Merge branch 'master' into specialized-shaderz

src/core/CPU/cpu_dynarmic.cpp

@@ -21,7 +21,6 @@ CPU::CPU(Memory& mem, Kernel& kernel, Emulator& emu) : mem(mem), emu(emu), sched
 void CPU::reset() {
 	setCPSR(CPSR::UserMode);
 	setFPSCR(FPSCR::MainThreadDefault);
-	env.totalTicks = 0;
 
 	cp15->reset();
 	cp15->setTLSBase(VirtualAddrs::TLSBase);  // Set cp15 TLS pointer to the main thread's thread-local storage

src/core/audio/dsp_core.cpp

@@ -0,0 +1,52 @@
+#include "audio/dsp_core.hpp"
+
+#include "audio/null_core.hpp"
+#include "audio/teakra_core.hpp"
+
+#include <algorithm>
+#include <cctype>
+#include <unordered_map>
+
+std::unique_ptr<Audio::DSPCore> Audio::makeDSPCore(DSPCore::Type type, Memory& mem, Scheduler& scheduler, DSPService& dspService) {
+	std::unique_ptr<DSPCore> core;
+
+	switch (type) {
+		case DSPCore::Type::Null: core = std::make_unique<NullDSP>(mem, scheduler, dspService); break;
+		case DSPCore::Type::Teakra: core = std::make_unique<TeakraDSP>(mem, scheduler, dspService); break;
+
+		default:
+			Helpers::warn("Invalid DSP core selected!");
+			core = std::make_unique<NullDSP>(mem, scheduler, dspService);
+			break;
+	}
+
+	mem.setDSPMem(core->getDspMemory());
+	return core;
+}
+
+Audio::DSPCore::Type Audio::DSPCore::typeFromString(std::string inString) {
+	// Transform to lower-case to make the setting case-insensitive
+	std::transform(inString.begin(), inString.end(), inString.begin(), [](unsigned char c) { return std::tolower(c); });
+
+	static const std::unordered_map<std::string, Audio::DSPCore::Type> map = {
+		{"null", Audio::DSPCore::Type::Null},
+		{"none", Audio::DSPCore::Type::Null},
+		{"lle", Audio::DSPCore::Type::Teakra},
+		{"teakra", Audio::DSPCore::Type::Teakra},
+	};
+
+	if (auto search = map.find(inString); search != map.end()) {
+		return search->second;
+	}
+
+	printf("Invalid DSP type. Defaulting to null\n");
+	return Audio::DSPCore::Type::Null;
+}
+
+const char* Audio::DSPCore::typeToString(Audio::DSPCore::Type type) {
+	switch (type) {
+		case Audio::DSPCore::Type::Null: return "null";
+		case Audio::DSPCore::Type::Teakra: return "teakra";
+		default: return "invalid";
+	}
+}

src/core/audio/miniaudio_device.cpp

@@ -0,0 +1,143 @@
+#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;
+	running = false;
+
+	// Probe for device and available backends and initialize audio
+	ma_backend backends[ma_backend_null + 1];
+	uint 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;
+
+			// 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 = channelCount;
+	deviceConfig.sampleRate = sampleRate;
+	//deviceConfig.periodSizeInFrames = 64;
+	//deviceConfig.periods = 16;
+	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);
+
+		// Wait until there's enough samples to pop
+		while (self->samples->size() < frameCount * channelCount) {
+			// If audio output is disabled from the emulator thread, make sure that this callback will return and not hang
+			if (!self->running) {
+				return;
+			}
+		}
+
+		self->samples->pop(output, frameCount * channelCount);
+	};
+
+	if (ma_device_init(&context, &deviceConfig, &device) != MA_SUCCESS) {
+		Helpers::warn("Unable to initialize audio device");
+		initialized = false;
+		return;
+	}
+
+	initialized = true;
+}
+
+void MiniAudioDevice::start() {
+	if (!initialized) {
+		Helpers::warn("MiniAudio device not initialized, won't start");
+		return;
+	}
+
+	// Ignore the call to start if the device is already running
+	if (!running) {
+		if (ma_device_start(&device) == MA_SUCCESS) {
+			running = true;
+		} else {
+			Helpers::warn("Failed to start audio device");
+		}
+	}
+}
+
+void MiniAudioDevice::stop() {
+	if (!initialized) {
+		Helpers::warn("MiniAudio device not initialized, can't start");
+		return;
+	}
+
+	if (running) {
+		running = false;
+
+		if (ma_device_stop(&device) != MA_SUCCESS) {
+			Helpers::warn("Failed to stop audio device");
+		} 
+	}
+}

