Implement audio output

.gitmodules

@@ -58,3 +58,6 @@
 [submodule "third_party/teakra"]
 	path = third_party/teakra
 	url = https://github.com/wwylele/teakra
+[submodule "third_party/miniaudio"]
+	path = third_party/miniaudio
+	url = https://github.com/mackron/miniaudio

CMakeLists.txt

@@ -54,6 +54,7 @@ include_directories(third_party/xxhash/include)
 include_directories(third_party/httplib)
 include_directories(third_party/stb)
 include_directories(third_party/opengl)
+include_directories(third_party/miniaudio)
 include_directories(third_party/mio/single_include)
 
 add_compile_definitions(NOMINMAX)             # Make windows.h not define min/max macros because third-party deps don't like it
@@ -89,7 +90,6 @@ include_directories(third_party/toml11)
 include_directories(third_party/glm)
 
 add_subdirectory(third_party/cmrc)
-add_subdirectory(third_party/teakra EXCLUDE_FROM_ALL)
 
 set(BOOST_ROOT "${CMAKE_SOURCE_DIR}/third_party/boost")
 set(Boost_INCLUDE_DIR "${CMAKE_SOURCE_DIR}/third_party/boost")
@@ -150,12 +150,13 @@ if(HOST_X64 OR HOST_ARM64)
 else()
     message(FATAL_ERROR "Currently unsupported CPU architecture")
 endif()
+add_subdirectory(third_party/teakra EXCLUDE_FROM_ALL)
 
 set(SOURCE_FILES src/emulator.cpp src/io_file.cpp src/config.cpp
 				 src/core/CPU/cpu_dynarmic.cpp src/core/CPU/dynarmic_cycles.cpp
 				 src/core/memory.cpp src/renderer.cpp src/core/renderer_null/renderer_null.cpp
 				 src/http_server.cpp src/stb_image_write.c src/core/cheats.cpp src/core/action_replay.cpp
-                 src/discord_rpc.cpp src/lua.cpp src/memory_mapped_file.cpp
+                 src/discord_rpc.cpp src/lua.cpp src/memory_mapped_file.cpp src/miniaudio.cpp
 )
 set(CRYPTO_SOURCE_FILES src/core/crypto/aes_engine.cpp)
 set(KERNEL_SOURCE_FILES src/core/kernel/kernel.cpp src/core/kernel/resource_limits.cpp
@@ -192,7 +193,9 @@ set(FS_SOURCE_FILES src/core/fs/archive_self_ncch.cpp src/core/fs/archive_save_d
 set(APPLET_SOURCE_FILES src/core/applets/applet.cpp src/core/applets/mii_selector.cpp src/core/applets/software_keyboard.cpp src/core/applets/applet_manager.cpp
                         src/core/applets/error_applet.cpp
 )
-set(AUDIO_SOURCE_FILES src/core/audio/dsp_core.cpp src/core/audio/null_core.cpp src/core/audio/teakra_core.cpp)
+set(AUDIO_SOURCE_FILES src/core/audio/dsp_core.cpp src/core/audio/null_core.cpp src/core/audio/teakra_core.cpp
+                       src/core/audio/miniaudio_device.cpp
+)
 set(RENDERER_SW_SOURCE_FILES src/core/renderer_sw/renderer_sw.cpp)
 
 # Frontend source files
@@ -250,6 +253,7 @@ set(HEADER_FILES include/emulator.hpp include/helpers.hpp include/termcolor.hpp
                  include/fs/archive_system_save_data.hpp include/lua_manager.hpp include/memory_mapped_file.hpp include/hydra_icon.hpp
                  include/PICA/dynapica/shader_rec_emitter_arm64.hpp include/scheduler.hpp include/applets/error_applet.hpp
                  include/audio/dsp_core.hpp include/audio/null_core.hpp include/audio/teakra_core.hpp
+                 include/audio/miniaudio_device.hpp include/ring_buffer.hpp
 )
 
 cmrc_add_resource_library(

include/audio/dsp_core.hpp

@@ -9,6 +9,7 @@
 #include "helpers.hpp"
 #include "logger.hpp"
 #include "scheduler.hpp"
+#include "ring_buffer.hpp"
 
 // The DSP core must have access to the DSP service to be able to trigger interrupts properly
 class DSPService;
@@ -23,16 +24,21 @@ namespace Audio {
 	static constexpr u64 lleSlice = 16384;
 
 	class DSPCore {
+		using Samples = Common::RingBuffer<s16, 1024>;
+
 	  protected:
 		Memory& mem;
 		Scheduler& scheduler;
 		DSPService& dspService;
 
