chore: dont allocate not-small ?Sprite array on stack

use memset like most other allocations in this emu

src/core/Bus.zig

@@ -169,7 +169,7 @@ fn openBus(self: *const Self, comptime T: type, address: u32) T {
         }
     };
 
-    return @truncate(T, rotr(u32, word, 8 * (address & 3)));
+    return @truncate(T, word);
 }
 
 pub fn read(self: *Self, comptime T: type, address: u32) T {

src/core/bus/Bios.zig

@@ -19,7 +19,7 @@ pub fn read(self: *Self, comptime T: type, r15: u32, addr: u32) T {
     }
 
     log.debug("Rejected read since r15=0x{X:0>8}", .{r15});
-    return @truncate(T, self._read(T, self.addr_latch + 8));
+    return @truncate(T, self._read(T, self.addr_latch));
 }
 
 pub fn dbgRead(self: *const Self, comptime T: type, r15: u32, addr: u32) T {

src/core/bus/dma.zig

@@ -11,6 +11,8 @@ const log = std.log.scoped(.DmaTransfer);
 const setHi = util.setHi;
 const setLo = util.setLo;
 
+const rotr = @import("../../util.zig").rotr;
+
 pub fn create() DmaTuple {
     return .{ DmaController(0).init(), DmaController(1).init(), DmaController(2).init(), DmaController(3).init() };
 }
@@ -99,6 +101,7 @@ fn DmaController(comptime id: u2) type {
 
         const sad_mask: u32 = if (id == 0) 0x07FF_FFFF else 0x0FFF_FFFF;
         const dad_mask: u32 = if (id != 3) 0x07FF_FFFF else 0x0FFF_FFFF;
+        const WordCount = if (id == 3) u16 else u14;
 
         /// Write-only. The first address in a DMA transfer. (DMASAD)
         /// Note: use writeSrc instead of manipulating src_addr directly
@@ -107,17 +110,19 @@ fn DmaController(comptime id: u2) type {
         /// Note: Use writeDst instead of manipulatig dst_addr directly
         dad: u32,
         /// Write-only. The Word Count for the DMA Transfer (DMACNT_L)
-        word_count: if (id == 3) u16 else u14,
+        word_count: WordCount,
         /// Read / Write. DMACNT_H
         /// Note: Use writeControl instead of manipulating cnt directly.
         cnt: DmaControl,
 
+        /// Internal. The last successfully read value
+        data_latch: u32,
         /// Internal. Currrent Source Address
         sad_latch: u32,
         /// Internal. Current Destination Address
         dad_latch: u32,
         /// Internal. Word Count
-        _word_count: if (id == 3) u16 else u14,
+        _word_count: WordCount,
 
         /// Some DMA Transfers are enabled during Hblank / VBlank and / or
         /// have delays. Thefore bit 15 of DMACNT isn't actually something
@@ -134,6 +139,8 @@ fn DmaController(comptime id: u2) type {
                 // Internals
                 .sad_latch = 0,
                 .dad_latch = 0,
+                .data_latch = 0,
+
                 ._word_count = 0,
                 .in_progress = false,
             };
@@ -158,7 +165,7 @@ fn DmaController(comptime id: u2) type {
                 // Reload Internals on Rising Edge.
                 self.sad_latch = self.sad;
                 self.dad_latch = self.dad;
-                self._word_count = if (self.word_count == 0) std.math.maxInt(@TypeOf(self._word_count)) else self.word_count;
+                self._word_count = if (self.word_count == 0) std.math.maxInt(WordCount) else self.word_count;
 
                 // Only a Start Timing of 00 has a DMA Transfer immediately begin
                 self.in_progress = new.start_timing.read() == 0b00;
@@ -181,11 +188,19 @@ fn DmaController(comptime id: u2) type {
             const offset: u32 = if (transfer_type) @sizeOf(u32) else @sizeOf(u16);
 
             const mask = if (transfer_type) ~@as(u32, 3) else ~@as(u32, 1);
+            const sad_addr = self.sad_latch & mask;
+            const dad_addr = self.dad_latch & mask;
 
