fix: 8-bit writes to WIN PPU registers

Advance Wars depends on these registers similar to Mario Kart's 8-bit
writes to Affine Background registers:

src/Gui.zig

@@ -180,11 +180,11 @@ const Audio = struct {
 
     export fn callback(userdata: ?*anyopaque, stream: [*c]u8, len: c_int) void {
         const apu = @ptrCast(*Apu, @alignCast(@alignOf(*Apu), userdata));
-        const written = SDL.SDL_AudioStreamGet(apu.stream, stream, len);
+        _ = SDL.SDL_AudioStreamGet(apu.stream, stream, len);
 
         // If we don't write anything, play silence otherwise garbage will be played
         // FIXME: I don't think this hack to remove DC Offset is acceptable :thinking:
-        if (written == 0) std.mem.set(u8, stream[0..@intCast(usize, len)], 0x40);
+        // if (written == 0) std.mem.set(u8, stream[0..@intCast(usize, len)], 0x40);
     }
 };
 

src/core/apu.zig

@@ -163,6 +163,8 @@ pub const Apu = struct {
     fs: FrameSequencer,
     capacitor: f32,
 
+    is_buffer_full: bool,
+
     pub fn init(sched: *Scheduler) Self {
         const apu: Self = .{
             .ch1 = ToneSweep.init(sched),
@@ -178,11 +180,12 @@ pub const Apu = struct {
             .bias = .{ .raw = 0x0200 },
 
             .sampling_cycle = 0b00,
-            .stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, 1 << 15, SDL.AUDIO_U16, 2, host_sample_rate) orelse unreachable,
+            .stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, 1 << 15, SDL.AUDIO_U16, 2, host_sample_rate).?,
             .sched = sched,
 
             .capacitor = 0,
             .fs = FrameSequencer.init(),
+            .is_buffer_full = false,
         };
 
         sched.push(.SampleAudio, apu.sampleTicks());
@@ -277,6 +280,13 @@ pub const Apu = struct {
     }
 
     pub fn sampleAudio(self: *Self, late: u64) void {
+        self.sched.push(.SampleAudio, self.sampleTicks() -| late);
+
+        // Whether the APU is busy or not is determined  by the main loop in emu.zig
+        // This should only ever be true (because this side of the emu is single threaded)
+        // When audio sync is disaabled
+        if (self.is_buffer_full) return;
+
         var left: i16 = 0;
         var right: i16 = 0;
 
@@ -325,28 +335,30 @@ pub const Apu = struct {
         left += bias;
         right += bias;
 
-        const tmp_left = std.math.clamp(@bitCast(u16, left), std.math.minInt(u11), std.math.maxInt(u11));
-        const tmp_right = std.math.clamp(@bitCast(u16, right), std.math.minInt(u11), std.math.maxInt(u11));
+        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));
 
         // Extend to 16-bit signed audio samples
-        const final_left = (tmp_left << 5) | (tmp_left >> 6);
-        const final_right = (tmp_right << 5) | (tmp_right >> 6);
+        const ext_left = (clamped_left << 5) | (clamped_left >> 6);
+        const ext_right = (clamped_right << 5) | (clamped_right >> 6);
 
-        if (self.sampling_cycle != self.bias.sampling_cycle.read()) {
-            const new_sample_rate = Self.sampleRate(self.bias.sampling_cycle.read());
-            log.info("Sample Rate changed from {}Hz to {}Hz", .{ Self.sampleRate(self.sampling_cycle), new_sample_rate });
+        // FIXME: This rarely happens
+        if (self.sampling_cycle != self.bias.sampling_cycle.read()) self.replaceSDLResampler();
 
-            // Sample Rate Changed, Create a new Resampler since i can't figure out how to change
-            // the parameters of the old one
-            const old = self.stream;
-            defer SDL.SDL_FreeAudioStream(old);
-
-            self.sampling_cycle = self.bias.sampling_cycle.read();
-            self.stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, @intCast(c_int, new_sample_rate), SDL.AUDIO_U16, 2, host_sample_rate) orelse unreachable;
+        _ = SDL.SDL_AudioStreamPut(self.stream, &[2]u16{ ext_left, ext_right }, 2 * @sizeOf(u16));
     }
 
