
-            0x0400_0030 => bus.ppu.aff_bg[1].writePaPb(value),
-            0x0400_0034 => bus.ppu.aff_bg[1].writePcPd(value),
-            0x0400_0038 => bus.ppu.aff_bg[1].setX(bus.ppu.dispstat.vblank.read(), value),
-            0x0400_003C => bus.ppu.aff_bg[1].setY(bus.ppu.dispstat.vblank.read(), value),
-            0x0400_0040 => bus.ppu.win.setH(value),
-            0x0400_0044 => bus.ppu.win.setV(value),
-            0x0400_0048 => bus.ppu.win.setIo(value),
-            0x0400_004C => log.debug("Wrote 0x{X:0>8} to MOSAIC", .{value}),
-            0x0400_0050 => {
-                bus.ppu.bldcnt.raw = @truncate(u16, value);
-                bus.ppu.bldalpha.raw = @truncate(u16, value >> 16);
-            },
-            0x0400_0054 => bus.ppu.bldy.raw = @truncate(u16, value),
+            0x0400_0000...0x0400_0054 => ppu.write(T, &bus.ppu, address, value),
             0x0400_0058...0x0400_005C => {}, // Unused
 
             // Sound
@@ -210,65 +196,28 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
 
             // Interrupts
             0x0400_0200 => bus.io.setIrqs(value),
-            0x0400_0204 => log.debug("Wrote 0x{X:0>8} to WAITCNT", .{value}),
+            0x0400_0204 => bus.io.waitcnt.set(@truncate(value)),
             0x0400_0208 => bus.io.ime = value & 1 == 1,
-            0x0400_020C...0x0400_021C => {}, // Unused
+            0x0400_0300 => {
+                bus.io.postflg = @enumFromInt(value & 1);
+                bus.io.haltcnt = if (value >> 15 & 1 == 0) .Halt else @panic("TODO: Implement STOP");
+            },
             else => util.io.write.undef(log, "Tried to write 0x{X:0>8}{} to 0x{X:0>8}", .{ value, T, address }),
         },
         u16 => switch (address) {
             // Display
-            0x0400_0000 => bus.ppu.dispcnt.raw = value,
-            0x0400_0004 => bus.ppu.dispstat.raw = value,
-            0x0400_0006 => {}, // vcount is read-only
-            0x0400_0008 => bus.ppu.bg[0].cnt.raw = value,
-            0x0400_000A => bus.ppu.bg[1].cnt.raw = value,
-            0x0400_000C => bus.ppu.bg[2].cnt.raw = value,
-            0x0400_000E => bus.ppu.bg[3].cnt.raw = value,
-            0x0400_0010 => bus.ppu.bg[0].hofs.raw = value, // TODO: Don't write out every HOFS / VOFS?
-            0x0400_0012 => bus.ppu.bg[0].vofs.raw = value,
-            0x0400_0014 => bus.ppu.bg[1].hofs.raw = value,
-            0x0400_0016 => bus.ppu.bg[1].vofs.raw = value,
-            0x0400_0018 => bus.ppu.bg[2].hofs.raw = value,
-            0x0400_001A => bus.ppu.bg[2].vofs.raw = value,
-            0x0400_001C => bus.ppu.bg[3].hofs.raw = value,
-            0x0400_001E => bus.ppu.bg[3].vofs.raw = value,
-            0x0400_0020 => bus.ppu.aff_bg[0].pa = @bitCast(i16, value),
-            0x0400_0022 => bus.ppu.aff_bg[0].pb = @bitCast(i16, value),
-            0x0400_0024 => bus.ppu.aff_bg[0].pc = @bitCast(i16, value),
-            0x0400_0026 => bus.ppu.aff_bg[0].pd = @bitCast(i16, value),
-            0x0400_0028 => bus.ppu.aff_bg[0].x = @bitCast(i32, setLo(u32, @bitCast(u32, bus.ppu.aff_bg[0].x), value)),
-            0x0400_002A => bus.ppu.aff_bg[0].x = @bitCast(i32, setHi(u32, @bitCast(u32, bus.ppu.aff_bg[0].x), value)),
-            0x0400_002C => bus.ppu.aff_bg[0].y = @bitCast(i32, setLo(u32, @bitCast(u32, bus.ppu.aff_bg[0].y), value)),
-            0x0400_002E => bus.ppu.aff_bg[0].y = @bitCast(i32, setHi(u32, @bitCast(u32, bus.ppu.aff_bg[0].y), value)),
-            0x0400_0030 => bus.ppu.aff_bg[1].pa = @bitCast(i16, value),
-            0x0400_0032 => bus.ppu.aff_bg[1].pb = @bitCast(i16, value),
-            0x0400_0034 => bus.ppu.aff_bg[1].pc = @bitCast(i16, value),
-            0x0400_0036 => bus.ppu.aff_bg[1].pd = @bitCast(i16, value),
-            0x0400_0038 => bus.ppu.aff_bg[1].x = @bitCast(i32, setLo(u32, @bitCast(u32, bus.ppu.aff_bg[1].x), value)),
-            0x0400_003A => bus.ppu.aff_bg[1].x = @bitCast(i32, setHi(u32, @bitCast(u32, bus.ppu.aff_bg[1].x), value)),
-            0x0400_003C => bus.ppu.aff_bg[1].y = @bitCast(i32, setLo(u32, @bitCast(u32, bus.ppu.aff_bg[1].y), value)),
-            0x0400_003E => bus.ppu.aff_bg[1].y = @bitCast(i32, setHi(u32, @bitCast(u32, bus.ppu.aff_bg[1].y), value)),
-            0x0400_0040 => bus.ppu.win.h[0].raw = value,
-            0x0400_0042 => bus.ppu.win.h[1].raw = value,
-            0x0400_0044 => bus.ppu.win.v[0].raw = value,
-            0x0400_0046 => bus.ppu.win.v[1].raw = value,
-            0x0400_0048 => bus.ppu.win.in.raw = value,
-            0x0400_004A => bus.ppu.win.out.raw = value,
-            0x0400_004C => log.debug("Wrote 0x{X:0>4} to MOSAIC", .{value}),
-            0x0400_0050 => bus.ppu.bldcnt.raw = value,
-            0x0400_0052 => bus.ppu.bldalpha.raw = value,
-            0x0400_0054 => bus.ppu.bldy.raw = value,
-            0x0400_004E, 0x0400_0056 => {}, // Not used
+            0x0400_0000...0x0400_0054 => ppu.write(T, &bus.ppu, address, value),
+            0x0400_0056 => {}, // Not used
 
             // Sound
-            0x0400_0060...0x0400_009E => apu.write(T, &bus.apu, address, value),
+            0x0400_0060...0x0400_00A6 => apu.write(T, &bus.apu, address, value),
 
             // Dma Transfers
             0x0400_00B0...0x0400_00DE => dma.write(T, &bus.dma, address, value),
 
             // Timers
             0x0400_0100...0x0400_010E => timer.write(T, &bus.tim, address, value),
-            0x0400_0114 => {}, // TODO: Gyakuten Saiban writes 0x8000 to 0x0400_0114
+            0x0400_0114 => {},
             0x0400_0110 => {}, // Not Used,
 
             // Serial Communication 1
@@ -292,27 +241,29 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
             // Interrupts
             0x0400_0200 => bus.io.ie.raw = value,
             0x0400_0202 => bus.io.irq.raw &= ~value,
-            0x0400_0204 => log.debug("Wrote 0x{X:0>4} to WAITCNT", .{value}),
+            0x0400_0204 => bus.io.waitcnt.set(value),
+            0x0400_0206 => {},
             0x0400_0208 => bus.io.ime = value & 1 == 1,
-            0x0400_0206, 0x0400_020A => {}, // Not Used
+            0x0400_020A => {},
+            0x0400_0300 => {
+                bus.io.postflg = @enumFromInt(value & 1);
+                bus.io.haltcnt = if (value >> 15 & 1 == 0) .Halt else @panic("TODO: Implement STOP");
+            },
             else => util.io.write.undef(log, "Tried to write 0x{X:0>4}{} to 0x{X:0>8}", .{ value, T, address }),
         },
         u8 => switch (address) {
             // Display
-            0x0400_0004 => bus.ppu.dispstat.raw = setLo(u16, bus.ppu.dispstat.raw, value),
-            0x0400_0005 => bus.ppu.dispstat.raw = setHi(u16, bus.ppu.dispstat.raw, value),
-            0x0400_0008 => bus.ppu.bg[0].cnt.raw = setLo(u16, bus.ppu.bg[0].cnt.raw, value),
-            0x0400_0009 => bus.ppu.bg[0].cnt.raw = setHi(u16, bus.ppu.bg[0].cnt.raw, value),
-            0x0400_000A => bus.ppu.bg[1].cnt.raw = setLo(u16, bus.ppu.bg[1].cnt.raw, value),
-            0x0400_000B => bus.ppu.bg[1].cnt.raw = setHi(u16, bus.ppu.bg[1].cnt.raw, value),
-            0x0400_0048 => bus.ppu.win.in.raw = setLo(u16, bus.ppu.win.in.raw, value),
-            0x0400_0049 => bus.ppu.win.in.raw = setHi(u16, bus.ppu.win.in.raw, value),
-            0x0400_004A => bus.ppu.win.out.raw = setLo(u16, bus.ppu.win.out.raw, value),
-            0x0400_0054 => bus.ppu.bldy.raw = setLo(u16, bus.ppu.bldy.raw, value),
+            0x0400_0000...0x0400_0055 => ppu.write(T, &bus.ppu, address, value),
 
             // Sound
             0x0400_0060...0x0400_00A7 => apu.write(T, &bus.apu, address, value),
 
+            // Dma Transfers
+            0x0400_00B0...0x0400_00DF => dma.write(T, &bus.dma, address, value),
+
+            // Timers
+            0x0400_0100...0x0400_010F => timer.write(T, &bus.tim, address, value),
+
             // Serial Communication 1
             0x0400_0120 => log.debug("Wrote 0x{X:0>2} to SIODATA32_L_L", .{value}),
             0x0400_0128 => log.debug("Wrote 0x{X:0>2} to SIOCNT_L", .{value}),
@@ -322,9 +273,16 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
             0x0400_0140 => log.debug("Wrote 0x{X:0>2} to JOYCNT_L", .{value}),
 
             // Interrupts
+            0x0400_0200, 0x0400_0201 => bus.io.ie.raw = setHalf(u16, bus.io.ie.raw, @truncate(address), value),
             0x0400_0202 => bus.io.irq.raw &= ~@as(u16, value),
+            0x0400_0203 => bus.io.irq.raw &= ~@as(u16, value) << 8, // TODO: Is this good?
+            0x0400_0204, 0x0400_0205 => bus.io.waitcnt.set(setHalf(u16, bus.io.waitcnt.raw, @truncate(address), value)),
+            0x0400_0206, 0x0400_0207 => {},
             0x0400_0208 => bus.io.ime = value & 1 == 1,
-            0x0400_0300 => bus.io.postflg = std.meta.intToEnum(PostFlag, value & 1) catch unreachable,
+            0x0400_0209 => {},
+            0x0400_020A, 0x0400_020B => {},
+
+            0x0400_0300 => bus.io.postflg = @enumFromInt(value & 1),
             0x0400_0301 => bus.io.haltcnt = if (value >> 7 & 1 == 0) .Halt else std.debug.panic("TODO: Implement STOP", .{}),
 
             0x0400_0410 => log.debug("Wrote 0x{X:0>2} to the common yet undocumented 0x{X:0>8}", .{ value, address }),
@@ -363,14 +321,22 @@ pub const DisplayControl = extern union {
 
 /// Read / Write
 pub const DisplayStatus = extern union {
+    /// read-only
     vblank: Bit(u16, 0),
+    /// read-only
     hblank: Bit(u16, 1),
+    // read-only
     coincidence: Bit(u16, 2),
     vblank_irq: Bit(u16, 3),
     hblank_irq: Bit(u16, 4),
     vcount_irq: Bit(u16, 5),
     vcount_trigger: Bitfield(u16, 8, 8),
     raw: u16,
+
+    pub fn set(self: *DisplayStatus, value: u16) void {
+        const mask: u16 = 0x00C7; // set bits are read-only
+        self.raw = (self.raw & mask) | (value & ~mask);
+    }
 };
 
 /// Read Only
@@ -384,10 +350,10 @@ const InterruptEnable = extern union {
     vblank: Bit(u16, 0),
     hblank: Bit(u16, 1),
     coincidence: Bit(u16, 2),
-    tm0_overflow: Bit(u16, 3),
-    tm1_overflow: Bit(u16, 4),
-    tm2_overflow: Bit(u16, 5),
-    tm3_overflow: Bit(u16, 6),
+    tim0: Bit(u16, 3),
+    tim1: Bit(u16, 4),
+    tim2: Bit(u16, 5),
+    tim3: Bit(u16, 6),
     serial: Bit(u16, 7),
     dma0: Bit(u16, 8),
     dma1: Bit(u16, 9),
@@ -400,7 +366,7 @@ const InterruptEnable = extern union {
 
 /// Read Only
 /// 0 = Pressed, 1 = Released
-const KeyInput = extern union {
+pub const KeyInput = extern union {
     a: Bit(u16, 0),
     b: Bit(u16, 1),
     select: Bit(u16, 2),
@@ -414,6 +380,32 @@ const KeyInput = extern union {
     raw: u16,
 };
 
+const AtomicKeyInput = struct {
+    const Self = @This();
+    const AtomicOrder = std.builtin.AtomicOrder;
+
+    inner: KeyInput,
+
+    pub fn init(value: KeyInput) Self {
+        return .{ .inner = value };
+    }
+
+    pub inline fn load(self: *const Self, comptime ordering: AtomicOrder) u16 {
+        return switch (ordering) {
+            .acq_rel, .release => @compileError("not supported for atomic loads"),
+            else => @atomicLoad(u16, &self.inner.raw, ordering),
+        };
+    }
+
+    pub inline fn fetchOr(self: *Self, value: u16, comptime ordering: AtomicOrder) void {
+        _ = @atomicRmw(u16, &self.inner.raw, .Or, value, ordering);
+    }
+
+    pub inline fn fetchAnd(self: *Self, value: u16, comptime ordering: AtomicOrder) void {
+        _ = @atomicRmw(u16, &self.inner.raw, .And, value, ordering);
+    }
+};
+
 // Read / Write
 pub const BackgroundControl = extern union {
     priority: Bitfield(u16, 0, 2),
@@ -462,6 +454,8 @@ pub const BldY = extern union {
     raw: u16,
 };
 
+const u8WriteKind = enum { Hi, Lo };
+
 /// Write-only
 pub const WinH = extern union {
     x2: Bitfield(u16, 0, 8),
@@ -471,6 +465,8 @@ pub const WinH = extern union {
 
 /// Write-only
 pub const WinV = extern union {
+    const Self = @This();
+
     y2: Bitfield(u16, 0, 8),
     y1: Bitfield(u16, 8, 8),
     raw: u16,
@@ -479,20 +475,20 @@ pub const WinV = extern union {
 pub const WinIn = extern union {
     w0_bg: Bitfield(u16, 0, 4),
     w0_obj: Bit(u16, 4),
-    w0_colour: Bit(u16, 5),
+    w0_bld: Bit(u16, 5),
     w1_bg: Bitfield(u16, 8, 4),
     w1_obj: Bit(u16, 12),
-    w1_colour: Bit(u16, 13),
+    w1_bld: Bit(u16, 13),
     raw: u16,
 };
 
 pub const WinOut = extern union {
     out_bg: Bitfield(u16, 0, 4),
     out_obj: Bit(u16, 4),
-    out_colour: Bit(u16, 5),
+    out_bld: Bit(u16, 5),
     obj_bg: Bitfield(u16, 8, 4),
     obj_obj: Bit(u16, 12),
-    obj_colour: Bit(u16, 13),
+    obj_bld: Bit(u16, 13),
     raw: u16,
 };
 
@@ -661,3 +657,24 @@ pub const SoundBias = extern union {
     sampling_cycle: Bitfield(u16, 14, 2),
     raw: u16,
 };
+
+/// Read / Write
+pub const WaitControl = extern union {
+    sram_cnt: Bitfield(u16, 0, 2),
+    s0_first: Bitfield(u16, 2, 2),
+    s0_second: Bit(u16, 4),
+    s1_first: Bitfield(u16, 5, 2),
+    s1_second: Bit(u16, 7),
+    s2_first: Bitfield(u16, 8, 2),
+    s2_second: Bit(u16, 10),
+    phi_out: Bitfield(u16, 11, 2),
+
+    prefetch_enable: Bit(u16, 14),
+    pak_kind: Bit(u16, 15),
+    raw: u16,
+
+    pub fn set(self: *WaitControl, value: u16) void {
+        const mask: u16 = 0x8000; // set bits are read-only
+        self.raw = (self.raw & mask) | (value & ~mask);
+    }
+};

src/core/bus/timer.zig

@@ -2,68 +2,102 @@ const std = @import("std");
 const util = @import("../../util.zig");
 
 const TimerControl = @import("io.zig").TimerControl;
-const Io = @import("io.zig").Io;
 const Scheduler = @import("../scheduler.zig").Scheduler;
-const Event = @import("../scheduler.zig").Event;
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("../Bus.zig");
 
-pub const TimerTuple = std.meta.Tuple(&[_]type{ Timer(0), Timer(1), Timer(2), Timer(3) });
+const handleInterrupt = @import("../cpu_util.zig").handleInterrupt;
+
+pub const TimerTuple = struct { Timer(0), Timer(1), Timer(2), Timer(3) };
 const log = std.log.scoped(.Timer);
 
+const getHalf = util.getHalf;
+const setHalf = util.setHalf;
+
 pub fn create(sched: *Scheduler) TimerTuple {
     return .{ Timer(0).init(sched), Timer(1).init(sched), Timer(2).init(sched), Timer(3).init(sched) };
 }
 
 pub fn read(comptime T: type, tim: *const TimerTuple, addr: u32) ?T {
-    const nybble = @truncate(u4, addr);
+    const nybble_addr: u4 = @truncate(addr);
 
     return switch (T) {
-        u32 => switch (nybble) {
+        u32 => switch (nybble_addr) {
             0x0 => @as(T, tim.*[0].cnt.raw) << 16 | tim.*[0].timcntL(),
             0x4 => @as(T, tim.*[1].cnt.raw) << 16 | tim.*[1].timcntL(),
             0x8 => @as(T, tim.*[2].cnt.raw) << 16 | tim.*[2].timcntL(),
             0xC => @as(T, tim.*[3].cnt.raw) << 16 | tim.*[3].timcntL(),
-            else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
+            else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
         },
-        u16 => switch (nybble) {
+        u16 => switch (nybble_addr) {
             0x0 => tim.*[0].timcntL(),
             0x2 => tim.*[0].cnt.raw,
+
             0x4 => tim.*[1].timcntL(),
             0x6 => tim.*[1].cnt.raw,
+
             0x8 => tim.*[2].timcntL(),
             0xA => tim.*[2].cnt.raw,
+
             0xC => tim.*[3].timcntL(),
             0xE => tim.*[3].cnt.raw,
-            else => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
+            else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
+        },
+        u8 => switch (nybble_addr) {
+            0x0, 0x1 => @truncate(tim.*[0].timcntL() >> getHalf(nybble_addr)),
+            0x2, 0x3 => @truncate(tim.*[0].cnt.raw >> getHalf(nybble_addr)),
+
+            0x4, 0x5 => @truncate(tim.*[1].timcntL() >> getHalf(nybble_addr)),
+            0x6, 0x7 => @truncate(tim.*[1].cnt.raw >> getHalf(nybble_addr)),
+
+            0x8, 0x9 => @truncate(tim.*[2].timcntL() >> getHalf(nybble_addr)),
+            0xA, 0xB => @truncate(tim.*[2].cnt.raw >> getHalf(nybble_addr)),
+
+            0xC, 0xD => @truncate(tim.*[3].timcntL() >> getHalf(nybble_addr)),
+            0xE, 0xF => @truncate(tim.*[3].cnt.raw >> getHalf(nybble_addr)),
         },
-        u8 => util.io.read.undef(T, log, "Tried to perform a {} read to 0x{X:0>8}", .{ T, addr }),
         else => @compileError("TIM: Unsupported read width"),
     };
 }
 
 pub fn write(comptime T: type, tim: *TimerTuple, addr: u32, value: T) void {
-    const nybble = @truncate(u4, addr);
+    const nybble_addr: u4 = @truncate(addr);
 
     return switch (T) {
-        u32 => switch (nybble) {
+        u32 => switch (nybble_addr) {
             0x0 => tim.*[0].setTimcnt(value),
             0x4 => tim.*[1].setTimcnt(value),
             0x8 => tim.*[2].setTimcnt(value),
             0xC => tim.*[3].setTimcnt(value),
             else => util.io.write.undef(log, "Tried to write 0x{X:0>8}{} to 0x{X:0>8}", .{ value, T, addr }),
         },
-        u16 => switch (nybble) {
+        u16 => switch (nybble_addr) {
             0x0 => tim.*[0].setTimcntL(value),
             0x2 => tim.*[0].setTimcntH(value),
+
             0x4 => tim.*[1].setTimcntL(value),
             0x6 => tim.*[1].setTimcntH(value),
+
             0x8 => tim.*[2].setTimcntL(value),
             0xA => tim.*[2].setTimcntH(value),
+
             0xC => tim.*[3].setTimcntL(value),
             0xE => tim.*[3].setTimcntH(value),
             else => util.io.write.undef(log, "Tried to write 0x{X:0>4}{} to 0x{X:0>8}", .{ value, T, addr }),
         },
-        u8 => util.io.write.undef(log, "Tried to write 0x{X:0>2}{} to 0x{X:0>8}", .{ value, T, addr }),
+        u8 => switch (nybble_addr) {
+            0x0, 0x1 => tim.*[0].setTimcntL(setHalf(u16, tim.*[0]._reload, nybble_addr, value)),
+            0x2, 0x3 => tim.*[0].setTimcntH(setHalf(u16, tim.*[0].cnt.raw, nybble_addr, value)),
+
+            0x4, 0x5 => tim.*[1].setTimcntL(setHalf(u16, tim.*[1]._reload, nybble_addr, value)),
+            0x6, 0x7 => tim.*[1].setTimcntH(setHalf(u16, tim.*[1].cnt.raw, nybble_addr, value)),
+
+            0x8, 0x9 => tim.*[2].setTimcntL(setHalf(u16, tim.*[2]._reload, nybble_addr, value)),
+            0xA, 0xB => tim.*[2].setTimcntH(setHalf(u16, tim.*[2].cnt.raw, nybble_addr, value)),
+
+            0xC, 0xD => tim.*[3].setTimcntL(setHalf(u16, tim.*[3]._reload, nybble_addr, value)),
+            0xE, 0xF => tim.*[3].setTimcntH(setHalf(u16, tim.*[3].cnt.raw, nybble_addr, value)),
+        },
         else => @compileError("TIM: Unsupported write width"),
     };
 }
@@ -97,11 +131,17 @@ fn Timer(comptime id: u2) type {
             };
         }
 
+        pub fn reset(self: *Self) void {
+            const scheduler = self.sched;
+
+            self.* = Self.init(scheduler);
+        }
+
         /// TIMCNT_L Getter
         pub fn timcntL(self: *const Self) u16 {
             if (self.cnt.cascade.read() or !self.cnt.enabled.read()) return self._counter;
 
-            return self._counter +% @truncate(u16, (self.sched.now() - self._start_timestamp) / self.frequency());
+            return self._counter +% @as(u16, @truncate((self.sched.now() - self._start_timestamp) / self.frequency()));
         }
 
         /// TIMCNT_L Setter
@@ -111,35 +151,52 @@ fn Timer(comptime id: u2) type {
 
         /// TIMCNT_L & TIMCNT_H
         pub fn setTimcnt(self: *Self, word: u32) void {
-            self.setTimcntL(@truncate(u16, word));
-            self.setTimcntH(@truncate(u16, word >> 16));
+            self.setTimcntL(@truncate(word));
+            self.setTimcntH(@truncate(word >> 16));
         }
 
         /// TIMCNT_H
         pub fn setTimcntH(self: *Self, halfword: u16) void {
             const new = TimerControl{ .raw = halfword };
 
-            // If Timer happens to be enabled, It will either be resheduled or disabled
-            self.sched.removeScheduledEvent(.{ .TimerOverflow = id });
+            if (self.cnt.enabled.read()) {
+                // timer was already enabled
 
