externals: Update dynarmic to latest

Updates dynarmic to its latest commit. Includes a fix for argument
limits while compiling with Clang 12.

dist/qt_themes/colorful_dark/icons/index.theme

@@ -1,7 +1,7 @@
 [Icon Theme]
 Name=colorful_dark
 Comment=Colorful theme (Dark style)
-Inherits=default
+Inherits=colorful
 Directories=16x16
  
 [16x16]

dist/qt_themes/colorful_midnight_blue/icons/index.theme

@@ -1,7 +1,7 @@
 [Icon Theme]
 Name=colorful_midnight_blue
 Comment=Colorful theme (Midnight Blue style)
-Inherits=default
+Inherits=colorful
 Directories=16x16
 
 [16x16]

dist/qt_themes/qdarkstyle_midnight_blue/style.qss

@@ -1257,10 +1257,6 @@ QComboBox::item:alternate {
   background: #19232D;
 }
 
-QComboBox::item:checked {
-  font-weight: bold;
-}
-
 QComboBox::item:selected {
   border: 0px solid transparent;
 }

src/audio_core/CMakeLists.txt

@@ -15,6 +15,8 @@ add_library(audio_core STATIC
     command_generator.cpp
     command_generator.h
     common.h
+    delay_line.cpp
+    delay_line.h
     effect_context.cpp
     effect_context.h
     info_updater.cpp

src/audio_core/command_generator.cpp

@@ -2,6 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <cmath>
+#include <numbers>
 #include "audio_core/algorithm/interpolate.h"
 #include "audio_core/command_generator.h"
 #include "audio_core/effect_context.h"
@@ -13,6 +15,20 @@ namespace AudioCore {
 namespace {
 constexpr std::size_t MIX_BUFFER_SIZE = 0x3f00;
 constexpr std::size_t SCALED_MIX_BUFFER_SIZE = MIX_BUFFER_SIZE << 15ULL;
+using DelayLineTimes = std::array<f32, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT>;
+
+constexpr DelayLineTimes FDN_MIN_DELAY_LINE_TIMES{5.0f, 6.0f, 13.0f, 14.0f};
+constexpr DelayLineTimes FDN_MAX_DELAY_LINE_TIMES{45.704f, 82.782f, 149.94f, 271.58f};
+constexpr DelayLineTimes DECAY0_MAX_DELAY_LINE_TIMES{17.0f, 13.0f, 9.0f, 7.0f};
+constexpr DelayLineTimes DECAY1_MAX_DELAY_LINE_TIMES{19.0f, 11.0f, 10.0f, 6.0f};
+constexpr std::array<f32, AudioCommon::I3DL2REVERB_TAPS> EARLY_TAP_TIMES{
+    0.017136f, 0.059154f, 0.161733f, 0.390186f, 0.425262f, 0.455411f, 0.689737f,
+    0.745910f, 0.833844f, 0.859502f, 0.000000f, 0.075024f, 0.168788f, 0.299901f,
+    0.337443f, 0.371903f, 0.599011f, 0.716741f, 0.817859f, 0.851664f};
+constexpr std::array<f32, AudioCommon::I3DL2REVERB_TAPS> EARLY_GAIN{
+    0.67096f, 0.61027f, 1.0f,     0.35680f, 0.68361f, 0.65978f, 0.51939f,
+    0.24712f, 0.45945f, 0.45021f, 0.64196f, 0.54879f, 0.92925f, 0.38270f,
+    0.72867f, 0.69794f, 0.5464f,  0.24563f, 0.45214f, 0.44042f};
 
 template <std::size_t N>
 void ApplyMix(s32* output, const s32* input, s32 gain, s32 sample_count) {
@@ -65,6 +81,154 @@ s32 ApplyMixDepop(s32* output, s32 first_sample, s32 delta, s32 sample_count) {
     }
 }
 
+float Pow10(float x) {
+    if (x >= 0.0f) {
+        return 1.0f;
+    } else if (x <= -5.3f) {
+        return 0.0f;
+    }
+    return std::pow(10.0f, x);
+}
+
+float SinD(float degrees) {
+    return std::sin(degrees * std::numbers::pi_v<float> / 180.0f);
+}
+
+float CosD(float degrees) {
+    return std::cos(degrees * std::numbers::pi_v<float> / 180.0f);
+}
+
+float ToFloat(s32 sample) {
+    return static_cast<float>(sample) / 65536.f;
+}
+
+s32 ToS32(float sample) {
+    constexpr auto min = -8388608.0f;
+    constexpr auto max = 8388607.f;
+    float rescaled_sample = sample * 65536.0f;
+    if (rescaled_sample < min) {
+        rescaled_sample = min;
+    }
+    if (rescaled_sample > max) {
+        rescaled_sample = max;
+    }
+    return static_cast<s32>(rescaled_sample);
+}
+
+constexpr std::array<std::size_t, 20> REVERB_TAP_INDEX_1CH{0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+                                                           0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+constexpr std::array<std::size_t, 20> REVERB_TAP_INDEX_2CH{0, 0, 0, 1, 1, 1, 1, 0, 0, 0,
+                                                           1, 1, 1, 0, 0, 0, 0, 1, 1, 1};
+
+constexpr std::array<std::size_t, 20> REVERB_TAP_INDEX_4CH{0, 0, 0, 1, 1, 1, 1, 2, 2, 2,
+                                                           1, 1, 1, 0, 0, 0, 0, 3, 3, 3};
+
+constexpr std::array<std::size_t, 20> REVERB_TAP_INDEX_6CH{4, 0, 0, 1, 1, 1, 1, 2, 2, 2,
+                                                           1, 1, 1, 0, 0, 0, 0, 3, 3, 3};
+
+template <std::size_t CHANNEL_COUNT>
+void ApplyReverbGeneric(I3dl2ReverbState& state,
+                        const std::array<const s32*, AudioCommon::MAX_CHANNEL_COUNT>& input,
+                        const std::array<s32*, AudioCommon::MAX_CHANNEL_COUNT>& output,
+                        s32 sample_count) {
+
+    auto GetTapLookup = []() {
+        if constexpr (CHANNEL_COUNT == 1) {
+            return REVERB_TAP_INDEX_1CH;
+        } else if constexpr (CHANNEL_COUNT == 2) {
+            return REVERB_TAP_INDEX_2CH;
+        } else if constexpr (CHANNEL_COUNT == 4) {
+            return REVERB_TAP_INDEX_4CH;
+        } else if constexpr (CHANNEL_COUNT == 6) {
+            return REVERB_TAP_INDEX_6CH;
+        }
+    };
+
+    const auto& tap_index_lut = GetTapLookup();
+    for (s32 sample = 0; sample < sample_count; sample++) {
+        std::array<f32, CHANNEL_COUNT> out_samples{};
+        std::array<f32, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> fsamp{};
+        std::array<f32, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> mixed{};
+        std::array<f32, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> osamp{};
+
+        // Mix everything into a single sample
+        s32 temp_mixed_sample = 0;
+        for (std::size_t i = 0; i < CHANNEL_COUNT; i++) {
+            temp_mixed_sample += input[i][sample];
+        }
+        const auto current_sample = ToFloat(temp_mixed_sample);
+        const auto early_tap = state.early_delay_line.TapOut(state.early_to_late_taps);
+
+        for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_TAPS; i++) {
+            const auto tapped_samp =
+                state.early_delay_line.TapOut(state.early_tap_steps[i]) * EARLY_GAIN[i];
+            out_samples[tap_index_lut[i]] += tapped_samp;
+
+            if constexpr (CHANNEL_COUNT == 6) {
+                // handle lfe
+                out_samples[5] += tapped_samp;
+            }
+        }
+
+        state.lowpass_0 = current_sample * state.lowpass_2 + state.lowpass_0 * state.lowpass_1;
+        state.early_delay_line.Tick(state.lowpass_0);
+
+        for (std::size_t i = 0; i < CHANNEL_COUNT; i++) {
+            out_samples[i] *= state.early_gain;
+        }
+
+        // Two channel seems to apply a latet gain, we require to save this
+        f32 filter{};
+        for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT; i++) {
+            filter = state.fdn_delay_line[i].GetOutputSample();
+            const auto computed = filter * state.lpf_coefficients[0][i] + state.shelf_filter[i];
+            state.shelf_filter[i] =
+                filter * state.lpf_coefficients[1][i] + computed * state.lpf_coefficients[2][i];
+            fsamp[i] = computed;
+        }
+
+        // Mixing matrix
+        mixed[0] = fsamp[1] + fsamp[2];
+        mixed[1] = -fsamp[0] - fsamp[3];
+        mixed[2] = fsamp[0] - fsamp[3];
+        mixed[3] = fsamp[1] - fsamp[2];
+
+        if constexpr (CHANNEL_COUNT == 2) {
+            for (auto& mix : mixed) {
+                mix *= (filter * state.late_gain);
+            }
+        }
+
+        for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT; i++) {
+            const auto late = early_tap * state.late_gain;
+            osamp[i] = state.decay_delay_line0[i].Tick(late + mixed[i]);
+            osamp[i] = state.decay_delay_line1[i].Tick(osamp[i]);
+            state.fdn_delay_line[i].Tick(osamp[i]);
+        }
+
+        if constexpr (CHANNEL_COUNT == 1) {
+            output[0][sample] = ToS32(state.dry_gain * ToFloat(input[0][sample]) +
+                                      (out_samples[0] + osamp[0] + osamp[1]));
+        } else if constexpr (CHANNEL_COUNT == 2 || CHANNEL_COUNT == 4) {
+            for (std::size_t i = 0; i < CHANNEL_COUNT; i++) {
+                output[i][sample] =
+                    ToS32(state.dry_gain * ToFloat(input[i][sample]) + (out_samples[i] + osamp[i]));
+            }
+        } else if constexpr (CHANNEL_COUNT == 6) {
+            const auto temp_center = state.center_delay_line.Tick(0.5f * (osamp[2] - osamp[3]));
+            for (std::size_t i = 0; i < 4; i++) {
+                output[i][sample] =
+                    ToS32(state.dry_gain * ToFloat(input[i][sample]) + (out_samples[i] + osamp[i]));
+            }
+            output[4][sample] =
+                ToS32(state.dry_gain * ToFloat(input[4][sample]) + (out_samples[4] + temp_center));
+            output[5][sample] =
+                ToS32(state.dry_gain * ToFloat(input[5][sample]) + (out_samples[5] + osamp[3]));
+        }
+    }
+}
+
 } // namespace
 
 CommandGenerator::CommandGenerator(AudioCommon::AudioRendererParameter& worker_params_,
@@ -271,11 +435,10 @@ void CommandGenerator::GenerateBiquadFilterCommandForVoice(ServerVoiceInfo& voic
         }
 
         // Generate biquad filter
-        //        GenerateBiquadFilterCommand(mix_buffer_count, biquad_filter,
-        //        dsp_state.biquad_filter_state,
-        //                                    mix_buffer_count + channel, mix_buffer_count +
-        //                                    channel, worker_params.sample_count,
-        //                                    voice_info.GetInParams().node_id);
+        // GenerateBiquadFilterCommand(mix_buffer_count, biquad_filter,
+        // dsp_state.biquad_filter_state,
+        //                            mix_buffer_count + channel, mix_buffer_count + channel,
+        //                            worker_params.sample_count, voice_info.GetInParams().node_id);
     }
 }
 
@@ -376,21 +539,54 @@ void CommandGenerator::GenerateEffectCommand(ServerMixInfo& mix_info) {
 
 void CommandGenerator::GenerateI3dl2ReverbEffectCommand(s32 mix_buffer_offset, EffectBase* info,
                                                         bool enabled) {
-    if (!enabled) {
+    auto* reverb = dynamic_cast<EffectI3dl2Reverb*>(info);
+    const auto& params = reverb->GetParams();
+    auto& state = reverb->GetState();
+    const auto channel_count = params.channel_count;
+
+    if (channel_count != 1 && channel_count != 2 && channel_count != 4 && channel_count != 6) {
         return;
     }
-    const auto& params = dynamic_cast<EffectI3dl2Reverb*>(info)->GetParams();
-    const auto channel_count = params.channel_count;
+
+    std::array<const s32*, AudioCommon::MAX_CHANNEL_COUNT> input{};
+    std::array<s32*, AudioCommon::MAX_CHANNEL_COUNT> output{};
+
+    const auto status = params.status;
     for (s32 i = 0; i < channel_count; i++) {
-        // TODO(ogniK): Actually implement reverb
-        /*
-        if (params.input[i] != params.output[i]) {
-            const auto* input = GetMixBuffer(mix_buffer_offset + params.input[i]);
-            auto* output = GetMixBuffer(mix_buffer_offset + params.output[i]);
-            ApplyMix<1>(output, input, 32768, worker_params.sample_count);
-        }*/
-        auto* output = GetMixBuffer(mix_buffer_offset + params.output[i]);
-        std::memset(output, 0, worker_params.sample_count * sizeof(s32));
+        input[i] = GetMixBuffer(mix_buffer_offset + params.input[i]);
+        output[i] = GetMixBuffer(mix_buffer_offset + params.output[i]);
+    }
+
+    if (enabled) {
+        if (status == ParameterStatus::Initialized) {
+            InitializeI3dl2Reverb(reverb->GetParams(), state, info->GetWorkBuffer());
+        } else if (status == ParameterStatus::Updating) {
+            UpdateI3dl2Reverb(reverb->GetParams(), state, false);
+        }
+    }
+
+    if (enabled) {
+        switch (channel_count) {
+        case 1:
+            ApplyReverbGeneric<1>(state, input, output, worker_params.sample_count);
+            break;
+        case 2:
+            ApplyReverbGeneric<2>(state, input, output, worker_params.sample_count);
+            break;
+        case 4:
+            ApplyReverbGeneric<4>(state, input, output, worker_params.sample_count);
+            break;
+        case 6:
+            ApplyReverbGeneric<6>(state, input, output, worker_params.sample_count);
+            break;
+        }
+    } else {
+        for (s32 i = 0; i < channel_count; i++) {
+            // Only copy if the buffer input and output do not match!
+            if ((mix_buffer_offset + params.input[i]) != (mix_buffer_offset + params.output[i])) {
+                std::memcpy(output[i], input[i], worker_params.sample_count * sizeof(s32));
+            }
+        }
     }
 }
 
@@ -528,6 +724,133 @@ s32 CommandGenerator::ReadAuxBuffer(AuxInfoDSP& recv_info, VAddr recv_buffer, u3
     return sample_count;
 }
 
+void CommandGenerator::InitializeI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbState& state,
+                                             std::vector<u8>& work_buffer) {
+    // Reset state
+    state.lowpass_0 = 0.0f;
+    state.lowpass_1 = 0.0f;
+    state.lowpass_2 = 0.0f;
+
+    state.early_delay_line.Reset();
+    state.early_tap_steps.fill(0);
+    state.early_gain = 0.0f;
+    state.late_gain = 0.0f;
+    state.early_to_late_taps = 0;
+    for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT; i++) {
+        state.fdn_delay_line[i].Reset();
+        state.decay_delay_line0[i].Reset();
+        state.decay_delay_line1[i].Reset();
+    }
+    state.last_reverb_echo = 0.0f;
+    state.center_delay_line.Reset();
+    for (auto& coef : state.lpf_coefficients) {
+        coef.fill(0.0f);
+    }
+    state.shelf_filter.fill(0.0f);
+    state.dry_gain = 0.0f;
+
+    const auto sample_rate = info.sample_rate / 1000;
+    f32* work_buffer_ptr = reinterpret_cast<f32*>(work_buffer.data());
+
+    s32 delay_samples{};
+    for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT; i++) {
+        delay_samples =
+            AudioCommon::CalculateDelaySamples(sample_rate, FDN_MAX_DELAY_LINE_TIMES[i]);
+        state.fdn_delay_line[i].Initialize(delay_samples, work_buffer_ptr);
+        work_buffer_ptr += delay_samples + 1;
+
+        delay_samples =
+            AudioCommon::CalculateDelaySamples(sample_rate, DECAY0_MAX_DELAY_LINE_TIMES[i]);
+        state.decay_delay_line0[i].Initialize(delay_samples, 0.0f, work_buffer_ptr);
+        work_buffer_ptr += delay_samples + 1;
+
+        delay_samples =
+            AudioCommon::CalculateDelaySamples(sample_rate, DECAY1_MAX_DELAY_LINE_TIMES[i]);
+        state.decay_delay_line1[i].Initialize(delay_samples, 0.0f, work_buffer_ptr);
+        work_buffer_ptr += delay_samples + 1;
+    }
+    delay_samples = AudioCommon::CalculateDelaySamples(sample_rate, 5.0f);
+    state.center_delay_line.Initialize(delay_samples, work_buffer_ptr);
+    work_buffer_ptr += delay_samples + 1;
+
+    delay_samples = AudioCommon::CalculateDelaySamples(sample_rate, 400.0f);
+    state.early_delay_line.Initialize(delay_samples, work_buffer_ptr);
+
+    UpdateI3dl2Reverb(info, state, true);
+}
+
+void CommandGenerator::UpdateI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbState& state,
+                                         bool should_clear) {
+
+    state.dry_gain = info.dry_gain;
+    state.shelf_filter.fill(0.0f);
+    state.lowpass_0 = 0.0f;
+    state.early_gain = Pow10(std::min(info.room + info.reflection, 5000.0f) / 2000.0f);
+    state.late_gain = Pow10(std::min(info.room + info.reverb, 5000.0f) / 2000.0f);
+
+    const auto sample_rate = info.sample_rate / 1000;
+    const f32 hf_gain = Pow10(info.room_hf / 2000.0f);
+    if (hf_gain >= 1.0f) {
+        state.lowpass_2 = 1.0f;
+        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 /
+                                                     static_cast<f32>(info.sample_rate)));
+        const auto c = std::sqrt(b * b - 4.0f * a * a);
+
+        state.lowpass_1 = (b - c) / (2.0f * a);
+        state.lowpass_2 = 1.0f - state.lowpass_1;
+    }
+    state.early_to_late_taps = AudioCommon::CalculateDelaySamples(
+        sample_rate, 1000.0f * (info.reflection_delay + info.reverb_delay));
+
+    state.last_reverb_echo = 0.6f * info.diffusion * 0.01f;
+    for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT; i++) {
+        const auto length =
+            FDN_MIN_DELAY_LINE_TIMES[i] +
+            (info.density / 100.0f) * (FDN_MAX_DELAY_LINE_TIMES[i] - FDN_MIN_DELAY_LINE_TIMES[i]);
+        state.fdn_delay_line[i].SetDelay(AudioCommon::CalculateDelaySamples(sample_rate, length));
+
+        const auto delay_sample_counts = state.fdn_delay_line[i].GetDelay() +
+                                         state.decay_delay_line0[i].GetDelay() +
+                                         state.decay_delay_line1[i].GetDelay();
+
+        float a = (-60.0f * static_cast<f32>(delay_sample_counts)) /
+                  (info.decay_time * static_cast<f32>(info.sample_rate));
+        float b = a / info.hf_decay_ratio;
+        float c = CosD(128.0f * 0.5f * info.hf_reference / static_cast<f32>(info.sample_rate)) /
+                  SinD(128.0f * 0.5f * info.hf_reference / static_cast<f32>(info.sample_rate));
+        float d = Pow10((b - a) / 40.0f);
+        float e = Pow10((b + a) / 40.0f) * 0.7071f;
+
+        state.lpf_coefficients[0][i] = e * ((d * c) + 1.0f) / (c + d);
+        state.lpf_coefficients[1][i] = e * (1.0f - (d * c)) / (c + d);
+        state.lpf_coefficients[2][i] = (c - d) / (c + d);
+
+        state.decay_delay_line0[i].SetCoefficient(state.last_reverb_echo);
+        state.decay_delay_line1[i].SetCoefficient(-0.9f * state.last_reverb_echo);
+    }
+
+    if (should_clear) {
+        for (std::size_t i = 0; i < AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT; i++) {
+            state.fdn_delay_line[i].Clear();
+            state.decay_delay_line0[i].Clear();
+            state.decay_delay_line1[i].Clear();
+        }
+        state.early_delay_line.Clear();
+        state.center_delay_line.Clear();
+    }
+
+    const auto max_early_delay = state.early_delay_line.GetMaxDelay();
+    const auto reflection_time = 1000.0f * (0.0098f * 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]);
+        state.early_tap_steps[tap] = std::min(length, max_early_delay);
+    }
+}
+
 void CommandGenerator::GenerateVolumeRampCommand(float last_volume, float current_volume,
                                                  s32 channel, s32 node_id) {
     const auto last = static_cast<s32>(last_volume * 32768.0f);

src/audio_core/command_generator.h

@@ -21,6 +21,8 @@ class ServerMixInfo;
 class EffectContext;
 class EffectBase;
 struct AuxInfoDSP;
+struct I3dl2ReverbParams;
+struct I3dl2ReverbState;
 using MixVolumeBuffer = std::array<float, AudioCommon::MAX_MIX_BUFFERS>;
 
 class CommandGenerator {
@@ -80,6 +82,9 @@ private:
     s32 ReadAuxBuffer(AuxInfoDSP& recv_info, VAddr recv_buffer, u32 max_samples, s32* out_data,
                       u32 sample_count, u32 read_offset, u32 read_count);
 
+    void InitializeI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbState& state,
+                               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);

src/audio_core/common.h

@@ -33,6 +33,29 @@ constexpr std::size_t TEMP_MIX_BASE_SIZE = 0x3f00; // TODO(ogniK): Work out this
 // and our const ends up being 0x3f04, the 4 bytes are most
 // likely the sample history
 constexpr std::size_t TOTAL_TEMP_MIX_SIZE = TEMP_MIX_BASE_SIZE + AudioCommon::MAX_SAMPLE_HISTORY;
+constexpr f32 I3DL2REVERB_MAX_LEVEL = 5000.0f;
+constexpr f32 I3DL2REVERB_MIN_REFLECTION_DURATION = 0.02f;
+constexpr std::size_t I3DL2REVERB_TAPS = 20;
+constexpr std::size_t I3DL2REVERB_DELAY_LINE_COUNT = 4;
+using Fractional = s32;
+
+template <typename T>
+constexpr Fractional ToFractional(T x) {
+    return static_cast<Fractional>(x * static_cast<T>(0x4000));
+}
+
+constexpr Fractional MultiplyFractional(Fractional lhs, Fractional rhs) {
+    return static_cast<Fractional>(static_cast<s64>(lhs) * rhs >> 14);
+}
+
+constexpr s32 FractionalToFixed(Fractional x) {
+    const auto s = x & (1 << 13);
+    return static_cast<s32>(x >> 14) + s;
+}
+
+constexpr s32 CalculateDelaySamples(s32 sample_rate_khz, float time) {
+    return FractionalToFixed(MultiplyFractional(ToFractional(sample_rate_khz), ToFractional(time)));
+}
 
 static constexpr u32 VersionFromRevision(u32_le rev) {
     // "REV7" -> 7

src/audio_core/delay_line.cpp

@@ -0,0 +1,104 @@
+#include <cstring>
+#include "audio_core/delay_line.h"
+
+namespace AudioCore {
+DelayLineBase::DelayLineBase() = default;
+DelayLineBase::~DelayLineBase() = default;
+
+void DelayLineBase::Initialize(s32 max_delay_, float* src_buffer) {
+    buffer = src_buffer;
+    buffer_end = buffer + max_delay_;
+    max_delay = max_delay_;
+    output = buffer;
+    SetDelay(max_delay_);
+    Clear();
+}
+
+void DelayLineBase::SetDelay(s32 new_delay) {
+    if (max_delay < new_delay) {
+        return;
+    }
+    delay = new_delay;
+    input = (buffer + ((output - buffer) + new_delay) % (max_delay + 1));
+}
+
+s32 DelayLineBase::GetDelay() const {
+    return delay;
+}
+
+s32 DelayLineBase::GetMaxDelay() const {
+    return max_delay;
+}
+
+f32 DelayLineBase::TapOut(s32 last_sample) {
+    const float* ptr = input - (last_sample + 1);
+    if (ptr < buffer) {
+        ptr += (max_delay + 1);
+    }
+
+    return *ptr;
+}
+
+f32 DelayLineBase::Tick(f32 sample) {
+    *(input++) = sample;
+    const auto out_sample = *(output++);
+
+    if (buffer_end < input) {
+        input = buffer;
+    }
+
+    if (buffer_end < output) {
+        output = buffer;
+    }
+
+    return out_sample;
+}
+
+float* DelayLineBase::GetInput() {
+    return input;
+}
+
+const float* DelayLineBase::GetInput() const {
+    return input;
+}
+
+f32 DelayLineBase::GetOutputSample() const {
+    return *output;
+}
+
+void DelayLineBase::Clear() {
+    std::memset(buffer, 0, sizeof(float) * max_delay);
+}
+
+void DelayLineBase::Reset() {
+    buffer = nullptr;
+    buffer_end = nullptr;
+    max_delay = 0;
+    input = nullptr;
+    output = nullptr;
+    delay = 0;
+}
+
+DelayLineAllPass::DelayLineAllPass() = default;
+DelayLineAllPass::~DelayLineAllPass() = default;
+
+void DelayLineAllPass::Initialize(u32 delay_, float coeffcient_, f32* src_buffer) {
+    DelayLineBase::Initialize(delay_, src_buffer);
+    SetCoefficient(coeffcient_);
+}
+
+void DelayLineAllPass::SetCoefficient(float coeffcient_) {
+    coefficient = coeffcient_;
+}
+
+f32 DelayLineAllPass::Tick(f32 sample) {
+    const auto temp = sample - coefficient * *output;
+    return coefficient * temp + DelayLineBase::Tick(temp);
+}
+
+void DelayLineAllPass::Reset() {
+    coefficient = 0.0f;
+    DelayLineBase::Reset();
+}
+
+} // namespace AudioCore

src/audio_core/delay_line.h

@@ -0,0 +1,46 @@
+#pragma once
+
+#include "common/common_types.h"
+
+namespace AudioCore {
+
+class DelayLineBase {
+public:
+    DelayLineBase();
+    ~DelayLineBase();
+
+    void Initialize(s32 max_delay_, float* src_buffer);
+    void SetDelay(s32 new_delay);
+    s32 GetDelay() const;
+    s32 GetMaxDelay() const;
+    f32 TapOut(s32 last_sample);
+    f32 Tick(f32 sample);
+    float* GetInput();
+    const float* GetInput() const;
+    f32 GetOutputSample() const;
+    void Clear();
+    void Reset();
+
+protected:
+    float* buffer{nullptr};
+    float* buffer_end{nullptr};
+    s32 max_delay{};
+    float* input{nullptr};
+    float* output{nullptr};
+    s32 delay{};
+};
+
+class DelayLineAllPass final : public DelayLineBase {
+public:
+    DelayLineAllPass();
+    ~DelayLineAllPass();
+
+    void Initialize(u32 delay, float coeffcient_, f32* src_buffer);
+    void SetCoefficient(float coeffcient_);
+    f32 Tick(f32 sample);
+    void Reset();
+
+private:
+    float coefficient{};
+};
+} // namespace AudioCore

src/audio_core/effect_context.cpp

@@ -90,6 +90,14 @@ s32 EffectBase::GetProcessingOrder() const {
     return processing_order;
 }
 
+std::vector<u8>& EffectBase::GetWorkBuffer() {
+    return work_buffer;
+}
+
+const std::vector<u8>& EffectBase::GetWorkBuffer() const {
+    return work_buffer;
+}
+
 EffectI3dl2Reverb::EffectI3dl2Reverb() : EffectGeneric(EffectType::I3dl2Reverb) {}
 EffectI3dl2Reverb::~EffectI3dl2Reverb() = default;
 
@@ -117,6 +125,12 @@ void EffectI3dl2Reverb::Update(EffectInfo::InParams& in_params) {
         usage = UsageState::Initialized;
         params.status = ParameterStatus::Initialized;
         skipped = in_params.buffer_address == 0 || in_params.buffer_size == 0;
+        if (!skipped) {
+            auto& cur_work_buffer = GetWorkBuffer();
+            // Has two buffers internally
+            cur_work_buffer.resize(in_params.buffer_size * 2);
+            std::fill(cur_work_buffer.begin(), cur_work_buffer.end(), 0);
+        }
     }
 }
 
@@ -129,6 +143,14 @@ void EffectI3dl2Reverb::UpdateForCommandGeneration() {
     GetParams().status = ParameterStatus::Updated;
 }
 
+I3dl2ReverbState& EffectI3dl2Reverb::GetState() {
+    return state;
+}
+
+const I3dl2ReverbState& EffectI3dl2Reverb::GetState() const {
+    return state;
+}
+
 EffectBiquadFilter::EffectBiquadFilter() : EffectGeneric(EffectType::BiquadFilter) {}
 EffectBiquadFilter::~EffectBiquadFilter() = default;
 

src/audio_core/effect_context.h

@@ -8,6 +8,7 @@
 #include <memory>
 #include <vector>
 #include "audio_core/common.h"
+#include "audio_core/delay_line.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
@@ -194,6 +195,8 @@ public:
     [[nodiscard]] bool IsEnabled() const;
     [[nodiscard]] s32 GetMixID() const;
     [[nodiscard]] s32 GetProcessingOrder() const;
+    [[nodiscard]] std::vector<u8>& GetWorkBuffer();
+    [[nodiscard]] const std::vector<u8>& GetWorkBuffer() const;
 
 protected:
     UsageState usage{UsageState::Invalid};
@@ -201,6 +204,7 @@ protected:
     s32 mix_id{};
     s32 processing_order{};
     bool enabled = false;
+    std::vector<u8> work_buffer{};
 };
 
 template <typename T>
@@ -212,7 +216,7 @@ public:
         return internal_params;
     }
 
-    const I3dl2ReverbParams& GetParams() const {
+    const T& GetParams() const {
         return internal_params;
     }
 
@@ -229,6 +233,27 @@ public:
     void UpdateForCommandGeneration() override;
 };
 
+struct I3dl2ReverbState {
+    f32 lowpass_0{};
+    f32 lowpass_1{};
+    f32 lowpass_2{};
+
+    DelayLineBase early_delay_line{};
+    std::array<u32, AudioCommon::I3DL2REVERB_TAPS> early_tap_steps{};
+    f32 early_gain{};
+    f32 late_gain{};
+
+    u32 early_to_late_taps{};
+    std::array<DelayLineBase, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> fdn_delay_line{};
+    std::array<DelayLineAllPass, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> decay_delay_line0{};
+    std::array<DelayLineAllPass, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> decay_delay_line1{};
+    f32 last_reverb_echo{};
+    DelayLineBase center_delay_line{};
+    std::array<std::array<f32, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT>, 3> lpf_coefficients{};
+    std::array<f32, AudioCommon::I3DL2REVERB_DELAY_LINE_COUNT> shelf_filter{};
+    f32 dry_gain{};
+};
+
 class EffectI3dl2Reverb : public EffectGeneric<I3dl2ReverbParams> {
 public:
     explicit EffectI3dl2Reverb();
@@ -237,8 +262,12 @@ public:
     void Update(EffectInfo::InParams& in_params) override;
     void UpdateForCommandGeneration() override;
 
+    I3dl2ReverbState& GetState();
+    const I3dl2ReverbState& GetState() const;
+
 private:
     bool skipped = false;
+    I3dl2ReverbState state{};
 };
 
 class EffectBiquadFilter : public EffectGeneric<BiquadFilterParams> {

src/common/CMakeLists.txt

@@ -168,7 +168,6 @@ add_library(common STATIC
     time_zone.cpp
     time_zone.h
     tree.h
-    uint128.cpp
     uint128.h
     uuid.cpp
     uuid.h

src/common/uint128.cpp

@@ -1,71 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#ifdef _MSC_VER
-#include <intrin.h>
-
-#pragma intrinsic(_umul128)
-#pragma intrinsic(_udiv128)
-#endif
-#include <cstring>
-#include "common/uint128.h"
-
-namespace Common {
-
-#ifdef _MSC_VER
-
-u64 MultiplyAndDivide64(u64 a, u64 b, u64 d) {
-    u128 r{};
-    r[0] = _umul128(a, b, &r[1]);
-    u64 remainder;
-#if _MSC_VER < 1923
-    return udiv128(r[1], r[0], d, &remainder);
-#else
-    return _udiv128(r[1], r[0], d, &remainder);
-#endif
-}
-
-#else
-
-u64 MultiplyAndDivide64(u64 a, u64 b, u64 d) {
-    const u64 diva = a / d;
-    const u64 moda = a % d;
-    const u64 divb = b / d;
-    const u64 modb = b % d;
-    return diva * b + moda * divb + moda * modb / d;
-}
-
-#endif
-
-u128 Multiply64Into128(u64 a, u64 b) {
-    u128 result;
-#ifdef _MSC_VER
-    result[0] = _umul128(a, b, &result[1]);
-#else
-    unsigned __int128 tmp = a;
-    tmp *= b;
-    std::memcpy(&result, &tmp, sizeof(u128));
-#endif
-    return result;
-}
-
-std::pair<u64, u64> Divide128On32(u128 dividend, u32 divisor) {
-    u64 remainder = dividend[0] % divisor;
-    u64 accum = dividend[0] / divisor;
-    if (dividend[1] == 0)
-        return {accum, remainder};
-    // We ignore dividend[1] / divisor as that overflows
-    const u64 first_segment = (dividend[1] % divisor) << 32;
-    accum += (first_segment / divisor) << 32;
-    const u64 second_segment = (first_segment % divisor) << 32;
-    accum += (second_segment / divisor);
-    remainder += second_segment % divisor;
-    if (remainder >= divisor) {
-        accum++;
-        remainder -= divisor;
-    }
-    return {accum, remainder};
-}
-
-} // namespace Common

src/common/uint128.h

@@ -4,19 +4,98 @@
 
 #pragma once
 
+#include <cstring>
 #include <utility>
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#pragma intrinsic(__umulh)
+#pragma intrinsic(_umul128)
+#pragma intrinsic(_udiv128)
+#else
+#include <x86intrin.h>
+#endif
+
 #include "common/common_types.h"
 
 namespace Common {
 
 // This function multiplies 2 u64 values and divides it by a u64 value.
-[[nodiscard]] u64 MultiplyAndDivide64(u64 a, u64 b, u64 d);
+[[nodiscard]] static inline u64 MultiplyAndDivide64(u64 a, u64 b, u64 d) {
+#ifdef _MSC_VER
+    u128 r{};
+    r[0] = _umul128(a, b, &r[1]);
+    u64 remainder;
+#if _MSC_VER < 1923
+    return udiv128(r[1], r[0], d, &remainder);
+#else
+    return _udiv128(r[1], r[0], d, &remainder);
+#endif
+#else
+    const u64 diva = a / d;
+    const u64 moda = a % d;
+    const u64 divb = b / d;
+    const u64 modb = b % d;
+    return diva * b + moda * divb + moda * modb / d;
+#endif
+}
 
