ci: ensure that submodules are updated recursively

Reviewed-on: #6

.github/workflows/main.yml

@@ -14,7 +14,8 @@ jobs:
   build:
     strategy:
       matrix:
-        os: [ubuntu-latest, windows-latest, macos-latest]
+        # os: [ubuntu-latest, windows-latest, macos-latest]
+        os: [ubuntu-latest, windows-latest]
     runs-on: ${{matrix.os}}
     steps:
       - uses: goto-bus-stop/setup-zig@v2
@@ -37,9 +38,9 @@ jobs:
             brew install sdl2
       - uses: actions/checkout@v3
         with:
-          submodules: true
+          submodules: recursive
       - name: build 
-        run: zig build -Drelease-safe
+        run: zig build -Doptimize=ReleaseSafe -Dcpu=baseline
       - name: upload
         uses: actions/upload-artifact@v3
         with:
@@ -50,7 +51,7 @@ jobs:
     steps: 
       - uses: actions/checkout@v3
         with:
-          submodules: true
+          submodules: recursive
       - uses: goto-bus-stop/setup-zig@v2
         with:
           version: master

.gitignore

@@ -11,8 +11,4 @@
 /lib/SDL2
 
 # Any Custom Scripts for Debugging purposes
-*.sh
-
-
-# Dear ImGui
-**/imgui.ini
+*.sh

.gitmodules

@@ -13,6 +13,6 @@
 [submodule "lib/zig-toml"]
 	path = lib/zig-toml
 	url = https://github.com/aeronavery/zig-toml
-[submodule "lib/zgui"]
-	path = lib/zgui
-	url = https://git.musuka.dev/paoda/zgui
+[submodule "lib/zba-gdbstub"]
+	path = lib/zba-gdbstub
+	url = https://git.musuka.dev/paoda/zba-gdbstub

.vscode/extensions.json

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

README.md

@@ -13,7 +13,7 @@ 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)
-- [ ] Immediate Mode GUI
+- [x] Immediate Mode GUI (see [this branch](https://git.musuka.dev/paoda/zba/src/branch/imgui))
 - [ ] Refactoring for easy-ish perf boosts
 
 ## Usage
@@ -77,7 +77,7 @@ arm7wrestler GBA Fixed | [destoer](https://github.com/destoer)
 
 ## Compiling
 
-Most recently built on Zig [v0.11.0-dev.987+a1d82352d](https://github.com/ziglang/zig/tree/a1d82352d)
+Most recently built on Zig [v0.11.0-dev.1580+a5b34a61a](https://github.com/ziglang/zig/tree/a5b34a61a)
 
 ### Dependencies
 
@@ -91,17 +91,17 @@ 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)
 `gl.zig` | <https://github.com/MasterQ32/zig-opengl>
 
-Use `git submodule update --init` from the project root to pull the git submodules `SDL.zig`, `zig-clap`, `known-folders`, `zig-toml` and `zig-datetime`
+Use `git submodule update --init` from the project root to pull the git relevant git submodules
 
 Be sure to provide SDL2 using:
 
 - Linux: Your distro's package manager
-- MacOS: ¯\\\_(ツ)_/¯
+- 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 -Drelease-fast`. The executable is located at `zig-out/bin/`.
+Once you've got all the dependencies, execute `zig build -Doptimize=ReleaseSafe`. The executable is located at `zig-out/bin/`.
 
 ## Controls
 

build.zig

@@ -2,61 +2,54 @@ const std = @import("std");
 const builtin = @import("builtin");
 
 const Sdk = @import("lib/SDL.zig/Sdk.zig");
-const zgui = @import("lib/zgui/build.zig");
+const Gdbstub = @import("lib/zba-gdbstub/build.zig");
 
 pub fn build(b: *std.build.Builder) void {
     // Minimum Zig Version
-    const min_ver = std.SemanticVersion.parse("0.11.0-dev.987+a1d82352d") catch return; // https://github.com/ziglang/zig/commit/19056cb68
+    const min_ver = std.SemanticVersion.parse("0.11.0-dev.1580+a5b34a61a") catch return; // https://github.com/ziglang/zig/commit/a5b34a61a
     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);
     }
 
-    // Standard target options allows the person running `zig build` to choose
-    // what target to build for. Here we do not override the defaults, which
-    // means any target is allowed, and the default is native. Other options
-    // for restricting supported target set are available.
     const target = b.standardTargetOptions(.{});
+    const optimize = b.standardOptimizeOption(.{});
 
-    // Standard release options allow the person running `zig build` to select
-    // between Debug, ReleaseSafe, ReleaseFast, and ReleaseSmall.
-    const mode = b.standardReleaseOptions();
+    const exe = b.addExecutable(.{
+        .name = "zba",
+        .root_source_file = .{ .path = "src/main.zig" },
+        .target = target,
+        .optimize = optimize,
+    });
 
-    const exe = b.addExecutable("zba", "src/main.zig");
     exe.setMainPkgPath("."); // Necessary so that src/main.zig can embed example.toml
-    exe.setTarget(target);
 
     // Known Folders (%APPDATA%, XDG, etc.)
-    exe.addPackagePath("known_folders", "lib/known-folders/known-folders.zig");
+    exe.addAnonymousModule("known_folders", .{ .source_file = .{ .path = "lib/known-folders/known-folders.zig" } });
 
     // DateTime Library
-    exe.addPackagePath("datetime", "lib/zig-datetime/src/main.zig");
+    exe.addAnonymousModule("datetime", .{ .source_file = .{ .path = "lib/zig-datetime/src/main.zig" } });
 
