feat: upgrade to zig v0.15.1

note: emu crashes for unknown reason

.github/workflows/main.yml

@@ -4,6 +4,7 @@ on:
   push:
     paths:
       - "**.zig"
+      - "dl_sdl2.ps1"
     branches:
       - main
   schedule:
@@ -14,45 +15,45 @@ jobs:
   build:
     strategy:
       matrix:
-        os: [ubuntu-latest, windows-latest, macos-latest]
+        os: [ubuntu-latest, windows-latest] # TODO: Figure out Apple Silicon macOS
     runs-on: ${{matrix.os}}
     steps:
       - uses: goto-bus-stop/setup-zig@v2
         with:
-          version: master
+          version: 0.13.0
+      - run: |
+            git config --global core.autocrlf false
+      - uses: actions/checkout@v3
+        with:
+          submodules: recursive
       - name: prepare-linux
         if: runner.os == 'Linux'
         run: |
-            sudo apt-get update
-            sudo apt-get install libsdl2-dev
+            sudo apt update
+            sudo apt install libgtk-3-dev libsdl2-dev
       - name: prepare-windows
         if: runner.os == 'Windows'
         run: |
-            vcpkg integrate install
-            vcpkg install sdl2:x64-windows
-            git config --global core.autocrlf false
+            .\dl_sdl2.ps1
       - name: prepare-macos
         if: runner.os == 'macOS'
         run: |
             brew install sdl2
-      - uses: actions/checkout@v3
-        with:
-          submodules: true
       - name: build 
-        run: zig build -Doptimize=ReleaseSafe
+        run: zig build -Doptimize=ReleaseSafe -Dcpu=baseline
       - name: upload
         uses: actions/upload-artifact@v3
         with:
           name: zba-${{matrix.os}}
-          path: zig-out/bin
+          path: zig-out
+
   lint:
     runs-on: ubuntu-latest
     steps: 
       - uses: actions/checkout@v3
         with:
-          submodules: true
+          submodules: recursive
       - uses: goto-bus-stop/setup-zig@v2
         with:
-          version: master
-      - run: zig fmt src/**/*.zig
-  
+          version: 0.13.0
+      - run: zig fmt --check {src,lib}/**/*.zig build.zig build.zig.zon

.gitignore

@@ -1,6 +1,7 @@
 /.vscode
 /bin
 **/zig-cache
+**/.zig-cache
 **/zig-out
 /docs
 **/*.log
@@ -11,4 +12,8 @@
 /lib/SDL2
 
 # Any Custom Scripts for Debugging purposes
-*.sh
+*.sh
+
+
+# Dear ImGui
+**/imgui.ini

.gitmodules

@@ -1,15 +0,0 @@
-[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

README.md

@@ -2,6 +2,8 @@
 
 A Game Boy Advance Emulator written in Zig ⚡!
 
+![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).
@@ -13,82 +15,34 @@ This is a simple (read: incomplete) for-fun long-term project. I hope to get "mo
 - [x] Affine Sprites
 - [ ] Windowing (see [this branch](https://git.musuka.dev/paoda/zba/src/branch/window))
 - [ ] Audio Resampler (Having issues with SDL2's)
-- [x] Immediate Mode GUI (see [this branch](https://git.musuka.dev/paoda/zba/src/branch/imgui))
 - [ ] Refactoring for easy-ish perf boosts
 
 ## Usage
 
-As it currently exists, ZBA is run from the terminal. In your console of choice, type `./zba --help` to see what you can do.
+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.
 
-I typically find myself typing `./zba -b ./bin/bios.bin ./bin/test/suite.gba` to see how badly my "cool new feature" broke everything else.
-
-Need a BIOS? 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)?
+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.
 
-## Tests
-
-GBA Tests | [jsmolka](https://github.com/jsmolka/)
---- | ---
-`arm.gba`,  `thumb.gba` | PASS
-`memory.gba`, `bios.gba` | PASS
-`flash64.gba`, `flash128.gba` | PASS
-`sram.gba` | PASS
-`none.gba` | PASS
-`hello.gba`, `shades.gba`, `stripes.gba` | PASS
-`nes.gba` | PASS
-
-GBARoms | [DenSinH](https://github.com/DenSinH/)
---- | ---
-`eeprom-test`, `flash-test` | PASS
-`midikey2freq` | PASS
-`swi-tests-random` | FAIL
-
-gba_tests | [destoer](https://github.com/destoer/)
---- | ---
-`cond_invalid.gba` | PASS
-`dma_priority.gba` | PASS
-`hello_world.gba` | PASS
-`if_ack.gba` | PASS
-`line_timing.gba` | FAIL
-`lyc_midline.gba` | FAIL
-`window_midframe.gba` | FAIL
-
-GBA Test Collection | [ladystarbreeze](https://github.com/ladystarbreeze)
---- | ---
-`retAddr.gba` | PASS
-`helloWorld.gba` | PASS
-`helloAudio.gba` | PASS
-
-FuzzARM | [DenSinH](https://github.com/DenSinH/)
---- | ---
-`main.gba` | PASS
-
-arm7wrestler GBA Fixed | [destoer](https://github.com/destoer)
---- | ---
-`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)
-
 ## Compiling
 
-Most recently built on Zig [v0.11.0-dev.1557+03cdb4fb5](https://github.com/ziglang/zig/tree/03cdb4fb5)
+Most recently built on Zig [v0.11.0](https://github.com/ziglang/zig/tree/0.11.0)
 
 ### Dependencies
 
 Dependency | Source
 --- | ---
-SDL.zig | <https://github.com/MasterQ32/SDL.zig>
-zig-clap | <https://github.com/Hejsil/zig-clap>
 known-folders | <https://github.com/ziglibs/known-folders>
-zig-toml | <https://github.com/aeronavery/zig-toml>
+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>
-`bitfields.zig` | [https://github.com/FlorenceOS/Florence](https://github.com/FlorenceOS/Florence/blob/aaa5a9e568/lib/util/bitfields.zig)
+`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 relevant git submodules
@@ -96,21 +50,47 @@ Use `git submodule update --init` from the project root to pull the git relevant
 Be sure to provide SDL2 using:
 
 - Linux: Your distro's package manager
-- MacOS: `brew` (install [this formula](https://formulae.brew.sh/formula/sdl2))
+- macOS: ¯\\\_(ツ)_/¯ (try [this formula](https://formulae.brew.sh/formula/sdl2)?)
 - Windows: [`vcpkg`](https://github.com/Microsoft/vcpkg) (install `sdl2:x64-windows`)
 
 `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 is located at `zig-out/bin/`.
+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

@@ -1,68 +1,154 @@
 const std = @import("std");
-const builtin = @import("builtin");
-const Sdk = @import("lib/SDL.zig/Sdk.zig");
-
-pub fn build(b: *std.build.Builder) void {
-    // Minimum Zig Version
-    const min_ver = std.SemanticVersion.parse("0.11.0-dev.1557+03cdb4fb5") catch return; // https://github.com/ziglang/zig/commit/03cdb4fb5
-    if (builtin.zig_version.order(min_ver).compare(.lt)) {
-        std.log.err("{s}", .{b.fmt("Zig v{} does not meet the minimum version requirement. (Zig v{})", .{ builtin.zig_version, min_ver })});
-        std.os.exit(1);
-    }
 
+// Although this function looks imperative, it does not perform the build
+// directly and instead it mutates the build graph (`b`) that will be then
+// executed by an external runner. The functions in `std.Build` implement a DSL
+// for defining build steps and express dependencies between them, allowing the
+// build runner to parallelize the build automatically (and the cache system to
+// know when a step doesn't need to be re-run).
+pub fn build(b: *std.Build) void {
+    // Standard target options allow 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(.{});
+    // Standard optimization options allow the person running `zig build` to select
+    // between Debug, ReleaseSafe, ReleaseFast, and ReleaseSmall. Here we do not
+    // set a preferred release mode, allowing the user to decide how to optimize.
     const optimize = b.standardOptimizeOption(.{});
+    // It's also possible to define more custom flags to toggle optional features
+    // of this build script using `b.option()`. All defined flags (including
+    // target and optimize options) will be listed when running `zig build --help`
+    // in this directory.
 
-    const exe = b.addExecutable(.{
-        .name = "zba",
-        .root_source_file = .{ .path = "src/main.zig" },
+    // This creates a module, which represents a collection of source files alongside
+    // some compilation options, such as optimization mode and linked system libraries.
+    // Zig modules are the preferred way of making Zig code available to consumers.
+    // addModule defines a module that we intend to make available for importing
+    // to our consumers. We must give it a name because a Zig package can expose
+    // multiple modules and consumers will need to be able to specify which
+    // module they want to access.
+    const exe_mod = b.addModule("zba", .{
+        // The root source file is the "entry point" of this module. Users of
+        // this module will only be able to access public declarations contained
+        // in this file, which means that if you have declarations that you
+        // intend to expose to consumers that were defined in other files part
+        // of this module, you will have to make sure to re-export them from
+        // the root file.
+        .root_source_file = b.path("src/main.zig"),
+        // Later on we'll use this module as the root module of a test executable
+        // which requires us to specify a target.
         .target = target,
-        .optimize = optimize,
+        .link_libc = true,
     });
 
-    exe.setMainPkgPath("."); // Necessary so that src/main.zig can embed example.toml
+    // Here we define an executable. An executable needs to have a root module
+    // which needs to expose a `main` function. While we could add a main function
+    // to the module defined above, it's sometimes preferable to split business
+    // business logic and the CLI into two separate modules.
+    //
+    // If your goal is to create a Zig library for others to use, consider if
+    // it might benefit from also exposing a CLI tool. A parser library for a
+    // data serialization format could also bundle a CLI syntax checker, for example.
+    //
+    // If instead your goal is to create an executable, consider if users might
+    // be interested in also being able to embed the core functionality of your
+    // program in their own executable in order to avoid the overhead involved in
+    // subprocessing your CLI tool.
+    //
+    // If neither case applies to you, feel free to delete the declaration you
+    // don't need and to put everything under a single module.
+    const exe = b.addExecutable(.{ .name = "zba", .root_module = exe_mod });
 
-    // Known Folders (%APPDATA%, XDG, etc.)
-    exe.addPackagePath("known_folders", "lib/known-folders/known-folders.zig");
+    const zgui = b.dependency("zgui", .{ .shared = false, .with_implot = true, .backend = .sdl3_opengl3 });
+    const sdl = b.dependency("sdl", .{ .target = target, .optimize = optimize, .preferred_linkage = .static });
+    const gl = @import("zigglgen").generateBindingsModule(b, .{ .api = .gl, .version = .@"3.3", .profile = .core });
 
-    // DateTime Library
-    exe.addPackagePath("datetime", "lib/zig-datetime/src/main.zig");
+    const sdl_lib = sdl.artifact("SDL3");
+    const zgui_lib = zgui.artifact("imgui");
 
-    // 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");
+    exe_mod.linkLibrary(sdl_lib);
+    exe_mod.linkLibrary(zgui_lib);
 
-    // Argument Parsing Library
-    exe.addPackagePath("clap", "lib/zig-clap/clap.zig");
+    exe_mod.addImport("gl", gl);
+    exe_mod.addImport("known_folders", b.dependency("known_folders", .{}).module("known-folders"));
+    exe_mod.addImport("datetime", b.dependency("datetime", .{}).module("datetime"));
+    exe_mod.addImport("clap", b.dependency("clap", .{}).module("clap"));
+    exe_mod.addImport("zba_util", b.dependency("zba_util", .{}).module("zba_util"));
+    exe_mod.addImport("arm32", b.dependency("arm32", .{}).module("arm32"));
+    exe_mod.addImport("gdbstub", b.dependency("zba_gdbstub", .{}).module("zba_gdbstub"));
+    exe_mod.addImport("nfd", b.dependency("nfdzig", .{}).module("nfd"));
+    exe_mod.addImport("zgui", zgui.module("root"));
+    exe_mod.addImport("bitjuggle", b.dependency("bitjuggle", .{}).module("bitjuggle"));
 
-    // TOML Library
-    exe.addPackagePath("toml", "lib/zig-toml/src/toml.zig");
+    // This declares intent for the executable to be installed into the
+    // install prefix when running `zig build` (i.e. when executing the default
+    // step). By default the install prefix is `zig-out/` but can be overridden
+    // by passing `--prefix` or `-p`.
+    b.installArtifact(exe);
 
-    // OpenGL 3.3 Bindings
-    exe.addPackagePath("gl", "lib/gl.zig");
+    // This creates a top level step. Top level steps have a name and can be
+    // invoked by name when running `zig build` (e.g. `zig build run`).
+    // This will evaluate the `run` step rather than the default step.
+    // For a top level step to actually do something, it must depend on other
+    // steps (e.g. a Run step, as we will see in a moment).
+    const run_step = b.step("run", "Run the app");
 
-    // Zig SDL Bindings: https://github.com/MasterQ32/SDL.zig
-    const sdk = Sdk.init(b, null);
-    sdk.link(exe, .dynamic);
-    exe.addPackage(sdk.getNativePackage("sdl2"));
+    // This creates a RunArtifact step in the build graph. A RunArtifact step
+    // invokes an executable compiled by Zig. Steps will only be executed by the
+    // runner if invoked directly by the user (in the case of top level steps)
+    // or if another step depends on it, so it's up to you to define when and
+    // how this Run step will be executed. In our case we want to run it when
+    // the user runs `zig build run`, so we create a dependency link.
+    const run_cmd = b.addRunArtifact(exe);
+    run_step.dependOn(&run_cmd.step);
 
-    exe.install();
-
-    const run_cmd = exe.run();
+    // By making the run step depend on the default step, it will be run from the
+    // installation directory rather than directly from within the cache directory.
     run_cmd.step.dependOn(b.getInstallStep());
+
+    // This allows the user to pass arguments to the application in the build
+    // command itself, like this: `zig build run -- arg1 arg2 etc`
     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(.{
-        .root_source_file = .{ .path = "src/main.zig" },
-        .target = target,
-        .optimize = optimize,
+    // Creates an executable that will run `test` blocks from the provided module.
+    // Here `mod` needs to define a target, which is why earlier we made sure to
+    // set the releative field.
+    const mod_tests = b.addTest(.{
+        .root_module = exe_mod,
     });
 
-    const test_step = b.step("test", "Run unit tests");
-    test_step.dependOn(&exe_tests.step);
+    // A run step that will run the test executable.
+    const run_mod_tests = b.addRunArtifact(mod_tests);
+
+    // Creates an executable that will run `test` blocks from the executable's
+    // root module. Note that test executables only test one module at a time,
+    // hence why we have to create two separate ones.
+    const exe_tests = b.addTest(.{
+        .root_module = exe.root_module,
+    });
+
+    // A run step that will run the second test executable.
+    const run_exe_tests = b.addRunArtifact(exe_tests);
+
+    // A top level step for running all tests. dependOn can be called multiple
+    // times and since the two run steps do not depend on one another, this will
+    // make the two of them run in parallel.
+    const test_step = b.step("test", "Run tests");
+    test_step.dependOn(&run_mod_tests.step);
+    test_step.dependOn(&run_exe_tests.step);
+
+    // Just like flags, top level steps are also listed in the `--help` menu.
+    //
+    // The Zig build system is entirely implemented in userland, which means
+    // that it cannot hook into private compiler APIs. All compilation work
+    // orchestrated by the build system will result in other Zig compiler
+    // subcommands being invoked with the right flags defined. You can observe
+    // these invocations when one fails (or you pass a flag to increase
+    // verbosity) to validate assumptions and diagnose problems.
+    //
+    // Lastly, the Zig build system is relatively simple and self-contained,
+    // and reading its source code will allow you to master it.
 }

build.zig.zon

@@ -0,0 +1,55 @@
+.{
+    .name = .zba,
+    .version = "0.1.0",
+    .paths = .{
+        "build.zig",
+        "build.zig.zon",
+        "lib/bitfield.zig",
+        "lib/gl.zig",
+        "src",
+    },
+    .minimum_zig_version = "0.15.1",
+    .fingerprint = 0xcb596c7fbdb20efc,
+    .dependencies = .{
+        // .tomlz = .{
+        //     .url = "git+https://github.com/paoda/tomlz#9a16dd53927ef2012478b6494bafb4475e44f4c9",
+        //     .hash = "12204f922cab84980e36b5c058d354ec0ee169bda401c8e0e80a463580349b476569",
+        // },
+
+        .zba_util = .{ .path = "../zba-util" },
+        .arm32 = .{ .path = "../arm32" },
+        .zba_gdbstub = .{ .path = "../zba-gdbstub" },
+        .known_folders = .{
+            .url = "git+https://github.com/ziglibs/known-folders.git#ab5cf5feb936fa3b72c95d3ad0c0c67791937ba1",
+            .hash = "known_folders-0.0.0-Fy-PJtTTAADUOhGKM0sxzG4eMkNQxRvx9e5dfHVyaeA3",
+        },
+        .nfdzig = .{
+            .url = "git+https://github.com/paoda/nfd-zig#0ad2a0c092ffba0c98613d619b82100c991f5ad6",
+            .hash = "nfdzig-0.1.0-11fxvN6IBgD5rvvfjrw1wPqibMsbUJ-h2ZcGR6FOEvrm",
+        },
+        .datetime = .{
+            .url = "git+https://github.com/frmdstryr/zig-datetime#3a39a21e6e34dcb0ade0ff828d0914d40ba535f3",
+            .hash = "datetime-0.8.0-cJNXzP_YAQBxQ5hkNNP6ScnG5XsqciJmeP5RVV4xwCBA",
+        },
+        .clap = .{
+            .url = "https://github.com/Hejsil/zig-clap/archive/refs/tags/0.11.0.tar.gz",
+            .hash = "clap-0.11.0-oBajB-HnAQDPCKYzwF7rO3qDFwRcD39Q0DALlTSz5H7e",
+        },
+        .sdl = .{
+            .url = "git+https://github.com/castholm/SDL.git#b1913e7c31ad72ecfd3ab04aeac387027754cfaf",
+            .hash = "sdl-0.3.0+3.2.22-7uIn9Pg3fwGG2IyIOPxxOSVe-75nUng9clt7tXGFLzMr",
+        },
+        .zigglgen = .{
+            .url = "git+https://github.com/castholm/zigglgen.git#a1e969b3e35818785fab0373124f50463793b28a",
+            .hash = "zigglgen-0.4.0-bmyqLQGMLwA0EPVmSm-Nc6Olb84zBzvwMjguxwKYLf1S",
+        },
+        .bitjuggle = .{
+            .url = "git+https://github.com/leecannon/zig-bitjuggle#80111f4f8c672aaea94a8a189ae2a7c8bbaf883f",
+            .hash = "bitjuggle-2.0.0-SJdU76dvAAARompHEhqKDiwZ4FE4FZ8eHvPvmz5JUOS0",
+        },
+        .zgui = .{
+            .url = "git+https://github.com/zig-gamedev/zgui#7fa8081c208885b85e3fdfc043cd9d9cb9559123",
+            .hash = "zgui-0.6.0-dev--L6sZL7tbQAPRLYrcQAVx0V49tPHAXNxclZ-v8IP4wLr",
+        },
+    },
+}

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

@@ -1,4 +1,4 @@
-[Host]
+[host]
 # Using nearest-neighbour scaling, how many times the native resolution 
 # of the game bow should the screen be?
 win_scale = 3
@@ -7,7 +7,7 @@ vsync = true
 # Mute ZBA
 mute = false
 
-[Guest]
+[guest]
 # Sync Emulation to Audio 
 audio_sync = true
 # Sync Emulation to Video
@@ -17,7 +17,7 @@ force_rtc = false
 # Skip BIOS
 skip_bios = false
 
-[Debug]
+[debug]
 # Enable detailed CPU logs
 cpu_trace = false
 # When false and builtin.mode == .Debug, ZBA will panic

lib/util/bitfield.zig

@@ -1,146 +0,0 @@
-const std = @import("std");
-
-fn PtrCastPreserveCV(comptime T: type, comptime PtrToT: type, comptime NewT: type) type {
-    return switch (PtrToT) {
-        *T => *NewT,
-        *const T => *const NewT,
-        *volatile T => *volatile NewT,
-        *const volatile T => *const volatile NewT,
-
-        else => @compileError("wtf you doing"),
-    };
-}
-
-fn BitType(comptime FieldType: type, comptime ValueType: type, comptime shamt: usize) type {
-    const self_bit: FieldType = (1 << shamt);
-
-    return extern struct {
-        bits: Bitfield(FieldType, shamt, 1),
-
-        pub fn set(self: anytype) void {
-            self.bits.field().* |= self_bit;
-        }
-
-        pub fn unset(self: anytype) void {
-            self.bits.field().* &= ~self_bit;
-        }
-
-        pub fn read(self: anytype) ValueType {
-            return @bitCast(ValueType, @truncate(u1, 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)) {
-                self.set();
-            } else {
-                self.unset();
-            }
-        }
-    };
-}
-
-// Original Bit Constructor
-// pub fn Bit(comptime FieldType: type, comptime shamt: usize) type {
-//     return BitType(FieldType, u1, shamt);
-// }
-
-pub fn Bit(comptime FieldType: type, comptime shamt: usize) type {
-    return BitType(FieldType, bool, shamt);
-}
-
-fn Boolean(comptime FieldType: type, comptime shamt: usize) type {
-    return BitType(FieldType, bool, shamt);
-}
-
-pub fn Bitfield(comptime FieldType: type, comptime shamt: usize, comptime num_bits: usize) type {
-    if (shamt + num_bits > @bitSizeOf(FieldType)) {
-        @compileError("bitfield doesn't fit");
-    }
-
-    const self_mask: FieldType = ((1 << num_bits) - 1) << shamt;
-
-    const ValueType = std.meta.Int(.unsigned, num_bits);
-
-    return extern struct {
-        dummy: FieldType,
-
-        fn field(self: anytype) PtrCastPreserveCV(@This(), @TypeOf(self), FieldType) {
-            return @ptrCast(PtrCastPreserveCV(@This(), @TypeOf(self), FieldType), self);
-        }
-
-        pub fn write(self: anytype, val: ValueType) void {
-            self.field().* &= ~self_mask;
-            self.field().* |= @intCast(FieldType, val) << shamt;
-        }
-
-        pub fn read(self: anytype) ValueType {
-            const val: FieldType = self.field().*;
-            return @intCast(ValueType, (val & self_mask) >> shamt);
-        }
-    };
-}
-
-test "bit" {
-    const S = extern union {
-        low: Bit(u32, 0),
-        high: Bit(u32, 1),
-        val: u32,
-    };
-
-    std.testing.expect(@sizeOf(S) == 4);
-    std.testing.expect(@bitSizeOf(S) == 32);
-
-    var s: S = .{ .val = 1 };
-
-    std.testing.expect(s.low.read() == 1);
-    std.testing.expect(s.high.read() == 0);
-
-    s.low.write(0);
-    s.high.write(1);
-
-    std.testing.expect(s.val == 2);
-}
-
-test "boolean" {
-    const S = extern union {
-        low: Boolean(u32, 0),
-        high: Boolean(u32, 1),
-        val: u32,
-    };
-
-    std.testing.expect(@sizeOf(S) == 4);
-    std.testing.expect(@bitSizeOf(S) == 32);
-
-    var s: S = .{ .val = 2 };
-
-    std.testing.expect(s.low.read() == false);
-    std.testing.expect(s.high.read() == true);
-
-    s.low.write(true);
-    s.high.write(false);
-
-    std.testing.expect(s.val == 1);
-}
-
-test "bitfield" {
-    const S = extern union {
-        low: Bitfield(u32, 0, 16),
-        high: Bitfield(u32, 16, 16),
-        val: u32,
-    };
-
-    std.testing.expect(@sizeOf(S) == 4);
-    std.testing.expect(@bitSizeOf(S) == 32);
-
-    var s: S = .{ .val = 0x13376969 };
-
-    std.testing.expect(s.low.read() == 0x6969);
-    std.testing.expect(s.high.read() == 0x1337);
-
-    s.low.write(0x1337);
-    s.high.write(0x6969);
-
-    std.testing.expect(s.val == 0x69691337);
-}

src/config.zig

@@ -1,5 +1,5 @@
 const std = @import("std");
-const toml = @import("toml");
+// const tomlz = @import("tomlz");
 
 const Allocator = std.mem.Allocator;
 