src/core/audio/null_core.cpp

@@ -0,0 +1,166 @@
+#include "audio/null_core.hpp"
+
+#include "services/dsp.hpp"
+
+namespace Audio {
+	namespace DSPPipeType {
+		enum : u32 {
+			Debug = 0,
+			DMA = 1,
+			Audio = 2,
+			Binary = 3,
+		};
+	}
+
+	void NullDSP::resetAudioPipe() {
+		// Hardcoded responses for now
+		// These are DSP DRAM offsets for various variables
+		// https://www.3dbrew.org/wiki/DSP_Memory_Region
+		static constexpr std::array<u16, 16> responses = {
+			0x000F,  // Number of responses
+			0xBFFF,  // Frame counter
+			0x9E92,  // Source configs
+			0x8680,  // Source statuses
+			0xA792,  // ADPCM coefficients
+			0x9430,  // DSP configs
+			0x8400,  // DSP status
+			0x8540,  // Final samples
+			0x9492,  // Intermediate mix samples
+			0x8710,  // Compressor
+			0x8410,  // Debug
+			0xA912,  // ??
+			0xAA12,  // ??
+			0xAAD2,  // ??
+			0xAC52,  // Surround sound biquad filter 1
+			0xAC5C   // Surround sound biquad filter 2
+		};
+
+		std::vector<u8>& audioPipe = pipeData[DSPPipeType::Audio];
+		audioPipe.resize(responses.size() * sizeof(u16));
+
+		// Push back every response to the audio pipe
+		size_t index = 0;
+		for (auto e : responses) {
+			audioPipe[index++] = e & 0xff;
+			audioPipe[index++] = e >> 8;
+		}
+	}
+
+	void NullDSP::reset() {
+		loaded = false;
+		for (auto& e : pipeData) {
+			e.clear();
+		}
+
+		// Note: Reset audio pipe AFTER resetting all pipes, otherwise the new data will be yeeted
+		resetAudioPipe();
+	}
+
+	void NullDSP::loadComponent(std::vector<u8>& data, u32 programMask, u32 dataMask) {
+		if (loaded) {
+			Helpers::warn("Loading DSP component when already loaded");
+		}
+
+		loaded = true;
+		scheduler.addEvent(Scheduler::EventType::RunDSP, scheduler.currentTimestamp + Audio::cyclesPerFrame);
+	}
+
+	void NullDSP::unloadComponent() {
+		if (!loaded) {
+			Helpers::warn("Audio: unloadComponent called without a running program");
+		}
+
+		loaded = false;
+		scheduler.removeEvent(Scheduler::EventType::RunDSP);
+	}
+
+	void NullDSP::runAudioFrame() {
+		// Signal audio pipe when an audio frame is done
+		if (dspState == DSPState::On) [[likely]] {
+			dspService.triggerPipeEvent(DSPPipeType::Audio);
+		}
+
+		scheduler.addEvent(Scheduler::EventType::RunDSP, scheduler.currentTimestamp + Audio::cyclesPerFrame);
+	}
+	
+	u16 NullDSP::recvData(u32 regId) {
+		if (regId != 0) {
+			Helpers::panic("Audio: invalid register in null frontend");
+		}
+
+		return dspState == DSPState::On;
+	}
+
+	void NullDSP::writeProcessPipe(u32 channel, u32 size, u32 buffer) {
+		enum class StateChange : u8 {
+			Initialize = 0,
+			Shutdown = 1,
+			Wakeup = 2,
+			Sleep = 3,
+		};
+
+		switch (channel) {
+			case DSPPipeType::Audio: {
+				if (size != 4) {
+					printf("Invalid size written to DSP Audio Pipe\n");
+					break;
+				}
+
+				// Get new state
+				const u8 state = mem.read8(buffer);
+				if (state > 3) {
+					log("WriteProcessPipe::Audio: Unknown state change type");
+				} else {
+					switch (static_cast<StateChange>(state)) {
+						case StateChange::Initialize:
+							// TODO: Other initialization stuff here
+							dspState = DSPState::On;
+							resetAudioPipe();
+							
+							dspService.triggerPipeEvent(DSPPipeType::Audio);
+							break;
+
+						case StateChange::Shutdown:
+							dspState = DSPState::Off;
+							break;
+
+						default: Helpers::panic("Unimplemented DSP audio pipe state change %d", state);
+					}
+				}
+				break;
+			}
+
+			case DSPPipeType::Binary:
+				Helpers::warn("Unimplemented write to binary pipe! Size: %d\n", size);
+
+				// This pipe and interrupt are normally used for requests like AAC decode
+				dspService.triggerPipeEvent(DSPPipeType::Binary);
+				break;
+
+			default: log("Audio::NullDSP: Wrote to unimplemented pipe %d\n", channel); break;
+		}
+	}
+
+	std::vector<u8> NullDSP::readPipe(u32 pipe, u32 peer, u32 size, u32 buffer) {
+		if (size & 1) Helpers::panic("Tried to read odd amount of bytes from DSP pipe");
+		if (pipe >= pipeCount || size > 0xffff) {
+			return {};
+		}
+
+		if (pipe != DSPPipeType::Audio) {
+			log("Reading from non-audio pipe! This might be broken, might need to check what pipe is being read from and implement writing to it\n");
+		}
+
+		std::vector<u8>& data = pipeData[pipe];
+		size = std::min<u32>(size, data.size());  // Clamp size to the maximum available data size
+
+		if (size == 0) {
+			return {};
+		}
+
+		// Return "size" bytes from the audio pipe and erase them
+		std::vector<u8> out(data.begin(), data.begin() + size);
+		data.erase(data.begin(), data.begin() + size);
+		return out;
+	}
+}  // namespace Audio