-            if (self.cnt.enabled.read() and (new.cascade.read() or !new.enabled.read())) {
-                // Either through the cascade bit or the enable bit, the timer has effectively been disabled
-                // The Counter should hold whatever value it should have been at when it was disabled
-                self._counter +%= @truncate(u16, (self.sched.now() - self._start_timestamp) / self.frequency());
+                // If enabled falling edge or cascade falling edge, timer is paused
+                if (!new.enabled.read() or (!self.cnt.cascade.read() and new.cascade.read())) {
+                    self.sched.removeScheduledEvent(.{ .TimerOverflow = id });
+
+                    // Counter should hold the value it stopped at meaning we have to calculate it now
+                    self._counter +%= @truncate((self.sched.now() - self._start_timestamp) / self.frequency());
+                }
+
+                // the timer has always been enabled, but the cascade bit which was blocking the timer has been unset
+                if (new.enabled.read() and (self.cnt.cascade.read() and !new.cascade.read())) {
+                    // we want to reschedule the timer event, however we won't reload the counter.
+                    // the invariant here is that self._counter holds the already calculated paused value
+
+                    self.rescheduleTimerExpire(0);
+                }
+            } else {
+                // the timer was previously disabeld
+
+                if (new.enabled.read()) {
+                    // timer should start counting (with a reloaded counter value)
+                    self._counter = self._reload;
+
+                    // if cascade happens to be set, the timer doesn't actually do anything though
+                    if (!new.cascade.read()) self.rescheduleTimerExpire(0);
+                }
             }
 
-            // The counter is only reloaded on the rising edge of the enable bit
-            if (!self.cnt.enabled.read() and new.enabled.read()) self._counter = self._reload;
-
-            // If Timer is enabled and we're not cascading, we need to schedule an overflow event
-            if (new.enabled.read() and !new.cascade.read()) self.rescheduleTimerExpire(0);
-
             self.cnt.raw = halfword;
         }
 
         pub fn onTimerExpire(self: *Self, cpu: *Arm7tdmi, late: u64) void {
             // Fire IRQ if enabled
-            const io = &cpu.bus.io;
+            const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+            const io = &bus_ptr.io;
 
             if (self.cnt.irq.read()) {
                 switch (id) {
@@ -149,33 +206,30 @@ fn Timer(comptime id: u2) type {
                     3 => io.irq.tim3.set(),
                 }
 
-                cpu.handleInterrupt();
+                handleInterrupt(cpu);
             }
 
             // DMA Sound Things
             if (id == 0 or id == 1) {
-                cpu.bus.apu.onDmaAudioSampleRequest(cpu, id);
+                bus_ptr.apu.onDmaAudioSampleRequest(cpu, id);
             }
 
             // Perform Cascade Behaviour
             switch (id) {
-                0 => if (cpu.bus.tim[1].cnt.cascade.read()) {
-                    cpu.bus.tim[1]._counter +%= 1;
-                    if (cpu.bus.tim[1]._counter == 0) cpu.bus.tim[1].onTimerExpire(cpu, late);
+                inline 0, 1, 2 => |idx| {
+                    const next = idx + 1;
+
+                    if (bus_ptr.tim[next].cnt.cascade.read()) {
+                        bus_ptr.tim[next]._counter +%= 1;
+                        if (bus_ptr.tim[next]._counter == 0) bus_ptr.tim[next].onTimerExpire(cpu, late);
+                    }
                 },
-                1 => if (cpu.bus.tim[2].cnt.cascade.read()) {
-                    cpu.bus.tim[2]._counter +%= 1;
-                    if (cpu.bus.tim[2]._counter == 0) cpu.bus.tim[2].onTimerExpire(cpu, late);
-                },
-                2 => if (cpu.bus.tim[3].cnt.cascade.read()) {
-                    cpu.bus.tim[3]._counter +%= 1;
-                    if (cpu.bus.tim[3]._counter == 0) cpu.bus.tim[3].onTimerExpire(cpu, late);
-                },
-                3 => {}, // There is no Timer for TIM3 to "cascade" to,
+                3 => {}, // THere is no timer for TIM3 to cascade to
             }
 
             // Reschedule Timer if we're not cascading
-            if (!self.cnt.cascade.read()) {
+            // TIM0 cascade value is N/A
+            if (id == 0 or !self.cnt.cascade.read()) {
                 self._counter = self._reload;
                 self.rescheduleTimerExpire(late);
             }

src/core/cpu.zig

@@ -1,667 +0,0 @@
-const std = @import("std");
-const util = @import("../util.zig");
-
-const Bus = @import("Bus.zig");
-const Bit = @import("bitfield").Bit;
-const Bitfield = @import("bitfield").Bitfield;
-const Scheduler = @import("scheduler.zig").Scheduler;
-const FilePaths = @import("../util.zig").FilePaths;
-const Logger = @import("../util.zig").Logger;
-
-const File = std.fs.File;
-
-// ARM Instructions
-pub const arm = struct {
-    pub const InstrFn = *const fn (*Arm7tdmi, *Bus, u32) void;
-    const lut: [0x1000]InstrFn = populate();
-
-    const processing = @import("cpu/arm/data_processing.zig").dataProcessing;
-    const psrTransfer = @import("cpu/arm/psr_transfer.zig").psrTransfer;
-    const transfer = @import("cpu/arm/single_data_transfer.zig").singleDataTransfer;
-    const halfSignedTransfer = @import("cpu/arm/half_signed_data_transfer.zig").halfAndSignedDataTransfer;
-    const blockTransfer = @import("cpu/arm/block_data_transfer.zig").blockDataTransfer;
-    const branch = @import("cpu/arm/branch.zig").branch;
-    const branchExchange = @import("cpu/arm/branch.zig").branchAndExchange;
-    const swi = @import("cpu/arm/software_interrupt.zig").armSoftwareInterrupt;
-    const swap = @import("cpu/arm/single_data_swap.zig").singleDataSwap;
-
-    const multiply = @import("cpu/arm/multiply.zig").multiply;
-    const multiplyLong = @import("cpu/arm/multiply.zig").multiplyLong;
-
-    /// Determine index into ARM InstrFn LUT
-    fn idx(opcode: u32) u12 {
-        return @truncate(u12, opcode >> 20 & 0xFF) << 4 | @truncate(u12, opcode >> 4 & 0xF);
-    }
-
-    // Undefined ARM Instruction handler
-    fn und(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-        const id = idx(opcode);
-        cpu.panic("[CPU/Decode] ID: 0x{X:0>3} 0x{X:0>8} is an illegal opcode", .{ id, opcode });
-    }
-
-    fn populate() [0x1000]InstrFn {
-        return comptime {
-            @setEvalBranchQuota(0xE000);
-            var ret = [_]InstrFn{und} ** 0x1000;
-
-            var i: usize = 0;
-            while (i < ret.len) : (i += 1) {
-                ret[i] = switch (@as(u2, i >> 10)) {
-                    0b00 => if (i == 0x121) blk: {
-                        break :blk branchExchange;
-                    } else if (i & 0xFCF == 0x009) blk: {
-                        const A = i >> 5 & 1 == 1;
-                        const S = i >> 4 & 1 == 1;
-                        break :blk multiply(A, S);
-                    } else if (i & 0xFBF == 0x109) blk: {
-                        const B = i >> 6 & 1 == 1;
-                        break :blk swap(B);
-                    } else if (i & 0xF8F == 0x089) blk: {
-                        const U = i >> 6 & 1 == 1;
-                        const A = i >> 5 & 1 == 1;
-                        const S = i >> 4 & 1 == 1;
-                        break :blk multiplyLong(U, A, S);
-                    } else if (i & 0xE49 == 0x009 or i & 0xE49 == 0x049) blk: {
-                        const P = i >> 8 & 1 == 1;
-                        const U = i >> 7 & 1 == 1;
-                        const I = i >> 6 & 1 == 1;
-                        const W = i >> 5 & 1 == 1;
-                        const L = i >> 4 & 1 == 1;
-                        break :blk halfSignedTransfer(P, U, I, W, L);
-                    } else if (i & 0xD90 == 0x100) blk: {
-                        const I = i >> 9 & 1 == 1;
-                        const R = i >> 6 & 1 == 1;
-                        const kind = i >> 4 & 0x3;
-                        break :blk psrTransfer(I, R, kind);
-                    } else blk: {
-                        const I = i >> 9 & 1 == 1;
-                        const S = i >> 4 & 1 == 1;
-                        const instrKind = i >> 5 & 0xF;
-                        break :blk processing(I, S, instrKind);
-                    },
-                    0b01 => if (i >> 9 & 1 == 1 and i & 1 == 1) und else blk: {
-                        const I = i >> 9 & 1 == 1;
-                        const P = i >> 8 & 1 == 1;
-                        const U = i >> 7 & 1 == 1;
-                        const B = i >> 6 & 1 == 1;
-                        const W = i >> 5 & 1 == 1;
-                        const L = i >> 4 & 1 == 1;
-                        break :blk transfer(I, P, U, B, W, L);
-                    },
-                    else => switch (@as(u2, i >> 9 & 0x3)) {
-                        // MSB is guaranteed to be 1
-                        0b00 => blk: {
-                            const P = i >> 8 & 1 == 1;
-                            const U = i >> 7 & 1 == 1;
-                            const S = i >> 6 & 1 == 1;
-                            const W = i >> 5 & 1 == 1;
-                            const L = i >> 4 & 1 == 1;
-                            break :blk blockTransfer(P, U, S, W, L);
-                        },
-                        0b01 => blk: {
-                            const L = i >> 8 & 1 == 1;
-                            break :blk branch(L);
-                        },
-                        0b10 => und, // COP Data Transfer
-                        0b11 => if (i >> 8 & 1 == 1) swi() else und, // COP Data Operation + Register Transfer
-                    },
-                };
-            }
-
-            return ret;
-        };
-    }
-};
-
-// THUMB Instructions
-pub const thumb = struct {
-    pub const InstrFn = *const fn (*Arm7tdmi, *Bus, u16) void;
-    const lut: [0x400]InstrFn = populate();
-
-    const processing = @import("cpu/thumb/data_processing.zig");
-    const alu = @import("cpu/thumb/alu.zig").fmt4;
-    const transfer = @import("cpu/thumb/data_transfer.zig");
-    const block_transfer = @import("cpu/thumb/block_data_transfer.zig");
-    const swi = @import("cpu/thumb/software_interrupt.zig").fmt17;
-    const branch = @import("cpu/thumb/branch.zig");
-
-    /// Determine index into THUMB InstrFn LUT
-    fn idx(opcode: u16) u10 {
-        return @truncate(u10, opcode >> 6);
-    }
-
-    /// Undefined THUMB Instruction Handler
-    fn und(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-        const id = idx(opcode);
-        cpu.panic("[CPU/Decode] ID: 0b{b:0>10} 0x{X:0>2} is an illegal opcode", .{ id, opcode });
-    }
-
-    fn populate() [0x400]InstrFn {
-        return comptime {
-            @setEvalBranchQuota(5025); // This is exact
-            var ret = [_]InstrFn{und} ** 0x400;
-
-            var i: usize = 0;
-            while (i < ret.len) : (i += 1) {
-                ret[i] = switch (@as(u3, i >> 7 & 0x7)) {
-                    0b000 => if (i >> 5 & 0x3 == 0b11) blk: {
-                        const I = i >> 4 & 1 == 1;
-                        const is_sub = i >> 3 & 1 == 1;
-                        const rn = i & 0x7;
-                        break :blk processing.fmt2(I, is_sub, rn);
-                    } else blk: {
-                        const op = i >> 5 & 0x3;
-                        const offset = i & 0x1F;
-                        break :blk processing.fmt1(op, offset);
-                    },
-                    0b001 => blk: {
-                        const op = i >> 5 & 0x3;
-                        const rd = i >> 2 & 0x7;
-                        break :blk processing.fmt3(op, rd);
-                    },
-                    0b010 => switch (@as(u2, i >> 5 & 0x3)) {
-                        0b00 => if (i >> 4 & 1 == 1) blk: {
-                            const op = i >> 2 & 0x3;
-                            const h1 = i >> 1 & 1;
-                            const h2 = i & 1;
-                            break :blk processing.fmt5(op, h1, h2);
-                        } else blk: {
-                            const op = i & 0xF;
-                            break :blk alu(op);
-                        },
-                        0b01 => blk: {
-                            const rd = i >> 2 & 0x7;
-                            break :blk transfer.fmt6(rd);
-                        },
-                        else => blk: {
-                            const op = i >> 4 & 0x3;
-                            const T = i >> 3 & 1 == 1;
-                            break :blk transfer.fmt78(op, T);
-                        },
-                    },
-                    0b011 => blk: {
-                        const B = i >> 6 & 1 == 1;
-                        const L = i >> 5 & 1 == 1;
-                        const offset = i & 0x1F;
-                        break :blk transfer.fmt9(B, L, offset);
-                    },
-                    else => switch (@as(u3, i >> 6 & 0x7)) {
-                        // MSB is guaranteed to be 1
-                        0b000 => blk: {
-                            const L = i >> 5 & 1 == 1;
-                            const offset = i & 0x1F;
-                            break :blk transfer.fmt10(L, offset);
-                        },
-                        0b001 => blk: {
-                            const L = i >> 5 & 1 == 1;
-                            const rd = i >> 2 & 0x7;
-                            break :blk transfer.fmt11(L, rd);
-                        },
-                        0b010 => blk: {
-                            const isSP = i >> 5 & 1 == 1;
-                            const rd = i >> 2 & 0x7;
-                            break :blk processing.fmt12(isSP, rd);
-                        },
-                        0b011 => if (i >> 4 & 1 == 1) blk: {
-                            const L = i >> 5 & 1 == 1;
-                            const R = i >> 2 & 1 == 1;
-                            break :blk block_transfer.fmt14(L, R);
-                        } else blk: {
-                            const S = i >> 1 & 1 == 1;
-                            break :blk processing.fmt13(S);
-                        },
-                        0b100 => blk: {
-                            const L = i >> 5 & 1 == 1;
-                            const rb = i >> 2 & 0x7;
-
-                            break :blk block_transfer.fmt15(L, rb);
-                        },
-                        0b101 => if (i >> 2 & 0xF == 0b1111) blk: {
-                            break :blk thumb.swi();
-                        } else blk: {
-                            const cond = i >> 2 & 0xF;
-                            break :blk branch.fmt16(cond);
-                        },
-                        0b110 => branch.fmt18(),
-                        0b111 => blk: {
-                            const is_low = i >> 5 & 1 == 1;
-                            break :blk branch.fmt19(is_low);
-                        },
-                    },
-                };
-            }
-
-            return ret;
-        };
-    }
-};
-
-const log = std.log.scoped(.Arm7Tdmi);
-
-pub const Arm7tdmi = struct {
-    const Self = @This();
-
-    r: [16]u32,
-    sched: *Scheduler,
-    bus: *Bus,
-    cpsr: PSR,
-    spsr: PSR,
-
-    /// Storage  for R8_fiq -> R12_fiq and their normal counterparts
-    /// e.g [r[0 + 8], fiq_r[0 + 8], r[1 + 8], fiq_r[1 + 8]...]
-    banked_fiq: [2 * 5]u32,
-
-    /// Storage for r13_<mode>, r14_<mode>
-    /// e.g. [r13, r14, r13_svc, r14_svc]
-    banked_r: [2 * 6]u32,
-
-    banked_spsr: [5]PSR,
-
-    logger: ?Logger,
-
-    pub fn init(sched: *Scheduler, bus: *Bus, log_file: ?std.fs.File) Self {
-        return Self{
-            .r = [_]u32{0x00} ** 16,
-            .sched = sched,
-            .bus = bus,
-            .cpsr = .{ .raw = 0x0000_001F },
-            .spsr = .{ .raw = 0x0000_0000 },
-            .banked_fiq = [_]u32{0x00} ** 10,
-            .banked_r = [_]u32{0x00} ** 12,
-            .banked_spsr = [_]PSR{.{ .raw = 0x0000_0000 }} ** 5,
-            .logger = if (log_file) |file| Logger.init(file) else null,
-        };
-    }
-
-    inline fn bankedIdx(mode: Mode, kind: BankedKind) usize {
-        const idx: usize = switch (mode) {
-            .User, .System => 0,
-            .Supervisor => 1,
-            .Abort => 2,
-            .Undefined => 3,
-            .Irq => 4,
-            .Fiq => 5,
-        };
-
-        return (idx * 2) + if (kind == .R14) @as(usize, 1) else 0;
-    }
-
-    inline fn bankedSpsrIndex(mode: Mode) usize {
-        return switch (mode) {
-            .Supervisor => 0,
-            .Abort => 1,
-            .Undefined => 2,
-            .Irq => 3,
-            .Fiq => 4,
-            else => std.debug.panic("[CPU/Mode] {} does not have a SPSR Register", .{mode}),
-        };
-    }
-
-    inline fn bankedFiqIdx(i: usize, mode: Mode) usize {
-        return (i * 2) + if (mode == .Fiq) @as(usize, 1) else 0;
-    }
-
-    pub inline fn hasSPSR(self: *const Self) bool {
-        const mode = getModeChecked(self, self.cpsr.mode.read());
-        return switch (mode) {
-            .System, .User => false,
-            else => true,
-        };
-    }
-
-    pub inline fn isPrivileged(self: *const Self) bool {
-        const mode = getModeChecked(self, self.cpsr.mode.read());
-        return switch (mode) {
-            .User => false,
-            else => true,
-        };
-    }
-
-    pub inline fn isHalted(self: *const Self) bool {
-        return self.bus.io.haltcnt == .Halt;
-    }
-
-    pub fn setCpsr(self: *Self, value: u32) void {
-        if (value & 0x1F != self.cpsr.raw & 0x1F) self.changeModeFromIdx(@truncate(u5, value & 0x1F));
-        self.cpsr.raw = value;
-    }
-
-    fn changeModeFromIdx(self: *Self, next: u5) void {
-        self.changeMode(getModeChecked(self, next));
-    }
-
-    pub fn setUserModeRegister(self: *Self, idx: usize, value: u32) void {
-        const current = getModeChecked(self, self.cpsr.mode.read());
-
-        switch (idx) {
-            8...12 => {
-                if (current == .Fiq) {
-                    self.banked_fiq[bankedFiqIdx(idx - 8, .User)] = value;
-                } else self.r[idx] = value;
-            },
-            13, 14 => switch (current) {
-                .User, .System => self.r[idx] = value,
-                else => {
-                    const kind = std.meta.intToEnum(BankedKind, idx - 13) catch unreachable;
-                    self.banked_r[bankedIdx(.User, kind)] = value;
-                },
-            },
-            else => self.r[idx] = value, // R0 -> R7  and R15
-        }
-    }
-
-    pub fn getUserModeRegister(self: *Self, idx: usize) u32 {
-        const current = getModeChecked(self, self.cpsr.mode.read());
-
-        return switch (idx) {
-            8...12 => if (current == .Fiq) self.banked_fiq[bankedFiqIdx(idx - 8, .User)] else self.r[idx],
-            13, 14 => switch (current) {
-                .User, .System => self.r[idx],
-                else => blk: {
-                    const kind = std.meta.intToEnum(BankedKind, idx - 13) catch unreachable;
-                    break :blk self.banked_r[bankedIdx(.User, kind)];
-                },
-            },
-            else => self.r[idx], // R0 -> R7  and R15
-        };
-    }
-
-    pub fn changeMode(self: *Self, next: Mode) void {
-        const now = getModeChecked(self, self.cpsr.mode.read());
-
-        // Bank R8 -> r12
-        var i: usize = 0;
-        while (i < 5) : (i += 1) {
-            self.banked_fiq[bankedFiqIdx(i, now)] = self.r[8 + i];
-        }
-
-        // Bank r13, r14, SPSR
-        switch (now) {
-            .User, .System => {
-                self.banked_r[bankedIdx(now, .R13)] = self.r[13];
-                self.banked_r[bankedIdx(now, .R14)] = self.r[14];
-            },
-            else => {
-                self.banked_r[bankedIdx(now, .R13)] = self.r[13];
-                self.banked_r[bankedIdx(now, .R14)] = self.r[14];
-                self.banked_spsr[bankedSpsrIndex(now)] = self.spsr;
-            },
-        }
-
-        // Grab R8 -> R12
-        i = 0;
-        while (i < 5) : (i += 1) {
-            self.r[8 + i] = self.banked_fiq[bankedFiqIdx(i, next)];
-        }
-
-        // Grab r13, r14, SPSR
-        switch (next) {
-            .User, .System => {
-                self.r[13] = self.banked_r[bankedIdx(next, .R13)];
-                self.r[14] = self.banked_r[bankedIdx(next, .R14)];
-            },
-            else => {
-                self.r[13] = self.banked_r[bankedIdx(next, .R13)];
-                self.r[14] = self.banked_r[bankedIdx(next, .R14)];
-                self.spsr = self.banked_spsr[bankedSpsrIndex(next)];
-            },
-        }
-
-        self.cpsr.mode.write(@enumToInt(next));
-    }
-
-    pub fn fastBoot(self: *Self) void {
-        self.r = std.mem.zeroes([16]u32);
-
-        self.r[0] = 0x08000000;
-        self.r[1] = 0x000000EA;
-        self.r[13] = 0x0300_7F00;
-        self.r[15] = 0x0800_0000;
-
-        self.banked_r[bankedIdx(.Irq, .R13)] = 0x0300_7FA0;
-        self.banked_r[bankedIdx(.Supervisor, .R13)] = 0x0300_7FE0;
-
-        self.cpsr.raw = 0x6000001F;
-    }
-
-    pub fn step(self: *Self) void {
-        if (self.cpsr.t.read()) {
-            const opcode = self.fetch(u16);
-            if (self.logger) |*trace| trace.mgbaLog(self, opcode);
-
-            thumb.lut[thumb.idx(opcode)](self, self.bus, opcode);
-        } else {
-            const opcode = self.fetch(u32);
-            if (self.logger) |*trace| trace.mgbaLog(self, opcode);
-
-            if (checkCond(self.cpsr, @truncate(u4, opcode >> 28))) {
-                arm.lut[arm.idx(opcode)](self, self.bus, opcode);
-            }
-        }
-    }
-
-    pub fn stepDmaTransfer(self: *Self) bool {
-        const dma0 = &self.bus.dma[0];
-        const dma1 = &self.bus.dma[1];
-        const dma2 = &self.bus.dma[2];
-        const dma3 = &self.bus.dma[3];
-
-        if (dma0.in_progress) {
-            dma0.step(self);
-            return true;
-        }
-
-        if (dma1.in_progress) {
-            dma1.step(self);
-            return true;
-        }
-
-        if (dma2.in_progress) {
-            dma2.step(self);
-            return true;
-        }
-
-        if (dma3.in_progress) {
-            dma3.step(self);
-            return true;
-        }
-
-        return false;
-    }
-
-    pub fn handleInterrupt(self: *Self) void {
-        const should_handle = self.bus.io.ie.raw & self.bus.io.irq.raw;
-
-        if (should_handle != 0) {
-            self.bus.io.haltcnt = .Execute;
-            // log.debug("An Interrupt was Fired!", .{});
-
-            // Either IME is not true or I in CPSR is true
-            // Don't handle interrupts
-            if (!self.bus.io.ime or self.cpsr.i.read()) return;
-            // log.debug("An interrupt was Handled!", .{});
-
-            // retAddr.gba says r15 on it's own is off by -04h in both ARM and THUMB mode
-            const r15 = self.r[15] + 4;
-            const cpsr = self.cpsr.raw;
-
-            self.changeMode(.Irq);
-            self.cpsr.t.write(false);
-            self.cpsr.i.write(true);
-
-            self.r[14] = r15;
-            self.spsr.raw = cpsr;
-            self.r[15] = 0x000_0018;
-        }
-    }
-
-    inline fn fetch(self: *Self, comptime T: type) T {
-        comptime std.debug.assert(T == u32 or T == u16); // Opcode may be 32-bit (ARM) or 16-bit (THUMB)
-        defer self.r[15] += if (T == u32) 4 else 2;
-
-        // FIXME: You better hope this is optimized out
-        const tick_cache = self.sched.tick;
-        defer self.sched.tick = tick_cache + Bus.fetch_timings[@boolToInt(T == u32)][@truncate(u4, self.r[15] >> 24)];
-
-        return self.bus.read(T, self.r[15]);
-    }
-
-    pub fn fakePC(self: *const Self) u32 {
-        return self.r[15] + 4;
-    }
-
-    pub fn panic(self: *const Self, comptime format: []const u8, args: anytype) noreturn {
-        var i: usize = 0;
-        while (i < 16) : (i += 4) {
-            const i_1 = i + 1;
-            const i_2 = i + 2;
-            const i_3 = i + 3;
-            std.debug.print("R{}: 0x{X:0>8}\tR{}: 0x{X:0>8}\tR{}: 0x{X:0>8}\tR{}: 0x{X:0>8}\n", .{ i, self.r[i], i_1, self.r[i_1], i_2, self.r[i_2], i_3, self.r[i_3] });
-        }
-        std.debug.print("cpsr: 0x{X:0>8} ", .{self.cpsr.raw});
-        prettyPrintPsr(&self.cpsr);
-
-        std.debug.print("spsr: 0x{X:0>8} ", .{self.spsr.raw});
-        prettyPrintPsr(&self.spsr);
-
-        if (self.cpsr.t.read()) {
-            const opcode = self.bus.dbgRead(u16, self.r[15] - 4);
-            const id = thumb.idx(opcode);
-            std.debug.print("opcode: ID: 0x{b:0>10} 0x{X:0>4}\n", .{ id, opcode });
-        } else {
-            const opcode = self.bus.dbgRead(u32, self.r[15] - 4);
-            const id = arm.idx(opcode);
-            std.debug.print("opcode: ID: 0x{X:0>3} 0x{X:0>8}\n", .{ id, opcode });
-        }
-
-        std.debug.print("tick: {}\n\n", .{self.sched.tick});
-
-        std.debug.panic(format, args);
-    }
-
-    fn prettyPrintPsr(psr: *const PSR) void {
-        std.debug.print("[", .{});
-
-        if (psr.n.read()) std.debug.print("N", .{}) else std.debug.print("-", .{});
-        if (psr.z.read()) std.debug.print("Z", .{}) else std.debug.print("-", .{});
-        if (psr.c.read()) std.debug.print("C", .{}) else std.debug.print("-", .{});
-        if (psr.v.read()) std.debug.print("V", .{}) else std.debug.print("-", .{});
-        if (psr.i.read()) std.debug.print("I", .{}) else std.debug.print("-", .{});
-        if (psr.f.read()) std.debug.print("F", .{}) else std.debug.print("-", .{});
-        if (psr.t.read()) std.debug.print("T", .{}) else std.debug.print("-", .{});
-        std.debug.print("|", .{});
-        if (getMode(psr.mode.read())) |mode| std.debug.print("{s}", .{modeString(mode)}) else std.debug.print("---", .{});
-
-        std.debug.print("]\n", .{});
-    }
-
-    fn modeString(mode: Mode) []const u8 {
-        return switch (mode) {
-            .User => "usr",
-            .Fiq => "fiq",
-            .Irq => "irq",
-            .Supervisor => "svc",
-            .Abort => "abt",
-            .Undefined => "und",
-            .System => "sys",
-        };
-    }
-
-    fn mgbaLog(self: *const Self, file: *const File, opcode: u32) !void {
-        const thumb_fmt = "{X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} cpsr: {X:0>8} | {X:0>4}:\n";
-        const arm_fmt = "{X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} {X:0>8} cpsr: {X:0>8} | {X:0>8}:\n";
-        var buf: [0x100]u8 = [_]u8{0x00} ** 0x100; // this is larger than it needs to be
-
-        const r0 = self.r[0];
-        const r1 = self.r[1];
-        const r2 = self.r[2];
-        const r3 = self.r[3];
-        const r4 = self.r[4];
-        const r5 = self.r[5];
-        const r6 = self.r[6];
-        const r7 = self.r[7];
-        const r8 = self.r[8];
-        const r9 = self.r[9];
-        const r10 = self.r[10];
-        const r11 = self.r[11];
-        const r12 = self.r[12];
-        const r13 = self.r[13];
-        const r14 = self.r[14];
-        const r15 = self.r[15];
-
-        const c_psr = self.cpsr.raw;
-
-        var log_str: []u8 = undefined;
-        if (self.cpsr.t.read()) {
-            if (opcode >> 11 == 0x1E) {
-                // Instruction 1 of a BL Opcode, print in ARM mode
-                const other_half = self.bus.dbgRead(u16, self.r[15]);
-                const bl_opcode = @as(u32, opcode) << 16 | other_half;
-
-                log_str = try std.fmt.bufPrint(&buf, arm_fmt, .{ r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15, c_psr, bl_opcode });
-            } else {
-                log_str = try std.fmt.bufPrint(&buf, thumb_fmt, .{ r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15, c_psr, opcode });
-            }
-        } else {
-            log_str = try std.fmt.bufPrint(&buf, arm_fmt, .{ r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15, c_psr, opcode });
-        }
-
-        _ = try file.writeAll(log_str);
-    }
-};
-
-pub fn checkCond(cpsr: PSR, cond: u4) bool {
-    return switch (cond) {
-        0x0 => cpsr.z.read(), // EQ - Equal
-        0x1 => !cpsr.z.read(), // NE - Not equal
-        0x2 => cpsr.c.read(), // CS - Unsigned higher or same
-        0x3 => !cpsr.c.read(), // CC - Unsigned lower
-        0x4 => cpsr.n.read(), // MI - Negative
-        0x5 => !cpsr.n.read(), // PL - Positive or zero
-        0x6 => cpsr.v.read(), // VS - Overflow
-        0x7 => !cpsr.v.read(), // VC - No overflow
-        0x8 => cpsr.c.read() and !cpsr.z.read(), // HI - unsigned higher
-        0x9 => !cpsr.c.read() or cpsr.z.read(), // LS - unsigned lower or same
-        0xA => cpsr.n.read() == cpsr.v.read(), // GE - Greater or equal
-        0xB => cpsr.n.read() != cpsr.v.read(), // LT - Less than
-        0xC => !cpsr.z.read() and (cpsr.n.read() == cpsr.v.read()), // GT - Greater than
-        0xD => cpsr.z.read() or (cpsr.n.read() != cpsr.v.read()), // LE - Less than or equal
-        0xE => true, // AL - Always
-        0xF => false, // NV - Never (reserved in ARMv3 and up, but seems to have not changed?)
-    };
-}
-
-pub const PSR = extern union {
-    mode: Bitfield(u32, 0, 5),
-    t: Bit(u32, 5),
-    f: Bit(u32, 6),
-    i: Bit(u32, 7),
-    v: Bit(u32, 28),
-    c: Bit(u32, 29),
-    z: Bit(u32, 30),
-    n: Bit(u32, 31),
-    raw: u32,
-};
-
-const Mode = enum(u5) {
-    User = 0b10000,
-    Fiq = 0b10001,
-    Irq = 0b10010,
-    Supervisor = 0b10011,
-    Abort = 0b10111,
-    Undefined = 0b11011,
-    System = 0b11111,
-};
-
-const BankedKind = enum(u1) {
-    R13 = 0,
-    R14,
-};
-
-fn getMode(bits: u5) ?Mode {
-    return std.meta.intToEnum(Mode, bits) catch null;
-}
-
-fn getModeChecked(cpu: *const Arm7tdmi, bits: u5) Mode {
-    return getMode(bits) orelse cpu.panic("[CPU/CPSR] 0b{b:0>5} is an invalid CPU mode", .{bits});
-}