-        _ = SDL.SDL_AudioStreamPut(self.stream, &[2]u16{ final_left, final_right }, 2 * @sizeOf(u16));
-        self.sched.push(.SampleAudio, self.sampleTicks() -| late);
+    fn replaceSDLResampler(self: *Self) void {
+        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 SDL.SDL_FreeAudioStream(old_stream);
+
+        self.sampling_cycle = self.bias.sampling_cycle.read();
+        self.stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, @intCast(c_int, sample_rate), SDL.AUDIO_U16, 2, host_sample_rate).?;
     }
 
     fn sampleTicks(self: *const Self) u64 {

src/core/bus/io.zig

@@ -302,6 +302,14 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
             0x0400_0009 => bus.ppu.bg[0].cnt.raw = (@as(u16, value) << 8) | (bus.ppu.bg[0].cnt.raw & 0xFF),
             0x0400_000A => bus.ppu.bg[1].cnt.raw = (bus.ppu.bg[1].cnt.raw & 0xFF00) | value,
             0x0400_000B => bus.ppu.bg[1].cnt.raw = (@as(u16, value) << 8) | (bus.ppu.bg[1].cnt.raw & 0xFF),
+            0x0400_0040 => bus.ppu.win.h[0].set(.Lo, value),
+            0x0400_0041 => bus.ppu.win.h[0].set(.Hi, value),
+            0x0400_0042 => bus.ppu.win.h[1].set(.Lo, value),
+            0x0400_0043 => bus.ppu.win.h[1].set(.Hi, value),
+            0x0400_0044 => bus.ppu.win.v[0].set(.Lo, value),
+            0x0400_0045 => bus.ppu.win.v[0].set(.Hi, value),
+            0x0400_0046 => bus.ppu.win.v[1].set(.Lo, value),
+            0x0400_0047 => bus.ppu.win.v[1].set(.Hi, value),
             0x0400_0048 => bus.ppu.win.setInL(value),
             0x0400_0049 => bus.ppu.win.setInH(value),
             0x0400_004A => bus.ppu.win.setOutL(value),
@@ -464,37 +472,57 @@ pub const BldY = extern union {
     raw: u16,
 };
 
+const u8WriteKind = enum { Hi, Lo };
+
 /// Write-only
 pub const WinH = extern union {
+    const Self = @This();
+
     x2: Bitfield(u16, 0, 8),
     x1: Bitfield(u16, 8, 8),
     raw: u16,
+
+    pub fn set(self: *Self, comptime K: u8WriteKind, value: u8) void {
+        self.raw = switch (K) {
+            .Hi => (@as(u16, value) << 8) | self.raw & 0xFF,
+            .Lo => (self.raw & 0xFF00) | value,
+        };
+    }
 };
 
 /// Write-only
 pub const WinV = extern union {
+    const Self = @This();
+
     y2: Bitfield(u16, 0, 8),
     y1: Bitfield(u16, 8, 8),
     raw: u16,
+
+    pub fn set(self: *Self, comptime K: u8WriteKind, value: u8) void {
+        self.raw = switch (K) {
+            .Hi => (@as(u16, value) << 8) | self.raw & 0xFF,
+            .Lo => (self.raw & 0xFF00) | value,
+        };
+    }
 };
 
 pub const WinIn = extern union {
     w0_bg: Bitfield(u16, 0, 4),
     w0_obj: Bit(u16, 4),
-    w0_colour: Bit(u16, 5),
+    w0_bld: Bit(u16, 5),
     w1_bg: Bitfield(u16, 8, 4),
     w1_obj: Bit(u16, 12),
-    w1_colour: Bit(u16, 13),
+    w1_bld: Bit(u16, 13),
     raw: u16,
 };
 
 pub const WinOut = extern union {
     out_bg: Bitfield(u16, 0, 4),
     out_obj: Bit(u16, 4),
-    out_colour: Bit(u16, 5),
+    out_bld: Bit(u16, 5),
     obj_bg: Bitfield(u16, 8, 4),
     obj_obj: Bit(u16, 12),
-    obj_colour: Bit(u16, 13),
+    obj_bld: Bit(u16, 13),
     raw: u16,
 };
 

src/core/cpu.zig

@@ -28,9 +28,14 @@ pub const arm = struct {
     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 = armIdx(opcode);
+        const id = idx(opcode);
         cpu.panic("[CPU/Decode] ID: 0x{X:0>3} 0x{X:0>8} is an illegal opcode", .{ id, opcode });
     }
 
@@ -120,9 +125,14 @@ pub const thumb = struct {
     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 = thumbIdx(opcode);
+        const id = idx(opcode);
         cpu.panic("[CPU/Decode] ID: 0b{b:0>10} 0x{X:0>2} is an illegal opcode", .{ id, opcode });
     }
 