@@ -7,6 +7,7 @@ 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 = .{},
@@ -58,29 +59,6 @@ pub fn load(allocator: Allocator, file_path: []const u8) !void {
     const contents = try config_file.readToEndAlloc(allocator, try config_file.getEndPos());
     defer allocator.free(contents);
 
-    var parser = try toml.parseFile(allocator, file_path);
-    defer parser.deinit();
-
-    const table = try parser.parse();
-    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 (host.Table.keys.get("mute")) |mute| state.host.mute = mute.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 (guest.Table.keys.get("skip_bios")) |skip| state.guest.skip_bios = skip.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;
-    }
+    // FIXME(2025-09-22): re-enable
+    // state = try tomlz.parser.decode(Config, allocator, contents);
 }

src/core/Bus.zig

@@ -1,6 +1,6 @@
 const std = @import("std");
 
-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");
@@ -13,13 +13,13 @@ const TimerTuple = @import("bus/timer.zig").TimerTuple;
 const Scheduler = @import("scheduler.zig").Scheduler;
 const FilePaths = @import("../util.zig").FilePaths;
 
-const io = @import("bus/io.zig");
+const _io = @import("bus/io.zig");
 const Allocator = std.mem.Allocator;
 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
@@ -81,7 +81,7 @@ pub fn init(self: *Self, allocator: Allocator, sched: *Scheduler, cpu: *Arm7tdmi
         .allocator = allocator,
     };
 
-    self.fillReadTable(read_table);
+    self.fillReadTable(@ptrCast(read_table));
 
     // Internal Display Memory behaves differently on 8-bit reads
     self.fillWriteTable(u32, write_tables[0]);
@@ -98,15 +98,54 @@ pub fn deinit(self: *Self) void {
     // 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(@ptrCast([*]const ?*anyopaque, self.read_table[0..])[0 .. 3 * table_len]);
+    self.allocator.free(@as([*]const ?*anyopaque, @ptrCast(self.read_table[0..]))[0 .. 3 * table_len]);
     self.* = undefined;
 }
 
+pub fn reset(self: *Self) void {
+    self.bios.reset();
+    self.ppu.reset();
+    self.apu.reset();
+    self.iwram.reset();
+    self.ewram.reset();
+
+    // https://github.com/ziglang/zig/issues/14705
+    {
+        comptime var i: usize = 0;
+        inline while (i < self.dma.len) : (i += 1) {
+            self.dma[0].reset();
+        }
+    }
+
+    // https://github.com/ziglang/zig/issues/14705
+    {
+        comptime var i: usize = 0;
+        inline while (i < self.tim.len) : (i += 1) {
+            self.tim[0].reset();
+        }
+    }
+
+    self.io.reset();
+}
+
+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);
+
+    // SAFETY: TODO: why do we know this is safe?
+    self.fillReadTable(@ptrCast(@constCast(self.read_table)));
+    self.fillWriteTable(u32, @constCast(self.write_tables[0]));
+    self.fillWriteTable(u8, @constCast(self.write_tables[1]));
+}
+
 fn fillReadTable(self: *Self, table: *[table_len]?*const anyopaque) void {
     const vramMirror = @import("ppu/Vram.zig").mirror;
 
-    for (table) |*ptr, i| {
-        const addr = @intCast(u32, page_size * i);
+    for (table, 0..) |*ptr, i| {
+        const addr: u32 = @intCast(page_size * i);
 
         ptr.* = switch (addr) {
             // General Internal Memory
@@ -128,12 +167,12 @@ fn fillReadTable(self: *Self, table: *[table_len]?*const anyopaque) void {
     }
 }
 
-fn fillWriteTable(self: *Self, comptime T: type, table: *[table_len]?*const anyopaque) void {
+fn fillWriteTable(self: *Self, comptime T: type, table: *[table_len]?*anyopaque) void {
     comptime std.debug.assert(T == u32 or T == u16 or T == u8);
     const vramMirror = @import("ppu/Vram.zig").mirror;
 
-    for (table) |*ptr, i| {
-        const addr = @intCast(u32, page_size * i);
+    for (table, 0..) |*ptr, i| {
+        const addr: u32 = @intCast(page_size * i);
 
         ptr.* = switch (addr) {
             // General Internal Memory
@@ -186,7 +225,7 @@ fn fillReadTableExternal(self: *Self, addr: u32) ?*anyopaque {
             // 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 (@truncate(u4, addr >> 24) == 0xD) return null;
+            if (@as(u4, @truncate(addr >> 24)) == 0xD) return null;
         }
     }
 
@@ -198,61 +237,12 @@ fn fillReadTableExternal(self: *Self, addr: u32) ?*anyopaque {
     return &self.pak.buf[masked_addr];
 }
 
-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;
-        const ptr = @ptrCast(Ptr, @alignCast(@alignOf(std.meta.Child(Ptr)), 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 dbgSlowRead(self: *const Self, comptime T: type, unaligned_address: u32) T {
-    const page = @truncate(u8, 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.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 readIo(self: *const Self, comptime T: type, address: u32) T {
-    return io.read(self, T, address) orelse self.openBus(T, address);
+    return _io.read(self, T, address) orelse self.openBus(T, address);
 }
 
 fn openBus(self: *const Self, comptime T: type, address: u32) T {
-    @setCold(true);
+    @branchHint(.cold);
     const r15 = self.cpu.r[15];
 
     const word = blk: {
@@ -265,7 +255,7 @@ fn openBus(self: *const Self, comptime T: type, address: u32) T {
         // the most recently fetched instruction by the pipeline
         if (!self.cpu.cpsr.t.read()) break :blk self.cpu.pipe.stage[1].?;
 
-        const page = @truncate(u8, r15 >> 24);
+        const page: u8 = @truncate(r15 >> 24);
 
         // PC + 2 = stage[0]
         // PC + 4 = stage[1]
@@ -274,7 +264,7 @@ fn openBus(self: *const Self, comptime T: type, address: u32) T {
         switch (page) {
             // EWRAM, PALRAM, VRAM, and Game ROM (16-bit)
             0x02, 0x05, 0x06, 0x08...0x0D => {
-                const halfword: u32 = @truncate(u16, self.cpu.pipe.stage[1].?);
+                const halfword: u32 = @as(u16, @truncate(self.cpu.pipe.stage[1].?));
                 break :blk halfword << 16 | halfword;
             },
 
@@ -285,8 +275,8 @@ fn openBus(self: *const Self, comptime T: type, address: u32) T {
                 const aligned = address & 3 == 0b00;
 
                 // TODO: What to do on PC + 6?
-                const high: u32 = if (aligned) self.dbgRead(u16, r15 + 4) else @truncate(u16, self.cpu.pipe.stage[1].?);
-                const low: u32 = @truncate(u16, self.cpu.pipe.stage[@boolToInt(aligned)].?);
+                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;
             },
@@ -297,8 +287,8 @@ fn openBus(self: *const Self, comptime T: type, address: u32) T {
                 // Unaligned: (PC + 4) | (PC + 2)
                 const aligned = address & 3 == 0b00;
 
-                const high: u32 = @truncate(u16, self.cpu.pipe.stage[1 - @boolToInt(aligned)].?);
-                const low: u32 = @truncate(u16, self.cpu.pipe.stage[@boolToInt(aligned)].?);
+                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;
             },
@@ -309,24 +299,23 @@ fn openBus(self: *const Self, comptime T: type, address: u32) T {
         }
     };
 
-    return @truncate(T, word);
+    return @truncate(word);
 }
 
 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 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[@boolToInt(T == u32)][@truncate(u4, unaligned_address >> 24)];
+    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;
-        const ptr = @ptrCast(Ptr, @alignCast(@alignOf(std.meta.Child(Ptr)), some_ptr));
+        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
@@ -336,10 +325,30 @@ pub fn read(self: *Self, comptime T: type, unaligned_address: u32) T {
     return self.slowRead(T, unaligned_address);
 }
 
-fn slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
-    @setCold(true);
+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);
 
-    const page = @truncate(u8, 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.dbgSlowRead(T, unaligned_address);
+}
+
+fn slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
+    @branchHint(.cold);
+
+    const page: u8 = @truncate(unaligned_address >> 24);
     const address = forceAlign(T, unaligned_address);
 
     return switch (page) {
@@ -366,6 +375,34 @@ fn slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
     };
 }
 
+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);
+
+    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);
 
@@ -380,36 +417,59 @@ fn readBackup(self: *const Self, comptime T: type, unaligned_address: u32) T {
 }
 
 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 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[@boolToInt(T == u32)][@truncate(u4, unaligned_address >> 24)];
+    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[@boolToInt(T == u8)][page]) |some_ptr| {
+    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;
-        const ptr = @ptrCast(Ptr, @alignCast(@alignOf(std.meta.Child(Ptr)), some_ptr));
+        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 @truncate(u8, unaligned_address >> 24) == 0x07) return;
+        if (T == u8 and @as(u8, @truncate(unaligned_address >> 24)) == 0x07) return;
 
         self.slowWrite(T, unaligned_address, value);
     }
 }
 
-fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
-    @setCold(true);
+/// 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);
 
-    const page = @truncate(u8, 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.dbgSlowWrite(T, unaligned_address, value);
+    }
+}
+
+fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    @branchHint(.cold);
+
+    const page: u8 = @truncate(unaligned_address >> 24);
     const address = forceAlign(T, unaligned_address);
 
     switch (page) {
@@ -417,7 +477,7 @@ fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) vo
         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),
+        0x04 => _io.write(self, T, address, value),
 
         // Internal Display Memory
         0x05 => self.ppu.palette.write(T, address, value),
@@ -426,7 +486,32 @@ fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) vo
 
         // External Memory (Game Pak)
         0x08...0x0D => self.pak.write(T, self.dma[3].word_count, address, value),
-        0x0E...0x0F => self.pak.backup.write(unaligned_address, @truncate(u8, rotr(T, value, 8 * rotateBy(T, unaligned_address)))),
+        0x0E...0x0F => self.pak.backup.write(unaligned_address, @truncate(rotr(T, value, 8 * rotateBy(T, unaligned_address)))),
+        else => {},
+    }
+}
+
+fn dbgSlowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    @branchHint(.cold);
+
+    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 => {},
     }
 }

src/core/apu.zig

@@ -1,20 +1,21 @@
 const std = @import("std");
-const SDL = @import("sdl2");
 const io = @import("bus/io.zig");
 const util = @import("../util.zig");
+const c = @import("../lib.zig").c;
 
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
+const LinearFifo = @import("../lib/fifo.zig").LinearFifo;
+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");
 const Wave = @import("apu/Wave.zig");
 const Noise = @import("apu/Noise.zig");
 
-const SoundFifo = std.fifo.LinearFifo(u8, .{ .Static = 0x20 });
+const SoundFifo = LinearFifo(u8, .{ .static = 0x20 });
 
 const getHalf = util.getHalf;
 const setHalf = util.setHalf;
-const intToBytes = util.intToBytes;
 
 const log = std.log.scoped(.APU);
 
@@ -22,7 +23,7 @@ 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_addr = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     return switch (T) {
         u32 => switch (byte_addr) {
@@ -73,27 +74,27 @@ pub fn read(comptime T: type, apu: *const Apu, addr: u32) ?T {
             else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
         },
         u8 => switch (byte_addr) {
-            0x60, 0x61 => @truncate(T, @as(u16, apu.ch1.sound1CntL()) >> getHalf(byte_addr)),
-            0x62, 0x63 => @truncate(T, apu.ch1.sound1CntH() >> getHalf(byte_addr)),
-            0x64, 0x65 => @truncate(T, apu.ch1.sound1CntX() >> getHalf(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(T, apu.ch2.sound2CntL() >> getHalf(byte_addr)),
+            0x68, 0x69 => @truncate(apu.ch2.sound2CntL() >> getHalf(byte_addr)),
             0x6A, 0x6B => 0x00,
-            0x6C, 0x6D => @truncate(T, apu.ch2.sound2CntH() >> getHalf(byte_addr)),
+            0x6C, 0x6D => @truncate(apu.ch2.sound2CntH() >> getHalf(byte_addr)),
             0x6E, 0x6F => 0x00,
-            0x70, 0x71 => @truncate(T, @as(u16, apu.ch3.sound3CntL()) >> getHalf(byte_addr)), // SOUND3CNT_L
-            0x72, 0x73 => @truncate(T, apu.ch3.sound3CntH() >> getHalf(byte_addr)),
-            0x74, 0x75 => @truncate(T, apu.ch3.sound3CntX() >> getHalf(byte_addr)), // SOUND3CNT_L
+            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(T, apu.ch4.sound4CntL() >> getHalf(byte_addr)),
+            0x78, 0x79 => @truncate(apu.ch4.sound4CntL() >> getHalf(byte_addr)),
             0x7A, 0x7B => 0x00,
-            0x7C, 0x7D => @truncate(T, apu.ch4.sound4CntH() >> getHalf(byte_addr)),
+            0x7C, 0x7D => @truncate(apu.ch4.sound4CntH() >> getHalf(byte_addr)),
             0x7E, 0x7F => 0x00,
-            0x80, 0x81 => @truncate(T, apu.soundCntL() >> getHalf(byte_addr)), // SOUNDCNT_L
-            0x82, 0x83 => @truncate(T, apu.soundCntH() >> getHalf(byte_addr)), // SOUNDCNT_H
-            0x84, 0x85 => @truncate(T, @as(u16, apu.soundCntX()) >> getHalf(byte_addr)),
+            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(T, apu.bias.raw >> getHalf(byte_addr)), // SOUNDBIAS
+            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),
@@ -106,7 +107,7 @@ pub fn read(comptime T: type, apu: *const Apu, addr: u32) ?T {
 }
 
 pub fn write(comptime T: type, apu: *Apu, addr: u32, value: T) void {
-    const byte_addr = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     if (byte_addr <= 0x81 and !apu.cnt.apu_enable.read()) return;
 
@@ -117,20 +118,20 @@ pub fn write(comptime T: type, apu: *Apu, addr: u32, value: T) void {
 
             switch (byte_addr) {
                 0x60 => apu.ch1.setSound1Cnt(value),
-                0x64 => apu.ch1.setSound1CntX(&apu.fs, @truncate(u16, value)),
+                0x64 => apu.ch1.setSound1CntX(&apu.fs, @truncate(value)),
 
-                0x68 => apu.ch2.setSound2CntL(@truncate(u16, value)),
-                0x6C => apu.ch2.setSound2CntH(&apu.fs, @truncate(u16, 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(u16, value)),
+                0x74 => apu.ch3.setSound3CntX(&apu.fs, @truncate(value)),
 
-                0x78 => apu.ch4.setSound4CntL(@truncate(u16, value)),
-                0x7C => apu.ch4.setSound4CntH(&apu.fs, @truncate(u16, 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(u16, value),
+                0x88 => apu.bias.raw = @truncate(value),
                 0x8C => {},
 
                 0x90, 0x94, 0x98, 0x9C => apu.ch3.wave_dev.write(T, apu.ch3.select, addr, value),
@@ -143,7 +144,7 @@ pub fn write(comptime T: type, apu: *Apu, addr: u32, value: T) void {
             if (byte_addr <= 0x81 and !apu.cnt.apu_enable.read()) return;
 
             switch (byte_addr) {
-                0x60 => apu.ch1.setSound1CntL(@truncate(u8, value)), // SOUND1CNT_L
+                0x60 => apu.ch1.setSound1CntL(@truncate(value)), // SOUND1CNT_L
                 0x62 => apu.ch1.setSound1CntH(value),
                 0x64 => apu.ch1.setSound1CntX(&apu.fs, value),
                 0x66 => {},
@@ -153,7 +154,7 @@ pub fn write(comptime T: type, apu: *Apu, addr: u32, value: T) void {
                 0x6C => apu.ch2.setSound2CntH(&apu.fs, value),
                 0x6E => {},
 
-                0x70 => apu.ch3.setSound3CntL(@truncate(u8, value)),
+                0x70 => apu.ch3.setSound3CntL(@truncate(value)),
                 0x72 => apu.ch3.setSound3CntH(value),
                 0x74 => apu.ch3.setSound3CntX(&apu.fs, value),
                 0x76 => {},
@@ -246,7 +247,7 @@ pub const Apu = struct {
 
     sampling_cycle: u2,
 
-    stream: *SDL.SDL_AudioStream,
+    stream: *c.SDL_AudioStream,
     sched: *Scheduler,
 
     fs: FrameSequencer,
@@ -271,25 +272,52 @@ pub const Apu = struct {
             .bias = .{ .raw = 0x0200 },
 
             .sampling_cycle = 0b00,
-            .stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, 1 << 15, host_format, 2, host_rate).?,
             .sched = sched,
+            .stream = undefined, // FIXME: bad practice
 
             .capacitor = 0,
             .fs = FrameSequencer.init(),
             .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();
@@ -303,8 +331,8 @@ pub const Apu = struct {
 
     /// SOUNDCNT
     fn setSoundCnt(self: *Self, value: u32) void {
-        if (self.cnt.apu_enable.read()) self.setSoundCntL(@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
@@ -351,18 +379,18 @@ pub const Apu = struct {
             // 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;
     }
@@ -423,8 +451,8 @@ pub const Apu = struct {
         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);
@@ -432,21 +460,33 @@ pub const Apu = struct {
 
         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));
+        const ret = c.SDL_PutAudioStreamData(self.stream, &[2]i16{ @bitCast(ext_left ^ 0x8000), @bitCast(ext_right ^ 0x8000) }, 2 * @sizeOf(i16));
+        if (!ret) @panic("TODO: Failed to put i16s into SDL Audio Queue");
     }
 
     fn replaceSDLResampler(self: *Self) void {
-        @setCold(true);
-        const sample_rate = Self.sampleRate(self.bias.sampling_cycle.read());
-        log.info("Sample Rate changed from {}Hz to {}Hz", .{ Self.sampleRate(self.sampling_cycle), sample_rate });
+        @branchHint(.cold);
+        _ = self;
 
-        // Sampling Cycle (Sample Rate) changed, Craete a new SDL Audio Resampler
-        // FIXME: Replace SDL's Audio Resampler with either a custom or more reliable one
-        const old_stream = self.stream;
-        defer SDL.SDL_FreeAudioStream(old_stream);
+        @panic("TODO: Implement Multiple Sample Rates...");
 
-        self.sampling_cycle = self.bias.sampling_cycle.read();
-        self.stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, @intCast(c_int, sample_rate), host_format, 2, host_rate).?;
+        // const sample_rate = Self.sampleRate(self.bias.sampling_cycle.read());
+        // log.info("Sample Rate changed from {}Hz to {}Hz", .{ Self.sampleRate(self.sampling_cycle), sample_rate });
+
+        // // Sampling Cycle (Sample Rate) changed, Craete a new SDL Audio Resampler
+        // // FIXME: Replace SDL's Audio Resampler with either a custom or more reliable one
+        // const old_stream = self.stream;
+        // defer c.SDL_DestroyAudioStream(old_stream);
+
+        // self.sampling_cycle = self.bias.sampling_cycle.read();
+
+        // var desired: c.SDL_AudioSpec = std.mem.zeroes(c.SDL_AudioSpec);
+        // desired.format = c.SDL_AUDIO_S16;
+        // desired.channels = 2;
+        // desired.freq = @intCast(sample_rate);
+
+        // const new_stream = c.SDL_OpenAudioDeviceStream(c.SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK, &desired, null, null) orelse @panic("TODO: Failed to replace SDL Audio Stream");
+        // self.stream = new_stream;
     }
 
     fn interval(self: *const Self) u64 {
@@ -494,18 +534,20 @@ 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);
         }
     }
 };
@@ -528,14 +570,19 @@ pub fn DmaSound(comptime kind: DmaSoundKind) type {
             };
         }
 
+        /// Used when resetting hte emulator (not emulation code)
+        fn reset(self: *Self) void {
+            self.* = Self.init();
+        }
+
         pub fn push(self: *Self, value: u32) void {
             if (!self.enabled) self.enable();
 
-            self.fifo.write(&intToBytes(u32, value)) catch |e| log.err("{} Error: {}", .{ kind, e });
+            self.fifo.write(std.mem.asBytes(&value)) catch |e| log.err("{} Error: {}", .{ kind, e });
         }
 
         fn enable(self: *Self) void {
-            @setCold(true);
+            @branchHint(.cold);
             self.enabled = true;
         }
 
@@ -544,11 +591,11 @@ 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;
         }
     };
 }
@@ -562,10 +609,14 @@ pub const FrameSequencer = struct {
     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

@@ -76,14 +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
@@ -99,8 +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

@@ -77,14 +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
@@ -100,8 +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

@@ -81,8 +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
@@ -111,14 +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
@@ -134,8 +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

@@ -64,8 +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
@@ -86,8 +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
@@ -98,8 +98,8 @@ 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

src/core/apu/device/Envelope.zig

@@ -3,17 +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.timer = 0;
-    self.vol = 0;
+    self.* = .{};
 }
 
 pub fn tick(self: *Self, nrx2: io.Envelope) void {

src/core/apu/device/Length.zig

@@ -1,13 +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.timer = 0;
+    self.* = .{};
 }
 
 pub fn tick(self: *Self, enabled: bool, ch_enable: *bool) void {

src/core/apu/device/Sweep.zig

@@ -3,26 +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.timer = 0;
-    self.enabled = false;
-    self.shadow = 0;
-    self.calc_performed = false;
+    self.* = .{};
 }
 
 pub fn tick(self: *Self, ch1: *ToneSweep) void {
@@ -36,8 +28,8 @@ pub fn tick(self: *Self, ch1: *ToneSweep) void {
             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);
             }

src/core/apu/signal/Wave.zig

@@ -18,7 +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 {
@@ -26,7 +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 {
@@ -70,7 +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

@@ -3,7 +3,7 @@ const std = @import("std");
 const Allocator = std.mem.Allocator;
 const log = std.log.scoped(.Bios);
 
-const rotr = @import("../../util.zig").rotr;
+const rotr = @import("zba_util").rotr;
 const forceAlign = @import("../Bus.zig").forceAlign;
 
 /// Size of the BIOS in bytes
@@ -27,7 +27,7 @@ pub fn read(self: *Self, comptime T: type, r15: u32, address: u32) T {
     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(T, rotr(u32, value, 8 * rotateBy(T, address)));
+    return @truncate(rotr(u32, value, 8 * rotateBy(T, address)));
 }
 
 fn rotateBy(comptime T: type, address: u32) u32 {
@@ -43,7 +43,7 @@ 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(T, rotr(u32, value, 8 * rotateBy(T, address)));
+    return @truncate(rotr(u32, value, 8 * rotateBy(T, address)));
 }
 
 /// Read without the GBA safety checks
@@ -51,30 +51,39 @@ 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"),
     };
 }
 
 pub fn write(_: *Self, comptime T: type, addr: u32, value: T) void {
-    @setCold(true);
+    @branchHint(.cold);
     log.debug("Tried to write {} 0x{X:} to 0x{X:0>8} ", .{ T, value, addr });
 }
 
 pub fn init(allocator: Allocator, maybe_path: ?[]const u8) !Self {
     if (maybe_path == null) return .{ .buf = null, .allocator = allocator };
-    const path = maybe_path.?;
+    const file_path = maybe_path.?;
 
     const buf = try allocator.alloc(u8, Self.size);
     errdefer allocator.free(buf);
 
-    const file = try std.fs.cwd().openFile(path, .{});
+    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 file_len = try file.readAll(buf);
-    if (file_len != Self.size) log.err("Expected BIOS to be {}B, was {}B", .{ Self.size, file_len });
+    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 });
+}
 
-    return Self{ .buf = buf, .allocator = allocator };
+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

@@ -1,7 +1,7 @@
 const std = @import("std");
 const config = @import("../../config.zig");
 
-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;
@@ -30,7 +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();
         }
     }
@@ -77,7 +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 {
@@ -94,7 +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();
         }
     }
@@ -139,7 +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
@@ -151,7 +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);
         }
     }
@@ -159,19 +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 }),
@@ -179,23 +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]),
     };
 }
 
@@ -213,25 +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

@@ -32,7 +32,7 @@ pub const Backup = struct {
     flash: Flash,
     eeprom: Eeprom,
 
-    const Kind = enum {
+    pub const Kind = enum {
         Eeprom,
         Sram,
         Flash,
@@ -77,7 +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) {
@@ -110,7 +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,
@@ -137,6 +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;
@@ -162,7 +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});
                 }
 
@@ -173,7 +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});
                 }
 

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

@@ -58,7 +58,9 @@ 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);
             }
         }
 