src/core/cpu/arm/block_data_transfer.zig

@@ -1,104 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-pub fn blockDataTransfer(comptime P: bool, comptime U: bool, comptime S: bool, comptime W: bool, comptime L: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u32) void {
-            const rn = @truncate(u4, opcode >> 16 & 0xF);
-            const rlist = opcode & 0xFFFF;
-            const r15 = rlist >> 15 & 1 == 1;
-
-            var count: u32 = 0;
-            var i: u5 = 0;
-            var first: u4 = 0;
-            var write_to_base = true;
-
-            while (i < 16) : (i += 1) {
-                const r = @truncate(u4, 15 - i);
-                if (rlist >> r & 1 == 1) {
-                    first = r;
-                    count += 1;
-                }
-            }
-
-            var start = cpu.r[rn];
-            if (U) {
-                start += if (P) 4 else 0;
-            } else {
-                start = start - (4 * count) + if (!P) 4 else 0;
-            }
-
-            var end = cpu.r[rn];
-            if (U) {
-                end = end + (4 * count) - if (!P) 4 else 0;
-            } else {
-                end -= if (P) 4 else 0;
-            }
-
-            var new_base = cpu.r[rn];
-            if (U) {
-                new_base += 4 * count;
-            } else {
-                new_base -= 4 * count;
-            }
-
-            var address = start;
-
-            if (rlist == 0) {
-                var und_addr = cpu.r[rn];
-                if (U) {
-                    und_addr += if (P) 4 else 0;
-                } else {
-                    und_addr -= 0x40 - if (!P) 4 else 0;
-                }
-
-                if (L) {
-                    cpu.r[15] = bus.read(u32, und_addr);
-                } else {
-                    bus.write(u32, und_addr, cpu.r[15] + 8);
-                }
-
-                cpu.r[rn] = if (U) cpu.r[rn] + 0x40 else cpu.r[rn] - 0x40;
-                return;
-            }
-
-            i = first;
-            while (i < 16) : (i += 1) {
-                if (rlist >> i & 1 == 1) {
-                    transfer(cpu, bus, r15, i, address);
-                    address += 4;
-
-                    if (W and !L and write_to_base) {
-                        cpu.r[rn] = new_base;
-                        write_to_base = false;
-                    }
-                }
-            }
-
-            if (W and L and rlist >> rn & 1 == 0) cpu.r[rn] = new_base;
-        }
-
-        fn transfer(cpu: *Arm7tdmi, bus: *Bus, r15_present: bool, i: u5, address: u32) void {
-            if (L) {
-                if (S and !r15_present) {
-                    // Always Transfer User mode Registers
-                    cpu.setUserModeRegister(i, bus.read(u32, address));
-                } else {
-                    const value = bus.read(u32, address);
-                    cpu.r[i] = if (i == 0xF) value & 0xFFFF_FFFC else value;
-                    if (S and i == 0xF) cpu.setCpsr(cpu.spsr.raw);
-                }
-            } else {
-                if (S) {
-                    // Always Transfer User mode Registers
-                    // This happens regardless if r15 is in the list
-                    const value = cpu.getUserModeRegister(i);
-                    bus.write(u32, address, value + if (i == 0xF) 8 else @as(u32, 0)); // PC is already 4 ahead to make 12
-                } else {
-                    bus.write(u32, address, cpu.r[i] + if (i == 0xF) 8 else @as(u32, 0));
-                }
-            }
-        }
-    }.inner;
-}

src/core/cpu/arm/branch.zig

@@ -1,22 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-const sext = @import("../../../util.zig").sext;
-
-pub fn branch(comptime L: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-            if (L) cpu.r[14] = cpu.r[15];
-            cpu.r[15] = cpu.fakePC() +% (sext(u32, u24, opcode) << 2);
-        }
-    }.inner;
-}
-
-pub fn branchAndExchange(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-    const rn = opcode & 0xF;
-    cpu.cpsr.t.write(cpu.r[rn] & 1 == 1);
-    cpu.r[15] = cpu.r[rn] & 0xFFFF_FFFE;
-}

src/core/cpu/arm/data_processing.zig

@@ -1,284 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-const rotateRight = @import("../barrel_shifter.zig").rotateRight;
-const execute = @import("../barrel_shifter.zig").execute;
-
-pub fn dataProcessing(comptime I: bool, comptime S: bool, comptime instrKind: u4) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-            const rd = @truncate(u4, opcode >> 12 & 0xF);
-            const rn = opcode >> 16 & 0xF;
-            const old_carry = @boolToInt(cpu.cpsr.c.read());
-
-            // If certain conditions are met, PC is 12 ahead instead of 8
-            if (!I and opcode >> 4 & 1 == 1) cpu.r[15] += 4;
-
-            const op1 = if (rn == 0xF) cpu.fakePC() else cpu.r[rn];
-
-            var op2: u32 = undefined;
-            if (I) {
-                const amount = @truncate(u8, (opcode >> 8 & 0xF) << 1);
-                op2 = rotateRight(S, &cpu.cpsr, opcode & 0xFF, amount);
-            } else {
-                op2 = execute(S, cpu, opcode);
-            }
-
-            // Undo special condition from above
-            if (!I and opcode >> 4 & 1 == 1) cpu.r[15] -= 4;
-
-            switch (instrKind) {
-                0x0 => {
-                    // AND
-                    const result = op1 & op2;
-                    cpu.r[rd] = result;
-                    setArmLogicOpFlags(S, cpu, rd, result);
-                },
-                0x1 => {
-                    // EOR
-                    const result = op1 ^ op2;
-                    cpu.r[rd] = result;
-                    setArmLogicOpFlags(S, cpu, rd, result);
-                },
-                0x2 => {
-                    // SUB
-                    cpu.r[rd] = armSub(S, cpu, rd, op1, op2);
-                },
-                0x3 => {
-                    // RSB
-                    cpu.r[rd] = armSub(S, cpu, rd, op2, op1);
-                },
-                0x4 => {
-                    // ADD
-                    cpu.r[rd] = armAdd(S, cpu, rd, op1, op2);
-                },
-                0x5 => {
-                    // ADC
-                    cpu.r[rd] = armAdc(S, cpu, rd, op1, op2, old_carry);
-                },
-                0x6 => {
-                    // SBC
-                    cpu.r[rd] = armSbc(S, cpu, rd, op1, op2, old_carry);
-                },
-                0x7 => {
-                    // RSC
-                    cpu.r[rd] = armSbc(S, cpu, rd, op2, op1, old_carry);
-                },
-                0x8 => {
-                    // TST
-                    if (rd == 0xF) {
-                        undefinedTestBehaviour(cpu);
-                        return;
-                    }
-
-                    const result = op1 & op2;
-                    setTestOpFlags(S, cpu, opcode, result);
-                },
-                0x9 => {
-                    // TEQ
-                    if (rd == 0xF) {
-                        undefinedTestBehaviour(cpu);
-                        return;
-                    }
-
-                    const result = op1 ^ op2;
-                    setTestOpFlags(S, cpu, opcode, result);
-                },
-                0xA => {
-                    // CMP
-                    if (rd == 0xF) {
-                        undefinedTestBehaviour(cpu);
-                        return;
-                    }
-
-                    cmp(cpu, op1, op2);
-                },
-                0xB => {
-                    // CMN
-                    if (rd == 0xF) {
-                        undefinedTestBehaviour(cpu);
-                        return;
-                    }
-
-                    cmn(cpu, op1, op2);
-                },
-                0xC => {
-                    // ORR
-                    const result = op1 | op2;
-                    cpu.r[rd] = result;
-                    setArmLogicOpFlags(S, cpu, rd, result);
-                },
-                0xD => {
-                    // MOV
-                    cpu.r[rd] = op2;
-                    setArmLogicOpFlags(S, cpu, rd, op2);
-                },
-                0xE => {
-                    // BIC
-                    const result = op1 & ~op2;
-                    cpu.r[rd] = result;
-                    setArmLogicOpFlags(S, cpu, rd, result);
-                },
-                0xF => {
-                    // MVN
-                    const result = ~op2;
-                    cpu.r[rd] = result;
-                    setArmLogicOpFlags(S, cpu, rd, result);
-                },
-            }
-        }
-    }.inner;
-}
-
-fn armSbc(comptime S: bool, cpu: *Arm7tdmi, rd: u4, left: u32, right: u32, old_carry: u1) u32 {
-    var result: u32 = undefined;
-    if (S and rd == 0xF) {
-        result = sbc(false, cpu, left, right, old_carry);
-        cpu.setCpsr(cpu.spsr.raw);
-    } else {
-        result = sbc(S, cpu, left, right, old_carry);
-    }
-
-    return result;
-}
-
-pub fn sbc(comptime S: bool, cpu: *Arm7tdmi, left: u32, right: u32, old_carry: u1) u32 {
-    // TODO: Make your own version (thanks peach.bot)
-    const subtrahend = @as(u64, right) -% old_carry +% 1;
-    const result = @truncate(u32, left -% subtrahend);
-
-    if (S) {
-        cpu.cpsr.n.write(result >> 31 & 1 == 1);
-        cpu.cpsr.z.write(result == 0);
-        cpu.cpsr.c.write(subtrahend <= left);
-        cpu.cpsr.v.write(((left ^ result) & (~right ^ result)) >> 31 & 1 == 1);
-    }
-
-    return result;
-}
-
-fn armSub(comptime S: bool, cpu: *Arm7tdmi, rd: u4, left: u32, right: u32) u32 {
-    var result: u32 = undefined;
-    if (S and rd == 0xF) {
-        result = sub(false, cpu, left, right);
-        cpu.setCpsr(cpu.spsr.raw);
-    } else {
-        result = sub(S, cpu, left, right);
-    }
-
-    return result;
-}
-
-pub fn sub(comptime S: bool, cpu: *Arm7tdmi, left: u32, right: u32) u32 {
-    const result = left -% right;
-
-    if (S) {
-        cpu.cpsr.n.write(result >> 31 & 1 == 1);
-        cpu.cpsr.z.write(result == 0);
-        cpu.cpsr.c.write(right <= left);
-        cpu.cpsr.v.write(((left ^ result) & (~right ^ result)) >> 31 & 1 == 1);
-    }
-
-    return result;
-}
-
-fn armAdd(comptime S: bool, cpu: *Arm7tdmi, rd: u4, left: u32, right: u32) u32 {
-    var result: u32 = undefined;
-    if (S and rd == 0xF) {
-        result = add(false, cpu, left, right);
-        cpu.setCpsr(cpu.spsr.raw);
-    } else {
-        result = add(S, cpu, left, right);
-    }
-
-    return result;
-}
-
-pub fn add(comptime S: bool, cpu: *Arm7tdmi, left: u32, right: u32) u32 {
-    var result: u32 = undefined;
-    const didOverflow = @addWithOverflow(u32, left, right, &result);
-
-    if (S) {
-        cpu.cpsr.n.write(result >> 31 & 1 == 1);
-        cpu.cpsr.z.write(result == 0);
-        cpu.cpsr.c.write(didOverflow);
-        cpu.cpsr.v.write(((left ^ result) & (right ^ result)) >> 31 & 1 == 1);
-    }
-
-    return result;
-}
-
-fn armAdc(comptime S: bool, cpu: *Arm7tdmi, rd: u4, left: u32, right: u32, old_carry: u1) u32 {
-    var result: u32 = undefined;
-    if (S and rd == 0xF) {
-        result = adc(false, cpu, left, right, old_carry);
-        cpu.setCpsr(cpu.spsr.raw);
-    } else {
-        result = adc(S, cpu, left, right, old_carry);
-    }
-
-    return result;
-}
-
-pub fn adc(comptime S: bool, cpu: *Arm7tdmi, left: u32, right: u32, old_carry: u1) u32 {
-    var result: u32 = undefined;
-    const did = @addWithOverflow(u32, left, right, &result);
-    const overflow = @addWithOverflow(u32, result, old_carry, &result);
-
-    if (S) {
-        cpu.cpsr.n.write(result >> 31 & 1 == 1);
-        cpu.cpsr.z.write(result == 0);
-        cpu.cpsr.c.write(did or overflow);
-        cpu.cpsr.v.write(((left ^ result) & (right ^ result)) >> 31 & 1 == 1);
-    }
-
-    return result;
-}
-
-pub fn cmp(cpu: *Arm7tdmi, left: u32, right: u32) void {
-    const result = left -% right;
-
-    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-    cpu.cpsr.z.write(result == 0);
-    cpu.cpsr.c.write(right <= left);
-    cpu.cpsr.v.write(((left ^ result) & (~right ^ result)) >> 31 & 1 == 1);
-}
-
-pub fn cmn(cpu: *Arm7tdmi, left: u32, right: u32) void {
-    var result: u32 = undefined;
-    const didOverflow = @addWithOverflow(u32, left, right, &result);
-
-    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-    cpu.cpsr.z.write(result == 0);
-    cpu.cpsr.c.write(didOverflow);
-    cpu.cpsr.v.write(((left ^ result) & (right ^ result)) >> 31 & 1 == 1);
-}
-
-fn setArmLogicOpFlags(comptime S: bool, cpu: *Arm7tdmi, rd: u4, result: u32) void {
-    if (S and rd == 0xF) {
-        cpu.setCpsr(cpu.spsr.raw);
-    } else {
-        setLogicOpFlags(S, cpu, result);
-    }
-}
-
-pub fn setLogicOpFlags(comptime S: bool, cpu: *Arm7tdmi, result: u32) void {
-    if (S) {
-        cpu.cpsr.n.write(result >> 31 & 1 == 1);
-        cpu.cpsr.z.write(result == 0);
-        // C set by Barrel Shifter, V is unaffected
-    }
-}
-
-fn setTestOpFlags(comptime S: bool, cpu: *Arm7tdmi, opcode: u32, result: u32) void {
-    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-    cpu.cpsr.z.write(result == 0);
-    // Barrel Shifter should always calc CPSR C in TST
-    if (!S) _ = execute(true, cpu, opcode);
-}
-
-fn undefinedTestBehaviour(cpu: *Arm7tdmi) void {
-    @setCold(true);
-    cpu.setCpsr(cpu.spsr.raw);
-}

src/core/cpu/arm/half_signed_data_transfer.zig

@@ -1,70 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-const sext = @import("../../../util.zig").sext;
-const rotr = @import("../../../util.zig").rotr;
-
-pub fn halfAndSignedDataTransfer(comptime P: bool, comptime U: bool, comptime I: bool, comptime W: bool, comptime L: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u32) void {
-            const rn = opcode >> 16 & 0xF;
-            const rd = opcode >> 12 & 0xF;
-            const rm = opcode & 0xF;
-            const imm_offset_high = opcode >> 8 & 0xF;
-
-            var base: u32 = undefined;
-            if (rn == 0xF) {
-                base = cpu.fakePC();
-                if (!L) base += 4;
-            } else {
-                base = cpu.r[rn];
-            }
-
-            var offset: u32 = undefined;
-            if (I) {
-                offset = imm_offset_high << 4 | rm;
-            } else {
-                offset = cpu.r[rm];
-            }
-
-            const modified_base = if (U) base +% offset else base -% offset;
-            var address = if (P) modified_base else base;
-
-            var result: u32 = undefined;
-            if (L) {
-                switch (@truncate(u2, opcode >> 5)) {
-                    0b01 => {
-                        // LDRH
-                        const value = bus.read(u16, address);
-                        result = rotr(u32, value, 8 * (address & 1));
-                    },
-                    0b10 => {
-                        // LDRSB
-                        result = sext(u32, u8, bus.read(u8, address));
-                    },
-                    0b11 => {
-                        // LDRSH
-                        result = if (address & 1 == 1) blk: {
-                            break :blk sext(u32, u8, bus.read(u8, address));
-                        } else blk: {
-                            break :blk sext(u32, u16, bus.read(u16, address));
-                        };
-                    },
-                    0b00 => unreachable, // SWP
-                }
-            } else {
-                if (opcode >> 5 & 0x01 == 0x01) {
-                    // STRH
-                    bus.write(u16, address, @truncate(u16, cpu.r[rd]));
-                } else unreachable; // SWP
-            }
-
-            address = modified_base;
-            if (W and P or !P) cpu.r[rn] = address;
-            if (L) cpu.r[rd] = result; // // This emulates the LDR rd == rn behaviour
-        }
-    }.inner;
-}

src/core/cpu/arm/multiply.zig

@@ -1,57 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-pub fn multiply(comptime A: bool, comptime S: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-            const rd = opcode >> 16 & 0xF;
-            const rn = opcode >> 12 & 0xF;
-            const rs = opcode >> 8 & 0xF;
-            const rm = opcode & 0xF;
-
-            const temp: u64 = @as(u64, cpu.r[rm]) * @as(u64, cpu.r[rs]) + if (A) cpu.r[rn] else 0;
-            const result = @truncate(u32, temp);
-            cpu.r[rd] = result;
-
-            if (S) {
-                cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                cpu.cpsr.z.write(result == 0);
-                // V is unaffected, C is *actually* undefined in ARMv4
-            }
-        }
-    }.inner;
-}
-
-pub fn multiplyLong(comptime U: bool, comptime A: bool, comptime S: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-            const rd_hi = opcode >> 16 & 0xF;
-            const rd_lo = opcode >> 12 & 0xF;
-            const rs = opcode >> 8 & 0xF;
-            const rm = opcode & 0xF;
-
-            if (U) {
-                // Signed (WHY IS IT U THEN?)
-                var result: i64 = @as(i64, @bitCast(i32, cpu.r[rm])) * @as(i64, @bitCast(i32, cpu.r[rs]));
-                if (A) result +%= @bitCast(i64, @as(u64, cpu.r[rd_hi]) << 32 | @as(u64, cpu.r[rd_lo]));
-
-                cpu.r[rd_hi] = @bitCast(u32, @truncate(i32, result >> 32));
-                cpu.r[rd_lo] = @bitCast(u32, @truncate(i32, result));
-            } else {
-                // Unsigned
-                var result: u64 = @as(u64, cpu.r[rm]) * @as(u64, cpu.r[rs]);
-                if (A) result +%= @as(u64, cpu.r[rd_hi]) << 32 | @as(u64, cpu.r[rd_lo]);
-
-                cpu.r[rd_hi] = @truncate(u32, result >> 32);
-                cpu.r[rd_lo] = @truncate(u32, result);
-            }
-
-            if (S) {
-                cpu.cpsr.z.write(cpu.r[rd_hi] == 0 and cpu.r[rd_lo] == 0);
-                cpu.cpsr.n.write(cpu.r[rd_hi] >> 31 & 1 == 1);
-                // C and V are set to meaningless values
-            }
-        }
-    }.inner;
-}