@@ -420,13 +430,13 @@ pub const Arm7tdmi = struct {
             const opcode = self.fetch(u16);
             if (cpu_logging) self.logger.?.mgbaLog(self, opcode);
 
-            thumb.lut[thumbIdx(opcode)](self, self.bus, opcode);
+            thumb.lut[thumb.idx(opcode)](self, self.bus, opcode);
         } else {
             const opcode = self.fetch(u32);
             if (cpu_logging) self.logger.?.mgbaLog(self, opcode);
 
             if (checkCond(self.cpsr, @truncate(u4, opcode >> 28))) {
-                arm.lut[armIdx(opcode)](self, self.bus, opcode);
+                arm.lut[arm.idx(opcode)](self, self.bus, opcode);
             }
         }
     }
@@ -517,11 +527,11 @@ pub const Arm7tdmi = struct {
 
         if (self.cpsr.t.read()) {
             const opcode = self.bus.dbgRead(u16, self.r[15] - 4);
-            const id = thumbIdx(opcode);
+            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 = armIdx(opcode);
+            const id = arm.idx(opcode);
             std.debug.print("opcode: ID: 0x{X:0>3} 0x{X:0>8}\n", .{ id, opcode });
         }
 
@@ -601,14 +611,6 @@ pub const Arm7tdmi = struct {
     }
 };
 
-inline fn armIdx(opcode: u32) u12 {
-    return @truncate(u12, opcode >> 20 & 0xFF) << 4 | @truncate(u12, opcode >> 4 & 0xF);
-}
-
-inline fn thumbIdx(opcode: u16) u10 {
-    return @truncate(u10, opcode >> 6);
-}
-
 pub fn checkCond(cpsr: PSR, cond: u4) bool {
     return switch (cond) {
         0x0 => cpsr.z.read(), // EQ - Equal

src/core/emu.zig

@@ -70,12 +70,21 @@ pub fn runFrame(sched: *Scheduler, cpu: *Arm7tdmi) void {
     }
 }
 
-fn syncToAudio(cpu: *const Arm7tdmi) void {
-    const stream = cpu.bus.apu.stream;
-    const min_sample_count = 0x800;
+fn syncToAudio(stream: *SDL.SDL_AudioStream, is_buffer_full: *bool) void {
+    const sample_size = 2 * @sizeOf(u16);
+    const max_buf_size: c_int = 0x400;
 
-    // Busy Loop while we wait for the Audio system to catch up
-    while (SDL.SDL_AudioStreamAvailable(stream) > (@sizeOf(u16) * 2) * min_sample_count) {}
+    // 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;
+
+    while (true) {
+        still_full = SDL.SDL_AudioStreamAvailable(stream) > sample_size * max_buf_size >> 1;
+        if (!sync_audio or !still_full) break;
+    }
 }
 
 pub fn runUnsynchronized(quit: *Atomic(bool), sched: *Scheduler, cpu: *Arm7tdmi, fps: ?*FpsTracker) void {
@@ -86,21 +95,21 @@ pub fn runUnsynchronized(quit: *Atomic(bool), sched: *Scheduler, cpu: *Arm7tdmi,
 
         while (!quit.load(.SeqCst)) {
             runFrame(sched, cpu);
-            if (sync_audio) syncToAudio(cpu);
+            syncToAudio(cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
 
             tracker.tick();
         }
     } else {
         while (!quit.load(.SeqCst)) {
             runFrame(sched, cpu);
-            if (sync_audio) syncToAudio(cpu);
+            syncToAudio(cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
         }
     }
 }
 
 pub fn runSynchronized(quit: *Atomic(bool), sched: *Scheduler, cpu: *Arm7tdmi, fps: ?*FpsTracker) void {
     log.info("Emulation thread w/ video sync", .{});
-    var timer = Timer.start() catch unreachable;
+    var timer = Timer.start() catch std.debug.panic("Failed to initialize std.timer.Timer", .{});
     var wake_time: u64 = frame_period;
 
     if (fps) |tracker| {
@@ -108,13 +117,14 @@ pub fn runSynchronized(quit: *Atomic(bool), sched: *Scheduler, cpu: *Arm7tdmi, f
 
         while (!quit.load(.SeqCst)) {
             runFrame(sched, cpu);
-            const new_wake_time = syncToVideo(&timer, wake_time);
+            const new_wake_time = blockOnVideo(&timer, wake_time);
 