@@ -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

@@ -3,7 +3,7 @@ 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 = struct { DmaController(0), DmaController(1), DmaController(2), DmaController(3) };
 const log = std.log.scoped(.DmaTransfer);
@@ -11,15 +11,16 @@ const log = std.log.scoped(.DmaTransfer);
 const getHalf = util.getHalf;
 const setHalf = util.setHalf;
 const setQuart = util.setQuart;
+const handleInterrupt = @import("../cpu_util.zig").handleInterrupt;
 
-const rotr = @import("../../util.zig").rotr;
+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_addr = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     return switch (T) {
         u32 => switch (byte_addr) {
@@ -54,19 +55,19 @@ pub fn read(comptime T: type, dma: *const DmaTuple, addr: u32) ?T {
         u8 => switch (byte_addr) {
             0xB0...0xB7 => null, // DMA0SAD, DMA0DAD
             0xB8, 0xB9 => 0x00, // DMA0CNT_L
-            0xBA, 0xBB => @truncate(T, dma.*[0].dmacntH() >> getHalf(byte_addr)),
+            0xBA, 0xBB => @truncate(dma.*[0].dmacntH() >> getHalf(byte_addr)),
 
             0xBC...0xC3 => null, // DMA1SAD, DMA1DAD
             0xC4, 0xC5 => 0x00, // DMA1CNT_L
-            0xC6, 0xC7 => @truncate(T, dma.*[1].dmacntH() >> getHalf(byte_addr)),
+            0xC6, 0xC7 => @truncate(dma.*[1].dmacntH() >> getHalf(byte_addr)),
 
             0xC8...0xCF => null, // DMA2SAD, DMA2DAD
             0xD0, 0xD1 => 0x00, // DMA2CNT_L
-            0xD2, 0xD3 => @truncate(T, dma.*[2].dmacntH() >> getHalf(byte_addr)),
+            0xD2, 0xD3 => @truncate(dma.*[2].dmacntH() >> getHalf(byte_addr)),
 
             0xD4...0xDB => null, // DMA3SAD, DMA3DAD
             0xDC, 0xDD => 0x00, // DMA3CNT_L
-            0xDE, 0xDF => @truncate(T, dma.*[3].dmacntH() >> getHalf(byte_addr)),
+            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 }),
         },
         else => @compileError("DMA: Unsupported read width"),
@@ -74,7 +75,7 @@ pub fn read(comptime T: type, dma: *const DmaTuple, addr: u32) ?T {
 }
 
 pub fn write(comptime T: type, dma: *DmaTuple, addr: u32, value: T) void {
-    const byte_addr = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     switch (T) {
         u32 => switch (byte_addr) {
@@ -195,6 +196,10 @@ fn DmaController(comptime id: u2) type {
             };
         }
 
+        pub fn reset(self: *Self) void {
+            self.* = Self.init();
+        }
+
         pub fn setDmasad(self: *Self, addr: u32) void {
             self.sad = addr & sad_mask;
         }
@@ -204,7 +209,7 @@ 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 {
@@ -228,14 +233,16 @@ 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);
@@ -253,10 +260,10 @@ fn DmaController(comptime id: u2) type {
                     self.data_latch = value << 16 | value;
                 }
 
-                cpu.bus.write(u16, dad_addr, @truncate(u16, rotr(u32, self.data_latch, 8 * (dad_addr & 3))));
+                cpu.bus.write(u16, dad_addr, @as(u16, @truncate(rotr(u32, self.data_latch, 8 * (dad_addr & 3)))));
             }
 
-            switch (@truncate(u8, sad_addr >> 24)) {
+            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
@@ -279,17 +286,17 @@ 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.write(true),
+                        1 => bus_ptr.io.irq.dma1.write(true),
+                        2 => bus_ptr.io.irq.dma2.write(true),
+                        3 => bus_ptr.io.irq.dma3.write(true),
                     }
 
-                    cpu.handleInterrupt();
+                    handleInterrupt(cpu);
                 }
 
                 // If we're not repeating, Fire the IRQs and disable the DMA
-                if (!self.cnt.repeat.read()) self.cnt.enabled.unset();
+                if (!self.cnt.repeat.read()) self.cnt.enabled.write(false);
 
                 // We want to disable our internal enabled flag regardless of repeat
                 // because we only want to step A DMA that repeats during it's specific
@@ -314,7 +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;
             }
         }
 
@@ -331,7 +338,7 @@ fn DmaController(comptime id: u2) type {
 
             // FIXME: Safe to just assume whatever DAD is set to is the FIFO Address?
             // self.dad_latch = fifo_addr;
-            self.cnt.repeat.set();
+            self.cnt.repeat.write(true);
             self._word_count = 4;
             self.in_progress = true;
         }
@@ -339,10 +346,7 @@ fn DmaController(comptime id: u2) type {
 }
 
 pub fn onBlanking(bus: *Bus, comptime kind: DmaKind) void {
-    comptime var i: usize = 0;
-    inline while (i < 4) : (i += 1) {
-        bus.dma[i].poll(kind);
-    }
+    inline for (0..4) |i| bus.dma[i].poll(kind);
 }
 
 const Adjustment = enum(u2) {

src/core/bus/gpio.zig

@@ -1,10 +1,14 @@
 const std = @import("std");
-const Bit = @import("bitfield").Bit;
+const Bit = @import("bitjuggle").Boolean;
 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();
@@ -27,8 +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,
             };
@@ -90,7 +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 => {},
         }
 
@@ -142,16 +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,
@@ -164,7 +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,
@@ -203,7 +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;
         }
@@ -286,20 +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 {
@@ -315,14 +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);
@@ -332,14 +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,
@@ -358,7 +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});
@@ -384,7 +390,7 @@ pub const Clock = struct {
                     }
                 }
 
-                break :blk @truncate(u4, ret.raw);
+                break :blk @truncate(ret.raw);
             },
         };
     }
@@ -397,11 +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.write(true);
+        handleInterrupt(self.cpu);
     }
 
     fn processCommand(self: *Self, raw_command: u8) State {
@@ -421,7 +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) {
@@ -449,16 +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

@@ -5,8 +5,8 @@ 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 Bit = @import("bitjuggle").Boolean;
+const Bitfield = @import("bitjuggle").Bitfield;
 const Bus = @import("../Bus.zig");
 
 const getHalf = util.getHalf;
@@ -38,9 +38,13 @@ pub const Io = struct {
         };
     }
 
+    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));
     }
 };
 
@@ -71,8 +75,8 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
             // Interrupts
             0x0400_0200 => @as(u32, bus.io.irq.raw) << 16 | bus.io.ie.raw,
             0x0400_0204 => bus.io.waitcnt.raw,
-            0x0400_0208 => @boolToInt(bus.io.ime),
-            0x0400_0300 => @enumToInt(bus.io.postflg),
+            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) {
@@ -92,7 +96,7 @@ 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.load(.Monotonic).raw,
+            0x0400_0130 => bus.io.keyinput.load(.monotonic),
 
             // Serial Communication 2
             0x0400_0134 => util.io.read.todo(log, "Read {} from RCNT", .{T}),
@@ -105,9 +109,9 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
             0x0400_0202 => bus.io.irq.raw,
             0x0400_0204 => bus.io.waitcnt.raw,
             0x0400_0206 => 0x0000,
-            0x0400_0208 => @boolToInt(bus.io.ime),
+            0x0400_0208 => @intFromBool(bus.io.ime),
             0x0400_020A => 0x0000,
-            0x0400_0300 => @enumToInt(bus.io.postflg),
+            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 }),
         },
@@ -137,13 +141,13 @@ pub fn read(bus: *const Bus, comptime T: type, address: u32) ?T {
             0x0400_015A, 0x0400_015B => 0x00,
 
             // Interrupts
-            0x0400_0200, 0x0400_0201 => @truncate(T, bus.io.ie.raw >> getHalf(@truncate(u8, address))),
-            0x0400_0202, 0x0400_0203 => @truncate(T, bus.io.irq.raw >> getHalf(@truncate(u8, address))),
-            0x0400_0204, 0x0400_0205 => @truncate(T, bus.io.waitcnt.raw >> getHalf(@truncate(u8, address))),
+            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(T, @as(u16, @boolToInt(bus.io.ime)) >> getHalf(@truncate(u8, address))),
+            0x0400_0208, 0x0400_0209 => @truncate(@as(u16, @intFromBool(bus.io.ime)) >> getHalf(@truncate(address))),
             0x0400_020A, 0x0400_020B => 0x00,
-            0x0400_0300 => @enumToInt(bus.io.postflg),
+            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 }),
@@ -192,10 +196,10 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
 
             // Interrupts
             0x0400_0200 => bus.io.setIrqs(value),
-            0x0400_0204 => bus.io.waitcnt.set(@truncate(u16, value)),
+            0x0400_0204 => bus.io.waitcnt.set(@truncate(value)),
             0x0400_0208 => bus.io.ime = value & 1 == 1,
             0x0400_0300 => {
-                bus.io.postflg = @intToEnum(PostFlag, value & 1);
+                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 }),
@@ -242,7 +246,7 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
             0x0400_0208 => bus.io.ime = value & 1 == 1,
             0x0400_020A => {},
             0x0400_0300 => {
-                bus.io.postflg = @intToEnum(PostFlag, value & 1);
+                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 }),
@@ -269,16 +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(u8, address), value),
+            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, @truncate(u16, bus.io.waitcnt.raw), @truncate(u8, address), value)),
+            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_0209 => {},
             0x0400_020A, 0x0400_020B => {},
 
-            0x0400_0300 => bus.io.postflg = @intToEnum(PostFlag, value & 1),
+            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 }),
@@ -346,10 +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),
@@ -362,7 +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),
@@ -378,7 +382,7 @@ const KeyInput = extern union {
 
 const AtomicKeyInput = struct {
     const Self = @This();
-    const Ordering = std.atomic.Ordering;
+    const AtomicOrder = std.builtin.AtomicOrder;
 
     inner: KeyInput,
 
@@ -386,18 +390,19 @@ const AtomicKeyInput = struct {
         return .{ .inner = value };
     }
 
-    pub inline fn load(self: *const Self, comptime ordering: Ordering) KeyInput {
-        return .{ .raw = switch (ordering) {
-            .AcqRel, .Release => @compileError("not supported for atomic loads"),
+    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 store(self: *Self, value: u16, comptime ordering: Ordering) void {
-        switch (ordering) {
-            .AcqRel, .Acquire => @compileError("not supported for atomic stores"),
-            else => @atomicStore(u16, &self.inner.raw, value, 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);
     }
 };
 

src/core/bus/timer.zig

@@ -3,7 +3,10 @@ const util = @import("../../util.zig");
 
 const TimerControl = @import("io.zig").TimerControl;
 const Scheduler = @import("../scheduler.zig").Scheduler;
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("../Bus.zig");
+
+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);
@@ -16,7 +19,7 @@ pub fn create(sched: *Scheduler) TimerTuple {
 }
 
 pub fn read(comptime T: type, tim: *const TimerTuple, addr: u32) ?T {
-    const nybble_addr = @truncate(u4, addr);
+    const nybble_addr: u4 = @truncate(addr);
 
     return switch (T) {
         u32 => switch (nybble_addr) {
@@ -41,24 +44,24 @@ pub fn read(comptime T: type, tim: *const TimerTuple, addr: u32) ?T {
             else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
         },
         u8 => switch (nybble_addr) {
-            0x0, 0x1 => @truncate(T, tim.*[0].timcntL() >> getHalf(nybble_addr)),
-            0x2, 0x3 => @truncate(T, tim.*[0].cnt.raw >> getHalf(nybble_addr)),
+            0x0, 0x1 => @truncate(tim.*[0].timcntL() >> getHalf(nybble_addr)),
+            0x2, 0x3 => @truncate(tim.*[0].cnt.raw >> getHalf(nybble_addr)),
 
-            0x4, 0x5 => @truncate(T, tim.*[1].timcntL() >> getHalf(nybble_addr)),
-            0x6, 0x7 => @truncate(T, tim.*[1].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(T, tim.*[2].timcntL() >> getHalf(nybble_addr)),
-            0xA, 0xB => @truncate(T, tim.*[2].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(T, tim.*[3].timcntL() >> getHalf(nybble_addr)),
-            0xE, 0xF => @truncate(T, tim.*[3].cnt.raw >> getHalf(nybble_addr)),
+            0xC, 0xD => @truncate(tim.*[3].timcntL() >> getHalf(nybble_addr)),
+            0xE, 0xF => @truncate(tim.*[3].cnt.raw >> getHalf(nybble_addr)),
         },
         else => @compileError("TIM: Unsupported read width"),
     };
 }
 
 pub fn write(comptime T: type, tim: *TimerTuple, addr: u32, value: T) void {
-    const nybble_addr = @truncate(u4, addr);
+    const nybble_addr: u4 = @truncate(addr);
 
     return switch (T) {
         u32 => switch (nybble_addr) {
@@ -128,11 +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
@@ -142,8 +151,8 @@ 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
@@ -158,7 +167,7 @@ fn Timer(comptime id: u2) type {
                     self.sched.removeScheduledEvent(.{ .TimerOverflow = id });
 
                     // Counter should hold the value it stopped at meaning we have to calculate it now
-                    self._counter +%= @truncate(u16, (self.sched.now() - self._start_timestamp) / self.frequency());
+                    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
@@ -185,22 +194,24 @@ fn Timer(comptime id: u2) type {
 
         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) {
-                    0 => io.irq.tim0.set(),
-                    1 => io.irq.tim1.set(),
-                    2 => io.irq.tim2.set(),
-                    3 => io.irq.tim3.set(),
+                    0 => io.irq.tim0.write(true),
+                    1 => io.irq.tim1.write(true),
+                    2 => io.irq.tim2.write(true),
+                    3 => io.irq.tim3.write(true),
                 }
 
-                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
@@ -208,9 +219,9 @@ fn Timer(comptime id: u2) type {
                 inline 0, 1, 2 => |idx| {
                     const next = idx + 1;
 
-                    if (cpu.bus.tim[next].cnt.cascade.read()) {
-                        cpu.bus.tim[next]._counter +%= 1;
-                        if (cpu.bus.tim[next]._counter == 0) cpu.bus.tim[next].onTimerExpire(cpu, late);
+                    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);
                     }
                 },
                 3 => {}, // THere is no timer for TIM3 to cascade to

src/core/cpu.zig

@@ -1,678 +0,0 @@
-const std = @import("std");
-
-const Bus = @import("Bus.zig");
-const Bit = @import("bitfield").Bit;
-const Bitfield = @import("bitfield").Bitfield;
-const Scheduler = @import("scheduler.zig").Scheduler;
-const Logger = @import("../util.zig").Logger;
-
-const File = std.fs.File;
-const log = std.log.scoped(.Arm7Tdmi);
-
-// 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;
-        };
-    }
-};
-
-pub const Arm7tdmi = struct {
-    const Self = @This();
-
-    r: [16]u32,
-    pipe: Pipeline,
-    sched: *Scheduler,
-    bus: *Bus,
-    cpsr: PSR,
-    spsr: PSR,
-
-    bank: Bank,
-
-    logger: ?Logger,
-
-    /// Bank of Registers from other CPU Modes
-    const Bank = struct {
-        /// Storage for r13_<mode>, r14_<mode>
-        /// e.g. [r13, r14, r13_svc, r14_svc]
-        r: [2 * 6]u32,
-
-        /// 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]...]
-        fiq: [2 * 5]u32,
-
-        spsr: [5]PSR,
-
-        const Kind = enum(u1) {
-            R13 = 0,
-            R14,
-        };
-
-        pub fn create() Bank {
-            return .{
-                .r = [_]u32{0x00} ** 12,
-                .fiq = [_]u32{0x00} ** 10,
-                .spsr = [_]PSR{.{ .raw = 0x0000_0000 }} ** 5,
-            };
-        }
-
-        inline fn regIdx(mode: Mode, kind: Kind) 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 spsrIdx(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 fiqIdx(i: usize, mode: Mode) usize {
-            return (i * 2) + if (mode == .Fiq) @as(usize, 1) else 0;
-        }
-    };
-
-    pub fn init(sched: *Scheduler, bus: *Bus, log_file: ?std.fs.File) Self {
-        return Self{
-            .r = [_]u32{0x00} ** 16,
-            .pipe = Pipeline.init(),
-            .sched = sched,
-            .bus = bus,
-            .cpsr = .{ .raw = 0x0000_001F },
-            .spsr = .{ .raw = 0x0000_0000 },
-            .bank = Bank.create(),
-            .logger = if (log_file) |file| Logger.init(file) else null,
-        };
-    }
-
-    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.bank.fiq[Bank.fiqIdx(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(Bank.Kind, idx - 13) catch unreachable;
-                    self.bank.r[Bank.regIdx(.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.bank.fiq[Bank.fiqIdx(idx - 8, .User)] else self.r[idx],
-            13, 14 => switch (current) {
-                .User, .System => self.r[idx],
-                else => blk: {
-                    const kind = std.meta.intToEnum(Bank.Kind, idx - 13) catch unreachable;
-                    break :blk self.bank.r[Bank.regIdx(.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.bank.fiq[Bank.fiqIdx(i, now)] = self.r[8 + i];
-        }
-
-        // Bank r13, r14, SPSR
-        switch (now) {
-            .User, .System => {
-                self.bank.r[Bank.regIdx(now, .R13)] = self.r[13];
-                self.bank.r[Bank.regIdx(now, .R14)] = self.r[14];
-            },
-            else => {
-                self.bank.r[Bank.regIdx(now, .R13)] = self.r[13];
-                self.bank.r[Bank.regIdx(now, .R14)] = self.r[14];
-                self.bank.spsr[Bank.spsrIdx(now)] = self.spsr;
-            },
-        }
-
-        // Grab R8 -> R12
-        i = 0;
-        while (i < 5) : (i += 1) {
-            self.r[8 + i] = self.bank.fiq[Bank.fiqIdx(i, next)];
-        }
-
-        // Grab r13, r14, SPSR
-        switch (next) {
-            .User, .System => {
-                self.r[13] = self.bank.r[Bank.regIdx(next, .R13)];
-                self.r[14] = self.bank.r[Bank.regIdx(next, .R14)];
-            },
-            else => {
-                self.r[13] = self.bank.r[Bank.regIdx(next, .R13)];
-                self.r[14] = self.bank.r[Bank.regIdx(next, .R14)];
-                self.spsr = self.bank.spsr[Bank.spsrIdx(next)];
-            },
-        }
-
-        self.cpsr.mode.write(@enumToInt(next));
-    }
-
-    /// 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(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.bank.r[Bank.regIdx(.Irq, .R13)] = 0x0300_7FA0;
-        self.bank.r[Bank.regIdx(.Supervisor, .R13)] = 0x0300_7FE0;
-
-        // self.cpsr.raw = 0x6000001F;
-        self.cpsr.raw = 0x0000_001F;
-
-        self.bus.bios.addr_latch = 0x0000_00DC + 8;
-    }
-
-    pub fn step(self: *Self) void {
-        defer {
-            if (!self.pipe.flushed) self.r[15] += if (self.cpsr.t.read()) 2 else @as(u32, 4);
-            self.pipe.flushed = false;
-        }
-
-        if (self.cpsr.t.read()) {
-            const opcode = @truncate(u16, self.pipe.step(self, u16) orelse return);
-            if (self.logger) |*trace| trace.mgbaLog(self, opcode);
-
-            thumb.lut[thumb.idx(opcode)](self, self.bus, opcode);
-        } else {
-            const opcode = self.pipe.step(self, u32) orelse return;
-            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 {
-        comptime var i: usize = 0;
-        inline while (i < 4) : (i += 1) {
-            if (self.bus.dma[i].in_progress) {
-                self.bus.dma[i].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;
-
-        // Return if IME is disabled, CPSR I is set or there is nothing to handle
-        if (!self.bus.io.ime or self.cpsr.i.read() or should_handle == 0) return;
-
-        // If Pipeline isn't full, we have a bug
-        std.debug.assert(self.pipe.isFull());
-
-        // log.debug("Handling Interrupt!", .{});
-        self.bus.io.haltcnt = .Execute;
-
-        // FIXME: This seems weird, but retAddr.gba suggests I need to make these changes
-        const ret_addr = self.r[15] - if (self.cpsr.t.read()) 0 else @as(u32, 4);
-        const new_spsr = self.cpsr.raw;
-
-        self.changeMode(.Irq);
-        self.cpsr.t.write(false);
-        self.cpsr.i.write(true);
-
-        self.r[14] = ret_addr;
-        self.spsr.raw = new_spsr;
-        self.r[15] = 0x0000_0018;
-        self.pipe.reload(self);
-    }
-
-    inline fn fetch(self: *Self, comptime T: type, address: u32) T {
-        comptime std.debug.assert(T == u32 or T == u16); // Opcode may be 32-bit (ARM) or 16-bit (THUMB)
-
-        // Bus.read will advance the scheduler. There are different timings for CPU fetches,
-        // so we want to undo what Bus.read will apply. We can do this by caching the current tick
-        // This is very dumb.
-        //
-        // FIXME: Please rework this
-        const tick_cache = self.sched.tick;
-        defer self.sched.tick = tick_cache + Bus.fetch_timings[@boolToInt(T == u32)][@truncate(u4, address >> 24)];
-
-        return self.bus.read(T, address);
-    }
-
-    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});
-        self.cpsr.toString();
-
-        std.debug.print("spsr: 0x{X:0>8} ", .{self.spsr.raw});
-        self.spsr.toString();
-
-        std.debug.print("pipeline: {??X:0>8}\n", .{self.pipe.stage});
-
-        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);
-    }
-};
-
-const condition_lut = [_]u16{
-    0xF0F0, // EQ - Equal
-    0x0F0F, // NE - Not Equal
-    0xCCCC, // CS - Unsigned higher or same
-    0x3333, // CC - Unsigned lower
-    0xFF00, // MI - Negative
-    0x00FF, // PL - Positive or Zero
-    0xAAAA, // VS - Overflow
-    0x5555, // VC - No Overflow
-    0x0C0C, // HI - unsigned hierh
-    0xF3F3, // LS - unsigned lower or same
-    0xAA55, // GE - greater or equal
-    0x55AA, // LT - less than
-    0x0A05, // GT - greater than
-    0xF5FA, // LE - less than or equal
-    0xFFFF, // AL - always
-    0x0000, // NV - never
-};
-
-pub inline fn checkCond(cpsr: PSR, cond: u4) bool {
-    const flags = @truncate(u4, cpsr.raw >> 28);
-
-    return condition_lut[cond] & (@as(u16, 1) << flags) != 0;
-}
-
-const Pipeline = struct {
-    const Self = @This();
-    stage: [2]?u32,
-    flushed: bool,
-
-    fn init() Self {
-        return .{
-            .stage = [_]?u32{null} ** 2,
-            .flushed = false,
-        };
-    }
-
-    pub fn isFull(self: *const Self) bool {
-        return self.stage[0] != null and self.stage[1] != null;
-    }
-
-    pub fn step(self: *Self, cpu: *Arm7tdmi, comptime T: type) ?u32 {
-        comptime std.debug.assert(T == u32 or T == u16);
-
-        const opcode = self.stage[0];
-        self.stage[0] = self.stage[1];
-        self.stage[1] = cpu.fetch(T, cpu.r[15]);
-
-        return opcode;
-    }
-
-    pub fn reload(self: *Self, cpu: *Arm7tdmi) void {
-        if (cpu.cpsr.t.read()) {
-            self.stage[0] = cpu.fetch(u16, cpu.r[15]);
-            self.stage[1] = cpu.fetch(u16, cpu.r[15] + 2);
-            cpu.r[15] += 4;
-        } else {
-            self.stage[0] = cpu.fetch(u32, cpu.r[15]);
-            self.stage[1] = cpu.fetch(u32, cpu.r[15] + 4);
-            cpu.r[15] += 8;
-        }
-
-        self.flushed = true;
-    }
-};
-
-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,
-
-    fn toString(self: PSR) void {
-        std.debug.print("[", .{});
-
-        if (self.n.read()) std.debug.print("N", .{}) else std.debug.print("-", .{});
-        if (self.z.read()) std.debug.print("Z", .{}) else std.debug.print("-", .{});
-        if (self.c.read()) std.debug.print("C", .{}) else std.debug.print("-", .{});
-        if (self.v.read()) std.debug.print("V", .{}) else std.debug.print("-", .{});
-        if (self.i.read()) std.debug.print("I", .{}) else std.debug.print("-", .{});
-        if (self.f.read()) std.debug.print("F", .{}) else std.debug.print("-", .{});
-        if (self.t.read()) std.debug.print("T", .{}) else std.debug.print("-", .{});
-        std.debug.print("|", .{});
-        if (getMode(self.mode.read())) |m| std.debug.print("{s}", .{m.toString()}) else std.debug.print("---", .{});
-
-        std.debug.print("]\n", .{});
-    }
-};
-
-const Mode = enum(u5) {
-    User = 0b10000,
-    Fiq = 0b10001,
-    Irq = 0b10010,
-    Supervisor = 0b10011,
-    Abort = 0b10111,
-    Undefined = 0b11011,
-    System = 0b11111,
-
-    fn toString(self: Mode) []const u8 {
-        return switch (self) {
-            .User => "usr",
-            .Fiq => "fiq",
-            .Irq => "irq",
-            .Supervisor => "svc",
-            .Abort => "abt",
-            .Undefined => "und",
-            .System => "sys",
-        };
-    }
-};
-
-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