     // Bitfield type from FlorenceOS: https://github.com/FlorenceOS/
-    // exe.addPackage(.{ .name = "bitfield", .path = .{ .path = "lib/util/bitfield.zig" } });
-    exe.addPackagePath("bitfield", "lib/util/bitfield.zig");
+    exe.addAnonymousModule("bitfield", .{ .source_file = .{ .path = "lib/util/bitfield.zig" } });
 
     // Argument Parsing Library
-    exe.addPackagePath("clap", "lib/zig-clap/clap.zig");
+    exe.addAnonymousModule("clap", .{ .source_file = .{ .path = "lib/zig-clap/clap.zig" } });
 
     // TOML Library
-    exe.addPackagePath("toml", "lib/zig-toml/src/toml.zig");
+    exe.addAnonymousModule("toml", .{ .source_file = .{ .path = "lib/zig-toml/src/toml.zig" } });
 
     // OpenGL 3.3 Bindings
-    exe.addPackagePath("gl", "lib/gl.zig");
+    exe.addAnonymousModule("gl", .{ .source_file = .{ .path = "lib/gl.zig" } });
+
+    // gdbstub
+    Gdbstub.link(exe);
 
     // Zig SDL Bindings: https://github.com/MasterQ32/SDL.zig
-    const sdk = Sdk.init(b);
+    const sdk = Sdk.init(b, null);
     sdk.link(exe, .dynamic);
-    exe.addPackage(sdk.getNativePackage("sdl2"));
+    exe.addModule("sdl2", sdk.getNativeModule());
 
-    // Dear ImGui bindings
-    const zgui_options = zgui.BuildOptionsStep.init(b, .{ .backend = .sdl2_opengl3 });
-    const zgui_pkg = zgui.getPkg(&.{zgui_options.getPkg()});
-    exe.addPackage(zgui_pkg);
-    zgui.link(exe, zgui_options);
-
-    exe.setBuildMode(mode);
     exe.install();
 
     const run_cmd = exe.run();
@@ -68,9 +61,11 @@ pub fn build(b: *std.build.Builder) void {
     const run_step = b.step("run", "Run the app");
     run_step.dependOn(&run_cmd.step);
 
-    const exe_tests = b.addTest("src/main.zig");
-    exe_tests.setTarget(target);
-    exe_tests.setBuildMode(mode);
+    const exe_tests = b.addTest(.{
+        .root_source_file = .{ .path = "src/main.zig" },
+        .target = target,
+        .optimize = optimize,
+    });
 
     const test_step = b.step("test", "Run unit tests");
     test_step.dependOn(&exe_tests.step);

src/core/Bus.zig

@@ -105,7 +105,7 @@ pub fn deinit(self: *Self) void {
 fn fillReadTable(self: *Self, table: *[table_len]?*const anyopaque) void {
     const vramMirror = @import("ppu/Vram.zig").mirror;
 
-    for (table) |*ptr, i| {
+    for (table, 0..) |*ptr, i| {
         const addr = @intCast(u32, page_size * i);
 
         ptr.* = switch (addr) {
@@ -132,7 +132,7 @@ fn fillWriteTable(self: *Self, comptime T: type, table: *[table_len]?*const anyo
     comptime std.debug.assert(T == u32 or T == u16 or T == u8);
     const vramMirror = @import("ppu/Vram.zig").mirror;
 
-    for (table) |*ptr, i| {
+    for (table, 0..) |*ptr, i| {
         const addr = @intCast(u32, page_size * i);
 
         ptr.* = switch (addr) {
@@ -198,55 +198,6 @@ 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);
 }
@@ -336,6 +287,27 @@ pub fn read(self: *Self, comptime T: type, unaligned_address: u32) T {
     return self.slowRead(T, unaligned_address);
 }
 
+pub fn dbgRead(self: *const Self, comptime T: type, unaligned_address: u32) T {
+    const bits = @typeInfo(std.math.IntFittingRange(0, page_size - 1)).Int.bits;
+    const page = unaligned_address >> bits;
+    const offset = unaligned_address & (page_size - 1);
+
+    // We're doing some serious out-of-bounds open-bus reads
+    if (page >= table_len) return self.openBus(T, unaligned_address);
+
+    if (self.read_table[page]) |some_ptr| {
+        // We have a pointer to a page, cast the pointer to it's underlying type
+        const Ptr = [*]const T;
+        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 slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
     @setCold(true);
 
@@ -366,6 +338,34 @@ 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 = @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 readBackup(self: *const Self, comptime T: type, unaligned_address: u32) T {
     const value = self.pak.backup.read(unaligned_address);
 
@@ -406,6 +406,31 @@ pub fn write(self: *Self, comptime T: type, unaligned_address: u32, value: T) vo
     }
 }
 
+/// Mostly Identical to `Bus.write`, slowmeme is handled by `Bus.dbgSlowWrite`
+pub fn dbgWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    const bits = @typeInfo(std.math.IntFittingRange(0, page_size - 1)).Int.bits;
+    const page = unaligned_address >> bits;
+    const offset = unaligned_address & (page_size - 1);
+
+    // We're doing some serious out-of-bounds open-bus writes, they do nothing though
+    if (page >= table_len) return;
+
+    if (self.write_tables[@boolToInt(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));
+
+        // 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;
+
+        self.dbgSlowWrite(T, unaligned_address, value);
+    }
+}
+
 fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
     @setCold(true);
 
@@ -431,6 +456,31 @@ fn slowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) vo
     }
 }
 