             // Spin to make up the difference of OS scheduler innacuracies
             // If we happen to also be syncing to audio, we choose to spin on
             // the amount of time needed for audio to catch up rather than
             // our expected wake-up time
-            if (sync_audio) syncToAudio(cpu) else spinLoop(&timer, wake_time);
+            syncToAudio(cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
+            if (!sync_audio) spinLoop(&timer, wake_time);
             wake_time = new_wake_time;
 
             tracker.tick();
@@ -122,16 +132,17 @@ pub fn runSynchronized(quit: *Atomic(bool), sched: *Scheduler, cpu: *Arm7tdmi, f
     } else {
         while (!quit.load(.SeqCst)) {
             runFrame(sched, cpu);
-            const new_wake_time = syncToVideo(&timer, wake_time);
-            // see above comment
-            if (sync_audio) syncToAudio(cpu) else spinLoop(&timer, wake_time);
+            const new_wake_time = blockOnVideo(&timer, wake_time);
 
+            // see above comment
+            syncToAudio(cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
+            if (!sync_audio) spinLoop(&timer, wake_time);
             wake_time = new_wake_time;
         }
     }
 }
 
-inline fn syncToVideo(timer: *Timer, wake_time: u64) u64 {
+inline fn blockOnVideo(timer: *Timer, wake_time: u64) u64 {
     // Use the OS scheduler to put the emulation thread to sleep
     const maybe_recalc_wake_time = sleep(timer, wake_time);
 
@@ -149,6 +160,8 @@ pub fn runBusyLoop(quit: *Atomic(bool), sched: *Scheduler, cpu: *Arm7tdmi) void
         runFrame(sched, cpu);
         spinLoop(&timer, wake_time);
 
+        syncToAudio(cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
+
         // Update to the new wake time
         wake_time += frame_period;
     }

src/core/ppu.zig

@@ -277,16 +277,17 @@ pub const Ppu = struct {
             aff_x += self.aff_bg[n - 2].pa;
             aff_y += self.aff_bg[n - 2].pc;
 
-            if (!shouldDrawBackground(n, self.bldcnt, &self.scanline, i)) continue;
+            const x = @bitCast(u32, ix);
+            const y = @bitCast(u32, iy);
+
+            const win_bounds = self.windowBounds(@truncate(u9, x), @truncate(u8, y));
+            if (!shouldDrawBackground(self, n, win_bounds, i)) continue;
 
             if (self.bg[n].cnt.display_overflow.read()) {
                 ix = if (ix > px_width) @rem(ix, px_width) else if (ix < 0) px_width + @rem(ix, px_width) else ix;
                 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 tile_id: u32 = self.vram.read(u8, screen_base + ((y / 8) * @bitCast(u32, tile_width) + (x / 8)));
             const row = y & 7;
             const col = x & 7;
@@ -296,7 +297,7 @@ pub const Ppu = struct {
 
             if (pal_id != 0) {
                 const bgr555 = self.palette.read(u16, pal_id * 2);
-                copyToBackgroundBuffer(n, self.bldcnt, &self.scanline, i, bgr555);
+                self.copyToBackgroundBuffer(n, win_bounds, i, bgr555);
             }
         }
 
@@ -305,7 +306,7 @@ pub const Ppu = struct {
         self.aff_bg[n - 2].y_latch.? += self.aff_bg[n - 2].pd; // PD is added to BGxY
     }
 
-    fn drawBackround(self: *Self, comptime n: u2) void {
+    fn drawBackground(self: *Self, comptime n: u2) void {
         // A Tile in a charblock is a byte, while a Screen Entry is a halfword
 
         const char_base = 0x4000 * @as(u32, self.bg[n].cnt.char_base.read());
@@ -325,10 +326,11 @@ pub const Ppu = struct {
 
         var i: u32 = 0;
         while (i < width) : (i += 1) {
-            if (!shouldDrawBackground(n, self.bldcnt, &self.scanline, i)) continue;
-
             const x = hofs + i;
 
