34 files changed, 659 insertions, 172 deletions

diff --git a/.gitmodules b/.gitmodules
index 4f4e8690b..e73ca99e3 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -31,3 +31,6 @@
 [submodule "opus"]
     path = externals/opus
     url = https://github.com/ogniK5377/opus.git
+[submodule "soundtouch"]
+        path = externals/soundtouch
+        url = https://github.com/citra-emu/ext-soundtouch.git
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 0f32ecfba..500d099fc 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -431,6 +431,9 @@ enable_testing()
 add_subdirectory(externals)
 add_subdirectory(src)
 
+# Set yuzu project as default StartUp Project in Visual Studio
+set_property(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} PROPERTY VS_STARTUP_PROJECT yuzu)
+
 
 # Installation instructions
 # =========================
diff --git a/externals/CMakeLists.txt b/externals/CMakeLists.txt
index 3d8e10c2b..53dcf1f1a 100644
--- a/externals/CMakeLists.txt
+++ b/externals/CMakeLists.txt
@@ -50,6 +50,9 @@ add_subdirectory(open_source_archives EXCLUDE_FROM_ALL)
 add_library(unicorn-headers INTERFACE)
 target_include_directories(unicorn-headers INTERFACE ./unicorn/include)
 
+# SoundTouch
+add_subdirectory(soundtouch)
+
 # Xbyak
 if (ARCHITECTURE_x86_64)
     # Defined before "dynarmic" above
diff --git a/externals/soundtouch b/externals/soundtouch
new file mode 160000
+Subproject 060181eaf273180d3a7e87349895bd0cb6ccbf4
diff --git a/src/audio_core/CMakeLists.txt b/src/audio_core/CMakeLists.txt
index 82e4850f7..c381dbe1d 100644
--- a/src/audio_core/CMakeLists.txt
+++ b/src/audio_core/CMakeLists.txt
@@ -17,6 +17,8 @@ add_library(audio_core STATIC
     sink_stream.h
     stream.cpp
     stream.h
+    time_stretch.cpp
+    time_stretch.h
 
     $<$<BOOL:${ENABLE_CUBEB}>:cubeb_sink.cpp cubeb_sink.h>
 )
@@ -24,6 +26,7 @@ add_library(audio_core STATIC
 create_target_directory_groups(audio_core)
 
 target_link_libraries(audio_core PUBLIC common core)
+target_link_libraries(audio_core PRIVATE SoundTouch)
 
 if(ENABLE_CUBEB)
     target_link_libraries(audio_core PRIVATE cubeb)
diff --git a/src/audio_core/cubeb_sink.cpp b/src/audio_core/cubeb_sink.cpp
index 5a1177d0c..79155a7a0 100644
--- a/src/audio_core/cubeb_sink.cpp
+++ b/src/audio_core/cubeb_sink.cpp
@@ -3,27 +3,23 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <atomic>
 #include <cstring>
-#include <mutex>
-
 #include "audio_core/cubeb_sink.h"
 #include "audio_core/stream.h"
+#include "audio_core/time_stretch.h"
 #include "common/logging/log.h"
+#include "common/ring_buffer.h"
+#include "core/settings.h"
 
 namespace AudioCore {
 
-class SinkStreamImpl final : public SinkStream {
+class CubebSinkStream final : public SinkStream {
 public:
-    SinkStreamImpl(cubeb* ctx, u32 sample_rate, u32 num_channels_, cubeb_devid output_device,
-                   const std::string& name)
-        : ctx{ctx}, num_channels{num_channels_} {
-
-        if (num_channels == 6) {
-            // 6-channel audio does not seem to work with cubeb + SDL, so we downsample this to 2
-            // channel for now
-            is_6_channel = true;
-            num_channels = 2;
-        }
+    CubebSinkStream(cubeb* ctx, u32 sample_rate, u32 num_channels_, cubeb_devid output_device,
+                    const std::string& name)
+        : ctx{ctx}, num_channels{std::min(num_channels_, 2u)}, time_stretch{sample_rate,
+                                                                            num_channels} {
 
         cubeb_stream_params params{};
         params.rate = sample_rate;
@@ -38,7 +34,7 @@ public:
 
         if (cubeb_stream_init(ctx, &stream_backend, name.c_str(), nullptr, nullptr, output_device,
                               &params, std::max(512u, minimum_latency),
-                              &SinkStreamImpl::DataCallback, &SinkStreamImpl::StateCallback,
+                              &CubebSinkStream::DataCallback, &CubebSinkStream::StateCallback,
                               this) != CUBEB_OK) {
             LOG_CRITICAL(Audio_Sink, "Error initializing cubeb stream");
             return;
@@ -50,7 +46,7 @@ public:
         }
     }
 
-    ~SinkStreamImpl() {
+    ~CubebSinkStream() {
         if (!ctx) {
             return;
         }
@@ -62,27 +58,32 @@ public:
         cubeb_stream_destroy(stream_backend);
     }
 
-    void EnqueueSamples(u32 num_channels, const std::vector<s16>& samples) override {
-        if (!ctx) {
+    void EnqueueSamples(u32 source_num_channels, const std::vector<s16>& samples) override {
+        if (source_num_channels > num_channels) {
+            // Downsample 6 channels to 2
+            std::vector<s16> buf;
+            buf.reserve(samples.size() * num_channels / source_num_channels);
+            for (size_t i = 0; i < samples.size(); i += source_num_channels) {
+                for (size_t ch = 0; ch < num_channels; ch++) {
+                    buf.push_back(samples[i + ch]);
+                }
+            }
+            queue.Push(buf);
             return;
         }
 
-        std::lock_guard lock{queue_mutex};
+        queue.Push(samples);
+    }
 
-        queue.reserve(queue.size() + samples.size() * GetNumChannels());
+    size_t SamplesInQueue(u32 num_channels) const override {
+        if (!ctx)
+            return 0;
 
-        if (is_6_channel) {
-            // Downsample 6 channels to 2
-            const size_t sample_count_copy_size = samples.size() * 2;
-            queue.reserve(sample_count_copy_size);
-            for (size_t i = 0; i < samples.size(); i += num_channels) {
-                queue.push_back(samples[i]);
-                queue.push_back(samples[i + 1]);
-            }
-        } else {
-            // Copy as-is
-            std::copy(samples.begin(), samples.end(), std::back_inserter(queue));
-        }
+        return queue.Size() / num_channels;
+    }
+
+    void Flush() override {
+        should_flush = true;
     }
 
     u32 GetNumChannels() const {
@@ -95,10 +96,11 @@ private:
     cubeb* ctx{};
     cubeb_stream* stream_backend{};
     u32 num_channels{};
-    bool is_6_channel{};
 
-    std::mutex queue_mutex;
-    std::vector<s16> queue;
+    Common::RingBuffer<s16, 0x10000> queue;
+    std::array<s16, 2> last_frame;
+    std::atomic<bool> should_flush{};
+    TimeStretcher time_stretch;
 
     static long DataCallback(cubeb_stream* stream, void* user_data, const void* input_buffer,
                              void* output_buffer, long num_frames);
@@ -144,38 +146,52 @@ CubebSink::~CubebSink() {
 SinkStream& CubebSink::AcquireSinkStream(u32 sample_rate, u32 num_channels,
                                          const std::string& name) {
     sink_streams.push_back(
-        std::make_unique<SinkStreamImpl>(ctx, sample_rate, num_channels, output_device, name));
+        std::make_unique<CubebSinkStream>(ctx, sample_rate, num_channels, output_device, name));
     return *sink_streams.back();
 }
 