@@ -1,111 +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);
-                    cpu.pipe.reload(cpu);
-                } else {
-                    bus.write(u32, und_addr, cpu.r[15] + 4);
-                }
-
-                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] = value;
-                    if (i == 0xF) {
-                        cpu.r[i] &= ~@as(u32, 3); // Align r15
-                        cpu.pipe.reload(cpu);
-
-                        if (S) 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) 4 else @as(u32, 0)); // PC is already 8 ahead to make 12
-                } else {
-                    bus.write(u32, address, cpu.r[i] + if (i == 0xF) 4 else @as(u32, 0));
-                }
-            }
-        }
-    }.inner;
-}

src/core/cpu/arm/branch.zig

@@ -1,26 +0,0 @@
-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] - 4;
-
-            cpu.r[15] +%= sext(u32, u24, opcode) << 2;
-            cpu.pipe.reload(cpu);
-        }
-    }.inner;
-}
-
-pub fn branchAndExchange(cpu: *Arm7tdmi, _: *Bus, opcode: u32) void {
-    const rn = opcode & 0xF;
-
-    const thumb = cpu.r[rn] & 1 == 1;
-    cpu.r[15] = cpu.r[rn] & if (thumb) ~@as(u32, 1) else ~@as(u32, 3);
-
-    cpu.cpsr.t.write(thumb);
-    cpu.pipe.reload(cpu);
-}

src/core/cpu/arm/data_processing.zig

@@ -1,185 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").arm.InstrFn;
-
-const exec = @import("../barrel_shifter.zig").exec;
-const ror = @import("../barrel_shifter.zig").ror;
-
-pub fn dataProcessing(comptime I: bool, comptime S: bool, comptime kind: 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
-            // TODO: Why these conditions?
-            if (!I and opcode >> 4 & 1 == 1) cpu.r[15] += 4;
-            const op1 = cpu.r[rn];
-
-            const amount = @truncate(u8, (opcode >> 8 & 0xF) << 1);
-            const op2 = if (I) ror(S, &cpu.cpsr, opcode & 0xFF, amount) else exec(S, cpu, opcode);
-
-            // Undo special condition from above
-            if (!I and opcode >> 4 & 1 == 1) cpu.r[15] -= 4;
-
-            var result: u32 = undefined;
-            var overflow: u1 = undefined;
-
-            // Perform Data Processing Logic
-            switch (kind) {
-                0x0 => result = op1 & op2, // AND
-                0x1 => result = op1 ^ op2, // EOR
-                0x2 => result = op1 -% op2, // SUB
-                0x3 => result = op2 -% op1, // RSB
-                0x4 => result = add(&overflow, op1, op2), // ADD
-                0x5 => result = adc(&overflow, op1, op2, old_carry), // ADC
-                0x6 => result = sbc(op1, op2, old_carry), // SBC
-                0x7 => result = sbc(op2, op1, old_carry), // RSC
-                0x8 => {
-                    // TST
-                    if (rd == 0xF)
-                        return undefinedTestBehaviour(cpu);
-
-                    result = op1 & op2;
-                },
-                0x9 => {
-                    // TEQ
-                    if (rd == 0xF)
-                        return undefinedTestBehaviour(cpu);
-
-                    result = op1 ^ op2;
-                },
-                0xA => {
-                    // CMP
-                    if (rd == 0xF)
-                        return undefinedTestBehaviour(cpu);
-
-                    result = op1 -% op2;
-                },
-                0xB => {
-                    // CMN
-                    if (rd == 0xF)
-                        return undefinedTestBehaviour(cpu);
-
-                    const tmp = @addWithOverflow(op1, op2);
-                    result = tmp[0];
-                    overflow = tmp[1];
-                },
-                0xC => result = op1 | op2, // ORR
-                0xD => result = op2, // MOV
-                0xE => result = op1 & ~op2, // BIC
-                0xF => result = ~op2, // MVN
-            }
-
-            // Write to Destination Register
-            switch (kind) {
-                0x8, 0x9, 0xA, 0xB => {}, // Test Operations
-                else => {
-                    cpu.r[rd] = result;
-                    if (rd == 0xF) {
-                        if (S) cpu.setCpsr(cpu.spsr.raw);
-                        cpu.pipe.reload(cpu);
-                    }
-                },
-            }
-
-            // Write Flags
-            switch (kind) {
-                0x0, 0x1, 0xC, 0xD, 0xE, 0xF => if (S and rd != 0xF) {
-                    // Logic Operation Flags
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    // C set by Barrel Shifter, V is unaffected
-
-                },
-                0x2, 0x3 => if (S and rd != 0xF) {
-                    // SUB, RSB Flags
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-
-                    if (kind == 0x2) {
-                        // SUB specific
-                        cpu.cpsr.c.write(op2 <= op1);
-                        cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                    } else {
-                        // RSB Specific
-                        cpu.cpsr.c.write(op1 <= op2);
-                        cpu.cpsr.v.write(((op2 ^ result) & (~op1 ^ result)) >> 31 & 1 == 1);
-                    }
-                },
-                0x4, 0x5 => if (S and rd != 0xF) {
-                    // ADD, ADC Flags
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    cpu.cpsr.c.write(overflow == 0b1);
-                    cpu.cpsr.v.write(((op1 ^ result) & (op2 ^ result)) >> 31 & 1 == 1);
-                },
-                0x6, 0x7 => if (S and rd != 0xF) {
-                    // SBC, RSC Flags
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-
-                    if (kind == 0x6) {
-                        // SBC specific
-                        const subtrahend = @as(u64, op2) -% old_carry +% 1;
-                        cpu.cpsr.c.write(subtrahend <= op1);
-                        cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                    } else {
-                        // RSC Specific
-                        const subtrahend = @as(u64, op1) -% old_carry +% 1;
-                        cpu.cpsr.c.write(subtrahend <= op2);
-                        cpu.cpsr.v.write(((op2 ^ result) & (~op1 ^ result)) >> 31 & 1 == 1);
-                    }
-                },
-                0x8, 0x9, 0xA, 0xB => {
-                    // Test Operation Flags
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-
-                    if (kind == 0xA) {
-                        // CMP specific
-                        cpu.cpsr.c.write(op2 <= op1);
-                        cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                    } else if (kind == 0xB) {
-                        // CMN specific
-                        cpu.cpsr.c.write(overflow == 0b1);
-                        cpu.cpsr.v.write(((op1 ^ result) & (op2 ^ result)) >> 31 & 1 == 1);
-                    } else {
-                        // TST, TEQ specific
-                        // Barrel Shifter should always calc CPSR C in TST
-                        if (!S) _ = exec(true, cpu, opcode);
-                    }
-                },
-            }
-        }
-    }.inner;
-}
-
-pub fn sbc(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 ret = @truncate(u32, left -% subtrahend);
-
-    return ret;
-}
-
-pub fn add(overflow: *u1, left: u32, right: u32) u32 {
-    const ret = @addWithOverflow(left, right);
-    overflow.* = ret[1];
-
-    return ret[0];
-}
-
-pub fn adc(overflow: *u1, left: u32, right: u32, old_carry: u1) u32 {
-    const tmp = @addWithOverflow(left, right);
-    const ret = @addWithOverflow(tmp[0], old_carry);
-    overflow.* = tmp[1] | ret[1];
-
-    return ret[0];
-}
-
-fn undefinedTestBehaviour(cpu: *Arm7tdmi) void {
-    @setCold(true);
-    cpu.setCpsr(cpu.spsr.raw);
-}

src/core/cpu/arm/half_signed_data_transfer.zig

@@ -1,53 +0,0 @@
-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;
-
-            const base = cpu.r[rn] + if (!L and rn == 0xF) 4 else @as(u32, 0);
-            const offset = if (I) imm_offset_high << 4 | rm else 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
-                        const value = bus.read(u16, address);
-                        result = if (address & 1 == 1) sext(u32, u8, @truncate(u8, value >> 8)) else sext(u32, u16, value);
-                    },
-                    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

@@ -1,29 +0,0 @@
-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

@@ -1,55 +0,0 @@
-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;
-
-            const base = cpu.r[rn];
-            const offset = if (I) shifter.immediate(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 = cpu.r[rd] + if (rd == 0xF) 4 else @as(u32, 0); // PC is 12 ahead
-                    bus.write(u8, address, @truncate(u8, value));
-                } else {
-                    // STR
-                    const value = cpu.r[rd] + if (rd == 0xF) 4 else @as(u32, 0);
-                    bus.write(u32, address, value);
-                }
-            }
-
-            address = modified_base;
-            if (W and P or !P) {
-                cpu.r[rn] = address;
-                if (rn == 0xF) cpu.pipe.reload(cpu);
-            }
-
-            if (L) {
-                // This emulates the LDR rd == rn behaviour
-                cpu.r[rd] = result;
-                if (rd == 0xF) cpu.pipe.reload(cpu);
-            }
-        }
-    }.inner;
-}

src/core/cpu/arm/software_interrupt.zig

@@ -1,23 +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 ret_addr = cpu.r[15] - 4;
-            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] = ret_addr; // Resume Execution
-            cpu.spsr.raw = cpsr; // Previous mode CPSR
-            cpu.r[15] = 0x0000_0008;
-            cpu.pipe.reload(cpu);
-        }
-    }.inner;
-}

src/core/cpu/barrel_shifter.zig

@@ -1,147 +0,0 @@
-const Arm7tdmi = @import("../cpu.zig").Arm7tdmi;
-const CPSR = @import("../cpu.zig").PSR;
-
-const rotr = @import("../../util.zig").rotr;
-
-pub fn exec(comptime S: bool, cpu: *Arm7tdmi, opcode: u32) u32 {
-    var result: u32 = undefined;
-    if (opcode >> 4 & 1 == 1) {
-        result = register(S, cpu, opcode);
-    } else {
-        result = immediate(S, cpu, opcode);
-    }
-
-    return result;
-}
-
-fn register(comptime S: bool, cpu: *Arm7tdmi, opcode: u32) u32 {
-    const rs_idx = opcode >> 8 & 0xF;
-    const rm = cpu.r[opcode & 0xF];
-    const rs = @truncate(u8, cpu.r[rs_idx]);
-
-    return switch (@truncate(u2, opcode >> 5)) {
-        0b00 => lsl(S, &cpu.cpsr, rm, rs),
-        0b01 => lsr(S, &cpu.cpsr, rm, rs),
-        0b10 => asr(S, &cpu.cpsr, rm, rs),
-        0b11 => ror(S, &cpu.cpsr, rm, rs),
-    };
-}
-
-pub fn immediate(comptime S: bool, cpu: *Arm7tdmi, opcode: u32) u32 {
-    const amount = @truncate(u8, opcode >> 7 & 0x1F);
-    const rm = cpu.r[opcode & 0xF];
-
-    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 = lsl(S, &cpu.cpsr, rm, amount),
-            0b01 => result = lsr(S, &cpu.cpsr, rm, amount),
-            0b10 => result = asr(S, &cpu.cpsr, rm, amount),
-            0b11 => result = ror(S, &cpu.cpsr, rm, amount),
-        }
-    }
-
-    return result;
-}
-
-pub fn lsl(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 lsr(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 asr(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 ror(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

@@ -1,108 +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 lsl = @import("../barrel_shifter.zig").lsl;
-const lsr = @import("../barrel_shifter.zig").lsr;
-const asr = @import("../barrel_shifter.zig").asr;
-const ror = @import("../barrel_shifter.zig").ror;
-
-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());
-
-            const op1 = cpu.r[rd];
-            const op2 = cpu.r[rs];
-
-            var result: u32 = undefined;
-            var overflow: u1 = undefined;
-
-            switch (op) {
-                0x0 => result = op1 & op2, // AND
-                0x1 => result = op1 ^ op2, // EOR
-                0x2 => result = lsl(true, &cpu.cpsr, op1, @truncate(u8, op2)), // LSL
-                0x3 => result = lsr(true, &cpu.cpsr, op1, @truncate(u8, op2)), // LSR
-                0x4 => result = asr(true, &cpu.cpsr, op1, @truncate(u8, op2)), // ASR
-                0x5 => result = adc(&overflow, op1, op2, carry), // ADC
-                0x6 => result = sbc(op1, op2, carry), // SBC
-                0x7 => result = ror(true, &cpu.cpsr, op1, @truncate(u8, op2)), // ROR
-                0x8 => result = op1 & op2, // TST
-                0x9 => result = 0 -% op2, // NEG
-                0xA => result = op1 -% op2, // CMP
-                0xB => {
-                    // CMN
-                    const tmp = @addWithOverflow(op1, op2);
-                    result = tmp[0];
-                    overflow = tmp[1];
-                },
-                0xC => result = op1 | op2, // ORR
-                0xD => result = @truncate(u32, @as(u64, op2) * @as(u64, op1)),
-                0xE => result = op1 & ~op2,
-                0xF => result = ~op2,
-            }
-
-            // Write to Destination Register
-            switch (op) {
-                0x8, 0xA, 0xB => {},
-                else => cpu.r[rd] = result,
-            }
-
-            // Write Flags
-            switch (op) {
-                0x0, 0x1, 0x2, 0x3, 0x4, 0x7, 0xC, 0xE, 0xF => {
-                    // Logic Operations
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    // C set by Barrel Shifter, V is unaffected
-                },
-                0x8, 0xA => {
-                    // Test Flags
-                    // CMN (0xB) is handled with ADC
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-
-                    if (op == 0xA) {
-                        // CMP specific
-                        cpu.cpsr.c.write(op2 <= op1);
-                        cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                    }
-                },
-                0x5, 0xB => {
-                    // ADC, CMN
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    cpu.cpsr.c.write(overflow == 0b1);
-                    cpu.cpsr.v.write(((op1 ^ result) & (op2 ^ result)) >> 31 & 1 == 1);
-                },
-                0x6 => {
-                    // SBC
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-
-                    const subtrahend = @as(u64, op2) -% carry +% 1;
-                    cpu.cpsr.c.write(subtrahend <= op1);
-                    cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                },
-                0x9 => {
-                    // NEG
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    cpu.cpsr.c.write(op2 <= 0);
-                    cpu.cpsr.v.write(((0 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                },
-                0xD => {
-                    // Multiplication
-                    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
-                },
-            }
-        }
-    }.inner;
-}

src/core/cpu/thumb/block_data_transfer.zig

@@ -1,101 +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 & ~@as(u32, 1);
-                    cpu.pipe.reload(cpu);
-                } 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);
-                    cpu.pipe.reload(cpu);
-                } else {
-                    bus.write(u32, address, cpu.r[15] + 2);
-                }
-
-                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
-            if (cond == 0xE or cond == 0xF)
-                cpu.panic("[CPU/THUMB.16] Undefined conditional branch with condition {}", .{cond});
-
-            if (!checkCond(cpu.cpsr, cond)) return;
-
-            cpu.r[15] +%= sext(u32, u8, opcode & 0xFF) << 1;
-            cpu.pipe.reload(cpu);
-        }
-    }.inner;
-}
-
-pub fn fmt18() InstrFn {
-    return struct {
-        // B but conditional
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            cpu.r[15] +%= sext(u32, u11, opcode & 0x7FF) << 1;
-            cpu.pipe.reload(cpu);
-        }
-    }.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 next_opcode = cpu.r[15] - 2;
-
-                cpu.r[15] = cpu.r[14] +% (offset << 1);
-                cpu.r[14] = next_opcode | 1;
-
-                cpu.pipe.reload(cpu);
-            } else {
-                // Instruction 1
-                const lr_offset = sext(u32, u11, offset) << 12;
-                cpu.r[14] = (cpu.r[15] +% lr_offset) & ~@as(u32, 1);
-            }
-        }
-    }.inner;
-}

src/core/cpu/thumb/data_processing.zig

@@ -1,199 +0,0 @@
-const Bus = @import("../../Bus.zig");
-const Arm7tdmi = @import("../../cpu.zig").Arm7tdmi;
-const InstrFn = @import("../../cpu.zig").thumb.InstrFn;
-
-const add = @import("../arm/data_processing.zig").add;
-
-const lsl = @import("../barrel_shifter.zig").lsl;
-const lsr = @import("../barrel_shifter.zig").lsr;
-const asr = @import("../barrel_shifter.zig").asr;
-
-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 lsl(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 lsr(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 asr(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;
-
-            // Write Flags
-            cpu.cpsr.n.write(result >> 31 & 1 == 1);
-            cpu.cpsr.z.write(result == 0);
-        }
-    }.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 rs = @as(u4, h2) << 3 | (opcode >> 3 & 0x7);
-            const rd = @as(u4, h1) << 3 | (opcode & 0x7);
-
-            const op1 = cpu.r[rd];
-            const op2 = cpu.r[rs];
-
-            var result: u32 = undefined;
-            var overflow: u1 = undefined;
-            switch (op) {
-                0b00 => result = add(&overflow, op1, op2), // ADD
-                0b01 => result = op1 -% op2, // CMP
-                0b10 => result = op2, // MOV
-                0b11 => {},
-            }
-
-            // Write to Destination Register
-            switch (op) {
-                0b01 => {}, // Test Instruction
-                0b11 => {
-                    // BX
-                    const is_thumb = op2 & 1 == 1;
-                    cpu.r[15] = op2 & ~@as(u32, 1);
-
-                    cpu.cpsr.t.write(is_thumb);
-                    cpu.pipe.reload(cpu);
-                },
-                else => {
-                    cpu.r[rd] = result;
-                    if (rd == 0xF) {
-                        cpu.r[15] &= ~@as(u32, 1);
-                        cpu.pipe.reload(cpu);
-                    }
-                },
-            }
-
-            // Write Flags
-            switch (op) {
-                0b01 => {
-                    // CMP
-                    cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                    cpu.cpsr.z.write(result == 0);
-                    cpu.cpsr.c.write(op2 <= op1);
-                    cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                },
-                0b00, 0b10, 0b11 => {}, // MOV and Branch Instruction
-            }
-        }
-    }.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);
-            const op1 = cpu.r[rs];
-            const op2: u32 = if (I) rn else cpu.r[rn];
-
-            if (is_sub) {
-                // SUB
-                const result = op1 -% op2;
-                cpu.r[rd] = result;
-
-                cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                cpu.cpsr.z.write(result == 0);
-                cpu.cpsr.c.write(op2 <= op1);
-                cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-            } else {
-                // ADD
-                var overflow: u1 = undefined;
-                const result = add(&overflow, op1, op2);
-                cpu.r[rd] = result;
-
-                cpu.cpsr.n.write(result >> 31 & 1 == 1);
-                cpu.cpsr.z.write(result == 0);
-                cpu.cpsr.c.write(overflow == 0b1);
-                cpu.cpsr.v.write(((op1 ^ result) & (op2 ^ result)) >> 31 & 1 == 1);
-            }
-        }
-    }.inner;
-}
-
-pub fn fmt3(comptime op: u2, comptime rd: u3) InstrFn {
-    return struct {
-        fn inner(cpu: *Arm7tdmi, _: *Bus, opcode: u16) void {
-            const op1 = cpu.r[rd];
-            const op2: u32 = opcode & 0xFF; // Offset
-
-            var overflow: u1 = undefined;
-            const result: u32 = switch (op) {
-                0b00 => op2, // MOV
-                0b01 => op1 -% op2, // CMP
-                0b10 => add(&overflow, op1, op2), // ADD
-                0b11 => op1 -% op2, // SUB
-            };
-
-            // Write to Register
-            if (op != 0b01) cpu.r[rd] = result;
-
-            // Write Flags
-            cpu.cpsr.n.write(result >> 31 & 1 == 1);
-            cpu.cpsr.z.write(result == 0);
-
-            switch (op) {
-                0b00 => {}, // MOV | C set by Barrel Shifter, V is unaffected
-                0b01, 0b11 => {
-                    // SUB, CMP
-                    cpu.cpsr.c.write(op2 <= op1);
-                    cpu.cpsr.v.write(((op1 ^ result) & (~op2 ^ result)) >> 31 & 1 == 1);
-                },
-                0b10 => {
-                    // ADD
-                    cpu.cpsr.c.write(overflow == 0b1);
-                    cpu.cpsr.v.write(((op1 ^ result) & (op2 ^ result)) >> 31 & 1 == 1);
-                },
-            }
-        }
-    }.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] & ~@as(u32, 2);
-            const right = (opcode & 0xFF) << 2;
-            cpu.r[rd] = left + right;
-        }
-    }.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

@@ -1,145 +0,0 @@
-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;
-
-            // Bit 1 of the PC intentionally ignored
-            cpu.r[rd] = bus.read(u32, (cpu.r[15] & ~@as(u32, 2)) + 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
-                        const value = bus.read(u16, address);
-                        cpu.r[rd] = if (address & 1 == 1) sext(u32, u8, @truncate(u8, value >> 8)) else sext(u32, u16, value);
-                    },
-                }
-            } 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

@@ -1,23 +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 ret_addr = cpu.r[15] - 2;
-            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] = ret_addr; // Resume Execution
-            cpu.spsr.raw = cpsr; // Previous mode CPSR
-            cpu.r[15] = 0x0000_0008;
-            cpu.pipe.reload(cpu);
-        }
-    }.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

@@ -1,13 +1,43 @@
 const std = @import("std");
-const SDL = @import("sdl2");
 const config = @import("../config.zig");
+const c = @import("../lib.zig").c;
 
 const Scheduler = @import("scheduler.zig").Scheduler;
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
-const FpsTracker = @import("../util.zig").FpsTracker;
+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 Atomic = std.atomic.Atomic;
+
+pub const Synchro = struct {
+    const AtomicBool = std.atomic.Value(bool);
+
+    // 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,
+    };
+
+    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;
@@ -24,7 +54,7 @@ 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 = @intToFloat(f64, clock_rate) / cycles_per_frame;
+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 {
+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(.Monotonic)) {
+            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(.Monotonic)) {
+            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;
 
@@ -87,12 +127,28 @@ 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.nextTimestamp();
             } else {
@@ -104,22 +160,35 @@ 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);
+fn audioSync(audio_sync: bool, stream: *c.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;
 
-    // Determine whether the APU is busy right at this moment
-    var still_full: bool = SDL.SDL_AudioStreamAvailable(stream) > sample_size * if (is_buffer_full.*) max_buf_size >> 1 else max_buf_size;
-    defer is_buffer_full.* = still_full; // Update APU Busy status right before exiting scope
+    _ = audio_sync;
+    _ = stream;
+    _ = is_buffer_full;
 
-    // If Busy is false, there's no need to sync here
-    if (!still_full) return;
+    _ = sample_size;
+    _ = max_buf_size;
 
-    while (true) {
-        still_full = SDL.SDL_AudioStreamAvailable(stream) > sample_size * max_buf_size >> 1;
-        if (!audio_sync or !still_full) break;
-    }
+    // TODO(paoda): re-enable
+
+    // // Determine whether the APU is busy right at this moment
+    // var still_full: bool = SDL.SDL_AudioStreamAvailable(stream) > sample_size * if (is_buffer_full.*) max_buf_size >> 1 else max_buf_size;
+    // defer is_buffer_full.* = still_full; // Update APU Busy status right before exiting scope
+
+    // // 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;
+    // }
 }
 
 fn videoSync(timer: *Timer, wake_time: u64) u64 {
@@ -146,10 +215,9 @@ fn sleep(timer: *Timer, wake_time: u64) ?u64 {
 
     const step = 2 * std.time.ns_per_ms; // Granularity of 2ms
     const times = sleep_for / step;
-    var i: usize = 0;
 
