HLE DSP: Initial mixer work

include/audio/dsp_shared_mem.hpp

@@ -324,8 +324,8 @@ namespace Audio::HLE {
 			BitField<15, 1, u32> outputBufferCountDirty;
 			BitField<16, 1, u32> masterVolumeDirty;
 
-			BitField<24, 1, u32> auxReturnVolume0Dirty;
-			BitField<25, 1, u32> auxReturnVolume1Dirty;
+			BitField<24, 1, u32> auxVolume0Dirty;
+			BitField<25, 1, u32> auxVolume1Dirty;
 			BitField<26, 1, u32> outputFormatDirty;
 			BitField<27, 1, u32> clippingModeDirty;
 			BitField<28, 1, u32> headphonesConnectedDirty;
@@ -337,7 +337,7 @@ namespace Audio::HLE {
 		/// The DSP has three intermediate audio mixers. This controls the volume level (0.0-1.0) for
 		/// each at the final mixer.
 		float_le masterVolume;
-		std::array<float_le, 2> auxReturnVolume;
+		std::array<float_le, 2> auxVolumes;
 
 		u16_le outputBufferCount;
 		u16 pad1[2];
@@ -422,7 +422,7 @@ namespace Audio::HLE {
 
 	struct DspStatus {
 		u16_le unknown;
-		u16_le dropped_frames;
+		u16_le droppedFrames;
 		u16 pad0[0xE];
 	};
 	ASSERT_DSP_STRUCT(DspStatus, 32);

include/audio/hle_core.hpp

@@ -95,8 +95,7 @@ namespace Audio {
 		DSPSource() { reset(); }
 	};
 
-	class HLE_DSP : public DSPCore {
-		// The audio frame types are public in case we want to use them for unit tests
+	class DSPMixer {
 	  public:
 		template <typename T, usize channelCount = 1>
 		using Sample = std::array<T, channelCount>;
@@ -113,6 +112,43 @@ namespace Audio {
 		template <typename T>
 		using QuadFrame = Frame<T, 4>;
 
+	  private:
+		using ChannelFormat = HLE::DspConfiguration::OutputFormat;
+		// The audio from each DSP voice is converted to quadraphonic and then fed into 3 intermediate mixing stages
+		// Two of these intermediate mixers (second and third) are used for effects, including custom effects done on the CPU
+		static constexpr usize mixerStageCount = 3;
+
+	  public:
+		ChannelFormat channelFormat = ChannelFormat::Stereo;
+		std::array<float, mixerStageCount> volumes;
+		std::array<bool, 2> enableAuxStages;
+
+		void reset() {
+			channelFormat = ChannelFormat::Stereo;
+
+			volumes.fill(0.0);
+			enableAuxStages.fill(false);
+		}
+	};
+
+	class HLE_DSP : public DSPCore {
+		// The audio frame types are public in case we want to use them for unit tests
+	  public:
+		template <typename T, usize channelCount = 1>
+		using Sample = DSPMixer::Sample<T, channelCount>;
+
+		template <typename T, usize channelCount>
+		using Frame = DSPMixer::Frame<T, channelCount>;
+
+		template <typename T>
+		using MonoFrame = DSPMixer::MonoFrame<T>;
+
+		template <typename T>
+		using StereoFrame = DSPMixer::StereoFrame<T>;
+
+		template <typename T>
+		using QuadFrame = DSPMixer::QuadFrame<T>;
+
 		using Source = Audio::DSPSource;
 		using SampleBuffer = Source::SampleBuffer;
 
@@ -131,6 +167,7 @@ namespace Audio {
 		std::array<Source, Audio::HLE::sourceCount> sources;  // DSP voices
 		Audio::HLE::DspMemory dspRam;
 
+		Audio::DSPMixer mixer;
 		std::unique_ptr<Audio::AAC::Decoder> aacDecoder;
 
 		void resetAudioPipe();
@@ -175,10 +212,13 @@ namespace Audio {
 
 		void handleAACRequest(const AAC::Message& request);
 		void updateSourceConfig(Source& source, HLE::SourceConfiguration::Configuration& config, s16_le* adpcmCoefficients);
+		void updateMixerConfig(HLE::SharedMemory& sharedMem);
 		void generateFrame(StereoFrame<s16>& frame);
 		void generateFrame(DSPSource& source);
 		void outputFrame();
-
+		// Perform the final mix, mixing the quadraphonic samples from all voices into the output audio frame
+		void performMix(Audio::HLE::SharedMemory& readRegion, Audio::HLE::SharedMemory& writeRegion);
+
 		// Decode an entire buffer worth of audio
 		void decodeBuffer(DSPSource& source);
 

src/core/audio/hle_core.cpp

@@ -76,6 +76,7 @@ namespace Audio {
 			source.reset();
 		}
 
+		mixer.reset();
 		// Note: Reset audio pipe AFTER resetting all pipes, otherwise the new data will be yeeted
 		resetAudioPipe();
 	}
@@ -250,6 +251,8 @@ namespace Audio {
 
 			source.isBufferIDDirty = false;
 		}
+
+		performMix(read, write);
 	}
 
 	void HLE_DSP::updateSourceConfig(Source& source, HLE::SourceConfiguration::Configuration& config, s16_le* adpcmCoefficients) {
@@ -465,6 +468,50 @@ namespace Audio {
 		}
 	}
 
+	void HLE_DSP::performMix(Audio::HLE::SharedMemory& readRegion, Audio::HLE::SharedMemory& writeRegion) {
+		updateMixerConfig(readRegion);
+		// TODO: Do the actual audio mixing
+
+		auto& dspStatus = writeRegion.dspStatus;
+		// Stub the DSP status. It's unknown what the "unknown" field is but Citra sets it to 0, so we do too to be safe
+		dspStatus.droppedFrames = 0;
+		dspStatus.unknown = 0;
+	}
+
+	void HLE_DSP::updateMixerConfig(Audio::HLE::SharedMemory& sharedMem) {
+		auto& config = sharedMem.dspConfiguration;
+		// No configs have been changed, so there's nothing to update
+		if (config.dirtyRaw == 0) {
+			return;
+		}
+
+		if (config.outputFormatDirty) {
+			mixer.channelFormat = config.outputFormat;
+		}
+
+		if (config.masterVolumeDirty) {
+			mixer.volumes[0] = config.masterVolume;
+		}
+
+		if (config.auxVolume0Dirty) {
+			mixer.volumes[1] = config.auxVolumes[0];
+		}
+
+		if (config.auxVolume1Dirty) {
+			mixer.volumes[2] = config.auxVolumes[1];
+		}
+
+		if (config.auxBusEnable0Dirty) {
+			mixer.enableAuxStages[0] = config.auxBusEnable[0] != 0;
+		}
+
+		if (config.auxBusEnable1Dirty) {
+			mixer.enableAuxStages[1] = config.auxBusEnable[1] != 0;
+		}
+
+		config.dirtyRaw = 0;
+	}
+
 	HLE_DSP::SampleBuffer HLE_DSP::decodePCM8(const u8* data, usize sampleCount, Source& source) {
 		SampleBuffer decodedSamples(sampleCount);