-long SinkStreamImpl::DataCallback(cubeb_stream* stream, void* user_data, const void* input_buffer,
-                                  void* output_buffer, long num_frames) {
-    SinkStreamImpl* impl = static_cast<SinkStreamImpl*>(user_data);
+long CubebSinkStream::DataCallback(cubeb_stream* stream, void* user_data, const void* input_buffer,
+                                   void* output_buffer, long num_frames) {
+    CubebSinkStream* impl = static_cast<CubebSinkStream*>(user_data);
     u8* buffer = reinterpret_cast<u8*>(output_buffer);
 
     if (!impl) {
         return {};
     }
 
-    std::lock_guard lock{impl->queue_mutex};
+    const size_t num_channels = impl->GetNumChannels();
+    const size_t samples_to_write = num_channels * num_frames;
+    size_t samples_written;
+
+    if (Settings::values.enable_audio_stretching) {
+        const std::vector<s16> in{impl->queue.Pop()};
+        const size_t num_in{in.size() / num_channels};
+        s16* const out{reinterpret_cast<s16*>(buffer)};
+        const size_t out_frames = impl->time_stretch.Process(in.data(), num_in, out, num_frames);
+        samples_written = out_frames * num_channels;
 
-    const size_t frames_to_write{
-        std::min(impl->queue.size() / impl->GetNumChannels(), static_cast<size_t>(num_frames))};
+        if (impl->should_flush) {
+            impl->time_stretch.Flush();
+            impl->should_flush = false;
+        }
+    } else {
+        samples_written = impl->queue.Pop(buffer, samples_to_write);
+    }
 
-    memcpy(buffer, impl->queue.data(), frames_to_write * sizeof(s16) * impl->GetNumChannels());
-    impl->queue.erase(impl->queue.begin(),
-                      impl->queue.begin() + frames_to_write * impl->GetNumChannels());
+    if (samples_written >= num_channels) {
+        std::memcpy(&impl->last_frame[0], buffer + (samples_written - num_channels) * sizeof(s16),
+                    num_channels * sizeof(s16));
+    }
 
-    if (frames_to_write < num_frames) {
-        // Fill the rest of the frames with silence
-        memset(buffer + frames_to_write * sizeof(s16) * impl->GetNumChannels(), 0,
-               (num_frames - frames_to_write) * sizeof(s16) * impl->GetNumChannels());
+    // Fill the rest of the frames with last_frame
+    for (size_t i = samples_written; i < samples_to_write; i += num_channels) {
+        std::memcpy(buffer + i * sizeof(s16), &impl->last_frame[0], num_channels * sizeof(s16));
     }
 
     return num_frames;
 }
 
-void SinkStreamImpl::StateCallback(cubeb_stream* stream, void* user_data, cubeb_state state) {}
+void CubebSinkStream::StateCallback(cubeb_stream* stream, void* user_data, cubeb_state state) {}
 
 std::vector<std::string> ListCubebSinkDevices() {
     std::vector<std::string> device_list;
diff --git a/src/audio_core/null_sink.h b/src/audio_core/null_sink.h
index f235d93e5..2ed0c83b6 100644
--- a/src/audio_core/null_sink.h
+++ b/src/audio_core/null_sink.h
@@ -21,6 +21,12 @@ public:
 private:
     struct NullSinkStreamImpl final : SinkStream {
         void EnqueueSamples(u32 /*num_channels*/, const std::vector<s16>& /*samples*/) override {}
+
+        size_t SamplesInQueue(u32 /*num_channels*/) const override {
+            return 0;
+        }
+
+        void Flush() override {}
     } null_sink_stream;
 };
 
diff --git a/src/audio_core/sink_stream.h b/src/audio_core/sink_stream.h
index 41b6736d8..4309ad094 100644
--- a/src/audio_core/sink_stream.h
+++ b/src/audio_core/sink_stream.h
@@ -25,6 +25,10 @@ public:
      * @param samples Samples in interleaved stereo PCM16 format.
      */
     virtual void EnqueueSamples(u32 num_channels, const std::vector<s16>& samples) = 0;
+
+    virtual std::size_t SamplesInQueue(u32 num_channels) const = 0;
+
+    virtual void Flush() = 0;
 };
 
 using SinkStreamPtr = std::unique_ptr<SinkStream>;
diff --git a/src/audio_core/stream.cpp b/src/audio_core/stream.cpp
index dbae75d8c..84dcdd98d 100644
--- a/src/audio_core/stream.cpp
+++ b/src/audio_core/stream.cpp
@@ -73,6 +73,7 @@ static void VolumeAdjustSamples(std::vector<s16>& samples) {
 void Stream::PlayNextBuffer() {
     if (!IsPlaying()) {
         // Ensure we are in playing state before playing the next buffer
+        sink_stream.Flush();
         return;
     }
 
@@ -83,6 +84,7 @@ void Stream::PlayNextBuffer() {
 
     if (queued_buffers.empty()) {
         // No queued buffers - we are effectively paused
+        sink_stream.Flush();
         return;
     }
 
@@ -90,6 +92,7 @@ void Stream::PlayNextBuffer() {
     queued_buffers.pop();
 
     VolumeAdjustSamples(active_buffer->Samples());
+
     sink_stream.EnqueueSamples(GetNumChannels(), active_buffer->GetSamples());
 
