Merge pull request #6498 from Kelebek1/Audio

[audio_core] Decouple audio update and processing, and process at variable rate

8 files changed, 180 insertions, 88 deletions

diff --git a/src/audio_core/audio_renderer.cpp b/src/audio_core/audio_renderer.cpp
index 80ffddb10..ccd5ca6cc 100644
--- a/src/audio_core/audio_renderer.cpp
+++ b/src/audio_core/audio_renderer.cpp
@@ -12,6 +12,7 @@
 #include "audio_core/voice_context.h"
 #include "common/logging/log.h"
 #include "common/settings.h"
+#include "core/core_timing.h"
 #include "core/memory.h"
 
 namespace {
@@ -68,7 +69,9 @@ namespace {
 } // namespace
 
 namespace AudioCore {
-AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory::Memory& memory_,
+constexpr s32 NUM_BUFFERS = 2;
+
+AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing_, Core::Memory::Memory& memory_,
                              AudioCommon::AudioRendererParameter params,
                              Stream::ReleaseCallback&& release_callback,
                              std::size_t instance_number)
@@ -77,7 +80,8 @@ AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory
       sink_context(params.sink_count), splitter_context(),
       voices(params.voice_count), memory{memory_},
       command_generator(worker_params, voice_context, mix_context, splitter_context, effect_context,
-                        memory) {
+                        memory),
+      core_timing{core_timing_} {
     behavior_info.SetUserRevision(params.revision);
     splitter_context.Initialize(behavior_info, params.splitter_count,
                                 params.num_splitter_send_channels);
@@ -86,16 +90,27 @@ AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory
     stream = audio_out->OpenStream(
         core_timing, params.sample_rate, AudioCommon::STREAM_NUM_CHANNELS,
         fmt::format("AudioRenderer-Instance{}", instance_number), std::move(release_callback));
-    audio_out->StartStream(stream);
-
-    QueueMixedBuffer(0);
-    QueueMixedBuffer(1);
-    QueueMixedBuffer(2);
-    QueueMixedBuffer(3);
+    process_event = Core::Timing::CreateEvent(
+        fmt::format("AudioRenderer-Instance{}-Process", instance_number),
+        [this](std::uintptr_t, std::chrono::nanoseconds) { ReleaseAndQueueBuffers(); });
+    for (s32 i = 0; i < NUM_BUFFERS; ++i) {
+        QueueMixedBuffer(i);
+    }
 }
 
 AudioRenderer::~AudioRenderer() = default;
 
+ResultCode AudioRenderer::Start() {
+    audio_out->StartStream(stream);
+    ReleaseAndQueueBuffers();
+    return ResultSuccess;
+}
+
+ResultCode AudioRenderer::Stop() {
+    audio_out->StopStream(stream);
+    return ResultSuccess;
+}
+
 u32 AudioRenderer::GetSampleRate() const {
     return worker_params.sample_rate;
 }
@@ -114,7 +129,7 @@ Stream::State AudioRenderer::GetStreamState() const {
 
 ResultCode AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_params,
                                               std::vector<u8>& output_params) {
-
+    std::scoped_lock lock{mutex};
     InfoUpdater info_updater{input_params, output_params, behavior_info};
 
     if (!info_updater.UpdateBehaviorInfo(behavior_info)) {
@@ -194,9 +209,6 @@ ResultCode AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_param
         LOG_ERROR(Audio, "Audio buffers were not consumed!");
         return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
     }
-
-    ReleaseAndQueueBuffers();
-
     return ResultSuccess;
 }
 