src/core/audio/teakra_core.cpp

@@ -0,0 +1,345 @@
+#include "audio/teakra_core.hpp"
+
+#include <algorithm>
+#include <chrono>
+#include <cstring>
+#include <thread>
+
+#include "services/dsp.hpp"
+
+using namespace Audio;
+
+struct Dsp1 {
+	// All sizes are in bytes unless otherwise specified
+	u8 signature[0x100];
+	u8 magic[4];
+	u32 size;
+	u8 codeMemLayout;
+	u8 dataMemLayout;
+	u8 pad[3];
+	u8 specialType;
+	u8 segmentCount;
+	u8 flags;
+	u32 specialStart;
+	u32 specialSize;
+	u64 zeroBits;
+
+	struct Segment {
+		u32 offs;     // Offset of the segment data
+		u32 dspAddr;  // Start of the segment in 16-bit units
+		u32 size;
+		u8 pad[3];
+		u8 type;
+		u8 hash[0x20];
+	};
+
+	Segment segments[10];
+};
+
+TeakraDSP::TeakraDSP(Memory& mem, Scheduler& scheduler, DSPService& dspService)
+	: DSPCore(mem, scheduler, dspService), pipeBaseAddr(0), running(false) {
+	// Set up callbacks for Teakra
+	Teakra::AHBMCallback ahbm;
+
+	// The AHBM read handlers read from paddrs rather than vaddrs which mem.read8 and the like use
+	// TODO: When we implement more efficient paddr accesses with a page table or similar, these handlers
+	// Should be made to properly use it, since this method is hacky and will segfault if given an invalid addr
+	ahbm.read8 = [&](u32 addr) -> u8 { return mem.getFCRAM()[addr - PhysicalAddrs::FCRAM]; };
+	ahbm.read16 = [&](u32 addr) -> u16 { return *(u16*)&mem.getFCRAM()[addr - PhysicalAddrs::FCRAM]; };
+	ahbm.read32 = [&](u32 addr) -> u32 { return *(u32*)&mem.getFCRAM()[addr - PhysicalAddrs::FCRAM]; };
+
+	ahbm.write8 = [&](u32 addr, u8 value) { mem.getFCRAM()[addr - PhysicalAddrs::FCRAM] = value; };
+	ahbm.write16 = [&](u32 addr, u16 value) { *(u16*)&mem.getFCRAM()[addr - PhysicalAddrs::FCRAM] = value; };
+	ahbm.write32 = [&](u32 addr, u32 value) { *(u32*)&mem.getFCRAM()[addr - PhysicalAddrs::FCRAM] = value; };
+
+	teakra.SetAHBMCallback(ahbm);
+	teakra.SetAudioCallback([](std::array<s16, 2> sample) { /* Do nothing */ });
+
+	// Set up event handlers. These handlers forward a hardware interrupt to the DSP service, which is responsible
+	// For triggering the appropriate DSP kernel events
+	// Note: It's important not to fire any events if "loaded" is false, ie if we haven't fully loaded a DSP component yet
+	teakra.SetRecvDataHandler(0, [&]() {
+		if (loaded) {
+			dspService.triggerInterrupt0();
+		}
+	});
+
+	teakra.SetRecvDataHandler(1, [&]() {
+		if (loaded) {
+			dspService.triggerInterrupt1();
+		}
+	});
+
+	auto processPipeEvent = [&](bool dataEvent) {
+		if (!loaded) {
+			return;
+		}
+
+		if (dataEvent) {
+			signalledData = true;
+		} else {
+			if ((teakra.GetSemaphore() & 0x8000) == 0) {
+				return;
+			}
+
+			signalledSemaphore = true;
+		}
+
+		if (signalledSemaphore && signalledData) {
+			signalledSemaphore = signalledData = false;
+
+			u16 slot = teakra.RecvData(2);
+			u16 side = slot & 1;
+			u16 pipe = slot / 2;
+
+			if (side != static_cast<u16>(PipeDirection::DSPtoCPU)) {
+				return;
+			}
+
+			if (pipe == 0) {
+				Helpers::warn("Pipe event for debug pipe: Should be ignored and the data should be flushed");
+			} else {
+				dspService.triggerPipeEvent(pipe);
+			}
+		}
+	};
+
+	teakra.SetRecvDataHandler(2, [processPipeEvent]() { processPipeEvent(true); });
+	teakra.SetSemaphoreHandler([processPipeEvent]() { processPipeEvent(false); });
+}
+
+void TeakraDSP::reset() {
+	teakra.Reset();
+	running = false;
+	loaded = false;
+	signalledData = signalledSemaphore = false;
+
+	audioFrameIndex = 0;
+}
+
+void TeakraDSP::setAudioEnabled(bool enable) {
+	if (audioEnabled != enable) {
+		audioEnabled = enable;
+
+		// Set the appropriate audio callback for Teakra
+		if (audioEnabled) {
+			teakra.SetAudioCallback([this](std::array<s16, 2> sample) {
+				audioFrame[audioFrameIndex++] = sample[0];
+				audioFrame[audioFrameIndex++] = sample[1];
+
+				// Push our samples at the end of an audio frame
+				if (audioFrameIndex >= audioFrame.size()) {
+					audioFrameIndex -= audioFrame.size();
+
+					// Wait until we've actually got room to do so
+					while (sampleBuffer.size() + 2 > sampleBuffer.Capacity()) {
+						std::this_thread::sleep_for(std::chrono::milliseconds{1});
+					}
+
+					sampleBuffer.push(audioFrame.data(), audioFrame.size());
+				}
+			});
+		} else {
+			teakra.SetAudioCallback([](std::array<s16, 2> sample) { /* Do nothing */ });
+		}
+	}
+}
+
+// https://github.com/citra-emu/citra/blob/master/src/audio_core/lle/lle.cpp
+void TeakraDSP::writeProcessPipe(u32 channel, u32 size, u32 buffer) {
+	size &= 0xffff;
+
+	PipeStatus status = getPipeStatus(channel, PipeDirection::CPUtoDSP);
+	bool needUpdate = false;  // Do we need to update the pipe status and catch up Teakra?
+
+	std::vector<u8> data;
+	data.reserve(size);
+
+	// Read data to write
+	for (int i = 0; i < size; i++) {
+		const u8 byte = mem.read8(buffer + i);
+		data.push_back(byte);
+	}
+	u8* dataPointer = data.data();
+
+	while (size != 0) {
+		if (status.isFull()) {
+			Helpers::warn("Teakra: Writing to full pipe");
+		}
+