     CoreTiming::ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
diff --git a/src/audio_core/time_stretch.cpp b/src/audio_core/time_stretch.cpp
new file mode 100644
index 000000000..da094c46b
--- /dev/null
+++ b/src/audio_core/time_stretch.cpp
@@ -0,0 +1,68 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cmath>
+#include <cstddef>
+#include "audio_core/time_stretch.h"
+#include "common/logging/log.h"
+
+namespace AudioCore {
+
+TimeStretcher::TimeStretcher(u32 sample_rate, u32 channel_count)
+    : m_sample_rate(sample_rate), m_channel_count(channel_count) {
+    m_sound_touch.setChannels(channel_count);
+    m_sound_touch.setSampleRate(sample_rate);
+    m_sound_touch.setPitch(1.0);
+    m_sound_touch.setTempo(1.0);
+}
+
+void TimeStretcher::Clear() {
+    m_sound_touch.clear();
+}
+
+void TimeStretcher::Flush() {
+    m_sound_touch.flush();
+}
+
+size_t TimeStretcher::Process(const s16* in, size_t num_in, s16* out, size_t num_out) {
+    const double time_delta = static_cast<double>(num_out) / m_sample_rate; // seconds
+
+    // We were given actual_samples number of samples, and num_samples were requested from us.
+    double current_ratio = static_cast<double>(num_in) / static_cast<double>(num_out);
+
+    const double max_latency = 1.0; // seconds
+    const double max_backlog = m_sample_rate * max_latency;
+    const double backlog_fullness = m_sound_touch.numSamples() / max_backlog;
+    if (backlog_fullness > 5.0) {
+        // Too many samples in backlog: Don't push anymore on
+        num_in = 0;
+    }
+
+    // We ideally want the backlog to be about 50% full.
+    // This gives some headroom both ways to prevent underflow and overflow.
+    // We tweak current_ratio to encourage this.
+    constexpr double tweak_time_scale = 0.05; // seconds
+    const double tweak_correction = (backlog_fullness - 0.5) * (time_delta / tweak_time_scale);
+    current_ratio *= std::pow(1.0 + 2.0 * tweak_correction, tweak_correction < 0 ? 3.0 : 1.0);
+
+    // This low-pass filter smoothes out variance in the calculated stretch ratio.
+    // The time-scale determines how responsive this filter is.
+    constexpr double lpf_time_scale = 2.0; // seconds
+    const double lpf_gain = 1.0 - std::exp(-time_delta / lpf_time_scale);
+    m_stretch_ratio += lpf_gain * (current_ratio - m_stretch_ratio);
+
+    // Place a lower limit of 5% speed.  When a game boots up, there will be
+    // many silence samples.  These do not need to be timestretched.
+    m_stretch_ratio = std::max(m_stretch_ratio, 0.05);
+    m_sound_touch.setTempo(m_stretch_ratio);
+
+    LOG_DEBUG(Audio, "{:5}/{:5} ratio:{:0.6f} backlog:{:0.6f}", num_in, num_out, m_stretch_ratio,
+              backlog_fullness);
+
+    m_sound_touch.putSamples(in, num_in);
+    return m_sound_touch.receiveSamples(out, num_out);
+}
+
+} // namespace AudioCore
diff --git a/src/audio_core/time_stretch.h b/src/audio_core/time_stretch.h
new file mode 100644
index 000000000..7e39e695e
--- /dev/null
+++ b/src/audio_core/time_stretch.h
@@ -0,0 +1,36 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <cstddef>
+#include <SoundTouch.h>
+#include "common/common_types.h"
+
+namespace AudioCore {
+
+class TimeStretcher {
+public:
+    TimeStretcher(u32 sample_rate, u32 channel_count);
+
+    /// @param in       Input sample buffer
+    /// @param num_in   Number of input frames in `in`
+    /// @param out      Output sample buffer
+    /// @param num_out  Desired number of output frames in `out`
+    /// @returns Actual number of frames written to `out`
+    size_t Process(const s16* in, size_t num_in, s16* out, size_t num_out);
+
+    void Clear();
+
+    void Flush();
+
+private:
+    u32 m_sample_rate;
+    u32 m_channel_count;
+    soundtouch::SoundTouch m_sound_touch;
+    double m_stretch_ratio = 1.0;
+};
+
+} // namespace AudioCore
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index f41946cc6..6a3f1fe08 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -71,6 +71,7 @@ add_library(common STATIC
     param_package.cpp
     param_package.h
     quaternion.h
+    ring_buffer.h
     scm_rev.cpp
     scm_rev.h
     scope_exit.h
diff --git a/src/common/ring_buffer.h b/src/common/ring_buffer.h
new file mode 100644
index 000000000..30d934a38
--- /dev/null
+++ b/src/common/ring_buffer.h
@@ -0,0 +1,111 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <array>
+#include <atomic>
+#include <cstddef>
+#include <cstring>
+#include <type_traits>
+#include <vector>
+#include "common/common_types.h"
+
+namespace Common {
+
+/// SPSC ring buffer
+/// @tparam T            Element type
+/// @tparam capacity     Number of slots in ring buffer
+/// @tparam granularity  Slot size in terms of number of elements
+template <typename T, size_t capacity, size_t granularity = 1>
+class RingBuffer {
+    /// A "slot" is made of `granularity` elements of `T`.
+    static constexpr size_t slot_size = granularity * 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<size_t>::max() / 2 / granularity);
+    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
+    size_t Push(const void* new_slots, size_t slot_count) {
+        const size_t write_index = m_write_index.load();
+        const size_t slots_free = capacity + m_read_index.load() - write_index;
+        const size_t push_count = std::min(slot_count, slots_free);
+
+        const size_t pos = write_index % capacity;
+        const size_t first_copy = std::min(capacity - pos, push_count);
+        const size_t second_copy = push_count - first_copy;
+
+        const char* in = static_cast<const char*>(new_slots);
+        std::memcpy(m_data.data() + pos * granularity, 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;
+    }
+
+    size_t Push(const std::vector<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
+    size_t Pop(void* output, size_t max_slots = ~size_t(0)) {
+        const size_t read_index = m_read_index.load();
+        const size_t slots_filled = m_write_index.load() - read_index;
+        const size_t pop_count = std::min(slots_filled, max_slots);
+
+        const size_t pos = read_index % capacity;
+        const size_t first_copy = std::min(capacity - pos, pop_count);
+        const size_t second_copy = pop_count - first_copy;
+
+        char* out = static_cast<char*>(output);
+        std::memcpy(out, m_data.data() + pos * granularity, 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(size_t max_slots = ~size_t(0)) {
+        std::vector<T> out(std::min(max_slots, capacity) * granularity);
+        const size_t count = Pop(out.data(), out.size() / granularity);
+        out.resize(count * granularity);
+        return out;
+    }
+
+    /// @returns Number of slots used
+    size_t Size() const {
+        return m_write_index.load() - m_read_index.load();
+    }
+
+    /// @returns Maximum size of ring buffer
+    constexpr 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.
+    alignas(128) std::atomic<size_t> m_read_index{0};
+    alignas(128) std::atomic<size_t> m_write_index{0};
+
+    std::array<T, granularity * capacity> m_data;
+};
+
+} // namespace Common
diff --git a/src/core/hle/kernel/errors.h b/src/core/hle/kernel/errors.h
index 4054d5db6..ad39c8271 100644
--- a/src/core/hle/kernel/errors.h
+++ b/src/core/hle/kernel/errors.h
@@ -21,6 +21,7 @@ enum {
     HandleTableFull = 105,
     InvalidMemoryState = 106,
     InvalidMemoryPermissions = 108,
+    InvalidThreadPriority = 112,
     InvalidProcessorId = 113,
     InvalidHandle = 114,
     InvalidCombination = 116,
@@ -36,7 +37,7 @@ enum {
 // WARNING: The kernel is quite inconsistent in it's usage of errors code. Make sure to always
 // double check that the code matches before re-using the constant.
 
-// TODO(bunnei): Replace these with correct errors for Switch OS
+// TODO(bunnei): Replace -1 with correct errors for Switch OS
 constexpr ResultCode ERR_HANDLE_TABLE_FULL(ErrorModule::Kernel, ErrCodes::HandleTableFull);
 constexpr ResultCode ERR_SESSION_CLOSED_BY_REMOTE(-1);
 constexpr ResultCode ERR_PORT_NAME_TOO_LONG(ErrorModule::Kernel, ErrCodes::TooLarge);
@@ -53,15 +54,16 @@ constexpr ResultCode ERR_INVALID_ADDRESS_STATE(ErrorModule::Kernel, ErrCodes::In
 constexpr ResultCode ERR_INVALID_MEMORY_PERMISSIONS(ErrorModule::Kernel,
                                                     ErrCodes::InvalidMemoryPermissions);
 constexpr ResultCode ERR_INVALID_HANDLE(ErrorModule::Kernel, ErrCodes::InvalidHandle);
+constexpr ResultCode ERR_INVALID_PROCESSOR_ID(ErrorModule::Kernel, ErrCodes::InvalidProcessorId);
 constexpr ResultCode ERR_INVALID_STATE(ErrorModule::Kernel, ErrCodes::InvalidState);
+constexpr ResultCode ERR_INVALID_THREAD_PRIORITY(ErrorModule::Kernel,
+                                                 ErrCodes::InvalidThreadPriority);
 constexpr ResultCode ERR_INVALID_POINTER(-1);
 constexpr ResultCode ERR_INVALID_OBJECT_ADDR(-1);
 constexpr ResultCode ERR_NOT_AUTHORIZED(-1);
 /// Alternate code returned instead of ERR_INVALID_HANDLE in some code paths.
 constexpr ResultCode ERR_INVALID_HANDLE_OS(-1);
 constexpr ResultCode ERR_NOT_FOUND(-1);
-constexpr ResultCode ERR_OUT_OF_RANGE(-1);
-constexpr ResultCode ERR_OUT_OF_RANGE_KERNEL(-1);
 constexpr ResultCode RESULT_TIMEOUT(ErrorModule::Kernel, ErrCodes::Timeout);
 /// Returned when Accept() is called on a port with no sessions to be accepted.
 constexpr ResultCode ERR_NO_PENDING_SESSIONS(-1);
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index 1c9373ed8..f500fd2e7 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -273,7 +273,11 @@ static void Break(u64 reason, u64 info1, u64 info2) {
 }
 