 // This function multiplies 2 u64 values and produces a u128 value;
-[[nodiscard]] u128 Multiply64Into128(u64 a, u64 b);
+[[nodiscard]] static inline u128 Multiply64Into128(u64 a, u64 b) {
+    u128 result;
+#ifdef _MSC_VER
+    result[0] = _umul128(a, b, &result[1]);
+#else
+    unsigned __int128 tmp = a;
+    tmp *= b;
+    std::memcpy(&result, &tmp, sizeof(u128));
+#endif
+    return result;
+}
 
-// This function divides a u128 by a u32 value and produces two u64 values:
-// the result of division and the remainder
-[[nodiscard]] std::pair<u64, u64> Divide128On32(u128 dividend, u32 divisor);
+[[nodiscard]] static inline u64 GetFixedPoint64Factor(u64 numerator, u64 divisor) {
+#ifdef __SIZEOF_INT128__
+    const auto base = static_cast<unsigned __int128>(numerator) << 64ULL;
+    return static_cast<u64>(base / divisor);
+#elif defined(_M_X64) || defined(_M_ARM64)
+    std::array<u64, 2> r = {0, numerator};
+    u64 remainder;
+#if _MSC_VER < 1923
+    return udiv128(r[1], r[0], divisor, &remainder);
+#else
+    return _udiv128(r[1], r[0], divisor, &remainder);
+#endif
+#else
+    // This one is bit more inaccurate.
+    return MultiplyAndDivide64(std::numeric_limits<u64>::max(), numerator, divisor);
+#endif
+}
+
+[[nodiscard]] static inline u64 MultiplyHigh(u64 a, u64 b) {
+#ifdef __SIZEOF_INT128__
+    return (static_cast<unsigned __int128>(a) * static_cast<unsigned __int128>(b)) >> 64;
+#elif defined(_M_X64) || defined(_M_ARM64)
+    return __umulh(a, b); // MSVC
+#else
+    // Generic fallback
+    const u64 a_lo = u32(a);
+    const u64 a_hi = a >> 32;
+    const u64 b_lo = u32(b);
+    const u64 b_hi = b >> 32;
+
+    const u64 a_x_b_hi = a_hi * b_hi;
+    const u64 a_x_b_mid = a_hi * b_lo;
+    const u64 b_x_a_mid = b_hi * a_lo;
+    const u64 a_x_b_lo = a_lo * b_lo;
+
+    const u64 carry_bit = (static_cast<u64>(static_cast<u32>(a_x_b_mid)) +
+                           static_cast<u64>(static_cast<u32>(b_x_a_mid)) + (a_x_b_lo >> 32)) >>
+                          32;
+
+    const u64 multhi = a_x_b_hi + (a_x_b_mid >> 32) + (b_x_a_mid >> 32) + carry_bit;
+
+    return multhi;
+#endif
+}
 
 } // namespace Common

src/common/wall_clock.cpp

@@ -2,6 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <cstdint>
+
 #include "common/uint128.h"
 #include "common/wall_clock.h"
 
@@ -18,7 +20,9 @@ using base_time_point = std::chrono::time_point<base_timer>;
 class StandardWallClock final : public WallClock {
 public:
     explicit StandardWallClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequency_)
-        : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, false) {
+        : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, false),
+          emulated_clock_factor{GetFixedPoint64Factor(emulated_clock_frequency, 1000000000)},
+          emulated_cpu_factor{GetFixedPoint64Factor(emulated_cpu_frequency, 1000000000)} {
         start_time = base_timer::now();
     }
 
@@ -41,16 +45,11 @@ public:
     }
 
     u64 GetClockCycles() override {
-        std::chrono::nanoseconds time_now = GetTimeNS();
-        const u128 temporary =
-            Common::Multiply64Into128(time_now.count(), emulated_clock_frequency);
-        return Common::Divide128On32(temporary, 1000000000).first;
+        return MultiplyHigh(GetTimeNS().count(), emulated_clock_factor);
     }
 
     u64 GetCPUCycles() override {
-        std::chrono::nanoseconds time_now = GetTimeNS();
-        const u128 temporary = Common::Multiply64Into128(time_now.count(), emulated_cpu_frequency);
-        return Common::Divide128On32(temporary, 1000000000).first;
+        return MultiplyHigh(GetTimeNS().count(), emulated_cpu_factor);
     }
 
     void Pause([[maybe_unused]] bool is_paused) override {
@@ -59,6 +58,8 @@ public:
 
 private:
     base_time_point start_time;
+    const u64 emulated_clock_factor;
+    const u64 emulated_cpu_factor;
 };
 
 #ifdef ARCHITECTURE_x86_64

src/common/x64/native_clock.cpp

@@ -8,68 +8,10 @@
 #include <mutex>
 #include <thread>
 
-#ifdef _MSC_VER
-#include <intrin.h>
-
-#pragma intrinsic(__umulh)
-#pragma intrinsic(_udiv128)
-#else
-#include <x86intrin.h>
-#endif
-
 #include "common/atomic_ops.h"
 #include "common/uint128.h"
 #include "common/x64/native_clock.h"
 
-namespace {
-
-[[nodiscard]] u64 GetFixedPoint64Factor(u64 numerator, u64 divisor) {
-#ifdef __SIZEOF_INT128__
-    const auto base = static_cast<unsigned __int128>(numerator) << 64ULL;
-    return static_cast<u64>(base / divisor);
-#elif defined(_M_X64) || defined(_M_ARM64)
-    std::array<u64, 2> r = {0, numerator};
-    u64 remainder;
-#if _MSC_VER < 1923
-    return udiv128(r[1], r[0], divisor, &remainder);
-#else
-    return _udiv128(r[1], r[0], divisor, &remainder);
-#endif
-#else
-    // This one is bit more inaccurate.
-    return MultiplyAndDivide64(std::numeric_limits<u64>::max(), numerator, divisor);
-#endif
-}
-
-[[nodiscard]] u64 MultiplyHigh(u64 a, u64 b) {
-#ifdef __SIZEOF_INT128__
-    return (static_cast<unsigned __int128>(a) * static_cast<unsigned __int128>(b)) >> 64;
-#elif defined(_M_X64) || defined(_M_ARM64)
-    return __umulh(a, b); // MSVC
-#else
-    // Generic fallback
-    const u64 a_lo = u32(a);
-    const u64 a_hi = a >> 32;
-    const u64 b_lo = u32(b);
-    const u64 b_hi = b >> 32;
-
-    const u64 a_x_b_hi = a_hi * b_hi;
-    const u64 a_x_b_mid = a_hi * b_lo;
-    const u64 b_x_a_mid = b_hi * a_lo;
-    const u64 a_x_b_lo = a_lo * b_lo;
-
-    const u64 carry_bit = (static_cast<u64>(static_cast<u32>(a_x_b_mid)) +
-                           static_cast<u64>(static_cast<u32>(b_x_a_mid)) + (a_x_b_lo >> 32)) >>
-                          32;
-
-    const u64 multhi = a_x_b_hi + (a_x_b_mid >> 32) + (b_x_a_mid >> 32) + carry_bit;
-
-    return multhi;
-#endif
-}
-
-} // namespace
-
 namespace Common {
 
 u64 EstimateRDTSCFrequency() {

src/core/CMakeLists.txt

@@ -19,7 +19,6 @@ add_library(core STATIC
     core.h
     core_timing.cpp
     core_timing.h
-    core_timing_util.cpp
     core_timing_util.h
     cpu_manager.cpp
     cpu_manager.h
@@ -148,7 +147,7 @@ add_library(core STATIC
     hle/kernel/client_session.h
     hle/kernel/code_set.cpp
     hle/kernel/code_set.h
-    hle/kernel/errors.h
+    hle/kernel/svc_results.h
     hle/kernel/global_scheduler_context.cpp
     hle/kernel/global_scheduler_context.h
     hle/kernel/handle_table.cpp
@@ -174,6 +173,7 @@ add_library(core STATIC
     hle/kernel/k_scheduler.h
     hle/kernel/k_scheduler_lock.h
     hle/kernel/k_scoped_lock.h
+    hle/kernel/k_scoped_resource_reservation.h
     hle/kernel/k_scoped_scheduler_lock_and_sleep.h
     hle/kernel/k_synchronization_object.cpp
     hle/kernel/k_synchronization_object.h
@@ -223,7 +223,6 @@ add_library(core STATIC
     hle/kernel/svc.cpp
     hle/kernel/svc.h
     hle/kernel/svc_common.h
-    hle/kernel/svc_results.h
     hle/kernel/svc_types.h
     hle/kernel/svc_wrap.h
     hle/kernel/time_manager.cpp
@@ -266,6 +265,7 @@ add_library(core STATIC
     hle/service/am/applets/software_keyboard.h
     hle/service/am/applets/web_browser.cpp
     hle/service/am/applets/web_browser.h
+    hle/service/am/applets/web_types.h
     hle/service/am/idle.cpp
     hle/service/am/idle.h
     hle/service/am/omm.cpp
@@ -400,6 +400,7 @@ add_library(core STATIC
     hle/service/hid/controllers/xpad.h
     hle/service/lbl/lbl.cpp
     hle/service/lbl/lbl.h
+    hle/service/ldn/errors.h
     hle/service/ldn/ldn.cpp
     hle/service/ldn/ldn.h
     hle/service/ldr/ldr.cpp

src/core/core_timing_util.cpp

@@ -1,84 +0,0 @@
-// Copyright 2008 Dolphin Emulator Project / 2017 Citra Emulator Project
-// Licensed under GPLv2+
-// Refer to the license.txt file included.
-
-#include "core/core_timing_util.h"
-
-#include <cinttypes>
-#include <limits>
-#include "common/logging/log.h"
-#include "common/uint128.h"
-#include "core/hardware_properties.h"
-
-namespace Core::Timing {
-
-constexpr u64 MAX_VALUE_TO_MULTIPLY = std::numeric_limits<s64>::max() / Hardware::BASE_CLOCK_RATE;
-
-s64 msToCycles(std::chrono::milliseconds ms) {
-    if (static_cast<u64>(ms.count() / 1000) > MAX_VALUE_TO_MULTIPLY) {
-        LOG_ERROR(Core_Timing, "Integer overflow, use max value");
-        return std::numeric_limits<s64>::max();
-    }
-    if (static_cast<u64>(ms.count()) > MAX_VALUE_TO_MULTIPLY) {
-        LOG_DEBUG(Core_Timing, "Time very big, do rounding");
-        return Hardware::BASE_CLOCK_RATE * (ms.count() / 1000);
-    }
-    return (Hardware::BASE_CLOCK_RATE * ms.count()) / 1000;
-}
-
-s64 usToCycles(std::chrono::microseconds us) {
-    if (static_cast<u64>(us.count() / 1000000) > MAX_VALUE_TO_MULTIPLY) {
-        LOG_ERROR(Core_Timing, "Integer overflow, use max value");
-        return std::numeric_limits<s64>::max();
-    }
-    if (static_cast<u64>(us.count()) > MAX_VALUE_TO_MULTIPLY) {
-        LOG_DEBUG(Core_Timing, "Time very big, do rounding");
-        return Hardware::BASE_CLOCK_RATE * (us.count() / 1000000);
-    }
-    return (Hardware::BASE_CLOCK_RATE * us.count()) / 1000000;
-}
-
-s64 nsToCycles(std::chrono::nanoseconds ns) {
-    const u128 temporal = Common::Multiply64Into128(ns.count(), Hardware::BASE_CLOCK_RATE);
-    return Common::Divide128On32(temporal, static_cast<u32>(1000000000)).first;
-}
-
-u64 msToClockCycles(std::chrono::milliseconds ns) {
-    const u128 temp = Common::Multiply64Into128(ns.count(), Hardware::CNTFREQ);
-    return Common::Divide128On32(temp, 1000).first;
-}
-
-u64 usToClockCycles(std::chrono::microseconds ns) {
-    const u128 temp = Common::Multiply64Into128(ns.count(), Hardware::CNTFREQ);
-    return Common::Divide128On32(temp, 1000000).first;
-}
-
-u64 nsToClockCycles(std::chrono::nanoseconds ns) {
-    const u128 temp = Common::Multiply64Into128(ns.count(), Hardware::CNTFREQ);
-    return Common::Divide128On32(temp, 1000000000).first;
-}
-
-u64 CpuCyclesToClockCycles(u64 ticks) {
-    const u128 temporal = Common::Multiply64Into128(ticks, Hardware::CNTFREQ);
-    return Common::Divide128On32(temporal, static_cast<u32>(Hardware::BASE_CLOCK_RATE)).first;
-}
-
-std::chrono::milliseconds CyclesToMs(s64 cycles) {
-    const u128 temporal = Common::Multiply64Into128(cycles, 1000);
-    u64 ms = Common::Divide128On32(temporal, static_cast<u32>(Hardware::BASE_CLOCK_RATE)).first;
-    return std::chrono::milliseconds(ms);
-}
-
-std::chrono::nanoseconds CyclesToNs(s64 cycles) {
-    const u128 temporal = Common::Multiply64Into128(cycles, 1000000000);
-    u64 ns = Common::Divide128On32(temporal, static_cast<u32>(Hardware::BASE_CLOCK_RATE)).first;
-    return std::chrono::nanoseconds(ns);
-}
-
-std::chrono::microseconds CyclesToUs(s64 cycles) {
-    const u128 temporal = Common::Multiply64Into128(cycles, 1000000);
-    u64 us = Common::Divide128On32(temporal, static_cast<u32>(Hardware::BASE_CLOCK_RATE)).first;
-    return std::chrono::microseconds(us);
-}
-
-} // namespace Core::Timing

src/core/core_timing_util.h

@@ -1,24 +1,59 @@
-// Copyright 2008 Dolphin Emulator Project / 2017 Citra Emulator Project
-// Licensed under GPLv2+
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
 #include <chrono>
+
 #include "common/common_types.h"
+#include "core/hardware_properties.h"
 
 namespace Core::Timing {
 
-s64 msToCycles(std::chrono::milliseconds ms);
-s64 usToCycles(std::chrono::microseconds us);
-s64 nsToCycles(std::chrono::nanoseconds ns);
-u64 msToClockCycles(std::chrono::milliseconds ns);
-u64 usToClockCycles(std::chrono::microseconds ns);
-u64 nsToClockCycles(std::chrono::nanoseconds ns);
-std::chrono::milliseconds CyclesToMs(s64 cycles);
-std::chrono::nanoseconds CyclesToNs(s64 cycles);
-std::chrono::microseconds CyclesToUs(s64 cycles);
+namespace detail {
+constexpr u64 CNTFREQ_ADJUSTED = Hardware::CNTFREQ / 1000;
+constexpr u64 BASE_CLOCK_RATE_ADJUSTED = Hardware::BASE_CLOCK_RATE / 1000;
+} // namespace detail
 
-u64 CpuCyclesToClockCycles(u64 ticks);
+[[nodiscard]] constexpr s64 msToCycles(std::chrono::milliseconds ms) {
+    return ms.count() * detail::BASE_CLOCK_RATE_ADJUSTED;
+}
+
+[[nodiscard]] constexpr s64 usToCycles(std::chrono::microseconds us) {
+    return us.count() * detail::BASE_CLOCK_RATE_ADJUSTED / 1000;
+}
+
+[[nodiscard]] constexpr s64 nsToCycles(std::chrono::nanoseconds ns) {
+    return ns.count() * detail::BASE_CLOCK_RATE_ADJUSTED / 1000000;
+}
+
+[[nodiscard]] constexpr u64 msToClockCycles(std::chrono::milliseconds ms) {
+    return static_cast<u64>(ms.count()) * detail::CNTFREQ_ADJUSTED;
+}
+
+[[nodiscard]] constexpr u64 usToClockCycles(std::chrono::microseconds us) {
+    return us.count() * detail::CNTFREQ_ADJUSTED / 1000;
+}
+
+[[nodiscard]] constexpr u64 nsToClockCycles(std::chrono::nanoseconds ns) {
+    return ns.count() * detail::CNTFREQ_ADJUSTED / 1000000;
+}
+
+[[nodiscard]] constexpr u64 CpuCyclesToClockCycles(u64 ticks) {
+    return ticks * detail::CNTFREQ_ADJUSTED / detail::BASE_CLOCK_RATE_ADJUSTED;
+}
+
+[[nodiscard]] constexpr std::chrono::milliseconds CyclesToMs(s64 cycles) {
+    return std::chrono::milliseconds(cycles / detail::BASE_CLOCK_RATE_ADJUSTED);
+}
+
+[[nodiscard]] constexpr std::chrono::nanoseconds CyclesToNs(s64 cycles) {
+    return std::chrono::nanoseconds(cycles * 1000000 / detail::BASE_CLOCK_RATE_ADJUSTED);
+}
+
+[[nodiscard]] constexpr std::chrono::microseconds CyclesToUs(s64 cycles) {
+    return std::chrono::microseconds(cycles * 1000 / detail::BASE_CLOCK_RATE_ADJUSTED);
+}
 
 } // namespace Core::Timing

src/core/frontend/applets/controller.h

@@ -31,6 +31,7 @@ struct ControllerParameters {
     bool allow_dual_joycons{};
     bool allow_left_joycon{};
     bool allow_right_joycon{};
+    bool allow_gamecube_controller{};
 };
 
 class ControllerApplet {

src/core/hle/kernel/client_port.cpp

@@ -4,11 +4,11 @@
 
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/server_port.h"
 #include "core/hle/kernel/session.h"
+#include "core/hle/kernel/svc_results.h"
 
 namespace Kernel {
 
@@ -21,7 +21,7 @@ std::shared_ptr<ServerPort> ClientPort::GetServerPort() const {
 
 ResultVal<std::shared_ptr<ClientSession>> ClientPort::Connect() {
     if (active_sessions >= max_sessions) {
-        return ERR_MAX_CONNECTIONS_REACHED;
+        return ResultMaxConnectionsReached;
     }
     active_sessions++;
 

src/core/hle/kernel/client_session.cpp

@@ -3,11 +3,11 @@
 // Refer to the license.txt file included.
 
 #include "core/hle/kernel/client_session.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/server_session.h"
 #include "core/hle/kernel/session.h"
+#include "core/hle/kernel/svc_results.h"
 #include "core/hle/result.h"
 
 namespace Kernel {
@@ -43,7 +43,7 @@ ResultCode ClientSession::SendSyncRequest(std::shared_ptr<KThread> thread,
                                           Core::Timing::CoreTiming& core_timing) {
     // Keep ServerSession alive until we're done working with it.
     if (!parent->Server()) {
-        return ERR_SESSION_CLOSED_BY_REMOTE;
+        return ResultSessionClosedByRemote;
     }
 
     // Signal the server session that new data is available

src/core/hle/kernel/errors.h

@@ -1,43 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "core/hle/result.h"
-
-namespace Kernel {
-
-// Confirmed Switch kernel error codes
-
-constexpr ResultCode ERR_MAX_CONNECTIONS_REACHED{ErrorModule::Kernel, 7};
-constexpr ResultCode ERR_INVALID_CAPABILITY_DESCRIPTOR{ErrorModule::Kernel, 14};
-constexpr ResultCode ERR_THREAD_TERMINATING{ErrorModule::Kernel, 59};
-constexpr ResultCode ERR_TERMINATION_REQUESTED{ErrorModule::Kernel, 59};
-constexpr ResultCode ERR_INVALID_SIZE{ErrorModule::Kernel, 101};
-constexpr ResultCode ERR_INVALID_ADDRESS{ErrorModule::Kernel, 102};
-constexpr ResultCode ERR_OUT_OF_RESOURCES{ErrorModule::Kernel, 103};
-constexpr ResultCode ERR_OUT_OF_MEMORY{ErrorModule::Kernel, 104};
-constexpr ResultCode ERR_HANDLE_TABLE_FULL{ErrorModule::Kernel, 105};
-constexpr ResultCode ERR_INVALID_ADDRESS_STATE{ErrorModule::Kernel, 106};
-constexpr ResultCode ERR_INVALID_CURRENT_MEMORY{ErrorModule::Kernel, 106};
-constexpr ResultCode ERR_INVALID_MEMORY_PERMISSIONS{ErrorModule::Kernel, 108};
-constexpr ResultCode ERR_INVALID_MEMORY_RANGE{ErrorModule::Kernel, 110};
-constexpr ResultCode ERR_INVALID_PROCESSOR_ID{ErrorModule::Kernel, 113};
-constexpr ResultCode ERR_INVALID_THREAD_PRIORITY{ErrorModule::Kernel, 112};
-constexpr ResultCode ERR_INVALID_HANDLE{ErrorModule::Kernel, 114};
-constexpr ResultCode ERR_INVALID_POINTER{ErrorModule::Kernel, 115};
-constexpr ResultCode ERR_INVALID_COMBINATION{ErrorModule::Kernel, 116};
-constexpr ResultCode RESULT_TIMEOUT{ErrorModule::Kernel, 117};
-constexpr ResultCode ERR_SYNCHRONIZATION_CANCELED{ErrorModule::Kernel, 118};
-constexpr ResultCode ERR_CANCELLED{ErrorModule::Kernel, 118};
-constexpr ResultCode ERR_OUT_OF_RANGE{ErrorModule::Kernel, 119};
-constexpr ResultCode ERR_INVALID_ENUM_VALUE{ErrorModule::Kernel, 120};
-constexpr ResultCode ERR_NOT_FOUND{ErrorModule::Kernel, 121};
-constexpr ResultCode ERR_BUSY{ErrorModule::Kernel, 122};
-constexpr ResultCode ERR_SESSION_CLOSED_BY_REMOTE{ErrorModule::Kernel, 123};
-constexpr ResultCode ERR_INVALID_STATE{ErrorModule::Kernel, 125};
-constexpr ResultCode ERR_RESERVED_VALUE{ErrorModule::Kernel, 126};
-constexpr ResultCode ERR_RESOURCE_LIMIT_EXCEEDED{ErrorModule::Kernel, 132};
-
-} // namespace Kernel

src/core/hle/kernel/handle_table.cpp

@@ -6,12 +6,12 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/svc_results.h"
 
 namespace Kernel {
 namespace {
@@ -33,7 +33,7 @@ HandleTable::~HandleTable() = default;
 ResultCode HandleTable::SetSize(s32 handle_table_size) {
     if (static_cast<u32>(handle_table_size) > MAX_COUNT) {
         LOG_ERROR(Kernel, "Handle table size {} is greater than {}", handle_table_size, MAX_COUNT);
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     // Values less than or equal to zero indicate to use the maximum allowable
@@ -53,7 +53,7 @@ ResultVal<Handle> HandleTable::Create(std::shared_ptr<Object> obj) {
     const u16 slot = next_free_slot;
     if (slot >= table_size) {
         LOG_ERROR(Kernel, "Unable to allocate Handle, too many slots in use.");
-        return ERR_HANDLE_TABLE_FULL;
+        return ResultHandleTableFull;
     }
     next_free_slot = generations[slot];
 
@@ -76,7 +76,7 @@ ResultVal<Handle> HandleTable::Duplicate(Handle handle) {
     std::shared_ptr<Object> object = GetGeneric(handle);
     if (object == nullptr) {
         LOG_ERROR(Kernel, "Tried to duplicate invalid handle: {:08X}", handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
     return Create(std::move(object));
 }
@@ -84,7 +84,7 @@ ResultVal<Handle> HandleTable::Duplicate(Handle handle) {
 ResultCode HandleTable::Close(Handle handle) {
     if (!IsValid(handle)) {
         LOG_ERROR(Kernel, "Handle is not valid! handle={:08X}", handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     const u16 slot = GetSlot(handle);

src/core/hle/kernel/hle_ipc.cpp

@@ -14,7 +14,6 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/k_readable_event.h"
@@ -26,6 +25,7 @@
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/svc_results.h"
 #include "core/hle/kernel/time_manager.h"
 #include "core/memory.h"
 

src/core/hle/kernel/k_address_arbiter.cpp

@@ -120,10 +120,10 @@ ResultCode KAddressArbiter::SignalAndIncrementIfEqual(VAddr addr, s32 value, s32
         s32 user_value{};
         if (!UpdateIfEqual(system, &user_value, addr, value, value + 1)) {
             LOG_ERROR(Kernel, "Invalid current memory!");
-            return Svc::ResultInvalidCurrentMemory;
+            return ResultInvalidCurrentMemory;
         }
         if (user_value != value) {
-            return Svc::ResultInvalidState;
+            return ResultInvalidState;
         }
 
         auto it = thread_tree.nfind_light({addr, -1});
@@ -189,10 +189,10 @@ ResultCode KAddressArbiter::SignalAndModifyByWaitingCountIfEqual(VAddr addr, s32
 
         if (!succeeded) {
             LOG_ERROR(Kernel, "Invalid current memory!");
-            return Svc::ResultInvalidCurrentMemory;
+            return ResultInvalidCurrentMemory;
         }
         if (user_value != value) {
-            return Svc::ResultInvalidState;
+            return ResultInvalidState;
         }
 
         while ((it != thread_tree.end()) && (count <= 0 || num_waiters < count) &&
@@ -221,11 +221,11 @@ ResultCode KAddressArbiter::WaitIfLessThan(VAddr addr, s32 value, bool decrement
         // Check that the thread isn't terminating.
         if (cur_thread->IsTerminationRequested()) {
             slp.CancelSleep();
-            return Svc::ResultTerminationRequested;
+            return ResultTerminationRequested;
         }
 
         // Set the synced object.
-        cur_thread->SetSyncedObject(nullptr, Svc::ResultTimedOut);
+        cur_thread->SetSyncedObject(nullptr, ResultTimedOut);
 
         // Read the value from userspace.
         s32 user_value{};
@@ -238,19 +238,19 @@ ResultCode KAddressArbiter::WaitIfLessThan(VAddr addr, s32 value, bool decrement
 
         if (!succeeded) {
             slp.CancelSleep();
-            return Svc::ResultInvalidCurrentMemory;
+            return ResultInvalidCurrentMemory;
         }
 
         // Check that the value is less than the specified one.
         if (user_value >= value) {
             slp.CancelSleep();
-            return Svc::ResultInvalidState;
+            return ResultInvalidState;
         }
 
         // Check that the timeout is non-zero.
         if (timeout == 0) {
             slp.CancelSleep();
-            return Svc::ResultTimedOut;
+            return ResultTimedOut;
         }
 
         // Set the arbiter.
@@ -288,29 +288,29 @@ ResultCode KAddressArbiter::WaitIfEqual(VAddr addr, s32 value, s64 timeout) {
         // Check that the thread isn't terminating.
         if (cur_thread->IsTerminationRequested()) {
             slp.CancelSleep();
-            return Svc::ResultTerminationRequested;
+            return ResultTerminationRequested;
         }
 
         // Set the synced object.
-        cur_thread->SetSyncedObject(nullptr, Svc::ResultTimedOut);
+        cur_thread->SetSyncedObject(nullptr, ResultTimedOut);
 
         // Read the value from userspace.
         s32 user_value{};
         if (!ReadFromUser(system, &user_value, addr)) {
             slp.CancelSleep();
-            return Svc::ResultInvalidCurrentMemory;
+            return ResultInvalidCurrentMemory;
         }
 
         // Check that the value is equal.
         if (value != user_value) {
             slp.CancelSleep();
-            return Svc::ResultInvalidState;
+            return ResultInvalidState;
         }
 
         // Check that the timeout is non-zero.
         if (timeout == 0) {
             slp.CancelSleep();
-            return Svc::ResultTimedOut;
+            return ResultTimedOut;
         }
 
         // Set the arbiter.

src/core/hle/kernel/k_condition_variable.cpp

@@ -92,10 +92,10 @@ ResultCode KConditionVariable::SignalToAddress(VAddr addr) {
         // Write the value to userspace.
         if (!WriteToUser(system, addr, std::addressof(next_value))) {
             if (next_owner_thread) {
-                next_owner_thread->SetSyncedObject(nullptr, Svc::ResultInvalidCurrentMemory);
+                next_owner_thread->SetSyncedObject(nullptr, ResultInvalidCurrentMemory);
             }
 
-            return Svc::ResultInvalidCurrentMemory;
+            return ResultInvalidCurrentMemory;
         }
     }
 
@@ -114,20 +114,20 @@ ResultCode KConditionVariable::WaitForAddress(Handle handle, VAddr addr, u32 val
             cur_thread->SetSyncedObject(nullptr, RESULT_SUCCESS);
 
             // Check if the thread should terminate.
-            R_UNLESS(!cur_thread->IsTerminationRequested(), Svc::ResultTerminationRequested);
+            R_UNLESS(!cur_thread->IsTerminationRequested(), ResultTerminationRequested);
 
             {
                 // Read the tag from userspace.
                 u32 test_tag{};
                 R_UNLESS(ReadFromUser(system, std::addressof(test_tag), addr),
-                         Svc::ResultInvalidCurrentMemory);
+                         ResultInvalidCurrentMemory);
 
                 // If the tag isn't the handle (with wait mask), we're done.
                 R_UNLESS(test_tag == (handle | Svc::HandleWaitMask), RESULT_SUCCESS);
 
                 // Get the lock owner thread.
                 owner_thread = kernel.CurrentProcess()->GetHandleTable().Get<KThread>(handle);
-                R_UNLESS(owner_thread, Svc::ResultInvalidHandle);
+                R_UNLESS(owner_thread, ResultInvalidHandle);
 
                 // Update the lock.
                 cur_thread->SetAddressKey(addr, value);
@@ -191,13 +191,13 @@ KThread* KConditionVariable::SignalImpl(KThread* thread) {
                 thread_to_close = owner_thread.get();
             } else {
                 // The lock was tagged with a thread that doesn't exist.
-                thread->SetSyncedObject(nullptr, Svc::ResultInvalidState);
+                thread->SetSyncedObject(nullptr, ResultInvalidState);
                 thread->Wakeup();
             }
         }
     } else {
         // If the address wasn't accessible, note so.
-        thread->SetSyncedObject(nullptr, Svc::ResultInvalidCurrentMemory);
+        thread->SetSyncedObject(nullptr, ResultInvalidCurrentMemory);
         thread->Wakeup();
     }
 
@@ -263,12 +263,12 @@ ResultCode KConditionVariable::Wait(VAddr addr, u64 key, u32 value, s64 timeout)
         KScopedSchedulerLockAndSleep slp{kernel, cur_thread, timeout};
 
         // Set the synced object.
-        cur_thread->SetSyncedObject(nullptr, Svc::ResultTimedOut);
+        cur_thread->SetSyncedObject(nullptr, ResultTimedOut);
 
         // Check that the thread isn't terminating.
         if (cur_thread->IsTerminationRequested()) {
             slp.CancelSleep();
-            return Svc::ResultTerminationRequested;
+            return ResultTerminationRequested;
         }
 
         // Update the value and process for the next owner.
@@ -302,7 +302,7 @@ ResultCode KConditionVariable::Wait(VAddr addr, u64 key, u32 value, s64 timeout)
             // Write the value to userspace.
             if (!WriteToUser(system, addr, std::addressof(next_value))) {
                 slp.CancelSleep();
-                return Svc::ResultInvalidCurrentMemory;
+                return ResultInvalidCurrentMemory;
             }
         }
 

src/core/hle/kernel/k_readable_event.cpp

@@ -6,7 +6,6 @@
 #include "common/assert.h"
 #include "common/common_funcs.h"
 #include "common/logging/log.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_thread.h"
@@ -47,7 +46,7 @@ ResultCode KReadableEvent::Reset() {
     KScopedSchedulerLock lk{kernel};
 
     if (!is_signaled) {
-        return Svc::ResultInvalidState;
+        return ResultInvalidState;
     }
 
     is_signaled = false;

src/core/hle/kernel/k_resource_limit.cpp

@@ -75,7 +75,7 @@ s64 KResourceLimit::GetFreeValue(LimitableResource which) const {
 ResultCode KResourceLimit::SetLimitValue(LimitableResource which, s64 value) {
     const auto index = static_cast<std::size_t>(which);
     KScopedLightLock lk(lock);
-    R_UNLESS(current_values[index] <= value, Svc::ResultInvalidState);
+    R_UNLESS(current_values[index] <= value, ResultInvalidState);
 
     limit_values[index] = value;
 

src/core/hle/kernel/k_scoped_resource_reservation.h

@@ -0,0 +1,67 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "core/hle/kernel/k_resource_limit.h"
+#include "core/hle/kernel/process.h"
+
+namespace Kernel {
+
+class KScopedResourceReservation {
+public:
+    explicit KScopedResourceReservation(std::shared_ptr<KResourceLimit> l, LimitableResource r,
+                                        s64 v, s64 timeout)
+        : resource_limit(std::move(l)), value(v), resource(r) {
+        if (resource_limit && value) {
+            success = resource_limit->Reserve(resource, value, timeout);
+        } else {
+            success = true;
+        }
+    }
+
+    explicit KScopedResourceReservation(std::shared_ptr<KResourceLimit> l, LimitableResource r,
+                                        s64 v = 1)
+        : resource_limit(std::move(l)), value(v), resource(r) {
+        if (resource_limit && value) {
+            success = resource_limit->Reserve(resource, value);
+        } else {
+            success = true;
+        }
+    }
+
+    explicit KScopedResourceReservation(const Process* p, LimitableResource r, s64 v, s64 t)
+        : KScopedResourceReservation(p->GetResourceLimit(), r, v, t) {}
+
+    explicit KScopedResourceReservation(const Process* p, LimitableResource r, s64 v = 1)
+        : KScopedResourceReservation(p->GetResourceLimit(), r, v) {}
+
+    ~KScopedResourceReservation() noexcept {
+        if (resource_limit && value && success) {
+            // resource was not committed, release the reservation.
+            resource_limit->Release(resource, value);
+        }
+    }
+
+    /// Commit the resource reservation, destruction of this object does not release the resource
+    void Commit() {
+        resource_limit = nullptr;
+    }
+
+    [[nodiscard]] bool Succeeded() const {
+        return success;
+    }
+
+private:
+    std::shared_ptr<KResourceLimit> resource_limit;
+    s64 value;
+    LimitableResource resource;
+    bool success;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_synchronization_object.cpp

@@ -40,20 +40,20 @@ ResultCode KSynchronizationObject::Wait(KernelCore& kernel, s32* out_index,
         // Check if the timeout is zero.
         if (timeout == 0) {
             slp.CancelSleep();
-            return Svc::ResultTimedOut;
+            return ResultTimedOut;
         }
 
         // Check if the thread should terminate.
         if (thread->IsTerminationRequested()) {
             slp.CancelSleep();
-            return Svc::ResultTerminationRequested;
+            return ResultTerminationRequested;
         }
 
         // Check if waiting was canceled.
         if (thread->IsWaitCancelled()) {
             slp.CancelSleep();
             thread->ClearWaitCancelled();
-            return Svc::ResultCancelled;
+            return ResultCancelled;
         }
 