+		// Calculate begin/end/size for write
+		const u16 writeEnd = status.isWrapped() ? (status.readPointer & PipeStatus::pointerMask) : status.byteSize;
+		const u16 writeBegin = status.writePointer & PipeStatus::pointerMask;
+		const u16 writeSize = std::min<u16>(u16(size), writeEnd - writeBegin);
+
+		if (writeEnd <= writeBegin) [[unlikely]] {
+			Helpers::warn("Teakra: Writing to pipe but end <= start");
+		}
+
+		// Write data to pipe, increment write and buffer pointers, decrement size
+		std::memcpy(getDataPointer(status.address * 2 + writeBegin), dataPointer, writeSize);
+		dataPointer += writeSize;
+		status.writePointer += writeSize;
+		size -= writeSize;
+
+		if ((status.writePointer & PipeStatus::pointerMask) > status.byteSize) [[unlikely]] {
+			Helpers::warn("Teakra: Writing to pipe but write > size");
+		}
+
+		if ((status.writePointer & PipeStatus::pointerMask) == status.byteSize) {
+			status.writePointer &= PipeStatus::wrapBit;
+			status.writePointer ^= PipeStatus::wrapBit;
+		}
+		needUpdate = true;
+	}
+
+	if (needUpdate) {
+		updatePipeStatus(status);
+		while (!teakra.SendDataIsEmpty(2)) {
+			runSlice();
+		}
+
+		teakra.SendData(2, status.slot);
+	}
+}
+
+std::vector<u8> TeakraDSP::readPipe(u32 channel, u32 peer, u32 size, u32 buffer) {
+	size &= 0xffff;
+
+	PipeStatus status = getPipeStatus(channel, PipeDirection::DSPtoCPU);
+
+	std::vector<u8> pipeData(size);
+	u8* dataPointer = pipeData.data();
+	bool needUpdate = false;  // Do we need to update the pipe status and catch up Teakra?
+
+	while (size != 0) {
+		if (status.isEmpty()) [[unlikely]] {
+			Helpers::warn("Teakra: Reading from empty pipe");
+			return pipeData;
+		}
+
+		// Read as many bytes as possible
+		const u16 readEnd = status.isWrapped() ? status.byteSize : (status.writePointer & PipeStatus::pointerMask);
+		const u16 readBegin = status.readPointer & PipeStatus::pointerMask;
+		const u16 readSize = std::min<u16>(u16(size), readEnd - readBegin);
+
+		// Copy bytes to the output vector, increment the read and vector pointers and decrement the size appropriately
+		std::memcpy(dataPointer, getDataPointer(status.address * 2 + readBegin), readSize);
+		dataPointer += readSize;
+		status.readPointer += readSize;
+		size -= readSize;
+
+		if ((status.readPointer & PipeStatus::pointerMask) > status.byteSize) [[unlikely]] {
+			Helpers::warn("Teakra: Reading from pipe but read > size");
+		}
+
+		if ((status.readPointer & PipeStatus::pointerMask) == status.byteSize) {
+			status.readPointer &= PipeStatus::wrapBit;
+			status.readPointer ^= PipeStatus::wrapBit;
+		}
+
+		needUpdate = true;
+	}
+
+	if (needUpdate) {
+		updatePipeStatus(status);
+		while (!teakra.SendDataIsEmpty(2)) {
+			runSlice();
+		}
+
+		teakra.SendData(2, status.slot);
+	}
+
+	return pipeData;
+}
+
+void TeakraDSP::loadComponent(std::vector<u8>& data, u32 programMask, u32 dataMask) {
+	// TODO: maybe move this to the DSP service
+	if (loaded) {
+		Helpers::warn("Loading DSP component when already loaded");
+		return;
+	}
+
+	teakra.Reset();
+	running = true;
+
+	u8* dspCode = teakra.GetDspMemory().data();
+	u8* dspData = dspCode + 0x40000;
+
+	Dsp1 dsp1;
+	std::memcpy(&dsp1, data.data(), sizeof(dsp1));
+
+	// TODO: verify DSP1 signature
+
+	// Load DSP segments to DSP RAM
+	// TODO: verify hashes
+	for (uint i = 0; i < dsp1.segmentCount; i++) {
+		auto& segment = dsp1.segments[i];
+		u32 addr = segment.dspAddr << 1;
+		u8* src = data.data() + segment.offs;
+		u8* dst = nullptr;
+
+		switch (segment.type) {
+			case 0:
+			case 1: dst = dspCode + addr; break;
+			default: dst = dspData + addr; break;
+		}
+
+		std::memcpy(dst, src, segment.size);
+	}
+
+	bool syncWithDsp = dsp1.flags & 0x1;
+	bool loadSpecialSegment = (dsp1.flags >> 1) & 0x1;
+
+	// TODO: how does the special segment work?
+	if (loadSpecialSegment) {
+		log("LoadComponent: special segment not supported");
+	}
+
+	if (syncWithDsp) {
+		// Wait for the DSP to reply with 1s in all RECV registers
+		for (int i = 0; i < 3; i++) {
+			do {
+				while (!teakra.RecvDataIsReady(i)) {
+					runSlice();
+				}
+			} while (teakra.RecvData(i) != 1);
+		}
+	}
+
+	// Retrieve the pipe base address
+	while (!teakra.RecvDataIsReady(2)) {
+		runSlice();
+	}
+	pipeBaseAddr = teakra.RecvData(2);
+	
+	// Schedule next DSP event
+	scheduler.addEvent(Scheduler::EventType::RunDSP, scheduler.currentTimestamp + Audio::lleSlice * 2);
+	loaded = true;
+}
+
+void TeakraDSP::unloadComponent() {
+	if (!loaded) {
+		Helpers::warn("Audio: unloadComponent called without a running program");
+		return;
+	}
+	loaded = false;
+	// Stop scheduling DSP events
+	scheduler.removeEvent(Scheduler::EventType::RunDSP);
+
+	// Wait for SEND2 to be ready, then send the shutdown command to the DSP
+	while (!teakra.SendDataIsEmpty(2)) {
+		runSlice();
+	}
+
+	teakra.SendData(2, 0x8000);
+
+	// Wait for shutdown to be acknowledged
+	while (!teakra.RecvDataIsReady(2)) {
+		runSlice();
+	}
+
+	// Read the value and discard it, completing shutdown
+	teakra.RecvData(2);
+	running = false;
+}