 /// Used to output a message on a debug hardware unit - does nothing on a retail unit
-static void OutputDebugString(VAddr address, s32 len) {
+static void OutputDebugString(VAddr address, u64 len) {
+    if (len == 0) {
+        return;
+    }
+
     std::string str(len, '\0');
     Memory::ReadBlock(address, str.data(), str.size());
     LOG_DEBUG(Debug_Emulated, "{}", str);
@@ -378,7 +382,7 @@ static ResultCode GetThreadPriority(u32* priority, Handle handle) {
 /// Sets the priority for the specified thread
 static ResultCode SetThreadPriority(Handle handle, u32 priority) {
     if (priority > THREADPRIO_LOWEST) {
-        return ERR_OUT_OF_RANGE;
+        return ERR_INVALID_THREAD_PRIORITY;
     }
 
     auto& kernel = Core::System::GetInstance().Kernel();
@@ -523,7 +527,7 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
     std::string name = fmt::format("unknown-{:X}", entry_point);
 
     if (priority > THREADPRIO_LOWEST) {
-        return ERR_OUT_OF_RANGE;
+        return ERR_INVALID_THREAD_PRIORITY;
     }
 
     SharedPtr<ResourceLimit>& resource_limit = Core::CurrentProcess()->resource_limit;
@@ -544,8 +548,8 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
     case THREADPROCESSORID_3:
         break;
     default:
-        ASSERT_MSG(false, "Unsupported thread processor ID: {}", processor_id);
-        break;
+        LOG_ERROR(Kernel_SVC, "Invalid thread processor ID: {}", processor_id);
+        return ERR_INVALID_PROCESSOR_ID;
     }
 
     auto& kernel = Core::System::GetInstance().Kernel();
diff --git a/src/core/hle/kernel/svc_wrap.h b/src/core/hle/kernel/svc_wrap.h
index 79c3fe31b..1eda5f879 100644
--- a/src/core/hle/kernel/svc_wrap.h
+++ b/src/core/hle/kernel/svc_wrap.h
@@ -222,9 +222,9 @@ void SvcWrap() {
     func((s64)PARAM(0));
 }
 
-template <void func(u64, s32 len)>
+template <void func(u64, u64 len)>
 void SvcWrap() {
-    func(PARAM(0), (s32)(PARAM(1) & 0xFFFFFFFF));
+    func(PARAM(0), PARAM(1));
 }
 
 template <void func(u64, u64, u64)>
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index 3d10d9af2..3f12a84dc 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -227,12 +227,12 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     // Check if priority is in ranged. Lowest priority -> highest priority id.
     if (priority > THREADPRIO_LOWEST) {
         LOG_ERROR(Kernel_SVC, "Invalid thread priority: {}", priority);
-        return ERR_OUT_OF_RANGE;
+        return ERR_INVALID_THREAD_PRIORITY;
     }
 
     if (processor_id > THREADPROCESSORID_MAX) {
         LOG_ERROR(Kernel_SVC, "Invalid processor id: {}", processor_id);
-        return ERR_OUT_OF_RANGE_KERNEL;
+        return ERR_INVALID_PROCESSOR_ID;
     }
 
     // TODO(yuriks): Other checks, returning 0xD9001BEA
diff --git a/src/core/hle/service/ns/pl_u.cpp b/src/core/hle/service/ns/pl_u.cpp
index da1c46d59..ac0eaaa8f 100644
--- a/src/core/hle/service/ns/pl_u.cpp
+++ b/src/core/hle/service/ns/pl_u.cpp
@@ -2,6 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
+#include <cstring>
+#include <vector>
+
 #include <FontChineseSimplified.h>
 #include <FontChineseTraditional.h>
 #include <FontExtendedChineseSimplified.h>
@@ -9,14 +13,19 @@
 #include <FontNintendoExtended.h>
 #include <FontStandard.h>
 
+#include "common/assert.h"
 #include "common/common_paths.h"
+#include "common/common_types.h"
 #include "common/file_util.h"
+#include "common/logging/log.h"
+#include "common/swap.h"
 #include "core/core.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/shared_memory.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ns/pl_u.h"
 
@@ -35,49 +44,41 @@ struct FontRegion {
     u32 size;
 };
 
-static constexpr std::array<std::pair<FontArchives, const char*>, 7> SHARED_FONTS{
+constexpr std::array<std::pair<FontArchives, const char*>, 7> SHARED_FONTS{
     std::make_pair(FontArchives::Standard, "nintendo_udsg-r_std_003.bfttf"),
     std::make_pair(FontArchives::ChineseSimple, "nintendo_udsg-r_org_zh-cn_003.bfttf"),
     std::make_pair(FontArchives::ChineseSimple, "nintendo_udsg-r_ext_zh-cn_003.bfttf"),
     std::make_pair(FontArchives::ChineseTraditional, "nintendo_udjxh-db_zh-tw_003.bfttf"),
     std::make_pair(FontArchives::Korean, "nintendo_udsg-r_ko_003.bfttf"),
     std::make_pair(FontArchives::Extension, "nintendo_ext_003.bfttf"),
-    std::make_pair(FontArchives::Extension, "nintendo_ext2_003.bfttf")};
+    std::make_pair(FontArchives::Extension, "nintendo_ext2_003.bfttf"),
+};
 
-static constexpr std::array<const char*, 7> SHARED_FONTS_TTF{"FontStandard.ttf",
-                                                             "FontChineseSimplified.ttf",
-                                                             "FontExtendedChineseSimplified.ttf",
-                                                             "FontChineseTraditional.ttf",
-                                                             "FontKorean.ttf",
-                                                             "FontNintendoExtended.ttf",
-                                                             "FontNintendoExtended2.ttf"};
+constexpr std::array<const char*, 7> SHARED_FONTS_TTF{
+    "FontStandard.ttf",
+    "FontChineseSimplified.ttf",
+    "FontExtendedChineseSimplified.ttf",
+    "FontChineseTraditional.ttf",
+    "FontKorean.ttf",
+    "FontNintendoExtended.ttf",
+    "FontNintendoExtended2.ttf",
+};
 
 // The below data is specific to shared font data dumped from Switch on f/w 2.2
 // Virtual address and offsets/sizes likely will vary by dump
-static constexpr VAddr SHARED_FONT_MEM_VADDR{0x00000009d3016000ULL};
-static constexpr u32 EXPECTED_RESULT{
-    0x7f9a0218}; // What we expect the decrypted bfttf first 4 bytes to be
-static constexpr u32 EXPECTED_MAGIC{
-    0x36f81a1e}; // What we expect the encrypted bfttf first 4 bytes to be
-static constexpr u64 SHARED_FONT_MEM_SIZE{0x1100000};
-static constexpr FontRegion EMPTY_REGION{0, 0};
-std::vector<FontRegion>
-    SHARED_FONT_REGIONS{}; // Automatically populated based on shared_fonts dump or system archives
-
-const FontRegion& GetSharedFontRegion(size_t index) {
-    if (index >= SHARED_FONT_REGIONS.size() || SHARED_FONT_REGIONS.empty()) {
-        // No font fallback
-        return EMPTY_REGION;
-    }
-    return SHARED_FONT_REGIONS.at(index);
-}
+constexpr VAddr SHARED_FONT_MEM_VADDR{0x00000009d3016000ULL};
+constexpr u32 EXPECTED_RESULT{0x7f9a0218}; // What we expect the decrypted bfttf first 4 bytes to be
+constexpr u32 EXPECTED_MAGIC{0x36f81a1e};  // What we expect the encrypted bfttf first 4 bytes to be
+constexpr u64 SHARED_FONT_MEM_SIZE{0x1100000};
+constexpr FontRegion EMPTY_REGION{0, 0};
 
 enum class LoadState : u32 {
     Loading = 0,
     Done = 1,
 };
 