+		Samples sampleBuffer;
+
 		MAKE_LOG_FUNCTION(log, dspLogger)
 
 	  public:
 		enum class Type { Null, Teakra };
-		DSPCore(Memory& mem, Scheduler& scheduler, DSPService& dspService) : mem(mem), scheduler(scheduler), dspService(dspService) {}
+		DSPCore(Memory& mem, Scheduler& scheduler, DSPService& dspService)
+			: mem(mem), scheduler(scheduler), dspService(dspService) {}
 
 		virtual void reset() = 0;
 		virtual void runAudioFrame() = 0;
@@ -49,6 +55,7 @@ namespace Audio {
 
 		static Audio::DSPCore::Type typeFromString(std::string inString);
 		static const char* typeToString(Audio::DSPCore::Type type);
+		Samples& getSamples() { return sampleBuffer; }
 	};
 
 	std::unique_ptr<DSPCore> makeDSPCore(DSPCore::Type type, Memory& mem, Scheduler& scheduler, DSPService& dspService);

include/audio/miniaudio_device.hpp

@@ -0,0 +1,28 @@
+#pragma once
+#include <string>
+#include <vector>
+
+#include "miniaudio.h"
+#include "ring_buffer.hpp"
+
+class MiniAudioDevice {
+	using Samples = Common::RingBuffer<ma_int16, 1024>;
+	// static constexpr ma_uint32 sampleRateIn = 32768; // 3DS sample rate
+	// static constexpr ma_uint32 sampleRateOut = 44100; // Output sample rate
+
+	ma_context context;
+	ma_device_config deviceConfig;
+	ma_device device;
+	ma_resampler resampler;
+	Samples* samples = nullptr;
+
+	bool initialized = false;
+	bool running = false;
+
+	std::vector<std::string> audioDevices;
+  public:
+	MiniAudioDevice();
+	// If safe is on, we create a null audio device
+	void init(Samples& samples, bool safe = false);
+	void start();
+};

include/emulator.hpp

@@ -8,6 +8,7 @@
 
 #include "PICA/gpu.hpp"
 #include "audio/dsp_core.hpp"
+#include "audio/miniaudio_device.hpp"
 #include "cheats.hpp"
 #include "config.hpp"
 #include "cpu.hpp"
@@ -47,6 +48,7 @@ class Emulator {
 	Scheduler scheduler;
 
 	Crypto::AESEngine aesEngine;
+	MiniAudioDevice audioDevice;
 	Cheats cheats;
 
 	// Variables to keep track of whether the user is controlling the 3DS analog stick with their keyboard