src/core/kernel/events.cpp

@@ -12,9 +12,9 @@ const char* Kernel::resetTypeToString(u32 type) {
 	}
 }
 
-Handle Kernel::makeEvent(ResetType resetType) {
+Handle Kernel::makeEvent(ResetType resetType, Event::CallbackType callback) {
 	Handle ret = makeObject(KernelObjectType::Event);
-	objects[ret].data = new Event(resetType);
+	objects[ret].data = new Event(resetType, callback);
 	return ret;
 }
 
@@ -42,8 +42,13 @@ bool Kernel::signalEvent(Handle handle) {
 			event->fired = false;
 		}
 	}
-	
+
 	rescheduleThreads();
+	// Run the callback for events that require a special callback
+	if (event->callback != Event::CallbackType::None) [[unlikely]] {
+		runEventCallback(event->callback);
+	}
+
 	return true;
 }
 
@@ -230,4 +235,12 @@ void Kernel::waitSynchronizationN() {
 	} else {
 		Helpers::panic("WaitSynchronizationN with waitAll");
 	}
+}
+
+void Kernel::runEventCallback(Event::CallbackType callback) {
+	switch (callback) {
+		case Event::CallbackType::None: break;
+		case Event::CallbackType::DSPSemaphore: serviceManager.getDSP().onSemaphoreEventSignal(); break;
+		default: Helpers::panic("Unimplemented special callback for kernel event!"); break;
+	}
 }

src/core/loader/ncch.cpp

@@ -32,7 +32,7 @@ bool NCCH::loadFromHeader(Crypto::AESEngine &aesEngine, IOFile& file, const FSIn
 	size = u64(*(u32*)&header[0x104]) * mediaUnit; // TODO: Maybe don't type pun because big endian will break
 	exheaderSize = *(u32*)&header[0x180];
 
-	const u64 programID = *(u64*)&header[0x118];
+	programID = *(u64*)&header[0x118];
 
 	// Read NCCH flags
 	secondaryKeySlot = header[0x188 + 3];

src/core/memory.cpp

@@ -15,7 +15,6 @@ using namespace KernelMemoryTypes;
 
