chore: debug read takes advantage of fastmem

deduplicate slowmem backup read handler

src/core/Bus.zig

@@ -197,8 +197,29 @@ fn fillTableExternalMemory(bus: *Self, addr: usize) ?*anyopaque {
     return &bus.pak.buf[masked_addr];
 }
 
-// TODO: Take advantage of fastmem here too?
 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 alignment = @alignOf(std.meta.Child(Ptr));
+        const ptr = @ptrCast(Ptr, @alignCast(alignment, 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);
 
@@ -210,29 +231,18 @@ pub fn dbgRead(self: *const Self, comptime T: type, unaligned_address: u32) T {
 
             break :blk self.openBus(T, address);
         },
-        0x02 => self.ewram.read(T, address),
+        0x02 => unreachable, // handled by fastmem
-        0x03 => self.iwram.read(T, address),
+        0x03 => unreachable, // handled by fastmem
         0x04 => self.readIo(T, address),
 
         // Internal Display Memory
-        0x05 => self.ppu.palette.read(T, address),
+        0x05 => unreachable, // handled by fastmem
-        0x06 => self.ppu.vram.read(T, address),
+        0x06 => unreachable, // handled by fastmem
-        0x07 => self.ppu.oam.read(T, address),
+        0x07 => unreachable, // handled by fastmem
 
         // External Memory (Game Pak)
         0x08...0x0D => self.pak.dbgRead(T, address),
-        0x0E...0x0F => blk: {
+        0x0E...0x0F => self.readBackup(T, unaligned_address),
-            const value = self.pak.backup.read(unaligned_address);
-
-            const multiplier = switch (T) {
-                u32 => 0x01010101,
-                u16 => 0x0101,
-                u8 => 1,
-                else => @compileError("Backup: Unsupported read width"),
-            };
-
-            break :blk @as(T, value) * multiplier;
-        },
         else => self.openBus(T, address),
     };
 }
@@ -352,7 +362,12 @@ fn slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
 
         // External Memory (Game Pak)
         0x08...0x0D => self.pak.read(T, address),
-        0x0E...0x0F => blk: {
+        0x0E...0x0F => self.readBackup(T, unaligned_address),
+        else => self.openBus(T, address),
+    };
+}
+
+fn readBackup(self: *const Self, comptime T: type, unaligned_address: u32) T {
     const value = self.pak.backup.read(unaligned_address);
 
     const multiplier = switch (T) {
@@ -362,10 +377,7 @@ fn slowRead(self: *Self, comptime T: type, unaligned_address: u32) T {
         else => @compileError("Backup: Unsupported read width"),
     };
 
-            break :blk @as(T, value) * multiplier;
+    return @as(T, value) * multiplier;
-        },
-        else => self.openBus(T, address),
-    };
 }
 
 pub fn write(self: *Self, comptime T: type, unaligned_address: u32, value: T) void {

src/core/apu.zig

@@ -15,10 +15,12 @@ const SoundFifo = std.fifo.LinearFifo(u8, .{ .Static = 0x20 });
 const getHalf = util.getHalf;
 const setHalf = util.setHalf;
 const intToBytes = util.intToBytes;
-const RingBuffer = util.RingBuffer;
 
 const log = std.log.scoped(.APU);
 
+pub const host_rate = @import("../platform.zig").sample_rate;
+pub const host_format = @import("../platform.zig").sample_format;
+
 pub fn read(comptime T: type, apu: *const Apu, addr: u32) ?T {
     const byte_addr = @truncate(u8, addr);
 
@@ -244,20 +246,17 @@ pub const Apu = struct {
 
     sampling_cycle: u2,
 
-    sample_queue: RingBuffer(u16),
+    stream: *SDL.SDL_AudioStream,
     sched: *Scheduler,
 
     fs: FrameSequencer,
     capacitor: f32,
 
+    is_buffer_full: bool,
+
     pub const Tick = enum { Length, Envelope, Sweep };
 
     pub fn init(sched: *Scheduler) Self {
-        const NUM_CHANNELS: usize = 2;
-
-        const allocator = std.heap.c_allocator;
-        const sample_buf = allocator.alloc(u16, 0x800 * NUM_CHANNELS) catch @panic("failed to allocate sample buffer");
-
         const apu: Self = .{
             .ch1 = ToneSweep.init(sched),
             .ch2 = Tone.init(sched),
@@ -272,11 +271,12 @@ pub const Apu = struct {
             .bias = .{ .raw = 0x0200 },
 