src/core/cpu/arm/psr_transfer.zig

@@ -1,59 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-const PSR = @import("../../cpu.zig").PSR;
-
-const log = std.log.scoped(.PsrTransfer);
-
-const rotr = @import("../../../util.zig").rotr;
-
-pub fn psrTransfer(comptime I: bool, comptime R: bool, comptime kind: u2) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-            switch (kind) {
-                0b00 => {
-                    // MRS
-                    const rd = opcode >> 12 & 0xF;
-
-                    if (R and !cpu.hasSPSR()) log.err("Tried to read SPSR from User/System Mode", .{});
-                    cpu.r[rd] = if (R) cpu.spsr.raw else cpu.cpsr.raw;
-                },
-                0b10 => {
-                    // MSR
-                    const field_mask = @truncate(u4, opcode >> 16 & 0xF);
-                    const rm_idx = opcode & 0xF;
-                    const right = if (I) rotr(u32, opcode & 0xFF, (opcode >> 8 & 0xF) * 2) else cpu.r[rm_idx];
-
-                    if (R and !cpu.hasSPSR()) log.err("Tried to write to SPSR in User/System Mode", .{});
-
-                    if (R) {
-                        // arm.gba seems to expect the SPSR to do somethign in SYS mode,
-                        // so we just assume that despite writing to the SPSR in USR or SYS mode
-                        // being UNPREDICTABLE, it just magically has a working SPSR somehow
-                        cpu.spsr.raw = fieldMask(&cpu.spsr, field_mask, right);
-                    } else {
-                        if (cpu.isPrivileged()) cpu.setCpsr(fieldMask(&cpu.cpsr, field_mask, right));
-                    }
-                },
-                else => cpu.panic("[CPU/PSR Transfer] Bits 21:220 of {X:0>8} are undefined", .{opcode}),
-            }
-        }
-    }.inner;
-}
-
-fn fieldMask(psr: *const PSR, field_mask: u4, right: u32) u32 {
-    // This bitwise ORs bits 3 and 0 of the field mask into a u2
-    // We do this because we only care about bits 7:0 and 31:28 of the CPSR
-    const bits = @truncate(u2, (field_mask >> 2 & 0x2) | (field_mask & 1));
-
-    const mask: u32 = switch (bits) {
-        0b00 => 0x0000_0000,
-        0b01 => 0x0000_00FF,
-        0b10 => 0xF000_0000,
-        0b11 => 0xF000_00FF,
-    };
-
-    return (psr.raw & ~mask) | (right & mask);
-}

src/core/cpu/arm/single_data_swap.zig

@@ -1,31 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-const rotr = @import("../../../util.zig").rotr;
-
-pub fn singleDataSwap(comptime B: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u32) void {
-            const rn = opcode >> 16 & 0xF;
-            const rd = opcode >> 12 & 0xF;
-            const rm = opcode & 0xF;
-
-            const address = cpu.r[rn];
-
-            if (B) {
-                // SWPB
-                const value = bus.read(u8, address);
-                bus.write(u8, address, @truncate(u8, cpu.r[rm]));
-                cpu.r[rd] = value;
-            } else {
-                // SWP
-                const value = rotr(u32, bus.read(u32, address), 8 * (address & 0x3));
-                bus.write(u32, address, cpu.r[rm]);
-                cpu.r[rd] = value;
-            }
-        }
-    }.inner;
-}

src/core/cpu/arm/single_data_transfer.zig

@@ -1,57 +0,0 @@
-const std = @import("std");
-const util = @import("../../../util.zig");
-
-const shifter = @import("../barrel_shifter.zig");
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-const rotr = @import("../../../util.zig").rotr;
-
-pub fn singleDataTransfer(comptime I: bool, comptime P: bool, comptime U: bool, comptime B: bool, comptime W: bool, comptime L: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u32) void {
-            const rn = opcode >> 16 & 0xF;
-            const rd = opcode >> 12 & 0xF;
-
-            var base: u32 = undefined;
-            if (rn == 0xF) {
-                base = cpu.fakePC();
-                if (!L) base += 4; // Offset of 12
-            } else {
-                base = cpu.r[rn];
-            }
-
-            const offset = if (I) shifter.immShift(false, cpu, opcode) else opcode & 0xFFF;
-
-            const modified_base = if (U) base +% offset else base -% offset;
-            var address = if (P) modified_base else base;
-
-            var result: u32 = undefined;
-            if (L) {
-                if (B) {
-                    // LDRB
-                    result = bus.read(u8, address);
-                } else {
-                    // LDR
-                    const value = bus.read(u32, address);
-                    result = rotr(u32, value, 8 * (address & 0x3));
-                }
-            } else {
-                if (B) {
-                    // STRB
-                    const value = if (rd == 0xF) cpu.r[rd] + 8 else cpu.r[rd];
-                    bus.write(u8, address, @truncate(u8, value));
-                } else {
-                    // STR
-                    const value = if (rd == 0xF) cpu.r[rd] + 8 else cpu.r[rd];
-                    bus.write(u32, address, value);
-                }
-            }
-
-            address = modified_base;
-            if (W and P or !P) cpu.r[rn] = address;
-            if (L) cpu.r[rd] = result; // This emulates the LDR rd == rn behaviour
-        }
-    }.inner;
-}

src/core/cpu/arm/software_interrupt.zig

@@ -1,22 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-pub fn armSoftwareInterrupt() InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, _: u32) void {
-            // Copy Values from Current Mode
-            const r15 = cpu.r[15];
-            const cpsr = cpu.cpsr.raw;
-
-            // Switch Mode
-            cpu.changeMode(.Supervisor);
-            cpu.cpsr.t.write(false); // Force ARM Mode
-            cpu.cpsr.i.write(true); // Disable normal interrupts
-
-            cpu.r[14] = r15; // Resume Execution
-            cpu.spsr.raw = cpsr; // Previous mode CPSR
-            cpu.r[15] = 0x0000_0008;
-        }
-    }.inner;
-}

src/core/cpu/barrel_shifter.zig

@@ -1,153 +0,0 @@
-const std = @import("std");
-
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
-const CPSR = @import("../cpu.zig").PSR;
-
-const rotr = @import("../../util.zig").rotr;
-
-pub fn execute(comptime S: bool, cpu: *Arm7tdmi, opcode: u32) u32 {
-    var result: u32 = undefined;
-    if (opcode >> 4 & 1 == 1) {
-        result = registerShift(S, cpu, opcode);
-    } else {
-        result = immShift(S, cpu, opcode);
-    }
-
-    return result;
-}
-
-fn registerShift(comptime S: bool, cpu: *Arm7tdmi, opcode: u32) u32 {
-    const rs_idx = opcode >> 8 & 0xF;
-    const rs = @truncate(u8, cpu.r[rs_idx]);
-
-    const rm_idx = opcode & 0xF;
-    const rm = if (rm_idx == 0xF) cpu.fakePC() else cpu.r[rm_idx];
-
-    return switch (@truncate(u2, opcode >> 5)) {
-        0b00 => logicalLeft(S, &cpu.cpsr, rm, rs),
-        0b01 => logicalRight(S, &cpu.cpsr, rm, rs),
-        0b10 => arithmeticRight(S, &cpu.cpsr, rm, rs),
-        0b11 => rotateRight(S, &cpu.cpsr, rm, rs),
-    };
-}
-
-pub fn immShift(comptime S: bool, cpu: *Arm7tdmi, opcode: u32) u32 {
-    const amount = @truncate(u8, opcode >> 7 & 0x1F);
-
-    const rm_idx = opcode & 0xF;
-    const rm = if (rm_idx == 0xF) cpu.fakePC() else cpu.r[rm_idx];
-
-    var result: u32 = undefined;
-    if (amount == 0) {
-        switch (@truncate(u2, opcode >> 5)) {
-            0b00 => {
-                // LSL #0
-                result = rm;
-            },
-            0b01 => {
-                // LSR #0 aka LSR #32
-                if (S) cpu.cpsr.c.write(rm >> 31 & 1 == 1);
-                result = 0x0000_0000;
-            },
-            0b10 => {
-                // ASR #0 aka ASR #32
-                result = @bitCast(u32, @bitCast(i32, rm) >> 31);
-                if (S) cpu.cpsr.c.write(result >> 31 & 1 == 1);
-            },
-            0b11 => {
-                // ROR #0 aka RRX
-                const carry: u32 = @boolToInt(cpu.cpsr.c.read());
-                if (S) cpu.cpsr.c.write(rm & 1 == 1);
-
-                result = (carry << 31) | (rm >> 1);
-            },
-        }
-    } else {
-        switch (@truncate(u2, opcode >> 5)) {
-            0b00 => result = logicalLeft(S, &cpu.cpsr, rm, amount),
-            0b01 => result = logicalRight(S, &cpu.cpsr, rm, amount),
-            0b10 => result = arithmeticRight(S, &cpu.cpsr, rm, amount),
-            0b11 => result = rotateRight(S, &cpu.cpsr, rm, amount),
-        }
-    }
-
-    return result;
-}
-
-pub fn logicalLeft(comptime S: bool, cpsr: *CPSR, rm: u32, total_amount: u8) u32 {
-    const amount = @truncate(u5, total_amount);
-    const bit_count: u8 = @typeInfo(u32).Int.bits;
-
-    var result: u32 = 0x0000_0000;
-    if (total_amount < bit_count) {
-        // We can perform a well-defined shift here
-        result = rm << amount;
-
-        if (S and total_amount != 0) {
-            const carry_bit = @truncate(u5, bit_count - amount);
-            cpsr.c.write(rm >> carry_bit & 1 == 1);
-        }
-    } else {
-        if (S) {
-            if (total_amount == bit_count) {
-                // Shifted all bits out, carry bit is bit 0 of rm
-                cpsr.c.write(rm & 1 == 1);
-            } else {
-                cpsr.c.write(false);
-            }
-        }
-    }
-
-    return result;
-}
-
-pub fn logicalRight(comptime S: bool, cpsr: *CPSR, rm: u32, total_amount: u32) u32 {
-    const amount = @truncate(u5, total_amount);
-    const bit_count: u8 = @typeInfo(u32).Int.bits;
-
-    var result: u32 = 0x0000_0000;
-    if (total_amount < bit_count) {
-        // We can perform a well-defined shift
-        result = rm >> amount;
-        if (S and total_amount != 0) cpsr.c.write(rm >> (amount - 1) & 1 == 1);
-    } else {
-        if (S) {
-            if (total_amount == bit_count) {
-                // LSR #32
-                cpsr.c.write(rm >> 31 & 1 == 1);
-            } else {
-                // All bits have been shifted out, including carry bit
-                cpsr.c.write(false);
-            }
-        }
-    }
-
-    return result;
-}
-
-pub fn arithmeticRight(comptime S: bool, cpsr: *CPSR, rm: u32, total_amount: u8) u32 {
-    const amount = @truncate(u5, total_amount);
-    const bit_count: u8 = @typeInfo(u32).Int.bits;
-
-    var result: u32 = 0x0000_0000;
-    if (total_amount < bit_count) {
-        result = @bitCast(u32, @bitCast(i32, rm) >> amount);
-        if (S and total_amount != 0) cpsr.c.write(rm >> (amount - 1) & 1 == 1);
-    } else {
-        // ASR #32 and ASR #>32 have the same result
-        result = @bitCast(u32, @bitCast(i32, rm) >> 31);
-        if (S) cpsr.c.write(result >> 31 & 1 == 1);
-    }
-
-    return result;
-}
-
-pub fn rotateRight(comptime S: bool, cpsr: *CPSR, rm: u32, total_amount: u8) u32 {
-    const result = rotr(u32, rm, total_amount);
-
-    if (S and total_amount != 0) {
-        cpsr.c.write(result >> 31 & 1 == 1);
-    }
-
-    return result;
-}

src/core/cpu/thumb/alu.zig

@@ -1,118 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-
-const adc = @import("../arm/data_processing.zig").adc;
-const sbc = @import("../arm/data_processing.zig").sbc;
-const sub = @import("../arm/data_processing.zig").sub;
-const cmp = @import("../arm/data_processing.zig").cmp;
-const cmn = @import("../arm/data_processing.zig").cmn;
-const setTestOpFlags = @import("../arm/data_processing.zig").setTestOpFlags;
-const setLogicOpFlags = @import("../arm/data_processing.zig").setLogicOpFlags;
-
-const logicalLeft = @import("../barrel_shifter.zig").logicalLeft;
-const logicalRight = @import("../barrel_shifter.zig").logicalRight;
-const arithmeticRight = @import("../barrel_shifter.zig").arithmeticRight;
-const rotateRight = @import("../barrel_shifter.zig").rotateRight;
-
-pub fn fmt4(comptime op: u4) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const rs = opcode >> 3 & 0x7;
-            const rd = opcode & 0x7;
-            const carry = @boolToInt(cpu.cpsr.c.read());
-
-            switch (op) {
-                0x0 => {
-                    // AND
-                    const result = cpu.r[rd] & cpu.r[rs];
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x1 => {
-                    // EOR
-                    const result = cpu.r[rd] ^ cpu.r[rs];
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x2 => {
-                    // LSL
-                    const result = logicalLeft(true, &cpu.cpsr, cpu.r[rd], @truncate(u8, cpu.r[rs]));
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x3 => {
-                    // LSR
-                    const result = logicalRight(true, &cpu.cpsr, cpu.r[rd], @truncate(u8, cpu.r[rs]));
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x4 => {
-                    // ASR
-                    const result = arithmeticRight(true, &cpu.cpsr, cpu.r[rd], @truncate(u8, cpu.r[rs]));
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x5 => {
-                    // ADC
-                    cpu.r[rd] = adc(true, cpu, cpu.r[rd], cpu.r[rs], carry);
-                },
-                0x6 => {
-                    // SBC
-                    cpu.r[rd] = sbc(true, cpu, cpu.r[rd], cpu.r[rs], carry);
-                },
-                0x7 => {
-                    // ROR
-                    const result = rotateRight(true, &cpu.cpsr, cpu.r[rd], @truncate(u8, cpu.r[rs]));
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x8 => {
-                    // TST
-                    const result = cpu.r[rd] & cpu.r[rs];
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0x9 => {
-                    // NEG
-                    cpu.r[rd] = sub(true, cpu, 0, cpu.r[rs]);
-                },
-                0xA => {
-                    // CMP
-                    cmp(cpu, cpu.r[rd], cpu.r[rs]);
-                },
-                0xB => {
-                    // CMN
-                    cmn(cpu, cpu.r[rd], cpu.r[rs]);
-                },
-                0xC => {
-                    // ORR
-                    const result = cpu.r[rd] | cpu.r[rs];
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0xD => {
-                    // MUL
-                    const temp = @as(u64, cpu.r[rs]) * @as(u64, cpu.r[rd]);
-                    const result = @truncate(u32, temp);
-                    cpu.r[rd] = result;
-
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    // V is unaffected, assuming similar behaviour to ARMv4 MUL C is undefined
-                },
-                0xE => {
-                    // BIC
-                    const result = cpu.r[rd] & ~cpu.r[rs];
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-                0xF => {
-                    // MVN
-                    const result = ~cpu.r[rs];
-                    cpu.r[rd] = result;
-                    setLogicOpFlags(true, cpu, result);
-                },
-            }
-        }
-    }.inner;
-}

src/core/cpu/thumb/block_data_transfer.zig

@@ -1,94 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-
-pub fn fmt14(comptime L: bool, comptime R: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            const count = @boolToInt(R) + countRlist(opcode);
-            const start = cpu.r[13] - if (!L) count * 4 else 0;
-
-            var end = cpu.r[13];
-            if (L) {
-                end += count * 4;
-            } else {
-                end -= 4;
-            }
-
-            var address = start;
-
-            var i: u4 = 0;
-            while (i < 8) : (i += 1) {
-                if (opcode >> i & 1 == 1) {
-                    if (L) {
-                        cpu.r[i] = bus.read(u32, address);
-                    } else {
-                        bus.write(u32, address, cpu.r[i]);
-                    }
-
-                    address += 4;
-                }
-            }
-
-            if (R) {
-                if (L) {
-                    const value = bus.read(u32, address);
-                    cpu.r[15] = value & 0xFFFF_FFFE;
-                } else {
-                    bus.write(u32, address, cpu.r[14]);
-                }
-                address += 4;
-            }
-
-            cpu.r[13] = if (L) end else start;
-        }
-    }.inner;
-}
-
-pub fn fmt15(comptime L: bool, comptime rb: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            var address = cpu.r[rb];
-            const end_address = cpu.r[rb] + 4 * countRlist(opcode);
-
-            if (opcode & 0xFF == 0) {
-                if (L) cpu.r[15] = bus.read(u32, address) else bus.write(u32, address, cpu.r[15] + 4);
-                cpu.r[rb] += 0x40;
-                return;
-            }
-
-            var i: u4 = 0;
-            var first_write = true;
-
-            while (i < 8) : (i += 1) {
-                if (opcode >> i & 1 == 1) {
-                    if (L) {
-                        cpu.r[i] = bus.read(u32, address);
-                    } else {
-                        bus.write(u32, address, cpu.r[i]);
-                    }
-
-                    if (!L and first_write) {
-                        cpu.r[rb] = end_address;
-                        first_write = false;
-                    }
-
-                    address += 4;
-                }
-            }
-
-            if (L and opcode >> rb & 1 != 1) cpu.r[rb] = address;
-        }
-    }.inner;
-}
-
-inline fn countRlist(opcode: u16) u32 {
-    var count: u32 = 0;
-
-    comptime var i: u4 = 0;
-    inline while (i < 8) : (i += 1) {
-        if (opcode >> (7 - i) & 1 == 1) count += 1;
-    }
-
-    return count;
-}

src/core/cpu/thumb/branch.zig

@@ -1,54 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-
-const checkCond = @import("../../cpu.zig").checkCond;
-const sext = @import("../../../util.zig").sext;
-
-pub fn fmt16(comptime cond: u4) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            // B
-            const offset = sext(u32, u8, opcode & 0xFF) << 1;
-
-            const should_execute = switch (cond) {
-                0xE, 0xF => cpu.panic("[CPU/THUMB.16] Undefined conditional branch with condition {}", .{cond}),
-                else => checkCond(cpu.cpsr, cond),
-            };
-
-            if (should_execute) {
-                cpu.r[15] = (cpu.r[15] + 2) +% offset;
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt18() InstrFn {
-    return struct {
-        // B but conditional
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const offset = sext(u32, u11, opcode & 0x7FF) << 1;
-            cpu.r[15] = (cpu.r[15] + 2) +% offset;
-        }
-    }.inner;
-}
-
-pub fn fmt19(comptime is_low: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            // BL
-            const offset = opcode & 0x7FF;
-
-            if (is_low) {
-                // Instruction 2
-                const old_pc = cpu.r[15];
-
-                cpu.r[15] = cpu.r[14] +% (offset << 1);
-                cpu.r[14] = old_pc | 1;
-            } else {
-                // Instruction 1
-                cpu.r[14] = (cpu.r[15] + 2) +% (sext(u32, u11, offset) << 12);
-            }
-        }
-    }.inner;
-}

src/core/cpu/thumb/data_processing.zig

@@ -1,152 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-const shifter = @import("../barrel_shifter.zig");
-
-const add = @import("../arm/data_processing.zig").add;
-const sub = @import("../arm/data_processing.zig").sub;
-const cmp = @import("../arm/data_processing.zig").cmp;
-const setLogicOpFlags = @import("../arm/data_processing.zig").setLogicOpFlags;
-
-const log = std.log.scoped(.Thumb1);
-
-pub fn fmt1(comptime op: u2, comptime offset: u5) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const rs = opcode >> 3 & 0x7;
-            const rd = opcode & 0x7;
-
-            const result = switch (op) {
-                0b00 => blk: {
-                    // LSL
-                    if (offset == 0) {
-                        break :blk cpu.r[rs];
-                    } else {
-                        break :blk shifter.logicalLeft(true, &cpu.cpsr, cpu.r[rs], offset);
-                    }
-                },
-                0b01 => blk: {
-                    // LSR
-                    if (offset == 0) {
-                        cpu.cpsr.c.write(cpu.r[rs] >> 31 & 1 == 1);
-                        break :blk @as(u32, 0);
-                    } else {
-                        break :blk shifter.logicalRight(true, &cpu.cpsr, cpu.r[rs], offset);
-                    }
-                },
-                0b10 => blk: {
-                    // ASR
-                    if (offset == 0) {
-                        cpu.cpsr.c.write(cpu.r[rs] >> 31 & 1 == 1);
-                        break :blk @bitCast(u32, @bitCast(i32, cpu.r[rs]) >> 31);
-                    } else {
-                        break :blk shifter.arithmeticRight(true, &cpu.cpsr, cpu.r[rs], offset);
-                    }
-                },
-                else => cpu.panic("[CPU/THUMB.1] 0b{b:0>2} is not a valid op", .{op}),
-            };
-
-            // Equivalent to an ARM MOVS
-            cpu.r[rd] = result;
-            setLogicOpFlags(true, cpu, result);
-        }
-    }.inner;
-}
-
-pub fn fmt5(comptime op: u2, comptime h1: u1, comptime h2: u1) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const src_idx = @as(u4, h2) << 3 | (opcode >> 3 & 0x7);
-            const dst_idx = @as(u4, h1) << 3 | (opcode & 0x7);
-
-            const src = if (src_idx == 0xF) (cpu.r[src_idx] + 2) & 0xFFFF_FFFE else cpu.r[src_idx];
-            const dst = if (dst_idx == 0xF) (cpu.r[dst_idx] + 2) & 0xFFFF_FFFE else cpu.r[dst_idx];
-
-            switch (op) {
-                0b00 => {
-                    // ADD
-                    const sum = add(false, cpu, dst, src);
-                    cpu.r[dst_idx] = if (dst_idx == 0xF) sum & 0xFFFF_FFFE else sum;
-                },
-                0b01 => cmp(cpu, dst, src), // CMP
-                0b10 => {
-                    // MOV
-                    cpu.r[dst_idx] = if (dst_idx == 0xF) src & 0xFFFF_FFFE else src;
-                },
-                0b11 => {
-                    // BX
-                    cpu.cpsr.t.write(src & 1 == 1);
-                    cpu.r[15] = src & 0xFFFF_FFFE;
-                },
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt2(comptime I: bool, is_sub: bool, rn: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const rs = opcode >> 3 & 0x7;
-            const rd = @truncate(u3, opcode);
-
-            if (is_sub) {
-                // SUB
-                cpu.r[rd] = if (I) blk: {
-                    break :blk sub(true, cpu, cpu.r[rs], rn);
-                } else blk: {
-                    break :blk sub(true, cpu, cpu.r[rs], cpu.r[rn]);
-                };
-            } else {
-                // ADD
-                cpu.r[rd] = if (I) blk: {
-                    break :blk add(true, cpu, cpu.r[rs], rn);
-                } else blk: {
-                    break :blk add(true, cpu, cpu.r[rs], cpu.r[rn]);
-                };
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt3(comptime op: u2, comptime rd: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const offset = @truncate(u8, opcode);
-
-            switch (op) {
-                0b00 => {
-                    // MOV
-                    cpu.r[rd] = offset;
-                    setLogicOpFlags(true, cpu, offset);
-                },
-                0b01 => cmp(cpu, cpu.r[rd], offset), // CMP
-                0b10 => cpu.r[rd] = add(true, cpu, cpu.r[rd], offset), // ADD
-                0b11 => cpu.r[rd] = sub(true, cpu, cpu.r[rd], offset), // SUB
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt12(comptime isSP: bool, comptime rd: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            // ADD
-            const left = if (isSP) cpu.r[13] else (cpu.r[15] + 2) & 0xFFFF_FFFD;
-            const right = (opcode & 0xFF) << 2;
-            const result = left + right;
-            cpu.r[rd] = result;
-        }
-    }.inner;
-}
-
-pub fn fmt13(comptime S: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            // ADD
-            const offset = (opcode & 0x7F) << 2;
-            cpu.r[13] = if (S) cpu.r[13] - offset else cpu.r[13] + offset;
-        }
-    }.inner;
-}