             if (transfer_type) {
-                cpu.bus.write(u32, self.dad_latch & mask, cpu.bus.read(u32, self.sad_latch & mask));
+                if (sad_addr >= 0x0200_0000) self.data_latch = cpu.bus.read(u32, sad_addr);
+                cpu.bus.write(u32, dad_addr, self.data_latch);
             } else {
-                cpu.bus.write(u16, self.dad_latch & mask, cpu.bus.read(u16, self.sad_latch & mask));
+                if (sad_addr >= 0x0200_0000) {
+                    const value: u32 = cpu.bus.read(u16, sad_addr);
+                    self.data_latch = value << 16 | value;
+                }
+
+                cpu.bus.write(u16, dad_addr, @truncate(u16, rotr(u32, self.data_latch, 8 * (dad_addr & 3))));
             }
 
             switch (sad_adj) {
@@ -266,7 +281,7 @@ fn DmaController(comptime id: u2) type {
     };
 }
 
-pub fn pollDmaOnBlank(bus: *Bus, comptime kind: DmaKind) void {
+pub fn onBlanking(bus: *Bus, comptime kind: DmaKind) void {
     bus.dma[0].poll(kind);
     bus.dma[1].poll(kind);
     bus.dma[2].poll(kind);

src/core/bus/io.zig

@@ -11,8 +11,8 @@ const Bus = @import("../Bus.zig");
 const DmaController = @import("dma.zig").DmaController;
 const Scheduler = @import("../scheduler.zig").Scheduler;
 
-const setHi = util.setLo;
-const setLo = util.setHi;
+const setHi = util.setHi;
+const setLo = util.setLo;
 
 const log = std.log.scoped(.@"I/O");
 
@@ -305,6 +305,14 @@ pub fn write(bus: *Bus, comptime T: type, address: u32, value: T) void {
             0x0400_0009 => bus.ppu.bg[0].cnt.raw = setHi(u16, bus.ppu.bg[0].cnt.raw, value),
             0x0400_000A => bus.ppu.bg[1].cnt.raw = setLo(u16, bus.ppu.bg[1].cnt.raw, value),
             0x0400_000B => bus.ppu.bg[1].cnt.raw = setHi(u16, bus.ppu.bg[1].cnt.raw, value),
+            0x0400_0040 => bus.ppu.win.h[0].raw = setLo(u16, bus.ppu.win.h[0].raw, value),
+            0x0400_0041 => bus.ppu.win.h[0].raw = setHi(u16, bus.ppu.win.h[0].raw, value),
+            0x0400_0042 => bus.ppu.win.h[1].raw = setLo(u16, bus.ppu.win.h[1].raw, value),
+            0x0400_0043 => bus.ppu.win.h[1].raw = setHi(u16, bus.ppu.win.h[1].raw, value),
+            0x0400_0044 => bus.ppu.win.v[0].raw = setLo(u16, bus.ppu.win.v[0].raw, value),
+            0x0400_0045 => bus.ppu.win.v[0].raw = setHi(u16, bus.ppu.win.v[0].raw, value),
+            0x0400_0046 => bus.ppu.win.v[1].raw = setLo(u16, bus.ppu.win.v[1].raw, value),
+            0x0400_0047 => bus.ppu.win.v[1].raw = setHi(u16, bus.ppu.win.v[1].raw, value),
             0x0400_0048 => bus.ppu.win.in.raw = setLo(u16, bus.ppu.win.in.raw, value),
             0x0400_0049 => bus.ppu.win.in.raw = setHi(u16, bus.ppu.win.in.raw, value),
             0x0400_004A => bus.ppu.win.out.raw = setLo(u16, bus.ppu.win.out.raw, value),
@@ -462,8 +470,12 @@ 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,
@@ -471,28 +483,37 @@ pub const WinH = extern union {
 
 /// Write-only
 pub const WinV = extern union {
+    const Self = @This();
+
     y2: Bitfield(u16, 0, 8),
     y1: Bitfield(u16, 8, 8),
     raw: u16,
+
+    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/arm/half_signed_data_transfer.zig