-void DecryptSharedFont(const std::vector<u32>& input, std::vector<u8>& output, size_t& offset) {
+static void DecryptSharedFont(const std::vector<u32>& input, std::vector<u8>& output,
+                              size_t& offset) {
     ASSERT_MSG(offset + (input.size() * sizeof(u32)) < SHARED_FONT_MEM_SIZE,
                "Shared fonts exceeds 17mb!");
     ASSERT_MSG(input[0] == EXPECTED_MAGIC, "Failed to derive key, unexpected magic number");
@@ -104,28 +105,52 @@ static void EncryptSharedFont(const std::vector<u8>& input, std::vector<u8>& out
     offset += input.size() + (sizeof(u32) * 2);
 }
 
+// Helper function to make BuildSharedFontsRawRegions a bit nicer
 static u32 GetU32Swapped(const u8* data) {
     u32 value;
     std::memcpy(&value, data, sizeof(value));
-    return Common::swap32(value); // Helper function to make BuildSharedFontsRawRegions a bit nicer
+    return Common::swap32(value);
 }
 
-void BuildSharedFontsRawRegions(const std::vector<u8>& input) {
-    unsigned cur_offset = 0; // As we can derive the xor key we can just populate the offsets based
-                             // on the shared memory dump
-    for (size_t i = 0; i < SHARED_FONTS.size(); i++) {
-        // Out of shared fonts/Invalid font
-        if (GetU32Swapped(input.data() + cur_offset) != EXPECTED_RESULT)
-            break;
-        const u32 KEY = GetU32Swapped(input.data() + cur_offset) ^
-                        EXPECTED_MAGIC; // Derive key withing inverse xor
-        const u32 SIZE = GetU32Swapped(input.data() + cur_offset + 4) ^ KEY;
-        SHARED_FONT_REGIONS.push_back(FontRegion{cur_offset + 8, SIZE});
-        cur_offset += SIZE + 8;
+struct PL_U::Impl {
+    const FontRegion& GetSharedFontRegion(size_t index) const {
+        if (index >= shared_font_regions.size() || shared_font_regions.empty()) {
+            // No font fallback
+            return EMPTY_REGION;
+        }
+        return shared_font_regions.at(index);
     }
-}
 
-PL_U::PL_U() : ServiceFramework("pl:u") {
+    void BuildSharedFontsRawRegions(const std::vector<u8>& input) {
+        // As we can derive the xor key we can just populate the offsets
+        // based on the shared memory dump
+        unsigned cur_offset = 0;
+
+        for (size_t i = 0; i < SHARED_FONTS.size(); i++) {
+            // Out of shared fonts/invalid font
+            if (GetU32Swapped(input.data() + cur_offset) != EXPECTED_RESULT) {
+                break;
+            }
+
+            // Derive key withing inverse xor
+            const u32 KEY = GetU32Swapped(input.data() + cur_offset) ^ EXPECTED_MAGIC;
+            const u32 SIZE = GetU32Swapped(input.data() + cur_offset + 4) ^ KEY;
+            shared_font_regions.push_back(FontRegion{cur_offset + 8, SIZE});
+            cur_offset += SIZE + 8;
+        }
+    }
+
+    /// Handle to shared memory region designated for a shared font
+    Kernel::SharedPtr<Kernel::SharedMemory> shared_font_mem;
+
+    /// Backing memory for the shared font data
+    std::shared_ptr<std::vector<u8>> shared_font;
+
+    // Automatically populated based on shared_fonts dump or system archives.
+    std::vector<FontRegion> shared_font_regions;
+};
+
+PL_U::PL_U() : ServiceFramework("pl:u"), impl{std::make_unique<Impl>()} {
     static const FunctionInfo functions[] = {
         {0, &PL_U::RequestLoad, "RequestLoad"},
         {1, &PL_U::GetLoadState, "GetLoadState"},
@@ -141,7 +166,7 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
     // Rebuild shared fonts from data ncas
     if (nand->HasEntry(static_cast<u64>(FontArchives::Standard),
                        FileSys::ContentRecordType::Data)) {
-        shared_font = std::make_shared<std::vector<u8>>(SHARED_FONT_MEM_SIZE);
+        impl->shared_font = std::make_shared<std::vector<u8>>(SHARED_FONT_MEM_SIZE);
         for (auto font : SHARED_FONTS) {
             const auto nca =
                 nand->GetEntry(static_cast<u64>(font.first), FileSys::ContentRecordType::Data);
@@ -177,12 +202,12 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
                 static_cast<u32>(offset + 8),
                 static_cast<u32>((font_data_u32.size() * sizeof(u32)) -
                                  8)}; // Font offset and size do not account for the header
-            DecryptSharedFont(font_data_u32, *shared_font, offset);
-            SHARED_FONT_REGIONS.push_back(region);
+            DecryptSharedFont(font_data_u32, *impl->shared_font, offset);
+            impl->shared_font_regions.push_back(region);
         }
 
     } else {
-        shared_font = std::make_shared<std::vector<u8>>(
+        impl->shared_font = std::make_shared<std::vector<u8>>(
             SHARED_FONT_MEM_SIZE); // Shared memory needs to always be allocated and a fixed size
 
         const std::string user_path = FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir);
@@ -206,8 +231,8 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
                         static_cast<u32>(offset + 8),
                         static_cast<u32>(ttf_bytes.size())}; // Font offset and size do not account
                                                              // for the header
-                    EncryptSharedFont(ttf_bytes, *shared_font, offset);
-                    SHARED_FONT_REGIONS.push_back(region);
+                    EncryptSharedFont(ttf_bytes, *impl->shared_font, offset);
+                    impl->shared_font_regions.push_back(region);
                 } else {
                     LOG_WARNING(Service_NS, "Unable to load font: {}", font_ttf);
                 }
@@ -222,8 +247,8 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
         if (file.IsOpen()) {
             // Read shared font data
             ASSERT(file.GetSize() == SHARED_FONT_MEM_SIZE);
-            file.ReadBytes(shared_font->data(), shared_font->size());
-            BuildSharedFontsRawRegions(*shared_font);
+            file.ReadBytes(impl->shared_font->data(), impl->shared_font->size());
+            impl->BuildSharedFontsRawRegions(*impl->shared_font);
         } else {
             LOG_WARNING(Service_NS,
                         "Shared Font file missing. Loading open source replacement from memory");
@@ -240,8 +265,8 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
             for (const std::vector<u8>& font_ttf : open_source_shared_fonts_ttf) {
                 const FontRegion region{static_cast<u32>(offset + 8),
                                         static_cast<u32>(font_ttf.size())};
-                EncryptSharedFont(font_ttf, *shared_font, offset);
-                SHARED_FONT_REGIONS.push_back(region);
+                EncryptSharedFont(font_ttf, *impl->shared_font, offset);
+                impl->shared_font_regions.push_back(region);
             }
         }
     }
@@ -275,7 +300,7 @@ void PL_U::GetSize(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_NS, "called, font_id={}", font_id);
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push<u32>(GetSharedFontRegion(font_id).size);
+    rb.Push<u32>(impl->GetSharedFontRegion(font_id).size);
 }
 