+fn dbgSlowWrite(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {
+    @setCold(true);
+
+    const page = @truncate(u8, unaligned_address >> 24);
+    const address = forceAlign(T, unaligned_address);
+
+    switch (page) {
+        // General Internal Memory
+        0x00 => self.bios.write(T, address, value),
+        0x02 => unreachable, // completely handled by fastmem
+        0x03 => unreachable, // completely handled by fastmem
+        0x04 => return, // FIXME: Let debug writes mess with I/O
+
+        // Internal Display Memory
+        0x05 => self.ppu.palette.write(T, address, value),
+        0x06 => self.ppu.vram.write(T, self.ppu.dispcnt, address, value),
+        0x07 => unreachable, // completely handled by fastmem
+
+        // External Memory (Game Pak)
+        0x08...0x0D => return, // FIXME: Debug Write to Backup/GPIO w/out messing with state
+        0x0E...0x0F => return, // FIXME: Debug Write to Backup w/out messing with state
+        else => {},
+    }
+}
+
 inline fn rotateBy(comptime T: type, address: u32) u32 {
     return switch (T) {
         u32 => address & 3,

src/core/bus/GamePak.zig

@@ -213,6 +213,7 @@ 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;

src/core/bus/backup.zig

@@ -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;

src/core/bus/dma.zig

@@ -339,10 +339,7 @@ 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

@@ -293,13 +293,13 @@ pub const Clock = struct {
         self.cpu.sched.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(u8, now.date.year - 2000));
+        self.month = @truncate(u5, bcd(now.date.month));
+        self.day = @truncate(u6, bcd(now.date.day));
+        self.weekday = @truncate(u3, bcd((now.date.weekday() + 1) % 7)); // API is Monday = 0, Sunday = 6. We want Sunday = 0, Saturday = 6
+        self.hour = @truncate(u6, bcd(now.time.hour));
+        self.minute = @truncate(u7, bcd(now.time.minute));
+        self.second = @truncate(u7, bcd(now.time.second));
     }
 
     fn step(self: *Self, value: Data) u4 {
@@ -449,16 +449,8 @@ 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

@@ -346,10 +346,10 @@ const InterruptEnable = extern union {
     vblank: Bit(u16, 0),
     hblank: Bit(u16, 1),
     coincidence: Bit(u16, 2),
-    tim0: Bit(u16, 3),
-    tim1: Bit(u16, 4),
-    tim2: Bit(u16, 5),
-    tim3: Bit(u16, 6),
+    tm0_overflow: Bit(u16, 3),
+    tm1_overflow: Bit(u16, 4),
+    tm2_overflow: Bit(u16, 5),
+    tm3_overflow: Bit(u16, 6),
     serial: Bit(u16, 7),
     dma0: Bit(u16, 8),
     dma1: Bit(u16, 9),

src/core/cpu.zig

@@ -39,13 +39,12 @@ pub const arm = struct {
     }
 
     fn populate() [0x1000]InstrFn {
-        return comptime {
+        comptime {
             @setEvalBranchQuota(0xE000);
-            var ret = [_]InstrFn{und} ** 0x1000;
+            var table = [_]InstrFn{und} ** 0x1000;
 
-            var i: usize = 0;
-            while (i < ret.len) : (i += 1) {
-                ret[i] = switch (@as(u2, i >> 10)) {
+            for (&table, 0..) |*handler, i| {
+                handler.* = switch (@as(u2, i >> 10)) {
                     0b00 => if (i == 0x121) blk: {
                         break :blk branchExchange;
                     } else if (i & 0xFCF == 0x009) blk: {
@@ -107,8 +106,8 @@ pub const arm = struct {
                 };
             }
 
-            return ret;
-        };
+            return table;
+        }
     }
 };
 
@@ -136,13 +135,12 @@ pub const thumb = struct {
     }
 
     fn populate() [0x400]InstrFn {
-        return comptime {
+        comptime {
             @setEvalBranchQuota(5025); // This is exact
-            var ret = [_]InstrFn{und} ** 0x400;
+            var table = [_]InstrFn{und} ** 0x400;
 
-            var i: usize = 0;
-            while (i < ret.len) : (i += 1) {
-                ret[i] = switch (@as(u3, i >> 7 & 0x7)) {
+            for (&table, 0..) |*handler, i| {
+                handler.* = 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;
@@ -230,8 +228,8 @@ pub const thumb = struct {
                 };
             }
 
-            return ret;
-        };
+            return table;
+        }
     }
 };
 
@@ -385,8 +383,7 @@ pub const Arm7tdmi = struct {
         const now = getModeChecked(self, self.cpsr.mode.read());
 
         // Bank R8 -> r12
-        var i: usize = 0;
-        while (i < 5) : (i += 1) {
+        for (0..5) |i| {
             self.bank.fiq[Bank.fiqIdx(i, now)] = self.r[8 + i];
         }
 
@@ -404,8 +401,7 @@ pub const Arm7tdmi = struct {
         }
 
         // Grab R8 -> R12
-        i = 0;
-        while (i < 5) : (i += 1) {
+        for (0..5) |i| {
             self.r[8 + i] = self.bank.fiq[Bank.fiqIdx(i, next)];
         }
 
@@ -470,8 +466,7 @@ pub const Arm7tdmi = struct {
     }
 
     pub fn stepDmaTransfer(self: *Self) bool {
-        comptime var i: usize = 0;
-        inline while (i < 4) : (i += 1) {
+        inline for (0..4) |i| {
             if (self.bus.dma[i].in_progress) {
                 self.bus.dma[i].step(self);
                 return true;
@@ -647,7 +642,7 @@ pub const PSR = extern union {
     }
 };
 