src/core/cpu/thumb/data_transfer.zig

@@ -1,148 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-
-const rotr = @import("../../../util.zig").rotr;
-const sext = @import("../../../util.zig").sext;
-
-pub fn fmt6(comptime rd: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            // LDR
-            const offset = (opcode & 0xFF) << 2;
-            cpu.r[rd] = bus.read(u32, (cpu.r[15] + 2 & 0xFFFF_FFFD) + offset);
-        }
-    }.inner;
-}
-
-pub fn fmt78(comptime op: u2, comptime T: bool) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            const ro = opcode >> 6 & 0x7;
-            const rb = opcode >> 3 & 0x7;
-            const rd = opcode & 0x7;
-
-            const address = cpu.r[rb] +% cpu.r[ro];
-
-            if (T) {
-                // Format 8
-                switch (op) {
-                    0b00 => {
-                        // STRH
-                        bus.write(u16, address, @truncate(u16, cpu.r[rd]));
-                    },
-                    0b01 => {
-                        // LDSB
-                        cpu.r[rd] = sext(u32, u8, bus.read(u8, address));
-                    },
-                    0b10 => {
-                        // LDRH
-                        const value = bus.read(u16, address);
-                        cpu.r[rd] = rotr(u32, value, 8 * (address & 1));
-                    },
-                    0b11 => {
-                        // LDRSH
-                        cpu.r[rd] = if (address & 1 == 1) blk: {
-                            break :blk sext(u32, u8, bus.read(u8, address));
-                        } else blk: {
-                            break :blk sext(u32, u16, bus.read(u16, address));
-                        };
-                    },
-                }
-            } else {
-                // Format 7
-                switch (op) {
-                    0b00 => {
-                        // STR
-                        bus.write(u32, address, cpu.r[rd]);
-                    },
-                    0b01 => {
-                        // STRB
-                        bus.write(u8, address, @truncate(u8, cpu.r[rd]));
-                    },
-                    0b10 => {
-                        // LDR
-                        const value = bus.read(u32, address);
-                        cpu.r[rd] = rotr(u32, value, 8 * (address & 0x3));
-                    },
-                    0b11 => {
-                        // LDRB
-                        cpu.r[rd] = bus.read(u8, address);
-                    },
-                }
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt9(comptime B: bool, comptime L: bool, comptime offset: u5) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            const rb = opcode >> 3 & 0x7;
-            const rd = opcode & 0x7;
-
-            if (L) {
-                if (B) {
-                    // LDRB
-                    const address = cpu.r[rb] + offset;
-                    cpu.r[rd] = bus.read(u8, address);
-                } else {
-                    // LDR
-                    const address = cpu.r[rb] + (@as(u32, offset) << 2);
-                    const value = bus.read(u32, address);
-                    cpu.r[rd] = rotr(u32, value, 8 * (address & 0x3));
-                }
-            } else {
-                if (B) {
-                    // STRB
-                    const address = cpu.r[rb] + offset;
-                    bus.write(u8, address, @truncate(u8, cpu.r[rd]));
-                } else {
-                    // STR
-                    const address = cpu.r[rb] + (@as(u32, offset) << 2);
-                    bus.write(u32, address, cpu.r[rd]);
-                }
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt10(comptime L: bool, comptime offset: u5) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            const rb = opcode >> 3 & 0x7;
-            const rd = opcode & 0x7;
-
-            const address = cpu.r[rb] + (@as(u6, offset) << 1);
-
-            if (L) {
-                // LDRH
-                const value = bus.read(u16, address);
-                cpu.r[rd] = rotr(u32, value, 8 * (address & 1));
-            } else {
-                // STRH
-                bus.write(u16, address, @truncate(u16, cpu.r[rd]));
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt11(comptime L: bool, comptime rd: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, bus: *Bus, opcode: u16) void {
-            const offset = (opcode & 0xFF) << 2;
-            const address = cpu.r[13] + offset;
-
-            if (L) {
-                // LDR
-                const value = bus.read(u32, address);
-                cpu.r[rd] = rotr(u32, value, 8 * (address & 0x3));
-            } else {
-                // STR
-                bus.write(u32, address, cpu.r[rd]);
-            }
-        }
-    }.inner;
-}

src/core/cpu/thumb/software_interrupt.zig

@@ -1,22 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-
-pub fn fmt17() InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, _: u16) void {
-            // Copy Values from Current Mode
-            const r15 = cpu.r[15];
-            const cpsr = cpu.cpsr.raw;
-
-            // Switch Mode
-            cpu.changeMode(.Supervisor);
-            cpu.cpsr.t.write(false); // Force ARM Mode
-            cpu.cpsr.i.write(true); // Disable normal interrupts
-
-            cpu.r[14] = r15; // Resume Execution
-            cpu.spsr.raw = cpsr; // Previous mode CPSR
-            cpu.r[15] = 0x0000_0008;
-        }
-    }.inner;
-}

src/core/cpu_util.zig

@@ -0,0 +1,75 @@
+const std = @import("std");
+
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bank = @import("arm32").Arm7tdmi.Bank;
+const Bus = @import("Bus.zig");
+
+pub inline fn isHalted(cpu: *const Arm7tdmi) bool {
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+    return bus_ptr.io.haltcnt == .Halt;
+}
+
+pub fn stepDmaTransfer(cpu: *Arm7tdmi) bool {
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+    inline for (0..4) |i| {
+        if (bus_ptr.dma[i].in_progress) {
+            bus_ptr.dma[i].step(cpu);
+            return true;
+        }
+    }
+
+    return false;
+}
+
+pub fn handleInterrupt(cpu: *Arm7tdmi) void {
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+    const should_handle = bus_ptr.io.ie.raw & bus_ptr.io.irq.raw;
+
+    // Return if IME is disabled, CPSR I is set or there is nothing to handle
+    if (!bus_ptr.io.ime or cpu.cpsr.i.read() or should_handle == 0) return;
+
+    // If Pipeline isn't full, we have a bug
+    std.debug.assert(cpu.pipe.isFull());
+
+    // log.debug("Handling Interrupt!", .{});
+    bus_ptr.io.haltcnt = .Execute;
+
+    // FIXME: This seems weird, but retAddr.gba suggests I need to make these changes
+    const ret_addr = cpu.r[15] - if (cpu.cpsr.t.read()) 0 else @as(u32, 4);
+    const new_spsr = cpu.cpsr.raw;
+
+    cpu.changeMode(.Irq);
+    cpu.cpsr.t.write(false);
+    cpu.cpsr.i.write(true);
+
+    cpu.r[14] = ret_addr;
+    cpu.spsr.raw = new_spsr;
+    cpu.r[15] = 0x0000_0018;
+    cpu.pipe.reload(cpu);
+}
+
+/// Advances state so that the BIOS is skipped
+///
+/// Note: This accesses the CPU's bus ptr so it only may be called
+/// once the Bus has been properly initialized
+///
+/// TODO: Make above notice impossible to do in code
+pub fn fastBoot(cpu: *Arm7tdmi) void {
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+    cpu.r = std.mem.zeroes([16]u32);
+
+    // cpu.r[0] = 0x08000000;
+    // cpu.r[1] = 0x000000EA;
+    cpu.r[13] = 0x0300_7F00;
+    cpu.r[15] = 0x0800_0000;
+
+    cpu.bank.r[Bank.regIdx(.Irq, .R13)] = 0x0300_7FA0;
+    cpu.bank.r[Bank.regIdx(.Supervisor, .R13)] = 0x0300_7FE0;
+
+    // cpu.cpsr.raw = 0x6000001F;
+    cpu.cpsr.raw = 0x0000_001F;
+
+    bus_ptr.bios.addr_latch = 0x0000_00DC + 8;
+}

src/core/emu.zig

@@ -2,29 +2,59 @@ const std = @import("std");
 const SDL = @import("sdl2");
 const config = @import("../config.zig");
 
-const Bus = @import("Bus.zig");
 const Scheduler = @import("scheduler.zig").Scheduler;
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
-const FpsTracker = @import("../util.zig").FpsTracker;
-const FilePaths = @import("../util.zig").FilePaths;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("Bus.zig");
+const Tracker = @import("../util.zig").FpsTracker;
+const Channel = @import("../util.zig").Queue;
+
+const stepDmaTransfer = @import("cpu_util.zig").stepDmaTransfer;
+const isHalted = @import("cpu_util.zig").isHalted;
 
 const Timer = std.time.Timer;
-const Thread = std.Thread;
-const Atomic = std.atomic.Atomic;
-const Allocator = std.mem.Allocator;
 
-// 228 Lines which consist of 308 dots (which are 4 cycles long)
-const cycles_per_frame: u64 = 228 * (308 * 4); //280896
-const clock_rate: u64 = 1 << 24; // 16.78MHz
+pub const Synchro = struct {
+    const AtomicBool = std.atomic.Value(bool);
 
-// TODO: Don't truncate this, be more accurate w/ timing
-// 59.6046447754ns (truncated to just 59ns)
-const clock_period: u64 = std.time.ns_per_s / clock_rate;
-const frame_period = (clock_period * cycles_per_frame);
+    // FIXME: This Enum ends up being really LARGE!!!
+    pub const Message = union(enum) {
+        rom_path: [std.fs.MAX_PATH_BYTES]u8,
+        bios_path: [std.fs.MAX_PATH_BYTES]u8,
+        restart: void,
+    };
 
-// 59.7275005696Hz
-pub const frame_rate = @intToFloat(f64, std.time.ns_per_s) /
-    ((@intToFloat(f64, std.time.ns_per_s) / @intToFloat(f64, clock_rate)) * @intToFloat(f64, cycles_per_frame));
+    paused: AtomicBool = AtomicBool.init(true), // FIXME: can ui_busy and paused be the same?
+    should_quit: AtomicBool = AtomicBool.init(false),
+
+    ch: Channel(Message),
+
+    pub fn init(allocator: std.mem.Allocator) !@This() {
+        const msg_buf = try allocator.alloc(Message, 1);
+        return .{ .ch = Channel(Message).init(msg_buf) };
+    }
+
+    pub fn deinit(self: *@This(), allocator: std.mem.Allocator) void {
+        allocator.free(self.ch.inner.buf);
+        self.* = undefined;
+    }
+};
+
+/// 4 Cycles in 1 dot
+const cycles_per_dot = 4;
+
+/// The GBA draws 228 Horizontal which each consist 308 dots
+/// (note: not all lines are visible)
+const cycles_per_frame = 228 * (308 * cycles_per_dot); //280896
+
+/// The GBA ARM7TDMI runs at 2^24 Hz
+const clock_rate = 1 << 24; // 16.78MHz
+
+/// The # of nanoseconds a frame should take
+const frame_period = (std.time.ns_per_s * cycles_per_frame) / clock_rate;
+
+/// Exact Value:  59.7275005696Hz
+/// The inverse of the frame period
+pub const frame_rate: f64 = @as(f64, @floatFromInt(clock_rate)) / cycles_per_frame;
 
 const log = std.log.scoped(.Emulation);
 
@@ -35,30 +65,37 @@ const RunKind = enum {
     LimitedFPS,
 };
 
-pub fn run(quit: *Atomic(bool), scheduler: *Scheduler, cpu: *Arm7tdmi, tracker: *FpsTracker) void {
-    const audio_sync = config.config().guest.audio_sync;
+pub fn run(cpu: *Arm7tdmi, scheduler: *Scheduler, tracker: *Tracker, sync: *Synchro) void {
+    const audio_sync = config.config().guest.audio_sync and !config.config().host.mute;
     if (audio_sync) log.info("Audio sync enabled", .{});
 
     if (config.config().guest.video_sync) {
-        inner(.LimitedFPS, audio_sync, quit, scheduler, cpu, tracker);
+        inner(.LimitedFPS, audio_sync, cpu, scheduler, tracker, sync);
     } else {
-        inner(.UnlimitedFPS, audio_sync, quit, scheduler, cpu, tracker);
+        inner(.UnlimitedFPS, audio_sync, cpu, scheduler, tracker, sync);
     }
 }
 
-fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), scheduler: *Scheduler, cpu: *Arm7tdmi, tracker: ?*FpsTracker) void {
+fn inner(comptime kind: RunKind, audio_sync: bool, cpu: *Arm7tdmi, scheduler: *Scheduler, tracker: ?*Tracker, sync: *Synchro) void {
     if (kind == .UnlimitedFPS or kind == .LimitedFPS) {
         std.debug.assert(tracker != null);
         log.info("FPS tracking enabled", .{});
     }
 
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+    // FIXME: audioSync accesses emulator state without any guarantees
+
     switch (kind) {
         .Unlimited, .UnlimitedFPS => {
             log.info("Emulation w/out video sync", .{});
 
-            while (!quit.load(.SeqCst)) {
+            while (!sync.should_quit.load(.monotonic)) {
+                handleChannel(cpu, &sync.ch);
+                if (sync.paused.load(.monotonic)) continue;
+
                 runFrame(scheduler, cpu);
-                audioSync(audio_sync, cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
+                audioSync(audio_sync, bus_ptr.apu.stream, &bus_ptr.apu.is_buffer_full);
 
                 if (kind == .UnlimitedFPS) tracker.?.tick();
             }
@@ -68,7 +105,10 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), schedule
             var timer = Timer.start() catch @panic("failed to initalize std.timer.Timer");
             var wake_time: u64 = frame_period;
 
-            while (!quit.load(.SeqCst)) {
+            while (!sync.should_quit.load(.monotonic)) {
+                handleChannel(cpu, &sync.ch);
+                if (sync.paused.load(.monotonic)) continue;
+
                 runFrame(scheduler, cpu);
                 const new_wake_time = videoSync(&timer, wake_time);
 
@@ -77,7 +117,7 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), schedule
                 // the amount of time needed for audio to catch up rather than
                 // our expected wake-up time
 
-                audioSync(audio_sync, cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
+                audioSync(audio_sync, bus_ptr.apu.stream, &bus_ptr.apu.is_buffer_full);
                 if (!audio_sync) spinLoop(&timer, wake_time);
                 wake_time = new_wake_time;
 
@@ -87,14 +127,30 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), schedule
     }
 }
 
+inline fn handleChannel(cpu: *Arm7tdmi, channel: *Channel(Synchro.Message)) void {
+    const message = channel.pop() orelse return;
+
+    switch (message) {
+        .rom_path => |path_buf| {
+            const path = std.mem.sliceTo(&path_buf, 0);
+            replaceGamepak(cpu, path) catch |e| log.err("failed to replace GamePak: {}", .{e});
+        },
+        .bios_path => |path_buf| {
+            const path = std.mem.sliceTo(&path_buf, 0);
+            replaceBios(cpu, path) catch |e| log.err("failed to replace BIOS: {}", .{e});
+        },
+        .restart => reset(cpu),
+    }
+}
+
 pub fn runFrame(sched: *Scheduler, cpu: *Arm7tdmi) void {
     const frame_end = sched.tick + cycles_per_frame;
 
     while (sched.tick < frame_end) {
-        if (!cpu.stepDmaTransfer()) {
-            if (cpu.isHalted()) {
+        if (!stepDmaTransfer(cpu)) {
+            if (isHalted(cpu)) {
                 // Fast-forward to next Event
-                sched.tick = sched.queue.peek().?.tick;
+                sched.tick = sched.nextTimestamp();
             } else {
                 cpu.step();
             }
@@ -105,6 +161,7 @@ pub fn runFrame(sched: *Scheduler, cpu: *Arm7tdmi) void {
 }
 
 fn audioSync(audio_sync: bool, stream: *SDL.SDL_AudioStream, is_buffer_full: *bool) void {
+    comptime std.debug.assert(@import("../platform.zig").sample_format == SDL.AUDIO_U16);
     const sample_size = 2 * @sizeOf(u16);
     const max_buf_size: c_int = 0x400;
 
@@ -115,6 +172,10 @@ fn audioSync(audio_sync: bool, stream: *SDL.SDL_AudioStream, is_buffer_full: *bo
     // If Busy is false, there's no need to sync here
     if (!still_full) return;
 
+    // TODO: Refactor!!!!
+    // while (SDL.SDL_AudioStreamAvailable(stream) > sample_size * max_buf_size >> 1)
+    //     std.atomic.spinLoopHint();
+
     while (true) {
         still_full = SDL.SDL_AudioStreamAvailable(stream) > sample_size * max_buf_size >> 1;
         if (!audio_sync or !still_full) break;
@@ -132,11 +193,10 @@ fn videoSync(timer: *Timer, wake_time: u64) u64 {
 
 // TODO: Better sleep impl?
 fn sleep(timer: *Timer, wake_time: u64) ?u64 {
-    // const step = std.time.ns_per_ms * 10; // 10ms
     const timestamp = timer.read();
 
     // ns_late is non zero if we are late.
-    const ns_late = timestamp -| wake_time;
+    var ns_late = timestamp -| wake_time;
 
     // If we're more than a frame late, skip the rest of this loop
     // Recalculate what our new wake time should be so that we can
@@ -144,19 +204,145 @@ fn sleep(timer: *Timer, wake_time: u64) ?u64 {
     if (ns_late > frame_period) return timestamp + frame_period;
     const sleep_for = frame_period - ns_late;
 
-    // // Employ several sleep calls in periods of 10ms
-    // // By doing this the behaviour should average out to be
-    // // more consistent
-    // const loop_count = sleep_for / step; // How many groups of 10ms
+    const step = 2 * std.time.ns_per_ms; // Granularity of 2ms
+    const times = sleep_for / step;
 
-    // var i: usize = 0;
-    // while (i < loop_count) : (i += 1) std.time.sleep(step);
+    for (0..times) |_| {
+        std.time.sleep(step);
 
-    std.time.sleep(sleep_for);
+        // Upon wakeup, check to see if this particular sleep was longer than expected
+        // if so we should exit early, but probably not skip a whole frame period
+        ns_late = timer.read() -| wake_time;
+        if (ns_late > frame_period) return null;
+    }
 
     return null;
 }
 
 fn spinLoop(timer: *Timer, wake_time: u64) void {
-    while (true) if (timer.read() > wake_time) break;
+    while (timer.read() < wake_time)
+        std.atomic.spinLoopHint();
+}
+
+pub const EmuThing = struct {
+    const Self = @This();
+    const Interface = @import("gdbstub").Emulator;
+    const Allocator = std.mem.Allocator;
+
+    pub const target =
+        \\<target version="1.0">
+        \\    <architecture>armv4t</architecture>
+        \\    <feature name="org.gnu.gdb.arm.core">
+        \\        <reg name="r0" bitsize="32" type="uint32"/>
+        \\        <reg name="r1" bitsize="32" type="uint32"/>
+        \\        <reg name="r2" bitsize="32" type="uint32"/>
+        \\        <reg name="r3" bitsize="32" type="uint32"/>
+        \\        <reg name="r4" bitsize="32" type="uint32"/>
+        \\        <reg name="r5" bitsize="32" type="uint32"/>
+        \\        <reg name="r6" bitsize="32" type="uint32"/>
+        \\        <reg name="r7" bitsize="32" type="uint32"/>
+        \\        <reg name="r8" bitsize="32" type="uint32"/>
+        \\        <reg name="r9" bitsize="32" type="uint32"/>
+        \\        <reg name="r10" bitsize="32" type="uint32"/>
+        \\        <reg name="r11" bitsize="32" type="uint32"/>
+        \\        <reg name="r12" bitsize="32" type="uint32"/>
+        \\        <reg name="sp" bitsize="32" type="data_ptr"/>
+        \\        <reg name="lr" bitsize="32"/>
+        \\        <reg name="pc" bitsize="32" type="code_ptr"/>
+        \\
+        \\        <reg name="cpsr" bitsize="32" regnum="25"/>
+        \\    </feature>
+        \\</target>
+    ;
+
+    // Game Pak SRAM isn't included
+    // TODO: Can i be more specific here?
+    pub const map =
+        \\ <memory-map version="1.0">
+        \\     <memory type="rom" start="0x00000000" length="0x00004000"/>
+        \\     <memory type="ram" start="0x02000000" length="0x00040000"/>
+        \\     <memory type="ram" start="0x03000000" length="0x00008000"/>
+        \\     <memory type="ram" start="0x04000000" length="0x00000400"/>
+        \\     <memory type="ram" start="0x05000000" length="0x00000400"/>
+        \\     <memory type="ram" start="0x06000000" length="0x00018000"/>
+        \\     <memory type="ram" start="0x07000000" length="0x00000400"/>
+        \\     <memory type="rom" start="0x08000000" length="0x02000000"/>
+        \\     <memory type="rom" start="0x0A000000" length="0x02000000"/>
+        \\     <memory type="rom" start="0x0C000000" length="0x02000000"/>
+        \\ </memory-map>
+    ;
+
+    cpu: *Arm7tdmi,
+    scheduler: *Scheduler,
+
+    pub fn init(cpu: *Arm7tdmi, scheduler: *Scheduler) Self {
+        return .{ .cpu = cpu, .scheduler = scheduler };
+    }
+
+    pub fn interface(self: *Self, allocator: Allocator) Interface {
+        return Interface.init(allocator, self);
+    }
+
+    pub fn read(self: *const Self, addr: u32) u8 {
+        return self.cpu.bus.dbgRead(u8, addr);
+    }
+
+    pub fn write(self: *Self, addr: u32, value: u8) void {
+        self.cpu.bus.dbgWrite(u8, addr, value);
+    }
+
+    pub fn registers(self: *const Self) *[16]u32 {
+        return &self.cpu.r;
+    }
+
+    pub fn cpsr(self: *const Self) u32 {
+        return self.cpu.cpsr.raw;
+    }
+
+    pub fn step(self: *Self) void {
+        const cpu = self.cpu;
+        const sched = self.scheduler;
+
+        // Is true when we have executed one (1) instruction
+        var did_step: bool = false;
+
+        // TODO: How can I make it easier to keep this in lock-step with runFrame?
+        while (!did_step) {
+            if (!stepDmaTransfer(cpu)) {
+                if (isHalted(cpu)) {
+                    // Fast-forward to next Event
+                    sched.tick = sched.queue.peek().?.tick;
+                } else {
+                    cpu.step();
+                    did_step = true;
+                }
+            }
+
+            if (sched.tick >= sched.nextTimestamp()) sched.handleEvent(cpu);
+        }
+    }
+};
+
+fn reset(cpu: *Arm7tdmi) void {
+    // @breakpoint();
+    cpu.sched.reset(); // Yes this is order sensitive, see the PPU reset for why
+    cpu.bus.reset();
+    cpu.reset();
+}
+
+fn replaceGamepak(cpu: *Arm7tdmi, file_path: []const u8) !void {
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+    try bus_ptr.replaceGamepak(file_path);
+    reset(cpu);
+}
+
+fn replaceBios(cpu: *Arm7tdmi, file_path: []const u8) !void {
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+    const allocator = bus_ptr.bios.allocator;
+    const bios_len = 0x4000;
+
+    bus_ptr.bios.buf = try allocator.alloc(u8, bios_len);
+    try bus_ptr.bios.load(file_path);
 }

src/core/ppu/Oam.zig

@@ -0,0 +1,44 @@
+const std = @import("std");
+
+const Allocator = std.mem.Allocator;
+
+const buf_len = 0x400;
+const Self = @This();
+
+buf: []u8,
+allocator: Allocator,
+
+pub fn read(self: *const Self, comptime T: type, address: usize) T {
+    const addr = address & 0x3FF;
+
+    return switch (T) {
+        u32, u16, u8 => std.mem.readInt(T, self.buf[addr..][0..@sizeOf(T)], .little),
+        else => @compileError("OAM: Unsupported read width"),
+    };
+}
+
+pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
+    const addr = address & 0x3FF;
+
+    switch (T) {
+        u32, u16 => std.mem.writeInt(T, self.buf[addr..][0..@sizeOf(T)], value, .little),
+        u8 => return, // 8-bit writes are explicitly ignored
+        else => @compileError("OAM: Unsupported write width"),
+    }
+}
+
+pub fn init(allocator: Allocator) !Self {
+    const buf = try allocator.alloc(u8, buf_len);
+    @memset(buf, 0);
+
+    return Self{ .buf = buf, .allocator = allocator };
+}
+
+pub fn reset(self: *Self) void {
+    @memset(self.buf, 0);
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.buf);
+    self.* = undefined;
+}

src/core/ppu/Palette.zig

@@ -0,0 +1,51 @@
+const std = @import("std");
+
+const Allocator = std.mem.Allocator;
+
+const buf_len = 0x400;
+const Self = @This();
+
+buf: []u8,
+allocator: Allocator,
+
+pub fn read(self: *const Self, comptime T: type, address: usize) T {
+    const addr = address & 0x3FF;
+
+    return switch (T) {
+        u32, u16, u8 => std.mem.readInt(T, self.buf[addr..][0..@sizeOf(T)], .little),
+        else => @compileError("PALRAM: Unsupported read width"),
+    };
+}
+
+pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
+    const addr = address & 0x3FF;
+
+    switch (T) {
+        u32, u16 => std.mem.writeInt(T, self.buf[addr..][0..@sizeOf(T)], value, .little),
+        u8 => {
+            const align_addr = addr & ~@as(u32, 1); // Aligned to Halfword boundary
+            std.mem.writeInt(u16, self.buf[align_addr..][0..@sizeOf(u16)], @as(u16, value) * 0x101, .little);
+        },
+        else => @compileError("PALRAM: Unsupported write width"),
+    }
+}
+
+pub fn init(allocator: Allocator) !Self {
+    const buf = try allocator.alloc(u8, buf_len);
+    @memset(buf, 0);
+
+    return Self{ .buf = buf, .allocator = allocator };
+}
+
+pub fn reset(self: *Self) void {
+    @memset(self.buf, 0);
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.buf);
+    self.* = undefined;
+}
+
+pub inline fn backdrop(self: *const Self) u16 {
+    return std.mem.readInt(u16, self.buf[0..2], .little);
+}

src/core/ppu/Vram.zig

@@ -0,0 +1,64 @@
+const std = @import("std");
+const io = @import("../bus/io.zig");
+
+const Allocator = std.mem.Allocator;
+
+const buf_len = 0x18000;
+const Self = @This();
+
+buf: []u8,
+allocator: Allocator,
+
+pub fn read(self: *const Self, comptime T: type, address: usize) T {
+    const addr = Self.mirror(address);
+
+    return switch (T) {
+        u32, u16, u8 => std.mem.readInt(T, self.buf[addr..][0..@sizeOf(T)], .little),
+        else => @compileError("VRAM: Unsupported read width"),
+    };
+}
+
+pub fn write(self: *Self, comptime T: type, dispcnt: io.DisplayControl, address: usize, value: T) void {
+    const mode: u3 = dispcnt.bg_mode.read();
+    const idx = Self.mirror(address);
+
+    switch (T) {
+        u32, u16 => std.mem.writeInt(T, self.buf[idx..][0..@sizeOf(T)], value, .little),
+        u8 => {
+            // Ignore write if it falls within the boundaries of OBJ VRAM
+            switch (mode) {
+                0, 1, 2 => if (0x0001_0000 <= idx) return,
+                else => if (0x0001_4000 <= idx) return,
+            }
+
+            const align_idx = idx & ~@as(u32, 1); // Aligned to a halfword boundary
+            std.mem.writeInt(u16, self.buf[align_idx..][0..@sizeOf(u16)], @as(u16, value) * 0x101, .little);
+        },
+        else => @compileError("VRAM: Unsupported write width"),
+    }
+}
+
+pub fn init(allocator: Allocator) !Self {
+    const buf = try allocator.alloc(u8, buf_len);
+    @memset(buf, 0);
+
+    return Self{ .buf = buf, .allocator = allocator };
+}
+
+pub fn reset(self: *Self) void {
+    @memset(self.buf, 0);
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.buf);
+    self.* = undefined;
+}
+
+pub fn mirror(address: usize) usize {
+    // Mirrored in steps of 128K (64K + 32K + 32K) (abcc)
+    const addr = address & 0x1FFFF;
+
+    // If the address is within 96K we don't do anything,
+    // otherwise we want to mirror the last 32K (addresses between 64K and 96K)
+    return if (addr < buf_len) addr else 0x10000 + (addr & 0x7FFF);
+}

src/core/scheduler.zig

@@ -1,7 +1,7 @@
 const std = @import("std");
 
