Merge pull request #10762 from CasualPokePlayer/fix_slow_audio_desyncs

Reduce gradual audio desyncing in dumps and apply the correct sample rate for GameCube audio

4 files changed, 69 insertions, 55 deletions

diff --git a/Source/Core/AudioCommon/Mixer.cpp b/Source/Core/AudioCommon/Mixer.cpp
index e0b7f28768..5baa02d25e 100644
--- a/Source/Core/AudioCommon/Mixer.cpp
+++ b/Source/Core/AudioCommon/Mixer.cpp
@@ -76,16 +76,18 @@ unsigned int Mixer::MixerFifo::Mix(short* samples, unsigned int numSamples,
   // advance indexR with sample position
   // remember fractional offset
 
-  float aid_sample_rate = static_cast<float>(m_input_sample_rate);
+  float aid_sample_rate =
+      FIXED_SAMPLE_RATE_DIVIDEND / static_cast<float>(m_input_sample_rate_divisor);
   if (consider_framelimit && emulationspeed > 0.0f)
   {
     float numLeft = static_cast<float>(((indexW - indexR) & INDEX_MASK) / 2);
 
-    u32 low_waterwark = m_input_sample_rate * timing_variance / 1000;
-    low_waterwark = std::min(low_waterwark, MAX_SAMPLES / 2);
+    u32 low_watermark = (FIXED_SAMPLE_RATE_DIVIDEND * timing_variance) /
+                        (static_cast<u64>(m_input_sample_rate_divisor) * 1000);
+    low_watermark = std::min(low_watermark, MAX_SAMPLES / 2);
 
     m_numLeftI = (numLeft + m_numLeftI * (CONTROL_AVG - 1)) / CONTROL_AVG;
-    float offset = (m_numLeftI - low_waterwark) * CONTROL_FACTOR;
+    float offset = (m_numLeftI - low_watermark) * CONTROL_FACTOR;
     if (offset > MAX_FREQ_SHIFT)
       offset = MAX_FREQ_SHIFT;
     if (offset < -MAX_FREQ_SHIFT)
@@ -258,9 +260,9 @@ void Mixer::PushSamples(const short* samples, unsigned int num_samples)
   m_dma_mixer.PushSamples(samples, num_samples);
   if (m_log_dsp_audio)
   {
-    int sample_rate = m_dma_mixer.GetInputSampleRate();
+    int sample_rate_divisor = m_dma_mixer.GetInputSampleRateDivisor();
     auto volume = m_dma_mixer.GetVolume();
-    m_wave_writer_dsp.AddStereoSamplesBE(samples, num_samples, sample_rate, volume.first,
+    m_wave_writer_dsp.AddStereoSamplesBE(samples, num_samples, sample_rate_divisor, volume.first,
                                          volume.second);
   }
 }
@@ -270,21 +272,21 @@ void Mixer::PushStreamingSamples(const short* samples, unsigned int num_samples)
   m_streaming_mixer.PushSamples(samples, num_samples);
   if (m_log_dtk_audio)
   {
-    int sample_rate = m_streaming_mixer.GetInputSampleRate();
+    int sample_rate_divisor = m_streaming_mixer.GetInputSampleRateDivisor();
     auto volume = m_streaming_mixer.GetVolume();
-    m_wave_writer_dtk.AddStereoSamplesBE(samples, num_samples, sample_rate, volume.first,
+    m_wave_writer_dtk.AddStereoSamplesBE(samples, num_samples, sample_rate_divisor, volume.first,
                                          volume.second);
   }
 }
 
 void Mixer::PushWiimoteSpeakerSamples(const short* samples, unsigned int num_samples,
-                                      unsigned int sample_rate)
+                                      unsigned int sample_rate_divisor)
 {
   short samples_stereo[MAX_SAMPLES * 2];
 
   if (num_samples < MAX_SAMPLES)
   {
-    m_wiimote_speaker_mixer.SetInputSampleRate(sample_rate);
+    m_wiimote_speaker_mixer.SetInputSampleRateDivisor(sample_rate_divisor);
 
     for (unsigned int i = 0; i < num_samples; ++i)
     {
@@ -301,19 +303,19 @@ void Mixer::PushGBASamples(int device_number, const short* samples, unsigned int
   m_gba_mixers[device_number].PushSamples(samples, num_samples);
 }
 
-void Mixer::SetDMAInputSampleRate(unsigned int rate)
+void Mixer::SetDMAInputSampleRateDivisor(unsigned int rate_divisor)
 {
-  m_dma_mixer.SetInputSampleRate(rate);
+  m_dma_mixer.SetInputSampleRateDivisor(rate_divisor);
 }
 