+            const win_bounds = self.windowBounds(@truncate(u9, x), @truncate(u8, 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));
@@ -353,7 +355,7 @@ pub const Ppu = struct {
 
             if (pal_id != 0) {
                 const bgr555 = self.palette.read(u16, pal_id * 2);
-                copyToBackgroundBuffer(n, self.bldcnt, &self.scanline, i, bgr555);
+                self.copyToBackgroundBuffer(n, win_bounds, i, bgr555);
             }
         }
     }
@@ -379,10 +381,10 @@ pub const Ppu = struct {
                 var layer: usize = 0;
                 while (layer < 4) : (layer += 1) {
                     self.drawSprites(@truncate(u2, layer));
-                    if (layer == self.bg[0].cnt.priority.read() and bg_enable & 1 == 1) self.drawBackround(0);
-                    if (layer == self.bg[1].cnt.priority.read() and bg_enable >> 1 & 1 == 1) self.drawBackround(1);
-                    if (layer == self.bg[2].cnt.priority.read() and bg_enable >> 2 & 1 == 1) self.drawBackround(2);
-                    if (layer == self.bg[3].cnt.priority.read() and bg_enable >> 3 & 1 == 1) self.drawBackround(3);
+                    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);
+                    if (layer == self.bg[3].cnt.priority.read() and bg_enable >> 3 & 1 == 1) self.drawBackground(3);
                 }
 
                 // Copy Drawn Scanline to Frame Buffer
@@ -407,8 +409,8 @@ pub const Ppu = struct {
                 var layer: usize = 0;
                 while (layer < 4) : (layer += 1) {
                     self.drawSprites(@truncate(u2, layer));
-                    if (layer == self.bg[0].cnt.priority.read() and bg_enable & 1 == 1) self.drawBackround(0);
-                    if (layer == self.bg[1].cnt.priority.read() and bg_enable >> 1 & 1 == 1) self.drawBackround(1);
+                    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);
                 }
 
@@ -533,6 +535,93 @@ pub const Ppu = struct {
         return self.palette.getBackdrop();
     }
 
+    fn copyToBackgroundBuffer(self: *Self, comptime n: u2, bounds: ?WindowBounds, i: usize, bgr555: u16) void {
+        if (self.bldcnt.mode.read() != 0b00) {
+            // Standard Alpha Blending
+            const a_layers = self.bldcnt.layer_a.read();
+            const is_blend_enabled = (a_layers >> n) & 1 == 1;
+
+            // If Alpha Blending is enabled and we've found an eligible layer for
+            // Pixel A, store the pixel in the bottom pixel buffer
+
+            const win_part = if (bounds) |win| blk: {
+                // Window Enabled
+                break :blk switch (win) {
+                    .win0 => self.win.in.w0_bld.read(),
+                    .win1 => self.win.in.w1_bld.read(),
+                    .out => self.win.out.out_bld.read(),
+                };
+            } else true;
+
+            if (win_part and is_blend_enabled) {
+                self.scanline.btm()[i] = bgr555;
+                return;
+            }
+        }
+
+        self.scanline.top()[i] = bgr555;
+    }
+
+    const WindowBounds = enum { win0, win1, out };
+
+    fn windowBounds(self: *Self, x: u9, y: u8) ?WindowBounds {
+        const win0 = self.dispcnt.win_enable.read() & 1 == 1;
+        const win1 = (self.dispcnt.win_enable.read() >> 1) & 1 == 1;
+        const winObj = self.dispcnt.obj_win_enable.read();
+
+        if (!(win0 or win1 or winObj)) return null;
+
+        if (win0 and self.win.inRange(0, x, y)) return .win0;
+        if (win1 and self.win.inRange(1, x, y)) return .win1;
+
+        return .out;
+    }
+
+    fn shouldDrawBackground(self: *Self, comptime n: u2, bounds: ?WindowBounds, i: usize) bool {
+        // If a pixel has been drawn on the top layer, it's because:
+        // 1. The pixel is to be blended with a pixel on the bottom layer
+        // 2. The pixel is not to be blended at all
+        // Also, if we find a pixel on the top layer we don't need to bother with this I think?
+        if (self.scanline.top()[i] != null) return false;
+
+        if (bounds) |win| {
+            switch (win) {
+                .win0 => if ((self.win.in.w0_bg.read() >> n) & 1 == 0) return false,
+                .win1 => if ((self.win.in.w1_bg.read() >> n) & 1 == 0) return false,
+                .out => if ((self.win.out.out_bg.read() >> n) & 1 == 0) return false,
+            }
+        }
+
+        if (self.scanline.btm()[i] != null) {
+            // The pixel found in the bottom layer is:
+            // 1. From a higher priority background
+            // 2. From a background that is marked for blending (Pixel A)
+
+            // If Alpha Blending isn't enabled, then we've already found a higher prio
+            // pixel, we can return early
+            if (self.bldcnt.mode.read() != 0b01) return false;
+
+            const b_layers = self.bldcnt.layer_b.read();
+
+            const win_part = if (bounds) |win| blk: {
+                // Window Enabled
+                break :blk switch (win) {
+                    .win0 => self.win.in.w0_bld.read(),
+                    .win1 => self.win.in.w1_bld.read(),
+                    .out => self.win.out.out_bld.read(),
+                };
+            } else true;
+
+            // If the Background is not marked for blending, we've already found
+            // a higher priority pixel, move on.
+
+            const is_blend_enabled = win_part and ((b_layers >> n) & 1 == 1);
+            if (!is_blend_enabled) return false;
+        }
+
+        return true;
+    }
+
     // TODO: Comment this + get a better understanding
     fn tilemapOffset(size: u2, x: u32, y: u32) u32 {
         // Current Row: (y % PIXEL_COUNT) / 8
@@ -794,6 +883,25 @@ const Window = struct {
         };
     }
 