 Memory::Memory(u64& cpuTicks, const EmulatorConfig& config) : cpuTicks(cpuTicks), config(config) {
 	fcram = new uint8_t[FCRAM_SIZE]();
-	dspRam = new uint8_t[DSP_RAM_SIZE]();
 
 	readTable.resize(totalPageCount, 0);
 	writeTable.resize(totalPageCount, 0);
@@ -525,4 +524,14 @@ void Memory::copySharedFont(u8* pointer) {
 	auto fonts = cmrc::ConsoleFonts::get_filesystem();
 	auto font = fonts.open("CitraSharedFontUSRelocated.bin");
 	std::memcpy(pointer, font.begin(), font.size());
+}
+
+std::optional<u64> Memory::getProgramID() {
+	auto cxi = getCXI();
+
+	if (cxi) {
+		return cxi->programID;
+	}
+
+	return std::nullopt;
 }

src/core/services/apt.cpp

@@ -84,7 +84,7 @@ void APTService::appletUtility(u32 messagePointer) {
 	u32 outputSize = mem.read32(messagePointer + 12);
 	u32 inputPointer = mem.read32(messagePointer + 20);
 
-	log("APT::AppletUtility(utility = %d, input size = %x, output size = %x, inputPointer = %08X) (Stubbed)\n", utility, inputSize, outputSize,
+	log("APT::AppletUtility(utility = %d, input size = %x, output size = %x, inputPointer = %08X)\n", utility, inputSize, outputSize,
 		inputPointer);
 
 	std::vector<u8> out(outputSize);
@@ -218,7 +218,7 @@ void APTService::initialize(u32 messagePointer) {
 }
 
 void APTService::inquireNotification(u32 messagePointer) {
-	log("APT::InquireNotification (STUBBED TO RETURN NONE)\n");
+	log("APT::InquireNotification\n");
 
 	mem.write32(messagePointer, IPC::responseHeader(0xB, 2, 0));
 	mem.write32(messagePointer + 4, Result::Success);

src/core/services/dsp.cpp

@@ -31,13 +31,8 @@ namespace Result {
 }
 
 void DSPService::reset() {
-	for (auto& e : pipeData)
-		e.clear();
-
-	// Note: Reset audio pipe AFTER resetting all pipes, otherwise the new data will be yeeted
-	resetAudioPipe();
 	totalEventCount = 0;
-	dspState = DSPState::Off;
+	semaphoreMask = 0;
 
 	semaphoreEvent = std::nullopt;
 	interrupt0 = std::nullopt;
@@ -48,40 +43,6 @@ void DSPService::reset() {
 	}
 }
 
-void DSPService::resetAudioPipe() {
-	// Hardcoded responses for now
-	// These are DSP DRAM offsets for various variables
-	// https://www.3dbrew.org/wiki/DSP_Memory_Region
-	static constexpr std::array<u16, 16> responses = {
-		0x000F, // Number of responses
-		0xBFFF, // Frame counter
-		0x9E92, // Source configs
-		0x8680, // Source statuses
-		0xA792, // ADPCM coefficients
-		0x9430, // DSP configs
-		0x8400, // DSP status
-		0x8540, // Final samples
-		0x9492, // Intermediate mix samples
-		0x8710, // Compressor
-		0x8410, // Debug
-		0xA912, // ??
-		0xAA12, // ??
-		0xAAD2, // ??
-		0xAC52, // Surround sound biquad filter 1
-		0xAC5C  // Surround sound biquad filter 2
-	};
-
-	std::vector<u8>& audioPipe = pipeData[DSPPipeType::Audio];
-	audioPipe.resize(responses.size() * sizeof(u16));
-
-	// Push back every response to the audio pipe
-	size_t index = 0;
-	for (auto e : responses) {
-		audioPipe[index++] = e & 0xff;
-		audioPipe[index++] = e >> 8;
-	}
-}
-
 void DSPService::handleSyncRequest(u32 messagePointer) {
 	const u32 command = mem.read32(messagePointer);
 	switch (command) {
@@ -117,8 +78,16 @@ void DSPService::loadComponent(u32 messagePointer) {
 	u32 size = mem.read32(messagePointer + 4);
 	u32 programMask = mem.read32(messagePointer + 8);
 	u32 dataMask = mem.read32(messagePointer + 12);
+	u32 buffer = mem.read32(messagePointer + 20);
+
+	std::vector<u8> data(size);
+	for (u32 i = 0; i < size; i++) {
+		data[i] = mem.read8(buffer + i);
+	}
 
 	log("DSP::LoadComponent (size = %08X, program mask = %X, data mask = %X\n", size, programMask, dataMask);
+	dsp->loadComponent(data, programMask, dataMask);
+
 	mem.write32(messagePointer, IPC::responseHeader(0x11, 2, 2));
 	mem.write32(messagePointer + 4, Result::Success);
 	mem.write32(messagePointer + 8, 1); // Component loaded
@@ -128,32 +97,12 @@ void DSPService::loadComponent(u32 messagePointer) {
 
 void DSPService::unloadComponent(u32 messagePointer) {
 	log("DSP::UnloadComponent\n");
+	dsp->unloadComponent();
+
 	mem.write32(messagePointer, IPC::responseHeader(0x12, 1, 0));
 	mem.write32(messagePointer + 4, Result::Success);
 }
 