@@ -220,10 +232,8 @@ void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
     command_generator.PostCommand();
     // Base sample size
     std::size_t BUFFER_SIZE{worker_params.sample_count};
-    // Samples
-    std::vector<s16> buffer(BUFFER_SIZE * stream->GetNumChannels());
-    // Make sure to clear our samples
-    std::memset(buffer.data(), 0, buffer.size() * sizeof(s16));
+    // Samples, making sure to clear
+    std::vector<s16> buffer(BUFFER_SIZE * stream->GetNumChannels(), 0);
 
     if (sink_context.InUse()) {
         const auto stream_channel_count = stream->GetNumChannels();
@@ -315,10 +325,24 @@ void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
 }
 
 void AudioRenderer::ReleaseAndQueueBuffers() {
-    const auto released_buffers{audio_out->GetTagsAndReleaseBuffers(stream)};
-    for (const auto& tag : released_buffers) {
-        QueueMixedBuffer(tag);
+    if (!stream->IsPlaying()) {
+        return;
     }
+
+    {
+        std::scoped_lock lock{mutex};
+        const auto released_buffers{audio_out->GetTagsAndReleaseBuffers(stream)};
+        for (const auto& tag : released_buffers) {
+            QueueMixedBuffer(tag);
+        }
+    }
+
+    const f32 sample_rate = static_cast<f32>(GetSampleRate());
+    const f32 sample_count = static_cast<f32>(GetSampleCount());
+    const f32 consume_rate = sample_rate / (sample_count * (sample_count / 240));
+    const s32 ms = (1000 / static_cast<s32>(consume_rate)) - 1;
+    const std::chrono::milliseconds next_event_time(std::max(ms / NUM_BUFFERS, 1));
+    core_timing.ScheduleEvent(next_event_time, process_event, {});
 }
 
 } // namespace AudioCore
diff --git a/src/audio_core/audio_renderer.h b/src/audio_core/audio_renderer.h
index 18567f618..88fdd13dd 100644
--- a/src/audio_core/audio_renderer.h
+++ b/src/audio_core/audio_renderer.h
@@ -6,6 +6,7 @@
 
 #include <array>
 #include <memory>
+#include <mutex>
 #include <vector>
 
 #include "audio_core/behavior_info.h"
@@ -45,6 +46,8 @@ public:
 
     [[nodiscard]] ResultCode UpdateAudioRenderer(const std::vector<u8>& input_params,
                                                  std::vector<u8>& output_params);
+    [[nodiscard]] ResultCode Start();
+    [[nodiscard]] ResultCode Stop();
     void QueueMixedBuffer(Buffer::Tag tag);
     void ReleaseAndQueueBuffers();
     [[nodiscard]] u32 GetSampleRate() const;
@@ -68,6 +71,9 @@ private:
     Core::Memory::Memory& memory;
     CommandGenerator command_generator;
     std::size_t elapsed_frame_count{};
+    Core::Timing::CoreTiming& core_timing;
+    std::shared_ptr<Core::Timing::EventType> process_event;
+    std::mutex mutex;
 };
 
 } // namespace AudioCore
diff --git a/src/audio_core/command_generator.cpp b/src/audio_core/command_generator.cpp
index 437cc5ccd..27437f1ea 100644
--- a/src/audio_core/command_generator.cpp
+++ b/src/audio_core/command_generator.cpp
@@ -795,7 +795,7 @@ void CommandGenerator::UpdateI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbSta
         state.lowpass_1 = 0.0f;
     } else {
         const auto a = 1.0f - hf_gain;
-        const auto b = 2.0f * (1.0f - hf_gain * CosD(256.0f * info.hf_reference /
+        const auto b = 2.0f * (2.0f - hf_gain * CosD(256.0f * info.hf_reference /
                                                      static_cast<f32>(info.sample_rate)));
         const auto c = std::sqrt(b * b - 4.0f * a * a);
 
@@ -843,7 +843,7 @@ void CommandGenerator::UpdateI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbSta
     }
 
     const auto max_early_delay = state.early_delay_line.GetMaxDelay();
-    const auto reflection_time = 1000.0f * (0.0098f * info.reverb_delay + 0.02f);
+    const auto reflection_time = 1000.0f * (0.9998f * info.reverb_delay + 0.02f);
     for (std::size_t tap = 0; tap < AudioCommon::I3DL2REVERB_TAPS; tap++) {
         const auto length = AudioCommon::CalculateDelaySamples(
             sample_rate, 1000.0f * info.reflection_delay + reflection_time * EARLY_TAP_TIMES[tap]);
@@ -1004,7 +1004,8 @@ void CommandGenerator::GenerateFinalMixCommand() {
 }
 