+const Arm7tdmi = @import("arm32").Arm7tdmi;
 const Bus = @import("Bus.zig");
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
 const Clock = @import("bus/gpio.zig").Clock;
 
 const Order = std.math.Order;
@@ -12,11 +12,11 @@ const log = std.log.scoped(.Scheduler);
 pub const Scheduler = struct {
     const Self = @This();
 
-    tick: u64,
+    tick: u64 = 0,
     queue: PriorityQueue(Event, void, lessThan),
 
     pub fn init(allocator: Allocator) Self {
-        var sched = Self{ .tick = 0, .queue = PriorityQueue(Event, void, lessThan).init(allocator, {}) };
+        var sched = Self{ .queue = PriorityQueue(Event, void, lessThan).init(allocator, {}) };
         sched.queue.add(.{ .kind = .HeatDeath, .tick = std.math.maxInt(u64) }) catch unreachable;
 
         return sched;
@@ -27,69 +27,73 @@ pub const Scheduler = struct {
         self.* = undefined;
     }
 
+    pub fn reset(self: *Self) void {
+        // `std.PriorityQueue` provides no reset function, so we will just create a new one
+        const allocator = self.queue.allocator;
+        self.queue.deinit();
+
+        var new_queue = PriorityQueue(Event, void, lessThan).init(allocator, {});
+        new_queue.add(.{ .kind = .HeatDeath, .tick = std.math.maxInt(u64) }) catch unreachable;
+
+        self.* = .{ .queue = new_queue };
+    }
+
     pub inline fn now(self: *const Self) u64 {
         return self.tick;
     }
 
     pub fn handleEvent(self: *Self, cpu: *Arm7tdmi) void {
-        if (self.queue.removeOrNull()) |event| {
-            const late = self.tick - event.tick;
+        const event = self.queue.remove();
+        const late = self.tick - event.tick;
 
-            switch (event.kind) {
-                .HeatDeath => {
-                    log.err("u64 overflow. This *actually* should never happen.", .{});
-                    unreachable;
-                },
-                .Draw => {
-                    // The end of a VDraw
-                    cpu.bus.ppu.drawScanline();
-                    cpu.bus.ppu.onHdrawEnd(cpu, late);
-                },
-                .TimerOverflow => |id| {
-                    switch (id) {
-                        0 => cpu.bus.tim[0].onTimerExpire(cpu, late),
-                        1 => cpu.bus.tim[1].onTimerExpire(cpu, late),
-                        2 => cpu.bus.tim[2].onTimerExpire(cpu, late),
-                        3 => cpu.bus.tim[3].onTimerExpire(cpu, late),
-                    }
-                },
-                .ApuChannel => |id| {
-                    switch (id) {
-                        0 => cpu.bus.apu.ch1.onToneSweepEvent(late),
-                        1 => cpu.bus.apu.ch2.onToneEvent(late),
-                        2 => cpu.bus.apu.ch3.onWaveEvent(late),
-                        3 => cpu.bus.apu.ch4.onNoiseEvent(late),
-                    }
-                },
-                .RealTimeClock => {
-                    const device = &cpu.bus.pak.gpio.device;
-                    if (device.kind != .Rtc or device.ptr == null) return;
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
 
-                    const clock = @ptrCast(*Clock, @alignCast(@alignOf(*Clock), device.ptr.?));
-                    clock.onClockUpdate(late);
-                },
-                .FrameSequencer => cpu.bus.apu.onSequencerTick(late),
-                .SampleAudio => cpu.bus.apu.sampleAudio(late),
-                .HBlank => cpu.bus.ppu.onHblankEnd(cpu, late), // The end of a HBlank
-                .VBlank => cpu.bus.ppu.onHdrawEnd(cpu, late), // The end of a VBlank
-            }
+        switch (event.kind) {
+            .HeatDeath => {
+                log.err("u64 overflow. This *actually* should never happen.", .{});
+                unreachable;
+            },
+            .Draw => {
+                // The end of a VDraw
+                bus_ptr.ppu.drawScanline();
+                bus_ptr.ppu.onHdrawEnd(cpu, late);
+            },
+            .TimerOverflow => |id| {
+                switch (id) {
+                    inline 0...3 => |idx| bus_ptr.tim[idx].onTimerExpire(cpu, late),
+                }
+            },
+            .ApuChannel => |id| {
+                switch (id) {
+                    0 => bus_ptr.apu.ch1.onToneSweepEvent(late),
+                    1 => bus_ptr.apu.ch2.onToneEvent(late),
+                    2 => bus_ptr.apu.ch3.onWaveEvent(late),
+                    3 => bus_ptr.apu.ch4.onNoiseEvent(late),
+                }
+            },
+            .RealTimeClock => {
+                const device = &bus_ptr.pak.gpio.device;
+                if (device.kind != .Rtc or device.ptr == null) return;
+
+                const clock: *Clock = @ptrCast(@alignCast(device.ptr.?));
+                clock.onClockUpdate(late);
+            },
+            .FrameSequencer => bus_ptr.apu.onSequencerTick(late),
+            .SampleAudio => bus_ptr.apu.sampleAudio(late),
+            .HBlank => bus_ptr.ppu.onHblankEnd(cpu, late), // The end of a HBlank
+            .VBlank => bus_ptr.ppu.onHdrawEnd(cpu, late), // The end of a VBlank
         }
     }
 
     /// Removes the **first** scheduled event of type `needle`
     pub fn removeScheduledEvent(self: *Self, needle: EventKind) void {
-        var it = self.queue.iterator();
-
-        var i: usize = 0;
-        while (it.next()) |event| : (i += 1) {
+        for (self.queue.items, 0..) |event, i| {
             if (std.meta.eql(event.kind, needle)) {
 
-                // This invalidates the iterator
+                // invalidates the slice we're iterating over
                 _ = self.queue.removeIndex(i);
 
-                // Since removing something from the PQ invalidates the iterator,
-                // this implementation can safely only remove the first instance of
-                // a Scheduled Event. Exit Early
+                // log.debug("Removed {?}@{}", .{ event.kind, event.tick });
                 break;
             }
         }

src/imgui.zig

@@ -0,0 +1,511 @@
+//! Namespace for dealing with ZBA's immediate-mode GUI
+//! Currently, ZBA uses zgui from https://github.com/michal-z/zig-gamedev
+//! which provides Zig bindings for https://github.com/ocornut/imgui under the hood
+
+const std = @import("std");
+const zgui = @import("zgui");
+const gl = @import("gl");
+const nfd = @import("nfd");
+const config = @import("config.zig");
+const emu = @import("core/emu.zig");
+
+const Gui = @import("platform.zig").Gui;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Scheduler = @import("core/scheduler.zig").Scheduler;
+const Bus = @import("core/Bus.zig");
+const Synchro = @import("core/emu.zig").Synchro;
+
+const RingBuffer = @import("zba-util").RingBuffer;
+const Dimensions = @import("platform.zig").Dimensions;
+
+const Allocator = std.mem.Allocator;
+const GLuint = gl.GLuint;
+
+const gba_width = @import("core/ppu.zig").width;
+const gba_height = @import("core/ppu.zig").height;
+
+const log = std.log.scoped(.Imgui);
+
+// two seconds worth of fps values into the past
+const histogram_len = 0x80;
+
+/// Immediate-Mode GUI State
+pub const State = struct {
+    title: [12:0]u8,
+
+    fps_hist: RingBuffer(u32),
+    should_quit: bool = false,
+    emulation: Emulation,
+
+    win_stat: WindowStatus = .{},
+
+    const WindowStatus = struct {
+        show_deps: bool = false,
+        show_regs: bool = false,
+        show_schedule: bool = false,
+        show_perf: bool = false,
+        show_palette: bool = false,
+    };
+
+    const Emulation = union(enum) {
+        Active,
+        Inactive,
+        Transition: enum { Active, Inactive },
+    };
+
+    /// if zba is initialized with a ROM already provided, this initializer should be called
+    /// with `title_opt` being non-null
+    pub fn init(allocator: Allocator, title_opt: ?*const [12]u8) !@This() {
+        const history = try allocator.alloc(u32, histogram_len);
+
+        return .{
+            .title = handleTitle(title_opt),
+            .emulation = if (title_opt == null) .Inactive else .{ .Transition = .Active },
+            .fps_hist = RingBuffer(u32).init(history),
+        };
+    }
+
+    pub fn deinit(self: *@This(), allocator: Allocator) void {
+        allocator.free(self.fps_hist.buf);
+        self.* = undefined;
+    }
+};
+
+pub fn draw(state: *State, sync: *Synchro, dim: Dimensions, cpu: *const Arm7tdmi, tex_id: GLuint) bool {
+    const scn_scale = config.config().host.win_scale;
+    const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+    zgui.backend.newFrame(@floatFromInt(dim.width), @floatFromInt(dim.height));
+
+    state.title = handleTitle(&bus_ptr.pak.title);
+
+    {
+        _ = zgui.beginMainMenuBar();
+        defer zgui.endMainMenuBar();
+
+        if (zgui.beginMenu("File", true)) {
+            defer zgui.endMenu();
+
+            if (zgui.menuItem("Quit", .{}))
+                state.should_quit = true;
+
+            if (zgui.menuItem("Insert ROM", .{})) blk: {
+                const file_path = tmp: {
+                    const path_opt = nfd.openFileDialog("gba", null) catch |e| {
+                        log.err("file dialog failed to open: {}", .{e});
+                        break :blk;
+                    };
+
+                    break :tmp path_opt orelse {
+                        log.warn("did not receive a file path", .{});
+                        break :blk;
+                    };
+                };
+                defer nfd.freePath(file_path);
+
+                log.info("user chose: \"{s}\"", .{file_path});
+
+                const message = tmp: {
+                    var msg: Synchro.Message = .{ .rom_path = undefined };
+                    @memcpy(msg.rom_path[0..file_path.len], file_path);
+                    break :tmp msg;
+                };
+
+                sync.ch.push(message) catch |e| {
+                    log.err("failed to send file path to emu thread: {}", .{e});
+                    break :blk;
+                };
+
+                state.emulation = .{ .Transition = .Active };
+            }
+
+            if (zgui.menuItem("Load BIOS", .{})) blk: {
+                const file_path = tmp: {
+                    const path_opt = nfd.openFileDialog("bin", null) catch |e| {
+                        log.err("file dialog failed to open: {}", .{e});
+                        break :blk;
+                    };
+
+                    break :tmp path_opt orelse {
+                        log.warn("did not receive a file path", .{});
+                        break :blk;
+                    };
+                };
+                defer nfd.freePath(file_path);
+
+                log.info("user chose: \"{s}\"", .{file_path});
+
+                const message = tmp: {
+                    var msg: Synchro.Message = .{ .bios_path = undefined };
+                    @memcpy(msg.bios_path[0..file_path.len], file_path);
+                    break :tmp msg;
+                };
+
+                sync.ch.push(message) catch |e| {
+                    log.err("failed to send file path to emu thread: {}", .{e});
+                    break :blk;
+                };
+            }
+        }
+
+        if (zgui.beginMenu("Emulation", true)) {
+            defer zgui.endMenu();
+
+            if (zgui.menuItem("Registers", .{ .selected = state.win_stat.show_regs }))
+                state.win_stat.show_regs = true;
+
+            if (zgui.menuItem("Palette", .{ .selected = state.win_stat.show_palette }))
+                state.win_stat.show_palette = true;
+
+            if (zgui.menuItem("Schedule", .{ .selected = state.win_stat.show_schedule }))
+                state.win_stat.show_schedule = true;
+
+            if (zgui.menuItem("Paused", .{ .selected = state.emulation == .Inactive })) {
+                state.emulation = switch (state.emulation) {
+                    .Active => .{ .Transition = .Inactive },
+                    .Inactive => .{ .Transition = .Active },
+                    else => state.emulation,
+                };
+            }
+
+            if (zgui.menuItem("Restart", .{}))
+                sync.ch.push(.restart) catch |e| log.err("failed to send restart req to emu thread: {}", .{e});
+        }
+
+        if (zgui.beginMenu("Stats", true)) {
+            defer zgui.endMenu();
+
+            if (zgui.menuItem("Performance", .{ .selected = state.win_stat.show_perf }))
+                state.win_stat.show_perf = true;
+        }
+
+        if (zgui.beginMenu("Help", true)) {
+            defer zgui.endMenu();
+
+            if (zgui.menuItem("Dependencies", .{ .selected = state.win_stat.show_deps }))
+                state.win_stat.show_deps = true;
+        }
+    }
+
+    {
+        const w: f32 = @floatFromInt(gba_width * scn_scale);
+        const h: f32 = @floatFromInt(gba_height * scn_scale);
+
+        const window_title = std.mem.sliceTo(&state.title, 0);
+        _ = zgui.begin(window_title, .{ .flags = .{ .no_resize = true, .always_auto_resize = true } });
+        defer zgui.end();
+
+        zgui.image(@ptrFromInt(tex_id), .{ .w = w, .h = h });
+    }
+
+    // TODO: Any other steps to respect the copyright of the libraries I use?
+    if (state.win_stat.show_deps) {
+        _ = zgui.begin("Dependencies", .{ .popen = &state.win_stat.show_deps });
+        defer zgui.end();
+
+        zgui.bulletText("known-folders by ziglibs", .{});
+        zgui.bulletText("nfd-zig by Fabio Arnold", .{});
+        {
+            zgui.indent(.{});
+            defer zgui.unindent(.{});
+
+            zgui.bulletText("nativefiledialog by Michael Labbe", .{});
+        }
+
+        zgui.bulletText("SDL.zig by Felix Queißner", .{});
+        {
+            zgui.indent(.{});
+            defer zgui.unindent(.{});
+
+            zgui.bulletText("SDL by Sam Lantinga", .{});
+        }
+
+        zgui.bulletText("tomlz by Matthew Hall", .{});
+        zgui.bulletText("zba-gdbstub by Rekai Musuka", .{});
+        zgui.bulletText("zba-util by Rekai Musuka", .{});
+        zgui.bulletText("zgui by Michal Ziulek", .{});
+        {
+            zgui.indent(.{});
+            defer zgui.unindent(.{});
+
+            zgui.bulletText("DearImGui by Omar Cornut", .{});
+        }
+        zgui.bulletText("zig-clap by Jimmi Holst Christensen", .{});
+        zgui.bulletText("zig-datetime by Jairus Martin", .{});
+
+        zgui.newLine();
+        zgui.bulletText("bitfield.zig by Hannes Bredberg and FlorenceOS contributors", .{});
+        zgui.bulletText("zig-opengl by Felix Queißner", .{});
+        {
+            zgui.indent(.{});
+            defer zgui.unindent(.{});
+
+            zgui.bulletText("OpenGL-Registry by The Khronos Group", .{});
+        }
+    }
+
+    if (state.win_stat.show_regs) {
+        _ = zgui.begin("Guest Registers", .{ .popen = &state.win_stat.show_regs });
+        defer zgui.end();
+
+        for (0..8) |i| {
+            zgui.text("R{}: 0x{X:0>8}", .{ i, cpu.r[i] });
+
+            zgui.sameLine(.{});
+
+            const padding = if (8 + i < 10) " " else "";
+            zgui.text("{s}R{}: 0x{X:0>8}", .{ padding, 8 + i, cpu.r[8 + i] });
+        }
+
+        zgui.separator();
+
+        widgets.psr("CPSR", cpu.cpsr);
+        widgets.psr("SPSR", cpu.spsr);
+
+        zgui.separator();
+
+        widgets.interrupts(" IE", bus_ptr.io.ie);
+        widgets.interrupts("IRQ", bus_ptr.io.irq);
+    }
+
+    if (state.win_stat.show_perf) {
+        _ = zgui.begin("Performance", .{ .popen = &state.win_stat.show_perf });
+        defer zgui.end();
+
+        const tmp = blk: {
+            var buf: [histogram_len]u32 = undefined;
+            const len = state.fps_hist.copy(&buf);
+
+            break :blk .{ buf, len };
+        };
+        const values = tmp[0];
+        const len = tmp[1];
+
+        if (len == values.len) _ = state.fps_hist.pop();
+
+        const sorted = blk: {
+            var buf: @TypeOf(values) = undefined;
+
+            @memcpy(buf[0..len], values[0..len]);
+            std.mem.sort(u32, buf[0..len], {}, std.sort.asc(u32));
+
+            break :blk buf;
+        };
+
+        const y_max: f64 = 2 * if (len != 0) @as(f64, @floatFromInt(sorted[len - 1])) else emu.frame_rate;
+        const x_max: f64 = @floatFromInt(values.len);
+
+        const y_args = .{ .flags = .{ .no_grid_lines = true } };
+        const x_args = .{ .flags = .{ .no_grid_lines = true, .no_tick_labels = true, .no_tick_marks = true } };
+
+        if (zgui.plot.beginPlot("Emulation FPS", .{ .w = 0.0, .flags = .{ .no_title = true, .no_frame = true } })) {
+            defer zgui.plot.endPlot();
+
+            zgui.plot.setupLegend(.{ .north = true, .east = true }, .{});
+            zgui.plot.setupAxis(.x1, x_args);
+            zgui.plot.setupAxis(.y1, y_args);
+            zgui.plot.setupAxisLimits(.y1, .{ .min = 0.0, .max = y_max, .cond = .always });
+            zgui.plot.setupAxisLimits(.x1, .{ .min = 0.0, .max = x_max, .cond = .always });
+            zgui.plot.setupFinish();
+
+            zgui.plot.plotLineValues("FPS", u32, .{ .v = values[0..len] });
+        }
+
+        const stats: struct { u32, u32, u32 } = blk: {
+            if (len == 0) break :blk .{ 0, 0, 0 };
+
+            const average: u32 = average: {
+                var sum: u32 = 0;
+                for (sorted[0..len]) |value| sum += value;
+
+                break :average @intCast(sum / len);
+            };
+            const median = sorted[len / 2];
+            const low = sorted[len / 100]; // 1% Low
+
+            break :blk .{ average, median, low };
+        };
+
+        zgui.text("Average: {:0>3} fps", .{stats[0]});
+        zgui.text(" Median: {:0>3} fps", .{stats[1]});
+        zgui.text(" 1% Low: {:0>3} fps", .{stats[2]});
+    }
+
+    if (state.win_stat.show_schedule) {
+        _ = zgui.begin("Schedule", .{ .popen = &state.win_stat.show_schedule });
+        defer zgui.end();
+
+        const scheduler = cpu.sched;
+
+        zgui.text("tick: {X:0>16}", .{scheduler.now()});
+        zgui.separator();
+
+        const sched_ptr: *Scheduler = @ptrCast(@alignCast(cpu.sched.ptr));
+        const Event = std.meta.Child(@TypeOf(sched_ptr.queue.items));
+
+        var items: [20]Event = undefined;
+        const len = @min(sched_ptr.queue.items.len, items.len);
+
+        @memcpy(items[0..len], sched_ptr.queue.items[0..len]);
+        std.mem.sort(Event, items[0..len], {}, widgets.eventDesc(Event));
+
+        for (items[0..len]) |event| {
+            zgui.text("{X:0>16} | {?}", .{ event.tick, event.kind });
+        }
+    }
+
+    if (state.win_stat.show_palette) {
+        _ = zgui.begin("Palette", .{ .popen = &state.win_stat.show_palette });
+        defer zgui.end();
+
+        widgets.paletteGrid(.Background, cpu);
+
+        zgui.sameLine(.{ .spacing = 20.0 });
+
+        widgets.paletteGrid(.Object, cpu);
+    }
+
+    // {
+    //     zgui.showDemoWindow(null);
+    // }
+
+    return true; // request redraw
+}
+
+const widgets = struct {
+    const PaletteKind = enum { Background, Object };
+
+    fn paletteGrid(comptime kind: PaletteKind, cpu: *const Arm7tdmi) void {
+        _ = zgui.beginGroup();
+        defer zgui.endGroup();
+
+        const address: u32 = switch (kind) {
+            .Background => 0x0500_0000,
+            .Object => 0x0500_0200,
+        };
+
+        for (0..0x100) |i| {
+            const offset: u32 = @truncate(i);
+            const bgr555 = cpu.bus.dbgRead(u16, address + offset * @sizeOf(u16));
+            widgets.colourSquare(bgr555);
+
+            if ((i + 1) % 0x10 != 0) zgui.sameLine(.{});
+        }
+        zgui.text(@tagName(kind), .{});
+    }
+
+    fn colourSquare(bgr555: u16) void {
+        // FIXME: working with the packed struct enum is currently broken :pensive:
+        const ImguiColorEditFlags_NoInputs: u32 = 1 << 5;
+        const ImguiColorEditFlags_NoPicker: u32 = 1 << 2;
+        const flags: zgui.ColorEditFlags = @bitCast(ImguiColorEditFlags_NoInputs | ImguiColorEditFlags_NoPicker);
+
+        const b: f32 = @floatFromInt(bgr555 >> 10 & 0x1f);
+        const g: f32 = @floatFromInt(bgr555 >> 5 & 0x1F);
+        const r: f32 = @floatFromInt(bgr555 & 0x1F);
+
+        var col = [_]f32{ r / 31.0, g / 31.0, b / 31.0 };
+
+        _ = zgui.colorEdit3("", .{ .col = &col, .flags = flags });
+    }
+
+    fn interrupts(comptime label: []const u8, int: anytype) void {
+        const h = 15.0;
+        const w = 9.0 * 2 + 3.5;
+        const ww = 9.0 * 3;
+
+        {
+            zgui.text(label ++ ":", .{});
+
+            zgui.sameLine(.{});
+            _ = zgui.selectable("VBL", .{ .w = w, .h = h, .selected = int.vblank.read() });
+
+            zgui.sameLine(.{});
+            _ = zgui.selectable("HBL", .{ .w = w, .h = h, .selected = int.hblank.read() });
+
+            zgui.sameLine(.{});
+            _ = zgui.selectable("VCT", .{ .w = w, .h = h, .selected = int.coincidence.read() });
+
+            {
+                zgui.sameLine(.{});
+                _ = zgui.selectable("TIM0", .{ .w = ww, .h = h, .selected = int.tim0.read() });
+
+                zgui.sameLine(.{});
+                _ = zgui.selectable("TIM1", .{ .w = ww, .h = h, .selected = int.tim1.read() });
+
+                zgui.sameLine(.{});
+                _ = zgui.selectable("TIM2", .{ .w = ww, .h = h, .selected = int.tim2.read() });
+
+                zgui.sameLine(.{});
+                _ = zgui.selectable("TIM3", .{ .w = ww, .h = h, .selected = int.tim3.read() });
+            }
+
+            zgui.sameLine(.{});
+            _ = zgui.selectable("SRL", .{ .w = w, .h = h, .selected = int.serial.read() });
+
+            {
+                zgui.sameLine(.{});
+                _ = zgui.selectable("DMA0", .{ .w = ww, .h = h, .selected = int.dma0.read() });
+
+                zgui.sameLine(.{});
+                _ = zgui.selectable("DMA1", .{ .w = ww, .h = h, .selected = int.dma1.read() });
+
+                zgui.sameLine(.{});
+                _ = zgui.selectable("DMA2", .{ .w = ww, .h = h, .selected = int.dma2.read() });
+
+                zgui.sameLine(.{});
+                _ = zgui.selectable("DMA3", .{ .w = ww, .h = h, .selected = int.dma3.read() });
+            }
+
+            zgui.sameLine(.{});
+            _ = zgui.selectable("KPD", .{ .w = w, .h = h, .selected = int.keypad.read() });
+
+            zgui.sameLine(.{});
+            _ = zgui.selectable("GPK", .{ .w = w, .h = h, .selected = int.game_pak.read() });
+        }
+    }
+
+    fn psr(comptime label: []const u8, register: anytype) void {
+        const Mode = @import("arm32").arm.Mode;
+
+        const maybe_mode = std.meta.intToEnum(Mode, register.mode.read()) catch null;
+        const mode = if (maybe_mode) |mode| mode.toString() else "???";
+        const w = 9.0;
+        const h = 15.0;
+
+        zgui.text(label ++ ": 0x{X:0>8}", .{register.raw});
+
+        zgui.sameLine(.{});
+        _ = zgui.selectable("N", .{ .w = w, .h = h, .selected = register.n.read() });
+
+        zgui.sameLine(.{});
+        _ = zgui.selectable("Z", .{ .w = w, .h = h, .selected = register.z.read() });
+
+        zgui.sameLine(.{});
+        _ = zgui.selectable("C", .{ .w = w, .h = h, .selected = register.c.read() });
+
+        zgui.sameLine(.{});
+        _ = zgui.selectable("V", .{ .w = w, .h = h, .selected = register.v.read() });
+
+        zgui.sameLine(.{});
+        zgui.text("{s}", .{mode});
+    }
+
+    fn eventDesc(comptime T: type) fn (void, T, T) bool {
+        return struct {
+            fn inner(_: void, left: T, right: T) bool {
+                return left.tick > right.tick;
+            }
+        }.inner;
+    }
+};
+
+fn handleTitle(title_opt: ?*const [12]u8) [12:0]u8 {
+    if (title_opt == null) return "[N/A Title]\x00".*; // No ROM present
+    const title = title_opt.?;
+
+    // ROM Title is an empty string (ImGui hates these)
+    if (title[0] == '\x00') return "[No Title]\x00\x00".*;
+
+    return title.* ++ [_:0]u8{};
+}

src/main.zig

@@ -4,54 +4,96 @@ const known_folders = @import("known_folders");
 const clap = @import("clap");
 
 const config = @import("config.zig");
+const emu = @import("core/emu.zig");
 