-std::vector<u8> DSPService::readPipe(u32 pipe, u32 size) {
-	if (size & 1) Helpers::panic("Tried to read odd amount of bytes from DSP pipe");
-	if (pipe >= pipeCount || size > 0xffff) {
-		return {};
-	}
-
-	if (pipe != DSPPipeType::Audio) {
-		log("Reading from non-audio pipe! This might be broken, might need to check what pipe is being read from and implement writing to it\n");
-	}
-
-	std::vector<u8>& data = pipeData[pipe];
-	size = std::min<u32>(size, u32(data.size())); // Clamp size to the maximum available data size
-
-	if (size == 0)
-		return {};
-
-	// Return "size" bytes from the audio pipe and erase them
-	std::vector<u8> out(data.begin(), data.begin() + size);
-	data.erase(data.begin(), data.begin() + size);
-	return out;
-}
-
 void DSPService::readPipeIfPossible(u32 messagePointer) {
 	u32 channel = mem.read32(messagePointer + 4);
 	u32 peer = mem.read32(messagePointer + 8);
@@ -162,7 +111,7 @@ void DSPService::readPipeIfPossible(u32 messagePointer) {
 	log("DSP::ReadPipeIfPossible (channel = %d, peer = %d, size = %04X, buffer = %08X)\n", channel, peer, size, buffer);
 	mem.write32(messagePointer, IPC::responseHeader(0x10, 2, 2));
 
-	std::vector<u8> data = readPipe(channel, size);
+	std::vector<u8> data = dsp->readPipe(channel, peer, size, buffer);
 	for (uint i = 0; i < data.size(); i++) {
 		mem.write8(buffer + i, data[i]);
 	}
@@ -176,22 +125,22 @@ void DSPService::recvData(u32 messagePointer) {
 	log("DSP::RecvData (register = %d)\n", registerIndex);
 	if (registerIndex != 0) Helpers::panic("Unknown register in DSP::RecvData");
 
-	// Return 0 if the DSP is running, otherwise 1
-	const u16 ret = dspState == DSPState::On ? 0 : 1;
+	const u16 data = dsp->recvData(registerIndex);
 
 	mem.write32(messagePointer, IPC::responseHeader(0x01, 2, 0));
 	mem.write32(messagePointer + 4, Result::Success);
-	mem.write16(messagePointer + 8, ret);
+	mem.write16(messagePointer + 8, data);
 }
 
 void DSPService::recvDataIsReady(u32 messagePointer) {
 	const u32 registerIndex = mem.read32(messagePointer + 4);
 	log("DSP::RecvDataIsReady (register = %d)\n", registerIndex);
-	if (registerIndex != 0) Helpers::panic("Unknown register in DSP::RecvDataIsReady");
+
+	bool isReady = dsp->recvDataIsReady(registerIndex);
 
 	mem.write32(messagePointer, IPC::responseHeader(0x02, 2, 0));
 	mem.write32(messagePointer + 4, Result::Success);
-	mem.write32(messagePointer + 8, 1); // Always return that the register is ready for now
+	mem.write32(messagePointer + 8, isReady ? 1 : 0);
 }
 
 DSPService::DSPEvent& DSPService::getEventRef(u32 type, u32 pipe) {
@@ -236,7 +185,6 @@ void DSPService::registerInterruptEvents(u32 messagePointer) {
 			mem.write32(messagePointer + 4, Result::Success);
 
 			totalEventCount++;
-			kernel.signalEvent(eventHandle);
 		}
 	}
 }
@@ -253,7 +201,7 @@ void DSPService::getSemaphoreEventHandle(u32 messagePointer) {
 	log("DSP::GetSemaphoreEventHandle\n");
 
 	if (!semaphoreEvent.has_value()) {
-		semaphoreEvent = kernel.makeEvent(ResetType::OneShot);
+		semaphoreEvent = kernel.makeEvent(ResetType::OneShot, Event::CallbackType::DSPSemaphore);
 	}
 
 	mem.write32(messagePointer, IPC::responseHeader(0x16, 1, 2));
@@ -267,6 +215,7 @@ void DSPService::setSemaphore(u32 messagePointer) {
 	const u16 value = mem.read16(messagePointer + 4);
 	log("DSP::SetSemaphore(value = %04X)\n", value);
 
+	dsp->setSemaphore(value);
 	mem.write32(messagePointer, IPC::responseHeader(0x7, 1, 0));
 	mem.write32(messagePointer + 4, Result::Success);
 }
@@ -275,6 +224,9 @@ void DSPService::setSemaphoreMask(u32 messagePointer) {
 	const u16 mask = mem.read16(messagePointer + 4);
 	log("DSP::SetSemaphoreMask(mask = %04X)\n", mask);
 