 s32 CommandGenerator::DecodePcm16(ServerVoiceInfo& voice_info, VoiceState& dsp_state,
-                                  s32 sample_count, s32 channel, std::size_t mix_offset) {
+                                  s32 sample_start_offset, s32 sample_end_offset, s32 sample_count,
+                                  s32 channel, std::size_t mix_offset) {
     const auto& in_params = voice_info.GetInParams();
     const auto& wave_buffer = in_params.wave_buffer[dsp_state.wave_buffer_index];
     if (wave_buffer.buffer_address == 0) {
@@ -1013,14 +1014,12 @@ s32 CommandGenerator::DecodePcm16(ServerVoiceInfo& voice_info, VoiceState& dsp_s
     if (wave_buffer.buffer_size == 0) {
         return 0;
     }
-    if (wave_buffer.end_sample_offset < wave_buffer.start_sample_offset) {
+    if (sample_end_offset < sample_start_offset) {
         return 0;
     }
-    const auto samples_remaining =
-        (wave_buffer.end_sample_offset - wave_buffer.start_sample_offset) - dsp_state.offset;
+    const auto samples_remaining = (sample_end_offset - sample_start_offset) - dsp_state.offset;
     const auto start_offset =
-        ((wave_buffer.start_sample_offset + dsp_state.offset) * in_params.channel_count) *
-        sizeof(s16);
+        ((dsp_state.offset + sample_start_offset) * in_params.channel_count) * sizeof(s16);
     const auto buffer_pos = wave_buffer.buffer_address + start_offset;
     const auto samples_processed = std::min(sample_count, samples_remaining);
 
@@ -1044,8 +1043,8 @@ s32 CommandGenerator::DecodePcm16(ServerVoiceInfo& voice_info, VoiceState& dsp_s
 }
 
 s32 CommandGenerator::DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_state,
-                                  s32 sample_count, [[maybe_unused]] s32 channel,
-                                  std::size_t mix_offset) {
+                                  s32 sample_start_offset, s32 sample_end_offset, s32 sample_count,
+                                  [[maybe_unused]] s32 channel, std::size_t mix_offset) {
     const auto& in_params = voice_info.GetInParams();
     const auto& wave_buffer = in_params.wave_buffer[dsp_state.wave_buffer_index];
     if (wave_buffer.buffer_address == 0) {
@@ -1054,7 +1053,7 @@ s32 CommandGenerator::DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_s
     if (wave_buffer.buffer_size == 0) {
         return 0;
     }
-    if (wave_buffer.end_sample_offset < wave_buffer.start_sample_offset) {
+    if (sample_end_offset < sample_start_offset) {
         return 0;
     }
 
@@ -1079,10 +1078,9 @@ s32 CommandGenerator::DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_s
     s32 coef1 = coeffs[idx * 2];
     s32 coef2 = coeffs[idx * 2 + 1];
 
-    const auto samples_remaining =
-        (wave_buffer.end_sample_offset - wave_buffer.start_sample_offset) - dsp_state.offset;
+    const auto samples_remaining = (sample_end_offset - sample_start_offset) - dsp_state.offset;
     const auto samples_processed = std::min(sample_count, samples_remaining);
-    const auto sample_pos = wave_buffer.start_sample_offset + dsp_state.offset;
+    const auto sample_pos = dsp_state.offset + sample_start_offset;
 
     const auto samples_remaining_in_frame = sample_pos % SAMPLES_PER_FRAME;
     auto position_in_frame = ((sample_pos / SAMPLES_PER_FRAME) * NIBBLES_PER_SAMPLE) +
@@ -1210,9 +1208,8 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
     }
 
     std::size_t temp_mix_offset{};
-    bool is_buffer_completed{false};
     auto samples_remaining = sample_count;
-    while (samples_remaining > 0 && !is_buffer_completed) {
+    while (samples_remaining > 0) {
         const auto samples_to_output = std::min(samples_remaining, min_required_samples);
         const auto samples_to_read = (samples_to_output * resample_rate + dsp_state.fraction) >> 15;
 