@@ -35,11 +35,8 @@ pub fn halfAndSignedDataTransfer(comptime P: bool, comptime U: bool, comptime I:
                     },
                     0b11 => {
                         // LDRSH
-                        result = if (address & 1 == 1) blk: {
-                            break :blk sext(u32, u8, bus.read(u8, address));
-                        } else blk: {
-                            break :blk sext(u32, u16, bus.read(u16, address));
-                        };
+                        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
                 }

src/core/cpu/thumb/data_transfer.zig

@@ -46,11 +46,8 @@ pub fn fmt78(comptime op: u2, comptime T: bool) InstrFn {
                     },
                     0b11 => {
                         // LDRSH
-                        cpu.r[rd] = if (address & 1 == 1) blk: {
-                            break :blk sext(u32, u8, bus.read(u8, address));
-                        } else blk: {
-                            break :blk sext(u32, u16, bus.read(u16, address));
-                        };
+                        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 {

src/core/ppu.zig

@@ -1,16 +1,19 @@
 const std = @import("std");
 const io = @import("bus/io.zig");
+const Bit = @import("bitfield").Bit;
+const Bitfield = @import("bitfield").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 EventKind = @import("scheduler.zig").EventKind;
 const Scheduler = @import("scheduler.zig").Scheduler;
 const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
-
-const Bit = @import("bitfield").Bit;
-const Bitfield = @import("bitfield").Bitfield;
+const FrameBuffer = @import("../util.zig").FrameBuffer;
 
 const Allocator = std.mem.Allocator;
-const log = std.log.scoped(.PPU);
-const pollDmaOnBlank = @import("bus/dma.zig").pollDmaOnBlank;
+const log = std.log.scoped(.Ppu);
 
 pub const width = 240;
 pub const height = 160;
@@ -48,14 +51,14 @@ pub const Ppu = struct {
         sched.push(.Draw, 240 * 4);
 
         const sprites = try allocator.create([128]?Sprite);
-        sprites.* = [_]?Sprite{null} ** 128;
+        std.mem.set(?Sprite, sprites, null);
 
         return Self{
             .vram = try Vram.init(allocator),
             .palette = try Palette.init(allocator),
             .oam = try Oam.init(allocator),
             .sched = sched,
-            .framebuf = try FrameBuffer.init(allocator),
+            .framebuf = try FrameBuffer.init(allocator, framebuf_pitch * height),
             .allocator = allocator,
 
             // Registers
@@ -274,16 +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;
@@ -293,7 +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);
             }
         }
 
@@ -302,7 +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());
@@ -322,10 +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));
@@ -350,7 +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);
             }
         }
     }
@@ -376,10 +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
@@ -404,8 +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);
                 }
 
@@ -483,7 +488,7 @@ pub const Ppu = struct {
                 while (i < width) : (i += 1) {
                     // If we're outside of the bounds of mode 5, draw the background colour
                     const bgr555 =
-                        if (scanline < m5_height and i < m5_width) self.vram.read(u16, vram_base + i * @sizeOf(u16)) else self.palette.getBackdrop();
+                        if (scanline < m5_height and i < m5_width) self.vram.read(u16, vram_base + i * @sizeOf(u16)) else self.palette.backdrop();
 
                     std.mem.writeIntNative(u32, self.framebuf.get(.Emulator)[fb_base + i * @sizeOf(u32) ..][0..@sizeOf(u32)], rgba888(bgr555));
                 }
@@ -527,7 +532,94 @@ pub const Ppu = struct {
         }
 
         if (maybe_top) |top| return top;
-        return self.palette.getBackdrop();
+        return self.palette.backdrop();
+    }
+
+    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
@@ -569,7 +661,7 @@ pub const Ppu = struct {
         // See if HBlank DMA is present and not enabled
 
         if (!self.dispstat.vblank.read())
-            pollDmaOnBlank(cpu.bus, .HBlank);
+            dma.onBlanking(cpu.bus, .HBlank);
 
         self.dispstat.hblank.set();
         self.sched.push(.HBlank, 68 * 4 -| late);
@@ -611,7 +703,7 @@ pub const Ppu = struct {
                 self.aff_bg[1].latchRefPoints();
 
                 // See if Vblank DMA is present and not enabled
-                pollDmaOnBlank(cpu.bus, .VBlank);
+                dma.onBlanking(cpu.bus, .VBlank);
             }
 
             if (scanline == 227) self.dispstat.vblank.unset();
@@ -620,158 +712,6 @@ pub const Ppu = struct {
     }
 };
 