+	dsp->setSemaphoreMask(mask);
+	semaphoreMask = mask;
+
 	mem.write32(messagePointer, IPC::responseHeader(0x17, 1, 0));
 	mem.write32(messagePointer + 4, Result::Success);
 }
@@ -285,51 +237,7 @@ void DSPService::writeProcessPipe(u32 messagePointer) {
 	const u32 buffer = mem.read32(messagePointer + 16);
 	log("DSP::writeProcessPipe (channel = %d, size = %X, buffer = %08X)\n", channel, size, buffer);
 
-	enum class StateChange : u8 {
-		Initialize = 0,
-		Shutdown = 1,
-		Wakeup = 2,
-		Sleep = 3,
-	};
-
-	switch (channel) {
-		case DSPPipeType::Audio: {
-			if (size != 4) {
-				printf("Invalid size written to DSP Audio Pipe\n");
-				break;
-			}
-
-			// Get new state
-			const u8 state = mem.read8(buffer);
-			if (state > 3) {
-				log("WriteProcessPipe::Audio: Unknown state change type");
-			} else {
-				switch (static_cast<StateChange>(state)) {
-					case StateChange::Initialize:
-						// TODO: Other initialization stuff here
-						dspState = DSPState::On;
-						resetAudioPipe();
-						break;
-
-					case StateChange::Shutdown:
-						dspState = DSPState::Off;
-						break;
-
-					default: Helpers::panic("Unimplemented DSP audio pipe state change %d", state);
-				}
-			}
-			break;
-		}
-
-		case DSPPipeType::Binary:
-			Helpers::warn("Unimplemented write to binary pipe! Size: %d\n", size);
-			break;
-
-		default:
-			log("DSP: Wrote to unimplemented pipe %d\n", channel);
-			break;
-	}
-
+	dsp->writeProcessPipe(channel, size, buffer);
 	mem.write32(messagePointer, IPC::responseHeader(0xD, 1, 0));
 	mem.write32(messagePointer + 4, Result::Success);
 }
@@ -354,12 +262,26 @@ void DSPService::invalidateDCache(u32 messagePointer) {
 	mem.write32(messagePointer + 4, Result::Success);
 }
 
-void DSPService::signalEvents() {
-	for (const DSPEvent& e : pipeEvents) {
-		if (e.has_value()) { kernel.signalEvent(e.value()); }
+void DSPService::triggerPipeEvent(int index) {
+	if (index < pipeCount && pipeEvents[index].has_value()) {
+		kernel.signalEvent(*pipeEvents[index]);
 	}
+}
 
-	if (semaphoreEvent.has_value()) { kernel.signalEvent(semaphoreEvent.value()); }
-	if (interrupt0.has_value()) { kernel.signalEvent(interrupt0.value()); }
-	if (interrupt1.has_value()) { kernel.signalEvent(interrupt1.value()); }
+void DSPService::triggerSemaphoreEvent() {
+	if (semaphoreEvent.has_value()) {
+		kernel.signalEvent(*semaphoreEvent);
+	}
+}
+
+void DSPService::triggerInterrupt0() {
+	if (interrupt0.has_value()) {
+		kernel.signalEvent(*interrupt0);
+	}
+}
+
+void DSPService::triggerInterrupt1() {
+	if (interrupt1.has_value()) {
+		kernel.signalEvent(*interrupt1);
+	}
 }

src/core/services/gsp_gpu.cpp

@@ -123,10 +123,6 @@ void GPUService::registerInterruptRelayQueue(u32 messagePointer) {
 }
 
 void GPUService::requestInterrupt(GPUInterrupt type) {
-	// HACK: Signal DSP events on GPU interrupt for now until we have the DSP since games need DSP events
-	// Maybe there's a better alternative?
-	kernel.signalDSPEvents();
-
 	if (sharedMem == nullptr) [[unlikely]] { // Shared memory hasn't been set up yet
 		return;
 	}
@@ -289,7 +285,7 @@ void GPUService::setBufferSwap(u32 messagePointer) {
 	info.displayFb = mem.read32(messagePointer + 28);  // Selects either framebuffer A or B
 
 	log("GSP::GPU::SetBufferSwap\n");
-	Helpers::panic("Untested GSP::GPU::SetBufferSwap call");
+	Helpers::warn("Untested GSP::GPU::SetBufferSwap call");
 
 	setBufferSwapImpl(screenId, info);
 	mem.write32(messagePointer, IPC::responseHeader(0x05, 1, 0));
@@ -529,4 +525,4 @@ void GPUService::importDisplayCaptureInfo(u32 messagePointer) {
 	mem.write32(messagePointer + 28, bottomScreenCapture.rightFramebuffer);
 	mem.write32(messagePointer + 32, bottomScreenCapture.format);
 	mem.write32(messagePointer + 36, bottomScreenCapture.stride);
-}
+}

src/core/services/nwm_uds.cpp

@@ -37,7 +37,8 @@ void NwmUdsService::initializeWithVersion(u32 messagePointer) {
 
 	initialized = true;
 
-	mem.write32(messagePointer + 4, Result::Success);
+	// Stubbed to fail temporarily, since some games will break trying to establish networks otherwise
+	mem.write32(messagePointer + 4, Result::FailurePlaceholder);
 	mem.write32(messagePointer + 8, 0);
 	mem.write32(messagePointer + 12, eventHandle.value());
 }