@@ -1229,24 +1226,38 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
             const auto& wave_buffer = in_params.wave_buffer[dsp_state.wave_buffer_index];
             // No more data can be read
             if (!dsp_state.is_wave_buffer_valid[dsp_state.wave_buffer_index]) {
-                is_buffer_completed = true;
                 break;
             }
 
             if (in_params.sample_format == SampleFormat::Adpcm && dsp_state.offset == 0 &&
                 wave_buffer.context_address != 0 && wave_buffer.context_size != 0) {
-                // TODO(ogniK): ADPCM loop context
+                memory.ReadBlock(wave_buffer.context_address, &dsp_state.context,
+                                 sizeof(ADPCMContext));
+            }
+
+            s32 samples_offset_start;
+            s32 samples_offset_end;
+            if (dsp_state.loop_count > 0 && wave_buffer.loop_start_sample != 0 &&
+                wave_buffer.loop_end_sample != 0 &&
+                wave_buffer.loop_start_sample <= wave_buffer.loop_end_sample) {
+                samples_offset_start = wave_buffer.loop_start_sample;
+                samples_offset_end = wave_buffer.loop_end_sample;
+            } else {
+                samples_offset_start = wave_buffer.start_sample_offset;
+                samples_offset_end = wave_buffer.end_sample_offset;
             }
 
             s32 samples_decoded{0};
             switch (in_params.sample_format) {
             case SampleFormat::Pcm16:
-                samples_decoded = DecodePcm16(voice_info, dsp_state, samples_to_read - samples_read,
-                                              channel, temp_mix_offset);
+                samples_decoded =
+                    DecodePcm16(voice_info, dsp_state, samples_offset_start, samples_offset_end,
+                                samples_to_read - samples_read, channel, temp_mix_offset);
                 break;
             case SampleFormat::Adpcm:
-                samples_decoded = DecodeAdpcm(voice_info, dsp_state, samples_to_read - samples_read,
-                                              channel, temp_mix_offset);
+                samples_decoded =
+                    DecodeAdpcm(voice_info, dsp_state, samples_offset_start, samples_offset_end,
+                                samples_to_read - samples_read, channel, temp_mix_offset);
                 break;
             default:
                 UNREACHABLE_MSG("Unimplemented sample format={}", in_params.sample_format);
@@ -1257,15 +1268,19 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
             dsp_state.offset += samples_decoded;
             dsp_state.played_sample_count += samples_decoded;
 
-            if (dsp_state.offset >=
-                    (wave_buffer.end_sample_offset - wave_buffer.start_sample_offset) ||
+            if (dsp_state.offset >= (samples_offset_end - samples_offset_start) ||
                 samples_decoded == 0) {
                 // Reset our sample offset
                 dsp_state.offset = 0;
                 if (wave_buffer.is_looping) {
-                    if (samples_decoded == 0) {
+                    dsp_state.loop_count++;
+                    if (wave_buffer.loop_count > 0 &&
+                        (dsp_state.loop_count > wave_buffer.loop_count || samples_decoded == 0)) {
                         // End of our buffer
-                        is_buffer_completed = true;
+                        voice_info.SetWaveBufferCompleted(dsp_state, wave_buffer);
+                    }
+
+                    if (samples_decoded == 0) {
                         break;
                     }
 