         // Add the waiters.
@@ -75,7 +75,7 @@ ResultCode KSynchronizationObject::Wait(KernelCore& kernel, s32* out_index,
 
         // Mark the thread as waiting.
         thread->SetCancellable();
-        thread->SetSyncedObject(nullptr, Svc::ResultTimedOut);
+        thread->SetSyncedObject(nullptr, ResultTimedOut);
         thread->SetState(ThreadState::Waiting);
         thread->SetWaitReasonForDebugging(ThreadWaitReasonForDebugging::Synchronization);
     }

src/core/hle/kernel/k_thread.cpp

@@ -18,7 +18,6 @@
 #include "core/core.h"
 #include "core/cpu_manager.h"
 #include "core/hardware_properties.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/k_condition_variable.h"
 #include "core/hle/kernel/k_resource_limit.h"
@@ -127,7 +126,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
 
     // Set core ID and wait result.
     core_id = phys_core;
-    wait_result = Svc::ResultNoSynchronizationObject;
+    wait_result = ResultNoSynchronizationObject;
 
     // Set priorities.
     priority = prio;
@@ -238,7 +237,7 @@ void KThread::Finalize() {
         while (it != waiter_list.end()) {
             // The thread shouldn't be a kernel waiter.
             it->SetLockOwner(nullptr);
-            it->SetSyncedObject(nullptr, Svc::ResultInvalidState);
+            it->SetSyncedObject(nullptr, ResultInvalidState);
             it->Wakeup();
             it = waiter_list.erase(it);
         }
@@ -447,7 +446,7 @@ ResultCode KThread::SetCoreMask(s32 core_id, u64 v_affinity_mask) {
         // If the core id is no-update magic, preserve the ideal core id.
         if (core_id == Svc::IdealCoreNoUpdate) {
             core_id = virtual_ideal_core_id;
-            R_UNLESS(((1ULL << core_id) & v_affinity_mask) != 0, Svc::ResultInvalidCombination);
+            R_UNLESS(((1ULL << core_id) & v_affinity_mask) != 0, ResultInvalidCombination);
         }
 
         // Set the virtual core/affinity mask.
@@ -526,7 +525,7 @@ ResultCode KThread::SetCoreMask(s32 core_id, u64 v_affinity_mask) {
                 if (GetStackParameters().is_pinned) {
                     // Verify that the current thread isn't terminating.
                     R_UNLESS(!GetCurrentThread(kernel).IsTerminationRequested(),
-                             Svc::ResultTerminationRequested);
+                             ResultTerminationRequested);
 
                     // Note that the thread was pinned.
                     thread_is_pinned = true;
@@ -604,7 +603,7 @@ void KThread::WaitCancel() {
             sleeping_queue->WakeupThread(this);
             wait_cancelled = true;
         } else {
-            SetSyncedObject(nullptr, Svc::ResultCancelled);
+            SetSyncedObject(nullptr, ResultCancelled);
             SetState(ThreadState::Runnable);
             wait_cancelled = false;
         }
@@ -663,12 +662,12 @@ ResultCode KThread::SetActivity(Svc::ThreadActivity activity) {
         // Verify our state.
         const auto cur_state = GetState();
         R_UNLESS((cur_state == ThreadState::Waiting || cur_state == ThreadState::Runnable),
-                 Svc::ResultInvalidState);
+                 ResultInvalidState);
 
         // Either pause or resume.
         if (activity == Svc::ThreadActivity::Paused) {
             // Verify that we're not suspended.
-            R_UNLESS(!IsSuspendRequested(SuspendType::Thread), Svc::ResultInvalidState);
+            R_UNLESS(!IsSuspendRequested(SuspendType::Thread), ResultInvalidState);
 
             // Suspend.
             RequestSuspend(SuspendType::Thread);
@@ -676,7 +675,7 @@ ResultCode KThread::SetActivity(Svc::ThreadActivity activity) {
             ASSERT(activity == Svc::ThreadActivity::Runnable);
 
             // Verify that we're suspended.
-            R_UNLESS(IsSuspendRequested(SuspendType::Thread), Svc::ResultInvalidState);
+            R_UNLESS(IsSuspendRequested(SuspendType::Thread), ResultInvalidState);
 
             // Resume.
             Resume(SuspendType::Thread);
@@ -698,7 +697,7 @@ ResultCode KThread::SetActivity(Svc::ThreadActivity activity) {
             if (GetStackParameters().is_pinned) {
                 // Verify that the current thread isn't terminating.
                 R_UNLESS(!GetCurrentThread(kernel).IsTerminationRequested(),
-                         Svc::ResultTerminationRequested);
+                         ResultTerminationRequested);
 
                 // Note that the thread was pinned and not current.
                 thread_is_pinned = true;
@@ -745,7 +744,7 @@ ResultCode KThread::GetThreadContext3(std::vector<u8>& out) {
         KScopedSchedulerLock sl{kernel};
 
         // Verify that we're suspended.
-        R_UNLESS(IsSuspendRequested(SuspendType::Thread), Svc::ResultInvalidState);
+        R_UNLESS(IsSuspendRequested(SuspendType::Thread), ResultInvalidState);
 
         // If we're not terminating, get the thread's user context.
         if (!IsTerminationRequested()) {
@@ -905,12 +904,11 @@ ResultCode KThread::Run() {
         KScopedSchedulerLock lk{kernel};
 
         // If either this thread or the current thread are requesting termination, note it.
-        R_UNLESS(!IsTerminationRequested(), Svc::ResultTerminationRequested);
-        R_UNLESS(!GetCurrentThread(kernel).IsTerminationRequested(),
-                 Svc::ResultTerminationRequested);
+        R_UNLESS(!IsTerminationRequested(), ResultTerminationRequested);
+        R_UNLESS(!GetCurrentThread(kernel).IsTerminationRequested(), ResultTerminationRequested);
 
         // Ensure our thread state is correct.
-        R_UNLESS(GetState() == ThreadState::Initialized, Svc::ResultInvalidState);
+        R_UNLESS(GetState() == ThreadState::Initialized, ResultInvalidState);
 
         // If the current thread has been asked to suspend, suspend it and retry.
         if (GetCurrentThread(kernel).IsSuspended()) {
@@ -962,7 +960,7 @@ ResultCode KThread::Sleep(s64 timeout) {
         // Check if the thread should terminate.
         if (IsTerminationRequested()) {
             slp.CancelSleep();
-            return Svc::ResultTerminationRequested;
+            return ResultTerminationRequested;
         }
 
         // Mark the thread as waiting.

src/core/hle/kernel/kernel.cpp

@@ -26,7 +26,6 @@
 #include "core/device_memory.h"
 #include "core/hardware_properties.h"
 #include "core/hle/kernel/client_port.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/k_scheduler.h"
@@ -39,6 +38,7 @@
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/service_thread.h"
 #include "core/hle/kernel/shared_memory.h"
+#include "core/hle/kernel/svc_results.h"
 #include "core/hle/kernel/time_manager.h"
 #include "core/hle/lock.h"
 #include "core/hle/result.h"
@@ -141,11 +141,17 @@ struct KernelCore::Impl {
         ASSERT(system_resource_limit->SetLimitValue(LimitableResource::Events, 700).IsSuccess());
         ASSERT(system_resource_limit->SetLimitValue(LimitableResource::TransferMemory, 200)
                    .IsSuccess());
-        ASSERT(system_resource_limit->SetLimitValue(LimitableResource::Sessions, 900).IsSuccess());
+        ASSERT(system_resource_limit->SetLimitValue(LimitableResource::Sessions, 933).IsSuccess());
 
-        if (!system_resource_limit->Reserve(LimitableResource::PhysicalMemory, 0x60000)) {
+        // Derived from recent software updates. The kernel reserves 27MB
+        constexpr u64 kernel_size{0x1b00000};
+        if (!system_resource_limit->Reserve(LimitableResource::PhysicalMemory, kernel_size)) {
             UNREACHABLE();
         }
+        // Reserve secure applet memory, introduced in firmware 5.0.0
+        constexpr u64 secure_applet_memory_size{0x400000};
+        ASSERT(system_resource_limit->Reserve(LimitableResource::PhysicalMemory,
+                                              secure_applet_memory_size));
     }
 
     void InitializePreemption(KernelCore& kernel) {
@@ -302,8 +308,11 @@ struct KernelCore::Impl {
         // Allocate slab heaps
         user_slab_heap_pages = std::make_unique<Memory::SlabHeap<Memory::Page>>();
 
+        constexpr u64 user_slab_heap_size{0x1ef000};
+        // Reserve slab heaps
+        ASSERT(
+            system_resource_limit->Reserve(LimitableResource::PhysicalMemory, user_slab_heap_size));
         // Initialize slab heaps
-        constexpr u64 user_slab_heap_size{0x3de000};
         user_slab_heap_pages->Initialize(
             system.DeviceMemory().GetPointer(Core::DramMemoryMap::SlabHeapBase),
             user_slab_heap_size);

src/core/hle/kernel/memory/memory_manager.cpp

@@ -8,9 +8,9 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/scope_exit.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/memory/memory_manager.h"
 #include "core/hle/kernel/memory/page_linked_list.h"
+#include "core/hle/kernel/svc_results.h"
 
 namespace Kernel::Memory {
 
@@ -95,7 +95,7 @@ ResultCode MemoryManager::Allocate(PageLinkedList& page_list, std::size_t num_pa
     // Choose a heap based on our page size request
     const s32 heap_index{PageHeap::GetBlockIndex(num_pages)};
     if (heap_index < 0) {
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     // TODO (bunnei): Support multiple managers
@@ -140,7 +140,7 @@ ResultCode MemoryManager::Allocate(PageLinkedList& page_list, std::size_t num_pa
 
     // Only succeed if we allocated as many pages as we wanted
     if (num_pages) {
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     // We succeeded!

src/core/hle/kernel/memory/page_table.cpp

@@ -6,8 +6,7 @@
 #include "common/assert.h"
 #include "common/scope_exit.h"
 #include "core/core.h"
-#include "core/hle/kernel/errors.h"
-#include "core/hle/kernel/k_resource_limit.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/address_space_info.h"
 #include "core/hle/kernel/memory/memory_block.h"
@@ -16,6 +15,7 @@
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/memory/system_control.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/svc_results.h"
 #include "core/memory.h"
 
 namespace Kernel::Memory {
@@ -141,7 +141,7 @@ ResultCode PageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_t
         (alias_region_size + heap_region_size + stack_region_size + kernel_map_region_size)};
     if (alloc_size < needed_size) {
         UNREACHABLE();
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     const std::size_t remaining_size{alloc_size - needed_size};
@@ -277,11 +277,11 @@ ResultCode PageTable::MapProcessCode(VAddr addr, std::size_t num_pages, MemorySt
     const u64 size{num_pages * PageSize};
 
     if (!CanContain(addr, size, state)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (IsRegionMapped(addr, size)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     PageLinkedList page_linked_list;
@@ -307,7 +307,7 @@ ResultCode PageTable::MapProcessCodeMemory(VAddr dst_addr, VAddr src_addr, std::
                                   MemoryAttribute::None, MemoryAttribute::IpcAndDeviceMapped));
 
     if (IsRegionMapped(dst_addr, size)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     PageLinkedList page_linked_list;
@@ -409,27 +409,25 @@ ResultCode PageTable::MapPhysicalMemory(VAddr addr, std::size_t size) {
         return RESULT_SUCCESS;
     }
 
-    auto process{system.Kernel().CurrentProcess()};
     const std::size_t remaining_size{size - mapped_size};
     const std::size_t remaining_pages{remaining_size / PageSize};
 
-    if (process->GetResourceLimit() &&
-        !process->GetResourceLimit()->Reserve(LimitableResource::PhysicalMemory, remaining_size)) {
-        return ERR_RESOURCE_LIMIT_EXCEEDED;
+    // Reserve the memory from the process resource limit.
+    KScopedResourceReservation memory_reservation(
+        system.Kernel().CurrentProcess()->GetResourceLimit(), LimitableResource::PhysicalMemory,
+        remaining_size);
+    if (!memory_reservation.Succeeded()) {
+        LOG_ERROR(Kernel, "Could not reserve remaining {:X} bytes", remaining_size);
+        return ResultResourceLimitedExceeded;
     }
 
     PageLinkedList page_linked_list;
-    {
-        auto block_guard = detail::ScopeExit([&] {
-            system.Kernel().MemoryManager().Free(page_linked_list, remaining_pages, memory_pool);
-            process->GetResourceLimit()->Release(LimitableResource::PhysicalMemory, remaining_size);
-        });
 
-        CASCADE_CODE(system.Kernel().MemoryManager().Allocate(page_linked_list, remaining_pages,
-                                                              memory_pool));
+    CASCADE_CODE(
+        system.Kernel().MemoryManager().Allocate(page_linked_list, remaining_pages, memory_pool));
 
-        block_guard.Cancel();
-    }
+    // We succeeded, so commit the memory reservation.
+    memory_reservation.Commit();
 
     MapPhysicalMemory(page_linked_list, addr, end_addr);
 
@@ -454,12 +452,12 @@ ResultCode PageTable::UnmapPhysicalMemory(VAddr addr, std::size_t size) {
     block_manager->IterateForRange(addr, end_addr, [&](const MemoryInfo& info) {
         if (info.state == MemoryState::Normal) {
             if (info.attribute != MemoryAttribute::None) {
-                result = ERR_INVALID_ADDRESS_STATE;
+                result = ResultInvalidCurrentMemory;
                 return;
             }
             mapped_size += GetSizeInRange(info, addr, end_addr);
         } else if (info.state != MemoryState::Free) {
-            result = ERR_INVALID_ADDRESS_STATE;
+            result = ResultInvalidCurrentMemory;
         }
     });
 
@@ -526,7 +524,7 @@ ResultCode PageTable::Map(VAddr dst_addr, VAddr src_addr, std::size_t size) {
         MemoryAttribute::Mask, MemoryAttribute::None, MemoryAttribute::IpcAndDeviceMapped));
 
     if (IsRegionMapped(dst_addr, size)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     PageLinkedList page_linked_list;
@@ -577,7 +575,7 @@ ResultCode PageTable::Unmap(VAddr dst_addr, VAddr src_addr, std::size_t size) {
     AddRegionToPages(dst_addr, num_pages, dst_pages);
 
     if (!dst_pages.IsEqual(src_pages)) {
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     {
@@ -626,11 +624,11 @@ ResultCode PageTable::MapPages(VAddr addr, PageLinkedList& page_linked_list, Mem
     const std::size_t size{num_pages * PageSize};
 
     if (!CanContain(addr, size, state)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (IsRegionMapped(addr, num_pages * PageSize)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     CASCADE_CODE(MapPages(addr, page_linked_list, perm));
@@ -768,7 +766,7 @@ ResultCode PageTable::SetHeapCapacity(std::size_t new_heap_capacity) {
 ResultVal<VAddr> PageTable::SetHeapSize(std::size_t size) {
 
     if (size > heap_region_end - heap_region_start) {
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     const u64 previous_heap_size{GetHeapSize()};
@@ -781,10 +779,14 @@ ResultVal<VAddr> PageTable::SetHeapSize(std::size_t size) {
 
         const u64 delta{size - previous_heap_size};
 
-        auto process{system.Kernel().CurrentProcess()};
-        if (process->GetResourceLimit() && delta != 0 &&
-            !process->GetResourceLimit()->Reserve(LimitableResource::PhysicalMemory, delta)) {
-            return ERR_RESOURCE_LIMIT_EXCEEDED;
+        // Reserve memory for the heap extension.
+        KScopedResourceReservation memory_reservation(
+            system.Kernel().CurrentProcess()->GetResourceLimit(), LimitableResource::PhysicalMemory,
+            delta);
+
+        if (!memory_reservation.Succeeded()) {
+            LOG_ERROR(Kernel, "Could not reserve heap extension of size {:X} bytes", delta);
+            return ResultResourceLimitedExceeded;
         }
 
         PageLinkedList page_linked_list;
@@ -794,12 +796,15 @@ ResultVal<VAddr> PageTable::SetHeapSize(std::size_t size) {
             system.Kernel().MemoryManager().Allocate(page_linked_list, num_pages, memory_pool));
 
         if (IsRegionMapped(current_heap_addr, delta)) {
-            return ERR_INVALID_ADDRESS_STATE;
+            return ResultInvalidCurrentMemory;
         }
 
         CASCADE_CODE(
             Operate(current_heap_addr, num_pages, page_linked_list, OperationType::MapGroup));
 
+        // Succeeded in allocation, commit the resource reservation
+        memory_reservation.Commit();
+
         block_manager->Update(current_heap_addr, num_pages, MemoryState::Normal,
                               MemoryPermission::ReadAndWrite);
 
@@ -816,17 +821,17 @@ ResultVal<VAddr> PageTable::AllocateAndMapMemory(std::size_t needed_num_pages, s
     std::lock_guard lock{page_table_lock};
 
     if (!CanContain(region_start, region_num_pages * PageSize, state)) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (region_num_pages <= needed_num_pages) {
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     const VAddr addr{
         AllocateVirtualMemory(region_start, region_num_pages, needed_num_pages, align)};
     if (!addr) {
-        return ERR_OUT_OF_MEMORY;
+        return ResultOutOfMemory;
     }
 
     if (is_map_only) {
@@ -1105,13 +1110,13 @@ constexpr ResultCode PageTable::CheckMemoryState(const MemoryInfo& info, MemoryS
                                                  MemoryAttribute attr) const {
     // Validate the states match expectation
     if ((info.state & state_mask) != state) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
     if ((info.perm & perm_mask) != perm) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
     if ((info.attribute & attr_mask) != attr) {
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     return RESULT_SUCCESS;
@@ -1138,14 +1143,14 @@ ResultCode PageTable::CheckMemoryState(MemoryState* out_state, MemoryPermission*
     while (true) {
         // Validate the current block
         if (!(info.state == first_state)) {
-            return ERR_INVALID_ADDRESS_STATE;
+            return ResultInvalidCurrentMemory;
         }
         if (!(info.perm == first_perm)) {
-            return ERR_INVALID_ADDRESS_STATE;
+            return ResultInvalidCurrentMemory;
         }
         if (!((info.attribute | static_cast<MemoryAttribute>(ignore_attr)) ==
               (first_attr | static_cast<MemoryAttribute>(ignore_attr)))) {
-            return ERR_INVALID_ADDRESS_STATE;
+            return ResultInvalidCurrentMemory;
         }
 
         // Validate against the provided masks

src/core/hle/kernel/process.cpp

@@ -14,9 +14,9 @@
 #include "core/device_memory.h"
 #include "core/file_sys/program_metadata.h"
 #include "core/hle/kernel/code_set.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/memory_block_manager.h"
@@ -39,6 +39,7 @@ namespace {
  */
 void SetupMainThread(Core::System& system, Process& owner_process, u32 priority, VAddr stack_top) {
     const VAddr entry_point = owner_process.PageTable().GetCodeRegionStart();
+    ASSERT(owner_process.GetResourceLimit()->Reserve(LimitableResource::Threads, 1));
     auto thread_res = KThread::Create(system, ThreadType::User, "main", entry_point, priority, 0,
                                       owner_process.GetIdealCoreId(), stack_top, &owner_process);
 
@@ -117,6 +118,9 @@ std::shared_ptr<Process> Process::Create(Core::System& system, std::string name,
 
     std::shared_ptr<Process> process = std::make_shared<Process>(system);
     process->name = std::move(name);
+
+    // TODO: This is inaccurate
+    // The process should hold a reference to the kernel-wide resource limit.
     process->resource_limit = std::make_shared<KResourceLimit>(kernel, system);
     process->status = ProcessStatus::Created;
     process->program_id = 0;
@@ -155,6 +159,9 @@ void Process::DecrementThreadCount() {
 }
 
 u64 Process::GetTotalPhysicalMemoryAvailable() const {
+    // TODO: This is expected to always return the application memory pool size after accurately
+    // reserving kernel resources. The current workaround uses a process-local resource limit of
+    // application memory pool size, which is inaccurate.
     const u64 capacity{resource_limit->GetFreeValue(LimitableResource::PhysicalMemory) +
                        page_table->GetTotalHeapSize() + GetSystemResourceSize() + image_size +
                        main_thread_stack_size};
@@ -248,8 +255,8 @@ ResultCode Process::Reset() {
     KScopedSchedulerLock sl{kernel};
 
     // Validate that we're in a state that we can reset.
-    R_UNLESS(status != ProcessStatus::Exited, Svc::ResultInvalidState);
-    R_UNLESS(is_signaled, Svc::ResultInvalidState);
+    R_UNLESS(status != ProcessStatus::Exited, ResultInvalidState);
+    R_UNLESS(is_signaled, ResultInvalidState);
 
     // Clear signaled.
     is_signaled = false;
@@ -264,6 +271,17 @@ ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata,
     system_resource_size = metadata.GetSystemResourceSize();
     image_size = code_size;
 
+    // Set initial resource limits
+    resource_limit->SetLimitValue(
+        LimitableResource::PhysicalMemory,
+        kernel.MemoryManager().GetSize(Memory::MemoryManager::Pool::Application));
+    KScopedResourceReservation memory_reservation(resource_limit, LimitableResource::PhysicalMemory,
+                                                  code_size + system_resource_size);
+    if (!memory_reservation.Succeeded()) {
+        LOG_ERROR(Kernel, "Could not reserve process memory requirements of size {:X} bytes",
+                  code_size + system_resource_size);
+        return ResultResourceLimitedExceeded;
+    }
     // Initialize proces address space
     if (const ResultCode result{
             page_table->InitializeForProcess(metadata.GetAddressSpaceType(), false, 0x8000000,
@@ -305,24 +323,22 @@ ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata,
         UNREACHABLE();
     }
 
-    // Set initial resource limits
-    resource_limit->SetLimitValue(
-        LimitableResource::PhysicalMemory,
-        kernel.MemoryManager().GetSize(Memory::MemoryManager::Pool::Application));
     resource_limit->SetLimitValue(LimitableResource::Threads, 608);
     resource_limit->SetLimitValue(LimitableResource::Events, 700);
     resource_limit->SetLimitValue(LimitableResource::TransferMemory, 128);
     resource_limit->SetLimitValue(LimitableResource::Sessions, 894);
-    ASSERT(resource_limit->Reserve(LimitableResource::PhysicalMemory, code_size));
 
     // Create TLS region
     tls_region_address = CreateTLSRegion();
+    memory_reservation.Commit();
 
     return handle_table.SetSize(capabilities.GetHandleTableSize());
 }
 
 void Process::Run(s32 main_thread_priority, u64 stack_size) {
     AllocateMainThreadStack(stack_size);
+    resource_limit->Reserve(LimitableResource::Threads, 1);
+    resource_limit->Reserve(LimitableResource::PhysicalMemory, main_thread_stack_size);
 
     const std::size_t heap_capacity{memory_usage_capacity - main_thread_stack_size - image_size};
     ASSERT(!page_table->SetHeapCapacity(heap_capacity).IsError());
@@ -330,8 +346,6 @@ void Process::Run(s32 main_thread_priority, u64 stack_size) {
     ChangeStatus(ProcessStatus::Running);
 
     SetupMainThread(system, *this, main_thread_priority, main_thread_stack_top);
-    resource_limit->Reserve(LimitableResource::Threads, 1);
-    resource_limit->Reserve(LimitableResource::PhysicalMemory, main_thread_stack_size);
 }
 
 void Process::PrepareForTermination() {
@@ -358,6 +372,11 @@ void Process::PrepareForTermination() {
     FreeTLSRegion(tls_region_address);
     tls_region_address = 0;
 
+    if (resource_limit) {
+        resource_limit->Release(LimitableResource::PhysicalMemory,
+                                main_thread_stack_size + image_size);
+    }
+
     ChangeStatus(ProcessStatus::Exited);
 }
 

src/core/hle/kernel/process_capability.cpp

@@ -6,10 +6,10 @@
 
 #include "common/bit_util.h"
 #include "common/logging/log.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/process_capability.h"
+#include "core/hle/kernel/svc_results.h"
 
 namespace Kernel {
 namespace {
@@ -123,13 +123,13 @@ ResultCode ProcessCapabilities::ParseCapabilities(const u32* capabilities,
             // If there's only one, then there's a problem.
             if (i >= num_capabilities) {
                 LOG_ERROR(Kernel, "Invalid combination! i={}", i);
-                return ERR_INVALID_COMBINATION;
+                return ResultInvalidCombination;
             }
 
             const auto size_flags = capabilities[i];
             if (GetCapabilityType(size_flags) != CapabilityType::MapPhysical) {
                 LOG_ERROR(Kernel, "Invalid capability type! size_flags={}", size_flags);
-                return ERR_INVALID_COMBINATION;
+                return ResultInvalidCombination;
             }
 
             const auto result = HandleMapPhysicalFlags(descriptor, size_flags, page_table);
@@ -159,7 +159,7 @@ ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& s
     const auto type = GetCapabilityType(flag);
 
     if (type == CapabilityType::Unset) {
-        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+        return ResultInvalidCapabilityDescriptor;
     }
 
     // Bail early on ignorable entries, as one would expect,
@@ -176,7 +176,7 @@ ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& s
         LOG_ERROR(Kernel,
                   "Attempted to initialize flags that may only be initialized once. set_flags={}",
                   set_flags);
-        return ERR_INVALID_COMBINATION;
+        return ResultInvalidCombination;
     }
     set_flags |= set_flag;
 
@@ -202,7 +202,7 @@ ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& s
     }
 
     LOG_ERROR(Kernel, "Invalid capability type! type={}", type);
-    return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    return ResultInvalidCapabilityDescriptor;
 }
 
 void ProcessCapabilities::Clear() {
@@ -225,7 +225,7 @@ ResultCode ProcessCapabilities::HandlePriorityCoreNumFlags(u32 flags) {
     if (priority_mask != 0 || core_mask != 0) {
         LOG_ERROR(Kernel, "Core or priority mask are not zero! priority_mask={}, core_mask={}",
                   priority_mask, core_mask);
-        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+        return ResultInvalidCapabilityDescriptor;
     }
 
     const u32 core_num_min = (flags >> 16) & 0xFF;
@@ -233,7 +233,7 @@ ResultCode ProcessCapabilities::HandlePriorityCoreNumFlags(u32 flags) {
     if (core_num_min > core_num_max) {
         LOG_ERROR(Kernel, "Core min is greater than core max! core_num_min={}, core_num_max={}",
                   core_num_min, core_num_max);
-        return ERR_INVALID_COMBINATION;
+        return ResultInvalidCombination;
     }
 
     const u32 priority_min = (flags >> 10) & 0x3F;
@@ -242,13 +242,13 @@ ResultCode ProcessCapabilities::HandlePriorityCoreNumFlags(u32 flags) {
         LOG_ERROR(Kernel,
                   "Priority min is greater than priority max! priority_min={}, priority_max={}",
                   core_num_min, priority_max);
-        return ERR_INVALID_COMBINATION;
+        return ResultInvalidCombination;
     }
 
     // The switch only has 4 usable cores.
     if (core_num_max >= 4) {
         LOG_ERROR(Kernel, "Invalid max cores specified! core_num_max={}", core_num_max);
-        return ERR_INVALID_PROCESSOR_ID;
+        return ResultInvalidCoreId;
     }
 
     const auto make_mask = [](u64 min, u64 max) {
@@ -269,7 +269,7 @@ ResultCode ProcessCapabilities::HandleSyscallFlags(u32& set_svc_bits, u32 flags)
 
     // If we've already set this svc before, bail.
     if ((set_svc_bits & svc_bit) != 0) {
-        return ERR_INVALID_COMBINATION;
+        return ResultInvalidCombination;
     }
     set_svc_bits |= svc_bit;
 
@@ -283,7 +283,7 @@ ResultCode ProcessCapabilities::HandleSyscallFlags(u32& set_svc_bits, u32 flags)
 
         if (svc_number >= svc_capabilities.size()) {
             LOG_ERROR(Kernel, "Process svc capability is out of range! svc_number={}", svc_number);
-            return ERR_OUT_OF_RANGE;
+            return ResultOutOfRange;
         }
 
         svc_capabilities[svc_number] = true;
@@ -321,7 +321,7 @@ ResultCode ProcessCapabilities::HandleInterruptFlags(u32 flags) {
         if (interrupt >= interrupt_capabilities.size()) {
             LOG_ERROR(Kernel, "Process interrupt capability is out of range! svc_number={}",
                       interrupt);
-            return ERR_OUT_OF_RANGE;
+            return ResultOutOfRange;
         }
 
         interrupt_capabilities[interrupt] = true;
@@ -334,7 +334,7 @@ ResultCode ProcessCapabilities::HandleProgramTypeFlags(u32 flags) {
     const u32 reserved = flags >> 17;
     if (reserved != 0) {
         LOG_ERROR(Kernel, "Reserved value is non-zero! reserved={}", reserved);
-        return ERR_RESERVED_VALUE;
+        return ResultReservedValue;
     }
 
     program_type = static_cast<ProgramType>((flags >> 14) & 0b111);
@@ -354,7 +354,7 @@ ResultCode ProcessCapabilities::HandleKernelVersionFlags(u32 flags) {
         LOG_ERROR(Kernel,
                   "Kernel version is non zero or flags are too small! major_version={}, flags={}",
                   major_version, flags);
-        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+        return ResultInvalidCapabilityDescriptor;
     }
 
     kernel_version = flags;
@@ -365,7 +365,7 @@ ResultCode ProcessCapabilities::HandleHandleTableFlags(u32 flags) {
     const u32 reserved = flags >> 26;
     if (reserved != 0) {
         LOG_ERROR(Kernel, "Reserved value is non-zero! reserved={}", reserved);
-        return ERR_RESERVED_VALUE;
+        return ResultReservedValue;
     }
 
     handle_table_size = static_cast<s32>((flags >> 16) & 0x3FF);
@@ -376,7 +376,7 @@ ResultCode ProcessCapabilities::HandleDebugFlags(u32 flags) {
     const u32 reserved = flags >> 19;
     if (reserved != 0) {
         LOG_ERROR(Kernel, "Reserved value is non-zero! reserved={}", reserved);
-        return ERR_RESERVED_VALUE;
+        return ResultReservedValue;
     }
 
     is_debuggable = (flags & 0x20000) != 0;

src/core/hle/kernel/server_port.cpp

@@ -5,11 +5,11 @@
 #include <tuple>
 #include "common/assert.h"
 #include "core/hle/kernel/client_port.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/server_port.h"
 #include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/svc_results.h"
 
 namespace Kernel {
 
@@ -18,7 +18,7 @@ ServerPort::~ServerPort() = default;
 
 ResultVal<std::shared_ptr<ServerSession>> ServerPort::Accept() {
     if (pending_sessions.empty()) {
-        return ERR_NOT_FOUND;
+        return ResultNotFound;
     }
 
     auto session = std::move(pending_sessions.back());

src/core/hle/kernel/session.cpp

@@ -4,15 +4,23 @@
 
 #include "common/assert.h"
 #include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/server_session.h"
 #include "core/hle/kernel/session.h"
 
 namespace Kernel {
 
 Session::Session(KernelCore& kernel) : KSynchronizationObject{kernel} {}
-Session::~Session() = default;
+Session::~Session() {
+    // Release reserved resource when the Session pair was created.
+    kernel.GetSystemResourceLimit()->Release(LimitableResource::Sessions, 1);
+}
 
 Session::SessionPair Session::Create(KernelCore& kernel, std::string name) {
+    // Reserve a new session from the resource limit.
+    KScopedResourceReservation session_reservation(kernel.GetSystemResourceLimit(),
+                                                   LimitableResource::Sessions);
+    ASSERT(session_reservation.Succeeded());
     auto session{std::make_shared<Session>(kernel)};
     auto client_session{Kernel::ClientSession::Create(kernel, session, name + "_Client").Unwrap()};
     auto server_session{Kernel::ServerSession::Create(kernel, session, name + "_Server").Unwrap()};
@@ -21,6 +29,7 @@ Session::SessionPair Session::Create(KernelCore& kernel, std::string name) {
     session->client = client_session;
     session->server = server_session;
 
+    session_reservation.Commit();
     return std::make_pair(std::move(client_session), std::move(server_session));
 }
 

src/core/hle/kernel/shared_memory.cpp

@@ -4,6 +4,7 @@
 
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/shared_memory.h"
@@ -13,7 +14,9 @@ namespace Kernel {
 SharedMemory::SharedMemory(KernelCore& kernel, Core::DeviceMemory& device_memory)
     : Object{kernel}, device_memory{device_memory} {}
 
-SharedMemory::~SharedMemory() = default;
+SharedMemory::~SharedMemory() {
+    kernel.GetSystemResourceLimit()->Release(LimitableResource::PhysicalMemory, size);
+}
 
 std::shared_ptr<SharedMemory> SharedMemory::Create(
     KernelCore& kernel, Core::DeviceMemory& device_memory, Process* owner_process,
@@ -21,6 +24,11 @@ std::shared_ptr<SharedMemory> SharedMemory::Create(
     Memory::MemoryPermission user_permission, PAddr physical_address, std::size_t size,
     std::string name) {
 
+    const auto resource_limit = kernel.GetSystemResourceLimit();
+    KScopedResourceReservation memory_reservation(resource_limit, LimitableResource::PhysicalMemory,
+                                                  size);
+    ASSERT(memory_reservation.Succeeded());
+
     std::shared_ptr<SharedMemory> shared_memory{
         std::make_shared<SharedMemory>(kernel, device_memory)};
 
@@ -32,6 +40,7 @@ std::shared_ptr<SharedMemory> SharedMemory::Create(
     shared_memory->size = size;
     shared_memory->name = name;
 