+    fn inRange(self: *const Self, comptime id: u1, x: u9, y: u8) bool {
+        const h = self.h[id];
+        const v = self.v[id];
+
+        const y1 = v.y1.read();
+        const y2 = if (y1 > v.y2.read()) 160 else std.math.min(160, v.y2.read());
+
+        if (y1 <= y and y < y2) {
+            // Within Y bounds
+            const x1 = h.x1.read();
+            const x2 = if (x1 > h.x2.read()) 240 else std.math.min(240, h.x2.read());
+
+            // Within X Bounds
+            return x1 <= x and x < x2;
+        }
+
+        return false;
+    }
+
     pub fn setH(self: *Self, value: u32) void {
         self.h[0].raw = @truncate(u16, value);
         self.h[1].raw = @truncate(u16, value >> 16);
@@ -1135,37 +1243,6 @@ fn alphaBlend(top: u16, btm: u16, bldalpha: io.BldAlpha) u16 {
     return (bld_b << 10) | (bld_g << 5) | bld_r;
 }
 
-fn shouldDrawBackground(comptime n: u2, bldcnt: io.BldCnt, scanline: *Scanline, i: usize) bool {
-    // If a pixel has been drawn on the top layer, it's because
-    // Either the pixel is to be blended with a pixel on the bottom layer
-    // or the pixel is not to be blended at all
-    // Consequentially, if we find a pixel on the top layer, there's no need
-    // to render anything I think?
-    if (scanline.top()[i] != null) return false;
-
-    if (scanline.btm()[i] != null) {
-        // The Pixel found in the Bottom layer is
-        // 1. From a higher priority
-        // 2. From a Backround that is marked for Blending (Pixel A)
-        //
-        // We now have to confirm whether this current Background can be used
-        // as Pixel B or not.
-
-        // If Alpha Blending isn't enabled, we've aready found a higher
-        // priority pixel to render. Move on
-        if (bldcnt.mode.read() != 0b01) return false;
-
-        const b_layers = bldcnt.layer_b.read();
-        const is_blend_enabled = (b_layers >> n) & 1 == 1;
-
-        // If the Background is not marked for blending, we've already found
-        // a higher priority pixel, move on.
-        if (!is_blend_enabled) return false;
-    }
-
-    return true;
-}
-
 fn shouldDrawSprite(bldcnt: io.BldCnt, scanline: *Scanline, x: u9) bool {
     if (scanline.top()[x] != null) return false;
 
@@ -1180,23 +1257,6 @@ fn shouldDrawSprite(bldcnt: io.BldCnt, scanline: *Scanline, x: u9) bool {
     return true;
 }
 
-fn copyToBackgroundBuffer(comptime n: u2, bldcnt: io.BldCnt, scanline: *Scanline, i: usize, bgr555: u16) void {
-    if (bldcnt.mode.read() != 0b00) {
-        // Standard Alpha Blending
-        const a_layers = bldcnt.layer_a.read();
-        const is_blend_enabled = (a_layers >> n) & 1 == 1;
-
-        // If Alpha Blending is enabled and we've found an eligible layer for
-        // Pixel A, store the pixel in the bottom pixel buffer
-        if (is_blend_enabled) {
-            scanline.btm()[i] = bgr555;
-            return;
-        }
-    }
-
-    scanline.top()[i] = bgr555;
-}
-
 fn copyToSpriteBuffer(bldcnt: io.BldCnt, scanline: *Scanline, x: u9, bgr555: u16) void {
     if (bldcnt.mode.read() != 0b00) {
         // Alpha Blending