@@ -1273,15 +1288,8 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
                         dsp_state.played_sample_count = 0;
                     }
                 } else {
-
                     // Update our wave buffer states
-                    dsp_state.is_wave_buffer_valid[dsp_state.wave_buffer_index] = false;
-                    dsp_state.wave_buffer_consumed++;
-                    dsp_state.wave_buffer_index =
-                        (dsp_state.wave_buffer_index + 1) % AudioCommon::MAX_WAVE_BUFFERS;
-                    if (wave_buffer.end_of_stream) {
-                        dsp_state.played_sample_count = 0;
-                    }
+                    voice_info.SetWaveBufferCompleted(dsp_state, wave_buffer);
                 }
             }
         }
diff --git a/src/audio_core/command_generator.h b/src/audio_core/command_generator.h
index 2ebb755b0..673e4fbef 100644
--- a/src/audio_core/command_generator.h
+++ b/src/audio_core/command_generator.h
@@ -86,10 +86,10 @@ private:
                                std::vector<u8>& work_buffer);
     void UpdateI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbState& state, bool should_clear);
     // DSP Code
-    s32 DecodePcm16(ServerVoiceInfo& voice_info, VoiceState& dsp_state, s32 sample_count,
-                    s32 channel, std::size_t mix_offset);
-    s32 DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_state, s32 sample_count,
-                    s32 channel, std::size_t mix_offset);
+    s32 DecodePcm16(ServerVoiceInfo& voice_info, VoiceState& dsp_state, s32 sample_start_offset,
+                    s32 sample_end_offset, s32 sample_count, s32 channel, std::size_t mix_offset);
+    s32 DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_state, s32 sample_start_offset,
+                    s32 sample_end_offset, s32 sample_count, s32 channel, std::size_t mix_offset);
     void DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* output, VoiceState& dsp_state,
                                s32 channel, s32 target_sample_rate, s32 sample_count, s32 node_id);
 
diff --git a/src/audio_core/info_updater.cpp b/src/audio_core/info_updater.cpp
index 4a5b1b4ab..9b4ca1851 100644
--- a/src/audio_core/info_updater.cpp
+++ b/src/audio_core/info_updater.cpp
@@ -189,9 +189,6 @@ bool InfoUpdater::UpdateVoices(VoiceContext& voice_context,
         if (voice_in_params.is_new) {
             // Default our values for our voice
             voice_info.Initialize();
-            if (channel_count == 0 || channel_count > AudioCommon::MAX_CHANNEL_COUNT) {
-                continue;
-            }
 
             // Zero out our voice states
             for (std::size_t channel = 0; channel < channel_count; channel++) {
diff --git a/src/audio_core/voice_context.cpp b/src/audio_core/voice_context.cpp
index 867b8fc6b..d8c954b60 100644
--- a/src/audio_core/voice_context.cpp
+++ b/src/audio_core/voice_context.cpp
@@ -66,7 +66,7 @@ void ServerVoiceInfo::Initialize() {
     in_params.last_volume = 0.0f;
     in_params.biquad_filter.fill({});
     in_params.wave_buffer_count = 0;
-    in_params.wave_bufffer_head = 0;
+    in_params.wave_buffer_head = 0;
     in_params.mix_id = AudioCommon::NO_MIX;
     in_params.splitter_info_id = AudioCommon::NO_SPLITTER;
     in_params.additional_params_address = 0;
@@ -75,7 +75,7 @@ void ServerVoiceInfo::Initialize() {
     out_params.played_sample_count = 0;
     out_params.wave_buffer_consumed = 0;
     in_params.voice_drop_flag = false;
-    in_params.buffer_mapped = false;
+    in_params.buffer_mapped = true;
     in_params.wave_buffer_flush_request_count = 0;
     in_params.was_biquad_filter_enabled.fill(false);
 
@@ -126,7 +126,7 @@ void ServerVoiceInfo::UpdateParameters(const VoiceInfo::InParams& voice_in,
     in_params.volume = voice_in.volume;
     in_params.biquad_filter = voice_in.biquad_filter;
     in_params.wave_buffer_count = voice_in.wave_buffer_count;
-    in_params.wave_bufffer_head = voice_in.wave_buffer_head;
+    in_params.wave_buffer_head = voice_in.wave_buffer_head;
     if (behavior_info.IsFlushVoiceWaveBuffersSupported()) {
         const auto in_request_count = in_params.wave_buffer_flush_request_count;
         const auto voice_request_count = voice_in.wave_buffer_flush_request_count;
@@ -185,14 +185,16 @@ void ServerVoiceInfo::UpdateWaveBuffers(
             wave_buffer.buffer_size = 0;
             wave_buffer.context_address = 0;
             wave_buffer.context_size = 0;
+            wave_buffer.loop_start_sample = 0;
+            wave_buffer.loop_end_sample = 0;
             wave_buffer.sent_to_dsp = true;
         }
 