-    while (i < times) : (i += 1) {
-        std.time.sleep(step);
+    for (0..times) |_| {
+        std.Thread.sleep(step);
 
         // 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
@@ -161,5 +229,129 @@ fn sleep(timer: *Timer, wake_time: u64) ?u64 {
 }
 
 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) Interface {
+        return Interface.init(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.zig

@@ -2,15 +2,16 @@ const std = @import("std");
 const io = @import("bus/io.zig");
 const util = @import("../util.zig");
 
-const Bit = @import("bitfield").Bit;
-const Bitfield = @import("bitfield").Bitfield;
+const Bit = @import("bitjuggle").Boolean;
+const Bitfield = @import("bitjuggle").Bitfield;
 const dma = @import("bus/dma.zig");
 
 const Oam = @import("ppu/Oam.zig");
 const Palette = @import("ppu/Palette.zig");
 const Vram = @import("ppu/Vram.zig");
 const Scheduler = @import("scheduler.zig").Scheduler;
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("Bus.zig");
 const FrameBuffer = @import("../util.zig").FrameBuffer;
 
 const Allocator = std.mem.Allocator;
@@ -20,12 +21,14 @@ const getHalf = util.getHalf;
 const setHalf = util.setHalf;
 const setQuart = util.setQuart;
 
+const handleInterrupt = @import("cpu_util.zig").handleInterrupt;
+
 pub const width = 240;
 pub const height = 160;
 pub const framebuf_pitch = width * @sizeOf(u32);
 
 pub fn read(comptime T: type, ppu: *const Ppu, addr: u32) ?T {
-    const byte_addr = @truncate(u8, addr);
+    const byte_addr: u8 = @truncate(addr);
 
     return switch (T) {
         u32 => switch (byte_addr) {
@@ -66,24 +69,24 @@ pub fn read(comptime T: type, ppu: *const Ppu, addr: u32) ?T {
             else => util.io.read.err(T, log, "unaligned {} read from 0x{X:0>8}", .{ T, addr }),
         },
         u8 => switch (byte_addr) {
-            0x00, 0x01 => @truncate(T, ppu.dispcnt.raw >> getHalf(byte_addr)),
+            0x00, 0x01 => @truncate(ppu.dispcnt.raw >> getHalf(byte_addr)),
             0x02, 0x03 => null,
-            0x04, 0x05 => @truncate(T, ppu.dispstat.raw >> getHalf(byte_addr)),
-            0x06, 0x07 => @truncate(T, ppu.vcount.raw >> getHalf(byte_addr)),
-            0x08, 0x09 => @truncate(T, ppu.bg[0].bg0Cnt() >> getHalf(byte_addr)),
-            0x0A, 0x0B => @truncate(T, ppu.bg[1].bg1Cnt() >> getHalf(byte_addr)),
-            0x0C, 0x0D => @truncate(T, ppu.bg[2].cnt.raw >> getHalf(byte_addr)),
-            0x0E, 0x0F => @truncate(T, ppu.bg[3].cnt.raw >> getHalf(byte_addr)),
+            0x04, 0x05 => @truncate(ppu.dispstat.raw >> getHalf(byte_addr)),
+            0x06, 0x07 => @truncate(ppu.vcount.raw >> getHalf(byte_addr)),
+            0x08, 0x09 => @truncate(ppu.bg[0].bg0Cnt() >> getHalf(byte_addr)),
+            0x0A, 0x0B => @truncate(ppu.bg[1].bg1Cnt() >> getHalf(byte_addr)),
+            0x0C, 0x0D => @truncate(ppu.bg[2].cnt.raw >> getHalf(byte_addr)),
+            0x0E, 0x0F => @truncate(ppu.bg[3].cnt.raw >> getHalf(byte_addr)),
             0x10...0x1F => null, // BGXHOFS/VOFS
             0x20...0x2F => null, // BG2 Rot/Scaling
             0x30...0x3F => null, // BG3 Rot/Scaling
             0x40...0x47 => null, // WINXH/V Registers
-            0x48, 0x49 => @truncate(T, ppu.win.getIn() >> getHalf(byte_addr)),
-            0x4A, 0x4B => @truncate(T, ppu.win.getOut() >> getHalf(byte_addr)),
+            0x48, 0x49 => @truncate(ppu.win.getIn() >> getHalf(byte_addr)),
+            0x4A, 0x4B => @truncate(ppu.win.getOut() >> getHalf(byte_addr)),
             0x4C, 0x4D => null, // MOSAIC
             0x4E, 0x4F => null,
-            0x50, 0x51 => @truncate(T, ppu.bld.getCnt() >> getHalf(byte_addr)),
-            0x52, 0x53 => @truncate(T, ppu.bld.getAlpha() >> getHalf(byte_addr)),
+            0x50, 0x51 => @truncate(ppu.bld.getCnt() >> getHalf(byte_addr)),
+            0x52, 0x53 => @truncate(ppu.bld.getAlpha() >> getHalf(byte_addr)),
             0x54, 0x55 => null, // BLDY
             else => util.io.read.err(T, log, "unexpected {} read from 0x{X:0>8}", .{ T, addr }),
         },
@@ -92,14 +95,14 @@ pub fn read(comptime T: type, ppu: *const Ppu, addr: u32) ?T {
 }
 
 pub fn write(comptime T: type, ppu: *Ppu, addr: u32, value: T) void {
-    const byte_addr = @truncate(u8, addr); // prefixed with 0x0400_00
+    const byte_addr: u8 = @truncate(addr); // prefixed with 0x0400_00
 
     switch (T) {
         u32 => switch (byte_addr) {
-            0x00 => ppu.dispcnt.raw = @truncate(u16, value),
+            0x00 => ppu.dispcnt.raw = @truncate(value),
             0x04 => {
-                ppu.dispstat.set(@truncate(u16, value));
-                ppu.vcount.raw = @truncate(u16, value >> 16);
+                ppu.dispstat.set(@truncate(value));
+                ppu.vcount.raw = @truncate(value >> 16);
             },
             0x08 => ppu.setAdjCnts(0, value),
             0x0C => ppu.setAdjCnts(2, value),
@@ -125,10 +128,10 @@ pub fn write(comptime T: type, ppu: *Ppu, addr: u32, value: T) void {
             0x4C => log.debug("Wrote 0x{X:0>8} to MOSAIC", .{value}),
 
             0x50 => {
-                ppu.bld.cnt.raw = @truncate(u16, value);
-                ppu.bld.alpha.raw = @truncate(u16, value >> 16);
+                ppu.bld.cnt.raw = @truncate(value);
+                ppu.bld.alpha.raw = @truncate(value >> 16);
             },
-            0x54 => ppu.bld.y.raw = @truncate(u16, value),
+            0x54 => ppu.bld.y.raw = @truncate(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_addr) {
@@ -151,19 +154,19 @@ pub fn write(comptime T: type, ppu: *Ppu, addr: u32, value: T) void {
             0x1C => ppu.bg[3].hofs.raw = value,
             0x1E => ppu.bg[3].vofs.raw = value,
 
-            0x20 => ppu.aff_bg[0].pa = @bitCast(i16, value),
-            0x22 => ppu.aff_bg[0].pb = @bitCast(i16, value),
-            0x24 => ppu.aff_bg[0].pc = @bitCast(i16, value),
-            0x26 => ppu.aff_bg[0].pd = @bitCast(i16, value),
-            0x28, 0x2A => ppu.aff_bg[0].x = @bitCast(i32, setHalf(u32, @bitCast(u32, ppu.aff_bg[0].x), byte_addr, value)),
-            0x2C, 0x2E => ppu.aff_bg[0].y = @bitCast(i32, setHalf(u32, @bitCast(u32, ppu.aff_bg[0].y), byte_addr, value)),
+            0x20 => ppu.aff_bg[0].pa = @bitCast(value),
+            0x22 => ppu.aff_bg[0].pb = @bitCast(value),
+            0x24 => ppu.aff_bg[0].pc = @bitCast(value),
+            0x26 => ppu.aff_bg[0].pd = @bitCast(value),
+            0x28, 0x2A => ppu.aff_bg[0].x = @bitCast(setHalf(u32, @as(u32, @bitCast(ppu.aff_bg[0].x)), byte_addr, value)),
+            0x2C, 0x2E => ppu.aff_bg[0].y = @bitCast(setHalf(u32, @as(u32, @bitCast(ppu.aff_bg[0].y)), byte_addr, value)),
 
-            0x30 => ppu.aff_bg[1].pa = @bitCast(i16, value),
-            0x32 => ppu.aff_bg[1].pb = @bitCast(i16, value),
-            0x34 => ppu.aff_bg[1].pc = @bitCast(i16, value),
-            0x36 => ppu.aff_bg[1].pd = @bitCast(i16, value),
-            0x38, 0x3A => ppu.aff_bg[1].x = @bitCast(i32, setHalf(u32, @bitCast(u32, ppu.aff_bg[1].x), byte_addr, value)),
-            0x3C, 0x3E => ppu.aff_bg[1].y = @bitCast(i32, setHalf(u32, @bitCast(u32, ppu.aff_bg[1].y), byte_addr, value)),
+            0x30 => ppu.aff_bg[1].pa = @bitCast(value),
+            0x32 => ppu.aff_bg[1].pb = @bitCast(value),
+            0x34 => ppu.aff_bg[1].pc = @bitCast(value),
+            0x36 => ppu.aff_bg[1].pd = @bitCast(value),
+            0x38, 0x3A => ppu.aff_bg[1].x = @bitCast(setHalf(u32, @as(u32, @bitCast(ppu.aff_bg[1].x)), byte_addr, value)),
+            0x3C, 0x3E => ppu.aff_bg[1].y = @bitCast(setHalf(u32, @as(u32, @bitCast(ppu.aff_bg[1].y)), byte_addr, value)),
 
             0x40 => ppu.win.h[0].raw = value,
             0x42 => ppu.win.h[1].raw = value,
@@ -202,20 +205,20 @@ pub fn write(comptime T: type, ppu: *Ppu, addr: u32, value: T) void {
             0x1E, 0x1F => ppu.bg[3].vofs.raw = setHalf(u16, ppu.bg[3].vofs.raw, byte_addr, value),
 
             // BG2 Rot/Scaling
-            0x20, 0x21 => ppu.aff_bg[0].pa = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[0].pa), byte_addr, value)),
-            0x22, 0x23 => ppu.aff_bg[0].pb = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[0].pb), byte_addr, value)),
-            0x24, 0x25 => ppu.aff_bg[0].pc = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[0].pc), byte_addr, value)),
-            0x26, 0x27 => ppu.aff_bg[0].pd = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[0].pd), byte_addr, value)),
-            0x28, 0x29, 0x2A, 0x2B => ppu.aff_bg[0].x = @bitCast(i32, setQuart(@bitCast(u32, ppu.aff_bg[0].x), byte_addr, value)),
-            0x2C, 0x2D, 0x2E, 0x2F => ppu.aff_bg[0].y = @bitCast(i32, setQuart(@bitCast(u32, ppu.aff_bg[0].y), byte_addr, value)),
+            0x20, 0x21 => ppu.aff_bg[0].pa = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[0].pa)), byte_addr, value)),
+            0x22, 0x23 => ppu.aff_bg[0].pb = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[0].pb)), byte_addr, value)),
+            0x24, 0x25 => ppu.aff_bg[0].pc = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[0].pc)), byte_addr, value)),
+            0x26, 0x27 => ppu.aff_bg[0].pd = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[0].pd)), byte_addr, value)),
+            0x28, 0x29, 0x2A, 0x2B => ppu.aff_bg[0].x = @bitCast(setQuart(@bitCast(ppu.aff_bg[0].x), byte_addr, value)),
+            0x2C, 0x2D, 0x2E, 0x2F => ppu.aff_bg[0].y = @bitCast(setQuart(@bitCast(ppu.aff_bg[0].y), byte_addr, value)),
 
             // BG3 Rot/Scaling
-            0x30, 0x31 => ppu.aff_bg[1].pa = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[1].pa), byte_addr, value)),
-            0x32, 0x33 => ppu.aff_bg[1].pb = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[1].pb), byte_addr, value)),
-            0x34, 0x35 => ppu.aff_bg[1].pc = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[1].pc), byte_addr, value)),
-            0x36, 0x37 => ppu.aff_bg[1].pd = @bitCast(i16, setHalf(u16, @bitCast(u16, ppu.aff_bg[1].pd), byte_addr, value)),
-            0x38, 0x39, 0x3A, 0x3B => ppu.aff_bg[1].x = @bitCast(i32, setQuart(@bitCast(u32, ppu.aff_bg[1].x), byte_addr, value)),
-            0x3C, 0x3D, 0x3E, 0x3F => ppu.aff_bg[1].y = @bitCast(i32, setQuart(@bitCast(u32, ppu.aff_bg[1].y), byte_addr, value)),
+            0x30, 0x31 => ppu.aff_bg[1].pa = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[1].pa)), byte_addr, value)),
+            0x32, 0x33 => ppu.aff_bg[1].pb = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[1].pb)), byte_addr, value)),
+            0x34, 0x35 => ppu.aff_bg[1].pc = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[1].pc)), byte_addr, value)),
+            0x36, 0x37 => ppu.aff_bg[1].pd = @bitCast(setHalf(u16, @as(u16, @bitCast(ppu.aff_bg[1].pd)), byte_addr, value)),
+            0x38, 0x39, 0x3A, 0x3B => ppu.aff_bg[1].x = @bitCast(setQuart(@bitCast(ppu.aff_bg[1].x), byte_addr, value)),
+            0x3C, 0x3D, 0x3E, 0x3F => ppu.aff_bg[1].y = @bitCast(setQuart(@bitCast(ppu.aff_bg[1].y), byte_addr, value)),
 
             // Window
             0x40, 0x41 => ppu.win.h[0].raw = setHalf(u16, ppu.win.h[0].raw, byte_addr, value),
@@ -266,7 +269,7 @@ pub const Ppu = struct {
         sched.push(.Draw, 240 * 4); // Add first PPU Event to Scheduler
 
         const sprites = try allocator.create([128]?Sprite);
-        std.mem.set(?Sprite, sprites, null);
+        @memset(sprites, null);
 
         return Self{
             .vram = try Vram.init(allocator),
@@ -290,6 +293,26 @@ pub const Ppu = struct {
         };
     }
 
+    pub fn reset(self: *Self) void {
+        self.sched.push(.Draw, 240 * 4);
+
+        self.vram.reset();
+        self.palette.reset();
+        self.oam.reset();
+        self.framebuf.reset();
+
+        self.win = Window.init();
+        self.bg = [_]Background{Background.init()} ** 4;
+        self.aff_bg = [_]AffineBackground{AffineBackground.init()} ** 2;
+        self.bld = Blend.create();
+        self.dispcnt = .{ .raw = 0x0000 };
+        self.dispstat = .{ .raw = 0x0000 };
+        self.vcount = .{ .raw = 0x0000 };
+
+        self.scanline.reset();
+        @memset(self.scanline_sprites, null);
+    }
+
     pub fn deinit(self: *Self) void {
         self.allocator.destroy(self.scanline_sprites);
         self.framebuf.deinit();
@@ -301,13 +324,13 @@ pub const Ppu = struct {
     }
 
     pub fn setBgOffsets(self: *Self, comptime n: u2, word: u32) void {
-        self.bg[n].hofs.raw = @truncate(u16, word);
-        self.bg[n].vofs.raw = @truncate(u16, word >> 16);
+        self.bg[n].hofs.raw = @truncate(word);
+        self.bg[n].vofs.raw = @truncate(word >> 16);
     }
 
     pub fn setAdjCnts(self: *Self, comptime n: u2, word: u32) void {
-        self.bg[n].cnt.raw = @truncate(u16, word);
-        self.bg[n + 1].cnt.raw = @truncate(u16, word >> 16);
+        self.bg[n].cnt.raw = @truncate(word);
+        self.bg[n + 1].cnt.raw = @truncate(word >> 16);
     }
 
     /// Search OAM for Sprites that might be rendered on this scanline
@@ -318,11 +341,11 @@ pub const Ppu = struct {
         search: while (i < self.oam.buf.len) : (i += 8) {
             // Attributes in OAM are 6 bytes long, with 2 bytes of padding
             // Grab Attributes from OAM
-            const attr0 = @bitCast(Attr0, self.oam.read(u16, i));
+            const attr0: Attr0 = @bitCast(self.oam.read(u16, i));
 
             // Only consider enabled Sprites
             if (attr0.is_affine.read() or !attr0.disabled.read()) {
-                const attr1 = @bitCast(Attr1, self.oam.read(u16, i + 2));
+                const attr1: Attr1 = @bitCast(self.oam.read(u16, i + 2));
                 const d = spriteDimensions(attr0.shape.read(), attr1.size.read());
 
                 // Account for double-size affine sprites
@@ -345,7 +368,7 @@ pub const Ppu = struct {
                 if (y_pos <= y and y < (y_pos + sprite_height)) {
                     for (self.scanline_sprites) |*maybe_sprite| {
                         if (maybe_sprite.* == null) {
-                            maybe_sprite.* = Sprite.init(attr0, attr1, @bitCast(Attr2, self.oam.read(u16, i + 4)));
+                            maybe_sprite.* = Sprite.init(attr0, attr1, @bitCast(self.oam.read(u16, i + 4)));
                             continue :search;
                         }
                     }
@@ -391,7 +414,7 @@ pub const Ppu = struct {
         const pc = self.oam.read(u16, base + 11 * @sizeOf(u16));
         const pd = self.oam.read(u16, base + 15 * @sizeOf(u16));
 
-        const matrix = @bitCast([4]i16, [_]u16{ pa, pb, pc, pd });
+        const matrix: [4]i16 = @bitCast([_]u16{ pa, pb, pc, pd });
 
         const sprite_width = sprite.width << if (double_size) 1 else 0;
         const sprite_height = sprite.height << if (double_size) 1 else 0;
@@ -402,7 +425,7 @@ pub const Ppu = struct {
         var i: u9 = 0;
         while (i < sprite_width) : (i += 1) {
             // TODO: Something is wrong here
-            const x = @truncate(u9, @bitCast(u16, sprite_x + i));
+            const x: u9 = @truncate(@as(u16, @bitCast(sprite_x + i)));
             if (x >= width) continue;
 
             if (!shouldDrawSprite(self.bld.cnt, &self.scanline, x)) continue;
@@ -424,11 +447,11 @@ pub const Ppu = struct {
             // Maybe this is the necessary check?
             if (rot_x >= sprite.width or rot_y >= sprite.height or rot_x < 0 or rot_y < 0) continue;
 
-            const tile_x = @bitCast(u16, rot_x);
-            const tile_y = @bitCast(u16, rot_y);
+            const tile_x: u16 = @bitCast(rot_x);
+            const tile_y: u16 = @bitCast(rot_y);
 
-            const col = @truncate(u3, tile_x);
-            const row = @truncate(u3, tile_y);
+            const col: u3 = @truncate(tile_x);
+            const row: u3 = @truncate(tile_y);
 
             // TODO: Finish that 2D Sprites Test ROM
             const tile_base = char_base + (tile_id * 0x20) + (row * tile_row_offset) + if (is_8bpp) col else col >> 1;
@@ -438,12 +461,12 @@ pub const Ppu = struct {
             const tile = self.vram.buf[tile_base + tile_offset];
             const pal_id: u16 = if (!is_8bpp) get4bppTilePalette(sprite.palBank(), col, tile) else tile;
 
-            const global_x = @truncate(u9, @bitCast(u16, local_x + sprite_x));
+            const global_x: u9 = @truncate(@as(u16, @bitCast(local_x + sprite_x)));
 
             // Sprite Palette starts at 0x0500_0200
             if (pal_id != 0) {
                 const bgr555 = self.palette.read(u16, 0x200 + pal_id * 2);
-                drawSpritePixel(self.bld.cnt, &self.scanline, @bitCast(Attr0, sprite.attr0), global_x, bgr555);
+                drawSpritePixel(self.bld.cnt, &self.scanline, @as(Attr0, @bitCast(sprite.attr0)), global_x, bgr555);
             }
         }
     }
@@ -467,7 +490,7 @@ pub const Ppu = struct {
         var i: u9 = 0;
         while (i < sprite.width) : (i += 1) {
             // TODO: Something is Wrong Here
-            const x = @truncate(u9, @bitCast(u16, sprite_x + i));
+            const x = @as(u9, @truncate(@as(u16, @bitCast(sprite_x + i))));
             if (x >= width) continue;
 
             if (!shouldDrawSprite(self.bld.cnt, &self.scanline, x)) continue;
@@ -481,11 +504,11 @@ pub const Ppu = struct {
             // Note that we flip the tile_pos not the (tile_pos % 8) like we do for
             // Background Tiles. By doing this we mirror the entire sprite instead of
             // just a specific tile (see how sprite.width and sprite.height are involved)
-            const tile_x = @intCast(u9, local_x) ^ if (sprite.hFlip()) (sprite.width - 1) else 0;
-            const tile_y = @intCast(u8, local_y) ^ if (sprite.vFlip()) (sprite.height - 1) else 0;
+            const tile_x = @as(u9, @intCast(local_x)) ^ if (sprite.hFlip()) (sprite.width - 1) else 0;
+            const tile_y = @as(u8, @intCast(local_y)) ^ if (sprite.vFlip()) (sprite.height - 1) else 0;
 
-            const col = @truncate(u3, tile_x);
-            const row = @truncate(u3, tile_y);
+            const col: u3 = @truncate(tile_x);
+            const row: u3 = @truncate(tile_y);
 
             // TODO: Finish that 2D Sprites Test ROM
             const tile_base = char_base + (tile_id * 0x20) + (row * tile_row_offset) + if (is_8bpp) col else col >> 1;
@@ -495,7 +518,7 @@ pub const Ppu = struct {
             const tile = self.vram.buf[tile_base + tile_offset];
             const pal_id: u16 = if (!is_8bpp) get4bppTilePalette(sprite.palBank(), col, tile) else tile;
 
-            const global_x = @truncate(u9, @bitCast(u16, local_x + sprite_x));
+            const global_x: u9 = @truncate(@as(u16, @bitCast(local_x + sprite_x)));
 
             // Sprite Palette starts at 0x0500_0200
             if (pal_id != 0) {
@@ -527,8 +550,8 @@ pub const Ppu = struct {
             aff_x += self.aff_bg[n - 2].pa;
             aff_y += self.aff_bg[n - 2].pc;
 
-            const _x = @truncate(u9, @bitCast(u32, ix));
-            const _y = @truncate(u8, @bitCast(u32, iy));
+            const _x: u9 = @truncate(@as(u32, @bitCast(ix)));
+            const _y: u8 = @truncate(@as(u32, @bitCast(iy)));
 
             const win_bounds = self.windowBounds(_x, _y);
             if (!shouldDrawBackground(self, n, win_bounds, i)) continue;
@@ -538,10 +561,10 @@ pub const Ppu = struct {
                 iy = if (iy > px_height) @rem(iy, px_height) else if (iy < 0) px_height + @rem(iy, px_height) else iy;
             } else if (ix > px_width or iy > px_height or ix < 0 or iy < 0) continue;
 
-            const x = @bitCast(u32, ix);
-            const y = @bitCast(u32, iy);
+            const x: u32 = @bitCast(ix);
+            const y: u32 = @bitCast(iy);
 
-            const tile_id: u32 = self.vram.read(u8, screen_base + ((y / 8) * @bitCast(u32, tile_width) + (x / 8)));
+            const tile_id: u32 = self.vram.read(u8, screen_base + ((y / 8) * @as(u32, @bitCast(tile_width)) + (x / 8)));
             const row = y & 7;
             const col = x & 7;
 
@@ -578,12 +601,12 @@ pub const Ppu = struct {
         while (i < width) : (i += 1) {
             const x = hofs + i;
 
-            const win_bounds = self.windowBounds(@truncate(u9, x), @truncate(u8, y));
+            const win_bounds = self.windowBounds(@truncate(x), @truncate(y));
             if (!shouldDrawBackground(self, n, win_bounds, i)) continue;
 
             // Grab the Screen Entry from VRAM
             const entry_addr = screen_base + tilemapOffset(size, x, y);
-            const entry = @bitCast(ScreenEntry, self.vram.read(u16, entry_addr));
+            const entry: ScreenEntry = @bitCast(self.vram.read(u16, entry_addr));
 