+    memory_reservation.Commit();
     return shared_memory;
 }
 

src/core/hle/kernel/svc.cpp

@@ -23,7 +23,6 @@
 #include "core/cpu_manager.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/k_address_arbiter.h"
 #include "core/hle/kernel/k_condition_variable.h"
@@ -31,6 +30,7 @@
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/k_synchronization_object.h"
 #include "core/hle/kernel/k_thread.h"
@@ -71,49 +71,49 @@ ResultCode MapUnmapMemorySanityChecks(const Memory::PageTable& manager, VAddr ds
                                       VAddr src_addr, u64 size) {
     if (!Common::Is4KBAligned(dst_addr)) {
         LOG_ERROR(Kernel_SVC, "Destination address is not aligned to 4KB, 0x{:016X}", dst_addr);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!Common::Is4KBAligned(src_addr)) {
         LOG_ERROR(Kernel_SVC, "Source address is not aligned to 4KB, 0x{:016X}", src_addr);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (size == 0) {
         LOG_ERROR(Kernel_SVC, "Size is 0");
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Size is not aligned to 4KB, 0x{:016X}", size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!IsValidAddressRange(dst_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Destination is not a valid address range, addr=0x{:016X}, size=0x{:016X}",
                   dst_addr, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (!IsValidAddressRange(src_addr, size)) {
         LOG_ERROR(Kernel_SVC, "Source is not a valid address range, addr=0x{:016X}, size=0x{:016X}",
                   src_addr, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (!manager.IsInsideAddressSpace(src_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Source is not within the address space, addr=0x{:016X}, size=0x{:016X}",
                   src_addr, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (manager.IsOutsideStackRegion(dst_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Destination is not within the stack region, addr=0x{:016X}, size=0x{:016X}",
                   dst_addr, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     if (manager.IsInsideHeapRegion(dst_addr, size)) {
@@ -121,7 +121,7 @@ ResultCode MapUnmapMemorySanityChecks(const Memory::PageTable& manager, VAddr ds
                   "Destination does not fit within the heap region, addr=0x{:016X}, "
                   "size=0x{:016X}",
                   dst_addr, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     if (manager.IsInsideAliasRegion(dst_addr, size)) {
@@ -129,7 +129,7 @@ ResultCode MapUnmapMemorySanityChecks(const Memory::PageTable& manager, VAddr ds
                   "Destination does not fit within the map region, addr=0x{:016X}, "
                   "size=0x{:016X}",
                   dst_addr, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     return RESULT_SUCCESS;
@@ -138,6 +138,7 @@ ResultCode MapUnmapMemorySanityChecks(const Memory::PageTable& manager, VAddr ds
 enum class ResourceLimitValueType {
     CurrentValue,
     LimitValue,
+    PeakValue,
 };
 
 ResultVal<s64> RetrieveResourceLimitValue(Core::System& system, Handle resource_limit,
@@ -146,7 +147,7 @@ ResultVal<s64> RetrieveResourceLimitValue(Core::System& system, Handle resource_
     const auto type = static_cast<LimitableResource>(resource_type);
     if (!IsValidResourceType(type)) {
         LOG_ERROR(Kernel_SVC, "Invalid resource limit type: '{}'", resource_type);
-        return ERR_INVALID_ENUM_VALUE;
+        return ResultInvalidEnumValue;
     }
 
     const auto* const current_process = system.Kernel().CurrentProcess();
@@ -157,14 +158,20 @@ ResultVal<s64> RetrieveResourceLimitValue(Core::System& system, Handle resource_
     if (!resource_limit_object) {
         LOG_ERROR(Kernel_SVC, "Handle to non-existent resource limit instance used. Handle={:08X}",
                   resource_limit);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
-    if (value_type == ResourceLimitValueType::CurrentValue) {
+    switch (value_type) {
+    case ResourceLimitValueType::CurrentValue:
         return MakeResult(resource_limit_object->GetCurrentValue(type));
+    case ResourceLimitValueType::LimitValue:
+        return MakeResult(resource_limit_object->GetLimitValue(type));
+    case ResourceLimitValueType::PeakValue:
+        return MakeResult(resource_limit_object->GetPeakValue(type));
+    default:
+        LOG_ERROR(Kernel_SVC, "Invalid resource value_type: '{}'", value_type);
+        return ResultInvalidEnumValue;
     }
-
-    return MakeResult(resource_limit_object->GetLimitValue(type));
 }
 } // Anonymous namespace
 
@@ -177,12 +184,12 @@ static ResultCode SetHeapSize(Core::System& system, VAddr* heap_addr, u64 heap_s
     if ((heap_size % 0x200000) != 0) {
         LOG_ERROR(Kernel_SVC, "The heap size is not a multiple of 2MB, heap_size=0x{:016X}",
                   heap_size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (heap_size >= 0x200000000) {
         LOG_ERROR(Kernel_SVC, "The heap size is not less than 8GB, heap_size=0x{:016X}", heap_size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     auto& page_table{system.Kernel().CurrentProcess()->PageTable()};
@@ -208,19 +215,19 @@ static ResultCode SetMemoryAttribute(Core::System& system, VAddr address, u64 si
 
     if (!Common::Is4KBAligned(address)) {
         LOG_ERROR(Kernel_SVC, "Address not page aligned (0x{:016X})", address);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (size == 0 || !Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Invalid size (0x{:X}). Size must be non-zero and page aligned.",
                   size);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!IsValidAddressRange(address, size)) {
         LOG_ERROR(Kernel_SVC, "Address range overflowed (Address: 0x{:016X}, Size: 0x{:016X})",
                   address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     const auto attributes{static_cast<Memory::MemoryAttribute>(mask | attribute)};
@@ -229,7 +236,7 @@ static ResultCode SetMemoryAttribute(Core::System& system, VAddr address, u64 si
         LOG_ERROR(Kernel_SVC,
                   "Memory attribute doesn't match the given mask (Attribute: 0x{:X}, Mask: {:X}",
                   attribute, mask);
-        return ERR_INVALID_COMBINATION;
+        return ResultInvalidCombination;
     }
 
     auto& page_table{system.Kernel().CurrentProcess()->PageTable()};
@@ -293,7 +300,7 @@ static ResultCode ConnectToNamedPort(Core::System& system, Handle* out_handle,
         LOG_ERROR(Kernel_SVC,
                   "Port Name Address is not a valid virtual address, port_name_address=0x{:016X}",
                   port_name_address);
-        return ERR_NOT_FOUND;
+        return ResultNotFound;
     }
 
     static constexpr std::size_t PortNameMaxLength = 11;
@@ -302,7 +309,7 @@ static ResultCode ConnectToNamedPort(Core::System& system, Handle* out_handle,
     if (port_name.size() > PortNameMaxLength) {
         LOG_ERROR(Kernel_SVC, "Port name is too long, expected {} but got {}", PortNameMaxLength,
                   port_name.size());
-        return ERR_OUT_OF_RANGE;
+        return ResultOutOfRange;
     }
 
     LOG_TRACE(Kernel_SVC, "called port_name={}", port_name);
@@ -311,11 +318,9 @@ static ResultCode ConnectToNamedPort(Core::System& system, Handle* out_handle,
     const auto it = kernel.FindNamedPort(port_name);
     if (!kernel.IsValidNamedPort(it)) {
         LOG_WARNING(Kernel_SVC, "tried to connect to unknown port: {}", port_name);
-        return ERR_NOT_FOUND;
+        return ResultNotFound;
     }
 
-    ASSERT(kernel.CurrentProcess()->GetResourceLimit()->Reserve(LimitableResource::Sessions, 1));
-
     auto client_port = it->second;
 
     std::shared_ptr<ClientSession> client_session;
@@ -340,7 +345,7 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
     std::shared_ptr<ClientSession> session = handle_table.Get<ClientSession>(handle);
     if (!session) {
         LOG_ERROR(Kernel_SVC, "called with invalid handle=0x{:08X}", handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     LOG_TRACE(Kernel_SVC, "called handle=0x{:08X}({})", handle, session->GetName());
@@ -405,7 +410,7 @@ static ResultCode GetProcessId(Core::System& system, u64* process_id, Handle han
         const Process* const owner_process = thread->GetOwnerProcess();
         if (!owner_process) {
             LOG_ERROR(Kernel_SVC, "Non-existent owning process encountered.");
-            return ERR_INVALID_HANDLE;
+            return ResultInvalidHandle;
         }
 
         *process_id = owner_process->GetProcessID();
@@ -415,7 +420,7 @@ static ResultCode GetProcessId(Core::System& system, u64* process_id, Handle han
     // NOTE: This should also handle debug objects before returning.
 
     LOG_ERROR(Kernel_SVC, "Handle does not exist, handle=0x{:08X}", handle);
-    return ERR_INVALID_HANDLE;
+    return ResultInvalidHandle;
 }
 
 static ResultCode GetProcessId32(Core::System& system, u32* process_id_low, u32* process_id_high,
@@ -438,7 +443,7 @@ static ResultCode WaitSynchronization(Core::System& system, s32* index, VAddr ha
         LOG_ERROR(Kernel_SVC,
                   "Handle address is not a valid virtual address, handle_address=0x{:016X}",
                   handles_address);
-        return ERR_INVALID_POINTER;
+        return ResultInvalidPointer;
     }
 
     static constexpr u64 MaxHandles = 0x40;
@@ -446,7 +451,7 @@ static ResultCode WaitSynchronization(Core::System& system, s32* index, VAddr ha
     if (handle_count > MaxHandles) {
         LOG_ERROR(Kernel_SVC, "Handle count specified is too large, expected {} but got {}",
                   MaxHandles, handle_count);
-        return ERR_OUT_OF_RANGE;
+        return ResultOutOfRange;
     }
 
     auto& kernel = system.Kernel();
@@ -459,7 +464,7 @@ static ResultCode WaitSynchronization(Core::System& system, s32* index, VAddr ha
 
         if (object == nullptr) {
             LOG_ERROR(Kernel_SVC, "Object is a nullptr");
-            return ERR_INVALID_HANDLE;
+            return ResultInvalidHandle;
         }
 
         objects[i] = object.get();
@@ -481,6 +486,7 @@ static ResultCode CancelSynchronization(Core::System& system, Handle thread_hand
     // Get the thread from its handle.
     const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
     std::shared_ptr<KThread> thread = handle_table.Get<KThread>(thread_handle);
+
     if (!thread) {
         LOG_ERROR(Kernel_SVC, "Invalid thread handle provided (handle={:08X})", thread_handle);
         return ResultInvalidHandle;
@@ -525,6 +531,7 @@ static ResultCode ArbitrateUnlock(Core::System& system, VAddr address) {
     LOG_TRACE(Kernel_SVC, "called address=0x{:X}", address);
 
     // Validate the input address.
+
     if (Memory::IsKernelAddress(address)) {
         LOG_ERROR(Kernel_SVC,
                   "Attempting to arbitrate an unlock on a kernel address (address={:08X})",
@@ -735,7 +742,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         if (info_sub_id != 0) {
             LOG_ERROR(Kernel_SVC, "Info sub id is non zero! info_id={}, info_sub_id={}", info_id,
                       info_sub_id);
-            return ERR_INVALID_ENUM_VALUE;
+            return ResultInvalidEnumValue;
         }
 
         const auto& current_process_handle_table =
@@ -744,7 +751,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         if (!process) {
             LOG_ERROR(Kernel_SVC, "Process is not valid! info_id={}, info_sub_id={}, handle={:08X}",
                       info_id, info_sub_id, handle);
-            return ERR_INVALID_HANDLE;
+            return ResultInvalidHandle;
         }
 
         switch (info_id_type) {
@@ -826,7 +833,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         }
 
         LOG_ERROR(Kernel_SVC, "Unimplemented svcGetInfo id=0x{:016X}", info_id);
-        return ERR_INVALID_ENUM_VALUE;
+        return ResultInvalidEnumValue;
     }
 
     case GetInfoType::IsCurrentProcessBeingDebugged:
@@ -836,13 +843,13 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
     case GetInfoType::RegisterResourceLimit: {
         if (handle != 0) {
             LOG_ERROR(Kernel, "Handle is non zero! handle={:08X}", handle);
-            return ERR_INVALID_HANDLE;
+            return ResultInvalidHandle;
         }
 
         if (info_sub_id != 0) {
             LOG_ERROR(Kernel, "Info sub id is non zero! info_id={}, info_sub_id={}", info_id,
                       info_sub_id);
-            return ERR_INVALID_COMBINATION;
+            return ResultInvalidCombination;
         }
 
         Process* const current_process = system.Kernel().CurrentProcess();
@@ -867,13 +874,13 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         if (handle != 0) {
             LOG_ERROR(Kernel_SVC, "Process Handle is non zero, expected 0 result but got {:016X}",
                       handle);
-            return ERR_INVALID_HANDLE;
+            return ResultInvalidHandle;
         }
 
         if (info_sub_id >= Process::RANDOM_ENTROPY_SIZE) {
             LOG_ERROR(Kernel_SVC, "Entropy size is out of range, expected {} but got {}",
                       Process::RANDOM_ENTROPY_SIZE, info_sub_id);
-            return ERR_INVALID_COMBINATION;
+            return ResultInvalidCombination;
         }
 
         *result = system.Kernel().CurrentProcess()->GetRandomEntropy(info_sub_id);
@@ -890,7 +897,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         if (info_sub_id != 0xFFFFFFFFFFFFFFFF && info_sub_id >= num_cpus) {
             LOG_ERROR(Kernel_SVC, "Core count is out of range, expected {} but got {}", num_cpus,
                       info_sub_id);
-            return ERR_INVALID_COMBINATION;
+            return ResultInvalidCombination;
         }
 
         const auto thread = system.Kernel().CurrentProcess()->GetHandleTable().Get<KThread>(
@@ -898,7 +905,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         if (!thread) {
             LOG_ERROR(Kernel_SVC, "Thread handle does not exist, handle=0x{:08X}",
                       static_cast<Handle>(handle));
-            return ERR_INVALID_HANDLE;
+            return ResultInvalidHandle;
         }
 
         const auto& core_timing = system.CoreTiming();
@@ -922,7 +929,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
 
     default:
         LOG_ERROR(Kernel_SVC, "Unimplemented svcGetInfo id=0x{:016X}", info_id);
-        return ERR_INVALID_ENUM_VALUE;
+        return ResultInvalidEnumValue;
     }
 }
 
@@ -945,22 +952,22 @@ static ResultCode MapPhysicalMemory(Core::System& system, VAddr addr, u64 size)
 
     if (!Common::Is4KBAligned(addr)) {
         LOG_ERROR(Kernel_SVC, "Address is not aligned to 4KB, 0x{:016X}", addr);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Size is not aligned to 4KB, 0x{:X}", size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (size == 0) {
         LOG_ERROR(Kernel_SVC, "Size is zero");
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!(addr < addr + size)) {
         LOG_ERROR(Kernel_SVC, "Size causes 64-bit overflow of address");
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     Process* const current_process{system.Kernel().CurrentProcess()};
@@ -968,21 +975,21 @@ static ResultCode MapPhysicalMemory(Core::System& system, VAddr addr, u64 size)
 
     if (current_process->GetSystemResourceSize() == 0) {
         LOG_ERROR(Kernel_SVC, "System Resource Size is zero");
-        return ERR_INVALID_STATE;
+        return ResultInvalidState;
     }
 
     if (!page_table.IsInsideAddressSpace(addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Address is not within the address space, addr=0x{:016X}, size=0x{:016X}", addr,
                   size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     if (page_table.IsOutsideAliasRegion(addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Address is not within the alias region, addr=0x{:016X}, size=0x{:016X}", addr,
                   size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     return page_table.MapPhysicalMemory(addr, size);
@@ -999,22 +1006,22 @@ static ResultCode UnmapPhysicalMemory(Core::System& system, VAddr addr, u64 size
 
     if (!Common::Is4KBAligned(addr)) {
         LOG_ERROR(Kernel_SVC, "Address is not aligned to 4KB, 0x{:016X}", addr);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Size is not aligned to 4KB, 0x{:X}", size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (size == 0) {
         LOG_ERROR(Kernel_SVC, "Size is zero");
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!(addr < addr + size)) {
         LOG_ERROR(Kernel_SVC, "Size causes 64-bit overflow of address");
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     Process* const current_process{system.Kernel().CurrentProcess()};
@@ -1022,21 +1029,21 @@ static ResultCode UnmapPhysicalMemory(Core::System& system, VAddr addr, u64 size
 
     if (current_process->GetSystemResourceSize() == 0) {
         LOG_ERROR(Kernel_SVC, "System Resource Size is zero");
-        return ERR_INVALID_STATE;
+        return ResultInvalidState;
     }
 
     if (!page_table.IsInsideAddressSpace(addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Address is not within the address space, addr=0x{:016X}, size=0x{:016X}", addr,
                   size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     if (page_table.IsOutsideAliasRegion(addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Address is not within the alias region, addr=0x{:016X}, size=0x{:016X}", addr,
                   size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     return page_table.UnmapPhysicalMemory(addr, size);
@@ -1206,23 +1213,23 @@ static ResultCode MapSharedMemory(Core::System& system, Handle shared_memory_han
 
     if (!Common::Is4KBAligned(addr)) {
         LOG_ERROR(Kernel_SVC, "Address is not aligned to 4KB, addr=0x{:016X}", addr);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (size == 0) {
         LOG_ERROR(Kernel_SVC, "Size is 0");
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Size is not aligned to 4KB, size=0x{:016X}", size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!IsValidAddressRange(addr, size)) {
         LOG_ERROR(Kernel_SVC, "Region is not a valid address range, addr=0x{:016X}, size=0x{:016X}",
                   addr, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     const auto permission_type = static_cast<Memory::MemoryPermission>(permissions);
@@ -1230,7 +1237,7 @@ static ResultCode MapSharedMemory(Core::System& system, Handle shared_memory_han
         Memory::MemoryPermission::ReadAndWrite) {
         LOG_ERROR(Kernel_SVC, "Expected Read or ReadWrite permission but got permissions=0x{:08X}",
                   permissions);
-        return ERR_INVALID_MEMORY_PERMISSIONS;
+        return ResultInvalidMemoryPermissions;
     }
 
     auto* const current_process{system.Kernel().CurrentProcess()};
@@ -1241,7 +1248,7 @@ static ResultCode MapSharedMemory(Core::System& system, Handle shared_memory_han
                   "Addr does not fit within the valid region, addr=0x{:016X}, "
                   "size=0x{:016X}",
                   addr, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     if (page_table.IsInsideHeapRegion(addr, size)) {
@@ -1249,7 +1256,7 @@ static ResultCode MapSharedMemory(Core::System& system, Handle shared_memory_han
                   "Addr does not fit within the heap region, addr=0x{:016X}, "
                   "size=0x{:016X}",
                   addr, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     if (page_table.IsInsideAliasRegion(addr, size)) {
@@ -1257,14 +1264,14 @@ static ResultCode MapSharedMemory(Core::System& system, Handle shared_memory_han
                   "Address does not fit within the map region, addr=0x{:016X}, "
                   "size=0x{:016X}",
                   addr, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     auto shared_memory{current_process->GetHandleTable().Get<SharedMemory>(shared_memory_handle)};
     if (!shared_memory) {
         LOG_ERROR(Kernel_SVC, "Shared memory does not exist, shared_memory_handle=0x{:08X}",
                   shared_memory_handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     return shared_memory->Map(*current_process, addr, size, permission_type);
@@ -1285,7 +1292,7 @@ static ResultCode QueryProcessMemory(Core::System& system, VAddr memory_info_add
     if (!process) {
         LOG_ERROR(Kernel_SVC, "Process handle does not exist, process_handle=0x{:08X}",
                   process_handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     auto& memory{system.Memory()};
@@ -1332,18 +1339,18 @@ static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_hand
     if (!Common::Is4KBAligned(src_address)) {
         LOG_ERROR(Kernel_SVC, "src_address is not page-aligned (src_address=0x{:016X}).",
                   src_address);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!Common::Is4KBAligned(dst_address)) {
         LOG_ERROR(Kernel_SVC, "dst_address is not page-aligned (dst_address=0x{:016X}).",
                   dst_address);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (size == 0 || !Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Size is zero or not page-aligned (size=0x{:016X})", size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!IsValidAddressRange(dst_address, size)) {
@@ -1351,7 +1358,7 @@ static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_hand
                   "Destination address range overflows the address space (dst_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   dst_address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (!IsValidAddressRange(src_address, size)) {
@@ -1359,7 +1366,7 @@ static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_hand
                   "Source address range overflows the address space (src_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   src_address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
@@ -1367,7 +1374,7 @@ static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_hand
     if (!process) {
         LOG_ERROR(Kernel_SVC, "Invalid process handle specified (handle=0x{:08X}).",
                   process_handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     auto& page_table = process->PageTable();
@@ -1376,7 +1383,7 @@ static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_hand
                   "Source address range is not within the address space (src_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   src_address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (!page_table.IsInsideASLRRegion(dst_address, size)) {
@@ -1384,7 +1391,7 @@ static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_hand
                   "Destination address range is not within the ASLR region (dst_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   dst_address, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     return page_table.MapProcessCodeMemory(dst_address, src_address, size);
@@ -1400,18 +1407,18 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
     if (!Common::Is4KBAligned(dst_address)) {
         LOG_ERROR(Kernel_SVC, "dst_address is not page-aligned (dst_address=0x{:016X}).",
                   dst_address);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!Common::Is4KBAligned(src_address)) {
         LOG_ERROR(Kernel_SVC, "src_address is not page-aligned (src_address=0x{:016X}).",
                   src_address);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (size == 0 || Common::Is4KBAligned(size)) {
         LOG_ERROR(Kernel_SVC, "Size is zero or not page-aligned (size=0x{:016X}).", size);
-        return ERR_INVALID_SIZE;
+        return ResultInvalidSize;
     }
 
     if (!IsValidAddressRange(dst_address, size)) {
@@ -1419,7 +1426,7 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
                   "Destination address range overflows the address space (dst_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   dst_address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (!IsValidAddressRange(src_address, size)) {
@@ -1427,7 +1434,7 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
                   "Source address range overflows the address space (src_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   src_address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
@@ -1435,7 +1442,7 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
     if (!process) {
         LOG_ERROR(Kernel_SVC, "Invalid process handle specified (handle=0x{:08X}).",
                   process_handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     auto& page_table = process->PageTable();
@@ -1444,7 +1451,7 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
                   "Source address range is not within the address space (src_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   src_address, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     if (!page_table.IsInsideASLRRegion(dst_address, size)) {
@@ -1452,7 +1459,7 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
                   "Destination address range is not within the ASLR region (dst_address=0x{:016X}, "
                   "size=0x{:016X}).",
                   dst_address, size);
-        return ERR_INVALID_MEMORY_RANGE;
+        return ResultInvalidMemoryRange;
     }
 
     return page_table.UnmapProcessCodeMemory(dst_address, src_address, size);
@@ -1515,8 +1522,13 @@ static ResultCode CreateThread(Core::System& system, Handle* out_handle, VAddr e
         return ResultInvalidPriority;
     }
 
-    ASSERT(process.GetResourceLimit()->Reserve(
-        LimitableResource::Threads, 1, system.CoreTiming().GetGlobalTimeNs().count() + 100000000));
+    KScopedResourceReservation thread_reservation(
+        kernel.CurrentProcess(), LimitableResource::Threads, 1,
+        system.CoreTiming().GetGlobalTimeNs().count() + 100000000);
+    if (!thread_reservation.Succeeded()) {
+        LOG_ERROR(Kernel_SVC, "Could not reserve a new thread");
+        return ResultResourceLimitedExceeded;
+    }
 
     std::shared_ptr<KThread> thread;
     {
@@ -1536,6 +1548,7 @@ static ResultCode CreateThread(Core::System& system, Handle* out_handle, VAddr e
     // Set the thread name for debugging purposes.
     thread->SetName(
         fmt::format("thread[entry_point={:X}, handle={:X}]", entry_point, *new_thread_handle));
+    thread_reservation.Commit();
 
     return RESULT_SUCCESS;
 }
@@ -1844,7 +1857,7 @@ static ResultCode ResetSignal(Core::System& system, Handle handle) {
 
     LOG_ERROR(Kernel_SVC, "invalid handle (0x{:08X})", handle);
 
-    return Svc::ResultInvalidHandle;
+    return ResultInvalidHandle;
 }
 
 static ResultCode ResetSignal32(Core::System& system, Handle handle) {
@@ -1860,18 +1873,18 @@ static ResultCode CreateTransferMemory(Core::System& system, Handle* handle, VAd
 
     if (!Common::Is4KBAligned(addr)) {
         LOG_ERROR(Kernel_SVC, "Address ({:016X}) is not page aligned!", addr);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!Common::Is4KBAligned(size) || size == 0) {
         LOG_ERROR(Kernel_SVC, "Size ({:016X}) is not page aligned or equal to zero!", size);
-        return ERR_INVALID_ADDRESS;
+        return ResultInvalidAddress;
     }
 
     if (!IsValidAddressRange(addr, size)) {
         LOG_ERROR(Kernel_SVC, "Address and size cause overflow! (address={:016X}, size={:016X})",
                   addr, size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     const auto perms{static_cast<Memory::MemoryPermission>(permissions)};
@@ -1879,10 +1892,17 @@ static ResultCode CreateTransferMemory(Core::System& system, Handle* handle, VAd
         perms == Memory::MemoryPermission::Write) {
         LOG_ERROR(Kernel_SVC, "Invalid memory permissions for transfer memory! (perms={:08X})",
                   permissions);
-        return ERR_INVALID_MEMORY_PERMISSIONS;
+        return ResultInvalidMemoryPermissions;
     }
 
     auto& kernel = system.Kernel();
+    // Reserve a new transfer memory from the process resource limit.
+    KScopedResourceReservation trmem_reservation(kernel.CurrentProcess(),
+                                                 LimitableResource::TransferMemory);
+    if (!trmem_reservation.Succeeded()) {
+        LOG_ERROR(Kernel_SVC, "Could not reserve a new transfer memory");
+        return ResultResourceLimitedExceeded;
+    }
     auto transfer_mem_handle = TransferMemory::Create(kernel, system.Memory(), addr, size, perms);
 
     if (const auto reserve_result{transfer_mem_handle->Reserve()}; reserve_result.IsError()) {
@@ -1894,6 +1914,7 @@ static ResultCode CreateTransferMemory(Core::System& system, Handle* handle, VAd
     if (result.Failed()) {
         return result.Code();
     }
+    trmem_reservation.Commit();
 
     *handle = *result;
     return RESULT_SUCCESS;
@@ -1989,7 +2010,6 @@ static ResultCode SetThreadCoreMask(Core::System& system, Handle thread_handle,
         LOG_ERROR(Kernel_SVC, "Unable to successfully set core mask (result={})", set_result.raw);
         return set_result;
     }
-
     return RESULT_SUCCESS;
 }
 
@@ -2002,8 +2022,17 @@ static ResultCode SetThreadCoreMask32(Core::System& system, Handle thread_handle
 static ResultCode SignalEvent(Core::System& system, Handle event_handle) {
     LOG_DEBUG(Kernel_SVC, "called, event_handle=0x{:08X}", event_handle);
 
+    auto& kernel = system.Kernel();
     // Get the current handle table.
-    const HandleTable& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
+    const HandleTable& handle_table = kernel.CurrentProcess()->GetHandleTable();
+
+    // Reserve a new event from the process resource limit.
+    KScopedResourceReservation event_reservation(kernel.CurrentProcess(),
+                                                 LimitableResource::Events);
+    if (!event_reservation.Succeeded()) {
+        LOG_ERROR(Kernel, "Could not reserve a new event");
+        return ResultResourceLimitedExceeded;
+    }
 
     // Get the writable event.
     auto writable_event = handle_table.Get<KWritableEvent>(event_handle);
@@ -2012,6 +2041,9 @@ static ResultCode SignalEvent(Core::System& system, Handle event_handle) {
         return ResultInvalidHandle;
     }
 
+    // Commit the successfuly reservation.
+    event_reservation.Commit();
+
     return writable_event->Signal();
 }
 
@@ -2043,7 +2075,7 @@ static ResultCode ClearEvent(Core::System& system, Handle event_handle) {
 
     LOG_ERROR(Kernel_SVC, "Event handle does not exist, event_handle=0x{:08X}", event_handle);
 
-    return Svc::ResultInvalidHandle;
+    return ResultInvalidHandle;
 }
 
 static ResultCode ClearEvent32(Core::System& system, Handle event_handle) {
@@ -2106,13 +2138,13 @@ static ResultCode GetProcessInfo(Core::System& system, u64* out, Handle process_
     if (!process) {
         LOG_ERROR(Kernel_SVC, "Process handle does not exist, process_handle=0x{:08X}",
                   process_handle);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     const auto info_type = static_cast<InfoType>(type);
     if (info_type != InfoType::Status) {
         LOG_ERROR(Kernel_SVC, "Expected info_type to be Status but got {} instead", type);
-        return ERR_INVALID_ENUM_VALUE;
+        return ResultInvalidEnumValue;
     }
 
     *out = static_cast<u64>(process->GetStatus());
@@ -2174,7 +2206,7 @@ static ResultCode SetResourceLimitLimitValue(Core::System& system, Handle resour
     const auto type = static_cast<LimitableResource>(resource_type);
     if (!IsValidResourceType(type)) {
         LOG_ERROR(Kernel_SVC, "Invalid resource limit type: '{}'", resource_type);
-        return ERR_INVALID_ENUM_VALUE;
+        return ResultInvalidEnumValue;
     }
 
     auto* const current_process = system.Kernel().CurrentProcess();
@@ -2185,16 +2217,16 @@ static ResultCode SetResourceLimitLimitValue(Core::System& system, Handle resour
     if (!resource_limit_object) {
         LOG_ERROR(Kernel_SVC, "Handle to non-existent resource limit instance used. Handle={:08X}",
                   resource_limit);
-        return ERR_INVALID_HANDLE;
+        return ResultInvalidHandle;
     }
 
     const auto set_result = resource_limit_object->SetLimitValue(type, static_cast<s64>(value));
     if (set_result.IsError()) {
-        LOG_ERROR(
-            Kernel_SVC,
-            "Attempted to lower resource limit ({}) for category '{}' below its current value ({})",
-            resource_limit_object->GetLimitValue(type), resource_type,
-            resource_limit_object->GetCurrentValue(type));
+        LOG_ERROR(Kernel_SVC,
+                  "Attempted to lower resource limit ({}) for category '{}' below its current "
+                  "value ({})",
+                  resource_limit_object->GetLimitValue(type), resource_type,
+                  resource_limit_object->GetCurrentValue(type));
         return set_result;
     }
 
@@ -2211,7 +2243,7 @@ static ResultCode GetProcessList(Core::System& system, u32* out_num_processes,
         LOG_ERROR(Kernel_SVC,
                   "Supplied size outside [0, 0x0FFFFFFF] range. out_process_ids_size={}",
                   out_process_ids_size);
-        return ERR_OUT_OF_RANGE;
+        return ResultOutOfRange;
     }
 
     const auto& kernel = system.Kernel();
@@ -2221,7 +2253,7 @@ static ResultCode GetProcessList(Core::System& system, u32* out_num_processes,
                                         out_process_ids, total_copy_size)) {
         LOG_ERROR(Kernel_SVC, "Address range outside address space. begin=0x{:016X}, end=0x{:016X}",
                   out_process_ids, out_process_ids + total_copy_size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     auto& memory = system.Memory();
@@ -2250,7 +2282,7 @@ static ResultCode GetThreadList(Core::System& system, u32* out_num_threads, VAdd
     if ((out_thread_ids_size & 0xF0000000) != 0) {
         LOG_ERROR(Kernel_SVC, "Supplied size outside [0, 0x0FFFFFFF] range. size={}",
                   out_thread_ids_size);
-        return ERR_OUT_OF_RANGE;
+        return ResultOutOfRange;
     }
 
     const auto* const current_process = system.Kernel().CurrentProcess();
@@ -2260,7 +2292,7 @@ static ResultCode GetThreadList(Core::System& system, u32* out_num_threads, VAdd
         !current_process->PageTable().IsInsideAddressSpace(out_thread_ids, total_copy_size)) {
         LOG_ERROR(Kernel_SVC, "Address range outside address space. begin=0x{:016X}, end=0x{:016X}",
                   out_thread_ids, out_thread_ids + total_copy_size);
-        return ERR_INVALID_ADDRESS_STATE;
+        return ResultInvalidCurrentMemory;
     }
 
     auto& memory = system.Memory();

src/core/hle/kernel/svc_results.h

@@ -1,4 +1,4 @@
-// Copyright 2020 yuzu emulator team
+// Copyright 2018 yuzu emulator team
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
@@ -6,21 +6,36 @@
 
 #include "core/hle/result.h"
 
-namespace Kernel::Svc {
+namespace Kernel {
 
+// Confirmed Switch kernel error codes
+
+constexpr ResultCode ResultMaxConnectionsReached{ErrorModule::Kernel, 7};
+constexpr ResultCode ResultInvalidCapabilityDescriptor{ErrorModule::Kernel, 14};
 constexpr ResultCode ResultNoSynchronizationObject{ErrorModule::Kernel, 57};
 constexpr ResultCode ResultTerminationRequested{ErrorModule::Kernel, 59};
+constexpr ResultCode ResultInvalidSize{ErrorModule::Kernel, 101};
 constexpr ResultCode ResultInvalidAddress{ErrorModule::Kernel, 102};
 constexpr ResultCode ResultOutOfResource{ErrorModule::Kernel, 103};
+constexpr ResultCode ResultOutOfMemory{ErrorModule::Kernel, 104};
+constexpr ResultCode ResultHandleTableFull{ErrorModule::Kernel, 105};
 constexpr ResultCode ResultInvalidCurrentMemory{ErrorModule::Kernel, 106};
+constexpr ResultCode ResultInvalidMemoryPermissions{ErrorModule::Kernel, 108};
+constexpr ResultCode ResultInvalidMemoryRange{ErrorModule::Kernel, 110};
 constexpr ResultCode ResultInvalidPriority{ErrorModule::Kernel, 112};
 constexpr ResultCode ResultInvalidCoreId{ErrorModule::Kernel, 113};
 constexpr ResultCode ResultInvalidHandle{ErrorModule::Kernel, 114};
+constexpr ResultCode ResultInvalidPointer{ErrorModule::Kernel, 115};
 constexpr ResultCode ResultInvalidCombination{ErrorModule::Kernel, 116};
 constexpr ResultCode ResultTimedOut{ErrorModule::Kernel, 117};
 constexpr ResultCode ResultCancelled{ErrorModule::Kernel, 118};
+constexpr ResultCode ResultOutOfRange{ErrorModule::Kernel, 119};
 constexpr ResultCode ResultInvalidEnumValue{ErrorModule::Kernel, 120};
+constexpr ResultCode ResultNotFound{ErrorModule::Kernel, 121};
 constexpr ResultCode ResultBusy{ErrorModule::Kernel, 122};
+constexpr ResultCode ResultSessionClosedByRemote{ErrorModule::Kernel, 123};
 constexpr ResultCode ResultInvalidState{ErrorModule::Kernel, 125};
+constexpr ResultCode ResultReservedValue{ErrorModule::Kernel, 126};
+constexpr ResultCode ResultResourceLimitedExceeded{ErrorModule::Kernel, 132};
 
-} // namespace Kernel::Svc
+} // namespace Kernel

src/core/hle/kernel/transfer_memory.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/process.h"
@@ -17,6 +18,7 @@ TransferMemory::TransferMemory(KernelCore& kernel, Core::Memory::Memory& memory)
 TransferMemory::~TransferMemory() {
     // Release memory region when transfer memory is destroyed
     Reset();
+    owner_process->GetResourceLimit()->Release(LimitableResource::TransferMemory, 1);
 }
 
 std::shared_ptr<TransferMemory> TransferMemory::Create(KernelCore& kernel,

src/core/hle/service/am/am.cpp

@@ -1047,20 +1047,21 @@ void IStorageAccessor::Write(Kernel::HLERequestContext& ctx) {
 
     const u64 offset{rp.Pop<u64>()};
     const std::vector<u8> data{ctx.ReadBuffer()};
+    const std::size_t size{std::min(data.size(), backing.GetSize() - offset)};
 