 void PL_U::GetSharedMemoryAddressOffset(Kernel::HLERequestContext& ctx) {
@@ -285,17 +310,18 @@ void PL_U::GetSharedMemoryAddressOffset(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_NS, "called, font_id={}", font_id);
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push<u32>(GetSharedFontRegion(font_id).offset);
+    rb.Push<u32>(impl->GetSharedFontRegion(font_id).offset);
 }
 
 void PL_U::GetSharedMemoryNativeHandle(Kernel::HLERequestContext& ctx) {
     // Map backing memory for the font data
-    Core::CurrentProcess()->vm_manager.MapMemoryBlock(
-        SHARED_FONT_MEM_VADDR, shared_font, 0, SHARED_FONT_MEM_SIZE, Kernel::MemoryState::Shared);
+    Core::CurrentProcess()->vm_manager.MapMemoryBlock(SHARED_FONT_MEM_VADDR, impl->shared_font, 0,
+                                                      SHARED_FONT_MEM_SIZE,
+                                                      Kernel::MemoryState::Shared);
 
     // Create shared font memory object
     auto& kernel = Core::System::GetInstance().Kernel();
-    shared_font_mem = Kernel::SharedMemory::Create(
+    impl->shared_font_mem = Kernel::SharedMemory::Create(
         kernel, Core::CurrentProcess(), SHARED_FONT_MEM_SIZE, Kernel::MemoryPermission::ReadWrite,
         Kernel::MemoryPermission::Read, SHARED_FONT_MEM_VADDR, Kernel::MemoryRegion::BASE,
         "PL_U:shared_font_mem");
@@ -303,7 +329,7 @@ void PL_U::GetSharedMemoryNativeHandle(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_NS, "called");
     IPC::ResponseBuilder rb{ctx, 2, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushCopyObjects(shared_font_mem);
+    rb.PushCopyObjects(impl->shared_font_mem);
 }
 
 void PL_U::GetSharedFontInOrderOfPriority(Kernel::HLERequestContext& ctx) {
@@ -316,9 +342,9 @@ void PL_U::GetSharedFontInOrderOfPriority(Kernel::HLERequestContext& ctx) {
     std::vector<u32> font_sizes;
 
     // TODO(ogniK): Have actual priority order
-    for (size_t i = 0; i < SHARED_FONT_REGIONS.size(); i++) {
+    for (size_t i = 0; i < impl->shared_font_regions.size(); i++) {
         font_codes.push_back(static_cast<u32>(i));
-        auto region = GetSharedFontRegion(i);
+        auto region = impl->GetSharedFontRegion(i);
         font_offsets.push_back(region.offset);
         font_sizes.push_back(region.size);
     }
diff --git a/src/core/hle/service/ns/pl_u.h b/src/core/hle/service/ns/pl_u.h
index 296c3db05..253f26a2a 100644
--- a/src/core/hle/service/ns/pl_u.h
+++ b/src/core/hle/service/ns/pl_u.h
@@ -5,7 +5,6 @@
 #pragma once
 
 #include <memory>
-#include "core/hle/kernel/shared_memory.h"
 #include "core/hle/service/service.h"
 
 namespace Service::NS {
@@ -23,11 +22,8 @@ private:
     void GetSharedMemoryNativeHandle(Kernel::HLERequestContext& ctx);
     void GetSharedFontInOrderOfPriority(Kernel::HLERequestContext& ctx);
 
-    /// Handle to shared memory region designated for a shared font
-    Kernel::SharedPtr<Kernel::SharedMemory> shared_font_mem;
-
-    /// Backing memory for the shared font data
-    std::shared_ptr<std::vector<u8>> shared_font;
+    struct Impl;
+    std::unique_ptr<Impl> impl;
 };
 
 } // namespace Service::NS
diff --git a/src/core/hle/service/prepo/prepo.cpp b/src/core/hle/service/prepo/prepo.cpp
index 3c43b8d8c..6a9eccfb5 100644
--- a/src/core/hle/service/prepo/prepo.cpp
+++ b/src/core/hle/service/prepo/prepo.cpp
@@ -1,36 +1,47 @@
-#include <cinttypes>
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
-#include "core/hle/kernel/event.h"
 #include "core/hle/service/prepo/prepo.h"
+#include "core/hle/service/service.h"
 
 namespace Service::PlayReport {
-PlayReport::PlayReport(const char* name) : ServiceFramework(name) {
-    static const FunctionInfo functions[] = {
-        {10100, nullptr, "SaveReport"},
-        {10101, &PlayReport::SaveReportWithUser, "SaveReportWithUser"},
-        {10200, nullptr, "RequestImmediateTransmission"},
-        {10300, nullptr, "GetTransmissionStatus"},
-        {20100, nullptr, "SaveSystemReport"},
-        {20200, nullptr, "SetOperationMode"},
-        {20101, nullptr, "SaveSystemReportWithUser"},
-        {30100, nullptr, "ClearStorage"},
-        {40100, nullptr, "IsUserAgreementCheckEnabled"},
-        {40101, nullptr, "SetUserAgreementCheckEnabled"},
-        {90100, nullptr, "GetStorageUsage"},
-        {90200, nullptr, "GetStatistics"},
-        {90201, nullptr, "GetThroughputHistory"},
-        {90300, nullptr, "GetLastUploadError"},
-    };
-    RegisterHandlers(functions);
-};
 
-void PlayReport::SaveReportWithUser(Kernel::HLERequestContext& ctx) {
-    // TODO(ogniK): Do we want to add play report?
-    LOG_WARNING(Service_PREPO, "(STUBBED) called");
+class PlayReport final : public ServiceFramework<PlayReport> {
+public:
+    explicit PlayReport(const char* name) : ServiceFramework{name} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {10100, nullptr, "SaveReport"},
+            {10101, &PlayReport::SaveReportWithUser, "SaveReportWithUser"},
+            {10200, nullptr, "RequestImmediateTransmission"},
+            {10300, nullptr, "GetTransmissionStatus"},
+            {20100, nullptr, "SaveSystemReport"},
+            {20200, nullptr, "SetOperationMode"},
+            {20101, nullptr, "SaveSystemReportWithUser"},
+            {30100, nullptr, "ClearStorage"},
+            {40100, nullptr, "IsUserAgreementCheckEnabled"},
+            {40101, nullptr, "SetUserAgreementCheckEnabled"},
+            {90100, nullptr, "GetStorageUsage"},
+            {90200, nullptr, "GetStatistics"},
+            {90201, nullptr, "GetThroughputHistory"},
+            {90300, nullptr, "GetLastUploadError"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void SaveReportWithUser(Kernel::HLERequestContext& ctx) {
+        // TODO(ogniK): Do we want to add play report?
+        LOG_WARNING(Service_PREPO, "(STUBBED) called");
 
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
 };
 
 void InstallInterfaces(SM::ServiceManager& service_manager) {
diff --git a/src/core/hle/service/prepo/prepo.h b/src/core/hle/service/prepo/prepo.h
index f5a6aba6d..0e7b01331 100644
--- a/src/core/hle/service/prepo/prepo.h
+++ b/src/core/hle/service/prepo/prepo.h
@@ -4,22 +4,12 @@
 