             // Calculate the Address of the Tile in the designated Charblock
             // We also take this opportunity to flip tiles if necessary
@@ -592,8 +615,8 @@ pub const Ppu = struct {
             // Calculate row and column offsets. Understand that
             // `tile_len`, `tile_row_offset` and `col` are subject to different
             // values depending on whether we are in 4bpp or 8bpp mode.
-            const row = @truncate(u3, y) ^ if (entry.v_flip.read()) 7 else @as(u3, 0);
-            const col = @truncate(u3, x) ^ if (entry.h_flip.read()) 7 else @as(u3, 0);
+            const row = @as(u3, @truncate(y)) ^ if (entry.v_flip.read()) 7 else @as(u3, 0);
+            const col = @as(u3, @truncate(x)) ^ if (entry.h_flip.read()) 7 else @as(u3, 0);
             const tile_addr = char_base + (tile_id * tile_len) + (row * tile_row_offset) + if (is_8bpp) col else col >> 1;
 
             const tile = self.vram.buf[tile_addr];
@@ -625,9 +648,8 @@ pub const Ppu = struct {
                 const framebuf_base = width * @as(usize, scanline);
                 if (obj_enable) self.fetchSprites();
 
-                var layer: usize = 0;
-                while (layer < 4) : (layer += 1) {
-                    self.drawSprites(@truncate(u2, layer));
+                for (0..4) |layer| {
+                    self.drawSprites(@truncate(layer));
                     if (layer == self.bg[0].cnt.priority.read() and bg_enable & 1 == 1) self.drawBackground(0);
                     if (layer == self.bg[1].cnt.priority.read() and bg_enable >> 1 & 1 == 1) self.drawBackground(1);
                     if (layer == self.bg[2].cnt.priority.read() and bg_enable >> 2 & 1 == 1) self.drawBackground(2);
@@ -640,9 +662,8 @@ pub const Ppu = struct {
                 const framebuf_base = width * @as(usize, scanline);
                 if (obj_enable) self.fetchSprites();
 
-                var layer: usize = 0;
-                while (layer < 4) : (layer += 1) {
-                    self.drawSprites(@truncate(u2, layer));
+                for (0..4) |layer| {
+                    self.drawSprites(@as(u2, @truncate(layer)));
                     if (layer == self.bg[0].cnt.priority.read() and bg_enable & 1 == 1) self.drawBackground(0);
                     if (layer == self.bg[1].cnt.priority.read() and bg_enable >> 1 & 1 == 1) self.drawBackground(1);
                     if (layer == self.bg[2].cnt.priority.read() and bg_enable >> 2 & 1 == 1) self.drawAffineBackground(2);
@@ -654,9 +675,8 @@ pub const Ppu = struct {
                 const framebuf_base = width * @as(usize, scanline);
                 if (obj_enable) self.fetchSprites();
 
-                var layer: usize = 0;
-                while (layer < 4) : (layer += 1) {
-                    self.drawSprites(@truncate(u2, layer));
+                for (0..4) |layer| {
+                    self.drawSprites(@truncate(layer));
                     if (layer == self.bg[2].cnt.priority.read() and bg_enable >> 2 & 1 == 1) self.drawAffineBackground(2);
                     if (layer == self.bg[3].cnt.priority.read() and bg_enable >> 3 & 1 == 1) self.drawAffineBackground(3);
                 }
@@ -668,10 +688,10 @@ pub const Ppu = struct {
                 const framebuf_base = width * @as(usize, scanline);
 
                 // FIXME: @ptrCast between slices changing the length isn't implemented yet
-                const vram_buf = @ptrCast([*]const u16, @alignCast(@alignOf(u16), self.vram.buf));
-                const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), self.framebuf.get(.Emulator)));
+                const vram_buf: [*]const u16 = @ptrCast(@alignCast(self.vram.buf));
+                const framebuf: [*]u32 = @ptrCast(@alignCast(self.framebuf.get(.Emulator)));
 
-                for (vram_buf[vram_base .. vram_base + width]) |bgr555, i| {
+                for (vram_buf[vram_base .. vram_base + width], 0..) |bgr555, i| {
                     framebuf[framebuf_base + i] = rgba888(bgr555);
                 }
             },
@@ -682,10 +702,10 @@ pub const Ppu = struct {
                 const framebuf_base = width * @as(usize, scanline);
 
                 // FIXME: @ptrCast between slices changing the length isn't implemented yet
-                const pal_buf = @ptrCast([*]const u16, @alignCast(@alignOf(u16), self.palette.buf));
-                const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), self.framebuf.get(.Emulator)));
+                const pal_buf: [*]const u16 = @ptrCast(@alignCast(self.palette.buf));
+                const framebuf: [*]u32 = @ptrCast(@alignCast(self.framebuf.get(.Emulator)));
 
-                for (self.vram.buf[vram_base .. vram_base + width]) |pal_id, i| {
+                for (self.vram.buf[vram_base .. vram_base + width], 0..) |pal_id, i| {
                     framebuf[framebuf_base + i] = rgba888(pal_buf[pal_id]);
                 }
             },
@@ -698,11 +718,10 @@ pub const Ppu = struct {
                 const framebuf_base = width * @as(usize, scanline);
 
                 // FIXME: @ptrCast between slices changing the length isn't implemented yet
-                const vram_buf = @ptrCast([*]const u16, @alignCast(@alignOf(u16), self.vram.buf));
-                const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), self.framebuf.get(.Emulator)));
+                const vram_buf: [*]const u16 = @ptrCast(@alignCast(self.vram.buf));
+                const framebuf: [*]u32 = @ptrCast(@alignCast(self.framebuf.get(.Emulator)));
 
-                var i: usize = 0;
-                while (i < width) : (i += 1) {
+                for (0..width) |i| {
                     const bgr555 = if (scanline < m5_height and i < m5_width) vram_buf[vram_base + i] else self.palette.backdrop();
                     framebuf[framebuf_base + i] = rgba888(bgr555);
                 }
@@ -716,9 +735,9 @@ pub const Ppu = struct {
         // If there are any nulls present in self.scanline.top() it means that no background drew a pixel there, so draw backdrop
 
         // FIXME: @ptrCast between slices changing the length isn't implemented yet
-        const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), self.framebuf.get(.Emulator)));
+        const framebuf: [*]u32 = @ptrCast(@alignCast(self.framebuf.get(.Emulator)));
 
-        for (self.scanline.top()) |maybe_top, i| {
+        for (self.scanline.top(), 0..) |maybe_top, i| {
             const maybe_btm = self.scanline.btm()[i];
 
             const bgr555 = self.getBgr555(maybe_top, maybe_btm);
@@ -728,7 +747,7 @@ pub const Ppu = struct {
         // Reset Current Scanline Pixel Buffer and list of fetched sprites
         // in prep for next scanline
         self.scanline.reset();
-        std.mem.set(?Sprite, self.scanline_sprites, null);
+        @memset(self.scanline_sprites, null);
     }
 
     fn getBgr555(self: *Self, maybe_top: Scanline.Pixel, maybe_btm: Scanline.Pixel) u16 {
@@ -984,36 +1003,39 @@ pub const Ppu = struct {
     }
 
     pub fn onHdrawEnd(self: *Self, cpu: *Arm7tdmi, late: u64) void {
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
         // Transitioning to a Hblank
         if (self.dispstat.hblank_irq.read()) {
-            cpu.bus.io.irq.hblank.set();
-            cpu.handleInterrupt();
+            bus_ptr.io.irq.hblank.write(true);
+            handleInterrupt(cpu);
         }
 
-        // See if HBlank DMA is present and not enabled
-
+        // If we're not also in VBlank, attempt to run any pending DMA Reqs
         if (!self.dispstat.vblank.read())
-            dma.onBlanking(cpu.bus, .HBlank);
+            dma.onBlanking(bus_ptr, .HBlank);
 
-        self.dispstat.hblank.set();
+        self.dispstat.hblank.write(true);
         self.sched.push(.HBlank, 68 * 4 -| late);
     }
 
     pub fn onHblankEnd(self: *Self, cpu: *Arm7tdmi, late: u64) void {
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
         // The End of a Hblank (During Draw or Vblank)
         const old_scanline = self.vcount.scanline.read();
         const scanline = (old_scanline + 1) % 228;
 
         self.vcount.scanline.write(scanline);
-        self.dispstat.hblank.unset();
+        self.dispstat.hblank.write(false);
 
         // Perform Vc == VcT check
         const coincidence = scanline == self.dispstat.vcount_trigger.read();
         self.dispstat.coincidence.write(coincidence);
 
         if (coincidence and self.dispstat.vcount_irq.read()) {
-            cpu.bus.io.irq.coincidence.set();
-            cpu.handleInterrupt();
+            bus_ptr.io.irq.coincidence.write(true);
+            handleInterrupt(cpu);
         }
 
         if (scanline < 160) {
@@ -1024,21 +1046,21 @@ pub const Ppu = struct {
             if (scanline == 160) {
                 self.framebuf.swap(); // Swap FrameBuffers
 
-                self.dispstat.vblank.set();
+                self.dispstat.vblank.write(true);
 
                 if (self.dispstat.vblank_irq.read()) {
-                    cpu.bus.io.irq.vblank.set();
-                    cpu.handleInterrupt();
+                    bus_ptr.io.irq.vblank.write(true);
+                    handleInterrupt(cpu);
                 }
 
                 self.aff_bg[0].latchRefPoints();
                 self.aff_bg[1].latchRefPoints();
 
                 // See if Vblank DMA is present and not enabled
-                dma.onBlanking(cpu.bus, .VBlank);
+                dma.onBlanking(bus_ptr, .VBlank);
             }
 
-            if (scanline == 227) self.dispstat.vblank.unset();
+            if (scanline == 227) self.dispstat.vblank.write(false);
             self.sched.push(.VBlank, 240 * 4 -| late);
         }
     }
@@ -1126,18 +1148,18 @@ const Window = struct {
     }
 
     pub fn setH(self: *Self, value: u32) void {
-        self.h[0].raw = @truncate(u16, value);
-        self.h[1].raw = @truncate(u16, value >> 16);
+        self.h[0].raw = @truncate(value);
+        self.h[1].raw = @truncate(value >> 16);
     }
 
     pub fn setV(self: *Self, value: u32) void {
-        self.v[0].raw = @truncate(u16, value);
-        self.v[1].raw = @truncate(u16, value >> 16);
+        self.v[0].raw = @truncate(value);
+        self.v[1].raw = @truncate(value >> 16);
     }
 
     pub fn setIo(self: *Self, value: u32) void {
-        self.in.raw = @truncate(u16, value);
-        self.out.raw = @truncate(u16, value >> 16);
+        self.in.raw = @truncate(value);
+        self.out.raw = @truncate(value >> 16);
     }
 };
 
@@ -1200,23 +1222,23 @@ const AffineBackground = struct {
     }
 
     pub fn setX(self: *Self, is_vblank: bool, value: u32) void {
-        self.x = @bitCast(i32, value);
-        if (!is_vblank) self.x_latch = @bitCast(i32, value);
+        self.x = @bitCast(value);
+        if (!is_vblank) self.x_latch = @bitCast(value);
     }
 
     pub fn setY(self: *Self, is_vblank: bool, value: u32) void {
-        self.y = @bitCast(i32, value);
-        if (!is_vblank) self.y_latch = @bitCast(i32, value);
+        self.y = @bitCast(value);
+        if (!is_vblank) self.y_latch = @bitCast(value);
     }
 
     pub fn writePaPb(self: *Self, value: u32) void {
-        self.pa = @bitCast(i16, @truncate(u16, value));
-        self.pb = @bitCast(i16, @truncate(u16, value >> 16));
+        self.pa = @bitCast(@as(u16, @truncate(value)));
+        self.pb = @bitCast(@as(u16, @truncate(value >> 16)));
     }
 
     pub fn writePcPd(self: *Self, value: u32) void {
-        self.pc = @bitCast(i16, @truncate(u16, value));
-        self.pd = @bitCast(i16, @truncate(u16, value >> 16));
+        self.pc = @bitCast(@as(u16, @truncate(value)));
+        self.pd = @bitCast(@as(u16, @truncate(value >> 16)));
     }
 
     // Every Vblank BG?X/Y registers are latched
@@ -1425,9 +1447,9 @@ fn alphaBlend(top: u16, btm: u16, bldalpha: io.BldAlpha) u16 {
     const btm_g = (btm >> 5) & 0x1F;
     const btm_b = (btm >> 10) & 0x1F;
 
-    const bld_r = std.math.min(31, (top_r * eva + btm_r * evb) >> 4);
-    const bld_g = std.math.min(31, (top_g * eva + btm_g * evb) >> 4);
-    const bld_b = std.math.min(31, (top_b * eva + btm_b * evb) >> 4);
+    const bld_r: u16 = @min(31, (top_r * eva + btm_r * evb) >> 4);
+    const bld_g: u16 = @min(31, (top_g * eva + btm_g * evb) >> 4);
+    const bld_b: u16 = @min(31, (top_b * eva + btm_b * evb) >> 4);
 
     return (bld_b << 10) | (bld_g << 5) | bld_r;
 }
@@ -1539,7 +1561,7 @@ const Scanline = struct {
 
     fn init(allocator: Allocator) !Self {
         const buf = try allocator.alloc(Pixel, width * 2); // Top & Bottom Scanline
-        std.mem.set(Pixel, buf, .unset);
+        @memset(buf, .unset);
 
         return .{
             // Top & Bototm Layers
@@ -1550,7 +1572,7 @@ const Scanline = struct {
     }
 
     fn reset(self: *Self) void {
-        std.mem.set(Pixel, self.buf, .unset);
+        @memset(self.buf, .unset);
     }
 
     fn deinit(self: *Self) void {

src/core/ppu/Oam.zig

@@ -12,7 +12,7 @@ pub fn read(self: *const Self, comptime T: type, address: usize) T {
     const addr = address & 0x3FF;
 
     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("OAM: Unsupported read width"),
     };
 }
@@ -21,7 +21,7 @@ pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
     const addr = address & 0x3FF;
 
     switch (T) {
-        u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
+        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"),
     }
@@ -29,11 +29,15 @@ pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
 
 pub fn init(allocator: Allocator) !Self {
     const buf = try allocator.alloc(u8, buf_len);
-    std.mem.set(u8, buf, 0);
+    @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

@@ -12,7 +12,7 @@ pub fn read(self: *const Self, comptime T: type, address: usize) T {
     const addr = address & 0x3FF;
 
     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("PALRAM: Unsupported read width"),
     };
 }
@@ -21,10 +21,10 @@ pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
     const addr = address & 0x3FF;
 
     switch (T) {
-        u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
+        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.writeIntSliceLittle(u16, self.buf[align_addr..][0..@sizeOf(u16)], @as(u16, value) * 0x101);
+            std.mem.writeInt(u16, self.buf[align_addr..][0..@sizeOf(u16)], @as(u16, value) * 0x101, .little);
         },
         else => @compileError("PALRAM: Unsupported write width"),
     }
@@ -32,16 +32,20 @@ pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
 
 pub fn init(allocator: Allocator) !Self {
     const buf = try allocator.alloc(u8, buf_len);
-    std.mem.set(u8, buf, 0);
+    @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.readIntNative(u16, self.buf[0..2]);
+    return std.mem.readInt(u16, self.buf[0..2], .little);
 }

src/core/ppu/Vram.zig

@@ -13,7 +13,7 @@ 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.readIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)]),
+        u32, u16, u8 => std.mem.readInt(T, self.buf[addr..][0..@sizeOf(T)], .little),
         else => @compileError("VRAM: Unsupported read width"),
     };
 }
@@ -23,7 +23,7 @@ pub fn write(self: *Self, comptime T: type, dispcnt: io.DisplayControl, address:
     const idx = Self.mirror(address);
 
     switch (T) {
-        u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[idx..][0..@sizeOf(T)], value),
+        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) {
@@ -32,7 +32,7 @@ pub fn write(self: *Self, comptime T: type, dispcnt: io.DisplayControl, address:
             }
 
             const align_idx = idx & ~@as(u32, 1); // Aligned to a halfword boundary
-            std.mem.writeIntSliceLittle(u16, self.buf[align_idx..][0..@sizeOf(u16)], @as(u16, value) * 0x101);
+            std.mem.writeInt(u16, self.buf[align_idx..][0..@sizeOf(u16)], @as(u16, value) * 0x101, .little);
         },
         else => @compileError("VRAM: Unsupported write width"),
     }
@@ -40,11 +40,15 @@ pub fn write(self: *Self, comptime T: type, dispcnt: io.DisplayControl, address:
 
 pub fn init(allocator: Allocator) !Self {
     const buf = try allocator.alloc(u8, buf_len);
-    std.mem.set(u8, buf, 0);
+    @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/scheduler.zig

@@ -1,6 +1,7 @@
 const std = @import("std");
 
-const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
+const Arm7tdmi = @import("arm32").Arm7tdmi;
+const Bus = @import("Bus.zig");
 const Clock = @import("bus/gpio.zig").Clock;
 
 const Order = std.math.Order;
@@ -11,11 +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;
@@ -26,6 +27,17 @@ 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;
     }
@@ -34,6 +46,8 @@ pub const Scheduler = struct {
         const event = self.queue.remove();
         const late = self.tick - event.tick;
 
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
+
         switch (event.kind) {
             .HeatDeath => {
                 log.err("u64 overflow. This *actually* should never happen.", .{});
@@ -41,45 +55,45 @@ pub const Scheduler = struct {
             },
             .Draw => {
                 // The end of a VDraw
-                cpu.bus.ppu.drawScanline();
-                cpu.bus.ppu.onHdrawEnd(cpu, late);
+                bus_ptr.ppu.drawScanline();
+                bus_ptr.ppu.onHdrawEnd(cpu, late);
             },
             .TimerOverflow => |id| {
                 switch (id) {
-                    inline 0...3 => |idx| cpu.bus.tim[idx].onTimerExpire(cpu, late),
+                    inline 0...3 => |idx| bus_ptr.tim[idx].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),
+                    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 = &cpu.bus.pak.gpio.device;
+                const device = &bus_ptr.pak.gpio.device;
                 if (device.kind != .Rtc or device.ptr == null) return;
 
-                const clock = @ptrCast(*Clock, @alignCast(@alignOf(*Clock), device.ptr.?));
+                const clock: *Clock = @ptrCast(@alignCast(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
+            .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 {
-        for (self.queue.items) |event, i| {
+        for (self.queue.items, 0..) |event, i| {
             if (std.meta.eql(event.kind, needle)) {
 
                 // invalidates the slice we're iterating over
                 _ = self.queue.removeIndex(i);
 
-                log.debug("Removed {?}@{}", .{ event.kind, event.tick });
+                // log.debug("Removed {?}@{}", .{ event.kind, event.tick });
                 break;
             }
         }

src/imgui.zig

@@ -0,0 +1,516 @@
+//! 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.uint;
+
+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(dim.width, 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();
+
+        // FIXME(paoda): go look at some documentation here
+        zgui.image(.{ .tex_data = null, .tex_id = @enumFromInt(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 ported by Fabio Arnold", .{});
+        {
+            zgui.indent(.{});
+            defer zgui.unindent(.{});
+
+            zgui.bulletText("nativefiledialog by Michael Labbe", .{});
+        }
+
+        zgui.bulletText("SDL ported by Carl Åstholm", .{});
+        {
+            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 ported by Michal Ziulek et al.", .{});
+        {
+            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 et al.", .{});
+        zgui.bulletText("zigglgen ported by Carl Åstholm", .{});
+        {
+            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;
+        _ = scheduler;
+
+        // zgui.text("tick: {X:0>16}", .{scheduler.now()});
+        zgui.text("tick: (TODO: FIX THIS)", .{});
+        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 });
+            _ = event;
+            zgui.text("TODO: Fix This", .{});
+        }
+    }
+
+    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/lib.zig

@@ -0,0 +1,28 @@
+pub const c = @cImport({
+    @cDefine("SDL_DISABLE_OLD_NAMES", {});
+    @cDefine("SDL_MAIN_HANDLED", {});
+
+    @cInclude("SDL3/SDL.h");
+    @cInclude("SDL3/SDL_main.h");
+});
+
+// https://github.com/castholm/zig-examples/blob/77a829c85b5ddbad673026d504626015db4093ac/opengl-sdl/main.zig#L200-L219
+pub inline fn errify(value: anytype) error{sdl_error}!switch (@typeInfo(@TypeOf(value))) {
+    .bool => void,
+    .pointer, .optional => @TypeOf(value.?),
+    .int => |info| switch (info.signedness) {
+        .signed => @TypeOf(@max(0, value)),
+        .unsigned => @TypeOf(value),
+    },
+    else => @compileError("unerrifiable type: " ++ @typeName(@TypeOf(value))),
+} {
+    return switch (@typeInfo(@TypeOf(value))) {
+        .bool => if (!value) error.sdl_error,
+        .pointer, .optional => value orelse error.sdl_error,
+        .int => |info| switch (info.signedness) {
+            .signed => if (value >= 0) @max(0, value) else error.sdl_error,
+            .unsigned => if (value != 0) value else error.sdl_error,
+        },
+        else => comptime unreachable,
+    };
+}