             .sampling_cycle = 0b00,
-            .sample_queue = RingBuffer(u16).init(sample_buf),
+            .stream = SDL.SDL_NewAudioStream(SDL.AUDIO_U16, 2, 1 << 15, host_format, 2, host_rate).?,
             .sched = sched,
 
             .capacitor = 0,
             .fs = FrameSequencer.init(),
+            .is_buffer_full = false,
         };
 
         sched.push(.SampleAudio, apu.interval());
@@ -370,6 +370,11 @@ pub const Apu = struct {
     pub fn sampleAudio(self: *Self, late: u64) void {
         self.sched.push(.SampleAudio, self.interval() -| 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;
 
@@ -425,7 +430,23 @@ pub const Apu = struct {
         const ext_left = (clamped_left << 5) | (clamped_left >> 6);
         const ext_right = (clamped_right << 5) | (clamped_right >> 6);
 
-        self.sample_queue.push(ext_left, ext_right) catch {};
+        if (self.sampling_cycle != self.bias.sampling_cycle.read()) self.replaceSDLResampler();
+
+        _ = SDL.SDL_AudioStreamPut(self.stream, &[2]u16{ ext_left, ext_right }, 2 * @sizeOf(u16));
+    }
+
+    fn replaceSDLResampler(self: *Self) void {
+        @setCold(true);
+        const sample_rate = Self.sampleRate(self.bias.sampling_cycle.read());
+        log.info("Sample Rate changed from {}Hz to {}Hz", .{ Self.sampleRate(self.sampling_cycle), sample_rate });
+
+        // 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), host_format, 2, host_rate).?;
     }
 
     fn interval(self: *const Self) u64 {

src/core/emu.zig

@@ -5,7 +5,6 @@ const config = @import("../config.zig");
 const Scheduler = @import("scheduler.zig").Scheduler;
 const Arm7tdmi = @import("cpu.zig").Arm7tdmi;
 const FpsTracker = @import("../util.zig").FpsTracker;
-const RingBuffer = @import("../util.zig").RingBuffer;
 
 const Timer = std.time.Timer;
 const Atomic = std.atomic.Atomic;
@@ -59,7 +58,7 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), schedule
 
             while (!quit.load(.Monotonic)) {
                 runFrame(scheduler, cpu);
-                audioSync(audio_sync, &cpu.bus.apu.sample_queue);
+                audioSync(audio_sync, cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
 
                 if (kind == .UnlimitedFPS) tracker.?.tick();
             }
@@ -78,7 +77,7 @@ fn inner(comptime kind: RunKind, audio_sync: bool, quit: *Atomic(bool), schedule
                 // the amount of time needed for audio to catch up rather than
                 // our expected wake-up time
 
-                audioSync(audio_sync, &cpu.bus.apu.sample_queue);
+                audioSync(audio_sync, cpu.bus.apu.stream, &cpu.bus.apu.is_buffer_full);
                 if (!audio_sync) spinLoop(&timer, wake_time);
                 wake_time = new_wake_time;
 
@@ -105,13 +104,22 @@ pub fn runFrame(sched: *Scheduler, cpu: *Arm7tdmi) void {
     }
 }
 
-fn audioSync(audio_sync: bool, sample_queue: *RingBuffer(u16)) void {
+fn audioSync(audio_sync: bool, stream: *SDL.SDL_AudioStream, is_buffer_full: *bool) void {
     comptime std.debug.assert(@import("../platform.zig").sample_format == SDL.AUDIO_U16);
-    // const sample_size = 2 * @sizeOf(u16);
+    const sample_size = 2 * @sizeOf(u16);
-    // const max_buf_size: c_int = 0x400;
+    const max_buf_size: c_int = 0x400;
 