-void Mixer::SetStreamInputSampleRate(unsigned int rate)
+void Mixer::SetStreamInputSampleRateDivisor(unsigned int rate_divisor)
 {
-  m_streaming_mixer.SetInputSampleRate(rate);
+  m_streaming_mixer.SetInputSampleRateDivisor(rate_divisor);
 }
 
-void Mixer::SetGBAInputSampleRates(int device_number, unsigned int rate)
+void Mixer::SetGBAInputSampleRateDivisors(int device_number, unsigned int rate_divisor)
 {
-  m_gba_mixers[device_number].SetInputSampleRate(rate);
+  m_gba_mixers[device_number].SetInputSampleRateDivisor(rate_divisor);
 }
 
 void Mixer::SetStreamingVolume(unsigned int lvolume, unsigned int rvolume)
@@ -335,7 +337,7 @@ void Mixer::StartLogDTKAudio(const std::string& filename)
 {
   if (!m_log_dtk_audio)
   {
-    bool success = m_wave_writer_dtk.Start(filename, m_streaming_mixer.GetInputSampleRate());
+    bool success = m_wave_writer_dtk.Start(filename, m_streaming_mixer.GetInputSampleRateDivisor());
     if (success)
     {
       m_log_dtk_audio = true;
@@ -372,7 +374,7 @@ void Mixer::StartLogDSPAudio(const std::string& filename)
 {
   if (!m_log_dsp_audio)
   {
-    bool success = m_wave_writer_dsp.Start(filename, m_dma_mixer.GetInputSampleRate());
+    bool success = m_wave_writer_dsp.Start(filename, m_dma_mixer.GetInputSampleRateDivisor());
     if (success)
     {
       m_log_dsp_audio = true;
@@ -414,19 +416,19 @@ void Mixer::RefreshConfig()
 
 void Mixer::MixerFifo::DoState(PointerWrap& p)
 {
-  p.Do(m_input_sample_rate);
+  p.Do(m_input_sample_rate_divisor);
   p.Do(m_LVolume);
   p.Do(m_RVolume);
 }
 
-void Mixer::MixerFifo::SetInputSampleRate(unsigned int rate)
+void Mixer::MixerFifo::SetInputSampleRateDivisor(unsigned int rate_divisor)
 {
-  m_input_sample_rate = rate;
+  m_input_sample_rate_divisor = rate_divisor;
 }
 
-unsigned int Mixer::MixerFifo::GetInputSampleRate() const
+unsigned int Mixer::MixerFifo::GetInputSampleRateDivisor() const
 {
-  return m_input_sample_rate;
+  return m_input_sample_rate_divisor;
 }
 
 void Mixer::MixerFifo::SetVolume(unsigned int lvolume, unsigned int rvolume)
@@ -445,5 +447,6 @@ unsigned int Mixer::MixerFifo::AvailableSamples() const
   unsigned int samples_in_fifo = ((m_indexW.load() - m_indexR.load()) & INDEX_MASK) / 2;
   if (samples_in_fifo <= 1)
     return 0;  // Mixer::MixerFifo::Mix always keeps one sample in the buffer.
-  return (samples_in_fifo - 1) * m_mixer->m_sampleRate / m_input_sample_rate;
+  return (samples_in_fifo - 1) * static_cast<u64>(m_mixer->m_sampleRate) *
+         m_input_sample_rate_divisor / FIXED_SAMPLE_RATE_DIVIDEND;
 }
diff --git a/Source/Core/AudioCommon/Mixer.h b/Source/Core/AudioCommon/Mixer.h
index f58c25f98a..7191fd6ab0 100644
--- a/Source/Core/AudioCommon/Mixer.h
+++ b/Source/Core/AudioCommon/Mixer.h
@@ -29,14 +29,14 @@ public:
   void PushSamples(const short* samples, unsigned int num_samples);
   void PushStreamingSamples(const short* samples, unsigned int num_samples);
   void PushWiimoteSpeakerSamples(const short* samples, unsigned int num_samples,
-                                 unsigned int sample_rate);
+                                 unsigned int sample_rate_divisor);
   void PushGBASamples(int device_number, const short* samples, unsigned int num_samples);
 