-pub const Mode = enum(u5) {
+const Mode = enum(u5) {
     User = 0b10000,
     Fiq = 0b10001,
     Irq = 0b10010,
@@ -656,7 +651,7 @@ pub const Mode = enum(u5) {
     Undefined = 0b11011,
     System = 0b11111,
 
-    pub fn toString(self: Mode) []const u8 {
+    fn toString(self: Mode) []const u8 {
         return switch (self) {
             .User => "usr",
             .Fiq => "fiq",

src/core/cpu/arm/single_data_transfer.zig

@@ -11,9 +11,7 @@ pub fn singleDataTransfer(comptime I: bool, comptime P: bool, comptime U: bool,
             const rn = opcode >> 16 & 0xF;
             const rd = opcode >> 12 & 0xF;
 
-            // rn is r15 and L is not set, the PC is 12 ahead
-            const base = cpu.r[rn] + if (!L and rn == 0xF) 4 else @as(u32, 0);
-
+            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;

src/core/cpu/thumb/block_data_transfer.zig

@@ -92,8 +92,7 @@ pub fn fmt15(comptime L: bool, comptime rb: u3) InstrFn {
 inline fn countRlist(opcode: u16) u32 {
     var count: u32 = 0;
 
-    comptime var i: u4 = 0;
-    inline while (i < 8) : (i += 1) {
+    inline for (0..8) |i| {
         if (opcode >> (7 - i) & 1 == 1) count += 1;
     }
 

src/core/emu.zig

@@ -4,7 +4,7 @@ const config = @import("../config.zig");
 
 const Scheduler = @import("scheduler.zig").Scheduler;
 const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
-const Tracker = @import("../util.zig").FpsTracker;
+const FpsTracker = @import("../util.zig").FpsTracker;
 
 const Timer = std.time.Timer;
 const Atomic = std.atomic.Atomic;
@@ -35,18 +35,18 @@ const RunKind = enum {
     LimitedFPS,
 };
 
-pub fn run(quit: *Atomic(bool), pause: *Atomic(bool), cpu: *Arm7tdmi, scheduler: *Scheduler, tracker: *Tracker) void {
+pub fn run(quit: *Atomic(bool), scheduler: *Scheduler, cpu: *Arm7tdmi, tracker: *FpsTracker) 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, pause, cpu, scheduler, tracker);
+        inner(.LimitedFPS, audio_sync, quit, scheduler, cpu, tracker);
     } else {
-        inner(.UnlimitedFPS, audio_sync, quit, pause, cpu, scheduler, tracker);
+        inner(.UnlimitedFPS, audio_sync, quit, scheduler, cpu, tracker);
     }
 }
 
-fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), pause: *Atomic(bool), cpu: *Arm7tdmi, scheduler: *Scheduler, tracker: ?*Tracker) void {
+fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), scheduler: *Scheduler, cpu: *Arm7tdmi, tracker: ?*FpsTracker) void {
     if (kind == .UnlimitedFPS or kind == .LimitedFPS) {
         std.debug.assert(tracker != null);
         log.info("FPS tracking enabled", .{});
@@ -57,8 +57,6 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), pause: *
             log.info("Emulation w/out video sync", .{});
 
             while (!quit.load(.Monotonic)) {
-                if (pause.load(.Monotonic)) continue;
-
                 runFrame(scheduler, cpu);
                 audioSync(audio_sync, cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
 
@@ -71,8 +69,6 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), pause: *
             var wake_time: u64 = frame_period;
 
             while (!quit.load(.Monotonic)) {
-                if (pause.load(.Monotonic)) continue;
-
                 runFrame(scheduler, cpu);
                 const new_wake_time = videoSync(&timer, wake_time);
 
@@ -150,9 +146,8 @@ 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) {
+    for (0..times) |_| {
         std.time.sleep(step);
 
         // Upon wakeup, check to see if this particular sleep was longer than expected
@@ -167,3 +162,59 @@ fn sleep(timer: *Timer, wake_time: u64) ?u64 {
 fn spinLoop(timer: *Timer, wake_time: u64) void {
     while (true) if (timer.read() > wake_time) break;
 }
+
+pub const EmuThing = struct {
+    const Self = @This();
+    const Interface = @import("gdbstub").Emulator;
+    const Allocator = std.mem.Allocator;
+
+    cpu: *Arm7tdmi,
+    scheduler: *Scheduler,
+
+    pub fn init(cpu: *Arm7tdmi, scheduler: *Scheduler) Self {
+        return .{ .cpu = cpu, .scheduler = scheduler };
+    }
+
+    pub fn interface(self: *Self, allocator: Allocator) Interface {
+        return Interface.init(allocator, self);
+    }
+
+    pub fn read(self: *const Self, addr: u32) u8 {
+        return self.cpu.bus.dbgRead(u8, addr);
+    }
+
+    pub fn write(self: *Self, addr: u32, value: u8) void {
+        self.cpu.bus.dbgWrite(u8, addr, value);
+    }
+
+    pub fn registers(self: *const Self) *[16]u32 {
+        return &self.cpu.r;
+    }
+
+    pub fn cpsr(self: *const Self) u32 {
+        return self.cpu.cpsr.raw;
+    }
+
+    pub fn step(self: *Self) void {
+        const cpu = self.cpu;
+        const sched = self.scheduler;
+
+        // Is true when we have executed one (1) instruction
+        var did_step: bool = false;
+
+        // TODO: How can I make it easier to keep this in lock-step with runFrame?
+        while (!did_step) {
+            if (!cpu.stepDmaTransfer()) {
+                if (cpu.isHalted()) {
+                    // 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);
+        }
+    }
+};