-    LOG_DEBUG(Service_AM, "called, offset={}, size={}", offset, data.size());
+    LOG_DEBUG(Service_AM, "called, offset={}, size={}", offset, size);
 
-    if (data.size() > backing.GetSize() - offset) {
+    if (offset > backing.GetSize()) {
         LOG_ERROR(Service_AM,
                   "offset is out of bounds, backing_buffer_sz={}, data_size={}, offset={}",
-                  backing.GetSize(), data.size(), offset);
+                  backing.GetSize(), size, offset);
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(ERR_SIZE_OUT_OF_BOUNDS);
         return;
     }
 
-    std::memcpy(backing.GetData().data() + offset, data.data(), data.size());
+    std::memcpy(backing.GetData().data() + offset, data.data(), size);
 
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
@@ -1070,11 +1071,11 @@ void IStorageAccessor::Read(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
 
     const u64 offset{rp.Pop<u64>()};
-    const std::size_t size{ctx.GetWriteBufferSize()};
+    const std::size_t size{std::min(ctx.GetWriteBufferSize(), backing.GetSize() - offset)};
 
     LOG_DEBUG(Service_AM, "called, offset={}, size={}", offset, size);
 
-    if (size > backing.GetSize() - offset) {
+    if (offset > backing.GetSize()) {
         LOG_ERROR(Service_AM, "offset is out of bounds, backing_buffer_sz={}, size={}, offset={}",
                   backing.GetSize(), size, offset);
 

src/core/hle/service/am/applets/controller.cpp

@@ -211,7 +211,8 @@ void Controller::Execute() {
     case ControllerSupportMode::ShowControllerFirmwareUpdate:
         UNIMPLEMENTED_MSG("ControllerSupportMode={} is not implemented",
                           controller_private_arg.mode);
-        [[fallthrough]];
+        ConfigurationComplete();
+        break;
     default: {
         ConfigurationComplete();
         break;

src/core/hle/service/hid/controllers/npad.cpp

@@ -21,6 +21,7 @@
 
 namespace Service::HID {
 constexpr s32 HID_JOYSTICK_MAX = 0x7fff;
+constexpr s32 HID_TRIGGER_MAX = 0x7fff;
 [[maybe_unused]] constexpr s32 HID_JOYSTICK_MIN = -0x7fff;
 constexpr std::size_t NPAD_OFFSET = 0x9A00;
 constexpr u32 BATTERY_FULL = 2;
@@ -48,6 +49,8 @@ Controller_NPad::NPadControllerType Controller_NPad::MapSettingsTypeToNPad(
         return NPadControllerType::JoyRight;
     case Settings::ControllerType::Handheld:
         return NPadControllerType::Handheld;
+    case Settings::ControllerType::GameCube:
+        return NPadControllerType::GameCube;
     default:
         UNREACHABLE();
         return NPadControllerType::ProController;
@@ -67,6 +70,8 @@ Settings::ControllerType Controller_NPad::MapNPadToSettingsType(
         return Settings::ControllerType::RightJoycon;
     case NPadControllerType::Handheld:
         return Settings::ControllerType::Handheld;
+    case NPadControllerType::GameCube:
+        return Settings::ControllerType::GameCube;
     default:
         UNREACHABLE();
         return Settings::ControllerType::ProController;
@@ -209,6 +214,13 @@ void Controller_NPad::InitNewlyAddedController(std::size_t controller_idx) {
         controller.assignment_mode = NpadAssignments::Single;
         controller.footer_type = AppletFooterUiType::JoyRightHorizontal;
         break;
+    case NPadControllerType::GameCube:
+        controller.style_set.gamecube.Assign(1);
+        // The GC Controller behaves like a wired Pro Controller
+        controller.device_type.fullkey.Assign(1);
+        controller.system_properties.is_vertical.Assign(1);
+        controller.system_properties.use_plus.Assign(1);
+        break;
     case NPadControllerType::Pokeball:
         controller.style_set.palma.Assign(1);
         controller.device_type.palma.Assign(1);
@@ -259,6 +271,7 @@ void Controller_NPad::OnInit() {
         style.joycon_right.Assign(1);
         style.joycon_dual.Assign(1);
         style.fullkey.Assign(1);
+        style.gamecube.Assign(1);
         style.palma.Assign(1);
     }
 
@@ -339,6 +352,7 @@ void Controller_NPad::RequestPadStateUpdate(u32 npad_id) {
     auto& pad_state = npad_pad_states[controller_idx].pad_states;
     auto& lstick_entry = npad_pad_states[controller_idx].l_stick;
     auto& rstick_entry = npad_pad_states[controller_idx].r_stick;
+    auto& trigger_entry = npad_trigger_states[controller_idx];
     const auto& button_state = buttons[controller_idx];
     const auto& analog_state = sticks[controller_idx];
     const auto [stick_l_x_f, stick_l_y_f] =
@@ -404,6 +418,17 @@ void Controller_NPad::RequestPadStateUpdate(u32 npad_id) {
         pad_state.left_sl.Assign(button_state[SL - BUTTON_HID_BEGIN]->GetStatus());
         pad_state.left_sr.Assign(button_state[SR - BUTTON_HID_BEGIN]->GetStatus());
     }
+
+    if (controller_type == NPadControllerType::GameCube) {
+        trigger_entry.l_analog = static_cast<s32>(
+            button_state[ZL - BUTTON_HID_BEGIN]->GetStatus() ? HID_TRIGGER_MAX : 0);
+        trigger_entry.r_analog = static_cast<s32>(
+            button_state[ZR - BUTTON_HID_BEGIN]->GetStatus() ? HID_TRIGGER_MAX : 0);
+        pad_state.zl.Assign(false);
+        pad_state.zr.Assign(button_state[R - BUTTON_HID_BEGIN]->GetStatus());
+        pad_state.l.Assign(button_state[ZL - BUTTON_HID_BEGIN]->GetStatus());
+        pad_state.r.Assign(button_state[ZR - BUTTON_HID_BEGIN]->GetStatus());
+    }
 }
 
 void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
@@ -418,6 +443,11 @@ void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8*
             &npad.joy_left_states,  &npad.joy_right_states, &npad.palma_states,
             &npad.system_ext_states};
 
+        // There is the posibility to have more controllers with analog triggers
+        const std::array<TriggerGeneric*, 1> controller_triggers{
+            &npad.gc_trigger_states,
+        };
+
         for (auto* main_controller : controller_npads) {
             main_controller->common.entry_count = 16;
             main_controller->common.total_entry_count = 17;
@@ -435,6 +465,21 @@ void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8*
             cur_entry.timestamp2 = cur_entry.timestamp;
         }
 
+        for (auto* analog_trigger : controller_triggers) {
+            analog_trigger->entry_count = 16;
+            analog_trigger->total_entry_count = 17;
+
+            const auto& last_entry = analog_trigger->trigger[analog_trigger->last_entry_index];
+
+            analog_trigger->timestamp = core_timing.GetCPUTicks();
+            analog_trigger->last_entry_index = (analog_trigger->last_entry_index + 1) % 17;
+
+            auto& cur_entry = analog_trigger->trigger[analog_trigger->last_entry_index];
+
+            cur_entry.timestamp = last_entry.timestamp + 1;
+            cur_entry.timestamp2 = cur_entry.timestamp;
+        }
+
         const auto& controller_type = connected_controllers[i].type;
 
         if (controller_type == NPadControllerType::None || !connected_controllers[i].is_connected) {
@@ -444,6 +489,7 @@ void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8*
 
         RequestPadStateUpdate(npad_index);
         auto& pad_state = npad_pad_states[npad_index];
+        auto& trigger_state = npad_trigger_states[npad_index];
 
         auto& main_controller =
             npad.fullkey_states.npad[npad.fullkey_states.common.last_entry_index];
@@ -456,6 +502,8 @@ void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8*
         auto& pokeball_entry = npad.palma_states.npad[npad.palma_states.common.last_entry_index];
         auto& libnx_entry =
             npad.system_ext_states.npad[npad.system_ext_states.common.last_entry_index];
+        auto& trigger_entry =
+            npad.gc_trigger_states.trigger[npad.gc_trigger_states.last_entry_index];
 
         libnx_entry.connection_status.raw = 0;
         libnx_entry.connection_status.is_connected.Assign(1);
@@ -524,6 +572,18 @@ void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8*
 
             libnx_entry.connection_status.is_right_connected.Assign(1);
             break;
+        case NPadControllerType::GameCube:
+            main_controller.connection_status.raw = 0;
+            main_controller.connection_status.is_connected.Assign(1);
+            main_controller.connection_status.is_wired.Assign(1);
+            main_controller.pad.pad_states.raw = pad_state.pad_states.raw;
+            main_controller.pad.l_stick = pad_state.l_stick;
+            main_controller.pad.r_stick = pad_state.r_stick;
+            trigger_entry.l_analog = trigger_state.l_analog;
+            trigger_entry.r_analog = trigger_state.r_analog;
+
+            libnx_entry.connection_status.is_wired.Assign(1);
+            break;
         case NPadControllerType::Pokeball:
             pokeball_entry.connection_status.raw = 0;
             pokeball_entry.connection_status.is_connected.Assign(1);
@@ -674,6 +734,7 @@ void Controller_NPad::OnMotionUpdate(const Core::Timing::CoreTiming& core_timing
                 right_sixaxis_entry.orientation = motion_devices[1].orientation;
             }
             break;
+        case NPadControllerType::GameCube:
         case NPadControllerType::Pokeball:
             break;
         }
@@ -1135,6 +1196,8 @@ bool Controller_NPad::IsControllerSupported(NPadControllerType controller) const
             return style.joycon_left;
         case NPadControllerType::JoyRight:
             return style.joycon_right;
+        case NPadControllerType::GameCube:
+            return style.gamecube;
         case NPadControllerType::Pokeball:
             return style.palma;
         default:

src/core/hle/service/hid/controllers/npad.h

@@ -51,6 +51,7 @@ public:
         JoyDual,
         JoyLeft,
         JoyRight,
+        GameCube,
         Pokeball,
     };
 
@@ -60,6 +61,7 @@ public:
         JoyconDual = 5,
         JoyconLeft = 6,
         JoyconRight = 7,
+        GameCube = 8,
         Pokeball = 9,
         MaxNpadType = 10,
     };
@@ -389,6 +391,25 @@ private:
     };
     static_assert(sizeof(SixAxisGeneric) == 0x708, "SixAxisGeneric is an invalid size");
 
+    struct TriggerState {
+        s64_le timestamp{};
+        s64_le timestamp2{};
+        s32_le l_analog{};
+        s32_le r_analog{};
+    };
+    static_assert(sizeof(TriggerState) == 0x18, "TriggerState is an invalid size");
+
+    struct TriggerGeneric {
+        INSERT_PADDING_BYTES(0x4);
+        s64_le timestamp;
+        INSERT_PADDING_BYTES(0x4);
+        s64_le total_entry_count;
+        s64_le last_entry_index;
+        s64_le entry_count;
+        std::array<TriggerState, 17> trigger{};
+    };
+    static_assert(sizeof(TriggerGeneric) == 0x1C8, "TriggerGeneric is an invalid size");
+
     struct NPadSystemProperties {
         union {
             s64_le raw{};
@@ -509,7 +530,9 @@ private:
         AppletFooterUiType footer_type;
         // nfc_states needs to be checked switchbrew does not match with HW
         NfcXcdHandle nfc_states;
-        INSERT_PADDING_BYTES(0xdef);
+        INSERT_PADDING_BYTES(0x8); // Mutex
+        TriggerGeneric gc_trigger_states;
+        INSERT_PADDING_BYTES(0xc1f);
     };
     static_assert(sizeof(NPadEntry) == 0x5000, "NPadEntry is an invalid size");
 
@@ -560,6 +583,7 @@ private:
     f32 sixaxis_fusion_parameter2{};
     bool sixaxis_at_rest{true};
     std::array<ControllerPad, 10> npad_pad_states{};
+    std::array<TriggerState, 10> npad_trigger_states{};
     bool is_in_lr_assignment_mode{false};
     Core::System& system;
 };

src/core/hle/service/ldn/errors.h

@@ -0,0 +1,13 @@
+// Copyright 2021 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/result.h"
+
+namespace Service::LDN {
+
+constexpr ResultCode ERROR_DISABLED{ErrorModule::LDN, 22};
+
+} // namespace Service::LDN

src/core/hle/service/ldn/ldn.cpp

@@ -6,6 +6,7 @@
 
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/result.h"
+#include "core/hle/service/ldn/errors.h"
 #include "core/hle/service/ldn/ldn.h"
 #include "core/hle/service/sm/sm.h"
 
@@ -103,7 +104,7 @@ public:
         : ServiceFramework{system_, "IUserLocalCommunicationService"} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, nullptr, "GetState"},
+            {0, &IUserLocalCommunicationService::GetState, "GetState"},
             {1, nullptr, "GetNetworkInfo"},
             {2, nullptr, "GetIpv4Address"},
             {3, nullptr, "GetDisconnectReason"},
@@ -138,13 +139,38 @@ public:
         RegisterHandlers(functions);
     }
 
-    void Initialize2(Kernel::HLERequestContext& ctx) {
+    void GetState(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_LDN, "(STUBBED) called");
-        // Result success seem make this services start network and continue.
-        // If we just pass result error then it will stop and maybe try again and again.
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(RESULT_UNKNOWN);
+
+        IPC::ResponseBuilder rb{ctx, 3};
+
+        // Indicate a network error, as we do not actually emulate LDN
+        rb.Push(static_cast<u32>(State::Error));
+
+        rb.Push(RESULT_SUCCESS);
     }
+
+    void Initialize2(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LDN, "called");
+
+        is_initialized = true;
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+private:
+    enum class State {
+        None,
+        Initialized,
+        AccessPointOpened,
+        AccessPointCreated,
+        StationOpened,
+        StationConnected,
+        Error,
+    };
+
+    bool is_initialized{};
 };
 
 class LDNS final : public ServiceFramework<LDNS> {

src/core/hle/service/ldr/ldr.cpp

@@ -11,10 +11,10 @@
 #include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/hle/ipc_helpers.h"
-#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/memory/system_control.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/svc_results.h"
 #include "core/hle/service/ldr/ldr.h"
 #include "core/hle/service/service.h"
 #include "core/loader/nro.h"
@@ -330,7 +330,7 @@ public:
             const VAddr addr{GetRandomMapRegion(page_table, size)};
             const ResultCode result{page_table.MapProcessCodeMemory(addr, baseAddress, size)};
 
-            if (result == Kernel::ERR_INVALID_ADDRESS_STATE) {
+            if (result == Kernel::ResultInvalidCurrentMemory) {
                 continue;
             }
 
@@ -361,7 +361,7 @@ public:
                 const ResultCode result{
                     page_table.MapProcessCodeMemory(addr + nro_size, bss_addr, bss_size)};
 
-                if (result == Kernel::ERR_INVALID_ADDRESS_STATE) {
+                if (result == Kernel::ResultInvalidCurrentMemory) {
                     continue;
                 }
 

src/core/hle/service/nfp/nfp.cpp

@@ -215,7 +215,7 @@ private:
         const auto& amiibo = nfp_interface.GetAmiiboBuffer();
         const TagInfo tag_info{
             .uuid = amiibo.uuid,
-            .uuid_length = static_cast<u8>(tag_info.uuid.size()),
+            .uuid_length = static_cast<u8>(amiibo.uuid.size()),
             .padding_1 = {},
             .protocol = 1, // TODO(ogniK): Figure out actual values
             .tag_type = 2,

src/core/hle/service/sockets/bsd.cpp

@@ -453,7 +453,8 @@ std::pair<s32, Errno> BSD::SocketImpl(Domain domain, Type type, Protocol protoco
         return {-1, Errno::MFILE};
     }
 
-    FileDescriptor& descriptor = file_descriptors[fd].emplace();
+    file_descriptors[fd] = FileDescriptor{};
+    FileDescriptor& descriptor = *file_descriptors[fd];
     // ENONMEM might be thrown here
 
     LOG_INFO(Service, "New socket fd={}", fd);
@@ -548,7 +549,8 @@ std::pair<s32, Errno> BSD::AcceptImpl(s32 fd, std::vector<u8>& write_buffer) {
         return {-1, Translate(bsd_errno)};
     }
 
-    FileDescriptor& new_descriptor = file_descriptors[new_fd].emplace();
+    file_descriptors[new_fd] = FileDescriptor{};
+    FileDescriptor& new_descriptor = *file_descriptors[new_fd];
     new_descriptor.socket = std::move(result.socket);
     new_descriptor.is_connection_based = descriptor.is_connection_based;
 

src/core/settings.h

@@ -181,12 +181,13 @@ struct Values {
     std::string motion_device;
     std::string udp_input_servers;
 
-    bool emulate_analog_keyboard;
-
+    bool mouse_panning;
+    float mouse_panning_sensitivity;
     bool mouse_enabled;
     std::string mouse_device;
     MouseButtonsRaw mouse_buttons;
 
+    bool emulate_analog_keyboard;
     bool keyboard_enabled;
     KeyboardKeysRaw keyboard_keys;
     KeyboardModsRaw keyboard_mods;

src/input_common/mouse/mouse_input.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2+
 // Refer to the license.txt file included.
 
+#include "core/settings.h"
 #include "input_common/mouse/mouse_input.h"
 
 namespace MouseInput {
@@ -32,10 +33,18 @@ void Mouse::UpdateThread() {
             info.motion.UpdateOrientation(update_time * 1000);
             info.tilt_speed = 0;
             info.data.motion = info.motion.GetMotion();
+            if (Settings::values.mouse_panning) {
+                info.last_mouse_change *= 0.96f;
+                info.data.axis = {static_cast<int>(16 * info.last_mouse_change.x),
+                                  static_cast<int>(16 * -info.last_mouse_change.y)};
+            }
         }
         if (configuring) {
             UpdateYuzuSettings();
         }
+        if (mouse_panning_timout++ > 20) {
+            StopPanning();
+        }
         std::this_thread::sleep_for(std::chrono::milliseconds(update_time));
     }
 }
@@ -65,8 +74,45 @@ void Mouse::PressButton(int x, int y, int button_) {
     mouse_info[button_index].data.pressed = true;
 }
 
-void Mouse::MouseMove(int x, int y) {
+void Mouse::StopPanning() {
     for (MouseInfo& info : mouse_info) {
+        if (Settings::values.mouse_panning) {
+            info.data.axis = {};
+            info.tilt_speed = 0;
+            info.last_mouse_change = {};
+        }
+    }
+}
+
+void Mouse::MouseMove(int x, int y, int center_x, int center_y) {
+    for (MouseInfo& info : mouse_info) {
+        if (Settings::values.mouse_panning) {
+            auto mouse_change =
+                (Common::MakeVec(x, y) - Common::MakeVec(center_x, center_y)).Cast<float>();
+            mouse_panning_timout = 0;
+
+            if (mouse_change.y == 0 && mouse_change.x == 0) {
+                continue;
+            }
+            const auto mouse_change_length = mouse_change.Length();
+            if (mouse_change_length < 3.0f) {
+                mouse_change /= mouse_change_length / 3.0f;
+            }
+
+            info.last_mouse_change = (info.last_mouse_change * 0.91f) + (mouse_change * 0.09f);
+
+            const auto last_mouse_change_length = info.last_mouse_change.Length();
+            if (last_mouse_change_length > 8.0f) {
+                info.last_mouse_change /= last_mouse_change_length / 8.0f;
+            } else if (last_mouse_change_length < 1.0f) {
+                info.last_mouse_change = mouse_change / mouse_change.Length();
+            }
+
+            info.tilt_direction = info.last_mouse_change;
+            info.tilt_speed = info.tilt_direction.Normalize() * info.sensitivity;
+            continue;
+        }
+
         if (info.data.pressed) {
             const auto mouse_move = Common::MakeVec(x, y) - info.mouse_origin;
             const auto mouse_change = Common::MakeVec(x, y) - info.last_mouse_position;

src/input_common/mouse/mouse_input.h

@@ -57,8 +57,10 @@ public:
      * Signals that mouse has moved.
      * @param x the x-coordinate of the cursor
      * @param y the y-coordinate of the cursor
+     * @param center_x the x-coordinate of the middle of the screen
+     * @param center_y the y-coordinate of the middle of the screen
      */
-    void MouseMove(int x, int y);
+    void MouseMove(int x, int y, int center_x, int center_y);
 
     /**
      * Signals that a motion sensor tilt has ended.
@@ -74,11 +76,13 @@ public:
 private:
     void UpdateThread();
     void UpdateYuzuSettings();
+    void StopPanning();
 
     struct MouseInfo {
         InputCommon::MotionInput motion{0.0f, 0.0f, 0.0f};
         Common::Vec2<int> mouse_origin;
         Common::Vec2<int> last_mouse_position;
+        Common::Vec2<float> last_mouse_change;
         bool is_tilting = false;
         float sensitivity{0.120f};
 
@@ -94,5 +98,6 @@ private:
     Common::SPSCQueue<MouseStatus> mouse_queue;
     bool configuring{false};
     bool update_thread_running{true};
+    int mouse_panning_timout{};
 };
 } // namespace MouseInput

src/input_common/mouse/mouse_poller.cpp

@@ -6,6 +6,7 @@
 #include <utility>
 
 #include "common/threadsafe_queue.h"
+#include "core/settings.h"
 #include "input_common/mouse/mouse_input.h"
 #include "input_common/mouse/mouse_poller.h"
 
@@ -71,7 +72,7 @@ public:
         std::lock_guard lock{mutex};
         const auto axis_value =
             static_cast<float>(mouse_input->GetMouseState(button).axis.at(axis));
-        return axis_value / (100.0f * range);
+        return axis_value * Settings::values.mouse_panning_sensitivity / (100.0f * range);
     }
 
     std::pair<float, float> GetAnalog(u32 analog_axis_x, u32 analog_axis_y) const {

src/input_common/sdl/sdl_impl.cpp

@@ -717,6 +717,13 @@ SDLState::SDLState() {
     if (SDL_SetHint(SDL_HINT_JOYSTICK_ALLOW_BACKGROUND_EVENTS, "1") == SDL_FALSE) {
         LOG_ERROR(Input, "Failed to set hint for background events with: {}", SDL_GetError());
     }
+// these hints are only defined on sdl2.0.9 or higher
+#if SDL_VERSION_ATLEAST(2, 0, 9)
+#if !SDL_VERSION_ATLEAST(2, 0, 12)
+    // There are also hints to toggle the individual drivers if needed.
+    SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI, "0");
+#endif
+#endif
 
     SDL_AddEventWatch(&SDLEventWatcher, this);
 

src/input_common/settings.h

@@ -340,6 +340,7 @@ enum class ControllerType {
     LeftJoycon,
     RightJoycon,
     Handheld,
+    GameCube,
 };
 
 struct PlayerInput {

src/tests/video_core/buffer_base.cpp

@@ -471,3 +471,79 @@ TEST_CASE("BufferBase: Unaligned page region query") {
     REQUIRE(buffer.IsRegionCpuModified(c + 4000, 1000));
     REQUIRE(buffer.IsRegionCpuModified(c + 4000, 1));
 }
+
+TEST_CASE("BufferBase: Cached write") {
+    RasterizerInterface rasterizer;
+    BufferBase buffer(rasterizer, c, WORD);
+    buffer.UnmarkRegionAsCpuModified(c, WORD);
+    buffer.CachedCpuWrite(c + PAGE, PAGE);
+    REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    buffer.FlushCachedWrites();
+    REQUIRE(buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    buffer.MarkRegionAsCpuModified(c, WORD);
+    REQUIRE(rasterizer.Count() == 0);
+}
+
+TEST_CASE("BufferBase: Multiple cached write") {
+    RasterizerInterface rasterizer;
+    BufferBase buffer(rasterizer, c, WORD);
+    buffer.UnmarkRegionAsCpuModified(c, WORD);
+    buffer.CachedCpuWrite(c + PAGE, PAGE);
+    buffer.CachedCpuWrite(c + PAGE * 3, PAGE);
+    REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    REQUIRE(!buffer.IsRegionCpuModified(c + PAGE * 3, PAGE));
+    buffer.FlushCachedWrites();
+    REQUIRE(buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    REQUIRE(buffer.IsRegionCpuModified(c + PAGE * 3, PAGE));
+    buffer.MarkRegionAsCpuModified(c, WORD);
+    REQUIRE(rasterizer.Count() == 0);
+}
+
+TEST_CASE("BufferBase: Cached write unmarked") {
+    RasterizerInterface rasterizer;
+    BufferBase buffer(rasterizer, c, WORD);
+    buffer.UnmarkRegionAsCpuModified(c, WORD);
+    buffer.CachedCpuWrite(c + PAGE, PAGE);
+    buffer.UnmarkRegionAsCpuModified(c + PAGE, PAGE);
+    REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    buffer.FlushCachedWrites();
+    REQUIRE(buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    buffer.MarkRegionAsCpuModified(c, WORD);
+    REQUIRE(rasterizer.Count() == 0);
+}
+
+TEST_CASE("BufferBase: Cached write iterated") {
+    RasterizerInterface rasterizer;
+    BufferBase buffer(rasterizer, c, WORD);
+    buffer.UnmarkRegionAsCpuModified(c, WORD);
+    buffer.CachedCpuWrite(c + PAGE, PAGE);
+    int num = 0;
+    buffer.ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
+    REQUIRE(num == 0);
+    REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    buffer.FlushCachedWrites();
+    REQUIRE(buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    buffer.MarkRegionAsCpuModified(c, WORD);
+    REQUIRE(rasterizer.Count() == 0);
+}
+
+TEST_CASE("BufferBase: Cached write downloads") {
+    RasterizerInterface rasterizer;
+    BufferBase buffer(rasterizer, c, WORD);
+    buffer.UnmarkRegionAsCpuModified(c, WORD);
+    REQUIRE(rasterizer.Count() == 64);
+    buffer.CachedCpuWrite(c + PAGE, PAGE);
+    REQUIRE(rasterizer.Count() == 63);
+    buffer.MarkRegionAsGpuModified(c + PAGE, PAGE);
+    int num = 0;
+    buffer.ForEachDownloadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
+    buffer.ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
+    REQUIRE(num == 0);
+    REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    REQUIRE(!buffer.IsRegionGpuModified(c + PAGE, PAGE));
+    buffer.FlushCachedWrites();
+    REQUIRE(buffer.IsRegionCpuModified(c + PAGE, PAGE));
+    REQUIRE(!buffer.IsRegionGpuModified(c + PAGE, PAGE));
+    buffer.MarkRegionAsCpuModified(c, WORD);
+    REQUIRE(rasterizer.Count() == 0);
+}

src/video_core/CMakeLists.txt

@@ -2,10 +2,8 @@ add_subdirectory(host_shaders)
 
 add_library(video_core STATIC
     buffer_cache/buffer_base.h
-    buffer_cache/buffer_block.h
+    buffer_cache/buffer_cache.cpp
     buffer_cache/buffer_cache.h
-    buffer_cache/map_interval.cpp
-    buffer_cache/map_interval.h
     cdma_pusher.cpp
     cdma_pusher.h
     command_classes/codecs/codec.cpp
@@ -152,8 +150,6 @@ add_library(video_core STATIC
     renderer_vulkan/vk_staging_buffer_pool.h
     renderer_vulkan/vk_state_tracker.cpp
     renderer_vulkan/vk_state_tracker.h
-    renderer_vulkan/vk_stream_buffer.cpp
-    renderer_vulkan/vk_stream_buffer.h
     renderer_vulkan/vk_swapchain.cpp
     renderer_vulkan/vk_swapchain.h
     renderer_vulkan/vk_texture_cache.cpp

src/video_core/buffer_cache/buffer_base.h

@@ -19,6 +19,7 @@ namespace VideoCommon {
 
 enum class BufferFlagBits {
     Picked = 1 << 0,
+    CachedWrites = 1 << 1,
 };
 DECLARE_ENUM_FLAG_OPERATORS(BufferFlagBits)
 
@@ -40,7 +41,7 @@ class BufferBase {
     static constexpr u64 BYTES_PER_WORD = PAGES_PER_WORD * BYTES_PER_PAGE;
 
     /// Vector tracking modified pages tightly packed with small vector optimization
-    union WrittenWords {
+    union WordsArray {
         /// Returns the pointer to the words state
         [[nodiscard]] const u64* Pointer(bool is_short) const noexcept {
             return is_short ? &stack : heap;
@@ -55,49 +56,59 @@ class BufferBase {
         u64* heap;     ///< Not-small buffers pointer to the storage
     };
 
-    struct GpuCpuWords {
-        explicit GpuCpuWords() = default;
-        explicit GpuCpuWords(u64 size_bytes_) : size_bytes{size_bytes_} {
+    struct Words {
+        explicit Words() = default;
+        explicit Words(u64 size_bytes_) : size_bytes{size_bytes_} {
             if (IsShort()) {
                 cpu.stack = ~u64{0};
                 gpu.stack = 0;
+                cached_cpu.stack = 0;
+                untracked.stack = ~u64{0};
             } else {
                 // Share allocation between CPU and GPU pages and set their default values
                 const size_t num_words = NumWords();
-                u64* const alloc = new u64[num_words * 2];
+                u64* const alloc = new u64[num_words * 4];
                 cpu.heap = alloc;
                 gpu.heap = alloc + num_words;
+                cached_cpu.heap = alloc + num_words * 2;
+                untracked.heap = alloc + num_words * 3;
                 std::fill_n(cpu.heap, num_words, ~u64{0});
                 std::fill_n(gpu.heap, num_words, 0);
+                std::fill_n(cached_cpu.heap, num_words, 0);
+                std::fill_n(untracked.heap, num_words, ~u64{0});
             }
             // Clean up tailing bits
-            const u64 last_local_page =
-                Common::DivCeil(size_bytes % BYTES_PER_WORD, BYTES_PER_PAGE);
+            const u64 last_word_size = size_bytes % BYTES_PER_WORD;
+            const u64 last_local_page = Common::DivCeil(last_word_size, BYTES_PER_PAGE);
             const u64 shift = (PAGES_PER_WORD - last_local_page) % PAGES_PER_WORD;
-            u64& last_word = cpu.Pointer(IsShort())[NumWords() - 1];
-            last_word = (last_word << shift) >> shift;
+            const u64 last_word = (~u64{0} << shift) >> shift;
+            cpu.Pointer(IsShort())[NumWords() - 1] = last_word;
+            untracked.Pointer(IsShort())[NumWords() - 1] = last_word;
         }
 
-        ~GpuCpuWords() {
+        ~Words() {
             Release();
         }
 
-        GpuCpuWords& operator=(GpuCpuWords&& rhs) noexcept {
+        Words& operator=(Words&& rhs) noexcept {
             Release();
             size_bytes = rhs.size_bytes;
             cpu = rhs.cpu;
             gpu = rhs.gpu;
+            cached_cpu = rhs.cached_cpu;
+            untracked = rhs.untracked;
             rhs.cpu.heap = nullptr;
             return *this;
         }
 
-        GpuCpuWords(GpuCpuWords&& rhs) noexcept
-            : size_bytes{rhs.size_bytes}, cpu{rhs.cpu}, gpu{rhs.gpu} {
+        Words(Words&& rhs) noexcept
+            : size_bytes{rhs.size_bytes}, cpu{rhs.cpu}, gpu{rhs.gpu},
+              cached_cpu{rhs.cached_cpu}, untracked{rhs.untracked} {
             rhs.cpu.heap = nullptr;
         }
 
-        GpuCpuWords& operator=(const GpuCpuWords&) = delete;
-        GpuCpuWords(const GpuCpuWords&) = delete;
+        Words& operator=(const Words&) = delete;
+        Words(const Words&) = delete;
 
         /// Returns true when the buffer fits in the small vector optimization
         [[nodiscard]] bool IsShort() const noexcept {
@@ -118,8 +129,17 @@ class BufferBase {
         }
 
         u64 size_bytes = 0;
-        WrittenWords cpu;
-        WrittenWords gpu;
+        WordsArray cpu;
+        WordsArray gpu;
+        WordsArray cached_cpu;
+        WordsArray untracked;
+    };
+
+    enum class Type {
+        CPU,
+        GPU,
+        CachedCPU,
+        Untracked,
     };
 
 public:
@@ -132,68 +152,93 @@ public:
     BufferBase& operator=(const BufferBase&) = delete;
     BufferBase(const BufferBase&) = delete;
 
+    BufferBase& operator=(BufferBase&&) = default;
+    BufferBase(BufferBase&&) = default;
+
     /// Returns the inclusive CPU modified range in a begin end pair
     [[nodiscard]] std::pair<u64, u64> ModifiedCpuRegion(VAddr query_cpu_addr,
                                                         u64 query_size) const noexcept {
         const u64 offset = query_cpu_addr - cpu_addr;
-        return ModifiedRegion<false>(offset, query_size);
+        return ModifiedRegion<Type::CPU>(offset, query_size);
     }
 
     /// Returns the inclusive GPU modified range in a begin end pair
     [[nodiscard]] std::pair<u64, u64> ModifiedGpuRegion(VAddr query_cpu_addr,
                                                         u64 query_size) const noexcept {
         const u64 offset = query_cpu_addr - cpu_addr;
-        return ModifiedRegion<true>(offset, query_size);
+        return ModifiedRegion<Type::GPU>(offset, query_size);
     }
 
     /// Returns true if a region has been modified from the CPU
     [[nodiscard]] bool IsRegionCpuModified(VAddr query_cpu_addr, u64 query_size) const noexcept {
         const u64 offset = query_cpu_addr - cpu_addr;
-        return IsRegionModified<false>(offset, query_size);
+        return IsRegionModified<Type::CPU>(offset, query_size);
     }
 
     /// Returns true if a region has been modified from the GPU
     [[nodiscard]] bool IsRegionGpuModified(VAddr query_cpu_addr, u64 query_size) const noexcept {
         const u64 offset = query_cpu_addr - cpu_addr;
-        return IsRegionModified<true>(offset, query_size);
+        return IsRegionModified<Type::GPU>(offset, query_size);
     }
 
     /// Mark region as CPU modified, notifying the rasterizer about this change
     void MarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 size) {
-        ChangeRegionState<true, true>(words.cpu, dirty_cpu_addr, size);
+        ChangeRegionState<Type::CPU, true>(dirty_cpu_addr, size);
     }
 
     /// Unmark region as CPU modified, notifying the rasterizer about this change
     void UnmarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 size) {
-        ChangeRegionState<false, true>(words.cpu, dirty_cpu_addr, size);
+        ChangeRegionState<Type::CPU, false>(dirty_cpu_addr, size);
     }
 