-const Palette = struct {
-    const palram_size = 0x400;
-    const Self = @This();
-
-    buf: []u8,
-    allocator: Allocator,
-
-    fn init(allocator: Allocator) !Self {
-        const buf = try allocator.alloc(u8, palram_size);
-        std.mem.set(u8, buf, 0);
-
-        return Self{
-            .buf = buf,
-            .allocator = allocator,
-        };
-    }
-
-    fn deinit(self: *Self) void {
-        self.allocator.free(self.buf);
-        self.* = undefined;
-    }
-
-    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)]),
-            else => @compileError("PALRAM: Unsupported read width"),
-        };
-    }
-
-    pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
-        const addr = address & 0x3FF;
-
-        switch (T) {
-            u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
-            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);
-            },
-            else => @compileError("PALRAM: Unsupported write width"),
-        }
-    }
-
-    fn getBackdrop(self: *const Self) u16 {
-        return self.read(u16, 0);
-    }
-};
-
-const Vram = struct {
-    const vram_size = 0x18000;
-    const Self = @This();
-
-    buf: []u8,
-    allocator: Allocator,
-
-    fn init(allocator: Allocator) !Self {
-        const buf = try allocator.alloc(u8, vram_size);
-        std.mem.set(u8, buf, 0);
-
-        return Self{
-            .buf = buf,
-            .allocator = allocator,
-        };
-    }
-
-    fn deinit(self: *Self) void {
-        self.allocator.free(self.buf);
-        self.* = undefined;
-    }
-
-    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)]),
-            else => @compileError("VRAM: Unsupported read width"),
-        };
-    }
-
-    pub fn write(self: *Self, comptime T: type, dispcnt: io.DisplayControl, address: usize, value: T) void {
-        const mode: u3 = dispcnt.bg_mode.read();
-        const idx = Self.mirror(address);
-
-        switch (T) {
-            u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[idx..][0..@sizeOf(T)], value),
-            u8 => {
-                // Ignore write if it falls within the boundaries of OBJ VRAM
-                switch (mode) {
-                    0, 1, 2 => if (0x0001_0000 <= idx) return,
-                    else => if (0x0001_4000 <= idx) return,
-                }
-
-                const align_idx = idx & ~@as(u32, 1); // Aligned to a halfword boundary
-                std.mem.writeIntSliceLittle(u16, self.buf[align_idx..][0..@sizeOf(u16)], @as(u16, value) * 0x101);
-            },
-            else => @compileError("VRAM: Unsupported write width"),
-        }
-    }
-
-    fn mirror(address: usize) usize {
-        // Mirrored in steps of 128K (64K + 32K + 32K) (abcc)
-        const addr = address & 0x1FFFF;
-
-        // If the address is within 96K we don't do anything,
-        // otherwise we want to mirror the last 32K (addresses between 64K and 96K)
-        return if (addr < vram_size) addr else 0x10000 + (addr & 0x7FFF);
-    }
-};
-
-const Oam = struct {
-    const oam_size = 0x400;
-    const Self = @This();
-
-    buf: []u8,
-    allocator: Allocator,
-
-    fn init(allocator: Allocator) !Self {
-        const buf = try allocator.alloc(u8, oam_size);
-        std.mem.set(u8, buf, 0);
-
-        return Self{
-            .buf = buf,
-            .allocator = allocator,
-        };
-    }
-
-    fn deinit(self: *Self) void {
-        self.allocator.free(self.buf);
-        self.* = undefined;
-    }
-
-    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)]),
-            else => @compileError("OAM: Unsupported read width"),
-        };
-    }
-
-    pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
-        const addr = address & 0x3FF;
-
-        switch (T) {
-            u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
-            u8 => return, // 8-bit writes are explicitly ignored
-            else => @compileError("OAM: Unsupported write width"),
-        }
-    }
-};
-
 const Window = struct {
     const Self = @This();
 
@@ -791,6 +731,25 @@ 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);
@@ -1087,37 +1046,6 @@ 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;
 