src/core/ppu.zig

@@ -625,8 +625,7 @@ 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) {
+                for (0..4) |layer| {
                     self.drawSprites(@truncate(u2, 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);
@@ -640,8 +639,7 @@ 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) {
+                for (0..4) |layer| {
                     self.drawSprites(@truncate(u2, 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);
@@ -654,8 +652,7 @@ 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) {
+                for (0..4) |layer| {
                     self.drawSprites(@truncate(u2, 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);
@@ -671,7 +668,7 @@ pub const Ppu = struct {
                 const vram_buf = @ptrCast([*]const u16, @alignCast(@alignOf(u16), self.vram.buf));
                 const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), 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);
                 }
             },
@@ -685,7 +682,7 @@ pub const Ppu = struct {
                 const pal_buf = @ptrCast([*]const u16, @alignCast(@alignOf(u16), self.palette.buf));
                 const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), 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]);
                 }
             },
@@ -701,8 +698,7 @@ pub const Ppu = struct {
                 const vram_buf = @ptrCast([*]const u16, @alignCast(@alignOf(u16), self.vram.buf));
                 const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), 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);
                 }
@@ -718,7 +714,7 @@ pub const Ppu = struct {
         // FIXME: @ptrCast between slices changing the length isn't implemented yet
         const framebuf = @ptrCast([*]u32, @alignCast(@alignOf(u32), 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);
@@ -843,7 +839,12 @@ pub const Ppu = struct {
                 const is_top_layer = (top_layer >> layer) & 1 == 1;
 
                 if (is_top_layer) {
-                    self.scanline.btm()[i] = Scanline.Pixel.from(.Background, bgr555); // this is intentional
+                    const pixel = self.scanline.btm()[i];
+
+                    // FIXME: Can't I do this check ealier? Test Amazing Mirror File Select, bld_demo.gba
+                    if (!pixel.isSet())
+                        self.scanline.btm()[i] = Scanline.Pixel.from(.Background, bgr555); // this is intentional
+
                     return;
                 }
             },
@@ -985,8 +986,7 @@ pub const Ppu = struct {
             cpu.handleInterrupt();
         }
 
-        // 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);
 

src/core/scheduler.zig

@@ -73,7 +73,7 @@ pub const Scheduler = struct {
 
     /// 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

src/main.zig

@@ -4,15 +4,20 @@ const known_folders = @import("known_folders");
 const clap = @import("clap");
 
 const config = @import("config.zig");
+const emu = @import("core/emu.zig");
 
 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 Atomic = std.atomic.Atomic;
+
 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
@@ -20,6 +25,7 @@ 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.
     \\
 );
@@ -85,14 +91,49 @@ pub fn main() void {
         cpu.fastBoot();
     }
 
-    // TODO: Just copy the title instead of grabbing a pointer to it
-    var gui = Gui.init(allocator, &bus.pak.title, &bus.apu) catch |e| exitln("failed to init gui: {}", .{e});
+    var quit = Atomic(bool).init(false);
+    var gui = Gui.init(&bus.pak.title, &bus.apu, width, height) catch |e| exitln("failed to init gui: {}", .{e});
     defer gui.deinit();
 
-    gui.run(&cpu, &scheduler) catch |e| exitln("failed to run gui thread: {}", .{e});
+    if (result.args.gdb) {
+        const Server = @import("gdbstub").Server;
+        const EmuThing = @import("core/emu.zig").EmuThing;
+
+        var wrapper = EmuThing.init(&cpu, &scheduler);
+        var emulator = wrapper.interface(allocator);
+        defer emulator.deinit();
+
+        log.info("Ready to connect", .{});
+
+        var server = Server.init(emulator) 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, &quit }) catch |e| exitln("gdb server thread crashed: {}", .{e});
+        defer thread.join();
+
+        gui.run(.{
+            .cpu = &cpu,
+            .scheduler = &scheduler,
+            .quit = &quit,
+        }) catch |e| exitln("main thread panicked: {}", .{e});
+    } else {
+        var tracker = FpsTracker.init();
+
+        const thread = std.Thread.spawn(.{}, emu.run, .{ &quit, &scheduler, &cpu, &tracker }) catch |e| exitln("emu thread panicked: {}", .{e});
+        defer thread.join();
+
+        gui.run(.{
+            .cpu = &cpu,
+            .scheduler = &scheduler,
+            .tracker = &tracker,
+            .quit = &quit,
+        }) 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 {
+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});
 

src/platform.zig

@@ -1,8 +1,6 @@
 const std = @import("std");
 const SDL = @import("sdl2");
 const gl = @import("gl");
-const zgui = @import("zgui");
-
 const emu = @import("core/emu.zig");
 const config = @import("config.zig");
 
@@ -10,19 +8,10 @@ const Apu = @import("core/apu.zig").Apu;
 const Arm7tdmi = @import("core/cpu.zig").Arm7tdmi;
 const Scheduler = @import("core/scheduler.zig").Scheduler;
 const FpsTracker = @import("util.zig").FpsTracker;
-const RingBuffer = @import("util.zig").RingBuffer;
 
 const gba_width = @import("core/ppu.zig").width;
 const gba_height = @import("core/ppu.zig").height;
 
-const GLuint = gl.GLuint;
-const GLsizei = gl.GLsizei;
-const SDL_GLContext = *anyopaque;
-const Allocator = std.mem.Allocator;
-
-const width = 1280;
-const height = 720;
-
 pub const sample_rate = 1 << 15;
 pub const sample_format = SDL.AUDIO_U16;
 