src/lib/fifo.zig

@@ -0,0 +1,548 @@
+// FIFO of fixed size items
+// Usually used for e.g. byte buffers
+
+const std = @import("std");
+const math = std.math;
+const mem = std.mem;
+const Allocator = mem.Allocator;
+const assert = std.debug.assert;
+const testing = std.testing;
+
+pub const LinearFifoBufferType = union(enum) {
+    /// The buffer is internal to the fifo; it is of the specified size.
+    static: usize,
+
+    /// The buffer is passed as a slice to the initialiser.
+    slice,
+
+    /// The buffer is managed dynamically using a `mem.Allocator`.
+    dynamic,
+};
+
+pub fn LinearFifo(
+    comptime T: type,
+    comptime buffer_type: LinearFifoBufferType,
+) type {
+    const autoalign = false;
+
+    const powers_of_two = switch (buffer_type) {
+        .static => std.math.isPowerOfTwo(buffer_type.static),
+        .slice => false, // Any size slice could be passed in
+        .dynamic => true, // This could be configurable in future
+    };
+
+    return struct {
+        allocator: if (buffer_type == .dynamic) Allocator else void,
+        buf: if (buffer_type == .static) [buffer_type.static]T else []T,
+        head: usize,
+        count: usize,
+
+        const Self = @This();
+        pub const Reader = std.io.Reader(*Self, error{}, readFn);
+        pub const Writer = std.io.Writer(*Self, error{OutOfMemory}, appendWrite);
+
+        // Type of Self argument for slice operations.
+        // If buffer is inline (Static) then we need to ensure we haven't
+        // returned a slice into a copy on the stack
+        const SliceSelfArg = if (buffer_type == .static) *Self else Self;
+
+        pub const init = switch (buffer_type) {
+            .static => initStatic,
+            .slice => initSlice,
+            .dynamic => initDynamic,
+        };
+
+        fn initStatic() Self {
+            comptime assert(buffer_type == .static);
+            return .{
+                .allocator = {},
+                .buf = undefined,
+                .head = 0,
+                .count = 0,
+            };
+        }
+
+        fn initSlice(buf: []T) Self {
+            comptime assert(buffer_type == .slice);
+            return .{
+                .allocator = {},
+                .buf = buf,
+                .head = 0,
+                .count = 0,
+            };
+        }
+
+        fn initDynamic(allocator: Allocator) Self {
+            comptime assert(buffer_type == .dynamic);
+            return .{
+                .allocator = allocator,
+                .buf = &.{},
+                .head = 0,
+                .count = 0,
+            };
+        }
+
+        pub fn deinit(self: Self) void {
+            if (buffer_type == .dynamic) self.allocator.free(self.buf);
+        }
+
+        pub fn realign(self: *Self) void {
+            if (self.buf.len - self.head >= self.count) {
+                mem.copyForwards(T, self.buf[0..self.count], self.buf[self.head..][0..self.count]);
+                self.head = 0;
+            } else {
+                var tmp: [4096 / 2 / @sizeOf(T)]T = undefined;
+
+                while (self.head != 0) {
+                    const n = @min(self.head, tmp.len);
+                    const m = self.buf.len - n;
+                    @memcpy(tmp[0..n], self.buf[0..n]);
+                    mem.copyForwards(T, self.buf[0..m], self.buf[n..][0..m]);
+                    @memcpy(self.buf[m..][0..n], tmp[0..n]);
+                    self.head -= n;
+                }
+            }
+            { // set unused area to undefined
+                const unused = mem.sliceAsBytes(self.buf[self.count..]);
+                @memset(unused, undefined);
+            }
+        }
+
+        /// Reduce allocated capacity to `size`.
+        pub fn shrink(self: *Self, size: usize) void {
+            assert(size >= self.count);
+            if (buffer_type == .dynamic) {
+                self.realign();
+                self.buf = self.allocator.realloc(self.buf, size) catch |e| switch (e) {
+                    error.OutOfMemory => return, // no problem, capacity is still correct then.
+                };
+            }
+        }
+
+        /// Ensure that the buffer can fit at least `size` items
+        pub fn ensureTotalCapacity(self: *Self, size: usize) !void {
+            if (self.buf.len >= size) return;
+            if (buffer_type == .dynamic) {
+                self.realign();
+                const new_size = if (powers_of_two) math.ceilPowerOfTwo(usize, size) catch return error.OutOfMemory else size;
+                self.buf = try self.allocator.realloc(self.buf, new_size);
+            } else {
+                return error.OutOfMemory;
+            }
+        }
+
+        /// Makes sure at least `size` items are unused
+        pub fn ensureUnusedCapacity(self: *Self, size: usize) error{OutOfMemory}!void {
+            if (self.writableLength() >= size) return;
+
+            return try self.ensureTotalCapacity(math.add(usize, self.count, size) catch return error.OutOfMemory);
+        }
+
+        /// Returns number of items currently in fifo
+        pub fn readableLength(self: Self) usize {
+            return self.count;
+        }
+
+        /// Returns a writable slice from the 'read' end of the fifo
+        fn readableSliceMut(self: SliceSelfArg, offset: usize) []T {
+            if (offset > self.count) return &[_]T{};
+
+            var start = self.head + offset;
+            if (start >= self.buf.len) {
+                start -= self.buf.len;
+                return self.buf[start .. start + (self.count - offset)];
+            } else {
+                const end = @min(self.head + self.count, self.buf.len);
+                return self.buf[start..end];
+            }
+        }
+
+        /// Returns a readable slice from `offset`
+        pub fn readableSlice(self: SliceSelfArg, offset: usize) []const T {
+            return self.readableSliceMut(offset);
+        }
+
+        pub fn readableSliceOfLen(self: *Self, len: usize) []const T {
+            assert(len <= self.count);
+            const buf = self.readableSlice(0);
+            if (buf.len >= len) {
+                return buf[0..len];
+            } else {
+                self.realign();
+                return self.readableSlice(0)[0..len];
+            }
+        }
+
+        /// Discard first `count` items in the fifo
+        pub fn discard(self: *Self, count: usize) void {
+            assert(count <= self.count);
+            { // set old range to undefined. Note: may be wrapped around
+                const slice = self.readableSliceMut(0);
+                if (slice.len >= count) {
+                    const unused = mem.sliceAsBytes(slice[0..count]);
+                    @memset(unused, undefined);
+                } else {
+                    const unused = mem.sliceAsBytes(slice[0..]);
+                    @memset(unused, undefined);
+                    const unused2 = mem.sliceAsBytes(self.readableSliceMut(slice.len)[0 .. count - slice.len]);
+                    @memset(unused2, undefined);
+                }
+            }
+            if (autoalign and self.count == count) {
+                self.head = 0;
+                self.count = 0;
+            } else {
+                var head = self.head + count;
+                if (powers_of_two) {
+                    // Note it is safe to do a wrapping subtract as
+                    // bitwise & with all 1s is a noop
+                    head &= self.buf.len -% 1;
+                } else {
+                    head %= self.buf.len;
+                }
+                self.head = head;
+                self.count -= count;
+            }
+        }
+
+        /// Read the next item from the fifo
+        pub fn readItem(self: *Self) ?T {
+            if (self.count == 0) return null;
+
+            const c = self.buf[self.head];
+            self.discard(1);
+            return c;
+        }
+
+        /// Read data from the fifo into `dst`, returns number of items copied.
+        pub fn read(self: *Self, dst: []T) usize {
+            var dst_left = dst;
+
+            while (dst_left.len > 0) {
+                const slice = self.readableSlice(0);
+                if (slice.len == 0) break;
+                const n = @min(slice.len, dst_left.len);
+                @memcpy(dst_left[0..n], slice[0..n]);
+                self.discard(n);
+                dst_left = dst_left[n..];
+            }
+
+            return dst.len - dst_left.len;
+        }
+
+        /// Same as `read` except it returns an error union
+        /// The purpose of this function existing is to match `std.io.Reader` API.
+        fn readFn(self: *Self, dest: []u8) error{}!usize {
+            return self.read(dest);
+        }
+
+        pub fn reader(self: *Self) Reader {
+            return .{ .context = self };
+        }
+
+        /// Returns number of items available in fifo
+        pub fn writableLength(self: Self) usize {
+            return self.buf.len - self.count;
+        }
+
+        /// Returns the first section of writable buffer.
+        /// Note that this may be of length 0
+        pub fn writableSlice(self: SliceSelfArg, offset: usize) []T {
+            if (offset > self.buf.len) return &[_]T{};
+
+            const tail = self.head + offset + self.count;
+            if (tail < self.buf.len) {
+                return self.buf[tail..];
+            } else {
+                return self.buf[tail - self.buf.len ..][0 .. self.writableLength() - offset];
+            }
+        }
+
+        /// Returns a writable buffer of at least `size` items, allocating memory as needed.
+        /// Use `fifo.update` once you've written data to it.
+        pub fn writableWithSize(self: *Self, size: usize) ![]T {
+            try self.ensureUnusedCapacity(size);
+
+            // try to avoid realigning buffer
+            var slice = self.writableSlice(0);
+            if (slice.len < size) {
+                self.realign();
+                slice = self.writableSlice(0);
+            }
+            return slice;
+        }
+
+        /// Update the tail location of the buffer (usually follows use of writable/writableWithSize)
+        pub fn update(self: *Self, count: usize) void {
+            assert(self.count + count <= self.buf.len);
+            self.count += count;
+        }
+
+        /// Appends the data in `src` to the fifo.
+        /// You must have ensured there is enough space.
+        pub fn writeAssumeCapacity(self: *Self, src: []const T) void {
+            assert(self.writableLength() >= src.len);
+
+            var src_left = src;
+            while (src_left.len > 0) {
+                const writable_slice = self.writableSlice(0);
+                assert(writable_slice.len != 0);
+                const n = @min(writable_slice.len, src_left.len);
+                @memcpy(writable_slice[0..n], src_left[0..n]);
+                self.update(n);
+                src_left = src_left[n..];
+            }
+        }
+
+        /// Write a single item to the fifo
+        pub fn writeItem(self: *Self, item: T) !void {
+            try self.ensureUnusedCapacity(1);
+            return self.writeItemAssumeCapacity(item);
+        }
+
+        pub fn writeItemAssumeCapacity(self: *Self, item: T) void {
+            var tail = self.head + self.count;
+            if (powers_of_two) {
+                tail &= self.buf.len - 1;
+            } else {
+                tail %= self.buf.len;
+            }
+            self.buf[tail] = item;
+            self.update(1);
+        }
+
+        /// Appends the data in `src` to the fifo.
+        /// Allocates more memory as necessary
+        pub fn write(self: *Self, src: []const T) !void {
+            try self.ensureUnusedCapacity(src.len);
+
+            return self.writeAssumeCapacity(src);
+        }
+
+        /// Same as `write` except it returns the number of bytes written, which is always the same
+        /// as `bytes.len`. The purpose of this function existing is to match `std.io.Writer` API.
+        fn appendWrite(self: *Self, bytes: []const u8) error{OutOfMemory}!usize {
+            try self.write(bytes);
+            return bytes.len;
+        }
+
+        pub fn writer(self: *Self) Writer {
+            return .{ .context = self };
+        }
+
+        /// Make `count` items available before the current read location
+        fn rewind(self: *Self, count: usize) void {
+            assert(self.writableLength() >= count);
+
+            var head = self.head + (self.buf.len - count);
+            if (powers_of_two) {
+                head &= self.buf.len - 1;
+            } else {
+                head %= self.buf.len;
+            }
+            self.head = head;
+            self.count += count;
+        }
+
+        /// Place data back into the read stream
+        pub fn unget(self: *Self, src: []const T) !void {
+            try self.ensureUnusedCapacity(src.len);
+
+            self.rewind(src.len);
+
+            const slice = self.readableSliceMut(0);
+            if (src.len < slice.len) {
+                @memcpy(slice[0..src.len], src);
+            } else {
+                @memcpy(slice, src[0..slice.len]);
+                const slice2 = self.readableSliceMut(slice.len);
+                @memcpy(slice2[0 .. src.len - slice.len], src[slice.len..]);
+            }
+        }
+
+        /// Returns the item at `offset`.
+        /// Asserts offset is within bounds.
+        pub fn peekItem(self: Self, offset: usize) T {
+            assert(offset < self.count);
+
+            var index = self.head + offset;
+            if (powers_of_two) {
+                index &= self.buf.len - 1;
+            } else {
+                index %= self.buf.len;
+            }
+            return self.buf[index];
+        }
+
+        /// Pump data from a reader into a writer.
+        /// Stops when reader returns 0 bytes (EOF).
+        /// Buffer size must be set before calling; a buffer length of 0 is invalid.
+        pub fn pump(self: *Self, src_reader: anytype, dest_writer: anytype) !void {
+            assert(self.buf.len > 0);
+            while (true) {
+                if (self.writableLength() > 0) {
+                    const n = try src_reader.read(self.writableSlice(0));
+                    if (n == 0) break; // EOF
+                    self.update(n);
+                }
+                self.discard(try dest_writer.write(self.readableSlice(0)));
+            }
+            // flush remaining data
+            while (self.readableLength() > 0) {
+                self.discard(try dest_writer.write(self.readableSlice(0)));
+            }
+        }
+
+        pub fn toOwnedSlice(self: *Self) Allocator.Error![]T {
+            if (self.head != 0) self.realign();
+            assert(self.head == 0);
+            assert(self.count <= self.buf.len);
+            const allocator = self.allocator;
+            if (allocator.resize(self.buf, self.count)) {
+                const result = self.buf[0..self.count];
+                self.* = Self.init(allocator);
+                return result;
+            }
+            const new_memory = try allocator.dupe(T, self.buf[0..self.count]);
+            allocator.free(self.buf);
+            self.* = Self.init(allocator);
+            return new_memory;
+        }
+    };
+}
+
+test "LinearFifo(u8, .Dynamic) discard(0) from empty buffer should not error on overflow" {
+    var fifo = LinearFifo(u8, .dynamic).init(testing.allocator);
+    defer fifo.deinit();
+
+    // If overflow is not explicitly allowed this will crash in debug / safe mode
+    fifo.discard(0);
+}
+
+test "LinearFifo(u8, .Dynamic)" {
+    var fifo = LinearFifo(u8, .dynamic).init(testing.allocator);
+    defer fifo.deinit();
+
+    try fifo.write("HELLO");
+    try testing.expectEqual(@as(usize, 5), fifo.readableLength());
+    try testing.expectEqualSlices(u8, "HELLO", fifo.readableSlice(0));
+
+    {
+        var i: usize = 0;
+        while (i < 5) : (i += 1) {
+            try fifo.write(&[_]u8{fifo.peekItem(i)});
+        }
+        try testing.expectEqual(@as(usize, 10), fifo.readableLength());
+        try testing.expectEqualSlices(u8, "HELLOHELLO", fifo.readableSlice(0));
+    }
+
+    {
+        try testing.expectEqual(@as(u8, 'H'), fifo.readItem().?);
+        try testing.expectEqual(@as(u8, 'E'), fifo.readItem().?);
+        try testing.expectEqual(@as(u8, 'L'), fifo.readItem().?);
+        try testing.expectEqual(@as(u8, 'L'), fifo.readItem().?);
+        try testing.expectEqual(@as(u8, 'O'), fifo.readItem().?);
+    }
+    try testing.expectEqual(@as(usize, 5), fifo.readableLength());
+
+    { // Writes that wrap around
+        try testing.expectEqual(@as(usize, 11), fifo.writableLength());
+        try testing.expectEqual(@as(usize, 6), fifo.writableSlice(0).len);
+        fifo.writeAssumeCapacity("6<chars<11");
+        try testing.expectEqualSlices(u8, "HELLO6<char", fifo.readableSlice(0));
+        try testing.expectEqualSlices(u8, "s<11", fifo.readableSlice(11));
+        try testing.expectEqualSlices(u8, "11", fifo.readableSlice(13));
+        try testing.expectEqualSlices(u8, "", fifo.readableSlice(15));
+        fifo.discard(11);
+        try testing.expectEqualSlices(u8, "s<11", fifo.readableSlice(0));
+        fifo.discard(4);
+        try testing.expectEqual(@as(usize, 0), fifo.readableLength());
+    }
+
+    {
+        const buf = try fifo.writableWithSize(12);
+        try testing.expectEqual(@as(usize, 12), buf.len);
+        var i: u8 = 0;
+        while (i < 10) : (i += 1) {
+            buf[i] = i + 'a';
+        }
+        fifo.update(10);
+        try testing.expectEqualSlices(u8, "abcdefghij", fifo.readableSlice(0));
+    }
+
+    {
+        try fifo.unget("prependedstring");
+        var result: [30]u8 = undefined;
+        try testing.expectEqualSlices(u8, "prependedstringabcdefghij", result[0..fifo.read(&result)]);
+        try fifo.unget("b");
+        try fifo.unget("a");
+        try testing.expectEqualSlices(u8, "ab", result[0..fifo.read(&result)]);
+    }
+
+    fifo.shrink(0);
+
+    {
+        try fifo.writer().print("{s}, {s}!", .{ "Hello", "World" });
+        var result: [30]u8 = undefined;
+        try testing.expectEqualSlices(u8, "Hello, World!", result[0..fifo.read(&result)]);
+        try testing.expectEqual(@as(usize, 0), fifo.readableLength());
+    }
+
+    {
+        try fifo.writer().writeAll("This is a test");
+        var result: [30]u8 = undefined;
+        try testing.expectEqualSlices(u8, "This", (try fifo.reader().readUntilDelimiterOrEof(&result, ' ')).?);
+        try testing.expectEqualSlices(u8, "is", (try fifo.reader().readUntilDelimiterOrEof(&result, ' ')).?);
+        try testing.expectEqualSlices(u8, "a", (try fifo.reader().readUntilDelimiterOrEof(&result, ' ')).?);
+        try testing.expectEqualSlices(u8, "test", (try fifo.reader().readUntilDelimiterOrEof(&result, ' ')).?);
+    }
+
+    {
+        try fifo.ensureTotalCapacity(1);
+        var in_fbs = std.io.fixedBufferStream("pump test");
+        var out_buf: [50]u8 = undefined;
+        var out_fbs = std.io.fixedBufferStream(&out_buf);
+        try fifo.pump(in_fbs.reader(), out_fbs.writer());
+        try testing.expectEqualSlices(u8, in_fbs.buffer, out_fbs.getWritten());
+    }
+}
+
+test LinearFifo {
+    inline for ([_]type{ u1, u8, u16, u64 }) |T| {
+        inline for ([_]LinearFifoBufferType{ LinearFifoBufferType{ .static = 32 }, .slice, .dynamic }) |bt| {
+            const FifoType = LinearFifo(T, bt);
+            var buf: if (bt == .slice) [32]T else void = undefined;
+            var fifo = switch (bt) {
+                .static => FifoType.init(),
+                .slice => FifoType.init(buf[0..]),
+                .dynamic => FifoType.init(testing.allocator),
+            };
+            defer fifo.deinit();
+
+            try fifo.write(&[_]T{ 0, 1, 1, 0, 1 });
+            try testing.expectEqual(@as(usize, 5), fifo.readableLength());
+
+            {
+                try testing.expectEqual(@as(T, 0), fifo.readItem().?);
+                try testing.expectEqual(@as(T, 1), fifo.readItem().?);
+                try testing.expectEqual(@as(T, 1), fifo.readItem().?);
+                try testing.expectEqual(@as(T, 0), fifo.readItem().?);
+                try testing.expectEqual(@as(T, 1), fifo.readItem().?);
+                try testing.expectEqual(@as(usize, 0), fifo.readableLength());
+            }
+
+            {
+                try fifo.writeItem(1);
+                try fifo.writeItem(1);
+                try fifo.writeItem(1);
+                try testing.expectEqual(@as(usize, 3), fifo.readableLength());
+            }
+
+            {
+                var readBuf: [3]T = undefined;
+                const n = fifo.read(&readBuf);
+                try testing.expectEqual(@as(usize, 3), n); // NOTE: It should be the number of items.
+            }
+        }
+    }
+}

src/main.zig

@@ -4,17 +4,21 @@ 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
@@ -22,6 +26,7 @@ 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.
+    \\ --gdb                Run ZBA from the context of a GDB Server
     \\<str>                 Path to the GBA GamePak ROM.
     \\
 );
@@ -29,7 +34,7 @@ const params = clap.parseParamsComptime(
 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();
 
@@ -56,7 +61,8 @@ pub fn main() void {
     defer allocator.free(config_path);
 
     // Parse CLI
-    const result = clap.parse(clap.Help, &params, clap.parsers.default, .{}) catch |e| exitln("failed to parse cli: {}", .{e});
+
+    const result = clap.parse(clap.Help, &params, clap.parsers.default, .{ .allocator = allocator }) catch |e| exitln("failed to parse cli: {}", .{e});
     defer result.deinit();
 
     // TODO: Move config file to XDG Config directory?
@@ -65,12 +71,29 @@ pub fn main() void {
 
     config.load(allocator, cfg_file_path) catch |e| exitln("failed to load config file: {}", .{e});
 
-    const paths = handleArguments(allocator, data_path, &result) catch |e| exitln("failed to handle cli arguments: {}", .{e});
-    defer if (paths.save) |path| allocator.free(path);
+    var paths = handleArguments(allocator, data_path, &result) catch |e| exitln("failed to handle cli arguments: {}", .{e});
+    defer paths.deinit(allocator);
 
-    const log_file = if (config.config().debug.cpu_trace) blk: {
-        break :blk std.fs.cwd().createFile("zba.log", .{}) catch |e| exitln("failed to create trace log file: {}", .{e});
-    } else null;
+    // 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
@@ -78,29 +101,80 @@ pub fn main() void {
     defer scheduler.deinit();
 
     var bus: Bus = undefined;
-    var cpu = Arm7tdmi.init(&scheduler, &bus, log_file);
+
+    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();
 
-    if (config.config().guest.skip_bios or result.args.skip or paths.bios == null) {
-        cpu.fastBoot();
+    if (config.config().guest.skip_bios or result.args.skip != 0 or paths.bios == null) {
+        @import("core/cpu_util.zig").fastBoot(&cpu);
     }
 
-    var gui = Gui.init(&bus.pak.title, &bus.apu, width, height) catch |e| exitln("failed to init gui: {}", .{e});
+    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();
 
-    gui.run(&cpu, &scheduler) catch |e| exitln("failed to run gui thread: {}", .{e});
+    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();
+        defer emulator.deinit(allocator);
+
+        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.warn("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 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 .{
@@ -123,7 +197,9 @@ fn configFilePath(allocator: Allocator, config_path: []const u8) ![]const u8 {
         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"));
+        // FIXME(2025-09-22): re-enable
+        // try config_file.writeAll(@embedFile("example.toml"));
+        try config_file.writeAll("");
     };
 
     return path;
@@ -144,17 +220,19 @@ fn ensureConfigDirExists(config_path: []const u8) !void {
     try dir.makePath("zba");
 }
 
-fn romPath(result: *const clap.Result(clap.Help, &params, clap.parsers.default)) []const u8 {
+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 => exitln("ZBA requires a path to a GamePak ROM", .{}),
+        0 => null,
+        1 => if (result.positionals[0]) |path| try allocator.dupe(u8, path) else null,
         else => exitln("ZBA received too many positional arguments.", .{}),
     };
 }
 
 fn exitln(comptime format: []const u8, args: anytype) noreturn {
-    const stderr = std.io.getStdErr().writer();
+    var buf: [1024]u8 = undefined;
+    var stderr = std.fs.File.stderr().writer(&buf).interface;
+
     stderr.print(format, args) catch {}; // Just exit already...
     stderr.writeByte('\n') catch {};
-    std.os.exit(1);
+    std.process.exit(1);
 }

src/platform.zig

@@ -1,262 +1,259 @@
 const std = @import("std");
-const SDL = @import("sdl2");
 const gl = @import("gl");
+const zgui = @import("zgui");
+const c = @import("lib.zig").c;
+
 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 gba_width = @import("core/ppu.zig").width;
 const gba_height = @import("core/ppu.zig").height;
 
-pub const sample_rate = 1 << 15;
-pub const sample_format = SDL.AUDIO_U16;
+const GLsizei = gl.sizei;
+const SDL_GLContext = *c.SDL_GLContextState;
+const Allocator = std.mem.Allocator;
 
-const default_title = "ZBA";
+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";
+
+const errify = @import("lib.zig").errify;
+
+var gl_procs: gl.ProcTable = undefined;
 
 pub const Gui = struct {
     const Self = @This();
-    const SDL_GLContext = *anyopaque; // SDL.SDL_GLContext is a ?*anyopaque
     const log = std.log.scoped(.Gui);
 
-    // zig fmt: off
-    const vertices: [32]f32 = [_]f32{
-        // Positions        // Colours      // Texture Coords
-         1.0, -1.0, 0.0,    1.0, 0.0, 0.0,  1.0, 1.0, // Top Right
-         1.0,  1.0, 0.0,    0.0, 1.0, 0.0,  1.0, 0.0, // Bottom Right
-        -1.0,  1.0, 0.0,    0.0, 0.0, 1.0,  0.0, 0.0, // Bottom Left
-        -1.0, -1.0, 0.0,    1.0, 1.0, 0.0,  0.0, 1.0, // Top Left
-    };
-
-    const indices: [6]u32 = [_]u32{
-        0, 1, 3, // First Triangle
-        1, 2, 3, // Second Triangle
-    };
-    // zig fmt: on
-
-    window: *SDL.SDL_Window,
+    window: *c.SDL_Window,
     ctx: SDL_GLContext,
-    title: []const u8,
     audio: Audio,
 
-    program_id: gl.GLuint,
+    state: imgui.State,
+    allocator: Allocator,
 
-    pub fn init(title: *const [12]u8, apu: *Apu, width: i32, height: i32) !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();
+    pub fn init(allocator: Allocator, apu: *Apu, title_opt: ?*const [12]u8) !Self {
+        c.SDL_SetMainReady();
 
-        const win_scale = @intCast(c_int, config.config().host.win_scale);
+        try errify(c.SDL_Init(c.SDL_INIT_VIDEO | c.SDL_INIT_AUDIO | c.SDL_INIT_EVENTS));
 
-        const window = SDL.SDL_CreateWindow(
-            default_title,
-            SDL.SDL_WINDOWPOS_CENTERED,
-            SDL.SDL_WINDOWPOS_CENTERED,
-            @as(c_int, width * win_scale),
-            @as(c_int, height * win_scale),
-            SDL.SDL_WINDOW_OPENGL | SDL.SDL_WINDOW_SHOWN,
-        ) orelse panic();
+        try errify(c.SDL_SetAppMetadata(window_title, "0.1.0", "moe.paoda.zba"));
+        try errify(c.SDL_GL_SetAttribute(c.SDL_GL_CONTEXT_MAJOR_VERSION, gl.info.version_major));
+        try errify(c.SDL_GL_SetAttribute(c.SDL_GL_CONTEXT_MINOR_VERSION, gl.info.version_minor));
+        try errify(c.SDL_GL_SetAttribute(c.SDL_GL_CONTEXT_PROFILE_MASK, c.SDL_GL_CONTEXT_PROFILE_CORE));
+        try errify(c.SDL_GL_SetAttribute(c.SDL_GL_CONTEXT_FLAGS, c.SDL_GL_CONTEXT_FORWARD_COMPATIBLE_FLAG));
 
-        const ctx = SDL.SDL_GL_CreateContext(window) orelse panic();
-        if (SDL.SDL_GL_MakeCurrent(window, ctx) < 0) panic();
+        const window: *c.SDL_Window = try errify(c.SDL_CreateWindow(window_title, default_dim.width, default_dim.height, c.SDL_WINDOW_OPENGL | c.SDL_WINDOW_RESIZABLE));
+        errdefer c.SDL_DestroyWindow(window);
 