         // Mark all our wave buffers as invalid
         for (std::size_t channel = 0; channel < static_cast<std::size_t>(in_params.channel_count);
              channel++) {
-            for (auto& is_valid : voice_states[channel]->is_wave_buffer_valid) {
-                is_valid = false;
+            for (std::size_t i = 0; i < AudioCommon::MAX_WAVE_BUFFERS; ++i) {
+                voice_states[channel]->is_wave_buffer_valid[i] = false;
             }
         }
     }
@@ -211,7 +213,7 @@ void ServerVoiceInfo::UpdateWaveBuffer(ServerWaveBuffer& out_wavebuffer,
                                        const WaveBuffer& in_wave_buffer, SampleFormat sample_format,
                                        bool is_buffer_valid,
                                        [[maybe_unused]] BehaviorInfo& behavior_info) {
-    if (!is_buffer_valid && out_wavebuffer.sent_to_dsp) {
+    if (!is_buffer_valid && out_wavebuffer.sent_to_dsp && out_wavebuffer.buffer_address != 0) {
         out_wavebuffer.buffer_address = 0;
         out_wavebuffer.buffer_size = 0;
     }
@@ -219,11 +221,40 @@ void ServerVoiceInfo::UpdateWaveBuffer(ServerWaveBuffer& out_wavebuffer,
     if (!in_wave_buffer.sent_to_server || !in_params.buffer_mapped) {
         // Validate sample offset sizings
         if (sample_format == SampleFormat::Pcm16) {
-            const auto buffer_size = in_wave_buffer.buffer_size;
-            if (in_wave_buffer.start_sample_offset < 0 || in_wave_buffer.end_sample_offset < 0 ||
-                (buffer_size < (sizeof(s16) * in_wave_buffer.start_sample_offset)) ||
-                (buffer_size < (sizeof(s16) * in_wave_buffer.end_sample_offset))) {
+            const s64 buffer_size = static_cast<s64>(in_wave_buffer.buffer_size);
+            const s64 start = sizeof(s16) * in_wave_buffer.start_sample_offset;
+            const s64 end = sizeof(s16) * in_wave_buffer.end_sample_offset;
+            if (0 > start || start > buffer_size || 0 > end || end > buffer_size) {
                 // TODO(ogniK): Write error info
+                LOG_ERROR(Audio,
+                          "PCM16 wavebuffer has an invalid size. Buffer has size 0x{:08X}, but "
+                          "offsets were "
+                          "{:08X} - 0x{:08X}",
+                          buffer_size, sizeof(s16) * in_wave_buffer.start_sample_offset,
+                          sizeof(s16) * in_wave_buffer.end_sample_offset);
+                return;
+            }
+        } else if (sample_format == SampleFormat::Adpcm) {
+            const s64 buffer_size = static_cast<s64>(in_wave_buffer.buffer_size);
+            const s64 start_frames = in_wave_buffer.start_sample_offset / 14;
+            const s64 start_extra = in_wave_buffer.start_sample_offset % 14 == 0
+                                        ? 0
+                                        : (in_wave_buffer.start_sample_offset % 14) / 2 + 1 +
+                                              (in_wave_buffer.start_sample_offset % 2);
+            const s64 start = start_frames * 8 + start_extra;
+            const s64 end_frames = in_wave_buffer.end_sample_offset / 14;
+            const s64 end_extra = in_wave_buffer.end_sample_offset % 14 == 0
+                                      ? 0
+                                      : (in_wave_buffer.end_sample_offset % 14) / 2 + 1 +
+                                            (in_wave_buffer.end_sample_offset % 2);
+            const s64 end = end_frames * 8 + end_extra;
+            if (in_wave_buffer.start_sample_offset < 0 || start > buffer_size ||
+                in_wave_buffer.end_sample_offset < 0 || end > buffer_size) {
+                LOG_ERROR(Audio,
+                          "ADPMC wavebuffer has an invalid size. Buffer has size 0x{:08X}, but "
+                          "offsets were "
+                          "{:08X} - 0x{:08X}",
+                          in_wave_buffer.buffer_size, start, end);
                 return;
             }
         }
@@ -239,29 +270,34 @@ void ServerVoiceInfo::UpdateWaveBuffer(ServerWaveBuffer& out_wavebuffer,
         out_wavebuffer.buffer_size = in_wave_buffer.buffer_size;
         out_wavebuffer.context_address = in_wave_buffer.context_address;
         out_wavebuffer.context_size = in_wave_buffer.context_size;
+        out_wavebuffer.loop_start_sample = in_wave_buffer.loop_start_sample;
+        out_wavebuffer.loop_end_sample = in_wave_buffer.loop_end_sample;
         in_params.buffer_mapped =
             in_wave_buffer.buffer_address != 0 && in_wave_buffer.buffer_size != 0;
         // TODO(ogniK): Pool mapper attachment
         // TODO(ogniK): IsAdpcmLoopContextBugFixed
+        if (sample_format == SampleFormat::Adpcm && in_wave_buffer.context_address != 0 &&
+            in_wave_buffer.context_size != 0 && behavior_info.IsAdpcmLoopContextBugFixed()) {
+        } else {
+            out_wavebuffer.context_address = 0;
+            out_wavebuffer.context_size = 0;
+        }
     }
 }
 