@@ -30,27 +19,9 @@ const default_title = "ZBA";
 
 pub const Gui = struct {
     const Self = @This();
+    const SDL_GLContext = *anyopaque; // SDL.SDL_GLContext is a ?*anyopaque
     const log = std.log.scoped(.Gui);
 
-    const State = struct {
-        fps_hist: RingBuffer(u32),
-        allocator: Allocator,
-
-        pub fn init(allocator: Allocator) !@This() {
-            const history = try allocator.alloc(u32, 0x400);
-
-            return .{
-                .fps_hist = RingBuffer(u32).init(history),
-                .allocator = allocator,
-            };
-        }
-
-        fn deinit(self: *@This()) void {
-            self.fps_hist.deinit(self.allocator);
-            self.* = undefined;
-        }
-    };
-
     // zig fmt: off
     const vertices: [32]f32 = [_]f32{
         // Positions        // Colours      // Texture Coords
@@ -71,22 +42,22 @@ pub const Gui = struct {
     title: []const u8,
     audio: Audio,
 
-    state: State,
-
     program_id: gl.GLuint,
 
-    pub fn init(allocator: Allocator, title: *const [12]u8, apu: *Apu) !Self {
+    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();
 
+        const win_scale = @intCast(c_int, config.config().host.win_scale);
+
         const window = SDL.SDL_CreateWindow(
             default_title,
             SDL.SDL_WINDOWPOS_CENTERED,
             SDL.SDL_WINDOWPOS_CENTERED,
-            width,
-            height,
+            @as(c_int, width * win_scale),
+            @as(c_int, height * win_scale),
             SDL.SDL_WINDOW_OPENGL | SDL.SDL_WINDOW_SHOWN,
         ) orelse panic();
 
@@ -96,59 +67,19 @@ pub const Gui = struct {
         gl.load(ctx, Self.glGetProcAddress) catch {};
         if (SDL.SDL_GL_SetSwapInterval(@boolToInt(config.config().host.vsync)) < 0) panic();
 
-        zgui.init(allocator);
-        zgui.plot.init();
-        zgui.backend.init(window, ctx, "#version 330 core");
-
-        // zgui.io.setIniFilename(null);
+        const program_id = try compileShaders();
 
         return Self{
             .window = window,
             .title = std.mem.sliceTo(title, 0),
             .ctx = ctx,
-            .program_id = try compileShaders(),
+            .program_id = program_id,
             .audio = Audio.init(apu),
-
-            .state = try State.init(allocator),
         };
     }
 
-    pub fn deinit(self: *Self) void {
-        self.audio.deinit();
-        self.state.deinit();
-
-        zgui.backend.deinit();
-        zgui.plot.deinit();
-        zgui.deinit();
-
-        gl.deleteProgram(self.program_id);
-        SDL.SDL_GL_DeleteContext(self.ctx);
-        SDL.SDL_DestroyWindow(self.window);
-        SDL.SDL_Quit();
-        self.* = undefined;
-    }
-
-    fn drawGbaTexture(self: *const Self, obj_ids: struct { GLuint, GLuint, GLuint }, tex_id: GLuint, buf: []const u8) void {
-        gl.bindTexture(gl.TEXTURE_2D, tex_id);
-        defer gl.bindTexture(gl.TEXTURE_2D, 0);
-
-        gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gba_width, gba_height, gl.RGBA, gl.UNSIGNED_INT_8_8_8_8, buf.ptr);
-
-        // Bind VAO, EBO. VBO not bound
-        gl.bindVertexArray(obj_ids[0]); // VAO
-        defer gl.bindVertexArray(0);
-
-        gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, obj_ids[2]); // EBO
-        defer gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, 0);
-
-        // Use compiled frag + vertex shader
-        gl.useProgram(self.program_id);
-        defer gl.useProgram(0);
-
-        gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_INT, null);
-    }
-
-    fn compileShaders() !GLuint {
+    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");
 
@@ -177,25 +108,20 @@ pub const Gui = struct {
     }
 
     // Returns the VAO ID since it's used in run()
-    fn genBufferObjects() struct { GLuint, GLuint, GLuint } {
-        var vao_id: GLuint = undefined;
-        var vbo_id: GLuint = undefined;
-        var ebo_id: GLuint = undefined;
-
+    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);
 
         gl.bindVertexArray(vao_id);
-        defer gl.bindVertexArray(0);
 
         gl.bindBuffer(gl.ARRAY_BUFFER, vbo_id);
-        defer gl.bindBuffer(gl.ARRAY_BUFFER, 0);
+        gl.bufferData(gl.ARRAY_BUFFER, @sizeOf(@TypeOf(vertices)), &vertices, gl.STATIC_DRAW);
 
         gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, ebo_id);
-        defer gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, 0);
-
-        gl.bufferData(gl.ARRAY_BUFFER, @sizeOf(@TypeOf(vertices)), &vertices, gl.STATIC_DRAW);
         gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, @sizeOf(@TypeOf(indices)), &indices, gl.STATIC_DRAW);
 
         // Position
@@ -211,209 +137,56 @@ pub const Gui = struct {
         return .{ vao_id, vbo_id, ebo_id };
     }
 
-    fn genGbaTexture(buf: []const u8) GLuint {
-        var tex_id: GLuint = undefined;
+    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);
-        defer gl.bindTexture(gl.TEXTURE_2D, 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);
 
         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;
     }
 
-    fn genOutTexture() GLuint {
-        var tex_id: GLuint = undefined;
-        gl.genTextures(1, &tex_id);
+    const RunOptions = struct {
+        quit: *std.atomic.Atomic(bool),
+        tracker: ?*FpsTracker = null,
+        cpu: *Arm7tdmi,
+        scheduler: *Scheduler,
+    };
 