-        gl.load(ctx, Self.glGetProcAddress) catch {};
-        if (SDL.SDL_GL_SetSwapInterval(@boolToInt(config.config().host.vsync)) < 0) panic();
+        const gl_ctx = try errify(c.SDL_GL_CreateContext(window));
+        errdefer errify(c.SDL_GL_DestroyContext(gl_ctx)) catch {};
 
-        const program_id = try compileShaders();
+        try errify(c.SDL_GL_MakeCurrent(window, gl_ctx));
+        errdefer errify(c.SDL_GL_MakeCurrent(window, null)) catch {};
+
+        if (!gl_procs.init(c.SDL_GL_GetProcAddress)) return error.gl_init_failed;
+
+        gl.makeProcTableCurrent(&gl_procs);
+        errdefer gl.makeProcTableCurrent(null);
+
+        try errify(c.SDL_GL_SetSwapInterval(@intFromBool(config.config().host.vsync)));
+
+        zgui.init(allocator);
+        zgui.plot.init();
+        zgui.backend.init(window, gl_ctx);
+
+        // zgui.io.setIniFilename(null);
 
         return Self{
             .window = window,
-            .title = std.mem.sliceTo(title, 0),
-            .ctx = ctx,
-            .program_id = program_id,
-            .audio = Audio.init(apu),
+            .ctx = gl_ctx,
+            .audio = try Audio.init(apu),
+
+            .allocator = allocator,
+            .state = try imgui.State.init(allocator, title_opt),
         };
     }
 
-    fn compileShaders() !gl.GLuint {
-        // TODO: Panic on Shader Compiler Failure + Error Message
-        const vert_shader = @embedFile("shader/pixelbuf.vert");
-        const frag_shader = @embedFile("shader/pixelbuf.frag");
+    pub fn deinit(self: *Self) void {
+        self.audio.deinit();
+        self.state.deinit(self.allocator);
 
-        const vs = gl.createShader(gl.VERTEX_SHADER);
-        defer gl.deleteShader(vs);
+        zgui.backend.deinit();
+        zgui.plot.deinit();
+        zgui.deinit();
 
-        gl.shaderSource(vs, 1, &[_][*c]const u8{vert_shader}, 0);
-        gl.compileShader(vs);
+        errify(c.SDL_GL_DestroyContext(self.ctx)) catch {};
+        c.SDL_DestroyWindow(self.window);
+        c.SDL_Quit();
 
-        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 program = gl.createProgram();
-        gl.attachShader(program, vs);
-        gl.attachShader(program, fs);
-        gl.linkProgram(program);
-
-        return program;
+        self.* = undefined;
     }
 
-    // Returns the VAO ID since it's used in run()
-    fn generateBuffers() struct { c_uint, c_uint, c_uint } {
-        var vao_id: c_uint = undefined;
-        var vbo_id: c_uint = undefined;
-        var ebo_id: c_uint = undefined;
-        gl.genVertexArrays(1, &vao_id);
-        gl.genBuffers(1, &vbo_id);
-        gl.genBuffers(1, &ebo_id);
+    const RunOptions = struct {
+        sync: *Synchro,
+        tracker: ?*FpsTracker = null,
+        cpu: *Arm7tdmi,
+        scheduler: *Scheduler,
+    };
 
-        gl.bindVertexArray(vao_id);
+    pub fn run(self: *Self, opt: RunOptions) !void {
+        const cpu = opt.cpu;
+        const tracker = opt.tracker;
+        const sync = opt.sync;
 
-        gl.bindBuffer(gl.ARRAY_BUFFER, vbo_id);
-        gl.bufferData(gl.ARRAY_BUFFER, @sizeOf(@TypeOf(vertices)), &vertices, gl.STATIC_DRAW);
+        const bus_ptr: *Bus = @ptrCast(@alignCast(cpu.bus.ptr));
 
-        gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, ebo_id);
-        gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, @sizeOf(@TypeOf(indices)), &indices, gl.STATIC_DRAW);
+        const vao_id = opengl_impl.vao();
+        defer gl.DeleteVertexArrays(1, vao_id[0..]);
 
-        // Position
-        gl.vertexAttribPointer(0, 3, gl.FLOAT, gl.FALSE, 8 * @sizeOf(f32), @intToPtr(?*anyopaque, 0)); // lmao
-        gl.enableVertexAttribArray(0);
-        // Colour
-        gl.vertexAttribPointer(1, 3, gl.FLOAT, gl.FALSE, 8 * @sizeOf(f32), @intToPtr(?*anyopaque, (3 * @sizeOf(f32))));
-        gl.enableVertexAttribArray(1);
-        // Texture Coord
-        gl.vertexAttribPointer(2, 2, gl.FLOAT, gl.FALSE, 8 * @sizeOf(f32), @intToPtr(?*anyopaque, (6 * @sizeOf(f32))));
-        gl.enableVertexAttribArray(2);
+        const emu_tex = opengl_impl.screenTex(bus_ptr.ppu.framebuf.get(.Renderer));
+        defer gl.DeleteTextures(1, emu_tex[0..]);
 
-        return .{ vao_id, vbo_id, ebo_id };
-    }
+        const out_tex = opengl_impl.outTex();
+        defer gl.DeleteTextures(1, out_tex[0..]);
 
-    fn generateTexture(buf: []const u8) c_uint {
-        var tex_id: c_uint = undefined;
-        gl.genTextures(1, &tex_id);
-        gl.bindTexture(gl.TEXTURE_2D, tex_id);
+        const fbo_id = try opengl_impl.frameBuffer(out_tex[0]);
+        defer gl.DeleteFramebuffers(1, fbo_id[0..]);
 
-        // gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_WRAP_S, gl.CLAMP_TO_EDGE);
-        // gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_WRAP_T, gl.CLAMP_TO_EDGE);
+        const prog_id = try opengl_impl.program(); // Dynamic Shaders?
+        defer gl.DeleteProgram(prog_id);
 
-        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);
-        // gl.generateMipmap(gl.TEXTURE_2D); // TODO: Remove?
-
-        return tex_id;
-    }
-
-    pub fn run(self: *Self, cpu: *Arm7tdmi, scheduler: *Scheduler) !void {
-        var quit = std.atomic.Atomic(bool).init(false);
-        var tracker = FpsTracker.init();
-
-        var buffer_ids = Self.generateBuffers();
-        defer {
-            gl.deleteBuffers(1, &buffer_ids[2]); // EBO
-            gl.deleteBuffers(1, &buffer_ids[1]); // VBO
-            gl.deleteVertexArrays(1, &buffer_ids[0]); // VAO
-        }
-        const vao_id = buffer_ids[0];
-
-        const tex_id = Self.generateTexture(cpu.bus.ppu.framebuf.get(.Renderer));
-        defer gl.deleteTextures(1, &tex_id);
-
-        const thread = try std.Thread.spawn(.{}, emu.run, .{ &quit, scheduler, cpu, &tracker });
-        defer thread.join();
-
-        var title_buf: [0x100]u8 = undefined;
+        var win_dim: Dimensions = default_dim;
 
         emu_loop: while (true) {
-            var event: SDL.SDL_Event = undefined;
-            while (SDL.SDL_PollEvent(&event) != 0) {
+            // 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: c.SDL_Event = undefined;
+
+            while (c.SDL_PollEvent(&event)) {
+                _ = zgui.backend.processEvent(&event);
+
                 switch (event.type) {
-                    SDL.SDL_QUIT => break :emu_loop,
-                    SDL.SDL_KEYDOWN => {
-                        const key_code = event.key.keysym.sym;
-                        var keyinput = cpu.bus.io.keyinput.load(.Monotonic);
+                    c.SDL_EVENT_QUIT => break :emu_loop,
+                    c.SDL_EVENT_KEY_DOWN => {
+                        // TODO: Make use of compare_and_xor?
+                        var keyinput: KeyInput = .{ .raw = 0x0000 };
 
-                        switch (key_code) {
-                            SDL.SDLK_UP => keyinput.up.unset(),
-                            SDL.SDLK_DOWN => keyinput.down.unset(),
-                            SDL.SDLK_LEFT => keyinput.left.unset(),
-                            SDL.SDLK_RIGHT => keyinput.right.unset(),
-                            SDL.SDLK_x => keyinput.a.unset(),
-                            SDL.SDLK_z => keyinput.b.unset(),
-                            SDL.SDLK_a => keyinput.shoulder_l.unset(),
-                            SDL.SDLK_s => keyinput.shoulder_r.unset(),
-                            SDL.SDLK_RETURN => keyinput.start.unset(),
-                            SDL.SDLK_RSHIFT => keyinput.select.unset(),
+                        switch (event.key.scancode) {
+                            c.SDL_SCANCODE_UP => keyinput.up.write(true),
+                            c.SDL_SCANCODE_DOWN => keyinput.down.write(true),
+                            c.SDL_SCANCODE_LEFT => keyinput.left.write(true),
+                            c.SDL_SCANCODE_RIGHT => keyinput.right.write(true),
+                            c.SDL_SCANCODE_X => keyinput.a.write(true),
+                            c.SDL_SCANCODE_Z => keyinput.b.write(true),
+                            c.SDL_SCANCODE_A => keyinput.shoulder_l.write(true),
+                            c.SDL_SCANCODE_S => keyinput.shoulder_r.write(true),
+                            c.SDL_SCANCODE_RETURN => keyinput.start.write(true),
+                            c.SDL_SCANCODE_RSHIFT => keyinput.select.write(true),
                             else => {},
                         }
 
-                        cpu.bus.io.keyinput.store(keyinput.raw, .Monotonic);
+                        bus_ptr.io.keyinput.fetchAnd(~keyinput.raw, .monotonic);
                     },
-                    SDL.SDL_KEYUP => {
-                        const key_code = event.key.keysym.sym;
-                        var keyinput = cpu.bus.io.keyinput.load(.Monotonic);
+                    c.SDL_EVENT_KEY_UP => {
+                        // FIXME(paoda): merge with above?
+                        // TODO: Make use of compare_and_xor?
+                        var keyinput: KeyInput = .{ .raw = 0x0000 };
 
-                        switch (key_code) {
-                            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(),
-                            SDL.SDLK_i => {
-                                comptime std.debug.assert(sample_format == SDL.AUDIO_U16);
-                                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 => {},
+                        switch (event.key.scancode) {
+                            c.SDL_SCANCODE_UP => keyinput.up.write(true),
+                            c.SDL_SCANCODE_DOWN => keyinput.down.write(true),
+                            c.SDL_SCANCODE_LEFT => keyinput.left.write(true),
+                            c.SDL_SCANCODE_RIGHT => keyinput.right.write(true),
+                            c.SDL_SCANCODE_X => keyinput.a.write(true),
+                            c.SDL_SCANCODE_Z => keyinput.b.write(true),
+                            c.SDL_SCANCODE_A => keyinput.shoulder_l.write(true),
+                            c.SDL_SCANCODE_S => keyinput.shoulder_r.write(true),
+                            c.SDL_SCANCODE_RETURN => keyinput.start.write(true),
+                            c.SDL_SCANCODE_RSHIFT => keyinput.select.write(true),
                             else => {},
                         }
 
-                        cpu.bus.io.keyinput.store(keyinput.raw, .Monotonic);
+                        bus_ptr.io.keyinput.fetchOr(keyinput.raw, .monotonic);
+                    },
+                    c.SDL_EVENT_WINDOW_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);
-            gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gba_width, gba_height, gl.RGBA, gl.UNSIGNED_INT_8_8_8_8, framebuf.ptr);
+            var zgui_redraw: bool = false;
 
-            gl.useProgram(self.program_id);
-            gl.bindVertexArray(vao_id);
-            gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_INT, null);
-            SDL.SDL_GL_SwapWindow(self.window);
+            switch (self.state.emulation) {
+                .Transition => |inner| switch (inner) {
+                    .Active => {
+                        sync.paused.store(false, .monotonic);
+                        if (!config.config().host.mute) try errify(c.SDL_PauseAudioDevice(c.SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK));
 
-            const dyn_title = std.fmt.bufPrintZ(&title_buf, "ZBA | {s} [Emu: {}fps] ", .{ self.title, tracker.value() }) catch unreachable;
-            SDL.SDL_SetWindowTitle(self.window, dyn_title.ptr);
+                        self.state.emulation = .Active;
+                    },
+                    .Inactive => {
+                        // Assert that double pausing is impossible
+                        try errify(c.SDL_ResumeAudioDevice(c.SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK));
+                        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[0]);
+                        defer gl.BindFramebuffer(gl.FRAMEBUFFER, 0);
+
+                        gl.Viewport(0, 0, gba_width, gba_height);
+                        opengl_impl.drawScreen(emu_tex[0], prog_id, vao_id[0], 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?
+
+                    try errify(c.SDL_LockAudioStream(self.audio.stream));
+                    defer errify(c.SDL_UnlockAudioStream(self.audio.stream)) catch @panic("TODO: FIXME");
+
+                    zgui_redraw = imgui.draw(&self.state, sync, win_dim, cpu, out_tex[0]);
+                },
+                .Inactive => zgui_redraw = imgui.draw(&self.state, sync, win_dim, cpu, out_tex[0]),
+            }
+
+            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();
+            }
+
+            try errify(c.SDL_GL_SwapWindow(self.window));
         }
 
-        quit.store(true, .Monotonic); // Terminate Emulator Thread
-    }
-
-    pub fn deinit(self: *Self) void {
-        self.audio.deinit();
-        gl.deleteProgram(self.program_id);
-        SDL.SDL_GL_DeleteContext(self.ctx);
-        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);
+        sync.should_quit.store(true, .monotonic);
     }
 };
 
@@ -264,68 +261,146 @@ const Audio = struct {
     const Self = @This();
     const log = std.log.scoped(.PlatformAudio);
 
-    device: SDL.SDL_AudioDeviceID,
+    stream: *c.SDL_AudioStream,
 
-    fn init(apu: *Apu) Self {
-        var have: SDL.SDL_AudioSpec = undefined;
-        var want: SDL.SDL_AudioSpec = std.mem.zeroes(SDL.SDL_AudioSpec);
-        want.freq = sample_rate;
-        want.format = sample_format;
-        want.channels = 2;
-        want.samples = 0x100;
-        want.callback = Self.callback;
-        want.userdata = apu;
+    fn init(apu: *Apu) !Self {
+        var desired: c.SDL_AudioSpec = std.mem.zeroes(c.SDL_AudioSpec);
+        desired.freq = sample_rate;
+        desired.format = c.SDL_AUDIO_S16LE;
+        desired.channels = 2;
 
-        std.debug.assert(sample_format == SDL.AUDIO_U16);
-        log.info("Host Sample Rate: {}Hz, Host Format: SDL.AUDIO_U16", .{sample_rate});
+        log.info("Host Sample Rate: {}Hz, Host Format: SDL_AUDIO_S16LE", .{sample_rate});
 
-        const device = SDL.SDL_OpenAudioDevice(null, 0, &want, &have, 0);
-        if (device == 0) panic();
+        const stream = try errify(c.SDL_OpenAudioDeviceStream(c.SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK, &desired, null, null));
+        errdefer c.SDL_DestroyAudioStream(stream);
 
-        if (!config.config().host.mute) {
-            SDL.SDL_PauseAudioDevice(device, 0); // Unpause Audio
-            log.info("Unpaused Device", .{});
-        }
-
-        return .{ .device = device };
+        apu.stream = stream;
+        return .{ .stream = stream };
     }
 
     fn deinit(self: *Self) void {
-        SDL.SDL_CloseAudioDevice(self.device);
+        c.SDL_DestroyAudioStream(self.stream);
         self.* = undefined;
     }
-
-    export fn callback(userdata: ?*anyopaque, stream: [*c]u8, len: c_int) void {
-        const T = *Apu;
-        const apu = @ptrCast(T, @alignCast(@alignOf(T), userdata));
-
-        _ = SDL.SDL_AudioStreamGet(apu.stream, stream, len);
-    }
 };
 
-const shader = struct {
-    const Kind = enum { vertex, fragment };
-    const log = std.log.scoped(.Shader);
+const opengl_impl = struct {
+    fn drawScreen(tex_id: gl.uint, prog_id: gl.uint, vao_id: gl.uint, buf: []const u8) void {
+        gl.BindTexture(gl.TEXTURE_2D, tex_id);
+        defer gl.BindTexture(gl.TEXTURE_2D, 0);
 
-    fn didCompile(id: gl.GLuint) bool {
-        var success: gl.GLint = undefined;
-        gl.getShaderiv(id, gl.COMPILE_STATUS, &success);
+        gl.TexSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gba_width, gba_height, gl.RGBA, gl.UNSIGNED_INT_8_8_8_8, buf.ptr);
 
-        if (success == 0) err(id);
+        // Bind VAO
+        gl.BindVertexArray(vao_id);
+        defer gl.BindVertexArray(0);
 
-        return success == 1;
+        // Use compiled frag + vertex shader
+        gl.UseProgram(prog_id);
+        defer gl.UseProgram(0);
+
+        gl.DrawArrays(gl.TRIANGLE_STRIP, 0, 3);
     }
 
-    fn err(id: gl.GLuint) void {
-        const buf_len = 512;
-        var error_msg: [buf_len]u8 = undefined;
+    fn program() !gl.uint {
+        const vert_shader: [1][*]const u8 = .{@embedFile("shader/pixelbuf.vert")};
+        const frag_shader: [1][*]const u8 = .{@embedFile("shader/pixelbuf.frag")};
 
-        gl.getShaderInfoLog(id, buf_len, 0, &error_msg);
-        log.err("{s}", .{std.mem.sliceTo(&error_msg, 0)});
+        const vs = gl.CreateShader(gl.VERTEX_SHADER);
+        defer gl.DeleteShader(vs);
+
+        gl.ShaderSource(vs, 1, vert_shader[0..], null);
+        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, frag_shader[0..], null);
+        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() [1]gl.uint {
+        var vao_id: [1]gl.uint = undefined;
+        gl.GenVertexArrays(1, vao_id[0..]);
+
+        return vao_id;
+    }
+
+    fn screenTex(buf: []const u8) [1]gl.uint {
+        var tex_id: [1]gl.uint = undefined;
+        gl.GenTextures(1, tex_id[0..]);
+
+        gl.BindTexture(gl.TEXTURE_2D, tex_id[0]);
+        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() [1]gl.uint {
+        var tex_id: [1]gl.uint = undefined;
+        gl.GenTextures(1, tex_id[0..]);
+
+        gl.BindTexture(gl.TEXTURE_2D, tex_id[0]);
+        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: gl.uint) ![1]gl.uint {
+        var fbo_id: [1]gl.uint = undefined;
+        gl.GenFramebuffers(1, fbo_id[0..]);
+
+        gl.BindFramebuffer(gl.FRAMEBUFFER, fbo_id[0]);
+        defer gl.BindFramebuffer(gl.FRAMEBUFFER, 0);
+
+        gl.FramebufferTexture(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, tex_id, 0);
+        gl.DrawBuffers(1, &.{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.uint) bool {
+            var success: [1]gl.int = undefined;
+            gl.GetShaderiv(id, gl.COMPILE_STATUS, success[0..]);
+
+            if (success[0] == 0) err(id);
+
+            return success[0] == 1;
+        }
+
+        fn err(id: gl.uint) void {
+            const buf_len = 512;
+            var error_msg: [buf_len]u8 = undefined;
+
+            gl.GetShaderInfoLog(id, buf_len, null, &error_msg);
+            log.err("{s}", .{std.mem.sliceTo(&error_msg, 0)});
+        }
+    };
 };
-
-fn panic() noreturn {
-    const str = @as(?[*:0]const u8, SDL.SDL_GetError()) orelse "unknown error";
-    @panic(std.mem.sliceTo(str, 0));
-}

src/shader/pixelbuf.frag

@@ -1,7 +1,6 @@
 #version 330 core
 out vec4 frag_color;
 
-in vec3 color;
 in vec2 uv;
 
 uniform sampler2D screen;

src/shader/pixelbuf.vert

@@ -1,13 +1,10 @@
 #version 330 core
-layout (location = 0) in vec3 pos;
-layout (location = 1) in vec3 in_color;
-layout (location = 2) in vec2 in_uv;
-
-out vec3 color;
 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() {
-	color = in_color;
-	uv = in_uv;
-	gl_Position = vec4(pos, 1.0);
-}
+	uv = uvs[gl_VertexID];
+	gl_Position = vec4(pos[gl_VertexID], 0.0, 1.0);
+}

src/util.zig

@@ -3,53 +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;
 
 const Allocator = std.mem.Allocator;
 
-// 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_amt = @intCast(Log2Int(T), @typeInfo(T).Int.bits - @typeInfo(U).Int.bits);
-
-    return @bitCast(T, @bitCast(iT, @as(ExtU, @truncate(U, value)) << shift_amt) >> shift_amt);
-}
-
-/// 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);
-}
-
 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, .Monotonic);
+            self.fps = self.count.swap(0, .monotonic);
             self.timer.reset();
         }
 
@@ -57,17 +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;
-}
-
 /// Creates a copy of a title with all Filesystem-invalid characters replaced
 ///
 /// e.g. POKEPIN R/S to POKEPIN R_S
@@ -82,9 +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 {
@@ -95,7 +69,7 @@ pub const io = struct {
         }
 
         pub fn undef(comptime T: type, comptime log: anytype, comptime format: []const u8, args: anytype) ?T {
-            @setCold(true);
+            @branchHint(.cold);
 
             const unhandled_io = config.config().debug.unhandled_io;
 
@@ -106,7 +80,7 @@ pub const io = struct {
         }
 
         pub fn err(comptime T: type, comptime log: anytype, comptime format: []const u8, args: anytype) ?T {
-            @setCold(true);
+            @branchHint(.cold);
 
             log.err(format, args);
             return null;
@@ -235,7 +209,7 @@ pub const audio = struct {
 
 /// 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 = @truncate(u2, addr);
+    const offset: u2 = @truncate(addr);
 
     return switch (offset) {
         0b00 => (left & 0xFFFF_FF00) | right,
@@ -249,11 +223,11 @@ pub inline fn setQuart(left: u32, addr: u8, right: u8) u32 {
 ///
 /// TODO: Support u16 reads of u32 values?
 pub inline fn getHalf(byte: u8) u4 {
-    return @truncate(u4, byte & 1) << 3;
+    return @as(u4, @truncate(byte & 1)) << 3;
 }
 
 pub inline fn setHalf(comptime T: type, left: T, addr: u8, right: HalfInt(T)) T {
-    const offset = @truncate(u1, addr >> if (T == u32) 1 else 0);
+    const offset: u1 = @truncate(addr >> if (T == u32) 1 else 0);
 
     return switch (T) {
         u32 => switch (offset) {
@@ -271,10 +245,10 @@ pub inline fn setHalf(comptime T: type, left: T, addr: u8, right: HalfInt(T)) T
 /// The Integer type which corresponds to T with exactly half the amount of bits
 fn HalfInt(comptime T: type) type {
     const type_info = @typeInfo(T);
-    comptime std.debug.assert(type_info == .Int); // Type must be an integer
-    comptime std.debug.assert(type_info.Int.bits % 2 == 0); // Type must have an even amount of bits
+    comptime std.debug.assert(type_info == .int); // Type must be an integer
+    comptime std.debug.assert(type_info.int.bits % 2 == 0); // Type must have an even amount of bits
 
-    return std.meta.Int(type_info.Int.signedness, type_info.Int.bits >> 1);
+    return std.meta.Int(type_info.int.signedness, type_info.int.bits >> 1);
 }
 
 /// Double Buffering Implementation
@@ -283,7 +257,7 @@ pub const FrameBuffer = struct {
 
     layers: [2][]u8,
     buf: []u8,
-    current: u1,
+    current: u1 = 0,
 
     allocator: Allocator,
 
@@ -292,18 +266,22 @@ pub const FrameBuffer = struct {
 
     pub fn init(allocator: Allocator, comptime len: comptime_int) !Self {
         const buf = try allocator.alloc(u8, len * 2);
-        std.mem.set(u8, buf, 0);
+        @memset(buf, 0);
 
         return .{
             // Front and Back Framebuffers
             .layers = [_][]u8{ buf[0..][0..len], buf[len..][0..len] },
             .buf = buf,
-            .current = 0,
 
             .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;
@@ -317,3 +295,31 @@ pub const FrameBuffer = struct {
         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();
+        }
+    };
+}