@@ -1132,23 +1060,6 @@ 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
@@ -1201,48 +1112,3 @@ const Scanline = struct {
         return self.layers[1];
     }
 };
-
-// Double Buffering Implementation
-const FrameBuffer = struct {
-    const Self = @This();
-
-    layers: [2][]u8,
-    buf: []u8,
-    current: u1,
-
-    allocator: Allocator,
-
-    // TODO: Rename
-    const Device = enum {
-        Emulator,
-        Renderer,
-    };
-
-    pub fn init(allocator: Allocator) !Self {
-        const framebuf_len = framebuf_pitch * height;
-        const buf = try allocator.alloc(u8, framebuf_len * 2);
-        std.mem.set(u8, buf, 0);
-
-        return .{
-            // Front and Back Framebuffers
-            .layers = [_][]u8{ buf[0..][0..framebuf_len], buf[framebuf_len..][0..framebuf_len] },
-            .buf = buf,
-            .current = 0,
-
-            .allocator = allocator,
-        };
-    }
-
-    fn deinit(self: *Self) void {
-        self.allocator.free(self.buf);
-        self.* = undefined;
-    }
-
-    pub fn swap(self: *Self) void {
-        self.current = ~self.current;
-    }
-
-    pub fn get(self: *Self, comptime dev: Device) []u8 {
-        return self.layers[if (dev == .Emulator) self.current else ~self.current];
-    }
-};

src/core/ppu/Oam.zig

@@ -0,0 +1,40 @@
+const std = @import("std");
+
+const Allocator = std.mem.Allocator;
+
+const buf_len = 0x400;
+const Self = @This();
+
+buf: []u8,
+allocator: Allocator,
+
+pub fn read(self: *const Self, comptime T: type, address: usize) T {
+    const addr = address & 0x3FF;
+
+    return switch (T) {
+        u32, u16, u8 => std.mem.readIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)]),
+        else => @compileError("OAM: Unsupported read width"),
+    };
+}
+
+pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
+    const addr = address & 0x3FF;
+
+    switch (T) {
+        u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
+        u8 => return, // 8-bit writes are explicitly ignored
+        else => @compileError("OAM: Unsupported write width"),
+    }
+}
+
+pub fn init(allocator: Allocator) !Self {
+    const buf = try allocator.alloc(u8, buf_len);
+    std.mem.set(u8, buf, 0);
+
+    return Self{ .buf = buf, .allocator = allocator };
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.buf);
+    self.* = undefined;
+}

src/core/ppu/Palette.zig

@@ -0,0 +1,47 @@
+const std = @import("std");
+
+const Allocator = std.mem.Allocator;
+
+const buf_len = 0x400;
+const Self = @This();
+
+buf: []u8,
+allocator: Allocator,
+
+pub fn read(self: *const Self, comptime T: type, address: usize) T {
+    const addr = address & 0x3FF;
+
+    return switch (T) {
+        u32, u16, u8 => std.mem.readIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)]),
+        else => @compileError("PALRAM: Unsupported read width"),
+    };
+}
+
+pub fn write(self: *Self, comptime T: type, address: usize, value: T) void {
+    const addr = address & 0x3FF;
+
+    switch (T) {
+        u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)], value),
+        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);
+        },
+        else => @compileError("PALRAM: Unsupported write width"),
+    }
+}
+
+pub fn init(allocator: Allocator) !Self {
+    const buf = try allocator.alloc(u8, buf_len);
+    std.mem.set(u8, buf, 0);
+
+    return Self{ .buf = buf, .allocator = allocator };
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.buf);
+    self.* = undefined;
+}
+
+pub fn backdrop(self: *const Self) u16 {
+    return self.read(u16, 0);
+}