-        gl.bindTexture(gl.TEXTURE_2D, tex_id);
-        defer gl.bindTexture(gl.TEXTURE_2D, 0);
+    pub fn run(self: *Self, opt: RunOptions) !void {
+        const cpu = opt.cpu;
+        const tracker = opt.tracker;
+        const quit = opt.quit;
 
-        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 genFrameBufObject(tex_id: c_uint) !GLuint {
-        var fbo_id: GLuint = undefined;
-        gl.genFramebuffers(1, &fbo_id);
-
-        gl.bindFramebuffer(gl.FRAMEBUFFER, fbo_id);
-        defer gl.bindFramebuffer(gl.FRAMEBUFFER, 0);
-
-        gl.framebufferTexture(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, tex_id, 0);
-
-        const draw_buffers: [1]GLuint = .{gl.COLOR_ATTACHMENT0};
-        gl.drawBuffers(1, &draw_buffers);
-
-        if (gl.checkFramebufferStatus(gl.FRAMEBUFFER) != gl.FRAMEBUFFER_COMPLETE)
-            return error.FrameBufferObejctInitFailed;
-
-        return fbo_id;
-    }
-
-    fn draw(self: *Self, tex_id: GLuint, cpu: *const Arm7tdmi) void {
-        const win_scale = config.config().host.win_scale;
-
-        {
-            const w = @intToFloat(f32, gba_width * win_scale);
-            const h = @intToFloat(f32, gba_height * win_scale);
-
-            _ = zgui.begin("Game Boy Advance Screen", .{ .flags = .{ .no_resize = true, .always_auto_resize = true } });
-            defer zgui.end();
-
-            zgui.image(@intToPtr(*anyopaque, tex_id), .{ .w = w, .h = h, .uv0 = .{ 0, 1 }, .uv1 = .{ 1, 0 } });
+        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];
 
-        {
-            _ = zgui.begin("Information", .{});
-            defer zgui.end();
+        const tex_id = Self.generateTexture(cpu.bus.ppu.framebuf.get(.Renderer));
+        defer gl.deleteTextures(1, &tex_id);
 
-            {
-                var i: usize = 0;
-                while (i < 8) : (i += 1) {
-                    zgui.text("R{}: 0x{X:0>8}", .{ i, cpu.r[i] });
-
-                    zgui.sameLine(.{});
-
-                    const prefix = if (8 + i < 10) " " else "";
-                    zgui.text("{s}R{}: 0x{X:0>8}", .{ prefix, 8 + i, cpu.r[8 + i] });
-                }
-            }
-
-            zgui.separator();
-
-            widgets.psr("CPSR", cpu.cpsr);
-            widgets.psr("SPSR", cpu.spsr);
-
-            zgui.separator();
-
-            widgets.interrupts(" IE", cpu.bus.io.ie); // space is for padding
-            widgets.interrupts("IRQ", cpu.bus.io.irq);
-        }
-
-        {
-            _ = zgui.begin("Performance", .{});
-            defer zgui.end();
-
-            const tmp = blk: {
-                var buf: [0x400]u32 = undefined;
-                const len = self.state.fps_hist.copy(&buf);
-
-                break :blk .{ buf, len };
-            };
-            const values = tmp[0];
-            const len = tmp[1];
-
-            if (len == values.len) _ = self.state.fps_hist.pop();
-
-            const sorted = blk: {
-                var buf: @TypeOf(values) = undefined;
-
-                std.mem.copy(u32, buf[0..len], values[0..len]);
-                std.sort.sort(u32, buf[0..len], {}, std.sort.asc(u32));
-
-                break :blk buf;
-            };
-
-            const y_max = 2 * if (len != 0) @intToFloat(f64, sorted[len - 1]) else emu.frame_rate;
-            const x_max = @intToFloat(f64, 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 = average: {
-                    var sum: u32 = 0;
-                    for (sorted[0..len]) |value| sum += value;
-
-                    break :average @intCast(u32, 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]});
-        }
-
-        {
-            _ = zgui.begin("Scheduler", .{});
-            defer zgui.end();
-
-            const scheduler = cpu.sched;
-
-            zgui.text("tick: {X:0>16}", .{scheduler.tick});
-            zgui.separator();
-
-            const Event = std.meta.Child(@TypeOf(scheduler.queue.items));
-
-            var items: [20]Event = undefined;
-            const len = scheduler.queue.len;
-
-            std.mem.copy(Event, &items, scheduler.queue.items);
-            std.sort.sort(Event, items[0..len], {}, widgets.eventDesc(Event));
-
-            for (items[0..len]) |event| {
-                zgui.text("{X:0>16} | {?}", .{ event.tick, event.kind });
-            }
-        }
-
-        {
-            zgui.showDemoWindow(null);
-        }
-    }
-
-    pub fn run(self: *Self, cpu: *Arm7tdmi, scheduler: *Scheduler) !void {
-        const obj_ids = Self.genBufferObjects();
-        defer gl.deleteBuffers(3, @as(*const [3]c_uint, &obj_ids));
-
-        const emu_tex = Self.genGbaTexture(cpu.bus.ppu.framebuf.get(.Renderer));
-        const out_tex = Self.genOutTexture();
-        defer gl.deleteTextures(2, &[_]c_uint{ emu_tex, out_tex });
-
-        const fbo_id = try Self.genFrameBufObject(out_tex);
-        defer gl.deleteFramebuffers(1, &fbo_id);
-
-        var tracker = FpsTracker.init();
-        var quit = std.atomic.Atomic(bool).init(false);
-        var pause = std.atomic.Atomic(bool).init(false);
-
-        const thread = try std.Thread.spawn(.{}, emu.run, .{ &quit, &pause, cpu, scheduler, &tracker });
-        defer thread.join();
-        defer quit.store(true, .Monotonic); // Terminate Emulator Thread
+        var title_buf: [0x100]u8 = undefined;
 
         emu_loop: while (true) {
             var event: SDL.SDL_Event = undefined;
 