+const Synchro = @import("core/emu.zig").Synchro;
 const Gui = @import("platform.zig").Gui;
 const Bus = @import("core/Bus.zig");
-const Arm7tdmi = @import("core/cpu.zig").Arm7tdmi;
 const Scheduler = @import("core/scheduler.zig").Scheduler;
 const FilePaths = @import("util.zig").FilePaths;
-
+const FpsTracker = @import("util.zig").FpsTracker;
 const Allocator = std.mem.Allocator;
+
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const IBus = @import("arm32").Bus;
+const IScheduler = @import("arm32").Scheduler;
+
 const log = std.log.scoped(.Cli);
-const width = @import("core/ppu.zig").width;
-const height = @import("core/ppu.zig").height;
 pub const log_level = if (builtin.mode != .Debug) .info else std.log.default_level;
 
 // CLI Arguments + Help Text
 const params = clap.parseParamsComptime(
     \\-h, --help            Display this help and exit.
+    \\-s, --skip            Skip BIOS.
     \\-b, --bios <str>      Optional path to a GBA BIOS ROM.
-    \\<str>                 Path to the GBA GamePak ROM
+    \\ --gdb                Run ZBA from the context of a GDB Server
+    \\<str>                 Path to the GBA GamePak ROM.
     \\
 );
 
-pub fn main() anyerror!void {
+pub fn main() void {
     // Main Allocator for ZBA
     var gpa = std.heap.GeneralPurposeAllocator(.{}){};
-    defer std.debug.assert(!gpa.deinit());
+    defer std.debug.assert(gpa.deinit() == .ok);
+
     const allocator = gpa.allocator();
 
-    // TODO: Make Error message not Linux Specific
-    const data_path = try known_folders.getPath(allocator, .data) orelse exit("Unable to Determine XDG Data Path", .{});
+    // Determine the Data Directory (stores saves)
+    const data_path = blk: {
+        const result = known_folders.getPath(allocator, .data);
+        const option = result catch |e| exitln("interrupted while determining the data folder: {}", .{e});
+        const path = option orelse exitln("no valid data folder found", .{});
+        ensureDataDirsExist(path) catch |e| exitln("failed to create folders under \"{s}\": {}", .{ path, e });
+
+        break :blk path;
+    };
     defer allocator.free(data_path);
 
-    const config_path = try configFilePath(allocator, data_path);
+    // Determine the Config Directory
+    const config_path = blk: {
+        const result = known_folders.getPath(allocator, .roaming_configuration);
+        const option = result catch |e| exitln("interreupted while determining the config folder: {}", .{e});
+        const path = option orelse exitln("no valid config folder found", .{});
+        ensureConfigDirExists(path) catch |e| exitln("failed to create required folder \"{s}\": {}", .{ path, e });
+
+        break :blk path;
+    };
     defer allocator.free(config_path);
 
-    const save_path = try savePath(allocator, data_path);
-    defer allocator.free(save_path);
+    // Parse CLI
 
-    try config.load(allocator, config_path);
-
-    // Handle CLI Input
-    const result = try clap.parse(clap.Help, &params, clap.parsers.default, .{});
+    const result = clap.parse(clap.Help, &params, clap.parsers.default, .{ .allocator = allocator }) catch |e| exitln("failed to parse cli: {}", .{e});
     defer result.deinit();
 
-    const paths = try handleArguments(allocator, data_path, &result);
-    defer if (paths.save) |path| allocator.free(path);
+    // TODO: Move config file to XDG Config directory?
+    const cfg_file_path = configFilePath(allocator, config_path) catch |e| exitln("failed to ready config file for access: {}", .{e});
+    defer allocator.free(cfg_file_path);
 
-    const cpu_trace = config.config().debug.cpu_trace;
-    const log_file: ?std.fs.File = if (cpu_trace) try std.fs.cwd().createFile("zba.log", .{}) else null;
+    config.load(allocator, cfg_file_path) catch |e| exitln("failed to load config file: {}", .{e});
+
+    var paths = handleArguments(allocator, data_path, &result) catch |e| exitln("failed to handle cli arguments: {}", .{e});
+    defer paths.deinit(allocator);
+
+    // if paths.bios is null, then we want to see if it's in the data directory
+    if (paths.bios == null) blk: {
+        const bios_path = std.mem.join(allocator, "/", &.{ data_path, "zba", "gba_bios.bin" }) catch |e| exitln("failed to allocate backup bios dir path: {}", .{e});
+        defer allocator.free(bios_path);
+
+        _ = std.fs.cwd().statFile(bios_path) catch |e| switch (e) {
+            error.FileNotFound => { // ZBA will crash on attempt to read BIOS but that's fine
+                log.err("file located at {s} was not found", .{bios_path});
+                break :blk;
+            },
+            else => exitln("error when checking \"{s}\": {}", .{ bios_path, e }),
+        };
+
+        paths.bios = allocator.dupe(u8, bios_path) catch |e| exitln("failed to duplicate path to bios: {}", .{e});
+    }
+
+    const log_file = switch (config.config().debug.cpu_trace) {
+        true => std.fs.cwd().createFile("zba.log", .{}) catch |e| exitln("failed to create trace log file: {}", .{e}),
+        false => null,
+    };
     defer if (log_file) |file| file.close();
 
     // TODO: Take Emulator Init Code out of main.zig
@@ -59,70 +101,134 @@ pub fn main() anyerror!void {
     defer scheduler.deinit();
 
     var bus: Bus = undefined;
-    var cpu = Arm7tdmi.init(&scheduler, &bus, log_file);
-    if (paths.bios == null) cpu.fastBoot();
 
-    try bus.init(allocator, &scheduler, &cpu, paths);
+    const ischeduler = IScheduler.init(&scheduler);
+    const ibus = IBus.init(&bus);
+
+    var cpu = Arm7tdmi.init(ischeduler, ibus);
+
+    bus.init(allocator, &scheduler, &cpu, paths) catch |e| exitln("failed to init zba bus: {}", .{e});
     defer bus.deinit();
 
-    var gui = Gui.init(&bus.pak.title, &bus.apu, width, height);
+    if (config.config().guest.skip_bios or result.args.skip != 0 or paths.bios == null) {
+        @import("core/cpu_util.zig").fastBoot(&cpu);
+    }
+
+    const title_ptr = if (paths.rom != null) &bus.pak.title else null;
+
+    // TODO: Just copy the title instead of grabbing a pointer to it
+    var gui = Gui.init(allocator, &bus.apu, title_ptr) catch |e| exitln("failed to init gui: {}", .{e});
     defer gui.deinit();
 
-    try gui.run(&cpu, &scheduler);
+    var sync = Synchro.init(allocator) catch |e| exitln("failed to allocate sync types: {}", .{e});
+    defer sync.deinit(allocator);
+
+    if (result.args.gdb != 0) {
+        const Server = @import("gdbstub").Server;
+        const EmuThing = @import("core/emu.zig").EmuThing;
+
+        var wrapper = EmuThing.init(&cpu, &scheduler);
+        var emulator = wrapper.interface(allocator);
+        defer emulator.deinit();
+
+        log.info("Ready to connect", .{});
+
+        var server = Server.init(
+            emulator,
+            .{ .memory_map = EmuThing.map, .target = EmuThing.target },
+        ) catch |e| exitln("failed to init gdb server: {}", .{e});
+        defer server.deinit(allocator);
+
+        log.info("Starting GDB Server Thread", .{});
+
+        const thread = std.Thread.spawn(.{}, Server.run, .{ &server, allocator, &sync.should_quit }) catch |e| exitln("gdb server thread crashed: {}", .{e});
+        defer thread.join();
+
+        gui.run(.{
+            .cpu = &cpu,
+            .scheduler = &scheduler,
+            .sync = &sync,
+        }) catch |e| exitln("main thread panicked: {}", .{e});
+    } else {
+        var tracker = FpsTracker.init();
+
+        const thread = std.Thread.spawn(.{}, emu.run, .{ &cpu, &scheduler, &tracker, &sync }) catch |e| exitln("emu thread panicked: {}", .{e});
+        defer thread.join();
+
+        gui.run(.{
+            .cpu = &cpu,
+            .scheduler = &scheduler,
+            .tracker = &tracker,
+            .sync = &sync,
+        }) catch |e| exitln("main thread panicked: {}", .{e});
+    }
 }
 
-pub fn handleArguments(allocator: Allocator, data_path: []const u8, result: *const clap.Result(clap.Help, &params, clap.parsers.default)) !FilePaths {
-    const rom_path = romPath(result);
-    log.info("ROM path: {s}", .{rom_path});
+fn handleArguments(allocator: Allocator, data_path: []const u8, result: *const clap.Result(clap.Help, &params, clap.parsers.default)) !FilePaths {
+    const rom_path = try romPath(allocator, result);
+    errdefer if (rom_path) |path| allocator.free(path);
 
-    const bios_path = result.args.bios;
-    if (bios_path) |path| log.info("BIOS path: {s}", .{path}) else log.info("No BIOS provided", .{});
+    const bios_path: ?[]const u8 = if (result.args.bios) |path| try allocator.dupe(u8, path) else null;
+    errdefer if (bios_path) |path| allocator.free(path);
 
-    const save_path = try savePath(allocator, data_path);
+    const save_path = try std.fs.path.join(allocator, &[_][]const u8{ data_path, "zba", "save" });
+
+    log.info("ROM path: {?s}", .{rom_path});
+    log.info("BIOS path: {?s}", .{bios_path});
     log.info("Save path: {s}", .{save_path});
 
-    return FilePaths{
+    return .{
         .rom = rom_path,
         .bios = bios_path,
         .save = save_path,
     };
 }
 
-fn configFilePath(allocator: Allocator, data_path: []const u8) ![]const u8 {
-    const path = try std.fs.path.join(allocator, &[_][]const u8{ data_path, "zba", "config.toml" });
+fn configFilePath(allocator: Allocator, config_path: []const u8) ![]const u8 {
+    const path = try std.fs.path.join(allocator, &[_][]const u8{ config_path, "zba", "config.toml" });
+    errdefer allocator.free(path);
 
     // We try to create the file exclusively, meaning that we err out if the file already exists.
     // All we care about is a file being there so we can just ignore that error in particular and
     // continue down the happy pathj
-    std.fs.accessAbsolute(path, .{}) catch {
-        const file_handle = try std.fs.createFileAbsolute(path, .{});
-        file_handle.close();
+    std.fs.accessAbsolute(path, .{}) catch |e| {
+        if (e != error.FileNotFound) return e;
+
+        const config_file = std.fs.createFileAbsolute(path, .{}) catch |err| exitln("failed to create \"{s}\": {}", .{ path, err });
+        defer config_file.close();
+
+        try config_file.writeAll(@embedFile("example.toml"));
     };
 
     return path;
 }
 
-fn savePath(allocator: Allocator, data_path: []const u8) ![]const u8 {
+fn ensureDataDirsExist(data_path: []const u8) !void {
     var dir = try std.fs.openDirAbsolute(data_path, .{});
     defer dir.close();
 
-    // Will either make the path recursively, or just exit early since it already exists
-    try dir.makePath("zba" ++ [_]u8{std.fs.path.sep} ++ "save");
-
-    // FIXME: Do we have to allocate? :sad:
-    return try std.fs.path.join(allocator, &[_][]const u8{ data_path, "zba", "save" });
+    // Will recursively create directories
+    try dir.makePath("zba" ++ std.fs.path.sep_str ++ "save");
 }
 
-fn romPath(result: *const clap.Result(clap.Help, &params, clap.parsers.default)) []const u8 {
+fn ensureConfigDirExists(config_path: []const u8) !void {
+    var dir = try std.fs.openDirAbsolute(config_path, .{});
+    defer dir.close();
+
+    try dir.makePath("zba");
+}
+
+fn romPath(allocator: Allocator, result: *const clap.Result(clap.Help, &params, clap.parsers.default)) !?[]const u8 {
     return switch (result.positionals.len) {
-        1 => result.positionals[0],
-        0 => exit("ZBA requires a path to a GamePak ROM\n", .{}),
-        else => exit("ZBA received too many positional arguments. \n", .{}),
+        0 => null,
+        1 => try allocator.dupe(u8, result.positionals[0]),
+        else => exitln("ZBA received too many positional arguments.", .{}),
     };
 }
 
-fn exit(comptime format: []const u8, args: anytype) noreturn {
+fn exitln(comptime format: []const u8, args: anytype) noreturn {
     const stderr = std.io.getStdErr().writer();
     stderr.print(format, args) catch {}; // Just exit already...
-    std.os.exit(1);
+    stderr.writeByte('\n') catch {};
+    std.process.exit(1);
 }

src/platform.zig

@@ -1,155 +1,260 @@
 const std = @import("std");
 const SDL = @import("sdl2");
+const gl = @import("gl");
+const zgui = @import("zgui");
+
 const emu = @import("core/emu.zig");
 const config = @import("config.zig");
+const imgui = @import("imgui.zig");
 
 const Apu = @import("core/apu.zig").Apu;
-const Arm7tdmi = @import("core/cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("core/Bus.zig");
 const Scheduler = @import("core/scheduler.zig").Scheduler;
 const FpsTracker = @import("util.zig").FpsTracker;
+const Synchro = @import("core/emu.zig").Synchro;
+const KeyInput = @import("core/bus/io.zig").KeyInput;
 
-const span = @import("util.zig").span;
+const gba_width = @import("core/ppu.zig").width;
+const gba_height = @import("core/ppu.zig").height;
 
-const pitch = @import("core/ppu.zig").framebuf_pitch;
-const default_title: []const u8 = "ZBA";
+const GLuint = gl.GLuint;
+const GLsizei = gl.GLsizei;
+const SDL_GLContext = *anyopaque;
+const Allocator = std.mem.Allocator;
+
+pub const Dimensions = struct { width: u32, height: u32 };
+const default_dim: Dimensions = .{ .width = 1280, .height = 720 };
+
+pub const sample_rate = 1 << 15;
+pub const sample_format = SDL.AUDIO_U16;
+
+const window_title = "ZBA";
 
 pub const Gui = struct {
     const Self = @This();
     const log = std.log.scoped(.Gui);
 
     window: *SDL.SDL_Window,
-    title: []const u8,
-    renderer: *SDL.SDL_Renderer,
-    texture: *SDL.SDL_Texture,
+    ctx: SDL_GLContext,
     audio: Audio,
 
-    pub fn init(title: *const [12]u8, apu: *Apu, width: i32, height: i32) Self {
-        const ret = SDL.SDL_Init(SDL.SDL_INIT_VIDEO | SDL.SDL_INIT_EVENTS | SDL.SDL_INIT_AUDIO | SDL.SDL_INIT_GAMECONTROLLER);
-        if (ret < 0) panic();
+    state: imgui.State,
+    allocator: Allocator,
 
-        const win_scale = @intCast(c_int, config.config().host.win_scale);
+    pub fn init(allocator: Allocator, apu: *Apu, title_opt: ?*const [12]u8) !Self {
+        if (SDL.SDL_Init(SDL.SDL_INIT_VIDEO | SDL.SDL_INIT_EVENTS | SDL.SDL_INIT_AUDIO) < 0) panic();
+        if (SDL.SDL_GL_SetAttribute(SDL.SDL_GL_CONTEXT_PROFILE_MASK, SDL.SDL_GL_CONTEXT_PROFILE_CORE) < 0) panic();
+        if (SDL.SDL_GL_SetAttribute(SDL.SDL_GL_CONTEXT_MAJOR_VERSION, 3) < 0) panic();
+        if (SDL.SDL_GL_SetAttribute(SDL.SDL_GL_CONTEXT_MAJOR_VERSION, 3) < 0) panic();
 
         const window = SDL.SDL_CreateWindow(
-            default_title.ptr,
+            window_title,
             SDL.SDL_WINDOWPOS_CENTERED,
             SDL.SDL_WINDOWPOS_CENTERED,
-            @as(c_int, width * win_scale),
-            @as(c_int, height * win_scale),
-            SDL.SDL_WINDOW_SHOWN,
+            default_dim.width,
+            default_dim.height,
+            SDL.SDL_WINDOW_OPENGL | SDL.SDL_WINDOW_SHOWN | SDL.SDL_WINDOW_RESIZABLE,
         ) orelse panic();
 
-        const renderer_flags = SDL.SDL_RENDERER_ACCELERATED | if (config.config().host.vsync) SDL.SDL_RENDERER_PRESENTVSYNC else 0;
-        const renderer = SDL.SDL_CreateRenderer(window, -1, @bitCast(u32, renderer_flags)) orelse panic();
+        const ctx = SDL.SDL_GL_CreateContext(window) orelse panic();
+        if (SDL.SDL_GL_MakeCurrent(window, ctx) < 0) panic();
 
-        const texture = SDL.SDL_CreateTexture(
-            renderer,
-            SDL.SDL_PIXELFORMAT_RGBA8888,
-            SDL.SDL_TEXTUREACCESS_STREAMING,
-            @as(c_int, width),
-            @as(c_int, height),
-        ) orelse panic();
+        gl.load(ctx, Self.glGetProcAddress) catch {};
+        if (SDL.SDL_GL_SetSwapInterval(@intFromBool(config.config().host.vsync)) < 0) panic();
+
+        zgui.init(allocator);
+        zgui.plot.init();
+        zgui.backend.init(window, ctx, "#version 330 core");
+
+        // zgui.io.setIniFilename(null);
 
         return Self{
             .window = window,
-            .title = span(title),
-            .renderer = renderer,
-            .texture = texture,
+            .ctx = ctx,
             .audio = Audio.init(apu),
+
+            .allocator = allocator,
+            .state = try imgui.State.init(allocator, title_opt),
         };
     }
 
-    pub fn run(self: *Self, cpu: *Arm7tdmi, scheduler: *Scheduler) !void {
-        var quit = std.atomic.Atomic(bool).init(false);
-        var tracker = FpsTracker.init();
+    pub fn deinit(self: *Self) void {
+        self.audio.deinit();
+        self.state.deinit(self.allocator);
 
-        const thread = try std.Thread.spawn(.{}, emu.run, .{ &quit, scheduler, cpu, &tracker });
-        defer thread.join();
+        zgui.backend.deinit();
+        zgui.plot.deinit();
+        zgui.deinit();
 
-        var title_buf: [0x100]u8 = [_]u8{0} ** 0x100;
+        SDL.SDL_GL_DeleteContext(self.ctx);
+        SDL.SDL_DestroyWindow(self.window);
+        SDL.SDL_Quit();
+
+        self.* = undefined;
+    }
+
+    const RunOptions = struct {
+        sync: *Synchro,
+        tracker: ?*FpsTracker = null,
+        cpu: *Arm7tdmi,
+        scheduler: *Scheduler,
+    };
+
+    pub fn run(self: *Self, opt: RunOptions) !void {
+        const cpu = opt.cpu;
+        const tracker = opt.tracker;
+        const sync = opt.sync;
+
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
+        const vao_id = opengl_impl.vao();
+        defer gl.deleteVertexArrays(1, &[_]GLuint{vao_id});
+
+        const emu_tex = opengl_impl.screenTex(bus_ptr.ppu.framebuf.get(.Renderer));
+        const out_tex = opengl_impl.outTex();
+        defer gl.deleteTextures(2, &[_]GLuint{ emu_tex, out_tex });
+
+        const fbo_id = try opengl_impl.frameBuffer(out_tex);
+        defer gl.deleteFramebuffers(1, &fbo_id);
+
+        const prog_id = try opengl_impl.program(); // Dynamic Shaders?
+        defer gl.deleteProgram(prog_id);
+
+        var win_dim: Dimensions = default_dim;
 
         emu_loop: while (true) {
+            // Outside of `SDL.SDL_QUIT` below, the DearImgui UI might signal that the program
+            // should exit, in which case we should also handle this
+            if (self.state.should_quit or sync.should_quit.load(.monotonic)) break :emu_loop;
+
             var event: SDL.SDL_Event = undefined;
             while (SDL.SDL_PollEvent(&event) != 0) {
+                _ = zgui.backend.processEvent(&event);
+
                 switch (event.type) {
                     SDL.SDL_QUIT => break :emu_loop,
                     SDL.SDL_KEYDOWN => {
-                        const io = &cpu.bus.io;
+                        // TODO: Make use of compare_and_xor?
                         const key_code = event.key.keysym.sym;
+                        var keyinput: KeyInput = .{ .raw = 0x0000 };
 
                         switch (key_code) {
-                            SDL.SDLK_UP => io.keyinput.up.unset(),
-                            SDL.SDLK_DOWN => io.keyinput.down.unset(),
-                            SDL.SDLK_LEFT => io.keyinput.left.unset(),
-                            SDL.SDLK_RIGHT => io.keyinput.right.unset(),
-                            SDL.SDLK_x => io.keyinput.a.unset(),
-                            SDL.SDLK_z => io.keyinput.b.unset(),
-                            SDL.SDLK_a => io.keyinput.shoulder_l.unset(),
-                            SDL.SDLK_s => io.keyinput.shoulder_r.unset(),
-                            SDL.SDLK_RETURN => io.keyinput.start.unset(),
-                            SDL.SDLK_RSHIFT => io.keyinput.select.unset(),
+                            SDL.SDLK_UP => keyinput.up.set(),
+                            SDL.SDLK_DOWN => keyinput.down.set(),
+                            SDL.SDLK_LEFT => keyinput.left.set(),
+                            SDL.SDLK_RIGHT => keyinput.right.set(),
+                            SDL.SDLK_x => keyinput.a.set(),
+                            SDL.SDLK_z => keyinput.b.set(),
+                            SDL.SDLK_a => keyinput.shoulder_l.set(),
+                            SDL.SDLK_s => keyinput.shoulder_r.set(),
+                            SDL.SDLK_RETURN => keyinput.start.set(),
+                            SDL.SDLK_RSHIFT => keyinput.select.set(),
                             else => {},
                         }
+
+                        bus_ptr.io.keyinput.fetchAnd(~keyinput.raw, .monotonic);
                     },
                     SDL.SDL_KEYUP => {
-                        const io = &cpu.bus.io;
+                        // TODO: Make use of compare_and_xor?
                         const key_code = event.key.keysym.sym;
+                        var keyinput: KeyInput = .{ .raw = 0x0000 };
 