 #pragma once
 
-#include <memory>
-#include <string>
-#include "core/hle/kernel/event.h"
-#include "core/hle/service/service.h"
+namespace Service::SM {
+class ServiceManager;
+}
 
 namespace Service::PlayReport {
 
-class PlayReport final : public ServiceFramework<PlayReport> {
-public:
-    explicit PlayReport(const char* name);
-    ~PlayReport() = default;
-
-private:
-    void SaveReportWithUser(Kernel::HLERequestContext& ctx);
-};
-
 void InstallInterfaces(SM::ServiceManager& service_manager);
 
 } // namespace Service::PlayReport
diff --git a/src/core/settings.h b/src/core/settings.h
index 08a16ef2c..0318d019c 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -148,6 +148,7 @@ struct Values {
 
     // Audio
     std::string sink_id;
+    bool enable_audio_stretching;
     std::string audio_device_id;
     float volume;
 
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index 3730e85b8..b0df154ca 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -120,6 +120,9 @@ TelemetrySession::TelemetrySession() {
     Telemetry::AppendOSInfo(field_collection);
 
     // Log user configuration information
+    AddField(Telemetry::FieldType::UserConfig, "Audio_SinkId", Settings::values.sink_id);
+    AddField(Telemetry::FieldType::UserConfig, "Audio_EnableAudioStretching",
+             Settings::values.enable_audio_stretching);
     AddField(Telemetry::FieldType::UserConfig, "Core_UseCpuJit", Settings::values.use_cpu_jit);
     AddField(Telemetry::FieldType::UserConfig, "Core_UseMultiCore",
              Settings::values.use_multi_core);
diff --git a/src/tests/CMakeLists.txt b/src/tests/CMakeLists.txt
index 4d74bb395..4e75a72ec 100644
--- a/src/tests/CMakeLists.txt
+++ b/src/tests/CMakeLists.txt
@@ -1,5 +1,6 @@
 add_executable(tests
     common/param_package.cpp
+    common/ring_buffer.cpp
     core/arm/arm_test_common.cpp
     core/arm/arm_test_common.h
     core/core_timing.cpp
diff --git a/src/tests/common/ring_buffer.cpp b/src/tests/common/ring_buffer.cpp
new file mode 100644
index 000000000..f3fe57839
--- /dev/null
+++ b/src/tests/common/ring_buffer.cpp
@@ -0,0 +1,130 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <cstddef>
+#include <numeric>
+#include <thread>
+#include <vector>
+#include <catch2/catch.hpp>
+#include "common/ring_buffer.h"
+
+namespace Common {
+
+TEST_CASE("RingBuffer: Basic Tests", "[common]") {
+    RingBuffer<char, 4, 1> buf;
+
+    // Pushing values into a ring buffer with space should succeed.
+    for (size_t i = 0; i < 4; i++) {
+        const char elem = static_cast<char>(i);
+        const size_t count = buf.Push(&elem, 1);
+        REQUIRE(count == 1);
+    }
+
+    REQUIRE(buf.Size() == 4);
+
+    // Pushing values into a full ring buffer should fail.
+    {
+        const char elem = static_cast<char>(42);
+        const size_t count = buf.Push(&elem, 1);
+        REQUIRE(count == 0);
+    }
+
+    REQUIRE(buf.Size() == 4);
+
+    // Popping multiple values from a ring buffer with values should succeed.
+    {
+        const std::vector<char> popped = buf.Pop(2);
+        REQUIRE(popped.size() == 2);
+        REQUIRE(popped[0] == 0);
+        REQUIRE(popped[1] == 1);
+    }
+
+    REQUIRE(buf.Size() == 2);
+
+    // Popping a single value from a ring buffer with values should succeed.
+    {
+        const std::vector<char> popped = buf.Pop(1);
+        REQUIRE(popped.size() == 1);
+        REQUIRE(popped[0] == 2);
+    }
+
+    REQUIRE(buf.Size() == 1);
+
+    // Pushing more values than space available should partially suceed.
+    {
+        std::vector<char> to_push(6);
+        std::iota(to_push.begin(), to_push.end(), 88);
+        const size_t count = buf.Push(to_push);
+        REQUIRE(count == 3);
+    }
+
+    REQUIRE(buf.Size() == 4);
+
+    // Doing an unlimited pop should pop all values.
+    {
+        const std::vector<char> popped = buf.Pop();
+        REQUIRE(popped.size() == 4);
+        REQUIRE(popped[0] == 3);
+        REQUIRE(popped[1] == 88);
+        REQUIRE(popped[2] == 89);
+        REQUIRE(popped[3] == 90);
+    }
+
+    REQUIRE(buf.Size() == 0);
+}
+
+TEST_CASE("RingBuffer: Threaded Test", "[common]") {
+    RingBuffer<char, 4, 2> buf;
+    const char seed = 42;
+    const size_t count = 1000000;
+    size_t full = 0;
+    size_t empty = 0;
+
+    const auto next_value = [](std::array<char, 2>& value) {
+        value[0] += 1;
+        value[1] += 2;
+    };
+
+    std::thread producer{[&] {
+        std::array<char, 2> value = {seed, seed};
+        size_t i = 0;
+        while (i < count) {
+            if (const size_t c = buf.Push(&value[0], 1); c > 0) {
+                REQUIRE(c == 1);
+                i++;
+                next_value(value);
+            } else {
+                full++;
+                std::this_thread::yield();
+            }
+        }
+    }};
+
+    std::thread consumer{[&] {
+        std::array<char, 2> value = {seed, seed};
+        size_t i = 0;
+        while (i < count) {
+            if (const std::vector<char> v = buf.Pop(1); v.size() > 0) {
+                REQUIRE(v.size() == 2);
+                REQUIRE(v[0] == value[0]);
+                REQUIRE(v[1] == value[1]);
+                i++;
+                next_value(value);
+            } else {
+                empty++;
+                std::this_thread::yield();
+            }
+        }
+    }};
+
+    producer.join();
+    consumer.join();
+
+    REQUIRE(buf.Size() == 0);
+    printf("RingBuffer: Threaded Test: full: %zu, empty: %zu\n", full, empty);
+}
+
+} // namespace Common
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 2db906ea5..58f2904ce 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -254,6 +254,15 @@ enum class TextureQueryType : u64 {
     BorderColor = 22,
 };
 
+enum class TextureProcessMode : u64 {
+    None = 0,
+    LZ = 1,  // Unknown, appears to be the same as none.
+    LB = 2,  // Load Bias.
+    LL = 3,  // Load LOD (LevelOfDetail)
+    LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB
+    LLA = 7  // Load LOD. The A is unknown, does not appear to differ with LL
+};
+
 enum class IpaInterpMode : u64 { Linear = 0, Perspective = 1, Flat = 2, Sc = 3 };
 enum class IpaSampleMode : u64 { Default = 0, Centroid = 1, Offset = 2 };
 