 void ServerVoiceInfo::WriteOutStatus(
     VoiceInfo::OutParams& voice_out, VoiceInfo::InParams& voice_in,
     std::array<VoiceState*, AudioCommon::MAX_CHANNEL_COUNT>& voice_states) {
-    if (voice_in.is_new) {
+    if (voice_in.is_new || in_params.is_new) {
         in_params.is_new = true;
         voice_out.wave_buffer_consumed = 0;
         voice_out.played_sample_count = 0;
         voice_out.voice_dropped = false;
-    } else if (!in_params.is_new) {
-        voice_out.wave_buffer_consumed = voice_states[0]->wave_buffer_consumed;
-        voice_out.played_sample_count = voice_states[0]->played_sample_count;
-        voice_out.voice_dropped = in_params.voice_drop_flag;
     } else {
-        voice_out.wave_buffer_consumed = 0;
-        voice_out.played_sample_count = 0;
-        voice_out.voice_dropped = false;
+        const auto& state = voice_states[0];
+        voice_out.wave_buffer_consumed = state->wave_buffer_consumed;
+        voice_out.played_sample_count = state->played_sample_count;
+        voice_out.voice_dropped = state->voice_dropped;
     }
 }
 
@@ -283,7 +319,8 @@ ServerVoiceInfo::OutParams& ServerVoiceInfo::GetOutParams() {
 
 bool ServerVoiceInfo::ShouldSkip() const {
     // TODO(ogniK): Handle unmapped wave buffers or parameters
-    return !in_params.in_use || (in_params.wave_buffer_count == 0) || in_params.voice_drop_flag;
+    return !in_params.in_use || in_params.wave_buffer_count == 0 || !in_params.buffer_mapped ||
+           in_params.voice_drop_flag;
 }
 
 bool ServerVoiceInfo::UpdateForCommandGeneration(VoiceContext& voice_context) {
@@ -381,7 +418,7 @@ bool ServerVoiceInfo::UpdateParametersForCommandGeneration(
 void ServerVoiceInfo::FlushWaveBuffers(
     u8 flush_count, std::array<VoiceState*, AudioCommon::MAX_CHANNEL_COUNT>& dsp_voice_states,
     s32 channel_count) {
-    auto wave_head = in_params.wave_bufffer_head;
+    auto wave_head = in_params.wave_buffer_head;
 
     for (u8 i = 0; i < flush_count; i++) {
         in_params.wave_buffer[wave_head].sent_to_dsp = true;
@@ -401,6 +438,17 @@ bool ServerVoiceInfo::HasValidWaveBuffer(const VoiceState* state) const {
     return std::find(valid_wb.begin(), valid_wb.end(), true) != valid_wb.end();
 }
 