include/ring_buffer.hpp

@@ -0,0 +1,117 @@
+// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <algorithm>
+#include <array>
+#include <atomic>
+#include <cstddef>
+#include <cstring>
+#include <new>
+#include <span>
+#include <type_traits>
+#include <vector>
+
+namespace Common {
+
+/// SPSC ring buffer
+/// @tparam T            Element type
+/// @tparam capacity     Number of slots in ring buffer
+template <typename T, std::size_t capacity>
+class RingBuffer {
+    /// A "slot" is made of a single `T`.
+    static constexpr std::size_t slot_size = sizeof(T);
+    // T must be safely memcpy-able and have a trivial default constructor.
+    static_assert(std::is_trivial_v<T>);
+    // Ensure capacity is sensible.
+    static_assert(capacity < std::numeric_limits<std::size_t>::max() / 2);
+    static_assert((capacity & (capacity - 1)) == 0, "capacity must be a power of two");
+    // Ensure lock-free.
+    static_assert(std::atomic_size_t::is_always_lock_free);
+
+public:
+    /// Pushes slots into the ring buffer
+    /// @param new_slots   Pointer to the slots to push
+    /// @param slot_count  Number of slots to push
+    /// @returns The number of slots actually pushed
+    std::size_t push(const void* new_slots, std::size_t slot_count) {
+        const std::size_t write_index = m_write_index.load();
+        const std::size_t slots_free = capacity + m_read_index.load() - write_index;
+        const std::size_t push_count = std::min(slot_count, slots_free);
+
+        const std::size_t pos = write_index % capacity;
+        const std::size_t first_copy = std::min(capacity - pos, push_count);
+        const std::size_t second_copy = push_count - first_copy;
+
+        const char* in = static_cast<const char*>(new_slots);
+        std::memcpy(m_data.data() + pos, in, first_copy * slot_size);
+        in += first_copy * slot_size;
+        std::memcpy(m_data.data(), in, second_copy * slot_size);
+
+        m_write_index.store(write_index + push_count);
+
+        return push_count;
+    }
+
+    std::size_t push(std::span<const T> input) {
+        return push(input.data(), input.size());
+    }
+
+    /// Pops slots from the ring buffer
+    /// @param output     Where to store the popped slots
+    /// @param max_slots  Maximum number of slots to pop
+    /// @returns The number of slots actually popped
+    std::size_t pop(void* output, std::size_t max_slots = ~std::size_t(0)) {
+        const std::size_t read_index = m_read_index.load();
+        const std::size_t slots_filled = m_write_index.load() - read_index;
+        const std::size_t pop_count = std::min(slots_filled, max_slots);
+
+        const std::size_t pos = read_index % capacity;
+        const std::size_t first_copy = std::min(capacity - pos, pop_count);
+        const std::size_t second_copy = pop_count - first_copy;
+
+        char* out = static_cast<char*>(output);
+        std::memcpy(out, m_data.data() + pos, first_copy * slot_size);
+        out += first_copy * slot_size;
+        std::memcpy(out, m_data.data(), second_copy * slot_size);
+
+        m_read_index.store(read_index + pop_count);
+
+        return pop_count;
+    }
+
+    std::vector<T> pop(std::size_t max_slots = ~std::size_t(0)) {
+        std::vector<T> out(std::min(max_slots, capacity));
+        const std::size_t count = Pop(out.data(), out.size());
+        out.resize(count);
+        return out;
+    }
+
+    /// @returns Number of slots used
+    [[nodiscard]] std::size_t size() const {
+        return m_write_index.load() - m_read_index.load();
+    }
+
+    /// @returns Maximum size of ring buffer
+    [[nodiscard]] constexpr std::size_t Capacity() const {
+        return capacity;
+    }
+
+private:
+    // It is important to align the below variables for performance reasons:
+    // Having them on the same cache-line would result in false-sharing between them.
+    // TODO: Remove this ifdef whenever clang and GCC support
+    //       std::hardware_destructive_interference_size.
+#ifdef __cpp_lib_hardware_interference_size
+    alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_read_index{0};
+    alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_write_index{0};
+#else
+    alignas(128) std::atomic_size_t m_read_index{0};
+    alignas(128) std::atomic_size_t m_write_index{0};
+#endif
+
+    std::array<T, capacity> m_data;
+};
+
+} // namespace Common