@@ -573,6 +582,7 @@ union Instruction {
         BitField<28, 1, u64> array;
         BitField<29, 2, TextureType> texture_type;
         BitField<31, 4, u64> component_mask;
+        BitField<55, 3, TextureProcessMode> process_mode;
 
         bool IsComponentEnabled(size_t component) const {
             return ((1ull << component) & component_mask) != 0;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index fb56decc0..32001e44b 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -116,7 +116,7 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // ABGR8U
     {GL_RGBA8, GL_RGBA, GL_BYTE, ComponentType::SNorm, false},                     // ABGR8S
     {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE, ComponentType::UInt, false},   // ABGR8UI
-    {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, ComponentType::UNorm, false},    // B5G6R5U
+    {GL_RGB8, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, ComponentType::UNorm, false},   // B5G6R5U
     {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, ComponentType::UNorm,
      false}, // A2B10G10R10U
     {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, ComponentType::UNorm, false}, // A1B5G5R5U
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 582c811e0..2d56370c7 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -1853,15 +1853,47 @@ private:
                     coord = "vec2 coords = vec2(" + x + ", " + y + ");";
                     texture_type = Tegra::Shader::TextureType::Texture2D;
                 }
+                // TODO: make sure coordinates are always indexed to gpr8 and gpr20 is always bias
+                // or lod.
+                const std::string op_c = regs.GetRegisterAsFloat(instr.gpr20);
 
                 const std::string sampler = GetSampler(instr.sampler, texture_type, false);
                 // Add an extra scope and declare the texture coords inside to prevent
                 // overwriting them in case they are used as outputs of the texs instruction.
+
                 shader.AddLine("{");
                 ++shader.scope;
                 shader.AddLine(coord);
-                const std::string texture = "texture(" + sampler + ", coords)";
+                std::string texture;
 
+                switch (instr.tex.process_mode) {
+                case Tegra::Shader::TextureProcessMode::None: {
+                    texture = "texture(" + sampler + ", coords)";
+                    break;
+                }
+                case Tegra::Shader::TextureProcessMode::LZ: {
+                    texture = "textureLod(" + sampler + ", coords, 0.0)";
+                    break;
+                }
+                case Tegra::Shader::TextureProcessMode::LB:
+                case Tegra::Shader::TextureProcessMode::LBA: {
+                    // TODO: Figure if A suffix changes the equation at all.
+                    texture = "texture(" + sampler + ", coords, " + op_c + ')';
+                    break;
+                }
+                case Tegra::Shader::TextureProcessMode::LL:
+                case Tegra::Shader::TextureProcessMode::LLA: {
+                    // TODO: Figure if A suffix changes the equation at all.
+                    texture = "textureLod(" + sampler + ", coords, " + op_c + ')';
+                    break;
+                }
+                default: {
+                    texture = "texture(" + sampler + ", coords)";
+                    LOG_CRITICAL(HW_GPU, "Unhandled texture process mode {}",
+                                 static_cast<u32>(instr.tex.process_mode.Value()));
+                    UNREACHABLE();
+                }
+                }
                 size_t dest_elem{};
                 for (size_t elem = 0; elem < 4; ++elem) {
                     if (!instr.tex.IsComponentEnabled(elem)) {
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index c43e79e78..d229225b4 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -95,6 +95,8 @@ void Config::ReadValues() {
 
     qt_config->beginGroup("Audio");
     Settings::values.sink_id = qt_config->value("output_engine", "auto").toString().toStdString();
+    Settings::values.enable_audio_stretching =
+        qt_config->value("enable_audio_stretching", true).toBool();
     Settings::values.audio_device_id =
         qt_config->value("output_device", "auto").toString().toStdString();
     Settings::values.volume = qt_config->value("volume", 1).toFloat();
@@ -230,6 +232,7 @@ void Config::SaveValues() {
 
     qt_config->beginGroup("Audio");
     qt_config->setValue("output_engine", QString::fromStdString(Settings::values.sink_id));
+    qt_config->setValue("enable_audio_stretching", Settings::values.enable_audio_stretching);
     qt_config->setValue("output_device", QString::fromStdString(Settings::values.audio_device_id));
     qt_config->setValue("volume", Settings::values.volume);
     qt_config->endGroup();
diff --git a/src/yuzu/configuration/configure_audio.cpp b/src/yuzu/configuration/configure_audio.cpp
index fbb813f6c..6ea59f2a3 100644
--- a/src/yuzu/configuration/configure_audio.cpp
+++ b/src/yuzu/configuration/configure_audio.cpp
@@ -46,6 +46,8 @@ void ConfigureAudio::setConfiguration() {
     }
     ui->output_sink_combo_box->setCurrentIndex(new_sink_index);
 
+    ui->toggle_audio_stretching->setChecked(Settings::values.enable_audio_stretching);
+
     // The device list cannot be pre-populated (nor listed) until the output sink is known.
     updateAudioDevices(new_sink_index);
 
@@ -67,6 +69,7 @@ void ConfigureAudio::applyConfiguration() {
     Settings::values.sink_id =
         ui->output_sink_combo_box->itemText(ui->output_sink_combo_box->currentIndex())
             .toStdString();
+    Settings::values.enable_audio_stretching = ui->toggle_audio_stretching->isChecked();
     Settings::values.audio_device_id =
         ui->audio_device_combo_box->itemText(ui->audio_device_combo_box->currentIndex())
             .toStdString();
diff --git a/src/yuzu/configuration/configure_audio.ui b/src/yuzu/configuration/configure_audio.ui
index ef67890dc..a29a0e265 100644
--- a/src/yuzu/configuration/configure_audio.ui
+++ b/src/yuzu/configuration/configure_audio.ui
@@ -31,6 +31,16 @@
         </item>
        </layout>
       </item>
+       <item>
+         <widget class="QCheckBox" name="toggle_audio_stretching">
+           <property name="toolTip">
+             <string>This post-processing effect adjusts audio speed to match emulation speed and helps prevent audio stutter. This however increases audio latency.</string>
+           </property>
+           <property name="text">
+             <string>Enable audio stretching</string>
+           </property>
+         </widget>
+       </item>
       <item>
        <layout class="QHBoxLayout">
         <item>
diff --git a/src/yuzu/game_list.cpp b/src/yuzu/game_list.cpp
index 8c6e16d47..3b3b551bb 100644
--- a/src/yuzu/game_list.cpp
+++ b/src/yuzu/game_list.cpp
@@ -366,7 +366,7 @@ void GameList::LoadCompatibilityList() {
     QJsonDocument json = QJsonDocument::fromJson(string_content.toUtf8());
     QJsonArray arr = json.array();
 
-    for (const QJsonValue& value : arr) {
+    for (const QJsonValueRef& value : arr) {
         QJsonObject game = value.toObject();
 
         if (game.contains("compatibility") && game["compatibility"].isDouble()) {
@@ -374,9 +374,9 @@ void GameList::LoadCompatibilityList() {
             QString directory = game["directory"].toString();
             QJsonArray ids = game["releases"].toArray();
 
-            for (const QJsonValue& value : ids) {
-                QJsonObject object = value.toObject();
-                QString id = object["id"].toString();
+            for (const QJsonValueRef& id_ref : ids) {
+                QJsonObject id_object = id_ref.toObject();
+                QString id = id_object["id"].toString();
                 compatibility_list.emplace(
                     id.toUpper().toStdString(),
                     std::make_pair(QString::number(compatibility), directory));
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index f00b5a66b..991abda2e 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -108,6 +108,8 @@ void Config::ReadValues() {
 
     // Audio
     Settings::values.sink_id = sdl2_config->Get("Audio", "output_engine", "auto");
+    Settings::values.enable_audio_stretching =
+        sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true);
     Settings::values.audio_device_id = sdl2_config->Get("Audio", "output_device", "auto");
     Settings::values.volume = sdl2_config->GetReal("Audio", "volume", 1);
 
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index 6ed9e7962..002a4ec15 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -150,6 +150,12 @@ swap_screen =
 # auto (default): Auto-select, null: No audio output, cubeb: Cubeb audio engine (if available)
 output_engine =
 
+# Whether or not to enable the audio-stretching post-processing effect.
+# This effect adjusts audio speed to match emulation speed and helps prevent audio stutter,
+# at the cost of increasing audio latency.
+# 0: No, 1 (default): Yes
+enable_audio_stretching =
+
 # Which audio device to use.
 # auto (default): Auto-select
 output_device =