                         switch (key_code) {
-                            SDL.SDLK_UP => io.keyinput.up.set(),
-                            SDL.SDLK_DOWN => io.keyinput.down.set(),
-                            SDL.SDLK_LEFT => io.keyinput.left.set(),
-                            SDL.SDLK_RIGHT => io.keyinput.right.set(),
-                            SDL.SDLK_x => io.keyinput.a.set(),
-                            SDL.SDLK_z => io.keyinput.b.set(),
-                            SDL.SDLK_a => io.keyinput.shoulder_l.set(),
-                            SDL.SDLK_s => io.keyinput.shoulder_r.set(),
-                            SDL.SDLK_RETURN => io.keyinput.start.set(),
-                            SDL.SDLK_RSHIFT => io.keyinput.select.set(),
-                            SDL.SDLK_i => log.err("Sample Count: {}", .{@intCast(u32, SDL.SDL_AudioStreamAvailable(cpu.bus.apu.stream)) / (2 * @sizeOf(u16))}),
-                            SDL.SDLK_j => log.err("Scheduler Capacity: {} | Scheduler Event Count: {}", .{ scheduler.queue.capacity(), scheduler.queue.count() }),
-                            SDL.SDLK_k => {
-                                // Dump IWRAM to file
-                                log.info("PC: 0x{X:0>8}", .{cpu.r[15]});
-                                log.info("LR: 0x{X:0>8}", .{cpu.r[14]});
-                                // const iwram_file = try std.fs.cwd().createFile("iwram.bin", .{});
-                                // defer iwram_file.close();
-
-                                // try iwram_file.writeAll(cpu.bus.iwram.buf);
-                            },
+                            SDL.SDLK_UP => keyinput.up.set(),
+                            SDL.SDLK_DOWN => keyinput.down.set(),
+                            SDL.SDLK_LEFT => keyinput.left.set(),
+                            SDL.SDLK_RIGHT => keyinput.right.set(),
+                            SDL.SDLK_x => keyinput.a.set(),
+                            SDL.SDLK_z => keyinput.b.set(),
+                            SDL.SDLK_a => keyinput.shoulder_l.set(),
+                            SDL.SDLK_s => keyinput.shoulder_r.set(),
+                            SDL.SDLK_RETURN => keyinput.start.set(),
+                            SDL.SDLK_RSHIFT => keyinput.select.set(),
                             else => {},
                         }
+
+                        bus_ptr.io.keyinput.fetchOr(keyinput.raw, .monotonic);
+                    },
+                    SDL.SDL_WINDOWEVENT => {
+                        if (event.window.event == SDL.SDL_WINDOWEVENT_RESIZED) {
+                            log.debug("window resized to: {}x{}", .{ event.window.data1, event.window.data2 });
+
+                            win_dim.width = @intCast(event.window.data1);
+                            win_dim.height = @intCast(event.window.data2);
+                        }
                     },
                     else => {},
                 }
             }
 
-            // Emulator has an internal Double Buffer
-            const framebuf = cpu.bus.ppu.framebuf.get(.Renderer);
-            _ = SDL.SDL_UpdateTexture(self.texture, null, framebuf.ptr, pitch);
-            _ = SDL.SDL_RenderCopy(self.renderer, self.texture, null, null);
-            SDL.SDL_RenderPresent(self.renderer);
+            var zgui_redraw: bool = false;
 
-            const dyn_title = std.fmt.bufPrint(&title_buf, "ZBA | {s} [Emu: {}fps] ", .{ self.title, tracker.value() }) catch unreachable;
-            SDL.SDL_SetWindowTitle(self.window, dyn_title.ptr);
+            switch (self.state.emulation) {
+                .Transition => |inner| switch (inner) {
+                    .Active => {
+                        sync.paused.store(false, .monotonic);
+                        if (!config.config().host.mute) SDL.SDL_PauseAudioDevice(self.audio.device, 0);
+
+                        self.state.emulation = .Active;
+                    },
+                    .Inactive => {
+                        // Assert that double pausing is impossible
+                        SDL.SDL_PauseAudioDevice(self.audio.device, 1);
+                        sync.paused.store(true, .monotonic);
+
+                        self.state.emulation = .Inactive;
+                    },
+                },
+                .Active => {
+                    // Add FPS count to the histogram
+                    if (tracker) |t| self.state.fps_hist.push(t.value()) catch {};
+
+                    // Draw GBA Screen to Texture
+                    {
+                        gl.bindFramebuffer(gl.FRAMEBUFFER, fbo_id);
+                        defer gl.bindFramebuffer(gl.FRAMEBUFFER, 0);
+
+                        gl.viewport(0, 0, gba_width, gba_height);
+                        opengl_impl.drawScreen(emu_tex, prog_id, vao_id, bus_ptr.ppu.framebuf.get(.Renderer));
+                    }
+
+                    // FIXME: We only really care about locking the audio device (and therefore writing silence)
+                    // since if nfd-zig is used the emu may be paused for way too long. Perhaps we should try and limit
+                    // spurious calls to SDL_LockAudioDevice?
+                    SDL.SDL_LockAudioDevice(self.audio.device);
+                    defer SDL.SDL_UnlockAudioDevice(self.audio.device);
+
+                    zgui_redraw = imgui.draw(&self.state, sync, win_dim, cpu, out_tex);
+                },
+                .Inactive => zgui_redraw = imgui.draw(&self.state, sync, win_dim, cpu, out_tex),
+            }
+
+            if (zgui_redraw) {
+                // Background Colour
+                const size = zgui.io.getDisplaySize();
+                gl.viewport(0, 0, @intFromFloat(size[0]), @intFromFloat(size[1]));
+                gl.clearColor(0, 0, 0, 1.0);
+                gl.clear(gl.COLOR_BUFFER_BIT);
+
+                zgui.backend.draw();
+            }
+
+            SDL.SDL_GL_SwapWindow(self.window);
         }
 
-        quit.store(true, .SeqCst); // Terminate Emulator Thread
+        sync.should_quit.store(true, .monotonic);
     }
 
-    pub fn deinit(self: *Self) void {
-        self.audio.deinit();
-        SDL.SDL_DestroyTexture(self.texture);
-        SDL.SDL_DestroyRenderer(self.renderer);
-        SDL.SDL_DestroyWindow(self.window);
-        SDL.SDL_Quit();
-        self.* = undefined;
+    fn glGetProcAddress(ctx: SDL.SDL_GLContext, proc: [:0]const u8) ?*anyopaque {
+        _ = ctx;
+        return SDL.SDL_GL_GetProcAddress(proc.ptr);
     }
 };
 
 const Audio = struct {
     const Self = @This();
     const log = std.log.scoped(.PlatformAudio);
-    const sample_rate = @import("core/apu.zig").host_sample_rate;
 
     device: SDL.SDL_AudioDeviceID,
 
@@ -157,17 +262,18 @@ const Audio = struct {
         var have: SDL.SDL_AudioSpec = undefined;
         var want: SDL.SDL_AudioSpec = std.mem.zeroes(SDL.SDL_AudioSpec);
         want.freq = sample_rate;
-        want.format = SDL.AUDIO_U16;
+        want.format = sample_format;
         want.channels = 2;
         want.samples = 0x100;
         want.callback = Self.callback;
         want.userdata = apu;
 
+        std.debug.assert(sample_format == SDL.AUDIO_U16);
+        log.info("Host Sample Rate: {}Hz, Host Format: SDL.AUDIO_U16", .{sample_rate});
+
         const device = SDL.SDL_OpenAudioDevice(null, 0, &want, &have, 0);
         if (device == 0) panic();
 
-        SDL.SDL_PauseAudioDevice(device, 0); // Unpause Audio
-
         return .{ .device = device };
     }
 
@@ -177,12 +283,9 @@ const Audio = struct {
     }
 
     export fn callback(userdata: ?*anyopaque, stream: [*c]u8, len: c_int) void {
-        const apu = @ptrCast(*Apu, @alignCast(@alignOf(*Apu), userdata));
-        _ = SDL.SDL_AudioStreamGet(apu.stream, stream, len);
+        const apu: *Apu = @ptrCast(@alignCast(userdata));
 
-        // If we don't write anything, play silence otherwise garbage will be played
-        // FIXME: I don't think this hack to remove DC Offset is acceptable :thinking:
-        // if (written == 0) std.mem.set(u8, stream[0..@intCast(usize, len)], 0x40);
+        _ = SDL.SDL_AudioStreamGet(apu.stream, stream, len);
     }
 };
 
@@ -190,3 +293,124 @@ fn panic() noreturn {
     const str = @as(?[*:0]const u8, SDL.SDL_GetError()) orelse "unknown error";
     @panic(std.mem.sliceTo(str, 0));
 }
+
+const opengl_impl = struct {
+    fn drawScreen(tex_id: GLuint, prog_id: GLuint, vao_id: GLuint, buf: []const u8) void {
+        gl.bindTexture(gl.TEXTURE_2D, tex_id);
+        defer gl.bindTexture(gl.TEXTURE_2D, 0);
+
+        gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gba_width, gba_height, gl.RGBA, gl.UNSIGNED_INT_8_8_8_8, buf.ptr);
+
+        // Bind VAO
+        gl.bindVertexArray(vao_id);
+        defer gl.bindVertexArray(0);
+
+        // Use compiled frag + vertex shader
+        gl.useProgram(prog_id);
+        defer gl.useProgram(0);
+
+        gl.drawArrays(gl.TRIANGLE_STRIP, 0, 3);
+    }
+
+    fn program() !GLuint {
+        const vert_shader = @embedFile("shader/pixelbuf.vert");
+        const frag_shader = @embedFile("shader/pixelbuf.frag");
+
+        const vs = gl.createShader(gl.VERTEX_SHADER);
+        defer gl.deleteShader(vs);
+
+        gl.shaderSource(vs, 1, &[_][*c]const u8{vert_shader}, 0);
+        gl.compileShader(vs);
+
+        if (!shader.didCompile(vs)) return error.VertexCompileError;
+
+        const fs = gl.createShader(gl.FRAGMENT_SHADER);
+        defer gl.deleteShader(fs);
+
+        gl.shaderSource(fs, 1, &[_][*c]const u8{frag_shader}, 0);
+        gl.compileShader(fs);
+
+        if (!shader.didCompile(fs)) return error.FragmentCompileError;
+
+        const prog = gl.createProgram();
+        gl.attachShader(prog, vs);
+        gl.attachShader(prog, fs);
+        gl.linkProgram(prog);
+
+        return prog;
+    }
+
+    fn vao() GLuint {
+        var vao_id: GLuint = undefined;
+        gl.genVertexArrays(1, &vao_id);
+
+        return vao_id;
+    }
+
+    fn screenTex(buf: []const u8) GLuint {
+        var tex_id: GLuint = undefined;
+        gl.genTextures(1, &tex_id);
+
+        gl.bindTexture(gl.TEXTURE_2D, tex_id);
+        defer gl.bindTexture(gl.TEXTURE_2D, 0);
+
+        gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.NEAREST);
+        gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.NEAREST);
+
+        gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, gba_width, gba_height, 0, gl.RGBA, gl.UNSIGNED_INT_8_8_8_8, buf.ptr);
+
+        return tex_id;
+    }
+
+    fn outTex() GLuint {
+        var tex_id: GLuint = undefined;
+        gl.genTextures(1, &tex_id);
+
+        gl.bindTexture(gl.TEXTURE_2D, tex_id);
+        defer gl.bindTexture(gl.TEXTURE_2D, 0);
+
+        gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.NEAREST);
+        gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.NEAREST);
+
+        gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, gba_width, gba_height, 0, gl.RGBA, gl.UNSIGNED_INT_8_8_8_8, null);
+
+        return tex_id;
+    }
+
+    fn frameBuffer(tex_id: GLuint) !GLuint {
+        var fbo_id: GLuint = undefined;
+        gl.genFramebuffers(1, &fbo_id);
+
+        gl.bindFramebuffer(gl.FRAMEBUFFER, fbo_id);
+        defer gl.bindFramebuffer(gl.FRAMEBUFFER, 0);
+
+        gl.framebufferTexture(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, tex_id, 0);
+        gl.drawBuffers(1, &@as(GLuint, gl.COLOR_ATTACHMENT0));
+
+        if (gl.checkFramebufferStatus(gl.FRAMEBUFFER) != gl.FRAMEBUFFER_COMPLETE)
+            return error.FrameBufferObejctInitFailed;
+
+        return fbo_id;
+    }
+
+    const shader = struct {
+        const log = std.log.scoped(.shader);
+
+        fn didCompile(id: gl.GLuint) bool {
+            var success: gl.GLint = undefined;
+            gl.getShaderiv(id, gl.COMPILE_STATUS, &success);
+
+            if (success == 0) err(id);
+
+            return success == 1;
+        }
+
+        fn err(id: gl.GLuint) void {
+            const buf_len = 512;
+            var error_msg: [buf_len]u8 = undefined;
+
+            gl.getShaderInfoLog(id, buf_len, 0, &error_msg);
+            log.err("{s}", .{std.mem.sliceTo(&error_msg, 0)});
+        }
+    };
+};

src/shader/pixelbuf.frag

@@ -0,0 +1,24 @@
+#version 330 core
+out vec4 frag_color;
+
+in vec2 uv;
+
+uniform sampler2D screen;
+
+void main() {
+	// https://near.sh/video/color-emulation
+	// Thanks to Talarubi + Near for the Colour Correction
+	// Thanks to fleur + mattrb for the Shader Impl
+
+	vec4 color = texture(screen, uv);
+	color.rgb = pow(color.rgb, vec3(4.0)); // LCD Gamma
+  
+	frag_color = vec4(
+		pow(vec3(
+		  	  0 * color.b +  50 * color.g + 255 * color.r,
+	     	 30 * color.b + 230 * color.g +  10 * color.r,
+			220 * color.b +  10 * color.g +  50 * color.r
+		) / 255, vec3(1.0 / 2.2)), // Out Gamma
+	1.0); 
+}
+

src/shader/pixelbuf.vert

@@ -0,0 +1,10 @@
+#version 330 core
+out vec2 uv;
+
+const vec2 pos[3] = vec2[3](vec2(-1.0f, -1.0f), vec2(-1.0f, 3.0f), vec2(3.0f, -1.0f));
+const vec2 uvs[3] = vec2[3](vec2( 0.0f,  0.0f), vec2( 0.0f, 2.0f), vec2(2.0f,  0.0f));
+
+void main() {
+	uv = uvs[gl_VertexID];
+	gl_Position = vec4(pos[gl_VertexID], 0.0, 1.0);
+}

src/util.zig

@@ -3,51 +3,32 @@ const builtin = @import("builtin");
 const config = @import("config.zig");
 
 const Log2Int = std.math.Log2Int;
-const Arm7tdmi = @import("core/cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
 
-// Sign-Extend value of type `T` to type `U`
-pub fn sext(comptime T: type, comptime U: type, value: T) T {
-    // U must have less bits than T
-    comptime std.debug.assert(@typeInfo(U).Int.bits <= @typeInfo(T).Int.bits);
-
-    const iT = std.meta.Int(.signed, @typeInfo(T).Int.bits);
-    const ExtU = if (@typeInfo(U).Int.signedness == .unsigned) T else iT;
-    const shift = @intCast(Log2Int(T), @typeInfo(T).Int.bits - @typeInfo(U).Int.bits);
-
-    return @bitCast(T, @bitCast(iT, @as(ExtU, @truncate(U, value)) << shift) >> shift);
-}
-
-/// See https://godbolt.org/z/W3en9Eche
-pub inline fn rotr(comptime T: type, x: T, r: anytype) T {
-    if (@typeInfo(T).Int.signedness == .signed)
-        @compileError("cannot rotate signed integer");
-
-    const ar = @intCast(Log2Int(T), @mod(r, @typeInfo(T).Int.bits));
-    return x >> ar | x << (1 +% ~ar);
-}
+const Allocator = std.mem.Allocator;
 
 pub const FpsTracker = struct {
     const Self = @This();
 
     fps: u32,
-    count: std.atomic.Atomic(u32),
+    count: std.atomic.Value(u32),
     timer: std.time.Timer,
 
     pub fn init() Self {
         return .{
             .fps = 0,
-            .count = std.atomic.Atomic(u32).init(0),
+            .count = std.atomic.Value(u32).init(0),
             .timer = std.time.Timer.start() catch unreachable,
         };
     }
 
     pub fn tick(self: *Self) void {
-        _ = self.count.fetchAdd(1, .Monotonic);
+        _ = self.count.fetchAdd(1, .monotonic);
     }
 
     pub fn value(self: *Self) u32 {
         if (self.timer.read() >= std.time.ns_per_s) {
-            self.fps = self.count.swap(0, .SeqCst);
+            self.fps = self.count.swap(0, .monotonic);
             self.timer.reset();
         }
 
@@ -55,68 +36,6 @@ pub const FpsTracker = struct {
     }
 };
 
-pub fn intToBytes(comptime T: type, value: anytype) [@sizeOf(T)]u8 {
-    comptime std.debug.assert(@typeInfo(T) == .Int);
-
-    var result: [@sizeOf(T)]u8 = undefined;
-
-    var i: Log2Int(T) = 0;
-    while (i < result.len) : (i += 1) result[i] = @truncate(u8, value >> i * @bitSizeOf(u8));
-
-    return result;
-}
-
-/// The Title from the GBA Cartridge is an Uppercase ASCII string which is
-/// null-padded to 12 bytes
-///
-/// This function returns a slice of the ASCII string without the null terminator(s)
-/// (essentially, a proper Zig/Rust/Any modern language String)
-pub fn span(title: *const [12]u8) []const u8 {
-    const end = std.mem.indexOfScalar(u8, title, '\x00');
-    return title[0 .. end orelse title.len];
-}
-
-test "span" {
-    var example: *const [12]u8 = "POKEMON_EMER";
-    try std.testing.expectEqualSlices(u8, "POKEMON_EMER", span(example));
-
-    example = "POKEMON_EME\x00";
-    try std.testing.expectEqualSlices(u8, "POKEMON_EME", span(example));
-
-    example = "POKEMON_EM\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKEMON_EM", span(example));
-
-    example = "POKEMON_E\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKEMON_E", span(example));
-
-    example = "POKEMON_\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKEMON_", span(example));
-
-    example = "POKEMON\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKEMON", span(example));
-
-    example = "POKEMO\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKEMO", span(example));
-
-    example = "POKEM\x00\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKEM", span(example));
-
-    example = "POKE\x00\x00\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POKE", span(example));
-
-    example = "POK\x00\x00\x00\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "POK", span(example));
-
-    example = "PO\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "PO", span(example));
-
-    example = "P\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "P", span(example));
-
-    example = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00";
-    try std.testing.expectEqualSlices(u8, "", span(example));
-}
-
 /// Creates a copy of a title with all Filesystem-invalid characters replaced
 ///
 /// e.g. POKEPIN R/S to POKEPIN R_S
@@ -131,9 +50,15 @@ pub fn escape(title: [12]u8) [12]u8 {
 }
 
 pub const FilePaths = struct {
-    rom: []const u8,
+    rom: ?[]const u8,
     bios: ?[]const u8,
-    save: ?[]const u8,
+    save: []const u8,
+
+    pub fn deinit(self: @This(), allocator: Allocator) void {
+        if (self.rom) |path| allocator.free(path);
+        if (self.bios) |path| allocator.free(path);
+        allocator.free(self.save);
+    }
 };
 
 pub const io = struct {
@@ -143,7 +68,9 @@ pub const io = struct {
             return 0;
         }
 
-        pub fn undef(comptime T: type, log: anytype, comptime format: []const u8, args: anytype) ?T {
+        pub fn undef(comptime T: type, comptime log: anytype, comptime format: []const u8, args: anytype) ?T {
+            @setCold(true);
+
             const unhandled_io = config.config().debug.unhandled_io;
 
             log.warn(format, args);
@@ -151,6 +78,13 @@ pub const io = struct {
 
             return null;
         }
+
+        pub fn err(comptime T: type, comptime log: anytype, comptime format: []const u8, args: anytype) ?T {
+            @setCold(true);
+
+            log.err(format, args);
+            return null;
+        }
     };
 
     pub const write = struct {
@@ -165,6 +99,7 @@ pub const io = struct {
 
 pub const Logger = struct {
     const Self = @This();
+    const FmtArgTuple = std.meta.Tuple(&.{ u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32 });
 
     buf: std.io.BufferedWriter(4096 << 2, std.fs.File.Writer),
 
@@ -176,6 +111,7 @@ pub const Logger = struct {
 
     pub fn print(self: *Self, comptime format: []const u8, args: anytype) !void {
         try self.buf.writer().print(format, args);
+        try self.buf.flush(); // FIXME: On panics, whatever is in the buffer isn't written to file
     }
 
     pub fn mgbaLog(self: *Self, cpu: *const Arm7tdmi, opcode: u32) void {
@@ -186,7 +122,7 @@ pub const Logger = struct {
         if (cpu.cpsr.t.read()) {
             if (opcode >> 11 == 0x1E) {
                 // Instruction 1 of a BL Opcode, print in ARM mode
-                const low = cpu.bus.dbgRead(u16, cpu.r[15]);
+                const low = cpu.bus.dbgRead(u16, cpu.r[15] - 2);
                 const bl_opcode = @as(u32, opcode) << 16 | low;
 
                 self.print(arm_fmt, Self.fmtArgs(cpu, bl_opcode)) catch @panic("failed to write to log file");
@@ -215,15 +151,13 @@ pub const Logger = struct {
             cpu.r[12],
             cpu.r[13],
             cpu.r[14],
-            cpu.r[15],
+            cpu.r[15] - if (cpu.cpsr.t.read()) 2 else @as(u32, 4),
             cpu.cpsr.raw,
             opcode,
         };
     }
 };
 
-const FmtArgTuple = std.meta.Tuple(&.{ u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32 });
-
 pub const audio = struct {
     const _io = @import("core/bus/io.zig");
 
@@ -273,22 +207,37 @@ pub const audio = struct {
     };
 };
 
-/// Sets the high bits of an integer to a value
-pub inline fn setHi(comptime T: type, left: T, right: HalfInt(T)) T {
-    return switch (T) {
-        u32 => (left & 0xFFFF_0000) | right,
-        u16 => (left & 0xFF00) | right,
-        u8 => (left & 0xF0) | right,
-        else => @compileError("unsupported type"),
+/// Sets a quarter (8) of the bits of the u32 `left` to the value of u8 `right`
+pub inline fn setQuart(left: u32, addr: u8, right: u8) u32 {
+    const offset: u2 = @truncate(addr);
+
+    return switch (offset) {
+        0b00 => (left & 0xFFFF_FF00) | right,
+        0b01 => (left & 0xFFFF_00FF) | @as(u32, right) << 8,
+        0b10 => (left & 0xFF00_FFFF) | @as(u32, right) << 16,
+        0b11 => (left & 0x00FF_FFFF) | @as(u32, right) << 24,
     };
 }
 
-/// sets the low bits of an integer to a value
-pub inline fn setLo(comptime T: type, left: T, right: HalfInt(T)) T {
+/// Calculates the correct shift offset for an aligned/unaligned u8 read
+///
+/// TODO: Support u16 reads of u32 values?
+pub inline fn getHalf(byte: u8) u4 {
+    return @as(u4, @truncate(byte & 1)) << 3;
+}
+
+pub inline fn setHalf(comptime T: type, left: T, addr: u8, right: HalfInt(T)) T {
+    const offset: u1 = @truncate(addr >> if (T == u32) 1 else 0);
+
     return switch (T) {
-        u32 => (left & 0x0000_FFFF) | @as(u32, right) << 16,
-        u16 => (left & 0x00FF) | @as(u16, right) << 8,
-        u8 => (left & 0x0F) | @as(u8, right) << 4,
+        u32 => switch (offset) {
+            0b0 => (left & 0xFFFF_0000) | right,
+            0b1 => (left & 0x0000_FFFF) | @as(u32, right) << 16,
+        },
+        u16 => switch (offset) {
+            0b0 => (left & 0xFF00) | right,
+            0b1 => (left & 0x00FF) | @as(u16, right) << 8,
+        },
         else => @compileError("unsupported type"),
     };
 }
@@ -301,3 +250,76 @@ fn HalfInt(comptime T: type) type {
 
     return std.meta.Int(type_info.Int.signedness, type_info.Int.bits >> 1);
 }
+
+/// Double Buffering Implementation
+pub const FrameBuffer = struct {
+    const Self = @This();
+
+    layers: [2][]u8,
+    buf: []u8,
+    current: u1 = 0,
+
+    allocator: Allocator,
+
+    // TODO: Rename
+    const Device = enum { Emulator, Renderer };
+
+    pub fn init(allocator: Allocator, comptime len: comptime_int) !Self {
+        const buf = try allocator.alloc(u8, len * 2);
+        @memset(buf, 0);
+
+        return .{
+            // Front and Back Framebuffers
+            .layers = [_][]u8{ buf[0..][0..len], buf[len..][0..len] },
+            .buf = buf,
+
+            .allocator = allocator,
+        };
+    }
+
+    pub fn reset(self: *Self) void {
+        @memset(self.buf, 0);
+        self.current = 0;
+    }
+
+    pub fn deinit(self: *Self) void {
+        self.allocator.free(self.buf);
+        self.* = undefined;
+    }
+
+    pub fn swap(self: *Self) void {
+        self.current = ~self.current;
+    }
+
+    pub fn get(self: *Self, comptime dev: Device) []u8 {
+        return self.layers[if (dev == .Emulator) self.current else ~self.current];
+    }
+};
+
+const RingBuffer = @import("zba-util").RingBuffer;
+
+// TODO: Lock Free Queue?
+pub fn Queue(comptime T: type) type {
+    return struct {
+        inner: RingBuffer(T),
+        mtx: std.Thread.Mutex = .{},
+
+        pub fn init(buf: []T) @This() {
+            return .{ .inner = RingBuffer(T).init(buf) };
+        }
+
+        pub fn push(self: *@This(), value: T) !void {
+            self.mtx.lock();
+            defer self.mtx.unlock();
+
+            try self.inner.push(value);
+        }
+
+        pub fn pop(self: *@This()) ?T {
+            self.mtx.lock();
+            defer self.mtx.unlock();
+
+            return self.inner.pop();
+        }
+    };
+}