   unsigned int GetSampleRate() const { return m_sampleRate; }
 
-  void SetDMAInputSampleRate(unsigned int rate);
-  void SetStreamInputSampleRate(unsigned int rate);
-  void SetGBAInputSampleRates(int device_number, unsigned int rate);
+  void SetDMAInputSampleRateDivisor(unsigned int rate_divisor);
+  void SetStreamInputSampleRateDivisor(unsigned int rate_divisor);
+  void SetGBAInputSampleRateDivisors(int device_number, unsigned int rate_divisor);
 
   void SetStreamingVolume(unsigned int lvolume, unsigned int rvolume);
   void SetWiimoteSpeakerVolume(unsigned int lvolume, unsigned int rvolume);
@@ -51,6 +51,9 @@ public:
   float GetCurrentSpeed() const { return m_speed.load(); }
   void UpdateSpeed(float val) { m_speed.store(val); }
 
+  // 54000000 doesn't work here as it doesn't evenly divide with 32000, but 108000000 does
+  static constexpr u64 FIXED_SAMPLE_RATE_DIVIDEND = 54000000 * 2;
+
 private:
   static constexpr u32 MAX_SAMPLES = 1024 * 4;  // 128 ms
   static constexpr u32 INDEX_MASK = MAX_SAMPLES * 2 - 1;
@@ -63,23 +66,24 @@ private:
   class MixerFifo final
   {
   public:
-    MixerFifo(Mixer* mixer, unsigned sample_rate, bool little_endian)
-        : m_mixer(mixer), m_input_sample_rate(sample_rate), m_little_endian(little_endian)
+    MixerFifo(Mixer* mixer, unsigned sample_rate_divisor, bool little_endian)
+        : m_mixer(mixer), m_input_sample_rate_divisor(sample_rate_divisor),
+          m_little_endian(little_endian)
     {
     }
     void DoState(PointerWrap& p);
     void PushSamples(const short* samples, unsigned int num_samples);
     unsigned int Mix(short* samples, unsigned int numSamples, bool consider_framelimit,
                      float emulationspeed, int timing_variance);
-    void SetInputSampleRate(unsigned int rate);
-    unsigned int GetInputSampleRate() const;
+    void SetInputSampleRateDivisor(unsigned int rate_divisor);
+    unsigned int GetInputSampleRateDivisor() const;
     void SetVolume(unsigned int lvolume, unsigned int rvolume);
     std::pair<s32, s32> GetVolume() const;
     unsigned int AvailableSamples() const;
 
   private:
     Mixer* m_mixer;
-    unsigned m_input_sample_rate;
+    unsigned m_input_sample_rate_divisor;
     bool m_little_endian;
     std::array<short, MAX_SAMPLES * 2> m_buffer{};
     std::atomic<u32> m_indexW{0};
@@ -93,11 +97,13 @@ private:
 
   void RefreshConfig();
 
-  MixerFifo m_dma_mixer{this, 32000, false};
-  MixerFifo m_streaming_mixer{this, 48000, false};
-  MixerFifo m_wiimote_speaker_mixer{this, 3000, true};
-  std::array<MixerFifo, 4> m_gba_mixers{MixerFifo{this, 48000, true}, MixerFifo{this, 48000, true},
-                                        MixerFifo{this, 48000, true}, MixerFifo{this, 48000, true}};
+  MixerFifo m_dma_mixer{this, FIXED_SAMPLE_RATE_DIVIDEND / 32000, false};
+  MixerFifo m_streaming_mixer{this, FIXED_SAMPLE_RATE_DIVIDEND / 48000, false};
+  MixerFifo m_wiimote_speaker_mixer{this, FIXED_SAMPLE_RATE_DIVIDEND / 3000, true};
+  std::array<MixerFifo, 4> m_gba_mixers{MixerFifo{this, FIXED_SAMPLE_RATE_DIVIDEND / 48000, true},
+                                        MixerFifo{this, FIXED_SAMPLE_RATE_DIVIDEND / 48000, true},
+                                        MixerFifo{this, FIXED_SAMPLE_RATE_DIVIDEND / 48000, true},
+                                        MixerFifo{this, FIXED_SAMPLE_RATE_DIVIDEND / 48000, true}};
   unsigned int m_sampleRate;
 
   bool m_is_stretching = false;
diff --git a/Source/Core/AudioCommon/WaveFile.cpp b/Source/Core/AudioCommon/WaveFile.cpp
index befcf6bfbd..dc689e74dd 100644
--- a/Source/Core/AudioCommon/WaveFile.cpp
+++ b/Source/Core/AudioCommon/WaveFile.cpp
@@ -2,6 +2,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "AudioCommon/WaveFile.h"
+#include "AudioCommon/Mixer.h"
 