     /// Mark region as modified from the host GPU
     void MarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 size) noexcept {
-        ChangeRegionState<true, false>(words.gpu, dirty_cpu_addr, size);
+        ChangeRegionState<Type::GPU, true>(dirty_cpu_addr, size);
     }
 
     /// Unmark region as modified from the host GPU
     void UnmarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 size) noexcept {
-        ChangeRegionState<false, false>(words.gpu, dirty_cpu_addr, size);
+        ChangeRegionState<Type::GPU, false>(dirty_cpu_addr, size);
+    }
+
+    /// Mark region as modified from the CPU
+    /// but don't mark it as modified until FlusHCachedWrites is called.
+    void CachedCpuWrite(VAddr dirty_cpu_addr, u64 size) {
+        flags |= BufferFlagBits::CachedWrites;
+        ChangeRegionState<Type::CachedCPU, true>(dirty_cpu_addr, size);
+    }
+
+    /// Flushes cached CPU writes, and notify the rasterizer about the deltas
+    void FlushCachedWrites() noexcept {
+        flags &= ~BufferFlagBits::CachedWrites;
+        const u64 num_words = NumWords();
+        const u64* const cached_words = Array<Type::CachedCPU>();
+        u64* const untracked_words = Array<Type::Untracked>();
+        u64* const cpu_words = Array<Type::CPU>();
+        for (u64 word_index = 0; word_index < num_words; ++word_index) {
+            const u64 cached_bits = cached_words[word_index];
+            NotifyRasterizer<false>(word_index, untracked_words[word_index], cached_bits);
+            untracked_words[word_index] |= cached_bits;
+            cpu_words[word_index] |= cached_bits;
+        }
     }
 
     /// Call 'func' for each CPU modified range and unmark those pages as CPU modified
     template <typename Func>
     void ForEachUploadRange(VAddr query_cpu_range, u64 size, Func&& func) {
-        ForEachModifiedRange<false, true>(query_cpu_range, size, func);
+        ForEachModifiedRange<Type::CPU>(query_cpu_range, size, func);
     }
 
     /// Call 'func' for each GPU modified range and unmark those pages as GPU modified
     template <typename Func>
     void ForEachDownloadRange(VAddr query_cpu_range, u64 size, Func&& func) {
-        ForEachModifiedRange<true, false>(query_cpu_range, size, func);
+        ForEachModifiedRange<Type::GPU>(query_cpu_range, size, func);
     }
 
     /// Call 'func' for each GPU modified range and unmark those pages as GPU modified
     template <typename Func>
     void ForEachDownloadRange(Func&& func) {
-        ForEachModifiedRange<true, false>(cpu_addr, SizeBytes(), func);
+        ForEachModifiedRange<Type::GPU>(cpu_addr, SizeBytes(), func);
     }
 
     /// Mark buffer as picked
@@ -206,6 +251,16 @@ public:
         flags &= ~BufferFlagBits::Picked;
     }
 
+    /// Increases the likeliness of this being a stream buffer
+    void IncreaseStreamScore(int score) noexcept {
+        stream_score += score;
+    }
+
+    /// Returns the likeliness of this being a stream buffer
+    [[nodiscard]] int StreamScore() const noexcept {
+        return stream_score;
+    }
+
     /// Returns true when vaddr -> vaddr+size is fully contained in the buffer
     [[nodiscard]] bool IsInBounds(VAddr addr, u64 size) const noexcept {
         return addr >= cpu_addr && addr + size <= cpu_addr + SizeBytes();
@@ -216,6 +271,11 @@ public:
         return True(flags & BufferFlagBits::Picked);
     }
 
+    /// Returns true when the buffer has pending cached writes
+    [[nodiscard]] bool HasCachedWrites() const noexcept {
+        return True(flags & BufferFlagBits::CachedWrites);
+    }
+
     /// Returns the base CPU address of the buffer
     [[nodiscard]] VAddr CpuAddr() const noexcept {
         return cpu_addr;
@@ -233,26 +293,48 @@ public:
     }
 
 private:
+    template <Type type>
+    u64* Array() noexcept {
+        if constexpr (type == Type::CPU) {
+            return words.cpu.Pointer(IsShort());
+        } else if constexpr (type == Type::GPU) {
+            return words.gpu.Pointer(IsShort());
+        } else if constexpr (type == Type::CachedCPU) {
+            return words.cached_cpu.Pointer(IsShort());
+        } else if constexpr (type == Type::Untracked) {
+            return words.untracked.Pointer(IsShort());
+        }
+    }
+
+    template <Type type>
+    const u64* Array() const noexcept {
+        if constexpr (type == Type::CPU) {
+            return words.cpu.Pointer(IsShort());
+        } else if constexpr (type == Type::GPU) {
+            return words.gpu.Pointer(IsShort());
+        } else if constexpr (type == Type::CachedCPU) {
+            return words.cached_cpu.Pointer(IsShort());
+        } else if constexpr (type == Type::Untracked) {
+            return words.untracked.Pointer(IsShort());
+        }
+    }
+
     /**
      * Change the state of a range of pages
      *
-     * @param written_words Pages to be marked or unmarked as modified
      * @param dirty_addr    Base address to mark or unmark as modified
      * @param size          Size in bytes to mark or unmark as modified
-     *
-     * @tparam enable            True when the bits will be set to one, false for zero
-     * @tparam notify_rasterizer True when the rasterizer has to be notified about the changes
      */
-    template <bool enable, bool notify_rasterizer>
-    void ChangeRegionState(WrittenWords& written_words, u64 dirty_addr,
-                           s64 size) noexcept(!notify_rasterizer) {
+    template <Type type, bool enable>
+    void ChangeRegionState(u64 dirty_addr, s64 size) noexcept(type == Type::GPU) {
         const s64 difference = dirty_addr - cpu_addr;
         const u64 offset = std::max<s64>(difference, 0);
         size += std::min<s64>(difference, 0);
         if (offset >= SizeBytes() || size < 0) {
             return;
         }
-        u64* const state_words = written_words.Pointer(IsShort());
+        u64* const untracked_words = Array<Type::Untracked>();
+        u64* const state_words = Array<type>();
         const u64 offset_end = std::min(offset + size, SizeBytes());
         const u64 begin_page_index = offset / BYTES_PER_PAGE;
         const u64 begin_word_index = begin_page_index / PAGES_PER_WORD;
@@ -268,13 +350,19 @@ private:
             u64 bits = ~u64{0};
             bits = (bits >> right_offset) << right_offset;
             bits = (bits << left_offset) >> left_offset;
-            if constexpr (notify_rasterizer) {
-                NotifyRasterizer<!enable>(word_index, state_words[word_index], bits);
+            if constexpr (type == Type::CPU || type == Type::CachedCPU) {
+                NotifyRasterizer<!enable>(word_index, untracked_words[word_index], bits);
             }
             if constexpr (enable) {
                 state_words[word_index] |= bits;
+                if constexpr (type == Type::CPU || type == Type::CachedCPU) {
+                    untracked_words[word_index] |= bits;
+                }
             } else {
                 state_words[word_index] &= ~bits;
+                if constexpr (type == Type::CPU || type == Type::CachedCPU) {
+                    untracked_words[word_index] &= ~bits;
+                }
             }
             page_index = 0;
             ++word_index;
@@ -291,7 +379,7 @@ private:
      * @tparam add_to_rasterizer True when the rasterizer should start tracking the new pages
      */
     template <bool add_to_rasterizer>
-    void NotifyRasterizer(u64 word_index, u64 current_bits, u64 new_bits) {
+    void NotifyRasterizer(u64 word_index, u64 current_bits, u64 new_bits) const {
         u64 changed_bits = (add_to_rasterizer ? current_bits : ~current_bits) & new_bits;
         VAddr addr = cpu_addr + word_index * BYTES_PER_WORD;
         while (changed_bits != 0) {
@@ -315,21 +403,20 @@ private:
      * @param query_cpu_range Base CPU address to loop over
      * @param size            Size in bytes of the CPU range to loop over
      * @param func            Function to call for each turned off region
-     *
-     * @tparam gpu               True for host GPU pages, false for CPU pages
-     * @tparam notify_rasterizer True when the rasterizer should be notified about state changes
      */
-    template <bool gpu, bool notify_rasterizer, typename Func>
+    template <Type type, typename Func>
     void ForEachModifiedRange(VAddr query_cpu_range, s64 size, Func&& func) {
+        static_assert(type != Type::Untracked);
+
         const s64 difference = query_cpu_range - cpu_addr;
         const u64 query_begin = std::max<s64>(difference, 0);
         size += std::min<s64>(difference, 0);
         if (query_begin >= SizeBytes() || size < 0) {
             return;
         }
-        const u64* const cpu_words = words.cpu.Pointer(IsShort());
+        u64* const untracked_words = Array<Type::Untracked>();
+        u64* const state_words = Array<type>();
         const u64 query_end = query_begin + std::min(static_cast<u64>(size), SizeBytes());
-        u64* const state_words = (gpu ? words.gpu : words.cpu).Pointer(IsShort());
         u64* const words_begin = state_words + query_begin / BYTES_PER_WORD;
         u64* const words_end = state_words + Common::DivCeil(query_end, BYTES_PER_WORD);
 
@@ -345,7 +432,8 @@ private:
         const u64 word_index_end = std::distance(state_words, last_modified_word);
 
         const unsigned local_page_begin = std::countr_zero(*first_modified_word);
-        const unsigned local_page_end = PAGES_PER_WORD - std::countl_zero(last_modified_word[-1]);
+        const unsigned local_page_end =
+            static_cast<unsigned>(PAGES_PER_WORD) - std::countl_zero(last_modified_word[-1]);
         const u64 word_page_begin = word_index_begin * PAGES_PER_WORD;
         const u64 word_page_end = (word_index_end - 1) * PAGES_PER_WORD;
         const u64 query_page_begin = query_begin / BYTES_PER_PAGE;
@@ -371,11 +459,13 @@ private:
             const u64 current_word = state_words[word_index] & bits;
             state_words[word_index] &= ~bits;
 
-            // Exclude CPU modified pages when visiting GPU pages
-            const u64 word = current_word & ~(gpu ? cpu_words[word_index] : 0);
-            if constexpr (notify_rasterizer) {
-                NotifyRasterizer<true>(word_index, word, ~u64{0});
+            if constexpr (type == Type::CPU) {
+                const u64 current_bits = untracked_words[word_index] & bits;
+                untracked_words[word_index] &= ~bits;
+                NotifyRasterizer<true>(word_index, current_bits, ~u64{0});
             }
+            // Exclude CPU modified pages when visiting GPU pages
+            const u64 word = current_word & ~(type == Type::GPU ? untracked_words[word_index] : 0);
             u64 page = page_begin;
             page_begin = 0;
 
@@ -416,17 +506,20 @@ private:
      * @param offset Offset in bytes from the start of the buffer
      * @param size   Size in bytes of the region to query for modifications
      */
-    template <bool gpu>
+    template <Type type>
     [[nodiscard]] bool IsRegionModified(u64 offset, u64 size) const noexcept {
-        const u64* const cpu_words = words.cpu.Pointer(IsShort());
-        const u64* const state_words = (gpu ? words.gpu : words.cpu).Pointer(IsShort());
+        static_assert(type != Type::Untracked);
+
+        const u64* const untracked_words = Array<Type::Untracked>();
+        const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
         const u64 word_end = std::min(word_begin + num_query_words, NumWords());
         const u64 page_limit = Common::DivCeil(offset + size, BYTES_PER_PAGE);
         u64 page_index = (offset / BYTES_PER_PAGE) % PAGES_PER_WORD;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index, page_index = 0) {
-            const u64 word = state_words[word_index] & ~(gpu ? cpu_words[word_index] : 0);
+            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
+            const u64 word = state_words[word_index] & ~off_word;
             if (word == 0) {
                 continue;
             }
@@ -445,13 +538,13 @@ private:
      *
      * @param offset Offset in bytes from the start of the buffer
      * @param size   Size in bytes of the region to query for modifications
-     *
-     * @tparam gpu True to query GPU modified pages, false for CPU pages
      */
-    template <bool gpu>
+    template <Type type>
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
-        const u64* const cpu_words = words.cpu.Pointer(IsShort());
-        const u64* const state_words = (gpu ? words.gpu : words.cpu).Pointer(IsShort());
+        static_assert(type != Type::Untracked);
+
+        const u64* const untracked_words = Array<Type::Untracked>();
+        const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
         const u64 word_end = std::min(word_begin + num_query_words, NumWords());
@@ -460,7 +553,8 @@ private:
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index) {
-            const u64 word = state_words[word_index] & ~(gpu ? cpu_words[word_index] : 0);
+            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
+            const u64 word = state_words[word_index] & ~off_word;
             if (word == 0) {
                 continue;
             }
@@ -488,8 +582,9 @@ private:
 
     RasterizerInterface* rasterizer = nullptr;
     VAddr cpu_addr = 0;
-    GpuCpuWords words;
+    Words words;
     BufferFlagBits flags{};
+    int stream_score = 0;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/buffer_block.h

@@ -1,62 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "common/common_types.h"
-
-namespace VideoCommon {
-
-class BufferBlock {
-public:
-    [[nodiscard]] bool Overlaps(VAddr start, VAddr end) const {
-        return (cpu_addr < end) && (cpu_addr_end > start);
-    }
-
-    [[nodiscard]] bool IsInside(VAddr other_start, VAddr other_end) const {
-        return cpu_addr <= other_start && other_end <= cpu_addr_end;
-    }
-
-    [[nodiscard]] std::size_t Offset(VAddr in_addr) const {
-        return static_cast<std::size_t>(in_addr - cpu_addr);
-    }
-
-    [[nodiscard]] VAddr CpuAddr() const {
-        return cpu_addr;
-    }
-
-    [[nodiscard]] VAddr CpuAddrEnd() const {
-        return cpu_addr_end;
-    }
-
-    void SetCpuAddr(VAddr new_addr) {
-        cpu_addr = new_addr;
-        cpu_addr_end = new_addr + size;
-    }
-
-    [[nodiscard]] std::size_t Size() const {
-        return size;
-    }
-
-    [[nodiscard]] u64 Epoch() const {
-        return epoch;
-    }
-
-    void SetEpoch(u64 new_epoch) {
-        epoch = new_epoch;
-    }
-
-protected:
-    explicit BufferBlock(VAddr cpu_addr_, std::size_t size_) : size{size_} {
-        SetCpuAddr(cpu_addr_);
-    }
-
-private:
-    VAddr cpu_addr{};
-    VAddr cpu_addr_end{};
-    std::size_t size{};
-    u64 epoch{};
-};
-
-} // namespace VideoCommon

src/video_core/buffer_cache/buffer_cache.cpp

@@ -0,0 +1,13 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/microprofile.h"
+
+namespace VideoCommon {
+
+MICROPROFILE_DEFINE(GPU_PrepareBuffers, "GPU", "Prepare buffers", MP_RGB(224, 128, 128));
+MICROPROFILE_DEFINE(GPU_BindUploadBuffers, "GPU", "Bind and upload buffers", MP_RGB(224, 128, 128));
+MICROPROFILE_DEFINE(GPU_DownloadMemory, "GPU", "Download buffers", MP_RGB(224, 128, 128));
+
+} // namespace VideoCommon

src/video_core/buffer_cache/map_interval.cpp

@@ -1,33 +0,0 @@
-// Copyright 2020 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <array>
-#include <cstddef>
-#include <memory>
-
-#include "video_core/buffer_cache/map_interval.h"
-
-namespace VideoCommon {
-
-MapIntervalAllocator::MapIntervalAllocator() {
-    FillFreeList(first_chunk);
-}
-
-MapIntervalAllocator::~MapIntervalAllocator() = default;
-
-void MapIntervalAllocator::AllocateNewChunk() {
-    *new_chunk = std::make_unique<Chunk>();
-    FillFreeList(**new_chunk);
-    new_chunk = &(*new_chunk)->next;
-}
-
-void MapIntervalAllocator::FillFreeList(Chunk& chunk) {
-    const std::size_t old_size = free_list.size();
-    free_list.resize(old_size + chunk.data.size());
-    std::transform(chunk.data.rbegin(), chunk.data.rend(), free_list.begin() + old_size,
-                   [](MapInterval& interval) { return &interval; });
-}
-
-} // namespace VideoCommon

src/video_core/buffer_cache/map_interval.h

@@ -1,93 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <cstddef>
-#include <memory>
-#include <vector>
-
-#include <boost/intrusive/set_hook.hpp>
-
-#include "common/common_types.h"
-#include "video_core/gpu.h"
-
-namespace VideoCommon {
-
-struct MapInterval : public boost::intrusive::set_base_hook<boost::intrusive::optimize_size<true>> {
-    MapInterval() = default;
-
-    /*implicit*/ MapInterval(VAddr start_) noexcept : start{start_} {}
-
-    explicit MapInterval(VAddr start_, VAddr end_, GPUVAddr gpu_addr_) noexcept
-        : start{start_}, end{end_}, gpu_addr{gpu_addr_} {}
-
-    bool IsInside(VAddr other_start, VAddr other_end) const noexcept {
-        return start <= other_start && other_end <= end;
-    }
-
-    bool Overlaps(VAddr other_start, VAddr other_end) const noexcept {
-        return start < other_end && other_start < end;
-    }
-
-    void MarkAsModified(bool is_modified_, u64 ticks_) noexcept {
-        is_modified = is_modified_;
-        ticks = ticks_;
-    }
-
-    boost::intrusive::set_member_hook<> member_hook_;
-    VAddr start = 0;
-    VAddr end = 0;
-    GPUVAddr gpu_addr = 0;
-    u64 ticks = 0;
-    bool is_written = false;
-    bool is_modified = false;
-    bool is_registered = false;
-    bool is_memory_marked = false;
-    bool is_sync_pending = false;
-};
-
-struct MapIntervalCompare {
-    constexpr bool operator()(const MapInterval& lhs, const MapInterval& rhs) const noexcept {
-        return lhs.start < rhs.start;
-    }
-};
-
-class MapIntervalAllocator {
-public:
-    MapIntervalAllocator();
-    ~MapIntervalAllocator();
-
-    MapInterval* Allocate() {
-        if (free_list.empty()) {
-            AllocateNewChunk();
-        }
-        MapInterval* const interval = free_list.back();
-        free_list.pop_back();
-        return interval;
-    }
-
-    void Release(MapInterval* interval) {
-        free_list.push_back(interval);
-    }
-
-private:
-    struct Chunk {
-        std::unique_ptr<Chunk> next;
-        std::array<MapInterval, 0x8000> data;
-    };
-
-    void AllocateNewChunk();
-
-    void FillFreeList(Chunk& chunk);
-
-    std::vector<MapInterval*> free_list;
-
-    Chunk first_chunk;
-
-    std::unique_ptr<Chunk>* new_chunk = &first_chunk.next;
-};
-
-} // namespace VideoCommon

src/video_core/command_classes/vic.cpp

@@ -110,12 +110,10 @@ void Vic::Execute() {
                                            converted_frame_buffer.get(), block_height, 0, 0);
 
             gpu.MemoryManager().WriteBlock(output_surface_luma_address, swizzled_data.data(), size);
-            gpu.Maxwell3D().OnMemoryWrite();
         } else {
             // send pitch linear frame
             gpu.MemoryManager().WriteBlock(output_surface_luma_address, converted_frame_buf_addr,
                                            linear_size);
-            gpu.Maxwell3D().OnMemoryWrite();
         }
         break;
     }
@@ -163,7 +161,6 @@ void Vic::Execute() {
         }
         gpu.MemoryManager().WriteBlock(output_surface_chroma_u_address, chroma_buffer.data(),
                                        chroma_buffer.size());
-        gpu.Maxwell3D().OnMemoryWrite();
         break;
     }
     default:

src/video_core/dirty_flags.cpp

@@ -12,13 +12,30 @@
 #define NUM(field_name) (sizeof(::Tegra::Engines::Maxwell3D::Regs::field_name) / (sizeof(u32)))
 
 namespace VideoCommon::Dirty {
-
+namespace {
 using Tegra::Engines::Maxwell3D;
 
-void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables) {
+void SetupDirtyVertexBuffers(Maxwell3D::DirtyState::Tables& tables) {
+    static constexpr std::size_t num_array = 3;
+    for (std::size_t i = 0; i < Maxwell3D::Regs::NumVertexArrays; ++i) {
+        const std::size_t array_offset = OFF(vertex_array) + i * NUM(vertex_array[0]);
+        const std::size_t limit_offset = OFF(vertex_array_limit) + i * NUM(vertex_array_limit[0]);
+
+        FillBlock(tables, array_offset, num_array, VertexBuffer0 + i, VertexBuffers);
+        FillBlock(tables, limit_offset, NUM(vertex_array_limit), VertexBuffer0 + i, VertexBuffers);
+    }
+}
+
+void SetupIndexBuffer(Maxwell3D::DirtyState::Tables& tables) {
+    FillBlock(tables[0], OFF(index_array), NUM(index_array), IndexBuffer);
+}
+
+void SetupDirtyDescriptors(Maxwell3D::DirtyState::Tables& tables) {
     FillBlock(tables[0], OFF(tic), NUM(tic), Descriptors);
     FillBlock(tables[0], OFF(tsc), NUM(tsc), Descriptors);
+}
 
+void SetupDirtyRenderTargets(Maxwell3D::DirtyState::Tables& tables) {
     static constexpr std::size_t num_per_rt = NUM(rt[0]);
     static constexpr std::size_t begin = OFF(rt);
     static constexpr std::size_t num = num_per_rt * Maxwell3D::Regs::NumRenderTargets;
@@ -41,5 +58,13 @@ void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tabl
         FillBlock(table, OFF(zeta), NUM(zeta), flag);
     }
 }
+} // Anonymous namespace
+
+void SetupDirtyFlags(Maxwell3D::DirtyState::Tables& tables) {
+    SetupDirtyVertexBuffers(tables);
+    SetupIndexBuffer(tables);
+    SetupDirtyDescriptors(tables);
+    SetupDirtyRenderTargets(tables);
+}
 
 } // namespace VideoCommon::Dirty

src/video_core/dirty_flags.h

@@ -30,6 +30,12 @@ enum : u8 {
     ColorBuffer7,
     ZetaBuffer,
 
+    VertexBuffers,
+    VertexBuffer0,
+    VertexBuffer31 = VertexBuffer0 + 31,
+
+    IndexBuffer,
+
     LastCommonEntry,
 };
 
@@ -47,6 +53,6 @@ void FillBlock(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables, std::size_
     FillBlock(tables[1], begin, num, index_b);
 }
 
-void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables);
+void SetupDirtyFlags(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables);
 
 } // namespace VideoCommon::Dirty

src/video_core/dma_pusher.cpp

@@ -23,8 +23,6 @@ void DmaPusher::DispatchCalls() {
     MICROPROFILE_SCOPE(DispatchCalls);
 
     gpu.SyncGuestHost();
-    // On entering GPU code, assume all memory may be touched by the ARM core.
-    gpu.Maxwell3D().OnMemoryWrite();
 
     dma_pushbuffer_subindex = 0;
 

src/video_core/engines/fermi_2d.cpp

@@ -18,8 +18,8 @@ Fermi2D::Fermi2D() {
 
 Fermi2D::~Fermi2D() = default;
 
-void Fermi2D::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) {
-    rasterizer = &rasterizer_;
+void Fermi2D::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
+    rasterizer = rasterizer_;
 }
 
 void Fermi2D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {

src/video_core/engines/fermi_2d.h

@@ -38,7 +38,7 @@ public:
     ~Fermi2D();
 
     /// Binds a rasterizer to this engine.
-    void BindRasterizer(VideoCore::RasterizerInterface& rasterizer);
+    void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
 
     /// Write the value to the register identified by method.
     void CallMethod(u32 method, u32 method_argument, bool is_last_call) override;

src/video_core/engines/kepler_compute.cpp

@@ -21,8 +21,8 @@ KeplerCompute::KeplerCompute(Core::System& system_, MemoryManager& memory_manage
 
 KeplerCompute::~KeplerCompute() = default;
 
-void KeplerCompute::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) {
-    rasterizer = &rasterizer_;
+void KeplerCompute::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
+    rasterizer = rasterizer_;
 }
 
 void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
@@ -39,7 +39,6 @@ void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_cal
     case KEPLER_COMPUTE_REG_INDEX(data_upload): {
         upload_state.ProcessData(method_argument, is_last_call);
         if (is_last_call) {
-            system.GPU().Maxwell3D().OnMemoryWrite();
         }
         break;
     }

src/video_core/engines/kepler_compute.h

@@ -46,7 +46,7 @@ public:
     ~KeplerCompute();
 
     /// Binds a rasterizer to this engine.
-    void BindRasterizer(VideoCore::RasterizerInterface& rasterizer);
+    void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
 
     static constexpr std::size_t NumConstBuffers = 8;
 

src/video_core/engines/kepler_memory.cpp

@@ -33,7 +33,6 @@ void KeplerMemory::CallMethod(u32 method, u32 method_argument, bool is_last_call
     case KEPLERMEMORY_REG_INDEX(data): {
         upload_state.ProcessData(method_argument, is_last_call);
         if (is_last_call) {
-            system.GPU().Maxwell3D().OnMemoryWrite();
         }
         break;
     }

src/video_core/engines/maxwell_3d.cpp

@@ -30,8 +30,8 @@ Maxwell3D::Maxwell3D(Core::System& system_, MemoryManager& memory_manager_)
 
 Maxwell3D::~Maxwell3D() = default;
 
-void Maxwell3D::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) {
-    rasterizer = &rasterizer_;
+void Maxwell3D::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
+    rasterizer = rasterizer_;
 }
 
 void Maxwell3D::InitializeRegisterDefaults() {
@@ -223,7 +223,6 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume
     case MAXWELL3D_REG_INDEX(data_upload):
         upload_state.ProcessData(argument, is_last_call);
         if (is_last_call) {
-            OnMemoryWrite();
         }
         return;
     case MAXWELL3D_REG_INDEX(fragment_barrier):
@@ -570,17 +569,18 @@ std::optional<u64> Maxwell3D::GetQueryResult() {
     }
 }
 
-void Maxwell3D::ProcessCBBind(std::size_t stage_index) {
+void Maxwell3D::ProcessCBBind(size_t stage_index) {
     // Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader stage.
-    auto& shader = state.shader_stages[stage_index];
-    auto& bind_data = regs.cb_bind[stage_index];
-
-    ASSERT(bind_data.index < Regs::MaxConstBuffers);
-    auto& buffer = shader.const_buffers[bind_data.index];
-
+    const auto& bind_data = regs.cb_bind[stage_index];
+    auto& buffer = state.shader_stages[stage_index].const_buffers[bind_data.index];
     buffer.enabled = bind_data.valid.Value() != 0;
     buffer.address = regs.const_buffer.BufferAddress();
     buffer.size = regs.const_buffer.cb_size;
+
+    const bool is_enabled = bind_data.valid.Value() != 0;
+    const GPUVAddr gpu_addr = is_enabled ? regs.const_buffer.BufferAddress() : 0;
+    const u32 size = is_enabled ? regs.const_buffer.cb_size : 0;
+    rasterizer->BindGraphicsUniformBuffer(stage_index, bind_data.index, gpu_addr, size);
 }
 
 void Maxwell3D::ProcessCBData(u32 value) {
@@ -635,7 +635,6 @@ void Maxwell3D::FinishCBData() {
 
     const u32 id = cb_data_state.id;
     memory_manager.WriteBlock(address, cb_data_state.buffer[id].data(), size);
-    OnMemoryWrite();
 
     cb_data_state.id = null_cb_data;
     cb_data_state.current = null_cb_data;

src/video_core/engines/maxwell_3d.h

@@ -55,7 +55,7 @@ public:
     ~Maxwell3D();
 
     /// Binds a rasterizer to this engine.
-    void BindRasterizer(VideoCore::RasterizerInterface& rasterizer);
+    void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
 
     /// Register structure of the Maxwell3D engine.
     /// TODO(Subv): This structure will need to be made bigger as more registers are discovered.
@@ -1314,8 +1314,7 @@ public:
 
                     GPUVAddr LimitAddress() const {
                         return static_cast<GPUVAddr>((static_cast<GPUVAddr>(limit_high) << 32) |
-                                                     limit_low) +
-                               1;
+                                                     limit_low);
                     }
                 } vertex_array_limit[NumVertexArrays];
 
@@ -1403,6 +1402,7 @@ public:
         };
 
         std::array<ShaderStageInfo, Regs::MaxShaderStage> shader_stages;
+
         u32 current_instance = 0; ///< Current instance to be used to simulate instanced rendering.
     };
 
@@ -1452,11 +1452,6 @@ public:
         return *rasterizer;
     }
 
-    /// Notify a memory write has happened.
-    void OnMemoryWrite() {
-        dirty.flags |= dirty.on_write_stores;
-    }
-
     enum class MMEDrawMode : u32 {
         Undefined,
         Array,
@@ -1478,7 +1473,6 @@ public:
         using Tables = std::array<Table, 2>;
 
         Flags flags;
-        Flags on_write_stores;
         Tables tables{};
     } dirty;
 
@@ -1541,7 +1535,7 @@ private:
     void FinishCBData();
 
     /// Handles a write to the CB_BIND register.
-    void ProcessCBBind(std::size_t stage_index);
+    void ProcessCBBind(size_t stage_index);
 
     /// Handles a write to the VERTEX_END_GL register, triggering a draw.
     void DrawArrays();

src/video_core/engines/maxwell_dma.cpp

@@ -60,9 +60,6 @@ void MaxwellDMA::Launch() {
         return;
     }
 
-    // All copies here update the main memory, so mark all rasterizer states as invalid.
-    system.GPU().Maxwell3D().OnMemoryWrite();
-
     if (is_src_pitch && is_dst_pitch) {
         CopyPitchToPitch();
     } else {

src/video_core/fence_manager.h

@@ -143,22 +143,26 @@ private:
     }
 
     bool ShouldWait() const {
+        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
         return texture_cache.ShouldWaitAsyncFlushes() || buffer_cache.ShouldWaitAsyncFlushes() ||
                query_cache.ShouldWaitAsyncFlushes();
     }
 
     bool ShouldFlush() const {
+        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
         return texture_cache.HasUncommittedFlushes() || buffer_cache.HasUncommittedFlushes() ||
                query_cache.HasUncommittedFlushes();
     }
 
     void PopAsyncFlushes() {
+        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
         texture_cache.PopAsyncFlushes();
         buffer_cache.PopAsyncFlushes();
         query_cache.PopAsyncFlushes();
     }
 
     void CommitAsyncFlushes() {
+        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
         texture_cache.CommitAsyncFlushes();
         buffer_cache.CommitAsyncFlushes();
         query_cache.CommitAsyncFlushes();

src/video_core/gpu.cpp

@@ -44,8 +44,8 @@ GPU::~GPU() = default;
 
 void GPU::BindRenderer(std::unique_ptr<VideoCore::RendererBase> renderer_) {
     renderer = std::move(renderer_);
+    rasterizer = renderer->ReadRasterizer();
 
-    VideoCore::RasterizerInterface& rasterizer = renderer->Rasterizer();
     memory_manager->BindRasterizer(rasterizer);
     maxwell_3d->BindRasterizer(rasterizer);
     fermi_2d->BindRasterizer(rasterizer);
@@ -171,7 +171,7 @@ void GPU::TickWork() {
         const std::size_t size = request.size;
         flush_requests.pop_front();
         flush_request_mutex.unlock();
-        renderer->Rasterizer().FlushRegion(addr, size);
+        rasterizer->FlushRegion(addr, size);
         current_flush_fence.store(fence);
         flush_request_mutex.lock();
     }
@@ -193,11 +193,11 @@ u64 GPU::GetTicks() const {
 }
 
 void GPU::FlushCommands() {
-    renderer->Rasterizer().FlushCommands();
+    rasterizer->FlushCommands();
 }
 
 void GPU::SyncGuestHost() {
-    renderer->Rasterizer().SyncGuestHost();
+    rasterizer->SyncGuestHost();
 }
 
 enum class GpuSemaphoreOperation {

src/video_core/gpu.h

@@ -366,6 +366,7 @@ protected:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
     std::unique_ptr<Tegra::CDmaPusher> cdma_pusher;
     std::unique_ptr<VideoCore::RendererBase> renderer;
+    VideoCore::RasterizerInterface* rasterizer = nullptr;
     const bool use_nvdec;
 
 private:

src/video_core/gpu_thread.cpp

@@ -38,6 +38,7 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
     }
 
     auto current_context = context.Acquire();
+    VideoCore::RasterizerInterface* const rasterizer = renderer.ReadRasterizer();
 
     CommandDataContainer next;
     while (state.is_running) {
@@ -52,13 +53,13 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
         } else if (const auto* data = std::get_if<SwapBuffersCommand>(&next.data)) {
             renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
         } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) {
-            renderer.Rasterizer().ReleaseFences();
+            rasterizer->ReleaseFences();
         } else if (std::holds_alternative<GPUTickCommand>(next.data)) {
             system.GPU().TickWork();
         } else if (const auto* flush = std::get_if<FlushRegionCommand>(&next.data)) {
-            renderer.Rasterizer().FlushRegion(flush->addr, flush->size);
+            rasterizer->FlushRegion(flush->addr, flush->size);
         } else if (const auto* invalidate = std::get_if<InvalidateRegionCommand>(&next.data)) {
-            renderer.Rasterizer().OnCPUWrite(invalidate->addr, invalidate->size);
+            rasterizer->OnCPUWrite(invalidate->addr, invalidate->size);
         } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
             return;
         } else {
@@ -84,6 +85,7 @@ ThreadManager::~ThreadManager() {
 void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
                                 Core::Frontend::GraphicsContext& context,
                                 Tegra::DmaPusher& dma_pusher, Tegra::CDmaPusher& cdma_pusher) {
+    rasterizer = renderer.ReadRasterizer();
     thread = std::thread(RunThread, std::ref(system), std::ref(renderer), std::ref(context),
                          std::ref(dma_pusher), std::ref(state), std::ref(cdma_pusher));
 }
@@ -129,12 +131,12 @@ void ThreadManager::FlushRegion(VAddr addr, u64 size) {
 }
 
 void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
-    system.Renderer().Rasterizer().OnCPUWrite(addr, size);
+    rasterizer->OnCPUWrite(addr, size);
 }
 
 void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
-    system.Renderer().Rasterizer().OnCPUWrite(addr, size);
+    rasterizer->OnCPUWrite(addr, size);
 }
 
 void ThreadManager::WaitIdle() const {

src/video_core/gpu_thread.h

@@ -27,6 +27,7 @@ class System;
 } // namespace Core
 
 namespace VideoCore {
+class RasterizerInterface;
 class RendererBase;
 } // namespace VideoCore
 