-    _ = audio_sync;
+    // Determine whether the APU is busy right at this moment
-    _ = sample_queue;
+    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 (!audio_sync or !still_full) break;
+    }
 }
 
 fn videoSync(timer: *Timer, wake_time: u64) u64 {

src/platform.zig

@@ -12,7 +12,7 @@ const FpsTracker = @import("util.zig").FpsTracker;
 const gba_width = @import("core/ppu.zig").width;
 const gba_height = @import("core/ppu.zig").height;
 
-pub const sample_rate = 1 << 16;
+pub const sample_rate = 1 << 15;
 pub const sample_format = SDL.AUDIO_U16;
 
 const default_title = "ZBA";
@@ -216,7 +216,7 @@ pub const Gui = struct {
                             SDL.SDLK_RSHIFT => keyinput.select.set(),
                             SDL.SDLK_i => {
                                 comptime std.debug.assert(sample_format == SDL.AUDIO_U16);
-                                log.err("Sample Count: {}", .{cpu.bus.apu.sample_queue.len() / 2});
+                                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 => {},
@@ -299,15 +299,7 @@ const Audio = struct {
         const T = *Apu;
         const apu = @ptrCast(T, @alignCast(@alignOf(T), userdata));
 
-        comptime std.debug.assert(sample_format == SDL.AUDIO_U16);
+        _ = SDL.SDL_AudioStreamGet(apu.stream, stream, len);
-        const sample_buf = @ptrCast([*]u16, @alignCast(@alignOf(u16), stream))[0 .. @intCast(u32, len) / @sizeOf(u16)];
-
-        var previous: u16 = 0x8000;
-        for (sample_buf) |*sample| {
-            if (apu.sample_queue.pop()) |value| previous = value;
-
-            sample.* = previous;
-        }
     }
 };
 

src/util.zig

@@ -275,110 +275,3 @@ fn HalfInt(comptime T: type) type {
 
     return std.meta.Int(type_info.Int.signedness, type_info.Int.bits >> 1);
 }
-
-const Mutex = std.Thread.Mutex;
-
-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,
-
-        mutex: Mutex,
-
-        const Error = error{buffer_full};
-
-        pub fn init(buf: []T) Self {
-            std.mem.set(T, buf, 0);
-
-            std.debug.assert(std.math.isPowerOfTwo(buf.len)); // capacity must be a power of two
-            std.debug.assert(buf.len <= max_capacity);
-
-            return .{ .read = 0, .write = 0, .buf = buf, .mutex = .{} };
-        }
-
-        pub fn push(self: *Self, left: T, right: T) Error!void {
-            self.mutex.lock();
-            defer self.mutex.unlock();
-
-            try self._push(left);
-            self._push(right) catch |e| {
-                self.write -= 1; // undo the previous write;
-                return e;
-            };
-        }
-
-        pub fn pop(self: *Self) ?T {
-            self.mutex.lock();
-            defer self.mutex.unlock();
-
-            return self._pop();
-        }
-
-        pub fn len(self: *Self) Index {
-            self.mutex.lock();
-            defer self.mutex.unlock();
-
-            return self._len();
-        }
-
-        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;
-        }
-
-        fn _pop(self: *Self) ?T {
-            if (self.isEmpty()) return null;
-            defer self.read += 1;
-
-            return self.buf[self.mask(self.read)];
-        }
-
-        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);
-        }
-    };
-}
-
-test "RingBuffer" {
-    const Queue = RingBuffer(u8);
-
-    var buf: [4]u8 = undefined;
-    var queue = Queue.init(&buf);
-
-    try queue.push(1, 2);
-    try std.testing.expectEqual(@as(?u8, 1), queue.pop());
-
-    try queue.push(3, 4);
-    try std.testing.expectError(Queue.Error.buffer_full, queue.push(5, 6));
-    try std.testing.expectEqual(@as(?u8, 2), queue.pop());
-
-    try queue.push(7, 8);
-
-    try std.testing.expectEqual(@as(?u8, 3), queue.pop());
-    try std.testing.expectEqual(@as(?u8, 4), queue.pop());
-    try std.testing.expectEqual(@as(?u8, 7), queue.pop());
-    try std.testing.expectEqual(@as(?u8, 8), queue.pop());
-    try std.testing.expectEqual(@as(?u8, null), queue.pop());
-}