 #include <string>
 
@@ -26,7 +27,7 @@ WaveFileWriter::~WaveFileWriter()
   Stop();
 }
 
-bool WaveFileWriter::Start(const std::string& filename, unsigned int HLESampleRate)
+bool WaveFileWriter::Start(const std::string& filename, u32 sample_rate_divisor)
 {
   // Ask to delete file
   if (File::Exists(filename))
@@ -65,7 +66,7 @@ bool WaveFileWriter::Start(const std::string& filename, unsigned int HLESampleRa
   if (basename.empty())
     SplitPath(filename, nullptr, &basename, nullptr);
 
-  current_sample_rate = HLESampleRate;
+  current_sample_rate_divisor = sample_rate_divisor;
 
   // -----------------
   // Write file header
@@ -78,7 +79,7 @@ bool WaveFileWriter::Start(const std::string& filename, unsigned int HLESampleRa
   Write(16);          // size of fmt block
   Write(0x00020001);  // two channels, uncompressed
 
-  const u32 sample_rate = HLESampleRate;
+  const u32 sample_rate = Mixer::FIXED_SAMPLE_RATE_DIVIDEND / sample_rate_divisor;
   Write(sample_rate);
   Write(sample_rate * 2 * 2);  // two channels, 16bit
 
@@ -114,8 +115,8 @@ void WaveFileWriter::Write4(const char* ptr)
   file.WriteBytes(ptr, 4);
 }
 
-void WaveFileWriter::AddStereoSamplesBE(const short* sample_data, u32 count, int sample_rate,
-                                        int l_volume, int r_volume)
+void WaveFileWriter::AddStereoSamplesBE(const short* sample_data, u32 count,
+                                        u32 sample_rate_divisor, int l_volume, int r_volume)
 {
   if (!file)
     ERROR_LOG_FMT(AUDIO, "WaveFileWriter - file not open.");
@@ -148,14 +149,14 @@ void WaveFileWriter::AddStereoSamplesBE(const short* sample_data, u32 count, int
     conv_buffer[2 * i + 1] = conv_buffer[2 * i + 1] * r_volume / 256;
   }
 
-  if (sample_rate != current_sample_rate)
+  if (sample_rate_divisor != current_sample_rate_divisor)
   {
     Stop();
     file_index++;
     std::ostringstream filename;
     filename << File::GetUserPath(D_DUMPAUDIO_IDX) << basename << file_index << ".wav";
-    Start(filename.str(), sample_rate);
-    current_sample_rate = sample_rate;
+    Start(filename.str(), sample_rate_divisor);
+    current_sample_rate_divisor = sample_rate_divisor;
   }
 
   file.WriteBytes(conv_buffer.data(), count * 4);
diff --git a/Source/Core/AudioCommon/WaveFile.h b/Source/Core/AudioCommon/WaveFile.h
index e95ed8eb73..350e1da9af 100644
--- a/Source/Core/AudioCommon/WaveFile.h
+++ b/Source/Core/AudioCommon/WaveFile.h
@@ -30,24 +30,28 @@ public:
   WaveFileWriter(WaveFileWriter&&) = delete;
   WaveFileWriter& operator=(WaveFileWriter&&) = delete;
 
-  bool Start(const std::string& filename, unsigned int HLESampleRate);
+  bool Start(const std::string& filename, u32 sample_rate_divisor);
   void Stop();
 
   void SetSkipSilence(bool skip) { skip_silence = skip; }
-  void AddStereoSamplesBE(const short* sample_data, u32 count, int sample_rate, int l_volume,
-                          int r_volume);  // big endian
+  // big endian
+  void AddStereoSamplesBE(const short* sample_data, u32 count, u32 sample_rate_divisor,
+                          int l_volume, int r_volume);
   u32 GetAudioSize() const { return audio_size; }
 
 private:
   static constexpr size_t BUFFER_SIZE = 32 * 1024;
 
-  File::IOFile file;
-  bool skip_silence = false;
-  u32 audio_size = 0;
-  std::array<short, BUFFER_SIZE> conv_buffer{};
   void Write(u32 value);
   void Write4(const char* ptr);
+
+  File::IOFile file;
   std::string basename;
-  int current_sample_rate;
-  int file_index = 0;
+  u32 file_index = 0;
+  u32 audio_size = 0;
+
+  u32 current_sample_rate_divisor;
+  std::array<short, BUFFER_SIZE> conv_buffer{};
+
+  bool skip_silence = false;
 };