+void ServerVoiceInfo::SetWaveBufferCompleted(VoiceState& dsp_state,
+                                             const ServerWaveBuffer& wave_buffer) {
+    dsp_state.is_wave_buffer_valid[dsp_state.wave_buffer_index] = false;
+    dsp_state.wave_buffer_consumed++;
+    dsp_state.wave_buffer_index = (dsp_state.wave_buffer_index + 1) % AudioCommon::MAX_WAVE_BUFFERS;
+    dsp_state.loop_count = 0;
+    if (wave_buffer.end_of_stream) {
+        dsp_state.played_sample_count = 0;
+    }
+}
+
 VoiceContext::VoiceContext(std::size_t voice_count_) : voice_count{voice_count_} {
     for (std::size_t i = 0; i < voice_count; i++) {
         voice_channel_resources.emplace_back(static_cast<s32>(i));
diff --git a/src/audio_core/voice_context.h b/src/audio_core/voice_context.h
index 70359cadb..e1050897b 100644
--- a/src/audio_core/voice_context.h
+++ b/src/audio_core/voice_context.h
@@ -60,10 +60,12 @@ struct WaveBuffer {
     u8 is_looping{};
     u8 end_of_stream{};
     u8 sent_to_server{};
-    INSERT_PADDING_BYTES(5);
+    INSERT_PADDING_BYTES(1);
+    s32 loop_count{};
     u64 context_address{};
     u64 context_size{};
-    INSERT_PADDING_BYTES(8);
+    u32 loop_start_sample{};
+    u32 loop_end_sample{};
 };
 static_assert(sizeof(WaveBuffer) == 0x38, "WaveBuffer is an invalid size");
 
@@ -76,6 +78,9 @@ struct ServerWaveBuffer {
     bool end_of_stream{};
     VAddr context_address{};
     std::size_t context_size{};
+    s32 loop_count{};
+    u32 loop_start_sample{};
+    u32 loop_end_sample{};
     bool sent_to_dsp{true};
 };
 
@@ -108,6 +113,7 @@ struct VoiceState {
     u32 external_context_size;
     bool is_external_context_used;
     bool voice_dropped;
+    s32 loop_count;
 };
 
 class VoiceChannelResource {
@@ -206,7 +212,7 @@ public:
         float last_volume{};
         std::array<BiquadFilterParameter, AudioCommon::MAX_BIQUAD_FILTERS> biquad_filter{};
         s32 wave_buffer_count{};
-        s16 wave_bufffer_head{};
+        s16 wave_buffer_head{};
         INSERT_PADDING_BYTES(2);
         BehaviorFlags behavior_flags{};
         VAddr additional_params_address{};
@@ -252,6 +258,7 @@ public:
     void FlushWaveBuffers(u8 flush_count,
                           std::array<VoiceState*, AudioCommon::MAX_CHANNEL_COUNT>& dsp_voice_states,
                           s32 channel_count);
+    void SetWaveBufferCompleted(VoiceState& dsp_state, const ServerWaveBuffer& wave_buffer);
 
 private:
     std::vector<s16> stored_samples;
diff --git a/src/core/hle/service/audio/audren_u.cpp b/src/core/hle/service/audio/audren_u.cpp
index 800feba6e..7583d68b2 100644
--- a/src/core/hle/service/audio/audren_u.cpp
+++ b/src/core/hle/service/audio/audren_u.cpp
@@ -96,7 +96,7 @@ private:
     void RequestUpdateImpl(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_Audio, "(STUBBED) called");
 
-        std::vector<u8> output_params(ctx.GetWriteBufferSize());
+        std::vector<u8> output_params(ctx.GetWriteBufferSize(), 0);
         auto result = renderer->UpdateAudioRenderer(ctx.ReadBuffer(), output_params);
 
         if (result.IsSuccess()) {
@@ -110,17 +110,19 @@ private:
     void Start(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_Audio, "(STUBBED) called");
 
-        IPC::ResponseBuilder rb{ctx, 2};
+        const auto result = renderer->Start();
 
-        rb.Push(ResultSuccess);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(result);
     }
 
     void Stop(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_Audio, "(STUBBED) called");
 
-        IPC::ResponseBuilder rb{ctx, 2};
+        const auto result = renderer->Stop();
 
-        rb.Push(ResultSuccess);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(result);
     }
 
     void QuerySystemEvent(Kernel::HLERequestContext& ctx) {