@@ -151,11 +152,12 @@ private:
     /// Pushes a command to be executed by the GPU thread
     u64 PushCommand(CommandData&& command_data);
 
-    SynchState state;
     Core::System& system;
-    std::thread thread;
-    std::thread::id thread_id;
     const bool is_async;
+    VideoCore::RasterizerInterface* rasterizer = nullptr;
+
+    SynchState state;
+    std::thread thread;
 };
 
 } // namespace VideoCommon::GPUThread

src/video_core/host_shaders/CMakeLists.txt

@@ -12,7 +12,6 @@ set(SHADER_FILES
     vulkan_blit_depth_stencil.frag
     vulkan_present.frag
     vulkan_present.vert
-    vulkan_quad_array.comp
     vulkan_quad_indexed.comp
     vulkan_uint8.comp
 )

src/video_core/host_shaders/vulkan_quad_array.comp

@@ -1,28 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#version 460 core
-
-layout (local_size_x = 1024) in;
-
-layout (std430, set = 0, binding = 0) buffer OutputBuffer {
-    uint output_indexes[];
-};
-
-layout (push_constant) uniform PushConstants {
-    uint first;
-};
-
-void main() {
-    uint primitive = gl_GlobalInvocationID.x;
-    if (primitive * 6 >= output_indexes.length()) {
-        return;
-    }
-
-    const uint quad_map[6] = uint[](0, 1, 2, 0, 2, 3);
-    for (uint vertex = 0; vertex < 6; ++vertex) {
-        uint index = first + primitive * 4 + quad_map[vertex];
-        output_indexes[primitive * 6 + vertex] = index;
-    }
-}

src/video_core/host_shaders/vulkan_uint8.comp

@@ -16,9 +16,16 @@ layout (std430, set = 0, binding = 1) writeonly buffer OutputBuffer {
     uint16_t output_indexes[];
 };
 
+uint AssembleIndex(uint id) {
+    // Most primitive restart indices are 0xFF
+    // Hardcode this to 0xFF for now
+    uint index = uint(input_indexes[id]);
+    return index == 0xFF ? 0xFFFF : index;
+}
+
 void main() {
     uint id = gl_GlobalInvocationID.x;
     if (id < input_indexes.length()) {
-        output_indexes[id] = uint16_t(input_indexes[id]);
+        output_indexes[id] = uint16_t(AssembleIndex(id));
     }
 }

src/video_core/memory_manager.cpp

@@ -21,8 +21,8 @@ MemoryManager::MemoryManager(Core::System& system_)
 
 MemoryManager::~MemoryManager() = default;
 
-void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) {
-    rasterizer = &rasterizer_;
+void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
+    rasterizer = rasterizer_;
 }
 
 GPUVAddr MemoryManager::UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size) {

src/video_core/memory_manager.h

@@ -72,7 +72,7 @@ public:
     ~MemoryManager();
 
     /// Binds a renderer to the memory manager.
-    void BindRasterizer(VideoCore::RasterizerInterface& rasterizer);
+    void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
 
     [[nodiscard]] std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr) const;
 
@@ -157,6 +157,8 @@ private:
 
     using MapRange = std::pair<GPUVAddr, size_t>;
     std::vector<MapRange> map_ranges;
+
+    std::vector<std::pair<VAddr, std::size_t>> cache_invalidate_queue;
 };
 
 } // namespace Tegra

src/video_core/rasterizer_interface.h

@@ -7,6 +7,7 @@
 #include <atomic>
 #include <functional>
 #include <optional>
+#include <span>
 #include "common/common_types.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/gpu.h"
@@ -49,6 +50,10 @@ public:
     /// Records a GPU query and caches it
     virtual void Query(GPUVAddr gpu_addr, QueryType type, std::optional<u64> timestamp) = 0;
 
+    /// Signal an uniform buffer binding
+    virtual void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
+                                           u32 size) = 0;
+
     /// Signal a GPU based semaphore as a fence
     virtual void SignalSemaphore(GPUVAddr addr, u32 value) = 0;
 

src/video_core/renderer_base.h

@@ -37,15 +37,11 @@ public:
                           std::unique_ptr<Core::Frontend::GraphicsContext> context);
     virtual ~RendererBase();
 
-    /// Initialize the renderer
-    [[nodiscard]] virtual bool Init() = 0;
-
-    /// Shutdown the renderer
-    virtual void ShutDown() = 0;
-
     /// Finalize rendering the guest frame and draw into the presentation texture
     virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
 
+    [[nodiscard]] virtual RasterizerInterface* ReadRasterizer() = 0;
+
     // Getter/setter functions:
     // ------------------------
 
@@ -57,14 +53,6 @@ public:
         return m_current_frame;
     }
 
-    [[nodiscard]] RasterizerInterface& Rasterizer() {
-        return *rasterizer;
-    }
-
-    [[nodiscard]] const RasterizerInterface& Rasterizer() const {
-        return *rasterizer;
-    }
-
     [[nodiscard]] Core::Frontend::GraphicsContext& Context() {
         return *context;
     }
@@ -98,7 +86,6 @@ public:
 
 protected:
     Core::Frontend::EmuWindow& render_window; ///< Reference to the render window handle.
-    std::unique_ptr<RasterizerInterface> rasterizer;
     std::unique_ptr<Core::Frontend::GraphicsContext> context;
     f32 m_current_fps = 0.0f; ///< Current framerate, should be set by the renderer
     int m_current_frame = 0;  ///< Current frame, should be set by the renderer

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -2,98 +2,208 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <memory>
+#include <span>
 
-#include <glad/glad.h>
-
-#include "common/assert.h"
-#include "common/microprofile.h"
 #include "video_core/buffer_cache/buffer_cache.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_device.h"
-#include "video_core/renderer_opengl/gl_rasterizer.h"
-#include "video_core/renderer_opengl/gl_resource_manager.h"
 
 namespace OpenGL {
+namespace {
+struct BindlessSSBO {
+    GLuint64EXT address;
+    GLsizei length;
+    GLsizei padding;
+};
+static_assert(sizeof(BindlessSSBO) == sizeof(GLuint) * 4);
 
-using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+constexpr std::array PROGRAM_LUT{
+    GL_VERTEX_PROGRAM_NV,   GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV,
+    GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV,
+};
+} // Anonymous namespace
 
-MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
+Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
+    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params) {}
 
-Buffer::Buffer(const Device& device_, VAddr cpu_addr_, std::size_t size_)
-    : BufferBlock{cpu_addr_, size_} {
-    gl_buffer.Create();
-    glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size_), nullptr, GL_DYNAMIC_DRAW);
-    if (device_.UseAssemblyShaders() || device_.HasVertexBufferUnifiedMemory()) {
-        glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_WRITE);
-        glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
+Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
+               VAddr cpu_addr_, u64 size_bytes_)
+    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(rasterizer_, cpu_addr_, size_bytes_) {
+    buffer.Create();
+    const std::string name = fmt::format("Buffer 0x{:x}", CpuAddr());
+    glObjectLabel(GL_BUFFER, buffer.handle, static_cast<GLsizei>(name.size()), name.data());
+    glNamedBufferData(buffer.handle, SizeBytes(), nullptr, GL_DYNAMIC_DRAW);
+
+    if (runtime.has_unified_vertex_buffers) {
+        glGetNamedBufferParameterui64vNV(buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &address);
     }
 }
 
-Buffer::~Buffer() = default;
-
-void Buffer::Upload(std::size_t offset, std::size_t data_size, const u8* data) {
-    glNamedBufferSubData(Handle(), static_cast<GLintptr>(offset),
-                         static_cast<GLsizeiptr>(data_size), data);
+void Buffer::ImmediateUpload(size_t offset, std::span<const u8> data) noexcept {
+    glNamedBufferSubData(buffer.handle, static_cast<GLintptr>(offset),
+                         static_cast<GLsizeiptr>(data.size_bytes()), data.data());
 }
 
-void Buffer::Download(std::size_t offset, std::size_t data_size, u8* data) {
-    MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
-    const GLsizeiptr gl_size = static_cast<GLsizeiptr>(data_size);
-    const GLintptr gl_offset = static_cast<GLintptr>(offset);
-    if (read_buffer.handle == 0) {
-        read_buffer.Create();
-        glNamedBufferData(read_buffer.handle, static_cast<GLsizeiptr>(Size()), nullptr,
-                          GL_STREAM_READ);
-    }
-    glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
-    glCopyNamedBufferSubData(gl_buffer.handle, read_buffer.handle, gl_offset, gl_offset, gl_size);
-    glGetNamedBufferSubData(read_buffer.handle, gl_offset, gl_size, data);
+void Buffer::ImmediateDownload(size_t offset, std::span<u8> data) noexcept {
+    glGetNamedBufferSubData(buffer.handle, static_cast<GLintptr>(offset),
+                            static_cast<GLsizeiptr>(data.size_bytes()), data.data());
 }
 
-void Buffer::CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
-                      std::size_t copy_size) {
-    glCopyNamedBufferSubData(src.Handle(), Handle(), static_cast<GLintptr>(src_offset),
-                             static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(copy_size));
-}
-
-OGLBufferCache::OGLBufferCache(VideoCore::RasterizerInterface& rasterizer_,
-                               Tegra::MemoryManager& gpu_memory_, Core::Memory::Memory& cpu_memory_,
-                               const Device& device_, OGLStreamBuffer& stream_buffer_,
-                               StateTracker& state_tracker)
-    : GenericBufferCache{rasterizer_, gpu_memory_, cpu_memory_, stream_buffer_}, device{device_} {
-    if (!device.HasFastBufferSubData()) {
+void Buffer::MakeResident(GLenum access) noexcept {
+    // Abuse GLenum's order to exit early
+    // GL_NONE (default) < GL_READ_ONLY < GL_READ_WRITE
+    if (access <= current_residency_access || buffer.handle == 0) {
         return;
     }
+    if (std::exchange(current_residency_access, access) != GL_NONE) {
+        // If the buffer is already resident, remove its residency before promoting it
+        glMakeNamedBufferNonResidentNV(buffer.handle);
+    }
+    glMakeNamedBufferResidentNV(buffer.handle, access);
+}
 
-    static constexpr GLsizeiptr size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
-    glCreateBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
-    for (const GLuint cbuf : cbufs) {
-        glNamedBufferData(cbuf, size, nullptr, GL_STREAM_DRAW);
+BufferCacheRuntime::BufferCacheRuntime(const Device& device_)
+    : device{device_}, has_fast_buffer_sub_data{device.HasFastBufferSubData()},
+      use_assembly_shaders{device.UseAssemblyShaders()},
+      has_unified_vertex_buffers{device.HasVertexBufferUnifiedMemory()},
+      stream_buffer{has_fast_buffer_sub_data ? std::nullopt : std::make_optional<StreamBuffer>()} {
+    GLint gl_max_attributes;
+    glGetIntegerv(GL_MAX_VERTEX_ATTRIBS, &gl_max_attributes);
+    max_attributes = static_cast<u32>(gl_max_attributes);
+    for (auto& stage_uniforms : fast_uniforms) {
+        for (OGLBuffer& buffer : stage_uniforms) {
+            buffer.Create();
+            glNamedBufferData(buffer.handle, BufferCache::SKIP_CACHE_SIZE, nullptr, GL_STREAM_DRAW);
+        }
+    }
+    for (auto& stage_uniforms : copy_uniforms) {
+        for (OGLBuffer& buffer : stage_uniforms) {
+            buffer.Create();
+            glNamedBufferData(buffer.handle, 0x10'000, nullptr, GL_STREAM_COPY);
+        }
+    }
+    for (OGLBuffer& buffer : copy_compute_uniforms) {
+        buffer.Create();
+        glNamedBufferData(buffer.handle, 0x10'000, nullptr, GL_STREAM_COPY);
     }
 }
 
-OGLBufferCache::~OGLBufferCache() {
-    glDeleteBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
+void BufferCacheRuntime::CopyBuffer(Buffer& dst_buffer, Buffer& src_buffer,
+                                    std::span<const VideoCommon::BufferCopy> copies) {
+    for (const VideoCommon::BufferCopy& copy : copies) {
+        glCopyNamedBufferSubData(
+            src_buffer.Handle(), dst_buffer.Handle(), static_cast<GLintptr>(copy.src_offset),
+            static_cast<GLintptr>(copy.dst_offset), static_cast<GLsizeiptr>(copy.size));
+    }
 }
 
-std::shared_ptr<Buffer> OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
-    return std::make_shared<Buffer>(device, cpu_addr, size);
+void BufferCacheRuntime::BindIndexBuffer(Buffer& buffer, u32 offset, u32 size) {
+    if (has_unified_vertex_buffers) {
+        buffer.MakeResident(GL_READ_ONLY);
+        glBufferAddressRangeNV(GL_ELEMENT_ARRAY_ADDRESS_NV, 0, buffer.HostGpuAddr() + offset,
+                               static_cast<GLsizeiptr>(size));
+    } else {
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer.Handle());
+        index_buffer_offset = offset;
+    }
 }
 
-OGLBufferCache::BufferInfo OGLBufferCache::GetEmptyBuffer(std::size_t) {
-    return {0, 0, 0};
+void BufferCacheRuntime::BindVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size,
+                                          u32 stride) {
+    if (index >= max_attributes) {
+        return;
+    }
+    if (has_unified_vertex_buffers) {
+        buffer.MakeResident(GL_READ_ONLY);
+        glBindVertexBuffer(index, 0, 0, static_cast<GLsizei>(stride));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, index,
+                               buffer.HostGpuAddr() + offset, static_cast<GLsizeiptr>(size));
+    } else {
+        glBindVertexBuffer(index, buffer.Handle(), static_cast<GLintptr>(offset),
+                           static_cast<GLsizei>(stride));
+    }
 }
 
-OGLBufferCache::BufferInfo OGLBufferCache::ConstBufferUpload(const void* raw_pointer,
-                                                             std::size_t size) {
-    DEBUG_ASSERT(cbuf_cursor < std::size(cbufs));
-    const GLuint cbuf = cbufs[cbuf_cursor++];
+void BufferCacheRuntime::BindUniformBuffer(size_t stage, u32 binding_index, Buffer& buffer,
+                                           u32 offset, u32 size) {
+    if (use_assembly_shaders) {
+        GLuint handle;
+        if (offset != 0) {
+            handle = copy_uniforms[stage][binding_index].handle;
+            glCopyNamedBufferSubData(buffer.Handle(), handle, offset, 0, size);
+        } else {
+            handle = buffer.Handle();
+        }
+        glBindBufferRangeNV(PABO_LUT[stage], binding_index, handle, 0,
+                            static_cast<GLsizeiptr>(size));
+    } else {
+        const GLuint base_binding = device.GetBaseBindings(stage).uniform_buffer;
+        const GLuint binding = base_binding + binding_index;
+        glBindBufferRange(GL_UNIFORM_BUFFER, binding, buffer.Handle(),
+                          static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size));
+    }
+}
 
-    glNamedBufferSubData(cbuf, 0, static_cast<GLsizeiptr>(size), raw_pointer);
-    return {cbuf, 0, 0};
+void BufferCacheRuntime::BindComputeUniformBuffer(u32 binding_index, Buffer& buffer, u32 offset,
+                                                  u32 size) {
+    if (use_assembly_shaders) {
+        GLuint handle;
+        if (offset != 0) {
+            handle = copy_compute_uniforms[binding_index].handle;
+            glCopyNamedBufferSubData(buffer.Handle(), handle, offset, 0, size);
+        } else {
+            handle = buffer.Handle();
+        }
+        glBindBufferRangeNV(GL_COMPUTE_PROGRAM_PARAMETER_BUFFER_NV, binding_index, handle, 0,
+                            static_cast<GLsizeiptr>(size));
+    } else {
+        glBindBufferRange(GL_UNIFORM_BUFFER, binding_index, buffer.Handle(),
+                          static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size));
+    }
+}
+
+void BufferCacheRuntime::BindStorageBuffer(size_t stage, u32 binding_index, Buffer& buffer,
+                                           u32 offset, u32 size, bool is_written) {
+    if (use_assembly_shaders) {
+        const BindlessSSBO ssbo{
+            .address = buffer.HostGpuAddr() + offset,
+            .length = static_cast<GLsizei>(size),
+            .padding = 0,
+        };
+        buffer.MakeResident(is_written ? GL_READ_WRITE : GL_READ_ONLY);
+        glProgramLocalParametersI4uivNV(PROGRAM_LUT[stage], binding_index, 1,
+                                        reinterpret_cast<const GLuint*>(&ssbo));
+    } else {
+        const GLuint base_binding = device.GetBaseBindings(stage).shader_storage_buffer;
+        const GLuint binding = base_binding + binding_index;
+        glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, buffer.Handle(),
+                          static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size));
+    }
+}
+
+void BufferCacheRuntime::BindComputeStorageBuffer(u32 binding_index, Buffer& buffer, u32 offset,
+                                                  u32 size, bool is_written) {
+    if (use_assembly_shaders) {
+        const BindlessSSBO ssbo{
+            .address = buffer.HostGpuAddr() + offset,
+            .length = static_cast<GLsizei>(size),
+            .padding = 0,
+        };
+        buffer.MakeResident(is_written ? GL_READ_WRITE : GL_READ_ONLY);
+        glProgramLocalParametersI4uivNV(GL_COMPUTE_PROGRAM_NV, binding_index, 1,
+                                        reinterpret_cast<const GLuint*>(&ssbo));
+    } else if (size == 0) {
+        glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding_index, 0, 0, 0);
+    } else {
+        glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding_index, buffer.Handle(),
+                          static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size));
+    }
+}
+
+void BufferCacheRuntime::BindTransformFeedbackBuffer(u32 index, Buffer& buffer, u32 offset,
+                                                     u32 size) {
+    glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, index, buffer.Handle(),
+                      static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size));
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -5,79 +5,157 @@
 #pragma once
 
 #include <array>
-#include <memory>
+#include <span>
 
+#include "common/alignment.h"
 #include "common/common_types.h"
+#include "common/dynamic_library.h"
 #include "video_core/buffer_cache/buffer_cache.h"
-#include "video_core/engines/maxwell_3d.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
-namespace Core {
-class System;
-}
-
 namespace OpenGL {
 
-class Device;
-class OGLStreamBuffer;
-class RasterizerOpenGL;
-class StateTracker;
+class BufferCacheRuntime;
 
-class Buffer : public VideoCommon::BufferBlock {
+class Buffer : public VideoCommon::BufferBase<VideoCore::RasterizerInterface> {
 public:
-    explicit Buffer(const Device& device_, VAddr cpu_addr_, std::size_t size_);
-    ~Buffer();
+    explicit Buffer(BufferCacheRuntime&, VideoCore::RasterizerInterface& rasterizer, VAddr cpu_addr,
+                    u64 size_bytes);
+    explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams);
 
-    void Upload(std::size_t offset, std::size_t data_size, const u8* data);
+    void ImmediateUpload(size_t offset, std::span<const u8> data) noexcept;
 
-    void Download(std::size_t offset, std::size_t data_size, u8* data);
+    void ImmediateDownload(size_t offset, std::span<u8> data) noexcept;
 
-    void CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
-                  std::size_t copy_size);
+    void MakeResident(GLenum access) noexcept;
 
-    GLuint Handle() const noexcept {
-        return gl_buffer.handle;
+    [[nodiscard]] GLuint64EXT HostGpuAddr() const noexcept {
+        return address;
     }
 
-    u64 Address() const noexcept {
-        return gpu_address;
+    [[nodiscard]] GLuint Handle() const noexcept {
+        return buffer.handle;
     }
 
 private:
-    OGLBuffer gl_buffer;
-    OGLBuffer read_buffer;
-    u64 gpu_address = 0;
+    GLuint64EXT address = 0;
+    OGLBuffer buffer;
+    GLenum current_residency_access = GL_NONE;
 };
 
-using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer>;
-class OGLBufferCache final : public GenericBufferCache {
+class BufferCacheRuntime {
+    friend Buffer;
+
 public:
-    explicit OGLBufferCache(VideoCore::RasterizerInterface& rasterizer,
-                            Tegra::MemoryManager& gpu_memory, Core::Memory::Memory& cpu_memory,
-                            const Device& device, OGLStreamBuffer& stream_buffer,
-                            StateTracker& state_tracker);
-    ~OGLBufferCache();
+    static constexpr u8 INVALID_BINDING = std::numeric_limits<u8>::max();
 
-    BufferInfo GetEmptyBuffer(std::size_t) override;
+    explicit BufferCacheRuntime(const Device& device_);
 
-    void Acquire() noexcept {
-        cbuf_cursor = 0;
+    void CopyBuffer(Buffer& dst_buffer, Buffer& src_buffer,
+                    std::span<const VideoCommon::BufferCopy> copies);
+
+    void BindIndexBuffer(Buffer& buffer, u32 offset, u32 size);
+
+    void BindVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size, u32 stride);
+
+    void BindUniformBuffer(size_t stage, u32 binding_index, Buffer& buffer, u32 offset, u32 size);
+
+    void BindComputeUniformBuffer(u32 binding_index, Buffer& buffer, u32 offset, u32 size);
+
+    void BindStorageBuffer(size_t stage, u32 binding_index, Buffer& buffer, u32 offset, u32 size,
+                           bool is_written);
+
+    void BindComputeStorageBuffer(u32 binding_index, Buffer& buffer, u32 offset, u32 size,
+                                  bool is_written);
+
+    void BindTransformFeedbackBuffer(u32 index, Buffer& buffer, u32 offset, u32 size);
+
+    void BindFastUniformBuffer(size_t stage, u32 binding_index, u32 size) {
+        if (use_assembly_shaders) {
+            const GLuint handle = fast_uniforms[stage][binding_index].handle;
+            const GLsizeiptr gl_size = static_cast<GLsizeiptr>(size);
+            glBindBufferRangeNV(PABO_LUT[stage], binding_index, handle, 0, gl_size);
+        } else {
+            const GLuint base_binding = device.GetBaseBindings(stage).uniform_buffer;
+            const GLuint binding = base_binding + binding_index;
+            glBindBufferRange(GL_UNIFORM_BUFFER, binding,
+                              fast_uniforms[stage][binding_index].handle, 0,
+                              static_cast<GLsizeiptr>(size));
+        }
     }
 
-protected:
-    std::shared_ptr<Buffer> CreateBlock(VAddr cpu_addr, std::size_t size) override;
+    void PushFastUniformBuffer(size_t stage, u32 binding_index, std::span<const u8> data) {
+        if (use_assembly_shaders) {
+            glProgramBufferParametersIuivNV(
+                PABO_LUT[stage], binding_index, 0,
+                static_cast<GLsizei>(data.size_bytes() / sizeof(GLuint)),
+                reinterpret_cast<const GLuint*>(data.data()));
+        } else {
+            glNamedBufferSubData(fast_uniforms[stage][binding_index].handle, 0,
+                                 static_cast<GLsizeiptr>(data.size_bytes()), data.data());
+        }
+    }
 
-    BufferInfo ConstBufferUpload(const void* raw_pointer, std::size_t size) override;
+    std::span<u8> BindMappedUniformBuffer(size_t stage, u32 binding_index, u32 size) noexcept {
+        const auto [mapped_span, offset] = stream_buffer->Request(static_cast<size_t>(size));
+        const GLuint base_binding = device.GetBaseBindings(stage).uniform_buffer;
+        const GLuint binding = base_binding + binding_index;
+        glBindBufferRange(GL_UNIFORM_BUFFER, binding, stream_buffer->Handle(),
+                          static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size));
+        return mapped_span;
+    }
+
+    [[nodiscard]] const GLvoid* IndexOffset() const noexcept {
+        return reinterpret_cast<const GLvoid*>(static_cast<uintptr_t>(index_buffer_offset));
+    }
+
+    [[nodiscard]] bool HasFastBufferSubData() const noexcept {
+        return has_fast_buffer_sub_data;
+    }
 
 private:
-    static constexpr std::size_t NUM_CBUFS = Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
-                                             Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram;
+    static constexpr std::array PABO_LUT{
+        GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV,          GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,
+        GL_TESS_EVALUATION_PROGRAM_PARAMETER_BUFFER_NV, GL_GEOMETRY_PROGRAM_PARAMETER_BUFFER_NV,
+        GL_FRAGMENT_PROGRAM_PARAMETER_BUFFER_NV,
+    };
 
     const Device& device;
 
-    std::size_t cbuf_cursor = 0;
-    std::array<GLuint, NUM_CBUFS> cbufs{};
+    bool has_fast_buffer_sub_data = false;
+    bool use_assembly_shaders = false;
+    bool has_unified_vertex_buffers = false;
+
+    u32 max_attributes = 0;
+
+    std::optional<StreamBuffer> stream_buffer;
+
+    std::array<std::array<OGLBuffer, VideoCommon::NUM_GRAPHICS_UNIFORM_BUFFERS>,
+               VideoCommon::NUM_STAGES>
+        fast_uniforms;
+    std::array<std::array<OGLBuffer, VideoCommon::NUM_GRAPHICS_UNIFORM_BUFFERS>,
+               VideoCommon::NUM_STAGES>
+        copy_uniforms;
+    std::array<OGLBuffer, VideoCommon::NUM_COMPUTE_UNIFORM_BUFFERS> copy_compute_uniforms;
+
+    u32 index_buffer_offset = 0;
 };
 
+struct BufferCacheParams {
+    using Runtime = OpenGL::BufferCacheRuntime;
+    using Buffer = OpenGL::Buffer;
+
+    static constexpr bool IS_OPENGL = true;
+    static constexpr bool HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS = true;
+    static constexpr bool HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT = true;
+    static constexpr bool NEEDS_BIND_UNIFORM_INDEX = true;
+    static constexpr bool NEEDS_BIND_STORAGE_INDEX = true;
+    static constexpr bool USE_MEMORY_MAPS = false;
+};
+
+using BufferCache = VideoCommon::BufferCache<BufferCacheParams>;
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_device.cpp

@@ -21,9 +21,7 @@
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 
 namespace OpenGL {
-
 namespace {
-
 // One uniform block is reserved for emulation purposes
 constexpr u32 ReservedUniformBlocks = 1;
 
@@ -197,11 +195,13 @@ bool IsASTCSupported() {
     const bool nsight = std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
     return nsight || HasExtension(extensions, "GL_EXT_debug_tool");
 }
-
 } // Anonymous namespace
 
-Device::Device()
-    : max_uniform_buffers{BuildMaxUniformBuffers()}, base_bindings{BuildBaseBindings()} {
+Device::Device() {
+    if (!GLAD_GL_VERSION_4_6) {
+        LOG_ERROR(Render_OpenGL, "OpenGL 4.6 is not available");
+        throw std::runtime_error{"Insufficient version"};
+    }
     const std::string_view vendor = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
     const std::string_view version = reinterpret_cast<const char*>(glGetString(GL_VERSION));
     const std::vector extensions = GetExtensions();
@@ -217,6 +217,9 @@ Device::Device()
             "Beta driver 443.24 is known to have issues. There might be performance issues.");
         disable_fast_buffer_sub_data = true;
     }
+
+    max_uniform_buffers = BuildMaxUniformBuffers();
+    base_bindings = BuildBaseBindings();
     uniform_buffer_alignment = GetInteger<size_t>(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT);
     shader_storage_alignment = GetInteger<size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT);
     max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);

src/video_core/renderer_opengl/gl_device.h

@@ -10,11 +10,9 @@
 
 namespace OpenGL {
 
-static constexpr u32 EmulationUniformBlockBinding = 0;
-
-class Device final {
+class Device {
 public:
-    struct BaseBindings final {
+    struct BaseBindings {
         u32 uniform_buffer{};
         u32 shader_storage_buffer{};
         u32 sampler{};

src/video_core/renderer_opengl/gl_fence_manager.cpp

@@ -47,7 +47,7 @@ void GLInnerFence::Wait() {
 
 FenceManagerOpenGL::FenceManagerOpenGL(VideoCore::RasterizerInterface& rasterizer_,
                                        Tegra::GPU& gpu_, TextureCache& texture_cache_,
-                                       OGLBufferCache& buffer_cache_, QueryCache& query_cache_)
+                                       BufferCache& buffer_cache_, QueryCache& query_cache_)
     : GenericFenceManager{rasterizer_, gpu_, texture_cache_, buffer_cache_, query_cache_} {}
 
 Fence FenceManagerOpenGL::CreateFence(u32 value, bool is_stubbed) {

src/video_core/renderer_opengl/gl_fence_manager.h

@@ -32,14 +32,13 @@ private:
 };
 
 using Fence = std::shared_ptr<GLInnerFence>;
-using GenericFenceManager =
-    VideoCommon::FenceManager<Fence, TextureCache, OGLBufferCache, QueryCache>;
+using GenericFenceManager = VideoCommon::FenceManager<Fence, TextureCache, BufferCache, QueryCache>;
 
 class FenceManagerOpenGL final : public GenericFenceManager {
 public:
-    explicit FenceManagerOpenGL(VideoCore::RasterizerInterface& rasterizer_, Tegra::GPU& gpu_,
-                                TextureCache& texture_cache_, OGLBufferCache& buffer_cache_,
-                                QueryCache& query_cache_);
+    explicit FenceManagerOpenGL(VideoCore::RasterizerInterface& rasterizer, Tegra::GPU& gpu,
+                                TextureCache& texture_cache, BufferCache& buffer_cache,
+                                QueryCache& query_cache);
 
 protected:
     Fence CreateFence(u32 value, bool is_stubbed) override;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -38,34 +38,21 @@
 namespace OpenGL {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+using GLvec4 = std::array<GLfloat, 4>;
 
 using Tegra::Engines::ShaderType;
 using VideoCore::Surface::PixelFormat;
 using VideoCore::Surface::SurfaceTarget;
 using VideoCore::Surface::SurfaceType;
 
-MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Format Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_VB, "OpenGL", "Vertex Buffer Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_Shader, "OpenGL", "Shader Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_UBO, "OpenGL", "Const Buffer Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_Index, "OpenGL", "Index Buffer Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_Texture, "OpenGL", "Texture Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_Framebuffer, "OpenGL", "Framebuffer Setup", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_Clears, "OpenGL", "Clears", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
-MICROPROFILE_DEFINE(OpenGL_PrimitiveAssembly, "OpenGL", "Prim Asmbl", MP_RGB(255, 100, 100));
+MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Management", MP_RGB(100, 255, 100));
 
 namespace {
 
-constexpr size_t NUM_CONST_BUFFERS_PER_STAGE = 18;
-constexpr size_t NUM_CONST_BUFFERS_BYTES_PER_STAGE =
-    NUM_CONST_BUFFERS_PER_STAGE * Maxwell::MaxConstBufferSize;
-constexpr size_t TOTAL_CONST_BUFFER_BYTES =
-    NUM_CONST_BUFFERS_BYTES_PER_STAGE * Maxwell::MaxShaderStage;
-
 constexpr size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
-constexpr size_t NUM_SUPPORTED_VERTEX_BINDINGS = 16;
 
 struct TextureHandle {
     constexpr TextureHandle(u32 data, bool via_header_index) {
@@ -101,20 +88,6 @@ TextureHandle GetTextureInfo(const Engine& engine, bool via_header_index, const
     return TextureHandle(engine.AccessConstBuffer32(shader_type, buffer, offset), via_header_index);
 }
 
-std::size_t GetConstBufferSize(const Tegra::Engines::ConstBufferInfo& buffer,
-                               const ConstBufferEntry& entry) {
-    if (!entry.IsIndirect()) {
-        return entry.GetSize();
-    }
-    if (buffer.size > Maxwell::MaxConstBufferSize) {
-        LOG_WARNING(Render_OpenGL, "Indirect constbuffer size {} exceeds maximum {}", buffer.size,
-                    Maxwell::MaxConstBufferSize);
-        return Maxwell::MaxConstBufferSize;
-    }
-
-    return buffer.size;
-}
-
 /// Translates hardware transform feedback indices
 /// @param location Hardware location
 /// @return Pair of ARB_transform_feedback3 token stream first and third arguments
@@ -147,14 +120,6 @@ void oglEnable(GLenum cap, bool state) {
     (state ? glEnable : glDisable)(cap);
 }
 
-void UpdateBindlessSSBOs(GLenum target, const BindlessSSBO* ssbos, size_t num_ssbos) {
-    if (num_ssbos == 0) {
-        return;
-    }
-    glProgramLocalParametersI4uivNV(target, 0, static_cast<GLsizei>(num_ssbos),
-                                    reinterpret_cast<const GLuint*>(ssbos));
-}
-
 ImageViewType ImageViewTypeFromEntry(const SamplerEntry& entry) {
     if (entry.is_buffer) {
         return ImageViewType::Buffer;
@@ -201,44 +166,28 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra
     : RasterizerAccelerated(cpu_memory_), gpu(gpu_), maxwell3d(gpu.Maxwell3D()),
       kepler_compute(gpu.KeplerCompute()), gpu_memory(gpu.MemoryManager()), device(device_),
       screen_info(screen_info_), program_manager(program_manager_), state_tracker(state_tracker_),
-      stream_buffer(device, state_tracker),
       texture_cache_runtime(device, program_manager, state_tracker),
       texture_cache(texture_cache_runtime, *this, maxwell3d, kepler_compute, gpu_memory),
+      buffer_cache_runtime(device),
+      buffer_cache(*this, maxwell3d, kepler_compute, gpu_memory, cpu_memory_, buffer_cache_runtime),
       shader_cache(*this, emu_window_, gpu, maxwell3d, kepler_compute, gpu_memory, device),
       query_cache(*this, maxwell3d, gpu_memory),
-      buffer_cache(*this, gpu_memory, cpu_memory_, device, stream_buffer, state_tracker),
       fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache),
       async_shaders(emu_window_) {
-    unified_uniform_buffer.Create();
-    glNamedBufferStorage(unified_uniform_buffer.handle, TOTAL_CONST_BUFFER_BYTES, nullptr, 0);
-
-    if (device.UseAssemblyShaders()) {
-        glCreateBuffers(static_cast<GLsizei>(staging_cbufs.size()), staging_cbufs.data());
-        for (const GLuint cbuf : staging_cbufs) {
-            glNamedBufferStorage(cbuf, static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize),
-                                 nullptr, 0);
-        }
-    }
     if (device.UseAsynchronousShaders()) {
         async_shaders.AllocateWorkers();
     }
 }
 
-RasterizerOpenGL::~RasterizerOpenGL() {
-    if (device.UseAssemblyShaders()) {
-        glDeleteBuffers(static_cast<GLsizei>(staging_cbufs.size()), staging_cbufs.data());
-    }
-}
+RasterizerOpenGL::~RasterizerOpenGL() = default;
 