-            while (SDL.SDL_PollEvent(&event) != 0) {
-                _ = zgui.backend.processEvent(&event);
+            // This might be true if the emu is running via a gdbstub server
+            // and the gdb stub exits first
+            if (quit.load(.Monotonic)) break :emu_loop;
 
+            while (SDL.SDL_PollEvent(&event) != 0) {
                 switch (event.type) {
                     SDL.SDL_QUIT => break :emu_loop,
                     SDL.SDL_KEYDOWN => {
@@ -451,6 +224,12 @@ pub const Gui = struct {
                             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 => {},
                             else => {},
                         }
 
@@ -460,36 +239,31 @@ pub const Gui = struct {
                 }
             }
 
-            // We Access non-atomic parts of the Emulator here
-            {
-                pause.store(true, .Monotonic);
-                defer pause.store(false, .Monotonic);
-
-                self.state.fps_hist.push(tracker.value()) catch {};
-
-                // Draw GBA Screen to Texture
-                {
-                    gl.bindFramebuffer(gl.FRAMEBUFFER, fbo_id);
-                    defer gl.bindFramebuffer(gl.FRAMEBUFFER, 0);
-
-                    const buf = cpu.bus.ppu.framebuf.get(.Renderer);
-                    gl.viewport(0, 0, gba_width, gba_height);
-                    self.drawGbaTexture(obj_ids, emu_tex, buf);
-                }
-
-                // Background Colour
-                const size = zgui.io.getDisplaySize();
-                gl.viewport(0, 0, @floatToInt(c_int, size[0]), @floatToInt(c_int, size[1]));
-                gl.clearColor(0, 0, 0, 1.0);
-                gl.clear(gl.COLOR_BUFFER_BIT);
-
-                zgui.backend.newFrame(width, height);
-                self.draw(out_tex, cpu);
-                zgui.backend.draw();
-            }
+            // 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);
 
+            gl.useProgram(self.program_id);
+            gl.bindVertexArray(vao_id);
+            gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_INT, null);
             SDL.SDL_GL_SwapWindow(self.window);
+
+            if (tracker) |t| {
+                const dyn_title = std.fmt.bufPrintZ(&title_buf, "ZBA | {s} [Emu: {}fps] ", .{ self.title, t.value() }) catch unreachable;
+                SDL.SDL_SetWindowTitle(self.window, dyn_title.ptr);
+            }
         }
+
+        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 {
@@ -567,95 +341,3 @@ fn panic() noreturn {
     const str = @as(?[*:0]const u8, SDL.SDL_GetError()) orelse "unknown error";
     @panic(std.mem.sliceTo(str, 0));
 }
-
-const widgets = struct {
-    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("core/cpu.zig").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;
-    }
-};

src/util.zig

@@ -317,78 +317,3 @@ pub const FrameBuffer = struct {
         return self.layers[if (dev == .Emulator) self.current else ~self.current];
     }
 };
-
-pub fn RingBuffer(comptime T: type) type {
-    return struct {
-        const Self = @This();
-        const Index = usize;
-        const max_capacity = (@as(Index, 1) << @typeInfo(Index).Int.bits - 1) - 1; // half the range of index type
-
-        const log = std.log.scoped(.RingBuffer);
-
-        read: Index,
-        write: Index,
-        buf: []T,
-
-        const Error = error{buffer_full};
-
-        pub fn init(buf: []T) Self {
-            std.debug.assert(std.math.isPowerOfTwo(buf.len)); // capacity must be a power of two
-            std.debug.assert(buf.len <= max_capacity);
-
-            std.mem.set(T, buf, 0);
-
-            return .{ .read = 0, .write = 0, .buf = buf };
-        }
-
-        pub fn deinit(self: *Self, allocator: Allocator) void {
-            allocator.free(self.buf);
-            self.* = undefined;
-        }
-
-        pub fn push(self: *Self, value: T) Error!void {
-            if (self.isFull()) return error.buffer_full;
-            defer self.write += 1;
-
-            self.buf[self.mask(self.write)] = value;
-        }
-
-        pub fn pop(self: *Self) ?T {
-            if (self.isEmpty()) return null;
-            defer self.read += 1;
-
-            return self.buf[self.mask(self.read)];
-        }
-
-        /// Returns the number of entries read
-        pub fn copy(self: *const Self, cpy: []T) Index {
-            const count = std.math.min(self.len(), cpy.len);
-            var start: Index = self.read;
-
-            for (cpy) |*v, i| {
-                if (i >= count) break;
-
-                v.* = self.buf[self.mask(start)];
-                start += 1;
-            }
-
-            return count;
-        }
-
-        fn len(self: *const Self) Index {
-            return self.write - self.read;
-        }
-
-        fn isFull(self: *const Self) bool {
-            return self.len() == self.buf.len;
-        }
-
-        fn isEmpty(self: *const Self) bool {
-            return self.read == self.write;
-        }
-
-        fn mask(self: *const Self, idx: Index) Index {
-            return idx & (self.buf.len - 1);
-        }
-    };
-}