src/core/ppu/Vram.zig

@@ -0,0 +1,60 @@
+const std = @import("std");
+const io = @import("../bus/io.zig");
+
+const Allocator = std.mem.Allocator;
+
+const buf_len = 0x18000;
+const Self = @This();
+
+buf: []u8,
+allocator: Allocator,
+
+pub fn read(self: *const Self, comptime T: type, address: usize) T {
+    const addr = Self.mirror(address);
+
+    return switch (T) {
+        u32, u16, u8 => std.mem.readIntSliceLittle(T, self.buf[addr..][0..@sizeOf(T)]),
+        else => @compileError("VRAM: Unsupported read width"),
+    };
+}
+
+pub fn write(self: *Self, comptime T: type, dispcnt: io.DisplayControl, address: usize, value: T) void {
+    const mode: u3 = dispcnt.bg_mode.read();
+    const idx = Self.mirror(address);
+
+    switch (T) {
+        u32, u16 => std.mem.writeIntSliceLittle(T, self.buf[idx..][0..@sizeOf(T)], value),
+        u8 => {
+            // Ignore write if it falls within the boundaries of OBJ VRAM
+            switch (mode) {
+                0, 1, 2 => if (0x0001_0000 <= idx) return,
+                else => if (0x0001_4000 <= idx) return,
+            }
+
+            const align_idx = idx & ~@as(u32, 1); // Aligned to a halfword boundary
+            std.mem.writeIntSliceLittle(u16, self.buf[align_idx..][0..@sizeOf(u16)], @as(u16, value) * 0x101);
+        },
+        else => @compileError("VRAM: Unsupported write width"),
+    }
+}
+
+pub fn init(allocator: Allocator) !Self {
+    const buf = try allocator.alloc(u8, buf_len);
+    std.mem.set(u8, buf, 0);
+
+    return Self{ .buf = buf, .allocator = allocator };
+}
+
+pub fn deinit(self: *Self) void {
+    self.allocator.free(self.buf);
+    self.* = undefined;
+}
+
+fn mirror(address: usize) usize {
+    // Mirrored in steps of 128K (64K + 32K + 32K) (abcc)
+    const addr = address & 0x1FFFF;
+
+    // If the address is within 96K we don't do anything,
+    // otherwise we want to mirror the last 32K (addresses between 64K and 96K)
+    return if (addr < buf_len) addr else 0x10000 + (addr & 0x7FFF);
+}

src/util.zig

@@ -5,6 +5,8 @@ const config = @import("config.zig");
 const Log2Int = std.math.Log2Int;
 const Arm7tdmi = @import("core/cpu.zig").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
@@ -165,6 +167,7 @@ pub const io = struct {
 
 pub const Logger = struct {
     const Self = @This();
+    const FmtArgTuple = std.meta.Tuple(&.{ u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32 });
 
     buf: std.io.BufferedWriter(4096 << 2, std.fs.File.Writer),
 
@@ -223,8 +226,6 @@ pub const Logger = struct {
     }
 };
 
-const FmtArgTuple = std.meta.Tuple(&.{ u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32 });
-
 pub const audio = struct {
     const _io = @import("core/bus/io.zig");
 
@@ -275,7 +276,7 @@ pub const audio = struct {
 };
 
 /// Sets the high bits of an integer to a value
-pub inline fn setHi(comptime T: type, left: T, right: HalfInt(T)) T {
+pub inline fn setLo(comptime T: type, left: T, right: HalfInt(T)) T {
     return switch (T) {
         u32 => (left & 0xFFFF_0000) | right,
         u16 => (left & 0xFF00) | right,
@@ -285,7 +286,7 @@ pub inline fn setHi(comptime T: type, left: T, right: HalfInt(T)) T {
 }
 
 /// sets the low bits of an integer to a value
-pub inline fn setLo(comptime T: type, left: T, right: HalfInt(T)) T {
+pub inline fn setHi(comptime T: type, left: T, right: HalfInt(T)) T {
     return switch (T) {
         u32 => (left & 0x0000_FFFF) | @as(u32, right) << 16,
         u16 => (left & 0x00FF) | @as(u16, right) << 8,
@@ -302,3 +303,44 @@ fn HalfInt(comptime T: type) type {
 
     return std.meta.Int(type_info.Int.signedness, type_info.Int.bits >> 1);
 }
+
+/// Double Buffering Implementation
+pub const FrameBuffer = struct {
+    const Self = @This();
+
+    layers: [2][]u8,
+    buf: []u8,
+    current: u1,
+
+    allocator: Allocator,
+
+    // TODO: Rename
+    const Device = enum { Emulator, Renderer };
+
+    pub fn init(allocator: Allocator, comptime len: comptime_int) !Self {
+        const buf = try allocator.alloc(u8, len * 2);
+        std.mem.set(u8, 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 deinit(self: *Self) void {
+        self.allocator.free(self.buf);
+        self.* = undefined;
+    }
+
+    pub fn swap(self: *Self) void {
+        self.current = ~self.current;
+    }
+
+    pub fn get(self: *Self, comptime dev: Device) []u8 {
+        return self.layers[if (dev == .Emulator) self.current else ~self.current];
+    }
+};