-void RasterizerOpenGL::SetupVertexFormat() {
+void RasterizerOpenGL::SyncVertexFormats() {
     auto& flags = maxwell3d.dirty.flags;
     if (!flags[Dirty::VertexFormats]) {
         return;
     }
     flags[Dirty::VertexFormats] = false;
 
-    MICROPROFILE_SCOPE(OpenGL_VAO);
-
     // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL. Enables
     // the first 16 vertex attributes always, as we don't know which ones are actually used until
     // shader time. Note, Tegra technically supports 32, but we're capping this to 16 for now to
@@ -274,55 +223,7 @@ void RasterizerOpenGL::SetupVertexFormat() {
     }
 }
 
-void RasterizerOpenGL::SetupVertexBuffer() {
-    auto& flags = maxwell3d.dirty.flags;
-    if (!flags[Dirty::VertexBuffers]) {
-        return;
-    }
-    flags[Dirty::VertexBuffers] = false;
-
-    MICROPROFILE_SCOPE(OpenGL_VB);
-
-    const bool use_unified_memory = device.HasVertexBufferUnifiedMemory();
-
-    // Upload all guest vertex arrays sequentially to our buffer
-    const auto& regs = maxwell3d.regs;
-    for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_BINDINGS; ++index) {
-        if (!flags[Dirty::VertexBuffer0 + index]) {
-            continue;
-        }
-        flags[Dirty::VertexBuffer0 + index] = false;
-
-        const auto& vertex_array = regs.vertex_array[index];
-        if (!vertex_array.IsEnabled()) {
-            continue;
-        }
-
-        const GPUVAddr start = vertex_array.StartAddress();
-        const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
-        ASSERT(end >= start);
-
-        const GLuint gl_index = static_cast<GLuint>(index);
-        const u64 size = end - start;
-        if (size == 0) {
-            glBindVertexBuffer(gl_index, 0, 0, vertex_array.stride);
-            if (use_unified_memory) {
-                glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, gl_index, 0, 0);
-            }
-            continue;
-        }
-        const auto info = buffer_cache.UploadMemory(start, size);
-        if (use_unified_memory) {
-            glBindVertexBuffer(gl_index, 0, 0, vertex_array.stride);
-            glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, gl_index,
-                                   info.address + info.offset, size);
-        } else {
-            glBindVertexBuffer(gl_index, info.handle, info.offset, vertex_array.stride);
-        }
-    }
-}
-
-void RasterizerOpenGL::SetupVertexInstances() {
+void RasterizerOpenGL::SyncVertexInstances() {
     auto& flags = maxwell3d.dirty.flags;
     if (!flags[Dirty::VertexInstances]) {
         return;
@@ -343,17 +244,7 @@ void RasterizerOpenGL::SetupVertexInstances() {
     }
 }
 
-GLintptr RasterizerOpenGL::SetupIndexBuffer() {
-    MICROPROFILE_SCOPE(OpenGL_Index);
-    const auto& regs = maxwell3d.regs;
-    const std::size_t size = CalculateIndexBufferSize();
-    const auto info = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
-    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, info.handle);
-    return info.offset;
-}
-
-void RasterizerOpenGL::SetupShaders() {
-    MICROPROFILE_SCOPE(OpenGL_Shader);
+void RasterizerOpenGL::SetupShaders(bool is_indexed) {
     u32 clip_distances = 0;
 
     std::array<Shader*, Maxwell::MaxShaderStage> shaders{};
@@ -410,11 +301,19 @@ void RasterizerOpenGL::SetupShaders() {
         const size_t stage = index == 0 ? 0 : index - 1;
         shaders[stage] = shader;
 
-        SetupDrawConstBuffers(stage, shader);
-        SetupDrawGlobalMemory(stage, shader);
         SetupDrawTextures(shader, stage);
         SetupDrawImages(shader, stage);
 
+        buffer_cache.SetEnabledUniformBuffers(stage, shader->GetEntries().enabled_uniform_buffers);
+
+        buffer_cache.UnbindGraphicsStorageBuffers(stage);
+        u32 ssbo_index = 0;
+        for (const auto& buffer : shader->GetEntries().global_memory_entries) {
+            buffer_cache.BindGraphicsStorageBuffer(stage, ssbo_index, buffer.cbuf_index,
+                                                   buffer.cbuf_offset, buffer.is_written);
+            ++ssbo_index;
+        }
+
         // Workaround for Intel drivers.
         // When a clip distance is enabled but not set in the shader it crops parts of the screen
         // (sometimes it's half the screen, sometimes three quarters). To avoid this, enable the
@@ -430,43 +329,26 @@ void RasterizerOpenGL::SetupShaders() {
     SyncClipEnabled(clip_distances);
     maxwell3d.dirty.flags[Dirty::Shaders] = false;
 
+    buffer_cache.UpdateGraphicsBuffers(is_indexed);
+
     const std::span indices_span(image_view_indices.data(), image_view_indices.size());
     texture_cache.FillGraphicsImageViews(indices_span, image_view_ids);
 
+    buffer_cache.BindHostGeometryBuffers(is_indexed);
+
     size_t image_view_index = 0;
     size_t texture_index = 0;
     size_t image_index = 0;
     for (size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) {
         const Shader* const shader = shaders[stage];
-        if (shader) {
-            const auto base = device.GetBaseBindings(stage);
-            BindTextures(shader->GetEntries(), base.sampler, base.image, image_view_index,
-                         texture_index, image_index);
-        }
-    }
-}
-
-std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
-    const auto& regs = maxwell3d.regs;
-
-    std::size_t size = 0;
-    for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
-        if (!regs.vertex_array[index].IsEnabled())
+        if (!shader) {
             continue;
-
-        const GPUVAddr start = regs.vertex_array[index].StartAddress();
-        const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
-
-        size += end - start;
-        ASSERT(end >= start);
+        }
+        buffer_cache.BindHostStageBuffers(stage);
+        const auto& base = device.GetBaseBindings(stage);
+        BindTextures(shader->GetEntries(), base.sampler, base.image, image_view_index,
+                     texture_index, image_index);
     }
-
-    return size;
-}
-
-std::size_t RasterizerOpenGL::CalculateIndexBufferSize() const {
-    return static_cast<std::size_t>(maxwell3d.regs.index_array.count) *
-           static_cast<std::size_t>(maxwell3d.regs.index_array.FormatSizeInBytes());
 }
 
 void RasterizerOpenGL::LoadDiskResources(u64 title_id, const std::atomic_bool& stop_loading,
@@ -475,6 +357,7 @@ void RasterizerOpenGL::LoadDiskResources(u64 title_id, const std::atomic_bool& s
 }
 
 void RasterizerOpenGL::Clear() {
+    MICROPROFILE_SCOPE(OpenGL_Clears);
     if (!maxwell3d.ShouldExecute()) {
         return;
     }
@@ -525,11 +408,9 @@ void RasterizerOpenGL::Clear() {
     }
     UNIMPLEMENTED_IF(regs.clear_flags.viewport);
 
-    {
-        auto lock = texture_cache.AcquireLock();
-        texture_cache.UpdateRenderTargets(true);
-        state_tracker.BindFramebuffer(texture_cache.GetFramebuffer()->Handle());
-    }
+    std::scoped_lock lock{texture_cache.mutex};
+    texture_cache.UpdateRenderTargets(true);
+    state_tracker.BindFramebuffer(texture_cache.GetFramebuffer()->Handle());
 
     if (use_color) {
         glClearBufferfv(GL_COLOR, regs.clear_buffers.RT, regs.clear_color);
@@ -541,7 +422,6 @@ void RasterizerOpenGL::Clear() {
     } else if (use_stencil) {
         glClearBufferiv(GL_STENCIL, 0, &regs.clear_stencil);
     }
-
     ++num_queued_commands;
 }
 
@@ -550,75 +430,12 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
 
     query_cache.UpdateCounters();
 
-    SyncViewport();
-    SyncRasterizeEnable();
-    SyncPolygonModes();
-    SyncColorMask();
-    SyncFragmentColorClampState();
-    SyncMultiSampleState();
-    SyncDepthTestState();
-    SyncDepthClamp();
-    SyncStencilTestState();
-    SyncBlendState();
-    SyncLogicOpState();
-    SyncCullMode();
-    SyncPrimitiveRestart();
-    SyncScissorTest();
-    SyncPointState();
-    SyncLineState();
-    SyncPolygonOffset();
-    SyncAlphaTest();
-    SyncFramebufferSRGB();
-
-    buffer_cache.Acquire();
-    current_cbuf = 0;
-
-    std::size_t buffer_size = CalculateVertexArraysSize();
-
-    // Add space for index buffer
-    if (is_indexed) {
-        buffer_size = Common::AlignUp(buffer_size, 4) + CalculateIndexBufferSize();
-    }
-
-    // Uniform space for the 5 shader stages
-    buffer_size =
-        Common::AlignUp<std::size_t>(buffer_size, 4) +
-        (sizeof(MaxwellUniformData) + device.GetUniformBufferAlignment()) * Maxwell::MaxShaderStage;
-
-    // Add space for at least 18 constant buffers
-    buffer_size += Maxwell::MaxConstBuffers *
-                   (Maxwell::MaxConstBufferSize + device.GetUniformBufferAlignment());
-
-    // Prepare the vertex array.
-    buffer_cache.Map(buffer_size);
-
-    // Prepare vertex array format.
-    SetupVertexFormat();
-
-    // Upload vertex and index data.
-    SetupVertexBuffer();
-    SetupVertexInstances();
-    GLintptr index_buffer_offset = 0;
-    if (is_indexed) {
-        index_buffer_offset = SetupIndexBuffer();
-    }
-
-    // Setup emulation uniform buffer.
-    if (!device.UseAssemblyShaders()) {
-        MaxwellUniformData ubo;
-        ubo.SetFromRegs(maxwell3d);
-        const auto info =
-            buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
-        glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, info.handle, info.offset,
-                          static_cast<GLsizeiptr>(sizeof(ubo)));
-    }
+    SyncState();
 
     // Setup shaders and their used resources.
-    auto lock = texture_cache.AcquireLock();
-    SetupShaders();
+    std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+    SetupShaders(is_indexed);
 
-    // Signal the buffer cache that we are not going to upload more things.
-    buffer_cache.Unmap();
     texture_cache.UpdateRenderTargets(false);
     state_tracker.BindFramebuffer(texture_cache.GetFramebuffer()->Handle());
     program_manager.BindGraphicsPipeline();
@@ -632,7 +449,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     if (is_indexed) {
         const GLint base_vertex = static_cast<GLint>(maxwell3d.regs.vb_element_base);
         const GLsizei num_vertices = static_cast<GLsizei>(maxwell3d.regs.index_array.count);
-        const GLvoid* offset = reinterpret_cast<const GLvoid*>(index_buffer_offset);
+        const GLvoid* const offset = buffer_cache_runtime.IndexOffset();
         const GLenum format = MaxwellToGL::IndexFormat(maxwell3d.regs.index_array.format);
         if (num_instances == 1 && base_instance == 0 && base_vertex == 0) {
             glDrawElements(primitive_mode, num_vertices, format, offset);
@@ -672,22 +489,22 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
 }
 
 void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
-    buffer_cache.Acquire();
-    current_cbuf = 0;
-
     Shader* const kernel = shader_cache.GetComputeKernel(code_addr);
 
-    auto lock = texture_cache.AcquireLock();
+    std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
     BindComputeTextures(kernel);
 
-    const size_t buffer_size = Tegra::Engines::KeplerCompute::NumConstBuffers *
-                               (Maxwell::MaxConstBufferSize + device.GetUniformBufferAlignment());
-    buffer_cache.Map(buffer_size);
-
-    SetupComputeConstBuffers(kernel);
-    SetupComputeGlobalMemory(kernel);
-
-    buffer_cache.Unmap();
+    const auto& entries = kernel->GetEntries();
+    buffer_cache.SetEnabledComputeUniformBuffers(entries.enabled_uniform_buffers);
+    buffer_cache.UnbindComputeStorageBuffers();
+    u32 ssbo_index = 0;
+    for (const auto& buffer : entries.global_memory_entries) {
+        buffer_cache.BindComputeStorageBuffer(ssbo_index, buffer.cbuf_index, buffer.cbuf_offset,
+                                              buffer.is_written);
+        ++ssbo_index;
+    }
+    buffer_cache.UpdateComputeBuffers();
+    buffer_cache.BindHostComputeBuffers();
 
     const auto& launch_desc = kepler_compute.launch_description;
     glDispatchCompute(launch_desc.grid_dim_x, launch_desc.grid_dim_y, launch_desc.grid_dim_z);
@@ -703,6 +520,12 @@ void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCore::QueryType type,
     query_cache.Query(gpu_addr, type, timestamp);
 }
 
+void RasterizerOpenGL::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
+                                                 u32 size) {
+    std::scoped_lock lock{buffer_cache.mutex};
+    buffer_cache.BindGraphicsUniformBuffer(stage, index, gpu_addr, size);
+}
+
 void RasterizerOpenGL::FlushAll() {}
 
 void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
@@ -711,19 +534,23 @@ void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
         return;
     }
     {
-        auto lock = texture_cache.AcquireLock();
+        std::scoped_lock lock{texture_cache.mutex};
         texture_cache.DownloadMemory(addr, size);
     }
-    buffer_cache.FlushRegion(addr, size);
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        buffer_cache.DownloadMemory(addr, size);
+    }
     query_cache.FlushRegion(addr, size);
 }
 
 bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size) {
+    std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
     if (!Settings::IsGPULevelHigh()) {
-        return buffer_cache.MustFlushRegion(addr, size);
+        return buffer_cache.IsRegionGpuModified(addr, size);
     }
     return texture_cache.IsRegionGpuModified(addr, size) ||
-           buffer_cache.MustFlushRegion(addr, size);
+           buffer_cache.IsRegionGpuModified(addr, size);
 }
 
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
@@ -732,11 +559,14 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
         return;
     }
     {
-        auto lock = texture_cache.AcquireLock();
+        std::scoped_lock lock{texture_cache.mutex};
         texture_cache.WriteMemory(addr, size);
     }
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        buffer_cache.WriteMemory(addr, size);
+    }
     shader_cache.InvalidateRegion(addr, size);
-    buffer_cache.InvalidateRegion(addr, size);
     query_cache.InvalidateRegion(addr, size);
 }
 
@@ -745,26 +575,35 @@ void RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
     if (addr == 0 || size == 0) {
         return;
     }
+    shader_cache.OnCPUWrite(addr, size);
     {
-        auto lock = texture_cache.AcquireLock();
+        std::scoped_lock lock{texture_cache.mutex};
         texture_cache.WriteMemory(addr, size);
     }
-    shader_cache.OnCPUWrite(addr, size);
-    buffer_cache.OnCPUWrite(addr, size);
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        buffer_cache.CachedWriteMemory(addr, size);
+    }
 }
 
 void RasterizerOpenGL::SyncGuestHost() {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    buffer_cache.SyncGuestHost();
     shader_cache.SyncGuestHost();
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        buffer_cache.FlushCachedWrites();
+    }
 }
 
 void RasterizerOpenGL::UnmapMemory(VAddr addr, u64 size) {
     {
-        auto lock = texture_cache.AcquireLock();
+        std::scoped_lock lock{texture_cache.mutex};
         texture_cache.UnmapMemory(addr, size);
     }
-    buffer_cache.OnCPUWrite(addr, size);
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        buffer_cache.WriteMemory(addr, size);
+    }
     shader_cache.OnCPUWrite(addr, size);
 }
 
@@ -799,14 +638,7 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
 }
 
 void RasterizerOpenGL::WaitForIdle() {
-    // Place a barrier on everything that is not framebuffer related.
-    // This is related to another flag that is not currently implemented.
-    glMemoryBarrier(GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT | GL_ELEMENT_ARRAY_BARRIER_BIT |
-                    GL_UNIFORM_BARRIER_BIT | GL_TEXTURE_FETCH_BARRIER_BIT |
-                    GL_SHADER_IMAGE_ACCESS_BARRIER_BIT | GL_COMMAND_BARRIER_BIT |
-                    GL_PIXEL_BUFFER_BARRIER_BIT | GL_TEXTURE_UPDATE_BARRIER_BIT |
-                    GL_BUFFER_UPDATE_BARRIER_BIT | GL_TRANSFORM_FEEDBACK_BARRIER_BIT |
-                    GL_SHADER_STORAGE_BARRIER_BIT | GL_QUERY_BUFFER_BARRIER_BIT);
+    glMemoryBarrier(GL_ALL_BARRIER_BITS);
 }
 
 void RasterizerOpenGL::FragmentBarrier() {
@@ -831,18 +663,21 @@ void RasterizerOpenGL::TickFrame() {
     num_queued_commands = 0;
 
     fence_manager.TickFrame();
-    buffer_cache.TickFrame();
     {
-        auto lock = texture_cache.AcquireLock();
+        std::scoped_lock lock{texture_cache.mutex};
         texture_cache.TickFrame();
     }
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        buffer_cache.TickFrame();
+    }
 }
 
 bool RasterizerOpenGL::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
                                              const Tegra::Engines::Fermi2D::Surface& dst,
                                              const Tegra::Engines::Fermi2D::Config& copy_config) {
     MICROPROFILE_SCOPE(OpenGL_Blits);
-    auto lock = texture_cache.AcquireLock();
+    std::scoped_lock lock{texture_cache.mutex};
     texture_cache.BlitImage(dst, src, copy_config);
     return true;
 }
@@ -854,7 +689,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
     }
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
 
-    auto lock = texture_cache.AcquireLock();
+    std::scoped_lock lock{texture_cache.mutex};
     ImageView* const image_view{texture_cache.TryFindFramebufferImageView(framebuffer_addr)};
     if (!image_view) {
         return false;
@@ -921,166 +756,6 @@ void RasterizerOpenGL::BindTextures(const ShaderEntries& entries, GLuint base_te
     }
 }
 
-void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, Shader* shader) {
-    static constexpr std::array PARAMETER_LUT{
-        GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV,          GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,
-        GL_TESS_EVALUATION_PROGRAM_PARAMETER_BUFFER_NV, GL_GEOMETRY_PROGRAM_PARAMETER_BUFFER_NV,
-        GL_FRAGMENT_PROGRAM_PARAMETER_BUFFER_NV,
-    };
-    MICROPROFILE_SCOPE(OpenGL_UBO);
-    const auto& stages = maxwell3d.state.shader_stages;
-    const auto& shader_stage = stages[stage_index];
-    const auto& entries = shader->GetEntries();
-    const bool use_unified = entries.use_unified_uniforms;
-    const std::size_t base_unified_offset = stage_index * NUM_CONST_BUFFERS_BYTES_PER_STAGE;
-
-    const auto base_bindings = device.GetBaseBindings(stage_index);
-    u32 binding = device.UseAssemblyShaders() ? 0 : base_bindings.uniform_buffer;
-    for (const auto& entry : entries.const_buffers) {
-        const u32 index = entry.GetIndex();
-        const auto& buffer = shader_stage.const_buffers[index];
-        SetupConstBuffer(PARAMETER_LUT[stage_index], binding, buffer, entry, use_unified,
-                         base_unified_offset + index * Maxwell::MaxConstBufferSize);
-        ++binding;
-    }
-    if (use_unified) {
-        const u32 index = static_cast<u32>(base_bindings.shader_storage_buffer +
-                                           entries.global_memory_entries.size());
-        glBindBufferRange(GL_SHADER_STORAGE_BUFFER, index, unified_uniform_buffer.handle,
-                          base_unified_offset, NUM_CONST_BUFFERS_BYTES_PER_STAGE);
-    }
-}
-
-void RasterizerOpenGL::SetupComputeConstBuffers(Shader* kernel) {
-    MICROPROFILE_SCOPE(OpenGL_UBO);
-    const auto& launch_desc = kepler_compute.launch_description;
-    const auto& entries = kernel->GetEntries();
-    const bool use_unified = entries.use_unified_uniforms;
-
-    u32 binding = 0;
-    for (const auto& entry : entries.const_buffers) {
-        const auto& config = launch_desc.const_buffer_config[entry.GetIndex()];
-        const std::bitset<8> mask = launch_desc.const_buffer_enable_mask.Value();
-        Tegra::Engines::ConstBufferInfo buffer;
-        buffer.address = config.Address();
-        buffer.size = config.size;
-        buffer.enabled = mask[entry.GetIndex()];
-        SetupConstBuffer(GL_COMPUTE_PROGRAM_PARAMETER_BUFFER_NV, binding, buffer, entry,
-                         use_unified, entry.GetIndex() * Maxwell::MaxConstBufferSize);
-        ++binding;
-    }
-    if (use_unified) {
-        const GLuint index = static_cast<GLuint>(entries.global_memory_entries.size());
-        glBindBufferRange(GL_SHADER_STORAGE_BUFFER, index, unified_uniform_buffer.handle, 0,
-                          NUM_CONST_BUFFERS_BYTES_PER_STAGE);
-    }
-}
-
-void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
-                                        const Tegra::Engines::ConstBufferInfo& buffer,
-                                        const ConstBufferEntry& entry, bool use_unified,
-                                        std::size_t unified_offset) {
-    if (!buffer.enabled) {
-        // Set values to zero to unbind buffers
-        if (device.UseAssemblyShaders()) {
-            glBindBufferRangeNV(stage, entry.GetIndex(), 0, 0, 0);
-        } else {
-            glBindBufferRange(GL_UNIFORM_BUFFER, binding, 0, 0, sizeof(float));
-        }
-        return;
-    }
-
-    // Align the actual size so it ends up being a multiple of vec4 to meet the OpenGL std140
-    // UBO alignment requirements.
-    const std::size_t size = Common::AlignUp(GetConstBufferSize(buffer, entry), sizeof(GLvec4));
-
-    const bool fast_upload = !use_unified && device.HasFastBufferSubData();
-
-    const std::size_t alignment = use_unified ? 4 : device.GetUniformBufferAlignment();
-    const GPUVAddr gpu_addr = buffer.address;
-    auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
-
-    if (device.UseAssemblyShaders()) {
-        UNIMPLEMENTED_IF(use_unified);
-        if (info.offset != 0) {
-            const GLuint staging_cbuf = staging_cbufs[current_cbuf++];
-            glCopyNamedBufferSubData(info.handle, staging_cbuf, info.offset, 0, size);
-            info.handle = staging_cbuf;
-            info.offset = 0;
-        }
-        glBindBufferRangeNV(stage, binding, info.handle, info.offset, size);
-        return;
-    }
-
-    if (use_unified) {
-        glCopyNamedBufferSubData(info.handle, unified_uniform_buffer.handle, info.offset,
-                                 unified_offset, size);
-    } else {
-        glBindBufferRange(GL_UNIFORM_BUFFER, binding, info.handle, info.offset, size);
-    }
-}
-
-void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, Shader* shader) {
-    static constexpr std::array TARGET_LUT = {
-        GL_VERTEX_PROGRAM_NV,   GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV,
-        GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV,
-    };
-    const auto& cbufs{maxwell3d.state.shader_stages[stage_index]};
-    const auto& entries{shader->GetEntries().global_memory_entries};
-
-    std::array<BindlessSSBO, 32> ssbos;
-    ASSERT(entries.size() < ssbos.size());
-
-    const bool assembly_shaders = device.UseAssemblyShaders();
-    u32 binding = assembly_shaders ? 0 : device.GetBaseBindings(stage_index).shader_storage_buffer;
-    for (const auto& entry : entries) {
-        const GPUVAddr addr{cbufs.const_buffers[entry.cbuf_index].address + entry.cbuf_offset};
-        const GPUVAddr gpu_addr{gpu_memory.Read<u64>(addr)};
-        const u32 size{gpu_memory.Read<u32>(addr + 8)};
-        SetupGlobalMemory(binding, entry, gpu_addr, size, &ssbos[binding]);
-        ++binding;
-    }
-    if (assembly_shaders) {
-        UpdateBindlessSSBOs(TARGET_LUT[stage_index], ssbos.data(), entries.size());
-    }
-}
-
-void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
-    const auto& cbufs{kepler_compute.launch_description.const_buffer_config};
-    const auto& entries{kernel->GetEntries().global_memory_entries};
-
-    std::array<BindlessSSBO, 32> ssbos;
-    ASSERT(entries.size() < ssbos.size());
-
-    u32 binding = 0;
-    for (const auto& entry : entries) {
-        const GPUVAddr addr{cbufs[entry.cbuf_index].Address() + entry.cbuf_offset};
-        const GPUVAddr gpu_addr{gpu_memory.Read<u64>(addr)};
-        const u32 size{gpu_memory.Read<u32>(addr + 8)};
-        SetupGlobalMemory(binding, entry, gpu_addr, size, &ssbos[binding]);
-        ++binding;
-    }
-    if (device.UseAssemblyShaders()) {
-        UpdateBindlessSSBOs(GL_COMPUTE_PROGRAM_NV, ssbos.data(), ssbos.size());
-    }
-}
-
-void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry,
-                                         GPUVAddr gpu_addr, size_t size, BindlessSSBO* ssbo) {
-    const size_t alignment{device.GetShaderStorageBufferAlignment()};
-    const auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
-    if (device.UseAssemblyShaders()) {
-        *ssbo = BindlessSSBO{
-            .address = static_cast<GLuint64EXT>(info.address + info.offset),
-            .length = static_cast<GLsizei>(size),
-            .padding = 0,
-        };
-    } else {
-        glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, info.handle, info.offset,
-                          static_cast<GLsizeiptr>(size));
-    }
-}
-
 void RasterizerOpenGL::SetupDrawTextures(const Shader* shader, size_t stage_index) {
     const bool via_header_index =
         maxwell3d.regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex;
@@ -1128,6 +803,30 @@ void RasterizerOpenGL::SetupComputeImages(const Shader* shader) {
     }
 }
 
+void RasterizerOpenGL::SyncState() {
+    SyncViewport();
+    SyncRasterizeEnable();
+    SyncPolygonModes();
+    SyncColorMask();
+    SyncFragmentColorClampState();
+    SyncMultiSampleState();
+    SyncDepthTestState();
+    SyncDepthClamp();
+    SyncStencilTestState();
+    SyncBlendState();
+    SyncLogicOpState();
+    SyncCullMode();
+    SyncPrimitiveRestart();
+    SyncScissorTest();
+    SyncPointState();
+    SyncLineState();
+    SyncPolygonOffset();
+    SyncAlphaTest();
+    SyncFramebufferSRGB();
+    SyncVertexFormats();
+    SyncVertexInstances();
+}
+
 void RasterizerOpenGL::SyncViewport() {
     auto& flags = maxwell3d.dirty.flags;
     const auto& regs = maxwell3d.regs;
@@ -1163,9 +862,11 @@ void RasterizerOpenGL::SyncViewport() {
         if (regs.screen_y_control.y_negate != 0) {
             flip_y = !flip_y;
         }
-        glClipControl(flip_y ? GL_UPPER_LEFT : GL_LOWER_LEFT,
-                      regs.depth_mode == Maxwell::DepthMode::ZeroToOne ? GL_ZERO_TO_ONE
-                                                                       : GL_NEGATIVE_ONE_TO_ONE);
+        const bool is_zero_to_one = regs.depth_mode == Maxwell::DepthMode::ZeroToOne;
+        const GLenum origin = flip_y ? GL_UPPER_LEFT : GL_LOWER_LEFT;
+        const GLenum depth = is_zero_to_one ? GL_ZERO_TO_ONE : GL_NEGATIVE_ONE_TO_ONE;
+        state_tracker.ClipControl(origin, depth);
+        state_tracker.SetYNegate(regs.screen_y_control.y_negate != 0);
     }
 
     if (dirty_viewport) {
@@ -1649,36 +1350,13 @@ void RasterizerOpenGL::BeginTransformFeedback(GLenum primitive_mode) {
     if (regs.tfb_enabled == 0) {
         return;
     }
-
     if (device.UseAssemblyShaders()) {
         SyncTransformFeedback();
     }
-
     UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::TesselationControl) ||
                      regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::TesselationEval) ||
                      regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::Geometry));
-
-    for (std::size_t index = 0; index < Maxwell::NumTransformFeedbackBuffers; ++index) {
-        const auto& binding = regs.tfb_bindings[index];
-        if (!binding.buffer_enable) {
-            if (enabled_transform_feedback_buffers[index]) {
-                glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, static_cast<GLuint>(index), 0, 0,
-                                  0);
-            }
-            enabled_transform_feedback_buffers[index] = false;
-            continue;
-        }
-        enabled_transform_feedback_buffers[index] = true;
-
-        auto& tfb_buffer = transform_feedback_buffers[index];
-        tfb_buffer.Create();
-
-        const GLuint handle = tfb_buffer.handle;
-        const std::size_t size = binding.buffer_size;
-        glNamedBufferData(handle, static_cast<GLsizeiptr>(size), nullptr, GL_STREAM_COPY);
-        glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, static_cast<GLuint>(index), handle, 0,
-                          static_cast<GLsizeiptr>(size));
-    }
+    UNIMPLEMENTED_IF(primitive_mode != GL_POINTS);
 
     // We may have to call BeginTransformFeedbackNV here since they seem to call different
     // implementations on Nvidia's driver (the pointer is different) but we are using
@@ -1692,23 +1370,7 @@ void RasterizerOpenGL::EndTransformFeedback() {
     if (regs.tfb_enabled == 0) {
         return;
     }
-
     glEndTransformFeedback();
-
-    for (std::size_t index = 0; index < Maxwell::NumTransformFeedbackBuffers; ++index) {
-        const auto& binding = regs.tfb_bindings[index];
-        if (!binding.buffer_enable) {
-            continue;
-        }
-        UNIMPLEMENTED_IF(binding.buffer_offset != 0);
-
-        const GLuint handle = transform_feedback_buffers[index].handle;
-        const GPUVAddr gpu_addr = binding.Address();
-        const std::size_t size = binding.buffer_size;
-        const auto info = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
-        glCopyNamedBufferSubData(handle, info.handle, 0, info.offset,
-                                 static_cast<GLsizeiptr>(size));
-    }
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -30,7 +30,6 @@
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state_tracker.h"
-#include "video_core/renderer_opengl/gl_stream_buffer.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 #include "video_core/shader/async_shaders.h"
 #include "video_core/textures/texture.h"
@@ -72,6 +71,7 @@ public:
     void DispatchCompute(GPUVAddr code_addr) override;
     void ResetCounter(VideoCore::QueryType type) override;
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
+    void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
     void FlushAll() override;
     void FlushRegion(VAddr addr, u64 size) override;
     bool MustFlushRegion(VAddr addr, u64 size) override;
@@ -119,27 +119,6 @@ private:
     void BindTextures(const ShaderEntries& entries, GLuint base_texture, GLuint base_image,
                       size_t& image_view_index, size_t& texture_index, size_t& image_index);
 
-    /// Configures the current constbuffers to use for the draw command.
-    void SetupDrawConstBuffers(std::size_t stage_index, Shader* shader);
-
-    /// Configures the current constbuffers to use for the kernel invocation.
-    void SetupComputeConstBuffers(Shader* kernel);
-
-    /// Configures a constant buffer.
-    void SetupConstBuffer(GLenum stage, u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
-                          const ConstBufferEntry& entry, bool use_unified,
-                          std::size_t unified_offset);
-
-    /// Configures the current global memory entries to use for the draw command.
-    void SetupDrawGlobalMemory(std::size_t stage_index, Shader* shader);
-
-    /// Configures the current global memory entries to use for the kernel invocation.
-    void SetupComputeGlobalMemory(Shader* kernel);
-
-    /// Configures a global memory buffer.
-    void SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry, GPUVAddr gpu_addr,
-                           size_t size, BindlessSSBO* ssbo);
-
     /// Configures the current textures to use for the draw command.
     void SetupDrawTextures(const Shader* shader, size_t stage_index);
 
@@ -152,6 +131,9 @@ private:
     /// Configures images in a compute shader.
     void SetupComputeImages(const Shader* shader);
 
+    /// Syncs state to match guest's
+    void SyncState();
+
     /// Syncs the viewport and depth range to match the guest state
     void SyncViewport();
 
@@ -215,6 +197,12 @@ private:
     /// Syncs the framebuffer sRGB state to match the guest state
     void SyncFramebufferSRGB();
 
+    /// Syncs vertex formats to match the guest state
+    void SyncVertexFormats();
+
+    /// Syncs vertex instances to match the guest state
+    void SyncVertexInstances();
+
     /// Syncs transform feedback state to match guest state
     /// @note Only valid on assembly shaders
     void SyncTransformFeedback();
@@ -225,19 +213,7 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
-    std::size_t CalculateVertexArraysSize() const;
-
-    std::size_t CalculateIndexBufferSize() const;
-
-    /// Updates the current vertex format
-    void SetupVertexFormat();
-
-    void SetupVertexBuffer();
-    void SetupVertexInstances();
-
-    GLintptr SetupIndexBuffer();
-
-    void SetupShaders();
+    void SetupShaders(bool is_indexed);
 
     Tegra::GPU& gpu;
     Tegra::Engines::Maxwell3D& maxwell3d;
@@ -249,12 +225,12 @@ private:
     ProgramManager& program_manager;
     StateTracker& state_tracker;
 
-    OGLStreamBuffer stream_buffer;
     TextureCacheRuntime texture_cache_runtime;
     TextureCache texture_cache;
+    BufferCacheRuntime buffer_cache_runtime;
+    BufferCache buffer_cache;
     ShaderCacheOpenGL shader_cache;
     QueryCache query_cache;
-    OGLBufferCache buffer_cache;
     FenceManagerOpenGL fence_manager;
 
     VideoCommon::Shader::AsyncShaders async_shaders;
@@ -262,20 +238,8 @@ private:
     boost::container::static_vector<u32, MAX_IMAGE_VIEWS> image_view_indices;
     std::array<ImageViewId, MAX_IMAGE_VIEWS> image_view_ids;
     boost::container::static_vector<GLuint, MAX_TEXTURES> sampler_handles;
-    std::array<GLuint, MAX_TEXTURES> texture_handles;
-    std::array<GLuint, MAX_IMAGES> image_handles;
-
-    std::array<OGLBuffer, Tegra::Engines::Maxwell3D::Regs::NumTransformFeedbackBuffers>
-        transform_feedback_buffers;
-    std::bitset<Tegra::Engines::Maxwell3D::Regs::NumTransformFeedbackBuffers>
-        enabled_transform_feedback_buffers;
-
-    static constexpr std::size_t NUM_CONSTANT_BUFFERS =
-        Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
-        Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram;
-    std::array<GLuint, NUM_CONSTANT_BUFFERS> staging_cbufs{};
-    std::size_t current_cbuf = 0;
-    OGLBuffer unified_uniform_buffer;
+    std::array<GLuint, MAX_TEXTURES> texture_handles{};
+    std::array<GLuint, MAX_IMAGES> image_handles{};
 
     /// Number of commands queued to the OpenGL driver. Resetted on flush.
     std::size_t num_queued_commands = 0;