src/core/audio/miniaudio_device.cpp

@@ -0,0 +1,119 @@
+#include "audio/miniaudio_device.hpp"
+
+#include "helpers.hpp"
+
+MiniAudioDevice::MiniAudioDevice() : initialized(false), running(false), samples(nullptr) {}
+
+void MiniAudioDevice::init(Samples& samples, bool safe) {
+	this->samples = &samples;
+
+	// Probe for device and available backends and initialize audio
+	ma_backend backends[ma_backend_null + 1];
+	unsigned count = 0;
+
+	if (safe) {
+		backends[0] = ma_backend_null;
+		count = 1;
+	} else {
+		bool found = false;
+		for (uint i = 0; i <= ma_backend_null; i++) {
+			ma_backend backend = ma_backend(i);
+			if (!ma_is_backend_enabled(backend)) {
+				continue;
+			}
+
+			backends[count++] = backend;
+			printf("Found audio backend: %s\n", ma_get_backend_name(backend));
+
+			// TODO: Make backend selectable here
+			found = true;
+			//count = 1;
+			//backends[0] = backend;
+		}
+
+		if (!found) {
+			initialized = false;
+
+			Helpers::warn("No valid audio backend found\n");
+			return;
+		}
+	}
+
+	if (ma_context_init(backends, count, nullptr, &context) != MA_SUCCESS) {
+		initialized = false;
+
+		Helpers::warn("Unable to initialize audio context");
+		return;
+	}
+	audioDevices.clear();
+
+	struct UserContext {
+		MiniAudioDevice* miniAudio;
+		ma_device_config& config;
+		bool found = false;
+	};
+	UserContext userContext = {.miniAudio = this, .config = deviceConfig};
+
+	ma_context_enumerate_devices(
+		&context,
+		[](ma_context* pContext, ma_device_type deviceType, const ma_device_info* pInfo, void* pUserData) -> ma_bool32 {
+			if (deviceType != ma_device_type_playback) {
+				return true;
+			}
+
+			UserContext* userContext = reinterpret_cast<UserContext*>(pUserData);
+			userContext->miniAudio->audioDevices.push_back(pInfo->name);
+
+			// TODO: Check if this is the device we want here
+			userContext->config.playback.pDeviceID = &pInfo->id;
+			userContext->found = true;
+			return true;
+		},
+		&userContext
+	);
+
+	if (!userContext.found) {
+		Helpers::warn("MiniAudio: Device not found");
+	}
+
+	deviceConfig = ma_device_config_init(ma_device_type_playback);
+	// The 3DS outputs s16 stereo audio @ 32768 Hz
+	deviceConfig.playback.format = ma_format_s16;
+	deviceConfig.playback.channels = 2;
+	deviceConfig.sampleRate = 32768;
+	//deviceConfig.periodSizeInFrames = 64;
+	//deviceConfig.periods = 2;
+	deviceConfig.pUserData = this;
+	deviceConfig.aaudio.usage = ma_aaudio_usage_game;
+	deviceConfig.wasapi.noAutoConvertSRC = true;
+
+	deviceConfig.dataCallback = [](ma_device* device, void* out, const void* input, ma_uint32 frameCount) {
+		auto self = reinterpret_cast<MiniAudioDevice*>(device->pUserData);
+		s16* output = reinterpret_cast<ma_int16*>(out);
+
+		self->samples->pop(output, frameCount);
+	};
+
+	if (ma_device_init(&context, &deviceConfig, &device) != MA_SUCCESS) {
+		Helpers::warn("Unable to initialize audio device");
+		initialized = false;
+		return;
+	}
+
+	initialized = true;
+	start();
+}
+
+void MiniAudioDevice::start() {
+	if (!initialized) {
+		Helpers::warn("MiniAudio device not initialize, won't start");
+		return;
+	}
+
+	if (ma_device_start(&device) == MA_SUCCESS) {
+		running = true;
+	} else {
+		running = false;
+		Helpers::warn("Failed to start audio device");
+	}
+}

src/core/audio/teakra_core.cpp

@@ -6,6 +6,10 @@
 #include "services/dsp.hpp"
 
 using namespace Audio;
+static constexpr u32 sampleRate = 32768;
+static constexpr u32 duration = 30;
+static s16 samples[sampleRate * duration * 2];
+static uint sampleIndex = 0;
 
 struct Dsp1 {
 	// All sizes are in bytes unless otherwise specified
@@ -51,10 +55,7 @@ TeakraDSP::TeakraDSP(Memory& mem, Scheduler& scheduler, DSPService& dspService)
 	ahbm.write32 = [&](u32 addr, u32 value) { *(u32*)&mem.getFCRAM()[addr - PhysicalAddrs::FCRAM] = value; };
 
 	teakra.SetAHBMCallback(ahbm);
-	teakra.SetAudioCallback([=](std::array<s16, 2> sample) {
+	teakra.SetAudioCallback([=](std::array<s16, 2> sample) { sampleBuffer.push(sample.data(), 2); });
-		//printf("%d %d\n", sample[0], sample[1]);
-		// NOP for now
-	});
 
 	// Set up event handlers. These handlers forward a hardware interrupt to the DSP service, which is responsible
 	// For triggering the appropriate DSP kernel events

src/emulator.cpp

@@ -25,10 +25,11 @@ Emulator::Emulator()
 #endif
 {
 	DSPService& dspService = kernel.getServiceManager().getDSP();
-
 	dsp = Audio::makeDSPCore(config.dspType, memory, scheduler, dspService);
 	dspService.setDSPCore(dsp.get());
 
+	audioDevice.init(dsp->getSamples());
+
 #ifdef PANDA3DS_ENABLE_DISCORD_RPC
 	if (config.discordRpcEnabled) {
 		discordRpc.init();

src/miniaudio.cpp

@@ -0,0 +1,7 @@
+// We do not need the ability to be able to encode or decode audio files for the time being
+// So we disable said functionality to make the executable smaller
+#define MA_NO_DECODING
+#define MA_NO_ENCODING
+#define MINIAUDIO_IMPLEMENTATION
+
+#include "miniaudio.h"

third_party/miniaudio

@@ -0,0 +1 @@
+Subproject commit 4a5b74bef029b3592c54b6048650ee5f972c1a48